Family Options Study

Size: px
Start display at page:

Download "Family Options Study"

Transcription

1 Fmily Options Study Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies U.S. Deprtment of Housing nd Urbn Development Office of Policy Development nd Reserch

2 Fmily Options Study Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies Prepred for U.S. Deprtment of Housing nd Urbn Development Office of Policy Development nd Reserch Prepred by Dniel Gubits Mrybeth Shinn Stephen Bell Michelle Wood Smuel Dstrup Cludi D. Solri Scott R. Brown Steven Brown Luren Dunton Winston Lin Debi McInnis Json Rodriguez Glen Svidge Brooke E. Spellmn Abt Assocites, Inc. In prtnership with Vnderbilt University July 2015 Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies i

3 ACKNOWLEDGMENTS The uthors of this report grtefully cknowledge the efforts of mny individuls who hve ssisted in crrying out the Fmily Options Study since it begn in They prticulrly thnk the Government Technicl Reviewer (GTR), Anne Fletcher, for her unwvering guidnce nd support throughout ll phses of the reserch. Other stff members t the U.S. Deprtment of Housing nd Urbn Development (HUD), Office of Policy Development nd Reserch (PD&R) hve provided vluble guidnce nd technicl oversight for the study. In prticulr, the uthors thnk Pul Dornn, the study s former GTR, for his substntil input into the study s design nd specifiction of study interventions. HUD s Office of Specil Needs Assistnce Progrms hs been n ctive prtner in the design nd implementtion of the Fmily Options Study, nd the reserch tem is grteful for tht office s guidnce nd support in developing the study design nd implementtion pln. In ddition, the uthors thnk Lynn Rodgers of PD&R for her ssistnce with cquiring HUD Public nd Indin Housing Informtion Center (PIC) nd Tennt Rentl Assistnce Certifiction System (TRACS) dt. Specil thnks re due to the fmilies prticipting in the Fmily Options Study who hve continued to shre their experiences nd open their lives to the study tem. The Fmily Options Study hs benefited from the contributions of number of technicl experts. The study design reflects guidnce from experts in rndom ssignment methodology nd from subject mtter experts knowledgeble bout nd close to the opertionl relities of the homeless ssistnce service delivery system. The uthors re especilly grteful to Ellen Bssuk (The Center for Socil Innovtion), Mrth Burt (independent consultnt), Dennis Culhne (University of Pennsylvni), Lrry Orr (The Johns Hopkins University), nd Beth Weitzmn (New York University) for their help in designing the study interventions nd the reserch pproch. Stephen Bell (Abt Assocites) nd Mrybeth Shinn (Vnderbilt University) re the study s Co-Principl Investigtors. Michelle Wood (Abt Assocites) is the Project Director. Dniel Gubits (Abt Assocites) is the study s Director of Anlysis. Implementing the study design required the efforts of lrge tem to recruit sites, develop site-specific implementtion plns, nd conduct rndom ssignment. Brooke Spellmn (Abt Assocites) led the site recruitment nd implementtion ctivities, with contributions from site reserch liisons Burt, Culhne, Shinn, Donld Chmberlin (independent consultnt), Wendy Vulton (formerly with The Center for Socil Innovtion), Mtt White (Abt Assocites), nd Suznne Zerger (formerly with The Center for Socil Innovtion). Styendr Ptrbnsh (formerly with Abt Assocites) designed the complex rndom ssignment lgorithm used to ssign fmilies, nd Sge Computing developed the rndomiztion softwre. Scott Brown (Vnderbilt University), Luren Dunton, Nichole Fiore, nd Meghn Henry (Abt Assocites) monitored rndom ssignment, collecting updtes bout the vilbility of intervention slots nd the sttus of fmilies enrolled in ech intervention. The Fmily Options Study involved n extensive field dt collection effort to collect informtion directly from study prticipnts t severl junctures. AbtSRBI led the prticipnt dt collection ctivities under the direction of Ricki Jrmon nd Brend Rodriquez, the study s survey director. Ashley Brdbury nd Brinn Roche (AbtSRBI) nd Debi McInnis (Abt Assocites) lso plyed key roles in coordinting nd monitoring dt collection ctivities. Kthy Gill nd Lynn Reneu (Abt SRBI) served s survey field mngers nd over - sw the 17 field interviewers working in the study sites. Smuel Dstrup (Abt Assocites) led the cost dt collection efforts, ssisted by Scott Brown, Burt, Dunton, Fiore, Ktherine Buck (Abt Assocites), nd Glen Svidge (Abt Assocites). Dunton nd Cludi Solri (Abt Assocites) mnged the collection nd nlysis of Homeless Mngement Informtion System (HMIS) dt from the 12 study sites, nd Steven Brown (formerly with Abt Assocites) coordinted nlysis of HUD dministrtive dt from PIC nd TRACS. Under the direction of Gubits, Scott Brown nd Steven Brown plyed key roles in estimting impcts using survey nd dministrtive dt. The study s co-project qulity dvisors, Jill Khdduri nd Jcob Klermn (Abt Assocites), provided technicl guidnce throughout ll phses of the project nd thoughtful nd constructive comments on this report. Missy Robinson (Abt Assocites) did n excellent job producing this complex document. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies ii

4 Acknowledgments The study would not be possible without the enthusism nd dediction of locl service providers, Continuum of Cre (CoC) leders, nd public housing gencies in the 12 prticipting communities: (1) Almed County, Cliforni; (2) Atlnt, Georgi; (3) Bltimore, Mrylnd; (4) Boston, Msschusetts; (5) the New Hven nd Bridgeport regions of Connecticut; (6) Denver, Colordo; (7) Honolulu, Hwii; (8) Knss City, Missouri; (9) Louisville, Kentucky; (10) Minnepolis, Minnesot; (11) Phoenix, Arizon; nd (12) Slt Lke City, Uth. The study tem is grteful for their mny contributions nd continued commitment to the study. The study tem would lso like to thnk the HMIS dministrtors in ech of the CoCs for providing HMIS dt for prticipting fmilies. A grnt to Vnderbilt University by the Eunice Kennedy Shriver Ntionl Institute of Child Helth nd Humn Development (NICHD) supported child dt collection nd nlysis for this study. DISCLAIMER The contents of this report re the views of the contrctor nd do not necessrily reflect the views or policies of the U.S. Deprtment of Housing nd Urbn Development or the U.S. Government. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies iii

5 FOREWORD Fmily homelessness hs stubbornly persisted despite more thn three decdes of federl investment. The 2010 relese of Opening Doors: Federl Strtegic Pln To Prevent nd End Homelessness estblished n mbitious gol to end homelessness mong children, fmilies, nd youth by Until now, HUD hd little empiricl evidence compring outcomes cross interventions to guide policy nd progrm decisions towrd this gol. The Fmily Options Study, lunched by HUD in 2008 (nd still under wy), is rigorously designed experimentl study intended to provide the strongest evidence possible bout the effectiveness nd reltive costs of four min interventions vilble to homeless fmilies permnent housing subsidy, project-bsed trnsitionl housing, community-bsed rpid re-housing, or usul cre. More thn 2,200 homeless fmilies, including more thn 5,000 children in 12 communities, were rndomly ssigned to one of four interventions. The fmilies re being trcked for minimum of 3 yers nd were extensively interviewed t bseline, 18 months fter rndom ssignment, nd gin 36 months fter rndom ssignment to ssess outcomes relted to housing stbility, fmily preservtion, dult well-being, child well-being, nd self-sufficiency. This report documents outcomes t 18 months, presenting striking evidence of the power of offering permnent subsidy to homeless fmily. Fmilies who were offered housing voucher experienced significnt reductions in subsequent homelessness, mobility, child seprtions, dult psychologicl distress, experiences of intimte prtner violence, school mobility mong children, nd food insecurity t 18 months. Moreover, the benefits of the voucher intervention were chieved t comprble cost to rpid re-housing nd emergency shelter nd t lower cost thn trnsitionl housing. The study design is both rigorous nd mbitious, nd the rndom ssignment nd subsequent contrsts in progrm use will provide strong bsis for informing future federl policy ddressing fmily homelessness. Results t 36 months will revel if the study findings re sustined over time. In the interim, this report provides unprecedented evidence tht housing vouchers mesurbly improve outcomes for homeless fmilies. It is impossible to thnk individully ll the people who hve contributed to this report nd the broder study since its inception in 2008, so I simply echo the pprecition expressed in the report s cknowledgments. I would be remiss, however, if I did not explicitly thnk Anne Fletcher for her mzing diligence s the Government Technicl Reviewer for this importnt but potentilly unwieldy study or if I did not remind reders of the true generosity of the fmilies who continue to prticipte in this study. Ktherine M. O Regn Assistnt Secretry for Policy Development & Reserch U.S. Deprtment of Housing nd Urbn Development Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies iv

6 TABLE OF CONTENTS List of Acronyms nd Abbrevitions... xv Executive Summry... xvi Study Interventions... xvi Rndom Assignment Design... xvii Dt Sources... xviii Previous Findings From the Fmily Options Study... xix Hypothesized Effects of the Interventions... xx Mening of Impct Comprisons... xxii Study Findings... xxii Conclusions... xxx Chpter 1. Introduction Bckground on the Homeless Services System The Continuum of Cre Progrmmtic Approches Evlution Design Bseline Chrcteristics of the Reserch Smple Orgniztion of the Report...10 Chpter 2. Implementing the Study Interventions Studied Site Selection Chrcteristics of Prticipting Sites Implementing Rndom Assignment...16 Chpter 3. Hypotheses About Intervention Effects (Conceptul Frmework) Chllenges Tht Fmilies Experiencing Homelessness Fce Chllenges Tht Homeless Mothers Fce Chllenges Tht Children Experiencing Homelessness Fce Heterogeneity Among Fmilies in Ptterns of Homelessness nd Use of Other Services Conceptul Frmework nd Hypothesized Effects for the Fmily Options Study Interventions Conceptul Rtionle for SUB nd CBRR Predictions Regrding Housing Stbility Predictions Regrding More Distl Outcomes Predictions Regrding Costs Hypotheses for Pirwise Comprisons Involving SUB, CBRR, nd UC Conceptul Rtionle for PBTH Predictions Regrding Housing Stbility Predictions Regrding Self-Sufficiency Predictions Regrding Adult Well-Being Predictions Regrding Distl Outcomes Predictions Regrding Costs Hypotheses for Pirwise Comprisons Wht Works for Whom? Predictions Regrding Wht Works for Whom...33 Chpter 4. Dt Sources nd Methodology Dt Sources Bseline Dt Collection Followup Dt Collection Construction of Progrm Usge Dt Methodology Impct Estimtion Model for Fmily nd Adult Outcomes Impct Estimtion Model for Child Well-Being Outcomes Impct Estimtion Model for Modertor Anlysis Strtegy for Addressing the Multiple Comprisons Problem...39 Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies v

7 Tble of Contents Chpter 5. Description of Usul Cre (UC) nd Outcomes Mesured in the Study The Emergency Shelter Experience of Usul Cre (UC) Fmilies UC Shelters in the Study Sites Housing Assistnce in UC Shelters Assessment of Fmilies in UC Shelters Supportive Services in UC Shelters Use of Other Homeless nd Housing Assistnce Progrms by Usul Cre (UC) Fmilies Outcomes for Fmilies Rndomly Assigned to Usul Cre (UC) Mesures of Housing Stbility Mesures of Fmily Preservtion Mesures of Adult Well-Being Mesures of Child Well-Being Mesures of Self-Sufficiency...58 Chpter 6. Impcts of Permnent Housing Subsidy (SUB) Compred With Usul Cre (UC) Permnent Housing Subsidy (SUB) Intervention Housing Assistnce in SUB Supportive Services in SUB Eligibility Criteri for SUB Progrm Use by Fmilies in the Permnent Housing Subsidy (SUB) Versus Usul Cre (UC) Comprison Impcts on Housing Stbility in the Permnent Housing Subsidy (SUB) Versus Usul Cre (UC) Comprison Impcts on Fmily Preservtion in the Permnent Housing Subsidy (SUB) Versus Usul Cre (UC) Comprison Impcts on Adult Well-Being in the Permnent Housing Subsidy (SUB) Versus Usul Cre (UC) Comprison Impcts on Child Well-Being in the Permnent Housing Subsidy (SUB) Versus Usul Cre (UC) Comprison Impcts on Self-Sufficiency in the Permnent Housing Subsidy (SUB) Versus Usul Cre (UC) Comprison Summry of the Permnent Housing Subsidy (SUB) Versus Usul Cre (UC) Comprison Across Domins...75 Chpter 7. Impcts of Community-Bsed Rpid Re- Housing (CBRR) Compred With Usul Cre (UC) Community-Bsed Rpid Re-Housing (CBRR) Intervention Housing Assistnce in CBRR Assessment of Fmily Needs in CBRR Supportive Services Provided in CBRR Eligibility Criteri for CBRR Progrm Use by Fmilies in the Community- Bsed Rpid Re-Housing (CBRR) Versus Usul Cre (UC) Comprison Impcts on Housing Stbility in the Community-Bsed Rpid Re-Housing (CBRR) Versus Usul Cre (UC) Comprison Impcts on Fmily Preservtion in the Community-Bsed Rpid Re-Housing (CBRR) Versus Usul Cre (UC) Comprison Impcts on Adult Well-Being in the Community-Bsed Rpid Re-Housing (CBRR) Versus Usul Cre (UC) Comprison Impcts on Child Well-Being in the Community-Bsed Rpid Re-Housing (CBRR) Versus Usul Cre (UC) Comprison Impcts on Self-Sufficiency in the Community-Bsed Rpid Re-Housing (CBRR) Versus Usul Cre (UC) Comprison Summry of the Community-Bsed Rpid Re-Housing (CBRR) Versus Usul Cre (UC) Comprison Across Domins...89 Chpter 8. Impcts of Project-Bsed Trnsitionl Housing (PBTH) Compred With Usul Cre (UC) Project-Bsed Trnsitionl Housing (PBTH) Intervention Housing Assistnce in PBTH Assessment of Fmily Needs in PBTH Supportive Services Provided in PBTH Eligibility Criteri for PBTH Progrm Rules in PBTH Progrm Use by Fmilies in the Project-Bsed Trnsitionl Housing (PBTH) Versus Usul Cre (UC) Comprison Impcts on Housing Stbility in the Project-Bsed Trnsitionl Housing (PBTH) Versus Usul Cre (UC) Comprison...97 Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies vi

8 Tble of Contents 8.4 Impcts on Fmily Preservtion in the Project-Bsed Trnsitionl Housing (PBTH) Versus Usul Cre (UC) Comprison Impcts on Adult Well-Being in the Project-Bsed Trnsitionl Housing (PBTH) Versus Usul Cre (UC) Comprison Impcts on Child Well-Being in the Project-Bsed Trnsitionl Housing (PBTH) Versus Usul Cre (UC) Comprison Impcts on Self-Sufficiency in the Project-Bsed Trnsitionl Housing (PBTH) Versus Usul Cre (UC) Comprison Summry of Project-Bsed Trnsitionl Housing (PBTH) Versus Usul Cre (UC) Comprison Across Domins Chpter 9. Impcts of Permnent Housing Subsidy (SUB) Compred With Community-Bsed Rpid Re-Housing (CBRR), SUB Compred With Project-Bsed Trnsi tionl Housing (PBTH), nd CBRR Compred With PBTH Anlysis Smples for Pirwise Comprisons The Permnent Housing Subsidy (SUB) Versus Community-Bsed Rpid Re-Housing (CBRR) Comprison Progrm Use by Fmilies in the SUB-Versus-CBRR Comprison Impcts of SUB Compred With CBRR The Permnent Housing Subsidy (SUB) Versus Project-Bsed Trnsitionl Housing (PBTH) Comprison Progrm Use by Fmilies in the SUB-Versus-PBTH Comprison Impcts of SUB Compred With PBTH The Community-Bsed Rpid Re-Housing (CBRR) Versus Project-Bsed Trnsitionl Housing (PBTH) Comprison Progrm Use by Fmilies in the CBRR-Versus-PBTH Comprison Impcts of CBRR Compred With PBTH Chpter 10. Impcts of Pooled Comprisons Chpter 11. Do Certin Interventions Work Better When Offered to Fmilies Who Fce Greter Difficulties? Descriptive Results Differentil Impcts Depending on Psychosocil Chllenges SUB Versus UC CBRR Versus UC PBTH Versus UC SUB Versus CBRR SUB Versus PBTH CBRR Versus PBTH Differentil Impcts Depending on Housing Brriers SUB Versus UC CBRR Versus UC PBTH Versus UC SUB Versus CBRR SUB Versus PBTH CBRR Versus PBTH Summry Chpter 12. Intervention Costs Introduction nd Summry Cost Dt Collection nd Anlysis Methodology Summry of Findings Cost of Permnent Housing Subsidy (SUB) Cost of Community-Bsed Rpid Re-Housing (CBRR) Cost of Project-Bsed Trnsitionl Housing (PBTH) Cost of Emergency Shelter Comprison of Costs Across Progrm Types Cost of All Progrm Use During the Followup Period by Fmilies in Ech Intervention Arm Monthly Cost of All Progrm Use t the Followup Survey by Fmilies in Ech Intervention Arm Chpter 13. Conclusions Mening of Impct Comprisons Usul Cre (UC) Permnent Housing Subsidy (SUB) Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies vii

9 Tble of Contents 13.4 Community-Bsed Rpid Re-Housing (CBRR) Project-Bsed Trnsitionl Housing (PBTH) Fmily Chllenges Implictions for Theory Questions for Longer Term Followup Summry References Additionl Reding Appendix A. Dt Sources nd Dtset Construction... A-1 Appendix B. Construction of Outcomes...B-1 Appendix C. Anlysis Methods... C-1 Appendix D. Anlysis of 18-Month Survey Nonresponse... D-1 Appendix E. Impcts on Use of Trnsitionl Housing...E-1 Appendix F. Impct Estimtes for Pooled Comprisons...F-1 Appendix G. Intervention Costs Methodology, Sites, nd Progrms... G-1 List of Exhibits Exhibit ES-1. Exhibit ES-2. Exhibit ES-3. Exhibit ES-4. Exhibit ES-5. Exhibit ES-6. Six Pirwise Comprisons Among the Four Interventions... xvii Totl Number of Fmilies Assigned to Ech Intervention nd Number of Followup Survey Respondents... xviii Dt Sources Used in the Anlysis of Short-Term Impcts... xviii Progrm Use Since Rndom Assignment for All Pirwise Impct Comprisons... xxiii Summry of Impcts for Six Policy Comprisons... xxv Averge Progrm Cost per Sty During the Followup Period Across Progrm Types... xxviii Exhibit 2-1. Exhibit 2-2. Exhibit 2-3. Exhibit 2-4. Exhibit 2-5. Exhibit 2-6. Exhibit 2-7. Exhibit 2-8. Intended Contrsts in Subsidy nd Services for the Fmily Options Interventions nd Usul Cre Group...11 Fmily Options Study Sites...12 Number of Progrms, by Site nd Intervention...13 Loction of Study Sites...14 Housing Mrket Chrcteristics of Study Sites...14 Homeless Popultion in Study Sites...15 Rndom Assignment Design...17 Interventions Avilble nd Prticipnt Enrollment by Assignment nd Site...17 Exhibit ES-7. Exhibit 1-1. Exhibit 1-2. Exhibit 1-3. Exhibit 1-4. Exhibit 1-5. Cost of Progrm Use Since Rndom Assignment for Ech Intervention Contrst... xxix Six Pirwise Comprisons Among the Experimentl Interventions...5 Fmily Chrcteristics: Fmily Composition...6 Fmily Chrcteristics: Housing Stbility nd History of Homelessness...7 Fmily Chrcteristics: Income Stbility nd Disbility...8 Fmily Chrcteristics: Brriers to Incresing Income or Finding Housing...9 Exhibit 3-1. Exhibit 3-2. Exhibit 3-3. Exhibit 4-1. Exhibit 4-2. Exhibit 5-1. Conceptul Intervention Model for SUB nd CBRR...22 Conceptul Intervention Model for PBTH...28 Hypotheticl Exmple in Which the Impct on the Outcome of Residentil Stbility of Intervention A Reltive to Intervention B Is Lrger for Fmilies With High Housing Brriers...32 Dt Sources Used in the Report...34 Progrm Types nd Their Dt Sources in the Progrm Usge Dt...37 Types of Living Spce Provided by UC Shelters...42 Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies viii

10 Tble of Contents Exhibit 5-2. Exhibit 5-3. Length of Time Spent in Emergency Shelters by UC Fmilies...43 UC Group Percent of Fmilies With t Lest 1-Night Sty in Emergency Shelter During Month, by Number of Months After RA...44 Exhibit 6-6. Exhibit 6-7. SUB Versus UC: Percent of Fmilies With t Lest 1 Night Sty in Emergency Shelter During Month, by Month After RA...70 SUB Versus UC: Impct on Length of Bseline Sty in Emergency Shelter...70 Exhibit 5-4. Fmily Rent Contributions nd Svings Requirements in UC Shelters...44 Exhibit 6-8. SUB Versus UC: Impcts on Fmily Preservtion...71 Exhibit 5-5. Exhibit 5-6. Exhibit 5-7. Types of Supportive Services Offered in UC Shelters nd How They Are Delivered, s Reported by Shelter Stff...45 Cse Mngement Intensity (rtio nd frequency)...46 Progrm Use Since Rndom Assignment for UC Group...47 Exhibit 6-9. Exhibit Exhibit SUB Versus UC: Impcts on Adult Well-Being...71 SUB Versus UC: Impcts on Child Well- Being Across Age Groups...72 SUB Versus UC: Impcts on Child Well- Being Developmentl Outcomes by Age Group...73 Exhibit 5-8. Fmily Options Study: Housing Stbility Outcomes...50 Exhibit SUB Versus UC: Impcts on Self-Sufficiency...74 Exhibit 5-9. Fmily Options Study: Fmily Preservtion Outcomes...51 Exhibit 7-1. Methods Used To Clculte CBRR Subsidy Amounts...78 Exhibit Exhibit Fmily Options Study: Adult Well- Being Outcomes...53 Fmily Options Study: Child Well- Being Outcomes Mesured for Children Across Age Groups...55 Exhibit 7-2. Exhibit 7-3. Types of Supportive Services Offered in CBRR Progrms nd How They Are Delivered...79 CBRR Cse Mngement Intensity (rtio nd frequency)...79 Exhibit Exhibit Fmily Options Study: Child Well- Being Developmentl Outcomes for Children in Specific Age Groups...57 Fmily Options Study: Self-Sufficiency Outcomes...60 Exhibit 7-4. Exhibit 7-5. CBRR Versus UC: Progrm Use Since Rndom Assignment...81 Number of Months of CBRR Receipt During Followup Period by CBRR Fmilies Who Ever Used CBRR...82 Exhibit 6-1. Exhibit 6-2. Exhibit 6-3. Exhibit 6-4. Exhibit 6-5. Subsidy Type Provided by Site...61 SUB Versus UC: Progrm Use Since Rndom Assignment...64 Number of Months of Subsidy Receipt During Followup Period by SUB Fmilies Who Ever Used SUB...66 SUB Versus UC: Impcts on Housing Stbility...67 SUB Versus UC: Percent of Fmilies With t Lest 1 Night Sty in Emergency Shelter During Month, by Month After RA...68 Exhibit 7-6. Exhibit 7-7. Exhibit 7-8. Exhibit 7-9. Exhibit CBRR Versus UC: Impcts on Housing Stbility...83 Housing Stbility Outcomes for the CBRR Rndom Assignment Group by Use of CBRR...84 CBRR Versus UC: Percent of Fmilies With t Lest 1 Night Sty in Emergency Shelter During Month, by Month After RA...85 CBRR Versus UC: Impct on Length of Bseline Sty in Emergency Shelter...85 CBRR Versus UC: Impcts on Fmily Preservtion...86 Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies ix

11 Tble of Contents Exhibit Exhibit Exhibit CBRR Versus UC: Impcts on Adult Well-Being...86 CBRR Versus UC: Impcts on Child Well-Being Across Age Groups...87 CBRR Versus UC: Impcts on Child Well-Being Developmentl Outcomes by Age Group...88 Exhibit Exhibit Exhibit 9-1. PBTH Versus UC: Impcts on Child Well-Being Developmentl Outcomes by Age Group PBTH Versus UC: Impcts on Self- Sufficiency Smple Sizes in the Six Pirwise Comprisons Exhibit CBRR Versus UC: Impcts on Self- Sufficiency...89 Exhibit 9-2. SUB Versus CBRR Progrm Use Since Rndom Assignment Exhibit 8-1. Exhibit 8-2. Exhibit 8-3. PBTH Housing Settings...91 Fmily Rent Contributions nd Svings Requirements in PBTH...91 Types of Supportive Services Offered in PBTH Progrms nd How They Are Delivered...92 Exhibit 9-3. Exhibit 9-4. Exhibit 9-5. SUB Versus CBRR: Impcts on Housing Stbility Impcts on Length of Bseline Sty in Emergency Shelter SUB Versus CBRR: Impcts on Fmily Preservtion Exhibit 8-4. PBTH Cse Mngement intensity (rtio nd frequency)...93 Exhibit 9-6. SUB Versus CBRR: Impcts on Adult Well-Being Exhibit 8-5. Exhibit 8-6. Exhibit 8-7. Types of Progrm Rules in PBTH...94 PBTH Versus UC: Progrm Use Since Rndom Assignment...95 Number of Months of Trnsitionl Housing Receipt During Followup Period by PBTH Fmilies Who Ever Used TH...97 Exhibit 9-7. Exhibit 9-8. Exhibit 9-9. SUB Versus CBRR: Impcts on Child Well-Being Across Age Groups SUB Versus CBRR: Impcts on Child Well-Being Developmentl Outcomes by Age Group SUB Versus CBRR: Impcts on Self- Sufficiency Exhibit 8-8. PBTH Versus UC: Impcts on Housing Stbility...98 Exhibit SUB Versus PBTH Progrm Use Since Rndom Assignment Exhibit 8-9. Exhibit Exhibit Exhibit Exhibit Exhibit PBTH Versus UC: Percent of Fmilies With t Lest 1 Night Sty in Emergency Shelter During Month, by Month After RA...99 Housing Stbility Outcomes for the PBTH Rndom Assignment Group by Use of TH PBTH Versus UC: Impct on Length of Bseline Sty in Emergency Shelter PBTH Versus UC: Impcts on Fmily Preservtion PBTH Versus UC: Impcts on Adult Well-Being PBTH Versus UC: Impcts on Child Well-Being Across Age Groups Exhibit Exhibit Exhibit Exhibit Exhibit Exhibit Exhibit SUB Versus PBTH: Impcts on Housing Stbility SUB Versus PBTH: Impcts on Fmily Preservtion SUB Versus PBTH: Impcts on Adult Well-Being SUB Versus PBTH: Impcts on Child Well-Being Across Age Groups SUB Versus PBTH: Impcts on Child Well-Being Developmentl Outcomes by Age Group SUB Versus PBTH: Impcts on Self- Sufficiency CBRR Versus PBTH Progrm Use Since Rndom Assignment Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies x

12 Tble of Contents Exhibit Exhibit Exhibit Exhibit Exhibit Exhibit Exhibit Exhibit Exhibit Exhibit Exhibit Exhibit Exhibit Exhibit Exhibit Exhibit Exhibit CBRR Versus PBTH: Impcts on Housing Stbility CBRR Versus PBTH: Impcts on Fmily Preservtion CBRR Versus PBTH: Impcts on Adult Well-Being CBRR Versus PBTH: Impcts on Child Well-Being Across Age Groups CBRR Versus PBTH: Impcts on Child Well-Being Developmentl Outcomes by Age Group CBRR Versus PBTH: Impcts on Self- Sufficiency Smple Sizes in the Four Pooled Comprisons Percentges of Adult Respondents Reporting Psychosocil Chllenges t the Time of Study Enrollment (for fmilies interviewed t 18 months) Percentge of Fmilies Reporting Tht Condition Ws Big or Smll Problem in Finding Plce to Live t the Time of Study Enrollment (for fmilies inter - viewed t 18 months) Impcts Moderted by Psychosocil Chllenges Index Impcts Moderted by Housing Brriers Index Progrms Included in the Cost Anlysis Averge Per-Fmily Monthly Cost of Supportive Services nd Housing or Shelter Across Progrm Types Summry Sttistics of Per-Fmily Monthly Progrm Cost by Progrm Type Averge Progrm Cost Per Sty During the Followup Period Across Progrm Types Summry of Cost of Progrm Use Since Rndom Assignment Summry of Monthly Cost of Progrm Use t the Time of the Followup Survey Exhibit Exhibit Exhibit SUB Progrm Cost Summry Sttistics for Fmilies Who Were Assigned To nd Took Up SUB Progrms Per-Fmily Monthly Progrm Costs for SUB Progrms SUB Per-Fmily Monthly Progrm Costs nd Fir Mrket Rents Exhibit Progrm-Level Cost Summry Sttistics for 12 CBRR Progrms Exhibit Per-Fmily Monthly Progrm Costs for CBRR Progrms Exhibit Progrm-Level Cost Summry Sttistics for 24 PBTH Progrms Exhibit Per-Fmily Monthly Progrm Costs for PBTH Progrms Exhibit Administrtive (pnel A) nd Prtner/In Kind (pnel B) Shre of Progrm Costs Exhibit Emergency Shelter Cost Summry Sttistics Exhibit Per-Fmily Monthly Progrm Costs for Emergency Shelter Progrms Exhibit Totl Monthly Costs nd Prtner/In-Kind Shre of Progrm Costs Exhibit Comprison of Cost Summry Sttistics Across Progrm Types Exhibit Cost of Progrm Use Since Rndom Assignment for Ech Intervention Contrst Exhibit Averge Per-Fmily Monthly Costs for Progrm Use t Time of the Followup Survey, by Comprison Exhibit A-1. Exhibit A-2. Exhibit A-3. Exhibit A-4. Smple Enrollment Period nd Number of Fmilies Enrolled by Intervention nd Site... A-1 Length of Time From Rndom Assignment to the 18-Month Followup Survey... A-2 Content of Prticipnt Dt Collected for Fmily Options Study 18-Month Impct Anlysis... A-4 Focl Child Smple Distribution by Site nd Intervention Group... A-7 Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies xi

13 Tble of Contents Exhibit A-5. Exhibit A-6. Exhibit A-7. Exhibit A-8. Exhibit A-9. Exhibit A-10. Exhibit B-1. Overll Fmily Options Study Survey Response Rtes... A-11 Survey Response Sttus for Fmily Options Study Surveys... A-12 HMIS Prticiption Rtes for Emergency Shelter nd Trnsitionl Housing Providers in the Study Sites, A-13 HMIS Mtch Rtes With the Fmily Options Smple, by Site... A-14 Smple Fmilies in PIC/TRACS Dt nd Those Assigned to the SUB Intervention in PIC/TRACS Dt, by Site... A-15 Dt Source Relibility, by Progrm Use Dt Item... A-16 Children s Completion Rtes for WJ III Letter-Word Identifiction nd Applied Problems Tests...B-13 Exhibit C-1. Confirmtory Hypothesis Tests... C-8 Exhibit D-1. Exhibit D-2. Exhibit D-3. Exhibit D-4. Exhibit D-5. Exhibit D-6. Exhibit D-7. Exhibit D-8. Exhibit D-9. Survey Nonresponse Incidence by Impct Comprison Adult Followup Survey... D-3 Chrcteristics Exmined in Bseline Equivlency Testing... D-4 Summry of Equivlence Testing in Impct Comprisons, Adult Survey... D-4 Equivlence t Bseline of Anlysis Smple for SUB Versus UC Impct Comprison, Adult Survey... D-5 Equivlence t Bseline of Anlysis Smple for CBRR Versus UC Impct Comprison, Adult Survey... D-6 Equivlence t Bseline of Anlysis Smple for PBTH Versus UC Impct Comprison, Adult Survey... D-7 Equivlence t Bseline of Anlysis Smple for SUB Versus CBRR Impct Comprison, Adult Survey... D-8 Equivlence t Bseline of Anlysis Smple for SUB Versus PBTH Impct Comprison, Adult Survey... D-9 Equivlence t Bseline of Anlysis Smple for CBRR Versus PBTH Impct Comprison, Adult Survey... D-10 Exhibit D-10. Equivlence t Bseline of Anlysis Smple for SUB + CBRR + PBTH Versus UC Impct Comprison, Adult Survey... D-11 Exhibit D-11. Equivlence t Bseline of Anlysis Smple for SUB + PBTH Versus CBRR Impct Comprison, Adult Survey... D-12 Exhibit D-12. Equivlence t Bseline of Anlysis Smple for SUB + CBRR Versus PBTH Impct Comprison, Adult Survey... D-13 Exhibit D-13. Equivlence t Bseline of Anlysis Smple for CBRR + PBTH Versus SUB Impct Comprison, Adult Survey... D-14 Exhibit D-14. Summry of Equivlence Testing of Respondents Versus Nonrespondents, by Assigned Intervention... D-15 Exhibit D-15. Summry of Equivlence Testing of Respondents Versus Nonrespondents, by Impct Comprison... D-16 Exhibit D-16. Equivlence t Bseline of Anlysis Smple for Respondents Versus Nonrespondents ssigned to SUB... D-17 Exhibit D-17. Equivlence t Bseline of Anlysis Smple for Respondents Versus Nonrespondents for CBRR... D-18 Exhibit D-18. Equivlence t Bseline of Anlysis Smple for Respondents Versus Nonrespondents for PBTH... D-19 Exhibit D-19. Equivlence t Bseline of Anlysis Smple for Respondents Versus Nonrespondents for UC... D-20 Exhibit D-20. Equivlence t Bseline of Anlysis Smple for Respondents Versus Nonrespondents for SUB Versus UC Impct Comprison, Adult Survey... D-21 Exhibit D-21. Equivlence t Bseline of Anlysis Smple for Respondents Versus Nonrespondents for CBRR Versus UC Impct Comprison, Adult Survey... D-22 Exhibit D-22. Equivlence t Bseline of Anlysis Smple for Respondents Versus Nonrespondents for PBTH Versus UC Impct Comprison, Adult Survey... D-23 Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies xii

14 Tble of Contents Exhibit D-23. Equivlence t Bseline of Anlysis Smple for Respondents Versus Nonrespondents for SUB Versus CBRR Impct Comprison, Adult Survey... D-24 Exhibit D-24. Equivlence t Bseline of Anlysis Smple for Respondents Versus Nonrespondents for SUB Versus PBTH Impct Comprison, Adult Survey... D-25 Exhibit D-25. Equivlence t Bseline of Anlysis Smple for Respondents Versus Nonrespondents for CBRR Versus PBTH Impct Comprison, Adult Survey... D-26 Exhibit D-26. Equivlence t Bseline of Anlysis Smple for Respondents Versus Nonrespondents for SUB + CBRR + PBTH Versus UC Impct Comprison, Adult Survey... D-27 Exhibit D-27. Equivlence t Bseline of Anlysis Smple for Respondents Versus Nonrespondents for SUB + PBTH Versus CBRR Impct Comprison, Adult Survey... D-28 Exhibit D-28. Equivlence t Bseline of Anlysis Smple for Respondents Versus Nonrespondents for SUB + CBRR Versus PBTH Impct Comprison, Adult Survey... D-29 Exhibit D-29. Equivlence t Bseline of Anlysis Smple for Respondents Versus Nonrespondents for PBTH + CBRR Versus SUB Impct Comprison, Adult Survey... D-30 Exhibit D-30. Executive Summry Impct Estimtes, Estimted Without Nonresponse Weights... D-32 Exhibit E-1. Exhibit E-2. Exhibit E-3. Exhibit E-4. Exhibit E-5. SUB Versus UC: Impcts on Use of ES nd TH...E-2 CBRR Versus UC: Impcts on Use of ES nd TH...E-2 PBTH Versus UC: Impcts on Use of ES nd TH...E-3 SUB Versus CBRR: Impcts on Use of ES nd TH...E-3 SUB Versus PBTH: Impcts on Use of ES nd TH...E-4 Exhibit E-6. Exhibit F-1. Exhibit F-2. Exhibit F-3. Exhibit F-4. Exhibit F-5. Exhibit F-6. Exhibit F-7. Exhibit F-8. Exhibit F-9. Exhibit F-10. Exhibit F-11. Exhibit F-12. Exhibit F-13. Exhibit F-14. Exhibit F-15. Exhibit F-16. CBRR Versus PBTH: Impcts on Use of ES nd TH...E-4 SUB + CBRR + PBTH Versus UC: Impcts on Housing Stbility...F-2 SUB + CBRR + PBTH Versus UC: Impcts on Fmily Preservtion...F-3 SUB + CBRR + PBTH Versus UC: Impcts on Adult Well-Being...F-3 SUB + CBRR + PBTH Versus UC: Impcts on Child Well-Being Across Age Groups...F-4 SUB + CBRR + PBTH Versus UC: Impcts on Child Well-Being Developmentl Outcomes by Age Group...F-5 SUB + CBRR + PBTH Versus UC: Impcts on Self-Sufficiency...F-6 SUB + CBRR Versus PBTH: Impcts on Housing Stbility...F-7 SUB + CBRR Versus PBTH: Impcts on Fmily Preservtion...F-8 SUB + CBRR Versus PBTH: Impcts on Adult Well-Being...F-8 SUB + CBRR Versus PBTH: Impcts on Child Well-Being Across Age Groups...F-9 SUB + CBRR Versus PBTH: Impcts on Child Well-Being Developmentl Outcomes by Age Group...F-10 SUB + CBRR Versus PBTH: Impcts on Self- Sufficiency...F-11 SUB + PBTH Versus CBRR: Impcts on Housing Stbility...F-12 SUB + PBTH Versus CBRR: Impcts on Fmily Preservtion...F-13 SUB + PBTH Versus CBRR: Impcts on Adult Well-Being...F-13 SUB + PBTH Versus CBRR: Impcts on Child Well-Being Across Age Groups...F-14 Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies xiii

15 Tble of Contents Exhibit F-17. SUB + PBTH Versus CBRR: Impcts on Child Well-Being Developmentl Outcomes by Age Group...F-15 Exhibit F-23. CBRR + PBTH Versus SUB: Impcts on Child Well-Being Developmentl Outcomes by Age Group...F-20 Exhibit F-18. SUB + PBTH Versus CBRR: Impcts on Self- Sufficiency...F-16 Exhibit F-24. CBRR + PBTH Versus SUB: Impcts on Self-Sufficiency...F-21 Exhibit F-19. Exhibit F-20. Exhibit F-21. Exhibit F-22. CBRR + PBTH Versus SUB: Impcts on Housing Stbility...F-17 CBRR + PBTH Versus SUB: Impcts on Fmily Preservtion...F-18 CBRR + PBTH Versus SUB: Impcts on Adult Well-Being...F-18 CBRR + PBTH Versus SUB: Impcts on Child Well-Being Across Age Groups...F-19 Exhibit G-1. Exhibit G-2. Exhibit G-3. Number of Progrms From Which Cost Dt Ws Collected nd Presence of SUB Intervention by Study Site... G-2 Cost Dt Collection Ctegories nd Associted Item Prompts... G-3 Determining Inclusion of Externl Services... G-4 Exhibit G-4. Vluing In-Kind Services... G-5 Exhibit G-5. Progrm Level Averge Per Fmily Cost Clcultions... G-7 Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies xiv

16 LIST OF ACRONYMS AND ABBREVIATIONS AHAR AMI ARRA ASQ-3 CBRR CDC CoC Annul Homeless Assessment Report Are Medin Income Americn Recovery nd Reinvestment Act Ages nd Stges Questionnire community-bsed rpid re-housing intervention Centers for Disese Control nd Prevention Continuum of Cre DAST-10 Drug Abuse Screening Test FDS FWER GED HCV HMIS HOME HPRP HTKS ITT NCS-R Finncil Dt Schedule fmilywise error rte generl eductionl development housing choice voucher Homeless Mngement Informtion System Home Observtion for Mesurement of the Environment (inventory) Homelessness Prevention nd Rpid Re-Housing Progrm Hed Toes Knees Shoulders (tsk) intention to tret Ntionl Comorbidity Survey Repliction NSHAPC Ntionl Survey of Homeless Assistnce Providers nd Clients PBTH project-bsed trnsitionl housing intervention PDS PHA PIC PSH PTSD RAPS4 SCHIP SDQ SHIFT SHP SNAP SSDI SSI STAIC SUB TANF TRACS UC WIC WJ III Posttrumtic Stress Dignostic Scle public housing gency Public nd Indin Housing Informtion Center permnent supportive housing post-trumtic stress disorder Rpid Alcohol Problems Screen Stte Children s Helth Insurnce Progrm Strengths nd Difficulties Questionnire Service nd Housing Interventions for Fmilies in Trnsition Supportive Housing Progrm Supplementl Nutrition Assistnce Progrm Socil Security Disbility Insurnce Supplementl Security Income Stte-Trit Anxiety Inventory for Children permnent housing subsidy intervention Temporry Assistnce for Needy Fmilies Tennt Rentl Assistnce Certifiction System usul cre Specil Supplementl Nutrition Progrm for Women, Infnts, nd Children Woodcock-Johnson III Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies xv

17 EXECUTIVE SUMMARY The U.S. Deprtment of Housing nd Urbn Develop - ment (HUD) undertook the Fmily Options Study to gther evidence bout which types of housing nd services interventions work best for homeless fmilies. The study compres the effects of three ctive interventions permnent housing subsidy (SUB), community-bsed rpid re-housing (CBRR), nd project-bsed trnsitionl housing (PBTH) to one nother nd to the usul cre (UC) vilble to homeless fmilies. SUB, CBRR, nd PBTH re distinguished from one nother by the durtion of housing ssistnce provided nd the type nd intensity of socil services offered. UC consists of emergency shelter nd housing or services tht fmilies ccess without immedite referrl to one of the three ctive interventions. From September 2010 through Jnury 2012, 2,282 fmilies enrolled in the Fmily Options Study cross 12 communities 1 fter spending t lest 7 dys in emergency shelter. After providing informed consent nd completing bseline survey, the fmilies were rndomly ssigned to one of the three ctive interventions or to UC. Rndom ssignment yielded groups of fmilies with no systemtic differences in bseline chrcteristics. Fmilies were free to tke up their ssigned interventions or to mke other rrngements, so fmilies used mix of progrms, often including progrms other thn the type to which they were ssigned. Nonetheless, ptterns of progrm use mong the groups of fmilies contrsted sub - stntilly, nd the study provides strong bsis for conclusions bout the reltive impcts of the interventions on severl spects of fmily well-being. The Fmily Options Study will follow the full set of 2,282 study fmilies for 36 months. The study tem conducted short trcking surveys with the fmilies 6, 12, nd 27 months fter rndom ssignment. The study tem lso conducted more extensive followup survey pproximtely 20 months fter rndom ssignment to collect detiled informtion bout fmily outcomes. Another followup survey will be conducted pproximtely 36 months fter rndom ssignment. The first of the extensive followup surveys chieved response rte of 81 percent, with 1,857 fmilies responding to the survey. 2 This report presents the short-term impcts of the interventions in five domins relted to fmily well-being: (1) housing stbility, (2) fmily preservtion, (3) dult well-being, (4) child well-being, nd (5) self-sufficiency. The report lso describes the reltive costs of the interventions bsed on progrm use during the first followup period. A subsequent report in 2016 will present impcts on study fmilies 36 months fter rndom ssignment long with intervention costs over the longer period. Study Interventions The study exmines four interventions: 1. Permnent housing subsidy, or SUB, usully housing choice voucher (HCV), could include ssistnce to find housing but no other supportive services. 2. Community-bsed rpid re-housing, or CBRR, provides temporry rentl ssistnce, potentilly renewble for up to 18 months, pired with limited, housing-focused services to help fmilies find nd rent conventionl, privte-mrket housing. 3. Project-bsed trnsitionl housing, or PBTH, provides temporry housing for up to 24 months in gency-controlled buildings or prtment units, pired with intensive supportive services. 4. Usul cre, or UC, is defined s ny housing or services tht fmily ccesses in the bsence of immedite referrl to the other interventions. This intervention typiclly includes t lest some dditionl sty in the emergency shelter from which fmilies were enrolled. The study tem nlyzed ll six possible contrsts mong these four interventions, s shown in Exhibit ES-1. The 1 The 12 communities prticipting in the study re Almed County, Cliforni; Atlnt, Georgi; Bltimore, Mrylnd; Boston, Msschusetts; Bridgeport nd New Hven, Connecticut; Denver, Colordo; Honolulu, Hwii; Knss City, Missouri; Louisville, Kentucky; Minnepolis, Minnesot; Phoenix, Arizon; nd Slt Lke City, Uth. 2 This report nlyzes short-term impcts of the interventions. The study tem ttempted to contct fmilies for the study s first followup survey beginning in the 18th month fter rndom ssignment. The medin time from rndom ssignment to the followup survey ws 20 months. Anlysis of progrm use nd cost of totl progrm use used dt over medin of 21 clendr months. Dt collection for the second followup survey ws completed in erly 2015 nd chieved 79-percent response rte. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies xvi

18 Executive Summry Exhibit ES-1. Six Pirwise Comprisons Among the Four Interventions order of the presenttion of findings for the vrious pirwise comprisons is reflected in the lphbetic ordering of the rrows (for exmple, discussion begins with Contrst A between SUB nd UC). Rndom Assignment Design To be eligible for the study, fmilies hd to include t lest one child ge 15 or younger nd hd to hve resided in emer - gency shelter for 7 or more dys. The study tem excluded fmilies who left shelter in fewer thn 7 dys becuse the interventions exmined my not be necessry for fmilies who cn resolve housing crisis quickly. As soon s ws fesible fter the 7-dy mrk, the evlution tem rndomly ssigned fmilies to SUB, CBRR, PBTH, or UC. Implementing the rndom ssignment design presented severl chllenges. In the originl design, ech fmily ws to hve hd chnce of being ssigned to ll four groups (SUB, CBRR, PBTH, or UC). A number of fctors prevented the study from being implemented exctly s plnned. First, 3 of the 12 sites were ble to provide only two of the three ctive interventions. Second, the rndom ssignment groups vilble to fmilies were confined to groups for which provider hd n vilble slot t the time of rndomiztion. Third, some service providers hd unique eligibility require - ments for fmilies. Before rndom ssignment, the study tem screened fmilies ginst the eligibility criteri of providers tht hd vilble slots. The purpose of this screening ws to minimize the likelihood of ssigning fmilies to interventions they would not be eligible to receive. As result, for n intervention option to be vilble to fmily undergoing rndom ssignment, t lest one slot needed to be vilble t n intervention provider for which the fmily ppered to meet provider-specific eligibility requirements bsed on preliminry screening. These fctors cumultively resulted in most study fmilies not hving ll four options vilble to them t rndom ssignment. Of the 2,282 fmilies enrolled in the study, 474 hd ll four rndomiztion options vilble, 1,544 fmilies hd three rndomiztion options, nd 264 fmilies hd two rn - domiztion options. All nlyses were conducted pirwise, contrsting n ctive intervention to nother ctive intervention or to UC. Only fmilies who were eligible for both interventions in pirwise comprison (for exmple, SUB Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies xvii

19 Executive Summry nd CBRR) nd rndomized to one of them were included in ech comprison. Hence, ech comprison cn be thought of s two-wy experiment between two well-mtched groups tht differ only in the intervention to which they were ssigned. Exhibit ES-2 shows the totl number of fmilies ssigned to ech intervention. The exhibit lso shows the number of fmilies who responded to the followup survey conducted pproximtely 20 months fter rndom ssignment; this set of fmilies is included in the impct nlyses in this report. Dt Sources The impct findings reported here bout the 1,857 fmilies re bsed on dt from severl sources described in Exhibit ES-3. Exhibit ES-2. Totl Number of Fmilies Assigned to Ech Intervention nd Number of Followup Survey Respondents Intervention Fmilies Assigned Fmilies Responding to the Followup Survey Response Rte (%) Permnent housing subsidy (SUB) Community-bsed rpid re-housing (CBRR) Project-bsed trnsitionl housing (PBTH) Usul cre (UC) Totl 2,282 1, Sources: Rndom ssignment records; Fmily Options Study 18-month followup survey Exhibit ES-3. Dt Sources Used in the Anlysis of Short-Term Impcts Study implementtion records Rndom ssignment enrollment dt Study fmilies Bseline survey Trcking surveys 18-month followup survey Child ssessments Child survey Study intervention providers Enrollment verifiction dt Progrm informtion Cost informtion Administrtive dt systems HMIS records HUD s PIC records HUD s TRACS records Rndom ssignment enrollment dt contin identifiers for enrolled fmilies, responses to eligibility screening questions, informtion bout intervention vilbility t the time of rndom ssignment, nd rndom ssignment result. The bseline survey conducted immeditely before rndom ssignment provides informtion bout the dult respondent nd the fmily. Trcking surveys conducted 6 nd 12 months fter rndom ssignment contin updted contct informtion nd detils bout fmily composition nd housing sttus. The 18-month followup survey conducted with dult respondents t medin durtion of 20 months mesures fmily outcomes. Adults reported on themselves nd up to two children, clled focl children, who were prt of the fmily t the time of study enrollment. Focl children were rndomly selected within specified ge groups. Child ssessments, which were conducted with focl children ges 3 yers, 6 months to 7 yers, 11 months in conjunction with the dult followup survey, mesure child well-being outcomes. The child survey conducted with focl children ges 8 to 17 yers in conjunction with the dult followup survey mesures child well-being outcomes. Enrollment verifiction dt collected from study providers mesure use of the ssigned intervention for ech fmily. Progrm informtion bout the housing nd services offered during the study period collected from intervention providers describes the interventions. Cost informtion collected from intervention providers mesures costs of overhed, rentl ssistnce, fcility opertions, supportive services, nd cpitl costs. Homeless Mngement Informtion System (HMIS) records provide indictors of study fmilies prticiption in homeless ssistnce progrms. HUD s Public nd Indin Housing Informtion Center (PIC) records mesure receipt of housing ssistnce from the Housing Choice Voucher progrm, public housing progrms, nd project-bsed voucher progrms. HUD s Tennt Rentl Assistnce Certifiction System (TRACS) records mesure receipt of housing ssistnce through project-bsed Section 8 progrms. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies xviii

20 Executive Summry Previous Findings From the Fmily Options Study A previous study report (Gubits et l., 2013) provides infor - mtion bout the bseline chrcteristics of the study smple nd insights regrding the homeless ssistnce system in the study communities. Bseline Chrcteristics of the Study Smple The bseline survey collected informtion bout ll 2,282 fmilies enrolled in the Fmily Options Study. 3 Fmily Composition The typicl fmily in the study consisted of n dult womn, 29 yers old, living with one or two of her children in n emergency shelter. At bseline, 30 percent of fmilies hd more thn one dult present. Nerly ll fmilies with two dults present were heded by couples. A plurlity of fmilies (43 percent) hd only one child younger thn ge 18 with them in the shelter, nother 30 percent hd two children present, nd 27 percent hd three or more child - ren present. One-hlf of the fmilies hd child younger thn ge 3 in the shelter, nd 10 percent of dult respondents reported tht they were pregnnt. Housing Stbility nd History of Homelessness Most fmilies in the study (79 percent) were not homeless immeditely before entering the shelter from which they were recruited into the study. About 63 percent of fmily heds in the study hd experienced homelessness t some other point in their lifetime, with 16 percent of dult respond - ents hving experienced homelessness s child. An even greter proportion (85 percent) indicted tht they lived doubled up t some point s n dult, defined in the survey s stying with fmily or friends becuse you couldn t find or fford plce of your own. Employment nd Other Sources of Income The employment, income, nd progrm prticiption of fmilies t bseline provide insight into the severity of income brriers tht fmilies fce in emergency shelters. Most fmily heds were not working t the time of rndom ssignment (83 percent), nd more thn one-hlf hd not worked for py in the previous 6 months. The medin nnul household income of ll fmilies in the study t bseline ws $7,410. Most fmilies in the study received some form of public ssistnce t the time of rndom ssignment. Most (88 percent) received ssistnce from the Supplementl Nutrition Assistnce Progrm, or SNAP, nd 41 percent received ssistnce from Temporry Assistnce for Needy Fmilies, or TANF. Most fmilies in the study (86 percent) reported receiving some combintion of Medicid benefits, stte helth insurnce benefits, nd the Stte Children s Helth Insurnce Progrm, or SCHIP. Other Brriers to Finding Housing or Incresing Income The bseline survey sked fmilies explicitly bout fctors tht would ffect their bility to find plce to live. Mny reported tht they either hd poor rentl history (26 percent hd been evicted) or hd never been leseholder (35 percent). In 11 percent of fmilies, the fmily hed hd previous felony conviction. Approximtely 22 percent of dult respondents gve survey responses tht indicted symptoms of post-trumtic stress disorder, or PTSD; 22 percent reported symptoms of serious psychologicl distress; nd 30 percent reported evidence of one or the other. A history of drug use in the yer before rndom ssignment ws indicted by the survey responses of 14 percent of the dult respondents; survey responses lso suggested lcohol buse within the pst yer for 11 percent of respondents. Intervention Eligibility Screening nd Fmily Decisions Gubits et l. (2013) lso exmined intervention vilbility nd fmily eligibility t rndomiztion for the 2,282 fmilies in the Fmily Options Study. Both vilbility of interventions nd fmily eligibility, ccording to screening before rndom ssignment, were most constrined for PBTH. CBRR ws more vilble thn SUB but hd slightly more restrictive eligibility requirements. All fmilies were eligible for UC by definition. For fmily to use the progrm to which it hd been ssigned, it hd to (1) pss n eligibility determintion conducted by the progrm to which it ws ssigned nd (2) choose to tke up the progrm. Gubits et l. (2013) found tht some of the fmilies who pssed the initil screening by the study were lter deemed ineligible by the progrms to which they were ssigned. Other fully eligible fmilies chose not to tke up the ssigned progrm. Compred with CBRR nd SUB, PBTH hd both the highest proportion of fmilies found ineligible 3 Gubits et l. (2013) compre the chrcteristics of Fmily Options Study fmilies with ntionl estimtes for homeless fmilies from HUD s 2010 Annul Homeless Assessment Report nd the 1996 Ntionl Survey of Homeless Assistnce Providers nd Clients. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies xix

21 Executive Summry fter rndom ssignment nd the highest proportion of fmilies who chose not to tke up the ssigned progrm. Considering both initil screening by the study nd lter eligibility screen - ing by progrms, SUB ws the most ccessible to fmilies nd PBTH ws the lest ccessible to fmilies. Gubits et l. (2013) concluded tht homeless ssistnce progrms in the study communities imposed eligibility criteri tht hmpered their bility to serve fmilies in shelter who needed the ssistnce. Even when progrms hd spce vil - ble, the progrms often screened out fmilies in shelter bsed on eligibility criteri such s insufficient income, substnce buse, criminl histories, nd other fctors tht presumbly contributed to the fmilies homelessness. Moreover, fmilies who re homeless do not lwys pursue the progrms offered to them, which suggests tht some progrms deliver ssistnce tht some fmilies perceive s less vluble to them thn other ssistnce vilble in their communities. Hypothesized Effects of the Interventions The study tem developed hypotheses bout the potentil effects of the interventions bsed on the conceptul frmework underlying the SUB, CBRR, nd PBTH interventions. The interventions reflect different implicit theories bout the nture of fmily homelessness nd the pproches best suited to ddress the problem. These implicit theories rise from different understndings of the origins of homelessness, the needs of homeless fmilies, the effect of fmily chllenges on chieving residentil stbility, nd the pproprite role of the homeless ssistnce system. Some theories posit tht household chllenges for exmple, trum, substnce use problems, mentl helth issues, lck of job skills must be ddressed first for fmilies to succeed in housing. Others posit tht progress on these issues is likely to be chieved only fter fmilies re stbilized in permnent housing. The different perceptions of the homeless ssistnce system s role result in different emphses mong three centrl gols of interventions for homeless fmilies: (1) ending the immedite episode of homelessness nd preventing returns to shelter; (2) fostering longer term residentil stbility; nd (3) promot - ing other outcomes, including self-sufficiency, fmily preservtion, nd dult nd child well-being. Conceptul Frmework for SUB nd CBRR It is pproprite to consider the conceptul rtionles for SUB nd CBRR together becuse proponents of both SUB nd CBRR believe tht the key gol of homeless interventions should be ending homelessness swiftly, reducing returns to shelter, nd restoring fmilies to housing stbility. This position follows from their view tht fmily homelessness is lrgely consequence of housing costs tht outstrip the incomes of poor fmilies, problem tht housing subsidies cn solve. Subsidies whether the permnent subsidies of the SUB intervention or temporry subsidies such s CBRR cn help fmilies obtin nd mintin stble housing. SUB ws not creted s response to homelessness. Insted, SUB lredy existed s n element of the broder socil sfety net t the time the homeless service system cme into being in the lte 1980s. Resource constrints men tht, outside the context of this study, SUB is rrely ccessible by fmilies t the outset of n episode of homelessness unless they lredy hve plce ner the top of witing list. By contrst, CBRR ws developed specificlly s response to homelessness. Becuse SUB is unlikely to become widely vilble to fmilies t the time they re experiencing homelessness, proponents of CBRR rgue tht limited resources dedicted to homelessness could be stretched to crete the best outcomes for the most people by mking subsidies temporry (Culhne, Metrux, nd Byrne, 2011). Proponents of CBRR emphsize restoring fmilies to conven - tionl housing s swiftly s possible (the rpid in rpid re-housing), thereby reducing time in shelter nd on the street, which they see s hrmful. In ddition, they focus on preventing returns to homelessness. Proponents of SUB focus more on long-term stbility nd question whether the short-term subsidies provided by CBRR re sufficient to foster such stbility. Proponents of CBRR rgue tht temporry subsidy my induce fmilies to strive to become economiclly self-sufficient sooner. Advoctes of both types of subsidies cknowledge tht home - less fmilies, like other poor fmilies, must contend with vriety of chllenges, but these dvoctes believe tht such chllenges re better ddressed by minstrem community gencies rther thn by specilized homeless services. Proponents of both types of subsidies rgue tht stble housing provides pltform from which fmilies cn ddress other problems on their own using community resources, if they need to nd choose to do so, while reserving scrce housing dollrs for housing. The stbility provided by either shortterm or permnent housing subsidy my hve rditing effects on other spects of fmily well-being. The study tem developed four hypotheses for comprisons involving SUB, CBRR, nd UC tht derive from this conceptul frmework. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies xx

22 Executive Summry Hypotheses for Comprisons Involving SUB, CBRR, nd UC SUB Versus UC Reltive to UC, SUB will reduce shelter use nd improve housing stbility nd my improve fmily preservtion, dult well-being, nd child well-being. CBRR Versus UC Reltive to UC, CBRR will reduce shelter use nd my improve housing stbility, employment nd ernings, fmily preservtion, dult well-being, nd child well-being. It will reduce the length of the shelter sty t the time of study entry nd my be less costly. SUB Versus CBRR Reltive to CBRR, SUB will reduce shelter use nd improve housing stbility nd my improve dult nd child well-being. Reltive to SUB, CBRR will reduce the length of the shelter sty t the time of study entry nd will be less costly. It my improve employment nd ernings. Conceptul Frmework for PBTH Proponents of PBTH hve different understnding of the origins of fmily homelessness nd the pproprite role of the homeless service system thn do proponents of SUB nd CBRR. Although the housing mrket is difficult for poor fmilies, most fmilies do not experience homelessness. Proponents of PBTH emphsize tht mny fmilies who do become homeless hve brriers in ddition to poverty tht mke it hrd for them to secure nd mintin housing. Thus, housing subsidies lone my be insufficient to ensure housing stbility nd other desirble outcomes, prticulrly for fmilies who hve been in shelter for t lest 7 dys (for exmple, Bssuk nd Geller, 2006). Fmily needs my rise from poverty, helth, disbility, or other problems tht led to homelessness to begin with or from the disruptive effects of homelessness on prents nd children. Proponents of PBTH believe tht by ddressing these brriers nd needs in supervised residentil setting, PBTH lys the best foundtion for ongoing stbility. Bsing their work on fmily needs, cse mngers coordinte the services (on site or by referrl) to ly the essentil groundwork for lter independence. Different PBTH progrms focus on different issues, but ll provide supportive services designed to reduce brriers to housing, enhnce prents well-being, nd bolster their bility to mnge in ordinry housing fter they leve progrms (Burt, 2010). Prctitioners gols for PBTH, s documented in the literture (for exmple, Burt, 2006), thus extend beyond housing stbility to dult well-being nd spects of fmily self-sufficiency. Although some PBTH progrms provide services directly to children, fmily preservtion nd child outcomes re usully seen s more distl outcomes. Given this conceptul frmework for PBTH, the study tem defined five hypotheses bout the potentil effects of PBTH when compred with UC, SUB, nd CBRR. Hypotheses for Comprisons Involving PBTH PBTH Versus UC Reltive to UC, PBTH will reduce shelter use nd improve housing stbility, employment, ernings, eduction, nd dult well-being nd my improve fmily preservtion nd child well-being. PBTH Versus SUB (From the perspective of PBTH proponents), reltive to SUB, PBTH will improve employment, ernings, eduction, nd dult well-being nd my improve long-term housing stbility, fmily preservtion, nd child well-being. (Stbility effects my not emerge t 18 months.) (From the perspective of SUB proponents) reltive to PBTH, SUB will reduce shelter use nd improve housing stbility nd my improve fmily preservtion, dult well-being, nd child well-being. PBTH Versus CBRR (From the perspective of PBTH proponents) reltive to CBRR, PBTH will improve employment, ernings, eduction, nd dult well-being nd my improve long-term stbility, fmily preservtion, nd child well-being. (Stbility effects my not emerge t 18 months.) (From the perspective of CBRR proponents) reltive to PBTH, CBRR will reduce shelter use nd my improve housing stbil - ity, fmily preservtion, dult well-being, child well-being, employment, nd ernings. It will reduce the length of time fmilies spend in plces not ment for humn hbittion nd in shelters, which re costly. Even if the longer housing subsidies of SUB or the more extensive socil services of PBTH re importnt for some fmilies, n importnt question is whether ll fmilies need such intensive involvement in the homelessness ssistnce system. Thus the study tem lso developed hypotheses tht the more intensive interventions would hve lrger effects on outcomes for fmilies who fced more housing brriers nd greter psychosocil chllenges. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies xxi

23 Executive Summry Mening of Impct Comprisons The inherent strength of the experimentl reserch design employed in the Fmily Options Study is the ssurnce tht the groups tht re creted through the rndom ssignment process will be similr to ech other. Becuse it is not possible to ccount for, or to use sttisticl methods to control for, ll the vribility tht my exist mong individul fmilies, rndomly ssigning lrge number of fmilies to different interventions is the most certin wy to ensure tht the groups will be comprble. The Fmily Options Study tests for the impcts of three different potentil emphses in federl or locl ssistnce policy to homeless fmilies; tht is, Wht impct would priority ccess to project-bsed trnsitionl housing (the PBTH rm of the experiment) hve on fmilies in shelter who re not ble to resolve their episodes of homelessness quickly? How does this policy compre with providing ccess to community-bsed rpid re-housing (the CBRR rm) nd to permnent housing subsidies (the SUB rm)? In ech cse, the corresponding policy question is, Wht impct would this policy emphsis hve on the outcomes of fmilies in shelter reltive to UC or nother policy emphsis? The followup dt for study prticipnts tell us wht would hppen when ech of these wys of trgeting offers nd ccess were pursued s federl or locl policy. The pirwise comprisons between ctive interventions show the impct of offering fmilies priority ccess to one intervention rther thn nother. The dt lso llow for the comprison of ech option with current policies tht do not crete priority ccess to ny prticulr form of housing ssistnce (tht is, the UC rm). The pirwise comprisons between ctive intervention rms nd UC show the impct of referring fmily to specific type of progrm compred with the impct of letting fmilies pursue ssistnce on their own. The nlysis in this report mesures the impct of hving been offered prticulr intervention regrdless of whether or not the fmily involved ctully received the intervention. The findings reflect the rel wy in which the homeless ssistnce system intercts with fmilies, in tht fmilies re offered n intervention rther thn mndted to ccept the ssistnce being offered. Whether fmilies prticipte in n ssigned progrm reflects the reltive desirbility nd ccessibility of the interventions for fmilies within the context of the other options they my choose to pursue on their own. As the report shows, substntil number of fmilies did not use the ctive intervention to which they were referred, nd some used other interventions. The full experimentl smple for given rm collectively shows how different forms of housing ssistnce re used when fmilies re given priority ccess to one prticulr progrm type while simultneously hving the freedom to use other forms of ssistnce vilble in their communities. Including ll the fmilies rndomly ssigned to UC similrly revels the rnge of progrms used when no priority ccess is provided. The progrms (including the interventions exmined in this study) tht UC fmilies used exist in communities nd would ech continue to exist even with stronger federl or locl push for only one of them. Thus, the full-smple comprisons between rndom ssignment rms known s intention to tret, or ITT, impct estimtes provide the best guide to policymkers in messy, complex world nd re reported here s the min study findings. 4 Study Findings Progrm Use During the Followup Period To ssess the impct of offering priority ccess to the inter - ventions, ech ctive intervention (SUB, CBRR, nd PBTH) ws compred with UC nd with ech of the other interventions (SUB versus CBRR, SUB versus PBTH, nd CBRR versus PBTH), resulting in six pirwise comprisons. This structure of reporting impct estimtes in ech of the six pirwise comprisons is used throughout the report. Exhibit ES-4 documents the progrm use of the 1,857 study fmilies who responded to the 18-month followup survey the smple for the impct nlysis. The exhibit shows the percentges of fmilies who ever prticipted in severl types of housing ssistnce progrms between rndom ssign - ment nd the followup survey response. The columns in Exhibit ES-4 re orgnized by pirwise comprison. The exhibit displys the number of fmilies included in ech 4 Policymkers my lso wnt to know the impct of prticulr type of homeless intervention on only the fmilies who prticipte in the intervention s informtion importnt to individul fmilies nd in guiding progrm improvement. The study is considering investigting such questions concerning the SUB nd CBRR interventions in two future ppers, subject to sttisticl limittions in isolting the direct effects of prticiption in the experimentl dt. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies xxii

24 Executive Summry Exhibit ES-4. Progrm Use Since Rndom Assignment for All Pirwise Impct Comprisons Type of Housing Assistnce Percent of Fmilies Who Ever Used Progrm Type From RA to 18-Month Followup Survey SUB UC CBRR UC PBTH UC SUB CBRR SUB PBTH CBRR PBTH Permnent housing subsidy (SUB) b Community-bsed rpid re-housing (CBRR) Trnsitionl housing c Permnent supportive housing Public housing Project-bsed vouchers/section 8 projects No use of homeless or housing progrms d N PBTH = project-bsed trnsitionl housing. UC = usul cre. RA = rndom ssignment. Notes: Percentges re regression djusted, controlling for site nd rndomiztion rtio. Percentges, mens, nd medins re weighted for survey nonresponse to represent full comprison smple. Percent of fmilies who ever used type of ssistnce progrm during the period from the month of RA to the month of 18-month followup survey response (medin period durtion: 21 months). Percentges do not dd to 100 percent becuse some fmilies use more thn one progrm type during the followup period. b Subsidy ssistnce is housing choice vouchers plus site-specific progrms offered to fmilies ssigned to SUB group in Connecticut nd Honolulu. c Trnsitionl housing includes both PBTH nd other forms of trnsitionl housing. d No use of homeless or housing progrms (ever used) indictes no use of the six progrm types in this exhibit during ny of the followup period nd no use of emergency shelter fter the first 6 months fter RA. No use in the month of the followup survey response indictes no use of ny of these seven progrm types. Source: Fmily Options Study Progrm Usge Dt comprison (see row 8) nd detils their progrm use during the followup period (see rows 1 to 7). It ccounts for six types of progrms: 1. Subsidy (tht is, the progrms tht comprise the SUB intervention in this study: HCVs, public housing in Honolulu, nd project-bsed vouchers in Bridgeport). 2. Rpid re-housing (tht is, CBRR). 3. Trnsitionl housing Permnent supportive housing. 5. Public housing in plces other thn Honolulu. 6. Project-bsed vouchers nd units in Section 8 projects. The experimentl contrsts in use of these progrms re depicted in the exhibit. Exhibit ES-4 shows tht the intervention ssignments creted substntil contrsts between groups, prticulrly in their use of progrms tht reflect the intended contrst (the shded boxes). For exmple, in the SUB-versus-UC comprison, 84 percent of fmilies ssigned to SUB used SUB, wheres only 12 percent of fmilies ssigned to UC used SUB. The durtions of ssistnce were lso longer for the ssigned interventions (not shown in the exhibit). In the SUB-versus-UC contrst, fmilies ssigned to SUB who used SUB did so for n verge of 16 months, wheres UC fmilies who used SUB without priority ccess used it for only 11 months, on verge. These findings generlly reflect the longer time it took UC fmilies to obtin ccess to SUB, if they did so t ll. Similrly, in the CBRR-versus-UC comprison, fmilies ssigned to CBRR who used CBRR did so for n verge of 8 months, wheres UC fmilies who used CBRR without priority ccess did so for n verge of 7 months. In the PBTH-versus-UC comprison, PBTH fmilies who used trnsitionl housing of ny kind did so for n verge of 12 months, wheres the UC fmilies who used trnsitionl housing without priority ccess did so for 8 months, on verge. Usul Cre The experiences of fmilies ssigned to UC inform policymkers bout wht typiclly hppens to fmilies (in the 12 study communities) who hve been in shelter for t lest 7 dys nd who do not receive priority ccess to designted ssistnce. These dt show tht, on verge, UC fmilies spent 4 months in emergency shelter during the followup period. For some fmilies ssigned to UC, the emergency shelter ws their only interction with the homeless or housing ssistnce systems. As Exhibit ES-4 shows, however, UC fmilies in ech comprison ultimtely found their wy to 5 Some trnsitionl housing progrms re bsed in projects or fcilities tht fmilies leve fter exiting the progrm. These progrms re studied here, hence the term project-bsed trnsitionl housing. Other trnsitionl housing progrms use residentil units in the community so tht fmilies cn trnsition in plce to unssisted housing without hving to move fter supports re no longer needed. Trnsition-in-plce progrms of this sort shre mny of the sme chrcteristics of CBRR, so we did not include them s progrms to which PBTH fmilies could be directed following rndom ssignment. This decision ws mde to provide stronger contrst between the PBTH nd CBRR interventions studied. Some PBTH progrms to which fmilies were ssigned provided units in the community (clled scttered-site units) without the opportunity to trnsition in plce. The Homeless Mngement Informtion System records, n importnt dt source for observing progrm use, unfortuntely do not distinguish between project-bsed nd trnsition-in-plce trnsitionl housing. Therefore, some of the trnsitionl housing use shown in Exhibit ES-4 my hve been in trnsition-in-plce units. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies xxiii

25 Executive Summry other types of ssistnce. Of the UC fmilies who responded to the followup survey (578, not shown in Exhibit ES-4), 18 percent received rpid re-housing, 25 percent received trnsitionl housing, nd 28 percent used some form of permnent subsidy (housing choice vouchers, public hous - ing, permnent supportive housing, project-bsed voucher, or ssistnce in Section 8 project). 6 Only 28 percent of UC fmilies did not use ny rpid rehousing, trnsitionl housing, or ny form of permnent subsidy during the followup period or emergency shelter fter the first 6 months beyond rndom ssignment. The outcomes for UC fmilies indicte tht they were not fring well 20 months fter study enrollment. One-hlf of UC fmilies reported hving spent t lest 1 night stying in shelter or plce not ment for humn hbittion or doubled up in the 6 months before the survey or hd sty in shelter in the yer preceding followup dt collection. In months 7 to 18 fter rndom ssignment, 28 percent hd styed in emergency shelter. In the 6 months before the survey, 15 percent of fmilies hd been seprted from child who ws with the fmily in shelter t study outset, nd 4 percent hd children in foster cre. Of UC fmily heds, 32 percent reported fir or poor helth nd 31 percent worked in the week before the followup survey. At the time of the survey, 15 percent reported lcohol dependence or substnce buse nd 12 percent hd experienced intimte prtner violence in the pst 6 months. More thn one-third of fmilies (36 percent) were food insecure. Impct Estimtes for Pirwise Comprisons Before seeing the results of the nlysis, the study tem pre - specified impcts on 18 key outcomes in the six pirwise comprisons to present in the executive summry. This step ws tken to prevent the selective presenttion of sttisticlly significnt results in the executive summry from mong the 73 outcomes exmined for ech comprison in the body of the report (438 impct estimtes). The outcomes deemed most centrl to the study nd those nticipted priori to be most likely to be ffected by the interventions were selected for this executive summry presenttion. Impcts on the full set of outcomes re presented in Chpters 6 through 9 of the report. Exhibit ES-5 reports estimted impcts for the 18 prespecified outcomes for ech pirwise comprison. The exhibit rows re orgnized into five pnels corresponding to ech outcome domin. The exhibit columns show the men vlue of ech outcome for the entire UC group, followed by impct estimtes for ech outcome in ech of the six pirwise comprisons. Asterisks to the right of the impct estimtes denote the sttisticl significnce of the estimtes, with more sterisks indicting higher levels of sttisticl significnce. Within ech domin, Exhibit ES-5 presents impcts on three or four outcomes. For the first four outcome domins (hous - ing stbility, fmily preservtion, dult well-being, nd child well-being) the outcomes re specified so tht lower vlues indicte improvements. Tht is, for these domins impct estimtes with negtive vlues indicte reductions in unfvor - ble outcomes or improvements for fmilies. For the selfsufficiency domin, the gols of the interventions re to chieve higher vlues for ech outcome. Thus, positive vlues for selfsufficiency impct estimtes indicte improvements. Detiled definitions for the full set of outcomes re in Chpter 5 nd Appendix B of the report. Now we turn to the results of the pirwise comprisons. The evidence of intervention effects is strongest for comprisons in which lrger number of impct estimtes re significntly different from zero. The study thus provides the strongest evidence of intervention effects cross the outcome domins for comprisons involving SUB. The number of significnt effects is higher nd the pttern of effects cross domins more consistent in the SUB-versus-UC, SUB-versus-PBTH, nd SUB-versus-CBRR comprisons thn is true for the other three pirwise comprisons. SUB Versus UC The most notble effect of SUB reltive to UC ws its reduction of stys in shelter nd plces not ment for humn hbittion nd reduction in doubled-up housing situtions in the 6 months before the followup survey. Assignment to SUB fter 7 dys in emergency shelter reduced subsequent shelter stys by nerly one-hlf. Assignment to SUB reduced by more thn one-hlf the proportion of fmilies who reported hving spent t lest 1 night in shelter or in plces not ment for humn hbittion in the 6 months before the followup survey. SUB lso produced lrge nd consistent effects cross every mesure of housing stbility nd doubling up (including those shown in Chpter 6 nd those selected for the executive summry). Compred with UC, SUB lso reduced the number of plces lived since rndom ssignment. Indirect benefits occurred for selected fmily preservtion indictors nd child nd dult well-being mesures. Reltive to UC, SUB led to improvements in fmily preservtion. For fmilies with child present t bseline, SUB reduced 6 In the entire study, 746 fmilies were rndomly ssigned to UC. Of these fmilies, 578 responded to the followup survey. Different subsets of these 578 fmilies form the comprison groups for SUB, CBRR, nd PBTH. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies xxiv

26 Executive Summry Exhibit ES-5. Summry of Impcts for Six Policy Comprisons Outcome Housing stbility (intervention gol: lower vlues) Men All UC Group SUB vs. UC CBRR vs. UC ITT Impct Estimtes PBTH vs. UC SUB vs. CBRR SUB vs. PBTH CBRR vs. PBTH At lest 1 night homeless or doubled up in pst 6 months or in shelter *** * 27.3*** 31.2*** 7.5 in pst 12 months b (%) At lest 1 night homeless or doubled up in pst 6 months (%) *** *** 27.3*** 9.1 Number of plces lived in pst 6 months *** *** 0.38*** 0.02 Any sty in emergency shelter in months 7 to 18 fter RA (%) *** ** 13.2*** 13.9*** 1.4 Fmily preservtion (intervention gol: lower vlues) Fmily hs t lest one child seprted in pst 6 months c (%) *** * 0.7 Spouse/prtner seprted in pst 6 months, of those with spouse/prtner ** present t RA d (%) [limited bse] Fmily hs no child reunified, of those fmilies with t lest one child bsent t RA e (%) [limited bse] ** 6.7 Adult well-being (intervention gol: lower vlues) Helth in pst 30 dys ws poor or fir (%) ** Psychologicl distress f 0.15*** *** 0.28*** Alcohol dependence or drug buse g (%) * * Experienced intimte prtner violence in pst 6 months (%) *** *** Child well-being (intervention gol: lower vlues) Number of schools ttended since RA h *** *** 0.16* Childcre or school bsences in lst month i * 0.13* Poor or fir helth (%) Behvior problems j Self-sufficiency (intervention gol: higher vlues) Work for py in week before survey (%) ** ** 6.8 Totl fmily income ($) 9, ,128** * 1,490* 18 Household is food secure (%) *** 6.1* * 7.7 Number of fmilies CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. UC = usul cre. ITT = intention to tret. RA = rndom ssignment. */**/*** Impct estimte is significntly different from zero t the.10,.05, nd.01 levels, respectively, using two-tiled t-test (not djusted for multiple comprisons). The definition of homeless in this report includes stys in emergency shelters nd plces not ment for humn hbittion. It excludes trnsitionl housing. Additionl impcts on the use of trnsitionl housing re provided in Appendix E. b After djustment for multiple comprisons, the impct on the confirmtory outcome is not sttisticlly significnt for the PBTH versus UC comprison nd is sttisticlly significnt t the.01 level for the SUB versus UC, SUB versus CBRR, nd SUB versus PBTH comprisons. c Percentge of fmilies in which child who ws with the fmily t bseline becme seprted from the fmily in the 6 months before the 18-month survey. d Percentge of fmilies in which spouse or prtner who ws with the fmily t bseline becme seprted from the fmily in the 6 months before the 18-month survey. e Percentge of fmilies in which t lest one child ws seprted from the fmily t bseline nd no child ws reunited with the fmily t the time of the 18-month survey. f Psychologicl distress is mesured with the Kessler 6 (K6) scle nd rnges from 0 to 24, with higher scores indicting greter distress. Impcts shown s stndrdized effect sizes. Effect sizes were stndrdized by dividing impcts by stndrd devition for the UC group. g Mesures evidence of lcohol dependence or drug buse using responses to the Rpid Alcohol Problems Screen (RAPS-4) nd six items from the Drug Abuse Screening Test (DAST-10). h Number of schools outcome is topcoded t four or more schools. i Absences outcome is defined s 0 = no bsences in pst month; 1 = one to two bsences; 2 = three to five bsences; 3 = six or more bsences. j Behvior problems outcome is mesured s the stndrdized Totl Difficulties score from the Strengths nd Difficulties Questionnire, or SDQ. Notes: Impct estimtes re regression djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Sources: Fmily Options Study 18-month followup survey nd Progrm Usge Dt subsequent child seprtions by two-fifths (10 percent in SUB fmilies compred with 17 percent in UC fmilies). SUB lso led to improvements in three of the four mesures of dult well-being preselected for the executive summry presenttion. SUB reduced psychologicl distress nd reduced evidence of lcohol nd drug problems. SUB lso hlved intimte prtner violence compred with UC. Assignment to SUB lso cused improvements in two of the child well-being mesures shown in Exhibit ES-5, both relted to schooling. Reltive to UC, SUB reduced the number of school bsences for focl children in the month before the survey nd lso reduced the number of schools tht the focl children ttended. The study finds no evidence tht SUB ffected the helth or behvior of focl children yet t this short-term followup point. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies xxv

27 Executive Summry SUB reduced work effort reltive to UC. SUB cused reductions in employment in the week before the survey by 6 percentge points compred with UC. This result is consistent with economic theory, given tht housing subsidies lessened the need for disposble income nd reduced the returns to work t the mrgin. SUB lso cused improvements in food security, however, incresing the percentge of households clssified s food secure by 10 percentge points reltive to UC. CBRR Versus UC Almost ll of the evidence suggests equivlent results for fmilies given priority ccess to CBRR nd fmilies ssigned to UC regrding housing stbility, fmily preservtion, nd dult nd child well-being. Most strikingly, reltive to UC, CBRR did not ffect subsequent stys in shelter or plces not ment for humn hbittion or housing stbility during the followup period. The reson for the lck of effects in this most strongly hypothesized re of potentil impct is uncler. Indictions bout consequences for children re limited, with CBRR leding to reduction in school or childcre bsences. Reltive to UC, CBRR did led to improved fmily income, with nnul income (for the clendr yer before the followup survey) for CBRR fmilies $1,128 more thn for UC fmilies. CBRR lso led to improvements in food security reltive to UC. PBTH Versus UC Compred with UC, PBTH hd effects only in the housing stbility domin. PBTH reduced the proportion of fmilies with sty in emergency shelter in months 7 to 18 fter rndom ssignment from 27 percent to 19 percent. Impcts were much smller thn those of SUB. The study yields no evidence of effects of PBTH reltive to UC in the other domins tht were exmined. The lck of impcts on dult well-being nd fmily selfsufficiency prticulrly mtter, given the emphsis plced by PBTH progrms on delivering supportive services in these res. None of the eight indictors exmined for results in this respect showed ny impct from PBTH, nor did PBTH provide better fmily preservtion or child well-being outcomes thn UC. Overll, the study did not find evidence tht the gols of PBTH s distinctive pproch to ssisting fmilies fcing unstble housing situtions were chieved reltive to UC. SUB Versus CBRR The most noteworthy effect of SUB reltive to CBRR ws in its reduction of stys in shelter or plces not ment for humn hbittion in the 6 months before the followup sur - vey nd in doubled-up housing situtions. The impcts cross these mesures were nerly s lrge s those for the SUBversus-UC comprison. The greter stbility fforded by the SUB ssistnce ws lso evidenced in reduction in the num - ber of plces lived in the pst 6 months reltive to CBRR. The other scttered effects shown by the SUB-versus-CBRR comprison mostly suggested more fvorble outcomes for fmilies ssigned to SUB. SUB reduced seprtions of spouses nd prtners, domestic violence, nd the number of schools ttended by focl children reltive to CBRR. Evidence suggests tht SUB cused reduction in work effort reltive to CBRR. SUB reduced the proportion of fmily heds who hd worked for py since study entry, the number of months worked since study entry, nd verge current ern - ings t the time of followup (not shown in the exhibit). Reltive to CBRR, ssignment to SUB lso reduced totl nnul fmily income in the yer before the survey. SUB Versus PBTH The comprison of SUB with PBTH yielded significnt impcts on 11 of 18 outcomes exmined. In most respects, the effects of SUB in comprison with UC were mirrored in the effects of SUB in comprison with PBTH. The most noteworthy effect of SUB reltive to PBTH ws in its greter prevention of stys in shelter or plces not ment for humn hbittion in the 6 months before the followup survey nd in doubledup housing situtions with impcts cross these mesures s lrge s those for the SUB-versus-UC comprison. The greter stbility fforded by the SUB ssistnce ws evidenced in reduction in the number of plces lived in the pst 6 months compred with PBTH. Although PBTH provides n lterntive plce of residence tht might be presumed stble, mny fmilies ssigned to PBTH either did not use PBTH progrm or hd left the progrm by the time of the survey. The SUB-versus-PBTH comprison showed effects on fmily preservtion, dult well-being, nd child well-being. The most notble effects of SUB reltive to PBTH re reduction in the proportion of fmilies with child seprtion in the pst 6 months, decrese in the psychologicl distress re - ported by fmily heds, nd reduction in the number of schools tht focl children ttended since rndom ssignment. In the self-sufficiency domin, the study found number of effects of SUB reltive to PBTH. SUB reduced the proportion of fmily heds who worked t the followup point (from 36 to 25 percent). Prtly s result of this lower work effort, SUB fmilies hd n verge nnul csh income of bout $1,500 less thn PBTH fmilies ($9,000 compred with $10,500). On the other hnd, the dditionl resources represented by the SUB housing ssistnce served to increse food security reltive to PBTH fmilies. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies xxvi

28 Executive Summry CBRR Versus PBTH For number of resons, the CBRR-versus-PBTH comprison offers weker test thn the other pirwise comprisons in the study. The number of fmilies in this comprison smple is the smllest of the pirwise comprisons nd so provides less sttisticl precision thn the other tests. The CBRR-versus-PBTH comprison does not yield strong pttern of effects in ny of the study domins. For the out - comes selected for the executive summry presenttion, the CBRR-versus-PBTH comprison yields sttisticlly significnt effects only in the dult well-being domin. Perhps sur - pris ingly, given the focus of PBTH on these domins, results for the three of four significnt tests for dult well-being ll f vored CBRR. In the dult well-being domin, CBRR ppers to hve reduced the proportion of dult respondents reporting poor or fir helth in the 30 dys before the survey nd lowered the mount of psychologicl distress for fmily heds reltive to PBTH. The percentge of dult respondents with evidence of lcohol dependence or drug buse ws lso lowered by CBRR reltive to PBTH. In result reported nd discussed further in Chpter 9, CBRR incresed the incidence of stys in shelters or plces not ment for humn hbittion reltive to PBTH. Do Certin Interventions Work Better When Applied to Fmilies Fcing Greter Difficulties? A centrl question motivting the Fmily Options Study is whether some interventions work better thn other interven - tions for fmilies with prticulr chrcteristics. As discussed previously, study findings hve shown tht, on verge, the SUB intervention hs substntil impcts reltive to the other interventions, not only for housing stbility but lso for out - comes in other domins. Do ll fmilies who experience home - lessness need deep permnent housing subsidy, however, or might some do s well on their own, in UC, or with the shorter nd often shllower subsidies of CBRR? Conversely, lthough on verge PBTH hd few impcts reltive to other interventions, might some fmilies who fce greter chllenges benefit more from its intensive socil services? The more generl form of this question is whether the reltive benefits of the longer term or more intensive interventions (SUB nd PBTH) might increse s fmilies reported difficulties increse. Becuse of the number of fmily chrcteristics tht could led to differentil effects of interventions, the study tem confined nlyses to exmintion of two brod ctegories of fmily chrcteristics, summrized in indices of psychosocil chllenges nd brriers to housing. The study tem exmined whether the impct of the interven - tions reltive to ech other nd to UC incresed s fmilies scores on these indices increse. It is cler tht fmilies in this study experience high levels of both psychosocil chllenges nd brriers to housing, which ws by design: the study enrolled fmilies only fter they hd spent t lest 7 dys in shelter. The exmintion of potentil modertor effects of difficulties of this sort does not provide evidence tht ny of the interventions studied work comprtively better for fmilies who hve greter psychosocil chllenges or housing brriers thn for fmilies who fce fewer difficulties. At this point the min study results on impcts cross ll fmilies provide the study s clerest guidnce for policy nd prctice. Intervention Costs The Fmily Options Study interventions were intended to vry in both intensity nd durtion. SUB progrms provided deep rentl subsidy such tht fmilies contributions to rent were limited to bout 30 percent of monthly djusted income. SUB did not provide supportive services, but the rentl sub - sidy ws for n indefinite durtion. PBTH progrms provided intensive housing nd services support for reltively long durtion. CBRR progrms provided short-term rentl subsidy with more limited supportive services, while emergency shelter progrms often offered intensive supportive services nd housing for limited time. The study tem compred the costs of the interventions using three mesures of cost: 1. Per fmily monthly progrm cost. 2. Progrm cost per sty during the followup period. 3. Cost of ll progrms used during the followup period. The first two mesures provide informtion on the reltive costs of funding different types of progrms. The third mes ure provides context for interpreting the impcts of priority ccess to the ctive interventions presented in the pirwise comprisons in Chpters 6 through 9. This mesure reflects the combined cost of ll homeless nd housing ssistnce progrms ccessed by fmilies in ech pirwise impct comprison. Per Fmily Monthly Progrm Cost Emergency shelter progrms hd the highest verge per fmily monthly progrm cost t slightly more thn $4,800. Supportive services mde up 63 percent of ES costs, the highest shre mong the four progrm types. PBTH progrms hd n verge cost of slightly more thn $2,700 per fmily per month, with supportive services constituting on verge 42 percent of PBTH progrm costs. SUB progrms cost on Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies xxvii

29 Executive Summry verge slightly less thn $1,200 per fmily per month. The cost of SUB consisted wholly of the cost of housing, becuse this intervention did not include supportive services. CBRR progrms hd the lowest per-fmily-per-month cost mong the progrm types, with progrm verge of slightly less thn $900. Housing costs mde up, on verge, 72 percent of CBRR progrm costs. The study tem found substntil vrition in the costs of the individul progrms tht mde up ech study intervention. PBTH nd ES progrms hd the gretest vrition, driven lrgely by vrition in supportive service costs but lso by vrition in cpitl costs nd dministrtive expenses. For the 24 PBTH progrms in the cost nlysis, verge per-fmily monthly progrm cost rnged from slightly more thn $1,260 to slightly less thn $6,300. Averge per-fmily monthly progrm cost for the 45 ES progrms rnged from slightly less thn $1,900 to nerly $9,200. Vrition in CBRR nd SUB costs cross progrms ws driven lrgely by housing costs. For the 12 CBRR progrms in the cost nlysis, verge per-fmily monthly progrm cost rnged from slightly more thn $550 to slightly less thn $1,400. Across the 10 sites with the SUB intervention, verge perfmily-per-month cost rnged from $770 to $2,100, lrgely reflecting differences in the locl cost of rentl housing. Progrm Costs per Sty During the Followup Period The study found different cost ordering when estimting the costs of the typicl durtion of ssistnce in ech inter - vention progrm type during the period from rndom ssignment to the followup survey for fmilies ssigned to tht progrm type (or in the cse of emergency shelter, fmilies ssigned to UC). This ssessment pplied the verge monthly per-fmily cost of ech intervention progrm to the totl time spent in the intervention progrms. Exhibit ES-6 shows verge costs of ech progrm per fmily who ws rndomly ssigned to nd used tht progrm type, ccounting for durtion of ssistnce. The costliest progrm during the followup period ws PBTH. The verge cost of housing nd support services in PBTH progrms for fmily who used PBTH ws slightly less thn $32,600 over n verge durtion of 13 months. Next, costs for SUB housing for fmilies who used SUB verged slightly more thn $18,800 for n verge durtion of 16 months. Emergency shelter costs were on verge slightly less thn $16,900 per fmily bsed on n verge length of sty of 4 months. Finlly, per-fmily CBRR costs for fmilies who used rpid re-housing verged slightly more thn $6,500 for n verge of 7 months of ssistnce. Totl Costs of Progrms Used Over the Followup Period Exhibit ES-4 shows tht fmilies ssigned to the four interventions used vriety of homeless nd housing ssistnce progrms during the followup period. The progrm use differed for ech pirwise comprison becuse different fmilies re included in ech comprison. The study tem combined informtion bout progrm use with per-fmily monthly progrm costs to estimte the totl costs of progrms used for ech intervention in the six pirwise comprisons. Exhibit ES-7 summrizes the results of this nlysis. The exhibit shows tht the totl progrm use of fmilies ssigned to SUB cost bout the sme s the totl progrm use of fmilies ssigned to UC nd slightly more thn for fmilies ssigned to CBRR. The cost of totl progrm use for SUB fmilies ws clerly less thn tht for PBTH fmilies, however. The ner Exhibit ES-6. Averge Progrm Cost per Sty During the Followup Period Across Progrm Types Averge progrm cost per sty during followup period $35,000 $30,000 $25,000 $20,000 $15,000 $10,000 $5,000 $0 $18,821 SUB (16 months) $6,578 CBRR (7 months) PBTH (13 months) Supportive services ES (4 months) Housing or shelter CBRR = community-bsed rpid re-housing. ES = emergency shelter. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. ITT = intention to tret. Note: The durtions reported in this exhibit re weighted to lign with the progrm-level cost dt nd so differ slightly from the durtions reported for CBRR nd PBTH in other exhibits. Sources: Fmily Options Study cost dt (CBRR, PBTH, nd ES); Fmily Options Study 18-month followup survey (CBRR nd PBTH); HUD Public nd Indin Housing Informtion Center, Tennt Rentl Assistnce Certifiction System, nd Finncil Dt Schedule records (SUB) $32,557 $16,829 Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies xxviii

30 Executive Summry Exhibit ES-7. Cost of Progrm Use Since Rndom Assignment for Ech Intervention Contrst Pnel A $35,000 Other SUB CBRR PBTH SUB vs. UC CBRR vs. UC PBTH vs. UC ES Cost of progrm use since rndom ssignment $30,000 $25,000 $20,000 $15,000 $10,000 $5,000 $30,832 $30,336 $27,605 $30,629 $30,817 $28,295 Pnel B $0 $35,000 SUB N = 530 UC N = 415 CBRR N = 455 UC N = 451 Assigned intervention PBTH N = 294 SUB vs. CBRR SUB vs. PBTH CBRR vs. PBTH UC N = 262 Cost of progrm use since rndom ssignment $30,000 $25,000 $20,000 $15,000 $10,000 $5,000 $31,158 $29,680 $27,864 $30,914 $22,524 $30,510 $0 SUB N = 381 CBRR N = 308 SUB N = 230 PBTH N = 187 CBRR N = 179 PBTH N = 197 Assigned intervention CBRR = community-bsed rpid rehousing. ES = emergency shelter. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. UC = usul cre. Notes: Averges re for ll 18-month survey respondents in ech rm of ech pirwise comprison nd re weighted for survey nonresponse to represent full comprison smple. Cost estimtes ssume site-specific verge cost per month bsed on the Fmily Options Study cost dt nd HUD dministrtive dt. The other ctegory includes permnent supportive housing, public housing, nd project-bsed ssistnce (project-bsed vouchers or Section 8 projects). Sources: Fmily Options Study cost dt; HUD Public nd Indin Housing Informtion Center, Tennt Rentl Assistnce Certifiction System, nd Finncil Dt Schedule records; Fmily Options Study Progrm Usge Dt equivlent cost of SUB s compred with UC ws driven both by decresed time in emergency shelter nd by decresed use of reltively more expensive PBTH progrms for fmilies ssigned to SUB. Similrly, the SUB nd CBRR costs of totl progrm use were nerly equivlent becuse the greter use of SUB progrms by SUB fmilies ws offset by the greter use of trnsitionl housing, emergency shelter, nd other progrms by CBRR fmilies. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies xxix

31 Executive Summry The SUB intervention usully lsts beyond the followup period for which we mesure both impcts nd costs in this report. Longer term housing ssistnce progrm costs re likely to chnge, nd progrm use by SUB fmilies my become reltively more costly. These subsequent costs will be ddressed in the 36-month report, long with impcts mesured t 36 months. Conclusions The Fmily Options Study s rndom ssignment design for mesuring intervention impcts is stronger design thn other studies of interventions for homeless fmilies. As result, the Fmily Options Study provides importnt new informtion bout wht hppens to fmilies who experience homelessness in the bsence of ny specil offers of ssistnce nd bout the impct of priority ccess to three types of progrms: SUB, PBTH, nd CBRR. The experimentl design of the study nd the contrsts in progrm use during the followup period provide solid foundtion for estimting the impcts of enhncing ccess to different kinds of ssistnce. The study provides the first cler evidence bout these effects nd thus cn serve s solid bsis for future policy decisionmking. Approximtely 20 months fter entry into shelter nd rndom ssignment, fmilies ssigned to SUB pper to be doing better thn the fmilies ssigned to CBRR, PBTH, nd UC. The benefits of priority ccess to SUB hve been chieved t comprble cost with tht of UC, slightly higher costs thn CBRR, nd t substntilly lower cost thn PBTH. Compred with those ssigned to UC, the fmilies rndomly ssigned to SUB on verge hve hd fewer negtive experiences (home - lessness, child seprtions, nd intimte prtner violence). SUB fmilies re lso somewht more likely to live in their own plce. Moreover, children in SUB fmilies move mong schools less, nd fmilies experience greter food security nd less economic stress. On the negtive side, heds of these fmilies exert less work effort. Fmilies given priority ccess to CBRR do bout s well s fmilies ssigned to UC, but they hve substntilly lower costs, minly becuse CBRR lowers the rte t which fmilies use costly trnsitionl hous - ing progrms. PBTH is more costly nd, t this point, hs few dvntges over other progrms. Further, no evidence suggests tht intervention impcts differ ccording to fmilies psychosocil chllenges or housing brriers whtever form of ctive ssistnce is prioritized. The 36-month followup nlysis will exmine whether these differences mong in - terventions continue to hold nd whether new differences emerge fter nother 16 months elpse. The study findings lend support for the underlying theoreticl model for SUB. The striking impcts of SUB in reducing subsequent stys in shelter nd plces not ment for humn hbittion provide support for the view tht, for most fmilies, homelessness is housing ffordbility problem tht cn be remedied with permnent housing subsidies without specilized homeless-specific psychosocil services. The findings lso provide further support for the more tenttive theoreticl proposition tht resolving homelessness would hve rditing impct, given the impcts found by this study of SUB on fmily preservtion, dult well-being, nd school stbility compred with the impcts of UC. The temporry housing subsidies of CBRR do not pper sufficient to improve housing stbility during the period studied nd hence hve little effect on outcomes presumed to emnte from chieving housing stbility. The study provides less support for the theoreticl model underlying PBTH. PBTH is intended to ddress the root cuses of homelessness by providing socil services pckged with housing ssistnce. The study does not provide evidence tht the intervention chieves this gol. PBTH led to modest reductions in homelessness when compred with UC, but it did not produce effects in other spects of fmily well-being. The Fmily Options Study is continuing to follow fmilies for 36 months fter study enrollment. This dditionl wve of dt collection will ddress number of importnt ques - tions. The 36-month nlysis will ddress whether the types of outcomes tht re improved by SUB t this point re dependent on contemporneous receipt of the housing ssistnce. Could effects fde if ssistnce ends? During the 20-month followup period reported here, 84 percent of SUB fmilies hd used SUB. By the time of the survey, SUB receipt hd fllen to 74 percent. The 36-month nlysis will exmine whether fmilies retin permnent housing ssistnce nd retin its benefits. On the other hnd, the reduced stress nd greter stbility observed for SUB fmilies t 20 months might yield dditionl benefits for dult nd child well-being over the longer term. Reductions in work effort in the short term might fde over the longer term s observed in the study of Effects of Housing Vouchers on Welfre Fmilies (Mills et l., 2006). The 36-month nlysis will nswer these questions. Similrly, the 36-month nlysis will exmine whether the focus of PBTH on ddressing psychosocil chllenges nd enhncing skills leds to benefits over the longer term tht were not evident t this point. The negtive outcomes of PBTH reltive to CBRR for dult well-being my be temporry, reflecting nxiety on the prt of PBTH fmilies tht benefits Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies xxx

32 Executive Summry were coming to n end (or hd recently ended). On the other hnd, PBTH housing stbility outcomes could fde fter ll fmilies hve left their PBTH progrms. At this point, the two mjor dvntges of CBRR over other interventions re the comprtively lower cost of CBRR nd the greter work effort observed mong fmilies ssigned to CBRR. Work effort could led fmilies to better economic outcomes in the future, with rditing benefits for other outcomes. In ny cse, if CBRR continues to hve similr outcomes to UC in most domins, but t lower cost, this result will be importnt. The reltive cost of the interventions seems prticulrly likely to chnge over time, becuse the SUB intervention usully lsts beyond the period for which we mesure both impcts nd costs in this report. Over the longer term, the continuing cost of SUB progrms my or my not continue to be offset by reductions in use of shelter nd other progrms. These future costs will be ddressed in the 36-month nlysis in conjunction with impcts mesured over the longer term. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies xxxi

33 CHAPTER 1. INTRODUCTION As prt of its mission to crete strong, sustinble, inclusive communities nd qulity ffordble homes for ll, the U.S. Deprtment of Housing nd Urbn Development (HUD) hs supported rnge of progrms to provide shelter nd services for fmilies experiencing homelessness. The Deprtment hs lso engged in prtnerships with other federl gencies to focus resources on erdicting homelessness. Opening Doors: Federl Strtegic Pln to Prevent nd End Homelessness, relesed in 2010 by the U.S. Intergency Council on Homelessness, rticultes this collective commitment nd lists four gols, one of which is to prevent nd end homelessness for fmilies, youth, nd children by 2020 (USICH, 2010). During 12-month period ending September 2013, more thn 150,000 fmilies with children in the United Sttes (495,714 people) styed in emergency shelters or trnsitionl housing progrms (HUD, 2014). People in fmilies ccounted for 35 percent of the totl sheltered homeless popultion (12-month estimte). In its effort to develop the best vilble evidence on which to bse policy decisions, HUD lunched the Fmily Options Study in 2008, wrding contrct to Abt Assocites, Vnderbilt University, nd severl other prtners. The purpose of the study is to obtin evidence to support decisionmking in the Deprtment s efforts to help fmilies leve homelessness nd to crete housing stbility nd other positive outcomes for fmilies who hve experienced homelessness. The Fmily Options Study mesures the reltive impcts of four interventions commonly used to help fmilies experiencing homelessness. The study investigtes the reltive effects of providing homeless fmilies with priority ccess to permnent housing subsidy (SUB), community-bsed rpid re-housing (CBRR), or project-bsed trnsitionl housing (PBTH). A smple of 2,282 fmilies ws rndomly ssigned to one of these three ctive interventions or to usul cre (UC) in which fmilies remined in emergency shelter with - out priority ccess to one of the ctive interventions. This report presents impct estimtes for the first 20 months fter ssignment to the interventions studied 7 nd lso presents informtion on the reltive costs of the three ctive interventions nd emergency shelter. This introductory chpter begins with description of the homeless services system. It then provides n overview of the design of the evlution. The chpter closes with description of the chrcteristics of the fmilies in the reserch smple t the time of enrollment nd n overview of the orgniztion of the reminder of the report. 1.1 Bckground on the Homeless Services System A rnge of progrmmtic pproches is used to ddress fmily homelessness. This section describes the governnce structures estblished in locl communities to ddress home - lessness nd the evolving progrms tht hve been used to provide fmilies with shelter nd to help them leve homelessness. Rther thn conducting demonstrtion to test new progrm model, the Fmily Options Study tested the impcts of types of progrms tht hve been employed by locl communities to ddress fmily homelessness The Continuum of Cre The 1987 McKinney-Vento Homeless Assistnce Act creted the foundtion for tody s homeless ssistnce systems. It specificlly funded the development of more sophisticted services thn were previously vilble for people experiencing homelessness (Burt et l., 2002). As result, shelter conditions improved, nd mny progrms dded services to ddress homeless fmilies brriers to mintining housing. The McKinney-Vento Act ws mended in 2009 to consolidte former homeless ssistnce grnt progrms into the Continuum of Cre (CoC) Progrm. Both the mended ct nd the CoC Progrm regultions formlly define the CoC, group of representtives from relevnt orgniztions within specified geogrphic re, nd the CoC s responsibilities, including homeless services system design, resource lloction, nd system mngement. 7 Two other reports provide informtion bout the Fmily Options Study. Gubits et l. (2012) described the reserch design nd nlysis pln. Gubits et l. (2013) documented study implementtion findings nd bseline chrcteristics of the reserch smple. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 1

34 Chpter 1. Introduction The term Continuum of Cre, or CoC, is used informlly to describe ll of these relted concepts: (1) the homeless services system itself, (2) the governnce structure tht leds the locl plnning nd mkes decisions bout centrlly llocted resources for the system, (3) the geogrphy covered by the system, nd (4) the federl CoC Progrm grnts tht HUD wrds nnully to fund prts of the system. The CoC Progrm interim rule requires the composition of the CoC to include representtives from orgniztions such s nonprofit homeless ssistnce providers, victim services providers, fith-bsed orgniztions, governments, businesses, dvoctes, public housing gencies, school districts, socil service providers, mentl helth gencies, hospitls, universities, ffordble housing developers, lw enforcement, orgniztions tht serve veterns, nd homeless nd formerly homeless individuls. 8 The members of the CoC must engge ll of these orgniztions to help them determine wht types of progrms re needed in their community. Although the representtion nd level of enggement of different types of stkeholders vry from one community to nother, it is universlly understood tht the CoC is the structure designted to led system-level discussions nd decisions bout strtegies for ddressing homelessness in the community Progrmmtic Approches Homeless ssistnce progrms funded by the CoC Progrm hve residentil nd nonresidentil components. Homeless ssistnce progrms generlly hve been grouped ccording to their residentil component rther thn the types of nonresidentil supportive services offered. The residentil progrms tht were prt of the homeless services system s of 2009 were ctegorized s emergency shelter, trnsitionl housing, or permnent supportive housing. Emergency nd trnsitionl housing re time-limited progrms tht rely on fmilies moving on to stble housing situtions, either subsidized or unsubsidized housing. Permnent supportive housing progrms offer permnent housing subsidies coupled with intensive services nd re vilble to fmilies only when prent hs qulifying disbility. These re brod ctegoriztions rther thn closely defined progrm models. Among nd within ech of these three progrm types, there hs been considerble vrition in qulity, housing structure nd loction, privcy nd independence for prticipnts, tenure, verge nd expected lengths of sty, services provided, rules, nd expected outcomes (Locke, Khdduri, nd O Hr, 2007). This study considers two of these three types of homeless ssistnce progrms: emergency shelter nd trnsitionl housing. The study s smple of fmilies ws recruited from emergency shelters, nd emergency shelters re the bsis for the usul cre (UC) rm of the study to which the ctive interventions, including project-bsed trnsitionl housing (PBTH), re compred. Emergency shelter nd trnsitionl housing re discussed further in the blnce of this section, s re two other types of progrms: rpid re-housing nd housing ssistnce progrms. Rpid re-housing hs received funding from the homeless services system nd from HUD, prticulrly since the enct - ment of the Homelessness Prevention nd Rpid Re-Housing Progrm (HPRP) in 2009, which provided substntil re - sources to locl progrms following tht model. 9 Also fundble under the CoC Progrm, it is the bsis of the communitybsed rpid re-housing (CBRR) ctive intervention included in this study. Housing ssistnce progrms re funded by HUD for broder group of low-income fmilies nd individuls, not specificlly focused on people experiencing homelessness. Thus they re outside the homeless services system, but they my be used to help fmilies who hve ex - perienced homelessness. They re the bsis of the permnent housing subsidies (SUB) intervention included in the study. Becuse permnent supportive housing focuses on dults with disbilities nd this study is not limited to fmilies with disbled dults, permnent supportive housing progrms were not selected s one of the interventions to be included in the Fmily Options Study. Emergency shelters typiclly serve s the first response to homelessness. Emergency shelters for fmilies frequently re open 24 hours dy nd provide shelter in congregte settings with communl sleeping nd eting spce. In some emergency shelters, however, fmilies my hve individul rooms or prtments. Shelters vry in the mount nd type of services they provide. Some shelters provide only bsic services (such s mels, showers, clothing, nd trnsporttion), wheres other shelters provide bsic services plus cse mngement nd referrls to specilized services (such s employment services or mentl helth nd substnce buse tretment). Throughout the country in 2013, 118,104 emergency shelter beds were vilble for people in homeless fmilies (HUD, 2013b). Ntionlly, bout one-qurter of fmilies leve shelter on their own within week, nd bout one-hlf leve within 8 CoC Progrm Interim Rule. CFR Prt 578.5(). 9 HPRP ws funded by the Americn Recovery nd Reinvestment Act (ARRA). Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 2

35 Chpter 1. Introduction month, consistent with the intention of the progrm to provide temporry shelter to people in crisis. 10 Trnsitionl housing progrms offer homeless fmilies plce to sty or rent subsidy with supportive services for longer period, generlly 6 to 24 months. Often fmilies re referred to trnsitionl housing from emergency shelter when shelter workers determine they need more intensive or longer term ssistnce nd meet eligibility criteri. Trnsitionl housing progrms my be rooms or prtments offered to severl fmilies in the sme building, termed project-bsed trnsitionl housing, or PBTH. Sometimes the housing is in clustered or scttered loctions where the progrm mintins the lese nd progrm prticipnts must leve upon completion of the progrm. This model is referred to s scttered-site trnsitionl housing. Sometimes the housing is in scttered loctions where fmilies rent their own prtments with temporry finncil ssistnce from the progrm nd where they cn sty fter the trnsitionl progrm ends, pying rent on their own. This model is clled trnsition-in-plce. 11 The 2013 Annul Homeless Assessment Report (AHAR) reports totl of 101,843 trnsitionl housing beds for people in homeless fmilies. This number represents the sum of beds in project-bsed progrms nd scttered-site progrms (seprte counts for the number of beds in the two types of trnsitionl housing do not exist). As expected, AHAR dt show tht stys in trnsitionl housing re longer thn those in emergency shelter. The medin vlue for fmily s sty in trnsitionl housing during single yer ws 151 nights in 2013 compred with 32 nights for emergency shelter (HUD, 2013b). 12 Similr to emergency shelters, services provided through trnsitionl housing vry substntilly from one progrm to nother. Services offered in trnsitionl housing my be more intensive thn the services offered in shelters nd my include cse mngement nd referrls, benefit cquisition nd retention, eduction nd employment services, nd mentl helth nd substnce buse tretment; they sometimes include fmily reunifiction, childcre, nd children s services. The gol of most trnsitionl housing progrms is to plce prticipnts in stble housing t progrm completion. Some trnsitionl housing progrms lso help fmilies ccess minstrem housing ssistnce funded outside the homeless services system. This study mesures the impcts of offering priority ccess to the project-bsed type of trnsitionl housing nd clls this intervention project-bsed trnsitionl housing (PBTH). 13 Fmilies experiencing homelessness my lso gin ccess to federlly funded housing ssistnce progrms for low-income households tht re funded by HUD nd operted outside the homeless services system. Housing ssistnce is provided in three wys. First, some households live in housing developments tht re owned nd operted by public housing gencies (PHAs) nd re known s public housing. Second, some households receive housing ssistnce through the Housing Choice Voucher (HCV) progrm. The HCV progrm provides tennt-bsed rentl subsidies tht fmilies cn use to rent mrket-rte housing in the community. Third, housing ssistnce is sometimes provided in privtely owned housing developments for which HUD provides rentl ssistnce through contrcts with privte owners. All three of these forms of housing ssistnce (1) re indefinitely renewble, s long s the fmily remins eligible, nd (2) hve common benefit structure tht cps fmilies monthly costs for rent nd utilities t pproximtely 30 percent of income. Housing ssistnce is often referred to s deep rentl subsidy. 14 This study mesures the impct of n offer of priority ccess to permnent housing ssistnce, usully Housing Choice Voucher, nd clls this progrm type permnent housing subsidy (SUB). Temporry rentl ssistnce is incresingly used to ssist fmilies experiencing homelessness. This type of ssistnce is referred to s rpid re-housing nd provides short-term subsidies (up to mximum of 18 months, with qurterly recertifiction of eligibility). These progrms provide some services, usully limited to ssistnce locting housing nd mintining self-sufficiency. The gol is to provide ech fmily with only the level nd length of ssistnce needed 10 Dt, which re from the 1-yer period from October 1, 2011, to September 30, 2012 (HUD, 2013), show tht, in 2012, 25 percent of people in fmilies styed 7 dys or fewer in emergency shelter, 53 percent styed from 1 to 6 months, nd 10 percent styed more thn 6 months in the reporting period. 11 Burt (2006) offers thorough description of the rnge of trnsitionl housing progrms. 12 Becuse AHAR uses 1-yer reporting period, PBTH stys tht lst longer thn 1 yer re truncted. As result, the ctul medin length of sty is likely higher thn the figure reported. 13 Tht is, most of the progrms studied in the PBTH intervention were project-bsed progrms, lso known s single site settings. A few progrms provided scttered-site trnsitionl housing, but ll progrms required fmilies to relocte t the end of progrm prticiption. Trnsition-in-plce progrms were excluded. See Chpter 8 for more detils. 14 The term deep rent subsidy is used to distinguish this type of housing ssistnce from the shllow rent subsidy provided in housing developments funded by the Low-Income Housing Tx Credit Progrm or the HOME Investment Prtnerships Progrm. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 3

36 Chpter 1. Introduction until the fmily cn py mrket rent. Towrd tht gol, sub - sidies re individully structured nd my be shllow (tht is, not necessrily reducing fmilies housing costs to s low s 30 percent of income) nd short term. This type of ssis - tnce ws funded t the federl level under the Homelessness Prevention nd Rpid Re-Housing Progrm (HPRP) s prt of the Americn Reinvestment nd Recovery Act of 2009 but ws bsed on erlier models implemented by some loclities (Burt, Person, nd Montgomery, 2005). It cn lso be funded under the CoC Progrm. This study mesures the impct of the priority offer of rpid re-housing nd clls this intervention community-bsed rpid re-housing (CBRR). 1.2 Evlution Design The objective of the Fmily Options Study is to provide evi - dence to help federl policymkers, community plnners, nd locl prctitioners mke decisions bout the best wys to ddress homelessness mong fmilies. The ultimte gol of the study is to determine wht interventions work best to promote housing stbility, fmily preservtion, self-sufficiency, nd dult nd child well-being for fmilies who hve become homeless. The evlution ws designed to ddress these reserch questions 1. Wht is the reltive effectiveness of homeless interventions in ensuring housing stbility of homeless fmilies? 2. Are the sme interventions tht re effective for short-term stbility of homeless fmilies effective for longer term stbility s well? Do some interventions promote fmily preservtion nd benefit children s well-being more thn other interventions? 5. Are different homeless interventions more effective for some ctegories of homeless fmilies thn for others? To ddress these reserch questions, the study uses n experi - mentl design. From September 2010 through Jnury 2012, the study tem recruited 2,282 homeless fmilies who hd been in emergency shelter for t lest 7 dys cross 12 sites. 16 These fmilies were rndomly ssigned to one of the three ctive interventions or to usul cre. However, not every fmily hd the chnce to be rndomly ssigned to ll three of the ctive interventions. Chpter 2 describes the rndom ssignment process in detil. Exhibit 1-1 shows the six pirwise contrsts mong the in - terventions. Fmilies were included in comprison only if they were eligible for both interventions being compred nd rndomized to one of them. Thus, for exmple, ll fmilies who were ineligible for ll PBTH progrms in site t the time of rndom ssignment were excluded from contrsts involving PBTH, mening contrsts C, E, nd F shown in Exhibit 1-1. This rndom ssignment design ssures tht comprisons of interventions involve well-mtched groups cross interventions. It follows tht ny observed differences in outcomes cn be ttributed to the differentil ssignment fmilies receive nd not to ny preexisting differences mong the fmilies. Gubits et l. (2013) verified the bseline equivlence of the pirwise comprisons using chrcteristics of fmilies t the time of rndom ssignment. 3. Wht is the reltive effectiveness of different homeless interventions in ensuring the well-being of homeless prents nd self-sufficiency of homeless fmilies? 15 The current report exmines impcts estimted 20 months fter enrollment. The study is collecting informtion on outcomes over longer, 36-month followup period. These longer term impcts will be nlyzed in 2015 nd reported in The 12 communities prticipting in the study re Almed County, Cliforni; Atlnt, Georgi; Bltimore, Mrylnd; Boston, Msschusetts; Bridgeport/New Hven, Connecticut; Denver, Colordo; Honolulu, Hwii; Knss City, Missouri; Louisville, Kentucky; Minnepolis, Minnesot; Phoenix, Arizon; nd Slt Lke City, Uth. Altogether, the study tem rndomly ssigned 2,307 fmilies. On reviewing bseline dt collected, however, the tem determined tht 25 fmilies did not stisfy the fmily eligibility requirement of hving t lest one child ge 15 or younger. They were thus enrolled in error. These 25 fmilies were removed from the reserch smple without skewing the sttisticl equivlence of the interventions. The full smple size t the time of the 20-month impct nlysis ws therefore 2,282 fmilies. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 4

37 Chpter 1. Introduction Exhibit 1-1. Six Pirwise Comprisons Among the Experimentl Interventions 1.3 Bseline Chrcteristics of the Reserch Smple At the time of enrollment in the study, ll fmilies completed bseline survey, providing informtion bout their household s chrcteristics. This section briefly reviews selected bseline chrcteristics to provide n overview of study fmilies. Gubits et l. (2013) provides more detiled description of the chrcteristics of the fmilies t the time of enrollment. To understnd how the fmilies in this study compre with the ntionl homeless fmily popultion, this section compres the smple with two ntionl estimtes of fmily homelessness. The most recent source of tht informtion is HUD s Annul Homeless Assessment Report (AHAR), which uses dministrtive dt collected loclly to produce ntionl estimtes of the number nd chrcteristics of sheltered homeless fmilies nd of people who re in shelter s indi - viduls. AHAR dt describe fmilies in shelter in 2010, when enrollment in the Fmily Options Study begn (HUD, 2012). An older (1996) source of informtion on homeless fmilies is the Ntionl Survey of Homeless Assistnce Providers nd Clients (NSHAPC; Burt et l., 1999). Although less current, NSHAPC provides survey-bsed, ntionlly representtive informtion tht is not vilble in the AHAR on the chrcteristics nd life histories of homeless fmilies. NSHAPC lso includes informtion on both sheltered nd unsheltered fmilies. Fmilies hd to sty in prticipting emergency shelter to be considered for enrollment in the Fmily Options Study (see Chpter 2 for more detils bout the enrollment process). Therefore, ny eligibility requirements tht emergency shelters plced on shelter entry lso shped the smple of fmilies who were included in the study. The most common restrictions, implemented by emergency shelter progrms in 9 of the 12 sites, relted to the composition of the fmily entering emergency shelter. Some progrms were not ble to ccommodte dult men or mrried couples in their progrms becuse the shelters provided congregte living situtions. Other progrms served only fmilies with children younger thn ge 5 or did not ccept dolescentged children. These eligibility requirements collectively limited the number of men, couples, nd older children in Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 5

38 Chpter 1. Introduction study fmilies. Another eligibility requirement relted to domestic violence. Some emergency shelter progrms would not ccept fmilies fleeing domestic violence, becuse of con - cerns bout their bility to ensure the sfety of the fmilies. The study tem would expect tht these requirements led to lower numbers of fmilies fcing domestic violence t bseline thn otherwise might hve been the cse. A typicl fmily in the study consisted of womn bout 29 yers old who hd one or two children with her in the shelter (see Exhibit 1-2). More thn one dult ws present in 30 percent of fmilies t bseline, nd, in most instnces, the second dult ws the spouse or prtner of the dult respondent. Ntionwide, 78 percent of dults in sheltered fmilies re women. The shre of men in sheltered fmilies hs incresed substntilly since 2007, probbly becuse incresing numbers of fmily shelters cn ccommodte them. A plurlity of fmilies in the study (43 percent) hd only one child younger thn ge 18 present, nd nother 30 percent hd two children with them in the shelter. One-hlf of fmilies included child younger thn ge 3, nd nerly 10 percent of dult respondents reported tht they were pregnnt t bseline. In study fmilies, older children were more likely thn younger children to be living seprtely from their prent who ws in emergency shelter t the time of enrollment. Chrcteristics of the study fmilies re similr to chrcteristics of homeless fmilies ntionwide. Mny fmilies who become homeless hve young children. Rtes of sheltered homelessness re higher for infnts nd other preschool children thn for ny other ge group: 0.8 percent of infnts younger thn 12 months nd 0.7 percent of children 1 to 5 yers styed in shelters nd trnsitionl housing progrms over the course of the yer. 17 Slightly over one-hlf of children in homeless fmilies re younger thn ge 6. About one-fourth of ll episodes of poverty in the United Sttes strt with the birth of child; the poverty Exhibit 1-2. Fmily Chrcteristics: Fmily Composition Fmily Composition Adults Fmily Chrcteristic Percent of Adult Respondents/ Percent of Fmilies/Yers Hispnic 20.2 White, non-hispnic 20.4 Africn-Americn, non-hispnic 40.9 Asin/Pcific Islnder, non-hispnic 7.2 Mixed, non-hispnic 11.2 Adult respondent is femle 91.8 Averge ge of dult respondent 30.8 yers Medin ge of dult respondent 29.0 yers Adult respondent is ge 24 or younger 27.4 Mle dult respondent with no femle wife/prtner present 3.8 Two or more dults present in shelter 29.8 Second dult: spouse or prtner 27.4 Spouse/prtner is prent of (t lest one) child with fmily 23.0 Second dult: dult child (ge 18 or older) 1.4 Age of dult respondent t rndom ssignment Less thn 21 yers old yers yers yers yers yers nd older 7.8 Number of children present in shelter 1 child children children or more children 11.2 At lest one child younger thn ge Mother is pregnnt 9.8 Source: Fmily Options Study bseline survey 17 Clculted from AHAR for persons in shelter between October 2011 nd September 2012 nd U.S. Census Sttistics for Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 6

39 Chpter 1. Introduction results from forgone ernings nd costs of cre nd from the need to stretch vilble income over more mouths to feed (Wldfogel, 2001). Thus, homelessness unsurprisingly is lso more common mong fmilies with newborn or preschool-ged child (Rog nd Gutmn, 1997; Weitzmn, 1989). The rcil chrcteristics of fmilies in the study smple re similr to those of homeless fmilies ntionwide, with n overrepresenttion of Africn-Americns when compred with the poverty popultion overll (HUD, 2012). Approximtely 41 percent of study fmilies re Africn-Americn nd not Hispnic or Ltino, nd 20 percent re Hispnic or Ltino (ll rces). About 20 percent of the study fmilies identified s White, non-hispnic/non-ltino. Another chrcteristic mesured by the bseline survey ws fmily s pst housing stbility nd history of homelessness. Exhibit 1-3 shows the bseline chrcteristics of the fmilies on these mesures. Most fmilies in the study were not home - less immeditely before entering the shelter from which they were recruited into the study. Only 21 percent described their preshelter living sitution in wy tht would be defined by HUD s homeless. 18 This rte is similr to the ntionl rte of 24 percent (HUD, 2012). Insted, most fmilies entered shelter from housing either their own housing unit or tht of friend or fmily member. About 63 percent of dult respondents in the study hd experienced homelessness t some other point in their lifetime, with bout 16 percent of dult respondents hving experienced homelessness s child. The mjority of dult respondents (85 percent) indi - cted tht they were doubled up t some point s n dult (defined s stying with fmily or friends becuse you couldn t find or fford plce of your own ). Ntionl figures re not vilble for compring the prior homelessness of the study smple to tht of ll homeless fmilies in The rte is greter, however, thn tht mes - ured in NSHAPC, which ws 50 percent (Burt et l., 1999). Prt of the difference my be explined by the fct tht the NSHAPC survey ws conducted bout 15 yers before this study s bseline enrollment period. Mny of the dults sur - veyed in NSHAPC hd come of ge t time when homeless - ness ws less common. In ddition, in n effort to trget the study to fmilies with t lest moderte needs, ll fmilies in this study hd been in shelter for t lest 7 dys. Of the dult respondents in this study s smple, 27 percent hd lived in foster cre, group home, or some institutionl setting s child. NSHAPC showed very similr ptterns Exhibit 1-3. Fmily Chrcteristics: Housing Stbility nd History of Homelessness Fmily Chrcteristic Housing instbility nd history of homelessness Housing immeditely before shelter sty Percent of Adult Respondents Owned or rented house or prtment 25.7 With friends or reltives, not pying rent 24.9 With friends or reltives, pying rent 21.1 Homeless 20.5 Hotel or motel, pid by self 4.2 Prtner s plce 2.9 Tretment or permnent housing progrm 1.1 Homeless history Previous episode of homelessness 62.9 Totl homelessness in life Medin: 6 months Doubled up history Ever doubled up s dult becuse could not py rent 84.7 Time doubled up pst 5 yers b Medin: 1 yer Childhood instbility Homeless s child 16.1 Foster cre, group home, or institution s child 27.1 Living situtions included in the definition of homeless re other emergency shelter (6.8 percent), voucher hotel or motel (4.0 percent), cr or vehicle (3.1 percent), trnsitionl housing (2.8 percent), domestic violence shelter (1.9 percent), nywhere outdoors. (1.6 percent), nd bndoned building (0.2 percent). b Time doubled up in pst 5 yers or time doubled up since ge 18 for those ges 18 to 22 yers. Source: Fmily Options Study bseline survey 18 Living situtions considered to indicte literl homelessness re emergency shelter, voucher hotel or motel, cr or vehicle, trnsitionl housing, domestic violence shelter, nywhere outdoors, nd bndoned building. This definition is consistent with the current HUD definition of homelessness, which includes living in emergency shelters, trnsitionl housing, or public or privte plces not designed for or ordinrily used s regulr sleeping ccommodtion for humn beings. See 24 CFR 91.5(1)(ii), the homeless definition finl rule. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 7

40 Chpter 1. Introduction of childhood housing instbility mong people who subsequently becme homeless, with bout 25 percent of the dults in homeless fmilies reporting tht they hd been in foster cre, group home, or nother institutionl setting s child (Burt et l., 2001). The bseline survey lso included questions bout the fmily s income nd employment sttus (see Exhibit 1-4). Most dult respondents in the study were not working t the time of enrollment (83 percent), nd more thn one-hlf hd not worked for py in the previous 6 months. Approximtely 45 percent hd not worked in more thn yer, nd 30 percent hd not worked in the pst 2 yers. For the 17 percent who were working t the time of enrollment, medin hours t their min job were 30 hours week. Looking t the employ - ment of ll dults in the fmily, bout 22 percent of fmilies hd one dult working (either the dult respondent or nother dult fmily member). The medin household income of ll fmilies in this study ws $7,410 t the bseline survey. Most fmilies in the study were receiving some form of public ssistnce t the time of the bseline survey. Eightyeight percent of fmilies in the study received Supplementl Nutrition Assistnce Progrm (SNAP), 41 percent received Temporry Assistnce for Needy Fmilies (TANF), nd 10 percent received Supplementl Security Income (SSI) for someone in the fmily. Mny fmilies in the study received Medicid benefits (60 percent), stte helth insurnce benefits (23 percent), or Stte Children s Helth Insurnce (SCHIP; 32 percent). About 86 percent of fmilies prticipted in t lest one of these helth insurnce progrms. Exhibit 1-4. Fmily Chrcteristics: Income Stbility nd Disbility Fmily Chrcteristic Income stbility nd disbility Employment history of dult respondents Percent of Adult Respondents/Percent of Fmilies No work pst 1 week 82.9 No work pst 6 months 57.1 No work pst 1 yer 45.0 No work pst 2 yers 30.3 Job chrcteristics for 17 percent of dult respondents who re working Ernings t min job Medin: $11,960 Hours per week t min job Medin: 30.0 Employment of dults in fmily One dult working for py 22.3 Two dults working for py 2.1 Totl fmily income during the pst yer 20th percentile $2,880 50th percentile (medin) $7,410 80th percentile $15,000 Public progrm prticiption SNAP (food stmps) receipt 87.8 TANF receipt 41.4 SSI receipt 9.5 UI receipt 7.2 Child support receipt 14.2 WIC receipt 36.2 Medicid receipt 60.0 Stte helth insurnce receipt 22.6 SCHIP receipt 32.4 At lest one of Medicid, stte helth insurnce, or SCHIP receipt 86.2 Disbility sttus Disbility nd/or disbled fmily member 38.7 Adult respondent hs disbility tht limits or prevents work 21.3 Nonhed ge 15+ hs disbility tht limits or prevents work 7.0 Child younger thn ge 15 hs disbility 17.2 SCHIP = Stte Children s Helth Insurnce Progrm. SNAP = Supplementl Nutrition Assistnce Progrm. SSI = Supplementl Security Income. TANF = Temporry Assistnce for Needy Fmilies. UI = unemployment insurnce. WIC = Specil Supplementl Nutrition Progrm for Women, Infnts, nd Children. Rows re not mutully exclusive. Source: Fmily Options Study bseline survey Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 8

41 Chpter 1. Introduction Fmilies in this study hve high rte of self-reported disbility. Thirty-nine percent of fmilies hve t lest one fmily member with self-reported disbility. About 21 percent of dult respondents sid tht they hve disbility tht would limit or prevent them working, nd 7 percent of fmilies hve fmily member ge 15 or older with disbility tht limits or prevents him or her from working. Study fmilies fced multiple brriers to incresing income or finding housing, s shown in Exhibit 1-5. Forty-nine percent of dult respondents in this study reported tht they hd experienced physicl buse or been thretened with violence by spouse or prtner t some time s dults. Other studies hve shown even higher rtes of domestic violence. For exmple, Bssuk et l. (1996) report results of study of homeless nd housed mothers receiving public ssistnce in Worcester, Msschusetts, indicting tht 63 percent of homeless mothers, nd lmost s high percentge of those who hd not become homeless (58 percent), reported tht they hd been severely physiclly ssulted by n intimte prtner s n dult. The Worcester study sked detiled series of questions bout such incidents s being slpped repetedly, hit with fist, hit with n object, or thretened with knife or gun, wheres the bseline survey for this study sked only one generl question bout physicl buse or threts of violence. Mentl helth nd substnce use issues re frequently iden - tified s brriers tht people experiencing homelessness fce. These issues re more prevlent mong homeless individuls thn they re mong homeless fmilies, wheres employment sttus nd broder economic chllenges re more often iden - tified s the centrl cuse of fmily homelessness (Rog nd Buckner, 2007). A history of drug use within the pst yer ws identified by 14 percent of dult respondents, nd 11 percent responded to survey questions in wy tht suggested lcohol buse within the pst yer. 19 These rtes re substntilly lower thn those reported to NSHAPC by homeless dults in fmilies (38 percent for drug use problems nd 18 percent for lcohol use problems within the pst yer; Burt et l., 2001). In nother study of homeless fmilies, Rog nd Buckner (2007) reported tht 12 percent of dult respondents hd used illicit drugs in the pst yer. Approximtely 22 percent of dult respondents gve survey responses tht indicte symptoms of post-trumtic stress disorder (PTSD), 22 percent reported symptoms of serious psychologicl distress, nd 30 percent reported evidence of one or the other. 20 In the Worcester study, Bssuk et l. (1996) reported similr rtes of PTSD symptoms for both the home - less fmilies (18 percent) nd housed welfre fmilies (16 percent). The rtes of PTSD nd serious psychologicl distress for homeless fmilies re substntilly higher thn ntionl Exhibit 1-5. Fmily Chrcteristics: Brriers to Incresing Income or Finding Housing Fmily Chrcteristic Brriers to incresing income or finding housing Exposure to violence nd mentl helth Percent of Adult Respondents/Percent of Fmilies Domestic violence by spouse or prtner s n dult 49.0 PTSD symptoms 21.6 Psychologicl distress 22.1 Previous housing history problems finding housing History of eviction Never leseholder Other brriers to housing 25.9 big or smll problem 34.8 big or smll problem Felony conviction of t lest one dult fmily member 14.2 Felony conviction of dult respondent 11.3 Felony conviction of nonhed fmily member 4.8 Drug buse 14.1 Alcohol buse 11.1 PTSD = post-trumtic stress disorder. Informtion ws collected on history of eviction nd never hving been leseholder only if the respondent thought these fctors presented problem in finding plce to live. Source: Fmily Options Study bseline survey 19 The bseline survey sked for responses to the four items in the Rpid Alcohol Problem Screen, or RAPS4 (Cherpitel, 2000). An ffirmtive nswer to ny of the four questions indictes n lcohol problem. The bseline survey lso sked sked for responses to seven items regrding use of illegl drugs, six of which re included in the Drug Abuse Screening Test, or DAST-10 (Skinner, 1982). An ffirmtive nswer to ny of these seven questions indictes drug problem. 20 About 14 percent of dult respondents hve both PTSD symptoms nd high psychologicl distress. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 9

42 Chpter 1. Introduction rtes of PTSD (5.2 percent for women nd 1.8 percent for men) (NCS-R, 2005) 21 nd serious psychologicl distress (3.9 percent for women nd 2.9 percent for men) (CDC, 2012). 22 Fmilies enrolled in the study lso reported tht they hd poor rentl history (26 percent hd been evicted) or tht they hd never been leseholder t ll (35 percent). 23 Some fmilies (14 percent) reported tht t lest one dult in the fmily hd been convicted of felony for drugs or other offenses. In 11 percent of fmilies, the dult respondent reported hving felony conviction. 1.4 Orgniztion of the Report This first chpter of the report hs provided n overview of fmily homelessness, the homeless services system, the evlution design, nd the bseline chrcteristics of the reserch smple. The blnce of this report is orgnized s follows. Chpter 2 describes the interventions studied nd the implementtion of the Fmily Options Study. Chpter 2 lso explins site selection nd exmines the chrcteristics of the study sites nd the process used to conduct rndom ssignment. Chpter 3 discusses the conceptul frmework of the interventions nd hypotheses bout their potentil effects. Chpter 4 presents the methodology nd dt sources. Chpter 5 describes the experiences of the usul cre group. It lso defines the outcomes derived from prticipnt surveys nd dministrtive dt tht re used to estimte intervention effects. Chpters 6 through 9 then present findings bout the impcts of the four interventions, orgnized by the six pirwise comprisons. In prticulr, Chpter 6 provides im - pct mesures for SUB compred to UC, for the five domins of housing stbility, fmily preservtion, dult well-being, child well-being, nd self-sufficiency. Chpter 7 presents findings from the comprison of CBRR to UC in the five domins nd Chpter 8 does so for the comprison of PBTH to UC. Chpter 9 turns to the other pirwise comprisons, reporting impcts of SUB compred to CBRR, SUB compred to PBTH, nd CBRR compred to PBTH. Chpter 10 discusses results bout the reltive impcts of groups of interventions bsed on pooled comprisons to illuminte other policy ques - tions. Chpter 11 explores the vribility of impcts cross types of fmilies, using indices relted to psychosocil chllenges nd housing brriers constructed for ech fmily. Chpter 12 describes the reltive costs of the interventions. Chpter 13 discusses study conclusions. Severl technicl ppendixes support the report. Appendix A provides detils bout the dt sources nd dtset construction. Appendix B discusses the construction of dult nd child well-being outcomes. Appendix C presents technicl detils regrding the smples nd nlysis methods. Survey nonresponse nlysis is documented in Appendix D. Appendix E contins supplementl tbles showing use of trnsitionl housing during the followup period. Appendix F presents exhibits showing the results of the pooled comprisons. Appendix G presents technicl detils bout the cost dt collection nd nlysis. 21 The sttistic for PTSD is the ntionl 12-month prevlence rte s mesured in the Ntionl Comorbidity Survey Repliction (NCS-R), which ws fielded in 2001 nd The NCS-R used different instrument to mesure PTSD thn wht ws used in the Fmily Options Study. 22 The sttistic for the ntionl rte of serious psychologicl distress is from the 2011 Ntionl Helth Interview Survey. This survey used the sme mesure of psychologicl distress tht ws used in the Fmily Options Study. 23 Percentges re of respondents who reported tht pst eviction or no rent history t ll presented big or smll problem for them in finding plce to live. The survey items did not cpture whether respondent hd pst eviction or hd no rent history t ll if the respondent did not think these fctors were problems in finding plce to live. Therefore, these percentges re lower bounds on the proportions of the respondent smple who hd history of eviction nd who hd never been leseholder. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 10

43 CHAPTER 2. IMPLEMENTING THE STUDY This chpter discusses the implementtion of the Fmily Options Study. It begins with n overview of the study interventions nd the contrsting fe - tures tht were envisioned in the study design. The next section ddresses site recruitment nd describes key chrcteristics of the 12 study sites. The reminder of the chpter then describes how the study tem implemented rndom ssignment. 2.1 Interventions Studied The Fmily Options Study exmines four interventions. The study tem collborted closely with HUD during the design phse of the study to determine wht types of interventions should be studied nd to define the distinguishing fetures. The study tem defined the interventions to include contrsts in the type nd durtion of housing ssistnce nd the presence of supportive services. The four interventions were defined s follows. 1. Subsidy (SUB) ws defined s permnent housing subsidy, usully housing choice voucher (HCV). SUB could hve included the ssistnce to find unit tht qulified for the voucher progrm tht might be vilble to nyone who receives voucher ssistnce, but it did not include other supportive services. 2. Community-bsed rpid re-housing (CBRR) ws intended to provide temporry rentl ssistnce for 2 to 6 months (potentilly renewble for up to 18 months) pired with limited, housing-focused services to help fmilies find nd rent conventionl, privte-mrket housing. 3. Project-bsed trnsitionl housing (PBTH) ws intended to provide temporry housing (for up to 24 months with ver - ge expected stys of 6 to 12 months) in gency-controlled buildings or prtment units pired with intensive supportive services. 4. Usul cre (UC) is defined s ny housing or services tht fmily ccesses in the bsence of immedite referrl to the other interventions. UC typiclly includes t lest some ddi - tionl sty in the emergency shelter from which fmilies were enrolled. The intended contrsts cross interventions in types of hous - ing subsidies nd of services, re shown in Exhibit 2-1. Detiled findings from the pirwise comprisons of ech ctive intervention (SUB, CBRR, nd PBTH) to UC re presented in Chpters 6, 7, nd 8. To ssess the nture of the housing nd services offered to fmilies ssigned to the interventions, the introductions to Chpters 6, 7, nd 8 describe how the interventions were implemented in the study sites. Those descriptions use informtion bout housing nd services col - lected from the progrms selected to operte the interventions. The intervention ssessments lso mke use of dt bout the extent to which fmilies received services from the ssigned intervention nd the durtion of tht ssistnce. 2.2 Site Selection After defining the distinguishing fetures of the study interventions, the study tem recruited sites. To select nd recruit the sites, the study tem cnvssed lrge group of Exhibit 2-1. Intended Contrsts in Subsidy nd Services for the Fmily Options Interventions nd Usul Cre Group Housing Subsidy Some Services Provided Some Hevy Light Permnent SUB Temporry PBTH CBRR None UC b CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. UC = usul cre. Outside of the hevy nd light distinction, the nture of services lso my differ between PBTH nd CBRR. CBRR focuses on services to help with locting housing, lesing up, nd settling in. By contrst, PBTH provides more comprehensive socil services, such s ssessments, provision of nd referrl to job-relted services, counseling, substnce buse tretment, nd fmily- nd child-oriented services. See Sections 5-1, 6-1, 7-1, nd 8-1 nd Chpter 12 for detils bout the types of services offered. b UC ws intended to involve other ssistnce tht fmilies ccessed on their own fter emergency shelter. In mny emergency shelters extensive services re provided. Informtion bout services offered in emergency shelters is provided in Section 5-1 nd in Chpter 12. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 11 None

44 Chpter 2. Implementing the Study Continuum of Cre (CoC) progrms cross the country to locte cities, counties, nd metropolitn res in which the number of fmilies entering emergency shelter ws considered sufficient to chieve enrollment gols nd where the intervention models defined for the study were present. Providers of SUB, CBRR, nd PBTH in selected communities hd to be willing to implement rndom ssignment evlution. Prticipting in the study ment tht progrms hd to gree to commit progrm slots to fmilies in the study nd to comply with rndom ssignment s the method of determining which fmilies would be referred to their progrms from prticipting emergency shelters. The tem worked closely with the CoC nd locl homeless system leders to collect informtion bout the homeless ssistnce system nd then negotited with progrm providers nd public housing gencies (PHAs) to determine whether sufficient number of progrm slots in ech intervention were vilble in the site to mke the study vible in the community. The study tem recruited 12 sites to conduct the study. The sites re displyed in Exhibit 2-2 with informtion bout the number of CoCs nd the geogrphy covered by ech site. By definition, ll sites hd the UC option vilble in their communities, becuse the study smple ws recruited from emergency shelters, nd UC ws defined s ssistnce tht fmilies ccessed fter 7-dy sty in emergency shelter with - out priority ccess to the ctive interventions (SUB, CBRR, nd PBTH). The study tem initilly sought to select sites tht hd ll three ctive interventions vilble. The study tem subsequently determined tht it would be cceptble to include some sites in which only two of the interventions were vilble. With ssistnce from HUD, the study tem ws ble to obtin greements with PHAs to mke housing subsidies vilble for rndomiztion to the SUB intervention in 10 of the 12 sites. HUD nd locl CoC stkeholders ssisted the study tem to secure CBRR slots for the study tht might not otherwise hve been vilble to fmilies in shelter. Thus, the study incresed the resources vilble to fmilies who were experiencing homelessness in emergency shelters in prticipting communities. Prt of the site recruitment process involved confirming tht ll progrms included in the study were good representtives of their defined intervention. The study tem strted by defining the distinguishing fetures of the interventions, Exhibit 2-2. Fmily Options Study Sites Site Nme Municipl Ares (cities/counties/geogrphic re) Included in the Study CoCs Included in the Site Almed County, Cliforni Berkeley CA-502 Oklnd/Almed County CoC Fremont Hywrd Oklnd Almed County Atlnt, Georgi Atlnt GA-500 Atlnt Tri-County CoC Bltimore, Mrylnd Bltimore MD-501 Bltimore City CoC Boston, Msschusetts Boston MA-500 Boston CoC Connecticut Bridgeport CT-503 Bridgeport/Strtford/Firfield CoC New Hven CT-501 New Hven CoC Norwlk CT-506 Norwlk/Firfield County CoC Stmford CT-508 Stmford/Greenwich CoC Firfield County Denver, Colordo Denver CO-503 Metropolitn Denver Homeless Inititive Honolulu, Hwii Islnd of Ohu HI-501 Honolulu CoC Knss City, Missouri Knss City MO-604 Knss City/Independence/Lee s Summit/Jckson County CoC Jckson County Louisville, Kentucky Louisville KY-501 Louisville/Jefferson County Jefferson County Minnepolis, Minnesot Minnepolis MN-500 Minnepolis/Hennepin County Hennepin County Phoenix, Arizon Phoenix AZ-502 Phoenix/Mes/Mricop County Regionl CoC Mricop County Slt Lke City, Uth Slt Lke City UT-500 Slt Lke City & County Slt Lke County CoC = Continuum of Cre. The Connecticut site includes multiple metropolitn res in the stte. Sources: Site recruitment dt nd HUD; CoC designtions reported reflect designtions in effect in September 2010 when study enrollment begn; since tht time, some CoCs hve been reorgnized nd renmed Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 12

45 Chpter 2. Implementing the Study such s the housing ssistnce subsidy durtion nd level nd presence of dedicted services linked to the housing ssistnce. The chllenge in this endevor ws tht shorthnd terms used by prctitioners nd reserchers, such s trnsitionl housing or supportive housing, do not reflect uniform pproches. In relity, s Rog nd Rndolph (2002) noted, even when progrms of prticulr type re specificlly chosen for study, their chrcteristics cn overlp considerbly with other progrms tht nominlly use n pproch lbeled in different wy. To ddress this chllenge, during initil site selection, the tem visited potentil study progrms (nd interviewed some by phone), collected dt on their opertions, nd completed n ssessment for ech cndidte progrm. The study tem selected progrms tht fit the study s definitions of the interventions bsed on these ssessments, rther thn bsed on progrms self-descriptions. Before selecting progrms to prticipte in the study, the study tem identified minimum requirements for progrm to be considered n exmple of ech intervention. Selecting progrms tht met these requirements ssured tht fmilies enrolled in the study would receive comprble levels of housing ssistnce nd service support within n intervention regrdless of site differences. Such comprbility in turn llowed for the evlution to test the outcomes ssocited with being rndomly ssigned to distinct interventions cross multiple sites. Overll, the dt collected from the prticipting progrms confirm tht the interventions were distinct from ech other in the wys intended by the study s design (see Sections 5-1, 6-1, 7-1, nd 8-1). Exhibit 2-3 tbultes the number of providers for ech intervention tht greed to prticipte in the study t ech site. As indicted in the exhibit, ll four interventions were offered in 9 sites. Two sites (Atlnt nd Bltimore) did not offer SUB nd one site (Boston) did not offer PBTH. Exhibit 2-3. Number of Progrms, by Site nd Intervention Site SUB CBRR PBTH UC Almed County, Cliforni Atlnt, Georgi Bltimore, Mrylnd Boston, Msschusetts Connecticut Denver, Colordo Honolulu, Hwii Knss City, Missouri Louisville, Kentucky Minnepolis, Minnesot Phoenix, Arizon Slt Lke City, Uth Totl CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. UC = usul cre. The Connecticut site includes multiple metropolitn res in the stte. Source: Study progrm dt 2.3 Chrcteristics of Prticipting Sites The 12 study sites represent diverse rnge of geogrphic loctions, size, popultion, nd housing nd lbor mrket chrcteristics. Gubits et l. (2013) provides detiled review of popultion, income, nd lbor mrket conditions in the prticipting sites. To provide context for the impct nlysis, this section discusses select housing mrket chrcteristics nd rtes of homelessness cross the 12 sites t the time study enrollment begn in Although not rndomly selected smple of communities, the sites re vried in geogrphy nd conditions tht re relted to homelessness. The sites re locted in ll four of the Census Bureudesignted regions in the country. Exhibit 2-4 displys the geogrphic coverge of the sites. Housing mrket chrcteristics offer insight into the con - ditions for obtining housing in ech of the 12 study sites (see Exhibit 2-5). The rentl vcncy rte serves s n indictor of how difficult it my be for fmily to obtin rentl housing. 24 Ten of the study sites begn enrollment in the fll of 2010; two sites (Bltimore nd Louisville) begn enrolling fmilies into the study during the spring of Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 13

46 Fmily Options Study: Short-Term Impcts of Housing nd Service Interventions Chpter for Homeless 2. Implementing Fmilies the Study Exhibit Loction Loction of Study of Study Sites Sites Exhibit 2-5. Housing Mrket Chrcteristics of Study Sites Housing mrket chrcteristics offer insight into the conditions for obtining housing in ech of the Site Rentl Vcncy Rte (% 2010) Medin Monthly Gross Rent ($ 2010) 12 study sites (see Exhibit 2-5). The rentl vcncy rte serves s n indictor of how difficult it my be Almed County, Cliforni 5.6 1,198 Atlnt, for Georgi fmily to obtin rentl housing Bltimore, Mrylnd Boston, Msschusetts 5.4 1,233 Connecticut b ,047 Denver, Colordo Honolulu, Hwii 6.1 1,171 Knss City, Missouri Louisville, Kentucky Minnepolis, Minnesot Phoenix, Arizon Slt Lke City, Uth Ares with lower rentl vcncy rtes re considered less likely to hve ffordble rentl housing. In 2010, the ntionwide rentl vcncy rte ws 8.2 percent. Among the 12 study sites, Boston hd the lowest rentl vcncy rte (5.4 percent) nd Atlnt hd the highest rentl vcncy rte (16.4 percent). Another 6 sites Almed County, Bltimore, Denver, Honolulu, Minnepolis, nd Slt Lke City hd vcncy rtes between 5 nd 8 percent, nd 3 hd vcncy rtes between 11 nd 14 percent. United Sttes Becuse these sites operted t the county level, the dt presented re for the county where the study site is locted. b The site includes the cities of Bridgeport nd New Hven, Connecticut. The figures shown re verges for the two cities. Source: 2010 Americn Community Service 1-Yer Estimtes, U.S. Census Bureu Ares with lower rentl vcncy rtes re considered less likely to hve ffordble rentl housing. In 2010, the ntion - wide rentl vcncy rte ws 8.2 percent. Among the 12 study sites, Boston hd the lowest rentl vcncy rte (5.4 percent) nd Atlnt hd the highest rentl vcncy rte (16.4 percent). Another 6 sites Almed County, Bltimore, Denver, Honolulu, Minnepolis, nd Slt Lke City hd vcncy rtes between 5 nd 8 percent, nd 3 hd vcncy rtes between 11 nd 14 percent. In 2010, the ntionl medin monthly gross rent ws $855. Of the 12 study sites, 6 Atlnt, Bltimore, Denver, Min - nepolis, Phoenix, nd Slt Lke City hd medin rents between $800 nd $900, similr to the ntionl rte. Another 2 sites Knss City nd Louisville hd rtes lower thn the ntionl verge, nd 4 sites Almed County, Boston, Connecticut, nd Honolulu hd medin rents bove $1,000, well bove the ntionl verge. Boston hd the highest medin rent of ll 12 sites, t $1,233. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 14

47 Chpter 2. Implementing the Study The prevlence of fmily homelessness lso differed mong the 12 study sites. To give sense of the extent of homelessness from one study site to nother, Exhibit 2-6 shows the proportion of the popultion tht ws homeless s reported in ech study site for the point-in-time counts conducted in Jnury 2011 s percent of the study site s totl popultion. Among the sites, Honolulu hd the highest rte of homelessness (1.26 percent of the popultion), while Phoenix hd the lowest (0.15 percent of the popultion). 25 Of the 12 study sites, 9 hd higher incidence of homelessness thn the ntionl rte of 0.20 percent. The exhibit lso shows the number of homeless fmilies (overll, not merely in the study) nd the number of people in these households in the pointin-time count. Boston hd the highest number of homeless fmilies (987 fmilies) nd Louisville hd the lowest (134 fmilies) reported. The high incidence in Boston my reflect Msschusetts right to shelter policy for homeless fmilies, mening tht ll fmilies who pply for shelter, lck lterntive housing options, nd whose income does not exceed 115 percent of the federl poverty line re entitled to shelter (Institute for Children nd Poverty, 2010). The right-to-shelter policy might lso increse lengths of sty in shelter in Boston. Ech site offers ssistnce to homeless fmilies through emer - gency shelter nd trnsitionl housing progrms. Exhibit 2-6 lso shows the level of ssistnce vilble s mesured by the number of emergency shelter nd trnsitionl housing beds tht re dedicted to ssisting people in homeless fmilies. 26 These figures provide n indiction of the locl homeless service system s size nd the reltive prevlence of emergency shelter nd trnsitionl housing in ech community s system for fmilies. One-third of the sites hd excess emergency shelter nd trnsitionl housing cpcity on the night of the point-in-time count in Jnury 2011, wheres the other two-thirds were using overflow cpcity or hd fmilies who were unsheltered (tht is, fmilies tht hd to sty in crs, on the streets, or in other privte or public plces not designed for or ordinrily used s regulr sleeping ccommodtion for humn beings). 27 Wheres roughly Exhibit 2-6. Homeless Popultion in Study Sites Site Totl Popultion in 2010 Homeless Popultion s Percentge of Totl Popultion (bsed on 2011 totl point-in-time person count) Totl Number of Homeless Fmilies (2011 point-intime count) Totl Number of Homeless Persons in Fmilies (2011 point-intime count) Number of Emergency Shelter Beds Dedicted to Homeless Fmilies (2011) Number of Trnsitionl Housing Beds Dedicted to Homeless Fmilies (2011) Almed County, Cliforni 1,510, , Atlnt, Georgi b 1,612, , ,489 Bltimore, Mrylnd 620, Boston, Msschusetts 617, ,926 2, Connecticut c 274, Denver, Colordo 600, , ,635 Honolulu, Hwii 337, , ,733 Knss City, Missouri 674, , Louisville, Kentucky 741, Minnepolis, Minnesot 1,152, ,572 1, Phoenix, Arizon 3,817, ,238 1,130 1,381 Slt Lke City, Uth 1,029, United Sttes 308,745, , , , ,364 County-level dt re presented for these sites becuse the study ws implemented in the county in which the metropolitn re is locted. b Represents the popultion of DeKlb nd Fulton Counties, becuse CoC GA-500 includes this lrger geogrphy, including Atlnt. c Represents the popultion of New Hven nd Firfield County, wheres the CoC dt represents the four CoCs tht prticipted in the study: CT 501 New Hven; CT 503 Bridgeport; CT-506 Norwlk-Firfield; nd CT-508 Stmford/Greenwich. Sources: 2010 Americn Community Service 1-yer estimtes; U.S. Census Bureu nd 2010 Decennil Census (totl popultion figures); 2011 CoC Housing Inventory Chrt nd Homeless Popultions nd Subpopultions Dt, HUD (HUD, 2011b) 25 The homeless dt from point-in-time counts re reported by the CoC nd thus do not lwys lign precisely with the geogrphy of the study site. 26 The housing inventory count of trnsitionl housing beds includes ll project-bsed nd scttered-site trnsitionl housing beds, becuse seprte counts for the two models do not exist. The study defined the PBTH intervention s trnsitionl housing in which the trnsitionl housing opertor mintined control of the housing nd required fmilies to move out of the ssisted unit into other housing t the completion of the progrm, excluding trnsition-in-plce models of trnsitionl housing. The informtion in Exhibit 2-6 overcounts the number of trnsitionl housing beds tht meet the definition of trnsitionl housing used for the study becuse trnsition in plce, which is not included in the definition of trnsitionl housing used for this study, is included in the counts provided in Exhibit CoCs re required to conduct point-in-time counts during the finl 10 dys of Jnury ech yer. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 15

48 Chpter 2. Implementing the Study equl numbers of emergency shelter beds nd trnsitionl housing beds re in plce for homeless fmilies ntionwide, 9 of the 12 study sites hve more trnsitionl housing beds thn emergency shelter beds. 2.4 Implementing Rndom Assignment This section describes the implementtion of rndom ssign - ment. The study tem excluded fmilies who left shelter in fewer thn 7 dys becuse the more intensive interventions considered in this study my not be necessry for fmilies who cn resolve housing crisis quickly. During those first 7 dys nd up until the point of rndom ssignment, it ws expected tht shelters would continue to provide ll services nd referrls they ordinrily provide to help fmilies leve shelter. As soon fter the 7-dy mrk s ws fesible, the evlution tem rndomly ssigned fmilies to the SUB, CBRR, PBTH, or UC interventions. In sites where permnent sup - portive housing (PSH) progrms were operting nd hd open slots, fmilies whom shelter stff deemed eligible for PSH were excluded from rndom ssignment. 28 These exclusions were mde only in the few sites where such progrms were vilble nd hd slots. If no PSH progrms hd openings in community, ll fmilies who met the other study criteri nd who greed to prticipte in the study were rndomly ssigned. Implementing the rndom ssignment design presented severl chllenges. In the originl design of the study, ech fmily ws to hve chnce of being ssigned to ll four groups (SUB, CBRR, PBTH, or UC). A number of fctors prevented the study from being implemented exctly s plnned. First, 3 of the 12 sites were ble to provide only two of the three ctive interventions (see Exhibit 2-3). Second, fmilies hd interventions vilble to them only where t lest one provider of the intervention type hd n vilble slot. Third, some service providers hd unique eligibility requirements for fmilies. Before rndom ssignment, the study tem screened fmilies for eligibility for the providers tht hd vilble slots. The purpose of this screening ws to minimize the likelihood of ssigning fmilies to interventions they would not be eligible to receive. 29 As result, for n intervention option to be vilble to fmily undergoing rndom ssignment, t lest one slot hd to be vilble t n intervention provider for which the fmily met providerspecific eligibility requirements. These fctors cumultively resulted in most study fmilies not hving ll four ssignment options vilble to them t the time of rndom ssignment. Of the 2,282 fmilies enrolled in the study, 474 hd ll four ssignment options vilble to them t rndom ssignment, 1,544 fmilies hd three ssignment options, nd 264 fmilies hd two ssignment options. To preserve the integrity of the experiment, fmilies were rndomly ssigned mong vilble interventions, nd ll nlyses re conducted pirwise, compring fmilies who were eligible for both interventions nd rndomized to one of them. Exhibit 2-7 illustrtes the rndom ssignment model the study tem used to llocte fmilies to ctive interventions or to usul cre. As shown t the top of the exhibit, the study popultion is ll fmilies who hd been in n emergency shelter for t lest 7 dys nd who hd t lest one child ge 15 or younger t enrollment. The ltter restriction ws imposed becuse children who rech dulthood by the time of the followup survey re not trgets for the study. In ech site, the study tem screened for eligibility those fmilies who met the bsic study eligibility requirements (presence in shelter for more thn 7 dys nd t lest one child ge 15 or younger) nd who provided informed consent. Exhibit 2-8 lists the 12 sites nd the number of fmilies ssigned to ech intervention in ech site. The next chpter discusses the hypothesized effects of the three ctive interventions SUB, CBRR, nd PBTH on fmily well-being in five domins: (1) housing stbility, (2) fmily preservtion, (3) dult well-being, (4) child well-being, nd (5) self-sufficiency. 28 PSH is vilble only to fmilies in which n dult hs qulifying disbility. 29 See Gubits et l. (2013) for detiled description of the eligibility screening conducted prior to rndom ssignment. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 16

49 Chpter 2. Implementing the Study Fmily Options Study: Short-Term Impcts of Housing nd Service Interventions for Homeless Fmilies Exhibit 2-7. Rndom Assignment Design Exhibit 2-7. Rndom Assignment Design CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. UC usul cre. CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. SUB = Notes: 2,307 fmilies completed the permnent bseline housing survey nd subsidy. were UC rndomly = usul ssigned. cre. The study tem subsequently determined tht 25 of these fmilies should not hve been enrolled in the study, s they did not hve t lest one child ge 15 or younger t enrollment. These 25 fmilies were removed from the impct nlysis smple. As shown t the top of the exhibit, the study popultion is ll fmilies who hd been in n emergency Exhibit 2-8. Interventions shelter Avilble for t lest 7 nd dys nd Prticipnt who hd t lest Enrollment one child ge by 15 or Assignment younger t enrollment. nd Site The ltter Site restriction ws imposed becuse children who rech dulthood by the time of the followup survey re not Totl Enrolled SUB CBRR PBTH UC trgets for the study. In ech site, the study tem screened for eligibility those fmilies who met the bsic Prticipnts Almed County, Cliforni study eligibility requirements 76 (presence in shelter 56 for more thn 7 dys nd t 49 lest one child ge 15 or Atlnt, Georgi younger) nd who provided informed consent. Exhibit lists the 12 sites nd 41 the number of fmilies ssigned to ech intervention in ech site. Bltimore, Mrylnd Boston, Msschusetts The next chpter discusses 64 the hypothesized effects 53of the three ctive interventions SUB, CBRR, nd Connecticut PBTH on fmily 47 well-being in five domins: (1) 73housing stbility, (2) fmily 18 preservtion, (3) dult Denver, Colordo well-being, (4) child 76well-being, nd (5) self-sufficiency Honolulu, Hwii Knss City, Missouri Louisville, Kentucky Short-Term Impcts pg Minnepolis, Minnesot Phoenix, Arizon Slt Lke City, Uth Totl ,282 CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. UC = usul cre. The Connecticut site includes multiple metropolitn res in the stte. Source: Rndom ssignment enrollment dt Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 17

50 CHAPTER 3. HYPOTHESES ABOUT INTERVENTION EFFECTS (CONCEPTUAL FRAMEWORK) This chpter of the report describes the conceptul frmework for the permnent housing subsidy (SUB), community-bsed rpid re-housing (CBRR), nd project-bsed trnsitionl housing (PBTH) interventions under study. The chpter lso posits specific hypotheses compring the three ctive interventions ech with the other nd with the usul cre (UC) intervention. The interventions reflect different implicit theories bout the nture of fmily homelessness nd pproches best suited to ddress the problem. These implicit theories rise from different understndings of (1) the origins of homelessness, (2) the needs of homeless fmilies, (3) the effect of fmily chllenges on chieving residentil stbility, nd (4) the pproprite role of the homeless ssistnce system. Some theories posit tht household chllenges for exmple, trum, substnce use problems, mentl helth issues, lck of job skills must be ddressed first for fmilies to succeed in housing. Others posit tht progress on these issues is likely to be chieved only fter fmilies re stbilized in permnent housing. The different perceptions of the homeless ssistnce system s role result in different emphses mong three centrl gols of interventions for homeless fmilies: (1) ending the immedite episode of homelessness nd preventing returns to shelter; (2) fostering longer term residentil stbility; nd (3) promoting other outcomes, including self-sufficiency, fmily preservtion, nd dult nd child well-being. To ly the bckground, this chpter begins with discussion of the chllenges tht mothers nd children who experience homelessness fce. Next, it presents some evidence tht homeless fmilies re not homogenous; rther, they differ in ptterns of homelessness nd interctions with other service systems. The chpter then lys out the rtionle for the interventions first by describing the rtionle for the two interventions with centrl focus on housing (SUB nd CBRR), nd then by describing the rtionle for PBTH, which focuses more on helping fmilies ddress other chllenges. The blnce of the chpter offers hypotheses bout the effects of ech intervention from the perspective of its proponents. The different rtionles for the interventions suggest different nd sometimes competing hypotheses bout which outcomes ech will ffect. The chpter lso offers hypotheses bout the reltive monetry costs of the pproches nd closes with discussion of possible differentil effects of interventions for fmilies with different chrcteristics; tht is, the question of wht works for whom. 3.1 Chllenges Tht Fmilies Experiencing Homelessness Fce This section begins by describing chllenges tht mothers experiencing homelessness fce. It synthesizes informtion from severl previous studies. Next, it describes chllenges tht their children encounter nd the extent to which these chllenges re unique to homelessness or re common to ll poor children. The finl portion of this section describes the heterogeneity in homeless fmilies Chllenges Tht Homeless Mothers Fce Numerous descriptive studies hve shown tht mothers who experience homelessness fce mny chllenges in ddition to residentil instbility. Some of these chllenges re due to poverty. Indeed, on the whole, mothers experiencing homelessness resemble their low-income housed counterprts, but they re younger nd hve younger children (Rog nd Buckner, 2007). Aside from living in poverty nd needing housing, the other chllenges tht homeless fmilies fce re complex. Reserchers nd service providers debte whether these chllenges re cuses, consequences, or simply correltes of homelessness. The descriptive studies in the literture cnnot ddress this debte, nor is this debte the focus of the Fmily Options Study. Rther, this study estimtes the impct of priority ccess to housing nd service interventions on fmilies homelessness nd residentil stbility, self-sufficiency, fmily preservtion or reunifiction, nd other spects of dult nd child well-being. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 18

51 Chpter 3. Hypotheses About Intervention Effects (Conceptul Frmework) Although mny mothers experiencing homelessness re similr to other women in poverty in tht they hve limited eduction nd work histories (Rog nd Buckner, 2007), mothers experiencing homelessness re substntilly poorer. For instnce, the most recent ntionl study of homeless fmilies, conducted in 1996 (the Ntionl Survey of Homeless Assistnce Providers nd Clients, or NSHAPC), reported the medin income for fmily experiencing homelessness ws only 46 percent of the federl poverty level t tht time (Burt et l., 1999). In ddition, compred with mothers of fmilies receiving public ssistnce, mothers experiencing homelessness come from much more difficult housing circumstnces including more frequent doubling up, overcrowding, nd mobility (Shinn et l., 1998). Mothers experiencing homelessness re likely to hve hd more difficult childhoods thn low-income housed mothers. They re more likely to hve grown up in poverty, to hve spent time in foster cre, nd, bsed on some studies, to hve hd more exposure to violence (Bssuk et l., 1997; Shinn et l., 1998). In study by Bssuk et l. (1996), more thn 90 percent of mothers experiencing homelessness hd suffered severe physicl nd sexul buse, domestic violence, or rndom violence t some point in their lives, lthough they did not differ gretly from other poor mothers in this respect (Rog nd Buckner, 2007). Mothers who re homeless multiple times, however, hve been exposed to more violence thn those who re homeless only once or thn poor women who do not experience homelessness (Bssuk, Perloff, nd Dwson, 2001). Rtes of dignosble mentl helth problems, including mjor depression, nxiety disorders, nd post-trumtic stress disorder (PTSD), re similr for homeless nd other poor women (Bssuk et l., 1996; Bssuk, Rubin, nd Lurit, 1998), but they re much higher in both groups thn in the generl popultion (Bssuk, Rubin, nd Lurit, 1998). Furthermore, levels of PTSD predicted continued residentil instbility 30 months fter progrm entry in the Service nd Housing Interventions for Fmilies in Trnsition (SHIFT) study of fmilies in shelters, trnsitionl housing, nd perm - nent supportive housing (PSH) in Upstte New York (Hyes, Zonneville, nd Bssuk, 2013). Mothers experiencing homel essness re more likely thn other poor mothers, but less likely thn single dults experiencing homelessness, to hve substnce use problems (Bssuk et l., 1997; Bssuk, Rubin, nd Lurit, 1998; Burt et l., 1999; Rog nd Buckner, 2007). Levels of disbility mong dults in fmilies experiencing homelessness re lso high 18.6 percent versus 8.1 percent in ll U.S. fmilies (HUD, 2013b). Findings regrding socil networks re conflicting. Mny studies found the networks of mothers experiencing homelessness to be weker or more conflicted thn those of housed women (for exmple, Bssuk et l., 1997). Other studies found tht, t the time of shelter request, women on the verge of homelessness reported more ties nd more recent contct with fmilies nd friends thn poor housed women (for exmple, Shinn et l., 1998) Chllenges Tht Children Experiencing Homelessness Fce As with their prents, children who experience homelessness fce multiple chllenges. Mny studies hve compred the chrcteristics of children in the midst of n episode of homelessness with those of children in low-income housed fmilies nd the U.S. child popultion s whole. Fewer studies hve ssessed whether the experience of homelessness my crete long-term dverse effects for children. Children whose fmilies become homeless experience high levels of both poverty nd instbility. Reserch on other popultions shows tht poverty nd residentil, school, nd fmilil instbility re ssocited with poor child djustment cross number of domins, including school performnce, behvior, nd self-regultion (for exmple, Adm, 2004; Betty, 2010; Chen, 2004; Evns, 2004; Fntuzzo et l., 2012; Herbers et l., 2012; Msten et l., 2012; McLoyd, 1998; Mehn nd Reynolds, 2004; Obrdović et l., 2009; Pribesh nd Downey, 1999; Voight, Shinn, nd Ntion, 2012; Yoshikw, Aber, nd Berdslee, 2012). Mny low-income children, housed nd homeless, live with only one prent, but children experiencing homelessness re prticulrly likely to become seprted or removed from their custodil prents. Prk et l. (2004) mtched gency records regrding homelessness nd child protective services to show tht, of 8,251 children who entered shelter with their prents for the first time in 1996 in New York City, 16 percent hd experienced out-of-home plcement within the next 5 yers. Cowl et l. (2002) interviewed fmilies in the sme city nd thus were ble to document informl nd forml seprtions. Their reserch found, over similr timefrme, tht 44 percent of group of 251 mothers ex - periencing homelessness hd been seprted from t lest one child compred with 8 percent of 292 housed mothers receiving public ssistnce. Only one-fifth of the 249 seprted children who hd been homeless returned to live with their mother by the end of the study period. Hyes, Zonneville, nd Bssuk (2013) documented even higher levels of sepr - tion, especilly for mothers in PSH. Seprtion of children from their prents hs immedite consequences for children, fmilies, nd the protective services system. Moreover, seprtion from prents in the fmily of origin is predictor of future homelessness in dults (Rog nd Buckner, 2007). Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 19

52 Chpter 3. Hypotheses About Intervention Effects (Conceptul Frmework) Thus, the Fmily Options Study cn mke importnt contributions by identifying whether prticulr interventions cn reduce rtes of child seprtion from fmilies who experience homelessness. Most studies tht exmine how homelessness ffects children consider only children who remin with their prents nd often exmine the children in the midst of n episode of homelessness. These findings hve evolved over time. Erly reserch on homeless children compred them with normtive dt on middle-clss children nd found lrge differences regrding helth, mentl helth, behvior development, nd cdemic performnce (Rfferty nd Shinn, 1991). Subsequent studies tht included comprison groups of low-income housed children found tht both children who experience homelessness nd children living in poverty fred worse on ll these dimensions thn middle-clss children or ntionl norms, but the poor nd homeless children differed less from ech other (Buckner, 2008). For exmple, Buckner nd Bssuk (1997) found tht essentilly the sme proportions of homeless nd housed, poor children older thn ge 9 hd psychitric disorder with impirment, nd tht the proportion nerly one-third ws fr higher thn the ntionl norms. Two studies of erly childhood development found differences between children experiencing homelessness nd housed, poor children on screening test bsed on prentl reports, but third study, bsed on stronger ssessment tool, did not. Differences in cdemic outcomes nd especilly in helth outcomes between children experiencing homelessness nd housed children living in poverty were more consistent, lthough still not uniform, nd gin both groups differed from the generl popultion. For exmple, Weinreb et l. (1998) found higher frequency of helth problems mong children experiencing homelessness thn mong housed chil - dren living in poverty. Msten et l. (1993) described the overll pttern s reflecting continuum of risk, with children living in poverty worse off thn their middle-clss peers nd children experiencing homelessness often worse off thn other poor children, lthough not lwys significntly so. Buckner (2008) suggested tht chnges in reserch findings from erlier to lter studies were not solely becuse of study design. The first fmilies to become homeless, when housing mrkets were reltively benign, were more vulnerble thn other poor fmilies, but, s more nd more poor fmilies hve become homeless, the differences hve gotten smller. He lso suggested tht shelter conditions in mny communities hve improved, nd school systems re doing better job of ccommodting children who re experiencing homelessness. Thus homelessness per se hs become less im - portnt thn some of the other stressors tht poor nd home - less children fce, such s exposure to violence in their homes nd communities (Buckner, Berdslee, nd Bssuk, 2004). Problems of children who become homeless my diminish over time. For exmple, Buckner et l. (1999) found in cross-sectionl nlysis tht children s psychitric symptoms peked t bout 4 months in shelter; therefter, children seemed to hve dpted to the shelter environment. In longitudinl study, Shinn et l. (2008) found few differences between housed children living in poverty nd children who hd experienced homelessness nd remined with their fmilies 5 yers fter shelter entry. Using bord of eduction records for children in the sme smple, Rfferty, Shinn, nd Weitzmn (2004) found tht, both before becoming homeless nd fter returning to residentil stbility, children who experienced homelessness with their fmilies did not differ significntly on stndrdized tests from continuously housed children living in poverty. While they were in shelter, how - ever, the performnce of the children experiencing homeless - ness ws significntly lower thn tht of the continuously housed, poor children. The children who hd experienced homelessness hd more school mobility nd greter grde repetition thn continuously housed, poor children Heterogeneity Among Fmilies in Ptterns of Homelessness nd Use of Other Services Ptterns of homelessness mong fmilies show considerble heterogeneity. Culhne, Metrux, et l. (2007) creted typology of these ptterns for fmilies entering shelter or trnsitionl housing for the first time, bsed on the durtion nd number of subsequent homeless spells recorded in dmin - istrtive dt. 30 About three-fourths of fmilies experienced temporry homelessness, mrked by single episodes of homelessness of short durtion, rnging from n verge of 33 dys in Columbus, Ohio, to 139 dys in New York City. Another 20 percent hd n verge of less thn 1.5 episodes of homelessness but longer durtions. These long styers verged 187 dys in Columbus nd 552 dys in New York. The remining 2 to 8 percent experienced episodic home - lessness chrcterized by multiple stys. Episodiclly homeless fmilies hd n verge of three episodes during 2- or 3-yer followup period, nd cumultive lengths of homelessness vried from 148 to 385 dys (Culhne et l., 2007). 30 The nlysis covered 2-yer observtion periods for the stte of Msschusetts nd city of Columbus, Ohio, nd 3-yer observtion periods for Phildelphi, Pennsylvni, nd New York City. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 20

53 Chpter 3. Hypotheses About Intervention Effects (Conceptul Frmework) Culhne et l. (2007) lso exmined how these ptterns were relted to use of other services, including psychitric nd substnce buse tretment; Supplementl Security Income, or SSI, disbility; Temporry Assistnce for Needy Fmilies (TANF); unemployment; nd foster cre. The tem - porry nd long-styer groups looked surprisingly similr in their use of services. It ws episodiclly homeless fmilies, those most likely to cycle in nd out of housing nd homelessness progrms, who hd the most intensive service needs. Thus, the reserchers posited tht lengths of sty for homeless fmilies re more relted to locl homelessness policy, progrm structure, nd funding thn fmily need. They suggested restructuring homeless services to find better mtch between household needs nd different pckges of rentl ssistnce nd services. The 2012 Annul Homeless Assessment Report, or AHAR, lso suggested considerble heterogeneity mong fmilies in ptterns of shelter use. One-fourth of fmilies (24.1 percent) who ccessed emergency shelter (in this cse, distinguished from trnsitionl housing) remined cumultive totl of 7 or fewer nights during the 12-month observtion period. Another 28.8 percent of fmilies used emergency shelters between 8 nd 30 dys. At the other end of the distribution, 10.3 percent of fmilies used shelter for 6 months or more (HUD, 2013). Existing reserch on the needs of fmilies experiencing homelessness, for the most prt, hs filed to cknowledge the heterogeneity mong the fmilies, which extends to fmily chllenges nd ptterns of shelter sty. Children who experience homelessness show similr heterogeneity, with some demonstrting problems cross multiple domins nd others showing resilience (Huntington, Buckner, nd Bssuk, 2008). To ddress this heterogeneity, this study will exmine whether the different interventions yield better outcomes for two types of fmilies: those with more psychosocil chllenges nd those who reported more brriers to housing t study entry. 3.2 Conceptul Frmework nd Hypothesized Effects for the Fmily Options Study Interventions The Fmily Options Study is the first experimentl evlution of the role tht these interventions SUB, CBRR, nd PBTH might ply in ending homelessness mong fmilies. Although experimentl evidence indictes (reviewed subsequently) tht subsidies cn prevent homelessness, there is t best qusi-experimentl evidence for the success of SUB nd CBRR in ending homelessness. This qusi-experimentl evidence needs to be interpreted with cre. When groups re not rndomly ssigned to tretments, the possibility lwys exists tht outcomes reflect preexisting differences mong groups rther thn impcts of the interventions. Consistent with the possibility tht some differences in outcomes re due to differences in the groups (or people) who receive ech tretment, the interim report on the enrollment phse of the Fmily Options Study showed tht CBRR nd especilly PBTH progrms were more selective bout the fmilies they would serve thn SUB progrms (Gubits et l., 2013). Thus preexisting differences, lso clled selection effects, re plusible lterntive explntions for mny of the findings we review in this chpter. 31 Most studies of PBTH lck ny comprison group, much less one tht hs been rndomly ssigned, so it is difficult to know wht would hve hppened in the bsence of intervention. In wht follows, we review the existing evidence to motivte our hypotheses nd to set the stge for our experimentl results, but we re cogniznt of the limittions of tht evidence. 3.3 Conceptul Rtionle for SUB nd CBRR This section considers the conceptul rtionle for SUB nd CBRR. Considering these two interventions together is p - proprite becuse proponents of both SUB nd CBRR believe tht the key gols of homeless interventions should be ending homelessness swiftly, reducing returns to shelter, nd restor - ing fmilies to housing stbility. This position follows from their view tht fmily homelessness is lrgely consequence of housing costs tht outstrip incomes of poor fmilies, problem tht housing subsidies cn solve. An episode of homelessness my be precipitted by unpredictble trigger events such s finncil crisis or domestic conflict (NAEH, 2012; O Flherty, 2009). Middle-clss fmilies my find lterntive housing quickly, but fmilies lredy living on the mrgins my need help in recovering. Subsidies, whether they re the permnent subsidies of the SUB intervention or temporry subsidies such s CBRR, cn help fmilies obtin nd mintin stble housing. Resource constrints men tht, outside the context of this study, SUB is rrely ccessible by fmilies t the outset of n 31 As noted previously, to preserve the integrity of the Fmily Options Study experiment, ll comprisons of interventions include only fmilies who were eligible for the interventions in given comprison nd were rndomly ssigned to one. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 21

54 Chpter 3. Hypotheses About Intervention Effects (Conceptul Frmework) episode of homelessness unless they lredy hve plce ner the top of witing list. SUB ws not creted s response to homelessness. Insted, SUB lredy existed s n element of the broder socil sfety net t the time the homeless services system cme into being in the lte 1980s. Housing ssistnce is intended for broder group of households thn those who experience homelessness, nd witing lists for housing choice vouchers, or HCVs, nd plcements in public housing re long. By contrst, CBRR ws developed specificlly s response to homelessness. Becuse SUB is unlikely to become widely vilble to fmilies t the time they re experiencing homelessness, proponents of CBRR rgue tht limited resources dedicted to homelessness could be stretched to crete the best outcomes for the most people by mking subsidies temporry (Culhne, Metrux, nd Byrne, 2011). Proponents of CBRR emphsize restoring fmilies to conventionl housing s swiftly s possible (the rpid in rpid re-housing), thereby reducing time in shelter nd on the street, which they see s hrmful. In ddition, they focus on preventing returns to homelessness. Proponents of SUB focus more on long-term stbility nd worry tht short-term subsidies provided by CBRR my not be sufficient to foster such stbility. Advoctes of both types of subsidies cknowledge tht homeless fmilies, like other poor fmilies, must contend with vriety of chllenges, but subsidy dvoctes believe tht such chllenges re better ddressed by minstrem community gencies thn by specilized homeless services. Chllenges do not strongly differentite most fmilies who experience homelessness from other poor fmilies who sty housed, t lest before they become homeless (Shinn et l., 1998, 2013). Homelessness nd housing instbility re likely to excerbte fmilies chllenges over time. Proponents of subsidies rgue tht stble housing provides pltform from which fmilies cn ddress other problems on their own, using community resources if they need to nd choose to do so, while reserving scrce housing dollrs for housing. This understnding is illustrted in Exhibit Predictions Regrding Housing Stbility Given this implicit rgument for SUB nd CBRR, wht pre - dictions would their proponents mke bout the effect of the interventions on housing stbility? Proponents of both SUB nd CBRR see the crisis of housing ffordbility s the root cuse of homelessness mong fmilies. In 2011, 11.8 million renter households in the United Sttes hd extremely low incomes, defined s less thn 30 percent of Are Medin Income (AMI). More thn one-hlf of them (52.7 percent) did not hve housing ssistnce nd either pid more thn one-hlf of their income for housing, hd severely indequte housing, or experienced both (Steffen et l., 2013). Further, young dults strting fmilies re not lwys ble to brek into the housing mrket tht is, they hve never lived in plce of their own. Americn Housing Survey dt suggest tht in 2011 households consisting of more thn one fmily totled 3.6 million, reflecting the drmtic growth of Fmily Options Study: Short-Term Impcts of Housing nd Service Interventions for Homeless Fmilies Exhibit 3-1. Exhibit Conceptul 3-1. Conceptul Intervention Intervention Model for SUB Model nd for CBRR SUB nd CBRR CBRR = community-bsed rpid re-housing. SUB = permnent housing subsidy. UC = usul cre. CBRR = community-bsed rpid re-housing. SUB = permnent housing subsidy. UC = usul cre. vs. = versus. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 22

55 Chpter 3. Hypotheses About Intervention Effects (Conceptul Frmework) sub fmilies during the recession. 32 Adults in the subfmilies, unsurprisingly, were less well educted nd more likely to be unemployed thn dults in one-fmily households (Eggers nd Moumen, 2013). Most fmilies who become homeless re probbly drwn from both subfmilies nd fmilies with extremely low incomes who py more thn one-hlf of their income for housing. Worst cse needs reports show tht unssisted renters with incomes t or below 30 percent of AMI, or roughly the poverty level, cnnot fford mrket-rte housing. Households tht become homeless re even poorer, on verge. As noted previously, in NSHAPC, the income for the medin homeless fmily ws 46 percent of the federl poverty level (Burt et l., 1999). The medin nnul fmily income of fmilies in the Fmily Options Study ws only $7,410 t the time of enrollment, or bout 15 percent of the ntionl medin household income. Although ntionl consensus clls for n end to homelessness, no similr consensus steps forwrd to ddress the broder problem of housing ffordbility. If this interprettion of the ntionl consensus is correct, the opertive question becomes this: Cn we ddress homelessness without ddress - ing the broder problem nd, if so, wht is the lest costly wy of doing so? Proponents of SUB believe tht, becuse fmilies who experience homelessness re very poor, they re likely to require long-term rentl subsidies to remin stble. Such fmilies re likely to hve ongoing difficulties ffording housing t mrket rtes without ssistnce. Although one-fourth of fmilies who experience homelessness exit shelter within 7 dys, fmilies in this study re likely to need more ssistnce. Fmilies hd to hve been in shelter for t lest 7 dys to enroll in the Fmily Options Study nd 63 percent of enrollees hd experienced previous episodes of homelessness. Such fmilies my need longer period of sub - sidy thn CBRR provides to regin nd mintin stbility. Strong evidence suggests tht ongoing subsidies tht hold housing costs for rent nd utilities to 30 percent of fmily income, s the SUB intervention does, both prevent nd end homelessness (Khdduri, 2008). The most rigorous evidence for subsidies s primry prevention for fmilies comes from the experimentl Effects of Housing Vouchers on Welfre Fmilies study (Mills et l., 2006; Wood, Turnhm, nd Mills, 2008). Reserchers rndomly ssigned 8,731 fmilies receiving TANF in six sites (Atlnt nd August, Georgi; Fresno, Cliforni; Houston, Texs; Los Angeles, Cliforni; nd Spokne, Wshington) to receive voucher or to be plced on the public housing witing list without voucher initilly. Using tretment-on-the-treted (TOT) nlysis, which exmined the effects of vouchers on those induced to use them by the experiment, the study showed tht housing vouchers prevented homelessness, both nrrowly nd brodly defined. In survey conducted 4.5 to 5 yers fter rndom ssignment, 12.5 percent of the control group, but only 3.3 percent of the experimentl group, reported living on the streets or in shelters in the previous yer, nd 44.8 percent of control households but only 9.3 percent of the experimentl households lcked plce of their own or hd lived with others in the previous yer. Fmilies in this study were TANF recipients but less thn 2 percent were homeless (in shelter or trnsitionl housing) t the time of rndom ssignment. Other nonexperimentl studies suggest tht housing subsidies effectively end homelessness for most fmilies who hve ex - perienced it. Subsidies reduced returns to shelter s mesured in dministrtive records in Phildelphi nd New York City (for exmple, Culhne, 1992; Wong, Culhne, nd Kuhn, 1997) but not in Georgi (Rodriguez, 2013). Subsidies lso enhnced stbility in survey dt from nonexperimentl study of 256 fmilies in New York City. Within 5 yers fter their initil shelter entry, 80 percent of fmilies who received housing subsidies were stble (defined s in their own plce without move for t lest yer) nd, on verge, the fmilies hd been in their current home for 3 yers. Only 18 percent of fmilies who did not receive subsidies ttined stbility. Although receipt of subsidies ws not rndomized, fmily chrcteristics t the time of shelter entry did little to predict receipt of subsidies (Shinn et l., 1998). In the bsence of sufficient long-term subsidies, rpid rehousing is n intervention involving temporry subsidies designed to chieve two importnt gols: (1) hstening exit from shelter s swiftly s possible nd (2) ssisting fmilies to lese up in mrket-rte housing. CBRR focuses on helping fmilies overcome whtever crisis precipitted their shelter entry to end their homelessness rpidly nd minimize the negtive consequences of homelessness for fmilies nd children (NAEH, 2012). Proponents lso rgue tht most fmilies do not need long periods of preprtion s in PBTH to be ble to live independently (NAEH, 2012). About two-thirds of households ntionwide (including both fmilies nd individuls) tht were literlly homeless upon entry into the Homelessness Prevention nd Rpid Re-Housing Progrm (HPRP) in its first 2 yers exited to housing tht ws deemed stble t the time of exit (64.2 percent of households who exited in yer 1, 64.9 percent of households who exited in yer 2; HUD, 2011, 2013c). 32 The number of households with subfmilies grew 22.6 percent from 2003 to 2009; dt chnges preclude computtion of percentge growth from 2009 to Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 23

56 Chpter 3. Hypotheses About Intervention Effects (Conceptul Frmework) No experimentl evidence exists regrding CBRR s effects on return to shelter, but there is suggestive evidence from multiple sources. Perhps the strongest evidence for the success of CBRR in preventing returns to homelessness comes from nonexperimentl study of more thn 9,000 households, roughly one-hlf of which were fmilies with children, tht trnsitioned out of homelessness in Georgi. Homeless Mngement Informtion System (HMIS) records showed tht, over 2-yer observtion period, 17.1 percent of those who exited from rpid re-housing returned to shelter compred with 47.5 percent who exited from emergency shelters with - out dditionl help from the homeless services system. This study did not find ny dditionl benefit of long-term subsidies (Rodriguez, 2013). Gle (2012) similrly documents returns to homelessness within 12 months mong fmilies with children in seven Continuums of Cre (CoCs) in four sttes. Tht study found tht returns to shelter were 4 percent for fmilies receiving rpid re-housing compred with 11 per - cent for fmilies receiving emergency shelter nd 9 percent for fmilies receiving trnsitionl housing (Gle, 2012). A third study of the 23 communities tht prticipted in the Rpid Re-Housing for Homeless Fmilies Demonstrtion progrm found tht only 10.4 percent of 483 fmilies returned to shelter in the yer following progrm exit (Finkel et l., forthcoming). This study hd no comprison group. Households in these studies were not rndomly ssigned to progrms, nd, to the extent tht rpid re-housing ws reserved for households deemed likely to become stble with ssistnce, tht might explin t lest prt of the substntil effects on returns to shelter. It is less cler tht CBRR leds to housing stbility s opposed to reducing literl homelessness nd returns to shelter. For exmple, in the evlution of the Rpid Re-Housing for Home - less Fmilies Demonstrtion progrm, 33 percent of 127 fmilies who prticipted in followup survey reported being doubled up in the yer fter exiting from rpid re-housing, nd 76 percent of 380 for whom ddresses could be verified hd moved during tht yer. These numbers re likely to be underestimtes becuse no ddress informtion could be found for 22 percent of prticipnts (Finkel et l., forthcoming). The high rtes of reloction nd doubling up suggest tht CBRR my not be sufficient to ensure these other spects of housing stbility nd tht SUB, n intervention tht is not prt of the homeless services system, my be superior to CBRR regrding these broder gols Predictions Regrding More Distl Outcomes Proponents of both permnent nd temporry housing subsidies (SUB nd CBRR) mke fewer clims bout the effects of subsidies on outcomes other thn housing stbility. Proponents of SUB rgue tht subsidies should stbilize fmilies in housing, thereby minimizing the hrmful consequences of homelessness. Given stble housing, fmilies cn ddress ny other problems they my experience using minstrem community resources or on their own. In dvocting rpid re-housing, CBRR proponents lso focus on reducing the length of time fmilies spend in emergency shelter or on the street with view to minimizing potentilly hrmful effects of shelter. The effects of subsidies on these more distl outcomes re likely indirect. Thus, dditionl hypotheses from the perspective of proponents of SUB nd CBRR re stted more tenttively. Fmily preservtion. Perhps the strongest clims for the vlue of SUB nd CBRR on outcomes other thn housing stbility re for fmily preservtion. Although severl studies hve documented extrordinrily high rtes of seprtion of children from fmilies who experience homelessness, very few studies hve explined why homelessness is ssocited with fmily seprtion. Those tht do re observtionl nd cnnot estblish cuslity. Prk et l. (2004) found tht mong fmilies with ny stys in homeless shelters in New York City, recurrent shelter entries nd longer stys were the strongest predictors of involvement with child protective services. Other predictors (from dministrtive records) were domestic buse nd hving fewer dults in the household, younger prent, nd more children. In smple of 292 homeless mothers in Upstte New York, fmilies with younger mothers nd more children hd more child protective ser - vices involvement within 30 months of progrm entry (Hyes, Zonneville, nd Bssuk, 2013). Additionl predictors were mternl mentl helth tretment, residentil instbility, unemployment, nd ttendnce t Alcoholics Anonymous or Nrcotics Anonymous (indictors for n lcohol or drug use problem). Cowl et l. (2002) found tht drug dependence, domestic violence, nd ny institutionl plcement of the mother were ssocited with both informl seprtions nd protective services involvement in both housed nd homeless fmilies in New York City, but homelessness ws by fr the most potent predictor. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 24

57 Chpter 3. Hypotheses About Intervention Effects (Conceptul Frmework) Moving beyond correltes, Prk et l. (2004) suggested the possibility of fishbowl effect in which fmilies in the home - less system re subject to heightened scrutiny from service providers, leding to reports to child protective services. Pst studies hve shown tht observers re more likely to rte behvior s busive when performed by low-income rther thn middle-income prents (McLoyd, 1990). Results in Prk et l. (2004) nd Cowl et l. (2002) suggest n dditionl explntion. In the Prk nd Cowl studies (but not in the Hyes, Zonneville, nd Bssuk [2013] study), rtes of protective services involvement nd more informl seprtions were low before the initil episode of homelessness. Child welfre involvement often begn only fter fmilies left shelter or between episodes of shelter sty. These findings suggest tht the stress, disruption, or heightened scrutiny cused by homelessness my be importnt nd lsting. Dt from 80 fmilies who took prt in smll qulittive field effort conducted s prt of the broder Fmily Options Study provide dditionl support for both the stress nd the scrutiny perspectives. Compred with independent living situtions, both shelters nd trnsitionl housing disrupted fmilies routines nd rituls. Fmilies were subject to surveillnce nd sometimes explicitly thretened with involvement with child protective services (Myberry et l., 2014). Whether seprtions re becuse of the stress nd disruption of homeless progrms or incresed scrutiny of prenting behvior, fmilies who re quickly rehoused in conventionl housing with SUB or CBRR my hve fewer seprtions. (Proponents of PBTH mke the opposite prediction, s discussed subsequently.) Child well-being. Reductions in instbility nd in prentl stress my enhnce child well-being more generlly. As cited previously, both residentil nd school instbility hve been shown to be risk fctors for children s cdemic performnce nd the centrl noncognitive skill of self-regultion. Residen - til stbility is likely to be closely linked to school stbility. To the extent tht SUB nd CBRR enhnce both forms of stbility, they should lso hve positive effects on dditionl child outcomes in these domins. A cvet is tht such out - comes might tke more thn 18 months the current followup point to develop. The discussion here bout effects on child well-being is more logicl thn empiricl. No studies of the effects of housing subsidies on homeless children hve been conducted, so we turn to the limited evidence for the impct of subsidies on poor children more generlly. The study of Effects of Housing Vouchers on Welfre Fmilies (Mills et l., 2006) found little reltionship between housing subsidies nd children s eductionl or behviorl outcomes. A lrge-scle study of rndomized housing voucher lottery in Chicgo tht used dministrtive records lso found very modest nd mostly nonsignificnt effects on school, rrest, nd helth outcomes. This study compred 18,347 children in fmilies offered vouchers vi the lottery with 48,263 chil - dren in fmilies who pplied for vouchers but did not receive them (Jcob, Kpustin, nd Ludwig, 2014). There were mixed results in the nonexperimentl Three-City Study (Coley et l., 2013), 6-yer longitudinl study of 2,400 low-income fmilies from moderte nd high-poverty neighborhoods in Boston, Msschusetts, Chicgo, Illinois, nd Sn Antonio, Texs. This study found ssocitions between housing qulity nd children s behvior but inconsistent ssocitions of outcomes with residentil mobility nd housing ssistnce. The study explined effects of housing qulity sttisticlly by ssocitions of qulity housing with lower prentl stress. In the Fmily Options Study, we might expect lrger effects on child outcomes, t lest by 36 months. Our smple is substntilly poorer nd less stble thn the poor fmilies in other studies, nd fmilies in our study fce the dditionl chllenge of homelessness. Thus the contrst between fmilies who do nd do not receive subsidies my be lrger. If subsidies hve lrger effects on reducing mobility, improving ccess to better housing, nd reducing prentl stress in our smple thn in other studies, then subsidies my be more likely to enhnce child outcomes. To the extent tht permnent subsidies provide greter stbility or reduce prentl stress more thn temporry subsidies, effects of SUB my be lrger thn effects of CBRR. Adult well-being. Reductions in instbility produced by SUB or CBRR might reduce dult psychologicl distress, but effects on other mesures of dult well-being, such s substnce use problems nd domestic violence, seem less likely. Proponents might rgue tht shortening shelter stys nd restoring fmilies to conventionl housing would minimize the hrm of homelessness. Prticulrly in the cse of SUB, ccess to permnent subsidies might reduce finncil stress nd thus enhnce dult well-being. In the study of Effects of Housing Vouchers on Welfre Fmilies, housing subsidies re - duced poverty nd hrdship, including residentil crowding nd homelessness, nd incresed spending on food. Qulit - tive findings suggest tht the subsidies reduced fer of home - lessness, even mong fmilies who hd not been homeless t the outset. The smller nd shorter term subsidies of CBRR would be expected to led to more modest reductions in stress. In fct, some qulittive evidence from erly interviews with fmilies in this study shows tht the uncertinty round continution of the CBRR subsidies cretes nxiety for fmilies while they re in CBRR (Fisher et l., 2014). Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 25

58 Chpter 3. Hypotheses About Intervention Effects (Conceptul Frmework) Self-sufficiency. Theory provides little reson to believe tht ongoing housing subsidies would enhnce employment rtes or led to gins in incomes. Rther, proponents of SUB see subsidies s prt of the socil sfety net tht supports both people who cnnot work nd low-wge workers. With subsidies viewed s prt of the socil sfety net, two considertions from neoclssicl economic theory clerly predict tht SUB should lower work effort. First, the vlue of SUB is equivlent to more income, implying less of need to work nd therefore less work (wht neoclssicl economic theory clls the income effect ). Second, the rent formul implies tht the vlue of SUB flls with income (tht is, the dditionl net income is 30 percent lower for fmilies receiving rentl ssistnce thn for other fmilies). In fct, combined with low hourly py, pyroll txes, nd similr benefit formuls for other progrms (for exmple, the Supple - mentl Nutrition Assistnce Progrm, or SNAP, nd TANF) nd costs of work (trnsporttion, childcre), the net increse in income from work will often be low nd sometimes neg - tive. Lower net py for ech hour of work would lso be expected to lower work effort (the substitution effect for neoclssicl economic theory). Evidence tht ernings disregrds in welfre progrms re not effective inducements to incresed work effort or income suggests tht the substitution effect my not be importnt, perhps becuse of imperfect recipient understnding (Mtsudir nd Blnk, 2013). Consistent with this theory, study in Chicgo tht took dvntge of n lloction of housing vouchers by lottery found persistent reduction in work effort by working-ge people (Jcob nd Ludwig, 2012). A study tht used rndom ssignment in lrger number of loctions cme to different conclusion, but this study ws bsed on nrrow set of households those tht were receiving TANF or were TANF eligible. The Welfre to Work Voucher demonstrtion pro - grm ws supposed to link housing vouchers to employment services for TANF nd TANF-eligible fmilies, but pressures to expend funds quickly ment tht, in prctice, there ws little link, so the study cn be viewed s pure test of the effect of vouchers, not of vouchers linked with services. (The nme of the evlution ws chnged to Effects of Housing Vou - chers on Welfre Fmilies.) The study found tht receipt of vouchers reduced employment in the first few qurters fter rndom ssignment but hd no effect over 3.5 yers. Vouchers lso incresed use of public ssistnce (Mills et l., 2006). Neither the Chicgo study nor the voucher study focused on people experiencing homelessness or on people with disbilities, nd it is hrd to predict whether vouchers would hve greter or lesser effect on work effort for people with greter disdvntges tht is, people more like the fmily heds in this study. In n observtionl study of veterns with mjor psychitric disorder or substnce buse dis - order, the (mostly mle) veterns who were housed with vouchers worked less nd hd lower incomes thn those who hd been ble to rent housing independently without vouchers in the period following rndom ssignment (Tsi, Ksprow, nd Rosenheck, 2011). The uthors suggest tht the vouchers reduced incentives to work, but the non - experimentl nture of the nlysis mkes the finding inconclusive. Proponents of CBRR might expect tht it would do more thn SUB to enhnce employment nd ernings. One gol of HPRP under which most CBRR ws funded ws to enble fmilies to chieve housing stbility by ttining self-sufficiency (HUD, 2009). The fct tht CBRR funds were time-limited my hve been further impetus to fmilies to generte sufficient income so they could sustin housing without help fter CBRR ssistnce ended. A qusiexperimentl study in Wshington Stte provides support for the impct of CBRR on self-sufficiency. Among workingge clients of the Deprtment of Socil nd Helth Services who becme homeless, the 1,537 who received rpid rehousing were more likely to be employed nd hd higher ernings in the yer following ssistnce thn crefully mtched homeless comprison group who did not receive ssistnce from ny shelter or trnsitionl housing progrm (Myfield, Blck, nd Felver, 2012). Thus CBRR my led to better income nd employment outcomes thn SUB nd UC Predictions Regrding Costs Proponents of CBRR note tht by giving fmilies only s much help s they need to return to permnent housing (Culhne, Metrux, nd Byrne, 2011), CBRR should reduce costs reltive to open-ended subsidies provided by SUB, permitting more fmilies to be served. Becuse the dily cost of remining in emergency shelter is likely to be higher thn the dily cost of CBRR, quick exits from shelter might mke CBRR less costly thn UC mesured over n equivlent time period. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 26

59 Chpter 3. Hypotheses About Intervention Effects (Conceptul Frmework) Hypotheses for Pirwise Comprisons Involving SUB, CBRR, nd UC In this section, we stte hypotheses bout SUB nd CBRR reltive to ech other nd to UC (see box), where UC consists of shelter nd whtever other resources fmilies cn cquire on their own. Hypotheses for Comprisons Involving SUB, CBRR, nd UC SUB Versus UC H1. Reltive to UC, SUB will reduce shelter use nd improve housing stbility nd my improve fmily preservtion, dult well-being, nd child well-being. CBRR Versus UC H2. Reltive to UC, CBRR will reduce shelter use nd my improve housing stbility, employment nd ernings, fmily preservtion, dult well-being, nd child well-being. It will reduce the length of the shelter sty t the time of study entry nd my be less costly. SUB Versus CBRR H3. Reltive to CBRR, SUB will reduce shelter use nd improve housing stbility nd my improve dult nd child well-being. H4. Reltive to SUB, CBRR will reduce the length of the shelter sty t the time of study entry nd will be less costly. It my improve employment nd ernings. We defer hypotheses involving PBTH to the next section, fter introducing the conceptul rtionle for tht intervention. Centrl predictions from ech perspective re indicted with will. Where evidence is quite limited or hypothesized effects re indirect, we use my. Note tht the Fmily Options Study exmines multiple indictors of ech of the key domins of housing stbility, self-sufficiency, fmily preservtion, nd dult nd child well-being, long with mesures of costs. Some hypotheses refer to entire domins, nd other hypotheses refer to specific indictors (such s homelessness or employment) within broder domins. In prticulr, predictions bout the reduction of homelessness include (mong other mesures) the confirmtory outcome of t lest 1 night homeless, doubled up in the pst 6 months (from survey dt), or t lest 1 night in n emergency shelter in the pst 12 months (primrily from HMIS records). 33 Outcome mesures re described in Chpter 5 nd Appendix B. 3.4 Conceptul Rtionle for PBTH Proponents of PBTH hve different understnding of the origins of fmily homelessness nd the proper role of the homeless service system thn proponents of SUB nd CBRR. Although the housing mrket is difficult for poor fmilies, most poor fmilies do not experience homelessness. Proponents of PBTH emphsize tht mny fmilies who do become homeless hve dditionl brriers tht mke it hrd for them to secure nd mintin housing. Thus, housing subsidies lone my be insufficient to ensure housing stbility nd other desirble outcomes (for exmple, Bssuk nd Geller, 2006). Fmily needs my rise from poverty, helth, disbility, or other problems tht led to homelessness to begin with or from the disruptive effects of homelessness on prents nd children. Proponents of PBTH believe tht by ddressing these brriers nd needs in supervised residentil setting, PBTH lys the best foundtion for ongoing stbility. For exmple, the high prevlence of domestic violence nd trum in homeless fmilies lives requires protected time nd trum-informed services in structured residentil setting for recovery. Fmilies struggling with mentl helth nd substnce use problems similrly need help to ddress these issues in n environment tht supports or requires sobriety. Brriers to work cn be overcome with job trining nd ssistnce with job serches, helping fmilies to increse employment nd ernings. Fmilies hving trouble with budgeting nd finncil mngement need clsses to enhnce these skills nd help with credit repir before they re redy to live independently. Services nd clsses lso support prenting. Bsed on fmily needs, cse mngers coordinte the services (on site or by referrl) to ly the essentil groundwork for lter independence. Different PBTH progrms focus on different issues, but ll of the progrms provide supportive services designed to reduce brriers to housing, enhnce prents well-being, nd bolster their bility to mnge in ordinry housing fter they leve progrms (Burt, 2010). Prctitioners gols for PBTH, s documented in the literture (for exmple, Burt, 2006), thus extend beyond housing stbility to dult well-being nd spects of fmily self-sufficiency. Although some PBTH progrms provide services directly to children, fmily preservtion nd child outcomes re usully seen s more distl outcomes, s shown in Exhibit 3-2. Note tht fmilies in emergency shelter often receive services similr to those provided in PBTH, which could reduce the estimted impcts of ll study interventions. We nticipte, however, tht exposure to these services will be briefer in UC thn in PBTH. 33 See Chpter 4 for detils bout the HMIS dt used in the study. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 27

60 fmily preservtion nd child outcomes re usully seen s more distl outcomes, s shown in Exhibit 3-2. FAMILY OPTIONS STUDY Note tht fmilies in emergency shelter Chpter often receive 3. Hypotheses services similr About to those Intervention provided Effects in PBTH, (Conceptul which Frmework) could reduce the estimted impcts of ll study interventions. We nticipte, however, tht exposure to these services will be briefer in UC thn in PBTH. Exhibit 3-2. Conceptul Exhibit 3-2. Intervention Conceptul Model Intervention for PBTH Model for PBTH PBTH = project-bsed trnsitionl housing. UC = usul cre. vs. = versus. PBTH = project-bsed trnsitionl housing. UC = usul cre. vs. = versus Predictions Regrding Housing In the study of Life After Trnsitionl Housing for Homeless Stbility Predictions Regrding Housing Stbility Fmilies, Burt (2010) interviewed 195 fmilies considered Proponents of Proponents PBTH rgue of tht PBTH trnsitionl rgue tht housing trnsitionl enhnces housing enhnces successful stbility grdutes by reducing of trnsitionl brriers to housing housing. progrms (both Multiple descriptive studies of trnsitionl housing or progrm-bsed of housing with nd supportive scttered-site services models) nd without in five communities stbility by reducing brriers to housing. Multiple descriptive cross five sttes. Fmilies were interviewed t the time of studies of trnsitionl housing or of housing with supportive progrm exit nd yer lter. Of these successful grdutes, services nd without time limit show strong housing out - 86 percent exited to their comes (s reviewed by Bssuk nd Geller, 2006; Bssuk et l., Short-Term own plce, Impcts including pg. some 40 who 2014). These studies rrely hve comprison groups, however, much less rndom ssignment, so it is hrd to know wht would hve hppened in the bsence of the services nd therefore the net impct of the intervention versus wht would hve hppened in the bsence of the intervention. Although not study of PBTH, the multisite Homeless Fm - ilies Progrm is one of the lrgest studies of combined housing nd services intervention (Rog nd Gutmn, 1997). In the Homeless Fmilies Progrm, in six sites with followup dt, 88 percent of 601 fmilies who received cse mngement services in ddition to housing subsidies remined in permnent housing for up to 18 months. In the one smll nonexperimentl study with counterfctul, 84 high-risk homeless fmilies who received intensive cse mngement plus housing subsidies were compred with 85 high-risk fmilies who received subsidies with less intensive services. At 1 yer, the fmilies receiving intensive cse mngement were slightly more likely to be in their originl prtment. Cse mngement services were prtilly confounded with subsidy type, however, which ws the strongest predictor of housing outcomes (Weitzmn nd Berry, 1994). styed in the sme plce they hd lived in while in trnsitionl housing. (This result will not be possible for the project-bsed trnsitionl housing PBTH to which this study rndomized fmilies.) Of those with followup dt, 60 percent were in their own plce without move for the yer fter exit, nd only 4 fmilies returned to homelessness. On the other hnd, move into permnent housing ws often mjor criterion for successful grdution, so numbers would hve been lower hd ll progrm prticipnts been included. Prticipt - ing progrms estimted tht 77 percent of the fmilies served were successful (Burt, 2010; Appendix F nd Chpter 5). The SHIFT study compred 120 fmilies recruited from 18 trnsitionl housing progrms with nonequivlent group of 129 fmilies recruited from 14 emergency shelters (nd 43 from PSH not considered further here) in four communities in Upstte New York (Hyes, Zonneville, nd Bssuk, 2013). The study counted fmilies s being stble if they did not move or if they moved only once, with subsidy. At 15 months fter progrm entry, fmilies from trnsitionl housing pro - grms were more stble (43 percent) thn fmilies from emergency shelters (14 percent), but, t 30 months, the groups did not differ (44 percent for trnsitionl housing, Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 28

61 Chpter 3. Hypotheses About Intervention Effects (Conceptul Frmework) 47 percent for emergency shelter). The low rtes of long-term housing stbility in both groups my be ssocited with the reltively high needs of the smple 48 percent of both groups of mothers met criteri for PTSD; 56 percent of women in emergency shelter nd 48 percent of those in trnsitionl housing hd lredy lived prt from one of their children t progrm entry. Proponents of PBTH conceptully would expect this intervention to foster higher rtes of long-term stbility thn UC, CBRR, nd possibly even SUB, becuse PBTH seeks to ddress brriers to housing stbility nd put fmilies on trck for better employment nd ernings. As noted previously, pro - ponents of SUB nd CBRR mke the opposite prediction. The only empiricl studies compring outcomes from PBTH ginst CBRR fvored CBRR. In studies bsed on lrge smples in Georgi (Rodriguez, 2013) nd on seven CoCs in four sttes (Gle, 2012), returns to homelessness were lower for CBRR cross 2-yer nd 1-yer followup periods, respectively. Agin, prticipnts were not rndomized to interventions, so results could reflect preexisting differences rther thn cusl effects of progrms. Predictions reltive to SUB re complicted by the fct tht PBTH progrms frequently ttempt to enroll their fmilies in housing subsidy progrms when they exit from trnsitionl housing; 53 percent of the fmilies exiting trnsitionl housing successfully in the Life After Trnsitionl Housing for Homeless Fmilies study (Burt, 2010) nd 78 percent in the Sound Fmilies Inititive (Northwest Institute for Children nd Fmilies, 2007) received subsidies. In the Life After Trnsitionl Housing for Homeless Fmilies study, these subsidies were ssocited with greter residentil stbility nd lso stbility in fmily composition. It seems likely, conceptully, tht PBTH followed by subsidy would led to more stbility thn PBTH lone. Becuse fmilies rndomized to PBTH in our study were not further rndomized to receive or not receive subsidies t the end of PBTH, we cnnot test this prediction. In the short term (t 18 months), predictions re further complicted by the fct tht PBTH cn lst up to 2 yers nd CBRR up to 18 months. To the extent tht fmilies remin in PBTH or CBRR interventions t the time of followup or hve only recently left them, little difference my emerge between these two interventions with respect to stbility t 18 months. Furthermore, the fct tht fmilies must move out of PBTH progrms might inflte the number of moves t 18 months. PBTH proponents would expect their efforts to strengthen fmilies to hve enduring benefits fter the sty in trnsitionl housing hs ended Predictions Regrding Self-Sufficiency Perhps the strongest evidence for the success of PBTH is the self-sufficiency outcomes of eduction nd employment. Two lrge-scle studies of trnsitionl housing found substn - til increses in employment between progrm entry nd progrm exit mong those fmilies who completed the progrm. The Sound Fmilies Inititive served 1,487 fmilies in three counties in Wshington Stte, with n verge sty of slightly more thn 12 months. Full-time cse mngers hd verge cselods of 15 fmilies nd lso offered referrls to offsite providers. Employment doubled from 22 percent t entry to 45 percent t exit, receipt of TANF decresed from 67 to 46 percent, nd 48 percent of fmilies incresed their income during this sme period of time (Northwest Institute for Children nd Fmilies, 2007). In the Life After Trnsitionl Housing for Homeless Fmilies study, smple of mothers deemed successful by their trnsitionl housing progrms incresed rtes of employment substntilly (from 18 to 61 percent) from progrm entry to progrm exit, gin with medin sty of bout yer (Burt, 2010). Mothers in two of five fmilies completed voctionl, trde, or business progrm; n dditionl 12 percent were involved in voctionl or business progrms when they left; nd hndful erned high school diplom or generl eductionl development, or GED, took college clsses, or even completed college degree. Although neither study hd comprison group nd lthough fmilies who exited the progrms rther thn successfully completing them were excluded, it seems unlikely tht chnges of this mgnitude would hve occurred without the progrms. On the other hnd, few fmilies rised their incomes bove the poverty threshold. Burt (2010: 90) concluded tht, despite the gins in eduction nd employment, period of time in trnsitionl housing progrm does not chnge the bsic relity of poor, reltively undereducted mothers erning power. Bssuk et l. (2014: 471) noted tht, even where prticipnts in housing nd service inter - ventions incresed employment levels over time, employ - ment tended to be chrcterized by multiple job chnges, periods of unemployment, nd low-wge prt-time work Predictions Regrding Adult Well-Being Adult well-being is centrl focus of prctitioners in mny PBTH progrms, s noted previously in the conceptul rtionle for the intervention. Thus, proponents would expect to see improvements, prticulrly in res of mentl helth nd substnce use. In the Life After Trnsitionl Housing for Homeless Fmilies study, dult well-being, especilly relted to mentl helth, improved between entry nd exit from trnsitionl housing (Burt, 2010). Agin, there ws no Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 29

62 Chpter 3. Hypotheses About Intervention Effects (Conceptul Frmework) counterfctul, so chnges seem hrd to ttribute to trnsitionl housing rther thn to the pssge of time, nd few chnges in child well-being occurred. Bssuk et l. (2014) found tht methodologicl flws precluded cler conclusions bout dult well-being in the studies they reviewed Predictions Regrding Distl Outcomes Most PBTH progrms focus primrily on the well-being of prents, bsed on the ssumption tht prentl well-being is centrl to child welfre. Some PBTH progrms lso provide fmilies with or link fmilies to services for children (especilly childcre or fter school services). Effects of PBTH on child outcomes nd fmily preservtion re thus likely to be indirect vi improvements in fmily stbility nd prents welfre nd prenting skills. Child well-being. Periods of homelessness nd the circumstnces tht led to them re often chotic s fmilies move from plce to plce. PBTH progrms provide stbility nd structure for the fmilies while they reside there, potentilly reducing school mobility. In the Sound Fmilies Inititive evlution, the percentge of school-ge children ttending two or more schools within single school yer dropped from more thn one-hlf (53 percent) t intke to less thn one-sixth (17 percent) t exit (Northwest Institute for Children nd Fmilies, 2007), nd similr reductions in school mobility were evident in the Life After Trnsitionl Housing for Homeless Fmilies study. Structured schedules, long with encourgement or monitoring by stff, might improve children s school ttendnce nd completion of homework, t lest while fmilies remin in the progrm. Supports for prenting, including reductions in stress, prenting clsses, nd modeling of disciplinry prctices, could enhnce prenting. Although positive effects on children re plusible, little empiricl evidence supports them, becuse of the observtionl designs nd lck of comprison groups in studies of trnsitionl housing nd other housing nd service interventions (Bssuk et l., 2014). Fmily preservtion. Trnsitionl housing cn lso help with fmily reunifiction. In the study of Life After Trnsitionl Housing for Homeless Fmilies, 42 percent of children living prt from their mother t trnsitionl housing entry rejoined the fmily during trnsitionl housing (Burt, 2010). Cregivers nd stff in the Sound Fmilies Inititive described trnsitionl housing progrms s centrl in helping to ensure prent-child reunifiction (Northwest Institute for Children nd Fmilies, 2007: 9), lthough the report did not quntify how often reunifiction hppened. Bssuk nd Geller s (2006) review suggests tht cse mngement services like those provided in PBTH progrms contribute to fmily preservtion nd reunifiction. How might tht occur? Stff in progrms my secure needed services for fmilies, model good disciplinry prctices, nd offer instruction in prenting tht should reduce seprtions to begin with. Stbilizing fmilies in residentil progrm my be enough to enble children who re with other fmily members to return to their prents. In the cse of more forml seprtions, PBTH stff my work with fmilies nd child protective services workers on requirements for reunifiction, such s supervised visits. Becuse stff re in position to monitor fmilies in PBTH, protective services workers my be more willing to llow children to sty with fmilies in situtions they regrd s potentilly risky or to return children who hve been seprted thn would be the cse in unsupervised settings. On the other hnd, PBTH my led to pproprite out-of-home plcements of children becuse of heightened PBTH stff involvement nd greter likelihood of observing children t risk in disorgnized fmilies. Bsing our ssessment on the Life After Trnsitionl Housing for Homeless Fmilies study (Burt, 2010), we predict net reduction in seprtions Predictions Regrding Costs The combintion of time-limited housing nd services provided by PBTH is likely to be more costly thn the timelimited subsidies provided by CBRR. The trdeoff between PBTH s provision of services (mking PBTH more costly thn SUB) nd the time limit on housing ssistnce (mking PBTH less costly thn SUB) mkes predictions relted to the overll costs of ech intervention more difficult. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 30

63 Chpter 3. Hypotheses About Intervention Effects (Conceptul Frmework) Hypotheses for Pirwise Comprisons Rooted in the predictions bout PBTH, the study tem developed five hypotheses for pirwise comprisons involving PBTH (see box). 34 Hypotheses for Comprisons Involving PBTH PBTH Versus UC H5. Reltive to UC, PBTH will reduce shelter use nd improve housing stbility, employment, ernings, eduction, nd dult well-being nd my improve fmily preservtion nd child well-being. PBTH Versus SUB H6. (From the perspective of PBTH proponents) reltive to SUB, PBTH will improve employment, ernings, eduction, nd dult well-being nd my improve longterm housing stbility, fmily preservtion, nd child wellbeing. (Stbility effects my not emerge t 18 months.) H7. (From the perspective of SUB proponents) reltive to PBTH, SUB will reduce shelter use nd improve housing stbility nd my improve fmily preservtion, dult wellbeing, nd child well-being. PBTH Versus CBRR H8. (From the perspective of PBTH proponents) reltive to CBRR, PBTH will improve employment, ernings, eduction, nd dult well-being nd my improve longterm stbility, fmily preservtion, nd child well-being. (Stbility effects my not emerge t 18 months.) H9. (From the perspective of CBRR proponents) reltive to PBTH, CBRR will reduce shelter use nd my improve housing stbility, fmily preservtion, dult well-being, child well-being, employment, nd ernings. 3.5 Wht Works for Whom? One gol of this study is to identify whether prticulr interventions re more effective for some groups of homeless fmilies thn for others. Becuse of the number of fmily chrcteristics tht could led to differentil effects of interventions, we confine nlyses to the exmintion of two brod ctegories of fmily chrcteristics, which we summrize in indices of psychosocil chllenges nd housing brriers. 35 The psychosocil chllenges index is count of the number of psychosocil chllenges reported by fmilies t the bseline survey immeditely before rndom ssignment, including helth nd mentl helth conditions, substnce use problems, PTSD, intimte prtner violence, felony conviction, history of foster cre or institutionl plcement s child, nd disbil - ity of the prent or child. Severl of these fctors predicted residentil instbility (cross progrm type) in the SHIFT study (Hyes, Zonneville, nd Bssuk, 2013). The housing brriers index is count of 15 potentil brriers including unemployment, lck of income, pst evictions or lese violtions, lck of trnsporttion, nd fmily composition tht fmilies reported t the time of the bseline survey to be t lest smll problems in trying to find housing. A similr list of brriers ws ssocited with incresed returns to homelessness for fmilies in the evlution of the Rpid Re-Housing for Homeless Fmilies Demonstrtion progrm (Finkel et l., forthcoming). The brriers list ws purged of items relted to disbility nd criminl justice tht overlpped with psychosocil chllenges. 34 Most mesures of housing stbility reflect either the time of the followup survey or the 6-month period of time before the followup survey, which ws conducted n verge of 20 months fter rndom ssignment. When mesuring outcomes for PBTH, complicting fctor is tht HUD progrm rules consider fmily to be homeless during sty in trnsitionl housing, wheres the outcome mesures used in this study consider only shelter use or unsheltered homelessness s n episode of homelessness. The study of impcts fter 36 months will not hve this problem, becuse virtully ll fmilies given priority ccess to PBTH will hve left trnsitionl housing by tht time. Further, wheres outside the context of this study, CBRR might reduce initil shelter stys reltive to PBTH (proponents set gol of getting fmilies bck into hosing in 30 dys), in this study, fmilies were not given priority ccess to PBTH unless the study identified n opening for which they ppered eligible. Fmilies ssigned to CBRR hd to identify housing unit in which to use the subsidy. Thus, it is plusible tht PBTH would led to fster exits from shelter in this study. 35 The dt collection nd nlysis pln clled for mesuring brriers to housing stbility by developing n empiricl index of fmily chrcteristics tht predicted the centrl study outcome (t lest 1 night homeless or doubled up in the pst 6 months or in shelter in the pst 12 months) in the UC group. Cndidte vribles for the index were those tht predicted returns to shelter in previous studies, such s young ge of fmily heds. The nlysis ws not successful in using those vribles to predict the outcome, however, perhps becuse of the multiprt chrcter of the outcome. Thus, for the nlyses of wht works for whom, we used the chrcteristics reported by the bseline survey respondent to be housing brriers rther thn fmily chrcteristics tht ctully predicted housing instbility. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 31

64 Chpter 3. Hypotheses About Intervention Effects (Conceptul Frmework) We expect fmilies who hve more psychosocil chllenges or housing brriers to hve worse outcomes thn other fmilies. The question in this section is whether one or nother of the interventions is prticulrly effective for fmilies with greter chllenges or brriers. Exhibit 3-3 illustrtes wht this result might look like for the effect of housing brriers on the outcome of residentil stbility. In the exhibit, ssign - ment to Intervention A versus Intervention B hs little effect on residentil stbility when fmilies fce few housing brriers (to the left of the exhibit); tht is, the lines for Intervention A nd Intervention B re close together. When fmilies hve high levels of housing brriers (to the right), however, Intervention A hs big impct reltive to Intervention B; tht is, the lines for Intervention A nd Intervention B re fr prt. Put differently, Intervention A mitigtes the dverse effects of housing brriers on the outcome of residentil stbility. Regrding the question of wht works for whom, proponents of SUB suggest tht it is likely to hve its lrgest effects in comprison with ll other interventions for fmilies who hve higher levels of housing brriers. By comprison with CBRR nd UC, SUB might lso hve lrger effects for fmilies fcing dditionl psychosocil chllenges. In nonexperimentl study in New York City, vriety of chllenges nd brriers predicted initil shelter entry mong fmilies receiving public ssistnce, but only housing subsidies nd mternl ge predicted housing stbility 5 yers lter mong the group tht experienced homelessness. Housing subsidies seem to hve countercted ll other risk fctors (Shinn et l., 1998). In the Effects of Housing Vouchers on Welfre Fmilies study, reductions in homelessness were especilly lrge for two vulnerble subgroups: those in which the hed of household ws unemployed nd those who were nering the end of eligibility for TANF t the time of rndom ssignment (Mills et l., 2006). Differentil effects for SUB re most likely for stbility outcomes but might lso extend to more distl outcomes. PBTH is the only intervention tht ddresses fmilies psycho - socil needs directly. Thus, we might expect PBTH to hve lrger effects reltive to ll other interventions for fmilies with higher scores on the psychosocil chllenges index. Pro - ponents of PBTH dditionlly rgue tht services ddress other brriers to housing, so PBTH might be more successful for fmilies with higher housing brriers. In the Life After Trnsitionl Housing for Homeless Fmilies study, personl chrcteristics mttered little to outcomes following trnsitionl housing stys, suggesting tht trnsitionl housing effectively countered risks ssocited with such chrcteristics (Burt, 2010). Differentil effects re hypothesized for ll three primry outcomes of PBTH stbility, self-sufficiency, nd dult well-being nd might extend to more distl out - comes. Exhibit 3-3. Hypotheticl Exmple in Which the Impct on the Outcome of Residentil Stbility of Intervention A Reltive to Intervention B Is Lrger for Fmilies With High Housing Brriers High Intervention A Intervention B Residentil stbility Low Low housing brriers High housing brriers Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 32

65 Chpter 3. Hypotheses About Intervention Effects (Conceptul Frmework) The rtionle for CBRR suggests tht, for most fmilies, only short-term support is necessry to return to housing stbility. Directions given to communities by HUD regrding HPRP tht funded most of the CBRR in this study suggested tht CBRR should go to fmilies who needed it to restore housing stbility but for whom reltively modest intervention would be sufficient. This directive implies belief tht CBRR my be less successful for fmilies with high scores on the housing brriers nd psychosocil chllenge indices. Some proponents of CBRR rgue, however, tht, while unble to prevent return to homelessness for some fmilies, CBRR might, over time, reduce the number nd durtion of episodes of homelessness. The Fmily Options Study ws not designed to test the impct of repeted offers of CBRR to the sme fmily. Differentil effects of the ctive interventions reltive to UC should be observble t 18 months. Differentil effects of the ctive interventions reltive to ech other re most likely to emerge fter ll fmilies hve exited the CBRR nd PBTH progrms. As result, differentil effects my be more visible t the time of the 36-month followup. In the cse of residentil stbility outcomes, contrsts between the ctive interventions t 18 months will be reduced to the extent tht fmilies continue to prticipte in PBTH nd CBRR progrms from 13 to 18 months fter rndom ssignment Predictions Regrding Wht Works for Whom The study tem developed two hypotheses bout the reltive effects of the interventions on fmilies with vrying levels of housing brriers nd psychosocil chllenge (see box). Hypotheses About Which Interventions Work for Whom From the Perspective of SUB Proponents H10. Reltive to ll other interventions, SUB will hve lrger effects on stbility for fmilies with higher scores on the housing brriers index nd my hve lrger effects on stbility for fmilies with higher scores on the psychosocil chllenges index. Differentil effects my extend to self-sufficiency, dult well-being, child well-being, nd fmily preservtion. From the Perspective of PBTH Proponents H11. Reltive to ll other interventions, PBTH will hve lrger effects on stbility, self-sufficiency, nd dult wellbeing for fmilies with higher scores on the psychosocil chllenges index nd my hve lrger effects on these outcomes for fmilies with higher scores on the housing brriers index. Differentil effects my extend to child well-being nd fmily preservtion. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 33

66 CHAPTER 4. DATA SOURCES AND METHODOLOGY This chpter describes the dt sources nd methodology employed in nlysis of the 20-month impct estimtes for the Fmily Options Study. The first section describes the dt sources used to (1) mesure pro - grm use by study fmilies, (2) nlyze the impcts of study interventions on fmilies, nd (3) describe study interventions nd their costs. The second section provides n overview of the method used to estimte the impcts of interventions. Additionl detil bout the impct methodology is provided in Appendix C. 4.1 Dt Sources The study hs collected dt from wide rnge of sources, including study fmilies, intervention providers, nd dministrtive dt systems. Exhibit 4-1 shows the sources of dt used in this report nd provides informtion bout how they were collected nd their content. Key spects of the dt collection re described in the subsections tht follow. Exhibit 4-1. Dt Sources Used in the Report (1 of 2) Dt Source Collection Process Dt Source Collects or Mesures From study implementtion Rndom ssignment enrollment dt (n = 2,282) From study fmilies Bseline survey (n = 2,282) 6- nd 12- month trcking surveys (n = 1,671; n = 1,632) 18-month followup dult survey (n = 1,857) 18-month followup child ssessments Recorded in web-bsed enrollment nd rndom ssignment tool, bsed on informtion entered by field interviewer nd point-in-time intervention vilbility In-person survey (40 minutes) conducted immeditely before rndom ssignment Completed for the full smple of fmilies rndomly ssigned Telephone survey (10 minutes) conducted 6 nd 12 months fter rndom ssignment In-person or telephone survey (60 minutes) conducted t lest 18 months fter rndom ssignment In-person child ssessments (50 minutes) conducted for focl children who were ges 3 yers, 6 months to 6 yers, 11 months Collection ttempted only if fmily hed responded to followup dult survey Nme, dte of birth, nd Socil Security number of fmily hed nd spouse or prtner Eligibility screening responses Intervention vilbility t rndom ssignment Rndom ssignment result Demogrphic chrcteristics Preshelter housing Housing brriers Homelessness history Employment Fmily composition Income nd income sources Fmily hed: physicl helth Fmily hed: mentl helth nd experiences of trum Fmily composition Current housing sttus Use of homeless nd housing progrms Current housing sttus Experience of homelessness Use of homeless nd housing progrms Housing qulity nd ffordbility of current unit Employment nd ernings Income nd income sources Mteril hrdship Fmily composition nd preservtion Adult well-being Child well-being (for up to two focl children) Receipt of services Verbl bility (Woodcock-Johnson III letter-word identifiction test; n = 876) Mth bility (Woodcock-Johnson III pplied problems test; n = 846) Self-regultion (Hed Toes Knees Shoulders ssessment; n = 780) Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 34

67 Chpter 4. Dt Sources nd Methodology Exhibit 4-1. Dt Sources Used in the Report (2 of 2) Dt Source Collection Process Dt Source Collects or Mesures From study fmilies (continued) 18-month followup child survey (n = 945) From study intervention providers Enrollment verifiction dt Progrm informtion Progrm cost informtion From dministrtive dt systems Homeless Mngement Informtion System (HMIS) HUD Public nd Indin Housing Informtion Center (PIC) HUD Tennt Rentl Assistnce Certifiction System (TRACS) From combintion of sources Progrm usge dt In-person or telephone survey (30 minutes) conducted for focl children who were ges 8 to 17 yers Collection ttempted only if fmily hed responded to followup dult survey Study tem verified (by telephone nd e-mil) whether fmilies enrolled in the progrms to which they were referred Conducted from September 2010 to September 2012 Study tem conducted site visits nd stff interviews Conducted from June 2011 to April 2012 Study tem conducted site visits nd stff interviews Collected udited expense sttements, progrm budgets, stffing lists, prtner commitment letters, nd progrm stff estimtes of costs not reflected in expense sttements Conducted from November 2012 to August 2013 Individul-level records collected from community nd government dministrtors of the Homeless Mngement Informtion System (HMIS; one or more per site) Individul-level dt collected from HUD Individul-level dt collected from HUD Combines dt from seven sources: enrollment verifiction; 6-, 12-, nd 18-month surveys; HMIS; HUD PIC; nd the Tennt Rentl Assistnce Certifiction System (TRACS) Mentl helth Experiences of trumtic events Substnce use School effort Arrests or police involvement Use of ssigned intervention Provider informtion Chrcteristics of housing ssistnce Chrcteristics of services Overhed costs Rentl ssistnce costs Fcility opertions costs Supportive services costs Cpitl costs Prticiption in homeless ssistnce progrms covered in HMIS (including emergency shelter, rpid re-housing, trnsitionl housing, nd permnent supportive housing) Receipt of housing ssistnce through HUD s Housing Choice Voucher (HCV) nd public housing progrms Receipt of housing ssistnce through projectbsed Section 8 progrms Prticiption in seven types of homeless nd housing ssistnce progrms (by clendr month fter rndom ssignment) Notes: All surveys conducted with fmily hed collected or updted fmily contct informtion for trcking purposes. Additionl informtion bout dt collection from study fmilies provided in Appendix A. Additionl informtion bout progrm cost dt collection provided in Appendix G Bseline Dt Collection Locl field interviewers enrolled fmilies into the study in person t the emergency shelters where the fmilies were stying. Enrollment begn in September 2010 nd ws completed in Jnury During enrollment, the interviewer would first inform the fmily bout the study. If the fmily consented to prticipte in the study, the interviewer would then sk eligibility screening questions for progrms tht hd vilble progrm slots. Finlly, the interviewer would dminister the bseline survey using Computer-Assisted Personl Interviewing, or CAPI, softwre. All fmily heds who consented to prticipte in the study completed the bseline survey before rndom ssignment. In fmilies with only one dult present, tht individul ws interviewed. For fmilies heded by couples, the study tem interviewed women. There were two resons for this preference: (1) Some homeless ssistnce progrms exclude men, nd in cses of fmily seprtions the children re more likely to remin with the mother; nd (2) some outcome mesures such s psychologicl distress hve different distributions for men nd women in the popultion t lrge, so this preference results in hving greter homogeneity in the smple Followup Dt Collection The followup dt collection effort ws conducted from July 2012 through October The study tem ttempted to contct fmilies for the study s followup survey beginning in the 18th month fter rndom ssignment. The medin time from rndom ssignment to the followup survey ws 20 months. The followup period referred to in the report thus covers period of 20 months fter rndom ssignment, lthough the survey is sometimes referred to s the 18-month followup survey. The dt collection consisted of three components. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 35

68 Chpter 4. Dt Sources nd Methodology Followup dult survey. Followup child ssessments. Followup child survey. The study ttempted to complete the dult survey with ll 2,282 fmily heds. For fmilies heded by couples, the sme dult interviewed t bseline ws interviewed t followup. Up to two focl children per fmily were rndomly selected during the dministrtion of the dult survey. 36 Informtion bout the well-being of the focl children ws collected from the prent in the dult survey nd directly from ppropritely ged children in either the child ssessments or the child survey. The dult survey needed to be completed before ttempts were mde to complete child ssessments nd the child survey with focl children (both so prentl consent could be given nd so focl children could be selected). Focl children needed to 1. Hve been either identified in the bseline survey s prt of the fmily or born fter rndom ssignment. 2. Hve been ges 1 to 17 yers t the time of the followup dult survey. 3. Hve been with the fmily t the time of the followup dult survey or hd enough contct with the fmily hed so tht their prent (the fmily hed) ws knowledgeble bout key spects of their lives. 37 The focl child selection process oversmpled children who were ges 3 to 17, nd with the fmily t both bseline nd followup, in order to mximize the number of children from whom dt were directly collected (in the child ssessments nd child survey) Construction of Progrm Usge Dt To construct new dtset with informtion bout progrm usge, the study tem combined dt from seven sources: enrollment verifiction; 6-, 12-, nd 18-month surveys; the Homeless Mngement Informtion System (HMIS); the HUD Public nd Indin Housing Informtion Center (PIC); nd the Tennt Rentl Assistnce Certifiction System (TRACS). The dtset contins monthly prticiption informtion for the entire period of observtion (from rndom ssignment to the time of the followup survey), spnning seven progrm types. Exhibit 4-2 shows the seven progrm types nd the dt sources for ech type. The dt re structured so tht they count 1 month of prticiption in progrm type if the fmily uses tht progrm type for t lest 1 night during the month. The dt contin 3,573 spells of progrm use for the 1,857 followup survey respondent fmilies. 39 The structure of the dt could bis estimtes of progrm durtion upwrd (reltive to true use) becuse (1) progrm entries re spred throughout the clendr month nd (2) progrm exits for emergency shelter, trnsitionl housing, nd permnent supportive housing re typiclly spred throughout the clendr month (rther thn occurring t the end of the clendr month). To ddress this upwrd bis, ech progrm spell is djusted slightly downwrd in mesures of durtion. (See Appendix A for full detils.) These dt re known to miss t lest some progrm use. The bseline sty in emergency shelter does not pper in the dt for 20 percent of survey respondent fmilies. The missing dt rte for subsequent stys in emergency shelter is unknown. The study tem expects tht HMIS records on community-bsed rpid re-housing, trnsitionl housing, nd permnent housing to be t lest s complete s the bseline emergency shelter records (t lest 80 percent). Becuse the dt on these three progrm types lso rely on enrollment verifiction (for the referred progrm) nd up to three selfreports, the study tem expects the vst mjority of progrm spells of these types to be cptured in the dt. The dt on use of subsidy, public housing, nd project-bsed vouchers or Section 8 projects should be essentilly complete becuse they re bsed on HUD dministrtive records. Addi - tionl detil bout the construction of the progrm usge dt is provided in Appendix A. The reminder of the chpter describes the methodology used to clculte the 20-month impct estimtes in the report. 36 The survey softwre rndomly selected the focl children immeditely fter the focl child screener, the first module in the dult survey. The subsequent module on child well-being then sked items bout the focl child or focl children who hd been selected. 37 About 530 children (out of bout 4,200 totl) screened for selection s focl children were living with the fmily hed less thn hlf the time. The fmily hed ws knowledgeble regrding only 60 of these 530 children. Thus, in ccordnce with the focl child selection protocol, these 60 children were selected s focl children (long with 2,724 other selected focl children who were living with the fmily hed t lest hlf of the time). During nlysis, however, it ws decided tht such smll number of children would not llow estimtes to generlize to the whole group of lrgely bsent children. Therefore, these 60 children were not included in impct nlyses. As result, the child impct results generlize only to children living with the fmily hed hlf of the time or more t the time of the dult survey. 38 The oversmpling criterion of being with the fmily t bseline ws included so tht oversmpled children would be directly ffected by the study s rndom ssignment. Children needed to be with the fmily t followup for the study tem to ttempt collection of child ssessments or the child survey. (The study did not ttempt to locte children seprted from the fmily.) 39 A spell of progrm use is continuous period of use, with single strting month nd single ending month. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 36

69 Chpter 4. Dt Sources nd Methodology Exhibit 4-2. Progrm Types nd Their Dt Sources in the Progrm Usge Dt Emergency shelter Progrm Type Subsidy (housing choice voucher) Community-bsed rpid re-housing Trnsitionl housing b Permnent supportive housing Public housing Project-bsed vouchers or Section 8 projects Dt Sources HMIS records 6-, 12-, nd 18-month surveys HUD PIC nd TRACS records Enrollment verifiction records (for referred progrm) 6-, 12-, nd 18-month surveys HMIS records Enrollment verifiction records (for referred progrm) 6-, 12-, nd 18-month surveys HMIS records Enrollment verifiction records (for referred progrm) 6-, 12-, nd 18-month surveys HMIS records 6-, 12-, nd 18-month surveys HUD PIC nd TRACS records 6-, 12-, nd 18-month surveys HUD PIC nd TRACS records 6-, 12-, nd 18-month surveys HMIS = Homeless Mngement Informtion System. HUD = U.S. Deprtment of Housing nd Urbn Development. PIC = Public nd Indin Housing Informtion Center. TRACS = Tennt Rentl Assistnce Certifiction System. The subsidy progrm type represents housing choice vouchers plus site-specific progrms tht fmilies were referred to when ssigned to the permnent housing subsidy (SUB) intervention. The site-specific non-housing choice voucher progrms were public housing in Honolulu, Hwii, nd project-bsed vouchers in Bridgeport, Connecticut. In other sites, these progrms re coded seprtely. b The trnsitionl housing progrm type represents both project-bsed nd scttered-site vrieties of trnsitionl housing, including trnsition-in-plce units. 4.2 Methodology This report presents seprte impct estimtes for ech of the 6 pirwise comprisons of single ssignment rm to nother ssignment rm, plus 4 dditionl comprisons of pooled ssignment rms to single ssignment rm (see Exhibit 1-1 nd Chpters 6 through 10). All 10 comprisons hve been nlyzed seprtely using the sme bsic estimtion model. Pirwise Comprisons SUB versus UC CBRR versus UC PBTH versus UC Pooled Comprisons. SUB versus CBRR SUB versus PBTH CBRR versus PBTH Wht is impct of ny kind of housing subsidy for homeless fmilies (SUB + CBRR + PBTH) compred to usul cre (UC)? Wht is the impct of housing subsidy with hevy services on homeless fmilies (PBTH) compred to housing subsidy with light or no services (SUB + CBRR)? Wht is the impct of interventions tht re more costly (PBTH + SUB) compred to less costly intervention (CBRR)? Wht is the impct of housing subsidy with no time limit (SUB) compred to time-limited housing subsidy (PBTH + CBRR)? The explntion of the estimtion model begins with some terminology tht describes how rndom ssignment ws implemented in this study. Enrollment nd rndom ssignment ws multistep process, s shown in Exhibit 2-7 (in Chpter 2). The PBTH, CBRR, nd (in some sites) the SUB interventions hd multiple service providers in ech site. Before rndom ssignment, the number of slots currently vilble t ll providers for ech of the interventions ws ssessed. An intervention ws deemed vilble if t lest one slot t one provider of tht intervention in the site ws currently vilble. After n intervention ws determined to be vilble, the interviewer sked the fmily series of questions to ssess provider-specific eligibility for the vilble interventions nd progrms. A fmily ws considered eligible for prticulr intervention if the household hed s responses to the eligibility questions showed tht the fmily met the eligibility requirements for t lest one provider of tht intervention tht currently hd n vilble slot. For exmple, some progrms required tht fmilies hve source of income tht would llow for them to py rent on their own within designted period of time. The study tem thus sked fmilies if they wnted to be considered for progrms with such n income requirement. Other progrms required fmilies to py monthly progrm fee, nd the screening question sked if fmilies wnted to be considered for progrms with this type of requirement. Other progrms required prticipnts to demonstrte sobriety, pss criminl bckground checks, or gree to prticipte in cse mngement or other services. The study tem sked Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 37

70 Chpter 4. Dt Sources nd Methodology screening questions for these questions tht scertined fmilies willingness to be considered for progrms with these requirements. To undergo rndom ssignment, fmily needed to be eligible for t lest one vilble intervention in ddition to UC. 40 Bsed on this pproch to rndom ssignment, ech fmily hs rndomiztion set. Rndomiztion set. The set of interventions to which it ws possible for fmily to be ssigned ws determined by considering both the vilbility of the intervention nd the ssessed eligibility of the fmily. In the study, ech fmily hs one of seven possible rndomiztion sets. These sets re {PBTH, SUB, CBRR, UC}, {PBTH, SUB, UC}, {PBTH, CBRR, UC}, {SUB, CBRR, UC}, {PBTH, UC}, {SUB, UC}, nd {CBRR, UC}. The rndomiztion set of ech fmily determines the pirwise comprisons in which the fmily is included. A fmily is included in the pirwise comprisons of its ssigned intervention with the other interventions in its rndomiztion set. For exmple, fmilies ssigned to PBTH with rndomiztion set {PBTH, SUB, UC} re included in these two pirwise comprisons: PBTH versus UC; nd SUB versus PBTH Impct Estimtion Model for Fmily nd Adult Outcomes For ech pirwise comprison, the study tem estimted impcts for the smple of fmilies who (1) hd both interventions in their rndomiztion set nd (2) were rndomly ssigned to one of the two interventions. The tem used mul - tivrite regression to increse the precision of our impct estimtes nd to djust for ny chnce imblnces between ssignment groups on bckground chrcteristics (Orr, 1999). Consider two interventions q nd r (for exmple, PBTH versus SUB), where the second option (r) is treted s the bse cse. Then, the impct on n outcome Y (for exmple, t lest 1 night homeless or doubled up during pst 6 months, working for py in week before survey, or dult psychologicl distress) of intervention q reltive to intervention r is estimted through eqution 1 for those fmilies who hd both options q nd r s possible ssignments, nd were ssigned to one of them. The estimtion eqution ws (1), where Y i = outcome Y for fmily i, T q,i = indictor vrible tht equls 1 if fmily i ws ssigned to intervention q, d q,r = verge impct of being ssigned to intervention q reltive to being ssigned to intervention r, X i = vector of bckground chrcteristics 41 of fmily i, I k,i = indictor vrible for site-ra regime 42 k for fmily i, e i = residul for fmily i (ssumed men-zero nd i.i.d. [independently nd identiclly distributed]), q,r = constnt term, nd b q,r, f q,r,k = other regression coefficients. The estimte of the impct prmeter d q,r is the intention-totret, or ITT, estimte. For the pirwise comprisons, it is n estimte of the verge effect of being offered intervention q rther thn intervention r. The verge effect is tken over ll fmilies in the q,r comprison, regrdless of whether fmilies ctully prticipted in the intervention to which they were ssigned. This model ssumes tht the true impct of intervention q reltive to intervention r is homogeneous cross sites. The impct prmeter d q,r is thus implicitly weighted verge of the point estimtes of site-level impcts, with ech sitelevel impct weighted by the number of fmilies in the site. A slight modifiction of this model is used to estimte impcts in the pooled comprisons. In tht modifiction, dditionl site-ra regime covrites re included, nd q represents being offered one of two or three interventions rther thn single intervention. Stndrd Errors The model described previously ws estimted using weighted lest squres (WLS) nd heteroskedsticity-consistent stnd - rd errors, lso known s robust stndrd errors (tht is, Huber-Eicker-White robust stndrd errors; see Greene, 2003; Huber, 1967; nd White 1980, 1984). Heteroskedstic residuls would rise if some types of fmilies hve higher vribility in their outcomes thn other fmilies or if the different interventions themselves influence this vribility. 40 Altogether, 183 of the screened fmilies were not eligible for ny vilble interventions besides UC. These fmilies were not enrolled in the study. 41 These bckground chrcteristics re listed in Appendix C. 42 Of the 12 sites, 10 hd single rndom ssignment regime during the 15-month study enrollment period. The remining 2 sites chnged rndom ssignment probbilities single time ech, creting 14 site-ra regime groups. The eqution includes 13 indictor vribles nd omits 1. These indictor vribles re included so tht the impct estimte is bsed on within-site comprisons. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 38

71 Chpter 4. Dt Sources nd Methodology Furthermore, this study uses the liner probbility model for binry outcomes, rther thn logit or probit model, becuse of the ese of interprettion of lest squres prmeter estimtes. The liner probbility model, however, induces heteroskedsticity (Angrist nd Pischke, 2008). To ddress this potentil heteroskedsticity, robust stndrd errors were estimted nd used in tests of sttisticl significnce. These stndrd errors re pproprite for mking inferences bout intervention effects for the sites in this study. The stndrd errors do not tke into ccount vribility in site-level effects, however, nd so re not pproprite for generlizing results to other sites. Adult Survey Nonresponse Weights The dult survey chieved n 81-percent response rte t followup. Nonresponse rises two concerns. First, nonresponse to followup survey used to mesure outcomes presents chllenge to the internl vlidity of the study if the intervention groups (tht is, PBTH, SUB, CBRR, nd UC) hve different ptterns of nonresponse. Second, followup survey nonresponse cn threten the generlizbility of results to the entire enrolled smple if survey nonrespondents differ from respondents, even if they do so symmetriclly cross rndomiztion rms. To ddress both of these issues, the nlysis tem prepred set of weights tht djust for dult survey nonresponse for ech pirwise comprison tht is bsed on fmily chrcteristics mesured in the bseline survey. 43 The weights were used in estimting impcts on ll fmily nd dult outcomes Impct Estimtion Model for Child Well-Being Outcomes The estimtion model for impcts on child well-being out - comes differs from the model described previously in two respects. First, the stndrd errors re modified to ccommo - dte the fct tht some child well-being impct regressions include two children from the sme fmily. To llow for correltion between impcts on children in the sme fmily, the model estimtes the robust stndrd errors clustered within fmily. Second, more complex weighting strtegy is used to ddress the process by which individul child observtions cme to be included in impct regressions. The child weights re the product of three components. 1. The dult survey nonresponse weight. 2. The inverse probbility of being selected s focl child. 3. A child survey nonresponse weight (conditionl on the prent being n dult survey respondent). The use of the child weights implies tht the child wellbeing impct estimtes for given comprison represent ll children who could hve been selected s focl children in ll the fmilies in the comprison. The second component implies tht n interviewed child from lrger fmily will get lrger weight thn n interviewed child from smller fmily Impct Estimtion Model for Modertor Anlysis The modertor nlysis ddressed in Chpter 11 presents evidence on whether the study interventions re more effective for fmilies with different levels of psychosocil needs or housing brriers. The estimtion model for the modertor nlysis is (2) where ll terms ppering in eqution 1 hve the sme definition, M i = potentil modertor index vrible (either psychosocil chllenges or housing brriers) for fmily i, p q,r = chnge in impct of being ssigned to intervention q reltive to being ssigned to intervention r ssocited with one-unit chnge in M index, nd γ q,r = other regression coefficient. The potentil modertor index vrible, M, is entered in the model both lone nd intercted with tretment, T. The test of sttisticl significnce of the p q,r coefficient serves s the test for whether impcts differ significntly ccording to the M index. Stndrd errors nd weights for fmily, dult, nd child outcomes re the sme s in the min impct estimtion Strtegy for Addressing the Multiple Comprisons Problem Sttement of the Problem Simply stted, the multiple comprisons problem is tht, s the number of hypothesis tests conducted grows, the likelihood of finding sttisticlly significnt impct somewhere mong the tested outcomes simply by chnce increses fr bove the desired risk level for producing flse positive results. This multiple comprisons problem is prticulrly 43 The construction of weights to ddress survey nonresponse is discussed in Little (1986). Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 39

72 Chpter 4. Dt Sources nd Methodology slient for the Fmily Options Study becuse the multiple rms, multiple domins, nd multiple outcomes generted n extremely lrge number of hypothesis tests in the min impct nlysis totl of 730 tests (10 comprisons 73 outcomes in the five outcome domins). Given this lrge number of tests, the probbility of finding n impct, even if there were no true impcts, ws quite lrge well bove the nominl 10-percent level. In prticulr, the probbility of finding t lest one significnt impct t the 0.10 level in k independent tests when ll true impcts re 0 is given by eqution 3. (3) Prob(min p 0.10 ll true impcts = 0) = k Thus, if 10 independent tests were performed, the probbility of finding t lest one significnt impct t the 0.10 level often tken s the litmus test for successful intervention when ll true impcts re equl to 0 is = 0.65; tht is, bout two-thirds of the time one would conclude n unsuccessful intervention is successful. When 20 independent tests re performed, the probbility is 0.88; tht is, nerly 9 times out of 10. In fct, with hundreds of tests, the nlysis is nerly certin to spuriously detect successful effect even if the intervention ws not truly successful for ny outcome. 44 Response to the Problem The study tem took two steps to ddress the multiple comprisons problem. 1. Adjust the stndrd of evidence used to declre subset of individul impct estimtes sttisticlly significnt. The study tem divided the hypothesis tests into smll set of seven confirmtory tests nd much lrger set of 723 explortory tests. The tem then used multiple comprisons procedure to djust the results of the seven confirmtory tests to mintin the integrity of the sttisticl inferences mde t the confirmtory level Prespecify impcts to present in the executive summry. The study tem prespecified the impcts on 18 key outcomes in the six pirwise comprisons (for 108 totl impct esti - mtes) to present in the executive summry before seeing the results. This step ws tken to prevent the selective pre - senttion of sttisticlly significnt results in the executive summry. The first step hinges on the definition nd implictions of confirmtory hypothesis tests. Following Schochet (2009), the tem defined confirmtory hypothesis tests s those tests tht ssess how strongly the study s pre-specified centrl hypotheses re supported by the dt (Schochet, 2009). Sttisticlly significnt findings from confirmtory hypothesis tests re considered definitive evidence of nonzero intervention impct, effectively ending debte on whether the intervention chieved n impct in the study sites. All other hypothesis test results re deemed explortory. For these tests, sttisticlly significnt impcts constitute suggestive evidence of possible intervention effects. Before beginning nlysis, HUD determined tht the housing stbility domin ws the most importnt outcome domin for the study. Therefore, the study tem designted seven hypothesis tests relted to housing stbility s confirmtory. These hypothesis tests were conducted for The six pirwise policy comprisons nd one pooled comprison (PBTH + SUB + CBRR versus UC). A single composite outcome of t lest one return to homelessness, constructed from two binry outcomes within the housing stbility domin. 1. At lest 1 night spent homeless or doubled up during the pst 6 months t the time of the followup survey (from the dult survey). 2. Any sty in emergency shelter in the pst 12 months t the time of the followup survey (from progrm sge dt, lrgely bsed on HMIS records). The six pirwise comprisons were included in order to ssess the reltive effectiveness of the interventions in contributing to housing stbility (thereby ddressing the study s first reserch question, stted in Section 1.4). The study tem lso included the pooled comprison of PBTH + SUB + CBRR versus UC becuse it provided evidence on whether housing subsidy of ny type improved housing stbility. Using two sources of dt to construct this outcome enbled us to mesure housing stbility s robustly s possible nd mde use of ll vilble dt on return to homelessness. 44 Although the study tem does not expect the hundreds of hypothesis tests performed in this report to be independent, the likelihood of t lest one spurious finding of sttisticl significnce will still be extremely high. 45 The multiple comprisons procedure used to djust the confirmtory test results ws the Westfll-Young resmpling procedure. This procedure is described in Appendix C. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 40

73 CHAPTER 5. DESCRIPTION OF USUAL CARE (UC) AND OUTCOMES MEASURED IN THE STUDY This chpter describes the fetures of the usul cre (UC) progrm environment with which the three ctive interventions of permnent housing subsidy (SUB), community-bsed rpid re-housing (CBRR), nd project-bsed trnsitionl housing (PBTH) re compred. It lso detils the experiences nd outcomes during the 20-month followup period of fmilies ssigned to remin in UC rther thn being referred to n ctive intervention fter 7 dys in emergency shelter. The first section of the chpter profiles the housing nd supportive services provided by the emergency shelters from which study prticipnts were drwn. The second section of the chpter presents informtion on the types of other progrms (besides emergency shelters) UC fmilies used during the followup period. The lst section of the chpter introduces the outcomes exmined in the impct nlysis nd presents benchmrk levels of these outcomes for UC fmilies ginst which the impcts of other interventions will be mesured. 5.1 The Emergency Shelter Experience of Usul Cre (UC) Fmilies Usul cre for this study consisted of whtever progrm services UC fmilies were ble to ccess on their own following sty in emergency shelter of t lest 7 dys, without specil referrl to one of the study s ctive interventions. All fmilies were recruited for the study from emergency shelters. For fmilies rndomly ssigned to UC, the study tem encourged shelter stff to not ctively guide fmilies to SUB, CBRR, nd PBTH providers in the community. Becuse fmilies ssigned to UC were not explicitly provided other ssistnce, ll UC fmilies remined in emergency shelter until they nvigted their wy out with or without the ssistnce of the emergency shelter stff, or until they reched the shelters length-of-sty limits. Across ll 12 study sites, 746 fmilies were rndomly ssigned to the UC intervention from 56 emer - gency shelters. 46 Of these 746 fmilies, 578 (77 percent) responded to the followup survey nd re therefore included in the impct nlysis in this report. For some fmilies ssigned to UC, emergency shelters were the fmilies only interction with the homeless ssistnce or housing subsidy system. Other UC fmilies found their wy into housing subsidy, rpid re-housing, nd trnsitionl housing progrms on their own (see Section 5.2). This section describes the emergency shelter prt of the usul cre service environment the shelters in which ll UC fmilies spent t lest 7 dys nd often longer, referred to here s UC shelters. 47 Other types of ssistnce UC fmilies received, s well s outcomes for UC fmilies, re described in subsequent sections of this chpter UC Shelters in the Study Sites The study tem ttempted to recruit study prticipnts from ll of the emergency shelters tht stkeholders in the study communities described s entry points into the homeless system in ech included city. In some cses, progrms my hve used the term emergency shelter to describe the ssistnce they provided, but in prctice their service model involved recruiting fmilies from other shelters. The study tem did not recruit fmilies into the study from these types of progrms (which function more like trnsitionl housing progrms) becuse this model ws not consistent with the study design. The study tem lso voided emergency shelters tht exclusively provided domestic violence ssistnce becuse most victim-service providers did not believe tht rndom ssignment to nonspecilized homeless ssistnce progrms ws pproprite for such clients. 48 As result 46 One fmily ws enrolled into the study from 57th shelter, but this fmily ws not ssigned to the UC group. Informtion in this chpter describing the housing nd services provided by UC shelters is bsed on 53 emergency shelters tht provided progrm dt. These 53 UC shelters served nerly ll the fmilies ssigned to UC (741 of 746, or 99 percent). 47 Fmilies ssigned to other interventions lso spent time in UC shelters, for t lest 7 dys before rndom ssignment nd often longer. 48 Two emergency shelters in the Connecticut site were domestic violence shelters, nd two of nine in Almed County, Cliforni, were entirely or prtly dedicted to domestic violence. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 41

74 Chpter 5. Description of Usul Cre (UC) nd Outcomes Mesured in the Study of these exclusions, but brod inclusion otherwise, UC encompsses ll gtewy shelters, or primry entry points, in site s homeless system. The number of UC shelters in ny site depends on the orgniztion of tht community s homeless system. For instnce, in Slt Lke City, Uth, the study recruited from the single primry shelter in its lrgely centrlized homeless system. In Almed County, Cliforni, on the other hnd, the study recruited from nine shelters, reflecting the lrge geogrphic re covered by the county nd the homeless system s generl decentrliztion. Mny UC shelters were prt of gencies tht operte other homeless ssistnce progrms in ddition to emergency shelter, such s trnsitionl housing. Under regulr opertions (outside of this study), the typicl prctice for these multiservice gencies would be to refer mny fmilies in shelter to their internl progrms insted of referring them to other providers. Under the protocol estblished for the study, emergency shelters greed to refer fmilies to ssis - tnce fter shelter bsed on the result of rndom ssignment. Bsed on HUD priorities for funding homeless ssistnce, very few HUD dollrs re used to support emergency shelters. As result, emergency shelters typiclly rely on brod rnge of other locl privte nd public funding sources. Shelters re therefore not governed by uniform stndrds nd requirements for progrm opertions, s is the cse for CBRR, SUB, nd to some extent PBTH. A gret del of similrity ws nonetheless found cross communities in the types of ssistnce provided in emergency shelters, nd lso found mny similrities between UC shelters nd PBTH progrms Housing Assistnce in UC Shelters The recruitment criteri for the study specified tht ll fmilies ssigned to UC hd spent t lest 7 dys in shelter before rndom ssignment. Most UC fmilies spent dditionl time in shelter fter rndom ssignment. UC fmilies lso used other types of housing nd homelessness ssistnce in the community during the 20-month followup period (see Sec - tion 5.2). This section describes the types of ssistnce pro - vided in UC shelters. As shown in Exhibit 5-1, more thn 80 percent of UC fmilies come from congregte shelter set - tings. These shelters consisted of congregte dorm settings (35 percent of fmilies); group homes in which fmilies hve privte bedroom with shred bthrooms nd kitchens (26 percent); nd fcilities with privte bedrooms nd bthrooms but shred kitchens nd other common spce (23 percent). Another 13 percent of fmilies styed in vrious prtmentbsed settings nd 4 percent in other settings. Length of Time Spent in Emergency Shelters by UC Fmilies The Progrm Usge Dt collected for the study from vrious sources indicte whether fmily spent t lest 1 night in emergency shelter during clendr month. 49 These dt show tht UC respondent fmilies styed in emergency shel - ter for n verge of 4 months nd medin of 2 months during the followup period. 50 Exhibit 5-2 shows the percent - ges of fmilies with vrious totl month durtions in emergency shelter during the followup period. About 53 percent of UC fmilies styed in emergency shelter for 3 or fewer Exhibit 5-1. Types of Living Spce Provided by UC Shelters Type of Living Spce Percent of Fmilies Assigned to UC in Shelters With This Setting (N = 722) Congregte dorms (shred bedrooms or sleeping spce) 34 Group homes (privte bedroom, shred kitchens nd bthrooms) 25 Fcilities (for exmple, motels) fmilies hve privte rooms with bthrooms but not kitchens 23 Fcility-bsed prtments 7 Aprtments clustered in lrger buildings not owned or controlled by the progrm 5 Other settings 5 UC = usul cre. Five UC shelters (representing 24 fmilies) did not provide this informtion nd were excluded from this exhibit. Note: Percentges re unweighted. Sources: Progrm dt forms; rndom ssignment records 49 See Appendix A for detils of how the sources of progrm usge were combined into single dtset. Missing dt on emergency shelter stys bis the counts of totl months spent in emergency shelter somewht downwrd. The bseline sty in emergency shelter does not pper in the dt from the Homeless Mngement Informtion Systems (HMIS) for 18.7 percent of UC respondent fmilies. The missing dt rte for subsequent stys in emergency shelter is unknown, but it is expected to be lower thn 18.7 percent becuse informtion on subsequent stys ws collected both from HMIS nd from followup surveys. 50 The men nd medin re computed with weights to djust for survey nonresponse, so tht the respondent fmilies represent ll 746 fmilies ssigned to UC. The length of the followup period is from the month of rndom ssignment to the month of the followup survey response (medin 21 clendr months included in followup period). Most fmilies hd only one spell in emergency shelter, wherein spells re seprted by t lest 1 clendr month with no emergency shelter sty. The weighted percentge of fmilies with 0 spells (tht is, missing dt on the spell when they were recruited from shelter t bseline nd lso no subsequent spell recorded in the HMIS progrm usge dt) is 12.1 percent, with 1 spell is 72.0 percent, with 2 spells is 13.6 percent, nd with 3 or more spells is 2.3 percent. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 42

75 Chpter 5. Description of Usul Cre (UC) nd Outcomes Mesured in the Study Exhibit 5-2. Length of Time Spent in Emergency Shelters by UC Fmilies Number of Months (with t lest 1 night sty) in Emergency Shelter During Followup Period Percent of UC Respondent Fmilies b (N = 578) c to to to to to to to to to to to to or more 3.7 UC = usul cre. The structure of the dt, which count month of shelter sty if t lest 1 night in the clendr month is in emergency shelter, could bis estimtes of shelter sty upwrd (reltive to true use) becuse shelter entries nd exits cn occur ny time during the clendr month. To ddress this upwrd bis, ech shelter spell is djusted slightly downwrd in mesures of durtion. (See Appendix A for full detils.) Missing dt on emergency shelter stys bis the counts of totl months spent in emergency shelter somewht downwrd. The bseline sty in emergency shelter does not pper in the dt from the Homeless Mngement Informtion System (HMIS) for 18.7 percent of UC respondent fmilies. The missing dt rte for subsequent stys in emergency shelter is unknown but is expected to be lower thn 18.7 percent, becuse informtion on lter stys ws collected both from HMIS nd from followup surveys. b Percentges re weighted for survey nonresponse to represent ll 746 UC fmilies. c Of UC respondent fmilies, 12.1 percent do not hve ny emergency shelter sty (t bseline or fter bseline) in the HMIS progrm usge dt, even though ll fmilies were recruited from emergency shelters. These fmilies re mong the 18.7 percent of UC respondent fmilies whose bseline sty in emergency shelter does not pper in the dt. The other 6.6 percent of UC respondent fmilies ( = 6.6) whose bseline sty in emergency shelter does not pper in the dt hve t lest one sty in emergency shelter fter bseline nd so re included in the rows below the months row. Source: Progrm Usge Dt months during the followup period. Another 23 percent styed 4 to 6 months during the followup period, nd 24 percent styed more thn 6 months. Exhibit 5-3 shows the percentge of UC fmilies who hve t lest 1 night in emergency shelter during the month for ech month reltive to rndom ssignment. A second line shows the percentge of UC fmilies who hve not left the initil sty in shelter ech month. The exhibit shows tht most fmilies hve left emergency shelter by the 3rd or 4th month fter rndom ssignment. The exhibit shows tht only bout 20 percent of UC fmilies re in emergency shelter by the 6th month fter rndom ssignment. Pst this point, the percentge drops slowly nd, in the 18th month fter rndom ssignment, bout 8 percent of UC fmilies re in emergency shelter. From Exhibit 5-2, which shows only 3.7 percent of UC fmilies in shelter for 18 or more months, it is possible to deduce tht little more thn one-hlf of the fmilies in shelter in the 18th month fter rndom ssignment hve returned to shelter fter deprture. Fmily Rent Contributions nd Svings Requirements in UC Shelters Some shelters require fmilies to py monthly progrm fee, either flt fee per month or fee bsed on income. Some shelters require fmilies to estblish svings while stying in the shelter nd work with fmilies to develop monthly svings gol. These requirements re intended to help fmilies develop finncil skills nd plnning needed to prepre for permnent housing. As shown in Exhibit 5-4, most fmilies were in shelters tht llowed fmilies to sty in shelter without pying rent or progrm fee, but more thn one-hlf were in shelters tht required them to sve money while they were enrolled in the progrm. Of the nerly onefourth of fmilies tht were required to py fee during their sty, most (85 percent) were in shelters tht required them to py flt mount (for exmple, $7 per dy), nd 15 percent were in progrms tht determined the mount bsed on percentge of fmily s income. UC progrms typiclly provided food for fmilies, but one-third of fmilies were in progrms tht did not provide ny food, nd 16 percent of fmilies were in progrms tht required them to provide t lest some of their own food. Some progrm stff indicted tht fmilies receiving the Supplementl Nutrition Assistnce Progrm (SNAP) hd to contribute portion of their SNAP benefits to offset the cost of the food provided by the progrm. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 43

76 Chpter 5. Description of Usul Cre (UC) nd Outcomes Mesured in the Study Exhibit 5-3. UC Group Percent of Fmilies With t Lest 1-Night Sty in Emergency Shelter During Month, by Number of Months After RA 100% 90% 80% 70% Percent of fmilies 60% 50% 40% 30% 20% All UC fmilies in emergency shelter 10% 0% UC fmilies who hve not left shelter since RA Number of months fter RA RA = rndom ssignment. UC = usul cre. Notes: Percentges re weighted for survey nonresponse to represent ll UC fmilies in the study. Missing dt on emergency shelter stys bis the percentges somewht downwrd. The bseline sty in emergency shelter does not pper in the dt for 18.7 percent of UC respondent fmilies. The missing dt rte for subsequent stys in emergency shelter is unknown. Source: Progrm Usge Dt Exhibit 5-4. Fmily Rent Contributions nd Svings Requirements in UC Shelters UC Shelter Progrm Fetures Are fmilies required to py progrm fee or rent? Percent of Fmilies Assigned to UC in Progrms With These Chrcteristics (N = 739) Yes 23% No 77 (If yes) How is the progrm fee/rent determined? Percentge of income 15% Flt mount bsed on fmily or unit size 85 Does the progrm require fmilies to set side some income ech month s svings Yes 60% No 40 Who is responsible for food for prticipting fmilies? Fmilies provide own food 32% Progrm provides food 52 Both 16 UC = usul cre. Four UC shelters (representing seven fmilies) did not provide this informtion nd were excluded from this exhibit. Note: Percentges re unweighted. Sources: Progrm dt forms; rndom ssignment records Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 44

77 Chpter 5. Description of Usul Cre (UC) nd Outcomes Mesured in the Study Assessment of Fmilies in UC Shelters All shelter progrms tht provided progrm dt indicted tht they conducted forml ssessment of fmilies upon entry into the shelter. These ssessments were described s comprehensive. Nerly ll fmilies ssigned to UC were recruited from shelters tht ssessed fmily needs relted to housing, self-sufficiency, employment, helth, mentl helth, substnce buse, nd child-specific needs. One-hlf of the fmilies ssigned to UC were recruited from progrms tht reported tht ssessments lso focused on life skills nd 44 percent were recruited from progrms tht ssessed prenting skills. Assessments typiclly resulted in forml service pln (or equivlent) with gols for the dults in the household, designed to help fmilies ddress their needs. Of fmilies ssigned to UC, 42 percent were in shelters tht lso worked with fmilies to develop gols for the children Supportive Services in UC Shelters In their role s the entry point to the homeless system, ll the UC shelters offered fmilies comprehensive ssessments, cse mngement, direct provision of mny supportive ser - vices, nd referrls to minstrem progrms or other progrms to meet dditionl fmily needs. Most shelters required pr - ticiption in services or ctivities s condition for remining in the shelter. Although shelter progrms indicted tht cse mngement emphsized plcement in permnent housing, stff lso indicted tht they considered cse mngement to encompss more thn housing tht cse mngement ws intended to meet client needs s defined jointly by fmilies nd cse mngers. As result, cse mngers t emergency shelters ppered to provide support nd referrls to ddress wide rnge of issues relted to housing, employment, helth, mentl helth, substnce buse, prenting, nd children. Exhibit 5-5 shows the rry of services tht shelter stff re - ported tht they offered nd the extent to which the service ws provided through cse mngement, by other progrm or gency stff beyond the cse mnger, or through forml rrngement with n externl gency tht ws gurnteed to provide the service becuse of shelter enrollment. Some progrms ddress prticulr service through only one method, wheres others use multiple methods. The second column of the exhibit shows the percentge of fmilies referred to progrms tht offer ech type of service. The subsequent columns report seprtely the percentge of fmilies referred to progrms tht provide tht service type through tht specific mens. In some cses, ddressing service through cse mngement mens provision of direct ssistnce by the cse mnger. In other cses, ddressing service through cse mngement mens tht the cse mngers provide referrls to other progrms, dvocte on behlf of the fmily to ccess cre, help remove brriers to cre, nd coch or support fmily s it ttempts to complete its gols relted to tht service need. Essentilly ll the shelter cse mngers worked with fmilies on housing serch nd plcement nd self-sufficiency issues. In ddition, cse mngement ws supplemented Exhibit 5-5. Types of Supportive Services Offered in UC Shelters nd How They Are Delivered, s Reported by Shelter Stff Types of Supportive Services Percent of Fmilies Assigned to UC in Shelters Tht Offered These Services (N = 739) Percent of Fmilies Assigned to UC in Shelters Tht Offer Services of This Type: Through Cse Mngement By Other Progrm or Agency Stff Through Forml Arrngements With Other Agencies Housing serch nd plcement ssistnce (%) Self-sufficiency (overll) (%) NA NA Childcre/fter-school cre Finncil mngement 5 6 Trnsporttion 7 0 Help obtining public benefits 1 0 Physicl helth cre (%) Employment trining (%) Child dvoccy (%) Life skills (%) Mentl helth cre (%) Prenting skills (%) Substnce buse (%) Fmily reunifiction (%) NA = dt not vilble. UC = usul cre. Four UC shelters (representing seven fmilies) did not provide this informtion nd were excluded from this exhibit. Note: Percentges re unweighted. Sources: Progrm dt; rndom ssignment records Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 45

78 Chpter 5. Description of Usul Cre (UC) nd Outcomes Mesured in the Study with other direct services in nerly ll progrms. Of UC fmilies, 87 percent were in shelters tht offered ccess to physicl helth cre, 77 percent were in shelters tht offered employment trining, 76 percent were in shelters offering child dvoccy, 76 percent were in shelters offering life skills, nd nerly s mny were in shelters tht offered ccess to mentl helth cre nd prenting services. Cse Mngement in UC Shelters Cse mngement ws considered by emergency shelter stff to be core prt of their progrms nd the primry mens of helping fmilies to resolve the crises tht resulted in homelessness. All but one shelter in the study reported providing cse mngement to fmilies. 51 Cse mngement vried widely cross the prticipting shelter progrms, s shown in Exhibit 5-6, but the verge cse mngement rtio ws 16 fmilies per shelter cse mnger. More thn one-third of UC fmilies (38 percent) were in shelters with cse mngers who worked with 10 or fewer fmilies t time, meeting with fmilies weekly, if not dily. Roughly the sme percentge of UC fmilies ws in shelters with cse mnger cselods between 11 nd 20 fmilies, wheres slightly more thn one-fourth of fmilies were in shelters with cselods of more thn 20 fmilies. Ninety-one percent of fmilies were in shelters in which they met with cse mngers weekly or more often (7 percent dily), nd even in the progrms with higher cselods, cse mngers reported meeting with fmilies biweekly. According to cse mngers, fmilies in emergency shelter most often met with their cse mngers for 30 to 45 minutes. Additionl detils bout ech emergency shelter s cse mngement re shown in Gubits et l. (2013), Appendix B-7. Cse mngement ws rrely offered fter fmilies moved out of shelters. Although some cse mngers sid they kept n open door nd my hve styed in touch with some fmilies, most progrms did not pper to provide significnt level of ssistnce fter progrm prticiption. Other Supportive Services in UC Shelters The primry service provided in shelter is cse mngement, but shelters offered other direct services, by other stff within the progrm, by nother progrm within the gency, or through forml rrngements with other gencies. Exhibit 5-5 shows tht the type of services offered vried from shelter to shelter, nd tht no other services were offered s widely s cse mngement. The most common services offered outside of cse mngement were physicl helth cre, mentl helth cre, employment trining, childcre, nd prenting clsses. Nerly one-hlf (47 percent) of the UC fmilies were in shel - ters tht offered direct helthcre services, most often through n onsite clinic operted by nother gency or through forml rrngement with n outside clinic. Other types of services were formlly offered t UC progrms serving fewer thn one-fourth of UC fmilies. In ddition to the services offered directly by the progrm or forml prtner, fmilies were lso routinely referred to other gencies to receive other types of services tht my be required to ddress their needs. 5.2 Use of Other Homeless nd Housing Assistnce Progrms by Usul Cre (UC) Fmilies For some fmilies ssigned to UC, emergency shelters were the fmilies only interction with progrms tht provide plce to sty or housing subsidy. Mny UC fmilies ulti - mtely found their wy to other types of ssistnce, however. Although the study tem knew wht types of progrms were vilble t the outset of the study, it ws not known how Exhibit 5-6. Cse Mngement Intensity (rtio nd frequency) Averge Number of Clients per Cse Mnger Frequency of Cse Mngement in UC Shelters (%) (N = 710) Dily Weekly or More Often Biweekly Monthly 10 or fewer clients (%) to 20 clients (%) to 30 clients (%) More thn 30 clients (%) Totl UC = usul cre. Seven progrms (representing 36 fmilies) did not provide this informtion nd were excluded from this exhibit. Some row nd column totls my not equl the sum of individul cells becuse of rounding. Note: Percentges re unweighted. Sources: Progrm dt forms; rndom ssignment records Totl 51 The one shelter tht did not provide cse mngement ws crisis center tht provided other supportive services nd ssisted fmilies in obtining cse mngement from other progrms. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 46

79 Chpter 5. Description of Usul Cre (UC) nd Outcomes Mesured in the Study much ssistnce UC fmilies would eventully obtin on their own. Dt collected for the study nswer this question. This section describes the extent to which UC fmilies used progrms other thn emergency shelter tht were vilble to them in their communities. Exhibit 5-7 shows the use of nonshelter homeless nd housing progrms by UC fmilies during the followup period. The exhibit tkes ccount of seven types of progrms. Subsidy (tht is, the progrms tht comprise the SUB intervention in this study: housing choice vouchers, or HCVs, plus public housing in Honolulu, Hwii, nd project-bsed vouchers in Bridgeport, Connecticut). Rpid re-housing (tht is, CBRR). Trnsitionl housing (both non- trnsition-in-plce nd trnsition-in-plce ). Permnent supportive housing. Public housing in plces other thn Honolulu. Project-bsed vouchers or units in Section 8 projects (in plces other thn Bridgeport where the SUB intervention ws provided using project-bsed vouchers). Emergency shelter. In ddition, the exhibit shows informtion on the proportion of UC fmilies who use none of these progrms during the entire followup period or emergency shelter fter the first 6 months fter rndom ssignment. The mesure of no progrm use is ment to exclude typicl-length bseline shelter sty. By the 7th month fter rndom ssignment, most UC fmilies hd deprted from emergency shelter. The first column of Exhibit 5-7 shows the percentge of UC fmilies who ever used type of progrm between the month of rndom ssignment nd the month of followup survey response. This column shows tht 18 percent of UC fmilies received rpid re-housing ssistnce nd one-fourth of UC fmilies received trnsitionl housing. Altogether, 28 percent of UC fmilies received some sort of permnent sub - sidy during the followup period ccessing either subsidy, public housing, permnent supportive housing, or projectbsed housing ssistnce progrms. A little more thn one-fourth (28 percent) of UC fmilies used none of the six progrm types during the followup period or emergency shelter following the first 6 months fter rndom ssignment. The second nd third columns of Exhibit 5-7 show the men nd medin number of months of progrm usge for those fmilies who used the progrm. The number of months of use of rpid re-housing (medin 6 months) nd trnsitionl housing (medin 8 months) re consistent with the expected durtions of these progrm types. It is interesting to note tht the higher men nd medin of the project-bsed vouchers or Section 8 projects row show tht fmilies were ble to ccess this type of ssistnce more quickly thn the other permnent types of ssistnce (subsidy, permnent Exhibit 5-7. Progrm Use Since Rndom Assignment for UC Group Type of Housing Assistnce Percent Ever Used From RA to 18-Month Followup Survey Number of Months Used From RA to 18-Month Followup Survey, If Ever Used Type of Housing Assistnce Men Medin Percent Used in Month of Followup Survey Response Subsidy (SUB) b Rpid re-housing (CBRR) Trnsitionl housing Permnent supportive housing Public housing Project-bsed vouchers or Section 8 projects Emergency shelter c No use of homeless or housing progrms d N UC = usul cre. RA = rndom ssignment. Percentge of fmilies who ever used type of ssistnce progrm during the period from the month of RA to the month of the 18-month followup survey response (medin period durtion: 21 months). Percentges do not dd to 100% becuse some fmilies use more thn one progrm type during the followup period. b Subsidy ssistnce is housing choice vouchers plus site-specific progrms offered to fmilies ssigned to SUB group in Connecticut nd Honolulu. c All fmilies were in emergency shelter t rndom ssignment. The percentge less thn 100 percent for ever used emergency shelter is due to missing dt on shelter use. d No use of homeless or housing progrms (ever used) indictes no use of the six progrm types in this tble during ny of the followup period nd no use of emergency shelter fter the first 6 months fter RA. No use in the month of followup survey response indictes no use of ny of these seven progrm types. Notes: Percentges re regression djusted, controlling for site nd rndomiztion rtio. Percentges, mens, nd medins re weighted for survey nonresponse to represent full comprison smple. Source: Progrm Usge Dt Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 47

80 Chpter 5. Description of Usul Cre (UC) nd Outcomes Mesured in the Study supportive housing, nd public housing). 52 It is not obvious why ccess to project-bsed vouchers or Section 8 projects ws fster thn ccess to these other types, lthough these could hve been units in ffordble housing developments tht were set side for homeless fmilies. 53 The fourth column of Exhibit 5-7 shows the percentge of fmilies who used progrm type in the month of followup survey response. Although the study tem expects tht mny outcomes in the report will be influenced by ssistnce received t ny point during the entire followup period, some outcomes will be most strongly influenced by ssistnce tht is received t the time of followup survey response. By the time of the followup survey, 10 percent of UC fmilies were still in trnsitionl housing progrms, nd 5 percent were in permnent supportive housing. Nerly one-fourth (23 percent) were using permnent housing ssistnce (subsidy, public housing, permnent supportive housing, nd project-bsed vouchers). Thus, the vst mjority of fmilies who got rpid re-housing ssistnce were no longer using it t the time of the followup survey, wheres most fmilies who ccessed permnent ssistnce were still receiving it. Trnsitionl housing fell in between. More thn one-hlf of UC fmilies (56 percent) were not using ny of the six progrm types nor emergency shelter t the time of their followup survey response. 5.3 Outcomes for Fmilies Rndomly Assigned to Usul Cre (UC) This section serves the dul purpose of (1) introducing the outcomes exmined in the impct nlysis nd (2) describing the outcomes of fmilies ssigned to UC who responded to the followup survey. The section is orgnized by outcome domin (housing stbility, fmily preservtion, dult wellbeing, child well-being, nd self-sufficiency). The subsections describe the outcomes in ech domin nd the dt sources used to mesure them. Ech subsection lso describes the chrcteristics of the fmilies in the UC group bsed on these outcome mesures. Appendix B provides dditionl technicl detils regrding the construction of the outcome mesures from survey nd dministrtive dt Mesures of Housing Stbility Understnding the extent to which the interventions llevite homelessness nd permit fmilies to remin stbly housed is one of the most importnt objectives of the study. The reserch tem defined severl outcomes relted to homelessness nd housing stbility, nd they used informtion from the followup survey nd Progrm Usge Dt to mesure these items. The reserch tem developed seven mesures relted to housing stbility nd homelessness experienced during the followup period At lest 1 night homeless or doubled up in pst 6 months (percent of fmilies). This outcome mesures the percentge of study fmilies who reported hving spent t lest 1 night in the 6 months before the followup survey either stying in shelter or plce not ment for humn hbittion, or living with friends or reltives becuse they could not find or fford plce of their own. This outcome is mesured from survey dt. At lest 1 night homeless in pst 6 months (percent of fmilies). This outcome mesures the percentge of fmilies who reported hving spent t lest 1 night in shelter or plce not ment for humn hbittion in the 6 months before the followup survey. 54 This mesure is bsed on responses to the followup survey. At lest 1 night doubled up in pst 6 months (percent of fmilies). This outcome mesures the percentge of fmilies who reported spending t lest 1 night in the 6 months before the survey living with friend or reltive becuse they could not find or fford plce of their own. This outcome is intended to mesure episodes in which fmilies reported doubled up living situtions resulting from economic hrdship. This outcome is mesured from responses to the survey. 52 Compring the percent ever used in the first column with the percent used in the followup survey month mkes it pprent tht the medin progrm user for ech of the permnent housing progrm types ws still prticipting in the followup month. This finding implies tht ny difference in medins hs to be becuse of differences in timing of progrm entry (rther thn differences in timing of exit). 53 Such setsides often re prt of the greement with the public housing gency (PHA) tht project-bses the vouchers or with the funder of the cpitl costs of n ffordble housing development. They my provide housing with services (tht is, supportive housing), or they my provide units without supportive services (tht is, the equivlent of SUB s defined for this study). Most of the fmilies who gined quick ccess to project-bsed housing ssistnce were in Knss City, Missouri; Atlnt, Georgi; nd Denver, Colordo. 54 See 24 CFR HUD defines homelessness s living in supervised publicly or privtely operted shelter designted to provide temporry living rrngement (including congregte shelters, trnsitionl housing, nd hotels nd motels pid for by chritble orgniztions or by federl, stte, or locl government progrms for low-income individuls). In the followup survey the study tem sked survey respondents if they hd spent t lest 1 night in the 6 months before the survey stying in shelters, institutions, or plces not typiclly used for sleeping such s on the street, in cr, in n bndoned building, or in bus or trin sttion. The survey question excluded stys in trnsitionl housing. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 48

81 Chpter 5. Description of Usul Cre (UC) nd Outcomes Mesured in the Study Any sty in emergency shelter in months 7 to 18 fter rndom ssignment (percent of fmilies). This mesure is the percentge of fmilies who spent t lest 1 night in emergency shelter during the period 7 to 18 months fter rndom ssignment. This time period is used to mesure return to emergency shelter fter the initil sty in emergency shelter. All fmilies were stying in emergency shelter t the time of rndom ssignment. This mesure is bsed on the Progrm Usge Dt, tken from the Homeless Mngement Informtion System (HMIS; described in Chpter 4). Number of dys homeless or doubled up in pst 6 months. This outcome mesures the verge number of dys spent in shelters or plces not ment for humn hbittion or doubled up in the 6 months before the survey. This outcome is mes - ured from survey dt. Number of dys homeless in pst 6 months. This outcome mesures the verge number of dys spent in shelters or plces not ment for humn hbittion in the 6 months prior to the survey. It is mesured from survey dt. Number of dys doubled up in pst 6 months. This out - come mesures the verge number of dys spent living with friends or reltives in the 6 months prior to the survey. It is mesured from survey dt. Confirmtory Outcome Becuse housing stbility is the centrl outcome domin for the study, the reserch tem designted smll set of impct comprisons nd hypothesis tests relted to housing stbility s the confirmtory set. For this purpose, the nlysts con - structed single composite outcome defined s t lest one return to homelessness from two binry outcomes, mesured from the followup survey nd from Progrm Usge Dt. At lest 1 night spent stying in shelter or plce not ment for humn hbittion or doubled up during the pst 6 months t the time of the survey (mesured from survey dt). Any sty in emergency shelter in the 12 months prior to the dte of the survey (mesured from Progrm Usge Dt). The nlysts used dt from the dult survey to construct outcomes pertining to the type of living rrngements t the time of the followup survey, number of plces lived, nd housing qulity. Living in own house or prtment t time of survey (percent of fmilies). Survey respondents re considered to hve independent housing if they rented or owned their own housing t the time of the survey (housing owned or rented by nonmrried prtner is not counted s living in the respondent s own house or prtment). This outcome mesures the percentge of fmilies who reported living in their own house or prtment, either with or without housing ssistnce. Living in own house or prtment t time of survey with no housing ssistnce (percent of fmilies). This outcome mesures the percentge of fmilies who reported living in their own house or prtment t the time of the survey nd were not receiving housing ssistnce. Living in own house or prtment t time of survey with housing ssistnce (percent of fmilies). This outcome mesures the percentge of fmilies who reported living in their own house or prtment t the time of the survey nd were receiving housing ssistnce to help py the rent. Number of plces lived in pst 6 months. This outcome mesures the number of plces the fmily lived in the 6 months before the survey. The study tem lso mesured impcts on two outcomes relted to the qulity of smple members housing t the time of the followup survey. The first ddresses the dequcy of fmilies living spce. The second mesures the physicl condition of the housing unit. Persons per room. This outcome mesures housing crowding using informtion collected from the dult respondent bout the number of rooms in the housing unit (not counting kitchens, hllwys, nd bthrooms) nd the number of people living in the housing unit. Housing situtions with more thn one person per room re considered crowded. Housing is fir or poor qulity (percent of fmilies). This outcome mesures the percentge of fmilies reporting tht the condition of their housing t the time of the survey ws fir or poor. 55 Housing Stbility of the UC Group Exhibit 5-8 shows the housing stbility outcomes for the UC group. The exhibit displys the verge vlue of ech outcome mesured for the 578 fmilies ssigned to the UC group who responded to the followup survey. 56 The 55 The housing qulity outcome is mesured with self-reported ssessments of housing condition. This outcome should not be interpreted s representing housing qulity s determined by outside inspections, such s those conducted s prt of HUD s Housing Qulity Stndrds process. 56 Outcome vlues re weighted for survey nonresponse, so the responses represent ll fmilies rndomly ssigned to UC. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 49

82 Chpter 5. Description of Usul Cre (UC) nd Outcomes Mesured in the Study Exhibit 5-8. Fmily Options Study: Housing Stbility Outcomes Homeless or doubled up during followup period Outcome Men Vlue Usul Cre Group (Stndrd Devition) At lest 1 night homeless or doubled up in pst 6 months or in shelter in pst 12 months (%) 50.1 (56.9) At lest 1 night homeless or doubled up in pst 6 months (%) 40.2 (55.8) At lest 1 night homeless in pst 6 months (%) 23.9 (48.5) At lest 1 night doubled up in pst 6 months (%) 31.4 (52.8) Any sty in emergency shelter in months 7 to 18 fter rndom ssignment (%) 27.8 (50.9) Number of dys homeless or doubled up in pst 6 months 52.7 (85.0) Number of dys homeless in pst 6 months 21.8 (56.8) Number of dys doubled up in pst 6 months 37.3 (74.3) Housing independence Living in own house or prtment t followup (%) 58.2 (56.1) Living in own house or prtment with no housing ssistnce (%) 34.2 (53.9) Living in own house or prtment with housing ssistnce (%) 24.1 (48.6) Number of plces lived Number of plces lived in pst 6 months 1.8 (1.4) Housing qulity Persons per room 1.6 (1.4) Housing qulity is poor or fir (%) 33.4 (53.6) The definition of homeless in this report includes stys in emergency shelters nd plces not ment for humn hbittion. It excludes trnsitionl housing. Notes: N = 578. See Appendix B for detils on outcome specifictions nd vlues. Mens nd stndrd devitions re weighted to djust for survey nonresponse. Sources: Fmily Options Study 18-month followup survey; Progrm Usge Dt UC group displys substntil housing instbility over the followup period. Bsed on responses to the survey, nerly one-fourth (24 percent) reported hving been homeless in the 6 months prior to the survey; with homeless defined s spending t lest 1 night in shelter or in plces not ment for humn hbittion. Bsed on Progrm Usge Dt, more thn one-fourth (28 percent) of UC fmilies hd styed in emergency shelter t some point in months 7 to 18 fter rndom ssignment. 57 UC fmilies lso reported experiences tht indicte degree of housing instbility. Nerly one-third of UC fmilies (31 percent) sid tht they spent t lest 1 night living with friends or reltives becuse they could not find or py for housing in the 6 months prior to the survey. Altogether, 40 percent of UC fmilies reported spending t lest 1 night homeless or doubled up in the 6 months prior to the survey ( shorter period thn covered by the dministrtive dt). Mesuring housing stbility using the confirmtory outcome reveled tht one-hlf (50 percent) of UC fmilies spent t lest 1 night homeless or doubled up in the 6 months prior to the survey or in emergency shelter in the 12 months prior to the survey. Perhps not surprisingly, given tht the Fmily Options Study smple entered the study while homeless, the housing instbility reported by the UC group is substntilly greter thn tht observed mong the control group in the study of Effects of Housing Vouchers on Welfre Fmilies (Mills et l., 2006). At the time of study enrollment, voucher study prticipnts current or former Temporry Assistnce for Needy Fmilies (TANF) recipients (or TANF eligible) on public housing gency (PHA) witing lists for HCVs were more stbly housed thn fmilies enrolled in the Fmily Options Study. For exmple, 25 percent of the control group in the voucher study reported not hving plce to sty in the yer before the followup survey, 7 percent reported hving lived on the streets or in shelter during tht time period, nd nother 18 percent sid they spent time living with friends or reltives when they did not hve their own plce to sty. At the time of the Fmily Options Study s followup survey, more thn one-hlf (58 percent) of the fmilies in the UC group reported living in their own house or prtment, 34 percent with no housing ssistnce. 58 The fmilies in the UC group reported hving lived in n verge of 1.8 plces in the 6 months prior to the survey. One-third of UC fmilies described the 57 Of the 27.8 percent of fmilies who styed in emergency shelter t some point in months 7 to 18 fter rndom ssignment, 15.0 percent hd returned to shelter fter deprture, nd the remining 12.8 percent hd n initil sty from bseline tht extended until t lest the 7th month fter rndom ssignment. 58 Of those who reported living in their own house or prtment t the time of the survey, 11 percent sid tht they lso hd been homeless or doubled up in the 6 months before the survey. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 50

83 Chpter 5. Description of Usul Cre (UC) nd Outcomes Mesured in the Study qulity of their housing t the time of the followup survey s fir or poor. Men occupncy ws 1.6 people per room. The UC group reported slightly higher housing qulity thn the voucher study control group, which hd 38 percent reporting housing in fir or poor condition t followup Mesures of Fmily Preservtion Fmily preservtion refers to seprtion nd reunifiction of fmily members. The study tem collected detiled informtion bout chnges in fmily composition during the followup period. The interviewers collected nmes nd ges of fmily members with the dult respondent in shelter t the time of rndom ssignment. Interviewers lso collected informtion bout fmily members the dult respondent considered prt of the fmily but who were seprted from the fmily t rndom ssignment. Then, during the followup survey, the study tem collected informtion on the wherebouts of ll fmily members reported t bseline. Informtion ws lso collected bout new fmily members who joined the fmily since the previous survey. The nlysis exmines impcts on five outcomes. The study tem used informtion on chnges in fmily composition to construct outcomes mesuring recent seprtions of fmily members who were present t bseline. Fmily hs t lest one child seprted in pst 6 months (percent of fmilies). This outcome mesures the percentge of fmilies for whom child who hd been with the fmily in shelter ws seprted from the fmily in the 6 months before the followup survey. This outcome includes both forml (tht is, by child welfre gency) nd informl seprtions from the fmily. Fmily hs t lest one foster cre plcement in the pst 6 months (percent of fmilies). This outcome mesures the percentge of fmilies who reported tht child ws in forml foster cre plcement in the 6 months prior to the survey. This outcome excludes informl rrngements in which child my hve styed with friends or fmily members. Spouse/prtner seprted in pst 6 months, of those with spouse/prtner present t rndom ssignment (percent of fmilies). This outcome mesures the percentge of fmilies in which spouse or prtner who hd been with the fmily in shelter t bseline ws seprted from the fmily in the 6 months prior to the survey. The tem lso constructed the following fmily reunifiction outcomes tht mesure the return of fmily members who hd been reported s seprted from the fmily t bseline. Fmily hs t lest one child reunified, of those fmilies with t lest one child bsent t rndom ssignment (percent of fmilies). This outcome mesures the percentge of fmilies in which child who hd been living prt from the fmily t bseline hd rejoined the fmily t the time of the followup survey. This outcome is mesured only for fmilies in which child ws seprted from the fmily t the time of rndom ssignment. Spouse or prtner reunified (percent of fmilies). This outcome mesures the percentge of fmilies in which spouse or prtner who ws seprted from the fmily t bseline hd rejoined the fmily t the time of the followup survey. This outcome is mesured only for fmilies in which spouse or prtner ws seprted from the fmily t the time of rndom ssignment. Fmily Preservtion in the UC Group Exhibit 5-9 presents the vlues of the fmily preservtion outcomes for the usul cre group. Across the UC group, 15 percent of fmilies hd child who ws with the fmily in shelter t the time of the bseline survey who ws seprted from the fmily t some point during the 6 months prior to the followup survey. Only 4 percent of the UC fmilies reported forml plcements in foster cre. The proportion Exhibit 5-9. Fmily Options Study: Fmily Preservtion Outcomes Outcome Current or recent seprtions of fmily members present t bseline Men Vlue (%) Usul Cre Group (Stndrd Devition) Fmily hs t lest one child seprted in pst 6 months (%) (N = 572) 15.4 (41.1) Fmily hs t lest one foster cre plcement in pst 6 months (%) (N = 573) 4.3 (22.9) Spouse/prtner seprted in pst 6 months, of those with spouse/prtner present t RA (%) (N = 161) 36.5 (55.0) Reunifiction of fmily members reported s seprted t bseline Fmily hs t lest one child reunified, of those fmilies with t lest one child bsent t RA (%) (N = 119) 27.1 (50.2) Spouse/prtner reunified, of those with spouse/prtner bsent t RA (%) (N = 54) 39.3 (57.0) RA = rndom ssignment. Notes: See Appendix B for detils on outcome specifictions nd vlues. Mens nd stndrd devitions re weighted to djust for survey nonresponse. Source: Fmily Options Study 18-month followup survey Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 51

84 Chpter 5. Description of Usul Cre (UC) nd Outcomes Mesured in the Study of UC fmilies with seprtions is lower thn the 44 percent of fmilies in the New York City study by Cowl et l. (2002) during longer 5-yer period nd the 44 percent of mothers in emergency shelter nd 39 percent of mothers in trnsitionl housing (during 15-month period) in the Service nd Housing Interventions for Fmilies in Trnsition (SHIFT) study in upstte New York (Hyes, Zonneville, nd Bssuk, 2013). It is fr higher, however, thn the 8 percent of poor fmilies who experienced seprtion during 5 yers in rndom smple of New York s public ssistnce cselod (Cowl et l., 2002). The proportion in foster cre in the UC group is more comprble with the proportion who first received child welfre plcement or preventive services in New York City in Prk et l. (2004); 4.3 percent during 1 yer nd 7.4 percent during 2 yers fter first entry into shelter. Thirty-seven percent of fmilies tht reported spouse or prtner present t bseline experienced the seprtion from tht spouse or prtner in the 6 months prior to the followup survey. Of the fmilies who hd spouse or prtner seprted t bseline, 39 percent reported tht the spouse or prtner hd rejoined the fmily. It is still the cse tht not ll shelters re ble to ccommodte couples, nd some tht cn do not ccommodte couples who re not mrried Mesures of Adult Well-Being The reserch tem included outcomes mesuring severl spects of well-being for the dult respondent in the study fmilies. The outcomes ddress physicl helth, mentl helth, symptoms of trum, substnce use, nd experience with domestic violence. Adult Physicl Helth Helth in pst 30 dys ws poor or fir (percent of fmilies). Adult respondents provided self-reported ssessment of their physicl helth in the followup survey. The outcome mesures the percentge of fmilies in which the dult respondent reported poor or fir helth (rther thn good or excellent helth) in the 30 dys before the survey. Adult Mentl Helth The tem mesured two outcomes relted to dult mentl helth. Gol-oriented thinking. This outcome is mesured with modified version of the Stte Hope Scle (Snyder et l., 1996). Scores rnge from 1 to 6, with higher scores representing higher levels of positive, gol-oriented thinking. Psychologicl distress. The study tem used the Kessler 6 (K6) scle to mesure nonspecific psychologicl distress in the 30 dys prior to the survey (Kessler et l., 2003). The scle rnges from 0 to 24, with higher scores indicting greter psychologicl distress. Adult Trum Symptoms Post-trumtic stress disorder (PTSD) symptoms (percent of fmilies). The outcome mesures the percentge of fmilies in which the dult respondent experienced symptoms of PTSD in the 30 dys prior to the survey. The study tem used responses to the 17 items bout PTSD symptoms from the Posttrumtic Stress Dignostic Scle, or PDS, to mke this determintion. Adult Substnce Use Alcohol dependence (percent of fmilies). The outcome mesures the percentge of fmilies in which the dult re - spondent displyed evidence of lcohol dependence bsed on self-reported informtion t the time of the survey. Adult respondents were sked to report on the four items in the Rpid Alcohol Problem Screen (RAPS4) (Cherpitel, 2000). An ffirmtive nswer to ny of the four items indictes evidence of n lcohol problem. Drug buse (percent of fmilies). The outcome mesures the percentge of fmilies in which the dult respondent showed evidence of drug problem bsed on self-reported informtion t the time of the survey. Evidence of drug problem ws mesured using six items from the Drug Abuse Screening Test, or DAST-10 (Skinner, 1982). An ffirmtive nswer to ny of the items indictes drug problem. Alcohol dependence or drug buse (percent of fmilies). This outcome mesures the percentge of fmilies in which the dult respondent displyed evidence of lcohol dependence or drug buse. Experience of Intimte Prtner Violence Experienced intimte prtner violence in 6 months prior to survey (percent of fmilies). The outcome mesures the percentge of dult respondents reporting experience of intimte prtner violence in the 6 months prior to the survey. Adult Well-Being in the UC Group Exhibit 5-10 shows the men vlues of the dult well-being outcomes for the UC group. Nerly one-third (32 percent) of the dult respondents described their helth s fir or poor. Across the UC group, 20 percent of dult respondents in UC fmilies experienced psychologicl distress nd 26 percent gve survey responses tht indicte symptoms of PTSD. This rte is slightly higher thn rtes of PTSD reported by Bssuk et l. (1996) for homeless fmilies (18 percent) nd housed welfre fmilies (16 percent). These rtes of PTSD Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 52

85 Chpter 5. Description of Usul Cre (UC) nd Outcomes Mesured in the Study Exhibit Fmily Options Study: Adult Well-Being Outcomes Mternl physicl helth Outcome Men Vlue Usul Cre Group (Stndrd Devition) Helth in pst 30 dys ws poor or fir (%) 31.5 (52.8) Mternl mentl helth Gol-oriented thinking 4.38 (1.19) Psychologicl distress score b 7.65 (6.55) Symptoms of serious psychologicl distress (%) 20.5 (45.9) Mternl trum symptoms Post-trumtic stress disorder (PTSD) symptoms in pst 30 dys (%) 25.9 (49.8) Mternl substnce use Alcohol dependence or drug buse c (%) 14.5 (40.0) Alcohol dependence c (%) 11.7 (36.6) Drug buse c (%) 5.6 (26.2) Experience of intimte prtner violence Experienced intimte prtner violence in pst 6 months (%) 11.6 (36.4) Gol-oriented thinking is mesured with modified version of the Stte Hope Scle nd rnges from 1 to 6, with higher scores indicting higher levels of positive, gol-oriented thinking. b Psychologicl distress is mesured with the Kessler 6 (K6) scle nd rnges from 0 to 24, with higher scores indicting greter distress. c Alcohol dependence is mesured with the Rpid Alcohol Problems Screen (RAPS4), nd drug buse is mesured with six items from the Drug Abuse Screening Test (DAST-10). Notes: N = 578. See Appendix B for detils on outcome specifictions nd vlues. Mens nd stndrd devitions re weighted to djust for survey nonresponse. Source: Fmily Options Study 18-month followup survey nd serious psychologicl distress for homeless fmilies re substntilly higher thn ntionl rtes of PTSD (5.2 percent for women nd 1.8 percent for men) 59 nd serious psychologicl distress (3.9 percent for women nd 2.9 percent for men) (CDC, 2012). 60 Alcohol buse in the 6 months prior to the survey ws suggested by dult survey respondents for 12 percent of UC fmilies, nd 6 percent of respondents gve survey responses tht indicted history of drug buse during the sme period. These rtes re substntilly lower thn reported (using dif - ferent mesures) by homeless dults in fmilies to the Ntionl Survey of Homeless Assistnce Providers nd Clients, or NSHAPC (38 percent for drug use problems nd 18 percent for lcohol use problems within the pst yer; Burt et l., 2001). In nother study of homeless fmilies, Rog nd Buckner (2007) reported tht 12 percent of dult respondents hd used illicit drugs in the pst yer. Regrding incidence of intimte prtner violence, 12 percent of the dult respondents in the UC group reported hving experienced intimte prtner violence in the 6 months before the survey. This percentge is substntilly lower thn fig - ures reported in other studies, but other studies report on experience of such violence over longer period of time. Gubits et l. (2013) reported tht, t enrollment, 49 percent of the Fmily Options Study smple reported hving experienced violence during their entire dulthood Mesures of Child Well-Being The study tem collected severl types of dt to mesure outcomes ssocited with child well-being. For ll focl children, prents reported on children s school or childcre enrollment, ttendnce, grdes, grde completion, experiences, behvior t school nd childcre, nd ttitudes bout school nd childcre. Prents lso reported on prosocil behviors nd emotionl nd behviorl problems of children with the Strengths nd Difficulties Questionnire (SDQ; Goodmn, 1997), long with children s helth, ccess to helth cre, nd sleep disruptions, which re ssocited with vriety of emotionl nd behviorl disorders (Dhl nd Hrvey, 2007). Additionl instruments were tuned to children s developmentl stge. Children 12 to 41 months of ge were ssessed with the Ages nd Stges Questionnire (ASQ-3) fmily of questionnires to mesure gross nd fine motor skills, socil development, communiction, nd problem solving, s observed by prents (Squires nd Bricker, 2009). The dult respondent completed the ASQ-3 form. Study 59 The sttistic for PTSD is the ntionl 12-month prevlence rte s mesured in the Ntionl Comorbidity Survey Repliction (NCS-R), which ws fielded in 2001 nd The NCSR used different instrument to mesure PTSD thn the Fmily Options Study (NCS-R, 2005). 60 The sttistic for ntionl rte of serious psychologicl distress is from the 2011 Ntionl Helth Interview Survey. This survey used the sme mesure of psychologicl distress s used in the Fmily Options Study (CDC, 2012b). Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 53

86 Chpter 5. Description of Usul Cre (UC) nd Outcomes Mesured in the Study stff ssessed children from 3 yers, 6 months to 7 yers, 11 months of ge with the Woodcock-Johnson III (WJ III) letter-word identifiction nd pplied problems scles (Wood - cock et l., 2001), which re erly indictors of verbl nd quntittive/nlytic skills, respectively. Children in this ge group lso completed the Hed Toes Knees Shoulders (HTKS) tsk. HTKS ssesses self-regultion, in which children must remember rules nd inhibit incorrect responses (for exmple, by following instructions to touch their hed when the interviewer sys touch your toes ). Finlly, surveys were conducted with children from 8 yers to 17 yers, 11 months of ge mesuring nxiety, fers, nd substnce use. This rry of mesures, long with prentl report, cptured the most likely mentl helth consequences of homelessness nd behviorl responses thereto. Prentl reports of behvior for this ge group included rrests or police involvement. Youth reported on school effort to supple - ment prentl reports of functioning in the key developmentl domin of school. Youth lso completed the Children s Hope Scle (Snyder et l., 1997), mesure of self-efficcy. Child Well-Being Mesures From Prent Report Across Age Groups The reserch tem used the prent-reported informtion on focl children to construct the following child well-being outcomes tht re mesured for children cross ge groups. Child Eduction Preschool or Hed Strt enrollment (percent of focl children). This outcome mesures enrollment in preschool or Hed Strt for children ges 1 yer, 6 months to 5 yers. It mesures the percentge of children ges 1 yer, 6 months to 5 yers who were enrolled in preschool or Hed Strt t the time of the followup survey. School enrollment (percent of focl children). This outcome is mesured for children ges 6 to 17 yers using the prent report. It mesures the percentge of children ges 6 to 17 who were enrolled in school t the time of the followup survey. Childcre or school bsences in pst month. This outcome is mesured from prent reports of the number of bsences from childcre or school in the month prior to the survey. The outcome is mesured using scle of 0 to 3, with 0 in - dicting no bsences nd 3 indicting 6 or more bsences. Number of schools ttended since rndom ssignment. This outcome is mesured from prentl reports nd indictes the totl number of schools child ttended since rndom ssignment. (Chnge in school could be becuse of grde completion, residentil move, or nother reson.) The outcome is mesured using scle of 1 to 4. Number of schools is topcoded t 4 or more schools. Grde completion not held bck (percent of focl chil - dren). This outcome is mesured from the prent survey for children ges 4 to 17 yers who were not reported to be in childcre or preschool t the time of the survey. The outcome mesures the percentge of these children who hve not repeted grde or been prevented from moving on to the next grde since rndom ssignment. Positive childcre or school experiences. This outcome mesures the prent s ssessment of the child s childcre or school experiences, rting them s mostly positive, both positive nd negtive, or mostly negtive. Positive childcre or school ttitudes. This outcome mesures the prent s ssessment of the child s ttitude towrd school or childcre. The prent ws sked to rte how much the child likes school or childcre. The outcome uses 5-point scle rnging from 1 (not t ll) to 5 (very much). School grdes. This outcome mesures the prent s ssessment of the child s grdes for the most recently completed term. The outcome uses 4-point scle rnging from 1 (mostly Ds nd Fs) to 4 (mostly As). Childcre or school conduct problems. This outcome mesures whether or not the prent reports hving been contcted by the child s school or childcre provider regrding the child s conduct problems or if the child ws suspended or expelled. Helth Poor or fir helth (percent of focl children). This outcome mesures the prent s ssessment of the child s helth t the time of the followup survey. The outcome mesures the percentge of children with poor or fir helth (rther thn good or excellent helth), bsed on the prent s ssessment. Well-child checkup in pst yer (percent of focl children). This outcome mesures the percentge of focl children who received physicl exmintion or well-child checkup in the yer prior to the survey, bsed on the prent s report. Child hs regulr source of helth cre (percent of focl children). This outcome mesures the percentge of focl children who hd regulr provider of helth cre t the time of the followup survey, bsed on the prent s report. Sleep problems. This outcome mesures the prent s report on the frequency of two indictors of sleep problems tiredness on wking nd tiredness during the dy. The out - come is mesured using score of 1 to 5, with higher vlues indicting greter frequency of these sleep problems. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 54

87 Chpter 5. Description of Usul Cre (UC) nd Outcomes Mesured in the Study Behviorl Strengths nd Chllenges Behvior problems. This outcome is mesured s the ntion - lly stndrdized score from the SDQ. The SDQ is behviorl nd personlity ssessment. The totl problem score mesures emotionl symptoms, conduct problems, hyperctivity, nd peer problems. Prosocil behvior. Prosocil behvior refers to positive ctions tht benefit others. Prosocil behvior is mesured for the study using the SDQ prosocil domin nd is ntion - lly stndrdized score. Chrcteristics of Focl Children in the UC Group Exhibit 5-11 displys the vlues of the outcomes described previously for focl children in the UC group. School nd Childcre Among children ges 1 yer, 6 months to 5 yers in the UC group, 35 percent were enrolled in center-bsed cre, preschool, or Hed Strt. For 4- to 17-yer-olds, 93 percent were enrolled in school. The percentge of school-ge youth (ges 6 through 17) enrolled (98.3 percent) is comprble with the percentge ntionlly (98.0 to 98.6 percent), lthough this study s smple includes more children t the lower end of the ge rnge. At younger nd older ges, fewer children in the UC group were enrolled in school thn their peers ntionlly; for 5- to 6-yer-olds no longer in childcre, 86.7 percent compred with 95.1 percent ntionlly; for 16- to 17-yer-olds, 91.1 percent compred with 95.7 percent ntionlly (Snyder nd Dillow, 2013: Tble 6). Children in the UC group scored 0.95 on verge for bsences from school or childcre in the pst month, wherein 0 indi - ctes no bsences nd 3 indictes 6 or more bsences. (During the summer, prents reported on the most recent month of enrollment.) Ntionl dt suggest somewht lower rtes of bsenteeism. Trnslting dt from the Ntionl Assessment of Eductionl Progress to our scle, verge numbers of bsences for fourth grders in 2011 were 0.71 (for ll children) nd 0.82 (for children eligible for free nd reduced-price lunch). For eighth grders, the comprble numbers were 0.82 nd 0.89 (Snyder nd Dillow, 2013: Tble 187). Exhibit Fmily Options Study: Child Well-Being Outcomes Mesured for Children Across Age Groups Child eduction Outcome Men Vlue Usul Cre Group (Stndrd Devition) Preschool or Hed Strt enrollment (%) 35.2 (60.0) School enrollment b (%) 93.0 (33.0) Childcre or school bsences in pst month c 0.95 (1.28) Number of schools ttended since rndom ssignment d 1.96 (1.15) Grde completion not held bck (%) 90.4 (38.1) Positive childcre or school experiences e 0.58 (0.72) Positive childcre or school ttitudes f 4.30 (1.28) School grdes g 2.93 (1.26) Childcre or school conduct problems h 0.24 (0.55) Child physicl helth Poor or fir helth (%) 4.6 (26.8) Well-child checkup in pst yer (%) 90.2 (38.1) Child hs regulr source of helth cre (%) 92.8 (33.0) Sleep problems i 2.08 (1.44) Child behviorl strengths nd chllenges Behvior problems j 0.58 (1.61) Prosocil behvior k 0.16 (1.41) Bse for preschool or Hed Strt enrollment is children ges 1 yer, 6 months to 5 yers. b Bse for school enrollment is children ges 4 to 17 yers. c Absences outcome is defined s 0 = No bsences in pst month, 1 = 1 to 2 bsences, 2 = 3 to 5 bsences, nd 3 = 6 or more bsences. d Number of schools outcome is topcoded t 4 or more schools. e Positive childcre or school experiences outcome is defined s 1 = mostly negtive experiences, 0 = both positive nd negtive experiences, nd 1 = mostly positive experiences. f Positive childcre or school ttitudes outcome is prent report of how much child likes school nd rnges from 1 to 5, with higher vlues indicting greter like of school. g School grdes outcome is defined s 1 = mostly Ds or Fs, 2 = mostly Cs, 3 = mostly Bs, nd 4 = mostly As. h Childcre or school conduct problems outcome is defined s 0 = no problems reported or 1 = prent contcted bout behvior or suspension or expulsion from school or childcre center. i Sleep problems outcome rnges from 1 to 5, with higher vlues indicting more frequent tiredness on wking nd during the dy. j Behvior problems outcome is mesured s the stndrdized Totl Difficulties score from the Strengths nd Difficulties Questionnire (SDQ). k Prosocil behvior is mesured s the stndrdized prosocil domin score from the Strengths nd Difficulties Questionnire (SDQ). Notes: Smple sizes vry by outcome. See Appendix B for detils on outcome specifictions nd vlues. Mens nd stndrd devitions re weighted to djust for survey nonresponse. Source: Fmily Options Study 18-month followup survey Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 55

88 Chpter 5. Description of Usul Cre (UC) nd Outcomes Mesured in the Study Children hd ttended 1.96 schools, on verge, since rn - dom ssignment, which mens one school chnge on verge in less thn 2 yers. As described in Chpter 3, school mobility is ssocited with lower levels of cdemic chievement. Only 90.4 percent of children hd completed ll the grdes in which they were enrolled; tht is, 9.6 percent hd been held bck since rndom ssignment. Prents reported tht children hd mostly positive experiences in school (0.58 on 3-point scle wherein 0 indictes both positive nd negtive experiences nd 1 indictes mostly positive experiences). Prents lso reported tht children liked school, verging 4.3 on scle wherein 4 is pretty much nd 5 is very much. Children s grdes, s reported by prents, verged 2.93 on scle wherein 2 is mostly Cs nd 3 is mostly Bs. Most children in the UC group (76 percent) hd no conduct problems t school, but 24 percent hd problems, some of them quite serious; percent hd been suspended or expelled from school nd n dditionl percent hd problems tht led the school to contct the prent. These problems, especilly suspension nd expulsion, vried by ge. Older children were more likely to be suspended or expelled (0.6 percent of 1- to 4-yer-old children, 8.0 percent of 5- to 12-yer-old children, nd 20.9 percent of 13- to 17- yer-old children). These proportions for school-ge children re fr higher thn the ntionl verges of 6.9 percent sus - pended nd 0.21 percent expelled (Snyder nd Dillow, 2013: Tble 193). Helth Prents reported tht 5 percent of children in the UC group were in fir or poor helth, comprble to 5 percent of poor children, but higher thn the 1 percent of nonpoor children younger thn ge 18 in the Ntionl Helth Interview Survey in Fmily Options Study children were somewht less likely to hve regulr source of medicl cre thn poor children generlly (93 versus 95 percent), however (Bloom, Jones, nd Freemn, 2013). Only 90.2 percent of UC children hd received well-child checkup in the pst yer. Prents reported tht children rrely hd trouble wking up or were tired during the dy (2.08 on 5-point scle wherein 2 indi - ctes rrely). Behviorl Strengths nd Chllenges Prents rted their children on the SDQ, stndrdized mes - ure of behviorl strengths nd chllenges. The reported scores re stndrdized by ge nd gender, so tht children cn be compred to their peers in ntionl smple. Children in the UC group scored mrkedly higher thn ntionl norms on behviorl problems (0.58 stndrd devitions in the ntionl dt) nd somewht lower (0.16 stndrd devitions) on prosocil behvior. Child Well-Being Outcomes for Specific Age Groups The study tem constructed the following child well-being outcomes mesured only for specific ge groups. Ages 1 Yer to 3 Yers, 6 Months Met developmentl milestones (percent of focl children). This outcome is defined s the percentge of focl children who scored bove the typicl developmentl thresholds on the five domins mesured in the ASQ-3. Low birth weight (percent of focl children born since rndom ssignment nd t lest 1 yer of ge t the time of the survey). This outcome is mesured for focl children born since rndom ssignment who were 1 yer old or older t the time of the followup survey. It mesures the percentge of these children whose birth weight ws below 5 pounds, 8 ounces, the threshold for low birth weight (CDC, 2014). The outcome uses prent reports of the birth weight for these children. Ages 3 Yers, 6 Months to 7 Yers Verbl bility. This outcome is mesured s the ntionlly stndrdized score from the WJ III letter-word identifiction test. Mth bility. This outcome is mesured s the ntionlly stndrdized score from the WJ III pplied problems test. Executive functioning (self-regultion). This outcome is mesured using the HTKS developmentl ssessment mesuring inhibitory control, ttention, nd working memory. Ages 8 to 17 Yers For focl children between the ges of 8 nd 17, the study tem mesured five outcomes from the child survey nd one from the prent report. Anxiety. This outcome is mesured using the A-Trit scle from the Stte-Trit Anxiety Inventory for Children, or STAIC (Spielberger et l., 1973). Scores rnge from 20 to 60, with higher scores indicting greter nxiety. Fers. This outcome is mesured using the Fers Scle (Rmirez, Msten, nd Sms, 1991). Scores rnge from 33 to 99, with higher scores indicting more fer. Children were sked to indicte the extent to which they hd 33 different fers. Substnce use. This outcome, which combines dt using 23 items from the Centers for Disese Control nd Prevention (CDC) 2011 Youth Risk Behvior Survey, mesures whether Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 56

89 Chpter 5. Description of Usul Cre (UC) nd Outcomes Mesured in the Study the child hd used tobcco, lcohol, or mrijun in the pst 30 dys or hd ever used other substnces cocine, inhlnts, or steroids (ges 8 to 17) or ecstsy, meth, heroin, controlled prescription drugs, or injected drugs (ges 13 to 17 only). Gol-oriented thinking. This outcome is mesured with modified version of the Children s Hope Scle (Snyder et l., 1997) which mesures positive, gol-oriented thinking. Scores rnge from 6 to 30, with higher scores indicting greter hope. School effort in pst month. On the child survey, respondents were sked to report on how hrd they worked in the month before the survey during the school dy nd on homework. The outcome mesure rnges from 1 to 4, with higher scores indicting greter effort t school nd on homework. Arrests or police involvement in pst 6 months (percent of focl children). This outcome is mesured using the prent report bout whether the child hd ny problems tht involved the police contcting the prent nd whether the child hd been rrested in the 6 months before the dult survey. Chrcteristics of Focl Children by Age-Specific Outcomes in the UC Group Exhibit 5-12 displys the vlues of the outcomes described previously for focl children in the UC group. Ages 1 Yer to 3 Yers, 6 Months By prent report, only 77 percent of children pssed the developmentl cutoff score in ll five domins of ASQ-3. Children were lest likely to meet ge stndrds for fine motor development nd most likely to meet stndrds for gross motor development, with performnce in the communiction, problem-solving, nd personl-socil domins flling in between. Prents reported tht 9 percent of bbies born since rndom ssignment hd low birth weight compred to ntionl figure in 2012 of 7.99 percent (CDC, 2014). Ages 3 Yers, 6 Months to 7 Yers Children ges 3 yers, 6 months to 7 yers were ssessed directly with two subscles of the WJ III test of cognitive bilities, with scores compred to ntionl ge norms. Given the lrge ssocition of fmily income with reding nd Exhibit Fmily Options Study: Child Well-Being Developmentl Outcomes for Children in Specific Age Groups Ages 1 yer to 3 yers, 6 months Outcome Men Vlue Usul Cre Group (Stndrd Devition) Met developmentl milestones (%) 77.2 (54.6) Low birth weight b (%) 9.2 (36.3) Ages 3 yers, 6 months to 7 yers Verbl bility c 0.21 (1.38) Mth bility d 0.22 (1.24) Executive functioning (self-regultion) e (22.87) Ages 8 to 17 yers Anxiety f (10.64) Fers g (21.63) Substnce use h (%) 7.71 (38.55) Gol-oriented thinking i (6.97) School effort in pst month j 2.81 (1.07) Arrests or police involvement in pst 6 months k (%) (41.34) Met developmentl milestones outcome is defined s scoring bove the typicl development cutoffs in ll domins of the Ages nd Stges Questionnire (ASQ-3). b Bse for low birthweight outcome (prent report) is children born since rndom ssignment who re t lest 1 yer old t followup. c Verbl bility outcome is the ntionlly stndrdized score from the Woodcock-Johnson III (WJ III) letter-word Identifiction test. d Mth bility outcome is the ntionlly stndrdized score from the WJ III pplied problems test. e Executive functioning outcome is the Hed Toes Knees Shoulders (HTKS) score nd rnges from 0 to 40, with higher scores indicting greter executive functioning. f Anxiety (child report) is mesured using the A-Trit scle from the Stte-Trit Anxiety Inventory for Children (STAIC). Scores rnge from 20 to 60, with higher scores indicting greter nxiety. g Fers outcome (child report) is the score from the Fers Scle nd rnges from 33 to 99, with higher scores indicting more fer. h Substnce use (child report) is mesured with 23 items from the Centers for Disese Control nd Prevention (CDC) 2011 Youth Risk Behvior Survey. i Gol-oriented thinking (child report) is mesured with modified version of the Children s Hope Scle nd rnges from 6 to 30, with higher scores indicting higher levels of positive, gol-oriented thinking. j School effort outcome (child report) rnges from 1 to 4, with higher scores indicting greter effort during school dy nd on homework. k Arrests or police involvement in pst 6 months is from prent report. Notes: Smple sizes vry by outcome. See Appendix B for detils on outcome specifictions nd vlues. Mens nd stndrd devitions re weighted to djust for survey nd ssessment nonresponse. Sources: Fmily Options Study 18-month followup survey; ASQ-3; WJ III; HTKS ssessment; Fmily Options Study 18-month child survey Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 57

90 Chpter 5. Description of Usul Cre (UC) nd Outcomes Mesured in the Study mth bility (for exmple, Miller, Votrub-Drzl, nd Setodji, 2013), it is not surprising tht UC children scored bout one-fifth of stndrd devition below ntionl norms on both letter-word identifiction ( mesure of verbl bility) nd pplied problems (n erly mesure of mth bility). Children lso completed the HTKS test of self-regultion or executive functioning, with men score of out of possible 40. The men score ws substntilly lower thn in normtive smple of lrgely middle-clss children in Michign nd Oregon (27.45 out of 40) of the sme men ge, lthough both the ge rnge nd the vribility of scores for children in our smple were lrger (fll scores from Ponitz et l., 2007). Children in our study lso scored lower thn the verge for low- nd moderte-income children ges 3 yers, 6 months to 8 yers, 6 months in the Supporting Helthy Mrrige Project (20.72) nd bout the sme s in the control group of n intervention study involving lowincome, multircil, multiethnic children in Sn Diego, Cli - forni (Lyzer, 2014) (12.2 in the Fmily Options Study versus 14.8 in the intervention study t 5 yers, nd 29.8 versus 25.5 t 6 yers). Ages 8 to 17 Yers Older children were surveyed bout broder rry of developmentl outcomes. Children s scores on the trit Anxiety mesure, generl mesure of worries, (men of 34.63) were somewht below those in the normtive smple of fourth grders from the test origintors (Spielberger, 1970; 36.3 for mles nd 38.1 for femles) nd further below those in lrge smple of disdvntged Blck children from lrge metropolitn school district (40.0 for mles nd for femles; Ppy nd Hedl, 1978). A score of 40 would reflect n verge nswer of sometimes on 3-point scle from hrdly ever to often cross ll items. Across 33 specific fers, rted on scle from not t ll to lot, children verged slightly below some, or ( consistent nswer of some would yield score of 66). The fers most commonly rted lot (by more thn one-hlf of the youth) were I worry bout my brothers nd sisters, I worry bout my prents, I worry bout myself, guns, nd dying. Lest often fered (less thn 15 percent lot ) were dogs, tht other (children/teens) will not wnt to (ply/ spend time) with the child, police, nd hving no friends. Substnce use in the UC group ws quite low by comprison with ntionl norms. (Our dt re self-report, but so re the corresponding ntionl norms from CDC, 2012). Only 8 percent of children ges 8 to 17 in the UC smple reported hving used tobcco, lcohol, or mrijun in the pst 30 dys or hd ever used more serious drugs. The CDC provides norms for children in grdes 9 through 12, to which the 13- to 17-yer-old youth in the study re compred. Study youth were less likely to hve smoked (8 versus 18 percent), used lcohol (11 versus 39 percent), or used mrijun (10 versus 23 percent) in the pst 30 dys. The CDC group is somewht older thn the group of study youth, but this dis - crepncy would not ccount for the difference. Study youth likely hd less disposble income thn the normtive smple. The mesure of gol-oriented thinking, version of the Children s Hope Scle (Snyder et l., 1997), mesures both belief in one s bility to solve problems nd rech gols nd belief bout esteem or efficcy. The version in this study used slightly different scoring formt thn the originl. The ver - ge score of indictes nswers closer to hving these beliefs most of the time, or 24, thn lot of the time, or 18. For school effort, children rted how hrd they worked on homework nd during the school dy, with n verge of 2.8 on 4-point scle wherein 2 is could hve done little better nd 3 is did bout s well s you could. Prents reported tht 12 percent of children ges 12 to 17 hd hd problem tht involved the police contcting the prent or hd been rrested in the 6 months prior to the survey Mesures of Self-Sufficiency The impct nlysis exmines reltive effects of the four interventions on severl outcomes relted to self-sufficiency. The study tem used the followup survey to construct five ctegories of self-sufficiency outcomes: (1) employment, (2) sources of income, (3) receipt of eduction or trining, (4) food security, nd (5) economic hrdship. Employment The study tem used responses to the dult followup survey to construct four outcomes regrding employment sttus. Work for py in week before survey (percent of fmilies). This outcome mesures the percentge of survey respondents who reported working for py in the week prior to the followup survey. Any work for py since rndom ssignment (percent of fmilies). This outcome mesures the percentge of survey respondents who reported working for py t ny time since rndom ssignment. Months worked for py since rndom ssignment (includes prtil months). This outcome is count of the months worked since rndom ssignment, including prtil months. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 58

91 Chpter 5. Description of Usul Cre (UC) nd Outcomes Mesured in the Study Hours of work per week t current min job. For dult respondents who hd more thn one job in the week prior to the survey, the min job is defined s the job t which she or he usully worked the most number of hours per week. Income Sources nd Amounts The study tem lso constructed outcomes tht mesure the percentge of fmilies who reported receiving income from the following sources in the month prior to the survey. Ernings. Temporry Assistnce for Needy Fmilies (TANF). Socil Security Disbility Insurnce (SSDI). Supplementl Security Income (SSI). Supplementl Nutrition Assistnce Progrm (SNAP). Specil Supplementl Nutrition Progrm for Women, Infnts, nd Children (WIC). The study tem lso constructed two other outcomes relted to income. Annulized ernings from the min job. This outcome mesures the nnulized vlue of current ernings from the job reported t the time of the survey. This vlue usully represents either the product of the reported hourly wge nd usul hours per week multiplied by 52 weeks or the reported usul weekly ernings multiplied by 52 weeks. Totl fmily income. This outcome mesures totl fmily income from ll sources for the clendr yer preceding the survey (2011 or 2012). Eduction nd Trining The nlysts constructed five outcomes pertining to prticiption in eduction nd trining ctivities during the followup period. The dult survey sked respondents whether they hd prticipted in ny eduction or trining ctivities since rndom ssignment nd, if so, how mny weeks they spent in such progrms. For up to six progrms reported, smple members reported on the type of progrm. The study tem used this informtion to construct the following eduction nd trining outcomes. Prticipted in ny school or trining lsting 2 weeks or more since rndom ssignment (percent of fmilies). Number of weeks in trining progrms since rndom ssignment. Prticipted in 2 weeks or more of school since rndom ssignment (percent of fmilies). This outcome mesures the percentge of fmilies in whom the dult respondent reported hving prticipted in school or cdemic trining. School or cdemic trining is defined s regulr high school directed towrd high school diplom, preprtion for generl eductionl development (GED) exmintion, 2-yer college, 4-yer college, or grdute courses. Prticipted in 2 weeks or more of bsic eduction since rndom ssignment (percent of fmilies). Bsic eduction is defined s nonvoctionl dult eduction such s bsic eduction, litercy trining, or English s second lnguge not directed towrd degree. Prticipted in 2 weeks or more of voctionl eduction or trining since rndom ssignment (percent of fmilies). Voctionl eduction or trining is defined s voctionl edu - ction outside college such s business or technicl schools, employer- or union-provided trining, or militry trining in voctionl skills (not militry skills). Food Security nd Hunger The nlysis exmines impcts of the interventions on food security for two outcomes. Household is food insecure (percent of fmilies). This outcome mesures the percentge of fmilies determined food insecure t the time of the followup survey ccording to criteri used by the U.S. Deprtment of Agriculture (USDA). 61 Food insecurity scle. This outcome mesures the food insecurity level of ech fmily bsed on responses to the USDA food security questions included on the followup survey. The food insecurity scle rnges from 0 to 6, with higher vlues indicting greter food insecurity. Economic Stress The nlysts lso mesured the economic hrdship reported by ech fmily t the time of the followup survey. The outcome, expressed s n economic stress scle, mesures the extent of hrdship using responses bout the frequency with which the fmily sid they experienced n inbility to fford medicl cre the fmily needed, clothing the fmily needed, leisure ctivities the fmily wnted, or rent. The economic stress scle lso tkes into ccount the dult respondent s ssessment of the fmily s monthly finnces; tht is, whether they usully hve some money left over t the end of the month, brely enough to mke ends meet, or not enough to mke ends meet. 61 See Nord, Andrews, nd Crlsen (2005). The ssessment of food insecurity is bsed on two USDA short-form metrics, which re scores ssigned to household bsed on nswers to six survey questions. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 59

92 Chpter 5. Description of Usul Cre (UC) nd Outcomes Mesured in the Study Self-Sufficiency Indictors for the UC Group Exhibit 5-13 displys the vlues of the self-sufficiency out - come mesures for the UC group. Nerly one-third (31 per - cent) of the dult respondents in these fmilies reported working for py in the week prior to the survey, nd threefifths (61 percent) sid tht they hd worked for py t some time since rndom ssignment. The dult respondents in the UC group spent n verge of 6.5 months working for py since rndom ssignment. They worked n verge of 9.6 hours per week t the current job, nd nnulized ernings from the current job verged $4,842. The 31 percent of UC fmilies who were working t the time of the survey worked n verge of 31 hours per week, hd hourly ernings of $10.13, 62 nd hd nnulized ernings for the current job of $16,350. UC fmilies report lower rtes of employment thn the Effects of Housing Vouchers on Welfre Fmilies study control group, where 47 percent of controls reported working t the point of followup. Employment for the UC group is similr to tht reported for homeless fmilies in NSHAPC, however. Dt from NSHAPC showed tht 71 percent of people in homeless fmilies did not work in the month before tht survey (Burt et l., 1999). For the fmilies in the UC group, income from ll sources verged just over $9,000 for the clendr yer prior to the survey, slightly higher thn wht ws reported in NSHAPC ($8,172 in 2011 dollrs) nd higher thn reported t bseline. Regrding sources of income in the month prior to the survey, the UC fmilies reported high rte of SNAP receipt Exhibit Fmily Options Study: Self-Sufficiency Outcomes Employment sttus Outcome Men Vlue Usul Cre Group (Stndrd Devition) Work for py in week before survey (%) 31.3 (52.7) Any work for py since RA (%) 60.6 (55.6) Months worked for py since RA 6.5 (8.5) Hours of work per week t current min job b 9.6 (17.8) Income sources nd mounts Annulized current ernings ($) 4,842 (10,438) Totl fmily income ($) 9,067 (8,777) Anyone in fmily hd ernings in pst month (%) 42.3 (56.2) Anyone in fmily received TANF in pst month (%) 30.6 (52.4) Anyone in fmily received SSDI in pst month (%) 7.4 (29.8) Anyone in fmily received SSI in pst month (%) 13.1 (38.4) Anyone in fmily received SNAP/food stmps in pst month (%) 83.4 (42.3) Anyone in fmily received WIC in pst month (%) 29.6 (51.9) Eduction nd trining Prticipted in 2 weeks or more of ny school or trining since RA (%) 25.4 (49.5) Number of weeks in school or trining progrms since RA 3.6 (10.8) Prticipted in 2 weeks or more of school since RA (%) 7.4 (29.8) Prticipted in 2 weeks or more of bsic eduction since RA (%) 1.6 (14.2) Prticipted in 2 weeks or more of voctionl eduction since RA (%) 6.8 (28.7) Food security/hunger Household is food insecure (%) 35.5 (54.4) Food insecurity scle c 1.73 (2.32) Economic stressors Economic stress scle d 0.05 (0.58) RA = rndom ssignment. SNAP = Supplementl Nutrition Assistnce Progrm. SSDI = Socil Security Disbility Insurnce. SSI = Supplementl Security Income. TANF = Temporry Assistnce for Needy Fmilies. WIC = Specil Supplementl Nutrition Progrm for Women, Infnts, nd Children. Number of months worked for py includes prtil clendr months. b Hours of work per week includes those not currently working (tht is, those with 0 hours of work per week). c Food insecurity scle rnges from 0 to 6, with higher vlues indicting higher food insecurity. d Economic stress scle rnges from 1 to 1, with higher vlues indicting higher economic stress. Notes: N = 578. See Appendix B for detils on outcome specifictions nd vlues. Mens nd stndrd devitions re weighted to djust for survey nonresponse. Source: Fmily Options Study 18-month followup survey 62 The verge hourly ernings were clculted for those who reported wges on n hourly, weekly, or biweekly bsis (representing 90 percent of those working for py t the time of the followup survey). Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 60

93 Chpter 5. Description of Usul Cre (UC) nd Outcomes Mesured in the Study (83 percent of fmilies reported receiving SNAP). Other sources of income reported were erned income (42 percent of fmilies), TANF (31 percent), WIC (30 percent), SSI (13 percent), nd SSDI (7 percent). One-fourth of the dult respondents in the usul cre fmilies sid tht they hd prticipted in 2 or more weeks of school or trining since rndom ssignment. On verge, fmilies in the UC group spent 3.6 weeks in eduction or trining ctivities. This prticiption in eduction nd trining is less thn reported by the voucher study control group. In the voucher study, 43 percent of control group members reported prticipting in eduction nd trining during tht study s much longer (4- to 5-yer) followup period. Despite the high rte of SNAP receipt reported by UC fmilies, more thn one-third (36 percent) met USDA criteri for food insecurity t the time of the followup survey. Food insecurity mong the voucher study control group ws even higher (42 percent), nd receipt of SNAP ws less (65 percent). Food insecurity verged 1.73 on scle of 0 to 6 (with higher vlues indicting greter food insecurity) for UC fmilies, lower thn for the voucher control group (3.29). Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 61

94 CHAPTER 6. IMPACTS OF PERMANENT HOUSING SUBSIDY (SUB) COMPARED WITH USUAL CARE (UC) This chpter presents estimtes of the impct of the permnent housing subsidy (SUB) intervention compred with outcomes of fmilies served by the usul cre (UC) homeless ssistnce systems in their communities. The gol is to determine whether being offered subsidy on priority bsis (tht is, not hving to enroll in nd rech the top of witing lists for subsidy ssistnce) increses fmilies housing stbility nd improves other fmily outcomes during 20-month followup intervl. The chpter begins with description of the SUB intervention s implemented in the study. It then shows the extent to which fmilies in both the SUB nd UC groups used permnent subsidies nd other housing nd service progrms vilble to them in the study sites. The next five sections present the effects of being offered the SUB intervention (s compred with UC) on outcomes within the five study domins housing stbility, fmily preservtion, dult well-being, child well-being, nd self-sufficiency. 6.1 Permnent Housing Subsidy (SUB) Intervention The SUB intervention provided indefinite rentl ssistnce, typiclly in privte-mrket housing. The intervention could include housing plcement ssistnce but not ongoing socil services. SUB ws vilble in 10 of the 12 study sites. The subsidies were provided by 18 locl nd stte public housing gencies (PHAs), with some sites hving more thn 1 pr tici - pting PHA. In totl, 599 fmilies were ssigned to SUB, rnging from 32 in Louisville, Kentucky, to 76 ech in Almed County, Cliforni, nd Denver, Colordo. Of these 599 fmilies, 530 (88 percent) responded to the 18-month followup survey nd so re included in the impct nlysis documented in this report Housing Assistnce in SUB All the housing ssistnce included in the SUB intervention is considered permnent; tht is, fmilies cn continue re - ceiving housing ssistnce s long s they remin eligible nd follow progrm rules, such s pying their shre of the rent nd living in housing tht psses housing qulity in - spection. In ll sites, recipients of the subsidy were subject to nnul recertifiction of income to determine the tennt s shre of the rent nd the mount of the housing ssistnce pyment mde to the owner of the housing. In 8 of the 10 sites (comprising 92 percent of fmily referrls), the SUB intervention ws tennt-bsed voucher provided by one or more PHAs through the Housing Choice Voucher (HCV) progrm, s shown in Exhibit 6-1. One of the sites offered permnent housing subsidies through public housing units (6 percent of fmily referrls), 63 nd nother offered project-bsed vouchers (3 percent of fmily referrls). 64 The HCV progrm is the federl government s lrgest housing ssistnce progrm, providing rentl subsidies to more thn 2 million households cross the country. 65 Prticipnts in the study who were ssigned to the SUB intervention, ccepted by the PHA, nd issued voucher were free to use the voucher to rent housing unit of their choice in the privte rentl mrket s long s it met HUD s Housing Qulity Stndrds 63 In Honolulu, Hwii, the subsidy intervention consisted of 33 units of public housing provided by the stte PHA nd 10 units of tennt-bsed rentl ssistnce provided by the city Deprtment of Community Services. Public housing units re owned nd mnged by the PHA. Like voucher progrm prticipnts, tennts in public housing py 30 percent of djusted monthly income for rent. The city s tennt-bsed rentl ssistnce progrm tht provided five SUB units for the study opertes much like the HCV progrm. 64 In Bridgeport, Connecticut, the subsidy intervention ws provided through 15 units of project-bsed vouchers. PHAs cn use up to 20 percent of their HCV progrm funding for project-bsed ssistnce, under which PHA enters into n ssistnce contrct with property owner for specified units nd for specified term. Recipients of this type of ssistnce lso py 30 percent of monthly income for rent Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 62

95 Chpter 6. Impcts of Permnent Housing Subsidy (SUB) Compred With Usul Cre (UC) Exhibit 6-1. Subsidy Type Provided by Site Type of Subsidy Number of Prticipting Subsidy Progrms With This Type of Subsidy Percent of Fmilies Assigned to Subsidy Intervention of This Type (N = 599) Tennt-bsed voucher Project-bsed voucher 1 3 Public housing 1 6 Note: Percentges do not dd to 100 becuse of rounding. Sources: Progrm dt; rndom ssignment enrollment dt nd hd rent tht the PHA determined to be resonble when compred with the rents of unssisted units in the sme housing mrket. The voucher ssistnce subsidized the monthly rent for the unit, nd the mount provided by the subsidy ws the pyment stndrd estblished by the PHA (or the unit s ctul rent, if lower) minus 30 percent of the fmily s djusted monthly income Supportive Services in SUB The SUB intervention ws intentionlly selected to provide n intervention without ongoing, designted, intensive sup - portive services. Upfront housing plcement ssistnce ws llowed, however. Of the 18 prticipting subsidy providers, only two progrms Honolulu, Hwii s public housing progrm nd Bridgeport, Connecticut s project-bsed voucher progrm indicted they provide ny cse mngement ser - vices. These two represented only 8 percent of the study referrls (nd lso 8 percent of SUB followup survey respon - dents). Only 20 percent of fmilies were referred to PHAs tht indicte tht they help pplicnts locte qulified housing units, 20 percent of fmilies were referred to PHAs tht provide ssistnce resolving conflicts with lndlords, nd n even smller percentge were referred to progrms tht provide moving ssistnce or help in lerning how to mintin the unit. PHAs did not lter their usul prctices for providing help to study fmilies. In some cses, emergency shelter stff ssisted fmilies ssigned to SUB to obtin ssistnce for pying PHA rrerges or move-in costs. Fmilies who receive voucher ssistnce cn, of course, ccess ny vilble services in the community on their own Eligibility Criteri for SUB All PHA-dministered subsidy progrms hve sttutory eligi - bility criteri tht require prospective fmilies to be ble to document U.S. citizenship or legl sttus, bsence of drugrelted criminl convictions, lck of previous evictions from federlly funded housing progrm, nd bsence of rrerges to the PHA. 67 In some cses, SUB progrms sked the reserch tem to dd eligibility screening criteri beyond those sttu - torily required, such s question relted to whether the fmily hd consistent source of income (two SUB progrms required this question), willingness to reside within the PHA s jurisdiction (two SUB progrms required this question), nd bility to py security deposits nd other strtup costs (one SUB progrm required this question). Exhibit 2-11 in Gubits et l. (2013) shows informtion on the percentge of fmilies referred to SUB progrms with these requirements. In some sites, PHAs hve residency requirements for the HCV progrm, so fmilies who receive voucher must use it in designted jurisdiction for specified period of time. For exmple, voucher recipients in the Oklnd Housing Authority s progrm could use the voucher only in the city of Oklnd, Cliforni, for the first yer of ssistnce. After tht time, prticipnts could move with the voucher to nother PHA jurisdiction. After enrolling in SUB progrm, tennts remin eligible for the subsidy ssistnce indefinitely. Their incomes re recertified nnully nd must remin low enough to qulify them for subsidy vlue greter thn $0, nd they must py their shre of the rent nd not engge in other lese violtions. The most common form of the subsidy, n HCV, is portble, nd the fmily my use it to move to different housing unit. 6.2 Progrm Use by Fmilies in the Permnent Housing Subsidy (SUB) Versus Usul Cre (UC) Comprison Ech impct comprison in the study my be thought of s distinct experiment or test, nd this chpter ddresses only the comprison between SUB nd UC, without reference to the fmilies who were rndomized to the community-bsed rpid re-housing (CBRR) nd project-bsed trnsitionl hous - ing (PBTH) interventions. In totl, 1,039 fmilies took prt 66 Pyment stndrds re djusted for the number of bedrooms in the unit. The ctul rent includes n estimte of the cost of utilities pid for by the tennt. Detils regrding the clcultion of housing ssistnce pyments under the HCV progrm re in 24 CFR Prt Although ll SUB progrms used these eligibility criteri, not ll SUB progrms sked the study tem to screen prospective study prticipnts for these items before rndom ssignment. This ccounts for the discrepncy between Exhibit 2-11 in Gubits et l. (2013) nd this sttement. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 63

96 Chpter 6. Impcts of Permnent Housing Subsidy (SUB) Compred With Usul Cre (UC) in the test of SUB versus UC. These fmilies ll hd the oppor - tunity to be ssigned to SUB or UC t the point of rndom ssignment nd were ssigned to one of these two interventions; 599 fmilies were ssigned to SUB nd 540 fmilies were ssigned to UC. 68 Of these 1,039 fmilies (599 SUB fmilies nd 540 UC fmilies), 530 SUB fmilies nd 415 UC fmilies (91 percent) responded to the 18-month followup survey. Therefore, 945 fmilies re included in the SUBversus-UC impct comprison reported in this chpter. The current section describes the extent to which the 530 SUB fmilies used the SUB intervention nd other types of homeless nd housing ssistnce during the followup period. Prllel informtion is presented for the 415 UC fmilies. Exhibit 6-2 shows the use by these 945 fmilies of seven types of homeless nd housing progrms. The first column shows the percentge of fmilies ssigned to SUB who ever used ech progrm type during the followup period. 69 The top row (shded in the exhibit) shows the tkeup of SUB by the fmilies ssigned to tht intervention; 84.2 percent of fmilies ssigned to SUB received the SUB intervention t some point during the followup period mening tht they successfully lesed housing unit with voucher or moved into n ssisted housing unit. 70 The second column correspondingly shows the percentge of fmilies ssigned to UC who ever used ech progrm type during the followup period. 71 The top row of the second column shows tht 12.4 percent of the UC fmilies received the SUB intervention during the followup period, presumbly through the regulr process of coming off witing lists nd lesing units. The next six rows of the exhibit show prticiption in other types of homeless nd housing ssistnce progrms. First, consider the other two interventions considered in this study. Of UC fmilies, 20 percent found their wy to rpid re-housing ssistnce during the followup period, nd 21 percent found their wy to trnsitionl housing. The use of progrms other thn SUB is lwys higher for the UC group thn for the SUB group, presumbly becuse the UC group did not hve the SUB intervention esily vilble nd so turned to other types of progrms. The eighth row shows the percentges of fmilies in the SUB nd UC groups who used none of the six types Exhibit 6-2. SUB Versus UC: Progrm Use Since Rndom Assignment Type of Housing Assistnce Percent Ever Used From RA to 18-Month Followup Survey Number of Months Used From RA to 18-Month Followup Survey, if Ever Used Type of Housing Assistnce Percent Used in Month of Followup Survey Response SUB UC SUB UC SUB UC Men Medin Men Medin Subsidy (SUB) b Rpid re-housing Trnsitionl housing Permnent supportive housing Public housing Project-bsed vouchers/section 8 projects Emergency shelter c No use of homeless or housing progrms d N SUB = permnent housing subsidy. UC = usul cre. RA = rndom ssignment. Percentge of fmilies who ever used type of ssistnce progrm during the period from the month of RA to the month of 18-month followup survey response (medin period durtion: 21 months). Percentges do not dd to 100 becuse some fmilies used more thn one progrm type during the followup period. b Subsidy ssistnce is housing choice vouchers plus site-specific progrms offered to fmilies ssigned to SUB group in Bridgeport, Connecticut, nd Honolulu, Hwii. c All fmilies were in emergency shelter t rndom ssignment. Percentges less thn 100 percent for ever used emergency shelter re becuse of missing dt on shelter use. d No use of homeless or housing progrms (ever used) indictes no use of the six progrm types in this tble during ny of the followup period nd no use of emergency shelter fter the first 6 months fter RA. No use in the month of followup survey response indictes no use of ny of these seven progrm types. Notes: Percentges re regression djusted, controlling for site nd rndomiztion rtio. Percentges, mens, nd medins re weighted for survey nonresponse to represent full comprison smple. Source: Fmily Options Study Progrm Usge Dt 68 In the entire study, 746 fmilies were rndomly ssigned to UC. Only 540 of these fmilies hd SUB vilble to them when they were rndomized, however. Therefore, only these 540 UC fmilies re prt of the SUB-versus-UC comprison smple. All 599 fmilies rndomly ssigned to SUB during the course of the study hd UC vilble to them, so ll re prt of the SUB-versus-UC comprison smple. 69 The followup period is from the clendr month of rndom ssignment through the clendr month of response to the 18-month followup survey. Therefore, the length of the followup period differs cross fmilies. This period lsts for medin of 20 months for the full smple. Anlysis of Progrm Usge Dt used dt for medin of 21 clendr months for the full smple. 70 All percentges, mens, nd medins in the exhibit re weighted to djust for survey nonresponse nd hence s best possible represent the full experimentl smple of 1,039 fmilies. The findings on progrm use re thus in line with similrly weighted impct estimtes provided subsequently in the chpter. 71 The percentges in the first six rows of these columns re not mutully exclusive becuse some fmilies use more thn one progrm type during the followup period. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 64

97 Chpter 6. Impcts of Permnent Housing Subsidy (SUB) Compred With Usul Cre (UC) of progrms shown on the exhibit during the followup period nd did not use emergency shelter from the seventh month fter rndom ssignment onwrd. About 5 percent of SUB fmilies nd 27 percent of UC fmilies fll into this group. The men nd medin number of months of use for ech progrm type re lso shown in the exhibit (third nd fourth columns for SUB fmilies, fifth nd sixth columns for UC fmilies) for only those fmilies who ever used given progrm type. 72 As one might expect, given tht housing subsidies were redily vilble to SUB fmilies, the number of months of SUB intervention use is higher for prticipting SUB fm - ilies (medin of 18 months) thn for the 12.4 percent of UC fmilies who received the SUB intervention by coming off witing lists for ssisted housing (medin of 11 months). The medin length for the SUB group implies tht the typicl SUB fmily who took up the intervention enrolled few months fter rndom ssignment nd kept the subsidy for the entire study period. The typicl UC fmily who received SUB did so bout 10 months fter rndom ssignment nd then kept the subsidy for the reminder of the study period. Wheres the previous columns consider ll experience from between rndomiztion nd the survey, the lst two columns consider the progrm use s of the month of the survey. Although most outcomes in the report re expected to be influenced by ssistnce received during the entire followup period, some outcomes will be prticulrly strongly influenced by ssistnce received t the time of followup survey response. The lst two columns of the exhibit show the percentges of SUB nd UC fmilies who received ech type of progrm in the clendr month of the followup survey response. The first row of the seventh column shows tht 75 percent of SUB fmilies received SUB ssistnce in the month they responded to the followup survey. The mjority of UC fmilies (56 percent) were not prticipting in home - less or housing progrm t the time they responded to the followup survey compred with only 18 percent of SUB fmilies. Some differences in the outcomes of SUB nd UC fmilies my reflect the fmilies current experience, rther thn lsting influence of ssistnce provided erlier in the followup period. As Exhibit 6-2 mkes cler, the SUB fmilies used rnge of progrms in ddition to the progrm to which they were referred by the study, which is consistent with the design of the study. Fmilies were not required to use the intervention to which they were ssigned nd were lso not forbidden from using other progrms tht were vilble to them in their community. The intent of the study ws to mximize use of the ssigned ctive intervention (in this cse, mximize use of the SUB intervention by the SUB fmilies) nd thereby to crete s wide contrst s possible between progrm mixes for the different ssignment groups (in this cse, SUB versus UC). As shown in the exhibit, the mix of progrms used ws very different for the SUB group thn for the UC group. The contrst in usge of SUB of 84.2 percent for SUB fmilies compred with 12.4 percent for UC fmilies is lrge. Adding together the four rows with permnent housing sub - sidies (SUB, permnent supportive housing, public housing, nd project-bsed ssistnce), the contrst between the two groups ws 87.1 compred with 29.6 percent, respectively, somewht nrrower gp but still quite lrge. This difference in the use of permnent housing subsidies by the SUB nd UC groups is lrge enough tht concerns bout the study s bility to detect n impct in the presence of nonprticiption nd crossover re miniml. Additionl detil bout the use of the SUB intervention by SUB fmilies is shown in Exhibit 6-3. This exhibit shows tht by fr most SUB fmilies who used the SUB intervention did so for 12 or more months. The reminder of the chpter reports estimted impcts in the vrious outcome domins tht if sttisticlly significnt cn be cuslly ttributed to the offer of permnent housing subsidy to the fmilies rndomly ssigned to SUB t the strt of the followup period s opposed to no such privileged ccess being vilble to UC fmilies. 72 Hence, 0 vlues re not fctored into the mens, nor do they pull downwrd the medins of the vrious distributions. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 65

98 Chpter 6. Impcts of Permnent Housing Subsidy (SUB) Compred With Usul Cre (UC) Exhibit 6-3. Number of Months of Subsidy Receipt During Followup Period by SUB Fmilies Who Ever Used SUB 40 Percent of SUB fmilies who ever used SUB Number of months of SUB receipt or more SUB = permnent housing subsidy. Percentges re weighted for survey nonresponse to represent ll fmilies in comprison smple. Note: N = 446. Source: Fmily Options Study Progrm Usge Dt 6.3 Impcts on Housing Stbility in the Permnent Housing Subsidy (SUB) Versus Usul Cre (UC) Comprison As discussed in Chpter 3, proponents of SUB view the crisis of housing ffordbility s the root cuse of homelessness mong fmilies. These observers believe tht, becuse fmilies who experience homelessness re very poor, they re likely to require long-term rentl subsidies to become stbly housed. The SUB-versus-UC comprison in the current study provides direct test of this clim by mesuring the effects of mking the SUB intervention esily vilble to fmilies compred with sitution in which permnent hous - ing subsidies re reltively difficult to ccess in the ner term. Wht do estimtes of the effects of SUB on housing stbility tell us? Exhibit 6-4 shows the experimentlly bsed evidence of mesured effects on homelessness, housing independence, residentil moves, nd housing qulity. All of the rows of the exhibit (nd other impct exhibits in the blnce of this report) hve the sme formt. The first three columns of the exhibit provide informtion bout the SUB fmilies the number of fmilies with dt on prticulr outcome nd the men vlue nd stndrd devition of the outcome. The next three columns provide the corresponding informtion for the UC fmilies included in this prticulr pirwise comprison. 73 The seventh column is the difference between the men vlue (or proportion) of the SUB fmilies nd the men vlue (or proportion) of the UC fmilies, referred to s the impct of SUB reltive to UC. 74 Asterisks to the right of this column denote the sttisticl significnce of the impct estimte, with more sterisks indicting higher levels of sttisticl significnce. The eighth column of the exhibit contins the stndrd error of the impct estimtes, which is used to test for sttisticl significnce nd cn be used to construct confidence intervl round the impct estimte. The lst column shows the stndrdized effect size of the impct, clculted by dividing the impct estimte by the stndrd devition of the outcome for ll fmilies ssigned to UC. 75 The stndrdized effect size is thus mesure of 73 The UC fmilies in this comprison re those who could hve been rndomized to SUB. The men vlues of outcomes for ll UC fmilies re shown in Chpter As explined in Chpter 4, the men vlues nd the impct estimte re regression djusted for bseline covrites. 75 The stndrd devitions for the entire UC group re shown in Chpter 5. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 66

99 Chpter 6. Impcts of Permnent Housing Subsidy (SUB) Compred With Usul Cre (UC) Exhibit 6-4. SUB Versus UC: Impcts on Housing Stbility Outcome Homelessness or doubled up during the followup period SUB UC ITT Impct Effect size N Men (SD) N Men (SD) Impct (SE) At lest 1 night homeless b or doubled up in pst 6 months or in shelter (41.2) (50.1) 28.0*** (3.1) 0.49 in pst 12 months (%) [Confirmtory] c At lest 1 night homeless b or doubled up in pst 6 months (%) (37.3) (49.0) 24.9*** (3.0) 0.45 At lest 1 night homeless b in pst 6 months (%) (31.5) (43.7) 15.9*** (2.6) 0.33 At lest 1 night doubled up in pst 6 months (%) (33.3) (45.6) 18.4*** (2.7) 0.35 Any sty in emergency shelter in months 7 to 18 fter RA (%) (34.9) (45.9) 12.9*** (2.6) 0.25 Number of dys homeless b or doubled up in pst 6 months (52.2) (74.6) 31.2*** (4.4) 0.37 Number of dys homeless b in pst 6 months (36.9) (51.3) 13.5*** (3.1) 0.24 Number of dys doubled up in pst 6 months (39.0) (63.5) 21.5*** (3.6) 0.29 Housing independence Living in own house or prtment t followup (%) (44.5) (49.7) 15.1*** (3.0) 0.27 Living in own house or prtment with no housing ssistnce (%) (29.3) (46.2) 23.0*** (2.8) 0.43 Living in own house or prtment with housing ssistnce (%) (48.2) (43.7) 38.1*** (3.0) 0.78 Number of plces lived Number of plces lived in pst 6 months (1.0) (1.2) 0.4*** (0.1) 0.26 Housing qulity Persons per room (0.8) (1.2) 0.4*** (0.1) 0.28 Housing qulity is poor or fir (%) (43.4) (47.6) 8.7*** (3.1) 0.16 SUB = permnent housing subsidy. UC = usul cre. ITT = intention-to-tret. RA = rndom ssignment. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire UC group. b The definition of homeless in this report includes stys in emergency shelters nd plces not ment for humn hbittion. It excludes trnsitionl housing. c After djustment for multiple comprisons, the impct on the confirmtory outcome is sttisticlly significnt t the.01 level for the SUB-versus-UC comprison. Notes: Impct estimtes nd outcome mens re regression djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Sources: Fmily Options Study 18-month followup survey; Progrm Usge Dt impct reltive to nturl vribility in the outcome. Such stndrdized effect sizes re conventionl wy to compre impct mgnitudes cross outcomes nd domins with different scles. For exmple, one my compre the stndrdized effect sizes for housing stbility outcomes in Exhibit 6-4 with those for other outcomes in other domins shown in this chpter. Stndrdized effect sizes my lso llow for the size of effects found in this study to be compred with the size of effects in other studies. Exhibit 6-4 shows tht the SUB intervention reduced the experience of subsequent stys in shelter or plces not ment for humn hbittion during the 20-month followup period by lrge mount. The first row of the exhibit shows evidence for the confirmtory outcome of the study: (1) t lest 1 night in shelter or plce not ment for humn hbittion or doubled up in the pst 6 months (from the followup survey), or (2) t lest 1 night in emergency shelter in the pst 12 months (from the study s Progrm Usge Dt). Of the fmilies in the UC smple, 50 percent experienced one of these two situtions. For the SUB group, tht proportion declined to 22 percent, representing reduction in homelessness of 28 percentge points nd hence eliminting more thn one-hlf of the home - lessness cptured by this mesure. This impct is highly sttisticlly significnt (even fter the djustment for multiple comprisons). 76 The following discussion ddresses estimted impcts for three outcomes constructed solely from survey dt: (1) t lest 1 night homeless or doubled up in the pst 6 months, (2) t lest 1 night homeless in the pst 6 months, nd (3) t lest 1 night doubled up in the pst 6 months. The impct estimtes in these three rows of the exhibit show tht, compred with UC, the SUB intervention cused sub - stntil, sttisticlly significnt reductions in ll three of these survey-bsed mesures of homelessness The study estimtes impcts on this confirmtory outcome for ech of the six pired comprisons nd four pooled comprisons. Seven of these estimtes hve been prespecified s confirmtory tests. A multiple comprison procedure is performed to compute djusted p-vlues for these tests to reduce the possibility of chnce findings of sttisticl significnce. The detils of this procedure re provided in Appendix C. 77 All impcts in this tble with the exception of the first row re considered explortory nd re not djusted for the presence of multiple comprisons. Likewise, ll impcts in other study domins re lso considered explortory. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 67

100 Chpter 6. Impcts of Permnent Housing Subsidy (SUB) Compred With Usul Cre (UC) The fifth row of Exhibit 6-4 shows the impct on ny sty in emergency shelter in months 7 to 18 fter rndom ssignment (mesured lrgely from dministrtive dt). 78 About 28 per - cent of UC fmilies spent t lest 1 night in emergency shelter during the yerlong period beginning 6 months fter rndom ssignment. Only 15 percent of SUB fmilies spent t lest 1 night in emergency shelter during this time, reduction of 13 percentge points. Thus, the shelter usge rte ws cut nerly in hlf by the vilbility of housing subsidy. Exhibit 6-5 provides more detiled chrcteriztion of the effect of SUB reltive to UC on emergency shelter stys. It shows month-by-month impcts on emergency shelter stys. As discussed in Chpter 5, the Progrm Usge Dt re miss - ing the initil shelter sty for bout 20 percent of fmilies. We hve no reson to believe, however, tht missing dt rtes re ssocited with rndom ssignment group (tht is, they re eqully likely to be missing for the SUB group s for the group ssigned to UC). The dt cn therefore be used to clculte estimtes of impcts without concern for bis. As cn be seen, gp between the shelter use of SUB nd UC fmilies begins to emerge in the third month fter study entry nd reches 7 percentge points by the fifth month, with 18 percent of SUB fmilies hving t lest 1 night in shelter in the fifth month fter rndom ssignment compred with UC rte of 25 percent. 79 A difference emerging in the third month is consistent with wht one might expect, given tht those ssigned to SUB needed to go through the process of using the progrm (hving their incomes verified, finding nd lesing unit) nd my hve remined in shelter during tht process. This gp of 6 to 9 percentge points remins through the 18th month. From the 10th month onwrd, the shre of SUB fmilies in shelter is much less thn one-hlf the proportion of UC fmilies, notble reduction. Exhibit 6-5. SUB Versus UC: Percent of Fmilies With t Lest 1 Night Sty in Emergency Shelter During Month, by Month After RA 100% 90% 80% 70% Percent of fmilies 60% 50% 40% 30% 20% UC 10% SUB 0% Month fter RA SUB = permnent housing subsidy. UC = usul cre. RA = rndom ssignment. Notes: Percentges re weighted for survey nonresponse to represent ll UC fmilies in the study. Missing dt on emergency shelter stys bises the percentges somewht downwrd. The bseline sty in emergency shelter does not pper in the dt for 18.7 percent of UC respondent fmilies. The missing dt rte for subsequent stys in emergency shelter is unknown. Source: Fmily Options Study Progrm Usge Dt 78 Outcomes regrding shelter stys re bsed on study Progrm Usge Dt (which re described in Chpter 4 nd Appendix A). 79 Although most of the fmilies using emergency shelter in the fifth month fter rndom ssignment hd not yet deprted from shelter, few hd deprted nd returned. The proportion of ll study fmilies in shelter who hd deprted nd returned increses with time since rndom ssignment. In the 13th month fter rndom ssignment, the proportion of SUB fmilies in shelter who hve deprted nd returned rises bove one-hlf. It does so for the UC fmilies in this comprison 1 month before. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 68

101 Chpter 6. Impcts of Permnent Housing Subsidy (SUB) Compred With Usul Cre (UC) The lst three homelessness outcomes exmined (see gin Exhibit 6-4) mesure the number of dys in the pst 6 months tht fmily ws homeless or doubled up. Assignment to SUB reduced time spent homeless or doubled up by n ver ge of 4 weeks in the pst 6 months reltive to UC. This difference is entirely ccounted for by the difference in pro - portions of fmilies who experienced homelessness or doubling up in the pst 6 months (shown in the second row of the exhibit). The 16 percent of SUB fmilies nd 41 percent of UC fmilies who experienced ny homelessness or doubling up were in these conditions for combined 18 weeks during the pst 6 months. 80 The housing independence outcomes in the next pnel of Exhibit 6-4 mesure whether fmily lived in its own house or prtment t followup, either with or without housing ssistnce. The SUB intervention incresed the proportion of fmilies living in their own dwelling plce from 58 to 73 percent reltive to UC. This difference is the net result of two opposing effects. As would be expected, the proportion of SUB fmilies living in their own plces without housing ssistnce (10 percent) is much lower thn the corresponding proportion for UC (33 percent). By contrst, nd more thn offsetting tht decline, the proportion of SUB fmilies living in their own plces with housing ssistnce (63 percent) is much higher thn the corresponding proportion for UC (25 percent). 81 The stbility offered by the SUB intervention lso reduced the verge number of plces lived in during the pst 6 months from 1.8 to Becuse this outcome hs lower bound of 1, the UC men of 1.8 compred with 1.4 for the SUB group mens tht the SUB intervention cut the number of moves during the finl 6 months of the followup period in hlf. The lst two rows in Exhibit 6-4 show how the SUB intervention ffects the nture of the housing occupied by study fmilies t the point of followup survey by considering the number of persons per room ( mesure of crowding) nd residence in poor qulity housing. Persons per room is stndrd proxy for overcrowding nd therefore for housing qulity. The SUB intervention reduced the number of persons per room from 1.6 to 1.2. The SUB intervention similrly reduced the proportion of fmilies living in units of poor or fir qulity from 34 to 25 percent. As SUB ppers to be highly effective in preventing subsequent stys in shelter or plces not ment for humn hbittion nd being doubled up, the reder my wonder whether the incidence of these situtions within the SUB group is entirely mong those fmilies who never used SUB. Exhibit 6-6 shows comprison of SUB fmilies who never took up their ssigned intervention with SUB fmilies who did use SUB. The right column of the top pnel shows tht smll proportion (bout 7 to 11 percent) of SUB fmilies who used SUB nevertheless experienced homelessness or being doubled up during the period between 7 nd 18 months fter rndom ssignment, presumbly fter the period of SUB use hd ended. 83 Wheres SUB is usully vilble to fmilies only fter sometimes lengthy wit, it ws vilble in this study on priority bsis to those fmilies ssigned to SUB. This priority ccess enbled SUB fmilies to exit shelter fster thn UC fmilies. As shown in Exhibit 6-7, SUB fmilies on verge hd shorter initil stys in emergency shelter thn UC fmilies (3.2 months for SUB compred with 3.7 months for UC). In sum, the SUB intervention hd strong, positive effect on housing stbility s compred with UC for every mesure considered. 80 Dividing the verge number of dys spent homeless or doubled up in the pst 6 months for SUB fmilies by the percentge who experienced either stte (20.4 dys/0.160 = dys) revels tht those who did experience either stte spent 128 dys on verge either homeless or doubled up in the pst 6 months. Performing the sme clcultion for UC fmilies (51.5 dys/0.409 = dys) revels tht UC fmilies who experienced either stte spent nerly the sme mount of time (126 dys) on verge either homeless or doubled up in the pst 6 months. 81 Although the survey response indictes tht 63 percent of the SUB fmilies were living in their own house or prtment with housing ssistnce t the time of the survey, the Progrm Usge Dt show tht the proportion of fmilies using SUB, public housing, or project-bsed vouchers in the survey month is 77 percent. This discrepncy between response to the survey item nd the Progrm Usge Dt (lrgely bsed on HUD dministrtive records for these progrm types) suggests some mesurement error in one or both of these dt sources. 82 Although this outcome is not techniclly the sme s the number of moves plus one (becuse it is possible for fmily to move out of plce (for exmple, housing unit shred with friends or reltives) nd then move bck into the sme unit during the 6-month period), its interprettion is essentilly the sme. 83 Although the difference in homelessness outcomes between those who did nd did not use SUB is consistent with strong, negtive impct on homelessness, these differences cnnot be cuslly ttributed to the use of SUB. Becuse the use of SUB ws not rndomly ssigned, comprisons re fundmentlly subject to selection bis (tht is, people who did not tke up SUB might hve been more likely to become homeless in the bsence of SUB). Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 69

102 Chpter 6. Impcts of Permnent Housing Subsidy (SUB) Compred With Usul Cre (UC) Exhibit 6-6. SUB Versus UC: Percent of Fmilies With t Lest 1 Night Sty in Emergency Shelter During Month, by Month After RA Outcome Fmilies Assigned to SUB Who Did Not Use SUB Fmilies Assigned to SUB Who Used SUB N = 84 N = 446 Homeless or doubled up during the followup period At lest 1 night homeless or doubled up in pst 6 months or in shelter in pst 12 months) (%) At lest 1 night homeless or doubled up in pst 6 months (%) At lest 1 night homeless in pst 6 months (%) At lest 1 night doubled up in pst 6 months (%) Any sty in emergency shelter in months 7 to 18 fter RA (%) Number of dys homeless or doubled up in pst 6 months Number of dys homeless in pst 6 months Number of dys doubled up in pst 6 months Housing independence Living in own house or prtment t followup (%) Living in own house or prtment with no housing ssistnce (%) Living in own house or prtment with housing ssistnce (%) Number of plces lived Number of plces lived in pst 6 months Housing qulity Persons per room Housing qulity is poor or fir (%) Length of sty in shelter Length (in months) of bseline sty in emergency shelter SUB = permnent housing subsidy. RA = rndom ssignment. Difference in mens is sttisticlly significnt t the.10 level. The length of bseline sty in emergency shelter outcome includes one-hlf of the month of rndom ssignment nd is topcoded t 18 months. The 21 percent of fmilies ssigned to SUB whose bseline shelter sty does not pper in the Progrm Usge Dt re not included in the nlysis. Notes: Mens re unweighted. See Chpter 5 nd Appendix B for outcome definitions. Sources: Fmily Options Study bseline survey; 18-month followup survey; Progrm Usge Dt Exhibit 6-7. SUB Versus UC: Impct on Length of Bseline Sty in Emergency Shelter Outcome N SUB UC ITT Impct Men (SD) N Men (SD) Impct (SE) Length (in months) of bseline sty in emergency shelter b ** 0.12 (3.1) (4.6) (0.2) SUB = permnent housing subsidy. UC = usul cre. ITT = intention-to-tret. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire UC group. b The length of bseline sty in emergency shelter outcome includes one-hlf of the month of rndom ssignment nd is topcoded t 18 months. The 20 percent of fmilies in this comprison whose bseline shelter sty does not pper in the Progrm Usge Dt re not included in the nlysis. Notes: Impct estimte nd outcome mens re regression djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Appendix B for outcome definition. Sources: Fmily Options Study 18-month followup survey; Progrm Usge Dt Effect size 6.4 Impcts on Fmily Preservtion in the Permnent Housing Subsidy (SUB) Versus Usul Cre (UC) Comprison As discussed in Chpter 3, ny effects of SUB on fmily pre servtion would be expected to be indirect, vi the effects on housing stbility. To test this possibility, Exhibit 6-8 reports estimted impcts on fmily preservtion from the SUB-versus-UC comprison. Indeed, evidence exists tht SUB substntilly reduced child seprtions, by two-fifths (from 16.9 to 9.8 percent of fmilies) mong children with the fmily t bseline. Children seprted from fmilies in both groups were typiclly living with reltives, rther thn living in forml foster cre plcements. Evidence exists tht SUB reduced foster cre plcements by more thn hlf, from 5.0 to 1.9 percent of fmilies. No evidence suggests, however, tht SUB incresed reunifictions of the smll number of children who were seprted from the fmily t bseline, or tht SUB impcted either seprtions or reunifictions of the spouse or prtner of the fmily hed. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 70

103 Chpter 6. Impcts of Permnent Housing Subsidy (SUB) Compred With Usul Cre (UC) Exhibit 6-8. SUB Versus UC: Impcts on Fmily Preservtion Outcome Current or recent seprtions of fmily members present t bseline SUB UC ITT Impct Effect Size N Men (SD) N Men (SD) Impct (SE) Fmily hs t lest one child seprted in pst 6 months (%) (30.0) (36.0) 7.1*** (2.4) 0.17 Fmily hs t lest one foster cre plcement in pst 6 months (%) (13.7) (19.9) 3.1** (1.4) 0.13 Spouse/prtner seprted in pst 6 months, of those with spouse/prtner present t RA (%) (47.6) (48.3) 0.7 (6.5) Reunifiction of fmily members reported s seprted t bseline Fmily hs t lest one child reunified, of those fmilies with t lest one child (47.9) (46.0) 5.0 (7.3) 0.10 bsent t RA (%) Spouse/prtner reunified, of those with spouse/prtner bsent t RA (%) (46.1) (49.0) 0.3 (12.3) SUB = permnent housing subsidy. UC = usul cre. ITT = intention-to-tret. RA = rndom ssignment. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire UC group. Notes: Impct estimtes nd outcome mens re regression djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Source: Fmily Options Study 18-month followup survey 6.5 Impcts on Adult Well-Being in the Permnent Housing Subsidy (SUB) Versus Usul Cre (UC) Comprison As discussed in Chpter 3, the theory nd gols of the SUB intervention compred with those of UC would not led us to expect substntil effects on dult well-being. Even so, Exhibit 6-9 provides evidence of severl impcts in this domin. SUB ppers to hve improved dult mentl helth reltive to UC, incresing positive gol-oriented thinking (the Stte Hope Scle) nd reducing psychologicl distress. Both of these impcts re estimted t round 0.15 of stndrd devition of the outcome mesure nd contrst with finding of no significnt impct on dults receiving vouchers being worried, tense, or nxious in the Effects of Housing Vouchers on Welfre Fmilies evlution (Mills et l., 2006). Physicl helth nd post-trumtic stress disorder (PTSD) symptoms were not ffected by SUB. This study provides some evidence tht permnent deep housing subsidy led to less lcohol dependence thn UC nd bout 25 percent less lcohol nd drug buse overll. Exhibit 6-9. SUB Versus UC: Impcts on Adult Well-Being Adult physicl helth Outcome SUB UC ITT Impct Effect Size N Men (SD) N Men (SD) Impct (SE) Helth in pst 30 dys ws poor or fir (%) (46.5) (46.4) 0.1 (2.9) Adult mentl helth Gol-oriented thinking b (1.06) (1.00) 0.18*** (0.07) 0.15 Psychologicl distress c (5.40) (5.65) 0.97*** (0.34) 0.15 Adult trum symptoms Post-trumtic stress disorder (PTSD) symptoms in pst 30 dys (%) (40.9) (42.8) 3.6 (2.6) 0.07 Adult substnce use Alcohol dependence or drug buse d (%) (32.3) (37.5) 4.5* (2.4) 0.11 Alcohol dependence d (%) (28.8) (34.2) 4.3* (2.2) 0.12 Drug buse d (%) (20.1) (25.9) 2.2 (1.5) 0.08 Experience of intimte prtner violence Experienced intimte prtner violence in pst 6 months (%) (23.5) (32.6) 6.7*** (2.0) 0.19 SUB = permnent housing subsidy. UC = usul cre. ITT = intention-to-tret. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire UC group. b Gol-oriented thinking is mesured with modified version of the Stte Hope Scle nd rnges from 1 to 6, with higher scores indicting higher levels of positive, gol-oriented thinking. c Psychologicl distress is mesured with the Kessler 6 (K6) scle nd rnges from 0 to 24, with higher scores indicting greter distress. d Alcohol dependence is mesured with the Rpid Alcohol Problems Screen (RAPS4), nd drug buse is mesured with six items from the Drug Abuse Screening Test (DAST-10). Notes: Impct estimtes nd outcome mens re regression djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Source: Fmily Options Study 18-month followup survey Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 71

104 Chpter 6. Impcts of Permnent Housing Subsidy (SUB) Compred With Usul Cre (UC) SUB lso produced mjor reduction in the shre of dults who experienced intimte prtner violence in the 6 months prior to the followup survey, cutting the incidence in hlf, from 12 to 6 percent. This effect is consistent with qulittive evidence from the Effects of Housing Vouchers on Welfre Fmilies evlution, where some femle fmily heds reported tht vouchers were helpful in escping difficult or busive domestic situtions. 6.6 Impcts on Child Well-Being in the Permnent Housing Subsidy (SUB) Versus Usul Cre (UC) Comprison As discussed in Chpter 3, ny effects of SUB on child wellbeing would be expected to be indirect, vi the substntil effect of SUB on housing stbility. Little evidence suggests tht such indirect effects hd emerged t this followup point. Across the 26 mesures depicted in Exhibits 6-10 nd 6-11, only 3 effects reched sttisticl significnce, or bout wht would be expected by chnce. The effect tht is lest likely to be due to chnce lone is on school mobility, both becuse of the higher level of sttisticl significnce nd becuse school mobility is ssocited with residentil mobility. 84 One fewer school move ws mde in 18 months for every five children in the SUB group compred with children in the UC group. The other smll effects (fewer bsences but more nxiety in the SUB group) re probbly best ttributed to chnce. The number of schools ttended indexes exposure to stble schooling rther thn n cdemic or behviorl outcome. Stble schooling might be expected to led to better long-term eductionl outcomes, lthough this shortterm followup period uncovered no evidence for tht. Exhibit SUB Versus UC: Impcts on Child Well-Being Across Age Groups Child eduction Outcome SUB UC ITT Impct Effect size N Men (SD) N Men (SD) Impct (SE) Preschool or Hed Strt enrollment b (%) (48.7) (48.3) 5.3 (4.7) 0.09 School enrollment c (%) (31.5) (29.8) 1.5 (2.3) 0.05 Childcre or school bsences in pst month d (0.94) (1.01) 0.15* (0.08) 0.12 Number of schools ttended since RA e (0.74) (0.89) 0.21*** (0.06) 0.18 Grde completion (not held bck) (%) (26.1) (29.1) 2.2 (2.1) 0.06 Positive childcre or school experiences f (0.53) (0.56) 0.03 (0.04) 0.04 Positive childcre or school ttitudes g (1.02) (1.02) 0.03 (0.07) 0.02 School grdes h (0.90) (0.96) 0.13 (0.08) 0.10 Childcre or school conduct problems i (0.41) (0.41) (0.03) Child physicl helth Poor or fir helth (%) (21.7) (20.7) 0.5 (1.4) 0.02 Well-child checkup in pst yer (%) (31.0) (28.8) 0.6 (2.4) 0.02 Child hs regulr source of helth cre (%) (24.8) (24.1) 1.1 (2.2) 0.03 Sleep problems j (1.13) (1.13) 0.06 (0.07) 0.04 Child behviorl strengths nd chllenges Behvior problems k (1.20) (1.26) 0.12 (0.09) 0.08 Prosocil behvior l (1.11) (1.12) 0.11 (0.08) 0.08 SUB = permnent housing subsidy. UC = usul cre. ITT = intention-to-tret. RA = rndom ssignment. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire UC group. b Bse for preschool or Hed Strt enrollment is children ges 1 yer, 6 months to 5 yers. c Bse for school enrollment is children ges 6 to 17 yers. d Absences outcome is defined s 0 = no bsences in pst month, 1 = 1 to 2 bsences, 2 = 3 to 5 bsences, 3 = 6 or more bsences. e Number of schools outcome is topcoded t 4 or more schools. f Positive childcre or school experiences outcome is defined s 1 = mostly negtive experiences, 0 = both positive nd negtive experiences, 1 = mostly positive experiences. g Positive childcre or school ttitudes outcome is prent report of how much child likes school nd rnges from 1 to 5, with higher vlues indicting greter like of school. h School grdes outcome is defined s 1 = mostly Ds or Fs, 2 = mostly Cs, 3 = mostly Bs, 4 = mostly As. i Childcre or school conduct problems outcome is defined s 0 = no conduct problems reported to prent, 1 = prent contcted bout conduct problems or suspension or expulsion from school or childcre center. j Sleep problems outcome rnges from 1 to 5, with higher vlues indicting more frequent tiredness upon wking nd during the dy. k Behvior problems outcome is mesured s the stndrdized Totl Difficulties score from the Strengths nd Difficulties Questionnire (SDQ). l Prosocil behvior is mesured s the stndrdized Pro-socil domin score from the SDQ. Notes: Impct estimtes nd outcome mens re regression djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Source: Fmily Options Study 18-month followup survey (prent report) 84 School mobility is modestly correlted with residentil mobility (r in the full smple = 0.15, p <.001), lthough they re mesured over different time frmes (6 months for residentil mobility; since rndom ssignment for school mobility). Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 72

105 Chpter 6. Impcts of Permnent Housing Subsidy (SUB) Compred With Usul Cre (UC) Exhibit SUB Versus UC: Impcts on Child Well-Being Developmentl Outcomes by Age Group Outcome Ages 1 yer to 3 yers, 6 months SUB UC ITT Impct Effect size N Men (SD) N Men (SD) Impct (SE) Met developmentl milestones b (%) (44.1) (43.6) 1.4 (5.5) 0.03 Low birth weight c (%) (36.2) (20.9) 8.1 (8.6) 0.22 Ages 3 yers, 6 months to 7 yers Verbl bility d (0.95) (0.99) 0.05 (0.10) 0.04 Mth bility e (0.95) (0.85) 0.07 (0.10) 0.05 Executive functioning (self-regultion) f (16.16) (16.04) 0.46 (1.30) 0.02 Ages 8 to 17 yers Anxiety g (7.66) (7.77) 1.51* (0.86) 0.14 Fers h (14.07) (14.83) 0.33 (1.61) 0.02 Substnce use i (%) (24.85) (27.14) 1.34 (2.55) 0.03 Gol-oriented thinking j (4.80) (5.27) (0.50) School effort in pst month k (0.80) (0.77) 0.10 (0.08) 0.09 Arrests or police involvement in pst 6 months l (%) (31.12) (28.88) 0.55 (4.78) SUB = permnent housing subsidy. UC = usul cre. ITT = intention-to-tret. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire UC group. b Met developmentl milestones outcome is defined s scoring bove the typicl development cutoffs in ll domins of the Ages nd Stges Questionnire (ASQ-3). c Bse for low birthweight outcome (prent report) is children born since rndom ssignment who re t lest 1 yer old t followup. d Verbl bility outcome is the ntionlly stndrdized score from the Woodcock-Johnson III (WJ III) letter-word identifiction scle. e Mth bility outcome is the ntionlly stndrdized score from the WJ III pplied problems test. f Executive functioning outcome is the Hed Toes Knees Shoulders (HTKS) score nd rnges from 0 to 40, with higher scores indicting greter executive functioning. g Anxiety (child report) is mesured using the A-Trit scle from the Stte-Trit Anxiety Inventory for Children (STAIC). Scores rnge from 20 to 60, with higher scores indicting greter nxiety. h Fers outcome (child report) is the score from the Fers Scle nd rnges from 33 to 99, with higher scores indicting more fer. i Substnce use (child report) is mesured with 23 items from the Centers for Disese Control nd Prevention (CDC) 2011 Youth Risk Behvior Survey. j Gol-oriented thinking (child report) is mesured with modified version of the Children s Hope Scle nd rnges from 6 to 30, with higher scores indicting higher levels of positive, gol-oriented thinking. k School effort outcome (child report) rnges from 1 to 4, with higher scores indicting greter effort during school dy nd on homework. l Arrest or police involvement in pst 6 months is from prent report. Notes: Impct estimtes nd outcome mens re regression djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Sources: Fmily Options Study 18-month followup survey (prent report); Fmily Options Study 18-month child survey (child report); ASQ-3; WJ III; HTKS Child well-being ws ssessed only for children who were with their prent t the time of the followup survey. It is possible tht the substntilly higher rtes of child seprtion in the UC group could led to n underestimte of effects of SUB on child well-being Impcts on Self-Sufficiency in the Permnent Housing Subsidy (SUB) Versus Usul Cre (UC) Comprison By incresing housing stbility, SUB might plusibly hve indirect effects on fmily self-sufficiency reltive to UC. In prticulr, the opportunity to obtin stble housing with shrply nd permnently lower burden for housing costs might enble dult fmily members to trnsfer ttention from stying housed to concentrting more on employment nd ernings nd even enhncing their skills through educ - tion nd trining prticiption. On the other hnd, the bility to obtin housing with limited out-of-pocket cost (30 percent of income) mkes vilble household finncil resources go further lessening the pressure to work, ern, nd cquire new skills nd eduction. Unlike the other ctive interventions studied in this report, SUB does not include cse mngement guidnce nd referrls to nonhousing services to enhnce efforts t work nd self-sufficiency. Exhibit 6-12 shows effects on self-sufficiency outcomes for the SUB-versus-UC comprison. Of the 20 outcomes, 10 hd sttisticlly significnt effects. The first 5 of these significnt effects show tht the lbor mrket enggement of SUB fmily heds ws lower thn tht of their counterprts 85 The study tem sked prents to report on children who were no longer with them, but very few were ble to do so, nd the results exclude their dt. It is plusible to ssume tht children who re no longer with their prents hve lower well-being, on verge, thn those who remin with their prents. Concerns bout child well-being could hve led to seprtions, prticulrly foster cre plcements, or seprtion from prents might cuse psychologicl or behviorl problems for children. If ll children living elsewhere could hve been included in the results, child outcomes in the UC group (which hs more children wy from prent) would likely be depressed more thn child outcomes in the SUB group. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 73

106 Chpter 6. Impcts of Permnent Housing Subsidy (SUB) Compred With Usul Cre (UC) Exhibit SUB Versus UC: Impcts on Self-Sufficiency Employment sttus Outcome SUB UC ITT Impct Effect size N Men (SD) N Men (SD) Impct (SE) Work for py in week before survey (%) (43.1) (45.8) 5.7** (2.9) 0.11 Any work for py since RA (%) (50.0) (48.7) 10.9*** (2.9) 0.20 Months worked for py since RA b (6.8) (7.4) 1.6*** (0.4) 0.19 Hours of work per week t current min job c (13.9) (14.7) 1.5 (0.9) 0.09 Income sources nd mounts Annulized current ernings ($) 524 3,568 (7,972) 408 4,299 (8,197) 731 (517) 0.07 Totl fmily income ($) 515 8,520 (6,990) 405 8,980 (7,355) 460 (488) 0.05 Anyone in fmily hd ernings in pst month (%) (47.6) (49.4) 8.8*** (3.1) 0.16 Anyone in fmily received TANF in pst month (%) (49.0) (47.8) 6.3** (3.0) 0.12 Anyone in fmily received SSDI in pst month (%) (24.2) (22.4) 0.0 (1.6) Anyone in fmily received SSI in pst month (%) (32.9) (31.7) 1.2 (1.6) 0.03 Anyone in fmily received SNAP/food stmps in pst month (%) (30.8) (36.4) 6.6*** (2.4) 0.15 Anyone in fmily received WIC in pst month (%) (47.4) (46.3) 2.8 (2.9) 0.05 Eduction nd trining Prticipted in 2 weeks or more of ny school or trining since RA (%) (44.1) (43.7) 1.1 (3.0) 0.02 Number of weeks in school or trining progrms since RA (10.8) (9.8) 0.3 (0.7) 0.03 Prticipted in 2 weeks or more of school since RA (%) (25.9) (25.5) 1.8 (2.0) 0.06 Prticipted in 2 weeks or more of bsic eduction since RA (%) (8.7) (14.6) 1.2* (0.7) 0.09 Prticipted in 2 weeks or more of voctionl eduction since RA (%) (25.9) (26.3) 0.7 (1.7) 0.02 Food security nd hunger Household is food insecure (%) (44.3) (48.0) 9.9*** (3.1) 0.18 Food insecurity scle d (1.97) (2.01) 0.40*** (0.13) 0.17 Economic stressors Economic stress scle e (0.47) (0.52) 0.16*** (0.03) 0.28 SUB = permnent housing subsidy. UC = usul cre. ITT = intention-to-tret. RA = rndom ssignment. SD = stndrd devition. SE = stndrd error. SNAP = Supplementl Nutrition Assistnce Progrm. SSDI = Socil Security Disbility Insurnce. SSI = Supplementl Security Income. TANF = Temporry Assistnce for Needy Fmilies. WIC = Specil Supplementl Nutrition Progrm for Women, Infnts, nd Children. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire UC group. b Number of months worked for py includes prtil clendr months. c Hours of work per week includes those not currently working (tht is, those with 0 hours of work per week). d Food insecurity scle rnges from 0 to 6, with higher vlues indicting higher food insecurity. e Economic stress scle rnges from 1 to 1, with higher vlues indicting higher economic stress. Notes: Impct estimtes nd outcome mens re regression djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Source: Fmily Options Study 18-month followup survey in UC. SUB cused declines in the shre who worked for py (t the time of the followup survey nd t ny time since rndom ssignment), months nd hours worked, nd shre of fmilies with n erner in the month prior to the followup survey. 86 The fifth row of the exhibit shows no corresponding effect on the nnulized current ernings of fmily heds, s would be suggested by the lower employment rte of SUB fmily heds. The lck of n effect ppers to be relted to the higher vribility of ernings, however, rther thn to SUB working fmily heds receiving higher verge wge thn UC working fmily heds The effect on verge hours per week, however, is lrgely driven by the difference in percentge of fmily heds who re working for py t the time of the survey. Dividing the verge hours of work per week t the current min job for SUB fmily heds by the percentge who worked for py in the week before the survey (7.10 hours per week/0.240 = 29.6 hours per week) revels tht those who were working did so n verge of 30 hours per week. Performing the sme clcultion for UC fmily heds (8.61 hours per week/0.297 = 29.0 hours per week) revels tht UC fmily heds who were working did so n verge of 29 hours per week, very similr to the SUB fmily heds. 87 If SUB working fmily heds hd the sme nnulized current ernings s UC working fmily heds, the difference in verge nnulized current ernings implied by the difference in percentge of fmily heds who re working would not be sttisticlly significnt. Therefore, the lck of sttisticl significnce on the impct estimte for nnulized current ernings should not be interpreted s evidence of higher verge ernings for SUB working fmily heds. The verge nnulized current ernings of UC working fmily heds is clculted s $4,302 per yer/0.297 = $14,485 per yer. Multiplying this figure by the 24 percent of SUB fmily heds who re working yields verge nnulized ernings of $3,476 for ll SUB fmilies, implying n verge effect of $3,476 $4,302 = -$826. Given our stndrd error of $517, the effect of -$826 would not be sttisticlly significnt. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 74

107 Chpter 6. Impcts of Permnent Housing Subsidy (SUB) Compred With Usul Cre (UC) These dmpening effects on lbor mrket enggement nd returns re not unexpected nd re consistent with economic theory. With more housing benefits, the need to work is less (the income effect), nd with higher tx rte (the 30 percent of income on housing), the net wge from ech dditionl hour of work is considerbly smller, mking other ctivities such s prenting, leisure, nd possibly eduction nd trining ctivities reltively more ttrctive (the substitution effect). The findings on employment nd hours re consistent with those reported by the study of the Effects of Housing Vouchers on Welfre Fmilies (Mills et l., 2006) nd by Jcob nd Ludwig (2012) in their Chicgo study. Employment reduc - tions ended fter few qurters in the former study but persisted longer in the ltter reserch. This difference mkes findings for the Fmily Options Study from future 36-month followup survey quite importnt for the SUB-versus-UC comprison; will the reduction in work effort cused by voucher receipt continue (s theory leds us to expect) or fde wy? We found no evidence tht SUB s reduction in work effort led to reduction in totl fmily income, lthough it is possible tht such n effect ws smller thn could be detected with this smple size. It is importnt to note tht the fmily income mesure does not include the vlue of the housing subsidy provided by the voucher or Supplementl Nutrition Assistnce Progrm (SNAP) benefits. In line with previous evidence from the Effects of Housing Vouchers on Welfre Fmilies evlution is the finding tht SUB incresed receipt of public ssistnce through Temporry Assistnce for Needy Fmilies (TANF) nd SNAP. SUB incresed fmily prticiption in ech of these progrms t the time of the followup survey, rising TANF receipt from 34 to 40 percent nd SNAP receipt from 84 to 90 percent. No evidence emerged of effects on the receipt rtes of Specil Supplementl Nutrition Progrm for Women, Infnts, nd Children (WIC), Socil Security Disbility Insurnce (SSDI), nd Supplementl Security Income (SSI) benefits. We lso found no evidence tht SUB incresed prticiption in eduction nd trining ctivities. 88 Some evidence emerged of improvements in broder mesures of well-being tht re not tied specificlly to employment. Compred with UC, SUB improved two indexes of food in - security. The vilbility of SUB lowered the shre of house - holds clssified s food insecure from 36 to 26 percent nd reduced scores on food insecurity scle by 0.17 stndrdized effect size. These results re not surprising. Totl household resources the sum of unchnged csh income nd noncsh housing ssistnce, which rose substntilly rose shrply. Some of the dditionl resources pper to hve been spent on food, decresing food insecurity. Economic stress (mesured through survey items tht sked bout frequency of not being ble to fford rent, medicl cre, clothing, nd so on) declined even more shrply becuse of SUB, flling from 0.16 points on scle from -1 to 1. Wheres ll other significnt impcts hve stndrdized effect sizes in the 0.10-to-0.20 rnge, this decline in economic stress hs stndrdized effect size of Summry of the Permnent Hous - ing Subsidy (SUB) Versus Usul Cre (UC) Comprison Across Domins For the SUB-versus-UC comprison, the study produced notble contrst in progrm use. Of ll fmilies ssigned to SUB, 84 percent received the SUB intervention compred with 12 percent of fmilies ssigned to UC. This contrst in progrm use led to striking differences between the experiences of SUB nd UC fmilies in severl res. The most notble effect of SUB reltive to UC ws its greter prevention of subsequent stys in shelter or plces not ment for humn hbittion. Of fmilies ssigned to SUB, 22 percent spent t lest 1 night homeless or doubled up in the 6 months prior to the followup survey or t lest 1 night in emergency shelter in the 12 months prior to the survey compred with 50 percent of fmilies ssigned to UC. Tht is, ssignment to SUB fter 7 dys in emergency shelter reduced subsequent homelessness by hlf. The SUB intervention lso cused substntil, sttisticlly significnt reductions in four other mesures of homelessness during the followup period: the proportion of fmilies who experienced (1) t lest 1 night in shelter or plce not ment for humn hbittion in the pst 6 months; (2) t lest 1 night doubled up in the pst 6 months; (3) t lest 1 night in shelter or plces not ment for humn hbittion or doubled up in the pst 6 months; nd (4) t lest one sty in emergency shelter in months 7 to 18 fter rndom ssignment. Compred with UC, SUB lso led to greter housing independence, with 73 percent of SUB fmilies living in their own house or prtment compred with 58 percent of UC fmilies. As expected, however, the proportion of SUB fmilies living in their own dwelling plce without housing ssistnce (10 percent) is much lower thn 88 In fct, the only significnt effect on five eduction nd trining outcomes showed tht SUB cused slight reduction in the proportion of fmily heds prticipting in bsic eduction (1 percent of SUB fmily heds compred with 2 percent of UC fmily heds). Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 75

108 Chpter 6. Impcts of Permnent Housing Subsidy (SUB) Compred With Usul Cre (UC) the corresponding proportion of UC fmilies (33 percent). SUB lso reduced the verge number of plces lived since rndom ssignment nd the proportion of fmilies living in poor or fir qulity housing. Indirect benefits occurred in terms of selected fmily preservtion indictors nd dult well-being mesures. Reltive to UC, SUB led to improvements in fmily preservtion. For fmilies with child present t bseline, SUB reduced subsequent child seprtions by two-fifths (10 percent in SUB fmilies compred with 17 percent in UC fmilies). SUB lso reduced foster cre plcements by more thn hlf (from 5 to 2 percent) reltive to UC. SUB lso led to improvements in four mesures of dult well-being. SUB incresed positive gol-oriented thinking, reduced psychologicl distress, nd reduced evidence of lcohol nd drug problems. SUB lso produced substntil reduction in intimte prtner violence reltive to UC. Incidence of intimte prtner violence ws 12 percent in UC fmilies compred with 6 percent in SUB fmilies. Among children who remined with their fmilies, little evidence exists tht SUB improved child well-being reltive to UC, prt from reduction in the number of schools ttended by focl children. No effects were detected on the physicl helth, behvior, or development of focl children. Importntly, SUB reduced fmily self-sufficiency reltive to UC. Lbor mrket enggement declined while dependence on public ssistnce (TANF nd SNAP) rose. SUB cused reduction in the proportion of fmily heds working t the time of the survey compred with UC (from 30 to 24 percent), reduction in the proportion with ny employment since rndom ssignment (from 61 to 50 percent), nd reduction in the verge number of months worked since rndom ssignment. These results re consistent with eco - nomic theory, given tht housing subsidies lessened the need for disposble income nd reduced the returns to working t the mrgin. Reltive to UC, SUB incresed receipt of public ssistnce through TANF nd SNAP (from 34 to 40 percent for TANF nd from 84 to 90 percent for SNAP). Even with the reduction in work effort, SUB fmilies ppered be in better finncil position thn UC fmilies. The dditionl resources represented by the housing subsidy led to n improvement in the food security of SUB fmilies (lowering the percentge of households clssified s food insecure from 36 to 26 percent, reltive to UC) nd decrese in economic stress. Overll, in severl wys fmilies ssigned to SUB pper to be doing better t this point thn fmilies ssigned to UC. Fmilies ssigned to SUB hve greter housing stbility, re less likely to hve hd child seprtions, nd experienced less psychologicl distress, less intimte prtner violence, less food insecurity, nd less economic stress thn their counterprts ssigned to UC. Chpter 9 reports on how SUB compres with the two other ctive interventions, CBRR nd PBTH. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 76

109 CHAPTER 7. IMPACTS OF COMMUNITY-BASED RAPID RE-HOUSING (CBRR) COMPARED WITH USUAL CARE (UC) This chpter presents the impcts observed in the community-bsed rpid re-housing (CBRR) intervention compred with outcomes of fmilies served by the usul cre (UC) homeless ssistnce systems in their communities. The gol is to determine whether offering temporry subsidies to help fmilies exit shelter rpidly increses their housing stbility nd improves other fmily outcomes during 20-month followup intervl. The chpter begins with description of the CBRR intervention s implemented in the study. It then shows the extent to which fmilies in both the CBRR nd UC groups used rpid re-housing nd other housing nd service progrms vilble to them in the study sites. The reminder of the chpter presents the effects of being offered the CBRR intervention (s compred with UC) on outcomes within the five study domins housing stbility, fmily preservtion, dult well-being, child well-being, nd self-sufficiency. 7.1 Community-Bsed Rpid Re-Housing (CBRR) Intervention The CBRR intervention provides progrm prticipnts with temporry rentl ssistnce nd limited services focused on housing serch ssistnce nd bsic service coordintion. In totl, 569 fmilies were referred to 27 CBRR progrms cross the 12 sites. 89 The number of fmilies ssigned to CBRR in ech site, rnging from 8 fmilies in Denver, Colordo, to 80 fmilies in Slt Lke City, Uth, is shown in Exhibit 2-8 in Chpter 2. Of the 569 fmilies, 455 (80 percent) responded to the 18-month followup survey nd so re included in the impct nlysis in this report. Nerly ll the CBRR providers in the Fmily Options Study were community-bsed nonprofit orgniztions. The only exceptions were in Louisville, Kentucky, nd Phoenix, Arizon, where CBRR ws provided by city government gencies. CBRR ws funded by the rpid re-housing component of the Home - lessness Prevention nd Rpid Re-Housing Progrm (HPRP) in ll the sites except one. 90, 91 The HPRP rpid re-housing funding could be used to provide rentl ssistnce (up to 18 months), security deposits, utility deposits nd pyments, help with moving costs, nd hotel nd motel vouchers. HPRP lso could fund cse mngement for prticipting fmilies. Any rentl ssistnce pid for with HPRP funds hd to meet rent resonbleness stndrds, nd units hd to pss hbitbility inspection. The inspection requirements were slightly less stringent thn the Housing Qulity Stndrds required for the Housing Choice Voucher (HCV) progrm form of permnent housing subsidy (SUB) Housing Assistnce in CBRR The CBRR intervention provided short-term rentl ssistnce (usully 7 to 8 months) to enble fmilies to rent privtemrket housing. The intention ws tht the prticipnts would remin in the unit tht they obtined with CBRR ssistnce fter the period of rentl ssistnce ended, pying the full rent on their own. 89 Much of the informtion describing CBRR in this section is bsed on the 16 CBRR progrms tht provided progrm dt. These 16 progrms represent 521 of the 569 totl CBRR referrls. More detil bout specific CBRR progrms is provided in Gubits et l. (2013), Appendixes A nd B. Originlly, 28 progrms were intended to prticipte in the study. The study referred fmilies to 27 of these progrms. The CBRR followup respondents represent 25 of the 27 progrms. 90 HPRP ws uthorized through the Americn Recovery nd Reinvestment Act of 2009 (ARRA). Across the ntion, communities received $1.5 billion in HPRP funding, one-time funding strem vilble for 3 yers from progrm inception, to provide homelessness prevention nd rpid re-housing ssistnce to individuls nd fmilies fcing homelessness. 91 In Boston, the CBRR intervention ws funded by the Stte of Msschusetts. The Boston progrms offered ssistnce very similr to HPRP, lthough rentl ssistnce could be provided for longer periods. The Minnepolis, Minnesot, nd Slt Lke City CBRR progrms supplemented HPRP funds with stte funds nd other ARRA funds, respectively. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 77

110 Chpter 7. Impcts of Community-Bsed Rpid Re-Housing (CBRR) Compred With Usul Cre (UC) HPRP regultions llowed for CBRR progrms to provide up to 18 months of ssistnce, but HUD encourged progrms to provide the lest mount of rentl ssistnce needed to stbilize fmilies in housing. Accordingly, HPRP regultions required CBRR progrms to recertify fmilies for income eligibility nd need every 3 months (HUD, 2009). Nerly 40 percent of fmilies ssigned to CBRR were referred to progrms tht lso provided ssistnce with pyment of rrers (either rentl rrers or bck pyments owed on utilities), nd most (84 percent) were referred to progrms tht provided ssistnce with strtup costs such s security deposits, utility setup costs, nd moving expenses. About two-thirds of fmilies ssigned to CBRR were referred to progrms tht typiclly provided 4 to 6 months of ssistnce. The other one-third of CBRR fmilies were referred to progrms tht typiclly provided more thn 6 months of ssistnce. Depth of CBRR Housing Assistnce The subsidy provided through CBRR represents substntil frction of monthly rent; however, the subsidy in CBRR ws rrely determined bsed on prticipnt contribution of fixed percentge of income, s is the cse in the HCV progrm nd other housing ssistnce progrms such s public housing. More thn one-hlf of fmilies were referred to CBRR progrms tht set the subsidy s fixed monthly mount, regrdless of monthly rent or fmily income. The fixed monthly mount ws typiclly determined by CBRR cse mngers bsed on dt collected through the client ssessment, considering stndrd set of criteri such s fmily income, debt, size, nd locl housing costs. For nother 19 percent of fmilies, the initil subsidy ws set t the discretion of the progrm, bsed on wht the progrm stff determined ws needed to get the fmily into housing, nd then the ongoing subsidy ws djusted bsed on formul. Rentl ssistnce for the remining 26 percent of fmilies ws clculted by formul, most often s percentge of the rent estblished by the lndlord. The stndrd housing ssistnce formul, rent minus percentge of income, ws used by progrms to which only 3 percent of CBRR fmilies were ssigned. In one of the sites, the CBRR progrm pid the full monthly rent. Some progrms estblished cps on the totl mount of subsidy pid to given fmily. Exhibit 7-1 summrizes the methods used by CBRR providers to clculte subsidies for fmilies in the study Assessment of Fmily Needs in CBRR All CBRR progrms indicted tht they conducted forml ssessment of study fmilies t the beginning of the progrm. In ddition, ll progrms reported tht they ressessed fmily needs s prt of the 3-month progrm recertifiction. The ssessments typiclly exmined fmily needs relted to hous - ing, self-sufficiency, nd employment, but three-fourths of fmilies were referred to CBRR progrms tht lso reported exploring helth, mentl helth, nd substnce buse issues in the ssessments. Slightly more thn one-hlf of fmilies were referred to CBRR progrms tht ssessed for children s needs. Other domins, such s prenting nd life skills, were formlly incorported into only hndful of progrms ssessments. In ll progrms, the ssessments resulted in forml service pln, with gols for the dults in the household designed to help fmilies obtin nd remin in stble housing nd to guide subsequent cse mngement nd referrls to other service progrms. Thirty percent of fmilies were referred to CBRR progrms tht lso worked with fmilies to develop gols for the children Supportive Services Provided in CBRR Prticipting CBRR progrms provided limited cse mngement with linkges to other progrms for dditionl support. CBRR services were hevily focused on housing nd selfsufficiency. Self-sufficiency services included help with budg - eting, obtining public benefits, eduction, trnsporttion, nd child cre. Most fmilies were referred to CBRR progrms in which the CBRR cse mnger took primry responsibility for providing housing serch nd plcement ssistnce. One progrm (representing 8 percent of fmily referrls) hd housing specilist to provide tht function insted. Exhibit 7-1. Methods Used To Clculte CBRR Subsidy Amounts CBRR Subsidy Clcultion Percent of Fmilies Assigned to CBRR Progrms With Ech Type of Subsidy Clcultion (N = 521) Subsidy mount is set by cse mngers on cse-by-cse bsis 55 Hybrid (initil subsidy set by cse mngers, then djusted bsed on formul relted to fmily contribution to rent) 19 Subsidy mount is percentge of rent estblished by the lndlord 23 Subsidy mount is difference between rent nd fmily contribution of some percent of income 3 CBRR = community-bsed rpid rehousing. Progrm dt were not collected from 11 progrms tht collectively hd 48 CBRR fmily referrls. Note: Percentges re unweighted. Sources: Progrm dt; rndom ssignment records Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 78

111 Chpter 7. Impcts of Community-Bsed Rpid Re-Housing (CBRR) Compred With Usul Cre (UC) Three-fourths of CBRR referrls were to progrms in which cse mngers provided self-sufficiency services, but smll proportion of fmilies were referred to progrms tht linked fmilies to other gencies for these services. No other types of services were provided to the sme extent. Slightly more thn one-third of fmilies were referred to pro - grms tht provided employment trining support. Other service res were explicitly offered t even fewer progrms. Exhibit 7-2 shows the rry of services offered in CBRR progrms nd the extent to which the services were provided through cse mngement, by other progrm or gency stff beyond the cse mnger, or through dedicted linkge with n externl gency tht ws gurnteed to provide the service becuse of CBRR enrollment. Dedicted linkges were rre. Cse Mngement Intensity in CBRR Exhibit 7-3 shows the reltive intensity of cse mngement for fmilies referred to CBRR both the rtio of clients to cse mngers nd the frequency of meetings. The verge cse mngement rtio for CBRR progrms in the study ws 36 fmilies per cse mnger. 92 Nerly one-hlf of clients were referred to progrms in which cse mnger worked with more thn 30 fmilies t time, nd most of the cse mngers with hevy cselods met with their clients monthly rther thn more frequently. About one-fifth of fmilies were referred to progrms with 11 to 20 clients per cse mnger, nd these cse mngers more often met with clients every 1 or 2 weeks or reported tht the frequency of meetings ws vrible. Some progrms indicted tht cse mngers met more frequently up front to support fmilies s they serched Exhibit 7-2. Types of Supportive Services Offered in CBRR Progrms nd How They Are Delivered Types of Supportive Services Percent of Fmilies Referred to CBRR Progrms Tht Offer These Services (N = 521) Percent of Fmilies Referred to CBRR Progrms Tht Offer Services of This Type Through Cse Mngement By Other Progrm or Agency Stff Through Dedicted Linkges With Other Agencies Housing serch nd plcement ssistnce Self-sufficiency (for exmple, finncil litercy, money mngement, help obtining public benefits, eduction, trnsporttion, childcre, nd fter-school cre) Employment nd trining Life skills Physicl helth cre Child dvoccy Prenting skills Mentl helth cre CBRR = community-bsed rpid rehousing. Progrm dt were not collected from 11 progrms tht collectively hd 48 CBRR fmily referrls. Note: Percentges re unweighted. Sources: Progrm dt; rndom ssignment records Exhibit 7-3. CBRR Cse Mngement Intensity (rtio nd frequency) Averge Number of Clients per Cse Mnger Percentge of Fmilies Referred to CBRR Progrms Tht Offer Cse Mngement in Ech of the Following Pckges (N = 464) Weekly Biweekly Monthly Qurterly Vrible 10 or fewer clients to 20 clients to 30 clients More thn 30 clients Vrible Totl CBRR = community-bsed rpid rehousing. Progrm dt on cse mngement rtios were not collected from 13 progrms tht collectively hd 105 CBRR fmily referrls. Note: Percentges re unweighted. Sources: Progrm dt; rndom ssignment records Totl 92 The verge cse mngement rtio is clculted s the weighted verge of the progrm s typicl point-in-time cselods (collected in interviews with progrm stff), wherein the weights re the number of fmilies referred to the progrms. Thus, the verge cse mngement rtio in CBRR ws derived by first multiplying ech progrm s cse mngement rtio by the number of people referred to tht progrm, then summing the products, nd then dividing the sum by the totl number of fmilies referred to CBRR. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 79

112 Chpter 7. Impcts of Community-Bsed Rpid Re-Housing (CBRR) Compred With Usul Cre (UC) for nd moved into housing, meeting less frequently fter the fmily ws housed. Visits of pproximtely 1 hour were firly typicl for CBRR progrms. 93 Cse mngement typiclly ended when the CBRR rentl ssistnce cesed Eligibility Criteri for CBRR CBRR progrms sometimes hd screening criteri tht excluded certin fmilies from prticipting. For exmple, those who were not working might be considered not redy for CBRR becuse they hd little chnce of being ble to fford their units when CBRR ssistnce ended. Just under 5 percent of study fmilies nswered bseline survey questions in wy tht screened them out from possible ssignment to CBRR, nd 10.4 percent of those ssigned to CBRR were determined ineligible by the progrms. Fmilies screened out t bseline re not included in the study smple, but those determined ineligible fter rndom ssignment re. Those fmilies deter - mined ineligible by progrms must be included in the impct estimtes to preserve the comprbility of the CBRR fmilies nd UC fmilies tht is, some of the UC fmilies might lso hve been screened out hd they been ssigned to CBRR. In order to continue to receive ssistnce for the period offered by the CBRR progrm to which they were referred, fmilies hd to hve incomes below certin thresholds, nd most CBRR progrms to which study fmilies were referred sked questions bout income every 3 months. Some CBRR progrms lso imposed dditionl progrm requirements with which fmilies hd to comply to mintin eligibility for ssistnce, such s working with cse mnger to chieve employment or increse erned income. 7.2 Progrm Use by Fmilies in the Community-Bsed Rpid Re-Housing (CBRR) Versus Usul Cre (UC) Comprison Ech impct comprison in the study my be thought of s distinct experiment or test, nd this chpter ddresses only the comprison between CBRR nd UC, without reference to the fmilies who were rndomized to the SUB or projectbsed trnsitionl housing (PBTH) interventions. In totl, 1,144 fmilies took prt in the test of CBRR versus UC. These fmilies ll hd the opportunity to be ssigned to either CBRR or UC t the point of rndom ssignment nd were ssigned to one of these two interventions 569 fmilies to CBRR nd 575 fmilies to UC. 94 Of these 1,144 fmilies, 79 percent (455 CBRR fmilies nd 451 UC fmilies) responded to the 18-month followup survey, nd thus 906 fmilies re included in the CBRR-versus-UC impct comprison reported in this chpter. This section describes the extent to which the 455 CBRR fmilies used the CBRR intervention nd other progrms tht were vilble in the community both within nd outside the homeless services system during the followup period. Prllel informtion is presented for the 451 UC fmilies. Exhibit 7-4 shows the use of seven types of homeless nd housing ssistnce progrms by these fmilies. The first column shows the percentge of fmilies ssigned to CBRR who ever used ech progrm type during the followup period. 95 The second row (shded in the exhibit) shows the tkeup of CBRR by the fmilies ssigned to tht intervention; 59.7 percent of fmilies referred to CBRR progrm received rpid re-housing ssistnce t some point during the followup period mening tht they followed up on the referrl, were deemed eligible by the progrm, found housing unit, nd received one of the types of temporry rentl ssistnce provided by CBRR. 69 The second column shows the percentge of fmilies ssigned to UC who ever used ech progrm type during the followup period. 97 The shded row of the second column shows tht 19.6 percent of the UC fmilies received rpid re-housing ssistnce during the followup period, despite not being given priority ccess to CBRR. Presumbly these fmilies lerned bout the vilbility of CBRR, perhps from friends or fmily members, 98 or they my hve lredy been clients of the community-bsed nonprofit orgniztions tht dmin - istered the locl CBRR progrms. 93 Additionl detils bout ech progrm s cse mngement re shown in Gubits et l. (2013), Appendix B In the entire study, 746 fmilies were rndomly ssigned to UC. Only 575 of these fmilies hd CBRR vilble to them when they were rndomized, however. Therefore, only these 575 UC fmilies re prt of the CBRR-versus-UC comprison smple. All 569 fmilies rndomly ssigned to CBRR during the course of the study hd UC vilble to them, so ll re prt of the CBRR-versus-UC comprison smple. 95 The followup period is from the clendr month of rndom ssignment through the clendr month of response to the 18-month followup survey. Therefore, the length of the followup period differs cross fmilies. This period lsts for medin of 21 clendr months for the full smple. 96 All percentges, mens, nd medins in the exhibit re weighted to djust for survey nonresponse nd hence s best possible represent the full experimentl smple of 1,144 fmilies. The findings on progrm use re thus in line with similrly weighted impct estimtes provided subsequently in the chpter. 97 The percentges in the first seven rows of these columns re not mutully exclusive becuse some fmilies use more thn one progrm type during the followup period. 98 Emergency shelter stff committed to not referring UC fmilies to ctive interventions to which they were not ssigned. This commitment my not hve been upheld in ll cses. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 80

113 Chpter 7. Impcts of Community-Bsed Rpid Re-Housing (CBRR) Compred With Usul Cre (UC) Exhibit 7-4. CBRR Versus UC: Progrm Use Since Rndom Assignment Type of Housing Assistnce Percent Ever Used From RA to 18-Month Followup Survey Number of Months Used From RA to 18-Month Followup Survey, if Ever Used Type of Housing Assistnce Percent Used in Month of Followup Survey Response CBRR UC CBRR UC CBRR UC Men Medin Men Medin Subsidy (SUB) b Rpid re-housing (CBRR) Trnsitionl housing Permnent supportive housing Public housing Project-bsed vouchers/section 8 projects Emergency shelter c No use of homeless or housing progrms d N UC = usul cre. RA = rndom ssignment. Percentge of fmilies who ever used type of ssistnce progrm during the period from the month of RA to the month of the 18-month followup survey response (medin period durtion: 21 months). Percentges do not dd to 100 becuse some fmilies used more thn one progrm type during the followup period. b Subsidy ssistnce is housing choice vouchers plus site-specific progrms offered to fmilies ssigned to permnent housing subsidy (SUB) group in Bridgeport, Connecticut nd Honolulu, Hwii. c All fmilies were in emergency shelter t rndom ssignment. Percentges less thn 100 percent for ever used emergency shelter re due to missing dt on shelter use. d No use of homeless or housing progrms (ever used) indictes no use of the six progrm types in this tble during ny of the followup period nd no use of emergency shelter fter the first 6 months fter RA. No use in the month of the followup survey response indictes no use of ny of these seven progrm types. Notes: Percentges re regression djusted, controlling for site nd rndomiztion rtio. Percentges, mens, nd medins re weighted for survey nonresponse to represent full comprison smple. Source: Fmily Options Study Progrm Usge Dt The first row of the exhibit, long with rows 3 through 6, show prticiption in other types of homeless nd housing ssistnce progrms. Twenty-eight percent of these UC fmilies (not shown in the exhibit) found their wy to SUB or other permnent housing progrms nd one-fourth found their wy to trnsitionl housing, despite the lck of preferen - til ccess to those progrms through the study. The use of progrms other thn rpid re-housing progrms is lwys higher for the UC group thn for the CBRR group, presumbly becuse the UC group ws not referred directly to the CBRR intervention nd so turned to other types of progrms. The seventh row shows the percentges of fmilies in the CBRR nd UC groups who used none of the six types of progrms during the followup period, nor used emergency shelter from the seventh month fter rndom ssignment onwrd. About 12 percent of CBRR fmilies nd 27 percent of UC fmilies fll into this group. The men nd medin number of months of use for ech progrm type re lso shown in the exhibit (third nd fourth columns for CBRR fmilies, fifth nd sixth columns for UC fmilies) for only those fmilies who ever used given progrm type. 99 The number of months of rpid re-housing ssistnce use (medin of 7 months) is higher for the fmilies who hd priority ccess to CBRR thn for the 19.1 percent of UC fmilies who received rpid re-housing ssistnce (medin of 5 months). Wheres the previous columns consider ll experience from between rndomiztion nd the point t which the survey ws dministered, the lst two columns consider the progrm use s of the month of the survey. Although the tem expects tht most outcomes in the report will be influenced by ssistnce received during the entire followup period, some outcomes will be prticulrly strongly influenced by ssistnce received t the time of followup survey response. The lst two columns of the exhibit show the percentges of CBRR nd UC fmilies who received ech type of progrm in the clendr month of the followup survey response. The first row of the seventh column shows tht the rpid re-housing ssistnce hd ended for most of the CBRR fmilies who ever received it nd for the UC fmilies who ever received it. The mjority of both CBRR (60 percent) nd UC fmilies (56 percent) were not prticipting in homeless or housing progrm t the time they responded to the followup survey. Thus, differences re not expected in the outcomes of CBRR nd UC fmilies in res tht reflect the fmilies current experience, but only in those tht reflect lsting influence of fmilies hving been offered temporry rentl ssistnce to help them leve homelessness. As Exhibit 7-4 mkes cler, the CBRR fmilies used rnge of progrms in ddition to the progrm to which they were referred by the study, which is consistent with the design of the study. Fmilies were not required to use the intervention 99 Hence, 0 vlues re not fctored into the mens, nor do they pull downwrd the medins of the vrious distributions. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 81

114 Chpter 7. Impcts of Community-Bsed Rpid Re-Housing (CBRR) Compred With Usul Cre (UC) to which they were ssigned nd lso were not forbidden from using other progrms tht were vilble to them in their community. The intent of the study ws to mximize use of the ssigned ctive intervention (in this cse, mximize use of the CBRR intervention by the CBRR fmilies) nd to crete the lrgest possible contrst between the progrm mixes of different ssignment groups (in this cse, CBRR versus UC). As shown in the exhibit, the use of CBRR ws quite different for the CBRR nd UC groups. The contrst in usge of CBRR 59.7 percent for CBRR fmilies nd 19.6 percent for UC fmilies is sizble, lthough smller thn the nlogous contrst between the SUB nd UC groups. As is conventionl in rndom ssignment nlyses, our gol is to estimte the intention-to-tret (ITT) impct tht is, the difference in impct by the progrm to which fmilies were ssigned, regrdless of whether they ctully used tht progrm (or some other progrm). This gol is consistent with the policy option of mking tretment vilble to fmily but without the bility to force fmily to use tht tretment. Becuse not ll fmilies rndomly ssigned to CBRR used CBRR, nd some fmilies ssigned to UC did use CBRR, the true ITT impct is likely smller thn it would hve been hd the gp in CBRR usge been wider (ssuming tht CBRR truly hs nonzero impct on fmilies who use it). In prticulr, the difference in the use of CBRR by the CBRR nd UC groups is nrrow enough, given the reltively smll smple size vil - ble for nlysis, tht the study my hve filed to detect s sttisticlly significnt one or more ITT impcts lrge enough to be of policy importnce. Additionl detil bout the use of the CBRR intervention by CBRR fmilies is shown in Exhibit 7-5. This exhibit shows tht nerly one-hlf (46 percent) of CBRR fmilies who used rpid re-housing did so for less thn 6 months, nd 81 percent did so for less thn 12 months. 100 These reltively short periods of use my be surprising, given tht the progrm rules permit use of CBRR for up to 18 months. They reflect the relity, however, of how the progrm ws being dministered in the study sites nd how fmilies were using it. The reminder of the chpter reports estimted impcts in the vrious outcome domins tht if sttisticlly significnt cn be cuslly ttributed to the offer of temporry housing subsidy to the fmilies rndomly ssigned to CBRR t the strt of the followup period s opposed to no such directed referrl or privileged ccess being provided to UC fmilies. Exhibit 7-5. Number of Months of CBRR Receipt During Followup Period by CBRR Fmilies Who Ever Used CBRR 12 Percent of CBRR fmilies who ever used RR Number of months of rpid re-housing receipt or more CBRR = community-bsed rpid re-housing. Percentges re weighted for survey nonresponse to represent ll fmilies in comprison smple. Note: N = 274. Source: Fmily Options Study Progrm Usge Dt 100 Exhibit 7-5 shows tht just over 4 percent of CBRR fmilies who used CBRR did so for 18 or more months. HPRP-funded rpid re-housing ws limited to 18 months of ssistnce. CBRR usge durtions longer thn 18 months re probbly best interpreted s using the mximum llowble ssistnce of 18 months. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 82

115 Chpter 7. Impcts of Community-Bsed Rpid Re-Housing (CBRR) Compred With Usul Cre (UC) 7.3 Impcts on Housing Stbility in the Community-Bsed Rpid Re-Housing (CBRR) Versus Usul Cre (UC) Comprison Proponents of CBRR shre with the proponents of SUB the view tht the crisis of housing ffordbility is the root cuse of homelessness mong fmilies, s ddressed in Chpter 3. Becuse SUB is unlikely to become widely vilble to fmilies t the time they re experiencing homelessness, proponents of CBRR rgue tht limited resources dedicted to homelessness could be stretched to crete the best outcomes for the most people by mking subsidies temporry. The CBRRversus-UC comprison offers evidence on whether temporry subsidies re n effective tool to improve housing stbility reltive to usul cre. Exhibit 7-6 shows the impcts of the offer of CBRR on home - lessness, housing independence, residentil moves, nd housing qulity. CBRR does not pper to reduce homelessness in this smple of fmilies. None of the eight impct estimtes for homelessness nd doubled up outcomes is sttisticlly different from 0. Assign - ment to CBRR hs no effect on the proportion of fmilies experiencing homelessness during the followup period in mesures bsed on the survey, on Progrm Usge Dt, or on combintion of the two. The estimtes lso provide no evidence of effects on the number of dys spent homeless or doubled up in the 6 months prior to the followup survey. It is possible tht the true ITT effects of CBRR my hve reduced subsequent stys in shelter or plces not ment for humn hbittion, but tht these impcts were too smll for this study to detect, given the smple size. The 95-percent confidence intervl (not shown in the exhibit) for the confirmtory impct on homelessness rnges from reduction in homelessness of 11.0 percentge points to n increse in homelessness of 3.0 percentge points, suggesting n importnt potentil rnge of beneficil impct. The housing independence outcomes in Exhibit 7-6 hve one sttisticlly significnt impct. Although ssignment to CBRR hs no effect reltive to UC on the proportion of fmilies living in their own house or prtment t the time of the followup survey, nd no effect on the proportion living on their own with housing ssistnce, CBRR increses the proportion of fmilies living on their own without housing ssistnce. Becuse no strong pttern of effects exists in fct, no other housing stbility impct ws sttisticlly significnt mong the severl tested to guide interprettion, this singulr result should be interpreted with cution. Exhibit 7-6. CBRR Versus UC: Impcts on Housing Stbility Outcome Homeless or doubled up during the followup period CBRR UC ITT Impct Effect size N Men (SD) N Men (SD) Impct (SE) At lest 1 night homeless b or doubled up in pst 6 months or (50.0) (50.0) 3.5 (3.6) 0.06 in shelter in pst 12 months (%) [Confirmtory] c At lest 1 night homeless b or doubled up in pst 6 months (%) (48.7) (49.2) 3.0 (3.5) 0.05 At lest 1 night homeless b in pst 6 months (%) (42.2) (42.9) 1.4 (3.0) 0.03 At lest 1 night doubled up in pst 6 months (%) (45.2) (46.0) 3.0 (3.2) 0.06 Any sty in emergency shelter in months 7 to 18 fter RA (%) (44.3) (45.7) 2.1 (3.1) 0.04 Number of dys homeless b or doubled up in pst 6 months (73.9) (74.0) 1.2 (5.4) Number of dys homeless b in pst 6 months (52.6) (47.7) 3.5 (3.7) 0.06 Number of dys doubled up in pst 6 months (61.7) (64.4) 3.8 (4.5) 0.05 Housing independence Living in own house or prtment t followup (%) (48.7) (49.6) 4.8 (3.3) 0.09 Living in own house or prtment with no housing ssistnce (%) (49.5) (47.3) 7.8** (3.3) 0.14 Living in own house or prtment with housing ssistnce (%) (38.8) (42.0) 3.2 (2.7) 0.07 Number of plces lived Number of plces lived in pst 6 months (1.0) (1.2) 0.1 (0.1) 0.07 Housing qulity Persons per room (1.2) (1.3) 0.1 (0.1) 0.08 Housing qulity is poor or fir (%) (46.3) (47.3) 2.7 (3.3) 0.05 CBRR = community-bsed rpid re-housing. UC = usul cre. ITT = intention-to-tret. RA = rndom ssignment. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire UC group. b The definition of homeless in this report includes stys in emergency shelters nd plces not ment for humn hbittion. It excludes trnsitionl housing. c After djustment for multiple comprisons, the impct on the confirmtory outcome is not sttisticlly significnt t the.10 level for the CBRR-versus-UC comprison. Notes: Impct estimtes nd outcome mens re regression djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Sources: Fmily Options Study 18-month followup survey; Progrm Usge Dt Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 83

116 Chpter 7. Impcts of Community-Bsed Rpid Re-Housing (CBRR) Compred With Usul Cre (UC) The lst three rows of Exhibit 7-6 show tht ssignment to CBRR hs no effect on the number of plces lived or on housing qulity reltive to UC. It my be tht priority ccess to CBRR helped fmilies with their housing stbility on verge but not enough to be sttisticlly detectble. The gin to the 59.7 percent of fmilies ssigned to CBRR who used rpid re-housing ssistnce my hve been wtered down by weker outcomes mong the 40.3 percent who did not use it. To check this hypotheticl possibility, Exhibit 7-7 compres the outcomes of fmilies ssigned to CBRR who never took up tht intervention with the corresponding outcomes for CBRR-ssigned fmilies who did use CBRR. For some of the outcomes in Exhibit 7-7 including ny sty in emergency shelter in months 7 to 18, living in own house or prtment t followup, nd length of continued use of shelter fter rndom ssignment results were indeed better for users of CBRR thn for nonusers. For other outcomes, however including t lest 1 night homeless in the pst 6 months nd t lest 1 night doubled up in the pst 6 months results were better for nonusers of CBRR or were similr between the two groups. These results collectively indicte tht the incidence of homelessness within the CBRR group ws not minly confined to those fmilies who never used CBRR. This finding mkes it less plusible tht the lck of significnt impcts for CBRR results from the weker contrst in the tkeup of CBRR between those rndomly ssigned to CBRR nd those ssigned to UC. 101 Additionl evidence on whether CBRR resulted in quicker exits from emergency shelter, one of the CBRR progrm s purposes, is presented in Exhibit 7-8. This nlysis returns to using experimentl contrsts rther thn nonexperimentl nlysis nd shows the month-by-month impcts of rndom ssignment to CBRR versus UC on the proportion of fmilies with t lest 1 night in emergency shelter. Compred with the UC group, lower proportion of the CBRR group is in emergency shelter in months 2 through 5 fter rndom Exhibit 7-7. Housing Stbility Outcomes for the CBRR Rndom Assignment Group by Use of CBRR Outcome Homeless or doubled up during the followup period Fmilies Assigned to CBRR Who Did Not Use CBRR N = 181 Fmilies Assigned to CBRR Who Used CBRR N = 274 At lest 1 night homeless or doubled up in pst 6 months or in shelter in pst 12 months (%) At lest 1 night homeless or doubled up in pst 6 months (%) At lest 1 night homeless in pst 6 months (%) At lest 1 night doubled up in pst 6 months (%) Any sty in emergency shelter in months 7 to 18 fter RA (%) Number of dys homeless or doubled up in pst 6 months Number of dys homeless in pst 6 months Number of dys doubled up in pst 6 months Housing independence Living in own house or prtment t followup (%) Living in own house or prtment with no housing ssistnce (%) Living in own house or prtment with housing ssistnce (%) Number of plces lived Number of plces lived in pst 6 months Housing qulity Persons per room Housing qulity is poor or fir (%) Length of sty in shelter Length (in months) of bseline sty in emergency shelter CBRR = community-bsed rpid re-housing. RA = rndom ssignment. Difference in mens is sttisticlly significnt t the.10 level. The length of bseline sty in emergency shelter outcome includes one-hlf of the month of rndom ssignment nd is topcoded t 18 months. The 17 percent of fmilies ssigned to CBRR whose bseline shelter sty does not pper in the Progrm Usge Dt re not included in the nlysis. Notes: Mens re unweighted. See Chpter 5 nd Appendix B for outcome definitions. Sources: Fmily Options Study bseline survey; 18-month followup survey; Progrm Usge Dt 101 Neither the differences nor the similrities between the groups outcomes cn be cuslly ttributed to the use of CBRR, becuse the use of CBRR ws not rndomly ssigned. Fmilies ssigned to CBRR who used CBRR my be systemticlly different from fmilies ssigned to CBRR who did not use CBRR. The direction of ny such bis is uncler. Fmilies who used CBRR my hve hd greter need, such tht, given use of the sme tretment, they would hve hd worse outcomes thn fmilies who did not use CBRR. Alterntively, fmilies who used CBRR my hve hd more cpcity to nvigte the system thn those who did not such tht, given the use of the sme tretment, they would hve hd better outcomes. Therefore, lthough being more likely to live in one s own house or prtment nd quicker exit from shelter my result from the use of CBRR, it cnnot be conclusively ruled out tht these differences re entirely explined by preexisting differences between the groups. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 84

117 Chpter 7. Impcts of Community-Bsed Rpid Re-Housing (CBRR) Compred With Usul Cre (UC) Exhibit 7-8. CBRR Versus UC: Percent of Fmilies With t Lest 1 Night Sty in Emergency Shelter During Month, by Month After RA 100% 90% 80% 70% Percent of fmilies 60% 50% 40% 30% 20% 10% CBRR UC 0% Month fter RA CBRR = community-bsed rpid re-housing. UC = usul cre. RA = rndom ssignment. Notes: Percentges re weighted for survey nonresponse to represent ll UC fmilies in the study. Missing dt on emergency shelter stys bises the percentges somewht downwrd. The bseline sty in emergency shelter does not pper in the dt for 18.7 percent of UC respondent fmilies. The missing dt rte for subsequent stys in emergency shelter is unknown. Source: Fmily Options Study Progrm Usge Dt ssignment, suggesting tht priority ccess to CBRR led to fster exit from shelter for some fmilies. However, the difference in use of emergency shelter is sttisticlly significnt only in 6 of the 18 months of observtion (months 2 to 5, 11, nd 12 t the.10 level; month 3 t the.05 level) nd fdes in the ltter prt of the followup period, suggesting tht nother group of CBRR fmilies remined in shelter rther thn leving soon fter being given priority ccess to CBRR (or left shelter nd returned). A sttisticl test of whether the offer of priority ccess to CBRR speeds exit from emergency shelter, shown in Exhibit 7-9, shows tht the length of the continued sty in shelter fter rndom ssignment is one-hlf month shorter thn for UC fmilies, nd this difference is sttisticlly significnt. Overll, it ppers tht the offer of priority ccess to CBRR hs virtully no effect on housing stbility during the followup period reltive to the mix of services tht were vilble to fmilies in the UC intervention. Exhibit 7-9. CBRR Versus UC: Impct on Length of Bseline Sty in Emergency Shelter Outcome N CBRR UC ITT Impct Men (SD) Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 85 N Men (SD) Impct (SE) Length (in months) of bseline sty in emergency shelter b * 0.11 (4.4) (4.6) (0.3) CBRR = community-bsed rpid re-housing. UC = usul cre. ITT = intention-to-tret. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire UC group. b The length of bseline sty in emergency shelter outcome includes one-hlf of the month of rndom ssignment nd is topcoded t 18 months. The 19 percent of fmilies in this comprison whose bseline shelter sty does not pper in the Progrm Usge Dt re not included in the nlysis. Notes: Impct estimte nd outcome mens re regression djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Appendix B for outcome definition. Sources: Fmily Options Study 18-month followup survey; Progrm Usge Dt Effect Size

118 Chpter 7. Impcts of Community-Bsed Rpid Re-Housing (CBRR) Compred With Usul Cre (UC) 7.4 Impcts on Fmily Preservtion in the Community-Bsed Rpid Re-Housing (CBRR) Versus Usul Cre (UC) Comprison As ddressed in Chpter 3, ny effects of ssignment to CBRR on fmily preservtion would be expected to be indirect, vi increses in housing stbility, but such effects were not evident. As shown in Exhibit 7-10, no evidence in fct exists of such effects on fmily seprtions mong children or mong spouses or prtners who were with the respondent t the time of the bseline survey, nor is there evidence of effects on reunifictions of the much smller number of fmily mem - bers who were seprted from the respondent t tht time. No effect ws detected on foster cre plcements. 7.5 Impcts on Adult Well-Being in the Community-Bsed Rpid Re-Housing (CBRR) Versus Usul Cre (UC) Comprison As ddressed in Chpter 3, the theory nd gols of the CBRR intervention compred with those of UC do not hypothesize importnt effects on dult well-being. Consistent with this expecttion, Exhibit 7-11 provides no evidence tht dults Exhibit CBRR Versus UC: Impcts on Fmily Preservtion Outcome Current or recent seprtions of fmily members present t bseline CBRR UC ITT Impct Effect Size N Men (SD) N Men (SD) Impct (SE) Fmily hs t lest one child seprted in pst 6 months (%) (34.8) (36.2) 2.0 (2.5) 0.05 Fmily hs t lest one foster cre plcement in pst 6 months (%) (15.5) (18.0) 1.1 (1.3) 0.05 Spouse/prtner seprted in pst 6 months, of those with spouse/prtner present t RA (%) (49.9) (48.9) 9.4 (6.5) 0.17 Reunifiction of fmily members reported s seprted t bseline Fmily hs t lest one child reunified, of those fmilies with t lest one child (47.7) (45.0) 6.1 (7.3) 0.12 bsent t RA (%) Spouse/prtner reunified, of those with spouse/prtner bsent t RA (%) (45.1) (47.9) 0.8 (12.0) CBRR = community-bsed rpid re-housing. UC = usul cre. ITT = intention-to-tret. RA = rndom ssignment. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire UC group. Notes: Impct estimtes nd outcome mens re regression djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Source: Fmily Options Study 18-month followup survey Exhibit CBRR Versus UC: Impcts on Adult Well-Being Adult physicl helth Outcome CBRR UC ITT Impct Effect Size N Men (SD) N Men (SD) Impct (SE) Helth in pst 30 dys ws poor or fir (%) (45.0) (46.8) 3.8 (3.1) 0.07 Adult mentl helth Gol-oriented thinking b (0.97) (1.07) 0.09 (0.07) 0.07 Psychologicl distress c (5.23) (5.74) 0.45 (0.37) 0.07 Adult trum symptoms Post-trumtic stress disorder (PTSD) symptoms in pst 30 dys (%) (42.2) (43.6) 0.1 (3.0) Adult substnce use Alcohol dependence or drug buse d (%) (31.9) (34.3) 3.1 (2.4) 0.08 Alcohol dependence d (%) (28.4) (31.2) 2.9 (2.2) 0.08 Drug buse d (%) (19.0) (23.4) 2.3 (1.5) 0.09 Experience of intimte prtner violence Experienced intimte prtner violence in pst 6 months (%) (31.1) (32.5) 1.1 (2.3) 0.03 CBRR = community-bsed rpid re-housing. UC = usul cre. ITT = intention-to-tret. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire UC group. b Gol-oriented thinking, mesured with modified version of the Stte Hope Scle, rnges from 1 to 6, with higher scores indicting higher levels of positive, gol-oriented thinking. c Psychologicl distress is mesured with the Kessler 6 (K6) scle nd rnges from 0 to 24, with higher scores indicting greter distress. d Alcohol dependence is mesured with the Rpid Alcohol Problems Screen (RAPS4), nd drug buse is mesured with six items from the Drug Abuse Screening Test (DAST-10). Notes: Impct estimtes nd outcome mens re regression djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. Chpter 5 nd Appendix B for outcome definitions. Source: Fmily Options Study 18-month followup survey Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 86

119 Chpter 7. Impcts of Community-Bsed Rpid Re-Housing (CBRR) Compred With Usul Cre (UC) benefited from priority ccess to this intervention cross ny of the outcomes mesured. In prticulr, no significnt effects pper for dult helth (physicl or mentl), trum nd violence, or substnce use nd buse. 7.6 Impcts on Child Well-Being in the Community-Bsed Rpid Re-Housing (CBRR) Versus Usul Cre (UC) Comprison As ddressed in Chpter 3, ny effects of priority ccess to CBRR on child well-being would be expected to be indirect, vi increses in housing stbility, but such effects were not evident. CBRR hs bout twice s mny significnt impcts (in 5 of 26 tests) s would be expected by chnce lone, but those impcts do not conform to ny pttern, with 2 effects fvoring children rndomized to CBRR nd 3 fvoring UC. As shown in Exhibit 7-12, 3 of 15 cross-ge outcomes re significnt, but in opposite directions. Fmilies rndomized to CBRR reported slightly lower school enrollment (3 percent - ge points) but lso fewer bsences (1 in 8 children hd one fewer bsence in the pst month) nd slightly better experiences in school (effect size of 0.1). As shown in Exhibit 7-13, 2 of 11 ge-specific impcts re significnt, both for the older 8- to 17-yer-old ge group. Children in the CBRR group reported more fers (3 more fers out of 33 rted lot rther thn some ). They lso reported lower school effort thn children in the UC group (one in five more UC children reported tht they did bout s well s they could rther thn tht they could hve done little better ). Although the result for low birth weight ppers lrge, very few children were born in the 18 months fter rndom ssignment, so the number of children included in the test is quite smll, nd the result Exhibit CBRR Versus UC: Impcts on Child Well-Being Across Age Groups Child eduction Outcome CBRR UC ITT Impct Effect Size N Men (SD) N Men (SD) Impct (SE) Preschool or Hed Strt enrollment b (%) (49.0) (48.2) 6.5 (5.5) 0.11 School enrollment c (%) (29.0) (24.3) 3.2* (1.8) 0.10 Childcre or school bsences in pst month d (0.93) (0.97) 0.13* (0.08) 0.10 Number of schools ttended since RA e (0.88) (0.89) 0.05 (0.07) 0.04 Grde completion (not held bck) (%) (24.4) (28.5) 2.0 (2.2) 0.05 Positive childcre or school experiences f (0.55) (0.57) 0.07* (0.04) 0.10 Positive childcre or school ttitudes g (0.98) (0.99) 0.06 (0.07) 0.04 School grdes h (0.97) (0.96) 0.04 (0.08) 0.03 Childcre or school conduct problems i (0.42) (0.43) 0.05 (0.03) 0.09 Child physicl helth Poor or fir helth (%) (21.7) (21.5) 0.1 (1.3) Well-child checkup in pst yer (%) (29.6) (27.1) 3.0 (2.3) 0.08 Child hs regulr source of helth cre (%) (23.2) (25.3) 0.7 (1.8) 0.02 Sleep problems j (1.07) (1.12) 0.09 (0.07) 0.06 Child behviorl strengths nd chllenges Behvior problems k (1.29) (1.27) 0.13 (0.10) 0.08 Prosocil behvior l (1.08) (1.06) 0.05 (0.08) 0.03 CBRR = community-bsed rpid re-housing. UC = usul cre. ITT = intention-to-tret. RA = rndom ssignment. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire UC group. b Bse for preschool or Hed Strt enrollment is children ges 1 yer, 6 months to 5 yers. c Bse for school enrollment is children ges 6 to 17 yers. d Absences outcome is defined s 0 = no bsences in pst month, 1 = 1 to 2 bsences, 2 = 3 to 5 bsences, 3 = 6 or more bsences. e Number of schools outcome is topcoded t 4 or more schools. f Positive childcre or school experiences outcome is defined s 1 = mostly negtive experiences, 0 = both positive nd negtive experiences, 1 = mostly positive experiences. g Positive childcre or school ttitudes outcome is prent report of how much child likes school nd rnges from 1 to 5, with higher vlues indicting greter like of school. h School grdes outcome is defined s 1 = mostly Ds or Fs, 2 = mostly Cs, 3 = mostly Bs, 4 = mostly As. i Childcre or school conduct problems outcome is defined s 0 = no conduct problems reported to prent, 1 = prent contcted bout conduct problems or suspension or expulsion from school or childcre center. j Sleep problems outcome rnges from 1 to 5, with higher vlues indicting more frequent tiredness upon wking nd during the dy. k Behvior problems outcome is mesured s the stndrdized Totl Difficulties score from the Strengths nd Difficulties Questionnire (SDQ). l Prosocil behvior is mesured s the stndrdized Prosocil domin score from the SDQ. Notes: Impct estimtes nd outcome mens re regression djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Source: Fmily Options Study 18-month followup survey (prent report) Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 87

120 Chpter 7. Impcts of Community-Bsed Rpid Re-Housing (CBRR) Compred With Usul Cre (UC) Exhibit CBRR Versus UC: Impcts on Child Well-Being Developmentl Outcomes by Age Group Outcome Ages 1 yer to 3 yers, 6 months CBRR UC ITT Impct Effect Size N Men (SD) N Men (SD) Impct (SE) Met developmentl milestones b (%) (40.2) (40.8) 6.3 (6.6) 0.12 Low birth weight c (%) (21.3) (28.8) 12.3 (10.0) 0.34 Ages 3 yers, 6 months to 7 yers Verbl bility d (1.02) (0.96) 0.05 (0.12) 0.03 Mth bility e (0.83) (0.78) (0.10) Executive functioning (self-regultion) f (15.83) (15.90) 0.22 (1.32) Ages 8 to 17 yers Anxiety g (7.60) (7.45) 1.01 (0.81) 0.09 Fers h (14.03) (13.93) 2.93* (1.50) 0.14 Substnce use i (%) (31.44) (29.36) 2.59 (2.96) 0.07 Gol-oriented thinking j (4.21) (5.00) 0.46 (0.46) 0.07 School effort in pst month k (0.82) (0.77) 0.20*** (0.08) 0.19 Arrests or police involvement in pst 6 months l (%) (30.75) (33.66) 3.59 (4.55) 0.09 CBRR = community-bsed rpid re-housing. UC = usul cre. ITT = intention-to-tret. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire UC group. b Met developmentl milestones outcome is defined s scoring bove the typicl development cutoffs in ll domins of the Ages nd Stges Questionnire (ASQ-3). c Bse for low birthweight outcome (prent report) is children born since rndom ssignment who re t lest 1 yer old t followup. d Verbl bility outcome is the ntionlly stndrdized score from the Woodcock-Johnson III (WJ III) letter-word identifiction scle. e Mth bility outcome is the ntionlly stndrdized score from the WJ III pplied problems test. f Executive functioning outcome is the Hed Toes Knees Shoulders (HTKS) score nd rnges from 0 to 40, with higher scores indicting greter executive functioning. g Anxiety (child report) is mesured using the A-Trit scle from the Stte-Trit Anxiety Inventory for Children (STAIC). Scores rnge from 20 to 60, with higher scores indicting greter nxiety. h Fers outcome (child report) is the score from the Fers Scle nd rnges from 33 to 99, with higher scores indicting more fer. i Substnce use (child report) is mesured with 23 items from the Centers for Disese Control nd Prevention (CDC) 2011 Youth Risk Behvior Survey. j Gol-oriented thinking (child report) is mesured with modified version of the Children s Hope Scle nd rnges from 6 to 30, with higher scores indicting higher levels of positive, gol-oriented thinking. k School effort outcome (child report) rnges from 1 to 4, with higher scores indicting greter effort during school dy nd on homework. l Arrest or police involvement in pst 6 months is from prent report. Notes: Impct estimtes nd outcome mens re regression djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Sources: Fmily Options Study 18-month followup survey (prent report); Fmily Options Study 18-month child survey (child report); ASQ-3; WJ III; HTKS is not sttisticlly significnt. Both the smll number of significnt effects nd the fct tht they do not consistently fvor one group suggest tht CBRR hd little or no effect on child outcomes reltive to UC. 7.7 Impcts on Self-Sufficiency in the Community-Bsed Rpid Re-Housing (CBRR) Versus Usul Cre (UC) Comprison By incresing housing stbility, CBRR my hve indirect effects on fmily self-sufficiency reltive to UC. In prticulr, spending 4 to 6 months in stble housing within fmilies own communities with shrply lower burden of housing costs could enble dult fmily members to concentrte more on employment nd ernings nd even enhnce their skills through eduction nd trining prticiption. In some CBRR progrms, cse mngement guidnce nd referrls my further enhnce efforts t work nd ccess to resources tht mke fmilies more self-sufficient. For the 18-month period following ssignment to CBRR, 20 indictors tell us whether priority ccess to CBRR in fct boosts fmily self-sufficiency compred with UC (see Exhibit 7-14). Effects pper to hve occurred for three self-sufficiency outcomes: (1) totl fmily income (incresed 12 percent), (2) shre of fmilies receiving the Supplementl Nutrition Assistnce Progrm (SNAP; incresed by 4 percentge points), nd (3) household food insecurity (6 fewer fmilies out of 100). In other self-sufficiency res, most notbly in the subdomins of employment nd eduction nd trining, the study tem found no evidence of impcts of CBRR. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 88

121 Chpter 7. Impcts of Community-Bsed Rpid Re-Housing (CBRR) Compred With Usul Cre (UC) Exhibit CBRR Versus UC: Impcts on Self-Sufficiency Employment sttus Outcome CBRR UC ITT Impct Effect Size N Men (SD) N Men (SD) Impct (SE) Work for py in week before survey (%) (47.9) (47.4) 0.1 (3.3) Any work for py since RA (%) (47.4) (48.3) 0.8 (3.1) Months worked for py since RA b (8.0) (7.7) 0.7 (0.5) 0.09 Hours of work per week t current min job c (16.6) (16.0) 0.4 (1.1) 0.02 Income sources nd mounts Annulized current ernings ($) 445 5,663 (9,638) 438 5,364 (9,625) 300 (640) 0.03 Totl fmily income ($) ,201 (7,704) 439 9,073 (8,010) 1,128** (505) 0.13 Anyone in fmily hd ernings in pst month (%) (49.9) (49.5) 2.5 (3.4) 0.04 Anyone in fmily received TANF in pst month (%) (46.9) (45.2) 3.7 (3.0) 0.07 Anyone in fmily received SSDI in pst month (%) (25.6) (27.8) 0.5 (1.7) 0.02 Anyone in fmily received SSI in pst month (%) (35.3) (34.0) 2.5 (2.1) 0.06 Anyone in fmily received SNAP/food stmps in pst month (%) (33.6) (37.9) 4.4* (2.5) 0.10 Anyone in fmily received WIC in pst month (%) (46.4) (46.2) 0.7 (3.1) Eduction nd trining Prticipted in 2 weeks or more of ny school or trining since RA (%) (45.0) (42.8) 2.4 (3.0) 0.05 Number of weeks in school or trining progrms since RA (10.1) (9.2) 0.2 (0.8) 0.02 Prticipted in 2 weeks or more of school since RA (%) (22.8) (25.0) 1.6 (1.6) 0.05 Prticipted in 2 weeks or more of bsic eduction since RA (%) (14.0) (12.4) 0.2 (0.8) 0.02 Prticipted in 2 weeks or more of voctionl eduction since RA (%) (25.6) (25.7) 1.1 (1.8) 0.04 Food security nd hunger Household is food insecure (%) (45.1) (47.8) 6.1* (3.3) 0.11 Food insecurity scle d (1.97) (2.06) 0.21 (0.14) 0.09 Economic stressors Economic stress scle e (0.49) (0.52) 0.05 (0.03) 0.09 CBRR = community-bsed rpid re-housing. UC = usul cre. ITT = intention-to-tret. RA = rndom ssignment. SD = stndrd devition. SE = stndrd error. SNAP = Supplementl Nutrition Assistnce Progrm. SSDI = Socil Security Disbility Insurnce. SSI = Supplementl Security Income. TANF = Temporry Assistnce for Needy Fmilies. WIC = Specil Supplementl Nutrition Progrm for Women, Infnts, nd Children. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire UC group. b Number of months worked for py includes prtil clendr months. c Hours of work per week includes those not currently working (tht is, those with 0 hours of work per week). d Food insecurity scle rnges from 0 to 6, with higher vlues indicting higher food insecurity. e Economic stress scle rnges from 1 to 1, with higher vlues indicting higher economic stress. Notes: Impct estimtes nd outcome mens re regression djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Source: Fmily Options Study 18-month followup survey 7.8 Summry of the Community- Bsed Rpid Re-Housing (CBRR) Versus Usul Cre (UC) Comprison Across Domins In the CBRR-versus-UC comprison, 60 percent of fmilies ssigned to CBRR nd 20 percent of fmilies ssigned to UC received CBRR. This contrst in progrm use did not led to notble differences in experiences between the CBRR nd UC fmilies. The vst mjority of the evidence involving dozens of outcomes in five domins suggests equivlent results for housing stbility, fmily preservtion, nd dult nd child well-being with or without privileged ccess to CBRR fter 7 dys in shelter. Assignment to CBRR my hve hd some consequences for children, but the indictions to this effect re limited nd mixed in direction. Also, priority ccess to CBRR my hve improved fmily income (nnul income for CBRR fmilies ws $1,128 higher thn for UC fmilies) nd food security during the followup period nd my hve incresed the likelihood of receiving SNAP, lthough these results could be spurious mong mny self-sufficiency indictors exmined. Most strikingly, reltive to UC, the study tem did not find evidence tht priority ccess to CBRR ffected housing stbility over the followup period. The reson for the lck of effects on the outcomes tht CBRR is intended to ffect directly is uncler. Higher prticiption in other homeless nd housing ssistnce progrms mong UC fmilies my hve diminished the impct of CBRR. Chpter 9 reports on how CBRR compres to the other two ctive interventions, SUB nd PBTH. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 89

122 CHAPTER 8. IMPACTS OF PROJECT-BASED TRANSITIONAL HOUSING (PBTH) COMPARED WITH USUAL CARE (UC) This chpter presents estimtes of the impct of the project-bsed trnsitionl housing (PBTH) interven - tion compred with outcomes of fmilies served by the usul cre (UC) homeless ssistnce systems in their communities. The gol is to determine whether offering priority ccess to unit in PBTH progrm increses fmilies housing stbility nd improves other fmily outcomes during 20-month followup intervl. The chpter begins with description of the PBTH intervention s implemented in the study. It then shows the extent to which fmilies in both the PBTH nd UC groups used trnsitionl housing nd other housing nd service progrms vilble to them in the study sites. The next five sections present the effects of being offered the PBTH intervention (s compred with UC) on outcomes within the five study domins housing stbility, fmily preservtion, dult well-being, child well-being, nd self-sufficiency. 8.1 Project-Bsed Trnsitionl Housing (PBTH) Intervention The PBTH intervention provides plce for fmilies to sty for finite period of time during which they re provided wide rry of services tht include cse mngement nd either direct provision of or referrl to services identified through n ssessment of fmily needs. PBTH ws offered to study fmilies in ll of the sites except one. 102 In totl, 368 fmilies were rndomly ssigned to this intervention (rnging from 4 in Minnepolis, Minnesot, to 66 in Honolulu, Hwii) nd referred to 46 different PBTH progrms. 103 Of these 368 fmilies, 294 (80 percent) responded to the 18-month followup survey nd so re included in the impct nlysis in this report. The objectives of trnsitionl housing re to prepre fmilies for permnent housing by providing cse mngement nd other services tht help overcome brriers to housing stbility nd to ddress other psychosocil needs tht the fmily my hve. For this study, the tem selected trnsitionl housing progrms tht provide housing primrily in project-bsed fcilities or housing units. The study s definition specified tht PBTH does not llow for fmilies to trnsition in plce in privte-mrket prtments or single-fmily homes, tking over responsibility for the housing unit s lese towrd or t the end of the progrm of trnsitionl ssistnce. The study excluded trnsition-in-plce progrms in order to generte strong contrst between PBTH nd community-bsed rpid re-housing (CBRR), which lso is time limited nd uses scttered-site housing units in which the fmily cn sty nd py the rent t the end of the CBRR progrm. The few progrms to which study fmilies were referred tht offer scttered-site trnsitionl housing require fmilies to relocte to other housing t progrm completion. PBTH progrms often receive funding from federl Suppor - tive Housing Progrm (SHP) grnts, which results in some consistency cross PBTH progrms. For instnce, the SHP grnt limits trnsitionl housing ssistnce to 24 months, funds brod rnge of supportive services, nd sets prme - ters for the wy in which progrms must clculte prticipnt rent contributions when they choose to require them. Not ll the PBTH progrms in the study receive funding from SHP grnts, however. Most hve wide rnge of funding sources, including privte foundtion grnts nd locl fundrising proceeds. Some progrms re fith bsed, nd mny of those progrms re completely privtely funded. All PBTH progrms in the study provide only temporry housing ssistnce. The study tem llowed ny time limit on tenure but specificlly sought progrms tht offered t lest 6 months of ssistnce. Nerly ll progrms provided mximum of 24 months of ssistnce. Progrms offering 102 PBTH ws not offered in Boston, Msschusetts. Also, PBTH ws very limited in Minnepolis, Minnesot, with only four fmilies rndomly ssigned to PBTH. 103 Much of the informtion describing PBTH in this section is bsed on the 30 PBTH progrms tht provided progrm dt. These 30 progrms represent 293 of the totl 368 PBTH referrls. More detil bout specific PBTH progrms is provided in Gubits et l. (2013), Appendixes A nd B. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 90

123 Chpter 8. Impcts of Project-Bsed Trnsitionl Housing (PBTH) Compred With Usul Cre (UC) referrls to permnent housing ssistnce t the end of the trnsitionl housing period were included in the PBTH in - tervention for the study, but not progrms tht gurnteed permnent ssistnce Housing Assistnce in PBTH PBTH progrms offered plce to sty nd receive services in vried physicl environments. As shown in Exhibit 8-1, nerly three-fourths of fmilies were referred to PBTH progrms tht provided fmilies with individul prtments (or occsion - lly single-fmily houses) during their prticiption in the progrm. One-fourth were referred to progrms tht provided privte sleeping rooms but shred kitchens or bthrooms. Fmily Pyments nd Svings Requirements in PBTH Most fmilies (92 percent) were referred to PBTH progrms tht required progrm fee or rent contribution from progrm prticipnts, bsed on 30 percent of their income (80 percent of progrms tht chrged fee) or other fctors such s fmily or unit size (20 percent of progrms tht chrged fee; see Exhibit 8-2). More thn one-hlf of the fmilies were referred to PBTH progrms tht required them to sve money while in the progrm. Fmilies in PBTH usully te their mels independently while enrolled in the progrm; 74 percent of fmilies referred to PBTH progrms were responsible for providing their own food while living in PBTH. The gency provided food in three progrms (representing 10 percent of referred PBTH fmilies) for which the progrms fcilities were former hotels where fmilies did not hve privte kitchen fcilities, nd 16 percent of fmilies ssigned to PBTH were referred to progrms tht provided some but not ll of the fmilies food. Some gencies commented tht, if fmilies were respon - sible for t lest one mel per dy, they would be eligible for Supplementl Nutrition Assistnce Progrm (SNAP) benefits Assessment of Fmily Needs in PBTH All PBTH progrms indicted tht the progrm conducted forml ssessment of study fmilies t the beginning of the progrm. 104 Progrm stff reported tht the ssessments Exhibit 8-1. PBTH Housing Settings Type of PBTH Percent of Fmilies Assigned to PBTH Progrms With Housing Units of This Type (N = 293) Seprte prtment with privte kitchen nd bthroom 73 Privte sleeping room but shred kitchen, bthroom, or both 27 PBTH = project-bsed trnsitionl housing. Progrm dt were not collected from 16 progrms tht collectively hd 75 PBTH fmily referrls. Note: Percentges re unweighted. Sources: Progrm dt; rndom ssignment records Exhibit 8-2. Fmily Rent Contributions nd Svings Requirements in PBTH PBTH Progrm Fetures Are fmilies required to py progrm fee or rent? Percent of Fmilies Referred to PBTH Progrms With These Chrcteristics (N = 293) Yes 92 No 8 (If yes) How is the progrm fee or rent determined? Percentge of income 80 Flt mount bsed on fmily or unit size 20 Does the progrm require fmilies to sve money while in the progrm? Yes 55 No 45 Who is responsible for food for prticipting fmilies? Fmilies provide own food 74 Progrm provides food 10 Both 16 PBTH = project-bsed trnsitionl housing. Progrm dt were not collected from 16 progrms tht collectively hd 75 PBTH fmily referrls. Note: Percentges re unweighted. Sources: Progrm dt; rndom ssignment records 104 As prt of progrm dt collection, the study tem sked progrm stff whether they conducted n ssessment of fmilies, when it occurred, which domins were ddressed s prt of the ssessment, whether stndrdized tool ws used to ensure tht ssessments were conducted nd reported systemticlly cross progrm stff, nd the extent to which ssessments resulted in gol setting nd service plns for fmily members. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 91

124 Chpter 8. Impcts of Project-Bsed Trnsitionl Housing (PBTH) Compred With Usul Cre (UC) covered brod rnge of topics, exploring fmily needs relted to housing; self-sufficiency; employment; physicl helth, mentl helth, nd substnce buse; child-specific needs; prenting; nd fmily life skills. A few progrms included other ssessment domins such s domestic violence, trum, debt burden, nd culturl needs, but these domins were not widespred res of ssessment. The ssessments ll resulted in forml service pln (or equivlent) with gols for the dults in the household, designed to help fmilies ddress their needs. Fifty-eight percent of fmilies were referred to progrms tht developed gols specificlly for the children Supportive Services Provided in PBTH Prticipting PBTH progrms provided comprehensive cse mngement nd provided mny supportive services directly, in some cses mking referrls to other progrms tht com - mitted to providing the service. Although ll PBTH progrms focused on ending fmily s homelessness through plcement in permnent housing, more thn 90 percent of fmilies were referred to PBTH progrms tht lso provided services relted to self-sufficiency (for exmple, finncil mngement, or help obtining public benefits) nd to employment nd trining. Exhibit 8-3 shows the wide rry of services offered in PBTH progrms nd the extent to which the service ws provided through cse mngement, by other progrm or gency stff beyond the cse mnger, or through forml rrngement with n externl gency tht gurnteed the fmily s ccess to the service becuse of PBTH enrollment. The second column of the exhibit shows the percentge of fmilies referred to progrms tht offered ech type of service. The subsequent columns report the percentge of fmilies referred to progrms tht provided the service s prt of cse mngement or in other wys. In some cses, ddressing service through cse mngement ment providing direct ssistnce by the cse mngers, wheres in other cses ddressing service through cse mngement ment tht the cse mngers provided referrls to other progrms, dvocted on behlf of the fmily to ccess the service, helped remove brriers to receiving the service, or coched nd supported fmily in their ttempts to obtin the service. In ddition to the nerly universl focus on self-sufficiency nd employment nd trining services, more thn 75 percent of PBTH fmilies were referred to progrms tht provided services to ddress life skills, mentl helth cre, prenting needs, nd physicl helth cre. About two-thirds of PBTH fmilies were referred to progrms tht provided child dvoccy nd cre relted to substnce buse. Cse Mngement Intensity in PBTH PBTH progrms considered cse mngement core prt of the intervention. PBTH providers often described the centrl focus of their progrms s cse mngement rther thn providing fmilies with plce to sty. The verge cse mngement rtio for PBTH progrms ws 20 fmilies per PBTH cse mnger. 105 Exhibit 8-4 shows tht nerly three-fifths of fmilies were referred to progrms in which Exhibit 8-3. Types of Supportive Services Offered in PBTH Progrms nd How They Are Delivered Types of Supportive Services Percent of Fmilies Referred to PBTH Progrms Tht Offer These Services (N = 293) Through Cse Mngement Percent of Fmilies Referred to PBTH Progrms Tht Offer Services of This Type By Other Progrm or Agency Stff Through Dedicted Linkges With Other Agencies Housing serch nd plcement ssistnce 100% 100% 16% 4% Self-sufficiency (overll) NA NA Childcre/fter-school cre Finncil litercy/money mngement 8 14 Help obtining public benefits 3 2 Trnsporttion 0 0 Employment nd trining Life skills Mentl helth cre Prenting skills Physicl helth cre Child dvoccy Substnce buse Fmily reunifiction PBTH = project-bsed trnsitionl housing. NA = dt not vilble. Progrm dt were not collected from 16 progrms tht collectively hd 75 PBTH fmily referrls. Note: Percentges re unweighted. Sources: Progrm dt; rndom ssignment records 105 See Chpter 7, footnote 5, for how the cse mngement rtio is mesured. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 92

125 Chpter 8. Impcts of Project-Bsed Trnsitionl Housing (PBTH) Compred With Usul Cre (UC) Exhibit 8-4. PBTH Cse Mngement Intensity (rtio nd frequency) Averge Number of Clients per Cse Mnger Percentge of Fmilies Referred to PBTH Progrms Tht Offer Cse Mngement in Ech of the Following Pckges (N = 285) Weekly or More Often Biweekly Monthly Qurterly 10 or fewer Clients to 20 clients to 30 clients More thn 30 clients Totl PBTH = project-bsed trnsitionl housing. Progrm dt on client rtios were not collected from 17 progrms tht collectively hd 83 PBTH fmily referrls. Note: Percentges re unweighted. Sources: Progrm dt; rndom ssignment records Totl cse mnger worked with 20 or fewer fmilies t time nd met with fmilies weekly if not more often. The other two-fifths were referred to progrms with lower intensity cse mngement, in which cse mngers hd ctive cselods of more thn 20 fmilies but generlly fewer thn 30. These progrms still often met with fmilies weekly, lthough some met only biweekly or monthly. Intensity cn lso be mesured by the mount of time tht cse mngers spent with fmilies t ech visit. Reported visit times vried from 15 to 90 minutes, with progrms reporting 1-hour cse mngement sessions most frequently. 106 The PBTH cse mngement rtios were higher nd the frequency of meetings ws lower thn for the UC shelters from which fmilies were recruited for the study, probbly reflecting the reltively longer period of time tht PBTH progrms expected to be working with fmilies. Cse mngement ws often offered for up to 6 months fter fmilies moved out of the PBTH progrm, but progrm stff generlly indicted tht postexit supportive contct ws initited by fmilies. These progrms sid tht they mintined n open-door policy for fmilies to contct them if desired but tht cse mngers did not initite continued regulr contct with fmilies Eligibility Criteri for PBTH Mny of the PBTH progrms estblished eligibility criteri to limit dmissions to the types of fmilies they deemed pproprite for PBTH ssistnce. PBTH progrms were firly restrictive in terms of which types of fmilies they trgeted. As prt of recruiting fmilies for this study, the study tem screened cndidtes to see which progrm they would qulify for, bsed on eligibility criteri provided by the prticipting progrms nd questions sked of the fmilies in the study s bseline survey. 107 Of the 1,564 fmilies considered for rn - dom ssignment to n vilble PBTH unit, nerly threefourths were screened for sobriety or willingness to engge in substnce buse tretment. More thn two-thirds of fmilies considered for PBTH were screened for minimum incomes or employment. 108 In ddition to screening relted to fmily behvior nd em - ployment, slightly more thn one-hlf of the fmilies screened for PBTH were sked questions to determine whether their households were the correct size for the vilble trnsitionl housing unit. Unit size criteri were n rtifct of the projectbsed nture of PBTH. For exmple, if two-bedroom unit ws vilble, fmily hd to hve the right number of people nd right mix of ges nd genders to be considered for the unit to void overcrowding or underuse. Fmily composition criteri limited enrollment for fmilies bsed on the types of people who were prt of their household. For instnce, some progrms excluded dult mles, mle children older thn ge 13, or ny children of either gender older thn certin ge. Fmily composition criteri sometimes reflected the chllenges of housing fmilies in congregte settings nd sometimes reflected progrm gols nd design. On the bsis of this screening, only 77 percent of study fmilies were eligible for rndom ssignment to PBTH. After rndom ssignment, 18 percent of those referred to PBTH progrms were screened out by progrms s ineligible despite 106 Additionl detils bout ech progrm s cse mngement re shown in Gubits et l. (2013), Appendix B The mjor ctegories of screening questions nd their reltive use in screening study fmilies for vilble PBTH, CBRR, nd permnent housing subsidy (SUB) openings re shown in Gubits et l. (2013), Exhibit Mny PBTH progrms (with 68 percent of fmilies referred) required fmilies to hve sufficient income to be ble to py their own rent in coming months. The progrms thus sked the study tem to limit referrls to fmilies who indicted (in responses to screening questions) hving some type of income, prticipting in Temporry Assistnce for Needy Fmilies (TANF), or being willing nd ble to obtin employment shortly fter enrolling in the progrm. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 93

126 Chpter 8. Impcts of Project-Bsed Trnsitionl Housing (PBTH) Compred With Usul Cre (UC) the previous screening by the reserch tem. Fmilies screened out t bseline re not included in the study smple, but those determined ineligible fter rndom ssignment re. Those fmilies determined ineligible by progrms must be included in the impct estimtes to preserve the comprbility of the PBTH nd UC fmilies tht is, some of the UC fmilies might lso hve been found ineligible by progrms hd they been ssigned to PBTH Progrm Rules in PBTH After they were enrolled in PBTH progrm, fmilies remined eligible for ssistnce (up to the mximum length of sty) s long s they complied with progrm rules. As shown in Exhibit 8-5, 52 percent of PBTH fmilies were referred to progrms tht imposed curfews, nd 42 percent of PBTH fmilies were referred to progrms tht limited overnight visitors, even though most PBTH is provided in privte prtment settings. 109 Eighty-seven percent of fmilies were referred to progrms tht required prticiption in services or ctivities in order to remin in the progrm. For instnce, PBTH progrms often required prticipnts to work with cse mnger to develop gols nd to identify nd pursue ctions needed to chieve them. Some progrms lso required prticiption in services such s money mngement clss, substnce buse ssessment, or group counseling sessions. 8.2 Progrm Use by Fmilies in the Project-Bsed Trnsitionl Housing (PBTH) Versus Usul Cre (UC) Comprison Ech impct comprison in the study my be thought of s distinct experiment or test, nd this chpter ddresses only the comprison between PBTH nd UC, without reference to the fmilies who were rndomized to the permnent housing subsidy (SUB) or community-bsed rpid re-housing (CBRR) interventions. In totl, 707 fmilies took prt in the test of PBTH versus UC. These fmilies ll hd the opportunity to be ssigned to PBTH or UC t the point of rndom ssignment nd were ssigned to one of these two interventions 368 fmilies to PBTH nd 339 fmilies to UC. 110 Seventy-nine percent of these 707 fmilies (294 PBTH fmilies nd 262 UC fmilies) responded to the 18-month followup survey, nd thus totl of 556 fmilies re included in the PBTHversus-UC impct comprison reported in this chpter. This section describes the extent to which the 294 PBTH fmilies used trnsitionl housing nd other types of homeless nd housing ssistnce during the followup period. Prllel informtion is presented for the 262 UC fmilies. The dt on progrm use do not distinguish between subtypes of trnsitionl housing nd include trnsition-in-plce ssistnce, so this section uses the bbrevition TH rther thn PBTH to describe the broder ctegory of ssistnce. Exhibit 8-5. Types of Progrm Rules in PBTH Types of Progrm Rules Percent of Fmilies Referred to PBTH Progrms With These Types of Rules (N = 293) Weekdy curfew 52 Weekend curfew 10 Limit on dytime visitors 9 Limit on overnight visitors 42 Complince with mndtory service requirements 87 PBTH = project-bsed trnsitionl housing. Progrm dt on client rtios were not collected from 17 progrms tht collectively hd 83 PBTH fmily referrls. Note: Percentges re unweighted. Sources: Progrm dt; rndom ssignment records 109 The percentges of study emergency shelters with these types of rules re shown in Gubits et l. (2013), Exhibit In generl, study emergency shelters were more likely to hve these types of rules thn study PBTH progrms. Most UC fmilies hd weekdy curfews (93 percent), limits on dytime visitors (70 percent), nd limits on overnight visitors (96 percent) in the study shelters from which they enrolled in the study. 110 In the entire study, 746 fmilies were rndomly ssigned to UC, but only 339 of these fmilies hd PBTH vilble to them when they were rndomized. Therefore, only these 339 UC fmilies re prt of the PBTH-versus-UC comprison smple. All 368 fmilies rndomly ssigned to PBTH during the course of the study hd UC vilble to them, so ll re prt of the PBTH-versus-UC comprison smple. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 94

127 Chpter 8. Impcts of Project-Bsed Trnsitionl Housing (PBTH) Compred With Usul Cre (UC) Exhibit 8-6 shows the use of seven types of homeless nd housing progrms by these fmilies. The first column shows the percentge of fmilies ssigned to PBTH who ever used ech progrm type during the followup period. 111 The third row (shded in the exhibit) shows the use of some type of TH by the fmilies ssigned to PBTH; 53.6 percent of fmilies in PBTH received TH ssistnce t some point during the followup period mening tht they either followed up on the referrl nd moved into the PBTH fcility or entered 112, 113 nother TH progrm. The second column shows the percentge of fmilies ssigned to UC who ever used ech progrm type during the followup period. 114 The shded row of the second column shows tht 29.1 percent of the UC fmilies received TH ssistnce during the followup period, despite not being given priority ccess to PBTH. Emergency shelter stff were requested by the study to not refer fmilies ssigned to UC to one of the ctive interventions. Nevertheless, s shown in the exhibit, fmilies were ble to lern bout trnsitionl housing pro - grms in their communities, nd these progrms hd progrm slots vilble t some point during the followup period. Rows 1, 2, nd 4 through 7 of the exhibit show prticiption in other types of homeless nd housing progrms. Twenty-five percent of these UC fmilies (not shown in the exhibit) found their wy to SUB or other permnent housing progrms during the followup period, presumbly through the regulr process of coming off witing lists nd lesing units, nd 12 percent found their wy to rpid re-housing ssistnce, despite the lck of preferentil ccess to those progrms through the study. The use of progrms other thn TH is generlly higher for the UC group thn for the PBTH group, presumbly becuse the UC group did not hve the PBTH intervention esily vilble nd so turned to other types of progrms. The eighth row shows the percentges of fmilies in the PBTH nd UC groups who used none of the seven Exhibit 8-6. PBTH Versus UC: Progrm Use Since Rndom Assignment Type of Housing Assistnce Percent Ever Used From RA to 18-Month Followup Survey Number of Months Used From RA to 18-Month Followup Survey, if Ever Used Type of Housing Assistnce Percent Used in Month of Followup Survey Response PBTH UC PBTH UC PBTH UC Men Medin Men Medin Subsidy (SUB) b Rpid re-housing (CBRR) Trnsitionl housing c Permnent supportive housing Public housing Project-bsed vouchers/section 8 projects Emergency shelter d No use of homeless or housing progrms e N PBTH = project-bsed trnsitionl housing. UC = usul cre. RA = rndom ssignment. Percentge of fmilies who ever used type of ssistnce progrm during the period from the month of RA to the month of the 18-month followup survey response (medin period durtion: 21 months). Percentges do not dd to 100 becuse some fmilies used more thn one progrm type during the followup period. b Subsidy ssistnce is housing choice vouchers plus site-specific progrms offered to fmilies ssigned to the permnent housing subsidy (SUB) group in Bridgeport, Connecticut nd Honolulu, Hwii. c Includes use of project-bsed, scttered-site, nd trnsition-in-plce trnsitionl housing. The inclusion of trnsition-in-plce ssistnce mkes this ctegory broder thn the study-defined PBTH intervention. d All fmilies were in emergency shelter t rndom ssignment. Percentges less thn 100 percent for ever used emergency shelter re due to missing dt on shelter use. e No use of homeless or housing progrms (ever used) indictes no use of the seven progrm types in this tble during ny of the followup period nd no use of emergency shelter fter the first 6 months fter RA. No use in the month of the followup survey response indictes no use of ny of these seven progrm types. Notes: Percentges re regression djusted, controlling for site nd rndomiztion rtio. Percentges, mens, nd medins re weighted for survey nonresponse to represent full comprison smple. Source: Fmily Options Study Progrm Usge Dt 111 The followup period is from the clendr month of rndom ssignment through the clendr month of response to the 18-month followup survey. Therefore, the length of the followup period differs cross fmilies. This period lsts for medin of 21 clendr months for the full smple. 112 All percentges, mens, nd medins in the exhibit re weighted to djust for survey nonresponse nd hence s best possible represent the full experimentl smple of 707 fmilies. The findings here on progrm use re thus in line with similrly weighted impct estimtes provided lter in the chpter. 113 The unweighted number of PBTH fmilies who used TH during the followup period is 159 fmilies. Of these, 92 fmilies were confirmed by enrollment verifiction to hve used the progrm to which they were referred by the study. It is not known how mny of the other 67 fmilies received PBTH or some other form of TH. 114 The percentges in the first six rows of these columns re not mutully exclusive becuse some fmilies use more thn one progrm type during the followup period. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 95

128 Chpter 8. Impcts of Project-Bsed Trnsitionl Housing (PBTH) Compred With Usul Cre (UC) types of progrms during the followup period, nor used emergency shelter from the seventh month fter rndom ssignment onwrd. About 20 percent of PBTH fmilies nd 30 percent of UC fmilies fll into this group. The men nd medin number of months of use for ech progrm type re lso shown in the exhibit (third nd fourth columns for PBTH fmilies, fifth nd sixth columns for UC fmilies) for only those fmilies who ever used given progrm type. 115 As one might expect, given tht TH ws redily vil - ble to PBTH fmilies, the number of months of TH use is higher for the fmilies who hd priority ccess to PBTH (medin of 12 months) thn for the 29.1 percent of UC fmilies who found their wy to TH (medin of 6 months) t some point during the followup period. The lst two columns of Exhibit 8-6 show the percentges of PBTH nd UC fmilies who received ech type of progrm in the clendr month of the followup survey response. Although the study tem expects most outcomes in the report to be influenced by ssistnce received during the entire followup period, some outcomes will be prticulrly strongly influenced by ssistnce received t the time of followup survey response. The shded row of the seventh column shows tht the TH ssistnce hd ended for bout 60 percent of the PBTH fmilies who ever received it nd for 66 percent of the UC fmilies who ever received it. 116 About one-fifth (22 percent) of PBTH fmilies were still in TH in the month of followup survey response compred with one-tenth of UC fmilies. The mjority of both PBTH nd UC fmilies (57 percent ech) were not using homeless or housing progrm t the time they responded to the followup survey. Thus, substntil differences re expected not in outcomes of PBTH nd UC fmilies in res tht reflect the fmilies current experience, but only in those tht reflect lsting influence of fmilies hving received the intensive cse mngement nd services provided by PBTH. As Exhibit 8-6 mkes cler, the PBTH fmilies used rnge of progrms in ddition to the progrm to which they were referred by the study, which is consistent with the design of the study. Fmilies were not required to use the intervention to which they were ssigned nd lso were not forbidden from using other progrms tht were vilble to them in their community. The intent of the study ws to mximize use of the ssigned ctive intervention (in this cse, mximize use of the PBTH intervention by the PBTH fmilies) nd to crete the lrgest possible contrst between the progrm mixes of different ssignment groups (in this cse, PBTH versus UC). As shown in the exhibit, the use of PBTH ws quite different for the PBTH nd UC groups. The contrst in usge of TH 53.6 percent for PBTH fmilies nd 29.1 percent for UC fmilies is sizble, lthough smller thn the nlogous contrst in either the SUB-versus-UC or CBRRversus-UC comprisons. As is conventionl in rndom ssignment nlyses, our gol is to estimte the intention-to-tret (ITT) impct tht is, the difference in impct by the progrm to which fmilies were ssigned, regrdless of whether they ctully used tht progrm (or some other progrm). This gol is consistent with the policy option of mking tretment vilble to fmily, but without the bility to force fmily to use tht tretment. Becuse not ll fmilies rndomly ssigned to PBTH used TH, nd some fmilies ssigned to UC did use TH, the true ITT impct is likely smller thn it would hve been hd the gp in TH usge been wider (ssuming tht TH truly hs nonzero impct on fmilies who use it). In prticulr, the difference in the use of TH by the PBTH nd UC groups is nrrow enough, given the reltively smll smple size vilble for nlysis, tht the study my hve filed to detect s sttisticlly significnt one or more ITT impcts lrge enough to be of policy importnce. Additionl detil bout the use of TH ssistnce by PBTH fmilies is shown in Exhibit 8-7. The exhibit shows tht bout one-fifth (21 percent) of PBTH fmilies who used TH did so for 18 or more months. The reminder of the chpter reports estimted impcts in the vrious outcome domins tht if sttisticlly significnt cn be cuslly ttributed to the offer of project-bsed trnsi - tionl housing to the fmilies rndomly ssigned to PBTH. 115 Hence, 0 vlues re not fctored into the mens, nor do they pull downwrd the medins of the vrious distributions. 116 Becuse 21.7 percent of PBTH fmilies were still using TH in the followup survey month, nd 53.6 percent hd used TH t some point during the followup period, it cn be clculted tht 1 (21.7/53.6) = 59.5 percent of PBTH fmilies who used TH t some point hd stopped using it by the survey month. A similr clcultion, 1 (9.8/29.1), yields 66.3 percent for UC fmilies. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 96

129 Chpter 8. Impcts of Project-Bsed Trnsitionl Housing (PBTH) Compred With Usul Cre (UC) Exhibit 8-7. Number of Months of Trnsitionl Housing Receipt During Followup Period by PBTH Fmilies Who Ever Used TH 25 Percent of PBTH fmilies who ever used TH Number of months of trnsitionl housing receipt or more PBTH = project-bsed trnsitionl housing. TH = trnsitionl housing. Percentges re weighted for survey nonresponse to represent ll fmilies in comprison smple. Note: N = 159. Source: Fmily Options Study Progrm Usge Dt 8.3 Impcts on Housing Stbility in the Project-Bsed Trnsitionl Housing (PBTH) Versus Usul Cre (UC) Comprison Proponents of PBTH emphsize tht most fmilies who become homeless hve brriers tht mke it hrd for them to secure nd mintin housing, s Chpter 3 ddresses. Thus, housing subsidies lone my be insufficient to ensure housing stbility nd other desirble outcomes. Fmily needs my rise from poverty, helth, disbility, or other problems tht led to homelessness to begin with or from the disruptive effects of homelessness on prents nd children. Proponents of PBTH believe tht by ddressing these brriers nd needs in supervised residentil setting, PBTH lys the best found - tion for ongoing stbility. The PBTH-versus-UC comprison offers evidence on whether this pproch is effective in the short term in improving fmily outcomes. It is importnt to keep in mind tht the followup period covered in this report is not quite long enough to observe the effect of PBTH fter progrm completion. About one-fifth of PBTH fmilies (nd one-tenth of UC fmilies) were still in TH in the month of followup response. Exhibit 8-8 shows the impcts of the offer of PBTH on home - lessness, housing independence, residentil moves, nd housing qulity. 117 The first row of the exhibit shows the impct on the confirmtory outcome of the study (homeless or doubled up in the pst 6 months, or sty in emergency shelter in the pst 12 months). The impct estimte indictes tht PBTH cuses 7.7-percentge-point reduction on the confirmtory outcome (from bout 52 to 44 percent). This estimte is sttisticlly significnt t the.10 level before the djustment for multiple comprisons but is not sttisticlly significnt fter djustment The homeless outcomes in this study diverge from the homeless definition finl rule in tht they do not include stys in trnsitionl housing in their definitions of being homeless. Additionl impcts on the use of trnsitionl housing during the followup period re provided in Appendix E. 118 The study estimtes impcts on this confirmtory outcome for ech of the six pired comprisons nd four pooled comprisons. Seven of these estimtes hve been prespecified s confirmtory tests. A multiple comprison procedure is performed to compute djusted p-vlues for these tests to reduce the possibility of chnce findings of sttisticl significnce. The detils of this procedure re provided in Appendix C. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 97

130 Chpter 8. Impcts of Project-Bsed Trnsitionl Housing (PBTH) Compred With Usul Cre (UC) Exhibit 8-8. PBTH Versus UC: Impcts on Housing Stbility Outcome Homeless or doubled up during the followup period PBTH UC ITT Impct Effect Size N Men (SD) N Men (SD) Impct (SE) At lest 1 night homeless b or doubled up in pst 6 months or (49.9) (50.1) 7.7* b (4.4) 0.14 in shelter in pst 12 months (%) [Confirmtory] c At lest 1 night homeless b or doubled up in pst 6 months (%) (48.8) (49.1) 4.6 (4.3) 0.08 At lest 1 night homeless b in pst 6 months (%) (39.1) (43.0) 7.3* (3.7) 0.15 At lest 1 night doubled up in pst 6 months (%) (47.5) (46.3) 0.1 (4.1) Any sty in emergency shelter in months 7 to 18 fter RA (%) (38.8) (44.9) 8.2** (3.6) 0.16 Number of dys homeless b or doubled up in pst 6 months (71.0) (75.1) 8.7 (6.7) 0.10 Number of dys homeless b in pst 6 months (41.8) (50.5) 7.7 (4.7) 0.14 Number of dys doubled up in pst 6 months (65.5) (64.4) 0.1 (5.8) Housing independence Living in own house or prtment t followup (%) (49.6) (49.0) 3.6 (4.7) 0.06 Living in own house or prtment with no housing ssistnce (%) (49.3) (48.1) 0.0 (4.6) Living in own house or prtment with housing ssistnce (%) (36.6) (43.0) 3.7 (3.7) 0.08 Number of plces lived Number of plces lived in pst 6 months (1.1) (1.2) 0.1 (0.1) 0.07 Housing qulity Persons per room (1.2) (1.2) 0.2 (0.1) 0.11 Housing qulity is poor or fir (%) (48.7) (47.3) 1.7 (4.4) 0.03 PBTH = project-bsed trnsitionl housing. UC = usul cre. ITT = intention-to-tret. RA = rndom ssignment. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire UC group. b The definition of homeless in this report includes stys in emergency shelters nd plces not ment for humn hbittion. It excludes trnsitionl housing. Additionl impcts on the use of trnsitionl housing in the followup period re provided in Appendix E. c After djustment for multiple comprisons, the impct on the confirmtory outcome is not sttisticlly significnt t the.10 level for the PBTH-versus-UC comprison. Notes: Impct estimtes nd outcome mens re regression djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Sources: Fmily Options Study 18-month followup survey; Progrm Usge Dt Two other homelessness outcomes show effects of PBTH reltive to UC: (1) 7.3-percentge-point reduction in expe - rience of homelessness in the pst 6 months, nd (2) n 8.2-percentge-point reduction in use of emergency shelter in months 7 to 18 fter rndom ssignment. The evidence of reductions in homelessness from two different dt sources (the followup survey nd the Progrm Usge Dt, the ltter of which reflects primrily Homeless Mngement Informtion System [HMIS] records) strengthens the finding tht PBTH reduced homeless stys in shelters or plces not ment for humn hbittion, even if not by n mount lrge enough to remin sttisticlly significnt fter the multiple comprisons djustment. PBTH ppers to hve no effect reltive to UC on the experience of being doubled up or the time spent doubled up in the 6 months before the survey response. Exhibit 8-9 shows the month-by-month impcts on the proportion of fmilies with t lest 1 night in emergency shelter during the month. This exhibit illustrtes tht somewht lower proportion of the PBTH group is in emergency shelter during the followup period s compred with the UC group. 119 The lst six rows of Exhibit 8-8 show tht, reltive to UC, PBTH hs no effect on the proportion of fmilies living in their own house or prtment (with or without ssistnce), no effect on number of moves, nd no effect on housing qulity. It my be tht these reltively modest effects of priority ccess to PBTH on housing stbility result in prt from the reltively low use of trnsitionl housing mong the PBTH ssignment group. The gin to the 53.6 percent of fmilies ssigned to PBTH who used trnsitionl housing TH ssistnce my hve been wtered down by weker outcomes mong the 46.4 percent who did not use it. To check this hypotheticl possi - bility, Exhibit 8-10 compres the outcomes of fmilies ssigned to PBTH who never took up TH with the corresponding outcomes for PBTH-ssigned fmilies who did use TH. 119 Sttisticlly significnt differences pper in 6 of the 18 months. The difference in month 2 is sttisticlly significnt t the.01 level; differences in months 4, 6, 12, nd 13 re sttisticlly significnt t the.05 level; nd the difference in month 17 is sttisticlly significnt t the.10 level. Becuse of missing dt on bseline stys, less thn 100 percent of ll PBTH nd UC fmilies re observed in shelter during month 0 (the month of rndom ssignment). The difference in the month 0 (significnt t the.10 level) should be considered chnce difference. The difference in unobserved initil shelter stys should be incresingly unrelted to differences observed lter in the followup period (s the initil shelter stys tht re unobserved in the dt become incresingly likely to hve ended s time elpses in the followup period). Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 98

131 Chpter 8. Impcts of Project-Bsed Trnsitionl Housing (PBTH) Compred With Usul Cre (UC) Exhibit 8-9. PBTH Versus UC: Percent of Fmilies With t Lest 1 Night Sty in Emergency Shelter During Month, by Month After RA 100% 90% 80% 70% Percent of fmilies 60% 50% 40% 30% 20% UC 10% PBTH 0% Month fter RA PBTH = project-bsed trnsitionl housing. UC = usul cre. RA = rndom ssignment. Notes: Percentges re weighted for survey nonresponse to represent ll UC fmilies in the study. Missing dt on emergency shelter stys bises the percentges somewht downwrd. The bseline sty in emergency shelter does not pper in the dt for 18.7 percent of UC respondent fmilies. The missing dt rte for subsequent stys in emergency shelter is unknown. Source: Fmily Options Study Progrm Usge Dt For some of the outcomes in Exhibit 8-10 including the percentge who reported t lest 1 night in shelter or plce not ment for humn hbittion or doubled up in the 6 months prior to the survey, number of dys homeless in the pst 6 months, nd number of plces lived in the pst 6 months results were indeed better for users of TH thn for nonusers, rising the possibility tht nonuse of TH could hve dmpened effects on these outcomes. For living in own house or prtment t followup, however, results were similr between the two groups. These results collectively indicte tht the incidence of homelessness within the PBTH group ws not minly confined to those fmilies who never used TH. This finding mkes it less plusible tht the impcts of PBTH would hve been substntilly lrger hd much wider contrst emerged in the use of TH between those rndomly ssigned to PBTH nd those ssigned to UC. 120 One outcome for which nonuse of TH my hve dmpened the effect of priority ccess to PBTH is the length of the initil sty in emergency shelter. The lst row of Exhibit 8-10 shows tht those fmilies who used TH spent n verge of 2.4 months in shelter fter study enrollment compred with 3.8 months for those who did not use TH, indicting correltion of TH use with shorter initil shelter stys. Exhibit 8-11 shows tht the effect of priority ccess to PBTH on the length of n initil shelter sty is not sttisticlly significnt. It is possible tht hd more PBTH fmilies used TH, the verge length of initil shelter sty for the entire PBTH group would hve been shorter thn tht for the UC group. Overll, it ppers tht PBTH leds to reduction in the proportion of fmilies experiencing homeless stys in shelters or plces not ment for humn hbittion reltive to UC t this erly followup point, when bout one-fifth of fmilies referred to PBTH were still stying in the PBTH fcility. PBTH does not pper to hve effects on other housing stbility outcomes t this point, however. 120 Neither the differences nor the similrities between the groups outcomes cn be cuslly ttributed to the use of TH, becuse the use of TH ws not rndomly ssigned. Fmilies ssigned to PBTH who used TH my be systemticlly different from fmilies ssigned to PBTH who did not use TH. The direction of ny such bis is uncler. Fmilies who used TH my hve hd greter need, such tht, given use of the sme tretment, they would hve hd worse outcomes thn fmilies who did not use TH. Alterntively, fmilies who used TH my hve hd more cpcity to nvigte the system thn those who did not such tht, given the use of the sme tretment, they would hve hd better outcomes. Therefore, it cnnot be conclusively ruled out tht differences in Exhibit 8-10 re entirely explined by preexisting differences between the groups. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 99

132 Chpter 8. Impcts of Project-Bsed Trnsitionl Housing (PBTH) Compred With Usul Cre (UC) Exhibit Housing Stbility Outcomes for the PBTH Rndom Assignment Group by Use of TH Outcome Homeless or doubled up during the followup period Fmilies Assigned to PBTH Who Did Not Use TH Fmilies Assigned to PBTH Who Used TH N = 135 N = 159 At lest 1 night homeless or doubled up in pst 6 months or in shelter in pst 12 months (%) At lest 1 night homeless or doubled up in pst 6 months (%) At lest 1 night homeless in pst 6 months (%) At lest 1 night doubled up in pst 6 months (%) Any sty in emergency shelter in months 7 to 18 fter RA (%) Number of dys homeless or doubled up in pst 6 months Number of dys homeless in pst 6 months Number of dys doubled up in pst 6 months Housing independence Living in own house or prtment t followup (%) Living in own house or prtment with no housing ssistnce (%) Living in own house or prtment with housing ssistnce (%) Number of plces lived Number of plces lived in pst 6 months Housing qulity Persons per room Housing qulity is poor or fir (%) Length of sty in shelter Length (in months) of bseline sty in emergency shelter PBTH = project-bsed trnsitionl housing. TH = trnsitionl housing. RA = rndom ssignment. Difference in mens is sttisticlly significnt t the.10 level. The length of bseline sty in emergency shelter outcome includes one-hlf of the month of rndom ssignment nd is topcoded t 18 months. The 23 percent of fmilies ssigned to PBTH whose bseline shelter sty does not pper in the Progrm Usge Dt re not included in the nlysis. Notes: Mens re unweighted. See Chpter 5 nd Appendix B for outcome definitions. Sources: Fmily Options Study bseline survey; 18-Month followup survey; Progrm Usge Dt Exhibit PBTH Versus UC: Impct on Length of Bseline Sty in Emergency Shelter Outcome N PBTH UC ITT Impct Men (SD) N Men (SD) Impct (SE) Length (in months) of bseline sty in emergency shelter b (3.6) (4.0) (0.4) PBTH = project-bsed trnsitionl housing. UC = usul cre. ITT = intention-to-tret. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire UC group. b The length of bseline sty in emergency shelter outcome includes one-hlf of the month of rndom ssignment nd is topcoded t 18 months. The 20 percent of fmilies in this comprison whose bseline shelter sty does not pper in the Progrm Usge Dt re not included in the nlysis. Notes: Impct estimte nd outcome mens re regression djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Appendix B for outcome definition. Sources: Fmily Options Study 18-month followup survey; Progrm Usge Dt Effect Size 8.4 Impcts on Fmily Preservtion in the Project-Bsed Trnsitionl Housing (PBTH) Versus Usul Cre (UC) Comprison Exhibit 8-12 shows the impcts of priority ccess to PBTH reltive to UC on fmily preservtion outcomes. As Chpter 3 ddresses, ny effects of ssignment to PBTH on fmily pres - ervtion would be expected to be indirect, vi housing stbility, self-sufficiency, nd dult well-being. PBTH s modest effects on homelessness did not pper to trnslte into effects on fmily seprtions mong children or mong spouses or prtners who were with the respondent t the time of the bseline survey, nor were there effects on reunifictions of the much smller number of fmily members who were seprted from the respondent t tht time. Priority ccess to PBTH hd no effect on foster cre plcements. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 100

133 Chpter 8. Impcts of Project-Bsed Trnsitionl Housing (PBTH) Compred With Usul Cre (UC) Exhibit PBTH Versus UC: Impcts on Fmily Preservtion Outcome Current or recent seprtions of fmily members present t bseline PBTH UC ITT Impct Effect Size N Men (SD) N Men (SD) Impct (SE) Fmily hs t lest one child seprted in pst 6 months (%) (35.6) (35.9) 0.6 (3.1) Fmily hs t lest one foster cre plcement in pst 6 months (%) (16.4) (19.3) 2.1 (1.9) 0.09 Spouse/prtner seprted in pst 6 months, of those with spouse/prtner present t RA (%) (44.6) (47.3) 1.2 (7.9) 0.02 Reunifiction of fmily members reported s seprted t bseline Fmily hs t lest one child reunified, of those fmilies with t lest one (47.3) (44.8) 1.9 (9.9) 0.04 child bsent t RA (%) Spouse/prtner reunified, of those with spouse/prtner bsent t RA (%) (50.9) (50.3) 11.0 (23.9) 0.19 PBTH = project-bsed trnsitionl housing. UC = usul cre. ITT = intention-to-tret. RA = rndom ssignment. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire UC group. Notes: Impct estimtes nd outcome mens re regression djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Source: Fmily Options Study 18-month followup survey 8.5 Impcts on Adult Well-Being in the Project-Bsed Trnsitionl Housing (PBTH) Versus Usul Cre (UC) Comprison Adult well-being is centrl focus of PBTH progrms, s noted in Chpter 3. Thus proponents would expect to see improvements, prticulrly in res of mentl helth nd substnce use. Exhibit 8-13 shows impcts of PBTH reltive to UC on dult well-being outcomes. Priority ccess to PBTH hs sttisticlly significnt impct on only one of the eight outcomes: 3.5-percentge-point reduction in drug buse. The intervention hd no effects on dult physicl or mentl helth or on trum symptoms, lcohol dependence, or expe - rience of intimte prtner violence. Given the lck of other effects, the impct on drug buse should be interpreted with cution. Overll, it ppers tht priority ccess to PBTH hs little effect on dult well-being compred with the mix of services vilble to the fmilies ssigned to UC. Exhibit PBTH Versus UC: Impcts on Adult Well-Being Adult physicl helth Outcome PBTH UC ITT Impct Effect Size N Men (SD) N Men (SD) Impct (SE) Helth in pst 30 dys ws poor or fir (%) (47.5) (46.9) 1.9 (4.0) 0.04 Adult mentl helth Gol-oriented thinking b (1.02) (0.98) 0.05 (0.09) 0.04 Psychologicl distress c (5.80) (5.75) 0.05 (0.50) Adult trum symptoms Post-trumtic stress disorder (PTSD) symptoms in pst 30 dys (%) (42.5) (42.9) 1.1 (3.8) 0.02 Adult substnce use Alcohol dependence or drug buse d (%) (36.4) (35.0) 0.5 (3.0) Alcohol dependence d (%) (32.4) (31.1) 0.3 (2.8) Drug buse d (%) (20.6) (25.4) 3.5* (1.9) 0.13 Experience of intimte prtner violence Experienced intimte prtner violence in pst 6 months (%) (28.9) (32.4) 1.1 (2.8) 0.03 PBTH = project-bsed trnsitionl housing. UC = usul cre. ITT = intention-to-tret. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire UC group. b Gol-oriented thinking is mesured with modified version of the Stte Hope Scle nd rnges from 1 to 6, with higher scores indicting higher levels of positive, gol-oriented thinking. c Psychologicl distress is mesured with the Kessler 6 (K6) scle nd rnges from 0 to 24, with higher scores indicting greter distress. d Alcohol dependence is mesured with the Rpid Alcohol Problems Screen (RAPS4), nd drug buse is mesured with six items from the Drug Abuse Screening Test (DAST-10). Notes: Impct estimtes nd outcome mens re regression djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Source: Fmily Options Study 18-month followup survey Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 101

134 Chpter 8. Impcts of Project-Bsed Trnsitionl Housing (PBTH) Compred With Usul Cre (UC) 8.6 Impcts on Child Well-Being in the Project-Bsed Trnsitionl Housing (PBTH) Versus Usul Cre (UC) Comprison As Chpter 3 ddresses, ny effects of PBTH on child well-being would be expected to be indirect. We found no evidence of such effects. Only 1 of 15 outcomes mesured cross ge groups (Exhibit 8-14) nd 0 of 11 ge-specific outcomes (Exhibit 8-15) ppered to reflect progrm impct. PBTH fmilies were 4 percentge points more likely to report tht children hd regulr source of helth cre. This result is probbly best interpreted s rndom vrition. Exhibit PBTH Versus UC: Impcts on Child Well-Being Across Age Groups Child eduction Outcome PBTH UC ITT Impct Effect N Men (SD) N Men (SD) Impct (SE) Size Preschool or Hed Strt enrollment b (%) (48.5) (48.6) 4.8 (6.9) 0.08 School enrollment c (%) (27.2) (28.6) 2.0 (2.4) 0.06 Childcre or school bsences in pst month d (0.92) (0.98) 0.06 (0.10) 0.05 Number of schools ttended since RA e (0.84) (0.90) 0.07 (0.09) 0.06 Grde completion (not held bck) (%) (27.3) (28.4) 0.7 (2.6) 0.02 Positive childcre or school experiences f (0.56) (0.57) 0.07 (0.05) 0.10 Positive childcre or school ttitudes g (0.99) (1.05) 0.07 (0.09) 0.05 School grdes h (0.87) (0.98) 0.10 (0.10) 0.08 Childcre or school conduct problems i (0.41) (0.43) 0.03 (0.04) 0.06 Child physicl helth Poor or fir helth (%) (22.3) (20.3) 2.5 (2.0) 0.09 Well-child checkup in pst yer (%) (27.8) (31.3) 2.5 (3.2) 0.07 Child hs regulr source of helth cre (%) (21.8) (25.5) 4.3* (2.3) 0.13 Sleep problems j (1.09) (1.15) 0.13 (0.10) 0.09 Child behviorl strengths nd chllenges Behvior problems k (1.13) (1.20) 0.13 (0.11) 0.08 Prosocil behvior l (1.11) (1.08) 0.07 (0.10) 0.05 PBTH = project-bsed trnsitionl housing. UC = usul cre. ITT = intention-to-tret. SD = stndrd devition. SE = stndrd error. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire UC group. b Bse for preschool or Hed Strt enrollment is children ges 1 yer, 6 months to 5 yers. c Bse for school enrollment is children ges 6 to 17 yers. d Absences outcome is defined s 0 = no bsences in pst month, 1 = 1 to 2 bsences, 2 = 3 to 5 bsences, 3 = 6 or more bsences. e Number of schools outcome is topcoded t 4 or more schools. f Positive childcre or school experiences outcome is defined s 1 = mostly negtive experiences, 0 = both positive nd negtive experiences, 1 = mostly positive experiences. g Positive childcre or school ttitudes outcome is prent report of how much child likes school nd rnges from 1 to 5, with higher vlues indicting greter like of school. h School grdes outcome is defined s 1 = mostly Ds or Fs, 2 = mostly Cs, 3 = mostly Bs, 4 = mostly As. i Childcre or school conduct problems outcome is defined s 0 = no conduct problems reported to prent, 1 = prent contcted bout conduct problems or suspension or expulsion from school or childcre center. j Sleep problems outcome rnges from 1 to 5, with higher vlues indicting more frequent tiredness upon wking nd during the dy. k Behvior problems outcome is mesured s the stndrdized Totl Difficulties score from the Strengths nd Difficulties Questionnire (SDQ). l Prosocil behvior is mesured s the stndrdized Prosocil domin score from the SDQ. Notes: Impct estimtes nd outcome mens re regression djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Source: Fmily Options Study 18-month followup survey (prent report) Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 102

135 Chpter 8. Impcts of Project-Bsed Trnsitionl Housing (PBTH) Compred With Usul Cre (UC) Exhibit PBTH Versus UC: Impcts on Child Well-Being Developmentl Outcomes by Age Group Outcome Ages 1 yer to 3 yers, 6 months PBTH UC ITT Impct Effect Size N Men (SD) N Men (SD) Impct (SE) Met developmentl milestones b (%) (45.8) (40.4) 2.5 (7.2) 0.05 Low birth weight c (%) (30.1) (22.9) 5.3 (12.3) 0.15 Ages 3 yers, 6 months to 7 yers Verbl bility d (0.84) (1.04) 0.19 (0.14) 0.14 Mth bility e (0.92) (0.98) 0.16 (0.14) 0.13 Executive functioning (self-regultion) f (16.11) (16.10) 2.45 (1.70) 0.11 Ages 8 to 17 yers Anxiety g (7.92) (7.05) 0.76 (1.00) 0.07 Fers h (14.90) (14.43) 1.70 (1.94) 0.08 Substnce use i (%) (31.99) (23.37) 1.57 (2.93) 0.04 Gol-oriented thinking j (5.19) (4.99) 0.61 (0.62) 0.09 School effort in pst month k (0.81) (0.78) 0.05 (0.09) 0.05 Arrests or police involvement in pst 6 months l (%) (31.80) (28.01) 2.34 (5.22) 0.06 PBTH = project-bsed trnsitionl housing. UC = usul cre. ITT = intention-to-tret. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire UC group. b Met developmentl milestones outcome is defined s scoring bove the typicl development cutoffs in ll domins of the Ages nd Stges Questionnire (ASQ-3). c Bse for low birthweight outcome (prent report) is children born since rndom ssignment who re t lest 1 yer old t followup. d Verbl bility outcome is the ntionlly stndrdized score from the Woodcock-Johnson III (WJ III) letter-word identifiction scle. e Mth bility outcome is the ntionlly stndrdized score from the WJ III pplied problems test. f Executive functioning outcome is the Hed Toes Knees Shoulders (HTKS) score nd rnges from 0 to 40, with higher scores indicting greter executive functioning. g Anxiety (child report) is mesured using the A-Trit scle from the Stte-Trit Anxiety Inventory for Children (STAIC). Scores rnge from 20 to 60, with higher scores indicting greter nxiety. h Fers outcome (child report) is the score from the Fers Scle nd rnges from 33 to 99, with higher scores indicting more fer. i Substnce use (child report) is mesured with 23 items from the Centers for Disese Control nd Prevention (CDC) 2011 Youth Risk Behvior Survey. j Gol-oriented thinking (child report) is mesured with modified version of the Children s Hope Scle nd rnges from 6 to 30, with higher scores indicting higher levels of positive, gol-oriented thinking. k School effort outcome (child report) rnges from 1 to 4, with higher scores indicting greter effort during school dy nd on homework. l Arrest or police involvement in pst 6 months is from prent report. Notes: Impct estimtes nd outcome mens re regression djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Sources: Fmily Options Study 18-month followup survey (prent report); Fmily Options Study 18-month child survey (child report); ASQ-3; WJ III; HTKS 8.7 Impcts on Self-Sufficiency in the Project-Bsed Trnsitionl Housing (PBTH) Versus Usul Cre (UC) Comprison Along with housing stbility nd dult well-being, selfsufficiency is lso centrl focus of PBTH progrms. Exhibit 8-16 shows impcts of PBTH reltive to UC on self-sufficiency outcomes. We found no sttisticlly significnt effects on ny of the 20 self-sufficiency outcomes. It ppers tht priority ccess to PBTH hd no effect on selfsufficiency reltive to the mix of services vilble in UC. 8.8 Summry of Project-Bsed Trnsitionl Housing (PBTH) Versus Usul Cre (UC) Comprison Across Domins For the PBTH-versus-UC comprison, 54 percent of fmilies ssigned to PBTH nd 29 percent of fmilies ssigned to UC ccessed trnsitionl housing during the followup period. This contrst in progrm use is smller thn for other comprisons of ctive interventions to UC; 121 however, it led to differences in experiences between PBTH nd UC fmilies in some res. Priority ccess to PBTH reduced the proportion 121 Although the tkeup rte for PBTH provides weker test of the intervention thn might be hoped for, the low tkeup of TH on the prt of mny fmilies ssigned to PBTH is of policy interest. It is not cler to wht extent this low tkeup represents fmilies declining progrms or progrms declining fmilies. Qulittive dt from smll number of fmilies in this study (80 in ll, 19 ssigned to PBTH) suggest tht both processes were importnt. When fmilies declined offers, the loction of progrms ws often n issue. Fmilies offered SUB nd CBRR hd more opportunity to live in neighborhoods of their choice, ner jobs, children s schools, nd support networks (Fisher et l., 2014). Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 103

136 Chpter 8. Impcts of Project-Bsed Trnsitionl Housing (PBTH) Compred With Usul Cre (UC) Exhibit PBTH Versus UC: Impcts on Self-Sufficiency Employment sttus Outcome PBTH UC ITT Impct Effect Size N Men (SD) N Men (SD) Impct (SE) Work for py in week before survey (%) (48.6) (47.0) 3.1 (4.1) 0.06 Any work for py since RA (%) (48.7) (48.9) 0.2 (3.9) Months worked for py since RA b (7.5) (8.0) 0.4 (0.6) 0.05 Hours of work per week t current min job c (17.3) (16.0) 1.2 (1.4) 0.07 Income sources nd mounts Annulized current ernings ($) 284 5,892 (9,756) 254 5,133 (9,131) 759 (822) 0.07 Totl fmily income ($) ,778 (9,356) 253 9,959 (7,812) 818 (729) 0.09 Anyone in fmily hd ernings in pst month (%) (50.1) (50.1) 1.3 (4.1) 0.02 Anyone in fmily received TANF in pst month (%) (45.6) (45.8) 1.0 (3.9) 0.02 Anyone in fmily received SSDI in pst month (%) (27.4) (28.9) 2.6 (2.1) 0.09 Anyone in fmily received SSI in pst month (%) (32.8) (37.2) 2.1 (2.4) 0.06 Anyone in fmily received SNAP/food stmps in pst month (%) (35.1) (37.8) 1.9 (3.2) 0.04 Anyone in fmily received WIC in pst month (%) (46.3) (45.1) 2.1 (3.9) 0.04 Eduction nd trining Prticipted in 2 weeks or more of ny school or trining since RA (%) (41.8) (43.0) 3.6 (3.9) 0.07 Number of weeks in school/trining progrms since RA (8.0) (10.2) 1.3 (0.8) 0.12 Prticipted in 2 weeks or more of school since RA (%) (24.0) (23.3) 0.6 (2.3) Prticipted in 2 weeks or more of bsic eduction since RA (%) (14.2) (12.3) 1.4 (1.3) 0.10 Prticipted in 2 weeks or more of voctionl eduction since RA (%) (22.7) (26.0) 3.6 (2.3) 0.13 Food security nd hunger Household is food insecure (%) (46.8) (47.5) 2.7 (4.2) 0.05 Food insecurity scle d (2.01) (1.98) 0.04 (0.18) 0.02 Economic stressors Economic stress scle e (0.48) (0.51) 0.05 (0.04) 0.09 PBTH = project-bsed trnsitionl housing. UC = usul cre. ITT = intention-to-tret. RA = rndom ssignment. SD = stndrd devition. SE = stndrd error. SNAP = Supplementl Nutrition Assistnce Progrm. SSDI = Socil Security Disbility Insurnce. SSI = Supplementl Security Income. TANF = Temporry Assistnce for Needy Fmilies. WIC = Specil Supplementl Nutrition Progrm for Women, Infnts, nd Children. */**/*** Impct estimte is significntly different from zero t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire UC group. b Number of months worked for py includes prtil clendr months. c Hours of work per week includes those not currently working (tht is, those with 0 hours of work per week). d Food insecurity scle rnges from 0 to 6, with higher vlues indicting higher food insecurity. e Economic stress scle rnges from 1 to 1, with higher vlues indicting higher economic stress. Notes: Impct estimtes nd outcome mens re regression djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Source: Fmily Options Study 18-month followup survey of fmilies who experienced t lest 1 night homeless or doubled up in the 12 months prior to the followup survey or t lest 1 night in shelter in the 6 months prior to the followup survey, from 52 percent for fmilies rndomized to UC, to 44 percent of fmilies ssigned to PBTH. In the four other domins, the vst mjority of indictors exmined revel equivlent results mong fmilies who were given priority ccess to PBTH nd fmilies who were ssigned to UC. The lck of impcts on dult well-being nd fmily self-sufficiency re noteworthy, given the emphsis plced by PBTH progrms on delivering help nd improvement in these domins. Only 2 of the 34 indictors exmined for results in this respect showed ny impct from PBTH. Priority ccess to PBTH lso did not provide better fmily preservtion or child well-being outcomes thn UC. Overll, the gols of PBTH s distinctive pproch to ssisting fmilies fcing unstble housing situtions were not chieved, even though degree of housing stbility ws delivered (presumbly during the period in which fmilies were stying in PBTH fcilities). One potentil reson for the lck of sttisticlly significnt effects in the PBTH-versus-UC comprison is tht services similr to those provided by PBTH were in mny cses vilble to fmilies in emergency shelter. Chpter 9 reports on how PBTH compres with the other two ctive interventions, SUB nd CBRR. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 104

137 CHAPTER 9. IMPACTS OF PERMANENT HOUSING SUBSIDY (SUB) COMPARED WITH COMMUNITY-BASED RAPID RE-HOUSING (CBRR), SUB COMPARED WITH PROJECT-BASED TRANSITIONAL HOUSING (PBTH), AND CBRR COMPARED WITH PBTH This chpter presents impct estimtes for the three pirwise comprisons tht do not involve usul cre (UC). These pirwise comprisons compre the ctive interventions of permnent housing subsidy (SUB), community-bsed rpid re-housing (CBRR), nd projectbsed trnsitionl housing (PBTH) with ech other. In ech of these comprisons, the gol is to determine whether the offer of n vilble plce in one ctive intervention leds to better, worse, or no different fmily outcomes thn those experienced by fmilies who were offered nother ctive intervention. The chpter begins with brief description of the nlysis smples for these comprisons. Next, it ddresses the SUB-versus-CBRR comprison, first showing the extent to which fmilies in both the SUB nd CBRR groups used housing nd service progrms vilble to them, nd then presenting effects on outcomes in the five study domins. The sme set of informtion is then presented in turn for the SUB-versus-PBTH nd CBRR-versus-PBTH comprisons. 9.1 Anlysis Smples for Pirwise Comprisons As ddressed previously in this report, fmilies in the study hd two, three, or four interventions vilble to them t the point of rndom ssignment. In this report, the term rndomiztion set denotes the set of interventions tht were vilble to fmily t rndom ssignment. For the impct nlysis, fmilies re included in pirwise comprisons of their ssigned intervention with ech of the other interventions in their rndomiztion set. This principle implies tht study fmily is included in one, two, or three pirwise com - prisons, depending on the size of its rndomiztion set. This principle lso implies tht the group of fmilies who represent given intervention differs somewht in different pirwise comprisons. For exmple, the group of SUB fmilies in the SUB-versus-CBRR comprison overlps with, but is not iden - ticl to, the group of SUB fmilies in the SUB-versus-PBTH comprison. This feture of the study is importnt to keep in mind when exmining impct estimtes in different pirwise comprisons, such s in this chpter. Exhibit 9-1 shows the number of fmilies who re included in the vrious pirwise comprisons tht comprise the entire study. Ech column of the exhibit shows the number of fmilies on both sides of prticulr comprison. Ech row shows how the number of fmilies representing prticulr intervention vries by pirwise comprison. If ll fmilies in the study hd ll four interventions vilble to them t the point of rndom ssignment, then the impct estimtes for lst three pirwise comprisons could be logiclly deduced from the three comprisons tht involve UC. 122 Becuse not ll fmilies hd rndomiztion sets with four interventions, it is possible for the impct estimtes in the lst three comprisons to devite from wht the first three 122 For exmple, the SUB-versus-CBRR impct on prticulr outcome would simply be the difference between the SUB-versus-UC nd CBRR-versus-UC impct estimtes. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 105

138 Chpter 9. Impcts of Permnent Housing Subsidy (SUB) Compred With Community-Bsed Rpid Re-Housing (CBRR), SUB Compred With Project-Bsed Trnsitionl Housing (PBTH), nd CBRR Compred With PBTH Exhibit 9-1. Smple Sizes in the Six Pirwise Comprisons Assigned Intervention Smple Size in Pirwise Comprison SUB Versus UC CBRR Versus UC PBTH Versus UC SUB Versus CBRR SUB Versus PBTH CBRR Versus PBTH SUB CBRR PBTH UC Totl CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. UC = usul cre. Smple sizes re numbers of fmilies who responded to the 18-month followup survey. Source: Fmily Options Study 18-month followup survey comprisons would suggest. As will be shown in this chpter, the signs of impct estimtes in the lst three comprisons re consistent with those suggested by the UC comprisons, but some mgnitudes of estimtes differ from the mgnitudes suggested by the comprisons with UC. The following sections ddress the results of the comprisons of the ctive interventions with ech other. 9.2 The Permnent Housing Subsidy (SUB) Versus Community-Bsed Rpid Re-Housing (CBRR) Comprison The SUB-versus-CBRR comprison contrsts the permnent housing subsidy of the SUB intervention with the temporry rentl ssistnce (usully lsting 7 to 8 months, but potentilly renewble for up to 18 months) of the CBRR intervention. Both these interventions provided priority ccess to the rentl subsidy, nd both required fmilies to engge in housing serch to locte suitble privte-mrket rentl unit. 123 The supportive services provided in the SUB intervention were limited to ssistnce with finding housing. The services offered by CBRR providers were lso focused on the housing serch. In ddition to cse mngement nd ssistnce with housing serch, CBRR progrms provided some level of self-sufficiency services for three-fourths of the fmilies ssigned to CBRR. One-fourth to one-third of the fmilies ssigned to CBRR hd ccess to broder rnge of services, including employment nd trining, life skills, nd physicl helth cre services. The subsidy in both of the interventions represented sub - stntil frction of monthly rent. Beyond the length of the subsidy, few differences in the dministrtion of the progrms re noteworthy. A stndrd formul set the subsidy mount in SUB, while subsidy determintion in CBRR vried mong providers, typiclly llowing for t lest some cse mnger discretion in setting the subsidy mount. To continue to receive CBRR ssistnce, fmilies hd to hve incomes below certin thresholds. Most CBRR progrms sked questions bout income every 3 months s prt of the recertifiction process to ssess continued need for ssistnce. This frequency ws much greter thn the nnul income recertifiction required by SUB providers Progrm Use by Fmilies in the SUB- Versus-CBRR Comprison Exhibit 9-2 shows the use of seven types of homeless nd housing progrms by the 381 SUB fmilies nd 308 CBRR fmilies nlyzed in the SUB-versus-CBRR comprison. 124 The first column shows the sme generl pttern of usge for these SUB fmilies s for ll SUB fmilies (see Exhibit 6-2 in Chpter 6). Likewise, the proportions of these CBRR fmilies shown in the second column re very similr to the proportions of ll CBRR fmilies (see Exhibit 7-4 in Chpter 7). The first two columns show tht 84 percent of fmilies ssigned to SUB used the SUB intervention nd 64 percent of CBRR fmilies used CBRR t some point during the period of obser - vtion. The numbers of months of pro grm use (in columns 3 through 6) nd the proportions using prticulr progrm in the month of followup survey response (in columns 7 through 8) re lso similr to those in the previous exhibits. A lrge difference exists in the proportion of fmilies prtici - pting in progrm during the month of the followup survey, with 82 percent of SUB fmilies prticipting in some progrm (most receiving SUB ssistnce) nd the mjority of CBRR fmilies (57 percent) not prticipting in ny progrm. 123 The miniml shre (bout 8 percent) of SUB fmilies provided with public housing in Honolulu, Hwii, or with project-bsed vouchers in Bridgeport, Connecticut, did not need to engge in housing serches. 124 The SUB-versus-CBRR comprison smple consists of 435 fmilies ssigned to SUB nd 382 fmilies ssigned to CBRR. Of these 817 fmilies, 381 SUB fmilies nd 308 CBRR fmilies (84 percent) responded to the 18-month followup survey. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 106

139 Chpter 9. Impcts of Permnent Housing Subsidy (SUB) Compred With Community-Bsed Rpid Re-Housing (CBRR), SUB Compred With Project-Bsed Trnsitionl Housing (PBTH), nd CBRR Compred With PBTH Exhibit 9-2. SUB Versus CBRR Progrm Use Since Rndom Assignment Type of Housing Assistnce Percent Ever Used From RA to 18-Month Followup Survey Number of Months Used From RA to 18-Month Followup Survey, if Ever Used Type of Housing Assistnce Percent Used in Month of Followup Survey Response SUB CBRR SUB CBRR SUB CBRR Men Medin Men Medin Subsidy (SUB) b Rpid re-housing (CBRR) Trnsitionl housing Permnent supportive housing Public housing Project-bsed vouchers/section 8 projects Emergency shelter c No use of homeless or housing progrms d N RA = rndom ssignment. Percentges of fmilies who ever used type of ssistnce progrm during the period from the month of RA to the month of 18-month Followup Survey response (medin period durtion: 21 months). Percentges do not dd to 100% becuse some fmilies use more thn one progrm type during the followup period. b Subsidy ssistnce is housing choice vouchers plus site-specific progrms offered to fmilies ssigned to SUB group in Bridgeport, Connecticut nd Honolulu, Hwii. c All fmilies were in emergency shelter t rndom ssignment. Percentges less thn 100 percent for ever used emergency shelter re due to missing dt on shelter use. d No use of homeless or housing progrms (ever used) indictes no use of the seven progrm types in this tble during ny of the followup period nd no use of emergency shelter fter the first 6 months fter RA. No use in the month of followup survey response indictes no use of ny of these seven progrm types. Notes: Percentges re regression-djusted, controlling for site nd rndomiztion rtio. Percentges, mens, nd, medins re weighted for survey nonresponse to represent full comprison smple. Source: Fmily Options Study Progrm Usge Dt Impcts of SUB Compred With CBRR As Chpter 3 ddressed, SUB proponents expect SUB to reduce homelessness nd improve housing stbility reltive to CBRR becuse they expect tht mny of the very poor fmilies who experience homelessness will need long-term rentl subsidies to remin stbly housed. The mgnitude of this expected difference hs been unknown before this study, however. Differentil effects on more distl outcomes re theorized to be dependent on the mgnitude of the hous - ing stbility effect. To the extent tht permnent subsidies provide greter residentil stbility or reduce prentl stress (stemming from moves or from fer of homelessness) more thn temporry subsidies, the benefits of SUB in other res such s child well-being nd fmily preservtion my be lrger thn effects of CBRR. Although CBRR ssistnce is temporry rther thn permnent, its emphsis on restoring fmilies to conventionl housing s swiftly s possible leds us to expect tht reltive to SUB, CBRR will reduce the length of the shelter sty t the time of study entry. The next section ddresses the experimentl impct evidence of how these interventions compre. Impcts on Housing Stbility in the SUB-Versus- CBRR Comprison Exhibit 9-3 shows the effect on housing stbility of being ssigned to SUB reltive to being ssigned to CBRR. The effects of SUB reltive to CBRR re fvorble, lrge, nd sttisticlly significnt on ll homelessness outcomes. The first row shows tht SUB reduced the confirmtory outcome of being homeless or doubled up in the pst 6 months or in emergency shelter in the pst 12 months from 50 to 22 percent reltive to CBRR. This represents 27-percentge-point reduction, mening tht ssignment to SUB resulted in reduc - tion of homelessness of more thn one-hlf when compred to the fmilies ssigned to CBRR. The next three rows show lrge reductions in the proportions of fmilies experiencing homelessness or doubling up in the pst 6 months. The fifth row shows tht SUB reduces the event of emergency shelter use during months 7 to 18 fter rndom ssignment by bout one-hlf (from 28 percent of fmilies to 15 percent). Row 6 of the exhibit shows tht SUB lso reduces the number of dys spent homeless or doubled up in the pst 6 months by bout 4 weeks reltive to CBRR. The second pnel of the exhibit shows tht SUB increses the proportion of fmilies living in their own house or prtment t followup (with or without ssistnce) from 64 to 72 percent reltive to CBRR. SUB fmilies re much less likely (34 percentge points) thn CBRR fmilies to be living in their own plce with no housing ssistnce nd much more likely (42 percentge points) to be living in their own plce with housing ssistnce. The third pnel of the exhibit shows tht SUB fmilies hd greter residentil stbility thn CBRR fmilies in the months before the survey, reducing the number of plces lived in the pst 6 months by 0.2 plces. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 107

140 Chpter 9. Impcts of Permnent Housing Subsidy (SUB) Compred With Community-Bsed Rpid Re-Housing (CBRR), SUB Compred With Project-Bsed Trnsitionl Housing (PBTH), nd CBRR Compred With PBTH Exhibit 9-3. SUB Versus CBRR: Impcts on Housing Stbility Outcome SUB CBRR ITT Impct Effect Size N Men (SD) N Men (SD) Impct (SE) Homeless or doubled up during the followup period At lest 1 night homeless b or doubled up in pst 6 months (41.7) (50.1) 27.3*** (3.8) 0.48 or in shelter in pst 12 months (%) [Confirmtory] c At lest 1 night homeless b or doubled up in pst 6 months (%) (38.4) (48.7) 20.9*** (3.6) 0.38 At lest 1 night homeless b in pst 6 months (%) (33.0) (43.9) 12.3*** (3.2) 0.25 At lest 1 night doubled up in pst 6 months (%) (34.4) (44.4) 15.4*** (3.2) 0.29 Any sty in emergency shelter in months 7 to 18 fter RA (%) (35.5) (45.8) 13.2*** (3.1) 0.26 Number of dys homeless b or doubled up in pst 6 months (53.5) (74.2) 30.7*** (5.4) 0.36 Number of dys homeless b in pst 6 months (37.2) (56.9) 14.9*** (4.1) 0.26 Number of dys doubled up in pst 6 months (38.9) (58.7) 20.0*** (4.3) 0.27 Housing independence Living in own house or prtment t followup (%) (45.1) (48.4) 7.9** (3.5) 0.14 Living in own house or prtment with no housing ssistnce (%) (29.7) (49.4) 33.7*** (3.4) 0.62 Living in own house or prtment with housing ssistnce (%) (48.6) (40.9) 41.9*** (3.4) 0.86 Number of plces lived Number of plces lived in pst 6 months (1.0) (1.0) 0.2*** (0.1) 0.17 Housing qulity Persons per room (0.8) (1.3) 0.3*** (0.1) 0.24 Housing qulity is poor or fir (%) (44.0) (47.1) 5.0 (3.7) 0.09 CBRR = community-bsed rpid re-housing. SUB = permnent housing subsidy. ITT = intention-to-tret. RA = rndom ssignment. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire usul cre group. b The definition of homeless in this report includes stys in emergency shelters nd plces not ment for humn hbittion. It excludes trnsitionl housing. c After djustment for multiple comprisons, the impct on the confirmtory outcome is sttisticlly significnt t the.01 level for the SUB versus-cbrr comprison. Notes: Impct estimtes nd outcome mens re regression djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Sources: Fmily Options Study 18-month followup survey; Progrm Usge Dt The bottom pnel of the exhibit shows tht SUB fmilies were living in less crowded conditions thn CBRR fmilies, with n verge of 1.2 persons per room compred with 1.5 persons per room for CBRR. No difference in housing qulity ws reported, however. Although CBRR proponents cknowledge the vlue of permnent housing ssistnce, they emphsize rpid re-housing s bility to shorten fmily stys in emergency shelter. Exhibit 9-4 presents sttisticl test of whether CBRR speeds exit from emergency shelter s compred with SUB. The first row of the exhibit shows tht the difference in the length of the initil shelter sty between the two groups (defined by to which progrm the fmily ws rndomly ssigned to receive priority ccess) is not sttisticlly significnt. Impcts on Fmily Preservtion in the SUB-Versus- CBRR Comprison As implied by Exhibit 3-1 in Chpter 3, ny differentil effects of SUB compred with CBRR of fmily preservtion Exhibit 9-4. Impcts on Length of Bseline Sty in Emergency Shelter Impct on Length (in months) of Bseline Sty First Assignment Group Second Assignment Group ITT impct Effect in Emergency Shelter for Comprison N Men (SD) N Men (SD) Impct (SE) Size b SUB versus CBRR (3.3) (4.8) 0.2 (0.3) 0.05 SUB versus PBTH (1.8) (3.5) 0.9*** (0.3) 0.19 CBRR versus PBTH (4.0) (3.8) 0.6 (0.5) 0.12 CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. ITT = intention-to-tret. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. The length of bseline sty in emergency shelter outcome includes one-hlf of the month of rndom ssignment nd is topcoded t 18 months. The 20 to 22 percent of fmilies in these comprisons whose bseline shelter sty does not pper in the Progrm Usge Dt re not included in the nlyses. b Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire usul cre group. Notes: Impct estimtes nd outcome mens re regression djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definition. Sources: Fmily Options Study 18-month followup survey; Progrm Usge Dt Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 108

141 Chpter 9. Impcts of Permnent Housing Subsidy (SUB) Compred With Community-Bsed Rpid Re-Housing (CBRR), SUB Compred With Project-Bsed Trnsitionl Housing (PBTH), nd CBRR Compred With PBTH would be expected to be indirect, vi the substntil differentil effects on housing stbility. Exhibit 9-5 shows the effects on fmily preservtion in the SUB-versus-CBRR comprison. One of five impct estimtes is sttisticlly significnt. Spouses or prtners were less likely to be seprted from ech other in the SUB s compred with the CBRR inter - vention, with nerly one-hlf of prtners present t bseline seprted t followup in CBRR compred with one-third in SUB. No evidence of other differences emerged regrding fmily seprtions or reunifictions. Impcts on Adult Well-Being in the SUB-Versus- CBRR Comprison As with fmily preservtion, ny differentil effects of SUB s compred with CBRR on dult well-being would be expected to be indirect, vi the substntil differentil effects on housing stbility. Exhibit 9-6 shows effects on dult well-being for this comprison. Two of the eight impct estimtes re sttisticlly significnt. SUB reduces the proportion of respon - dents who experienced post-trumtic stress disorder (PTSD) symptoms in the pst 30 dys from 25 percent for CBRR Exhibit 9-5. SUB Versus CBRR: Impcts on Fmily Preservtion Outcome Current or recent seprtions of fmily members present t bseline SUB CBRR ITT Impct Effect Size N Men (SD) N Men (SD) Impct (SE) Fmily hs t lest one child seprted in pst 6 months (%) (32.8) (33.8) 1.2 (2.7) 0.03 Fmily hs t lest one foster cre plcement in pst 6 months (%) (15.3) (16.0) 0.3 (1.4) Spouse/prtner seprted in pst 6 months, of those with spouse/prtner present t RA (%) (48.1) (50.1) 15.7** (7.9) 0.28 Reunifiction of fmily members reported s seprted t bseline Fmily hs t lest one child reunified, of those fmilies with t lest one child (46.6) (45.4) 2.4 (10.1) 0.05 bsent t RA (%) Spouse/prtner reunified, of those with spouse/prtner bsent t RA (%) (48.0) (44.1) 5.7 (14.8) 0.10 CBRR = community-bsed rpid re-housing. SUB = permnent housing subsidy. ITT = intention-to-tret. RA = rndom ssignment. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire usul cre group. Notes: Impct estimtes nd outcome mens re regression djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Source: Fmily Options Study 18-month followup survey Exhibit 9-6. SUB Versus CBRR: Impcts on Adult Well-Being Adult physicl helth Outcome SUB CBRR ITT Impct Effect Size N Men (SD) N Men (SD) Impct (SE) Helth in pst 30 dys ws poor or fir (%) (45.9) (44.8) 0.5 (3.4) Adult mentl helth Gol-oriented thinking b (1.07) (0.95) 0.10 (0.08) 0.08 Psychologicl distress c (5.54) (5.23) 0.49 (0.39) 0.07 Adult trum symptoms Post-trumtic stress disorder (PTSD) symptoms in pst 30 dys (%) (40.6) (42.2) 5.8* (3.2) 0.12 Adult substnce use Alcohol dependence or drug buse d (%) (32.1) (32.6) 0.4 (2.6) Alcohol dependence d (%) (28.6) (28.4) 0.2 (2.3) Drug buse d (%) (20.2) (20.1) 0.0 (1.6) Experience of intimte prtner violence Experienced intimte prtner violence in pst 6 months (%) (23.4) (31.9) 6.7*** (2.4) 0.18 CBRR = community-bsed rpid re-housing. SUB = permnent housing subsidy. ITT = intention-to-tret. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire usul cre group. b Gol-oriented thinking is mesured with modified version of the Stte Hope Scle nd rnges from 1 to 6, with higher scores indicting higher levels of positive, gol-oriented thinking. c Psychologicl distress is mesured with the Kessler 6 (K6) scle nd rnges from 0 to 24, with higher scores indicting greter distress. d Alcohol dependence is mesured with the Rpid Alcohol Problems Screen (RAPS4), nd drug buse is mesured with six items from the Drug Abuse Screening Test (DAST-10). Notes: Impct estimtes nd outcome mens re regression djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Source: Fmily Options Study 18-month followup survey. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 109

142 Chpter 9. Impcts of Permnent Housing Subsidy (SUB) Compred With Community-Bsed Rpid Re-Housing (CBRR), SUB Compred With Project-Bsed Trnsitionl Housing (PBTH), nd CBRR Compred With PBTH fmilies to 19 percent. SUB lso cut in hlf the proportion of fmily heds who experienced intimte prtner violence in the pst 6 months reltive to CBRR (from 13 to 6 percent). This effect my rise becuse the permnent subsidy llows for fmily heds to be more selective bout who lives in the household. 125 No effects on dult physicl helth, mentl helth, or substnce use were found. Impcts on Child Well-Being in the SUB-Versus- CBRR Comprison Exhibit 9-7 shows effects on child well-being outcomes mesured cross ll ges. Given the strong impct of SUB by comprison with CBRR on residentil stbility, it is not sur - prising tht SUB hs n effect on school mobility (pproximtely one fewer school move for every four children in fmilies ssigned to SUB). Only 1 other effect (for 2 out of 15) ppers in the cross-ge outcomes, lso in the school domin (first pnel of Exhibit 9-7) children in the SUB intervention hd slightly better grdes. Becuse this impct did not pper in the SUB-versus-UC comprison, it is perhps best thought of s rndom noise. No effects were found on child physicl helth or on behviorl strengths nd chllenges. Only 2 of 11 ge-specific outcomes shown in Exhibit 9-8 show effects, both for the older 8- to 17-yer-old ge group. Children in the SUB group were less thn one-hlf s likely to use substnces (4 versus 11 percent) nd reported greter gol-oriented thinking (effect size of 0.16) thn children in the UC group. Agin, these impcts were not evident in the Exhibit 9-7. SUB Versus CBRR: Impcts on Child Well-Being Across Age Groups Child eduction Outcome SUB CBRR ITT Impct Effect Size N Men (SD) N Men (SD) Impct (SE) Preschool or Hed Strt enrollment b (%) (48.8) (47.9) 2.3 (5.5) 0.04 School enrollment c (%) (29.3) (31.1) 2.2 (2.7) 0.07 Childcre or school bsences in pst month d (0.94) (0.93) 0.02 (0.09) 0.02 Number of schools ttended since RA e (0.75) (0.87) 0.25*** (0.08) 0.22 Grde completion (not held bck) (%) (24.1) (22.3) 1.7 (2.5) 0.04 Positive childcre or school experiences f (0.53) (0.53) 0.03 (0.05) 0.05 Positive childcre or school ttitudes g (1.00) (0.98) (0.08) School grdes h (0.91) (0.96) 0.17* (0.09) 0.14 Childcre or school conduct problems i (0.41) (0.41) 0.02 (0.03) 0.04 Child physicl helth Poor or fir helth (%) (22.0) (21.1) 0.3 (1.7) Well-child checkup in pst yer (%) (30.9) (28.5) 0.0 (2.7) Child hs regulr source of helth cre (%) (25.6) (25.4) 0.4 (2.2) Sleep problems j (1.12) (1.07) 0.09 (0.08) 0.06 Child behviorl strengths nd chllenges Behvior problems k (1.21) (1.27) 0.10 (0.11) 0.06 Prosocil behvior l (1.10) (1.12) 0.10 (0.10) 0.07 CBRR = community-bsed rpid re-housing. SUB = permnent housing subsidy. ITT = intention-to-tret. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire usul cre group. b Bse for preschool or Hed Strt enrollment is children ges 1 yer, 6 months to 5 yers. c Bse for school enrollment is children ges 6 to 17 yers. d Absences outcome is defined s 0 = no bsences in pst month, 1 = 1 to 2 bsences, 2 = 3 to 5 bsences, 3 = 6 or more bsences. e Number of schools outcome is topcoded t 4 or more schools. f Positive childcre or school experiences outcome is defined s 1 = mostly negtive experiences, 0 = both positive nd negtive experiences, 1 = mostly positive experiences. g Positive childcre or school ttitudes outcome is prent report of how much child likes school nd rnges from 1 to 5, with higher vlues indicting greter like of school. h School grdes outcome is defined s 1 = mostly Ds or Fs, 2 = mostly Cs, 3 = mostly Bs, 4 = mostly As. i Childcre or school conduct problems outcome is defined s 0 = no conduct problems reported to prent, 1 = prent contcted bout conduct problems or suspension or expulsion from school or childcre center. j Sleep problems outcome rnges from 1 to 5, with higher vlues indicting more frequent tiredness upon wking nd during the dy. k Behvior problems outcome is mesured s the stndrdized Totl Difficulties score from the Strengths nd Difficulties Questionnire (SDQ). l Prosocil behvior is mesured s the stndrdized Prosocil domin score from the SDQ. Notes: Impct estimtes nd outcome mens re regression djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Source: Fmily Options Study 18-month followup survey (prent report) 125 This explntion is consistent with qulittive evidence from the Effects of Housing Vouchers on Welfre Fmilies study, where voucher holders reported tht the subsidy enbled them to escpe unhelthy interctions with household members (Mills et l., 2006). Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 110

143 Chpter 9. Impcts of Permnent Housing Subsidy (SUB) Compred With Community-Bsed Rpid Re-Housing (CBRR), SUB Compred With Project-Bsed Trnsitionl Housing (PBTH), nd CBRR Compred With PBTH Exhibit 9-8. SUB Versus CBRR: Impcts on Child Well-Being Developmentl Outcomes by Age Group Outcome Ages 1 yer to 3 yers, 6 months SUB CBRR ITT Impct Effect Size N Men (SD) N Men (SD) Impct (SE) Met developmentl milestones b (%) (43.6) (39.6) 7.8 (6.1) 0.14 Low birth weight c (%) (39.0) (23.6) 19.0 (12.6) 0.52 Ages 3 yers, 6 months to 7 yers Verbl bility d (0.92) (0.95) 0.09 (0.12) 0.06 Mth bility e (0.94) (0.83) (0.13) Executive functioning (self-regultion) f (15.94) (15.79) 0.89 (1.39) 0.04 Ages 8 to 17 yers Anxiety g (7.65) (7.66) 1.42 (1.17) 0.13 Fers h (14.21) (14.26) 0.97 (2.06) 0.04 Substnce use i (%) (21.57) (31.25) 7.37** (3.18) 0.19 Gol-oriented thinking j (4.73) (4.27) 1.12* (0.61) 0.16 School effort in pst month k (0.80) (0.80) 0.15 (0.10) 0.14 Arrests or police involvement in pst 6 months l (%) (32.38) (32.47) 5.79 (6.19) 0.14 CBRR = community-bsed rpid re-housing. SUB = permnent housing subsidy. ITT = intention-to-tret. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire usul cre group. b Met developmentl milestones outcome is defined s scoring bove the typicl development cutoffs in ll domins of the Ages nd Stges Questionnire (ASQ-3). c Bse for low birthweight outcome (prent report) is children born since rndom ssignment who re t lest 1 yer old t followup. d Verbl bility outcome is the ntionlly stndrdized score from the Woodcock-Johnson III (WJ III) letter-word identifiction scle. e Mth bility outcome is the ntionlly stndrdized score from the WJ III pplied problems test. f Executive functioning outcome is the Hed Toes Knees Shoulders (HTKS) score nd rnges from 0 to 40, with higher scores indicting greter executive functioning. g Anxiety (child report) is mesured using the A-Trit scle from the Stte-Trit Anxiety Inventory for Children (STAIC). Scores rnge from 20 to 60, with higher scores indicting greter nxiety. h Fers outcome (child report) is the score from the Fers Scle nd rnges from 33 to 99, with higher scores indicting more fer. i Substnce use (child report) is mesured with 23 items from the Centers for Disese Control nd Prevention (CDC) 2011 Youth Risk Behvior Survey. j Gol-oriented thinking (child report) is mesured with modified version of the Children s Hope Scle nd rnges from 6 to 30, with higher scores indicting higher levels of positive, gol-oriented thinking. k School effort outcome (child report) rnges from 1 to 4, with higher scores indicting greter effort during school dy nd on homework. l Arrest or police involvement in pst 6 months is from prent report. Notes: Impct estimtes nd outcome mens re regression djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Sources: Fmily Options Study 18-month followup survey (prent report); Fmily Options Study 18-month child survey (child report); ASQ-3; WJ III; HTKS SUB-versus-UC comprison. Although the result for low birth weight ppers lrge, very few children were born in the 18 months fter rndom ssignment, so the numbers of children included in the test re quite smll, nd the result is not sttisticlly significnt. All 4 of the significnt differences out of 26 tests regrding child well-being fvored the SUB group, but only the impct on school mobility nd preschool or Hed Strt Progrm enrollment ws pprent in the SUB-versus-UC comprison. The others re, therefore, probbly best interpreted s rndom vrition. Impcts on Self-Sufficiency in the SUB-Versus- CBRR Comprison Exhibit 9-9 shows effects on self-sufficiency outcomes in the SUB-versus-CBRR comprison. Sttisticlly significnt effects were found on 5 of the 20 outcomes in the exhibit. Similr to the SUB-versus-UC results, SUB hd negtive effects on the proportion of fmily heds who worked since rndom ssignment (reduction of 9 percentge points) nd on the number of months worked for py since rndom ssignment (reduction of 2 months). The pprent reduction in lbor supply during the followup period is consistent with theory nd previous studies, s Section 6.7 ddressed. Although no significnt reduction occurred in the proportion of fmily heds working t followup, reduction in current ernings did occur, suggesting tht SUB ws still reducing lbor supply somewht t the point of followup. The tem found sttisticlly significnt effect on totl in - come, with SUB reducing nnul income in the yer prior to the survey by $978, but no effects on sources of income t the time of the survey. Likewise, the results indicte no effects on prticiption in eduction nd trining progrms or on food security. The impct estimte on the economic stress scle shows tht SUB reduced economic stress in the 6 months before the survey by stndrdized effect size of 0.18 compred with economic stress under CBRR. This result is not surprising, given tht the two groups hd similr csh income nd tht the housing subsidy represented n dditionl severl Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 111

144 Chpter 9. Impcts of Permnent Housing Subsidy (SUB) Compred With Community-Bsed Rpid Re-Housing (CBRR), SUB Compred With Project-Bsed Trnsitionl Housing (PBTH), nd CBRR Compred With PBTH Exhibit 9-9. SUB Versus CBRR: Impcts on Self-Sufficiency Employment sttus Outcome SUB CBRR ITT Impct Effect Size N Men (SD) N Men (SD) Impct (SE) Work for py in week before survey (%) (43.9) (46.9) 4.1 (3.5) 0.08 Any work for py since RA (%) (50.0) (48.2) 8.9** (3.6) 0.16 Months worked for py since RA b (7.0) (7.9) 2.1*** (0.6) 0.24 Hours of work per week t current min job c (14.4) (15.8) 1.8 (1.2) 0.10 Income sources nd mounts Annulized current ernings ($) 377 4,083 (8,471) 307 5,256 (9,386) 1,172* (682) 0.11 Totl fmily income ($) 369 8,400 (6,958) 294 9,378 (6,900) 978* (586) 0.11 Anyone in fmily hd ernings in pst month (%) (48.0) (49.5) 5.3 (3.7) 0.09 Anyone in fmily received TANF in pst month (%) (48.9) (47.7) 4.3 (3.6) 0.08 Anyone in fmily received SSDI in pst month (%) (26.6) (24.1) 1.3 (2.0) 0.04 Anyone in fmily received SSI in pst month (%) (33.8) (34.0) 1.1 (2.3) 0.03 Anyone in fmily received SNAP/food stmps in pst month (%) (31.7) (35.7) 3.5 (2.8) 0.08 Anyone in fmily received WIC in pst month (%) (47.2) (47.0) 2.5 (3.4) 0.05 Eduction nd trining Prticipted in 2 weeks or more of ny school or trining since RA (%) (44.1) (43.4) 0.8 (3.5) 0.02 Number of weeks in school/trining progrms since RA (10.5) (10.4) 0.3 (0.8) 0.03 Prticipted in 2 weeks or more of school since RA (%) (26.2) (21.6) 2.3 (1.9) 0.08 Prticipted in 2 weeks or more of bsic eduction since RA (%) (8.9) (14.9) 1.7 (1.0) 0.12 Prticipted in 2 weeks or more of voctionl eduction since RA (%) (25.3) (27.8) 2.3 (2.1) 0.08 Food security nd hunger Household is food insecure (%) (44.2) (45.6) 4.4 (3.6) 0.08 Food insecurity scle d (1.98) (1.99) 0.22 (0.16) 0.10 Economic stressors Economic stress scle e (0.48) (0.48) 0.10*** (0.04) 0.18 CBRR = community-bsed rpid re-housing. SUB = permnent housing subsidy. ITT = intention-to-tret. RA = rndom ssignment. SD = stndrd devition. SE = stndrd error. SNAP = Supplementl Nutrition Assistnce Progrm. SSDI = Socil Security Disbility Insurnce. SSI = Supplementl Security Income. TANF = Temporry Assistnce for Needy Fmilies. WIC = Specil Supplementl Nutrition Progrm for Women, Infnts, nd Children. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire usul cre group. b Number of months worked for py includes prtil clendr months. c Hours of work per week includes those not currently working (tht is, those with 0 hours of work per week). d Food insecurity scle rnges from 0 to 6, with higher vlues indicting higher food insecurity. e Economic stress scle rnges from 1 to 1, with higher vlues indicting higher economic stress. Notes: Impct estimtes nd outcome mens re regression djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Source: Fmily Options Study 18-month followup survey thou snd dollrs of resources vilble nnully to the 76 percent of SUB fmilies who still held the voucher t followup. Summry of SUB-Versus-CBRR Comprison Across Domins For the SUB-versus-CBRR comprison, the study engineered notble contrst in progrm use, with 84 percent of SUB fmilies receiving the SUB intervention nd 64 percent of CBRR fmilies receiving CBRR. This contrst in progrm pr - ticiption led to mrked differences between the experiences of SUB nd CBRR fmilies in severl res. The most noteworthy effect of priority ccess to SUB reltive to CBRR ws in its greter prevention of homelessness. Only 18 percent of SUB fmilies, compred with 39 percent of CBRR fmilies, spent t lest 1 night stying in shelter or plce not ment for humn hbittion or doubled up in the 6 months prior to followup survey completion. Also, only 15 percent of SUB fmilies, compred with 28 percent of CBRR fmilies, spent t lest 1 night in emergency shelter in months 7 to 18 fter study entry. On verge, SUB fmilies spent bout 4 weeks fewer thn CBRR fmilies being homeless or doubled up during the pst 6 months (22 compred with 52 dys). As compred with CBRR, SUB lso led to greter housing independence, incresing the proportion of fmilies living in their own house or prtment t followup from 64 to 72 per - cent nd reducing crowding. The greter stbility fforded by the SUB ssistnce ws evidenced in reduction in the number of plces lived in the pst 6 months nd reduction in the number of schools ttended by focl children since rndom ssignment. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 112

145 Chpter 9. Impcts of Permnent Housing Subsidy (SUB) Compred With Community-Bsed Rpid Re-Housing (CBRR), SUB Compred With Project-Bsed Trnsitionl Housing (PBTH), nd CBRR Compred With PBTH The SUB-versus-CBRR comprison does not yield strong pttern of effects in fmily preservtion, dult well-being, child well-being, or self-sufficiency. The multitude of sttisticl tests performed in these domins is likely to produce few sttisticlly significnt estimtes simply by chnce. The reserch tem finds more credible those significnt effects on outcomes tht lso hve significnt effects in the SUB-versus- UC comprison. Thus, few of the significnt impcts in domins other thn housing stbility re worth noting. In the dult well-being domin, SUB reduced the mount of intimte prtner violence experienced by fmily heds by hlf reltive to CBRR. In the self-sufficiency domin, SUB reduced the proportion of fmily heds who work during the followup period from 62 to 53 percent nd reduced the verge number of months worked by 2 months. On the other hnd, the dditionl resources represented by the SUB housing ssistnce served to reduce economic stress for fmilies in the pst 6 months of the followup period. Overll, do the fmilies ssigned to SUB pper to be doing better t this point thn the fmilies ssigned to CBRR? In short, yes. The SUB fmilies, on verge, hve certinly hd fewer negtive experiences thn the CBRR fmilies (reductions in homelessness, doubling up, nd intimte prtner violence). The SUB fmilies re somewht more likely to live in their own plce, the children move mong schools less, nd the fmily heds experience less economic stress. The wys in which SUB fmilies pper to be doing better thn CBRR fmilies, however, seem dependent on contemporneous receipt of the housing ssistnce. The short-term results provide no direct evidence of effects on outcomes tht might outlst the housing ssistnce (such s trining or work history outcomes for prents or school, helth, or developmentl outcomes for children). The next wve of the study will revel whether ny of these other effects emerge t the 3-yer followup point. 9.3 The Permnent Housing Subsidy (SUB) Versus Project-Bsed Trnsitionl Housing (PBTH) Comprison The SUB-versus-PBTH comprison contrsts the permnent housing subsidy of the SUB intervention with the temporry housing (up to 24 months with verge stys of bout 12 months during this followup period) in gency-controlled units pired with intensive supportive services of the PBTH intervention. As described in Chpter 3, these interventions represent distinct pproches to ddressing fmily homeless - ness. Proponents of PBTH emphsize tht most fmilies who become homeless hve dditionl brriers tht mke it hrd for them to secure nd mintin housing. Consistent with this perspective, PBTH progrms offer comprehensive cse mngement nd provide mny supportive services directly. These services re entirely bsent from the SUB intervention Progrm Use by Fmilies in the SUB- Versus-PBTH Comprison Exhibit 9-10 shows the use of seven types of homeless nd housing progrms by the 230 SUB fmilies nd 187 PBTH fmilies nlyzed in the SUB-versus-PBTH comprison. 126 The first column shows some modest differences in the generl pttern of usge for these SUB fmilies s compred with ll SUB fmilies (shown in Exhibit 6-2 in Chpter 6). 127 The proportions of these PBTH fmilies shown in the second column re very similr to the proportions of ll PBTH fm - ilies (shown in Exhibit 8-6 in Chpter 8). The first two columns show tht 83 percent of fmilies ssigned to SUB used the SUB intervention nd 52 percent of PBTH fmilies used trnsitionl housing. The numbers of months of progrm use (in columns 3 through 6) nd the proportions using the progrm in the month of followup survey response (in columns 7 through 8) re lrgely similr to those in the previous exhibits. As result, the proportion of fmilies prticipting in some progrm during the survey month exhibits lrge difference, with 79 percent of SUB fmilies doing so (most receiving SUB ssistnce) nd the mjority of PBTH fmilies (58 percent) not prticipting in ny progrm. 126 The SUB-versus-PBTH comprison smple consists of 256 fmilies ssigned to SUB nd 240 fmilies ssigned to PBTH. Of these 496 fmilies, 230 SUB fmilies nd 187 CBRR fmilies (84 percent) responded to the 18-month followup survey. 127 When compred with ll SUB fmilies, this subset of SUB fmilies is somewht less likely to ever use rpid re-housing (7 percent rther thn 13 percent). Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 113

146 Chpter 9. Impcts of Permnent Housing Subsidy (SUB) Compred With Community-Bsed Rpid Re-Housing (CBRR), SUB Compred With Project-Bsed Trnsitionl Housing (PBTH), nd CBRR Compred With PBTH Exhibit SUB Versus PBTH Progrm Use Since Rndom Assignment Type of Housing Assistnce Percent Ever Used From RA to 18-Month Followup Survey Number of Months Used From RA to 18-Month Followup Survey, if Ever Used Type of Housing Assistnce Percent Used in Month of Followup Survey Response SUB PBTH SUB PBTH SUB PBTH Men Medin Men Medin Subsidy (SUB) b Rpid re-housing (CBRR) Trnsitionl housing Permnent supportive housing Public housing Project-bsed vouchers/section 8 projects Emergency shelter c No use of homeless or housing progrms d N PBTH = project-bsed trnsitionl housing. RA = rndom ssignment. Percentge of fmilies who ever used type of ssistnce progrm during the period from the month of RA to the month of the 18-month followup survey response (medin period durtion: 21 months). Percentges do not dd to 100 becuse some fmilies used more thn one progrm type during the followup period. b Subsidy ssistnce is housing choice vouchers plus site-specific progrms offered to fmilies ssigned to the permnent housing subsidy (SUB) group in Bridgeport, Connecticut nd Honolulu, Hwii. c All fmilies were in emergency shelter t rndom ssignment. Percentges less thn 100 percent for ever used emergency shelter re becuse of missing dt on shelter use. d No use of homeless or housing progrms (ever used) indictes no use of the seven progrm types in this tble during ny of the followup period nd no use of emergency shelter fter the first 6 months fter RA. No use in the month of the followup survey response indictes no use of ny of these seven progrm types. Notes: Percentges re regression djusted, controlling for site nd rndomiztion rtio. Percentges, mens, nd, medins re weighted for survey nonresponse to represent full comprison smple. Source: Fmily Options Study Progrm Usge Dt Impcts of SUB Compred With PBTH As Chpter 3 ddressed, SUB nd PBTH proponents hve divergent views bout the pckge of housing ssistnce nd services tht homeless fmilies need. From the perspective of SUB proponents, SUB is expected to reduce homelessness nd improve housing stbility reltive to PBTH, nd by this mens it my improve fmily preservtion, dult wellbeing, nd child well-being. These expecttions re becuse SUB provides permnent housing ssistnce. On the other hnd, PBTH proponents expect tht PBTH will improve longterm housing stbility, employment, ernings, eduction, nd dult well-being reltive to SUB nd tht it my improve fmily preservtion nd child well-being. These expecttions of PBTH proponents re becuse PBTH ddresses brriers to housing stbility nd ttempts to put fmilies on trck for better employment nd ernings. The next section ddresses the experimentl short-term impct evidence on these divergent expecttions. Impcts on Housing Stbility in the SUB-Versus- PBTH Comprison Exhibit 9-11 shows the effect on housing stbility of being ssigned to SUB reltive to being ssigned to PBTH. The effects of SUB reltive to PBTH re fvorble, lrge, nd sttisticlly significnt on nerly ll homelessness outcomes. 128 SUB reduced the confirmtory outcome of being homeless or doubled up in the pst 6 months or in emergency shelter in the pst 12 months from 49 to 17 percent reltive to PBTH. This rte represents reduction of 31 percentge points, which is nerly two-thirds of the recent experience of homelessness for the PBTH fmilies. The next three rows show lrge reductions in the proportions of fmilies experiencing subsequent stys in shelter or plces not ment for humn hbittion nd doubling up in the pst 6 months. The pre - ventive effect of SUB on doubling up ws prticulrly strong; 34 percent of PBTH fmilies experienced doubling up in the 6 months before the survey compred with only 9 percent of SUB fmilies. SUB, reltive to PBTH, lso reduced the event of emergency shelter use during months 7 to 18 fter rndom ssignment by two-thirds (from 21 to 7 percent of fmilies) nd the number of dys spent homeless or doubled up in the pst 6 months by 5 weeks. The lst two rows of the first pnel show tht this reduction is lrgely becuse of reduction in the number of dys spent doubled up rther thn the number of dys spent homeless. The second pnel of the exhibit shows tht SUB increses the proportion of fmilies living in their own house or prtment t followup (with or without ssistnce) from 55 to 81 percent reltive to PBTH. This effect is due in prt to the 21 percent of PBTH fmilies who re still living in trnsitionl 128 The homeless outcomes in this study diverge from the homeless definition finl rule in tht they do not include stys in trnsitionl housing in their definitions of being homeless. Additionl impcts on the use of trnsitionl housing during the followup period re provided in Appendix E. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 114

147 Chpter 9. Impcts of Permnent Housing Subsidy (SUB) Compred With Community-Bsed Rpid Re-Housing (CBRR), SUB Compred With Project-Bsed Trnsitionl Housing (PBTH), nd CBRR Compred With PBTH Exhibit SUB Versus PBTH: Impcts on Housing Stbility Outcome Homeless or doubled up during the followup period SUB PBTH ITT Impct Effect Size N Men (SD) N Men (SD) Impct (SE) At lest 1 night homeless b or doubled up in pst 6 months (39.5) (50.0) 31.2*** (5.0) 0.55 or in shelter in pst 12 months (%) [Confirmtory] c At lest 1 night homeless b or doubled up in pst 6 months (%) (36.5) (48.5) 27.3*** (4.7) 0.49 At lest 1 night homeless b in pst 6 months (%) (30.1) (38.7) 13.3*** (3.8) 0.27 At lest 1 night doubled up in pst 6 months (%) (32.3) (46.6) 24.8*** (4.4) 0.47 Any sty in emergency shelter in months 7 to 18 fter RA (%) (25.5) (39.1) 13.9*** (3.6) 0.27 Number of dys homelessb b or doubled up in pst 6 months (51.1) (72.3) 34.9*** (7.0) 0.41 Number of dys homeless b in pst 6 months (35.5) (37.6) 6.1 (4.4) 0.11 Number of dys doubled up in pst 6 months (40.8) (68.6) 34.8*** (6.5) 0.47 Housing independence Living in own house or prtment t followup (%) (40.4) (49.5) 26.6*** (5.1) 0.47 Living in own house or prtment with no housing ssistnce (%) (31.7) (49.0) 22.9*** (4.7) 0.43 Living in own house or prtment with housing ssistnce (%) (46.6) (39.2) 49.5*** (4.5) 1.02 Number of plces lived Number of plces lived in pst 6 months (0.8) (1.1) 0.4*** (0.1) 0.27 Housing qulity Persons per room (0.7) (1.0) 0.4*** (0.1) 0.26 Housing qulity is poor or fir (%) (41.9) (49.0) 18.2*** (4.7) 0.34 PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. ITT = intention-to-tret. RA = rndom ssignment. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire usul cre group. b The definition of homeless in this report includes stys in emergency shelters nd plces not ment for humn hbittion. It excludes trnsitionl housing. c After djustment for multiple comprisons, the impct on the confirmtory outcome is sttisticlly significnt t the.01 level for the SUB-versus-PBTH comprison. Notes: Impct estimtes nd outcome mens re regression djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Sources: Fmily Options Study 18-month followup survey; Progrm Usge Dt housing t the time of the followup survey. SUB fmilies re much less likely (23 percentge points) thn PBTH fmilies to be living in their own plce with no housing ssistnce nd much more likely to be living in their own plce with housing ssistnce (50 percentge points). The third pnel of the exhibit shows tht SUB fmilies hd greter residentil stbility thn PBTH fmilies in the months before the survey. SUB reduced the number of plces lived in the pst 6 months by 0.4 plces. Some of this effect is likely becuse some PBTH fmilies moved out of trnsitionl housing t the end of their progrm prticiption. The bottom pnel of the exhibit shows tht SUB fmilies were living in less crowded conditions thn PBTH fmilies, with n verge of 1.3 persons per room compred with 1.7 persons per room for PBTH. SUB fmilies lso reported better housing qulity t the time of the survey thn did fmilies ssigned to PBTH. Impcts on Fmily Preservtion in the SUB-Versus- PBTH Comprison Exhibit 9-12 shows the effects on fmily preservtion in the SUB-versus-PBTH comprison. Results here re vried in direction. SUB reduced seprtions from children who were with the fmily t bseline by two-fifths (from 15 to 9 percent of fmilies) nd more thn doubled reunifictions with the smller number of children who were seprted from the fmily t bseline (from 17 to 45 percent) compred with the PBTH intervention. On the other hnd, the even smller number of spouses nd prtners seprted t bseline were more thn twice s likely to be reunited in the PBTH intervention (58.6 versus 21.3 percent). Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 115

148 Chpter 9. Impcts of Permnent Housing Subsidy (SUB) Compred With Community-Bsed Rpid Re-Housing (CBRR), SUB Compred With Project-Bsed Trnsitionl Housing (PBTH), nd CBRR Compred With PBTH Exhibit SUB Versus PBTH: Impcts on Fmily Preservtion Outcome Current or recent seprtions of fmily members present t bseline SUB PBTH ITT Impct Effect Size N Men (SD) N Men (SD) Impct (SE) Fmily hs t lest one child seprted in pst 6 months (%) (28.5) (34.9) 6.3* (3.5) 0.15 Fmily hs t lest one foster cre plcement in pst 6 months (%) (11.5) (16.3) 0.9 (1.2) 0.04 Spouse/prtner seprted in pst 6 months, of those with spouse/prtner present t RA (%) (47.5) (46.6) 5.1 (9.4) 0.09 Reunifiction of fmily members reported s seprted t bseline Fmily hs t lest one child reunified, of those fmilies with t lest one (48.9) (45.4) 27.7** (12.0) 0.55 child bsent t RA (%) Spouse/prtner reunified, of those with spouse/prtner bsent t RA (%) (47.0) (51.3) 37.3* (20.2) 0.65 PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. ITT = intention-to-tret. RA = rndom ssignment. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire usul cre group. Notes: Impct estimtes nd outcome mens re regression djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Source: Fmily Options Study 18-month followup survey Impcts on Adult Well-Being in the SUB-Versus-PBTH Comprison Exhibit 9-13 shows short-term effects on dult well-being outcomes for the SUB-versus-PBTH comprison. SUB hd sttisticlly significnt impcts on the two mentl helth outcomes reltive to PBTH, leding to n increse in positive, gol-oriented thinking (stndrdized effect size of 0.17) nd decrese in psychologicl distress (stndrdized effect size of 0.23). Reltive to PBTH, SUB hd no sttisticlly significnt effects on physicl helth, trum symptoms, substnce use, or experience of intimte prtner violence. Exhibit SUB Versus PBTH: Impcts on Adult Well-Being Adult physicl helth Outcome SUB PBTH ITT Impct Effect Size N Men (SD) N Men (SD) Impct (SE) Helth in pst 30 dys ws poor or fir (%) (46.6) (47.4) 5.3 (4.7) 0.10 Adult mentl helth Gol-oriented thinking b (1.05) (1.04) 0.20* (0.11) 0.17 Psychologicl distress c (5.15) (6.01) 1.51*** (0.54) 0.23 Adult trum symptoms Post-trumtic stress disorder (PTSD) symptoms in pst 30 dys (%) (41.7) (43.8) 5.1 (4.3) 0.10 Adult substnce use Alcohol dependence or drug buse d (%) (32.8) (38.7) 4.7 (4.1) 0.12 Alcohol dependence d (%) (28.2) (36.3) 5.8 (3.9) 0.16 Drug buse d (%) (24.0) (17.7) 3.1 (2.1) 0.12 Experience of intimte prtner violence Experienced intimte prtner violence in pst 6 months (%) (24.7) (28.8) 3.9 (3.1) 0.11 PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. ITT = intention-to-tret. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire usul cre group. b Gol-oriented thinking is mesured with modified version of the Stte Hope Scle nd rnges from 1 to 6, with higher scores indicting higher levels of positive, gol-oriented thinking. c Psychologicl distress is mesured with the Kessler 6 (K6) scle nd rnges from 0 to 24, with higher scores indicting greter distress. d Alcohol dependence is mesured with the Rpid Alcohol Problems Screen (RAPS4), nd drug buse is mesured with six items from the Drug Abuse Screening Test (DAST-10). Notes: Impct estimtes nd outcome mens re regression djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Source: Fmily Options Study 18-month followup survey Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 116

149 Chpter 9. Impcts of Permnent Housing Subsidy (SUB) Compred With Community-Bsed Rpid Re-Housing (CBRR), SUB Compred With Project-Bsed Trnsitionl Housing (PBTH), nd CBRR Compred With PBTH Impcts on Child Well-Being in the SUB-Versus- PBTH Comprison Only 3 of 15 cross-ge impcts (Exhibit 9-14) nd 0 of 11 ge-specific impcts (Exhibit 9-15) reched sttisticl significnce. The first pnel of Exhibit 9-14 shows tht, reltive to PBTH, SUB decresed school mobility but incresed the proportion of children with reported childcre or school conduct problems (by 9 percentge points). The third pnel of Exhibit 9-14 shows tht SUB lso reduced prent-rted prosocil behvior (stndrdized effect size of 0.16). Child well-being ws ssessed only for children who were with their prent t the time of the followup survey. To the extent tht out-of-home plcements either reflect or cuse problems in conduct, selection could ccount, in prt, for the ltter two impcts. Both the smll number of significnt effects (3 of 26) which could esily be by chnce nd the fct tht they do not consistently fvor one group suggest tht SUB hd little or no effect on child outcomes reltive to PBTH. Exhibit SUB Versus PBTH: Impcts on Child Well-Being Across Age Groups Child eduction Outcome SUB PBTH ITT Impct Effect Size N Men (SD) N Men (SD) Impct (SE) Preschool or Hed Strt enrollment b (%) (49.2) (49.2) 3.3 (7.7) 0.05 School enrollment c (%) (30.8) (28.8) 2.1 (3.2) 0.06 Childcre or school bsences in pst month d (0.92) (0.91) 0.12 (0.11) 0.10 Number of schools ttended since RA e (0.73) (0.82) 0.16* (0.09) 0.14 Grde completion (not held bck) (%) (23.1) (20.1) 1.1 (3.0) 0.03 Positive childcre or school experiences f (0.53) (0.53) 0.07 (0.06) 0.10 Positive childcre or school ttitudes g (1.04) (0.99) 0.11 (0.12) 0.09 School grdes h (0.93) (0.88) 0.05 (0.11) 0.04 Childcre or school conduct problems i (0.43) (0.39) 0.09* (0.05) 0.16 Child physicl helth Poor or fir helth (%) (23.6) (25.4) 1.2 (3.0) 0.04 Well-child checkup in pst yer (%) (32.2) (27.3) 3.2 (3.0) 0.08 Child hs regulr source of helth cre (%) (24.1) (20.4) 1.5 (2.0) 0.04 Sleep problems j (1.10) (1.12) 0.13 (0.11) 0.09 Child behviorl strengths nd chllenges Behvior problems k (1.17) (1.13) 0.14 (0.13) 0.09 Prosocil behvior l (1.06) 224 (1.07) 0.23* (0.12) 0.16 PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. ITT = intention-to-tret. RA = rndom ssignment. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire usul cre group. b Bse for preschool or Hed Strt enrollment is children ges 1 yer, 6 months to 5 yers. c Bse for school enrollment is children ges 6 to 17 yers. d Absences outcome is defined s 0 = no bsences in pst month, 1 = 1 to 2 bsences, 2 = 3 to 5 bsences, 3 = 6 or more bsences. e Number of schools outcome is topcoded t 4 or more schools. f Positive childcre or school experiences outcome is defined s 1 = mostly negtive experiences, 0 = both positive nd negtive experiences, 1 = mostly positive experiences. g Positive childcre or school ttitudes outcome is prent report of how much child likes school nd rnges from 1 to 5, with higher vlues indicting greter like of school. h School grdes outcome is defined s 1 = mostly Ds or Fs, 2 = mostly Cs, 3 = mostly Bs, 4 = mostly As. i Childcre or school conduct problems outcome is defined s 0 = no conduct problems reported to prent, 1 = prent contcted bout conduct problems or suspension or expulsion from school or childcre center. j Sleep problems outcome rnges from 1 to 5, with higher vlues indicting more frequent tiredness upon wking nd during the dy. k Behvior problems outcome is mesured s the stndrdized Totl Difficulties score from the Strengths nd Difficulties Questionnire (SDQ). l Prosocil behvior is mesured s the stndrdized Prosocil domin score from the SDQ. Notes: Impct estimtes nd outcome mens re regression djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Source: Fmily Options Study 18-month followup survey (prent report) Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 117

150 Chpter 9. Impcts of Permnent Housing Subsidy (SUB) Compred With Community-Bsed Rpid Re-Housing (CBRR), SUB Compred With Project-Bsed Trnsitionl Housing (PBTH), nd CBRR Compred With PBTH Exhibit SUB Versus PBTH: Impcts on Child Well-Being Developmentl Outcomes by Age Group Outcome Ages 1 yer to 3 yers, 6 months SUB PBTH ITT Impct Effect Size N Men (SD) N Men (SD) Impct (SE) Met developmentl milestones b (%) (46.5) (44.2) 2.2 (9.4) 0.04 Low birth weight c (%) (30.8) (36.3) 3.5 (9.3) 0.10 Ages 3 yers, 6 months to 7 yers Verbl bility d (1.00) (0.81) 0.16 (0.15) 0.12 Mth bility e (0.98) (0.93) 0.08 (0.15) 0.06 Executive functioning (self-regultion) f (16.16) (16.39) 1.71 (1.63) 0.07 Ages 8 to 17 yers Anxiety g (6.68) (7.35) 1.51 (1.37) 0.14 Fers h (13.73) (14.37) 0.43 (2.15) 0.02 Substnce use i (%) (32.24) (35.20) 2.60 (4.55) 0.07 Gol-oriented thinking j (4.77) (5.55) 0.52 (1.00) 0.08 School effort in pst month k (0.82) (0.78) 0.06 (0.16) 0.06 Arrests or police involvement in pst 6 months l (%) (27.15) (22.07) 2.11 (6.39) 0.05 PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. ITT = intention-to-tret. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire usul cre group. b Met developmentl milestones outcome is defined s scoring bove the typicl development cutoffs in ll domins of the Ages nd Stges Questionnire (ASQ-3). c Bse for low birthweight outcome (prent report) is children born since rndom ssignment who re t lest 1 yer old t followup. d Verbl bility outcome is the ntionlly stndrdized score from the Woodcock-Johnson III (WJ III) letter-word identifiction scle. e Mth bility outcome is the ntionlly stndrdized score from the WJ III pplied problems test. f Executive functioning outcome is the Hed Toes Knees Shoulders (HTKS) score nd rnges from 0 to 40, with higher scores indicting greter executive functioning. g Anxiety (child report) is mesured using the A-Trit scle from the Stte-Trit Anxiety Inventory for Children (STAIC). Scores rnge from 20 to 60, with higher scores indicting greter nxiety. h Fers outcome (child report) is the score from the Fers Scle nd rnges from 33 to 99, with higher scores indicting more fer. i Substnce use (child report) is mesured with 23 items from the Centers for Disese Control nd Prevention (CDC) 2011 Youth Risk Behvior Survey. j Gol-oriented thinking (child report) is mesured with modified version of the Children s Hope Scle nd rnges from 6 to 30, with higher scores indicting higher levels of positive, gol-oriented thinking. k School effort outcome (child report) rnges from 1 to 4, with higher scores indicting greter effort during school dy nd on homework. l Arrest or police involvement in pst 6 months is from prent report. Notes: Impct estimtes nd outcome mens re regression djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Sources: Fmily Options Study 18-month followup survey (prent report); Fmily Options Study 18-month child survey (child report); ASQ-3; WJ III; HTKS Impcts on Self-Sufficiency in the SUB-Versus- PBTH Comprison Exhibit 9-16 shows sttisticlly significnt effects on 11 of the 20 self-sufficiency outcomes in the SUB-versus-PBTH comprison. Similr to the results of other comprisons in - volving SUB, SUB hd negtive effects on mesures of work effort. Reltive to PBTH, SUB reduced the proportion of fmily heds who worked for py in the week before the survey (from 36 to 25 percent) nd who performed ny work for py since study entry (from 60 to 50 percent). The reduction in lbor supply during the followup period is consistent with theory nd previous studies, s Section 6.7 ddressed. The fourth row shows tht SUB lso hd negtive effect on hours of work t current min job. This effect, however, is entirely driven by the negtive effect on current employment shown in the first row rther thn representing n independent effect on work hours for those who were working. 129 The second pnel of the exhibit shows tht the verge nnul csh income reported by SUB fmilies (bout $9,000) ws lower thn tht reported by PBTH fmilies (bout $10,500). Two of the impcts on income sources lso reflect lower csh income for SUB fmilies. SUB fmilies were less likely to hve hd nyone with ernings (by 9 percentge points) or with Socil Security Disbility Insurnce (SSDI) benefits (by 4 percentge points) thn PBTH fmilies. The third pnel of the exhibit shows surprising result in its first row. A greter proportion of SUB fmilies prticipted in t lest 2 weeks of school or trining during the followup period (30 percent of SUB fmilies compred with 21 percent of PBTH fmilies). This result is surprising becuse most PBTH providers incorported some kind of employment trining into their progrms (lthough this trining my hve been shorter thn 2 weeks). 129 In fct, those SUB fmily heds who were working t followup worked n verge of 32.3 hours per week t their min job compred with 30.9 hours per week for working PBTH fmily heds. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 118

151 Chpter 9. Impcts of Permnent Housing Subsidy (SUB) Compred With Community-Bsed Rpid Re-Housing (CBRR), SUB Compred With Project-Bsed Trnsitionl Housing (PBTH), nd CBRR Compred With PBTH Exhibit SUB Versus PBTH: Impcts on Self-Sufficiency Employment sttus Outcome SUB PBTH ITT Impct Effect Size N Men (SD) N Men (SD) Impct (SE) Work for py in week before survey (%) (42.7) (48.2) 11.0** (4.5) 0.21 Any work for py since RA (%) (50.1) (49.0) 10.7** (4.7) 0.19 Months worked for py since RA b (6.8) (7.1) 0.8 (0.6) 0.10 Hours of work per week t current min job c (14.4) (17.0) 3.1** (1.5) 0.17 Income sources nd mounts Annulized current ernings ($) 226 4,112 (8,800) 182 5,859 (10,023) 1,747** (833) 0.17 Totl fmily income ($) 225 8,993 (6,983) ,483 (9,210) 1,490* (767) 0.17 Anyone in fmily hd ernings in pst month (%) (48.4) (50.1) 9.2* (4.7) 0.16 Anyone in fmily received TANF in pst month (%) (48.5) (47.4) 2.8 (4.5) 0.05 Anyone in fmily received SSDI in pst month (%) (21.4) (28.0) 4.1* (2.4) 0.14 Anyone in fmily received SSI in pst month (%) (32.8) (32.9) 0.3 (2.5) Anyone in fmily received SNAP/food stmps in pst month (%) (33.3) (33.5) 1.7 (3.3) 0.04 Anyone in fmily received WIC in pst month (%) (47.7) (47.0) 0.5 (4.2) Eduction nd trining Prticipted in 2 weeks or more of ny school or trining since RA (%) (46.2) (41.1) 8.5* (4.5) 0.17 Number of weeks in school/trining progrms since RA (11.2) (8.0) 1.5 (0.9) 0.14 Prticipted in 2 weeks or more of school since RA (%) (27.0) (23.6) 2.6 (2.5) 0.09 Prticipted in 2 weeks or more of bsic eduction since RA (%) (9.3) (10.3) 0.6 (1.0) 0.04 Prticipted in 2 weeks or more of voctionl eduction since RA (%) (27.6) (22.6) 1.4 (2.5) 0.05 Food security nd hunger Household is food insecure (%) (44.0) (45.4) 7.9* (4.5) 0.14 Food insecurity scle d (1.89) (2.00) 0.47** (0.20) 0.20 Economic stressors Economic stress scle e (0.45) (0.48) 0.15*** (0.04) 0.26 PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. ITT = intention-to-tret. RA = rndom ssignment. SD = stndrd devition. SE = stndrd error. SNAP = Supplementl Nutrition Assistnce Progrm. SSDI = Socil Security Disbility Insurnce. SSI = Supplementl Security Income. TANF = Temporry Assistnce for Needy Fmilies. WIC = Specil Supplementl Nutrition Progrm for Women, Infnts, nd Children. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire usul cre group. b Number of months worked for py includes prtil clendr months. c Hours of work per week includes those not currently working (tht is, those with 0 hours of work per week). d Food insecurity scle rnges from 0 to 6, with higher vlues indicting higher food insecurity. e Economic stress scle rnges from 1 to 1, with higher vlues indicting higher economic stress. Notes: Impct estimtes nd outcome mens re regression djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Source: Fmily Options Study 18-month followup survey The fourth pnel of the exhibit shows tht, reltive to PBTH, SUB reduced the proportion of fmilies who were food inse - cure from 32 to 24 percent nd reduced the verge level of food insecurity by stndrdized effect size of.20. Likewise, the finl row of the exhibit shows tht SUB reduced economic stress by stndrdized effect size of These reductions in food insecurity nd economic stress suggest tht the ddi - tionl severl thousnd dollrs of resources represented by SUB ech yer more thn offset the verge $1,500 reduction in csh income. Summry of SUB-Versus-PBTH Comprison Across Domins For the SUB-versus-PBTH comprison, the rndomiztion resulted in notble contrst in progrm use, with 83 percent of SUB fmilies receiving the SUB intervention nd 52 percent of PBTH fmilies receiving trnsitionl housing. This contrst in progrm prticiption led to mrked differences between the experiences of SUB nd PBTH fmilies in severl res. The most noteworthy effect of SUB reltive to PBTH ws in its greter prevention of homelessness. Only 13 percent of SUB fmilies, compred with 41 percent of PBTH fmilies, spent t lest 1 night in n emergency shelter, plce not ment for humn hbittion, or doubled up in the pst 6 months. Also, only 7 percent of SUB fmilies, compred with 21 percent of PBTH fmilies, spent t lest 1 night in emergency shelter in months 7 to 18 fter rndom ssignment. On verge, SUB fmilies spent 5 weeks fewer thn PBTH fmilies being homeless or doubled up during the pst 6 months (16 compred with 51 dys). Most of this effect ws reduction of the number of dys spent doubled up. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 119

152 Chpter 9. Impcts of Permnent Housing Subsidy (SUB) Compred With Community-Bsed Rpid Re-Housing (CBRR), SUB Compred With Project-Bsed Trnsitionl Housing (PBTH), nd CBRR Compred With PBTH As compred with PBTH, SUB lso led to greter housing independence, incresing the proportion of fmilies living in their own house or prtment t followup from 55 to 81 percent. Prt of this effect resulted from the fct tht 21 percent of PBTH fmilies were still living in trnsitionl housing (nd so not living in their own plces) t the point of followup. The greter stbility fforded by the SUB ssistnce ws evidenced in reduction in the number of plces lived in the pst 6 months nd reduction in the number of schools ttended by focl children since study entry. The SUB-versus-PBTH comprison does not yield strong pttern of effects in the fmily preservtion, dult well-being, or child well-being domins. Among these domins, the two most notble effects of SUB reltive to PBTH re 6-percentgepoint reduction in the proportion of fmilies with child seprtion in the pst 6 months (from 15 to 9 percent) nd decrese in the psychologicl distress reported by fmily heds (stndrdized effect size of 0.23). In the self-sufficiency domin, the tem found number of effects of SUB reltive to PBTH. SUB reduced the proportion of fmily heds who worked during the followup period from 60 to 50 percent nd reduced the proportion who worked t the followup point from 36 to 25 percent. Prtly s result of this lower work effort, SUB fmilies hd n verge nnul csh income of bout $1,500 less thn PBTH fmilies ($9,000 compred with $10,500). On the other hnd, the dditionl resources represented by the SUB housing ssistnce more thn offset the smller csh income nd served to reduce food insecurity nd economic stress for fmilies in the lst 6 months of the followup period (reltive to PBTH fmilies). Also, higher proportion of SUB fmily heds prticipted in school or trining during the followup period (30 percent compred with 21 percent for PBTH). Overll, do the fmilies ssigned to SUB pper to be doing better t this point thn the fmilies ssigned to PBTH? In mny wys, yes. The SUB fmilies hve mny fewer experiences of being homeless nd doubling up. They re much more likely thn PBTH fmilies to live in their own plce, re more food secure, hve children who move mong schools less, nd hve fmily heds who experience less psychologicl distress nd economic stress. As in the SUB-versus-CBRR comprison, however, the wys in which SUB fmilies pper to be doing better thn PBTH fmilies seem dependent on contemporneous receipt of the housing ssistnce. The results in the short term provide little direct evidence of effects on outcomes tht might outlst the housing ssistnce (such s work history outcomes for prents or school, helth, or developmentl outcomes for children). 9.4 The Community-Bsed Rpid Re-Housing (CBRR) Versus Project- Bsed Trnsitionl Housing (PBTH) Comprison The CBRR-versus-PBTH comprison contrsts the temporry privte-mrket rentl ssistnce nd modest cse mngement of the CBRR intervention with the temporry, gencycontrolled housing pired with intensive supportive services of the PBTH intervention. Although both interventions provide temporry ssistnce, the length of ssistnce differs. CBRR ssistnce is usully provided for 7 to 8 months nd is potentilly renewble for up to 18 months. PBTH provides housing for up to 24 months, with verge stys of 12 months during this followup period. As described in Chpter 3, these interventions represent distinct pproches to ddressing fmily homelessness. Proponents of PBTH emphsize tht most fmilies who become homeless hve dditionl brriers tht mke it hrd for them to secure nd mintin housing. Consistent with this perspective, PBTH progrms offer comprehensive cse mngement nd provide mny supportive services directly. Some services beyond housing serch were offered to bout three-fourths of CBRR fmilies, generlly with lower intensity thn in PBTH. Employment nd trining services were offered by nerly ll PBTH progrms (representing 92 percent of ssigned PBTH fmilies), but only by minority of CBRR progrms (representing 37 percent of ssigned CBRR fmilies). The CBRR cse mngers typiclly served bout 36 fmilies ech, roughly twice s mny fmilies s the typicl PBTH cse mnger Progrm Use by Fmilies in the CBRR- Versus-PBTH Comprison Exhibit 9-17 shows the use of seven types of homeless nd housing progrms by the 179 CBRR fmilies nd 197 PBTH fmilies nlyzed in the CBRR-versus-PBTH comprison. 130 The first column shows some differences in the generl pttern of usge for these CBRR fmilies s compred with ll CBRR fmilies (shown in Exhibit 7-4). Compred with the proportion mong ll fmilies rndomly ssigned to CBRR, lower proportion of the CBRR fmilies in this pirwise comprison used rpid re-housing (51 compred with 60 percent); higher proportion used trnsitionl housing 130 The CBRR-versus-PBTH comprison smple consists of 232 fmilies ssigned to CBRR nd 239 fmilies ssigned to PBTH. Of these 471 fmilies, 179 CBRR fmilies nd 197 PBTH fmilies (80 percent) responded to the 18-month followup survey. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 120

153 Chpter 9. Impcts of Permnent Housing Subsidy (SUB) Compred With Community-Bsed Rpid Re-Housing (CBRR), SUB Compred With Project-Bsed Trnsitionl Housing (PBTH), nd CBRR Compred With PBTH Exhibit CBRR Versus PBTH Progrm Use Since Rndom Assignment Type of Housing Assistnce Percent Ever Used From RA to 18-Month Followup Survey Number of Months Used From RA to 18-Month Followup Survey, if Ever Used Type of Housing Assistnce Percent Used in Month of Followup Survey Response CBRR PBTH CBRR PBTH CBRR PBTH Men Medin Men Medin Subsidy (SUB) b Rpid re-housing (CBRR) Trnsitionl housing Permnent supportive housing Public housing Project-bsed vouchers/section 8 projects Emergency shelter c No use of homeless or housing progrms d N CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. RA = rndom ssignment. Percentge of fmilies who ever used type of ssistnce progrm during the period from the month of RA to the month of the 18-month followup survey response (medin period durtion: 21 months). Percentges do not dd to 100 becuse some fmilies used more thn one progrm type during the followup period. b Subsidy ssistnce is housing choice vouchers plus site-specific progrms offered to fmilies ssigned to the permnent housing subsidy (SUB) group in Bridgeport, Connecticut nd Honolulu, Hwii. c All fmilies were in emergency shelter t rndom ssignment. Percentges less thn 100 percent for ever used emergency shelter re becuse of missing dt on shelter use. d No use of homeless or housing progrms (ever used) indictes no use of the seven progrm types in this tble during ny of the followup period nd no use of emergency shelter fter the first 6 months fter RA. No use in the month of the followup survey response indictes no use of ny of these seven progrm types. Notes: Percentges re regression djusted, controlling for site nd rndomiztion rtio. Percentges, mens, nd medins re weighted for survey nonresponse to represent full comprison smple. Source: Fmily Options Study Progrm Usge Dt (24 compred with 19 percent); nd higher proportion used none of these progrms during the followup period (19 compred with 12 percent). The lower tkeup rte of rpid re-housing mens tht, unfortuntely, this set of CBRR fmilies provides weker test of CBRR in the CBRR-versus- PBTH comprison thn the tests in the CBRR-versus-UC nd SUB-versus-CBRR comprisons. The proportions of PBTH fmilies shown in the second column re similr to the proportions of ll PBTH fmilies shown in Exhibit 8-6. The first two columns show tht 51 percent of fmilies ssigned to CBRR used rpid re-housing (wheres only 13 percent of PBTH fmilies did so) nd 55 percent of PBTH fmilies used trnsitionl housing (wheres only 24 percent of CBRR fmilies did so). The numbers of months of progrm use (in columns 3 through 6) nd the proportions using the progrm in the month of followup survey response (in columns 7 through 8) re lrgely similr to those in the previous exhibits. By the followup survey month, less thn one-hlf of both groups re prticipting in ny progrm. Of the CBRR fmilies, 35 percent re prticipting in some progrm in the survey month compred with 41 percent of the PBTH fmilies. 131 Only 5 percent of CBRR fmilies were still receiving rpid re-housing in the survey month compred with 22 percent of PBTH fmilies who were still receiving trnsitionl housing t this point Impcts of CBRR Compred With PBTH As discussed in Chpter 3, CBRR nd PBTH proponents hve divergent views bout the pckge of housing ssistnce nd services tht homeless fmilies need. From the perspective of CBRR proponents, CBRR is expected to reduce homelessness reltive to PBTH nd my improve housing stbility, fmily preservtion, dult well-being, child well-being, em - ployment, nd ernings. CBRR is lso expected to reduce the length of the shelter sty from the time of study entry. These expecttions reflect the fct tht CBRR provides rpidly delivered housing ssistnce towrd the gol of recipients quickly exiting shelter for community-bsed housing, thus trgeting wht CBRR proponents think should be the min objective of homeless interventions. On the other hnd, PBTH proponents expect tht PBTH will improve long-term housing stbility, employment, ernings, eduction, nd dult well-being reltive to CBRR nd my improve fmily preservtion nd child well-being. These expecttions of PBTH proponents re becuse PBTH ddresses brriers to housing stbility nd ttempts to put fmilies on trck for better employment nd ernings. The next section ddresses the experimentl short-term impct evidence on these divergent expecttions. 131 These proportions of fmilies prticipting in ny progrm re clculted by subtrcting from 100 percent the proportions with no use of progrms in the survey month (shown in the exhibit). Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 121

154 Chpter 9. Impcts of Permnent Housing Subsidy (SUB) Compred With Community-Bsed Rpid Re-Housing (CBRR), SUB Compred With Project-Bsed Trnsitionl Housing (PBTH), nd CBRR Compred With PBTH Impcts on Housing Stbility in the CBRR-Versus- PBTH Comprison Exhibit 9-18 shows the effect on housing stbility of being ssigned to CBRR reltive to being ssigned to PBTH. The first pnel of the exhibit shows some evidence tht ssignment to CBRR leds to greter use of emergency shelter nd stys in plces not ment for humn hbittion thn ssignment to PBTH. The third row shows tht 25 percent of CBRR fmilies experienced homelessness defined in this wy in the pst 6 months compred with only 15 percent of PBTH fmilies. CBRR lso incresed the number of dys in the pst 6 months spent homeless or doubled up from 40 to 61, or bout 3 weeks. These outcomes re both bsed on responses to the followup survey. Given tht trnsitionl housing is not included in the study s definition of homelessness, the fct tht 22 percent of PBTH fmilies were still using trnsitionl housing t the followup point when nerly ll CBRR fmilies hd finished use of CBRR by this time likely influenced these results. The outcome mesuring ny sty in emergency shelter in months 7 to 18 fter rndom ssignment is lrgely bsed on dministrtive Homeless Mngement Informtion System (HMIS) dt. This outcome does not show difference in emergency shelter usge between CBRR nd PBTH fmilies. The bottom three pnels of the exhibit show no sttisticlly significnt differences between CBRR nd PBTH fmilies in housing independence or number of residentil moves. Fmilies ssigned to CBRR were less likely to report poor or fir housing qulity thn were fmilies ssigned to PBTH (27 compred with 37 percent). Exhibit 9-4 presents sttisticl test of whether CBRR speeds exit from emergency shelter s compred with PBTH. The third row of the exhibit shows no evidence of such n effect. The difference in verge lengths of bseline shelter sty between the two groups is not sttisticlly significnt. Impcts on Fmily Preservtion in the CBRR-Versus- PBTH Comprison Exhibit 9-19 shows the effects on fmily preservtion in the CBRR-versus-PBTH comprison. We found no evidence of differentil effects of these interventions on fmily seprtions or reunifictions, lthough the numbers, prticulrly in the cse of reunifictions of fmily members seprted t the time of the bseline survey, were too smll to yield strong test. Exhibit CBRR Versus PBTH: Impcts on Housing Stbility Outcome Homeless or doubled up during the followup period CBRR PBTH ITT Impct Effect Size N Men (SD) N Men (SD) Impct (SE) At lest 1 night homeless b or doubled up in pst 6 months (50.1) (49.7) 7.5 (5.7) 0.13 or in shelter in pst 12 months (%) [Confirmtory] c At lest 1 night homeless b or doubled up in pst 6 months (%) (49.6) (48.3) 9.1 (5.6) 0.16 At lest 1 night homeless b in pst 6 months (%) (43.5) (37.4) 9.7** (4.8) 0.20 At lest 1 night doubled up in pst 6 months (%) (47.5) (47.3) 3.4 (5.5) 0.06 Any sty in emergency shelter in months 7 to 18 fter RA (%) (41.8) (38.7) 1.4 (4.4) 0.03 Number of dys homeless b or doubled up in pst 6 months (77.8) (68.3) 21.2** (8.8) 0.25 Number of dys homeless b in pst 6 months (52.9) (41.8) 12.7** (6.0) 0.22 Number of dys doubled up in pst 6 months (69.4) (62.5) 10.0 (8.1) 0.14 Housing independence Living in own house or prtment t followup (%) (49.1) (49.6) 3.0 (6.1) 0.05 Living in own house or prtment with no housing ssistnce (%) (49.5) (49.7) 1.5 (5.9) 0.03 Living in own house or prtment with housing ssistnce (%) (38.9) (34.3) 4.5 (4.1) 0.09 Number of plces lived Number of plces lived in pst 6 months (1.0) (1.1) 0.0 (0.1) 0.02 Housing qulity Persons per room (1.4) (1.3) 0.0 (0.1) 0.03 Housing qulity is poor or fir (%) (43.9) (48.5) 9.7* (5.3) 0.18 CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. ITT = intention-to-tret. RA = rndom ssignment. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire usul cre group. b The definition of homeless in this report includes stys in emergency shelters nd plces not ment for humn hbittion. It excludes trnsitionl housing. c After djustment for multiple comprisons, the impct on the confirmtory outcome is not sttisticlly significnt t the.10 level for the CBRR-versus-PBTH comprison. Notes: Impct estimtes nd outcome mens re regression djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Sources: Fmily Options Study 18-month followup survey; Progrm Usge Dt Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 122

155 Chpter 9. Impcts of Permnent Housing Subsidy (SUB) Compred With Community-Bsed Rpid Re-Housing (CBRR), SUB Compred With Project-Bsed Trnsitionl Housing (PBTH), nd CBRR Compred With PBTH Exhibit CBRR Versus PBTH: Impcts on Fmily Preservtion Outcome Current or recent seprtions of fmily members present t bseline CBRR PBTH ITT Impct Effect Size N Men (SD) N Men (SD) Impct (SE) Fmily hs t lest one child seprted in pst 6 months (%) (34.3) (34.7) 0.7 (3.6) 0.02 Fmily hs t lest one foster cre plcement in pst 6 months (%) (18.1) (14.2) 1.6 (1.9) 0.07 Spouse/prtner seprted in pst 6 months, of those with spouse/prtner present t RA (%) (48.4) (44.2) 7.1 (8.7) 0.13 Reunifiction of fmily members reported s seprted t bseline Fmily hs t lest one child reunified, of those fmilies with t lest one child (48.7) (48.6) 6.7 (12.2) 0.13 bsent t RA (%) Spouse/prtner reunified, of those with spouse/prtner bsent t RA (%) (47.9) (50.7) 12.9 (23.7) 0.23 CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. ITT = intention-to-tret. RA = rndom ssignment. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire usul cre group. Notes: Impct estimtes nd outcome mens re regression djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Source: Fmily Options Study 18-month followup survey Impcts on Adult Well-Being in the CBRR-Versus- PBTH Comprison Exhibit 9-20 shows sttisticlly significnt effects on 3 of the 8 dult well-being outcomes in the CBRR-versus-PBTH comprison. All three of these significnt effects hve CBRR producing more fvorble outcomes thn PBTH, which is surprising becuse PBTH progrms hve n explicit focus on dult well-being. The first row shows tht CBRR reduces the proportion of fmily heds with poor or fir physicl helth in the pst 30 dys by 11 percentge points (from 34 to 23 percent). The second pnel of the exhibit shows tht CBRR reduces the verge level of psychologicl distress by stndrdized effect size of 0.28 reltive to PBTH. The third sttisticlly significnt effect is shown in the fourth pnel of the exhibit, with CBRR reducing the proportion of fmily heds reporting lcohol dependence or drug buse from 16 to 10 percent. Exhibit CBRR Versus PBTH: Impcts on Adult Well-Being Adult physicl helth Outcome CBRR PBTH ITT Impct Effect Size N Men (SD) N Men (SD) Impct (SE) Helth in pst 30 dys ws poor or fir (%) (43.6) (46.8) 11.3** (4.8) 0.21 Adult mentl helth Gol-oriented thinking b (0.92) (1.03) 0.16 (0.11) 0.13 Psychologicl distress c (5.04) (5.87) 1.84*** (0.58) 0.28 Adult trum symptoms Post-trumtic stress disorder (PTSD) symptoms in pst 30 dys (%) (41.0) (42.4) 3.3 (4.6) 0.07 Adult substnce use Alcohol dependence or drug buse d (%) (32.3) (36.0) 6.8* (3.9) 0.17 Alcohol dependence d (%) (27.8) (31.6) 5.3 (3.5) 0.14 Drug buse d (%) (20.7) (22.0) 1.2 (2.1) 0.04 Experience of intimte prtner violence Experienced intimte prtner violence in pst 6 months (%) (29.4) (28.9) 1.1 (3.2) 0.03 CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. ITT = intention-to-tret. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire usul cre group. b Gol-oriented thinking is mesured with modified version of the Stte Hope Scle nd rnges from 1 to 6, with higher scores indicting higher levels of positive, gol-oriented thinking. c Psychologicl distress is mesured with the Kessler 6 (K6) scle nd rnges from 0 to 24, with higher scores indicting greter distress. d Alcohol dependence is mesured with the Rpid Alcohol Problems Screen (RAPS4), nd drug buse is mesured with six items from the Drug Abuse Screening Test (DAST-10). Notes: Impct estimtes nd outcome mens re regression djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Source: Fmily Options Study 18-month followup survey Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 123

156 Chpter 9. Impcts of Permnent Housing Subsidy (SUB) Compred With Community-Bsed Rpid Re-Housing (CBRR), SUB Compred With Project-Bsed Trnsitionl Housing (PBTH), nd CBRR Compred With PBTH The effect on psychologicl distress ppers consistent with results found in other pirwise comprisons. In prticulr, PBTH fmilies on verge report more psychologicl distress thn either CBRR or SUB fmilies. On the other hnd, the study tem interprets the effects on physicl helth nd sub - stnce use with some cution. The men CBRR vlues for these outcomes pper somewht more fvorble thn the men vlues for the entire CBRR group. In prllel mnner, the men PBTH vlues for these outcomes in this comprison smple pper somewht less fvorble thn the men vlues for the entire PBTH group. We found no suggestion in other pirwise comprisons tht CBRR fmilies hve prticulrly good physicl helth or low substnce use, nor tht PBTH fmilies hve prticulrly poor physicl helth or high substnce use. No effects of CBRR reltive to PBTH re observed on either the proportion of fmily heds with PTSD symptoms or the proportion of fmily heds experiencing intimte prtner violence. Impcts on Child Well-Being in the CBRR-Versus- PBTH Comprison As Chpter 3 ddressed, differentil effects of CBRR nd PBTH on child well-being would be expected to be indirect, vi effects on housing stbility, self-sufficiency, nd dult well-being, which were modest nd in different directions. Only 2 of 15 cross-ge outcomes (Exhibit 9-21) nd 2 of 11 ge-specific outcomes (Exhibit 9-22) ppered to reflect progrm impct. Of these outcomes, 1 cross-ge effect fvored PBTH, for which school enrollment ws 6.6 percentge points higher, nd the other fvored CBRR, for which preschool or Hed Strt Progrm enrollment ws 16.8 percentge points higher. The 2 ge-specific effects fvored CBRR; 15 percent more of children ges 1 yer to 3 yers, 6 months met ll developmentl milestones on the Ages nd Stges Questionnire (ASQ-3; pnel 1 of Exhibit 9-22), nd children ges 8 to 17 yers reported slightly lower levels of trit nxiety (pnel 3). Results for the developmentl screening test (effect Exhibit CBRR Versus PBTH: Impcts on Child Well-Being Across Age Groups Child eduction Outcome CBRR PBTH ITT Impct Effect Size N Men (SD) N Men (SD) Impct (SE) Preschool or Hed Strt enrollment b (%) (49.4) (48.0) 16.8** (7.8) 0.28 School enrollment c (%) (34.1) (27.4) 6.6** (3.1) 0.20 Childcre or school bsences in pst month d (0.90) (0.94) 0.14 (0.11) 0.11 Number of schools ttended since RA e (0.94) (0.85) (0.11) Grde completion (not held bck) (%) (24.6) (28.2) 2.8 (3.3) 0.07 Positive childcre or school experiences f (0.56) (0.57) 0.03 (0.06) 0.04 Positive childcre or school ttitudes g (1.00) (1.02) 0.11 (0.10) 0.09 School grdes h (0.99) (0.84) 0.09 (0.12) 0.07 Childcre or school conduct problems i (0.44) (0.43) 0.03 (0.04) 0.05 Child physicl helth Poor or fir helth (%) (17.4) (21.2) 4.4 (2.7) 0.16 Well-child checkup in pst yer (%) (29.2) (28.2) 0.6 (3.0) Child hs regulr source of helth cre (%) (21.1) (21.9) 1.2 (2.1) 0.04 Sleep problems j (1.04) (1.07) 0.02 (0.11) Child behviorl strengths nd chllenges Behvior problems k (1.27) (1.16) 0.02 (0.13) Prosocil behvior l (1.11) 249 (1.13) 0.09 (0.12) 0.06 CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. ITT = intention-to-tret. RA = rndom ssignment. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire usul cre group. b Bse for preschool or Hed Strt enrollment is children ges 1 yer, 6 months to 5 yers. C Bse for school enrollment is children ges 6 to 17 yers. d Absences outcome is defined s 0 = no bsences in pst month, 1 = 1 to 2 bsences, 2 = 3 to 5 bsences, 3 = 6 or more bsences. e Number of schools outcome is topcoded t 4 or more schools. f Positive childcre or school experiences outcome is defined s 1 = mostly negtive experiences, 0 = both positive nd negtive experiences, 1 = mostly positive experiences. g Positive childcre or school ttitudes outcome is prent report of how much child likes school nd rnges from 1 to 5, with higher vlues indicting greter like of school. h School grdes outcome is defined s 1 = mostly Ds or Fs, 2 = mostly Cs, 3 = mostly Bs, 4 = mostly As. i Childcre or school conduct problems outcome is defined s 0 = no conduct problems reported to prent, 1 = prent contcted bout conduct problems or suspension or expulsion from school or childcre center. j Sleep problems outcome rnges from 1 to 5, with higher vlues indicting more frequent tiredness upon wking nd during the dy. k Behvior problems outcome is mesured s the stndrdized Totl Difficulties score from the Strengths nd Difficulties Questionnire (SDQ). l Prosocil behvior is mesured s the stndrdized Prosocil domin score from the SDQ. Notes: Impct estimtes nd outcome mens re regression djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Source: Fmily Options Study 18-month followup survey (prent report) Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 124

157 Chpter 9. Impcts of Permnent Housing Subsidy (SUB) Compred With Community-Bsed Rpid Re-Housing (CBRR), SUB Compred With Project-Bsed Trnsitionl Housing (PBTH), nd CBRR Compred With PBTH Exhibit CBRR Versus PBTH: Impcts on Child Well-Being Developmentl Outcomes by Age Group Outcome Ages 1 yer to 3 yers, 6 months CBRR PBTH ITT Impct Effect Size N Men (SD) N Men (SD) Impct (SE) Met developmentl milestones b (%) (39.8) (44.4) 15.3** (7.4) 0.28 Low birth weight c (%) (28.9) (25.8) 1.9 (8.3) 0.05 Ages 3 yers, 6 months to 7 yers Verbl bility d (1.06) (0.88) 0.04 (0.21) 0.03 Mth bility e (0.87) (0.92) 0.19 (0.16) 0.16 Executive functioning (self-regultion) f (16.27) (15.83) 1.07 (1.76) 0.05 Ages 8 to 17 yers Anxiety g (7.00) (8.06) 2.61* (1.41) 0.25 Fers h (13.53) (14.77) 1.88 (2.24) 0.09 Substnce use i (%) (33.10) (32.24) 1.53 (4.61) 0.04 Gol-oriented thinking j (4.19) (4.97) 1.12 (0.68) 0.16 School effort in pst month k (0.82) (0.82) 0.18 (0.13) 0.17 Arrests or police involvement in pst 6 months l (%) (23.76) (34.28) 8.57 (6.09) 0.21 CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. ITT = intention-to-tret. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire usul cre group. b Met developmentl milestones outcome is defined s scoring bove the typicl development cutoffs in ll domins of the Ages nd Stges Questionnire (ASQ-3). c Bse for low birthweight outcome (prent report) is children born since rndom ssignment who re t lest 1 yer old t followup. d Verbl bility outcome is the ntionlly stndrdized score from the Woodcock-Johnson III (WJ III) letter-word identifiction scle. e Mth bility outcome is the ntionlly stndrdized score from the WJ III pplied problems test. f Executive functioning outcome is the Hed Toes Knees Shoulders (HTKS) score nd rnges from 0 to 40, with higher scores indicting greter executive functioning. g Anxiety (child report) is mesured using the A-Trit scle from the Stte-Trit Anxiety Inventory for Children (STAIC). Scores rnge from 20 to 60, with higher scores indicting greter nxiety. h Fers outcome (child report) is the score from the Fers Scle nd rnges from 33 to 99, with higher scores indicting more fer. i Substnce use (child report) is mesured with 23 items from the Centers for Disese Control nd Prevention (CDC) 2011 Youth Risk Behvior Survey. j Gol-oriented thinking (child report) is mesured with modified version of the Children s Hope Scle nd rnges from 6 to 30, with higher scores indicting higher levels of positive, gol-oriented thinking. k School effort outcome (child report) rnges from 1 to 4, with higher scores indicting greter effort during school dy nd on homework. l Arrest or police involvement in pst 6 months is from prent report. Notes: Impct estimtes nd outcome mens re regression djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Sources: Fmily Options Study 18-month followup survey (prent report); Fmily Options Study 18-month child survey (child report); ASQ-3; WJ III; HTKS size of 0.28) re lrge enough to be importnt, but without more consistent evidence of effects they re probbly best interpreted s rndom vrition. Impcts on Self-Sufficiency in the CBRR-Versus- PBTH Comprison Exhibit 9-23 shows sttisticlly significnt effects on 4 of the 20 self-sufficiency outcomes in the CBRR-versus-PBTH comprison. No sttisticlly significnt differences emerged in the employment outcomes of CBRR nd PBTH fmily heds or in verge fmily income or income sources. The third pnel contins some puzzling results. CBRR incresed the proportion of fmily heds who prticipted in ny type of school or trining nd the number of weeks in school or trining since rndom ssignment reltive to PBTH. CBRR lso decresed the proportion of fmily heds who prticipted in bsic eduction nd voctionl eduction, however. The contrdictory evidence leds us to believe these results re by chnce, unlikely to be replicted in other studies. The tem therefore concludes tht CBRR hd little or no effect on self-sufficiency outcomes reltive to PBTH. Summry of CBRR-Versus-PBTH Comprison Across Domins For number of resons, the CBRR-versus-PBTH comprison offers weker test thn the other pirwise comprisons in the study. The number of fmilies in this comprison smple is the lowest of the pirwise comprisons nd so provides less sttisticl power thn the other tests. 132 In ddition, tkeup rtes for the ssigned interventions 55 percent for PBTH fmilies nd 51 percent for CBRR fmilies re lower thn for other comprisons. 132 The smller comprison smple is in lrge prt becuse of the greter selectivity of PBTH progrms, leding to the bsence of PBTH from the rndomiztion sets of 356 fmilies. See Gubits et l. (2013), Exhibit 3-5, for more informtion on the reltive selectivity of PBTH, CBRR, nd SUB progrms. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 125

158 Chpter 9. Impcts of Permnent Housing Subsidy (SUB) Compred With Community-Bsed Rpid Re-Housing (CBRR), SUB Compred With Project-Bsed Trnsitionl Housing (PBTH), nd CBRR Compred With PBTH Exhibit CBRR Versus PBTH: Impcts on Self-Sufficiency Employment sttus Outcome CBRR PBTH ITT Impct Effect Size N Men (SD) N Men (SD) Impct (SE) Work for py in week before survey (%) (46.5) (48.8) 6.8 (5.3) 0.13 Any work for py since RA (%) (48.7) (48.7) 2.9 (5.0) 0.05 Months worked for py since RA b (8.3) (7.8) 1.2 (0.8) 0.14 Hours of work per week t current min job c (16.9) (17.6) 1.4 (1.9) 0.08 Income sources nd mounts Annulized current ernings ($) 174 5,181 (8,720) 191 6,603 (9,466) 1,423 (1,000) 0.14 Totl fmily income ($) ,784 (8,527) ,801 (8,912) 18 (922) Anyone in fmily hd ernings in pst month (%) (50.1) (50.1) 1.2 (5.1) 0.02 Anyone in fmily received TANF in pst month (%) (45.5) (44.0) 7.5 (5.1) 0.14 Anyone in fmily received SSDI in pst month (%) (27.8) (29.6) 0.5 (2.9) 0.02 Anyone in fmily received SSI in pst month (%) (34.8) (33.9) 3.4 (3.2) 0.09 Anyone in fmily received SNAP/Food Stmps in pst month (%) (30.9) (36.5) 4.1 (3.9) 0.10 Anyone in fmily received WIC in pst month (%) (47.7) (45.2) 4.0 (4.8) 0.08 Eduction nd trining Prticipted in 2 weeks or more of ny school or trining since RA (%) (45.1) (42.1) 9.1* (5.0) 0.18 Number of weeks in school/trining progrms since RA (10.7) (7.1) 1.9* (1.1) 0.17 Prticipted in 2 weeks or more of school since RA (%) (23.1) (23.0) 0.2 (2.7) Prticipted in 2 weeks or more of bsic eduction since RA (%) (7.5) (17.2) 3.7** (1.8) 0.26 Prticipted in 2 weeks or more of voctionl eduction since RA (%) (14.9) (23.0) 3.9* (2.1) 0.14 Food security nd hunger Household is food insecure (%) (44.1) (47.5) 7.7 (5.3) 0.14 Food insecurity scle d (1.93) (2.05) 0.24 (0.23) 0.10 Economic stressors Economic stress scle e (0.48) (0.49) 0.04 (0.06) 0.06 CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. ITT = intention-to-tret. RA = rndom ssignment. SD = stndrd devition. SE = stndrd error. SNAP = Supplementl Nutrition Assistnce Progrm. SSDI = Socil Security Disbility Insurnce. SSI = Supplementl Security Income. TANF = Temporry Assistnce for Needy Fmilies. WIC = Specil Supplementl Nutrition Progrm for Women, Infnts, nd Children. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire usul cre group. b Number of months worked for py includes prtil clendr months. c Hours of work per week includes those not currently working (tht is, those with 0 hours of work per week). d Food insecurity scle rnges from 0 to 6, with higher vlues indicting higher food insecurity. e Economic stress scle rnges from 1 to 1, with higher vlues indicting higher economic stress. Notes: Impct estimtes nd outcome mens re regression djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Source: Fmily Options Study 18-month followup survey The CBRR-versus-PBTH comprison does not yield strong pttern of effects in ny of the five study domins. In the housing stbility domin, CBRR ppers to hve incresed the incidence of homelessness (not including stys in trnsitionl housing) somewht reltive to PBTH, ccording to the survey dt. This effect is likely influenced by the fct tht 22 percent of PBTH fmilies were still using trnsitionl housing t followup. No similr effect emerges in the use of emergency shelters (bsed on progrm use dt), however. No effects pper on the proportion of fmilies living in their own house or prtment t followup or the number of residentil moves in the pst 6 months. In the dult well-being domin, CBRR ppers to hve lowered the mount of psychologicl distress for fmily heds reltive to PBTH. This result seems inconsistent with the incresed report of emergency shelter use nd stys in plces not ment for humn hbittion in the 6 months before the survey. Given the generl pucity of sttisticlly significnt results for this comprison, nd the inconsistency of the results tht do chieve sttisticl significnce, the study tem hesittes to drw strong conclusions for this comprison. Overll, the experiences of fmilies ssigned to CBRR do not pper to differ gretly from those ssigned to PBTH. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 126

159 CHAPTER 10. IMPACTS OF POOLED COMPARISONS In ddition to conducting the six pirwise comprisons, the study tem combined ssignment groups in vrious wys to exmine dditionl comprisons. The following four comprisons were identified s being of interest to HUD. 1. Wht is the impct of ny kind of housing subsidy for homeless fmilies (permnent housing subsidy [SUB] + communitybsed rpid re-housing [CBRR] + project-bsed trnsitionl housing [PBTH]) compred with the impct of the usul cre (UC) offered in the community? 2. Wht is the impct of housing subsidy with hevy services provided to homeless fmilies (PBTH) compred with the impct of housing subsidies with light or no services (SUB + CBRR)? 3. Wht is the impct of interventions tht re more costly (SUB + PBTH) compred with the impct of less costly intervention (CBRR)? 4. Wht is the impct of housing subsidy with no time limit (SUB) compred with the impct of time-limited housing subsidies (CBRR + PBTH)? One benefit of pooling interventions in impct comprisons is tht it provides lrger smple sizes for nlysis. A fmily ws included in pooled comprison if its rndomiztion set included t lest one intervention on ech side of the impct comprison. For exmple, fmily ws included in the SUB + PBTH comprison with CBRR if it hd the opportunity to be ssigned to CBRR nd to either SUB or PBTH. 133 Exhibit 10-1 shows the number of fmilies tht re included in the comprisons used to ddress the preceding questions. Our exmintion of the impct results from the four pooled comprisons unexpectedly yielded little useful informtion on the questions posed. Insted, ll the estimtes pper to be dominted by the reltively lrge effects of the SUB intervention when compred with ny of the other ssignment sttuses, no mtter how the different rndom ssignment rms re grouped. Therefore, we do not discuss the results of the pooled comprisons here in the body of the report; rther, we provide the results in Appendix F with no dditionl discussion. Exhibit Smple Sizes in the Four Pooled Comprisons Assigned Intervention Smple Size in Pooled Comprison SUB + CBRR + PBTH vs. UC SUB + CBRR vs. PBTH SUB + PBTH vs. CBRR CBRR + PBTH vs. SUB SUB CBRR PBTH UC 578 Totl 1, CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. UC = usul cre. Smple sizes re number of fmilies who responded to the 18-month followup survey. Source: Fmily Options Study 18-month followup survey 133 The rndomiztion sets tht provided the opportunity to be ssigned to CBRR nd to either SUB or PBTH were {SUB, CBRR, PBTH, UC}, {SUB, CBRR, UC}, nd {PBTH, CBRR, UC}. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 127

160 CHAPTER 11. DO CERTAIN INTERVENTIONS WORK BETTER WHEN OFFERED TO FAMILIES WHO FACE GREATER DIFFICULTIES? Previous chpters of this report hve exmined which interventions work best cross ll the fmilies in the study. This chpter sks whether some interventions work better thn others for fmilies with prticulr chrc - teristics. For exmple, previous chpters hve shown tht, on verge, the permnent housing subsidy (SUB) intervention hs substntil impcts reltive to ll other interventions, not only for housing stbility but lso for outcomes in other domins. But do ll fmilies who experience homelessness need deep permnent housing subsidy, or might some do s well on their own in usul cre (UC) or with the shorter nd often shllower subsidies of community-bsed rpid re-housing (CBRR)? The hypotheses set forth in Chpter 3 suggest tht the dvntge of SUB might increse s fmilies brriers to housing increse, becuse SUB overcomes those brriers. Similrly, previous chpters hve shown tht, on verge, project-bsed trnsitionl housing (PBTH) hs only modest impcts confined to housing stbility in comprison with UC. Surprisingly, if nything, PBTH hs worse impcts on dult well-being reltive to CBRR. But might this be consequence of offering PBTH to fmilies who do not need such intensive services in supervised setting? Chpter 3 suggests tht PBTH might hve greter benefits for fmilies who fce more of the psychosocil chllenges thn PBTH is designed to ddress. The more generl form of this question is whether the reltive benefits of the longer term or more intensive interventions (SUB nd PBTH) might increse s fmilies reported difficulties increse. To evlute this possibility, we creted indices of psychosocil chllenges nd housing brriers nd exmined whether the impct of the interventions reltive to ech other nd to UC increses s fmilies scores on these indices increse. The psychosocil chllenge index is count of the number of psychosocil chllenges reported by fmilies t the bseline survey just before rndom ssignment. We considered 9 potentil chllenges, including helth nd mentl helth conditions, substnce use problems, post-trumtic stress disorder (PTSD), intimte prtner violence, felony conviction, history of foster cre or institutionl plcement s child, nd disbility of the prent or child. Severl of these fctors predicted residentil instbility (cross progrm type) in the Service nd Housing Interventions for Fmilies in Trnsition (SHIFT) study (Hyes, Zonneville, nd Bssuk, 2013). Fmilies reported n verge of 2.20 psychosocil chllenges. The housing brriers index is count of the number of 15 potentil brriers including unemployment, lck of income, pst evictions or lese violtions, lck of trnsporttion, nd fmily composition tht fmilies reported t the time of the bseline survey to be t lest smll problems in trying to find housing. (We omitted brriers involving criminl ctivity nd disbility tht overlpped with the psychosocil chllenge index.) A similr list of brriers ws ssocited with incresed returns to homelessness for fmilies in the Evlution of the Rpid Rehousing for Homeless Fmilies Demonstrtion Progrm (Finkel et l., forthcoming). Fmilies reported n verge of 6.52 housing brriers. The two indices were positively correlted, suggesting tht they reflect seprte but relted mesures of difficulties tht fmilies fce Descriptive Results As shown in Exhibit 11-1, the most common psychosocil chllenges reported by the dult respondent living in shelter before rndomiztion were experience of domestic violence (48.9 percent), poor or fir helth (29.5 percent), nd hving been in foster cre or n institution s child (25.3 percent). 134 The correltion is r = 0.25, p <.01 in the full smple. Stndrd devitions were 1.86 for psychosocil chllenges nd 2.56 for housing brriers. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 128

161 Chpter 11. Do Certin Interventions Work Better When Offered to Fmilies Who Fce Greter Difficulties? Exhibit Percentges of Adult Respondents Reporting Psychosocil Chllenges t the Time of Study Enrollment (for fmilies interviewed t 18 months) Psychosocil Chllenges Percent Domestic violence 48.9% Poor or fir helth 29.5 In foster cre or institution s child 25.3 PTSD 22.4 Severe psychologicl distress 21.8 Disbility 20.7 Drug or lcohol use 20.4 Child with disbility 17.0 Pst felony 10.6 PTSD = post-trumtic stress disorder. Source: Fmily Options Study bseline survey As shown in Exhibit 11-2, nerly ll fmilies reported insufficient income to py rent nd insufficient income to py security deposit s problems, nd most reported not being employed, hving poor credit history, nd lcking trnsporttion s problems t the time they enrolled in the study Differentil Impcts Depending on Psychosocil Chllenges Exhibit 11-3 illustrtes possible differentil impcts of in - terventions bsed on number of psychosocil chllenges by showing the estimted size of the impct of ech contrst (for exmple, SUB versus UC) t the 20th nd 80th percentiles of chllenge. The 20th percentile is 0 chllenges (tht is, more thn 20 percent of fmilies reported no chllenges) nd the 80th percentile is 4.0 chllenges. The sterisks in the exhibit reflect whether the vrition in impct by level of ech psychosocil chllenge is sttisticlly significnt. For exmple, the first row considers impcts for the confirmtory outcome of t lest 1 night homeless or doubled up in the pst 6 months or t lest 1 night in shelter in the pst 12 months. In the first pir of columns, the impct of SUB in comprison with UC is estimted to be reduction of percentge points in this outcome for those with low numbers of chllenge nd percentge points for those with high numbers of chllenges. The verge effect is very lrge nd significnt, s reported previously in Chpter 6. The two impct estimtes re nerly the sme, however the vrition in impct bsed on number of psychosocil chllenges is trivil. As in other chpters of this report, we consider findings tht re sttisticlly significnt t the.10 level or better s evidence of differences in impct mgnitude for fmilies with different degrees of chllenges. Moreover, s in other chpters, we tke into ccount the number of sttisticlly significnt findings reltive to the number of tests conducted. For ech policy comprison, such s SUB versus UC, we test for vrition in impct for ech of the 18 key outcome vribles included in the executive summry, nd so might expect 2 of the 18 tests conducted to rech sttisticl significnce on the bsis of Exhibit Percentge of Fmilies Reporting Tht Condition Ws Big or Smll Problem in Finding Plce to Live t the Time of Study Enrollment (for fmilies interviewed t 18 months) Housing Brriers Percent Reporting Big or Smll Problem Not enough income to py rent 96.7% Inbility to py security deposit or first/lst month s rent 94.2 Not currently employed 79.9 Poor credit history 73.3 Lck of trnsporttion to look for housing 65.5 No reference from pst lndlords 43.9 Pst eviction 39.6 No rent history t ll 39.0 Recently moved to community nd no locl rent history 32.6 Problems with pst lndlords 19.6 Three or more children in the household 17.8 Rcil discrimintion 17.4 Pst lese violtions 16.8 Someone in the household less thn 21 yers old 8.5 Teengers in the household 5.5 Source: Fmily Options Study bseline survey Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 129

162 Chpter 11. Do Certin Interventions Work Better When Offered to Fmilies Who Fce Greter Difficulties? Exhibit Impcts Moderted by Psychosocil Chllenges Index Outcome SUB vs. UC CBRR vs. UC PBTH vs. UC SUB vs. CBRR SUB vs. PBTH CBRR vs. PBTH Impct t Low vs. High Chllenge Housing stbility At lest 1 night homeless or doubled up in pst 6 months or in shelter in pst 12 months (%) At lest 1 night homeless or doubled up in pst 6 months (%) Number of plces lived in pst 6 months Any sty in emergency shelter in months 7 to 18 fter RA (%) Fmily preservtion Fmily hs t lest one child seprted in pst 6 months b (%) Spouse/prtner seprted in pst 6 months, of those with spouse/ prtner present t RA c (%) Fmily hs t lest one child reunified, of those fmilies with t lest one child bsent t RA d (%) Adult well-being Low High Low High Low High Low High Low High Low High * * * ** Helth in pst 30 dys ws poor * or fir (%) Psychologicl distress e Alcohol dependence or drug buse f (%) Experienced intimte prtner violence in pst 6 months (%) ** * Child well-being Number of schools ttended since *** ** RA g Childcre or school bsences in * *** lst month h Poor or fir helth (%) * ** Behvior problems i Self-sufficiency Work for py in week before ** survey (%) Totl fmily income ($) $1,146 $176 $682 $1,379 $649 $1,374 $843 $1,184 $1,536 $1,886 $1,176 $1,490 Household is food insecure (%) CBRR = community-bsed rpid rehousing. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. UC = usul cre. RA = rndom ssignment. */**/*** Impct mgnitude vries significntly with level of psychologicl chllenge t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. The definition of homeless in this report includes stys in emergency shelters nd plces not ment for humn hbittion. It excludes trnsitionl housing. b Percentge of fmilies in which child who ws with the fmily t bseline becme seprted from the fmily in the 6 months before the 18-month survey. c Percentge of fmilies in which spouse or prtner who ws with the fmily t bseline becme seprted from the fmily in the 6 months before the 18-month survey. d Percentge of fmilies in which t lest one child ws seprted from the fmily t bseline nd no child ws reunited with the fmily t the time of the 18-month survey. e Psychologicl distress is mesured with the Kessler 6 (K6) scle nd rnges from 0 to 24, with higher scores indicting greter distress. Impcts shown s stndrdized effect sizes. Effect sizes were stndrdized by dividing impcts by stndrd devition for the UC group. f Mesures evidence of lcohol dependence or drug buse using responses to the Rpid Alcohol Problems Screen (RAPS-4) nd six items from the Drug Abuse Screening Test (DAST-10). g Number of schools outcome is topcoded t four or more schools. h Absences outcome is defined s 0 = no bsences in pst month; 1 = 1 2 bsences; 2 = 3 5 bsences; 3 = 6 or more bsences. i Behvior problems outcome is mesured s the stndrdized Totl Difficulties score from the Strengths nd Difficulties Questionnire (SDQ). Notes: The Low estimte is clculted t the 20th percentile of the modertor in the full smple nd the High estimte is clculted t the 80th percentile of the modertor. Impct men difference estimtes re regression-djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Source: Fmily Options Study 18-month followup survey Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 130

163 Chpter 11. Do Certin Interventions Work Better When Offered to Fmilies Who Fce Greter Difficulties? chnce lone. Where more thn 2 test results exceed the.10 threshold of significnce for given policy comprison but only modestly more we lso consider whether the sttisticlly significnt findings conform to the hypothesis tht the more intensive intervention in given comprison will hve lrger impcts t higher levels of chllenges or brriers. If this pttern does not hold, the existence of modestly more thn the number of significnt findings expected by chnce is not credited s evidence of rel impct vrition SUB Versus UC SUB, s the more intensive intervention, would be expected to hve greter impct reltive to UC for fmilies with higher numbers of psychosocil chllenges. With only 2 sttisticlly significnt test findings mong 18 tests conducted the num ber expected by chnce lone if no true impct vrition occurred we cnnot conclude tht impcts of SUB in com - prison with UC differ for fmilies with different numbers of psychologicl chllenges. Insted, the results re best inter - preted s rndom vrition. In generl, the levels of psychosocil chllenge tht fmilies experience do not moderte the substntil verge differences between SUB nd UC shown in previous chpters CBRR Versus UC CBRR, s the lest intensive of the ctive interventions, would be expected to work better in comprison with UC for fmilies with low psychosocil chllenges rther thn those with high chllenges. With only 2 sttisticlly significnt test findings mong 18 tests conducted fewer thn the number expected by chnce lone if no true impct vrition occurred we cnnot conclude tht the size of impcts fmilies experience from CBRR compred with UC vry by their number of psychosocil chllenges. In prticulr, no evidence shows tht CBRR worked better for fmilies with lower numbers of chllenges PBTH Versus UC As the more intensive intervention, nd one tht ddresses psychosocil chllenges directly, PBTH would be expected to hve greter impct reltive to UC for fmilies with higher numbers of these chllenges, but this ws not the cse. No differentil effects met the threshold of sttisticl significnce for ny of the 18 outcomes exmined SUB Versus CBRR SUB, s the more intensive intervention, would be expected to hve greter impct thn CBRR for fmilies with higher numbers of psychosocil chllenges even if SUB does not ddress these chllenges directly. Of 5 sttisticlly significnt results (of 18 tests totl), 2 point in the opposite direction: SUB s effects grows less fvorble reltive to CBRR s the num - ber of chllenges increses for the confirmtory outcome (ny sty in emergency shelter in the 12 months before the survey or sty in plce not ment for humn hbittion or doubling up in the 6 months before the survey) nd child seprtions from the fmily. In fct, the positive point estimte for the ltter result implies tht SUB cuses more child seprtions reltive to CBRR for fmilies with high number of chllenges, which is counterintuitive. Of the 3 significnt results tht conform to the expected pttern (1) experience of intimte prtner violence, (2) number of schools ttended, nd (3) children s poor or fir helth one seems to be contrdictory (for high-chllenge fmilies, SUB hs smller effect on homelessness thn CBRR but lrger reduction in the number of schools ttended) nd nother involves per - verse effect (for low-chllenge fmilies, SUB worsens children s helth reltive to CBRR). This inconsistent pttern of effects prevents us from concluding tht the size of impcts fmilies experience from SUB compred with CBRR vries by number of psychosocil chllenge. These results should be interpreted s spurious, the result of rndom vrition in the dt SUB Versus PBTH Becuse PBTH ddresses psychosocil needs more directly thn SUB, proponents would expect it to be especilly beneficil for fmilies with higher numbers of these chllenges. Proponents of SUB mke the opposite prediction. With no sttisticlly significnt test finding mong 18 tests conducted, the evidence does not confirm either point of view. We con - clude tht psychologicl chllenges do not moderte the size of impcts fmilies experience from PBTH compred with SUB CBRR Versus PBTH As the more intensive intervention, PBTH would be expected to hve greter impct reltive to CBRR for fmilies with higher psychosocil chllenges. Of 5 sttisticlly significnt findings (of 18 tests totl), 2 point in the opposite direction: CBRR s effects grow more fvorble reltive to PBTH s the number of chllenges increses for child seprtions nd child bsences from childcre or school. The effect of PBTH grows more fvorble reltive to CBRR s chllenges increse in 3 significnt results: (1) the confirmtory homeless out - come, (2) the number of schools ttended, nd (3) poor or fir child helth. For helth, however, the result simply brings the two interventions to prity (t high levels of chllenge) from n initil dvntge fvoring CBRR (t low levels of chllenge). With only 2 of 5 results conforming well to the postulted pttern, we re cutious in interpreting the Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 131

164 Chpter 11. Do Certin Interventions Work Better When Offered to Fmilies Who Fce Greter Difficulties? evidence s nything other thn chnce differences tht hppened to occur in this instnce t more thn the expected rte (5 in 18 times, rther thn 2 in 18) Differentil Impcts Depending on Housing Brriers Exhibit 11-4 illustrtes possible differentil impcts of interventions bsed on fmilies numbers of housing brriers by showing the estimted size of the impct of ech contrst (for exmple, SUB versus UC) t the 20th nd 80th percentiles of brriers. For housing brriers, the 20th percentile is 4.0 brriers nd the 80th percentile 8.57 brriers. As in the previous section on psychosocil chllenges, the sterisks reflect whether the vrition in impct by level of housing brriers is sttisticlly significnt. Also, s in tht section, we consider both the number of sttisticlly significnt findings nd their ptterns in interpreting whether results show rel evidence of vrition in impct SUB Versus UC SUB, s the more intensive intervention, would be expected to hve greter impct reltive to UC for fmilies with higher housing brriers, but this expecttion ws not the cse. None of the 18 comprisons were sttisticlly significnt. The substntil verge differences between SUB nd UC shown in previous chpters hold cross numbers of housing brriers CBRR Versus UC CBRR, s the lest intensive of the ctive interventions, would be expected to work best in comprison with UC for fmilies with lower brriers to housing. Some question whether limited-term subsidies re sufficient to overcome higher housing brriers. With only 2 sttisticlly significnt test findings mong 18 tests conducted the number expected by chnce lone we cnnot conclude tht these suppositions re true. Insted, we find no evidence tht housing brriers moderte the size of impcts fmilies experience from CBRR compred with UC. The overll pttern of effects does not vry by number of housing brriers PBTH Versus UC As the more intensive intervention, PBTH would be expected to hve greter impct reltive to UC for fmilies with higher housing brriers. With only 1 sttisticlly significnt test find - ing mong 18 tests conducted less thn the number expected by chnce lone if no true impct vrition occur red we cnnot conclude tht the size of impcts fmilies experience from PBTH compred with UC depends on the number of brriers they fce SUB Versus CBRR SUB, s the more intensive intervention, would be expected to hve greter impct thn CBRR for fmilies with higher housing brriers. With only 1 sttisticlly significnt test finding mong 18 tests conducted fewer thn expected by chnce lone we cnnot conclude tht the size of impcts fmilies experience from SUB compred with CBRR differ for fmilies with different numbers of housing brriers SUB Versus PBTH Becuse housing subsidies overcome mny brriers to hous - ing, proponents of SUB would expect it to be especilly beneficil for fmilies with higher levels of these brriers. Proponents of PBTH mke the opposite prediction. The 2 sttisticlly significnt results out of 18 fil to confirm either proposition nd re best interpreted s chnce CBRR Versus PBTH As the more intensive intervention, PBTH would be expected to hve greter impct reltive to CBRR for fmilies with higher housing brriers. Of 18 results here, 3 rech sttisticl significnce, but the direction of effects is not consistent. CBRR s effects, not PBTH s, grow more fvorble s brriers increse for one outcome child bsences from childcre or school. The opposite pttern holds PBTH s effects growing more fvorble reltive to CBRR s chllenges increse for 3 other outcomes: 2 correlted homelessness outcomes spending t lest 1 night homeless or doubled up in the pst 6 months nd the confirmtory outcome. We interpret these results s simply rndom vrition becuse they re not hugely more common thn expected due to chnce (3 in 18 times, rther thn 2 in 18) nd do not point consistently to the superiority of one intervention or nother for different groups of fmilies Summry It is cler tht fmilies in this study experience high numbers of psychosocil chllenges nd even higher numbers of brri - ers to housing. This result ws by design: the study enrolled fmilies only fter they hd spent t lest 7 dys in shelter. At the sme time, the exmintion of potentil modertor effects of difficulties of this sort does not provide evidence tht ny of the interventions studied works comprtively better for fmilies who hve greter psychosocil chllenges or housing brriers thn for fmilies who fce fewer difficulties. We cnnot completely rule out the possibility of differentil effects doing so would require lrger smple sizes thn re vilble in the study. At this point, however, the min study results on impcts cross ll fmilies provide the study s clerest guidnce for policy nd prctice. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 132

165 Chpter 11. Do Certin Interventions Work Better When Offered to Fmilies Who Fce Greter Difficulties? Exhibit Impcts Moderted by Housing Brriers Index Outcome SUB vs. UC CBRR vs. UC PBTH vs. UC SUB vs. CBRR SUB vs. PBTH CBRR vs. PBTH Impct t Low vs. High Housing Brriers Housing stbility At lest 1 night homeless or doubled up in pst 6 months or in shelter in pst 12 months (%) At lest 1 night homeless or doubled up in pst 6 months (%) Number of plces lived in pst 6 months Any sty in emergency shelter in months 7 to 18 fter RA (%) Fmily preservtion Fmily hs t lest one child seprted in pst 6 months b (%) Spouse/prtner seprted in pst 6 months, of those with spouse/ prtner present t RA c (%) Fmily hs t lest one child reunified, of those fmilies with t lest one child bsent t RA d (%) Adult well-being Low High Low High Low High Low High Low High Low High * * * Helth in pst 30 dys ws poor or fir (%) Psychologicl distress e ** * Alcohol dependence or drug buse f (%) Experienced intimte prtner violence in pst 6 months (%) Child well-being Number of schools ttended since RA g Childcre or school bsences in *** ** lst month h Poor or fir helth (%) Behvior problems i Self-sufficiency Work for py in week before * survey (%) Totl fmily income ($) $779 $111 $1,193 $1,020 $973 $1,071 $1,515 $693 $1,658 $1,424 $217 $389 Household is food insecure (%) * CBRR = community-bsed rpid rehousing. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. UC = usul cre. RA = rndom ssignment. */**/*** Impct mgnitude vries significntly with level of housing brriers t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. The definition of homeless in this report includes stys in emergency shelters nd plces not ment for humn hbittion. It excludes trnsitionl housing. b Percentge of fmilies in which child who ws with the fmily t bseline becme seprted from the fmily in the 6 months before the 18-month survey. c Percentge of fmilies in which spouse or prtner who ws with the fmily t bseline becme seprted from the fmily in the 6 months before the 18-month survey. d Percentge of fmilies in which t lest one child ws seprted from the fmily t bseline nd no child ws reunited with the fmily t the time of the 18-month survey. e Psychologicl distress is mesured with the Kessler 6 (K6) scle nd rnges from 0 to 24, with higher scores indicting greter distress. Impcts shown s stndrdized effect sizes. Effect sizes were stndrdized by dividing impcts by stndrd devition for the UC group. f Mesures evidence of lcohol dependence or drug buse using responses to the Rpid Alcohol Problems Screen (RAPS-4) nd six items from the Drug Abuse Screening Test (DAST-10). g Number of schools outcome is topcoded t four or more schools. h Absences outcome is defined s 0 = no bsences in pst month; 1 = one two bsences; 2 = three five bsences; 3 = six or more bsences. i Behvior problems outcome is mesured s the stndrdized Totl Difficulties score from the Strengths nd Difficulties Questionnire, or SDQ. Notes: The Low estimte is clculted t the 20th percentile of the modertor in the full smple nd the High estimte is clculted t the 80th percentile of the modertor. Impct men difference estimtes re regression-djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Source: Fmily Options Study 18-month followup survey Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 133

166 CHAPTER 12. INTERVENTION COSTS This chpter of the report documents the costs of providing the housing or shelter nd supportive services in the progrms ssocited with the inter - ventions in the Fmily Options Study, including the cost of remining in emergency shelter. For decisionmkers de - signing nd implementing policy to ddress homelessness mong fmilies, understnding the reltive costs of the ctive interventions in this study is criticl complement to un - derstnding their reltive impcts. This chpter begins by introducing the concepts nd methods used to nlyze nd describe progrm costs nd then providing high-level summry of the cost estimtes (Section 12.1). To ssess the reltive costs of the interventions, it is crucil to understnd the cost per month for ech progrm, but lso to look t costs from two other perspectives: (1) the cost dur - ing the period the fmily uses the progrm (the progrm sty ) nd (2) the overll cost to ll providers of shelter nd housing ssistnce of giving fmilies priority ccess to prticulr type of progrm. Following the Introduction nd Summry section, the next sections (12.2 through 12.5) present in detil the costs per month of the progrms tht provided the permnent housing subsidy (SUB), communitybsed rpid re-housing (CBRR), nd project-bsed trnsitionl housing (PBTH) interventions nd of emergency shelter. Those sections lso present the verge costs per progrm sty for the fmilies ssigned to the ctive interventions nd the costs for the sty in emergency shelter tht followed rndom ssign - ment of fmilies to usul cre (UC). Section 12.6 compres monthly costs nd costs per sty cross the three ctive in - terventions nd emergency shelter. Section 12.7 compres the costs of ll use of shelter nd housing progrms in the followup period fter rndom ssignment by fmilies given priority ccess to ech intervention. Finlly, Section 12.8 compres the verge costs per month of ll use of shelter nd housing progrms used t the time of the followup survey by fmilies given priority ccess to ech intervention Introduction nd Summry The objective of the Fmily Options Study is to provide evi - dence to support decisions of policymkers, plnners, nd prctitioners ddressing homelessness mong fmilies. Although much of the study is focused on estimting reltive effects of different types of interventions, such estimtes re only one input into decisions bout homelessness policy. Another input is the cost of the interventions. Becuse of differences in the type of housing or shelter provided, the durtion of ssistnce, nd the rnge nd intensity of supportive services offered, the progrms ssocited with ech ctive intervention vry in cost. With respect to durtion, housing ssistnce provided by CBRR nd PBTH progrms is temporry; subsidies provided by SUB progrms re indefinite, s long s regulr eligibility requirements for the subsidy re mintined. Regrding the intensity of supportive services, progrms without supportive services will usully require fewer resources for given durtion thn those offer - ing such services. This chpter reports on the costs of providing the housing nd services in the progrms ssocited with the three ctive interventions: CBRR, PBTH, nd SUB. The chpter lso reports costs for emergency shelter progrms to provide informtion on the cost of the initil sty in emergency shelter during which fmilies were enrolled in the study nd lso ny subsequent return to shelter during the period between rndom ssignment nd the followup survey. 135 We present nlyses for four concepts of costs: (1) per-fmily monthly progrm cost, (2) progrm cost per sty during the followup period, (3) cost of ll progrm use during the followup period, nd (4) monthly cost of ll progrm use t the time of the followup survey. 1. Per-fmily monthly progrm cost: The cost of ll resources used to provide shelter or housing nd services to fmily during the course of month becuse they re receiving 135 The UC intervention includes whtever housing subsidies or supportive services fmilies were ble to obtin without study ssistnce. Becuse it ws not fesible to determine the extent nd costs of ny ssistnce beyond wht ws provided by the emergency shelter progrm, for per-fmily monthly progrm costs nd progrm cost per sty during the followup period, we report costs ssocited with emergency shelter only rther thn ll costs ssocited with the UC intervention in prticulr, for ssistnce tht fmilies my hve received fter leving shelter. The third cost mesure, the cost of ll progrm use during the followup period does estimte the cost of other progrm use for ll intervention types, including UC. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 134

167 Chpter 12. Intervention Costs ssistnce through prticulr progrm. We estimte perfmily monthly progrm costs of CBRR, PBTH, nd emergency shelter progrms by ctloging nd vluing the housing or shelter dimension of ech progrm (cpitl nd operting costs or rent) nd lso the services dimension (the personnel, spce, nd mterils used to provide services) in ech of 81 study progrms. We estimte costs of SUB in ech of the 10 sites providing SUB using dministrtive dt. Household-level dministrtive dt provide informtion on household-level monthly housing ssistnce pyments (HAPs) nd housing uthority finncil reports supply site-level costs of dministering the Housing Choice Voucher progrm Progrm cost per sty during the followup period: The cost of ll resources used to provide shelter or housing nd supportive services to fmily by the progrm to which they were given priority ccess during the time between rndom ssignment nd the fmily s followup survey. 137 We estimte progrm costs per sty by multiplying per-fmily monthly progrm costs by the verge number of months of ssistnce received by fmilies ssigned to tht progrm type in the progrm s site. Combining the monthly cost mesure of intensity of ssistnce with the durtion of receipt provides single mesure of the mount of housing or shelter nd services provided to fmilies t study intervention. 3. Cost of ll progrm use during the followup period: The cost of ll progrm use during the followup period ccounts for costs of ll progrms fmilies used during the followup period. Fmilies given priority ccess to prticulr type of progrm through rndom ssignment nonetheless used multiple progrms both the progrm type to which they were rndomized nd other progrm types. Rndom ssignment to one progrm mkes it more likely tht fmily will use tht progrm, but lso either more or less likely tht the fmily will use other housing or shelter progrms (see Exhibits 6-2, 7-4, nd 8-6). When compring progrm costs, it is useful for decisionmkers within the homeless services nd housing ssistnce systems to consider the costs of ll progrm use during the followup period for which this study estimtes impcts. As in the cost per sty, this cost includes the expense of providing housing or shelter nd relted ssis - tnce (services) to study fmilies during the time between rndom ssignment nd the followup survey. We estimte the cost of ll progrm use by multiplying the verge sitelevel per-fmily monthly progrm cost for ech progrm type by the number of months of ssistnce of ech respective type provided to ech fmily, s observed in the Fmily Options Study Progrm Usge Dt. The cost of ll progrm use during the followup period is the sum of these monthly costs times months of ssistnce. This chpter reports verges of this mount (clculted using the sme nonresponse weights used in the impct nlyses) for ech of the study s pirwise comprisons. Thus, this estimte provides totl cost of housing or shelter nd services tht reflects the different mixes of progrm types used tht resulted from fmily s being provided priority ccess to prticulr progrm type Monthly cost of ll progrm use t the followup survey: The monthly cost of ll progrm use t the followup survey considers the verge per-fmily monthly progrm cost of progrms from which fmilies were receiving ssistnce t the time of the followup survey. Initil rndom ssignment to one progrm type my mke it more or less likely tht the fmily will use other housing or shelter progrms in the medium nd long terms. As result, giving fmilies priority ccess to prticulr progrm type tody cn chnge the cost of ssistnce they receive months nd yers into the future. These subsequent costs will be nlyzed for longer timefrme nd reported in the 36-month report long with impcts mesured t 36 months. This chpter reports verges of this point-in-time cost clculted for ech of the study s pirwise comprisons. Ech of the four cost mesures provides useful, but different, informtion to decisionmker evluting the reltive costs nd benefits of CBRR, PBTH, nd SUB. A decisionmker con - sidering the reltive cost of funding one prticulr progrm 136 Gubits et l. (2013) described the type nd extent of services linked to the housing or shelter provided by the progrms prticipting in ech intervention. This cost nlysis dds to tht informtion by estimting the vlue of the resources tht progrms nd their prtners expend on those services per fmily per month, which cn serve s one mesure of the depth or intensity of the services. 137 The length of time between rndom ssignment nd the followup survey vried cross fmilies, with medin of 20 months. Weighted verges of durtion of ssistnce receipt re clculted t both the study level nd site level from the Progrm Usge Dt (see Chpter 4, Exhibit 4-2) using nonresponse weights. All observed months of use of progrm of the type to which fmily ws rndomly ssigned re counted for fmilies ssigned to CBRR, PBTH, nd SUB. Emergency shelter durtions re clculted s the verge durtion in the first observed shelter sty, the sty during which fmilies were rndomly ssigned. Subsequent returns to shelter re not included in this cost mesure, but they re cptured in the cost of ll progrm use during the followup period. 138 Mny fmilies ccessed shelter or housing nd relted services from progrms not in the cost study. All stys t progrms tht mtched type from study were vlued t the site-level verge of the per-fmily monthly progrm cost. So, ssistnce from ny rpid re-housing progrm ws vlued t the sites verge CBRR per-fmily monthly progrm cost, nd ssistnce from ny trnsitionl housing progrm ws vlued t the site s verge PBTH per-fmily monthly progrm cost. Study fmilies lso received ssistnce from progrms the study clssifies s permnent supportive housing, public housing, or project-bsed housing ssistnce (project-bsed vouchers or Section 8 projects), for which we hve no direct cost estimtes. Under the ssumption tht they hve similr progrm nd cost structures, we use site-level verge PBTH costs to estimte permnent supportive housing costs nd SUB costs to estimte public housing nd project-bsed ssistnce costs. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 135

168 Chpter 12. Intervention Costs type versus nother would use per-fmily monthly costs nd progrm costs per sty to ssess the costs ssocited with serving fmily with prticulr type of intervention nd over given length of time. For exmple, the nlysis below in Section shows tht the ongoing cost per month of providing fmily with CBRR, on verge, is lower thn the cost of providing either PBTH or SUB nd is substntilly lower thn the cost of continued sty in emergency shelter. Costs per sty within the followup period fter rndom s - signment dd the considertion tht ongoing costs re limited to the durtion of the ssistnce. Continuing our exmple, CBRR ssistnce is of shorter durtion thn PBTH or SUB, nd the per-sty cost of providing fmily with CBRR is sub - stntilly lower thn costs for the longer ssistnce episodes observed for SUB nd PBTH. (These computtions, however, ignore the possibility tht, t the end of this sty in this pro - grm, the individul might begin sty in some other progrm. The third nd fourth mesures ddress tht possibility nd lso the possibility tht someone ssigned to one progrm uses nother progrm insted.) The third mesure, the cost of ll progrm use during the followup period, incorportes the relity tht fmily rndomly ssigned to one progrm might nevertheless use different progrm t some point over the followup period. The intervention to which fmily is ssigned lso hs n indirect effect on the fmily s use of progrms not ssocited with the intervention. For exmple, this study finds tht SUB results in improved housing stbility, mening fewer subsequent stys in costly emergency shelters. As result of voiding those stys in emergency shelter, the estimted per-fmily cost of ll progrm use since rndom ssignment is only $1,478 higher for fmilies ssigned to SUB s compred with fmilies ssigned to CBRR, even though costs per sty for CBRR re less thn one-hlf s lrge s costs per sty for SUB. This third mesure of costs represents the totl cost of the housing nd services provided by the homeless services nd housing ssistnce systems to study fmilies given priority ccess to ech of the ctive interventions. The costs cn then be ssessed in the context of the reltive impcts of the interventions (reported in Chpters 6 through 9), which the study mesured during the sme period. The fourth mesure is first look estimte of how the third mesure of ll progrm use might continue to ccumulte during longer time horizon. If fmilies were to continue to use the mix of progrms they re observed using t the time of the followup survey, then ech month the verge cost of ll progrm use by fmilies ssigned to ech intervention would grow by this monthly mount. This pproch to estimting costs is different from previous studies tht clculte the costs of homelessness. Mny studies in recent decdes sought to compre the cost of supportive housing for chroniclly homeless individuls or fmilies with minstrem helthcre nd public sfety costs of mnging this popultion in the bsence of supportive housing. 139 By contrst, this nlysis focuses on the costs incurred, not by other systems or services, but by the progrms tht constitute the interventions in the study. Thus, it is very different from this cost offset literture. Insted, this study mkes creful distinction between services tht re n integrl prt of the progrm nd other services tht the fmilies rndomized to ny of the four tretment groups might hve received from minstrem systems or from specilized systems but not becuse they were prticipting in the SUB, CBRR, nd PBTH interventions. A complete cost-benefit nlysis, which is not prt of the study design, would include estimtes of cost offsets to other systems nd of ll costs of services tht study prticipnts my hve received from providers tht were not involved with the study, nd it would lso include n ttempt to monetize the benefits ssocited with differences in impcts Cost Dt Collection nd Anlysis Methodology In clculting the costs of the SUB, CBRR, PBTH, nd emer - gency shelter progrms, the study tem included ll resources tht re used to provide the housing or shelter nd services tht re prt of the progrms tht provided the interventions. Thus, the study tem included services provided by prtners tht re not in the progrms budget, when those services re n integrl prt of the progrm for exmple, becuse prticipnts in the progrm hve preferentil ccess to the services. The cost concept lso includes the monetry equivlent of in-kind dontions of services nd mterils nd, for housing nd shelter, we include cpitl costs incurred. The study tem hd two ims in selecting progrms to in - clude in the cost nlysis. First, progrms tht served the most study fmilies were selected so tht cost estimtes would reflect ssistnce study fmilies ctully received. Second, progrms of ech intervention type offered t ech site were included so tht cost estimtes would reflect vrition in the housing or shelter nd services provided cross progrms. 139 An introduction to nd overview of this literture is provided in Culhne et l. (2007). Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 136

169 Chpter 12. Intervention Costs Exhibit 12-1 shows the counts of progrms included in the cost nlysis by intervention type. The progrms selected for the cost nlysis represent more thn 85 percent of study fmilies who ccepted study referrl to CBRR nd PBTH progrms nd more thn 90 percent of fmilies ssigned to UC from shelter progrms. The issue of selecting progrms did not rise for SUB becuse dministrtive dt were vilble for ll the SUB progrms in the study. For CBRR, PBTH, nd emergency shelter progrms, the study collected costs t the progrm level nd normlized the costs by the number of fmilies served by the progrm (s opposed to tying prticulr housing units or shelter beds nd specific supportive services to study fmilies directly). These progrm-level costs cn be thought of s the verge cost of providing housing or shelter nd services to typicl fmily served by the progrm. The primry source for cost dt for CBRR, PBTH, nd emergency shelter progrms ws udited expense sttements. Progrm budgets, stffing lists, prtner commitment letters, nd progrm stff estimtes of lbor nd mteril costs of ny services not reflected in expense sttements supplemented these sttements. The study tem developed SUB progrm costs using HUD Public nd Indin Housing Informtion Center (PIC) nd Tennt Rentl Assistnce Certifiction System (TRACS) dt covering ll study fmilies receiving SUB ssistnce nd Finncil Dt Schedule (FDS) dt for ll housing gencies providing the SUB intervention. 140 Becuse of differences in the underlying dt vilble for different progrm types, verges re clculted slightly differently for SUB thn for CBRR, PBTH, nd emergency shelter progrms. For CBRR, PBTH, nd emergency shelter, study- nd site-level verge per-fmily monthly costs re bsed on progrm-level cost dt ssessed through the study tem s primry dt collection. To reflect ctul ssistnce received, these verges re clculted weighting by the number of study fmilies tht enrolled in the progrms. For SUB dt, both study- nd site-level verge per-fmily monthly progrm costs re clculted directly from ssistnce pyments for fmilies ssigned to SUB observed in dministrtive dt. To clculte verge progrm cost per sty during the followup period, progrm-level monthly costs (nd site-level for SUB) re multiplied by verge months of ssistnce of the relevnt progrm type received by fmilies ssigned to tht type who received some ssistnce. 141 These progrm-level estimtes re gin verged, weighting by the number of study fmilies who enrolled in progrms (or tht re observed in SUB dministrtive dt). As described in the previous section, the verge cost of ll progrm use during the followup period is clculted from household-level dt using site-level cost estimtes for ech progrm type nd individul fmily months of ssistnce receipt. Averges re reported for ech of the study s pirwise comprison using the sme nonresponse weights used in the impct nlyses. This chpter breks costs into two brod ctegories: Housing or shelter costs refer to the rentl cost either observed or estimted of the spce used to provide housing or shelter nd progrm services, nd to ny mintennce or other fcility opertion costs (including durble items such s furnishings). The rentl cost is net of ny rent pyments mde by the fmily. Supportive services costs refer to the cost of ny services other thn housing or shelter provided s n integrl prt of the progrm, including cse mngement, nd ny csh or in-kind ssistnce (for exmple, mels provided in emergency shelters). Two other ctegories of costs re mesured nd included in the clcultion of housing or shelter costs nd of services costs. Additionl detil is shown for these two types of costs becuse they provide informtion on typicl progrm structures: Exhibit Progrms Included in the Cost Anlysis Intervention Number of Progrms in Cost Anlysis SUB 10 sites (dministrtive dt) CBRR 12 PBTH 24 ES 45 CBRR = community-bsed rpid re-housing. ES = emergency shelter. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. Notes: For the 10 sites with the SUB intervention, verge costs re clculted from household-level dministrtive dt by site for fmilies who received services from ll 18 SUB providers. SUB ws not vilble in Atlnt or Bltimore. Source: Fmily Options Study cost dt 140 PIC nd TRACS provide dt on housing subsidies. FDS mesures dministrtive costs for SUB. 141 Becuse these site-level verge durtions re clculted using Progrm Usge Dt tht is linked to the followup survey, these verges re clculted using the sme survey nonresponse weights used to estimte impcts. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 137

170 Chpter 12. Intervention Costs Administrtive nd overhed costs include mngement slries; legl, ccounting, nd other professionl services; nd progrm support costs such s insurnce premiums nd gency nd ssocition fees. Administrtive nd overhed costs re divided mong supportive services nd housing or shelter costs ccording to the cost types reltive shre of totl costs so tht they re included in the two brod ctegories. In-kind nd prtner costs include ny costs of housing or shelter nd supportive services provided to fmilies becuse they prticipte in progrm. These costs re not provided by the progrm itself, nd, s result, re not included in progrm finncil sttements. Common exmples include onsite helth or mentl helth providers funded by n outside gency, community volunteers providing vriety of services, nd consumer goods donted to progrm clients. The impor - tnce of these costs vries widely from progrm to progrm. When present, they typiclly re prt of the cost of supportive services provided by progrm. In some cses, however, housing or shelter costs include the costs of lbor, such s hndymn services, or of fcilities used regulrly for progrm ctivities tht were provided in kind. In other cses, ccounting, legl, or dministrtive services were provided in kind or by prtners. In ech cse, the study tem pportioned the cost to the pproprite ctegory. Further detil on cost collection nd estimtion methodology is in Appendix G Summry of Findings The Fmily Options Study interventions were intended to vry in both intensity nd durtion. SUB progrms provide deep housing subsidy but limited supportive services for n indefinite durtion. PBTH progrms provide deep housing nd services support for reltively long durtion. CBRR progrms provide short-term housing subsidy with little supportive services. Finlly, emergency shelter progrms often offer supportive services nd housing for very limited time. This section summrizes findings for ech of the four cost mesures: per-fmily monthly progrm cost, progrm cost per sty during the followup period, cost of ll progrm use during the followup period, nd monthly cost of ll pro - grm use t the followup survey. Subsequent sections provide dditionl detil nd explntion for these reported costs. Per-Fmily Monthly Progrm Costs Monthly costs of serving typicl fmily vry considerbly by progrm type. Exhibit 12-2 presents the verge perfmily monthly progrm cost for ech type of progrm. SUB progrms, on verge, cost slightly less thn $1,200 per fmily per month, which consists wholly of the cost of housing, becuse this intervention provides no supportive services. CBRR progrms hve the lowest per-fmily monthly progrm cost mong the progrm types, with progrm verge of slightly less thn $900. Housing costs, on verge, mke up 72 percent of CBRR progrm costs. Exhibit Averge Per-Fmily Monthly Cost of Supportive Services nd Housing or Shelter Across Progrm Types $6,000 Supportive services Housing or shelter $5,000 $4,819 Averge monthly cost per fmily $4,000 $3,000 $2,000 $1,000 $1,162 $878 $2,706 $0 SUB CBRR PBTH ES CBRR = community-bsed rpid re-housing. ES = emergency shelter. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. Sources: Fmily Options Study cost dt (CBRR, PBTH, nd ES); HUD Public nd Indin Housing Informtion Center, Tennt Rentl Assistnce Certifiction System, nd Finncil Dt Schedule records (SUB) Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 138

171 Chpter 12. Intervention Costs Emergency shelter progrms hve the highest per-fmily monthly progrm cost for both supportive services nd housing or shelter, which, on verge, totl slightly more thn $4,800. Supportive services mke up 63 percent of emergency shelter costs, the highest shre mong the four progrm types. The higher monthly cost of housing or shelter for emergency shelter progrms reflects both progrm structure nd the pproch to clssifying costs. Emergency shelters tend to hve higher per-fmily levels of fcility stffing nd expenditure for mintennce nd mterils thn PBTH progrms or thn wht is reflected in rents pid by CBRR nd SUB progrms. In ddition, housing or shelter costs include the cpitl cost vlue of ll physicl spce provided by the progrm, including fcilities such s clssrooms, cse mngement offices, kitchens, nd dedicted childcre centers. 142 PBTH progrms hve n verge per-fmily monthly progrm cost of slightly more thn $2,700, with supportive services, on verge, constituting 42 percent of PBTH progrm costs. Within ech study intervention, the study tem lso found substntil vrition in the costs of the individul progrms. Exhibit 12-3 shows this vrition mong per-fmily monthly progrm costs for ech progrm type. PBTH nd emergency shelter progrms hve substntil vrition, driven lrgely by vrition in supportive services costs but lso by vrition in cpitl costs nd dministrtive expenses. For the 24 PBTH progrms in the cost nlysis, per-fmily monthly progrm cost rnges from slightly more thn $1,260 to slightly less thn $6,300. Per-fmily monthly progrm cost for the 45 emergency shelter progrms rnges from $1,900 to slightly more thn $9,000. Vrition in CBRR nd SUB costs cross progrms is driven lrgely by housing costs. For the 12 CBRR progrms in the cost nlysis, per-fmily monthly progrm cost rnges from slightly more thn $550 to slightly less thn $1,400. Across the 10 sites with the SUB intervention, verge per-fmily monthly progrm cost rnges from $770 to $2,100, lrgely reflecting differences in the locl cost of rentl housing. Progrm Costs Per Sty During the Followup Period We now turn to progrm costs per sty during the followup period. This cost concept reflects differences in durtion of ssistnce nd shows different cost ordering of the progrms. Exhibit 12-4 reports progrm costs per sty for ech progrm type. This cost concept ccounts for durtion between rndom ssignment nd the followup survey, mesured for fmilies who were ssigned to nd enrolled in progrms of ech type. Using this mesure, the costliest progrm type is PBTH, t n verge of slightly more thn $32,500 with n verge durtion of 13 months. 143 The next most costly progrm is SUB, with n verge of slightly more thn $18,800 with n verge durtion of 16 months. Emergency shelter progrm costs per sty verged slightly more thn $16,800 with n verge length of sty of 4 months. 144 Finlly, perfmily totl costs for fmilies who used rpid re-housing ver - ged slightly more thn $6,500 for n verge of 7 months of ssistnce. Exhibit Summry Sttistics of Per-Fmily Monthly Progrm Cost by Progrm Type Progrm Type Per-Fmily Monthly Progrm Cost Summry Sttistic Fmilies Progrms Men Minimum 25th Percentile Medin 75th Percentile Mximum ($) ($) ($) ($) ($) ($) SUB $1,162 $770 $844 $1,085 $1,370 $2,095 CBRR ,388 PBTH ,706 1,261 1,738 2,352 3,535 6,292 ES ,819 1,888 3,907 4,352 5,786 9,170 CBRR = community-bsed rpid re-housing. ES = emergency shelter. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. Notes: CBRR, PBTH, nd ES sttistics re clculted from progrm-level dt, weighted by the number of study fmilies tht enrolled in the progrm. SUB sttistics re clculted from household-level dt. Sources: Fmily Options Study cost dt (CBRR, PBTH, nd ES); HUD Public nd Indin Housing Informtion Center, Tennt Rentl Assistnce Certifiction System, nd Finncil Dt Schedule records (SUB) 142 This study s finding of higher monthly costs for fmily shelter progrms thn for other homelessness ssistnce progrm types is consistent with previous estimtes reported in HUD s Costs Associted With First-Time Homelessness for Fmilies nd Individuls (Spellmn, et l., 2010), which found emergency shelter monthly costs for fmilies were higher thn trnsitionl housing costs nd higher thn the locl Fir Mrket Rent (FMR) in three of four cities. By contrst, shelters serving individuls hd costs tht, on verge, were equl to or substntilly lower thn trnsitionl housing costs nd the FMR. 143 Averge durtions for progrm costs per sty during the followup period re clculted differently thn durtions reported in Chpters 6 through 9 for PBTH nd CBRR progrms. Averge durtions for progrm costs per sty re the verges, weighted by the number of study fmilies enrolled in progrms represented in the cost study, of site-level verge durtions for fmilies enrolled in progrms of the type to which they were ssigned. 144 This verge is the length fter rndom ssignment of the initil shelter sty for fmilies ssigned to UC. (See Chpter 5, Exhibit 5-2.) Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 139

172 Chpter 12. Intervention Costs Exhibit Averge Progrm Cost Per Sty During the Followup Period Across Progrm Types $35,000 $32,557 Supportive services Housing or shelter Averge progrm cost per sty during followup period $30,000 $25,000 $20,000 $15,000 $10,000 $5,000 $18,821 $6,578 $16,829 $0 SUB (16 months) CBRR (7 months) PBTH (13 months) ES (4 months) CBRR = community-bsed rpid re-housing. ES = emergency shelter. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. Note: The durtions reported in this exhibit re weighted to lign with the progrm-level cost dt nd so differ slightly from the durtions reported for CBRR nd PBTH in other exhibits. Sources: Fmily Options Study cost dt (CBRR, PBTH, nd ES); Fmily Options Study 18-month followup survey (CBRR nd PBTH); HUD Public nd Indin Housing Informtion Center, Tennt Rentl Assistnce Certifiction System, nd Finncil Dt Schedule records (SUB). These estimted costs per sty incorporte costs of ssistnce provided only through the end of the followup period (pproximtely 21 clendr months fter rndom ssignment). Costs will continue to ccrue fter the end of the followup period for this study, nd these longer term costs will differ cross interventions. CBRR nd PBTH re explicitly temporry progrms. For nerly ll study prticipnts, the term of CBRR or PBTH hd ended or would end shortly by the end of the followup period. By contrst, SUB is long-term intervention. Thus, during longer time horizon, the costs per sty in CBRR, PBTH, nd emergency shelter re unlikely to rise much; the cost per sty of SUB is likely to increse substntilly. Of the fmilies who used the SUB intervention, 89 percent were still receiving SUB t the followup survey. Our monthly estimtes imply tht costs for fmily who remins in SUB for 5 yers, s n exmple, would totl more thn $60,000 (using 5-percent discount rte). Some literture describes the length of time tht fmilies use housing vouchers. For exmple, Thompson (2007) reports tht nonelderly/ nondisbled households with children tht enter the voucher progrm receive ssistnce for n verge of 3.8 yers (s reported in 2002). This men includes 30 percent of fmilies ccessing ssistnce for less thn 1 yer. The liter ture does not include reported durtion bsed on homelessness experience, however. One could rgue such fmilies would hve longer length of sty thn typicl fmilies becuse they hve more needs or shorter length of sty (becuse they hve more brriers to mintining housing). Cost of All Progrm Use During the Followup Period We now consider the totl costs of ll types of shelter or housing ssistnce used by fmilies in the different intervention groups during the followup period fter rndom ssignment. Reltive to monthly costs or per-sty costs, these costs re more homogenous. As observed in the Progrm Use Since Rndom Assignment exhibits in Chpters 6 through 9, rndom ssignment to prticulr intervention chnged usge ptterns for other types of shelter progrms or housing ssistnce. For exmple, ssignment to the SUB intervention reduced use of both emergency shelter nd PBTH reltive to ssignment to ech of the other intervention rms. Exhibit 12-5 reports estimtes of the totl costs of ll housing ssistnce progrms (both those ssigned nd other ssistnce fmilies ccessed) tht resulted from these vrious ptterns. In this pproch, costs re tllied from rndom ssignment through the dte of the followup survey. Furthermore, this pproch considers ll shelter nd housing progrms provid - ing services to fmilies. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 140

173 Chpter 12. Intervention Costs Exhibit Summry of Cost of Progrm Use Since Rndom Assignment Contrst Estimted Averge Totl Cost ($) of Observed Housing Assistnce of ll Types for Fmilies in Comprison Assigned To SUB PBTH CBRR UC SUB vs. UC $30,832 $30,336 CBRR vs. UC $27,605 $30,629 PBTH vs. UC $30,817 $28,295 SUB vs. CBRR $31,158 $29,680 SUB vs. PBTH $27,864 $30,914 CBRR vs. PBTH $30,510 $22,524 CBRR = community-bsed rpid re-housing. ES = emergency shelter. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. UC = usul cre. Sources: Fmily Options Study cost dt; Progrm Usge Dt Using this cost concept, fmilies ssigned to SUB hve verge costs similr to fmilies ssigned to UC, slightly higher verge costs thn fmilies ssigned to CBRR, nd clerly lower costs thn fmilies ssigned to PBTH. This pttern is very different from the pttern shown in Exhibit 12-4, which by only tking the much longer progrm stys for SUB into ccount shows much higher costs for SUB thn for CBRR. The only slightly higher cost of SUB s compred with UC for the cost of totl progrm use is driven both by decresed time in emergency shelter nd by decresed use of reltively more expensive PBTH progrms by SUB fmilies s compred with UC fmilies. Similrly, the SUB nd CBRR interventions differ by only $1,500 (5 percent) in costs becuse the greter use of SUB progrms by fmilies ssigned to the SUB intervention is offset by the greter use of CBRR, TH, permnent supportive housing, nd emergency shelter progrms by fmilies ssigned to the CBRR intervention. As with cost per sty, n importnt cvet must be pplied to these estimtes of the totl cost of ll use of shelter nd housing ssistnce progrms during the followup period. The SUB intervention is likely to lst beyond tht followup period for most fmilies. Thus, SUB my become reltively more costly, even fter tking into ccount use of progrms other thn those to which fmilies were rndomly ssigned. These subsequent costs will be nlyzed for the longer timefrme nd reported in the 36-month report long with impcts mesured t 36 months. Monthly Cost of All Progrm Use t the Followup Survey Exhibit 12-6 reports the cost of progrm use t the time of the followup survey for ech ssigned progrm type of ech pirwise impct comprison. The cost of progrm use t the followup survey the verge per-fmily monthly progrm cost of progrms tht study fmilies were using t followup follows similr pttern cross tretment rms s the cost of progrm use since rn - dom ssignment. For exmple, becuse more fmilies ssigned to UC re in reltively costly emergency shelters t the followup survey s compred with more fmilies receiving SUB ssistnce mong fmilies ssigned to SUB, the cost of progrm use t the followup survey is nerly equl for the two tretment rms. The extent to which these ongoing cost levels persist will be n importnt topic of the cost nlysis for the 36-month report for the Fmily Options Study. Following this high-level summry of the cost findings, the next four sections of this chpter provide more indepth review of the per-fmily monthly progrm costs nd progrm costs per sty during the followup period for the SUB, CBRR, nd PBTH interventions nd for emergency shelter progrms from which fmilies were ssigned. Exhibit Summry of Monthly Cost of Progrm Use t the Time of the Followup Survey Contrst Estimted Averge Totl Monthly Cost ($) of Observed Housing Assistnce of ll Types for Fmilies in Comprison Assigned To SUB PBTH CBRR UC SUB vs. UC $1,086 $1,066 CBRR vs. UC $895 $1,098 PBTH vs. UC $1,009 $1,012 SUB vs. CBRR $1,081 $979 SUB vs. PBTH $1,065 $977 CBRR vs. PBTH $718 $989 CBRR = community-bsed rpid re-housing. ES = emergency shelter. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. UC = usul cre. Sources: Fmily Options Study cost dt; Progrm Usge Dt Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 141

174 Chpter 12. Intervention Costs 12.2 Cost of Permnent Housing Subsidy (SUB) Permnent housing subsidy progrms (SUB) hd per-fmily monthly progrm costs tht verged $1,162 (Exhibit 12-7). Locl rentl mrkets lrgely determine this cost. This verge cost ws nerly s low s the verge monthly cost of CBRR progrms of $878, which ws lower becuse the verge mount of the CBRR ssistnce in some progrms ws set well below the Fir Mrket Rent (FMR) nd the tennt income-bsed housing ssistnce pyment, or HAP, tht determines SUB costs. In ddition, ssistnce in some CBRR progrms ws phsed out during number of months (lowering verge monthly ssistnce). SUB progrms cost per sty during the followup period re nerly three times those of CBRR progrms (with n verge durtion of 7 months), however, becuse the verge durtion of SUB progrms within the followup period is slightly more thn 16 months, t slightly more thn $18,800. On the other hnd, SUB progrms cost bout one-hlf the cost of PBTH progrms on monthly bsis nd for the ver - ge progrm sty. This difference is lrgely becuse no sup - portive services re provided with SUB. Averge housing costs re lso lower for SUB progrms, however, becuse some PBTH progrms re locted in reltively high-rent res nd mintennce nd stffing costs (for exmple, front desk nd security) pper to be higher for some PBTH progrms thn for typicl smll- to medium-sized multifmily prtment building. Becuse of the greter cost per month, PBTH progrm stys re more costly thn SUB progrm stys, despite the longer durtion of progrm stys for SUB (verging 16 versus 13 months for PBTH). Exhibit 12-8 shows the verge cost of SUB progrms per month for ech site in which SUB ws vilble. Costs of SUB progrms re mde up of the HAP mde on behlf of ech fmily nd public housing gency (PHA) dministrtive costs, both of which vry cross sites. HAPs re determined s the difference between the PHAs pyment stndrd nd the income-bsed tennt pyment. Pyment stndrds re set by PHAs to be within 90 nd 110 percent of HUD s published FMR, with exceptions pproved by HUD (HUD-PIH, 2004). Administrtive costs rnged from 6 to 18 percent cross sites. Exhibit 12-9, sctter plot of monthly costs ginst FMR, shows tht, cross the 10 sites, SUB progrm costs re highly correlted with locl rentl mrket conditions. Progrm costs per sty during the followup period for SUB sites re clculted s the site-level per-fmily monthly progrm cost multiplied by the site-level observed verge months of ssistnce received between rndom ssignment nd the followup survey. Fmilies received, on verge, 16 months of SUB ssistnce in this period. Of the fmilies who used SUB, 89 percent remined in SUB t the time of the followup survey. Thus, nerly ll the vrition in durtion within the followup period is result of the time it took for fmilies to lese up (mke the trnsition from shelter to moving into housing unit with voucher subsidy). Exhibit SUB Progrm Cost Summry Sttistics for Fmilies Who Were Assigned To nd Took Up SUB Progrms Averge Over ll Fmilies Receiving SUB Rnge of Site Averges Across 10 SUB Sites SUB per-fmily monthly progrm cost ($) 1, to 2,095 SUB progrm cost per sty during the followup period ($) 18,821 12,304 to 32,449 Durtion in SUB during the followup period (months) to 18.2 Housing shre (%) 100 No vrition Supportive services shre (%) 0 No vrition Prtner nd in-kind shre (%) 0 No vrition Administrtive nd overhed cost shre (included in both housing nd supportive services costs) b (%) 9 6 to 18 SUB = permnent housing subsidy. Durtion clculted for fmilies ssigned to SUB tht received some SUB ssistnce. b Administrtive nd overhed costs re clculted for ech site from site-level dministrtive dt (Finncil Dt Schedule [FDS]) records. Housing nd supportive service shres dd to 100; prtner nd in-kind shre nd dministrtive shre re included in housing nd supportive services. Note: Averges re clculted directly from fmily-level records. Sources: HUD Public nd Indin Housing Informtion Center records; Tennt Rentl Assistnce Certifiction System records; FDS records Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 142

175 Chpter 12. Intervention Costs Exhibit Per-Fmily Monthly Progrm Costs for SUB Progrms Louisville Knss City Study sites with SUB intervention Slt Lke City Minnepolis Phoenix Denver Connecticut Almed Boston Honolulu $0 $500 $1,000 $1,500 $2,000 $2,500 Per-fmily monthly progrm cost SUB = permnent housing subsidy. Sources: HUD Public nd Indin Housing Informtion Center records; Tennt Rentl Assistnce Certifiction System records; Finncil Dt Schedule records Exhibit SUB Per-Fmily Monthly Progrm Costs nd Fir Mrket Rents $2,500 $2,000 Per-fmily monthly progrm cost $1,500 $1,000 $500 $0 $0 $200 $400 $600 $800 $1,000 $1,200 $1,400 $1,600 $1,800 $2, FMR two-bedroom SUB = permnent housing subsidy. FMR = Fir Mrket Rent. Note: The fitted line hs been dded to this sctter plot to mke the strong positive correltion between SUB progrm costs nd FMRs obvious. Sources: HUD Public nd Indin Housing Informtion Center records (costs); Tennt Rentl Assistnce Certifiction System records (costs); Finncil Dt Schedule records (costs); huduser.gov FMR documenttion Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 143

176 Chpter 12. Intervention Costs 12.3 Cost of Community-Bsed Rpid Re-Housing (CBRR) Community-bsed rpid re-housing progrms hve per-fmily monthly progrm costs tht verge $878, the lowest mong the CBRR, PBTH, nd SUB progrm types. Summry sttistics for the 12 CBRR progrms included in the cost nlysis re reported in Exhibit The key fetures of the CBRR intervention short-term rentl ssistnce with limited cse mngement re reflected in the cost dt: CBRR progrms hd the shortest durtion of 7 months mong the ctive interventions used by fmilies. Rentl ssistnce represents the bulk of progrm costs, t n verge of 72 percent. The cost dt collection confirmed tht cse mngement, typi - clly limited to housing serch ssistnce, is nerly lwys the only supportive service provided by CBRR. The dministr - tive shre of costs for CBRR progrms (which we pportion between housing or shelter nd supportive services costs) verges 11 percent. Some economies of scle re pprent when providing CBRR ssistnce becuse, in generl, progrms with higher per-fmily monthly progrm costs or tht ssisted greter number of fmilies hve lower dministrtive cost shres. Housing costs drive most of the vrition in costs cross CBRR progrms. Exhibit depicts the per-fmily monthly Exhibit Progrm-Level Cost Summry Sttistics for 12 CBRR Progrms Averge Across Progrms Rnge Across Progrms CBRR per-fmily monthly progrm cost ($) to 1,388 CBRR progrm cost per sty during the followup period ($) 6,578 3,509 to 15,738 Durtion in CBRR during the followup period (months) to 13.5 (cross sites) Housing shre to 87 Supportive services shre to 13 Prtner nd in-kind shre (included in housing nd supportive services costs s relevnt) Two sites: 1, 3 0 to 3 Ten sites: 0 Administrtive nd overhed cost shre (included in both housing nd supportive services costs) (%) 11 3 to 26 CBRR = community-bsed rpid re-housing. Durtion for study fmilies who were ssigned to nd enrolled in CBRR. Notes: Averges re weighted bsed on the number of study fmilies enrolled in ech CBRR progrm clculted from study enrollment dt. Housing nd supportive services shres dd to 100; prtner nd in-kind shre nd dministrtive shre re included in housing nd supportive services. Sources: Fmily Options Study cost dt; Progrm Usge Dt Exhibit Per-Fmily Monthly Progrm Costs for CBRR Progrms Supportive services costs Housing costs 12 CBRR progrms $0 $200 $400 $600 $800 $1,000 Per-fmily monthly progrm cost $1,200 $1,400 $1,600 CBRR = community-bsed rpid re-housing. Note: Site nmes re suppressed to preserve progrm nonymity. Source: Fmily Options Study cost dt Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 144

177 Chpter 12. Intervention Costs progrm costs for the 12 CBRR progrms for which the study tem obtined costs, one progrm for ech site in which fmilies were rndomized to CBRR. The CBRR progrms in the study were funded lrgely by temporrily vilble Homelessness Prevention nd Rpid Re-Housing Progrm (HPRP) funds, which did not require tht per-fmily ssistnce be bsed on locl rents. 145 Mny progrms included rent, however, long with household income in the formul used to determine ssistnce. As result, monthly housing ssistnce for CBRR progrms tends to be greter in sites with higher rentl prices (mesured by re FMR). The reltionship between FMR nd monthly verge costs is not nerly s strong for CBRR s it is for SUB, however, becuse not ll progrms determined ssistnce bsed on rent, nd becuse ssistnce ws often phsed out over number of months. CBRR ws intended to provide short-term rent subsidy. Consistent with tht intention, mong the CBRR-ssigned fmilies who received CBRR ssistnce, 81 percent received 1 yer or less of ssistnce (s of the followup survey). 146 Becuse CBRR hs the lowest monthly cost mong ll progrm types nd the shortest durtion mong CBRR, PBTH, nd SUB, t slightly more thn $6,500, CBRR is the lest costly intervention studied in terms of the progrm cost per sty during the followup period. As shown in Exhibit 12-5, however, nd discussed further in the following section, when ll the housing nd service progrms used by fmilies ssigned to CBRR re considered, the cost dvntge of the CBRR intervention over the SUB intervention in the followup period nerly disppers. Similrly, when counting ll progrms used, the cost dvntge of CBRR over PBTH nd UC is gretly reduced Cost of Project-Bsed Trnsitionl Housing (PBTH) PBTH progrms hve per-fmily monthly progrm costs of $2,706, second to emergency shelter nd more thn double the monthly costs per fmily of CBRR nd SUB. Among fmilies who used their ssigned PBTH progrms, progrm costs per sty during the followup period verged slightly more thn $32,500, the highest of ny ctive intervention. Averge monthly housing or shelter costs for PBTH re higher thn costs for CBRR nd SUB, but the bulk of the cost differentil is driven by the greter supportive services tht PBTH progrms provide. As reported with other cost summry sttistics in Exhibit 12-12, supportive services represent, on verge, 42 percent of PBTH costs. The supportive services costs included here re for services provided directly by the progrm or its prtners or volunteers to resident fmilies becuse they were in the PBTH progrm. Minstrem services tht SUB, CBRR, or PBTH fmilies received outside of the progrms to which they were ssigned re not included here. It is possible tht PBTH services sub - stituted for some of the minstrem services tht were used by CBRR or SUB fmilies, but which were not included in the costs we ttribute to those progrms. Minstrem services re those provided to fmilies becuse they re vilble in Exhibit Progrm-Level Cost Summry Sttistics for 24 PBTH Progrms Averge Across Progrms Rnge Across Progrms PBTH per-fmily monthly progrm cost ($) 2,706 1,261 to 6,292 PBTH progrm cost per sty during the followup period ($) 32,557 16,066 to 60,002 Durtion in PBTH during the followup period (months) to 17.1 (cross sites) Housing shre (%) to 84 Supportive services shre (%) to 16 Prtner nd in-kind shre (included in housing or shelter or supportive services cost s relevnt) (%) 8 0 to 44 Administrtive nd overhed cost shre (included in both housing nd supportive services costs) (%) 14 3 to 39 PBTH = project-bsed trnsitionl housing. Durtion in PBTH progrms in the first 18 months fter rndom ssignment clculted for study fmilies who were ssigned to nd enrolled in PBTH. Notes: Averges re weighted bsed on the number of study fmilies enrolled in ech PBTH progrm clculted from study enrollment dt. Housing nd supportive service shres dd to 100; prtner nd in-kind shre nd dministrtive shre re included in housing nd supportive services. Sources: Fmily Options Study cost dt; Progrm Usge Dt 145 HPRP ws uthorized through the Americn Recovery nd Reinvestment Act of Across the ntion, communities received $1.5 billion in HPRP funding, one-time funding strem vilble for 3 yers from progrm inception, to provide both homelessness prevention nd rpid re-housing ssistnce to individuls nd fmilies fcing homelessness. 146 Bsed on progrm usge dt, only 10 percent of CBRR-ssigned fmilies who received CBRR ssistnce were still receiving this ssistnce s of the followup survey. This proportion compres with 89 percent for SUB nd 40 percent for PBTH. In Chpters 5 through 9, durtion is reported by contrst for the time before the followup survey. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 145

178 Chpter 12. Intervention Costs the community nd becuse the fmilies re eligible, but not becuse they re prticipting in prticulr housing or shelter progrm. It lso is possible, however, tht the intensive cse mngement tht is prt of the higher PBTH supportive services costs led to more referrls tht incresed PBTH residents use of minstrem services compred with use of those services by SUB or CBRR fmilies. Exhibit shows how per-fmily monthly progrm cost vries cross the PBTH progrms. Most of the vrition in totl costs results from vrition in the extent of supportive services provided, which rnge from 16 to 77 percent of totl costs. Supportive services include some level of cse mngement for ll PBTH progrms. In cost dt collection, the study tem identified brod nd vried rry of dditionl supportive services tht PBTH progrms nd their prtners provided. Exmples include onsite mentl helth nd substnce buse therpy; childcre, tutoring, nd mentoring; clothing, toiletries, nd food; nd holidy gifts nd event tickets. Although vrition exists cross progrms in the cost of the housing portion of PBTH, no cler reltionship is pprent between locl FMR nd either totl PBTH costs or the housing/ shelter portion of costs. Insted, vrition in housing or shelter costs ppers to rise from differences in the ctul spce provided to fmilies in PBTH progrms nd stffing nd mintennce costs ssocited with progrm fcilities. Most progrms house fmilies on dedicted cmpus with vrying mixes of full-time mintennce, front desk, nd security stff. Some of these fcilities re locted in reltively high-cost neighborhoods, while others re locted in res with lower rentl rtes. A few progrms rent housing units tht re geogrphiclly dispersed within the locl community, often referred to s scttered-site units. As depicted in Exhibit 12-14, the shre of dministrtive costs nd of prtner nd in-kind services vry substntilly cross progrms nd hs positive correltion with per-fmily monthly progrm costs for PBTH. A positive reltionship is lso observed between supportive services shre nd both Exhibit Per-Fmily Monthly Progrm Costs for PBTH Progrms Supportive services costs Housing costs 24 PBTH progrms in cost nlysis $0 $1,000 $2,000 $3,000 $4,000 Per-fmily monthly progrm cost $5,000 $6,000 $7,000 PBTH = project-bsed trnsitionl housing. Note: Site nmes re suppressed to preserve progrm nonymity. Source: Fmily Options Study cost dt Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 146

179 Chpter 12. Intervention Costs Exhibit Administrtive (pnel A) nd Prtner/In Kind (pnel B) Shre of Progrm Costs Pnel A $7,000 $6,000 Per-fmily monthly progrm cost $5,000 $4,000 $3,000 $2,000 $1,000 $0 0% 5% 10% 15% 20% 25% Administrtive shre of costs 30% 35% 40% 45% Pnel B $7,000 $6,000 Per-fmily monthly progrm cost $5,000 $4,000 $3,000 $2,000 $1,000 $0 0% 5% 10% 15% 20% 25% 30% 35% 40% 45% 50% Prtner/in kind shre of costs Note: The fitted line hs been dded to this sctter plot to mke the positive correltions obvious. Source: Fmily Options Study cost dt Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 147

180 Chpter 12. Intervention Costs prtner nd in-kind shre nd dministrtive cost shre. So, in generl, progrms with more supportive services require greter dministrtive overhed, perhps becuse they more often leverge prtner orgniztions nd dontions to provide incresed supportive services. Looking t the underlying dt, no cler ssocition is pprent between the number of fmilies served t time by the progrm nd verge cost per household, dministrtive cost shre, or in-kind shre. This finding suggests tht cost vrition for PBTH progrms is lrgely driven by the type nd mount of housing nd supportive services provided rther thn by the number of households served. PBTH progrms provide time-limited ssistnce. The limits re typiclly longer thn in CBRR progrms nd shorter thn the period of time fmilies typiclly continue to qulify for nd use vouchers. Durtion of ssistnce within the followup period fter rndom ssignment verged 12 months for fm - ilies who were ssigned to nd used PBTH, nerly 4 months less thn the verge durtion for fmilies in SUB during the sme period. Mny PBTH progrms enroll fmilies for longer thn 18 months nd, indeed, 40 percent of study fmilies who used PBTH were still enrolled in the progrms s of the followup survey. Becuse monthly per-fmily costs of PBTH re more thn double monthly per-fmily costs for SUB, the verge progrm cost per sty during the followup period for PBTH of more thn $32,500 is substntilly more thn tht of SUB during the sme period Cost of Emergency Shelter Emergency shelters hve the highest verge cost per fmily per month more thn $4,800. The emergency shelter progrms lso hd the gretest vrition cross progrms in totl cost, in the importnce of prtner services nd inkind dontions to progrm ctivities, nd in the shre represented by dministrtive nd overhed costs. Exhibit reports overview sttistics for emergency shelter progrms. Exhibit shows the wide vrition cross progrms in both totl verge per-fmily monthly progrm cost nd in the shelter nd supportive services shres of costs. As with PBTH progrms, differences for the shelter component of costs hve little reltion to FMR differences cross sites, but insted re determined by vrition in shelter structure (for exmple, individul units versus congregte beds) nd loction nd by vritions in fcility-relted stffing nd mintennce costs. Although dministrtive cost shres rnge from 3 to 38 percent of totl costs, only smll positive correltion is seen between the dministrtive cost shre nd totl cost. Emergency shelters lso differ gretly in the importnce of prtner nd in-kind costs to the progrm structure. For onethird of progrms, prtner nd in-kind dontions represent t lest 20 percent of totl progrm costs. Exhibit shows the positive reltionship between prtner nd in-kind cost shre nd totl verge monthly cost per fmily. Prtner services nd in-kind dontions re nerly lwys supportive services (including direct ssistnce), but the study tem lso observed prtner services nd in-kind dontions in dministrtive (for exmple, ccounting services) nd fcility opertions (for exmple, hndymn services) tht were ppropritely ctegorized s dministrtive nd housing or shelter costs, respectively. Prtner nd donted supportive services rnged from medicl services nd counseling to summer cmp nd therpy niml sessions. In-kind goods provided to fmilies in shelters rnged from food nd clothes to concert tickets. The study tem used the durtion (fter rndom ssignment) of fmilies ssigned to UC s initil shelter sty to clculte progrm cost per sty during the followup sty for emergency shelter progrms. This initil sty verged 4 months fter rndom ssignment. Exhibit Emergency Shelter Cost Summry Sttistics Averge Across Progrms Rnge Across Progrms ES per-fmily monthly progrm cost ($) 4,819 1,888 to 9,170 ES progrm cost per sty fter rndom ssignment for the initil shelter sty ($) 16,829 4,366 to 51,637 Durtion in ES in initil shelter sty (fter rndom ssignment) (months) to 9.0 (cross sites) Shelter shre (%) to 96 Supportive services shre (%) to 4 Prtner nd in-kind shre (included in shelter or supportive services cost s relevnt) (%) 15 0 to 59 Administrtive nd overhed cost shre (included in both housing nd supportive services costs) (%) 16 3 to 38 ES = emergency shelter. Durtions in ES progrms in this tble re clculted for fmilies ssigned to usul cre. Notes: Averges re weighted bsed on the number of study fmilies enrolled in ech ES progrm clculted from study enrollment dt. Housing nd supportive services shres dd to 100; prtner nd in-kind shre nd dministrtive shre re including in housing nd supportive services. Sources: Fmily Options Study cost dt; Progrm Usge Dt Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 148

181 Chpter 12. Intervention Costs Exhibit Per-Fmily Monthly Progrm Costs for Emergency Shelter Progrms Supportive services costs Shelter costs 45 ES progrms in cost study $0 $1,000 $2,000 $3,000 $4,000 Note: Progrm nmes re suppressed to preserve progrm nonymity. Source: Fmily Options Study cost dt $5,000 $6,000 $7,000 $8,000 $9,000 $10,000 Per-fmily monthly progrm cost Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 149

182 Chpter 12. Intervention Costs Exhibit Totl Monthly Costs nd Prtner/In-Kind Shre of Progrm Costs $10,000 $9,000 $8,000 Per-fmily monthly progrm cost $7,000 $6,000 $5,000 $4,000 $3,000 $2,000 $1,000 $0 0% 10% 20% 30% 40% 50% 60% 70% Prtner/in-kind shre of costs Note: The fitted line hs been dded to this sctter plot to mke the close positive correltion obvious. Source: Fmily Options Study cost dt 12.6 Comprison of Costs Across Progrm Types Differences in the nture of the CBRR, PBTH, SUB, nd emergency shelter progrms re reflected in the differences in verge costs cross the progrms. Exhibit reports summry sttistics for the four progrm types. CBRR nd SUB provide ssistnce for privte mrket housing. In both cses, but prticulrly for SUB, housing ssistnce is driven by locl housing mrket conditions, s mesured by FMR. Even though CBRR provides some supportive services in the form of housing plcement nd limited cse mngement ssistnce, CBRR costs re lower thn SUB costs on permonth verge bsis, becuse CBRR ssistnce is sometimes fixed mount tht is less thn the typicl HAPs provided by vouchers in the sme site, nd, in mny cses, the subsidy declines the longer the fmily is in CBRR. CBRR lso costs less thn the SUB per-progrm sty, becuse the ssistnce lsts, on verge, slightly more thn 7 months per fmily. By contrst, SUB provides rentl ssistnce up to the PHA pyment stndrds (tied to FMR) less the household s expected contribution (typiclly 30 percent of fmilies incomes every month) nd hs no time limit. The verge durtion of SUB for fmilies included in the SUB intervention ws 16 months during the followup period. PBTH nd emergency shelter re similr to ech other nd distinct from CBRR nd SUB in tht they provide mix of housing or shelter nd supportive services. In fct, mny PBTH nd emergency shelter progrms tht study tem members visited for cost dt collection re operted by the sme gency; in number of instnces, shelter nd PBTH progrms were distinguished only by length of sty, with fmilies in both progrms ccessing the sme supportive services nd living in the sme fcility. Other emergency shelters re distinct in providing congregte shelter or shred rooms for sleeping, wheres PBTH (nd SUB nd CBRR) lrgely provide fmilies with privte units. Prtner nd inkind resources represent greter shre of costs, on verge, for emergency shelter progrms thn for PBTH progrms. In generl, the study tem found tht PBTH relied more thn did emergency shelter progrms on prtner orgniztions to provide professionl services such s counseling services or mentoring, wheres emergency shelter progrms were more likely to use volunteer nd in-kind resources. PBTH progrms provided housing nd supportive services, on verge, for nerly 13.0 months during the followup period, wheres initil shelter stys persisted, on verge, for 3.5 months fter rndom ssignment. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 150

183 Chpter 12. Intervention Costs Exhibit Comprison of Cost Summry Sttistics Across Progrm Types SUB CBRR PBTH ES Housing or shelter shres (%) Supportive services shre (%) Prtner nd in-kind shre (included in shelter or supportive services cost s relevnt) (%) 0 Two sites: 1, Ten sites: 0 Administrtive nd overhed cost shre (included in both housing nd supportive services costs) (%) Per-fmily monthly progrm cost ($) 1, ,706 4,819 Durtion of ssistnce in study progrm during the followup period (months) Progrm cost per sty during the followup period ($) 18,821 6,578 32,557 16,829 CBRR = community-bsed rpid re-housing. ES = emergency shelter. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. Notes: Durtions in respective progrm types re clculted for ll fmilies who were ssigned to nd enrolled in SUB, CBRR, nd PBTH. For PBTH nd CBRR progrms, durtions re verges of site-level durtion verges, weighted by the number of fmilies enrolling in progrms in the cost study. ES durtion is bsed on initil length of shelter sty for fmilies ssigned to UC. Housing nd supportive services shres dd to 100; prtner nd in-kind shre nd dministrtive shre re included in housing nd supportive services. Sources: Fmily Options Study cost dt (CBRR, PBTH, nd ES); Fmily Options Study 18-month followup survey (CBRR nd PBTH); HUD Public nd Indin Housing Informtion Center records (SUB); Tennt Rentl Assistnce Certifiction System records (SUB); Finncil Dt Schedule records (SUB) Cost of All Progrm Use During the Followup Period by Fmilies in Ech Intervention Arm Thus fr, this chpter hs presented estimtes of verge perfmily monthly progrm costs bsed on dt collected nd observed for the progrms tht provided the three ctive study interventions nd emergency shelter. The chpter hs lso compred verge progrm costs per sty during the followup period, which djusts monthly costs with site-level verges of number of months spent in the respective progrm types by fmilies ssigned to ech intervention. This section now turns to estimtes of the cost of ll use of shelter nd housing ssistnce progrms during the followup period. These estimtes cn be thought of s the costs of chieving the reltive impcts of the interventions reported in Chpters 6 through 9. They tke ccount of the extent to which fmilies ssigned to ech intervention used tht type of progrm nd lso used different types of ssistnce. Assignment to receive priority ccess to prticulr type of progrm both incresed the rte t which fmilies used tht progrm nd ffected the rte t which fmilies used other types of shelter nd housing ssistnce progrms during the followup period. These estimtes use the per-fmily monthly progrm costs, together with the observed ptters of progrm usge reported in Chpters 6 through 9, to construct estimtes of totl costs of the mix of homeless or housing ssistnce progrms tht served study fmilies in ech of the intervention rms in the period between rndom ssignment nd the followup survey. Progrm use dt mesured the number of months ech fmily received seven types of homeless or housing ssistnce pro - grms. The seven types of progrms re: subsidy, rpid rehousing, trnsitionl housing, emergency shelter, permnent supportive housing, public housing, nd project-bsed housing ssistnce (project-bsed vouchers or Section 8 projects). (See Chpters 6 through 9 for more informtion on progrm use, prticulrly the respective Progrm Use Since Rndom Assignment exhibits.) Trnslting the number of months of ssistnce received into the cost of ll housing nd service progrms used since rndom ssignment requires dditionl ssumptions. This requirement is minly becuse mny fmilies ccessed shelter or housing nd relted services from progrms not in the cost study. We used the following ssumptions. First, we used site-level per-fmily monthly progrm costs for ech of our four progrm types s cost estimtes for month of ssistnce t ny progrm of tht type. This pproch trets, for exmple, ll trnsitionl housing progrms in site s hving the sme per-fmily monthly progrm cost s the PBTH cost we estimte in this chpter. Second, the study dt trck fmilies use of permnent sup - portive housing, public housing, nd project-bsed housing ssistnce tht ws not ssocited with the study or included in the cost nlysis. Under the ssumption tht they hve similr progrm nd cost structures, the estimtes reported in this section use site-level PBTH costs s proxy for the cost of permnent supportive housing nd SUB costs s proxy for the costs of public housing nd project-bsed housing ssistnce. The study tem exmined these costs of ll progrm use ssocited with the combintion of ssistnce tht the fmilies received for ech of the six pirwise comprisons (see Exhibit 1-1 for n overview of the pirwise comprisons): SUB versus UC. CBRR versus UC. PBTH versus UC. SUB versus CBRR. SUB versus PBTH. CBRR versus PBTH. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 151

184 Chpter 12. Intervention Costs As reported in Chpters 6 through 9, different sets of fmilies took prt in ech of the study s pirwise comprisons. Exhibit presents the verge cost of ll progrm use during the followup period for the fmilies in ech ssignment rm of ech comprison. Looking cross ll the pirwise comprisons, whenever fmilies re ssigned to SUB or PBTH, costs of SUB or PBTH interventions represent the highest shre of ll progrm costs. Substntil emergency shelter costs remin even in these instnces, however, nd costs of emergency shelter represent the highest cost Exhibit Cost of Progrm Use Since Rndom Assignment for Ech Intervention Contrst Pnel A $35,000 Other SUB CBRR PBTH SUB vs. UC CBRR vs. UC PBTH vs. UC ES Cost of progrm use since rndom ssignment $30,000 $25,000 $20,000 $15,000 $10,000 $5,000 $30,832 $30,336 $27,605 $30,629 $30,817 $28,295 $0 SUB N = 530 UC N = 415 CBRR N = 455 UC N = 451 PBTH N = 294 UC N = 262 Assigned intervention Pnel B $35,000 SUB vs. CBRR SUB vs. PBTH CBRR vs. PBTH Cost of progrm use since rndom ssignment $30,000 $25,000 $20,000 $15,000 $10,000 $5,000 $31,158 $29,680 $27,864 $30,914 $22,524 $30,510 $0 SUB N = 381 CBRR N = 308 SUB N = 230 PBTH N = 187 CBRR N = 179 PBTH N = 197 Assigned intervention CBRR = community-bsed rpid re-housing. ES = emergency shelter. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. Notes: Averges re for ll 18-month survey respondents in ech rm of ech pirwise comprison, nd re weighted for survey nonresponse to represent full comprison smple. Cost estimtes ssume site-specific verge cost per month bsed on the Fmily Options Study cost dt nd HUD dministrtive dt. The other ctegory includes permnent supportive housing, public housing, nd project-bsed ssistnce (project-bsed vouchers or Section 8 projects). Sources: Fmily Options Study cost dt; HUD Public nd Indin Housing Informtion Center records (SUB); Tennt Rentl Assistnce Certifiction System records (SUB); Finncil Dt Schedule records (SUB); Fmily Options Study Progrm Usge dt Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 152

185 Chpter 12. Intervention Costs shre whenever fmilies re not ssigned to SUB or PBTH. This finding illustrtes tht, when fmilies hve been in shelter for 7 dys or more, substntil shelter costs re ssocited with ssisting ll fmilies up until the time they leve the emergency shelter either to use progrm to which they were given priority ccess or to go somewhere else. Totl costs for the verge fmily in the SUB-versus-UC pirwise comprison group re shown in the fr left set of stcked br chrts in Pnel A of Exhibit During the period between rndom ssignment nd the followup survey, costs of homeless nd housing ssistnce progrms provided to UC prticipnts were similr to costs for ssistnce to SUB prticipnts. When we include the cost of ll progrm use for ech set of fmilies fmilies ssigned to SUB versus fmilies ssigned to UC the totl verge cost per fmily ssigned to SUB is only $500 more thn for fmilies ssigned to UC. The decresed use of progrms providing PBTH, CBRR, nd emergency shelter outweighs most of the incresed use of SUB. In other comprisons, fmilies ssigned to SUB hve slightly higher verge costs thn fmilies ssigned to CBRR, nd they clerly hve lower costs thn fmilies ssigned to PBTH (see Pnel B of Exhibit 12-19). In the three contrsts contining the PBTH intervention, the high monthly cost of PBTH progrms results in higher verge cost of ll progrms used for fmilies ssigned to PBTH compred with other interventions. On the other hnd, in ech of the three comprisons involving the CBRR intervention, fmilies ssigned to CBRR hve the lower verge cost of ll progrms used, reflected the lower monthly cost of the CBRR intervention. For CBRR versus UC, CBRR fmilies hve n verge cost of ll progrms used tht is $3,000 lower tht those ssigned to UC, while for SUB versus CBRR, CBRR fmilies verge cost of ll progrms used is $1,500 lower thn those ssigned to SUB Monthly Cost of All Progrm Use t the Followup Survey by Fmilies in Ech Intervention Arm Exhibit shows the monthly costs of ll progrm use t the followup survey for ech pirwise comprison. This nlysis uses the per-fmily monthly progrm cost for ech type of progrm nd informtion bout the mix of progrm types fmilies were using t the time of the followup survey. As discussed in Chpters 6 through 9, the mix of progrms used during the month of the followup survey is different thn wht is observed during the entire followup period. For exmple, in the SUB-versus-UC comprison, 84 percent of fmilies ssigned to SUB used SUB t some point during the followup period nd 75 percent were using SUB t the time of the survey (see Exhibit 6-2). Of the fmilies ssigned to SUB, 13 percent used rpid re-housing during the followup period, but none were using rpid re-housing during the month of the survey. Altogether, 6 percent of fmilies ssigned to SUB used trnsitionl housing t some time during the followup period, but only 1 percent of fmilies were using trnsitionl housing t the time of the survey. Among fmilies ssigned to UC, 12 percent used SUB during the followup period nd 11 percent were using SUB in the month of the survey; 20 percent used rpid re-housing during the followup period but only 3 percent used rpid re-housing during the month of the survey. Similrly, 21 percent of fmilies ssigned to UC used trnsitionl housing during the followup period but only 8 percent were using this type of ssistnce during the month of the survey. Do the reltive costs of progrms used by fmilies in ech intervention rm lso differ when we consider only the month of the followup survey? Exhibit shows tht in contrsts involving SUB nd CBRR, the reltive costs of the progrm use ssocited with ech intervention rm re similr for the entire followup period nd in the month of the followup survey. In the month of the followup survey, costs of progrm use for fmilies ssigned to CBRR re lower thn for fmilies ssigned to SUB (by $102), UC (by $203), nd PBTH (by $271). This pttern is consistent with comprisons of the costs of progrm use during the entire followup period, in which fmilies ssigned to CBRR hd lower totl costs of ll progrm use thn did fmilies ssigned to ny of the other interventions. This find - ing reflects the greter use of nd the lower per-fmily monthly progrm cost of CBRR compred with other interventions. In ll contrsts involving SUB, the costs of progrm use during the month of the followup survey re slightly higher for fmilies ssigned to SUB. Compred with fmilies ssigned to UC, costs of progrm use for fmilies ssigned to SUB were $20 higher in the month of the followup survey. The costs of progrm use for fmilies ssigned to SUB were $88 higher thn for fmilies ssigned to PBTH nd were $102 higher thn for fmilies ssigned to CBRR. These differences in costs re reltively smll becuse the costs ssocited with the greter use of SUB during the month of the followup survey were offset by the use of the higher cost emergency shelter nd PBTH mong fmilies ssigned to the other interventions. For the PBTH-versus-UC nd SUB-versus-PBTH comprisons, the reltive costs for the followup survey month re different thn for the entire followup period. For exmple, in the PBTHversus-UC comprison, the costs of progrm use during the full followup period re $2,432 higher for fmilies ssigned Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 153

186 Chpter 12. Intervention Costs Exhibit Averge Per-Fmily Monthly Costs for Progrm Use t Time of the Followup Survey, by Comprison Pnel A $1,200 Other SUB CBRR PBTH SUB vs. UC CBRR vs. UC PBTH vs. UC ES $1,000 $1,086 $1,066 $1,098 $1,009 $1,012 Cost of progrm use in survey month $800 $600 $400 $895 $200 $0 SUB N = 530 UC N = 415 CBRR N = 455 UC N = 451 PBTH N = 294 UC N = 262 Assigned intervention Pnel B $1,200 SUB vs. CBRR SUB vs. PBTH CBRR vs. PBTH $1,081 $1,065 $1,000 $979 $977 $989 Cost of progrm use in survey month $800 $600 $400 $200 $718 $0 SUB N = 381 CBRR N = 308 SUB N = 230 PBTH N = 187 CBRR N = 179 PBTH N = 197 Assigned intervention CBRR = community-bsed rpid re-housing. ES = emergency shelter. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. Notes: Averges re for ll 18-month survey respondents in ech rm of ech pirwise comprison nd re weighted for survey nonresponse to represent full comprison smple. Cost estimtes ssume site-specific verge cost per month bsed on the Fmily Options Study cost dt nd HUD dministrtive dt. The other ctegory includes permnent supportive housing, public housing, nd project-bsed ssistnce (project-bsed vouchers or Section 8 projects). Sources: Fmily Options Study cost dt; HUD Public nd Indin Housing Informtion Center records (SUB); Tennt Rentl Assistnce Certifiction System records (SUB); Finncil Dt Schedule records (SUB); Fmily Options Study Progrm Usge Dt Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 154

187 Chpter 12. Intervention Costs to PBTH but re nerly equivlent in the month of the fol - lowup survey. Compred with SUB, the costs of progrm use for fmilies ssigned to PBTH re $3,063 higher during the entire followup period but re $88 less in the month of the followup survey. Compred with CBRR, fmilies ssigned to PBTH hve higher costs of progrm use in the month of the survey nd during the entire followup period. It is not cler how expected future costs of homeless or housing ssistnce will compre cross the interventions. Importntly, the SUB intervention usully lsts beyond the followup period reported here. Costs of providing SUB re indefinite nd will likely continue to grow t nerly the permonth cost. Fmilies ssigned to UC or the other interventions, however, my experience greter housing instbility thn their counterprts ssigned to SUB. This instbility could result in higher future costs from subsequent use of reltively more expensive shelter nd trnsitionl housing progrms. The costs of progrm use during longer (36-month) period will be exmined in the 36-month report for the Fmily Options Study long with the impcts mesured during the longer period. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 155

188 CHAPTER 13. CONCLUSIONS HUD lunched the Fmily Options Study in 2008 to fill gp in knowledge bout which housing nd services interventions work best for fmilies experiencing homelessness. This report provides the first rigorous evidence bout the reltive effects of priority ccess to permnent housing subsidies (SUB), community-bsed rpid re-housing (CBRR), project-bsed trnsitionl housing (PBTH), nd usul cre (UC) tht is, leving fmilies to find their wy out of shelter without priority ccess to one of the three ctive interventions. Nerly 2,300 fmilies in 12 sites cross the country were rndomly ssigned to one of these four tretment rms fter spending t lest 7 dys in emergency shelter. Rndom ssignment yielded well-mtched groups of fmilies, with no systemtic differences in bseline chrcteristics. Fmilies were free to tke up their ssigned interventions or mke other rrngements on their own, so fmilies in ech tretment rm used mix of progrms. Nonetheless, the study generted substntil contrsts in progrm use during the followup period tht is, the progrm or set of progrms fmilies used ws influenced strongly by the intervention to which fmilies were rndomly ssigned. Rndom ssignment nd the subsequent contrsts in progrm use provide strong bsis for drwing conclusions bout the reltive impcts of the lterntive interventions on severl spects of fmily well-being. Wht do the findings from the study tell us bout these interventions nd their effects? Ech of the four sets of fmilies creted through rndom ssignment supply importnt informtion for policy, informtion summrized in turn in this chpter for the UC, SUB, CBRR, nd PBTH interventions. The most importnt lessons from the study emerge from comprisons between the interventions, which tell us how effective given pproch to homelessness ssistnce is by contrst with n lterntive pproch. Of prticulr interest is how the ctive interventions PBTH, CBRR, nd SUB tht offer priority ccess to prticulr forms of ssistnce compred with llowing fmilies to nvigte the UC system on their own. To interpret these comprisons, the chpter begins by ddressing how the Fmily Options Study, in its design nd in the progrm use ptterns tht emerged within tht design, informs policy Mening of Impct Comprisons The inherent strength of the experimentl reserch design employed in the Fmily Options Study is the ssurnce tht the groups tht re creted through the rndom ssignment process will be similr to ech other. Becuse it is not possible to ccount for, or to use sttisticl methods to control for, ll the vribility tht my exist mong individul fmilies, rndomly ssigning lrge number of fmilies to different interventions is the most certin wy to ensure tht the groups will be comprble. The Fmily Options Study tests for the impcts of three dif - ferent potentil emphses in federl or locl ssistnce policy to homeless fmilies; (1) Wht impct would priority ccess to project-bsed trnsitionl housing (the PBTH rm of the experiment) hve on fmilies in shelter who re not ble to resolve their episodes of homelessness quickly? (2) How does this impct compre with the impct of providing ccess to community-bsed rpid re-housing (the CBRR rm)? (3) How does this impct compre with the impct of providing ccess to permnent housing subsidy (the SUB rm)? In ech cse, the corresponding policy question is, wht impct would this policy emphsis hve on the outcomes of fmilies in shelter, reltive to usul cre or nother policy emphsis? The followup dt for study prticipnts tell us wht would hppen were ech of these wys of trgeting offers nd ccess pursued s federl or locl policy. The pirwise comprisons between ctive interventions show the impct of offering fmilies priority ccess to one intervention rther thn n - other. The dt lso llow for the comprison of ech of these options with current policies tht do not crete priority ccess to ny prticulr form of housing ssistnce (tht is, the UC rm). The pirwise comprisons between ctive intervention rms nd UC show the impct of referring fmily to specific type of progrm compred with the impct of letting fmilies pursue ssistnce on their own. The nlysis in this report mesures the impct of hving been offered prticulr intervention regrdless of whether the fmily involved ctully received the intervention. The findings reflect the rel wy in which the homeless ssistnce system intercts with fmilies, in tht fmilies re offered n Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 156

189 Chpter 13. Conclusions intervention rther thn mndted to ccept the ssistnce being offered. Whether fmilies prticipte in n ssigned progrm reflects the reltive desirbility nd ccessibility of the interventions for fmilies within the context of the other options they my choose to pursue on their own. As is shown in the report, substntil number of fmilies did not use the ctive intervention to which they were referred, nd some used other interventions. The full experimentl smple for given rm collectively shows how different forms of housing ssistnce re used when fmilies re given priority ccess to one prticulr progrm type while simultneously hving the freedom to use other forms of ssistnce vilble in their communities. Including ll the fmilies rndomly ssigned to UC similrly revels the rnge of progrms used when no priority ccess is provided. The progrms used by UC fmilies (including the interventions exmined in this study) exist in communities nd would ech continue to exist even with stronger federl or locl push for only one of them. Thus, the full-smple comprisons between rndom ssignment rms known s intention-to-tret (ITT) impct estimtes provide the best guide to policymkers in messy, complex world nd re reported here s the min study findings. All this sid, evidence of the effects of prticulr progrm type on fmilies who ctully use tht pproch (for exmple, the effect of CBRR on the fmilies who use CBRR) compred with equivlent fmilies who do not use the pproch would hve high vlue to the homeless ssistnce field. Such infor - mtion is importnt not becuse ny federl or locl policy ction could ctully crete such contrst for the popultion of ll shelter-housed fmilies, but becuse efforts to improve prticulr intervention model need to be bsed on knowledge of wht ctully prticipting in tht model does for fmilies s compred with not prticipting. Furthermore, n individul fmily s choice mong the options vilble in its community is best guided by hed-to-hed contrsts in the results to be expected if the fmily ctully prticipted in one progrm type versus nother. For these resons, n - lyses of the effect of tretment on the treted (using qusiexperimentl methods not structured exclusively round full-smple rndom ssignment comprisons) will be impor - tnt to consumers of the study s results. The evlution will seek to provide such informtion in future project reports prticulrly regrding the SUB nd CBRR interventions building off the core ITT nlyses nd policy informtion presented here but subject to sttisticl limittions in isolting the direct effects of prticiption in the experimentl dt Usul Cre (UC) Emergency shelters in this study were the entry points into homeless ssistnce in ech site. Fmilies rndomly ssigned to UC typiclly remined in emergency shelter nd sought whtever ssistnce ws vilble in the community. The experiences of UC fmilies reflect how the homeless services system works when fmilies in shelter re not given priority ccess to nother homeless or housing ssistnce progrm. The study provides vluble informtion bout wht types of ssistnce fmilies use without specil offers of ssistnce nd how fmilies in shelter for t lest 7 dys fre over time. UC fmilies (tht is, fmilies to whom rndom ssignment did not give priority ccess to ny ctive intervention) spent substntil periods of time in emergency shelter fter rndom ssignment. UC fmilies spent n verge of 4 months in emergency shelter fter rndom ssignment (nerly ll immeditely fter rndom ssignment). More thn one-hlf (53 percent) of UC fmilies spent 3 or fewer months in emergency shelter, 23 percent spent 4 to 6 months, nd 24 percent spent more thn 6 months in emergency shelter during the followup period. 147 Emergency shelters offered rnge of supportive services. The shelters provided rnge of supportive services in primrily congregte settings (dorms or other group living situtions). All the shelters offered comprehensive needs ssessments, cse mngement, supportive services, nd referrls to other progrms. Shelters in some instnces lso offered supportive services such s ccess to physicl helth cre, employment trining, child dvoccy, life skills trining, mentl helth cre, nd prenting services. UC fmilies prticipted in homeless nd housing ssis - tnce progrms t firly high rtes. Some fmilies ssigned to UC did not use ny other form of homeless or housing ssistnce besides shelters, but most did. In prticulr, 28 percent of UC fmilies ccessed some form of permnent subsidy, 25 percent received trnsitionl housing, nd 18 percent received rpid re-housing. UC fmilies were not fring well 20 months fter study enrollment. One-hlf of UC fmilies reported being homeless or doubled up in the 6 months before the survey or hd sty in shelter in the yer preceding followup dt collection. In months 7 through 18 fter rndom ssignment, 28 percent of UC fmilies styed in emergency shelter. In the 6 months before the survey, 15 percent of fmilies hd been seprted 147 Anlysis of progrm use nd cost of totl progrm use used dt during medin of 21 clendr months. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 157

190 Chpter 13. Conclusions from child who ws with the fmily in shelter t study outset, nd 4 percent hd children in foster cre. Fir or poor helth ws reported by 32 percent of UC fmily heds, nd 31 percent worked in the week before the followup survey. At the time of the survey, 15 percent reported lcohol depen - dence or substnce buse, nd 12 percent hd experienced intimte prtner violence in the pst 6 months. More thn one-third of fmilies were food insecure. Most children hd experienced school move since rndom ssignment. Monthly costs for emergency shelter were substntil. The study found tht the emergency shelter progrms used by the UC fmilies cost slightly more thn $4,800 per month per fmily. Of this totl, 63 percent were for supportive services. Altogether, costs of ll the homeless nd housing progrms nd ssocited services tht fmilies ssigned to the UC group ccessed whether in shelter or in ctive progrms were bout $30,000 during the followup period in the comprisons involving UC Permnent Housing Subsidy (SUB) In most cses, the fmilies ssigned to the SUB ctive interven - tion were given priority ccess to housing choice voucher, or HCV, nd they my hve been offered housing serch ssistnce (they were not offered ongoing socil services). This type of ssistnce is not generlly ccessible to fmilies while in emergency shelter unless they rech the top of witing lists for subsidies during tht period. Wht does the Fmily Options Study tell us bout permnent subsidies for homeless fmilies? When SUB is vilble to fmilies in shelter, they tke it up t high rtes nd continue to use it for sustined period. SUB progrms were the lest likely of the ctive interventions to exclude fmilies becuse of eligibility rules. For exmple, only 2 percent of fmilies in the study were not given the opportunity to be rndomly ssigned to SUB becuse of nswers to screening questions sked before rndom ssignment. Of the fmilies rndomly ssigned to SUB, however, 11 percent were subsequently found to be ineligible. Altogether, 84 percent of respondent fmilies ssigned to SUB used SUB t some point during the followup period, for n verge of 16 months. Smller numbers of fmilies ssigned to SUB used CBRR (13 percent) nd trnsitionl housing (6 percent), with some overlp mong the three groups. 148 Some fmilies ssigned to SUB used other forms of permnent subsidy to which they did not hve priority ccess (for exmple, public housing or permnent supportive housing [PSH]), bringing the totl who used ny form of permnent subsidy to 87 percent. Compred with CBRR, PBTH, nd UC, SUB cused striking improvements in housing stbility. Priority ccess to SUB reduced the incidence of subsequent stys in shelter or plces not ment for humn hbittion by one-hlf when compred with priority ccess to CBRR or PBTH or with UC lone. SUB lso led to notble improvements in other spects of housing stbility reltive to the other interventions, reducing the incidence of doubling up, subsequent emergency shelter stys, housing crowding, nd number of plces lived during the followup period. The benefits of SUB extended beyond housing stbility, especilly when compred with UC. The benefits of priority ccess to SUB extended beyond housing stbility, with re - ductions in child seprtions reltive to UC nd PBTH nd reductions in foster cre plcements reltive to UC. SUB lso reduced psychologicl distress reltive to UC nd PBTH nd reduced reported lcohol nd drug problems reltive to UC. SUB reduced intimte prtner violence when compred with UC or CBRR. Finlly, SUB lowered the number of schools ttended by focl children reltive to the ll the other interventions. SUB reduced lbor mrket enggement but improved food security nd reduced economic stress. Assignment to SUB led to reductions in employment during the followup period reltive to ll other interventions nd reduced employment in the week before the followup survey compred with ssignment to UC nd PBTH. For exmple, 61 percent of the UC group hd worked for py t some point fter rndom ssignment, but only 50 percent of the SUB group hd done so. Reltive to CBRR nd PBTH, SUB reduced current ernings nd reduced nnul fmily income in the yer before the survey. Compred with UC, SUB incresed the proportion of fmilies receiving Temporry Assistnce for Needy Fmilies, or TANF, nd the Supplementl Nutrition Assistnce Progrm (SNAP) nd reduced the proportion of fmilies with income from ernings in the month before the survey. Compred with PBTH nd UC, SUB led to improvements in food security. Fmilies ssigned to SUB reported less economic stress in the 6 months before the interview compred with reports from CBRR, PBTH, nd UC fmilies. 148 For exmple, the sme fmily my hve used PBTH nd SUB t different points during the followup period. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 158

191 Chpter 13. Conclusions The benefits of SUB were chieved t lower cost thn UC nd PBTH nd t only slightly greter cost thn CBRR. On verge, SUB progrms cost slightly less thn $1,200 per fmily per month, lower thn the corresponding monthly costs for emergency shelter nd PBTH but higher thn the monthly cost for CBRR. SUB fmilies used less emergency shelter nd PBTH during the followup period thn fmilies ssigned to those interventions. As result, totl costs of progrm use by fmilies ssigned to SUB were clerly less thn those of PBTH fmilies (by bout $3,100) nd only slightly higher thn those ssigned to UC (by $500). Totl cost of progrm use for fmilies ssigned to SUB ws $1,500 more thn for fmilies ssigned to CBRR. The nerly equivlent cost of progrm use for SUB s compred with UC during the followup period ws driven both by decresed time in reltively more expensive emergency shelter nd by decresed use of reltively more expensive PBTH progrms Community-Bsed Rpid Re-Housing (CBRR) The CBRR intervention offered short-term rentl ssistnce lsting up to 18 months (medin length of use ws 7 months) to rent privte-mrket housing. CBRR lso offered limited cse mngement services focused on housing nd selfsufficiency. CBRR typiclly received funding from the Home - lessness Prevention nd Rpid Re-Housing Progrm, or HPRP. Wht do the findings from this study tell us bout this intervention? Tkeup of CBRR ws reltively low. Of fmilies rndomly ssigned to CBRR, 60 percent used CBRR during the 20-month followup period, much lower thn the 84-percent tkeup rte for SUB in tht rndom ssignment rm. Fmilies ssigned to CBRR in some instnces lso used trnsitionl housing (19 percent) nd multiple forms of permnent sub sidies (23 percent cross ll types of permnent subsidy). Qulittive reserch suggested tht the short durtion or uncertinty bout the durtion of ssistnce mde some fmilies reluctnt to use CBRR ssistnce (Fisher et l., 2014). CBRR led to more rpid deprtures from emergency shelter thn UC. One of the gols of CBRR ws to promote more rpid exit from emergency shelter into housing. The impcts mesured by the study provide evidence tht this gol ws chieved, reltive to UC. Assignment to CBRR leds to slightly fster exit from emergency shelter thn ssignment to UC, with no significnt difference compred with shelter exits in SUB or PBTH. CBRR ws equivlent to UC nd less effective thn SUB in preventing subsequent stys in shelters nd plces not ment for humn hbittion nd in improving other spects of housing stbility. The study found tht priority ccess to CBRR ws equivlent to leving fmilies in UC nd substntilly less effective thn priority ccess to SUB in reducing subsequent stys in shelters or plces not ment for humn hbittion nd in improving other spects of housing stbility. Modestly worse results for CBRR thn for PBTH my result from the fct tht 22 percent of fmilies ssigned to PBTH were still in PBTH progrms t the time of the followup survey, but only 6 percent of CBRR fmilies were still receiving CBRR. Priority ccess to CBRR hd little impct on other outcomes compred with priority ccess to UC, with the exception of incresing fmily income, receipt of SNAP, nd food security. As noted previously, SUB hd slightly better outcomes thn CBRR regrding seprtion of spouses, intimte prtner violence, children s school outcomes, nd economic stress, but CBRR improved work effort, nnu - lized ernings, nd nnul fmily income compred with SUB. Perhps surprisingly given the dditionl services pro vided by PBTH, CBRR enhnced severl spects of dult well-being reltive to PBTH, reducing psychologicl distress, lcohol dependence or drug buse, nd the frequency of fir or poor helth. CBRR hs the lowest monthly cost of the ctive interven tions studied, nd totl costs during the 21 months fter rndom ssignment for those ssigned to CBRR were slightly lower thn those of fmilies ssigned to SUB. CBRR progrms hd lower per-fmily, per-month cost thn PBTH nd SUB, verging slightly less thn $900. Housing costs comprised, on verge, 72 percent of these costs. Totl costs of ll pro - grms used by prticipnts ssigned to the CBRR intervention during the followup period were on verge lower thn totl costs for ll the other interventions. CBRR totl costs were only 5 percent lower thn SUB, however, becuse the greter use of SUB progrms by fmilies in the SUB rndomiztion rm ws offset by the greter use of trnsitionl housing, PSH, nd emergency shelter progrms by CBRR fmilies. Costs of progrm use for fmilies ssigned to CBRR were $3,000 lower thn for UC, $8,000 lower thn for PBTH, nd $1,500 lower thn for SUB over 21 months Project-Bsed Trnsitionl Housing (PBTH) The PBTH intervention offered housing for up to 24 months, coupled with wide rry of socil services. The study focused primrily on housing provided in fcility-bsed settings (lthough some PBTH fmilies were referred to progrms with scttered-site units from which they were required to move when ssistnce ended). PBTH progrms offered Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 159

192 Chpter 13. Conclusions comprehensive cse mngement, referrl to outside providers, nd direct services in severl res nd lso offered ccess to employment trining, life skills, mentl helth cre, prenting skills, nd physicl helth cre. The scope of needs ddressed in PBTH progrms ws similr to tht of emergency shelters. During the followup period, 54 percent of respondent fmilies ssigned to PBTH used tht form of ssistnce for n verge of 12 months. The Fmily Options Study provides severl lessons bout the PBTH model. Tkeup of PBTH ws reltively low. PBTH providers were more selective thn either SUB or CBRR providers regrding the fmilies they would serve. Only 77 percent of fmilies considered for the study pssed the initil screening for PBTH tht took plce before rndom ssignment, nd 18 percent of those who pssed nd were ssigned to PBTH were subsequently screened out by the PBTH progrms s ineligible. Of the fmilies ssigned to PBTH, 54 percent used some form of trnsitionl housing during the followup period. This low level of tkeup reflects combintion of fmily choices nd progrm eligibility requirements, with some fmilies considered ineligible by the progrms to which they were ssigned nd some fmilies choosing not to use PBTH ssistnce. Qulittive interviews suggest tht the fixed loction of PBTH units my hve been brrier when the ssigned loction ws not close to fmilies schools, work, trnsporttion, nd support networks, or when fmilies perceived the progrms to be in bd neighborhoods (Fisher et l., 2014). Fmilies ssigned to PBTH lso used CBRR (10 percent) nd vrious forms of permnent subsidy (21 percent, including SUB nd other forms of permnent subsidy). PBTH reduced stys in emergency shelter nd on the street compred with UC but did not led to other effects. Compred with UC, priority ccess to PBTH reduced the proportion of fmilies who reported stys in shelters or plces not ment for humn hbittion in the 6 months before the survey. This finding my reflect the fct tht bout 22 percent of fmilies who were ssigned to nd who used PBTH were still in PBTH t the time of the followup survey. In four other domins, however, most indictors exmined revel equivlent results with or without ssignment to PBTH fter shelter (tht is, compred with UC ssignment). PBTH costs less thn shelters on per-fmily, per-month bsis, but totl costs for PBTH fmilies during the period fter rndom ssignment were the highest mong ll interventions, including UC. PBTH progrms cost slightly more thn $2,700 per fmily per month, with supportive services constituting 42 percent of these costs lower monthly cost thn emergency shelter but higher thn SUB nd CBRR. The cost of totl progrm use for PBTH fmilies during the followup period ws substntilly higher thn for UC (by $2,500), SUB (by $3,100), nd CBRR (by $8,000) Fmily Chllenges Fmilies prticipting in the Fmily Options Study experienced numerous housing brriers nd psychosocil chllenges. The study did not yield evidence during the 20-month period of followup tht ny of the interventions studied works comprtively better for fmilies who hve greter psychosocil chllenges or housing brriers thn for fmilies who fce fewer difficulties. Thus, the min study results on impcts cross ll fmilies provide the study s clerest guidnce for policy in the medium term Implictions for Theory In ddition to findings on the effects of priority ccess to the three ctive interventions (reltive to UC nd reltive to ech other) nd on intervention costs, the study is lso informtive bout the theories underlying the ctive interventions. This section drws out those implictions. Study findings lend support for the underlying theoreticl model for SUB. The striking impcts of SUB in reducing subsequent stys in shelters or plces not ment for humn hbittion provide support for the view tht homelessness is for mny fmilies housing ffordbility problem tht cn be remedied with permnent subsidies without specilized homeless-specific psychosocil services. The findings lso provide support for the theoreticl proposition tht resolving homelessness hs rditing impct, given the mesured impcts of SUB on fmily preservtion nd dult well-being compred with those of UC. The temporry housing subsidies of CBRR were not strong enough to improve housing stbility of fmilies given priority ccess to tht intervention in the followup period studied, lthough the study provides evidence tht fmilies ssigned to CBRR left shelter sooner thn fmilies who remined in UC. With no mrked improvement in housing stbility, CBRR hd little effect on other outcomes presumed to emnte from enhnced stbility. Few study findings to dte support the theoreticl model underlying PBTH. PBTH is intended to ddress psychosocil chllenges nd brriers to housing by providing socil services. The study does not provide evidence tht the intervention meets this gol. Although priority ccess to PBTH led to reductions in homelessness when compred with UC, it did not produce effects in other spects of fmily well-being. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 160

193 Chpter 13. Conclusions 13.8 Questions for Longer Term Followup The Fmily Options Study is continuing to follow fmilies through 36 months fter study enrollment. This dditionl wve of dt collection will provide informtion on number of importnt questions. The 36-month nlysis will ddress whether the types of outcomes tht re improved by rndom ssignment to SUB re dependent on contemporneous receipt of the housing ssistnce. Could effects fde if ssistnce ends? During the followup period, 84 percent of SUB fmilies hd used SUB. At the end of this period SUB use hd fllen to 75 percent, lthough 77 percent received some form of permnent subsidy t tht point. The 36-month nlysis will exmine whether fmilies retin different forms of permnent housing ssistnce nd their benefits during longer intervl. It is lso possible tht the reduced stress nd greter stbility observed for SUB fmilies t 20 months will yield dditionl benefits for dult nd child-well-being during the longer term tht re not yet evident. Reductions in work effort in the short term could persist or s observed in the Effects of Housing Vouchers on Welfre Fmilies study (Mills et l., 2006) my fde during the longer term. The 36-month nlysis will lso exmine whether the focus of PBTH on ddressing psychosocil chllenges nd enhncing skills leds to benefits during the longer term tht re not evident fter 20 months. The chllenges to dult well-being reltive to CBRR my be temporry, reflecting nxiety on the prt of PBTH fmilies tht benefits re coming to n end (or hve recently ended). When ll fmilies hve left the PBTH progrms to which they were given priority ccess, they my return to emergency shelter or experience other forms of housing instbility. As of 20 months fter rndom ssignment of fmilies who hd been in emergency shelter for t lest week, the two mjor dvntges of CBRR over other interventions re greter work effort reltive to SUB nd, considering the cost of ll progrm use during the followup period, lower cost reltive to PBTH, UC, nd SUB. Work effort could led fmilies to better economic outcomes in the future with rditing benefits for other outcomes. In ny cse, if CBRR continues to hve similr outcomes to UC in most domins, but t lower cost, tht finding will be importnt. The reltive costs of homeless ssistnce in the different interventions re likely to chnge over time. Prticiption in SUB often lsts beyond the period for which the study tem mesured impcts nd costs in this report. The continuing costs of SUB progrms my or my not be offset by cost sv ings through reductions in use of shelter nd other progrms. These future costs will be ddressed in the 36-month nlysis in conjunction with impcts mesured during the longer term Summry The Fmily Options Study s rndom ssignment design for mesuring intervention impcts is stronger design thn tht of other studies of interventions for homeless fmilies. As result, the Fmily Options Study provides importnt new informtion bout wht hppens to fmilies who experience homelessness in the bsence of ny specil offers of ssistnce nd bout the impct of priority ccess to three prticulr interventions, SUB, PBTH, nd CBRR. The experimentl design of the study nd the contrsts in progrm use during the followup period provide solid foundtion for estimting the impcts of enhncing ccess to different kinds of ssistnce. The study provides the first cler evidence bout these effects nd thus cn serve s solid bsis for future policy decisionmking. Approximtely 20 months fter entry into shelter nd rndom ssignment, fmilies ssigned to SUB pper to be doing better thn the fmilies ssigned to CBRR, PBTH, nd UC. The benefits of priority ccess to SUB hve been chieved t cost comprble with tht of UC, slightly higher costs thn CBRR, nd t substntilly lower cost thn PBTH. Compred with those ssigned to UC, the fmilies rndomly ssigned to SUB on verge hve hd fewer negtive experiences (stys in shelters or plces not ment for humn hbittion, doubling up, child seprtions, nd intimte prtner violence). SUB fmilies re lso somewht more likely to live in their own plce. Moreover, children in SUB fmilies move mong schools less, nd fmilies experience greter food security nd less economic stress. On the negtive side, heds of these fmilies exert less work effort. Fmilies given priority ccess to CBRR do bout s well s fmilies ssigned to UC but hve substntilly lower cost, minly becuse CBRR lowers the rte t which fmilies use costly trnsitionl housing progrms. PBTH is more costly nd t this point hs few dvntges over other progrms. Furthermore, no evidence indictes tht intervention impcts differ ccording to fmilies psychosocil chllenges or housing brriers whtever form of ctive ssistnce is prioritized. The 36-month followup nlysis will exmine whether these differences mong in - terventions continue to hold nd whether new differences emerge fter nother 16 months elpse. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 161

194 REFERENCES Adm, Emm K Beyond Qulity: Prentl nd Residentil Stbility nd Children s Adjustment, Current Directions in Psychologicl Science 13: Angrist, Joshu D., nd Jören-Steffen Pischke Mostly Hrmless Econometrics: An Empiricist s Compnion. Princeton, NJ: Princeton University Press. Bssuk, Ellen L., John C. Buckner, Lind F. Weinreb, Angel Browne, Shri S. Bssuk, Ree Dwson, nd Jennifer N. Perloff Homelessness in Femle-Heded Fmilies: Childhood nd Adult Risk nd Protective Fctors, Americn Journl of Public Helth 87 (2): Bssuk, Ellen L., Crmel J. DeCndi, Alexnder Tsertsvdze, nd Molly K. Richrd The Effectiveness of Housing Interventions nd Housing nd Service Interventions on Ending Fmily Homelessness: A Systemtic Review, Americn Journl of Orthopsychitry 84 (5): Bssuk, Ellen L., nd Stephnie Geller The Role of Housing nd Services in Ending Fmily Homelessness, Housing Policy Debte 17 (4): Bssuk, Ellen L., Jennifer N. Perloff, nd Ree Dwson Multiply Homeless Fmilies: The Insidious Impct of Violence, Housing Policy Debte 12 (2): Bssuk, Ellen L., Lenore L. Rubin, nd Alison M. Lurit Chrcteristics of Sheltered Homeless Fmilies, Americn Journl of Public Helth 76 (9): Bssuk, Ellen L., Lind F. Weinreb, John C. Buckner, Angel Browne, Amy Slomon, nd Shri S. Bssuk The Chrcteristics nd Needs of Sheltered Homeless nd Low-Income Housed Mothers, Americn Journl of the Americn Medicl Assocition 276 (8): Betty, Alexndr S Student Mobility: Exploring the Impct of Frequent Moves on Achivement: Summry of Work - shop. Wshington, DC: Ntionl Reserch Council nd Institute of Medicine, Committee on the Impct of Mobility nd Chnge on the Lives of Young Children, School, nd Neighborhoods. Bloom, Brbr, Lindsey I. Jones, nd Gulnur Freemn Summry Helth Sttistics for U.S. Children: Ntionl Helth Interview Survey, 2012, Ntionl Center for Helth Sttistics, Vitl Helth Stt 10 (258): Buckner, John C Understnding the Impct of Home lessness on Children: Chllenges nd Future Reserch Directions, Americn Behviorl Scientist 51: Buckner, John C., nd Ellen L. Bssuk Mentl Dis - orders nd Service Utiliztions Among Youths From Homeless nd Low-Income Housed Fmilies, Journl of Americn Acdemy of Child nd Adolescent Psychitry 36 (7): Buckner, John C., Ellen L. Bssuk, Lind Weinreb, nd Mrgret Brooks Homelessness nd Its Reltion to the Mentl Helth nd Behvior of Low-Income School Aged Children, Developmentl Psychology 35: Buckner, John C., Willim R. Berdslee, nd Ellen L. Bssuk Exposure to Violence nd Low-Income Children s Mentl Helth: Direct, Moderted, nd Medited Reltions, Americn Journl of Orthopsychitry 74: Burt, Mrth R Chrcteristics of Trnsitionl Housing for Homeless Fmilies: Finl Report. Prepred for the U.S. Deprtment of Housing nd Urbn Development, Office of Policy Development nd Reserch. Wshington, DC: Government Printing Office Life After Trnsitionl Housing for Homeless Fmilies. Prepred for the U.S. Deprtment of Housing nd Urbn Development, Office of Policy Development nd Reserch. Wshington, DC: Government Printing Office. Burt, Mrth R., Ludn Y. Aron, Toby Dougls, Jesse Vlente, Edgr Lee, nd Britt Iwen Homelessness: Progrms nd the People They Serve. Findings of the Ntionl Survey of Homeless Assistnce Providers nd Clients. Prepred for the Intergency Council on Homelessness. Wshington, DC: Government Printing Office. Burt, Mrth R., Ludn Y. Aron, nd Edgr Lee, with Jesse Vlente Helping Americ s Homeless: Emergency Shelter or Affordble Housing? Wshington, DC: Urbn Institute Press. Burt, Mrth R., Crol L. Person, nd Ann Elizbeth Mont - gomery Strtegies for Preventing Homelessness. Prepred for the U.S. Deprtment of Housing nd Urbn Development, Office of Policy Development nd Reserch. Wshington, DC: Government Printing Office. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 162

195 References Burt, Mrth R., Dve Pollck, Abby Soslnd, Kelly S. Mikelson, Elizbeth Drpp, Kristy Greenwlt, nd Ptrick Shrkey Evlution of Continuums of Cre for Homeless People. Wshington DC: Urbn Institute; Firfx, VA: ICF Consulting. Centers for Disese Control nd Prevention (CDC) Stte nd Locl Youth Risk Behvior Survey. hs_questionnire.pdf Ntionl Center for Helth Sttistics. Vitl Stts. VitlStts_Births.htm. Chen, Edith Why Socioeconomic Sttus Affects the Helth of Children: A Psychosocil Perspective, Current Directions in Psychologicl Science 13 (3): Cherpitel, Cheryl J A Brief Screening Instrument for Problem Drinking in the Emergency Room: The RAPS-4, Journl of Studies on Alcohol 61 (3): Coley, Rebekh Levine, Tm Leventhl, Alici Doyle Lynch, nd Meliss Kull Reltions Between Housing Chrcteristics nd the Well-Being of Low-Income Children nd Adolescents, Developmentl Psychology 49 (9): Cowl, Kristin, Mrybeth Shinn, Beth C. Weitzmn, Dniel Stojnovic, nd Lriss Lby Mother-Child Seprtions Among Homeless nd Housed Fmilies Receiving Public Assistnce in New York City, Americn Journl of Community Psychology 30 (5): Culhne, Dennis P The Qundries of Shelter Reform: An Apprisl of Efforts to Mnge Homelessness, Socil Service Review 66: Culhne, Dennis P., Stephen Metrux, nd Thoms Byrne A Prevention-Centered Approch to Homelessness Assistnce: A Prdigm Shift? Housing Policy Debte 21 (2): Culhne, Dennis P., Stephen Metrux, Jun Min Prk, Mrynne Schretzmn, nd Jesse Vlente Testing Typology of Fmily Homelessness Bsed on Ptterns of Public Shelter Utiliztion in Four U.S. Jurisdictions: Implictions for Policy nd Progrm Plnning, Housing Policy Debte 18 (1): Culhne, Dennis P., Wyne D. Prker, Brbr Poppe, Kennen S. Gross, nd Ezr Sykes Accountbility, Cost-Effectiveness, nd Progrm Performnce: Progress Since Developed for the Ntionl Symposium on Homelessness Reserch. Wshington, DC: Government Printing Office. Dhl, Ronld E., nd Alison G. Hrvey Sleep in Children nd Adolescents With Behviorl nd Emotionl Disorders, Sleep Medicine Clinics 2: Eggers, Fredrick J., nd Foud Moumen Anlysis of Trends in Household Composition Using Americn Housing Survey Dt. Prepred for the U.S. Deprtment of Housing nd Urbn Development, Office of Policy Development nd Reserch. Wshington, DC: Government Printing Office. Evns, Gry The Environment of Child Poverty, Americn Psychologist 55: Fntuzzo, John W., Whitney A. LeBoeuf, Chin-Chin Chen, Hether L. Rouse, nd Dennis P. Culhne The Unique nd Combined Effects of Homelessness nd School Mobility on the Eductionl Outcomes of Young Children, Eductionl Resercher 41 (9): Finkel, Meryl, Meghn Henry, Ntlie Mtthews, Brooke Spellmn, nd Dennis Culhne. Forthcoming. Evlution of the Rpid Re-housing Demonstrtion Progrm: Finl Report. Prepred for the U.S. Deprtment of Housing nd Urbn Development, Office of Policy Development nd Reserch. Wshington, DC: Government Printing Office. Fisher, Benjmin W., Lindsy S. Myberry, Mrybeth Shinn, nd Jill Khdduri Leving Homelessness Behind: Housing Decisions Among Fmilies Exiting Shelter, Housing Policy Debte 24 (2): Gle, Ktherine The Promise nd the Prctice of Rpid Rehousing. Housing/ServicesAndProgrms/Progrms/Homeless/ HomelessFmilies/Mterils.spx. Goodmn, Robert N The Strengths nd Difficulties Questionnire: A Reserch Note, Journl of Child Psychology nd Psychitry 38: Greene, Willim H Econometric Anlysis: Fifth Edition. Upper Sddle River, NJ: Prentice-Hll. Gubits, Dniel, Brooke Spellmn, Luren Dunton, Scott Brown, nd Michelle Wood Interim Report: Fmily Options Study. Prepred for the U.S. Deprtment of Housing nd Urbn Development, Office of Policy Development nd Reserch. Wshington, DC: Government Printing Office. Gubits, Dniel, Brooke Spellmn, Debi McInnis, Stephen Bell, nd Michelle Wood Fmily Options Study: Revised Dt Collection nd Anlysis Pln. Prepred for the U.S. Deprt - ment of Housing nd Urbn Development, Office of Policy Development nd Reserch. Wshington, DC: Government Printing Office. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 163

196 References Hyes, Mureen A., Megn Zonneville, nd Ellen L. Bssuk The SHIFT Study: Finl Report. Wlthm, MA: The Ntionl Center on Fmily Homelessness. Herbers, Jnelle E., J.J. Cutuli, Lur M. Supkoff, Dvid Heistd, Chi-Keug Chn, Elizbeth Hinz, nd Ann S. Msten Erly Reding Skills nd Acdemic Achievement Trjectories of Students Fcing Poverty, Homelessness nd High Residentil Mobility, Eductionl Resercher 41 (9): Huber, Peter J The Behvior of Mximum Likelihood Estimtes Under Nonstndrd Conditions. In Proceedings of the Fifth Berkeley Symposium on Mthemticl Sttistics nd Probbility, edited by Lucien M. Le Cm, Jerzy Neymn, nd Elizbeth L. Scott. Berkeley, CA: University of Cliforni Press, Vol. 1: Huntington, Nichols, John C. Buckner, nd Ellen L. Bssuk Adpttion in Homeless Children: An Empiricl Exmintion Using Cluster Anlysis, Americn Behviorl Scientist 51 (6): Institute for Children nd Poverty Ntionl Survey of Progrms nd Services for Homeless Fmilies. Jcob, Brin, Mx Kpustin, nd Jens Ludwig Humn Cpitl Effects of Anti-Poverty Progrms: Evidence From Rndomized Housing Voucher Lottery. NBER Working Pper Jcob, Brin A., nd Jens Ludwig The Effects of Housing Assistnce on Lbor Supply: Evidence from Vou - cher Lottery, Americn Economic Review 102 (1): Kessler, Ronld C., Peggy R. Brker, Lis J. Colpe, Joel F. Epstein, Joe C. Gfroerer, Ev Hiripi, Mry J. Howes, Shron- Lise T. Normnd, Ronld W. Mnderscheid, Elen E. Wlters, nd Aln M. Zslvsky Screening for Serious Mentl Illness in the Generl Popultion, Archives of Generl Psychitry 60 (2): Khdduri, Jill Housing Vouchers Are Criticl for Ending Fmily Homelessness. Wshington, DC: Homelessness Reserch Institute. Lyzer, Crolyn Personl communiction. Little, Roderick J.A Survey Non-Response Adjustments for Estimtes of Mens, Interntionl Sttisticl Review 54 (2): Locke, Gretchen, Jill Khdduri, nd Ann O Hr Housing Models. In Towrd Ending Homelessness: The 2007 Ntionl Symposium on Homelessness Reserch, edited by Gretchen Locke, Jill Khdduri, nd Ann O Hr. Wshington, DC: U.S. Deprtment of Housing nd Urbn Development, Office of Policy Development nd Reserch. Lowenstein, Amy E., Noemi Altmn, Ptrici M. Chou, Kristen Fucett, Adm Greeney, Dniel Gubits, Jorgen Hrris, JoAnn Hsueg, Erik Lundquist, Chrles Michlopoulos, nd Vinh Q. Nguyen A Fmily-Strengthening Progrm for Low-Income Fmilies: Finl Impcts from the Supportive Helthy Mrrige Evlution, Technicl Supplement. OPRE Report B. Wshington, DC: U.S. Deprtment of Helth nd Humn Services, Administrtion of Children nd Fmilies, Office of Plnning, Reserch nd Evlution. Msten, Ann S., Jnette E. Herbers, Christopher D. Desjrdins, J.J. Cutuli, Christopher M. McCormick, Julinn K. Spienz, Jeffrey D. Long, nd Philip D. Zelzo Executive Func - tion Skills nd School Success in Young Children Experiencing Homelessness, Eductionl Resercher 41 (9): Msten, Ann S., Donn Miliotis, Sndr A. Grhm-Bermnn, MryLouise Rmirez, nd Jennifer Neemnn Children in Homeless Fmilies: Risks to Mentl Helth nd Development, Journl of Consulting nd Clinicl Psychology 61: Mtsudir, Jordn D., nd Rebecc M. Blnk The Impct of Ernings Disregrds on the Behvior of Low- Income Fmilies, Journl of Policy Anlysis nd Mngement 33 (1): Myberry, Lindsy S., Mrybeth Shinn, Jessic Gibbons Benton, nd Jsmine Wise Fmilies Experiencing Housing Instbility: The Effects of Housing Progrms on Fmily Routines nd Rituls, Americn Journl of Orthopsychitry 84 (1): Myfield, Jim, Cllie Blck, nd Brbr E.M. Felver Employment Outcomes Associted With Rpid Re-Housing Assistnce for Homeless DSHS Clients in Wshington Stte. Settle: Wshington Stte Deprtment of Socil nd Helth Services. McGrew, Kevin S., Fredrick A. Shrnk, nd Richrd W. Woodcock Technicl Mnul. Woodcock-Johnson Normtive Updte. Rolling Medows, IL: Riverside Publishing. McLoyd, Vonnie. C The Impct of Economic Hrdship on Blck Fmilies nd Children: Psychologicl Distress, Prenting, nd Socioemotionl Development, Child Development 61 (2): Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 164

197 References Socioeconomic Disdvntge nd Child Development, Americn Psychologist 53: Mehn, Mjid, nd Arthur J. Reynolds School Mobility nd Achievement: A Met-Anlysis, Children nd Youth Services Review 26: Miller, Porti, Elizbeth Votrub-Drzl, nd Clude M. Setodji Fmily Income nd Erly Achievement Across the Urbn-Rurl Continuum, Developmentl Psychology 49 (8): Mills, Gregory, Dniel Gubits, Lrry Orr, Dvid Long, Judith Feins, Bubul Kul, Michelle Wood, Amy Jones, Cloudburst Consulting, nd the QED Group Effects of Housing Vouchers on Welfre Fmilies: Finl Report. Prepred for the U.S. Deprtment of Housing nd Urbn Development, Office of Policy Development nd Reserch. Wshington, DC: Government Printing Office. Ntionl Allince to End Homelessness (NAEH) Rpid Re-Housing Successes. librry/entry/rpid-re-housing-successes. Ntionl Comorbidity Survey Repliction (NCS-R) Ntionl Comorbidity Survey (NCS): Lifetime nd 12 Months Prevlence Estimtes from the NCS-R nd NCS-A. Prevlence_Estimtes.pdf. Nord, Mrk, Mrgret Andrews, nd Steven Crlsen Household Food Security in the United Sttes, ERS Food Assistnce nd Nutrition Report No. ERR11. Wshington, DC: Government Printing Office. Northwest Institute for Children nd Fmilies Evlution of the Sound Fmilies Inititive: Finl Findings Summry: A Closer Look t Fmilies Lives During nd After Supportive Trnsitionl Housing. Settle: University of Wshington School of Socil Work. Obrdović, Jolen, Jeffrey D. Long, J.J. Cutuli, Chi-Keung Chn, Elizbeth Hinz, Dvid Heistd, nd Ann S. Msten Acdemic Achievement of Homeless nd Highly Mobile Children in n Urbn School District: Longitudinl Evidence on Risk, Growth, nd Resilience, Development nd Psychopthology 21 (02): O Flherty, Brendn Homelessness As Bd Luck: Implic - tions for Reserch nd Policy. New York: Columbi Universtiy. Orr, Lrry L Socil Experiments: Evluting Public Pro - grms With Experimentl Methods. Thousnd Oks, CA: SAGE Publictions. Ppy, Jmes, nd John Hedl, Jr Pyschometric Chr - cteristics nd Norms for Disdvntged Third nd Fourth Grde Children on Stte-Trit Anxiety Inventory for Children, Journl of Abnorml Child Psychology 6 (1): Prk, Jung M., Stephen Metrux, Gbriel Brodbr, nd Dennis P. Culhne Child Welfre Involvement Among Children in Homeless Fmilies, Child Welfre 83: Ponitz, Clire E. Cmeron, Megn E. McClellnd, Abigil M. Jewkes, Crol McDonld Connor, Crrie L. Frris, nd Fredrick J. Morrison Touch Your Toes! Developing Direct Mesure of Behviorl Regultion in Erly Childhood, Erly Childhood Reserch Qurterly 23: Pribesh, Shn, nd Dougls B. Downey Why Are Residentil nd School Moves Associted With Poor School Performnce? Demogrphy 36 (4): Pum, Michel J., Robert B. Olsen, Stephen H. Bell, nd Cristofer Price Wht To Do When Dt Are Missing in Group Rndomized Controlled Trils. NCEE Wshington, DC: U.S. Deprtment of Eduction, Institute of Eduction Sciences, Ntionl Center for Eduction Evlution nd Regionl Assistnce. Rfferty, Yvonne, nd Mrybeth Shinn The Impct of Homelessness on Children, Americn Psychologist 46: Rfferty, Yvonne, Mrybeth Shinn, nd Beth C. Weitzmn Acdemic Achievement Among Formerly Homeless Adolescents nd Their Continuously Housed Peers, Journl of School Psychology 42 (3): Rmirez, Mrylouise, Ann S. Msten, nd D.M. Sms Fers in Homeless Children. Pper presented t the Biennil Meeting for the Society for Reserch in Child Development, Settle. Rodriguez, Json Homelessness Recurrence in Georgi. Atlnt: Georgi Deprtment of Community Affirs, Stte Housing Trust Fund for the Homeless. Rog, Debr J., nd John C. Buckner Towrd Understnding Homelessness: The 2007 Ntionl Symposium on Homelessness Reserch. Wshington, DC: U.S. Deprtment of Helth nd Humn Services, Assistnt Secretry for Pln - ning nd Evlution; U.S. Deprtment of Housing nd Urbn Development, Office of Policy Development nd Reserch. Rog, Debr J., nd Mrjorie Gutmn The Homeless Fmilies Progrm: A Summry of Key Findings. In To Improve Helth nd Helth Cre: The Robert Wood Johnson Foundtion Anthology, edited by Stephen L. Iscs nd Jmes R. Knickmn. Sn Frncisco: Jossey-Bss Publishers: Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 165

198 References Rog, Debr J., nd Frnces L. Rndolph A Multisite Evlution of Supported Housing: Lessons Lerned From Cross-Site Collbortion, New Directions for Evlution 94: Schochet, Peter An Approch for Addressing the Multiple Testing Problem in Socil Policy Impct Evlutions, Evlution Review 33 (6): Shinn, Mrybeth, Andrew L. Greer, Jy Binbridge, Jonthn Kwon, nd Sr Zuiderveen Efficient Trgeting of Homelessness Prevention Services for Fmilies, Americn Journl of Public Helth 103 (S2): S324 S330. Shinn, Mrybeth, Judith S. Schteingrt, Nthniel Chioke Willims, Jennifer Crlin-Mthis, Nncy Bilo-Krgis, Rchel Becker-Klein, nd Beth C. Weitzmn Long-Term Associtions of Homelessness with Children s Well-Being, Americn Behviorl Scientist 51: Shinn, Mrybeth, Beth C. Weitzmn, Dniel Stojnovic, Jmes R. Knickmn, Lueil Jimenez, Lis Duchon, Susn Jmes, nd Dvid H. Krntz Predictors of Homelessness Among Fmilies in New York City: From Shelter Request to Housing Stbility, Americn Journl of Public Helth 88 (11): Skinner, Hrvey A The Drug Abuse Screening Test, Addictive Behvior 7 (4): Snyder, C. Richrd, Betsy Hoz, Willim E. Pelhm, Michel Rpoff, Lenne Wre, Michel Dnovsky, Lori Highberger, Howrd Ribenstein, nd Kndy J. Sthl The Develop - ment nd Vlidtion of the Children s Hope Scle, Journl of Peditric Psychology 22 (3): Snyder, C. Richrd, Lori M. Irving, nd John R. Anderson Hope nd Helth: Mesuring the Will nd the Wys. In The Hndbook of Socil nd Clinicl Psychology: The Helth Perspective, edited by C. Richrd Synder nd Donelson R. Forsyth. Elmsford, NY: Pergmon Press: Snyder, C. Richrd, Susie C. Sympson, Florence C. Ybsco, Tyrone F. Borders, Michel A. Bbyk, nd Rymond L. Higgins Development nd Vlidtion of the Stte Hope Scle, Journl of Personlity nd Socil Psychology 70: Snyder, Thoms D., nd Slly A. Dillow Digest of Eduction Sttistics NCES U.S. Deprtment of Eduction, Institute of Eduction Sciences, Ntionl Center for Eduction Sttistics. Wshington, DC: Government Printing Office. Spellmn, Brooke, Jill Khdduri, Brin Sokol, nd Joshu Leopold Costs Associted With First-Time Homelessness for Fmilies nd Individuls. Prepred for the U.S. Deprtment of Housing nd Urbn Development, Office of Policy Development nd Reserch. Wshington, DC: Government Printing Office. Spielberger, Chrles D., Richrd L. Gorsuch, Robert E. Lushene, Peter R. Vgg, nd G.A. Jcobs Stte-Trit Anxiety Inventory for Children. Plo Alto, CA: Mind Grden. Squires, Jne, nd Dine D. Bricker Ages & Stges Questionnires: A Prent-Completed Child Monitoring System, 3rd ed. Bltimore: Brookes Publishing. Steffen, Brry L., Shun Bucholtz, Mrge Mrtin, Dvid A. Vndenbroucke, nd Yunn-Gnn Dvid Yo Worst Cse Housing Needs 2011: Report to Congress. Prepred for the U.S. Deprtment of Housing nd Urbn Development, Office of Policy Development nd Reserch. Wshington, DC: Government Printing Office. Thompson, Dinne T Evluting Length of Sty in Assisted Housing Progrms: A Methodologicl Note, Cityscpe 9 (1): Tsi, Jck, Wesley Ksprow, nd Robert Rosenheck Exiting Homelessness Without Voucher: A Comprison of Independently Housed nd Other Homeless Veterns, Psychologicl Services 8 (2): U.S. Census Bureu Decennil Census Americn Community Survey 1-Yer Estimtes. dt_documenttion/dt_min/. U.S. Deprtment of Housing nd Urbn Development (HUD) Notice of Alloctions, Appliction Procedures, nd Requirements for Homelessness Prevention nd Rpid Re-Housing Progrm Grntees Under the Americn Recovery nd Reinvestment Act of hudexchnge.info/resource/1882/hprp-notice-june / Homelessness Prevention nd Rpid Re- Housing Progrm: Yer 1 Summry. info/resources/documents/hprp_yer1summry.pdf b Housing Inventory Chrt nd Homeless Popultion nd Subpopultion Dt. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 166

199 References The 2011 Annul Homeless Assessment Report (AHAR) to Congress. Wshington, DC: U.S. Deprt ment of Housing nd Urbn Development. hudexchnge.info/resource/1872/2010-nnul-homelessssessment-report-hr/ The 2012 Annul Homeless Assessment Report (AHAR) to Congress: Volume II. Wshington, DC: U.S. Deprtment of Housing nd Urbn Development estimtes-of-homelessness-in-the-us/ b. The 2013 Annul Homeless Assessment Report (AHAR) to Congress: Volume I. Wshington, DC: U.S. Deprtment of Housing nd Urbn Development. pit-estimtes-of-homelessness/ c. Homelessness Prevention nd Rpid Re- Housing Progrm (HPRP): Yer 2 Summry. onecpd.info/resources/documents/prp_yer2summry.pdf d. U.S. Rentl Housing Finnce Survey The 2013 Annul Homeless Assessment Report (AHAR) to Congress: Volume II. Wshington, DC: U.S. Deprt - ment of Housing nd Urbn Development b. Residentil Housing Finnce Survey. U.S. Deprtment of Housing nd Urbn Development, Office of Public nd Indin Housing (HUD-PIH) Form HUD Instruction Booklet. gov/hudportl/documents/huddoc?id=50058i.pdf. U.S. Intergency Council on Homelessness (USICH) Opening Doors: Federl Strtegic Pln To Prevent nd End Homelessness. spx?docid=1dc1e97f c37c02. Voight, Adm, Mrybeth Shinn, nd Mury Ntion The Longitudinl Effects of Residentil Mobility on the Acdemic Achievement of Urbn Elementry nd Middle School Students, Eductionl Resercher 41 (9): Wldfogel, Jne Interntionl Policies Towrd Prentl Leve nd Child Cre, The Future of Children 11 (1): Weinreb, Lind, Robert Goldberg, Ellen L. Bssuk, nd Jennifer Perloff Determinnts of Helth nd Service Use Ptterns in Homeless nd Low-Income Housed Children, Peditrics 102: 562. Weitzmn, Beth C Pregnncy nd Childbirth: Risk Fctors for Homelessness? Fmily Plnning Perspectives 21 (4): Weitzmn, Beth C., nd Crolyn Berry Formerly Homeless Fmilies nd the Trnsition to Permnent Housing: High-Risk Fmilies nd the Role of Intensive Cse Mngement Services. Finl Report to the Edn McConnell Clrk Foundtion. New York: New York University, Robert F. Wgner Grdute School of Public Service, Helth Reserch Progrm. White, Hlbert A Heteroskedsticity-Consistent Covrince Mtrix Estimtor nd Direct Test for Heteroskedsticity, Econometric 48: Asymptotic Theory for Econometricins. Orlndo, FL: Acdemic Press. Wong, Yin-Ling, Dennis P. Culhne, nd Rndll Kuhn Predictors of Exit nd Reentry Among Fmily Shelter Users in New York City, Socil Service Review 71 (3): Wood, Michelle, Jennifer Turnhm, nd Gregory Mills Housing Affordbility nd Fmily Well-Being: Results from the Housing Voucher Evlution, Housing Policy Debte 19 (2): Woodcock, R.W., K.S. McGrew, nd N. Mther Woodcock-Johnson III Tests of Achievement. Rolling Medows, IL: Riverside World Helth Orgniztion (WHO) Interntionl Sttisticl Clssifiction of Diseses nd Relted Helth Problems, 10th Revision Vol. 2. Instruction Mnul. pdf?u=1. Yoshikw, Hirokzv, J. Lwrence Aber, nd Willim R. Berdslee The Effects of Poverty on the Mentl, Emotionl, nd Behviorl Helth of Children nd Youth: Implictions for Prevention, Americn Psychologist 67: Youth in Mind Strengths nd Difficulties Questionnires. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 167

200 ADDITIONAL READING Allison, Pul D Missing Dt. University Pper No Thousnd Oks, CA: SAGE Publictions. Angrist, Joshu D., Guido W. Imbens, nd Donld B. Rubin Identifiction of Cusl Effects Using Instrumentl Vribles, Journl of the Americn Sttisticl Assocition 91: Angrist, Joshu D., nd Aln B. Krueger The Effect of Age t School Entry on Eductionl Attinment: An Ap - pliction of Instrumentl Vribles With Moments From Two Smples, Journl of the Americn Sttisticl Assocition 87 (418): Split-Smple Instrumentl Vribles Estimtes of the Return to Schooling, Journl of Business nd Economic Sttistics 13 (2): Burt, Mrth R., nd Brooke Spellmn Chnging Homeless nd Minstrem Service Systems: Essentil Approches to Ending Homelessness. Developed for the Ntionl Symposium on Homelessness Reserch. Wshington, DC: Government Printing Office Rog, Debr J., Kimberly McCombs-Thornton, Arin M. Gilbert-Mongelli, Consuelo Brito, nd C. Scott Holupk Implementtion of the Homeless Fmilies Progrm: 2. Chrcteristics, Strengths, nd Needs of Prticipnt Fmilies, Americn Journl of Orthopsychitry 65 (4): Shinn, Mrybeth Most Homeless Fmilies Just Need Affordble Housing. Wshington, DC: Ntionl Allince to End Homelessness. U.S. Intergency Council on Homelessness (USICH) Opening Doors: Homelessness Among Fmilies. homelessness_mong_fmilies_fct_sheet/. Westfll, Peter H., Rndll Tobis, nd Russell Wolfinger Multiple Comprisons nd Multiple Tests Using SAS. Cry, NC: SAS Institute. Yudko, Errol, Olg Lozhkin, nd Adrin Fouts A Comprehensive Review of the Pyschometric Properties of the Drug Abuse Screening Test, Journl of Substnce Abuse Tretment 32: Glzermn, Steven, Dn Levy, nd Dvid Myers Nonexperimentl Versus Experimentl Estimtes of Ernings Impcts. Unpublished pper submitted to the Annls of the Acdemy of Politicl nd Socil Sciences. Short-Term Impcts of Housing nd Services Interventions for Homeless Fmilies 168

201 Appendix A. Dt Sources nd Dtset Construction APPENDIX A. DATA SOURCES AND DATASET CONSTRUCTION This ppendix of the Fmily Options Study short-term impcts report describes the dt sources, dt collection procedures, response rtes, nd dt processing procedures used in the Fmily Options Study. The study enrolled 2,282 fmilies cross 12 sites between September 2010 nd Jnury Exhibit A-1 shows the timing of smple enrollment nd the enrollment numbers by site nd intervention group. Of the 2,282 fmilies who enrolled in the study, 1,857 completed the 18-month dult survey. Exhibit A-1. Smple Enrollment Period nd Number of Fmilies Enrolled by Intervention nd Site Fmilies Rndomly Assigned, by Intervention (N) Site Enrollment Period CBRR PBTH SUB UC Totl (N) Almed County Sep 2010 Jn Atlnt Oct 2010 Jn Bltimore Mr 2011 Jn Boston Feb 2011 Jn Connecticut* Oct 2010 Dec Denver Jn 2011 Jn Honolulu Oct 2010 Jn Knss City Oct 2010 Jn Louisville Apr 2011 Jn Minnepolis Nov 2010 Jn Phoenix Oct 2010 Dec Slt Lke City Sep 2010 Oct ,282 Totl CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. UC = usul cre. *Includes the cities of New Hven nd Bridgeport, Connecticut. Source: Rndom ssignment records A.1 Timing of Followup The period of observtion during which intervention effects were mesured is the dte of rndom ssignment to the 18 month followup survey for the 1,857 fmilies who completed n dult survey. A minimum of 18 months elpsed between rndom ssignment nd the followup survey for ll fmilies who completed the survey.1 Of ll survey completers, At lest 18 months pssed from the dte of rndom ssignment before the interviewer begn to contct fmilies to conduct the followup survey. Fmilies who were difficult to locte hd longer period of time elpse between the dtes of rndom ssignment nd the 18-month followup survey. A -1

202 Appendix A. Dt Sources nd Dtset Construction percent completed the survey within 4 months of initil contct for the 18-month followup survey. The nlysis period, during which ll impcts were estimted, ws thus between 18 nd 24 months fter rndom ssignment for most fmilies (Exhibit A-2). Exhibit A-2. Length of Time From Rndom Assignment to the 18-Month Followup Survey Durtion (months) From 18 to 19 Percent of Fmilies 7.4 From 19 to From 20 to From 21 to From 22 to From 23 to From 24 to or more 6.3 Medin: 20.4 months (612 dys) Men: 21.1 months (633 dys) Notes: N = 1,857. Percents re unweighted. Includes four fmilies who were interviewed between 17 months, 3 weeks nd 18 months. Source: Fmily Options Study 18-month followup dt A.2 Rndom Assignment Dt The study tem creted secure website to support the enrollment nd rndom ssignment of fmilies into the Fmily Options Study. Locl site interviewers used the rndom ssignment website to Document tht the dult respondent provided informed consent. Enter the personl identifiers for the dult respondent nd spouse/prtner, if pplicble. Check tht intervention providers hd openings vilble in their progrms, mking it possible to conduct rndom ssignment. Document tht the bseline survey ws complete. Rndomly ssign the fmily to vilble intervention groups. Gubits et l. (2013) provides further detils bout the enrollment process. Eligibility Determintion In n ttempt to mximize the likelihood tht fmilies ssigned to one of the interventions would be ccepted by the intervention provider, the study tem conducted screening before rndom ssignment. The study tem collected the eligibility requirements of ech provider nd used those requirements to develop eligibility screening questions tht were dministered fter informed consent but before rndom ssignment. The study tem sked the fmily only the eligibility screening questions relevnt to the intervention providers in tht site with openings vilble. A fmily ws eligible for rndom ssignment to n intervention if the dult respondent s nswers to the screener questions met the eligibility requirements for t lest one prticipting provider of tht intervention with n opening t the time of rndom ssignment. The screener questions improved the likelihood tht fmilies would be eligible for the ssigned intervention.2 The study tem retined dt on the eligibility screening responses. Gubits et l. (2013) provides further detils bout eligibility determintion. 2 After rndom ssignment to one of the study interventions nd referrl to n intervention provider, fmilies were required to complete the provider s regulr eligibility determintion process, including, for some progrms, criminl bckground checks, drug testing, nd income verifiction. A -2

203 Appendix A. Dt Sources nd Dtset Construction A.3 Bseline Dt As prt of the study enrollment process, ll fmilies who provided informed consent nd who were determined eligible for vilble interventions completed bseline survey. The bseline survey covered number of topics, including fmily composition, demogrphic chrcteristics, housing stbility, history of homelessness, employment, income, nd helth. The study tem collected bseline survey dt in the shelter where the fmily ws stying t the time of rndom ssignment. The study tem used Computer-Assisted Personl Interviewing (CAPI) technology to dminister the survey, which took n verge of 40 minutes per survey. The covrites, discussed in Appendix C.1, were derived from the bseline survey responses. Gubits et l. (2013) provides further detils bout the bseline survey nd dt collection. A.4 Enrollment Verifiction Dt The study tem collected informtion from prticipting providers to document enrollment in the ssigned intervention. The study tem contcted prticipting providers regulrly (weekly or monthly) to inquire bout the sttus of fmilies who hd been referred to their progrms. This informtion is referred to s the enrollment verifiction dt. The study tem collected the following informtion: Whether the study fmily mde contct with the provider to which they were referred. Whether the fmily ws ccepted by the progrm (enrolled). Whether the fmily ctully moved into housing unit using tht ssistnce. For fmilies who were ccepted by the progrm but did not move in, why they did not use the housing ssistnce. The clls were mde throughout the enrollment period September 2011 to Jnury 2012 nd continued through September 30, 2012, 9 months fter the lst fmily enrolled. These enrollment verifiction dt were used in the Progrm Usge Dt file, discussed in Section A.13. Gubits et l. (2013) provides further detils bout the enrollment verifiction process. A.5 6- nd 12-Month Trcking Surveys During the followup period the study tem conducted brief trcking surveys 6 nd 12 months fter enrollment. These surveys lsted n verge of 10 minutes nd provided updted contct informtion for the dult respondent nd secondry contcts. The trcking surveys lso cptured dt on the current living sitution, receipt of housing ssistnce, nd fmily composition for ech fmily. Locl site interviewers dministered the trcking surveys using CAPI technology. In most sites, the interviewer ws the sme person who dministered the bseline survey. Becuse the trcking surveys were reltively short, most prticipnts opted to complete the survey by telephone rther thn in person. A.6 18-Month Followup Adult Survey nd Focl Child Selection The 18-month followup dt collection hd number of components: dult survey, child survey, nd child ssessments. The 18-month followup dult survey collected informtion bout housing stbility, fmily preservtion, dult well-being, child well-being, nd self-sufficiency. Exhibit A-3 offers detiled list of topics included in the dult survey. The dult survey took n verge of 60 minutes to complete. The dult survey contined child screener module nd prent-on-child module. The child screener module ws designed to help the study tem identify eligible children for selection s focl children. Up to two focl children were selected for ech fmily. The dult survey lso included prent-on-child module in which the prent provided informtion bout the focl children. The section tht follows describes the process of focl child selection nd the prent-on-child module. A -3

204 Appendix A. Dt Sources nd Dtset Construction Exhibit A-3. Content of Prticipnt Dt Collected for Fmily Options Study 18-Month Impct Anlysis Adult Survey topics Child 12 to 41 Months Child 3 Yers, 6 Months to 7 Yers, 11 Months Housing sitution, qulity nd ffordbility Employment Income Eduction nd trining Economic hrdship Food security Fmily composition Fmily seprtion nd reunifiction Physicl helth Behviorl helth Substnce use Service receipt Child eductionb Child helthb Child behviorb Fmily routines Fers Anxiety Life events Substnce use School ttendnce Grdes School problems School effort Developmentl milestones (prent report)c Assessments Child 8 Yers to 17 yers, 11 Months Executive functioning (self-regultion)d Woodcock Johnson III Letter-word identifiction (reding) test Applied Problems (mthemtics) test Children nd youth ges 8 to 17 completed survey. The dult respondent ws sked questions bout ech focl child. c The dult respondent provided informtion on developmentl milestones using the Ages nd Stges Questionnire (ASQ-3). d Children ges 3 yers, 6 months to 7 yers, 11 months completed the Hed Toes Knees Shoulders (HTKS) ssessment, which mesures executive functioning. Source: Fmily Options Study 18-month followup survey b Focl Child Selection To nlyze the reltive impcts of the permnent housing subsidy (SUB), community-bsed rpid re-housing (CBRR), project-bsed trnsitionl housing (PBTH), nd usul cre (UC) interventions on child well-being, the study tem selected up to two focl children for ech fmily who completed the dult survey. The study tem then collected informtion bout the focl children to mesure child well-being outcomes. The items collected depended on the ge of the focl children. This section describes the process for selecting focl children. Two types of children were considered for focl child selection. First, ll children identified t bseline those in shelter with the dult t rndom ssignment nd those who were prt of the fmily but not in shelter with the dult t enrollment were eligible for focl child selection. If focl child smple selection hd been restricted to children identified t bseline, the focl child smple could hve been defined before the strt of the 18-month dt collection. The study tem expnded the focl child selection criteri, however, to include children who were born fter rndom ssignment. The study tem referred to these children s newborns. Becuse newborns could not be identified before the strt of the 18-month dt collection, the study tem dministered the focl child selection screener during the 18-month survey. To ensure tht ll newborns hd chnce to be selected s focl children, the study tem generted rndomly ordered list of ll the children identified t bseline nd two slots for up to two newborn children nd ssigned ech child number. This number determined the order in which the children were screened for focl child selection. A -4

205 Appendix A. Dt Sources nd Dtset Construction To be selected s focl child, ech child hd to first meet these two criteri. 1. The child ws one of the following:. Listed s child on the household roster from the bseline survey.3 b. Identified s newborn, by the dult respondent in the focl child screener section.4 2. The child ws t lest 12 months of ge but younger thn 18 yers of ge, s confirmed in the focl child screener. After the potentil child ws confirmed eligible for selection bsed on the first two criteri, the screener determined if the dult respondent ws knowledgeble enough bout the child s ctivities in the pst 30 dys to nswer the prent-on-child module. The screener mde this determintion using the next series of questions, indicting the third criterion for selection eligibility. 3. The child ws one of the following:. Living in the sme household s the dult respondent t lest hlf of the time or ll of the time t the 18-month followup point.5 b. The prent spent time with the child frequently nd ws t lest somewht fmilir with the child s ctivities.6 The preceding criteri constitute the miniml selection criteri. If possible, the first focl child would lso meet these dditionl criteri. 4. The child s ge t the followup survey ws greter thn or equl to 3 yers, 6 months. 5. The child ws living with the prent in the shelter t bseline. 6. The child ws living in the sme household s the prent t lest hlf of the time or ll of the time t the 18-month followup survey.7 Potentil focl children were then clssified into one of three types. Type 1: Met ll the miniml criteri; ws t lest 3 yer, 6 months of ge but younger thn 18 yers of ge; ws living in the shelter with the dult respondent t enrollment; nd lived with the respondent t lest hlf of the time or ll of the time t the time of the 18-month survey. Type 2: Met the miniml criteri, but did not meet the dditionl criteri. Type 3: Did not meet the miniml criteri. The focl child selection process worked s follows. If the prent hd ny Type 1 children, the study tem rndomly selected one s Focl Child A. Next, if the prent hd ny other Type 1 or Type 2 children, the study tem rndomly selected one s Focl Child B. If the prent hd no Type 1 children but did hve t lest one Type 2 child, then the study tem did not select Focl Child A but rndomly selected one Type 2 child s Focl Child B. Next, if the prent hd ny other Type 2 children, the study tem rndomly selected one s Focl Child C. 3 On the bseline survey, the tem collected children s ges but not dtes of birth, nd ll children on the household roster were ge 17 or younger. The roster included children who the dult respondent thought were prt of the fmily, even if they were not in the shelter with the respondent. All rndomly ssigned fmilies hd t lest one child ge 15 or younger. 4 Screener question 1: Between [rndom ssignment dte] nd [6 months before tody s dte] hve you (given birth to/fthered) child? 5 The point-in-time question to the prent ws, Do you currently live in the sme household s [child nme]...? It ws not question bout the entire period between bseline nd followup. 6 Two criteri hd to be stisfied. First, the prent spent 1 or more hours dy with the child t lest few times week during the month before the followup survey. Second, during tht month, the prent lwys, usully, or sometimes knew t lest two of the following: (1) how the child spent time when not in school or childcre, (2) which other kids the child spent time with, (3) whether the child hd finished her/his schoolwork or studying, nd (4) which TV progrms the child wtched. 7 This criterion is the sme s criterion 3. The difference is tht the miniml criteri ccept either 3 or 3b, wheres the dditionl criteri require 3. A -5

206 Appendix A. Dt Sources nd Dtset Construction If the prent hd only Type 3 children, focl child ws not selected. Focl Child A criteri excluded newborns nd focl children who were not living with the respondent t lest hlf of the time. This exclusion helped to mximize the number of fmilies in which focl child selection included t lest one focl child for whom direct child dt collection (child ssessments or child survey) ws possible. The focl child screening (confirmtion/collection of dte of birth nd collection of informtion for other criteri) ws performed for ech child in turn, following the rndomly ordered list, until two focl children were selected. After two focl children were selected, the focl child screening cesed. Therefore, collection of informtion for screening criteri other thn dte of birth ws not performed for every child in the respondent study fmilies. Within the 1,857 households in which n dult survey ws completed, the study tem selected 2,784 focl children.8 About 530 children screened for selection s focl children were living with the fmily hed less thn hlf of the time (out of bout 4,200 totl children screened). Of those children, the fmily hed ws knowledgeble bout only 60 of the children. In ccordnce with the focl child selection protocol, those 60 children were selected s focl children (long with 2,724 other selected focl children who were living with the fmily hed t lest hlf of the time). During nlysis, however, it ws decided tht such smll number of children would not llow for estimtes to generlize to the whole group of lrgely bsent children. Therefore, the 60 children were not included in impct nlyses. As result, the child impct results generlize only to children living with the fmily hed hlf of the time or more t the time of the dult survey. Exhibit A-4 shows the focl child smple sizes by ge group nd intervention group. Detils on vritions in focl child selection nd corresponding child weights re included in Appendix C.3. 8 No focl child ws selected in 130 fmilies, minly becuse children were no longer residing with the respondent nd the respondent did not know enough bout the child s ctivities during the previous 30 dys to respond properly nd lso becuse the children ged out of the ge rnge by the time of the followup survey. A -6

207 Appendix A. Dt Sources nd Dtset Construction Exhibit A-4. Focl Child Smple Distribution, by Site nd Intervention Group Focl Child Age (N) Site Nme Almed County Intervention Group 12 to 41 Months 3 Yers, 6 Months to 7 Yers, 11 Months 8 Yers to 17 Yers, 11 Months Totl Focl Child Smple (N) CBRR PBTH SUB UC Totl Atlnt CBRR PBTH UC Totl Bltimore CBRR PBTH UC Totl Boston CBRR SUB UC Totl Connecticut* CBRR PBTH SUB UC Totl Denver CBRR PBTH SUB UC Totl Honolulu CBRR PBTH SUB UC Totl Knss City CBRR PBTH SUB UC Totl Louisville CBRR PBTH SUB UC Totl Minnepolis CBRR PBTH SUB UC Totl Phoenix CBRR PBTH SUB UC Totl Slt Lke City CBRR PBTH SUB UC Totl Overll CBRR PBTH SUB UC Grnd Totl 576 1,079 1,129 2,784 CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. UC = usul cre. *Includes the cities of New Hven nd Bridgeport, Connecticut. Sources: Fmily Options Study bseline survey; followup survey A -7

208 Appendix A. Dt Sources nd Dtset Construction Prent-on-Child Module in Adult Survey In the prent-on-child module, the dult respondent provided informtion bout school ttendnce, cdemic performnce, behvior, helth, nd fmily routines. All focl children were between the ges of 12 months nd 17 yers, 11 months. If focl child s CAPI-clculted ge ws 12 to 41 months, the interviewer dministered the Ages nd Stges Questionnire (ASQ-3) to the dult s prt of the prent-on-child module. If focl child s CAPI-clculted ge ws 3 yers to 17 yers, 11 months, the interviewer dministered the Strengths nd Difficulties Questionnire (SDQ). The ASQ-3 is fmily of questionnires tht ssess gross nd fine motor skills, socil development, communiction, nd problem solving s observed by prents (Squires nd Bricker, 2009). Adult respondents completed the ASQ-3 for children ges 12 to 41 months.9 The ASQ-3 took n verge of 10 minutes to complete. The questionnire ws self-dministered for in-person dult surveys. For surveys conducted by phone, the questionnire ws dministered to the dult by phone. Detils on the component questions of the ASQ-3 nd scoring re in Appendix B.4. The SDQ is behviorl nd personlity ssessment. The questionnire ddresses child emotionl symptoms, conduct problems, hyperctivity, peer problems, nd prosocil behvior. Adult respondents completed this questionnire for focl children between ges 3 yers nd 17 yers, 11 months. Detils on the component questions of the SDQ nd scoring re in Appendix B.5. A.7 18-Month Followup Child Survey The study tem dministered the 18-month followup child survey to focl children between ges 8 yers nd 17 yers, 11 months.10 Interviewers dministered the survey using CAPI, with surveys lsting 30 minutes on verge. Surveys were conducted either in person or by telephone. The child survey sked questions bout nxiety using the A-Trit scle from the Stte-Trit Anxiety Inventory for Children, or STAIC (Spielberger et l., 1973); fers (Rmirez, Msten, nd Sms, 1991); substnce use (CDC, 2012); school ttendnce, effort, nd disciplinry problems; nd gol-oriented thinking using modified version of the Children s Hope Scle (Snyder et l., 1997). Detils on the component questions nd scoring re in Appendix B. A.8 18-Month Followup Child Assessments The study tem dministered two child ssessment tests to focl children between the ges of 3 yers, 6 months nd 7 yers, 11 months.11 These ssessments were the Woodcock-Johnson III (WJ III) nd the Hed Toes Knees Shoulders (HTKS) ssessments. The WJ III ssessment consisted of two tests. The first ws the Letter-Word Identifiction test nd the second ws the Applied Problems test (McGrew, Shrnk, nd Woodcock, 2007). These tests re subtests of eductionl chievement from the broder WJIII bttery of tests mesuring verbl nd quntittive/nlytic skills. The WJ III tests were dministered in person nd took n verge of 30 minutes to complete per child. Detils on scoring the WJ III tests re in Appendix B.5. The HTKS ssessment (Pontiz et l., 2007) mesures self-regultion, in which children must remember rules nd inhibit incorrect responses. HTKS ws conducted in person nd seprtely for ech focl child. The HTKS took n verge of 15 minutes per child to dminister. Detils on scoring HTKS re in Appendix B.5. 9 The CAPI progrm instructed interviewers to dminister the ASQ-3 during the dult survey bsed on whether CAPI clculted the focl child s ge to be 12 to 41 months. The CAPI ge clcultion ws bsed on the child s dte of birth nd the dult survey dte. The interviewers used the ASQ-3 online ge clcultor to determine which questionnire to dminister. 10 The CAPI progrm used the dte of birth for selected focl children nd the dult survey dte to clculte the selected focl child s ge in months nd yers. 11 The CAPI progrm used the dte of birth for selected focl children nd the dult survey dte to clculte the selected focl child s ge in months nd yers. A -8

209 Appendix A. Dt Sources nd Dtset Construction A.9 18-Month Followup HOME Inventory Interviewers on the study tem completed n observtion form bsed on subset of questions from the Home Observtion for Mesurement of the Environment (HOME) inventory. The HOME inventory questions used were bsed on observtiononly items bout prent-child interctions.12 Interviewers conducting in-person surveys completed this form s prt of the 18-month followup dt collection. The questions on the HOME inventory form sked bout interctions between the dult tking the dult survey nd ech focl child.13 For purposes of the HOME inventory tool, the interviewer ws instructed to explicitly prise ech child during the dult survey nd observe the dult respondent s rection. The interviewer prised ech focl child up to four times throughout the survey until the dult expressed rection. The HOME inventory dt were not used in nlyses in this report. A.10 Qulittive Surveys In 2011, the study tem conducted indepth surveys with 80 fmilies (20 fmilies from ech intervention group) in four sites: Almed County, Cliforni; Bridgeport nd New Hven, Connecticut; Knss City, Missouri; nd Phoenix, Arizon. Surveys were dministered in person 3 to 10 months fter rndom ssignment (6.4 months on verge), usully in the respondent s plce of residence. The qulittive dt collection ws designed to collect informtion to nswer the following questions. 1. How do fmilies mke housing decisions? 2. Wht re fmilies experiences (chllenges) nvigting the housing service system? 3. Wht explins seprtions of prents from children nd prtners from ech other? 4. How do housing situtions influence fmily processes? A tem of two interviewers met with fmilies to conduct in-depth surveys, lsting n verge of 2 hours. The surveys covered the following topics: 1. Current housing sitution, stisfction with current sitution, housing pyments, nd number of ddresses since study enrollment. 2. Subsidy use. 3. Eligibility nd tkeup of ssigned intervention, including resons for not using the ssigned intervention ssistnce. 4. Service receipt nd stisfction. 5. Household composition child nd spouse seprtions nd reunifictions. 6. Fmily processes nd rituls. Surveys were udio recorded nd ccompnied by hndwritten notes. Survey summries were trnscribed using NVivo softwre nd were coded. The qulittive dt identified fctors tht influenced the fmily s housing decisions s they left shelter (Fisher et l., 2014). These qulittive dt were used to interpret impct findings presented in this report. A.11 Additionl Detils About Surveys nd Dt Collection This section provides dditionl detils on the surveys nd dt collection process nd results. The following section summrizes the topics covered in ech of the forementioned survey instruments, the household members supplying different dt components, nd implementtion strtegies for the 18-month followup dt collection. This section lso discusses the study tem s efforts to mximize response rtes, incentives to prticipte in the study, nd summry of overll response rtes. 12 The full HOME inventory contins three types of items: (1) items sked bout during survey, (2) items either sked bout during survey or observed, nd (3) items bsed only on observtion. The form ws bsed on observtion-only HOME inventory items tht were relted to prent-child interctions. 13 In most cses for in-person surveys, the dult nd focl child were present, but, in some cses, the dult nd child components were not conducted on the sme dy nd interctions my not hve been observed. A -9

210 Appendix A. Dt Sources nd Dtset Construction Exhibit A-3 in n erlier section of this ppendix provides n overview of survey topics by study fmily members contributing informtion dults; children ges 12 to 41 months; children 3 yers, 6 months to 7 yers, 11 months; nd children 8 yers to 17 yers, 11 months. Implementtion of 18-Month Dt Collection The 18-month followup dt collection begn in July 2012 nd concluded in October Field interviewers conducted the followup surveys. Field mngers re-contcted 10 percent of ll respondents nd dministered brief vlidtion questionnire to ssure tht the survey ws done with the correct respondent nd tht the interviewer followed proper protocols. The 18-month followup dult survey dt collection process included 7. Locting reviewing contct history. 8. Adult informed consent renewing consent for the dult respondent. 9. Adult survey dministrtion including focl child selection. If no focl child selected, dt collection concluded here. If t lest one focl child selected then dt collection continued. Focl child dt collection steps included 1. Prentl permission required before interviewers could contct focl children. 2. Child ssent if focl child ws ges 8 yers to 17 yers, 11 months. 3. Child dt collection. Prent completes ASQ-3 nd SDQ (focl child ges 12 to 41 months). Child ssessments with focl child ges 3 yers, 6 months to 7 yers, 11 months. Child survey with focl child ges 8 yers to 17 yers, 11 months. 4. HOME inventory by the interviewer. At the time of the survey, interviewers first renewed consent with the dult smple member. Study prticipnts completed prticiption greement when they enrolled in the study, providing their informed consent to prticipte in the reserch study. The tem renewed consent with prticipting fmilies t 18 months to remind them of the voluntry nture of prticiption, the study requirements, nd risks of prticiption. The renewed consent form introduced the child dt collection component. When consent ws renewed, interviewers dministered the dult survey to the respondent. Although the survey instrument ws designed to be conducted in person, 759 (40.9 percent) of dult respondents chose to do the survey by phone. In-person surveys were conducted in vriety of loctions, both inside nd outside the respondent s residence. Locl interviewers completed ll dult surveys using lptops equipped with CAPI technology. If t lest one focl child ws selected for the study, interviewers reviewed the prentl permission form with dult respondents fter completing the dult survey. Adult respondents could decline study prticiption for focl children independently of whether they grnted permission for nother focl child in the fmily. If the focl child ws 12 to 41 months of ge, the dult respondent ws sked to complete the ASQ-3 nd ll items in the prent-on-child module of the dult survey pertining to children in tht ge rnge. 14 A smll number of child cses were finlized in November A-10

211 Appendix A. Dt Sources nd Dtset Construction If focl child ws ge 3 yers, 6 months to 7 yers, 11 months, the interviewer mde n ppointment to meet with the dult respondent nd the focl child to conduct the child ssessments. After obtining dult permission, interviewers lso requested focl child permission to conduct the child ssessments. In totl, 41 ttempts to dminister the child ssessments were refused by either the dult or focl child. All child ssessments were completed in person. Interviewers directly contcted focl children ged 8 to 17 yers, 11 months to mke n ppointment to dminister the child survey. Before beginning the child survey, interviewers reviewed the child ssent form with the respondent nd, if ssent ws grnted, proceeded to conduct the survey. Not ll focl children decided to prticipte. In totl, 91 ttempts to dminister the child survey were refused by either the prent or focl child. Like the dult survey, substntil percentge of the older focl children (45.0 percent) preferred to do the survey by telephone. Efforts To Improve Response Rtes The study tem used vriety of methods to mintin current contct informtion on study fmilies, with n effort to minimize prticipnt burden. Study fmilies were contcted qurterly. The contcts rnged from cll (3 months fter rndom ssignment) to miling (t 9 nd 15 months fter rndom ssignment) to more intensive trcking survey (t 6 nd 12 months fter rndom ssignment). Incentives All respondents received n incentive pyment in pprecition for their time spent to complete the dt collection. Adults who completed the bseline survey received $35 money order. Ech time the dult prticipnt responded to trcking effort, they received $15 money order. Adults who completed the 18-month followup survey received $50 money order. Adults lso received $15 money order on behlf of ech child who completed the child ssessments or the child survey. Response Rtes Exhibit A-5 shows the overll response rtes for ech prticipnt dt collection effort. The completion rte represents the number of completed surveys s percentge of the totl cses ttempted. The nlytic response rte is pplicble only to the focl child dt collection components (child ssessment nd survey components). The finl enrollment for the study ws 2,282 fmilies, which ws the smple bse for ll the dt collection efforts. The 18 month followup dult survey chieved n 81 percent completion rte. During the 18-month followup period, 11 dult respondents were confirmed decesed. Exhibit A-5. Overll Fmily Options Study Survey Response Rtes Smple Relesed (N) Cses Completed (N) Completion Rte (%) Anlytic Response Rte (%) Bseline 2,282 2, month trcking 2,282 1, month trcking 2,282 1, month dult 2,282 1, Ages nd Stges Hed Toes Knees Shoulders 1, WJ III Letter-Word 1, WJ III Applied Problems 1, Child survey 1, WJ III = Woodcock-Johnson III. Notes: The child ssessment dt collection nlytic response rte t 18 months is product of its completion rte for focl children nd the 81.4 percent response rte for fmily heds. Similrly, the child survey dt collection t 18 months is product of its 83.8 completion rte for focl children nd the 81.4 percent response rte for fmily heds. Sources: Fmily Options Study bseline survey; 6-month trcking survey; 12-month trcking survey; dult survey; child ssessments; child survey; HOME observtions. A-11

212 Appendix A. Dt Sources nd Dtset Construction Child dt collection could be done only fter n dult survey ws completed becuse focl child selection occurred s prt of the dult survey. Further, the dult respondent hd to give prentl permission before the child dt collection could commence. The child completion rtes re bsed on the number of completed child components s percentge of the focl children selected in households with completed dult survey. Becuse n dult survey ws completed with only 81 percent of the dult smple, the child dt collection nlytic response rtes re lower, fter djusting for the households without completed dult surveys. Exhibit A-6 shows the number nd percentge of fmilies who responded to one, two, three, or four surveys. More thn onehlf (55.2 percent) of the smple responded to ll four survey efforts, nd 79 percent of the smple responded to t lest three of the four survey efforts. Exhibit A-6. Survey Response Sttus for Fmily Options Study Surveys 6-Month Trcking Survey 12-Month Trcking Survey Yes No Yes No Yes No Yes Yes No Yes Yes Yes No Yes Yes Yes Yes Yes Yes Yes Bseline Survey 18-Month Adult Survey Fmilies (N) % No No No Yes No Yes No No Yes Yes No Yes 1, , Totl Sources: Fmily Options Study bseline survey; 6-month trcking survey; 12-month trcking survey; 18-month followup survey. A.12 Administrtive Dt The Fmily Options Study drew on two min dministrtive dt sources: (1) the Homeless Mngement Informtion System (HMIS) dt from ech study site nd (2) HUD s Public nd Indin Housing Informtion Center (PIC) nd Tennt Rentl Assistnce Certifiction System (TRACS) dt. Ech source of dministrtive dt is described below. HMIS Dt An HMIS15 is the electronic informtion system designted by the locl Continuum of Cre (CoC) progrm to record dt on ll people served within CoC s shelter, housing, nd service system for individuls nd fmilies experiencing homelessness. Agencies collect informtion directly from people they serve nd enter the dt into their CoC s HMIS. Exhibit A-7 shows the HMIS prticiption rtes for the CoCs contining our smple sites bsed on informtion reported by communities to HUD in the spring of HMIS bed prticiption refers to the percentge of beds tht re covered in the HMIS. Thus, for exmple, dt on clients stying in 83 percent of the beds designted for fmilies in emergency shelters in Almed County tht re prticipting in the study re included in HMIS. The study tem used HMIS records to mesure use of emergency shelter, rpid re-housing, trnsitionl housing, nd permnent supportive housing (PSH) nd the length of time fmilies spent in these housing progrms. 15 See for more informtion on HMIS. A-12

213 Appendix A. Dt Sources nd Dtset Construction Exhibit A-7. HMIS Prticiption Rtes for Emergency Shelter nd Trnsitionl Housing Providers in the Study Sites, 2011 HMIS Bed Prticiption Rtes for All Providers in the CoC (%) Study Site ES TH HMIS Bed Prticiption Rtes for Providers in the Study (%) ES TH Almed County Atlnt Bltimore Boston NA Louisville Minnepolis Connecticut* Denver Honolulu Knss City Phoenix Slt Lke City ES = emergency shelter. TH = trnsitionl housing. CoC = Continuum of Cre. NA = not vilble. * This study site comprises four CoCs in the New Hven/Bridgeport, Connecticut re; therefore, the figures reported for CoC coverge represent the rnge of coverge levels in these four CoCs. Note: All beds (tht is, progrm slots) enumerted in this tble re considered beds for fmilies by the CoCs. Source: HUD Homeless Dt Exchnge, or HDX. HMIS dt elements supplied by sites were Progrm entry dte. Progrm exit dte. Progrm nme. Progrm type. Providers enter new HMIS record for every new entry into progrm. Thus, for people who receive more thn one episode of ssistnce, HMIS contins multiple records per person. Providers sk clients entering progrms to provide personlly identifying informtion, but clients re not required to comply to receive services. Exhibit A-8 shows the number nd percentge of fmilies in the smple tht were identified in the site s HMIS. A fmily is considered mtched in the HMIS if t lest one progrm record ws found for the hed of household in the HMIS dt received nd time of progrm use occurred fter the rndom ssignment dte. The study tem gthered supplementry shelter progrm-use dt for Minnepolis, Minnesot, nd Boston, Msschusetts. Hennepin County provided records for emergency shelters tht were not covered in the Minnepolis HMIS. The Stte of Msschusetts Office of Community Development provided records of Emergency Assistnce progrm use tht were not covered in the Boston HMIS. In ddition, the Connecticut HMIS dt were provided by the Bridgeport CoC nd New Hven CoC seprtely. A-13

214 Appendix A. Dt Sources nd Dtset Construction Exhibit A-8. HMIS Mtch Rtes With the Fmily Options Smple, by Site Site Almed Atlnt Bltimore Boston Connecticut* Denver Honolulu Knss City Louisville Minnepolis Phoenix Slt Lke City All Sites Originl Smple (N) Smple Fmilies in HMIS (N) Smple Fmilies in HMIS (%) ,282 2, *Includes the cities of New Hven nd Bridgeport, Connecticut. Source: Homeless Mngement Informtion Systems (HMIS) PIC nd TRACS Dt Files The study tem used two HUD dministrtive dt files, PIC nd TRACS. PIC dt were used to mesure smple members receipt of housing ssistnce from one of three progrms public housing, the Housing Choice Voucher (HCV) progrm, nd project-bsed voucher ssistnce.16 TRACS dt were used to trck informtion regrding progrm entry nd exit for project-bsed Section 8 progrms.17 PIC nd TRACS dt mesure use of the SUB intervention. HUD provided the dt in 20 PIC nd 20 TRACS extrcts with qurterly dt. The qurterly extrcts cover the period from Mrch 2009 through December 2013 nd cover the effective dte period from Jnury 1, 2008, until December 31, The PIC extrcts contined 57 vribles nd 889 individuls in 759 fmilies. The TRACS extrcts contined 41 vribles nd 80 individuls in 74 fmilies. Together, the 40 qurterly extrcts ccounted for 959 individuls in 827 fmilies. For fmilies using vouchers, the study tem used PIC dt to identify the dte on which the household begn to receive rentl ssistnce, referred to s the lese-up dte. The progrm dmission dte, effective dte of the ction, progrm type code, nd progrm ction code were the mjor vribles used from the PIC extrcts to determine the timing of new dmissions (versus nnul reexmintions, interim reexmintions, or other ctions).18 Depending on the type of ction recorded in PIC, the dte of progrm dmission is either the sme s or erlier thn the effective dte of ction. For new dmissions, the dte of dmission nd the effective dte of ction re normlly the sme. When they differ, the effective dte is considered the better indictor of lese up, becuse the effective dte refers to either the signing of the lese or the ctul occupncy of the unit (s opposed to, for instnce, the issunce of the voucher to the prticipnt). Among records of ctions other thn new dmissions, some effective dtes fell up to yer fter the dte of rndom ssignment.19 Therefore, for ction types other thn new dmission, the dte of progrm dmission ws consistently used s the dte of lese up. Exhibit A-9 summrizes the number of smple fmilies who were mtched in PIC/TRACS nd the number ssigned to SUB mtched in PIC/TRACS. 16 HUD Form describes the full list of vribles vilble in the PIC dt nd is ccessible on line t 17 Documenttion on the TRACS dt is ccessible on line t 18 An ction is the dministrtive trnsction tht triggers the completion of the HUD Form tht is submitted to PIC. The includes 14 ction codes: new dmission, nnul reexmintion, interim reexmintion, portbility move-in, portbility move-out, end of prticiption, other chnge of unit, Fmily Self-Sufficiency (FSS)/Welfre to Work (WtW) ddendum, nnul reexmintion, issunce of voucher, expirtion of voucher, flt rent nnul updte, nnul Housing Qulity Stndrds (HQS) inspection, nd historicl djustment. 19 This lpse in time might occur, for instnce, if the housing uthority (HA) simply filed to record the new dmission but recorded subsequent ction. A-14

215 Appendix A. Dt Sources nd Dtset Construction Exhibit A-9. Smple Fmilies in PIC/TRACS Dt nd Those Assigned to the SUB Intervention in PIC/TRACS Dt, by Site Smple Fmilies in PIC/TRACS (N) All Fmilies Assigned to SUB (N) All Fmilies Assigned to SUB in PIC/TRACS (N) Almed County Atlnt NA Bltimore NA Boston Percent SUB-Assigned Fmilies in PIC/TRACS Connecticut* Denver Honolulu Knss City Louisville Minnepolis Phoenix Slt Lke City All Sites PIC = Public nd Indin Housing Informtion Center. SUB = permnent housing subsidy. TRACS = Tennt Rentl Assistnce Certifiction System. *Includes the cities of New Hven nd Bridgeport, Connecticut Source: PIC/TRACS dt A.13 Progrm Usge Dt File The study used severl types of informtion to understnd wht ssistnce people received during the followup period. Enrollment verifiction dt, trcking nd followup surveys, nd dministrtive dt, ech described previously in Appendix A, were combined to form Progrm Usge Dt file, with informtion bout housing progrm use for every month strting t the dte of rndom ssignment through the dte of the dult survey. This section describes how the study tem creted the Progrm Usge Dt file, which contins dt on the 1,857 fmilies who responded to the 18-month followup dult survey. The study tem gthered fmily-level informtion on progrm entry dtes, progrm exit dtes, nd progrm types using six dt sources throughout the course of the Fmily Options Study. The dt sources re 1. Enrollment verifiction dt month trcking survey month trcking survey month followup dult survey. 5. HMIS. 6. PIC/TRACS. Ech dt source hs informtion bout progrm use since the dte of rndom ssignment. The dt sources vry in the mount of time they cover. For exmple, the dministrtive dt nd the 18-month survey cover the full nlysis period, but the trcking surveys nd enrollment verifiction cover only prt of the period. The study tem considered dt from ll six sources when compiling fmily histories of progrm use. In mny cses, the sme instnce of progrm use by fmily ws recorded in more thn one dt source. In some of these cses, the multiple dt sources were in complete greement. In other cses, the dt sources hd discrepnt informtion bout entry dtes, exit dtes, nd/or progrm type. To resolve conflicting informtion cross dt sources, the study tem devised system of decision rules. The fundmentl rule for clening the dt ws tht two instnces of progrm use could not overlp, forcing the study tem to clen dtes tht A-15

216 Appendix A. Dt Sources nd Dtset Construction indicted the fmily ws in two or more progrms simultneously. The study tem rnked the dt sources in the order believed to contin the most to lest relible progrm use informtion. Perceived relibility of the dt sources vried by dt item progrm entry dte, exit dte, nd type. Exhibit A-10 summrizes the relibility rtings. The study tem considered the progrm entry dte from the enrollment verifiction dt most relible becuse the tem collected these dt directly from the prticipting provider specificlly bout the study fmilies. The dministrtive dt HMIS nd PIC/TRACS were lso treted s highly relible, second to the enrollment verifiction. The dministrtive dt re mintined by communities nd HUD. Progrm entry dte informtion from the trcking surveys nd 18-month followup survey ws considered less relible becuse of humn recll error. If entry dte informtion ws vilble only in the surveys, the study tem considered progrm entry dtes closest to survey dtes s more relible thn other older entry dtes becuse they would hve lower recll error. The study tem considered the progrm exit dte from the dministrtive dt to be most relible. The exit dte from the enrollment verifiction dt ws considered lest relible becuse dt were not collected for long enough period to record n exit dte. The trcking surveys lso contined missing exit dte informtion if the fmily ws in housing progrm t the time of those surveys nd could suffer from recll error. The18-month followup informtion on exit dtes covered the full study period but could still suffer from recll error. The study tem considered the progrm type dt from the enrollment verifiction s the most relible becuse these providers were involved in the study to represent n intervention progrm type. Progrm type dt from dministrtive sources were considered to lso be highly relible, second to the enrollment verifiction dt. The study tem worked closely with the HMIS dministrtors to ccurtely code progrms. Dt from PIC/TRACS were lso considered highly relible becuse dt re mintined by HUD. Progrm type informtion in the 18-month followup nd trcking surveys were considered to be lest relible becuse of recll error nd likely lck of knowledge of the progrm type beyond the nme of the progrm. Bsing its nlysis on these nd other site-specific rules, the study tem mnully determined which records nd informtion were preserved tht most ccurtely reflected the progrm use history of fmily. These dt were converted into the Progrm Usge Dt file, which contined one record per fmily. The Progrm Usge Dt file contined series of monthly indictor binry vribles reflecting the period from the month of rndom ssignment through the month of the lst 18 month survey end dte in the study. The study tem prepred seprte set of indictor vribles for severl progrm types: emergency shelter, trnsitionl housing, community-bsed rpid re-housing, HCVs, PSH, public housing, nd Section 8 projects/project-bsed vouchers. An indictor vrible ws set equl to 1 to indicte tht the study fmily used prticulr progrm type t lest 1 dy during the month, or it ws set equl to 0 to indicte no use of tht progrm type in tht month. The Progrm Usge Dt file ws constructed to complement the outcomes reported in the 18-month followup dult survey. Therefore, informtion provided by dministrtive dt beyond the month of followup survey response ws not incorported into the file. Exhibit A-10. Dt Source Relibility, by Progrm Use Dt Item Progrm Use Dt Item Progrm entry dte Higher Relibility Enrollment verifiction Lower Relibility 18-month followup survey; trcking surveys HMIS; PIC/TRACS HMIS; PIC/TRACS Progrm exit dte 18-month followup survey; trcking surveys; enrollment verifiction Progrm type Enrollment verifiction 18-month followup survey; trcking surveys HMIS; PIC/TRACS HMIS = Homeless Mngement Informtion System. PIC = Public nd Indin Housing Informtion Center. TRACS = Tennt Rentl Assistnce Certifiction System. Sources: Enrollment verifiction dt; 6-month trcking survey; 12-month trcking survey; 18-month followup survey; HMIS; PIC/TRACS A-16

217 Appendix B. Construction of Outcomes APPENDIX B. CONSTRUCTION OF OUTCOMES This ppendix of the Fmily Options Study short-term impcts report describes how the study tem constructed outcomes mesures from the 18-month followup survey nd dministrtive dt. It supplements informtion in Chpter 5. The section is orgnized by outcome domin: housing stbility, fmily preservtion, dult well-being, child well-being, nd self-sufficiency. B.1 Mesures of Housing Stbility Homelessness during followup period. The study tem developed seven mesures relted to homelessness experienced during the 18-month followup period. 1. At lest 1 night homeless or doubled up during pst 6 months. This binry vrible is constructed from responses to Questions A9 nd A11. It mesures the percentge of study fmilies who reported spending t lest 1 night during the 6 months before the 18-month followup survey either homeless (residing in shelter or institution or stying in plce not typiclly used for sleeping, such s the street, cr, bndoned building, or trin sttion) or living with friends or reltives becuse they could not find or fford plce of their own. The survey item explicitly excluded trnsitionl housing. Although trnsitionl housing is included in the federl definition of homelessness, this outcome does not reflect stys in trnsitionl housing. Dt were missing on this mesure for less thn 0.1 percent of cses. 2. At lest 1 night homeless during pst 6 months. This binry vrible is constructed from responses to Question A9. It mesures the percentge of fmilies who sid they spent t lest 1 night homeless during the 6 months before the 18-month followup survey. This outcome does not reflect stys in trnsitionl housing. Dt were missing on this mesure for 0.1 percent of cses. 3. At lest 1 night doubled up during pst 6 months. This binry vrible is constructed from responses to Question A11. It mesures the percentge of fmilies who sid they spent t lest 1 night living with friends or reltives during the 6 months before the 18-month followup survey. No cses hd missing dt on this mesure. 4. Any sty in emergency shelter in months 7 to 18 fter rndom ssignment. This binry vrible is constructed using Progrm Usge Dt. Cses indicting enrollment in emergency shelter in the period from 7 to 18 months fter rndom ssignment re coded s 1. Homeless Mngement Informtion System, or HMIS, records were the source of most (87 percent) of the spell dt for emergency shelter. The surveys were the other sources of emergency shelter spell dt. No cses hd missing dt on this mesure. 5. Number of dys homeless or doubled up during pst 6 months. This continuous vrible is constructed from responses to Questions A101 to A103 nd A121 to A123. The outcome counts the totl number of dys spent homeless or doubled up in the 6 months before the 18-month followup survey. This outcome does not reflect stys in trnsitionl housing. Dt were missing on this mesure for less thn 0.3 percent of cses. B -1

218 Appendix B. Construction of Outcomes 6. Number of dys homeless during pst 6 months. This continuous vrible is constructed from responses to Questions A101 to A103. The outcome counts the totl number of dys spent homeless in the 6 months before the 18-month followup survey. This outcome does not reflect stys in trnsitionl housing. Dt were missing on this mesure for 0.7 percent of cses. 7. Number of dys doubled up during pst 6 months. This continuous vrible is constructed from responses to Questions A121 to A123. The outcome counts the totl number of dys spent doubled up in the 6 months before the survey. Dt were missing on this mesure for less thn 0.1 percent of cses. Homelessness during the followup period: confirmtory outcome. The study tem lso constructed single composite binry outcome defined s recent experience of homelessness from two binry outcomes within the housing stbility domin. 1. At lest 1 night spent homeless or doubled up during the pst 6 months t the time of the 18-month followup survey. This outcome does not reflect stys in trnsitionl housing. 2. Any sty in emergency shelter in the 12 months prior to the dte of the 18-month survey mesured from Progrm Usge Dt. If either of the two binry outcomes were coded s 1, the composite confirmtory outcome ws lso coded s 1. Dt were missing on this mesure for less thn 0.1 percent of cses. Housing independence. The study tem used dt from the dult survey to construct three outcomes pertining to the type of living rrngements t the time of the 18-month followup survey. 1. Living in own house or prtment t time of survey. This binry vrible is constructed from responses to Question A4, which sks if the respondent is currently living in house or prtment tht he or she owns or rents. The interviewer instructed the respondent not to include his or her prent s or gurdin s home or prtment. Dt were missing on this mesure for 2.3 percent of cses. 2. Living in own house or prtment t time of survey with no housing ssistnce. This binry vrible is bsed on responses to Question A4 bout living sitution nd Questions A7 nd A8 bout receipt of housing ssistnce t the time of the 18-month followup survey. The outcome is ssigned vlue of 1 for respondents who nswered Question A4 to indicte they were living in house or prtment tht they own or rent nd who nswered no to Questions A7 nd A8 bout whether they received housing ssistnce. Dt were missing on this mesure for 2.3 percent of cses. 3. Living in own house or prtment t time of survey with housing ssistnce. This binry vrible is constructed from responses to Question A4 bout living sitution nd Questions A7 nd A8 bout receipt of housing ssistnce t the time of the 18-month followup survey. The outcome is ssigned vlue of 1 for respondents who nswered Question A4 to indicte they were living in house or prtment tht they own or rent nd who nswered yes to Questions A7 nd A8 bout whether they received housing ssistnce. Dt were missing on this mesure for 2.3 percent of cses. Number of plces lived. The reserch lso mesured n outcome relted to housing instbility during the followup period using the prent survey. Number of plces lived/styed during pst 6 months. This continuous vrible is constructed from responses to Question A13. The outcome mesures the count of plces lived during the 6 months before the prent survey. The vrible is topcoded t five plces. A vlue of 6 mens tht the dult respondent reporting hving lived in five or more plces in the 6 months before the survey. Dt were missing on this mesure for 0.2 percent of cses. B -2

219 Appendix B. Construction of Outcomes Housing qulity. The housing domin lso includes two outcomes mesuring the qulity of smple members housing t the time of the 18-month followup survey. 1. Persons per room. This continuous vrible is bsed on responses to Question B7 (number of rooms not including kitchens, bthrooms, nd hllwys) nd Questions B1A_1 to B1A_14 nd B3 nd B3b (number of persons living with the dult respondent). The outcome ws constructed by dividing the number of people by the number of rooms. Dt were missing on this mesure for 1.8 percent of cses. 2. Housing qulity is poor. This binry vrible is bsed on responses to Question B8. Respondents were sked to rte the condition of their current house or prtment s either excellent, good, fir, or poor. Dt were missing on this mesure for 0.8 percent of cses. B.2 Mesures of Fmily Preservtion The study tem collected detiled informtion bout the composition of the study fmilies nd chnges in fmily composition tht occurred during the followup period. The study tem specificlly collected nmes nd ges of fmily members with the dult respondent in shelter t the time of enrollment nd of fmily members who were not with the dult respondent t enrollment but whom the dult respondent considered to be prt of the fmily. Then, t the 18-month followup survey, the study tem collected informtion on the wherebouts of ll fmily members reported t bseline nd bout new fmily members who hd joined the fmily since the previous survey. The study tem used this informtion to construct outcomes mesuring recent seprtions of fmily members who were present t bseline. The three outcomes re 1. Fmily hs t lest one child seprted in pst 6 months (percent of fmilies). This binry vrible is constructed from items in Section D of the 18-month followup survey nd from informtion gthered from Section E of the bseline survey. This vrible mesures the percentge of fmilies in which child who hd been with the fmily t bseline becme seprted in the 6 months before the 18-month followup survey. Dt were missing on this mesure for 1.5 percent of cses. 2. Fmily hs t lest one foster cre plcement in the pst 6 months (percent of fmilies). This binry vrible is constructed from items in Section D of the 18-month followup survey. It mesures the percentge of fmilies in which ny children were living in foster cre in the 6 months before the survey. Dt were missing on this mesure for 1.2 percent of cses. 3. Spouse/prtner seprted in pst 6 months, of those with spouse/prtner present t rndom ssignment (percent of fmilies). This binry vrible is constructed from items in Section D of the 18-month followup survey nd Section E of the bseline survey. This outcome mesures the percentge of fmilies in which spouse or prtner who ws with the fmily t bseline becme seprted in the 6 months before the 18-month followup survey. The tem lso constructed the following two fmily reunifiction outcomes tht mesure reunifiction of fmily members who hd been reported s seprted from the fmily t bseline. 1. Fmily hs t lest one child reunified, of those fmilies with t lest one child bsent t rndom ssignment (percent of fmilies). This binry vrible is constructed from Section D of the 18-month followup survey nd Section E of the bseline survey. It mesures the percentge of fmilies in which child who ws not with the fmily t bseline ws residing with the fmily t the time of the 18-month followup survey. 2. Spouse or prtner reunified (percent of fmilies). This binry vrible is tken from Section D of the 18 month followup survey nd Section E of the bseline survey. It mesures the percentge of fmilies in which spouse or prtner who ws not with the fmily t bseline ws residing with the fmily t the time of the 18 month followup survey. B -3

220 Appendix B. Construction of Outcomes B.3 Mesures of Adult Well-Being The study tem used the dult survey to construct the following outcomes mesuring severl spects of well-being for the dult respondent: dult physicl helth, dult behviorl helth, dult trum symptoms, dult substnce buse, nd experience of intimte prtner violence. Adult physicl helth is fir or poor. This binry vrible is constructed from responses to Question E1. The dult respondent ws sked to report on overll helth in the 30 dys before the survey. Response options were excellent, very good, good, fir, or poor. The outcome vlue is 1 if respondents rted helth s fir or poor. Dt were missing on this mesure for 0.1 percent of cses. Gol-oriented thinking. The dult survey collected six items of the Stte Hope Scle (Snyder et l., 1996) in Questions E2 to E2f. Prticipnts respond to ech item using 6-point scle. 1 = definitely flse. 2 = mostly flse. 3 = somewht flse. 4 = somewht true. 5 = mostly true. 6 = definitely true. The items re 1. If I should find myself in jm, I could think of mny wys to get out of it. 2. At the present time, I m energeticlly pursuing my gols. 3. There re lots of wys round ny problem tht I m fcing now. 4. Right now I see myself s being pretty successful. 5. I cn think of mny wys to rech my current gols. 6. At this time, I m meeting the gols tht I hve set for myself. The study tem creted score for the Stte Hope Scle if the respondent nswered t lest four of the six items. For ech respondent, the nlysts verged the responses given. This process yields mesures rnging from 1 to 6 with higher scores indicting greter hope. The dult Stte Hope Scle mesures Snyder s cognitive model of hope which defines hope s positive motivtionl stte tht is bsed on n interctively derived sense of successful () gency (gol-directed energy), nd (b) pthwys (plnning to meet gols) (Snyder, Irving, nd Anderson, 1991: 287). Dt were missing on this mesure for 0.6 percent of cses. Psychologicl Distress. This continuous vrible is the Kessler 6 Psychologicl Distress Scle (Kessler et l., 2003). It is derived from six survey items (Questions E31 to E36). The respondents were sked how much of the time in the pst 30 dys they hd felt ech of six mesures of distress 1. Nervous? 2. Hopeless? 3. Restless or fidgety? 4. So depressed tht nothing could cheer you up? 5. Tht everything ws n effort? 6. Worthless? B -4

221 Appendix B. Construction of Outcomes Responses options were 1 = ll of the time. 2 = some of the time. 3 = little of the time. 4 = none of the time. The study tem creted score for ech respondent if the respondent nswered t lest four of the items. Scores were reversed such tht response of ll of the time = 4, most of the time = 3, some of the time = 2, little of the time = 1, nd none of the time = 0. The scores were summed, creting continuous indictor of psychologicl distress. The score rnges from 0 to 24 with higher vlues indicting greter distress. Dt were missing on this mesure for 0.4 percent of cses. Exhibit 5-10 in Chpter 5 reports the percent of fmily heds reporting symptoms of serious psychologicl distress. Tht mesure ws derived from the continuous distress scle using cutoff of 13 (tht is, scores of 13 nd over were coded to indicte serious psychologicl distress). This score ws the optiml cutoff point for the generl popultion smple in the Kessler et l. (2003) vlidtion study. Post-trumtic stress disorder (PTSD) symptoms. This binry outcome is constructed from responses to Questions E4 to E4q. It mesures the presence of PTSD symptoms in dult respondents in the month before the survey. These questions re the Posttrumtic Stress Dignostic Scle (PDS) ssessment tht is designed to id in the detection nd dignosis of PTSD. The PDS ssessment prllels DSM-IV dignostic criteri for PTSD dignosis nd my be dministered repetedly over time to help monitor chnges in symptoms. Respondents were sked to report on how much ech of the following items hd bothered them in the month before the survey. Subset 1 E4. Repeted, disturbing memories, thoughts, or imges of stressful experience? E4b. Repeted, disturbing drems of stressful experience? E4c. Suddenly cting or feeling s if stressful experiences were hppening gin (s if you were reliving it)? E4d. Feeling very upset when something reminded you of stressful experience? E4e. Hving physicl rections (for exmple, hert pounding, trouble brething, or sweting) when something reminded you of stressful experience? Subset 2 E4f. Avoid thinking bout or tlking bout the stressful experiences or void hving feelings relted to it? E4g. Avoid ctivities or situtions becuse they remind you of stressful experience? E4h. Trouble remembering importnt prts of the stressful experience? E4i. Loss of interest in things tht you used to enjoy? E4j. Feeling distnt or cut off from other people? E4k. Feeling emotionlly numb or being unble to hve loving feelings for those close to you? E4l. Feeling s if your future will somehow be cut short? B -5

222 Appendix B. Construction of Outcomes Subset 3 E4m. Trouble flling or stying sleep? E4n. Feeling irritble or hving ngry outbursts? E4o. Hving difficulty concentrting? E4p. Being super lert or wtchful on gurd? E4q. Feeling jumpy or esily strtled? Responses options were 1 = not t ll. 2 = little bit. 3 = modertely. 4 = quite bit. 5 = extremely. The PTSD outcome ws creted if the respondent nswered t lest 12 of the 17 items. The first step in scoring the responses ws to ssess if the respondent ws symptomtic on ech item. Responses of 3 = modertely, 4 = quite bit, or 5 = extremely to ny of the items indicte the respondent is symptomtic nd receive vlue of 1. If respondent nswered 1 = not t ll or 2 = little bit to n item they were ssessed s not symptomtic nd received vlue of 0. The 17 items were then divided into subscles. Subscle B: sum of 5 items in Subset 1: e. Subscle C: sum of 7 items in Subset 2: f l. Subscle D: sum of 5 items in Subset 3: m q. To receive vlue of 1 for the PTSD binry vrible, the respondent hd to be symptomtic on one or more items in subscle B, three or more items in subscle C, nd on two or more items in subscle D. This mesure of PTSD ws lso collected t bseline nd serves s covrite scored in the sme mnner (see Appendix C.1). Dt were missing on this mesure for less thn 0.8 percent of cses. Adult substnce buse. The study tem mesured three outcomes relted to substnce buse. The first is binry vrible indicting lcohol dependence, the second is binry vrible indicting drug buse, nd the third is binry vrible indicting lcohol dependence or drug buse. Alcohol dependence is constructed from responses to Questions E5 through E8 in the 18-month followup survey, which sked the following four items in the Rpid Alcohol Problems Screen (RAPS4; Cherpitel, 2000). E5. Do you sometimes tke drink in the morning when you first get up? E6. During the pst 6 months, hs friend or fmily member ever told you bout things you sid or did while you were drinking tht you could not remember? E7. During the pst 6 months, hve you hd feeling of guilt or remorse fter drinking? E8. During the pst 6 months, hve you filed to do wht ws normlly expected of you becuse of drinking? An ffirmtive nswer to ny of the items indictes n lcohol problem. Dt were missing on this mesure for 0.3 percent of cses. The Rpid Alcohol Problems Screen is five-item instrument, derived from other screens, tht is designed to mximize sensitivity while mintining good specificity. The RAPS4, further refinement of the five-item B -6

223 Appendix B. Construction of Outcomes instrument, sks if n individul felt guilt fter their drinking (Remorse), could not remember things sid or did fter drinking (Amnesi), filed to do wht ws normlly expected fter drinking (Perform), or hd morning drink (Strter). The RAPS4 indictes lcohol buse for the Fmily Options Study. The outcome indicting drug buse is constructed from responses to Questions E10 through E10h. The survey instrument included six items regrding use of illegl drugs, ll of which re prt of the short version of the Drug Abuse Screening Test, or DAST-10 (Skinner, 1982; Yudko, Lozhkin, nd Fouts, 2007). The following six items were sked of respondents in reltion to the 6 months before the survey dte.20 E10. Hve you used more thn one drug t time? E10b. Hve you hd blckouts or flshbcks s result of drug use? E10e. Hve you ever not spent time with your fmily or missed work becuse of drug use? E10f. Hve you engged in illegl ctivities in order to obtin drugs? E10g. Hve you ever experienced withdrwl symptoms s result of hevy drug intke? E10h. Hve you hd medicl problems s result of drug use (for exmple memory loss, heptitis, convulsions, bleeding?) An ffirmtive nswer to ny of these six items indictes drug problem. Dt were missing on this mesure for 0.4 percent of cses. If the respondent is determined to hve n lcohol problem or drug problem, the outcome lcohol dependence or drug buse is ssigned vlue of 1. Dt were missing on this mesure for 0.4 percent of cses. Experienced intimte prtner violence in the pst 6 months. This binry vrible is bsed on responses to Question E11. E11 sks if, in the pst 6 months, the dult respondent hs been physiclly bused or thretened with violence by person with whom she/he ws romnticlly involved, such s spouse, boy/girlfriend, or prtner. Dt were missing on this mesure for 0.3 percent of cses. B.4 Mesures of Child Well-Being The study tem used informtion reported by the prent bout the focl children to construct the following child well-being outcomes cross ll ge groups. Eduction Preschool or Hed Strt enrollment. This binry vrible is bsed on responses to Questions F10 nd F11_2 on the prent survey. The dult respondent ws first sked bout the number of different childcre rrngements or schools the child hd been in for t lest 10 hours week since the rndom ssignment dte (F10). If the child hd t lest one rrngement, the respondent ws then sked to describe the type of cre from the following list: Fmily-bsed cre in someone s home with other children. School- or center-bsed cre. Childcre provided in my home. If the response ws school- or center-bsed cre, then the indictor for preschool or Hed Strt enrollment ws coded s yes. 20 Respondents were lso sked bout two dditionl items relted to drug use tht re not used in the DAST-10 nd were not used to crete the drug buse outcome. E10c sked, Hve your friends or reltives known or suspected tht you used drugs? nd E10d sked, Hve you ever lost friends becuse of drugs? B -7

224 Appendix B. Construction of Outcomes School enrollment. This binry vrible is bsed on responses to Questions F6, F8, nd F9. The dult respondent ws sked bout enrollment of children ges 4 to 17 yers in school (ges 6 to 17) or childcre (ges 4 to younger thn 6). If the respondent nswered yes or volunteered informtion tht the child is homeschooled or on summer/school vction nd the response to the child s highest grde or yer of school completed ws not tht they were currently in ny form of childcre or preschool (F8), then enrollment ws indicted s yes. Also, if the respondent indicted tht the child s highest grde or yer of school tht he or she ever completed ws the 12th grde nd the child received high school diplom, then school enrollment ws lso indicted s yes. Childcre or school bsences (pst month). This vrible is treted s continuous bsed on responses to Question F13 for children ges 1 yer, 6 months to 17 yers, 11 months. Respondents were sked to report on the number of dys the child missed school in the pst month (or the lst month of school if the survey is conducted during the summer (F6_4)). The outcome is mesured using scle of 0 to 3: 0 = no bsences, 1 = 1 to 2 bsences, 2 = 3 to 5 bsences, 3 = 6 or more bsences. Number of schools ttended since rndom ssignment. This continuous vrible is bsed on responses to Question F12. The number of schools is top coded t four or more schools. The vlues of the outcome re 1 = one school. 2 = two schools. 3 = three schools. 4 = four or more schools. Grde completion. This binry vrible is bsed on responses to Question F12b. It mesures whether child ge 4 to 17 yers repeted grde since the time of rndom ssignment. Positive childcre or school experiences. This continuous vrible is bsed on responses to F17, in which prents ssess the child s childcre or school experiences, using these rtings 1 = mostly positive. 0 = both positive nd negtive. -1 = mostly negtive. Positive childcre or school ttitudes. This continuous vrible is constructed from responses to Question F16 in which the prent ssesses the child s ttitude towrd school or childcre. The question is worded s follows: How much does your child like school/childcre? The response options were 1 = not t ll. 2 = not very much. 3 = some. 4 = pretty much. 5 = very much. School grdes. This continuous outcome ws constructed using responses to Question F12c. The prent ws sked to describe the child s grdes in the most recently completed school term. The outcome uses 4-point scle with the following vlues. 1 = mostly Ds nd Fs. 2 = mostly Cs. 3 = mostly Bs. 4 = mostly As. B -8

225 Appendix B. Construction of Outcomes Childcre or school conduct problems. This binry vrible is bsed on responses to Questions F14 nd F16. The outcome mesures whether the prent hs received reports of conduct problems from the childcre center or school or whether the child hs been suspended or expelled from the childcre provider or school. The outcome uses the following vlues. 0 = no clls to prent. 1 = prent got reports of bd conduct or suspension/expulsion. Helth Poor or fir helth (t the time of the survey). This binry outcome is bsed on responses to Question F18, which sks the prent to ssess the child s helth t the time of the survey. Allowble responses were excellent, very good, good, fir, or poor. The outcome hs vlue of 1 if responses were fir or poor. Well-child check (in 12 months before the survey). This binry outcome is bsed on responses to Question F19. Prents re sked whether outcome mesures the percentge of focl children who received physicl exmintion or well-child checkup in the yer before the survey, bsed on the prent s report. Regulr source of (helth) cre. This binry outcome is bsed on responses to Question F18. Sleep problems. This continuous vrible is bsed on responses to Questions F26i, F26j, nd F26k. The prent is sked to report on the frequency of two indictors of sleep problems tiredness on wking nd tiredness during the dy. The llowble responses re 1 = lmost lwys. 2 = most dys. 3 = sometimes. 4 = rrely. 5 = lmost never. The questions re worded s follows: Would you sy tht [CHILD] hs difficulty wking up in the morning? [CHILD] hs difficulty wking up on school dys? [CHILD] is tired during the dy? To crete the outcome, the study tem reverse-coded the response options (for exmple, lmost never = 1; lmost lwys = 5). The vlue of the outcome thus rnges from 1 to 5 with higher vlues indicting greter tiredness on wking nd during the dy. Low birth weight. This binry outcome is mesured only for children born since rndom ssignment who were t lest 1 yer old t the time of the prent survey. The mesure is clculted using responses to Question F2, in which the prent ws sked to report the birth weight of the child in pounds nd ounces. Birth weights below 5 pounds, 8 ounces were coded s 1 to indicte low birth weight (WHO, 2010: 152). Behviorl Strengths nd Chllenges Behviorl problems. This continuous vrible is mesured s the stndrdized totl difference score from the Strengths nd Difficulties Questionnire (SDQ). The SDQ is behviorl nd personlity ssessment. The totl problem score mesures emotionl symptoms, conduct problems, hyperctivity, nd peer problems. The outcome is mesured using responses to Questions F21 (for 3-yer-olds), F22 (4- to 10-yer-olds), nd F23 (11- to 17-yer olds) on the prent survey. Prents indicted whether series of sttements were not true, somewht true, or certinly true for the child during the 6 months before the survey. B -9

226 Appendix B. Construction of Outcomes Prosocil behvior. This continuous vrible is mesured s the stndrdized prosocil domin score from the SDQ. The SDQ test ws dministered to prents during the prent survey, sking bout ll focl children between ges 3 yers nd 17 yers, 11 months. Prents were sked the sme questions for ll focl children, lthough the wording of some questions vried depending on the ge group: 3-yer-olds (Question F21 y), 4- to 10-yer-olds (Question F22 y), nd 11- to 17-yer-olds (Question F23 y). The SDQ contins 25 questions split into five sections: emotionl symptoms, conduct problems, hyperctivity, peer problems, nd prosocil. Ech section sks set of five descriptions relted to the forementioned section hedings. The responses to the descriptions re 1 = not true, 2 = somewht true, nd 3 = certinly true. /The items from ech section follow. Emotionl symptoms Often unhppy, depressed, or terful. Often complins of hedches, stomch-ches, or sickness. Mny worries or often seems worried. For children ges 11 to 17: Nervous in new situtions; esily loses confidence. For children ges 3 to 10: Nervous or clingy in new situtions; esily loses confidence. Mny fers; esily scred. Conduct problems For children ges 4 to 17: Often lies or chets. For children ge 3: Often rgumenttive with dults. Often loses temper. Generlly well behved; usully does wht dults request. For children ges 11 to 17: Often fights with other youth or bullies them. For children ges 3 to 10: Often fights with other children or bullies them. For children ges 4 to 17: Stels from home, school, or elsewhere. For children ge 3: Cn be spiteful to others. Hyperctivity Esily distrcted; concentrtion wnders. Restless, overctive; cnnot sty still for long. Constntly fidgeting or squirming. For children ges 4 to 17: Thinks things out before cting. For children ge 3: Cn stop nd think things out before cting. For children ges 4 to 17: Good ttention spn; sees work through to the end. For children ge 3: Good ttention spn; sees tsks through to the end. Peer problems Generlly liked by other youth/children. For youth ges 11 to 17: Would rther be lone thn with other youth. For children ges 3 to 10: Rther solitry; prefers to ply lone. B-10

227 Appendix B. Construction of Outcomes Hs t lest one good friend. Picked on or bullied by other youth/children. Gets long better with dults thn with other children. Prosocil Considerte of other people s feelings. For children ges 11 to 17: Shres redily with other youth; for exmple, shres books, gmes, nd food. For children ges 3 to 10: Shres redily with other children; for exmple, shres toys, trets, nd pencils. Helpful if someone is hurt, upset, or feeling ill. Kind to younger children. Often offers to help others (prents, techers, other children). The descriptions re more often worded with negtive vlence not true = 0 points, somewht true = 1 point, nd certinly true = 2 points such tht more negtive behviors or tendencies re given more points. Some questions re worded with positive vlence, however, such s Generlly liked by other children. Questions of positive vlence re scored so tht not true is given 2 points, somewht true is given 1 point, nd certinly true is given 0 points. This scoring rrngement is such tht higher scores for the Totl Difficulties score indicte more negtive behvior. Ech of the five sections is given n individul score rnging from 0 to 10. Only the first four sections re included in the overll Totl Difficulties score; the prosocil scle is excluded. The Totl Difficulties score hs rnge from 0 to 40. To compute totl score, t lest two-thirds (three of five) of the questions within ech domin hd to be nswered. If one or two items were missing within domin, the verge score of those items ws multiplied by 5 to impute the totl score for tht domin. If more thn three items in ny domin were missing, both the domin score nd the totl problems score were counted s missing. The scores were lso stndrdized by ge nd gender. Ages 1 Yer to 3 Yers, 6 Months Met developmentl milestones. This outcome is defined s scoring bove the typicl developmentl cutoffs in ll five domins of the Ages nd Stges Questionnire (ASQ-3). This binry vrible indictes if the child pssed ll five domins in the ASQ-3 corresponding to his or her ge. The study used 12 versions of the ASQ-3 for children ges 11 to 44 months. The ASQ-3 is developmentl ssessment for children of ges from 1 month to 5 yers, 6 months. The study tem dministered the test to the prents, sking bout ll focl children between the ges of 1 yer nd 3 yers, 6 months (12 to 41 months). The test ws typiclly dministered directly fter the prent survey. The study tem dministered 12 versions of the test cross the ge groups. 1. The 12-month test for focl children ges 11 months through 12 months, 30 dys. 2. The 14-month test for focl children ges 13 months through 14 months, 30 dys. 3. The 16-month test for focl children ges 15 months through 16 months, 30 dys. 4. The 18-month test for focl children ges 17 months through 18 months, 30 dys. 5. The 20-month test for focl children ges 19 months through 20 months, 30 dys. 6. The 22-month test for focl children ges 21 months through 22 months, 30 dys. 7. The 24-month test for focl children ges 23 months through 25 months, 15 dys. 8. The 27-month test for focl children ges 25 months nd 16 dys through 28 months, 15 dys. B-11

228 Appendix B. Construction of Outcomes 9. The 30-month test for focl children ges 28 months nd 16 dys through 31 months, 15 dys. 10. The 33-month test for focl children ges 31 months nd 16 dys through 34 months, 15 dys. 11. The 36-month test for focl children ges 34 months nd 16 dys through 38 months, 30 dys. 12. The 42-month test for focl children ges 39 months through 44 months, 30 dys. The test is structured the sme for ech version of the test, lthough the questions differ. Ech test hs six sections: (1) communiction, (2) gross motor, (3) fine motor, (4) problemsolving, (5) personl-socil, nd (6) overll. The first five sections hve six questions tht cn be nswered yes, sometimes, or not yet. The lst section Overll hs between 8 nd 10 open-ended questions tht re not included in the finl score. Ech yes nswer receives 10 points, sometimes nswer receives 5 points, nd not yet nswer receives 0 points. The scores for ech section rnge from 0 to 60. A rw score ws clculted seprtely for ech of the five sections. A section is scored when t lest four of the six questions in the section re nswered. When ll six questions re nswered, the scores from those six vilble nswers re summed together. When section hs four or five nswers, the missing scores re ssigned vlue derived from the verge of the vilble scores, nd ll six scores re summed. The ASQ-3 hs ntionl norms for scores for ech domin by ech ge version nd rw score tht cn be used s dignostic cutoff point for the domin set t two stndrd devitions below the men. Scores greter thn the cutoff re considering pssing, wheres scores t or below the cutoff my indicte the potentil presence of developmentl dely in tht domin nd cn be used for mking referrl for more comprehensive ssessment. The tem clculted z-scores for ech domin by subtrcting the domin rw score from the ntionl domin men (for the pproprite ge version) nd dividing by the ntionl domin stndrd devition. For the outcome mesure, rw scores in ech domin were compred with the cutoff scores nd scores bove the cutoff were counted s pssing. The outcome then ws whether children pssed ll vlid domins. If children were missing one of the five domins, they were still included in the outcome nd were ssessed on whether they pssed ll four nonmissing domins. Nerly one-fourth (23.6 percent) of cses hd discrepncies between the ge of the child in our survey nd the version of the ASQ-3 dministered. In 98 percent of these cses, the structure of the ASQ-3 survey llowed for imputtion of scores, resulting in finl missing dt rte of 2 percent becuse of ge-version discrepncies (detils bout imputtion nd missing dt re vilble on request). Ages 3 Yers, 6 Months to 7 Yers Verbl bility. This outcome is mesured s the ntionlly stndrdized score from the Woodcock-Johnson III (WJ III) Letter-Word Identifiction test. Mth bility. This outcome is mesured s the ntionlly stndrdized score from the WJ III Applied Problems test. The study tem dministered two tests from the WJ III tests of chievement to eligible smple children ges 3 to 7 yers the Letter-Word Identifiction test, with 76 possible questions, nd the Applied Problems test, with 63 possible questions. The interviewers begn tests t Question1 regrdless of ge. Interviewers did not clculte finl rw scores in the field becuse it is subject to error. The nlysis tem independently clculted rw scores. For both the Letter-Word Identifiction nd Applied Problems tests, the study tem clculted rw scores bsed on series of rules. First, children were llowed to refuse the test either t the beginning or during the test. Refusls were coded s missing test scores. The rw score ws sum of ll the correct nswers, strting t Question 1, until the child nswered six consecutive questions incorrectly. Ech question ws weighted the sme, with vlue of 1 indicting correct response nd vlue of 0 indicting n incorrect response. Exhibit B-1 shows the finl distribution of the number of children who completed the WJ III tests. B-12

229 Appendix B. Construction of Outcomes Exhibit B-1. Children s Completion Rtes for WJ III Letter-Word Identifiction nd Applied Problems Tests WJ III Test Distribution n % Totl children completed WJ III Both tests (Letter-Word Identifiction nd Applied Problems) completed Only Letter-Word Identifiction test completed Only Applied Problems test completed WJ III = Woodcock-Johnson III. Source: Woodcock-Johnson Tests of Acdemic Achievement The study tem entered rw scores into the WJ III Compuscore softwre to clculte z-scores tht re ge nd gender djusted.21 Rw scores were entered into Woodcock-Johnson III Test of Acdemic Achievement Form A for Stndrd Bttery tests 1 nd 10, where test 1 is the Letter-Word Identifiction test nd test 10 is the Applied Problems test. The study tem exported the key informtion into n Excel file nd entered rw scores into the WJ III Compuscore softwre tht converts rw scores into the z-scores used in report nlyses. Dt from the Excel file were copied nd psted rther thn entered mnully into Compuscore to reduce dt entry error. Within the Compuscore softwre, the following informtion ws entered: first nme, gender, study identifiction number, nd dte of birth, nd dte of testing. On exporting dt from the Compuscore softwre, the Norm Bsis of ge nd the Stndrd Set of scores were selected for inclusion in the export, resulting in comm-delimited file tht includes, mong the stndrd set of scores, the z-score of ech test for ech child. Executive functioning (self-regultion). This outcome is mesured with the score on the Hed Toes Knees Shoulders (HTKS) ssessment. HTKS is development ssessment testing inhibitory control, ttention, nd working memory. The study tem dministered this test directly to focl children ges 3 yers, 6 months to 7 yers, 11 months. All children were given the sme test, regrdless of ge. The test consists of 20 questions divided into two prts. The test begn with demonstrtion of the exercise in which the children were instructed to touch their toes when they were told to touch your hed nd to touch their hed when told to touch your toes, in effect doing the opposite of wht they were told. After some prctice nd repeted reminders to mke sure tht the children understood the instructions, the ssessment begn. The first 10 questions instructed the children to touch your hed or touch your toes. The responses to touch your hed would be for the child to touch his or her toes (the correct response), to motion towrd touching his or her hed nd then correct him/herself nd touch his or her toes ( self-corrected response), or to touch his or her hed (the incorrect response). Ech correct response received 2 points, ech self-corrected response received 1 point, nd ech incorrect response received 0 points. If child hd 6 or more incorrect responses in the first 10 questions, the test ws discontinued. Children who nswered 5 or fewer of the 10 questions incorrectly moved to the second set of 10 questions. For the second prt, children were instructed to touch their knees when told to touch your shoulders nd to touch their shoulders when told to touch your knees. This principle of doing the opposite of wht is told ws the sme for this prt of the test the only chnge ws the body prt touched. The children were tken through second demonstrtion in which they prcticed touching their knees when told to touch their shoulders nd touching their shoulders when instructed to touch their knees. After prctice nd only one reminder, the second set of 10 questions begn. 21 Reference the WJ III technicl mnul for detils bout the z-score (McGrew, Shrnk, nd Woodcock, 2007). B-13

230 Appendix B. Construction of Outcomes In the second set of 10 questions, children received ll four instructions touching the hed when instructed to touch the toes nd vice vers nd touching the shoulders when instructed to touch the knees nd vice vers, dding to the complexity of the test. The scoring ws the sme, with correct responses receiving 2 points, self-corrected responses receiving 1 point, nd incorrect responses receiving 0 points. Missing vlues were imputed if two-thirds of the overll questions children were dministered were nonmissing. The scores ssigned to these missing vlues were the verge from the nswered items multiplied by the totl number of trils ech child ws eligible for, so tht, if child were dministered 10 trils, the imputtion would be the verge score of the nswered items times 10. Children could receive totl score of between 0 nd 40 points from the 20 questions. Ages 8 to 17 Yers The study tem collected informtion from children nd youth ges 8 to 17 yers on the child survey from which severl outcome mesures were constructed. The outcomes re described in this section. Anxiety. Question A1 on the child survey is used to crete n indictor of nxiety. Question A1 is the A-Trit scle from the Stte-Trit Anxiety Inventory for Children, or STAIC (Spielberger et l., 1973). Scores rnge from 20 to 60, with higher scores indicting greter nxiety. The scle is proprietry, so only prtil list is shown here. Respondents reported on the frequency with which they felt severl items using these response options 1 = hrdly ever. 2 = sometimes. 3 = often. Exmples of the items re I worry bout mking mistkes. I hve trouble deciding wht to do. I worry bout things tht my hppen. Fers. This outcome is bsed on responses to Questions B1 to B33 of the child survey. Respondents reported on the extent to which they hd fers using these response options. 1 = not t ll. 2 = some. 3 = lot. Questions B1 to B33 sked bout the following fers: spiders, getting sick, being robbed, hving no friends, dogs, wht will hppen in the future, hving no plce to live, something bd hppening to people in my fmily, snkes, getting bd grdes, people fighting, being tesed, wht other people think of me, being hit by cr or truck, drug delers, being lone, flunking school, gngs, being lost, rts, tht other children/tweens will not wnt to ply/spend time with me, police, hving no plce to sleep, dying, nightmres, being hungry, hving no food to et, being sent to the principl, guns, fire, losing my fvorite stuff, I worry bout my prents, I worry bout my brothers nd sisters, I worry bout myself. The response scores were summed, yielding totl scores rnging from 33 to 99, with higher scores indicting greter fer. Substnce use. This outcome hs vlues of 0 to 2 nd is bsed on responses to Questions D6 to D23 on the child survey. The outcome is mesured with 23 items from the Centers for Disese Control nd Prevention, or CDC, 2011 Youth Risk Behvior Survey. This outcome mesures whether the child hd used tobcco, lcohol, or mrijun in the pst 30 dys or hd ever used other substnces (cocine, inhlnts, steroids ges 8 to 17 or ecstsy, meth, heroin, controlled prescription drugs, or injected drugs ges 13 to 17 only). B-14

231 Appendix B. Construction of Outcomes Gol-oriented thinking (positive thinking). This outcome is mesured using responses to Questions G1 to G6 on the child survey. These items re modified version of the Children s Hope Scle. Scores rnge from 6 to 30 with higher scores indicting greter hope. Respondents indicted how frequently they felt six items using these response options 1 = none of the time. 2 = little of the time. 3 = lot of the time. 4 = most of the time. 5 = ll of the time. The six items were 1. You think you re doing pretty well. 2. You cn think of mny wys to get the things in life tht re most importnt to you. 3. You re doing just s well s other kids your ge. 4. When you hve problem, you cn come up with lots of wys to solve it. 5. You think the things you hve done in the pst will help you in the future. 6. Even when others wnt to quit, you know you cn find wys to solve the problem. School effort. This outcome is constructed from responses to Questions E4 nd E5. Respondents were sked two questions bout their school effort in the month before the survey. Response options were 1 = could hve done lot better. 2 = could hve done little better. 3 = did bout s well s you could. 4 = did very well; could not hve done better. The questions re worded likes this In the lst month, how hrd hve you worked on your homework? In the lst month, how hrd hve you tried to work during the school dy? Arrests or police involvement. This binry outcome is constructed from responses from the prent to Questions F24 nd F25 on the prent survey. Prents were sked whether the child hd been rrested in the 6 months before the survey nd whether the focl child hd hd ny problems tht involved the police contcting the prent in the 6 months before the survey. B.5 Mesures of Self-Sufficiency The impct nlysis exmines effects of the four interventions on severl outcomes relted to self-sufficiency of smple members. These outcomes pertin to employment sttus (dult respondent), income sources (fmily), prticiption in eduction nd trining (dult respondent), food security (fmily), nd economic hrdship (fmily). B-15

232 Appendix B. Construction of Outcomes Employment sttus. Adult respondents reported on work ctivity in the week preceding the 18-month followup survey nd whether they hd worked t ny time since rndom ssignment. Those who hd worked since rndom ssignment nswered questions bout the number of jobs held since rndom ssignment nd the number of months worked since rndom ssignment. Smple members who reported hving worked for py in the week before the 18 month followup survey were sked to provide detils bout the number of hours worked per week nd ernings t the min job. The nlysts used this informtion to construct five outcomes. 1. Work for py in the week before the survey. This binry vrible is bsed on responses to Question C1. Dt were missing on this mesure for less thn 0.1 percent of cses. 2. Any work for py since rndom ssignment. This binry vrible is bsed on responses to Question C2. Dt were missing on this mesure for 0.1 percent of cses. 3. Clendr months worked for py since rndom ssignment. This continuous vrible is bsed on responses to Question C4. Respondents who reported tht they hd not worked since rndom ssignment were ssumed to hve worked 0 months since rndom ssignment. Dt were missing on this mesure for 1.2 percent of cses. 4. Number of hours worked per week t the min job. This continuous vrible is bsed on responses to Question C5.22 For dult respondents who sid they did not work in the week before the survey, the number of hours worked ws ssumed to be 0. Dt were missing on this mesure for 0.3 percent of cses. 5. Annulized ernings from the min job. This continuous vrible is bsed on responses to Questions C6 through C11 bout wges pid t the min job. The outcome mesures the nnulized ernings from the min job. For dult respondents who sid they did not work in the week before the survey, the nnulized ernings re ssumed to be 0. Dt were missing on this mesure for 2.1 percent of cses. Income sources/mounts. Question C12 on the 18-month followup survey sked whether the respondent or nyone in the respondent s fmily received income from vrious sources or vrious types of government ssistnce in the month before the survey. The nlysts used responses to these questions to construct binry outcomes to indicte receipt of the following types of income or ssistnce for the fmily. Dt were missing on ech of these mesures for 0.1 percent of cses. Ernings. Temporry Assistnce for Needy Fmilies, or TANF. Socil Security Disbility Insurnce, or SSDI. Supplementl Security Income, or SSI. Supplementl Nutrition Assistnce Progrm, or SNAP. Specil Supplementl Nutrition Progrm for Women, Infnts, nd Children, or WIC. Smple members lso were sked in Questions C13 through C19 to estimte totl nnul income for the fmily from ll sources for the most recently completed clendr yer (2011 or 2012) preceding the 18-month followup survey. The nlysts used responses to these questions to construct continuous outcome vrible mesuring totl nnul fmily income for the fmily. Dt were missing on this mesure for 3 percent of cses. 22 If the respondent reported hving more thn one job, the interviewers instructed the respondent to provide the number of hours worked t the min job. The min job ws defined s the job t which the respondent usully worked the most number of hours per week. B-16

233 Appendix B. Construction of Outcomes Eduction nd trining. The study tem used the dult survey to construct severl outcomes mesuring prticiption in eduction nd trining in the followup period. Adult respondents indicted whether they hd prticipted in ny eduction or trining ctivities since rndom ssignment. If so, they reported on the number of weeks they spent in trining. For up to six progrms reported, smple members reported on the type of progrm, using the following response options. 1 = regulr high school (HS), directed to HS diplom. 2 = preprtion for generl eductionl development (GED) exm. 3 = 2-yer college directed towrd degree. 4 = 4-yer college directed towrd degree. 5 = grdute courses. 6 = college courses not directed towrd degree. 7 = voctionl eduction outside college (business or technicl schools, employer or union-provided trining, or militry trining in voctionl but not militry skills). 8 = nonvoctionl dult eduction not directed towrd degree (bsic eduction, litercy trining, English s second lnguge). 9 = job serch ssistnce, job finding, orienttion to the world of work. The eduction nd trining outcomes re described s follows. Prticipted in ny school or trining lsting 2 weeks or more since rndom ssignment. This binry vrible is constructed from responses to Question C24. Dt were missing on this mesure for 0.2 percent of cses. Number of weeks in trining progrms since rndom ssignment. This continuous vrible is bsed on responses to Question C27. Dt were missing on this mesure for 0.9 percent of cses. Prticipted in school/cdemic trining lsting 2 weeks or more since rndom ssignment. This binry vrible is bsed on responses to Question C26. School or cdemic trining is defined s regulr high school directed towrd high school diplom, preprtion for GED exm, 2-yer college, 4-yer college, or grdute courses. Dt were missing on this mesure for 0.2 percent of cses. Prticipted in bsic eduction lsting 2 weeks or more since rndom ssignment (percentge of fmilies). This binry vrible is constructed from responses to Question C26. Bsic eduction is defined s nonvoctionl dult eduction (such s bsic eduction, litercy trining, or English s second lnguge) not directed towrd degree. Dt were missing on this mesure for 0.2 percent of cses. Prticipted in voctionl eduction/trining lsting 2 weeks or more since rndom ssignment (percentge of fmilies). This binry vrible is constructed from responses to Question C26. Voctionl eduction/trining is defined s voctionl eduction outside college such s business or technicl schools, employer- or union-provided trining, or militry trining in voctionl skills (not militry skills). Dt were missing on this mesure for 0.2 percent of cses. Food security/hunger. The study tem lso mesured food security on the dult survey in Questions C28 through C32. The first outcome is binry vrible tht equls 1 if household ws food insecure ccording to criteri used by the U.S. Deprtment of Agriculture (USDA).23 Survey respondents were sked series of questions used by USDA nd the U.S. Census Bureu to mesure food security. More thn two ffirmtive nswers to these questions men tht household is considered insecure. No cses hd missing dt on this mesure. For the second 23 See Nord, Andrews, nd Crlsen (2005). Our ssessment of food insecurity is bsed on two USDA short form metrics, which re scores ssigned to households bsed on nswers to six survey questions. B-17

234 Appendix B. Construction of Outcomes mesure, survey nswers re trnslted into food security scle, nd households with high scores on this scle re determined to be insecure. Dt were missing on this mesure for 0.4 percent of cses. Respondents were sked to indicte whether ech of the following food security sttements were often true, sometimes true, or never true in the 30 dys before the survey. We couldn t fford to et blnced mels. The food tht (I/we) bought just didn t lst, nd (I/we) didn t hve money to get more. Respondents were lso sked whether in the pst 30 dys they or other dults in the household hd ever cut the size of mels becuse they did not hve enough money for food. The respondents were lso sked whether in the 30 dys before the survey they hd ever eten less thn they thought they should becuse they did not hve enough money to buy food, if they hd ever been hungry but didn t et becuse they couldn t fford enough food, nd if they or other dults in the household ever did not et for whole dy becuse they did not hve enough money for food. Economic stressors. The nlysts lso mesured the economic hrdship reported by ech fmily t the time of the 18-month followup survey. Questions 33 through 33d sked the dult respondents to report on the frequency with which the fmily experienced four items relted to economic hrdship in the 6 months before the survey. The response options were (1 = never; 2 = once in while; 3 = firly often; nd 4 = very often). The question ws worded s follows. How often does it hppen tht you do not hve enough money to fford The kind of medicl cre your fmily should hve? The kind of clothing your fmily should hve? The leisure ctivities your fmily wnts? Your rent? Question C34 sked how the fmily s finnces usully work out t the end of the month, with these possibly response codes: 1 = some money left over. 2 = just enough money to mke ends meet. 3 = not enough money to mke ends meet. For both questions, higher vlues indicte higher economic stress. The economic stress outcome is clculted for cses in which four of the five items (Question 33 through 33d nd Question 34) re nonmissing. For Questions 33 through 33d, the responses were recoded into scle rnging from less economiclly stressed to more economiclly stressed where 1 = -1, 2 = -0.33, 3 = 0.33, 4 = 1. For Question 24, responses were recoded s 1 = -1, 2 = 0, 3 = 1. The nonmissing recoded responses were then verged. The economic stress scle rnges from -1 to 1, with higher vlues indicting higher economic stress. Dt were missing on this mesure for 0.5 percent of cses. B.6 Progrm Use Outcomes The study tem used the Progrm Usge Dt to crete the outcomes reported in the report s progrm use exhibits. The study tem creted ech type of progrm use outcome for ech of seven progrm types: (1) permnent housing subsidy (SUB), (2) community-bsed rpid re-housing (CBRR), (3) trnsitionl housing, (4) permnent supportive housing (PSH), (5) public housing, (6) project-bsed vouchers/section 8 projects, nd (7) emergency shelter. Ever used prticulr progrm type. These series of binry vribles were coded s 1 if ny monthly binry indictor from the clendr month of rndom ssignment to the clendr month of the 18-month followup survey response indicted use of the progrm type. B-18

235 Appendix B. Construction of Outcomes Used prticulr progrm type in the survey month. These series of binry vribles were coded s 1 if the monthly binry indictor from the clendr month of the 18-month followup survey response indicted use of the progrm type. Number of months of use of prticulr progrm type. These continuous vribles were defined using ssumptions bout how fmilies use the vrious homeless nd housing progrms. Becuse the monthly indictor vribles in the Progrm Usge Dt re coded s 1 if prticulr progrm ws used t lest 1 night of prticulr clendr month, simply counting the number indictor vribles equl to 1 would systemticlly inflte mesures of progrm use. The study tem ssumed Entry to ll progrm types could hppen t ny time during the month. Exits from emergency shelter, trnsitionl housing, nd PSH could hppen t ny time during the month. Exits from SUB, CBRR, public housing, nd project-bsed vouchers/section 8 projects lwys hppened t the end of the month, becuse ssistnce is provided in monthly increments. These ssumptions were developed nd confirmed with prctitioners in the field. In ddition, the study tem ssumed ll stys in the followup survey month (for ll progrm types) extended to the end of the month, becuse the end of the observtion window ws n rtifct of dt collection. A single sty of progrm type ws identified in the dt by month indictors before nd fter sty with no use of tht prticulr progrm type. Using these ssumptions s bsis for correcting counts ment For stys longer thn 1 month Subtrcting 1 month from counts of clendr months for emergency shelter, trnsitionl housing, nd PSH stys. Subtrcting 1/2 month for stys in SUB, CBRR, public housing, nd project-bsed vouchers/section 8 projects. For stys tht lsted single clendr month Stys for emergency shelter, trnsitionl housing, nd PSH were shortened to 1/4 month (1/4 month is the expected length ssuming tht entry nd exit re eqully likely t ny point in the month). Stys for SUB, CBRR, public housing, nd project-bsed vouchers/section 8 projects were shortened to 1/2 month (1/2 month is the expected length ssuming tht entry is eqully likely t ny point in the month nd tht exit occurs t the end of the month). Appendix E shows impct estimtes for dditionl progrm use outcomes. The outcomes tht mesure ny use of prticulr progrm type (or types) during months 0 to 18 or months 7 to 18 re coded s 1 if ny monthly binry indictor during the relevnt time period indicted use of the progrm type (or types). The outcomes tht mesure number of months of emergency shelter or trnsitionl housing during months 0 to 18, number of months of emergency shelter during months 0 to 18, nd number of months of trnsitionl housing during months 0 to 18 re creted in nerly identicl mnner to the number of months outcomes described previously. The only difference is how stys tht include the 18th month fter rndom ssignment re counted. If the 19th month indictor showed use of the relevnt progrm type(s), then it ws ssumed tht the sty continued through the end of the 18th month. If the 19th month indictor showed no use of the relevnt progrm type(s), then it ws ssumed tht the sty ended t some point during the month. For stys of more thn 1 month tht included the 18th month, either full month in the 18th month ws counted (if the sty continued to the 19th month) or 1/2 month in the 18th month ws counted (if the sty did not continue to the 19th month). For single-clendr-month stys in the 18th month, either 1/2 month in the 18th month ws counted (if the sty continued to the 19th month) or 1/4 month in the 18th month ws counted (if the sty did not continue to the 19th month). B-19

236 Appendix C. Anlysis Methods APPENDIX C. ANALYSIS METHODS The impct estimtion models for fmily, dult, nd child outcomes re described in Chpter 4 of the Fmily Options Study short-term impcts report. This ppendix provides dditionl detils bout the covrites used in the impct models, imputtion of missing dt, fmily/dult weights, child weights, nd the multiple comprisons djustment for confirmtory hypothesis tests. C.1 Covrites Covrites in the impct models improved the precision of the estimtes. Becuse individuls were rndomly ssigned to control nd tretment groups, the ddition of these covrites does not ffect the expected vlue of the estimte itself. All covrites hd to be chrcteristics tht were known (or determined) before rndomiztion. In selecting covrites, the study tem considered (1) the importnce of the vrible in predicting the outcomes of interest, (2) the extent of vrition on the vrible for the smple, nd (3) the completeness of the dt. A full set of covrites mesured in the bseline survey ws included in the impct models for housing stbility, dult well-being, nd self-sufficiency outcomes. Becuse of smller smple sizes, more limited sets of covrites were included in the impct models for fmily preservtion nd child outcomes. The superscript indictes those covrites included in the impct model for fmily preservtion outcomes. The superscript b indictes those covrites included in the impct model for child outcomes. Continuous Vribles Age of fmily hed t bseline (liner), ge squred (qudrtic)b. Number of children with fmily in shelter. Annulized current ernings. Fmily income (liner ctegories: $0 to < $5,000; $5,000 to < $10,000; $10,000 to < $15,000; $15,000 to < $20,000; $20,000 to < $25,000; $25,000; nd income ctegories squred). Totl yers styed with fmily or friends becuse of economic necessity in pst 5 yers s n dult. Totl yers previously homeless in your life before entering the current shelter. Binry Vribles Rce/ethnicity (ctegories: White; Blck or Africn-Americn; Asin, Ntive Hwiin, or other Pcific Islnder; Hispnic or Ltino; other)b. Gender. Mritl sttus (ctegories: divorced; mrried; single/never mrried; widowed). Children of certin ge group (ctegories: fmily hs child younger thn ge 1; fmily hs child between ge 1 nd 5 yers; fmily hs child between ge 13 nd 17 yers). Children not with fmily in shelter t bseline (ctegories: ny child; two or more children). C -1

237 Appendix C. Anlysis Methods Number of children with fmily in shelter (ctegories: one child; two children; three children; four or more children). Hve spouse or prtner tht is with the fmily in shelter t bseline. Hve spouse or prtner tht is not with the fmily in shelter t bseline. Pregnnt t bseline. Any helth problems (hs self-reported poor helth; hs dibetes; hs nemi; hs high blood pressure; hs hert disese; hd stroke; hs heptitis/liver problems; hs rthritis, rheumtism, joint problems; hs chest infection, cold, cough, bronchitis; hs pneumoni; hs tuberculosis; hs cncer; hs problems wlking, lost limb, or other mobility impirment; hs gonorrhe, syphilis, herpes, chlmydi, other sexully trnsmitted diseses; is HIV (Humn Immunodeficiency Virus) positive; hs AIDS (Acquired Immune Deficiency Syndrome); uses drugs intrvenously; hs other medicl condition). Severe psychologicl distress t bseline. Post-trumtic stress disorder (PTSD) symptoms t bseline. A fmily member hs disbility tht limits or prevents the household hed from working for py. Fmily hed hs disbility tht limits or prevents working for py. Substnce buse problems (drug or lcohol). Highest level of eduction (ctegories: less thn high school diplom; high school diplom; more thn high school diplom or generl eductionl development). Working for py t bseline. Ever worked for py. Unemployment (ctegories: no work in the pst 6 months; no work in the pst 24 months). Receipt of vrious types of public ssistnce t bseline (ctegories: ny helth insurnce Medicid, stte helth insurnce, Stte Children s Helth Insurnce Progrm; Supplementl Nutrition Assistnce Progrm, or SNAP; Supplementl Security Income (SSI) or Socil Security Disbility Income (SSDI); Temporry Assistnce for Needy Fmilies, or TANF; unemployment insurnce; Specil Supplementl Nutrition Progrm for Women, Infnts, nd Children). Fmily income is under $5,000. Owned or rented own house or prtment before entering shelter. Number of months since fmily hd regulr plce to sty nd months squred. Previously styed with fmily or friends becuse of economic necessity. Previously experienced homelessness. Pst evictions, lese violtions, or problems with lndlord. Ever convicted of felony. Ever been in foster cre s child (foster home, group home, or ny other kind of institution). Ever homeless s child. Ever experienced intimte prtner violence in dulthood. Ineligible fmilies (ctegories: 1 to 7, indicting the fmily ws not eligible for n vilble tretment group)b. Site loction rndom ssignment regime interction terms (ctegories: Almed County; Atlnt; Bltimore; Boston; Connecticut; Denver; Honolulu; Knss City; Louisville-1; Louisville-2; Minnepolis; Phoenix; Slt Lke City-1; Slt Lke City-2)b. C -2

238 Appendix C. Anlysis Methods Binry Vribles for Child Outcomes Only Focl child ge. Focl child gender. In ddition to including these sets of covrites, the impct models for the pooled comprisons (permnent housing subsidy [SUB] + community-bsed rpid re-housing [CBRR] + project-bsed trnsitionl housing [PBTH] vs. usul cre [UC], SUB + CBRR vs. PBTH, SUB + PBTH vs. CBRR, CBRR + PBTH vs. SUB) included interction terms between site/rndom ssignment regime nd rndomiztion set (to correctly control for differing rndom ssignment rtios cross sites nd ssignment groups). Becuse of very smll smple sizes, the covrite set for the low birth weight child outcome ws limited to prent s ge t bseline, focl child gender, nd (for pooled comprisons only) the interction terms between site/rndom ssignment regime nd rndomiztion set. Finlly, few outcomes included one or two dditionl covrites to control for closely relted bseline vribles (when these bseline vribles were not lredy included in the min covrite set). Outcome: nyone in fmily hd ernings in pst month; extr covrite: nyone in fmily hd ernings t bseline. Outcomes: nyone in fmily received SSI in pst month, nyone in fmily received SSDI in pst month; extr covrite: nyone in fmily received SSI t bseline. Outcome: dult helth in pst 30 dys ws poor or fir; extr covrites: dult helth in pst 30 dys ws poor, dult helth in pst 30 dys ws fir. Outcomes: lcohol dependence or drug buse, lcohol dependence, drug buse; extr covrites: drug buse t bseline, behviorl helth problem t bseline. Outcomes: gol-oriented thinking, psychologicl distress, PTSD symptoms in pst 30 dys; extr covrite: behviorl helth problem t bseline. C.2 Missing Dt nd Imputtions Although respondents were sked to complete ll questions from the bseline survey, some dt in the Fmily Options Study remined missing. Overll, most covrites used in the imputtion models hd no missing dt. Only eight of the covrites hd more thn 1 percent missing dt nd no covrites hd more thn 5 percent missing. To ccount for missing dt on covrites, the study tem used single stochstic imputtion using SAS s PROC MI to impute missing covrite vlues.24 This method ssigns vlues to missing vribles using regression model tht predicts the vlue of the missing vrible bsed on other chrcteristics of the smple member nd the responses of other study prticipnts who re similr. The chrcteristics used in the imputtion include ll covrites used in the impct model. C.3 Fmily/Adult Weights The study chieved n 81-percent response rte for the 18-month followup survey. Nonresponse rises two concerns. First, nonresponse to followup survey used to mesure outcomes presents chllenge to the internl vlidity of the study if the intervention groups (tht is, SUB, CBRR, PBTH, nd UC) hve different ptterns of nonresponse. Second, followup survey nonresponse cn threten the generlizbility of results to the entire enrolled smple if survey nonrespondents differ from respondents, even if they do so symmetriclly cross rndomiztion rms. 24 Single stochstic imputtion dds rndom error term to every imputed vlue so tht the dt do not hve rtificilly low vribility. This vrying component is rndomly drwn from distribution with the sme vrince s the observed vlues. C -3

239 Appendix C. Anlysis Methods To ddress both of these issues, the study tem prepred 10 sets of weights tht djusted for dult survey nonresponse to the 18-month survey: 1 set for ech pirwise nd pooled comprison.25 The weights were used in the impct regressions for the outcomes in this report tht re defined t the fmily level nd t the dult respondent level. These weights were constructed by, (1) for ech intervention group within pirwise comprison (or ech side of pooled comprison), seprtely regressing dummy vrible for survey response on the sme bseline chrcteristics included in the impct model nd using the results to generte propensity to respond for ech fmily26; (2) for ech intervention group within pirwise comprison (or ech side of pooled comprison), dividing ech group into quintiles bsed on its modeled propensity; (3) within ech intervention group-quintile, the totl number of smple fmilies in the quintile divided by the number of respondent fmilies in the quintile clculted the weights for respondents. This lst step rises the representtion of respondent fmilies to the level of the full smple in the weighted dt, thereby restoring the composition of the nlysis dt to tht of the full smple on the fctors used to estimte propensities to respond. C.4 Child Weights The study tem prepred 50 sets of weights to be used for estimting impcts on child outcomes: 5 sets for ech of the 10 pirwise nd pooled comprisons. The 5 sets of weights correspond to the 5 types of dt used to construct child outcomes: 1. Prent-report survey dt (from the 18-month dult survey). 2. Ages nd Stges Questionnire (ASQ-3) dt. 3. Woodcock-Johnson III (WJ III) ssessment dt. 4. Hed Toes Knees Shoulders (HTKS) ssessment dt. 5. Child survey dt. The weights for the prent-reported outcomes were clculted s! i = F"'%&, )( #+*)(+#!#%$ℎt; ℎ$%!S"%" ($'& "$#ℎti, where! i = the child weight for prent-reported outcomes for child j in fmily i. F"'%&, )( #+*)(+#!#%$ℎt; = the fmily/dult nonresponse weight for fmily i (described in Section C.3). ℎ$%!S"%" ($'& "$#ℎti = the inverse probbility of being selected s focl child for child j in fmily i. (The focl child selection process is described in Appendix A, Section A.6.) The construction of weights to ddress survey nonresponse is discussed in Little (1986). 26 The purpose of the nonresponse regressions ws purely predictive, rther thn inferentil, which implied tht the number of covrites in the model ws not of concern (s it ws in the impct regressions). Thus, rther thn using single stochstic imputtion to ddress missing covrite vlues for the nonresponse regressions, ll missing vlues were imputed s the vlue 0. Then, in ddition to the impct model bseline covrites, the regression models lso included dummy vribles tht indicted when vlues for covrites were missing. 27 Section A.6 notes tht fter two focl children were selected for fmily, the focl child screening cesed. Therefore, collection of informtion for screening criteri other thn dte of birth ws not performed for every child in the study fmilies. For nonscreened children, the study tem used other informtion collected in the survey bout whether ech child ws currently living with the fmily to determine ex-post eligibility for selection (to clculte selection probbilities for selected children). It ws ssumed tht children currently living with the fmily would be eligible for focl child selection (if ge ws in trgeted rnge), nd it ws ssumed tht children not currently living with the fmily would be ineligible (regrdless of ge). The ssumption of ineligibility for unscreened children not currently living with the fmily ws bsed on the fct tht most screened children who were not currently living with the fmily did not meet the extr criteri necessry for eligibility: for 89 percent of these children, the prent either did not regulrly spend time with the child or ws not t lest somewht fmilir with the child s ctivities. C -4

240 Appendix C. Anlysis Methods The weights for other types of outcomes were clculted s [ " #!c ]i =! i ℎ$%!N'& ")('&)" "$#ℎt[ " #!c ]i, where [ " #!c ]i = the child weight for [dt source] (either ASQ-3, WJ III, HTKS, or child survey) for child j in fmily i. ℎ$%!N'& ")('&)" "$#ℎt[ " #!c ]i = the child nonresponse weight for [dt source] for child j in fmily i. The child nonresponse weights were clculted in three-step process: (1) for ech intervention group within pirwise comprison (or ech side of pooled comprison), seprtely regressing dummy vrible for unit response to the questionnire, ssessment, or survey on limited set of predictors28 nd using the results to generte propensity to respond for ech child to the prticulr instrument; (2) for ech intervention group within pirwise comprison (or ech side of pooled comprison), dividing the group into quintiles bsed on its modeled propensity; (3) within ech intervention group-quintile, clculting the nonresponse weight for the respondents the weighted totl number of focl children in the quintile divided by the weighted number of respondent children in the quintile, where the weights were the child selection weights (inverse probbility of focl child selection). The construction of the child weights from fmily nonresponse weights, focl child selection weights, nd child nonresponse weights implies tht, for ll child outcomes, the respondent smples re weighted to represent ll the ppropritely ged children in ll study fmilies.29 C.5 Multiple Comprisons Adjustment for Confirmtory Hypothesis Tests Sttement of the Problem Simply stted, the multiple comprisons problem is tht, s the number of hypothesis tests conducted grows, the likelihood of finding sttisticlly significnt impct somewhere mong the tested outcomes simply by chnce increses fr bove the desired risk level for producing flse positive results. This multiple comprisons problem is prticulrly slient for the Fmily Options Study becuse the number of hypothesis tests performed is extremely lrge. Becuse the study design is bsed on four intervention rms, the study exmines impcts in six pirwise comprisons nd four pooled comprisons. For ech of these comprisons, the study looks t five outcome domins (housing stbility, self-sufficiency, dult well-being, child well-being, nd fmily preservtion), with ech domin contining severl outcome vribles. The multiple rms, multiple domins, nd multiple outcomes cumultively generte n extremely lrge number of hypothesis tests. Given this lrge number of tests, the probbility of finding n impct, even in the cse of no true impcts, is quite lrge, well bove the nominl 10-percent level. In prticulr, the probbility of finding t lest one significnt impct t the.10 level in k independent tests when ll true impcts re 0 is given by Eqution The reltively smll smple sizes for ech collection instrument necessitted smller set of predictor vribles thn tht used to crete fmily/dult nonresponse weights. The predictors included: child s ge, child s gender, prent respondent s ge, prent respondent s gender, prent s rce/ethnicity (ctegories: White; Blck or Africn-Americn; Asin, Ntive Hwiin or other Pcific Islnder; Hispnic or Ltino; other), children not with fmily in shelter t bseline (ctegories: ny child; two or more children), children of certin ge group (ctegories: fmily hs child younger thn 1 yer, child ges 1 to 5 yers, child ges 13 to 17 yers), prent s substnce buse problems (drug or lcohol), prent ever convicted of felony, fmily income ctegory, fmily income under $5,000, number of children with the fmily t bseline, whether the dult respondent hs spouse or prtner t bseline (either in shelter or not in shelter), prent hd previously experienced homelessness, prent working for py t bseline, nd site loction rndom ssignment regime interction terms. 29 An implicit ssumption in this weighting method is tht, within n dult survey response propensity quintile, the distribution of numbers nd ges of children in the fmilies who did not respond t ll to the 18-month dult survey is the sme s tht of the fmilies who did respond to the 18-month dult survey. C -5

241 Appendix C. Anlysis Methods P! min p.10!e!" $# = 0 = k. (1) Thus, if 10 independent tests re performed, then the probbility of finding t lest one significnt impct t the.10 level often tken s the litmus test for successful intervention when ll true impcts re equl to 0 is =0.65; tht is, bout two-thirds of the time one would conclude n unsuccessful intervention is successful. When 20 independent tests re performed, the probbility is 0.88; tht is, nerly 9 times out of 10. In fct, with hundreds of tests, it is nerly certin to spuriously detect successful intervention, even if the intervention ws not truly successful for ny outcome. This probbility of finding t lest one significnt impct (or more generlly, rejecting t lest one null hypothesis) when ll true impcts equl 0 (or more generlly, when ll null hypotheses re true) in fmily of k tests is clled the fmilywise error rte (FWER). In generl, the FWER decreses s the k test sttistics used become more correlted (tht is, the outcome mesures tested become more closely relted), leding to somewht less risk of flse positive conclusions thn indicted in the previous numericl estimtes. Mny multiple comprison djustment procedures hve been devised to keep the FWER t or below the desired level (such s 0.05 or 0.10), some of which tke ccount of correltion mong outcomes. Study Response to the Problem The study tem took two steps to ddress the multiple comprisons problem. 1. Adjust the stndrd of evidence used to declre subset of individul impct estimtes sttisticlly significnt. The study tem divided the hypothesis tests into smll set of 7 confirmtory tests nd much lrger set of 723 explortory tests. The tem then used multiple comprisons procedure to djust the results of the 7 confirmtory tests to mintin the integrity of the sttisticl inferences mde t the confirmtory level. 2. Prespecify impcts to present in the executive summry. The study tem prespecified the impcts on 18 key outcomes in the 6 pirwise comprisons (for 108 totl impct estimtes) to present in the executive summry before seeing the results. This step ws tken to prevent the selective presenttion of sttisticlly significnt results in the executive summry. The first step hinges on the definition nd implictions of confirmtory hypothesis tests. Following Schochet (2009), we defined confirmtory hypothesis tests s those tests tht ssess how strongly the study s prespecified centrl hypotheses re supported by the dt, (Schochet, 2009: 549). Sttisticlly significnt findings from confirmtory hypothesis tests re considered definitive evidence of nonzero intervention impct, effectively ending debte on whether the intervention chieved n impct in the study sites. All other hypothesis test results re deemed explortory. For these tests, sttisticlly significnt impcts constitute suggestive evidence of possible intervention effects. Before beginning nlysis, HUD determined tht the housing stbility domin is the most importnt outcome domin for the study. Therefore, the study tem designted seven hypothesis tests relted to housing stbility s confirmtory. These hypothesis tests were conducted for The six pirwise policy comprisons nd one pooled comprison (PBTH + SUB + CBRR versus UC). A single composite outcome indicting sty in emergency shelter or plce not ment for humn hbittion or n experience of doubling up. This outcome ws constructed from two binry outcomes within the housing stbility domin. At lest 1 night spent in emergency shelter or plce not ment for humn hbittion or doubled up during the pst 6 months t the time of the followup survey (from the dult survey). Any sty in emergency shelter in the pst 12 months t the time of the followup survey (from Progrm Usge Dt, lrgely bsed on Homeless Mngement Informtion System, or HMIS, records). The six pirwise comprisons were included to ssess the reltive effectiveness of the interventions in contributing to housing stbility (thereby ddressing the study s first reserch question stted in Section 1.4). The study tem lso included the pooled comprison of PBTH + SUB + CBRR versus UC becuse it provided evidence on whether C -6

242 Appendix C. Anlysis Methods housing subsidy of ny type improved housing stbility. Using two sources of dt to construct this outcome enbled the study tem to mesure housing stbility s robustly s possible nd mde use of ll vilble dt on return to homelessness. Implementing the Multiple Comprisons Procedure The p-vlues on the seven impct coefficients were djusted to ccount for the presence of seven confirmtory tests. The tem chose the Westfll-Young resmpling method s the procedure to control the FWER t.10 level for the seven tests.30 This procedure ws chosen for the dditionl sttisticl power (reltive to Bonferroni-type methods) it ws expected to provide in tests of binry outcome vrible. The Westfll-Young resmpling method involves ressigning ech study fmily to the interventions in its rndomiztion set (using the originl ssignment probbilities in effect for the fmily t rndom ssignment) mny times to form mny smple replictes. For ech replicte, the seven impcts on the confirmtory outcome were reclculted, s follows. In nottion, let A, B, C, D, E, F, G = seven impct estimtes on the confirmtory outcome. p, p, p, pj, p;, p;, p = p-vlues from t-tests of impct estimtes. These vlues re the rw, undjusted p-vlues for ech impct estimte. The impct estimtes were then plced in the order of their rw p-vlues. IMPACT1, IMPACT2, IMPACT3, IMPACT4, IMPACT5, IMPACT6, IMPACT7 = the impct estimtes in order of their rw p-vlues. IMPACT1 is the impct estimte with the smllest rw p-vlue nd IMPACT7 is the impct estimte with the lrgest rw p-vlue. j j j j j j p :!"i #n, p!"i #z, p!"i #3, p!"i #4, p!"i #s, p!"i #6, p!"i #? = rw p-vlues in order from smllest to lrgest. Then, some lrge number R (the study used 20,000) permuttion replictes were formed. Within ech replicte smple, study fmilies were ressigned to the interventions in their rndomiztion sets using the originl probbilities. For ech replicte, the seven impcts were estimted, producing seven p-vlues. Next, the djusted p-vlues were clculted s follows P!"# P!"# P!"# P!"# P!"# 30 N#! " f $ #"! & % whe e i $l = $z = mx P!"# $l, $3 = mx P!"# $z, $4 = mx P!"# $3, $S = mx P!"# $4, l p!"# $l,,p!"# $7 < p!"# $1 R N#! " f $ #"! & % whe e. N#! " f $ #"! & % whe e. e e e e e e P!"# $2,,p!"# $7 < P!"# $2 R P!"# $3,,p!"# $7 < P!"# $3 R N#! " f $ #"! & % whe e. N#! " f $ #"! & % whe e. P!"# $4,,p!"# $7 < P!"# $4 R e e e P!"# $s,p!"# $6,P!"# $7 < P!"# $S : Westfll-Young methods re described in Westfll, Tobis, nd Wolfinger (2011). C -7

243 Appendix C. Anlysis Methods P!"# P!"# $6 = mx P!"# $s, $7 = mx P!"# $6,. p!"# $6,p!"# $? < p!"# $6 N#! " f $ #"! & % whe e i R. N#! " f $ #"! & % whe e p!"# $? < p!"# $? R ; where p!"a # is the p-vlue for n impct estimte in prticulr replicte. Exhibit C-1 shows the undjusted nd djusted p-vlues for the study s seven confirmtory hypothesis tests. Exhibit C-1. Confirmtory Hypothesis Tests ITT Impct on At Lest 1 Night Homeless or Doubled Up (pst 6 months) or in Shelter (pst 12 months) (%) Pirwise or Pooled Comprison SUB vs. UC Impct (SE) p-vlue (undjusted) p-vlue (djusted for multiple comprisons) 28.0 (3.1) < 01 < 01 CBRR vs. UC 3.5 (3.6) PBTH vs. UC 7.7 (4.4) SUB vs. CBRR 27.3 (3.8) < 01 < 01 SUB vs. PBTH 31.2 (5.0) < 01 < (5.7) (2.6) < 01 < 01 CBRR vs. PBTH SUB + CBRR + PBTH vs. UC CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. UC = usul cre. ITT = intention to tret. SE = stndrd error. The definition of "homeless" in this report includes stys in emergency shelters nd plces not ment for humn hbittion. It excludes trnsitionl housing. Notes: Impct estimtes re regression-djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definition. Sources: Fmily Options Study 18-month followup survey; Progrm Usge Dt C -8

244 Appendix D. Anlysis of 18-Month Survey Nonresponse APPENDIX D. ANALYSIS OF 18-MONTH SURVEY NONRESPONSE D.1 Introduction The impct estimtes in the Fmily Options Study were bsed on outcome mesures derived lrgely from the 18 month followup surveys. This ppendix nlyzes the extent to which survey nonresponse influenced these estimtes. Not ll prticipnts completed the followup survey, which successfully gthered informtion for 1,857 of the 2,282 fmilies who enrolled in the study. This ppendix ddresses whether, in light of this nonresponse, impct estimtes re (1) internlly vlid in the sense tht the fmilies in the rms of ech impct comprison remin comprble nd (2) likely vlid for the entire study smple fter weighting to ccount for nonresponse. Blnce in Impct Comprison Arms After Nonresponse The Fmily Options Study rndomly ssigned fmilies to study interventions so tht differences in outcomes mong fmilies who received different interventions would be ttributble to ssignment to the intervention. Gubits et l. (2013) presented evidence confirming tht rndom ssignment successfully produced equivlent smples when compring the tretment rms within ech of the six pirwise impct comprisons in the study. This equivlence testing ws conducted on ll fmilies prticipting in the study, however. It is possible tht whether fmily responded to the followup survey ws influenced by the tretment to which they were ssigned in wys tht could disrupt this blnce. This possibility, in turn, is indictive of whether fmilies in ech rm of the impct comprisons re comprble sometimes referred to s the study s internl vlidity. We ssess the extent to which nonresponse ffected internl vlidity by ddressing the following two questions. 1. Wht were the response rtes for the Fmily Options Study 18-month followup survey, nd how did they vry between pirwise comprison smples? 2. Did the nlysis smple remin blnced for ech impct comprison fter nonresponse? Respondents Versus Nonrespondents Survey nonresponse my lso be relted to prticipnt chrcteristics such tht fmilies who respond to the survey re not comprble with fmilies who do not. If this difference ws the cse, nd if the study findings differ on the sme chrcteristics tht relte to survey respondents, then the study findings my not be pplicble to the entire smple including nonrespondents. To ssess the extent to which findings re likely pplicble to nonrespondents in ddition to respondents, we ddress the two following questions. 1. Do respondents nd nonrespondents hve systemtic differences in observble bseline chrcteristics? 2. How were the min results of this report ffected by the use of nonresponse nlysis weights? Overview of Findings In generl, the nlysis presented in this ppendix (1) indictes tht the impct results in the Fmily Options Study remin internlly vlid fter survey nonresponse nd (2) provides nondefinitive evidence tht the impct results my be pplicble to the entire study smple. We find tht response rtes do vry bsed on the tretment to which fmilies were ssigned. Response rtes were slightly lower for fmilies ssigned to the usul cre (UC) intervention, D -1

245 Appendix D. Anlysis of 18-Month Survey Nonresponse rnging from 77.0 to 78.6 percent, depending on the impct comprison, s compred with fmilies ssigned to the three ctive interventions. Prticipnts ssigned to the permnent housing subsidy (SUB) intervention hd the gretest frequency of survey response, with n overll response rte of 88.5 percent. These differences indicte the importnce of our second nlysis ssessing internl vlidity comprison of bseline chrcteristics for ech rm of ech impct comprison. Here we find tht, lthough nonresponse ptterns somewht degrded the bseline equivlence smples s reported in Gubits et l. (2013) for comprisons involving the community-bsed rpid re housing (CBRR) intervention, omnibus test results including ll our comprison chrcteristics suggest no systemtic differences between sides of the impct comprisons, with the exception of the CBRR-versus-UC comprison. Turning to our nlysis relevnt to the pplicbility of study findings to the entire bseline smple, we find some evidence tht bseline chrcteristics do predict survey response, which suggests tht respondents nd nonrespondents my be systemticlly different. This finding in prt motivtes the use of survey nonresponse weights, s described in Section In this ppendix, we present estimtes clculted without the nonresponse weights for the study s hedline outcomes. Substntive differences between impct estimtes clculted with nd without the nonresponse weights would indicte tht impcts for nonrespondents (which cnnot be estimted) my differ from those estimted in the study for respondents. The estimtes did not vry substntilly from the weighted estimtes. Although not definitive, this finding serves s evidence tht the impct results my be pplicble to the entire study smple. D.2 Blnce in Impct Comprison Arms After Nonresponse This section presents two nlyses tht ddress the thret to the internl vlidity of the study s impct findings of survey nonresponse. To ssess the extent to which the rms in ech impct comprison remin comprble fter nonresponse, this section first reports nd compres response rtes for ech tretment rm of ech impct comprison. Next, the section presents n nlysis of the blnce on bseline chrcteristics for ech impct comprison within the nlysis smple of survey respondents. D.2.1 Survey Nonresponse Wht were response rtes for the Fmily Options Study 18-month followup survey, nd how did the rtes vry between pirwise comprison smples? Exhibit D-1 reports the number of respondents to the followup survey by impct comprison. Results bsed on rw response rtes for ech impct comprison suggest tht ssignment to CBRR, project-bsed trnsitionl housing (PBTH), or SUB incresed the propensity to respond to the followup survey, s UC prticipnt response rtes rnged from 77.0 to 78.6 percent. Prticipnts ssigned to SUB were most likely to respond to the followup survey, with n overll response rte of 88.5 percent. For ech impct comprison, the study tem tests for sttisticlly significnt difference between the two comprison rms in the response rtes. The tem found sttisticlly significnt difference in the response rtes of the two rms in three of the six pirwise comprisons (those including SUB) nd in three of the four pooled comprisons. In these cses, it is prticulrly relevnt to test for differences in bseline chrcteristics cross the tretment rms of the comprisons. D -2

246 Appendix D. Anlysis of 18-Month Survey Nonresponse Exhibit D-1. Survey Nonresponse Incidence by Impct Comprison Adult Followup Survey Bseline Fmilies Totl pirwise comprisons Adult Surveys Completed Response Rte (%) 2,282 1,857 SUB versus ** UC CBRR versus UC PBTH versus UC SUB versus ** CBRR SUB versus ** PBTH CBRR versus PBTH Chi-sq 81.4 Pooled Comprisons SUB + CBRR + PBTH versus 1,536 1,279 UC ** 77.5 SUB + PBTH versus * CBRR SUB + CBRR versus PBTH CBRR + PBTH versus ** SUB CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. UC = usul cre. Note: Significntly different response rtes re indicted by for p-vlue **<.01 nd * <.05. Sources: Fmily Options Bseline Survey; 18-month followup survey D.2.2 Equivlence t Bseline of Anlysis Smple by Impct Comprison Did the nlysis smple remin blnced for ech impct comprison fter nonresponse? The second step in the nlysis of the comprbility of both rms of ech impct comprison is comprison of bseline chrcteristics. If the blnce in observble chrcteristics between groups t bseline remined fter nonresponse, survey nonresponse ws not relted to observble chrcteristics nd therefore ws unlikely to be relted to unobservble chrcteristics. In tht cse, impct estimtes remined vlid comprison of the effect of receiving different interventions on the prticulr outcome for the survey respondent popultion. Exhibit D-2 lists the bseline chrcteristics tht re compred within ech impct comprison. These chrcteristics were the sme bseline chrcteristics used to demonstrte bseline equivlence in Gubits et l. (2013), nd were chosen becuse they were either mjor demogrphic chrcteristics or they were bseline mesures in the study s five outcome domins. This section reports results from sttisticl tests performed to determine both if groups being compred differed on the ech of the bseline chrcteristics described previously nd if the combined set of chrcteristics suggested the groups differed (n omnibus F-test). As review of the bseline equivlence findings of the full bseline smple, Gubits et l. (2013) reported sttisticlly significnt differences in the SUB-versus-UC nd CBRR-versus-UC comprisons in eductionl ttinment nd in the PBTH-versus-CBRR group in ge of household hed. Only the CBRR-versus-UC comprison hd p-vlue of the omnibus F-test tht indicted sttisticlly significnt differences in the two groups t the.05 level. This result suggested n unlucky division of fmilies into the CBRR nd UC interventions. Differences in mens nd percentges for individul vribles, however, were not substntively lrge. D -3

247 Appendix D. Anlysis of 18-Month Survey Nonresponse Exhibit D-2. Chrcteristics Exmined in Bseline Equivlency Testing Bseline Chrcteristic t the Time of Rndom Assignment Age of household hed Previously convicted of felony Gender Fmily nnul income Mritl sttus Previously homeless (before current spell) Rce/ethnicity Previously lived in doubled up housing Eductionl ttinment Number of brriers in finding housing Number of dults in fmily Household hed hs child under 18 living elsewhere Number of children in fmily Number of mjor life chllenges fcedb Worked for py in lst week Brriers to finding housing were reported by dult respondent s big problems in finding housing. The mximum number of brriers ws 19. The 19 possible brriers were (1) not hving enough income to py rent, (2) inbility to py security deposit or first/lst month s rent, (3) lck of trnsporttion to look for housing, (4) poor credit history, (5) rcil discrimintion, (6) not being currently employed, (7) no rent history t ll, (8) recently moved to community nd no locl rent history, (9) no reference from pst lndlords, (10) pst eviction, (11) problems with pst lndlords, (12) pst lese violtions, (13) hving problems with police, (14) hving criminl record or bckground, (15) hving felony drug record, (16) hving three or more children in the household, (17) hving teengers in the household, (18) hving someone in the household under 21 yers old, nd (19) hving someone in the household who hs disbility. b The seven mjor life chllenges mesured were (1) psychologicl distress, (2) post-trumtic stress disorder, (3) felony conviction, (4) experience of domestic violence, (5) childhood seprtion (foster cre, group home, or institutionliztion), (6) self-reported medicl condition, nd (7) substnce buse. Source: Fmily Options bseline survey reproduced from Exhibit 4-9 of Gubits et l. (2013) As reported in Exhibit D-3, survey nonresponse ltered these bseline equivlence results to some extent. Age nd mjor brriers to finding housing were dded to the chrcteristics, which differed t bseline in the CBRR-versusUC comprison, nd the omnibus F-test continued to indicte the responding smples differed on observble bseline chrcteristics. Four bseline chrcteristics were individully sttisticlly different cross groups for the PBTH-versus-CBRR comprison, lthough the omnibus F-test ws not sttisticlly significnt. SUB versus CBRR nd ech of the pooled comprison lso hd either one or two chrcteristics tht hd sttisticlly significnt difference cross groups. Among the pooled comprison groups, however, only CBRR + SUB versus PBTH hd sttisticlly significnt omnibus F-test. Tken together, these comprisons indicte tht, lthough nonresponse ptterns somewht degrded the bseline equivlence smples s reported in Gubits et l. (2013) for comprisons involving CBRR, omnibus tests results including ll our comprison chrcteristics suggests no systemtic differences between sides of the impct comprisons, with the exception of the CBRR-versus-UC comprison. Recll tht this comprison ws found to hve n unlucky drw with sttisticlly significnt but reltively smll-in-mgnitude difference in bseline chrcteristics for the entire study smple. Exhibits D-4 through D-13 report the summry sttistics for bseline chrcteristics for ech rm of ech comprison which, together, is the informtion summrized in Exhibit D-3. Exhibit D-3. Summry of Equivlence Testing in Impct Comprisons, Adult Survey Pirwise Impct Comprison SUB versus UC CBRR versus UC PBTH versus UC SUB versus CBRR SUB versus PBTH CBRR versus PBTH Number of Chrcteristics With Significnt Differences (out of 15; α = 0.10) Chrcteristic(s) With Significnt Difference p-vlue of Omnibus F-test Eductionl ttinment Age, eductionl ttinment, mjor brrier None Child living elsewhere None Mritl sttus, eductionl ttinment, number of dults, ever doubled up SUB + CBRR + PBTH versus UC 2 Eductionl ttinment, mjor brrier SUB + PBTH versus CBRR 1 Ever doubled up SUB + CBRR versus PBTH 1 Ever doubled up CBRR + PBTH versus SUB 2 Gender, child living elsewhere CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. UC = usul cre. Sources: Fmily Options Bseline Survey; 18-month followup survey D -4

248 Appendix D. Anlysis of 18-Month Survey Nonresponse Exhibit D-4. Equivlence t Bseline of Anlysis Smple for SUB Versus UC Impct Comprison, Adult Survey Chrcteristic SUB UC Difference Significnce Level Strs Number of fmilies Age of household hed t RA (percent) Less thn 21 yers old yers yers yers yers yers nd older Men ge (yers) Gender (percent) Femle Mle Mritl sttus (percent) Single (never mrried/widowed/ seprted/divorced) Mrried or mrrige-like sitution Rce/ethnicity (percent) Blck/Africn Americn, not Hispnic White, not Hispnic Hispnic Other Eductionl ttinment (percent) Less thn high school diplom ** High school diplom/ged More thn high school diplom Number of dults in fmily (percent) 1 dult or more dults Number of children in fmily (percent) 1 child children children children or more Missing dt Worked for py lst week (percent) Ever convicted of felony (percent) Fmily nnul income (percent) Less thn $5, $5,000 9, $10,000 14, $15,000 19, $20,000 24, $25,000 or more Missing dt Ever been homeless before (percent) Ever been doubled up before (percent) Mjor brrier to finding housing Child under 18 living elsewhere (percent) Number of mjor life chllengesb F-test on ll chrcteristics except site F vlue = F-test p-vlue = SUB = permnent housing subsidy. UC = usul cre. GED = generl eductionl development. RA = rndom ssignment. *.10 level. **.05 level. ***.01 level. Brriers to finding housing were reported by fmily heds s "big problems" in finding housing. The mximum number of brriers is 19. b The seven mjor life chllenges mesured re: psychologicl distress, Post Trumtic Stress Disorder, felony conviction, experience of domestic violence, childhood seprtion (foster cre, group home, or institutionliztion), medicl condition, nd substnce buse. Notes: Chi-squre tests used to test the difference between groups for ll chrcteristics except men ge. t-test used for men ge. Sources: Fmily Options Bseline Survey; 18-month followup survey D -5

249 Appendix D. Anlysis of 18-Month Survey Nonresponse Exhibit D-5. Equivlence t Bseline of Anlysis Smple for CBRR Versus UC Impct Comprison, Adult Survey Chrcteristic CBRR UC Difference Significnce Level Strs Number of fmilies Age of household hed t RA (percent) Less thn 21 yers old * yers yers yers yers yers nd older Men ge (yers) Gender (percent) Femle Mle Mritl sttus (percent) Single (never mrried/widowed/ seprted/divorced) Mrried or mrrige-like sitution Rce/ethnicity (percent) Blck/Africn Americn, not Hispnic White, not Hispnic Hispnic Other Eductionl ttinment (percent) Less thn high school diplom ** High school diplom/ged More thn high school diplom Number of dults in fmily (percent) 1 dult or more dults Number of children in fmily (percent) 1 child children children children or more Missing dt Worked for py lst week (percent) Ever convicted of felony (percent) Fmily nnul income (percent) Less thn $5, $5,000 9, $10,000 14, $15,000 19, $20,000 24, $25,000 or more Missing dt Ever been homeless before (percent) Ever been doubled up before (percent) Mjor brrier to finding housing ** Child under 18 living elsewhere (percent) Number of mjor life chllengesb F-test on ll chrcteristics except site F vlue = F-test p-vlue = CBRR = community-bsed rpid re-housing. UC = usul cre. GED = generl eductionl development. RA = rndom ssignment. *.10 level. **.05 level. ***.01 level. Brriers to finding housing were reported by fmily heds s "big problems" in finding housing. The mximum number of brriers is 19. b The seven mjor life chllenges mesured re: psychologicl distress, Post Trumtic Stress Disorder, felony conviction, experience of domestic violence, childhood seprtion (foster cre, group home, or institutionliztion), medicl condition, nd substnce buse. Notes: Chi-squre tests used to test the difference between groups for ll chrcteristics except men ge. t-test used for men ge. Sources: Fmily Options Bseline Survey; 18-month followup survey D -6

250 Appendix D. Anlysis of 18-Month Survey Nonresponse Exhibit D-6. Equivlence t Bseline of Anlysis Smple for PBTH Versus UC Impct Comprison, Adult Survey Chrcteristic PBTH UC Difference Significnce Level Strs Number of fmilies Age of household hed t RA (percent) Less thn 21 yers old yers yers yers yers yers nd older Men ge (yers) Gender (percent) Femle Mle Mritl sttus (percent) Single (never mrried/widowed/ seprted/divorced) Mrried or mrrige-like sitution Rce/ethnicity (percent) Blck/Africn Americn, not Hispnic White, not Hispnic Hispnic Other Eductionl ttinment (percent) Less thn high school diplom High school diplom/ged More thn high school diplom Number of dults in fmily (percent) 1 dult or more dults Number of children in fmily (percent) 1 child children children children or more Missing dt Worked for py lst week (percent) Ever convicted of felony (percent) Fmily nnul income (percent) Less thn $5, $5,000 9, $10,000 14, $15,000 19, $20,000 24, $25,000 or more Missing dt Ever been homeless before (percent) Ever been doubled up before (percent) Mjor brrier to finding housing Child under 18 living elsewhere (percent) Number of mjor life chllengesb F-test on ll chrcteristics except site F vlue = F-test p-vlue = PBTH = project-bsed trnsitionl housing. UC = usul cre. GED = generl eductionl development. RA = rndom ssignment. *.10 level. **.05 level. ***.01 level. Brriers to finding housing were reported by fmily heds s "big problems" in finding housing. The mximum number of brriers is 19. b The seven mjor life chllenges mesured re: psychologicl distress, Post Trumtic Stress Disorder, felony conviction, experience of domestic violence, childhood seprtion (foster cre, group home, or institutionliztion), medicl condition, nd substnce buse. Notes: Chi-squre tests used to test the difference between groups for ll chrcteristics except men ge. t-test used for men ge. Sources: Fmily Options Bseline Survey; 18-month followup survey D -7

251 Appendix D. Anlysis of 18-Month Survey Nonresponse Exhibit D-7. Equivlence t Bseline of Anlysis Smple for SUB Versus CBRR Impct Comprison, Adult Survey Chrcteristic SUB CBRR Difference Significnce Level Strs Number of fmilies Age of household hed t RA (percent) Less thn 21 yers old yers yers yers yers yers nd older Men ge (yers) Gender (percent) Femle Mle Mritl sttus (percent) Single (never mrried/widowed/ seprted/divorced) Mrried or mrrige-like sitution Rce/ethnicity (percent) Blck/Africn Americn, not Hispnic White, not Hispnic Hispnic Other Eductionl ttinment (percent) Less thn high school diplom High school diplom/ged More thn high school diplom Number of dults in fmily (percent) 1 dult or more dults Number of children in fmily (percent) 1 child children children children or more Missing dt Worked for py lst week (percent) Ever convicted of felony (percent) Fmily nnul income (percent) Less thn $5, $5,000 9, $10,000 14, $15,000 19, $20,000 24, $25,000 or more Missing dt Ever been homeless before (percent) Ever been doubled up before (percent) Mjor brrier to finding housing Child under 18 living elsewhere (percent) * Number of mjor life chllengesb F-test on ll chrcteristics except site F vlue = F-test p-vlue = CBRR = community-bsed rpid re-housing. SUB = permnent housing subsidy. GED = generl eductionl development. RA = rndom ssignment. *.10 level. **.05 level. ***.01 level. Brriers to finding housing were reported by fmily heds s "big problems" in finding housing. The mximum number of brriers is 19. b The seven mjor life chllenges mesured re: psychologicl distress, Post Trumtic Stress Disorder, felony conviction, experience of domestic violence, childhood seprtion (foster cre, group home, or institutionliztion), medicl condition, nd substnce buse. Notes: Chi-squre tests used to test the difference between groups for ll chrcteristics except men ge. t-test used for men ge. Sources: Fmily Options Bseline Survey; 18-month followup survey D -8

252 Appendix D. Anlysis of 18-Month Survey Nonresponse Exhibit D-8. Equivlence t Bseline of Anlysis Smple for SUB Versus PBTH Impct Comprison, Adult Survey Chrcteristic SUB PBTH Difference Significnce Level Strs Number of fmilies Age of household hed t RA (percent) Less thn 21 yers old yers yers yers yers yers nd older Men ge (yers) Gender (percent) Femle Mle Mritl sttus (percent) Single (never mrried/widowed/ seprted/divorced) Mrried or mrrige-like sitution Rce/ethnicity (percent) Blck/Africn Americn, not Hispnic White, not Hispnic Hispnic Other Eductionl ttinment (percent) Less thn high school diplom High school diplom/ged More thn high school diplom Number of dults in fmily (percent) 1 dult or more dults Number of children in fmily (percent) 1 child children children children or more Missing dt Worked for py lst week (percent) Ever convicted of felony (percent) Fmily nnul income (percent) Less thn $5, $5,000 9, $10,000 14, $15,000 19, $20,000 24, $25,000 or more Missing dt Ever been homeless before (percent) Ever been doubled up before (percent) Mjor brrier to finding housing Child under 18 living elsewhere (percent) Number of mjor life chllengesb F-test on ll chrcteristics except site F vlue = F-test p-vlue = PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. GED = generl eductionl development. RA = rndom ssignment. *.10 level. **.05 level. ***.01 level. Brriers to finding housing were reported by fmily heds s "big problems" in finding housing. The mximum number of brriers is 19. b The seven mjor life chllenges mesured re: psychologicl distress, Post Trumtic Stress Disorder, felony conviction, experience of domestic violence, childhood seprtion (foster cre, group home, or institutionliztion), medicl condition, nd substnce buse. Notes: Chi-squre tests used to test the difference between groups for ll chrcteristics except men ge. t-test used for men ge. Sources: Fmily Options Bseline Survey; 18-month followup survey D -9

253 Appendix D. Anlysis of 18-Month Survey Nonresponse Exhibit D-9. Equivlence t Bseline of Anlysis Smple for CBRR Versus PBTH Impct Comprison, Adult Survey Chrcteristic CBRR PBTH Difference Significnce Level Strs Number of fmilies Age of household hed t RA (percent) Less thn 21 yers old yers yers yers yers yers nd older Men ge (yers) Gender (percent) Femle Mle Mritl sttus (percent) * Single (never mrried/widowed/ seprted/divorced) Mrried or mrrige-like sitution Rce/ethnicity (percent) Blck/Africn Americn, not Hispnic White, not Hispnic Hispnic Other Eductionl ttinment (percent) Less thn high school diplom ** High school diplom/ged More thn high school diplom Number of dults in fmily (percent) 1 dult or more dults * Number of children in fmily (percent) 1 child children children children or more Missing dt Worked for py lst week (percent) Ever convicted of felony (percent) Fmily nnul income (percent) Less thn $5, $5,000 9, $10,000 14, $15,000 19, $20,000 24, $25,000 or more Missing dt Ever been homeless before (percent) Ever been doubled up before (percent) Mjor brrier to finding housing Child under 18 living elsewhere (percent) Number of mjor life chllengesb F-test on ll chrcteristics except site F vlue = F-test p-vlue = CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. GED = generl eductionl development. RA = rndom ssignment. *.10 level. **.05 level. ***.01 level. Brriers to finding housing were reported by fmily heds s "big problems" in finding housing. The mximum number of brriers is 19. b The seven mjor life chllenges mesured re: psychologicl distress, Post Trumtic Stress Disorder, felony conviction, experience of domestic violence, childhood seprtion (foster cre, group home, or institutionliztion), medicl condition, nd substnce buse. Notes: Chi-squre tests used to test the difference between groups for ll chrcteristics except men ge. t-test used for men ge. Sources: Fmily Options Bseline Survey; 18-month followup survey D-10

254 Appendix D. Anlysis of 18-Month Survey Nonresponse Exhibit D-10. Equivlence t Bseline of Anlysis Smple for SUB + CBRR + PBTH Versus UC Impct Comprison, Adult Survey Chrcteristic SUB, CBRR, PBTH UC Difference Significnce Level Strs Number of fmilies 1, Age of household hed t RA (percent) Less thn 21 yers old yers yers yers yers yers nd older Men ge (yers) Gender (percent) Femle Mle Mritl sttus (percent) Single (never mrried/widowed/ seprted/divorced) Mrried or mrrige-like sitution Rce/ethnicity (percent) Blck/Africn Americn, not Hispnic White, not Hispnic Hispnic Other Eductionl ttinment (percent) Less thn high school diplom ** High school diplom/ged More thn high school diplom Number of dults in fmily (percent) 1 dult or more dults Number of children in fmily (percent) 1 child children children children or more Missing dt Worked for py lst week (percent) Ever convicted of felony (percent) Fmily nnul income (percent) Less thn $5, $5,000 9, $10,000 14, $15,000 19, $20,000 24, $25,000 or more Missing dt Ever been homeless before (percent) Ever been doubled up before (percent) Mjor brrier to finding housing * Child under 18 living elsewhere (percent) Number of mjor life chllengesb F-test on ll chrcteristics except site F vlue = F-test p-vlue = CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. UC = usul cre. GED = generl eductionl development. RA = rndom ssignment. *.10 level. **.05 level. ***.01 level. Brriers to finding housing were reported by fmily heds s "big problems" in finding housing. The mximum number of brriers is 19. b The seven mjor life chllenges mesured re: psychologicl distress, Post Trumtic Stress Disorder, felony conviction, experience of domestic violence, childhood seprtion (foster cre, group home, or institutionliztion), medicl condition, nd substnce buse. Notes: Chi-squre tests used to test the difference between groups for ll chrcteristics except men ge. t-test used for men ge. Sources: Fmily Options Bseline Survey; 18-month followup survey D-11

255 Appendix D. Anlysis of 18-Month Survey Nonresponse Exhibit D-11. Equivlence t Bseline of Anlysis Smple for SUB + PBTH Versus CBRR Impct Comprison, Adult Survey Chrcteristic SUB, PBTH CBRR Difference Significnce Level Strs Number of fmilies Age of household hed t RA (percent) Less thn 21 yers old yers yers yers yers yers nd older Men ge (yers) Gender (percent) Femle Mle Mritl sttus (percent) Single (never mrried/widowed/ seprted/divorced) Mrried or mrrige-like sitution Rce/ethnicity (percent) Blck/Africn Americn, not Hispnic White, not Hispnic Hispnic Other Eductionl ttinment (percent) Less thn high school diplom High school diplom/ged More thn high school diplom Number of dults in fmily (percent) 1 dult or more dults Number of children in fmily (percent) 1 child children children children or more Missing dt Worked for py lst week (percent) Ever convicted of felony (percent) Fmily nnul income (percent) Less thn $5, $5,000 9, $10,000 14, $15,000 19, $20,000 24, $25,000 or more Missing dt Ever been homeless before (percent) Ever been doubled up before (percent) * Mjor brrier to finding housing Child under 18 living elsewhere (percent) Number of mjor life chllengesb F-test on ll chrcteristics except site F vlue = F-test p-vlue = CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. GED = generl eductionl development. RA = rndom ssignment. *.10 level. **.05 level. ***.01 level. Brriers to finding housing were reported by fmily heds s "big problems" in finding housing. The mximum number of brriers is 19. b The seven mjor life chllenges mesured re: psychologicl distress, Post Trumtic Stress Disorder, felony conviction, experience of domestic violence, childhood seprtion (foster cre, group home, or institutionliztion), medicl condition, nd substnce buse. Notes: Chi-squre tests used to test the difference between groups for ll chrcteristics except men ge. t-test used for men ge. Sources: Fmily Options Bseline Survey; 18-month followup survey D-12

256 Appendix D. Anlysis of 18-Month Survey Nonresponse Exhibit D-12. Equivlence t Bseline of Anlysis Smple for SUB + CBRR Versus PBTH Impct Comprison, Adult Survey Chrcteristic SUB, CBRR PBTH Difference Significnce Level Strs Number of fmilies Age of household hed t RA (percent) Less thn 21 yers old yers yers yers yers yers nd older Men ge (yers) Gender (percent) Femle Mle Mritl sttus (percent) Single (never mrried/widowed/ seprted/divorced) Mrried or mrrige-like sitution Rce/ethnicity (percent) Blck/Africn Americn, not Hispnic White, not Hispnic Hispnic Other Eductionl ttinment (percent) Less thn high school diplom High school diplom/ged More thn high school diplom Number of dults in fmily (percent) 1 dult or more dults Number of children in fmily (percent) 1 child children children children or more Missing dt Worked for py lst week (percent) Ever convicted of felony (percent) Fmily nnul income (percent) Less thn $5, $5,000 9, $10,000 14, $15,000 19, $20,000 24, $25,000 or more Missing dt Ever been homeless before (percent) Ever been doubled up before (percent) * Mjor brrier to finding housing Child under 18 living elsewhere (percent) Number of mjor life chllengesb F-test on ll chrcteristics except site F vlue = F-test p-vlue = CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. GED = generl eductionl development. RA = rndom ssignment. *.10 level. **.05 level. ***.01 level. Brriers to finding housing were reported by fmily heds s "big problems" in finding housing. The mximum number of brriers is 19. b The seven mjor life chllenges mesured re: psychologicl distress, Post Trumtic Stress Disorder, felony conviction, experience of domestic violence, childhood seprtion (foster cre, group home, or institutionliztion), medicl condition, nd substnce buse. Notes: Chi-squre tests used to test the difference between groups for ll chrcteristics except men ge. t-test used for men ge. Sources: Fmily Options Bseline Survey; 18-month followup survey D-13

257 Appendix D. Anlysis of 18-Month Survey Nonresponse Exhibit D-13. Equivlence t Bseline of Anlysis Smple for CBRR + PBTH Versus SUB Impct Comprison, Adult Survey Chrcteristic CBRR, PBTH SUB Difference Significnce Level Strs Number of fmilies Age of household hed t RA (percent) Less thn 21 yers old yers yers yers yers yers nd older Men ge (yers) Gender (percent) Femle * Mle Mritl sttus (percent) Single (never mrried/widowed/ seprted/divorced) Mrried or mrrige-like sitution Rce/ethnicity (percent) Blck/Africn Americn, not Hispnic White, not Hispnic Hispnic Other Eductionl ttinment (percent) Less thn high school diplom High school diplom/ged More thn high school diplom Number of dults in fmily (percent) 1 dult or more dults Number of children in fmily (percent) 1 child children children children or more Missing dt Worked for py lst week (percent) Ever convicted of felony (percent) Fmily nnul income (percent) Less thn $5, $5,000 9, $10,000 14, $15,000 19, $20,000 24, $25,000 or more Missing dt Ever been homeless before (percent) Ever been doubled up before (percent) Mjor brrier to finding housing Child under 18 living elsewhere (percent) * Number of mjor life chllengesb F-test on ll chrcteristics except site F vlue = F-test p-vlue = CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. GED = generl eductionl development. RA = rndom ssignment. *.10 level. **.05 level. ***.01 level. Brriers to finding housing were reported by fmily heds s "big problems" in finding housing. The mximum number of brriers is 19. b The seven mjor life chllenges mesured re: psychologicl distress, Post Trumtic Stress Disorder, felony conviction, experience of domestic violence, childhood seprtion (foster cre, group home, or institutionliztion), medicl condition, nd substnce buse. Notes: Chi-squre tests used to test the difference between groups for ll chrcteristics except men ge. t-test used for men ge. Sources: Fmily Options Bseline Survey; 18-month followup survey D-14

258 Appendix D. Anlysis of 18-Month Survey Nonresponse D.3 Respondents Versus Nonrespondents This section provides evidence regrding whether the impcts estimted on the smple of survey respondents re pplicble to the entire study popultion, including nonrespondents. We present results of two nlyses. The first ssesses whether respondents nd nonrespondents hve systemtic differences in observble bseline chrcteristics. The second compres unweighted impct estimtes with the weighted impct estimtes presented in the body of the report. D.3.1 Do Respondents Differ From Nonrespondents on Bseline Chrcteristics? Do respondents nd nonrespondents hve systemtic differences in observble bseline chrcteristics? Exhibit D-14 summrizes the results of tests compring bseline chrcteristics for the nlysis smple nd the smple of nonrespondents for ech intervention. Averge vlues or incidences of four bseline chrcteristics differed between fmilies who responded to the followup survey for both fmilies ssigned to SUB nd fmilies ssigned to UC. The joint F-test on ll chrcteristics in regression indicted tht the set of bseline chrcteristics were jointly significnt predictors for fmilies ssigned to both CBRR nd UC. Exhibit D-15 summrizes the results of tests compring response nd nonresponse popultions. Among the pirwise comprison smples, ll but PBTH versus UC hd t lest three bseline chrcteristic verges tht differed for response s opposed to nonresponse, nd t lest four chrcteristics differed for ech of the pooled comprisons smples. Menwhile, omnibus F-tests indicted tht bseline chrcteristics were generlly predictive of nonresponse cross the comprison smples, with ll but two meeting.10 p-vlue threshold. Tken together, these tests suggest tht respondents seemed to systemticlly differ from nonrespondents. The mgnitude of these sttisticlly significnt differences is reported in Exhibits D-16 through D-29. When differences were sttisticlly significnt reltive to nonrespondents, prticipnts who responded to the followup survey were more often in younger ge ctegories, more often femle, less often White, more often worked for py in the previous week, more often hd been previously homeless or doubled up, nd less often hd children living elsewhere. Exhibit D-14. RA Result Summry of Equivlence Testing of Respondents Versus Nonrespondents, by Assigned Intervention Number of Bseline Chrcteristics With Significnt Differences Between Response nd Nonresponse Smple (out of 15; α = 0.10) Chrcteristic(s) With Significnt Difference p-vlue of Omnibus F-test SUB 4 Age, Gender, Prior Homeless, Child Elsewhere 0.17 CBRR 2 Doubled Up, Child Elsewhere 0.06 PBTH 2 Rce/Ethnicity, Prior Homeless 0.40 UC 4 Rce/Ethnicity, Work for Py, Doubled Up, Mjor Brrier 0.03 CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. UC = usul cre. Sources: Fmily Options Bseline Survey; 18-month followup survey D-15

259 Appendix D. Anlysis of 18-Month Survey Nonresponse Exhibit D-15. Summry of Equivlence Testing of Respondents Versus Nonrespondents, by Impct Comprison Pirwise Impct Comprison Smple (response nd nonresponse compred for both tretment rms) Number of Chrcteristics With Significnt Differences (out of 15; α = 0.10) Chrcteristic(s) With Significnt difference p-vlue of omnibus F-test SUB versus UC 4 Age, Rce/Ethnicity, Prior Homeless, Mjor Brrier 0.11 CBRR versus UC 4 Rce/Ethnicity, Doubled Up, Child Elsewhere, Mjor Brrier PBTH versus UC 1 Rce/Ethnicity 0.05 SUB versus CBRR 6 Age, Gender, Rce/Ethnicity, Prior Homeless, Doubled Up, Child Elsewhere 0.06 SUB versus PBTH 3 Gender, Rce/Ethnicity, Prior Homeless 0.03 CBRR versus PBTH 3 Gender, Doubled Up, Chllenges 0.15 SUB + CBRR + PBTH versus UC 5 Gender, Rce/Ethnicity, Prior Homeless, Doubled Up, Mjor Brrier SUB + PBTH versus CBRR 5 Gender, Rce/Ethnicity, Prior Homeless, Doubled Up, Child Elsewhere 0.07 SUB + CBRR versus PBTH 4 Gender, Rce/Ethnicity, Prior Homeless, Doubled Up 0.02 PBTH + CBRR versus SUB 5 Gender, Rce/Ethnicity, Prior Homeless, Doubled Up, Child Elsewhere CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. UC = usul cre. Sources: Fmily Options Bseline Survey; 18-month followup survey D-16

260 Appendix D. Anlysis of 18-Month Survey Nonresponse Exhibit D-16. Equivlence t Bseline of Anlysis Smple for Respondents Versus Nonrespondents Assigned to SUB Chrcteristic Response Nonresponse Difference Significnce Level Strs Number of fmilies % Age of household hed t RA (percent) Less thn 21 yers old *** yers yers yers yers yers nd older Men ge (yers) Gender (percent) Femle ** Mle Mritl sttus (percent) Single (never mrried/widowed/ seprted/divorced) Mrried or mrrige-like sitution Rce/ethnicity (percent) Blck/Africn Americn, not Hispnic White, not Hispnic Hispnic Other Eductionl ttinment (percent) Less thn high school diplom High school diplom/ged More thn high school diplom Number of dults in fmily (percent) 1 dult or more dults Number of children in fmily (percent) 1 child children children children or more Missing dt Worked for py lst week (percent) Ever convicted of felony (percent) Fmily nnul income (percent) Less thn $5, $5,000 9, $10,000 14, $15,000 19, $20,000 24, $25,000 or more Missing dt Ever been homeless before (percent) ** Ever been doubled up before (percent) Mjor brrier to finding housing Child under 18 living elsewhere (percent) * Number of mjor life chllengesb F-test on ll chrcteristics except site F vlue = F-test p-vlue = SUB = permnent housing subsidy. GED = generl eductionl development. RA = rndom ssignment. *.10 level. **.05 level. ***.01 level. Brriers to finding housing were reported by fmily heds s "big problems" in finding housing. The mximum number of brriers is 19. b The seven mjor life chllenges mesured re: psychologicl distress, Post Trumtic Stress Disorder, felony conviction, experience of domestic violence, childhood seprtion (foster cre, group home, or institutionliztion), medicl condition, nd substnce buse. Notes: Chi-squre tests used to test the difference between groups for ll chrcteristics except men ge. t-test used for men ge. Sources: Fmily Options Bseline Survey; 18-month followup survey D-17

261 Appendix D. Anlysis of 18-Month Survey Nonresponse Exhibit D-17. Equivlence t Bseline of Anlysis Smple for Respondents Versus Nonrespondents for CBRR Chrcteristic Response Nonresponse Difference Significnce Level Strs Number of fmilies % Age of household hed t RA (percent) Less thn 21 yers old yers yers yers yers yers nd older Men ge (yers) Gender (percent) Femle Mle Mritl sttus (percent) Single (never mrried/widowed/ seprted/divorced) Mrried or mrrige-like sitution Rce/ethnicity (percent) Blck/Africn Americn, not Hispnic White, not Hispnic Hispnic Other Eductionl ttinment (percent) Less thn high school diplom High school diplom/ged More thn high school diplom Number of dults in fmily (percent) 1 dult or more dults Number of children in fmily (percent) 1 child children children children or more Missing dt Worked for py lst week (percent) Ever convicted of felony (percent) Fmily nnul income (percent) Less thn $5, $5,000 9, $10,000 14, $15,000 19, $20,000 24, $25,000 or more Missing dt Ever been homeless before (percent) Ever been doubled up before (percent) *** Mjor brrier to finding housing Child under 18 living elsewhere (percent) ** Number of mjor life chllengesb F-test on ll chrcteristics except site F vlue = F-test p-vlue = CBRR = community-bsed rpid re-housing. GED = generl eductionl development. RA = rndom ssignment. *.10 level. **.05 level. ***.01 level. Brriers to finding housing were reported by fmily heds s "big problems" in finding housing. The mximum number of brriers is 19. b The seven mjor life chllenges mesured re: psychologicl distress, Post Trumtic Stress Disorder, felony conviction, experience of domestic violence, childhood seprtion (foster cre, group home, or institutionliztion), medicl condition, nd substnce buse. Notes: Chi-squre tests used to test the difference between groups for ll chrcteristics except men ge. t-test used for men ge. Sources: Fmily Options Bseline Survey; 18-month followup survey D-18

262 Appendix D. Anlysis of 18-Month Survey Nonresponse Exhibit D-18. Equivlence t Bseline of Anlysis Smple for Respondents Versus Nonrespondents for PBTH Chrcteristic Response Nonresponse Difference Significnce Level Strs Number of fmilies % Age of household hed t RA (percent) Less thn 21 yers old yers yers yers yers yers nd older Men ge (yers) Gender (percent) Femle Mle Mritl sttus (percent) Single (never mrried/widowed/ seprted/divorced) Mrried or mrrige-like sitution Rce/ethnicity (percent) Blck/Africn Americn, not Hispnic *** White, not Hispnic Hispnic Other Eductionl ttinment (percent) Less thn high school diplom High school diplom/ged More thn high school diplom Number of dults in fmily (percent) 1 dult or more dults Number of children in fmily (percent) 1 child children children children or more Missing dt Worked for py lst week (percent) Ever convicted of felony (percent) Fmily nnul income (percent) Less thn $5, $5,000 9, $10,000 14, $15,000 19, $20,000 24, $25,000 or more Missing dt Ever been homeless before (percent) * Ever been doubled up before (percent) Mjor brrier to finding housing Child under 18 living elsewhere (percent) Number of mjor life chllengesb F-test on ll chrcteristics except site F vlue = F-test p-vlue = PBTH = project-bsed trnsitionl housing. GED = generl eductionl development. RA = rndom ssignment. *.10 level. **.05 level. ***.01 level. Brriers to finding housing were reported by fmily heds s "big problems" in finding housing. The mximum number of brriers is 19. b The seven mjor life chllenges mesured re: psychologicl distress, Post Trumtic Stress Disorder, felony conviction, experience of domestic violence, childhood seprtion (foster cre, group home, or institutionliztion), medicl condition, nd substnce buse. Notes: Chi-squre tests used to test the difference between groups for ll chrcteristics except men ge. t-test used for men ge. Sources: Fmily Options Bseline Survey; 18-month followup survey D-19

263 Appendix D. Anlysis of 18-Month Survey Nonresponse Exhibit D-19. Equivlence t Bseline of Anlysis Smple for Respondents Versus Nonrespondents for UC Chrcteristic Response Nonresponse Difference Significnce Level Strs Number of fmilies % Age of household hed t RA (percent) Less thn 21 yers old yers yers yers yers yers nd older Men ge (yers) Gender (percent) Femle Mle Mritl sttus (percent) Single (never mrried/widowed/ seprted/divorced) Mrried or mrrige-like sitution Rce/ethnicity (percent) Blck/Africn Americn, not Hispnic *** White, not Hispnic Hispnic Other Eductionl ttinment (percent) Less thn high school diplom High school diplom/ged More thn high school diplom Number of dults in fmily (percent) 1 dult or more dults Number of children in fmily (percent) 1 child children children children or more Missing dt Worked for py lst week (percent) ** Ever convicted of felony (percent) Fmily nnul income (percent) Less thn $5, $5,000 9, $10,000 14, $15,000 19, $20,000 24, $25,000 or more Missing dt Ever been homeless before (percent) Ever been doubled up before (percent) * Mjor brrier to finding housing *** Child under 18 living elsewhere (percent) Number of mjor life chllengesb F-test on ll chrcteristics except site F vlue = F-test p-vlue = UC = usul cre. GED = generl eductionl development. RA = rndom ssignment. *.10 level. **.05 level. ***.01 level. Brriers to finding housing were reported by fmily heds s "big problems" in finding housing. The mximum number of brriers is 19. b The seven mjor life chllenges mesured re: psychologicl distress, Post Trumtic Stress Disorder, felony conviction, experience of domestic violence, childhood seprtion (foster cre, group home, or institutionliztion), medicl condition, nd substnce buse. Notes: Chi-squre tests used to test the difference between groups for ll chrcteristics except men ge. t-test used for men ge. Sources: Fmily Options Bseline Survey; 18-month followup survey D-20

264 Appendix D. Anlysis of 18-Month Survey Nonresponse Exhibit D-20. Equivlence t Bseline of Anlysis Smple for Respondents Versus Nonrespondents for SUB Versus UC Impct Comprison, Adult Survey Chrcteristic Response Nonresponse Difference Significnce Level Strs Number of fmilies % Age of household hed t RA (percent) Less thn 21 yers old * yers yers yers yers yers nd older Men ge (yers) Gender (percent) Femle Mle Mritl sttus (percent) Single (never mrried/widowed/ seprted/divorced) Mrried or mrrige-like sitution Rce/ethnicity (percent) Blck/Africn Americn, not Hispnic *** White, not Hispnic Hispnic Other Eductionl ttinment (percent) Less thn high school diplom High school diplom/ged More thn high school diplom Number of dults in fmily (percent) 1 dult or more dults Number of children in fmily (percent) 1 child children children children or more Missing dt Worked for py lst week (percent) Ever convicted of felony (percent) Fmily nnul income (percent) Less thn $5, $5,000 9, $10,000 14, $15,000 19, $20,000 24, $25,000 or more Missing dt Ever been homeless before (percent) * Ever been doubled up before (percent) Mjor brrier to finding housing ** Child under 18 living elsewhere (percent) Number of mjor life chllengesb F-test on ll chrcteristics except site F vlue = F-test p-vlue = SUB = permnent housing subsidy. UC = usul cre. GED = generl eductionl development. RA = rndom ssignment. *.10 level. **.05 level. ***.01 level. Brriers to finding housing were reported by fmily heds s "big problems" in finding housing. The mximum number of brriers is 19. b The seven mjor life chllenges mesured re: psychologicl distress, Post Trumtic Stress Disorder, felony conviction, experience of domestic violence, childhood seprtion (foster cre, group home, or institutionliztion), medicl condition, nd substnce buse. Notes: Chi-squre tests used to test the difference between groups for ll chrcteristics except men ge. t-test used for men ge. Sources: Fmily Options Bseline Survey; 18-month followup survey D-21

265 Appendix D. Anlysis of 18-Month Survey Nonresponse Exhibit D-21. Equivlence t Bseline of Anlysis Smple for Respondents Versus Nonrespondents for CBRR Versus UC Impct Comprison, Adult Survey Chrcteristic Response Nonresponse Difference Significnce Level Strs Number of fmilies % Age of household hed t RA (percent) Less thn 21 yers old yers yers yers yers yers nd older Men ge (yers) Gender (percent) Femle Mle Mritl sttus (percent) Single (never mrried/widowed/ seprted/divorced) Mrried or mrrige-like sitution Rce/ethnicity (percent) Blck/Africn Americn, not Hispnic ** White, not Hispnic Hispnic Other Eductionl ttinment (percent) Less thn high school diplom High school diplom/ged More thn high school diplom Number of dults in fmily (percent) 1 dult or more dults Number of children in fmily (percent) 1 child children children children or more Missing dt Worked for py lst week (percent) Ever convicted of felony (percent) Fmily nnul income (percent) Less thn $5, $5,000 9, $10,000 14, $15,000 19, $20,000 24, $25,000 or more Missing dt Ever been homeless before (percent) Ever been doubled up before (percent) *** Mjor brrier to finding housing *** Child under 18 living elsewhere (percent) ** Number of mjor life chllengesb F-test on ll chrcteristics except site F vlue = F-test p-vlue = 3 CBRR = community-bsed rpid re-housing. UC = usul cre. GED = generl eductionl development. RA = rndom ssignment. *.10 level. **.05 level. ***.01 level. Brriers to finding housing were reported by fmily heds s "big problems" in finding housing. The mximum number of brriers is 19. b The seven mjor life chllenges mesured re: psychologicl distress, Post Trumtic Stress Disorder, felony conviction, experience of domestic violence, childhood seprtion (foster cre, group home, or institutionliztion), medicl condition, nd substnce buse. Notes: Chi-squre tests used to test the difference between groups for ll chrcteristics except men ge. t-test used for men ge. Sources: Fmily Options Bseline Survey; 18-month followup survey D-22

266 Appendix D. Anlysis of 18-Month Survey Nonresponse Exhibit D-22. Equivlence t Bseline of Anlysis Smple for Respondents Versus Nonrespondents for PBTH Versus UC Impct Comprison, Adult Survey Chrcteristic Response Nonresponse Difference Significnce Level Strs Number of fmilies % Age of household hed t RA (percent) Less thn 21 yers old yers yers yers yers yers nd older Men ge (yers) Gender (percent) Femle Mle Mritl sttus (percent) Single (never mrried/widowed/ seprted/divorced) Mrried or mrrige-like sitution Rce/ethnicity (percent) Blck/Africn Americn, not Hispnic *** White, not Hispnic Hispnic Other Eductionl ttinment (percent) Less thn high school diplom High school diplom/ged More thn high school diplom Number of dults in fmily (percent) 1 dult or more dults Number of children in fmily (percent) 1 child children children children or more Missing dt Worked for py lst week (percent) Ever convicted of felony (percent) Fmily nnul income (percent) Less thn $5, $5,000 9, $10,000 14, $15,000 19, $20,000 24, $25,000 or more Missing dt Ever been homeless before (percent) Ever been doubled up before (percent) Mjor brrier to finding housing Child under 18 living elsewhere (percent) Number of mjor life chllengesb F-test on ll chrcteristics except site F vlue = F-test p-vlue = PBTH = project-bsed trnsitionl housing. UC = usul cre. GED = generl eductionl development. RA = rndom ssignment. *.10 level. **.05 level. ***.01 level. Brriers to finding housing were reported by fmily heds s "big problems" in finding housing. The mximum number of brriers is 19. b The seven mjor life chllenges mesured re: psychologicl distress, Post Trumtic Stress Disorder, felony conviction, experience of domestic violence, childhood seprtion (foster cre, group home, or institutionliztion), medicl condition, nd substnce buse. Notes: Chi-squre tests used to test the difference between groups for ll chrcteristics except men ge. t-test used for men ge. Sources: Fmily Options Bseline Survey; 18-month followup survey D-23

267 Appendix D. Anlysis of 18-Month Survey Nonresponse Exhibit D-23. Equivlence t Bseline of Anlysis Smple for Respondents Versus Nonrespondents for SUB Versus CBRR Impct Comprison, Adult Survey Chrcteristic Response Nonresponse Difference Significnce Level Strs Number of fmilies % Age of household hed t RA (percent) Less thn 21 yers old * yers yers yers yers yers nd older Men ge (yers) Gender (percent) Femle * Mle Mritl sttus (percent) Single (never mrried/widowed/ seprted/divorced) Mrried or mrrige-like sitution Rce/ethnicity (percent) Blck/Africn Americn, not Hispnic ** White, not Hispnic Hispnic Other Eductionl ttinment (percent) Less thn high school diplom High school diplom/ged More thn high school diplom Number of dults in fmily (percent) 1 dult or more dults Number of children in fmily (percent) 1 child children children children or more Missing dt Worked for py lst week (percent) Ever convicted of felony (percent) Fmily nnul income (percent) Less thn $5, $5,000 9, $10,000 14, $15,000 19, $20,000 24, $25,000 or more Missing dt 0.02 Ever been homeless before (percent) *** Ever been doubled up before (percent) *** Mjor brrier to finding housing Child under 18 living elsewhere (percent) *** Number of mjor life chllengesb F-test on ll chrcteristics except site F vlue = F-test p-vlue = CBRR = community-bsed rpid re-housing. SUB = permnent housing subsidy. GED = generl eductionl development. RA = rndom ssignment. *.10 level. **.05 level. ***.01 level. Brriers to finding housing were reported by fmily heds s "big problems" in finding housing. The mximum number of brriers is 19. b The seven mjor life chllenges mesured re: psychologicl distress, Post Trumtic Stress Disorder, felony conviction, experience of domestic violence, childhood seprtion (foster cre, group home, or institutionliztion), medicl condition, nd substnce buse. Notes: Chi-squre tests used to test the difference between groups for ll chrcteristics except men ge. t-test used for men ge. Sources: Fmily Options Bseline Survey; 18-month followup survey D-24

268 Appendix D. Anlysis of 18-Month Survey Nonresponse Exhibit D-24. Equivlence t Bseline of Anlysis Smple for Respondents Versus Nonrespondents for SUB Versus PBTH Impct Comprison, Adult Survey Chrcteristic Response Nonresponse Difference Significnce Level Strs Number of fmilies % Age of household hed t RA (percent) Less thn 21 yers old yers yers yers yers yers nd older Men ge (yers) Gender (percent) Femle * Mle Mritl sttus (percent) Single (never mrried/widowed/ seprted/divorced) Mrried or mrrige-like sitution Rce/ethnicity (percent) Blck/Africn Americn, not Hispnic *** White, not Hispnic Hispnic Other Eductionl ttinment (percent) Less thn high school diplom High school diplom/ged More thn high school diplom Number of dults in fmily (percent) 1 dult or more dults Number of children in fmily (percent) 1 child children children children or more Missing dt Worked for py lst week (percent) Ever convicted of felony (percent) Fmily nnul income (percent) Less thn $5, $5,000 9, $10,000 14, $15,000 19, $20,000 24, $25,000 or more Missing dt Ever been homeless before (percent) *** Ever been doubled up before (percent) Mjor brrier to finding housing Child under 18 living elsewhere (percent) Number of mjor life chllengesb F-test on ll chrcteristics except site F vlue = F-test p-vlue = PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. GED = generl eductionl development. RA = rndom ssignment. *.10 level. **.05 level. ***.01 level. Brriers to finding housing were reported by fmily heds s "big problems" in finding housing. The mximum number of brriers is 19. b The seven mjor life chllenges mesured re: psychologicl distress, Post Trumtic Stress Disorder, felony conviction, experience of domestic violence, childhood seprtion (foster cre, group home, or institutionliztion), medicl condition, nd substnce buse. Notes: Chi-squre tests used to test the difference between groups for ll chrcteristics except men ge. t-test used for men ge. Sources: Fmily Options Bseline Survey; 18-month followup survey D-25

269 Appendix D. Anlysis of 18-Month Survey Nonresponse Exhibit D-25. Equivlence t Bseline of Anlysis Smple for Respondents Versus Nonrespondents for CBRR Versus PBTH Impct Comprison, Adult Survey Chrcteristic Response Nonresponse Difference Significnce Level Strs Number of fmilies % Age of household hed t RA (percent) Less thn 21 yers old yers yers yers yers yers nd older Men ge (yers) Gender (percent) Femle ** Mle Mritl sttus (percent) Single (never mrried/widowed/ seprted/divorced) Mrried or mrrige-like sitution Rce/ethnicity (percent) Blck/Africn Americn, not Hispnic White, not Hispnic Hispnic Other Eductionl ttinment (percent) Less thn high school diplom High school diplom/ged More thn high school diplom Number of dults in fmily (percent) 1 dult or more dults Number of children in fmily (percent) 1 child children children children or more Missing dt Worked for py lst week (percent) Ever convicted of felony (percent) Fmily nnul income (percent) Less thn $5, $5,000 9, $10,000 14, $15,000 19, $20,000 24, $25,000 or more Missing dt Ever been homeless before (percent) Ever been doubled up before (percent) *** Mjor brrier to finding housing Child under 18 living elsewhere (percent) Number of mjor life chllengesb ** F-test on ll chrcteristics except site F vlue = F-test p-vlue = CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. GED = generl eductionl development. RA = rndom ssignment. *.10 level. **.05 level. ***.01 level. Brriers to finding housing were reported by fmily heds s "big problems" in finding housing. The mximum number of brriers is 19. b The seven mjor life chllenges mesured re: psychologicl distress, Post Trumtic Stress Disorder, felony conviction, experience of domestic violence, childhood seprtion (foster cre, group home, or institutionliztion), medicl condition, nd substnce buse. Notes: Chi-squre tests used to test the difference between groups for ll chrcteristics except men ge. t-test used for men ge. Sources: Fmily Options Bseline Survey; 18-month followup survey D-26

270 Appendix D. Anlysis of 18-Month Survey Nonresponse Exhibit D-26. Equivlence t Bseline of Anlysis Smple for Respondents Versus Nonrespondents for SUB + CBRR + PBTH Versus UC Impct Comprison, Adult Survey Chrcteristic Response Nonresponse Difference Significnce Level Strs Number of fmilies 1, % Age of household hed t RA (percent) Less thn 21 yers old yers yers yers yers yers nd older Men ge (yers) Gender (percent) Femle ** Mle Mritl sttus (percent) Single (never mrried/widowed/ seprted/divorced) Mrried or mrrige-like sitution Rce/ethnicity (percent) Blck/Africn Americn, not Hispnic *** White, not Hispnic Hispnic Other Eductionl ttinment (percent) Less thn high school diplom High school diplom/ged More thn high school diplom Number of dults in fmily (percent) 1 dult or more dults Number of children in fmily (percent) 1 child children children children or more Missing dt Worked for py lst week (percent) Ever convicted of felony (percent) Fmily nnul income (percent) Less thn $5, $5,000 9, $10,000 14, $15,000 19, $20,000 24, $25,000 or more Missing dt Ever been homeless before (percent) ** Ever been doubled up before (percent) *** Mjor brrier to finding housing ** Child under 18 living elsewhere (percent) Number of mjor life chllengesb F-test on ll chrcteristics except site F vlue = F-test p-vlue = 0 CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. UC = usul cre. GED = generl eductionl development. RA = rndom ssignment. *.10 level. **.05 level. ***.01 level. Brriers to finding housing were reported by fmily heds s "big problems" in finding housing. The mximum number of brriers is 19. b The seven mjor life chllenges mesured re: psychologicl distress, Post Trumtic Stress Disorder, felony conviction, experience of domestic violence, childhood seprtion (foster cre, group home, or institutionliztion), medicl condition, nd substnce buse. Notes: Chi-squre tests used to test the difference between groups for ll chrcteristics except men ge. t-test used for men ge. Sources: Fmily Options Bseline Survey; 18-month followup survey D-27

271 Appendix D. Anlysis of 18-Month Survey Nonresponse Exhibit D-27. Equivlence t Bseline of Anlysis Smple for Respondents Versus Nonrespondents for SUB + PBTH Versus CBRR Impct Comprison, Adult Survey Chrcteristic Response Nonresponse Difference Significnce Level Strs Number of fmilies % Age of household hed t RA (percent) Less thn 21 yers old yers yers yers yers yers nd older Men ge (yers) Gender (percent) Femle *** Mle Mritl sttus (percent) Single (never mrried/widowed/ seprted/divorced) Mrried or mrrige-like sitution Rce/ethnicity (percent) Blck/Africn Americn, not Hispnic *** White, not Hispnic Hispnic Other Eductionl ttinment (percent) Less thn high school diplom High school diplom/ged More thn high school diplom Number of dults in fmily (percent) 1 dult or more dults Number of children in fmily (percent) 1 child children children children or more Missing dt Worked for py lst week (percent) Ever convicted of felony (percent) Fmily nnul income (percent) Less thn $5, $5,000 9, $10,000 14, $15,000 19, $20,000 24, $25,000 or more Missing dt Ever been homeless before (percent) *** Ever been doubled up before (percent) *** Mjor brrier to finding housing Child under 18 living elsewhere (percent) ** Number of mjor life chllengesb F-test on ll chrcteristics except site F vlue = F-test p-vlue = CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. GED = generl eductionl development. RA = rndom ssignment. *.10 level. **.05 level. ***.01 level. Brriers to finding housing were reported by fmily heds s "big problems" in finding housing. The mximum number of brriers is 19. b The seven mjor life chllenges mesured re: psychologicl distress, Post Trumtic Stress Disorder, felony conviction, experience of domestic violence, childhood seprtion (foster cre, group home, or institutionliztion), medicl condition, nd substnce buse. Notes: Chi-squre tests used to test the difference between groups for ll chrcteristics except men ge. t-test used for men ge. Sources: Fmily Options Bseline Survey; 18-month followup survey D-28

272 Appendix D. Anlysis of 18-Month Survey Nonresponse Exhibit D-28. Equivlence t Bseline of Anlysis Smple for Respondents Versus Nonrespondents for SUB + CBRR Versus PBTH Impct Comprison, Adult Survey Chrcteristic Response Nonresponse Difference Significnce Level Strs Number of fmilies % Age of household hed t RA (percent) Less thn 21 yers old yers yers yers yers yers nd older Men ge (yers) Gender (percent) Femle ** Mle Mritl sttus (percent) Single (never mrried/widowed/ seprted/divorced) Mrried or mrrige-like sitution Rce/ethnicity (percent) Blck/Africn Americn, not Hispnic ** White, not Hispnic Hispnic Other Eductionl ttinment (percent) Less thn high school diplom High school diplom/ged More thn high school diplom Number of dults in fmily (percent) 1 dult or more dults Number of children in fmily (percent) 1 child children children children or more Missing dt Worked for py lst week (percent) Ever convicted of felony (percent) Fmily nnul income (percent) Less thn $5, $5,000 9, $10,000 14, $15,000 19, $20,000 24, $25,000 or more Missing dt Ever been homeless before (percent) *** Ever been doubled up before (percent) *** Mjor brrier to finding housing Child under 18 living elsewhere (percent) Number of mjor life chllengesb F-test on ll chrcteristics except site F vlue = F-test p-vlue = 9 CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. GED = generl eductionl development. RA = rndom ssignment. *.10 level. **.05 level. ***.01 level. Brriers to finding housing were reported by fmily heds s "big problems" in finding housing. The mximum number of brriers is 19. b The seven mjor life chllenges mesured re: psychologicl distress, Post Trumtic Stress Disorder, felony conviction, experience of domestic violence, childhood seprtion (foster cre, group home, or institutionliztion), medicl condition, nd substnce buse. Notes: Chi-squre tests used to test the difference between groups for ll chrcteristics except men ge. t-test used for men ge. Sources: Fmily Options Bseline Survey; 18-month followup survey D-29

273 Appendix D. Anlysis of 18-Month Survey Nonresponse Exhibit D-29. Equivlence t Bseline of Anlysis Smple for Respondents Versus Nonrespondents for PBTH + CBRR Versus SUB Impct Comprison, Adult Survey Chrcteristic Response Nonresponse Difference Significnce Level Strs Number of fmilies % Age of household hed t RA (percent) Less thn 21 yers old yers yers yers yers yers nd older Men ge (yers) Gender (percent) Femle *** Mle Mritl sttus (percent) Single (never mrried/widowed/ seprted/divorced) Mrried or mrrige-like sitution Rce/ethnicity (percent) Blck/Africn Americn, not Hispnic *** White, not Hispnic Hispnic Other Eductionl ttinment (percent) Less thn high school diplom High school diplom/ged More thn high school diplom Number of dults in fmily (percent) 1 dult or more dults Number of children in fmily (percent) 1 child children children children or more Missing dt Worked for py lst week (percent) Ever convicted of felony (percent) Fmily nnul income (percent) Less thn $5, $5,000 9, $10,000 14, $15,000 19, $20,000 24, $25,000 or more Missing dt Ever been homeless before (percent) *** Ever been doubled up before (percent) ** Mjor brrier to finding housing Child under 18 living elsewhere (percent) ** Number of mjor life chllengesb F-test on ll chrcteristics except site F vlue = F-test p-vlue = 7 CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. GED = generl eductionl development. RA = rndom ssignment. *.10 level. **.05 level. ***.01 level. Brriers to finding housing were reported by fmily heds s "big problems" in finding housing. The mximum number of brriers is 19. b The seven mjor life chllenges mesured re: psychologicl distress, Post Trumtic Stress Disorder, felony conviction, experience of domestic violence, childhood seprtion (foster cre, group home, or institutionliztion), medicl condition, nd substnce buse. Notes: Chi-squre tests used to test the difference between groups for ll chrcteristics except men ge. t-test used for men ge. Sources: Fmily Options Bseline Survey; 18-month followup survey D-30

274 Appendix D. Anlysis of 18-Month Survey Nonresponse D.3.2 Unweighted Impct Estimtes How were the min results of this report ffected by the use of nonresponse nlysis weights? As discussed in Appendix C, the study tem used nonresponse weights to produce ll estimtes in this report. This methodology is motivted in prt by the finding in the previous section tht for most impct comprisons, respondents nd nonrespondents differed systemticlly on number of bseline chrcteristics. Survey nonresponse weights djust impct estimtes such tht the nlysis smple reflects the observble chrcteristics of the bseline smple. This djustment represents correction, however, only insofr s impcts vry with observble prticipnt chrcteristics. The impct models lso controlled for observble bseline chrcteristics. Together, these mesures djust the impct estimtes to reflect potentil differences between the groups in ech comprison tht could hve been induced by nonresponse. In this section, we look for evidence of such vrition by compring the study s hedline impct estimtes to estimtes tht do not use survey nonresponse weights. The results re lso presented for completeness for reders interested in the undjusted estimtes. Exhibit D-30 presents these undjusted estimtes, which re comprble to Exhibit ES-5 in the executive summry. Chnges to coefficient signs, mgnitudes, nd sttisticl significnce re miniml. In every cse, the 95-percent confidence intervls ssocited with impct coefficients estimted without nonresponse weights included the coefficient estimtes in the min weighted specifiction, with intervls lrgely overlpping. Chnges of sign (tht is, from positive to negtive) were limited to coefficients tht were both close in mgnitude to zero nd estimted s sttisticlly insignificnt. Chnges in sttisticl significnce were minor, with the exception of the impct estimte for child reunifiction in SUB versus PBTH, which ws sttisticlly significnt t the.05 level when nonresponse weights were used but ws not significnt t the.10 level without weights. D-31

275 Appendix D. Anlysis of 18-Month Survey Nonresponse Exhibit D-30. Executive Summry Impct Estimtes, Estimted Without Nonresponse Weights Men Outcome All UC Group ITT Impct Estimtes SUB vs. UC CBRR vs. UC PBTH vs. UC SUB vs. CBRR SUB vs. PBTH CBRR vs. PBTH Housing stbility (intervention gol: lower vlues) At lest one night homeless or doubled up (pst 6 months) or in shelter (pst 12 months) (%) [Confirmtory] *** * 27.9 *** 28.7 *** 7.4 At lest one night homeless or doubled up in pst 6 months (%) *** *** 25.3 *** *** 0.38 *** 14.3 *** 12.1 *** 2.9 Number of plces lived in pst 6 months *** *** ** Any child seprted in pst 6 monthsb (%) ** Spouse/prtner seprted in pst 6 months c (%) [limited bse] No child reunified, of those with t lest one child seprted t bseline d (%) [limited bse] Any sty in emergency shelter in months 7 to 18 fter RA (%) Fmily preservtion (intervention gol: lower vlues) * ** ** Adult well-being (intervention gol: lower vlues) Helth in pst 30 dys ws poor or fir (%) K6 Psychologicl Distress Scle e 1.1 *** *** 1.7 *** Alcohol dependence or drug buse f (%) * Experienced intimte prtner violence in pst 6 months (%) *** *** *** *** ** ** *** 8.4 9, Child well-being (intervention gol: lower vlues) School mobility (number of schools since RA) g Childcre or school bsences h Child helth in pst 30 dys ws poor or fir (%) Strengths nd difficulties: totl problem score (z-score) i Self sufficiency (intervention gol: higher vlues) Work for py in week before survey (%) Totl fmily income ($) 1,101 ** * Household is food secure (%) *** ,632 ** Number of fmilies CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. UC = usul cre. ITT = intention-to-tret. RA = rndom ssignment. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test (not djusted for multiple comprisons). After djustment for multiple comprisons, the impct on the confirmtory outcome is not sttisticlly significnt for the PBTH versus UC comprison nd is sttisticlly significnt t the.01 level for the SUB versus UC, SUB versus CBRR, nd SUB versus PBTH comprisons. b Mesures the percentge of fmilies in which child who ws with the fmily t bseline becme seprted from the fmily in the six months prior to the 18-month survey. c Mesures the percentge of fmilies in which spouse or prtner who ws with the fmily t bseline becme seprted from the fmily in the 6 months prior to the 18-month survey. d Mesures the percentge of fmilies in which t lest one child ws seprted from the fmily t bseline nd no child ws reunited with the fmily t the time of the 18-month survey. e Mesures psychologicl distress using the Kessler-6 scle. Higher scores indicte greter distress. Impcts shown s stndrdized effect sizes. Effect sizes were stndrdized by dividing impcts by stndrd devition for the UC group. f Mesures evidence of lcohol dependence or drug buse using responses to the Rpid Alcohol Problems Screen (RAPS-4) nd 6 items from the Drug Abuse Screening Test (DAST-10). g Mesures re 1 = 1 school; 2 = 2 schools; 3 = 3 schools; 4 = 4 schools ttended since rndom ssignment. h Mesures re 0 = no bsences; 1 = 1-2 bsences; 2 = 3-5 bsences; 3 = 6 or more bsences in the month prior to the 18-month survey. i This mesure is the score on the Strengths nd Difficulties behviorl nd personlity ssessment. The totl problem score mesures emotionl symptoms, conduct problems, hyperctivity, nd peer problems. The score rnges from 0 to 40 with higher scores indicting more negtive behvior. Notes: Impct estimtes re regression-djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Sources: Fmily Options Study 18-Month followup survey; nd progrm usge dt D-32

276 Appendix E. Impcts on Use of Trnsitionl Housing APPENDIX E. IMPACTS ON USE OF TRANSITIONAL HOUSING After the reserch tem submitted the drft report of the Fmily Options Study, HUD requested tht the tem estimte impcts on seven dditionl outcomes relted to use of trnsitionl housing nd emergency shelter during the followup period. These outcomes more closely mesure the impct of rndom ssignment to contrsting interventions on homelessness s mesured in the McKinney-Vento Homeless Assistnce Act nd Opening Doors: Federl Strtegic Pln To Prevent nd End Homelessness. The dditionl outcomes re 1. Any use of trnsitionl housing in months 0 to 18 fter rndom ssignment. 2. Number of months using emergency shelter or trnsitionl housing in months 0 to 18 fter rndom ssignment. 3. Number of months using emergency shelter in months 0 to 18 fter rndom ssignment. 4. Number of months using trnsitionl housing in months 0 to 18 fter rndom ssignment. 5. Any use of emergency shelter or trnsitionl housing in months 7 to 18 fter rndom ssignment. 6. Any use of emergency shelter in months 7 to 18 fter rndom ssignment (this outcome is lso included in the housing stbility nlyses in Chpters 6 through 9 of the report). 7. Any use of trnsitionl housing in months 7 to 18 fter rndom ssignment. The new outcomes re mesured with Progrm Usge Dt. They differ from outcomes in the nlyses of housing stbility provided in Chpters 6 through 9, which did not consider use of trnsitionl housing, either seprtely or in conjunction with the use of emergency shelter. This ppendix includes six exhibits, one for ech pirwise comprison: (1) permnent housing subsidy (SUB) versus usul cre (UC), (2) community-bsed rpid re-housing (CBRR) versus UC, (3) project-bsed trnsitionl housing (PBTH) versus UC, (4) SUB versus CBRR, (5) SUB versus PBTH, nd (6) CBRR versus PBTH. For comprisons involving PBTH (Exhibits E-3, E-5, nd E-6), note tht impcts on the use of trnsitionl housing in prt reflect tkeup of the ssigned intervention encourged nd fcilitted for the PBTH group but not the other group included in the comprison by the study design. E -1

277 Appendix E. Impcts on Use of Trnsitionl Housing Exhibit E-1. SUB Versus UC: Impcts on Use of ES nd TH Outcome SUB N Men Any use of TH during months 0 18 Any use of trnsitionl housing during months 0 18 fter RA (%) Number of months of ES nd TH use during months 0 18 UC ITT Impct (SD) N Men (SD) Impct (SE) Effect Size (24.9) (40.4) 14.9 *** (2.4) 0.31 Number of months of emergency shelter (3.7) (5.7) 2.1 *** (0.3) 0.31 nd trnsitionl housing use during months 0 18 fter RA Number of months of emergency shelter (3.1) (4.7) 0.9 *** (0.2) 0.18 use during months 0 18 fter RA Number of months of trnsitionl housing (2.0) (3.8) 1.2 *** (0.2) 0.24 use during months 0 18 fter RA Any use of ES or TH during months 7 18 Any use of emergency shelter or trnsitionl (37.4) (49.4) 22.6 *** (3.0) 0.40 housing during months 7 18 fter RA (%) Any use of emergency shelter during (34.9) (45.9) 12.9 *** (2.6) 0.25 months 7 18 fter RA (%) Any use of trnsitionl housing during (18.6) (36.6) 13.4 *** (2.1) 0.30 months 7 18 fter RA (%) SUB = permnent housing subsidy. UC = usul cre. ES = emergency shelter. ITT = intention-to-tret. RA = rndom ssignment. SD = stndrd devition. SE = stndrd error. TH = trnsitionl housing. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire UC group. Notes: Impct estimtes nd outcome mens re regression-djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Appendix B for outcome definitions. Source: Fmily Options Study Progrm Usge dt Exhibit E-2. CBRR Versus UC: Impcts on Use of ES nd TH Outcome CBRR N Men UC (SD) N Men ITT Impct (SD) Impct (SE) Effect Size Any use of TH during months (37.7) (43.1) 5.6 ** (2.7) 0.11 Any use of trnsitionl housing during months 0 18 fter RA (%) Number of months of ES nd TH use during months (5.5) (5.9) 0.9 ** (0.4) 0.14 Number of months of emergency shelter nd trnsitionl housing use during months 0 18 fter RA (4.3) (4.6) 0.4 (0.3) 0.08 Number of months of emergency shelter use during months 0 18 fter RA (3.6) (4.4) 0.5 ** (0.3) 0.10 Number of months of trnsitionl housing use during months 0 18 fter RA Any use of ES or TH during months (47.9) (49.6) 6.1 * (3.4) 0.11 Any use of emergency shelter or trnsitionl housing during months 7 18 fter RA (%) (44.3) (45.7) 2.1 (3.1) 0.04 Any use of emergency shelter during months 7 18 fter RA (%) (33.4) (39.0) 4.2 * (2.5) 0.10 Any use of trnsitionl housing during months 7 18 fter RA (%) CBRR = community-bsed rpid re-housing. UC = usul cre. ES = emergency shelter. ITT = intention-to-tret. RA = rndom ssignment. SD = stndrd devition. SE = stndrd error. TH = trnsitionl housing. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire UC group. Notes: Impct estimtes nd outcome mens re regression-djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Appendix B for outcome definitions. Source: Fmily Options Study Progrm Usge dt E -2

278 Appendix E. Impcts on Use of Trnsitionl Housing Exhibit E-3. PBTH Versus UC: Impcts on Use of ES nd TH Outcome PBTH N Men UC (SD) N Men ITT Impct (SD) Impct (SE) Effect Size Any use of TH during months (49.9) (45.8) 24.2 *** (4.3) 0.50 Any use of trnsitionl housing during months 0 18 fter RA (%) Number of months of ES nd TH use during months (6.8) (5.9) 3.1 *** (0.5) 0.46 Number of months of emergency shelter nd trnsitionl housing use during months 0 18 fter RA (3.5) (4.1) 0.7 ** (0.3) 0.14 Number of months of emergency shelter use during months 0 18 fter RA (6.6) (4.7) 3.8 *** (0.5) 0.75 Number of months of trnsitionl housing use during months 0 18 fter RA Any use of ES or TH during months (49.7) (49.8) 14.1 *** (4.5) 0.25 Any use of emergency shelter or trnsitionl housing during months 7 18 fter RA (%) (38.8) (44.9) 8.2 ** (3.6) 0.16 Any use of emergency shelter during months 7 18 fter RA (%) (49.9) (41.3) 25.4 *** (4.1) 0.57 Any use of trnsitionl housing during months 7 18 fter RA (%) PBTH = project-bsed trnsitionl housing. UC = usul cre. ES = emergency shelter. ITT = intention-to-tret. RA = rndom ssignment. SD = stndrd devition. SE = stndrd error. TH = trnsitionl housing. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire UC group. Notes: Impct estimtes nd outcome mens re regression-djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Appendix B for outcome definitions. Source: Fmily Options Study Progrm Usge dt Exhibit E-4. SUB Versus CBRR: Impcts on Use of ES nd TH Outcome SUB N Men CBRR (SD) N Men ITT Impct (SD) Impct (SE) Effect Size Any use of TH during months (23.4) (35.4) 9.5 *** (2.4) 0.19 Any use of trnsitionl housing during months 0 18 fter RA (%) Number of months of ES nd TH use during months (3.7) (5.5) 1.4 *** (0.3) 0.21 Number of months of emergency shelter nd trnsitionl housing use during months 0 18 fter RA (3.3) (4.7) 0.6 ** (0.3) 0.13 Number of months of emergency shelter use during months 0 18 fter RA (1.7) (3.3) 0.7 *** (0.2) 0.15 Number of months of trnsitionl housing use during months 0 18 fter RA Any use of ES or TH during months (37.7) (48.5) 19.5 *** (3.4) 0.35 Any use of emergency shelter or trnsitionl housing during months 7 18 fter RA (%) (35.5) (45.8) 13.2 *** (3.1) 0.26 Any use of emergency shelter during months 7 18 fter RA (%) (17.5) (32.2) 9.2 *** (2.1) 0.21 Any use of trnsitionl housing during months 7 18 fter RA (%) CBRR = community-bsed rpid re-housing. SUB = permnent housing subsidy. ES = emergency shelter. ITT = intention-to-tret. RA = rndom ssignment. SD = stndrd devition. SE = stndrd error. TH = trnsitionl housing. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire UC group. Notes: Impct estimtes nd outcome mens re regression-djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Appendix B for outcome definitions. Source: Fmily Options Study Progrm Usge dt E -3

279 Appendix E. Impcts on Use of Trnsitionl Housing Exhibit E-5. SUB Versus PBTH: Impcts on Use of ES nd TH Outcome SUB N Men PBTH (SD) N Men ITT Impct (SD) Impct (SE) Effect Size Any use of TH during months (28.2) (50.1) 44.3 *** (4.1) 0.91 Any use of trnsitionl housing during months 0 18 fter RA (%) Number of months of ES nd TH use during months (3.3) (6.8) 5.9 *** (0.5) 0.89 Number of months of emergency shelter nd trnsitionl housing use during months 0 18 fter RA (2.0) (3.5) 0.7 ** (0.3) 0.15 Number of months of emergency shelter use during months 0 18 fter RA (2.5) (6.7) 5.1 *** (0.5) 1.02 Number of months of trnsitionl housing use during months 0 18 fter RA Any use of ES or TH during months (31.2) (50.0) 45.4 *** (4.3) 0.81 Any use of emergency shelter or trnsitionl housing during months 7 18 fter RA (%) (25.5) (39.1) 13.9 *** (3.6) 0.27 Any use of emergency shelter during months 7 18 fter RA (%) (21.4) (49.5) 37.4 *** (3.9) 0.84 Any use of trnsitionl housing during months 7 18 fter RA (%) PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. ES = emergency shelter. ITT = intention-to-tret. RA = rndom ssignment. SD = stndrd devition. SE = stndrd error. TH = trnsitionl housing. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire UC group. Notes: Impct estimtes nd outcome mens re regression-djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Appendix B for outcome definitions. Source: Fmily Options Study Progrm Usge dt Exhibit E-6. CBRR Versus PBTH: Impcts on Use of ES nd TH Outcome CBRR N Men PBTH (SD) N Men ITT Impct (SD) Impct (SE) Effect Size Any use of TH during months (43.8) (50.0) 28.9 *** (5.5) 0.59 Any use of trnsitionl housing during months 0 18 fter RA (%) Number of months of ES nd TH use during months (5.8) (6.8) 3.6 *** (0.7) 0.54 Number of months of emergency shelter nd trnsitionl housing use during months 0 18 fter RA (3.9) (3.7) 0.1 (0.4) Number of months of emergency shelter use during months 0 18 fter RA (4.3) (6.5) 3.5 *** (0.7) 0.70 Number of months of trnsitionl housing use during months 0 18 fter RA Any use of ES or TH during months (47.1) (49.9) 23.1 *** (5.3) 0.41 Any use of emergency shelter or trnsitionl housing during months 7 18 fter RA (%) (41.8) (38.7) 1.4 (4.4) 0.03 Any use of emergency shelter during months 7 18 fter RA (%) (38.4) (49.9) 27.6 *** (5.2) 0.62 Any use of trnsitionl housing during months 7 18 fter RA (%) CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. ES = emergency shelter. ITT = intention-to-tret. RA = rndom ssignment. SD = stndrd devition. SE = stndrd error. TH = trnsitionl housing. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire UC group. Notes: Impct estimtes nd outcome mens re regression-djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Appendix B for outcome definitions. Source: Fmily Options Study Progrm Usge dt E -4

280 Appendix F. Impct Estimtes for Pooled Comprisons APPENDIX F. IMPACT ESTIMATES FOR POOLED COMPARISONS F -1

281 Appendix F. Impct Estimtes for Pooled Comprisons Exhibit F-1. SUB + CBRR + PBTH Versus UC: Impcts on Housing Stbility UC ITT Impct Effect Size N Men (SD) N Men (SD) Impct (SE) Homelessness or doubled up during the followup period At lest 1 night homelessb or doubled up 1, (48.2) (50.0) 13.6 *** (2.6) 0.24 (pst 6 months) or in shelter (pst c 12 months) (%) [Confirmtory] At lest 1 night homelessb or doubled up 1, (45.6) (49.1) 11.6 *** (2.5) 0.21 in pst 6 months (%) At lest 1 night homelessb in pst 1, (37.7) (42.9) 8.2 *** (2.2) months (%) At lest 1 night doubled up in pst 1, (42.2) (46.1) 8.5 *** (2.3) months (%) Any sty in emergency shelter in months 1, (39.7) (45.2) 7.6 *** (2.2) to 18 fter rndom ssignment (%) Number of dys homelessb or doubled 1, (66.5) (74.8) 16.0 *** (3.8) 0.19 up in pst 6 months Number of dys homelessb in pst 1, (44.4) (49.9) 6.5 ** (2.6) months Number of dys doubled up in pst 1, (55.7) (65.1) 11.5 *** (3.2) months Housing independence Living in own house or prtment t 1, (47.6) (49.5) 7.3 *** (2.5) 0.13 followup (%) Living in own house or prtment with 1, (45.0) (47.1) 6.3 *** (2.4) 0.12 no housing ssistnce (%) Living in own house or prtment with 1, (48.3) (42.9) 13.6 *** (2.3) 0.28 housing ssistnce (%) Number of plces lived Number of plces lived in pst 6 months 1, (1.0) (1.2) 0.2 *** (0.1) 0.13 Housing qulity Persons per room 1, (1.1) (1.2) 0.2 *** (0.1) (21.4) (49.5) 37.4 *** (3.9) 0.84 Any use of trnsitionl housing during months 7 18 fter RA (%) CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. UC = usul cre. ITT = intention-to-tret. RA = rndom ssignment. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire UC group. b The definition of homeless in this report includes stys in emergency shelters nd plces not ment for humn hbittion. It excludes trnsitionl housing. c After djustment for multiple comprisons, the impct on the confirmtory outcome is sttisticlly significnt t the.01 level for the SUB + CBRR + PBTH versus UC comprison. Notes: Impct estimtes nd outcome mens re regression-djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Sources: Fmily Options Study 18-month followup survey; Progrm Usge dt Outcome SUB + CBRR + PBTH F -2

282 Appendix F. Impct Estimtes for Pooled Comprisons Exhibit F-2. SUB + CBRR + PBTH Versus UC: Impcts on Fmily Preservtion UC ITT Impct Effect Size N Men (SD) N Men (SD) Impct (SE) Current or recent seprtions of fmily members present t bseline Fmily hs t lest one child seprted 1, (33.2) (35.6) 3.0 (1.9) 0.07 in pst 6 months (%) Fmily hs t lest one foster cre 1, (15.0) (19.2) 2.0 * (1.0) 0.09 plcement in pst 6 months (%) Spouse/prtner seprted in pst (48.0) (48.3) 2.6 (4.8) months, of those with spouse/ prtner present t RA (%) Reunifiction of fmily members reported s seprted t bseline Fmily hs t lest one child reunified, of (47.5) (45.0) 5.3 (5.9) 0.11 those fmilies with t lest one child bsent t RA (%) Spouse/prtner reunified, of those with (47.2) (48.7) 3.2 (9.6) 0.06 spouse/prtner bsent t RA (%) CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. UC = usul cre. ITT = intention-to-tret. RA = rndom ssignment. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire UC group. Notes: Impct estimtes nd outcome mens re regression-djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Sources: Fmily Options Study 18-month followup survey Outcome SUB + CBRR + PBTH Exhibit F-3. SUB + CBRR + PBTH Versus UC: Impcts on Adult Well-Being Outcome SUB + CBRR + PBTH N Men (SD) UC N Men ITT Impct (SD) Impct (SE) Effect Size Adult physicl helth Helth in pst 30 dys ws poor or fir (%) 1, (46.2) (46.6) 0.3 (2.2) Adult mentl helth Gol-oriented thinkingb 1, (1.02) (1.03) 0.08 (0.05) 0.07 Psychologicl distressc 1, (5.46) (5.69) 0.49 * (0.28) 0.07 Adult trum symptoms Post-trumtic stress disorder (PTSD) 1, (41.7) (43.5) 2.3 (2.2) 0.05 symptoms in pst 30 dys (%) Adult substnce use Alcohol dependence or drug bused (%) 1, (33.2) (35.3) 2.6 (1.8) 0.06 Alcohol dependenced (%) 1, (29.5) (32.1) 2.4 (1.7) 0.07 Drug bused (%) 1, (19.8) (23.6) 2.1 * (1.1) 0.08 Experience of intimte prtner violence Experienced intimte prtner violence in 1, (27.7) (32.5) 3.2 ** (1.6) 0.09 pst 6 months (%) CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. UC = usul cre. ITT = intention-to-tret. RA = rndom ssignment. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire UC group. b Gol-oriented thinking is mesured with modified version of the Stte Hope Scle nd rnges from 1 to 6, with higher scores indicting higher levels of positive, gol-oriented thinking. c Psychologicl distress is mesured with the Kessler-6 (K6) scle nd rnges from 0 to 24, with higher scores indicting greter distress. d Alcohol dependence is mesured with the Rpid Alcohol Problems Screen-4 (RAPS4) nd drug buse is mesured with 6 items from the Drug Abuse Screening Test-10 (DAST-10). Notes: Impct estimtes nd outcome mens re regression-djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Sources: Fmily Options Study 18-month followup survey F -3

283 Appendix F. Impct Estimtes for Pooled Comprisons Exhibit F-4. SUB + CBRR + PBTH Versus UC: Impcts on Child Well-Being Across Age Groups Outcome SUB + CBRR + PBTH N Men (SD) UC N Men ITT Impct (SD) Impct (SE) Effect Size Child eduction Preschool or Hed Strt enrollment (%)b (48.7) (48.3) 4.8 (4.0) 0.08 School enrollment (%)c 1, (29.6) (26.6) 1.8 (1.5) 0.05 Child cre or school bsences in pst 1, (0.93) (0.99) 0.10 (0.06) 0.07 monthd Number of schools ttended since 1, (0.82) (0.88) 0.11 ** (0.05) 0.10 rndom ssignmente Grde completion (not held bck) (%) 1, (25.8) (29.1) 1.6 (1.6) 0.04 Positive childcre or school experiencesf 1, (0.54) (0.56) 0.05 * (0.03) 0.07 Positive childcre or school ttitudesg 1, (1.00) (1.00) 0.05 (0.05) 0.04 School grdesh (0.92) (0.96) 0.10 (0.06) 0.08 Child cre or school conduct problemsi 1, (0.42) (0.43) 0.02 (0.02) 0.04 Child physicl helth Poor or fir helth (%) 1, (21.8) (21.6) 0.5 (1.0) 0.02 Well-child check-up in pst yer (%) 1, (29.8) (28.6) 0.4 (1.7) Child hs regulr source of helth cre (%) 1, (23.6) (24.9) 1.3 (1.5) 0.04 Sleep problemsj 1, (1.10) (1.12) 0.02 (0.06) Child behviorl strengths nd chllenges Behvior problemsk 1, (1.22) (1.24) 0.12 (0.07) 0.07 Prosocil behviorl 1, (1.10) (1.08) 0.04 (0.06) 0.03 CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. UC = usul cre. ITT = intention-to-tret. RA = rndom ssignment. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire UC group. b Bse for preschool or Hed Strt enrollment is children ge 1.5 to 5 yers. c Bse for school enrollment is children ge 6 to 17. d Absences outcome is defined s 0 = No bsences in pst month, 1 = 1 to 2 bsences, 2 = 3 to 5 bsences, 3 = 6 or more bsences. e Number of schools outcome is topcoded t 4 or more schools. f Positive childcre or school experiences outcome is defined s -1 = mostly negtive experiences, 0 = Both positive nd negtive experiences, 1 = mostly positive experiences. g Positive childcre or school ttitudes outcome is prent report of how much child likes school nd rnges from 1 to 5, with higher vlues indicting greter like of school. h School grdes outcome is defined s 1 = mostly Ds or Fs, 2 = mostly Cs, 3 = mostly Bs, 4 = mostly As. i Childcre or school conduct problems outcome is defined s 0 = No conduct problems reported to prent, 1 = Prent contcted bout conduct problems or suspension or expulsion from school or childcre center. j Sleep problems outcome rnges from 1 to 5, with higher vlues indicting more frequent tiredness upon wking nd during the dy. k Behvior problems outcome is mesured s the stndrdized Totl Difficulties score from the Strengths nd Difficulties Questionnire (SDQ). l Prosocil behvior is mesured s the stndrdized Prosocil domin score from the Strengths nd Difficulties Questionnire (SDQ). Notes: Impct estimtes nd outcome mens re regression-djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Source: Fmily Options Study 18-month followup survey F -4

284 Appendix F. Impct Estimtes for Pooled Comprisons Exhibit F-5. SUB + CBRR + PBTH Versus UC: Impcts on Child Well-Being Developmentl Outcomes by Age Group Outcome SUB + CBRR + PBTH N Men (SD) UC N Men ITT Impct (SD) Impct (SE) Effect Size Ages 1 yer to 3 yers, 6 months Met developmentl milestones (%)b (43.3) (40.6) 2.3 (4.9) 0.04 Low birth weight (%)c (31.3) (24.6) 1.0 (9.7) 0.03 Ages 3 yers, 6 months to 7 yers Verbl bilityd (0.95) (1.01) 0.05 (0.09) 0.03 Mth bilitye (0.90) (0.88) 0.05 (0.08) 0.04 Executive functioning (self-regultion)f (16.04) (16.07) 0.45 (1.03) 0.02 Ages 8 to 17 yers Anxietyg (7.70) (7.58) 1.17 * (0.63) 0.11 Fersh (14.28) (14.48) 0.75 (1.18) 0.03 Substnce use (%)i (29.21) (28.16) 0.67 (2.02) 0.02 Gol-oriented thinkingj (4.70) (5.08) 0.07 (0.37) School effort in pst monthk (0.81) (0.77) 0.09 (0.06) 0.09 Arrests or police involvement in pst (31.04) (31.60) 1.67 (3.67) months (%)l CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. UC = usul cre. ITT = intention-to-tret. RA = rndom ssignment. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire UC group. b Met developmentl milestones outcome is defined s scoring bove the typicl development cutoffs in ll domins of the ASQ-3. c Bse for low birthweight outcome (prent report) is children born since rndom ssignment who re t lest 1 yer old t followup. d Verbl bility outcome is the ntionlly stndrdized score from the WJIII Letter-Word Identifiction test. e Mth bility outcome is the ntionlly stndrdized score from the WJIII Applied Problems test. f Executive functioning outcome is the HTKS score nd rnges from 0 to 40, with higher scores indicting greter executive functioning. g Anxiety (child report) is mesured using the A-Trit scle from the Stte-Trit Anxiety Inventory for Children (STAIC). Scores rnge from 20 to 60, with higher scores indicting greter nxiety. h Fers outcome (child report) is the score from the Fers Scle nd rnges from 33 to 99, with higher scores indicting more fer. i Substnce use (child report) is mesured with 23 items from the Centers for Disese Control nd Prevention 2011 Youth Risk Behvior Survey. j Gol-oriented thinking (child report) is mesured with modified version of the Children's Hope Scle nd rnges from 6 to 30, with higher scores indicting higher levels of positive, gol-oriented thinking. k School effort outcome (child report) rnges from 1 to 4, with higher scores indicting greter effort during school dy nd on homework. l Arrest or police involvement in pst 6 months is from prent report. Notes: Impct estimtes nd outcome mens re regression-djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Sources: Fmily Options Study 18-month followup survey (prent report); Ages nd Stges Questionnire (ASQ-3); Woodcock-Johnson III Assessment (WJIII); Hed, Toes, Knees, nd Shoulders (HTKS) Assessment; Fmily Options Study 18-Month Child Survey (child report) F -5

285 Appendix F. Impct Estimtes for Pooled Comprisons Exhibit F-6. SUB + CBRR + PBTH Versus UC: Impcts on Self-Sufficiency Outcome SUB + CBRR + PBTH N Men (SD) UC N Men ITT Impct (SD) Impct (SE) Effect Size Employment sttus Work for py in week before 1, (46.5) (46.8) 1.1 (2.3) 0.02 survey (%) Any work for py since RA (%) 1, (49.2) (48.6) 3.7 (2.3) 0.07 Months worked for py since RAb 1, (7.5) (7.6) 0.4 (0.3) 0.05 Hours of work per week t current 1, (15.9) (15.6) 0.1 (0.8) min jobc Income sources/mounts Annulized current ernings ($) 1,253 4,922 (9,066) 565 4,935 (9,317) 13 (448) Totl fmily income ($) 1,235 9,636 (7,892) 562 9,080 (7,592) 556 (390) 0.06 Anyone in fmily hd ernings in pst 1, (49.4) (49.5) 2.7 (2.5) 0.05 month (%) Anyone in fmily received TANF in 1, (47.7) (46.6) 4.3 * (2.3) 0.08 pst month (%) Anyone in fmily received SSDI in 1, (25.5) (25.7) 0.6 (1.2) 0.02 pst month (%) Anyone in fmily received SSI in pst 1, (33.7) (33.5) 1.2 (1.4) 0.03 month (%) Anyone in fmily received SNAP/Food 1, (32.9) (37.1) 4.8 ** (1.9) 0.11 Stmps in pst month (%) Anyone in fmily received WIC in pst 1, (46.8) (45.8) 0.5 (2.2) month (%) Eduction nd trining Prticipted in 2+ weeks of ny 1, (43.9) (43.6) 0.3 (2.3) school or trining since RA (%) Number of weeks in school/trining 1, (10.0) (9.6) 0.1 (0.5) progrms since RA Prticipted in 2+ weeks of school 1, (24.4) (25.4) 1.1 (1.4) 0.04 since RA (%) Prticipted in 2+ weeks of bsic 1, (12.1) (13.7) 0.2 (0.6) 0.02 eduction since RA (%) Prticipted in 2+ weeks of voctionl 1, (25.1) (25.7) 0.8 (1.3) 0.03 eduction since RA (%) Food security/hunger Household is food insecure (%) 1, (45.2) (47.9) 5.7 ** (2.4) 0.10 Food insecurity scled 1, (1.98) (2.03) 0.21 ** (0.10) 0.09 Economic stressors Economic stress sclee 1, (0.48) (0.51) 0.10 *** (0.03) 0.17 CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. UC = usul cre. ITT = intention-to-tret. RA = rndom ssignment. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire UC group. b Number of months worked for py includes prtil clendr months. c Hours of work per week includes those not currently working (i.e., those with 0 hours of work per week). d Food insecurity scle rnges from 0 to 6, with higher vlues indicting higher food insecurity. e Economic stress scle rnges from -1 to 1, with higher vlues indicting higher economic stress. Notes: Impct estimtes nd outcome mens re regression-djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Source: Fmily Options Study 18-month followup survey F -6

286 Appendix F. Impct Estimtes for Pooled Comprisons Exhibit F-7. SUB + CBRR Versus PBTH: Impcts on Housing Stbility SUB + CBRR PBTH ITT Impct Effect Size N Men (SD) N Men (SD) Impct (SE) Homelessness or doubled up during the followup period At lest 1 night homelessb or doubled up (47.0) (49.9) 13.7 *** (4.1) 0.24 (pst 6 months) or in shelter (pst c 12 months) (%) [Confirmtory] At lest 1 night homelessb or doubled up (44.8) (48.9) 12.2 *** (3.9) 0.22 in pst 6 months (%) At lest 1 night homelessb in pst (37.3) (39.2) 5.7 * (3.3) months (%) At lest 1 night doubled up in pst (41.1) (47.6) 12.1 *** (3.7) months (%) Any sty in emergency shelter in months (34.4) (38.7) 6.0 ** (3.0) to 18 fter rndom ssignment (%) Number of dys homelessb or doubled (66.9) (71.2) 11.4 * (5.9) 0.13 up in pst 6 months Number of dys homelessb in pst (44.5) (42.0) 1.2 (3.7) months Number of dys doubled up in pst (57.0) (65.8) 13.3 ** (5.3) months Housing independence Living in own house or prtment t (45.4) (49.6) 15.3 *** (4.3) 0.27 followup (%) Living in own house or prtment with (43.2) (49.3) 13.6 *** (4.0) 0.25 no housing ssistnce (%) Living in own house or prtment with (49.9) (36.7) 28.8 *** (3.7) 0.59 housing ssistnce (%) Number of plces lived Number of plces lived in pst 6 months (0.9) (1.1) 0.3 *** (0.1) 0.20 Housing qulity Persons per room (1.1) (1.2) 0.2 ** (0.1) (42.8) (48.6) 13.7 *** (3.8) 0.25 Any use of trnsitionl housing during months 7 18 fter RA (%) CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. UC = usul cre. ITT = intention-to-tret. RA = rndom ssignment. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire UC group. b The definition of homeless in this report includes stys in emergency shelters nd plces not ment for humn hbittion. It excludes trnsitionl housing. c The seven confirmtory sttisticl tests for the study do not include the impct on this outcome in the SUB + CBRR versus PBTH comprison. Notes: Impct estimtes nd outcome mens re regression-djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Sources: Fmily Options Study 18-month followup survey; Progrm Usge dt Outcome F -7

287 Appendix F. Impct Estimtes for Pooled Comprisons Exhibit F-8. SUB + CBRR Versus PBTH: Impcts on Fmily Preservtion SUB + CBRR PBTH ITT Impct Effect Size N Men (SD) N Men (SD) Impct (SE) Current or recent seprtions of fmily members present t bseline Fmily hs t lest one child seprted (31.3) (35.8) 3.0 (2.7) 0.07 in pst 6 months (%) Fmily hs t lest one foster cre (14.8) (16.5) 0.1 (1.2) plcement in pst 6 months (%) Spouse/prtner seprted in pst (47.8) (44.8) 2.9 (6.9) months, of those with spouse/ prtner present t RA (%) Reunifiction of fmily members reported s seprted t bseline Fmily hs t lest one child reunified, (48.5) (47.3) 12.9 (9.1) 0.26 of those fmilies with t lest one child bsent t RA (%) Spouse/prtner reunified, of those with (46.8) (50.8) 25.3 (16.5) 0.44 spouse/prtner bsent t RA (%) CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. UC = usul cre. ITT = intention-to-tret. RA = rndom ssignment. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire UC group. Notes: Impct estimtes nd outcome mens re regression-djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Sources: Fmily Options Study 18-month followup survey Outcome Exhibit F-9. SUB + CBRR Versus PBTH: Impcts on Adult Well-Being Outcome SUB + CBRR N Men PBTH (SD) N Men ITT Impct (SD) Impct (SE) Effect Size Adult physicl helth Helth in pst 30 dys ws poor or fir (%) (45.4) (47.6) 5.1 (3.6) 0.10 Adult mentl helth Gol-oriented thinkingb (1.00) (1.02) 0.20 ** (0.09) 0.17 Psychologicl distressc (5.10) (5.82) 1.35 *** (0.41) 0.21 Adult trum symptoms Post-trumtic stress disorder (PTSD) (41.3) (42.6) 1.9 (3.3) 0.04 symptoms in pst 30 dys (%) Adult substnce use Alcohol dependence or drug bused (%) (32.5) (36.5) 3.7 (3.0) 0.09 Alcohol dependenced (%) (28.0) (32.6) 3.5 (2.8) 0.10 Drug bused (%) (22.6) (20.7) 1.3 (1.6) 0.05 Experience of intimte prtner violence Experienced intimte prtner violence in (26.9) (28.6) 1.9 (2.4) 0.05 pst 6 months (%) CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. UC = usul cre. ITT = intention-to-tret. RA = rndom ssignment. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire UC group. b Gol-oriented thinking is mesured with modified version of the Stte Hope Scle nd rnges from 1 to 6, with higher scores indicting higher levels of positive, gol-oriented thinking. c Psychologicl distress is mesured with the Kessler-6 (K6) scle nd rnges from 0 to 24, with higher scores indicting greter distress. d Alcohol dependence is mesured with the Rpid Alcohol Problems Screen-4 (RAPS4) nd drug buse is mesured with 6 items from the Drug Abuse Screening Test-10 (DAST-10). Notes: Impct estimtes nd outcome mens re regression-djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Sources: Fmily Options Study 18-month followup survey F -8

288 Appendix F. Impct Estimtes for Pooled Comprisons Exhibit F-10. SUB + CBRR Versus PBTH: Impcts on Child Well-Being Across Age Groups Outcome SUB + CBRR N Men PBTH (SD) N Men ITT Impct (SD) Impct (SE) Effect Size Child eduction Preschool or Hed Strt enrollment (%)b (49.1) (48.8) 4.9 (6.4) 0.08 School enrollment (%)c (32.4) (27.2) 5.1 ** (2.5) 0.15 Child cre or school bsences in pst (0.91) (0.92) 0.13 (0.09) 0.10 monthd Number of schools ttended since (0.84) (0.84) 0.08 (0.08) 0.07 rndom ssignmente Grde completion (not held bck) (%) (23.8) (27.4) 2.5 (2.6) 0.07 Positive childcre or school experiencesf (0.54) (0.56) 0.02 (0.04) 0.03 Positive childcre or school ttitudesg (1.02) (0.99) 0.02 (0.08) 0.02 School grdesh (0.96) (0.88) 0.03 (0.09) 0.02 Child cre or school conduct problemsi (0.44) (0.41) 0.05 (0.04) 0.09 Child physicl helth Poor or fir helth (%) (21.1) (22.4) 1.4 (2.2) 0.05 Well-child check-up in pst yer (%) (30.9) (28.0) 1.3 (2.6) 0.03 Child hs regulr source of helth cre (%) (22.8) (21.9) 1.2 (1.7) 0.04 Sleep problemsj (1.08) (1.09) 0.10 (0.08) 0.07 Child behviorl strengths nd chllenges Behvior problemsk (1.22) (1.14) 0.06 (0.10) 0.04 Prosocil behviorl (1.08) (1.11) 0.10 (0.09) 0.07 CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. UC = usul cre. ITT = intention-to-tret. RA = rndom ssignment. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire UC group. b Bse for preschool or Hed Strt enrollment is children ge 1.5 to 5 yers. c Bse for school enrollment is children ge 6 to 17. d Absences outcome is defined s 0 = No bsences in pst month, 1 = 1 to 2 bsences, 2 = 3 to 5 bsences, 3 = 6 or more bsences. e Number of schools outcome is topcoded t 4 or more schools. f Positive childcre or school experiences outcome is defined s -1 = mostly negtive experiences, 0 = Both positive nd negtive experiences, 1 = mostly positive experiences. g Positive childcre or school ttitudes outcome is prent report of how much child likes school nd rnges from 1 to 5, with higher vlues indicting greter like of school. h School grdes outcome is defined s 1 = mostly Ds or Fs, 2 = mostly Cs, 3 = mostly Bs, 4 = mostly As. i Childcre or school conduct problems outcome is defined s 0 = No conduct problems reported to prent, 1 = Prent contcted bout conduct problems or suspension or expulsion from school or childcre center. j Sleep problems outcome rnges from 1 to 5, with higher vlues indicting more frequent tiredness upon wking nd during the dy. k Behvior problems outcome is mesured s the stndrdized Totl Difficulties score from the Strengths nd Difficulties Questionnire (SDQ). l Prosocil behvior is mesured s the stndrdized Prosocil domin score from the Strengths nd Difficulties Questionnire (SDQ). Notes: Impct estimtes nd outcome mens re regression-djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Source: Fmily Options Study 18-month followup survey (prent report) F -9

289 Appendix F. Impct Estimtes for Pooled Comprisons Exhibit F-11. SUB + CBRR Versus PBTH: Impcts on Child Well-Being Developmentl Outcomes by Age Group Outcome SUB + CBRR N Men PBTH (SD) N Men ITT Impct (SD) Impct (SE) Effect Size Ages 1 yer to 3 yers, 6 months Met developmentl milestones (%)b (44.4) (45.6) 5.7 (7.3) 0.10 Low birth weight (%)c (29.6) (30.8) 6.3 (13.3) 0.17 Ages 3 yers, 6 months to 7 yers Verbl bilityd (1.02) (0.85) 0.05 (0.12) 0.04 Mth bilitye (0.94) (0.91) 0.02 (0.11) 0.02 Executive functioning (self-regultion)f (16.18) (16.14) 0.02 (1.27) Ages 8 to 17 yers Anxietyg (6.85) (7.85) 0.18 (1.07) 0.02 Fersh (13.70) (14.94) 0.09 (1.62) Substnce use (%)i (32.59) (32.19) 3.78 (3.88) 0.10 Gol-oriented thinkingj (4.48) (5.20) 0.70 (0.58) 0.10 School effort in pst monthk (0.82) (0.81) 0.13 (0.10) 0.12 Arrests or police involvement in pst (25.41) (31.80) 6.99 (4.65) months (%)l CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. UC = usul cre. ITT = intention-to-tret. RA = rndom ssignment. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire UC group. b Met developmentl milestones outcome is defined s scoring bove the typicl development cutoffs in ll domins of the ASQ-3. c Bse for low birthweight outcome (prent report) is children born since rndom ssignment who re t lest 1 yer old t followup. d Verbl bility outcome is the ntionlly stndrdized score from the WJIII Letter-Word Identifiction test. e Mth bility outcome is the ntionlly stndrdized score from the WJIII Applied Problems test. f Executive functioning outcome is the HTKS score nd rnges from 0 to 40, with higher scores indicting greter executive functioning. g Anxiety (child report) is mesured using the A-Trit scle from the Stte-Trit Anxiety Inventory for Children (STAIC). Scores rnge from 20 to 60, with higher scores indicting greter nxiety. h Fers outcome (child report) is the score from the Fers Scle nd rnges from 33 to 99, with higher scores indicting more fer. i Substnce use (child report) is mesured with 23 items from the Centers for Disese Control nd Prevention 2011 Youth Risk Behvior Survey. j Gol-oriented thinking (child report) is mesured with modified version of the Children's Hope Scle nd rnges from 6 to 30, with higher scores indicting higher levels of positive, gol-oriented thinking. k School effort outcome (child report) rnges from 1 to 4, with higher scores indicting greter effort during school dy nd on homework. l Arrest or police involvement in pst 6 months is from prent report. Notes: Impct estimtes nd outcome mens re regression-djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Sources: Fmily Options Study 18-month followup survey (prent report); Ages nd Stges Questionnire (ASQ-3); Woodcock-Johnson III Assessment (WJIII); Hed, Toes, Knees, nd Shoulders (HTKS) Assessment; Fmily Options Study 18-Month Child Survey (child report) F-10

290 Appendix F. Impct Estimtes for Pooled Comprisons Exhibit F-12. SUB + CBRR Versus PBTH: Impcts on Self-Sufficiency Outcome SUB + CBRR N Men PBTH (SD) N Men ITT Impct (SD) Impct (SE) Effect Size Employment sttus Work for py in week before (44.5) (48.6) 8.8 ** (3.7) 0.17 survey (%) Any work for py since RA (%) (49.8) (48.7) 6.5 * (3.6) 0.12 Months worked for py since RAb (7.7) (7.6) 0.0 (0.5) Hours of work per week t current (15.6) (17.3) 2.3 * (1.3) 0.13 min jobc Income sources/mounts Annulized current ernings ($) 400 4,491 (8,774) 281 6,109 (9,744) 1,618 ** (722) 0.16 Totl fmily income ($) ,142 (7,731) ,749 (9,394) 607 (685) 0.07 Anyone in fmily hd ernings in pst (49.4) (50.1) 4.3 (3.8) 0.08 month (%) Anyone in fmily received TANF in (47.3) (45.6) 3.3 (3.5) 0.06 pst month (%) Anyone in fmily received SSDI in (24.4) (27.6) 1.7 (1.9) 0.06 pst month (%) Anyone in fmily received SSI in pst (33.6) (33.0) 1.5 (2.2) 0.04 month (%) Anyone in fmily received SNAP/Food (32.2) (35.2) 3.5 (2.7) 0.08 Stmps in pst month (%) Anyone in fmily received WIC in pst (47.7) (46.3) 3.9 (3.5) 0.07 month (%) Eduction nd trining Prticipted in 2+ weeks of ny (45.7) (41.7) 8.1 ** (3.6) 0.16 school or trining since RA (%) Number of weeks in school/trining (11.0) (8.0) 1.8 ** (0.7) 0.16 progrms since RA Prticipted in 2+ weeks of school (25.3) (24.1) 1.2 (1.9) 0.04 since RA (%) Prticipted in 2+ weeks of bsic (8.6) (14.2) 2.0 * (1.1) 0.14 eduction since RA (%) Prticipted in 2+ weeks of voctionl (23.1) (22.1) 0.2 (1.9) eduction since RA (%) Food security/hunger Household is food insecure (%) (44.0) (46.8) 7.8 ** (3.7) 0.14 Food insecurity scled (1.90) (2.01) 0.37 ** (0.16) 0.16 Economic stressors Economic stress sclee (0.47) (0.48) 0.07 * (0.04) 0.11 CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. UC = usul cre. ITT = intention-to-tret. RA = rndom ssignment. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire UC group. b Number of months worked for py includes prtil clendr months. c Hours of work per week includes those not currently working (i.e., those with 0 hours of work per week). d Food insecurity scle rnges from 0 to 6, with higher vlues indicting higher food insecurity. e Economic stress scle rnges from -1 to 1, with higher vlues indicting higher economic stress. Notes: Impct estimtes nd outcome mens re regression-djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Source: Fmily Options Study 18-month followup survey F-11

291 Appendix F. Impct Estimtes for Pooled Comprisons Exhibit F-13. SUB + PBTH Versus CBRR: Impcts on Housing Stbility SUB + PBTH CBRR ITT Impct Effect Size N Men (SD) N Men (SD) Impct (SE) Homelessness or doubled up during the followup period At lest 1 night homelessb or doubled up (45.7) (50.1) 19.6 *** (3.4) 0.35 (pst 6 months) or in shelter (pst c 12 months) (%) [Confirmtory] At lest 1 night homelessb or doubled up (42.9) (48.7) 14.7 *** (3.2) 0.26 in pst 6 months (%) At lest 1 night homelessb in pst (34.6) (42.5) 9.1 *** (2.8) months (%) At lest 1 night doubled up in pst (40.3) (45.0) 8.7 *** (3.0) months (%) Any sty in emergency shelter in months (36.6) (44.9) 9.7 *** (2.8) to 18 fter rndom ssignment (%) Number of dys homelessb or doubled (59.8) (74.4) 23.6 *** (4.9) 0.28 up in pst 6 months Number of dys homelessb in pst (38.8) (53.6) 10.9 *** (3.5) months Number of dys doubled up in pst (49.5) (62.2) 14.7 *** (4.1) months Housing independence Living in own house or prtment t (47.0) (48.7) 3.7 (3.2) 0.07 followup (%) Living in own house or prtment with (40.1) (49.4) 23.2 *** (3.1) 0.43 no housing ssistnce (%) Living in own house or prtment with (50.0) (39.9) 27.1 *** (3.0) 0.56 housing ssistnce (%) Number of plces lived Number of plces lived in pst 6 months (1.1) (1.0) 0.1 (0.1) 0.07 Housing qulity Persons per room (1.0) (1.3) 0.2 *** (0.1) (45.9) (46.0) 0.6 (3.2) Any use of trnsitionl housing during months 7 18 fter RA (%) CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. UC = usul cre. ITT = intention-to-tret. RA = rndom ssignment. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire UC group. b The definition of homeless in this report includes stys in emergency shelters nd plces not ment for humn hbittion. It excludes trnsitionl housing. c The seven confirmtory sttisticl tests for the study do not include the impct on this outcome in the SUB + PBTH versus CBRR comprison. Notes: Impct estimtes nd outcome mens re regression-djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Sources: Fmily Options Study 18-month followup survey; Progrm Usge dt Outcome F-12

292 Appendix F. Impct Estimtes for Pooled Comprisons Exhibit F-14. SUB + PBTH Versus CBRR: Impcts on Fmily Preservtion SUB + PBTH CBRR ITT Impct Effect Size N Men (SD) N Men (SD) Impct (SE) Current or recent seprtions of fmily members present t bseline Fmily hs t lest one child seprted (33.5) (33.9) 0.3 (2.4) in pst 6 months (%) Fmily hs t lest one foster cre (14.9) (15.7) 0.3 (1.2) plcement in pst 6 months (%) Spouse/prtner seprted in pst (46.8) (49.7) 13.8 ** (6.6) months, of those with spouse/ prtner present t RA (%) Reunifiction of fmily members reported s seprted t bseline Fmily hs t lest one child reunified, (47.1) (47.0) 4.1 (8.0) 0.08 of those fmilies with t lest one child bsent t RA (%) Spouse/prtner reunified, of those with (48.5) (45.2) 9.6 (13.6) 0.17 spouse/prtner bsent t RA (%) CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. UC = usul cre. ITT = intention-to-tret. RA = rndom ssignment. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire UC group. Notes: Impct estimtes nd outcome mens re regression-djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Sources: Fmily Options Study 18-month followup survey Outcome Exhibit F-15. SUB + PBTH Versus CBRR: Impcts on Adult Well-Being Outcome SUB + PBTH N Men CBRR (SD) N Men ITT Impct (SD) Impct (SE) Effect Size Adult physicl helth Helth in pst 30 dys ws poor or fir (%) (46.1) (44.5) 2.6 (2.9) 0.05 Adult mentl helth Gol-oriented thinkingb (1.06) (0.95) (0.07) Psychologicl distressc (5.69) (5.17) 0.20 (0.34) 0.03 Adult trum symptoms Post-trumtic stress disorder (PTSD) (41.2) (42.0) 3.4 (2.8) 0.07 symptoms in pst 30 dys (%) Adult substnce use Alcohol dependence or drug bused (%) (33.5) (33.5) 0.0 (2.2) Alcohol dependenced (%) (29.6) (29.8) 0.6 (1.9) 0.02 Drug bused (%) (20.8) (20.2) 0.4 (1.4) 0.02 Experience of intimte prtner violence Experienced intimte prtner violence in (25.4) (31.7) 5.8 *** (2.1) 0.16 pst 6 months (%) CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. UC = usul cre. ITT = intention-to-tret. RA = rndom ssignment. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire UC group. b Gol-oriented thinking is mesured with modified version of the Stte Hope Scle nd rnges from 1 to 6, with higher scores indicting higher levels of positive, gol-oriented thinking. c Psychologicl distress is mesured with the Kessler-6 (K6) scle nd rnges from 0 to 24, with higher scores indicting greter distress. d Alcohol dependence is mesured with the Rpid Alcohol Problems Screen-4 (RAPS4) nd drug buse is mesured with 6 items from the Drug Abuse Screening Test-10 (DAST-10). Notes: Impct estimtes nd outcome mens re regression-djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Sources: Fmily Options Study 18-month followup survey F-13

293 Appendix F. Impct Estimtes for Pooled Comprisons Exhibit F-16. SUB + PBTH Versus CBRR: Impcts on Child Well-Being Across Age Groups Outcome SUB + PBTH N Men CBRR (SD) N Men ITT Impct (SD) Impct (SE) Effect Size Child eduction Preschool or Hed Strt enrollment (%)b (48.5) (48.8) 5.3 (4.9) 0.09 School enrollment (%)c (28.6) (30.5) 3.0 (2.2) 0.09 Child cre or school bsences in pst (0.94) (0.92) 0.09 (0.08) 0.07 monthd Number of schools ttended since (0.80) (0.88) 0.13 * (0.07) 0.11 rndom ssignmente Grde completion (not held bck) (%) (25.7) (24.1) 0.6 (2.1) 0.02 Positive childcre or school experiencesf (0.54) (0.55) 0.04 (0.04) 0.05 Positive childcre or school ttitudesg (1.01) (0.98) 0.06 (0.07) 0.05 School grdesh (0.89) (0.97) 0.14 * (0.08) 0.11 Child cre or school conduct problemsi (0.42) (0.42) (0.03) Child physicl helth Poor or fir helth (%) (21.7) (20.3) 0.7 (1.3) 0.03 Well-child check-up in pst yer (%) (30.0) (28.8) 0.0 (2.3) Child hs regulr source of helth cre (%) (24.4) (24.4) 0.1 (1.7) Sleep problemsj (1.11) (1.06) 0.02 (0.07) Child behviorl strengths nd chllenges Behvior problemsk (1.19) (1.29) 0.05 (0.09) 0.03 Prosocil behviorl (1.11) (1.08) 0.02 (0.08) CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. UC = usul cre. ITT = intention-to-tret. RA = rndom ssignment. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire UC group. b Bse for preschool or Hed Strt enrollment is children ge 1.5 to 5 yers. c Bse for school enrollment is children ge 6 to 17. d Absences outcome is defined s 0 = No bsences in pst month, 1 = 1 to 2 bsences, 2 = 3 to 5 bsences, 3 = 6 or more bsences. e Number of schools outcome is topcoded t 4 or more schools. f Positive childcre or school experiences outcome is defined s -1 = mostly negtive experiences, 0 = Both positive nd negtive experiences, 1 = mostly positive experiences. g Positive childcre or school ttitudes outcome is prent report of how much child likes school nd rnges from 1 to 5, with higher vlues indicting greter like of school. h School grdes outcome is defined s 1 = mostly Ds or Fs, 2 = mostly Cs, 3 = mostly Bs, 4 = mostly As. i Childcre or school conduct problems outcome is defined s 0 = No conduct problems reported to prent, 1 = Prent contcted bout conduct problems or suspension or expulsion from school or childcre center. j Sleep problems outcome rnges from 1 to 5, with higher vlues indicting more frequent tiredness upon wking nd during the dy. k Behvior problems outcome is mesured s the stndrdized Totl Difficulties score from the Strengths nd Difficulties Questionnire (SDQ). l Prosocil behvior is mesured s the stndrdized Prosocil domin score from the Strengths nd Difficulties Questionnire (SDQ). Notes: Impct estimtes nd outcome mens re regression-djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Source: Fmily Options Study 18-month followup survey (prent report) F-14

294 Appendix F. Impct Estimtes for Pooled Comprisons Exhibit F-17. SUB + PBTH Versus CBRR: Impcts on Child Well-Being Developmentl Outcomes by Age Group Outcome SUB + PBTH N Men CBRR (SD) N Men ITT Impct (SD) Impct (SE) Effect Size Ages 1 yer to 3 yers, 6 months Met developmentl milestones (%)b (43.7) (40.4) 11.4 ** (5.0) 0.21 Low birth weight (%)c (35.1) (22.4) 12.6 (9.5) 0.35 Ages 3 yers, 6 months to 7 yers Verbl bilityd (0.91) (1.00) 0.11 (0.12) 0.08 Mth bilitye (0.93) (0.84) 0.03 (0.10) 0.03 Executive functioning (self-regultion)f (15.88) (15.97) 0.48 (1.22) 0.02 Ages 8 to 17 yers Anxietyg (7.81) (7.57) 1.10 (0.96) 0.10 Fersh (14.43) (14.01) 0.45 (1.60) 0.02 Substnce use (%)i (26.05) (32.05) 5.37 * (2.86) 0.14 Gol-oriented thinkingj (4.81) (4.25) 1.38 *** (0.47) 0.20 School effort in pst monthk (0.81) (0.82) 0.20 ** (0.08) 0.19 Arrests or police involvement in pst (33.00) (31.66) 6.05 (4.74) months (%)l CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. UC = usul cre. ITT = intention-to-tret. RA = rndom ssignment. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire UC group. b Met developmentl milestones outcome is defined s scoring bove the typicl development cutoffs in ll domins of the ASQ-3. c Bse for low birthweight outcome (prent report) is children born since rndom ssignment who re t lest 1 yer old t followup. d Verbl bility outcome is the ntionlly stndrdized score from the WJIII Letter-Word Identifiction test. e Mth bility outcome is the ntionlly stndrdized score from the WJIII Applied Problems test. f Executive functioning outcome is the HTKS score nd rnges from 0 to 40, with higher scores indicting greter executive functioning. g Anxiety (child report) is mesured using the A-Trit scle from the Stte-Trit Anxiety Inventory for Children (STAIC). Scores rnge from 20 to 60, with higher scores indicting greter nxiety. h Fers outcome (child report) is the score from the Fers Scle nd rnges from 33 to 99, with higher scores indicting more fer. i Substnce use (child report) is mesured with 23 items from the Centers for Disese Control nd Prevention 2011 Youth Risk Behvior Survey. j Gol-oriented thinking (child report) is mesured with modified version of the Children's Hope Scle nd rnges from 6 to 30, with higher scores indicting higher levels of positive, gol-oriented thinking. k School effort outcome (child report) rnges from 1 to 4, with higher scores indicting greter effort during school dy nd on homework. l Arrest or police involvement in pst 6 months is from prent report. Notes: Impct estimtes nd outcome mens re regression-djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Sources: Fmily Options Study 18-month followup survey (prent report); Ages nd Stges Questionnire (ASQ-3); Woodcock-Johnson III Assessment (WJIII); Hed, Toes, Knees, nd Shoulders (HTKS) Assessment; Fmily Options Study 18-Month Child Survey (child report) F-15

295 Appendix F. Impct Estimtes for Pooled Comprisons Exhibit F-18. SUB + PBTH Versus CBRR: Impcts on Self-Sufficiency Outcome SUB + PBTH N Men CBRR (SD) N Men ITT Impct (SD) Impct (SE) Effect Size Employment sttus Work for py in week before (46.0) (47.4) 1.5 (3.1) 0.03 survey (%) Any work for py since RA (%) (49.7) (48.0) 6.0 ** (3.1) 0.11 Months worked for py since RAb (7.3) (8.0) 1.8 *** (0.5) 0.21 Hours of work per week t current (15.7) (16.5) 1.1 (1.1) 0.06 min jobc Income sources/mounts Annulized current ernings ($) 568 4,946 (8,872) 393 5,487 (9,479) 541 (605) 0.05 Totl fmily income ($) 560 9,213 (7,734) 379 9,918 (7,664) 705 (525) 0.08 Anyone in fmily hd ernings in pst (49.1) (49.9) 4.1 (3.2) 0.07 month (%) Anyone in fmily received TANF in (47.7) (46.5) 2.0 (3.0) 0.04 pst month (%) Anyone in fmily received SSDI in (27.6) (26.0) 0.5 (1.7) 0.02 pst month (%) Anyone in fmily received SSI in pst (33.8) (33.7) 1.4 (2.1) 0.04 month (%) Anyone in fmily received SNAP/Food (33.5) (34.3) 0.6 (2.3) Stmps in pst month (%) Anyone in fmily received WIC in pst (46.6) (47.1) 0.4 (3.0) month (%) Eduction nd trining Prticipted in 2+ weeks of ny (43.4) (44.7) 1.1 (3.0) 0.02 school or trining since RA (%) Number of weeks in school/trining (9.4) (10.6) 0.6 (0.7) 0.06 progrms since RA Prticipted in 2+ weeks of school (25.1) (21.9) 2.0 (1.7) 0.07 since RA (%) Prticipted in 2+ weeks of bsic (12.4) (14.1) 0.2 (0.9) eduction since RA (%) Prticipted in 2+ weeks of voctionl (24.5) (26.0) 0.7 (1.7) 0.02 eduction since RA (%) Food security/hunger Household is food insecure (%) (45.4) (44.6) 1.5 (3.2) 0.03 Food insecurity scled (2.00) (1.93) 0.05 (0.14) 0.02 Economic stressors Economic stress sclee (0.49) (0.49) 0.06 * (0.03) 0.10 CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. UC = usul cre. ITT = intention-to-tret. RA = rndom ssignment. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire UC group. b Number of months worked for py includes prtil clendr months. c Hours of work per week includes those not currently working (i.e., those with 0 hours of work per week). d Food insecurity scle rnges from 0 to 6, with higher vlues indicting higher food insecurity. e Economic stress scle rnges from -1 to 1, with higher vlues indicting higher economic stress. Notes: Impct estimtes nd outcome mens re regression-djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Source: Fmily Options Study 18-month followup survey F-16

296 Appendix F. Impct Estimtes for Pooled Comprisons Exhibit F-19. CBRR + PBTH Versus SUB: Impcts on Housing Stbility CBRR + PBTH SUB ITT Impct Effect Size N Men (SD) N Men (SD) Impct (SE) Homelessness or doubled up during the followup period At lest 1 night homelessb or doubled up (50.0) (41.2) 27.1 *** (3.1) 0.48 (pst 6 months) or in shelter (pst 12 months) (%) [Confirmtory]c At lest 1 night homelessb or doubled up (48.6) (37.4) 21.6 *** (2.9) 0.39 in pst 6 months (%) At lest 1 night homelessb in pst (42.1) (31.6) 11.4 *** (2.5) months (%) At lest 1 night doubled up in pst (45.3) (33.5) 17.0 *** (2.7) months (%) Any sty in emergency shelter in months (43.7) (34.2) 11.7 *** (2.4) to 18 fter rndom ssignment (%) Number of dys homelessb or doubled (73.4) (51.6) 29.1 *** (4.3) 0.34 up in pst 6 months Number of dys homelessb in pst (51.1) (36.7) 9.8 *** (3.0) months Number of dys doubled up in pst (62.7) (38.1) 22.8 *** (3.7) months Housing independence Living in own house or prtment t (48.8) (44.1) 12.4 *** (3.0) 0.22 followup (%) Living in own house or prtment with (49.2) (29.8) 31.5 *** (2.8) 0.58 no housing ssistnce (%) Living in own house or prtment with (40.3) (48.1) 44.1 *** (2.9) 0.91 housing ssistnce (%) Number of plces lived Number of plces lived in pst 6 months (1.0) (1.0) 0.3 *** (0.1) 0.20 Housing qulity Persons per room (1.2) (0.8) 0.3 *** (0.1) (47.9) (43.3) 11.0 *** (3.1) 0.20 Any use of trnsitionl housing during months 7 18 fter RA (%) CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. UC = usul cre. ITT = intention-to-tret. RA = rndom ssignment. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire UC group. b The definition of homeless in this report includes stys in emergency shelters nd plces not ment for humn hbittion. It excludes trnsitionl housing. c The seven confirmtory sttisticl tests for the study do not include the impct on this outcome in the CBRR + PBTH versus SUB comprison. Notes: Impct estimtes nd outcome mens re regression-djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Sources: Fmily Options Study 18-month followup survey; Progrm Usge dt Outcome F-17

297 Appendix F. Impct Estimtes for Pooled Comprisons Exhibit F-20. CBRR + PBTH Versus SUB: Impcts on Fmily Preservtion CBRR + PBTH SUB ITT Impct Effect Size N Men (SD) N Men (SD) Impct (SE) Current or recent seprtions of fmily members present t bseline Fmily hs t lest one child seprted (34.2) (31.1) 2.6 (2.2) 0.06 in pst 6 months (%) Fmily hs t lest one foster cre (16.1) (14.3) 0.4 (1.0) 0.02 plcement in pst 6 months (%) Spouse/prtner seprted in pst (49.0) (47.3) 12.1 * (6.2) months, of those with spouse/ prtner present t RA (%) Reunifiction of fmily members reported s seprted t bseline Fmily hs t lest one child reunified, (45.1) (47.8) 6.5 (8.5) 0.13 of those fmilies with t lest one child bsent t RA (%) Spouse/prtner reunified, of those with (48.6) (46.6) 1.2 (13.4) 0.02 spouse/prtner bsent t RA (%) CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. UC = usul cre. ITT = intention-to-tret. RA = rndom ssignment. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire UC group. Notes: Impct estimtes nd outcome mens re regression-djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Sources: Fmily Options Study 18-month followup survey Outcome Exhibit F-21. CBRR + PBTH Versus SUB: Impcts on Adult Well-Being Outcome CBRR + PBTH N Men (SD) SUB N Men ITT Impct (SD) Impct (SE) Effect Size Adult physicl helth Helth in pst 30 dys ws poor or fir (%) (45.8) (46.4) 1.1 (2.9) 0.02 Adult mentl helth Gol-oriented thinkingb (0.98) (1.07) 0.12 * (0.07) 0.10 Psychologicl distressc (5.57) (5.42) 0.87 *** (0.32) 0.13 Adult trum symptoms Post-trumtic stress disorder (PTSD) (42.8) (41.1) 5.1 * (2.7) 0.10 symptoms in pst 30 dys (%) Adult substnce use Alcohol dependence or drug bused (%) (35.1) (32.2) 2.1 (2.2) 0.05 Alcohol dependenced (%) (31.7) (28.3) 2.6 (2.0) 0.07 Drug bused (%) (19.2) (20.8) 1.3 (1.4) 0.05 Experience of intimte prtner violence Experienced intimte prtner violence in (30.7) (23.6) 5.5 *** (1.9) 0.15 pst 6 months (%) CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. UC = usul cre. ITT = intention-to-tret. RA = rndom ssignment. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire UC group. b Gol-oriented thinking is mesured with modified version of the Stte Hope Scle nd rnges from 1 to 6, with higher scores indicting higher levels of positive, gol-oriented thinking. c Psychologicl distress is mesured with the Kessler-6 (K6) scle nd rnges from 0 to 24, with higher scores indicting greter distress. d Alcohol dependence is mesured with the Rpid Alcohol Problems Screen-4 (RAPS4) nd drug buse is mesured with 6 items from the Drug Abuse Screening Test-10 (DAST-10). Notes: Impct estimtes nd outcome mens re regression-djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Sources: Fmily Options Study 18-month followup survey F-18

298 Appendix F. Impct Estimtes for Pooled Comprisons Exhibit F-22. CBRR + PBTH Versus SUB: Impcts on Child Well-Being Across Age Groups Outcome CBRR + PBTH N Men SUB (SD) N Men ITT Impct (SD) Impct (SE) Effect Size Child eduction Preschool or Hed Strt enrollment (%)b (48.3) (48.7) 1.2 (4.9) 0.02 School enrollment (%)c (30.2) (30.9) 0.2 (2.1) Child cre or school bsences in pst (0.92) (0.95) (0.08) monthd Number of schools ttended since (0.85) (0.74) 0.17 *** (0.06) 0.15 rndom ssignmente Grde completion (not held bck) (%) (21.4) (24.4) 0.9 (1.9) 0.02 Positive childcre or school experiencesf (0.53) (0.53) 0.04 (0.04) 0.06 Positive childcre or school ttitudesg (0.98) (1.02) 0.05 (0.07) 0.04 School grdesh (0.93) (0.91) 0.16 ** (0.07) 0.13 Child cre or school conduct problemsi (0.40) (0.42) 0.04 (0.03) 0.07 Child physicl helth Poor or fir helth (%) (22.8) (21.4) 0.3 (1.6) Well-child check-up in pst yer (%) (28.0) (31.4) 1.6 (2.2) 0.04 Child hs regulr source of helth cre (%) (23.6) (24.8) 0.9 (1.7) 0.03 Sleep problemsj (1.09) (1.12) 0.08 (0.07) 0.06 Child behviorl strengths nd chllenges Behvior problemsk (1.22) (1.20) 0.09 (0.09) 0.05 Prosocil behviorl (1.10) (1.09) 0.11 (0.08) 0.08 CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. UC = usul cre. ITT = intention-to-tret. RA = rndom ssignment. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire UC group. b Bse for preschool or Hed Strt enrollment is children ge 1.5 to 5 yers. c Bse for school enrollment is children ge 6 to 17. d Absences outcome is defined s 0 = No bsences in pst month, 1 = 1 to 2 bsences, 2 = 3 to 5 bsences, 3 = 6 or more bsences. e Number of schools outcome is topcoded t 4 or more schools. f Positive childcre or school experiences outcome is defined s -1 = mostly negtive experiences, 0 = Both positive nd negtive experiences, 1 = mostly positive experiences. g Positive childcre or school ttitudes outcome is prent report of how much child likes school nd rnges from 1 to 5, with higher vlues indicting greter like of school. h School grdes outcome is defined s 1 = mostly Ds or Fs, 2 = mostly Cs, 3 = mostly Bs, 4 = mostly As. i Childcre or school conduct problems outcome is defined s 0 = No conduct problems reported to prent, 1 = Prent contcted bout conduct problems or suspension or expulsion from school or childcre center. j Sleep problems outcome rnges from 1 to 5, with higher vlues indicting more frequent tiredness upon wking nd during the dy. k Behvior problems outcome is mesured s the stndrdized Totl Difficulties score from the Strengths nd Difficulties Questionnire (SDQ). l Prosocil behvior is mesured s the stndrdized Prosocil domin score from the Strengths nd Difficulties Questionnire (SDQ). Notes: Impct estimtes nd outcome mens re regression-djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Source: Fmily Options Study 18-month followup survey (prent report) F-19

299 Appendix F. Impct Estimtes for Pooled Comprisons Exhibit F-23. CBRR + PBTH Versus SUB: Impcts on Child Well-Being Developmentl Outcomes by Age Group Outcome CBRR + PBTH N Men SUB (SD) N Men ITT Impct (SD) Impct (SE) Effect Size Ages 1 yer to 3 yers, 6 months Met developmentl milestones (%)b (41.3) (44.1) 4.3 (5.4) 0.08 Low birth weight (%)c (29.6) (37.0) 9.5 (9.0) 0.26 Ages 3 yers, 6 months to 7 yers Verbl bilityd (0.90) (0.97) 0.05 (0.09) 0.04 Mth bilitye (0.87) (0.94) 0.04 (0.09) 0.03 Executive functioning (self-regultion)f (16.09) (16.13) (1.08) Ages 8 to 17 yers Anxietyg (7.57) (7.50) 0.99 (0.84) 0.09 Fersh (14.27) (14.09) 0.70 (1.50) 0.03 Substnce use (%)i (32.71) (25.76) 6.20 ** (2.79) 0.16 Gol-oriented thinkingj (4.77) (4.69) 0.90 * (0.50) 0.13 School effort in pst monthk (0.79) (0.80) 0.12 (0.09) 0.11 Arrests or police involvement in pst (29.81) (31.78) 4.15 (5.02) months (%)l CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. UC = usul cre. ITT = intention-to-tret. RA = rndom ssignment. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire UC group. b Met developmentl milestones outcome is defined s scoring bove the typicl development cutoffs in ll domins of the ASQ-3. c Bse for low birthweight outcome (prent report) is children born since rndom ssignment who re t lest 1 yer old t followup. d Verbl bility outcome is the ntionlly stndrdized score from the WJIII Letter-Word Identifiction test. e Mth bility outcome is the ntionlly stndrdized score from the WJIII Applied Problems test. f Executive functioning outcome is the HTKS score nd rnges from 0 to 40, with higher scores indicting greter executive functioning. g Anxiety (child report) is mesured using the A-Trit scle from the Stte-Trit Anxiety Inventory for Children (STAIC). Scores rnge from 20 to 60, with higher scores indicting greter nxiety. h Fers outcome (child report) is the score from the Fers Scle nd rnges from 33 to 99, with higher scores indicting more fer. i Substnce use (child report) is mesured with 23 items from the Centers for Disese Control nd Prevention 2011 Youth Risk Behvior Survey. j Gol-oriented thinking (child report) is mesured with modified version of the Children's Hope Scle nd rnges from 6 to 30, with higher scores indicting higher levels of positive, gol-oriented thinking. k School effort outcome (child report) rnges from 1 to 4, with higher scores indicting greter effort during school dy nd on homework. l Arrest or police involvement in pst 6 months is from prent report. Notes: Impct estimtes nd outcome mens re regression-djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Sources: Fmily Options Study 18-month followup survey (prent report); Ages nd Stges Questionnire (ASQ-3); Woodcock-Johnson III Assessment (WJIII); Hed, Toes, Knees, nd Shoulders (HTKS) Assessment; Fmily Options Study 18-Month Child Survey (child report) F-20

300 Appendix F. Impct Estimtes for Pooled Comprisons Exhibit F-24. CBRR + PBTH Versus SUB: Impcts on Self-Sufficiency Outcome CBRR + PBTH N Men SUB (SD) N Men ITT Impct (SD) Impct (SE) Effect Size Employment sttus Work for py in week before (47.4) (43.3) 7.4 ** (2.9) 0.14 survey (%) Any work for py since RA (%) (48.5) (50.0) 8.9 *** (3.0) 0.16 Months worked for py since RAb (7.6) (6.8) 1.6 *** (0.4) 0.19 Hours of work per week t current (16.3) (14.1) 2.7 *** (1.0) 0.15 min jobc Income sources/mounts Annulized current ernings ($) 489 5,580 (9,622) 485 3,933 (8,053) 1,647 *** (575) 0.16 Totl fmily income ($) ,051 (7,893) 477 8,644 (7,002) 1,407 *** (510) 0.16 Anyone in fmily hd ernings in pst (49.7) (47.9) 8.3 *** (3.0) 0.15 month (%) Anyone in fmily received TANF in (47.5) (48.8) 4.5 (3.0) 0.09 pst month (%) Anyone in fmily received SSDI in (25.7) (24.8) 0.1 (1.7) pst month (%) Anyone in fmily received SSI in pst (33.6) (33.5) 1.2 (1.8) 0.03 month (%) Anyone in fmily received SNAP/Food (34.9) (31.9) 4.0 * (2.3) 0.09 Stmps in pst month (%) Anyone in fmily received WIC in pst (47.0) (47.5) 2.5 (2.8) 0.05 month (%) Eduction nd trining Prticipted in 2+ weeks of ny (42.5) (44.5) 3.0 (2.9) 0.06 school or trining since RA (%) Number of weeks in school/trining (9.6) (10.7) 0.6 (0.6) 0.05 progrms since RA Prticipted in 2+ weeks of school (22.3) (26.1) 2.7 * (1.6) 0.09 since RA (%) Prticipted in 2+ weeks of bsic (13.4) (9.0) 1.2 (0.8) 0.09 eduction since RA (%) Prticipted in 2+ weeks of voctionl (26.0) (26.1) 1.0 (1.7) 0.04 eduction since RA (%) Food security/hunger Household is food insecure (%) (45.5) (44.4) 4.7 (2.9) 0.09 Food insecurity scled (1.99) (1.97) 0.24 * (0.13) 0.10 Economic stressors Economic stress sclee (0.48) (0.47) 0.11 *** (0.03) 0.19 CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. UC = usul cre. ITT = intention-to-tret. RA = rndom ssignment. SD = stndrd devition. SE = stndrd error. */**/*** Impct estimte is significntly different from 0 t the.10,.05, nd.01 levels, respectively, using two-tiled t-test. Effect size column shows stndrdized effect sizes, which were clculted by dividing impct by stndrd devition for the entire UC group. b Number of months worked for py includes prtil clendr months. c Hours of work per week includes those not currently working (i.e., those with 0 hours of work per week). d Food insecurity scle rnges from 0 to 6, with higher vlues indicting higher food insecurity. e Economic stress scle rnges from -1 to 1, with higher vlues indicting higher economic stress. Notes: Impct estimtes nd outcome mens re regression-djusted for bseline chrcteristics nd re weighted to djust for survey nonresponse. See Chpter 5 nd Appendix B for outcome definitions. Source: Fmily Options Study 18-month followup survey F-21

301 Appendix G. Intervention Costs Methodology, Sites, nd Progrms APPENDIX G. INTERVENTION COSTS METHODOLOGY, SITES, AND PROGRAMS This ppendix of the Fmily Options Study provides greter detil on the cost nlysis methodology nd the sites nd progrms in the cost nlysis. The pproch to collecting nd nlyzing cost dt ws guided by the intended use of the cost estimtes s context for interpreting the study s impct estimtes. The study tem clculted the perfmily monthly progrm cost for subset of progrms tht served the vst mjority of study fmilies who used ech intervention. Costs re clculted using primry dt collected by the study for trnsitionl housing, rpid re housing, nd emergency shelter progrms. Costs for vouchers nd public housing provided by the permnent housing subsidy (SUB) intervention re clculted from dministrtive dt. The study tem sought to cpture costs of ll progrm inputs consistently cross ech progrm type. The first four sections of this ppendix detil the process of determining the per-fmily monthly progrm cost for ech progrm in the cost study, which this ppendix clls the progrm-level per-fmily monthly progrm cost. Section G.1 reviews the progrms in the cost study. Section G.2 describes the elements of the per-fmily monthly progrm cost relevnt to the project-bsed trnsitionl housing (PBTH), community-bsed rpid re-housing (CBRR), nd emergency shelter progrms. Section G.3 reviews our pproch to clculting the per-fmily monthly progrm cost for the SUB intervention. Section G.4 reviews the ctul clcultion of the per-fmily monthly progrm cost. Section G.5 reviews how we verge these progrm-level per-fmily monthly progrm costs to rrive t the study s first cost concept of verge per-fmily monthly progrm cost for ech progrm type for the entire study nd for study sites. This section lso reviews how per-fmily monthly progrm costs re combined with study fmilies observed progrm usge to clculte the other three cost concepts: (1) progrm cost per sty during the followup period, (2) cost of ll progrm use during the followup period, nd (3) monthly cost of ll progrm use t the 18-month followup survey. G.1 Progrms in the Cost Study The study focuses on distinct set of housing nd shelter progrms, selected first to prticipte in the Fmily Options Study nd then gin for the cost nlysis bsed lrgely on the extent of prticipnts use of the intervention t individul progrms. The study tem collected cost dt t lrge subset of CBRR, PBTH, nd emergency shelter progrms prticipting in the study cross the 12 study sites. The study tem purposefully selected progrms from ech study site tht hd high number of referrls nd enrollments by study fmilies or, in the cse of emergency shelter, tht hd high number of fmilies recruited for the study. We ttempted to include ll study progrms in which t lest one study fmily enrolled nd received the intervention nd either (1) t lest five study fmilies were referred to the progrm or (2) the number of fmilies ws mong the top three for the progrm type in the site. As such, the estimtes were more likely to include lrger progrms tht were ble to mke spces vilble to study fmilies nd tht study fmilies were willing to ccept. The progrms selected for the cost nlysis represented more thn 85 percent of study fmilies who ccepted study referrl to CBRR nd PBTH, nd more thn 90 percent of fmilies ssigned to usul cre (UC) from shelter progrms. G -1

302 Appendix G. Intervention Costs Methodology, Sites, nd Progrms Reported cost estimtes re bsed on detiled reviews of 81 CBRR, PBTH, nd emergency shelter progrms providing housing or services to homeless fmilies cross the 12 Fmily Options Study sites, plus dministrtive dt (both t the household nd public housing gency [PHA] level) covering the 10 sites providing the SUB intervention. The dt collected from these progrms represent vluble contribution to the understnding of the cost of providing services to homeless fmilies using ech of the intervention types nd emergency shelter. Exhibit G-1 reports the number of progrms by site where we collected cost dt long with indicting which of the sites included the SUB intervention, for which comprehensive dministrtive dt were used to estimte costs. As described in Gubits et l. (2013), fmilies were enrolled in the study from September 2010 through Jnury Then, depending on which intervention fmily ws ssigned to, fmilies received housing nd services from the progrm ssocited with their intervention for either dys or weeks (shelter), months (CBRR nd PBTH), or yers (PBTH or SUB). We designed the cost nlysis to be representtive of the pproximte timefrme when fmilies were most likely to be enrolled in the progrms. The study tem collected cost dt for the full-yer period tht best ligned with the time period in which most fmilies were rndomly ssigned to the progrm nd overlpped with the progrm s finncil recordkeeping. Most often (48 progrms) this period ws fiscl yer We djusted ll costs using locl consumer price index mesures of infltion so tht ll estimtes re reported in 2013 dollrs. Exhibit G-1. Site Number of Progrms From Which Cost Dt Ws Collected nd Presence of SUB Intervention by Study Site Number of CBRR Progrms Number of PBTH Progrms Number of ES Progrms SUB Site Almed County Yes Atlnt 1* 1 2 No Bltimore No Boston 1 5 Yes Connecticut* Yes Denver Yes Honolulu Yes Knss City 1* 3 3 Yes Louisville Yes Minnepolis 1 1 Yes Phoenix Yes Slt Lke City Yes Totl CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. SUB = permnent housing subsidy. ES = emergency shelter. RA = rndom ssignment. * This study site comprises four CoCs in the New Hven/Bridgeport, Connecticut re. Pooled dt from multiple rpid re-housing progrms were reported for Atlnt nd Knss City. PBTH ws not offered in Boston, nd enrollment in PBTH in Minnepolis did not support cost-dt collection. SUB ws not offered in Atlnt nd Bltimore. Source: Fmily Options Study cost dt G.2 Elements of Progrm-Level per-fmily Monthly Progrm Cost The methodology ws designed to collect comprehensive progrm costs. Both to ensure tht the study tem collected ll costs ssocited with proving progrm services nd to llow for nlysis compring progrm structure, the study tem collected costs in five high-level cost ctegories: (1) overhed, (2) rentl ssistnce, (3) fcility opertions, (4) supportive services, nd (5) cpitl costs. Exhibit G-2 shows the cost ctegories nd how typicl line items were sorted within ctegories. G -2

303 Appendix G. Intervention Costs Methodology, Sites, nd Progrms Exhibit G-2. Cost Dt Collection Ctegories nd Associted Item Prompts Agency Overhed Administrtive stff Slries Fringe Other stffrelted costs Advertising Audit Accounting Legl Mngement fee Lesing & utilities (prorted gency) Miscellneous office expenses Pyroll txes Indirect or llocted costs Supportive Services Housing or Shelter Progrm/Assistnce Expenses Fcility Operting Costs (project-bsed progrms) Supportive services stff Slries Fringe Other stffrelted costs Stff trnsporttion Progrm supplies Activities Housing plcement Employment serch Direct support Food Clothing Furniture Trnsporttion Eduction Csh Other Rentl Assistnce for Client Housing Lese of client housing Electric Gs Wter/sewer Other Fmily contributions Housing opertions stff Slries Fringe Other Trsh removl Lndscping Exterminting Pinting nd decorting Property Insurnce Rel estte txes Repirs Supplies Furnishing equipment Other Property Vlue or Lese Expenses Mrket rte lese of client, progrm, nd dministrtive property Estimte of rentl rte of owned or donted property Fcility rent (or cpitl cost) for spce used for supportive services The study tem collected cost dt for CBRR, PBTH, nd emergency shelter progrms from the progrms directly. The study tem reviewed these ctegories nd ll progrm ctivities with key progrm opertions nd gency ccounting stff using stndrdized protocol nd dt collection tools tht were dpted from previous studies, including HUD s Cost of Homelessness Study (Spellmn et l., 2010). Our primry source of cost informtion for ll inputs except cpitl costs nd in-kind nd prtner costs ws udited expense sttements. These sttements were supplemented by progrm budgets, stffing lists, prtner commitment letters, nd progrm stff estimtes of lbor nd mteril costs of ny services not reflected in expense sttements. To further ensure ll progrm services were recognized s costs, interviews lso reviewed progrm clendrs nd cse mngement pproches for ech ge group of progrm residents. To clrify nd simplify comprisons cross intervention progrms for reporting our findings in Chpter 12, we collpsed the five cost-dt collection ctegories into two brod designtions: (1) housing or shelter nd (2) supportive services. 1. Housing or shelter refers to the rentl cost either observed or estimted of the spce used to provide housing or shelter nd progrm services nd lso ny mintennce or other fcility opertion costs (including durble items such s furnishings). This rentl cost is net of ny rent pyments mde by the fmily. 2. Supportive services costs refer to ny services other thn shelter or housing provided s n integrl prt of the progrm, including cse mngement, nd ny csh or in-kind ssistnce (for exmple, mels provided in emergency shelters). Additionl detil is reported in Chpter 12 for two other ctegories (1) dministrtive nd overhed costs nd (2) in-kind nd prtner costs becuse they provide informtion on typicl progrm structures. 1. Administrtive nd overhed costs include mngement slries; legl, ccounting, nd other professionl services; nd progrm support costs, such s insurnce premiums nd gency nd ssocition fees. Administrtive nd overhed costs re divided mong supportive services nd housing nd shelter costs ccording to the cost types reltive shre of totl costs so tht they re included in the two brod ctegories. G -3

304 Appendix G. Intervention Costs Methodology, Sites, nd Progrms 2. In-kind nd prtner costs include ny costs of housing or shelter or supportive services provided to fmilies becuse they prticipte in progrm. These costs re not provided by the progrm, nd, s result, re not included in progrm finncil sttements. Common exmples include onsite helth or mentl helth providers funded by n outside gency, community volunteers providing vriety of services, nd consumer goods donted to progrm clients. The importnce of these costs vries widely from progrm to progrm. When present, they typiclly re prt of the cost of supportive services provided by progrm. In some cses, however, housing or shelter costs include the costs of lbor, such s hndymn services, or of fcilities used regulrly for progrm ctivities tht were provided in kind. In other cses, ccounting, legl, or dministrtive services were provided in kind or by prtners. In ech cse, the study tem pportioned the cost to the pproprite ctegory. The next section provides dditionl detil regrding the ssessment of in-kind nd prtner costs nd cpitl costs nd detil bout our tretment of prticipnt contributions. G.2.1 In-Kind nd Prtner Costs We crefully reviewed ll services tht were provided to fmilies becuse they were progrm clients to ensure tht we were cpturing ll inputs consistently cross progrms nd cross interventions, whether the ssistnce ws provided directly by the progrm or in kind by prtner or volunteer. The nlysis, however, does not include the costs of every socil service ccessible to or encountered by progrm prticipnts. The study tem did not gther costs for services provided or mde vilble by virtue of fmilies housing or socioeconomic sttus lone, regrdless of their enrollment in prticulr progrm. Rther, costs were included for housing or shelter, goods, nd services provided to fmilies explicitly becuse they were clients of the study progrm. The nlysis included only the prtner services nd in-kind ssistnce tht resulted from dedicted reltionship with the progrm nd were ccessed by t lest 20 percent of eligible fmilies (s determined by interviews with progrm stff). In ddition, the nlysis included ny services tht progrms cited s criticl to the mission or core pproch, regrdless of prticiption rtes. Exhibit G-3 provides exmples of services tht our pproch did nd did not include when identifying prtner nd in-kind costs. When inputs were provided by n externl prtner or through in kind dontions, the study tem estimted the vlue of the progrm input using the following resources, s vilble Progrm documenttion (such s n udit estimte of in-kind services vlue). Costs of similr service or item pid for by the progrm or by nother progrm t the site. Externl documenttion, such s Bureu of Lbor Sttistics dt on locl wges or publicly listed costs for the goods or services. Exmples of prtner nd in-kind services nd resources used to determine n ssocited cost re listed in Exhibit G-4. Exhibit G-3. Determining Inclusion of Externl Services Included On Site A helth clinic sends doctor nd nurse prctitioner to the progrm site to conduct weekly screenings nd checkups. Included Off Site A Hed Strt school reserves nd gurntees priority slots for ll ppropritely ged children from the progrm. Not Included On Site A volunteer orgniztion holds prenting clsses in common re, but less thn 20 percent of prents ttend. Not Included Off Site If client expresses interest, cse mngers refer her to n externl job trining progrm run by the locl workforce bord. G -4

305 Appendix G. Intervention Costs Methodology, Sites, nd Progrms Exhibit G-4. Vluing In-Kind Services Exmple Estimtion Approch Prtnering orgniztion sends substnce-buse counselor to progrm to hold weekly meetings with residents. Progrm hs n equivlent socil worker with substncebuse speciliztion on stff. This hourly wge rte is pplied to level of effort by prtner stff. Helth clinic sends tem of nurse prctitioners monthly to offer bsic preventtive helth cre nd checkups. Clinic sends leverge letter detiling the estimted costs of the nurses time nd medicl supplies. Locl church contributes move-in pcket for ech fmily, including furniture nd kitchenwre. Progrm estimtes the vlue of ech pcket t $500 per client. YMCA nerby holds five slots in its summer cmp for children stying in progrm s shelter. YMCA website lists costs for summer cmp ctivities. Are nonprofit orgniztion sends dentist to conduct exms for ll prents nd children once per yer. Bureu of Lbor Sttistics medin wge for dentist in tht stte is pplied to the level of effort by the dentist. YMCA = Young Men s Christin Assocition. The pproch used to vluing prtner nd in-kind stff is conservtive in tht it pplies estimted lbor rtes to prtner nd in-kind level of effort, but it does not pply n overhed cost multiplier to tht rte. The logic of this pproch is n ssumption tht the progrm s existing overhed infrstructure could bsorb ny dditionl dministrtive cost ssocited with directly funding this prtner or in-kind service were the service to be provided by the progrm. G.2.2 Cpitl Costs Accounting for cpitl costs ssocited with the physicl spce used to provide client housing nd progrm services is criticl to estblishing comprble costs within nd cross progrms. For some progrms this ccounting ws firly strightforwrd. A few progrms rent client housing units, progrm services, spce, or dministrtive office spce t mrket rtes, providing direct mrket-rte estimtes of the cost of fcilities used by the progrm. At other progrms, where client client-housing fcilities re unique nd different from typicl housing stock, estimting cost of ongoing occupncy of the spce used for the progrm ws more nunced. For exmple, congregte shelter in n otherwise industril neighborhood or in the bsement of downtown historic religious snctury is not comprble with ny spce tht would be rented t mrket rte. Other spce is provided t n in-kind discount; for exmple, office spce owned by municiplity is lesed t effectively no cost to the progrm. For cses in which mrket rtes were not pid by the orgniztion directly, the study tem took n opportunity-cost pproch to vluing spce used to provide housing nd services. We specificlly relied on the following resources to derive nnul cpitl cost estimtes, s vilble:31 Progrm s or donor s stted nnul vlue of the occupied spce. A 5-percent nnul cost of cpitl pplied to totl property vlue estimte; for exmple, property vlue estimte from recent pprisl or sles price-bsed insurnce estimte. HUD s published Fir Mrket Rent (FMR) for comprble number of units s the spce used by the progrm, djusted for mintennce costs. Redily vilble estimtes of mrket rent for similr nerby properties, djusted for mintennce costs; for exmple, similr units or properties listed publicly for rent next door or resonble rentl estimtes from similr neighboring properties provided by Zillow. 31 In developing this pproch, we used multiple methods for fcilities where dt were vilble. Estimtes for the sme property resulting from different methods were remrkbly similr (typiclly within 5 to 10 percent), which gives us confidence tht prticulr choice of property vlution method does not introduce mteril vrition into our cost estimtes. G -5

306 Appendix G. Intervention Costs Methodology, Sites, nd Progrms When rentl vlue ws imputed, either using comprble properties or locl FMR, n estimte of net rent the opportunity cost of interest ws imputed s 55 percent of the gross rent. This defltion ws bsed on n nlysis of dt from the Residentil Housing Finnce Survey (RHFS; HUD, 2014b). Our nlysis of the RHFS indicted n verge 45-percent expense rtio for mrket-rte rentl properties with between 5 nd 49 units the relevnt size properties for the size of buildings used by progrms in our nlysis. G.2.3 Prticipnt Contributions Of the 45 shelters in the cost nlysis, 8 required clients to py some set mount or percent of income s rent. Of the 24 trnsitionl housing progrms, 21 required client rentl pyments.32 This nlysis reports progrm costs net of these contributions. Annul progrm costs were reduced specificlly by the mount of ny client contributions before determining per-fmily verges, which conceptully is consistent with reporting the net costs to society of providing ssistnce. From prcticl stndpoint, this procedure llows for closer comprison cross progrm types. The ctul fmily rentl pyments for fmilies receiving rpid re-housing or subsidy ssistnce were unknown (For SUB, we know households expected rent contribution, but we do not observe ctul pyment to lndlords.) Reducing costs by tennt rent for PBTH nd emergency shelter progrms mkes the resulting estimtes more comprble with CBRR nd SUB costs. As such, costs reported in Chpter 12 were net of ny client contributions. The pttern of client contributions is similr cross sites, with the exception of Hwii PBTH, where fmily contributions verged slightly less thn $480 per fmily per month cross the five PBTH progrms. Among the 8 remining sites (16 progrms) in which PBTH fmilies py some rent, the verge monthly per-fmily rent is $144. For the 8 emergency shelters t which clients py some rent or progrm fee, the verge per-fmily per-month contribution is $105. All costs were reported net of ny fmily contributions. G.3 Costs of SUB Intervention The pproch to collecting cost dt for the SUB intervention differed from tht outlined previously for progrms providing CBRR, PBTH, nd emergency shelter. Becuse the SUB intervention consisted of permnent housing ssistnce provided by PHAs, mostly in the form of housing choice voucher, dministrtive dt sources vilble to HUD were n efficient source of cost informtion. The cost of housing ssistnce ws clculted directly from household-level dministrtive dt for ech study fmily ssigned to SUB who received SUB ssistnce. These HUD Public nd Indin Housing Informtion Center (PIC) dt contin move-in dtes nd housing ssistnce pyments (HAPs) mde on behlf of the fmily. To determine sitelevel nd overll verge HAP, we verged these household HAPs, weighting households by the number of dys of ssistnce received. These site-level HAPs were then inflted to ccount for dministrtive costs. Annul PHA finncil reports provided informtion on the totl dministrtive costs for the voucher progrm in ech of the 18 PHAs tht prticipted in the Fmily Options Study.33 G.4 Clculting Costs For CBRR, PBTH, nd emergency shelter progrms, the study tem collected or clculted costs for the entire progrm, typiclly for fiscl yer. Trnslting this totl progrm cost into per-fmily unit cost required determining the number of fmilies ssisted t time for PBTH nd emergency shelter progrms, nd, for CBRR progrms, the verge number of months of ssistnce provided to ech fmily. For PBTH nd emergency shelter progrms, 32 A few progrms return mndtory client svings to clients s csh on their exit. Becuse funds dispersed re exctly offset by previous client contributions, we do not include these trnsctions s progrm costs. 33 For fmilies receiving public housing in Honolulu from the SUB intervention, we imputed the cost of providing the public housing unit using the Honolulu FMR discounted to reflect typicl mintennce costs together with verge observed mintennce nd ctul dministrtive costs. G -6

307 Appendix G. Intervention Costs Methodology, Sites, nd Progrms P#" #!- e e e "!# m!ℎl "#! # c! = T!"! "#! # c! 1 N#! " f #!"! $! " d! e. The number of fmilies served t time is determined from progrm reports of the number of units nd occupncy rte for unit-bsed fcilities nd the number of beds typicl fmily size nd occupncy rte for congregte fcilities. For rpid re-housing progrms, the study tem divided totl progrm costs by the totl number of months of rent supported by the progrm for ll fmilies in the progrm during the period for which costs were collected. In some cses, we received progrm dministrtive dt from which we clculted the number of months of ssistnce directly. For other providers, progrms provided us totl progrm expenditures, totl number of households ssisted, nd verge number of months of housing provided by the ssistnce. Exhibit G-5 reviews the cost clcultion for ech progrm type. This per-fmily monthly progrm cost for ech progrm in the cost study is used to clculte ech of our four cost concepts; (1) verge per-fmily monthly progrm cost, (2) progrm cost per sty during the followup period, (3) cost of ll progrm use during the followup period, nd (4) monthly progrm cost of ll progrm use t the 18 month followup survey. Exhibit G-5. Progrm Level Averge per Fmily Cost Clcultions Progrm Type Progrm Level Averge per Fmily Monthly Cost Clcultion ES (Totl nnul progrm cost)/12)/typicl number of fmilies ssisted t time CBRR (Totl nnul progrm costs)/totl fmily-months of rent subsidized with funds PBTH (Totl nnul progrm cost)/12)/typicl number of fmilies served t time SUB Voucher Averge observed per fmily per month rentl subsidy for smple fmilies scled by PHA dministrtive cost rte SUB Public Housing Averge per fmily per month costs with imputed rentl vlue for smple fmilies scled by PHA dministrtive cost rte CBRR = community-bsed rpid re-housing. PBTH = project-bsed trnsitionl housing. ES = emergency shelter. PHA = public housing uthority. By fmily-month, we men rent is subsidy for one fmily for 1 month. G.4.1 Per-Fmily Monthly Progrm Cost Chpter 12 reports per-fmily monthly progrm cost verged cross ll progrms included in the cost nlysis. As noted previously, this group of progrms is subset of ll progrms in the Fmily Options Study. One im in selecting progrms ws to produce cost estimtes tht reflected the ssistnce tht study fmilies ctully received. To be consistent with this im to clculte verge per-fmily monthly progrm costs we weighted progrm-level per-fmily monthly progrm costs by the number of study fmilies who ctully enrolled in the progrm fter being referred to the progrm by the study for PBTH nd CBRR progrms. To be consistent with this pproch for emergency shelter progrms, we weighted progrm-level costs by the number of fmilies ssigned to the UC group t ech shelter in the cost study. SUB costs were verged directly from fmily-level dt, weighted by the number of dys fmily received ssistnce. G.4.2 Progrm Cost per Sty During the Followup Period Progrm costs per sty during the followup period re per-fmily monthly progrm costs scled up by the durtion of ssistnce provided. The study uses 18-month followup survey dt to ssess the durtion of ssistnce the fmilies receive. Study dt do not connect this informtion on durtion of ssistnce to prticulr progrms, however, including those for which the study tem collected cost dt. To clculte progrm cost per sty during the followup, therefore, we first clculted verge durtion in ech progrm type by fmilies ssigned to the intervention ssocited with the progrm type for ech site. This verging used the sme survey nonresponse weights used to clculte the study impcts. These site-level verge durtions were multiplied by progrm-level costs to rrive t n estimted progrm-level progrm cost per sty during the followup period. These progrm-level estimtes re then verged cross ll progrms to rrive t verge progrm cost per sty during the followup G -7

308 Appendix G. Intervention Costs Methodology, Sites, nd Progrms period. Agin, the verge ws clculted weighting by the number of fmilies who enrolled in ech progrm (or were referred to UC for emergency shelter progrms). For SUB progrms, the verge durtion for ll fmilies ssigned to SUB cross the study or t individul sites (for site-level progrm cost-per-sty estimtes) ws multiplied by the relevnt clculted verge per-fmily monthly progrm cost. G.4.3 Cost of ll Progrm Use During the Followup Period The per-fmily monthly progrm cost is clculted from progrm-level dt points, nd progrm cost per sty during the followup period is clculted from progrm-level dt points multiplied by site-level verge durtion of ssistnce. By contrst, we estimte the cost of ll progrm use during the followup period from fmily-level dt points multiplied by site-level per-fmily monthly progrm cost estimtes. This estimtion is becuse verge costs of ll progrm use during the followup period re clculted for ech rm of the six pirwise comprisons in the study concept tht is bsed on fmilies ssigned to intervention rms s opposed to fmilies referred to prticulr progrms. This pproch requires two dditionl ssumptions. First, site-level per-fmily monthly progrm costs for ech of our four progrm types re used s cost estimtes for month of ssistnce t ny progrm of tht type. For exmple, ll trnsitionl housing progrms in site hve the sme per-fmily monthly progrm cost s the site-level verge PBTH cost we estimte using progrm-level costs reported in Chpter 12. Second, the study dt trck fmilies use of permnent supportive housing (PSH), public housing, nd project-bsed housing ssistnce tht ws not ssocited with the study or included in the cost nlysis. Under the ssumption tht they hve similr progrm nd cost structures, the estimtes reported in this section use site-level PBTH costs s proxy for the cost of PSH nd SUB costs s proxy for the costs of public housing nd project-bsed housing ssistnce. To clculte the verge cost of ll progrm use during the followup period for the fmilies in ech rm of ech pirwise comprison, we first multiplied ech fmily s observed durtion in ech progrm type with the site-level verge per-fmily monthly progrm cost of providing tht type of ssistnce. After summing ll progrm types fmily uses during the followup period, we verged this fmily-level cost of ll progrm use during the followup period over ll fmilies in ech tretment rm, using the sme nonresponse weights used in the impct nlysis. G.4.4 Cost of ll Progrm Use t the Time of the Followup Survey To clculte the cost of ll progrm use t the time of the followup survey, we verge the site-level verge perfmily monthly progrm cost for the progrm type in the site in which fmily is receiving ssistnce (if ny) t the time of the followup survey. As with our per-fmily monthly cost mesure, site-level verge per-fmily monthly progrm costs re clculted with weights for the number of study fmilies who ccepted n ssignment to the progrms. Associting site-level progrm type costs with ssistnce study fmilies re receiving t the time of the followup survey requires the sme two ssumptions described previously for cost of ll progrm use during the followup period. First, site-level per-fmily monthly progrm costs for ech of our four progrm types re used s cost estimtes for month of ssistnce t ny progrm of tht type. Second, the study dt trck fmilies use of PSH, public housing, nd project-bsed housing ssistnce tht ws not ssocited with the study or included in the cost nlysis. We clculte the cost of ll progrm use t the time of the followup survey by verging these site-level cost estimtes tht we ssocite with the ssistnce fmilies re receiving. These verges re clculted over fmilies in ech rm of ech impct comprison using the sme survey nonresponse used in the impct nlysis. G -8

Health insurance exchanges What to expect in 2014

Health insurance exchanges What to expect in 2014 Helth insurnce exchnges Wht to expect in 2014 33096CAEENABC 02/13 The bsics of exchnges As prt of the Affordble Cre Act (ACA or helth cre reform lw), strting in 2014 ALL Americns must hve minimum mount

More information

Health insurance marketplace What to expect in 2014

Health insurance marketplace What to expect in 2014 Helth insurnce mrketplce Wht to expect in 2014 33096VAEENBVA 06/13 The bsics of the mrketplce As prt of the Affordble Cre Act (ACA or helth cre reform lw), strting in 2014 ALL Americns must hve minimum

More information

Health insurance exchanges What to expect in 2014

Health insurance exchanges What to expect in 2014 Helth insurnce exchnges Wht to expect in 2014 33096CAEENABC 11/12 The bsics of exchnges As prt of the Affordble Cre Act (ACA or helth cre reform lw), strting in 2014 ALL Americns must hve minimum mount

More information

Small Businesses Decisions to Offer Health Insurance to Employees

Small Businesses Decisions to Offer Health Insurance to Employees Smll Businesses Decisions to Offer Helth Insurnce to Employees Ctherine McLughlin nd Adm Swinurn, June 2014 Employer-sponsored helth insurnce (ESI) is the dominnt source of coverge for nonelderly dults

More information

MAX. As an increasingly larger share of Medicaid enrollees MEDICAID POLICY BRIEF

MAX. As an increasingly larger share of Medicaid enrollees MEDICAID POLICY BRIEF MAX CENTERS FOR MEDICARE & MEDICAID SERVICES MEDICAID POLICY BRIEF Brief 14 December 2012 The Avilbility nd Usbility of Behviorl Helth Orgniztion Encounter Dt in MAX 2009 Jessic Nysenbum, Ellen Bouchery,

More information

Utilization of Smoking Cessation Benefits in Medicaid Managed Care, 2009-2013

Utilization of Smoking Cessation Benefits in Medicaid Managed Care, 2009-2013 Utiliztion of Smoking Cesstion Benefits in Medicid Mnged Cre, 2009-2013 Office of Qulity nd Ptient Sfety New York Stte Deprtment of Helth Jnury 2015 Introduction According to the New York Stte Tocco Control

More information

ClearPeaks Customer Care Guide. Business as Usual (BaU) Services Peace of mind for your BI Investment

ClearPeaks Customer Care Guide. Business as Usual (BaU) Services Peace of mind for your BI Investment ClerPeks Customer Cre Guide Business s Usul (BU) Services Pece of mind for your BI Investment ClerPeks Customer Cre Business s Usul Services Tble of Contents 1. Overview...3 Benefits of Choosing ClerPeks

More information

DlNBVRGH + Sickness Absence Monitoring Report. Executive of the Council. Purpose of report

DlNBVRGH + Sickness Absence Monitoring Report. Executive of the Council. Purpose of report DlNBVRGH + + THE CITY OF EDINBURGH COUNCIL Sickness Absence Monitoring Report Executive of the Council 8fh My 4 I.I...3 Purpose of report This report quntifies the mount of working time lost s result of

More information

Active & Retiree Plan: Trustees of the Milwaukee Roofers Health Fund Coverage Period: 06/01/2015-05/31/2016 Summary of Benefits and Coverage:

Active & Retiree Plan: Trustees of the Milwaukee Roofers Health Fund Coverage Period: 06/01/2015-05/31/2016 Summary of Benefits and Coverage: Summry of Benefits nd Coverge: Wht this Pln Covers & Wht it Costs Coverge for: Single & Fmily Pln Type: NPOS This is only summry. If you wnt more detil bout your coverge nd costs, you cn get the complete

More information

Enterprise Risk Management Software Buyer s Guide

Enterprise Risk Management Software Buyer s Guide Enterprise Risk Mngement Softwre Buyer s Guide 1. Wht is Enterprise Risk Mngement? 2. Gols of n ERM Progrm 3. Why Implement ERM 4. Steps to Implementing Successful ERM Progrm 5. Key Performnce Indictors

More information

An Undergraduate Curriculum Evaluation with the Analytic Hierarchy Process

An Undergraduate Curriculum Evaluation with the Analytic Hierarchy Process An Undergrdute Curriculum Evlution with the Anlytic Hierrchy Process Les Frir Jessic O. Mtson Jck E. Mtson Deprtment of Industril Engineering P.O. Box 870288 University of Albm Tuscloos, AL. 35487 Abstrct

More information

Lump-Sum Distributions at Job Change, p. 2

Lump-Sum Distributions at Job Change, p. 2 Jnury 2009 Vol. 30, No. 1 Lump-Sum Distributions t Job Chnge, p. 2 E X E C U T I V E S U M M A R Y Lump-Sum Distributions t Job Chnge GROWING NUMBER OF WORKERS FACED WITH ASSET DECISIONS AT JOB CHANGE:

More information

Anthem Blue Cross Life and Health Insurance Company University of Southern California Custom Premier PPO 800/20%/20%

Anthem Blue Cross Life and Health Insurance Company University of Southern California Custom Premier PPO 800/20%/20% Anthem Blue Cross Life nd Helth Insurnce Compny University of Southern Cliforni Custom Premier 800/20%/20% Summry of Benefits nd Coverge: Wht this Pln Covers & Wht it Costs Coverge Period: 01/01/2015-12/31/2015

More information

JaERM Software-as-a-Solution Package

JaERM Software-as-a-Solution Package JERM Softwre-s--Solution Pckge Enterprise Risk Mngement ( ERM ) Public listed compnies nd orgnistions providing finncil services re required by Monetry Authority of Singpore ( MAS ) nd/or Singpore Stock

More information

How To Get Low Wage Workers Covered By Insurance Through Their Employer

How To Get Low Wage Workers Covered By Insurance Through Their Employer Stephen H. Long M. Susn Mrquis Low-Wge Workers nd Helth Insurnce Coverge: Cn Policymkers Trget Them through Their Employers? Mny policy inititives to increse helth insurnce coverge would subsidize employers

More information

Humana Critical Illness/Cancer

Humana Critical Illness/Cancer Humn Criticl Illness/Cncer Criticl illness/cncer voluntry coverges py benefits however you wnt With our criticl illness nd cncer plns, you'll receive benefit fter serious illness or condition such s hert

More information

Policy Brief. Receipt of Public Benefits and Private Support among Low-income Households with Children after the Great Recession.

Policy Brief. Receipt of Public Benefits and Private Support among Low-income Households with Children after the Great Recession. Policy Brief #31, April 2012 The Ntionl Poverty Center s Policy Brief series summrizes key cdemic reserch findings, highlighting implictions for policy. The NPC encourges the dissemintion of this publiction

More information

2015 EDITION. AVMA Report on Veterinary Compensation

2015 EDITION. AVMA Report on Veterinary Compensation 2015 EDITION AVMA Report on Veterinry Compenstion AVMA Report on Veterinry Compenstion 2015 EDITION Copyright 2015 by the All rights reserved. ISBN-13: 978-1-882691-31-9 AVMA Report on Veterinry Compenstion

More information

A National Look at the High School Counseling Office

A National Look at the High School Counseling Office A Ntionl Look t the High School Counseling Office Wht Is It Doing nd Wht Role Cn It Ply in Fcilitting Students Pths to College? by Alexndri Wlton Rdford, Nicole Ifill, nd Terry Lew Introduction Between

More information

a GAO-03-568 GAO COLLEGE COMPLETION Additional Efforts Could Help Education with Its Completion Goals Report to Congressional Requesters

a GAO-03-568 GAO COLLEGE COMPLETION Additional Efforts Could Help Education with Its Completion Goals Report to Congressional Requesters GAO United Sttes Generl Accounting Office Report to Congressionl Requesters My 2003 COLLEGE COMPLETION Additionl Efforts Could Help Eduction with Its Completion Gols GAO-03-568 My 2003 COLLEGE COMPLETION

More information

Vendor Rating for Service Desk Selection

Vendor Rating for Service Desk Selection Vendor Presented By DATE Using the scores of 0, 1, 2, or 3, plese rte the vendor's presenttion on how well they demonstrted the functionl requirements in the res below. Also consider how efficient nd functionl

More information

Economics Letters 65 (1999) 9 15. macroeconomists. a b, Ruth A. Judson, Ann L. Owen. Received 11 December 1998; accepted 12 May 1999

Economics Letters 65 (1999) 9 15. macroeconomists. a b, Ruth A. Judson, Ann L. Owen. Received 11 December 1998; accepted 12 May 1999 Economics Letters 65 (1999) 9 15 Estimting dynmic pnel dt models: guide for q mcroeconomists b, * Ruth A. Judson, Ann L. Owen Federl Reserve Bord of Governors, 0th & C Sts., N.W. Wshington, D.C. 0551,

More information

Small Business Cloud Services

Small Business Cloud Services Smll Business Cloud Services Summry. We re thick in the midst of historic se-chnge in computing. Like the emergence of personl computers, grphicl user interfces, nd mobile devices, the cloud is lredy profoundly

More information

GAO HIGHER EDUCATION. Improved Tax Information Could Help Families Pay for College. Report to the Committee on Finance, U.S.

GAO HIGHER EDUCATION. Improved Tax Information Could Help Families Pay for College. Report to the Committee on Finance, U.S. GAO United Sttes Government Accountbility Office Report to the Committee on Finnce, U.S. Sente My 2012 HIGHER EDUCATION Improved Tx Informtion Could Help Fmilies Py for College GAO-12-560 My 2012 HIGHER

More information

San Mateo County ACCEL Adult-Education College and Career Educational Leadership AB 86 Adult Education Consortium Project Management Plan 24, 2014 -

San Mateo County ACCEL Adult-Education College and Career Educational Leadership AB 86 Adult Education Consortium Project Management Plan 24, 2014 - A Sn Mteo County ACCEL Adult-Eduction College nd Creer Eductionl Ledership AB 86 Adult Eduction Consortium Project Mngement Pln - Februry 24, 2014 - This project mngement pln presents ACCEL s process frmework

More information

STATE OF MONTANA Developomental Disabilities Program Comprehensive Evaluation Hi-Line Home Programs, Inc Adult Services

STATE OF MONTANA Developomental Disabilities Program Comprehensive Evaluation Hi-Line Home Programs, Inc Adult Services Dtes of Review: FY '09 Evluttor(s): S. Crpenter DESK REVIEW: Accredittion: Acredittion is no longer required by the stte contrct. Significnt Events from the Agency: Developomentl Disbilities Progrm Comprehensive

More information

How To Set Up A Network For Your Business

How To Set Up A Network For Your Business Why Network is n Essentil Productivity Tool for Any Smll Business TechAdvisory.org SME Reports sponsored by Effective technology is essentil for smll businesses looking to increse their productivity. Computer

More information

Small Business Networking

Small Business Networking Why network is n essentil productivity tool for ny smll business Effective technology is essentil for smll businesses looking to increse the productivity of their people nd processes. Introducing technology

More information

Contextualizing NSSE Effect Sizes: Empirical Analysis and Interpretation of Benchmark Comparisons

Contextualizing NSSE Effect Sizes: Empirical Analysis and Interpretation of Benchmark Comparisons Contextulizing NSSE Effect Sizes: Empiricl Anlysis nd Interprettion of Benchmrk Comprisons NSSE stff re frequently sked to help interpret effect sizes. Is.3 smll effect size? Is.5 relly lrge effect size?

More information

Recognition Scheme Forensic Science Content Within Educational Programmes

Recognition Scheme Forensic Science Content Within Educational Programmes Recognition Scheme Forensic Science Content Within Eductionl Progrmmes one Introduction The Chrtered Society of Forensic Sciences (CSoFS) hs been ccrediting the forensic content of full degree courses

More information

GAO IRS AUDIT RATES. Rate for Individual Taxpayers Has Declined But Effect on Compliance Is Unknown

GAO IRS AUDIT RATES. Rate for Individual Taxpayers Has Declined But Effect on Compliance Is Unknown GAO United Sttes Generl Accounting Office Report to the Chirmn, Subcommittee on Oversight, Committee on Wys nd Mens, House of Representtives April 2001 IRS AUDIT RATES Rte for Individul Txpyers Hs Declined

More information

Treatment Spring Late Summer Fall 0.10 5.56 3.85 0.61 6.97 3.01 1.91 3.01 2.13 2.99 5.33 2.50 1.06 3.53 6.10 Mean = 1.33 Mean = 4.88 Mean = 3.

Treatment Spring Late Summer Fall 0.10 5.56 3.85 0.61 6.97 3.01 1.91 3.01 2.13 2.99 5.33 2.50 1.06 3.53 6.10 Mean = 1.33 Mean = 4.88 Mean = 3. The nlysis of vrince (ANOVA) Although the t-test is one of the most commonly used sttisticl hypothesis tests, it hs limittions. The mjor limittion is tht the t-test cn be used to compre the mens of only

More information

Why is the NSW prison population falling?

Why is the NSW prison population falling? NSW Bureu of Crime Sttistics nd Reserch Bureu Brief Issue pper no. 80 September 2012 Why is the NSW prison popultion flling? Jcqueline Fitzgerld & Simon Corben 1 Aim: After stedily incresing for more thn

More information

VoIP for the Small Business

VoIP for the Small Business Reducing your telecommunictions costs VoIP (Voice over Internet Protocol) offers low cost lterntive to expensive trditionl phone services nd is rpidly becoming the communictions system of choice for smll

More information

Small Business Networking

Small Business Networking Why network is n essentil productivity tool for ny smll business Effective technology is essentil for smll businesses looking to increse the productivity of their people nd processes. Introducing technology

More information

July 2005, NCJ 209588 Substance Dependence, Abuse, and Treatment of Jail Inmates, 2002. Highlights. No dependence or abuse 53 47 32.

July 2005, NCJ 209588 Substance Dependence, Abuse, and Treatment of Jail Inmates, 2002. Highlights. No dependence or abuse 53 47 32. U.S. Deprtment of Justice Office of Justice Progrms Bureu of Justice Sttistics Specil Report By Jennifer C. Krerg nd Doris J. Jmes BJS Sttisticins In 2002 more thn two-thirds of jil inmtes were found to

More information

How To Network A Smll Business

How To Network A Smll Business Why network is n essentil productivity tool for ny smll business Effective technology is essentil for smll businesses looking to increse the productivity of their people nd processes. Introducing technology

More information

Voluntary Prekindergarten Parent Handbook

Voluntary Prekindergarten Parent Handbook Voluntry Prekindergrten Prent Hndbook Wht is Voluntry Prekindergrten (VPK)? Voluntry Prekindergrten (VPK) is legisltively mndted progrm designed to prepre every four-yer-old in Florid for kindergrten nd

More information

Test Management using Telelogic DOORS. Francisco López Telelogic DOORS Specialist

Test Management using Telelogic DOORS. Francisco López Telelogic DOORS Specialist Test Mngement using Telelogic DOORS Frncisco López Telelogic DOORS Specilist Introduction Telelogic solution for Requirements Mngement DOORS Requirements mngement nd trcebility pltform for complex systems

More information

Small Business Networking

Small Business Networking Why network is n essentil productivity tool for ny smll business Effective technology is essentil for smll businesses looking to increse the productivity of their people nd business. Introducing technology

More information

VoIP for the Small Business

VoIP for the Small Business Reducing your telecommunictions costs Reserch firm IDC 1 hs estimted tht VoIP system cn reduce telephony-relted expenses by 30%. Voice over Internet Protocol (VoIP) hs become vible solution for even the

More information

COMPUTER SECURITY CS 470. Catalog Description. Course Objectives. Course Materials

COMPUTER SECURITY CS 470. Catalog Description. Course Objectives. Course Materials COMPUTER SECURITY CS 470 Ctlog Description PREREQUISITE: CS 350. Study of network security rchitectures nd models, cryptogrphy, uthentiction nd uthoriztion protocols, secure ppliction nd systems development,

More information

How To Reduce Telecommunictions Costs

How To Reduce Telecommunictions Costs Reducing your telecommunictions costs Reserch firm IDC 1 hs estimted tht VoIP system cn reduce telephony-relted expenses by 30%. Voice over Internet Protocol (VoIP) hs become vible solution for even the

More information

VoIP for the Small Business

VoIP for the Small Business Reducing your telecommunictions costs TechAdvisory.org SME Reports sponsored by Cybernut Solutions provides outsourced IT support from welth of knowledgeble technicins nd system dministrtors certified

More information

How To Get A Free Phone Line From A Cell Phone To A Landline For A Business

How To Get A Free Phone Line From A Cell Phone To A Landline For A Business Reducing your telecommunictions costs Reserch firm IDC 1 hs estimted tht VoIP system cn reduce telephony-relted expenses by 30%. Voice over Internet Protocol (VoIP) hs become vible solution for even the

More information

VoIP for the Small Business

VoIP for the Small Business VoIP for the Smll Business Reducing your telecommunictions costs Reserch firm IDC 1 hs estimted tht VoIP system cn reduce telephony-relted expenses by 30%. Voice over Internet Protocol (VoIP) hs become

More information

PROFILES. Physical Education Profiles. Physical Education and Physical Activity Practices and Policies Among Secondary Schools at Select US Sites

PROFILES. Physical Education Profiles. Physical Education and Physical Activity Practices and Policies Among Secondary Schools at Select US Sites PROFILES 2012 Physicl Eduction Profiles Physicl Eduction nd Physicl Activity Prctices nd Policies Among Secondry Schools t Select US s Ntionl Center for Chronic Disese Prevention nd Helth Promotion Division

More information

persons withdrawing from addiction is given by summarizing over individuals with different ages and numbers of years of addiction remaining:

persons withdrawing from addiction is given by summarizing over individuals with different ages and numbers of years of addiction remaining: COST- BENEFIT ANALYSIS OF NARCOTIC ADDICTION TREATMENT PROGRAMS with Specil Reference to Age Irving Leveson,l New York City Plnning Commission Introduction Efforts to del with consequences of poverty,

More information

VoIP for the Small Business

VoIP for the Small Business Reducing your telecommunictions costs Reserch firm IDC 1 hs estimted tht VoIP system cn reduce telephony-relted expenses by 30%. Voice over Internet Protocol (VoIP) hs become vible solution for even the

More information

Corporate Compliance vs. Enterprise-Wide Risk Management

Corporate Compliance vs. Enterprise-Wide Risk Management Corporte Complince vs. Enterprise-Wide Risk Mngement Brent Sunders, Prtner (973) 236-4682 November 2002 Agend Corporte Complince Progrms? Wht is Enterprise-Wide Risk Mngement? Key Differences Why Will

More information

Unit 29: Inference for Two-Way Tables

Unit 29: Inference for Two-Way Tables Unit 29: Inference for Two-Wy Tbles Prerequisites Unit 13, Two-Wy Tbles is prerequisite for this unit. In ddition, students need some bckground in significnce tests, which ws introduced in Unit 25. Additionl

More information

Small Business Networking

Small Business Networking Why network is n essentil productivity tool for ny smll business Effective technology is essentil for smll businesses looking to increse the productivity of their people nd business. Introducing technology

More information

VoIP for the Small Business

VoIP for the Small Business Reducing your telecommunictions costs Reserch firm IDC 1 hs estimted tht VoIP system cn reduce telephony-relted expenses by 30%. Voice over Internet Protocol (VoIP) hs become vible solution for even the

More information

STATUS OF LAND-BASED WIND ENERGY DEVELOPMENT IN GERMANY

STATUS OF LAND-BASED WIND ENERGY DEVELOPMENT IN GERMANY Yer STATUS OF LAND-BASED WIND ENERGY Deutsche WindGurd GmbH - Oldenburger Strße 65-26316 Vrel - Germny +49 (4451)/9515 - [email protected] - www.windgurd.com Annul Added Cpcity [MW] Cumultive Cpcity [MW]

More information

Long Term Financial Planning

Long Term Financial Planning Long Term Finncil Plnning Overview of New Pltform City of Knss City, Missouri 2013 2018 Long term finncil plnning combines finncil forecsting with finncil strtegizing to identify future chllenges nd opportunities

More information

VoIP for the Small Business

VoIP for the Small Business VoIP for the Smll Business Reducing your telecommunictions costs Reserch firm IDC 1 hs estimted tht VoIP system cn reduce telephony-relted expenses by 30%. Voice over Internet Protocol (VoIP) hs become

More information

Techniques for Requirements Gathering and Definition. Kristian Persson Principal Product Specialist

Techniques for Requirements Gathering and Definition. Kristian Persson Principal Product Specialist Techniques for Requirements Gthering nd Definition Kristin Persson Principl Product Specilist Requirements Lifecycle Mngement Elicit nd define business/user requirements Vlidte requirements Anlyze requirements

More information

Portfolio approach to information technology security resource allocation decisions

Portfolio approach to information technology security resource allocation decisions Portfolio pproch to informtion technology security resource lloction decisions Shivrj Knungo Deprtment of Decision Sciences The George Wshington University Wshington DC 20052 [email protected] Abstrct This

More information

2001 Attachment Sequence No. 118

2001 Attachment Sequence No. 118 Form Deprtment of the Tresury Internl Revenue Service Importnt: Return of U.S. Persons With Respect to Certin Foreign Prtnerships Attch to your tx return. See seprte instructions. Informtion furnished

More information

Influence of Playing Experience and Coaching Education on Coaching Efficacy among Malaysian Youth Coaches

Influence of Playing Experience and Coaching Education on Coaching Efficacy among Malaysian Youth Coaches World Applied Sciences Journl 30 (Innovtion Chllenges in Multidiciplinry Reserch & Prctice): 414-419, 2014 ISSN 1818-4952 IDOSI Publictions, 2014 DOI: 10.5829/idosi.wsj.2014.30.icmrp.198 Influence of Plying

More information

VoIP for the Small Business

VoIP for the Small Business VoIP for the Smll Business Reducing your telecommunictions costs Reserch firm IDC 1 hs estimted tht VoIP system cn reduce telephony-relted expenses by 30%. Voice over Internet Protocol (VoIP) hs become

More information

Quick Reference Guide: One-time Account Update

Quick Reference Guide: One-time Account Update Quick Reference Guide: One-time Account Updte How to complete The Quick Reference Guide shows wht existing SingPss users need to do when logging in to the enhnced SingPss service for the first time. 1)

More information

Helicopter Theme and Variations

Helicopter Theme and Variations Helicopter Theme nd Vritions Or, Some Experimentl Designs Employing Pper Helicopters Some possible explntory vribles re: Who drops the helicopter The length of the rotor bldes The height from which the

More information

PROPERTY AND CASUALTY INSURANCE. Effects of the Nonadmitted and Reinsurance Reform Act of 2010

PROPERTY AND CASUALTY INSURANCE. Effects of the Nonadmitted and Reinsurance Reform Act of 2010 United Sttes Government Accountbility Office Report to Congressionl Committees Jnury 2014 PROPERTY AND CASUALTY INSURANCE Effects of the Nondmitted nd Reinsurnce Reform Act of 2010 GAO-14-136 Jnury 2014

More information

Introducing Kashef for Application Monitoring

Introducing Kashef for Application Monitoring WextWise 2010 Introducing Kshef for Appliction The Cse for Rel-time monitoring of dtcenter helth is criticl IT process serving vriety of needs. Avilbility requirements of 6 nd 7 nines of tody SOA oriented

More information

GAO HIGHER EDUCATION. Federal Science, Technology, Engineering, and Mathematics Programs and Related Trends

GAO HIGHER EDUCATION. Federal Science, Technology, Engineering, and Mathematics Programs and Related Trends GAO United Sttes Government Accountbility Office Report to the Chirmn, Committee on Rules, House of Representtives October 2005 HIGHER EDUCATION Federl Science, Technology, Engineering, nd Mthemtics Progrms

More information

Balanced Scorecard. Linking Strategy to Actions. KPMG Swiss Practice Benchmarking Congress, Bürgenstock May 28 th, 1997, Roger Jaquet

Balanced Scorecard. Linking Strategy to Actions. KPMG Swiss Practice Benchmarking Congress, Bürgenstock May 28 th, 1997, Roger Jaquet Blnced Scorecrd Linking Strtegy to Actions KPMG Swiss Prctice Benchmrking Congress, Bürgenstock My 28 th, 1997, Roger Jquet Contents ❶ Objectives 1 ❶ Wht s the historicl bckground of the BSC? 2-4 ❶ BSC

More information

Reasoning to Solve Equations and Inequalities

Reasoning to Solve Equations and Inequalities Lesson4 Resoning to Solve Equtions nd Inequlities In erlier work in this unit, you modeled situtions with severl vriles nd equtions. For exmple, suppose you were given usiness plns for concert showing

More information

2013 Flax Weed Control Trial

2013 Flax Weed Control Trial 2013 Flx Weed Control Tril Dr. Hether Drby, UVM Extension Agronomist Susn Monhn, Conner Burke, Eric Cummings, nd Hnnh Hrwood UVM Extension Crops nd Soils Technicins 802-524-6501 Visit us on the web: http://www.uvm.edu/extension/cropsoil

More information

Learner-oriented distance education supporting service system model and applied research

Learner-oriented distance education supporting service system model and applied research SHS Web of Conferences 24, 02001 (2016) DOI: 10.1051/ shsconf/20162402001 C Owned by the uthors, published by EDP Sciences, 2016 Lerner-oriented distnce eduction supporting service system model nd pplied

More information

In addition, the following elements form an integral part of the Agency strike prevention plan:

In addition, the following elements form an integral part of the Agency strike prevention plan: UNITED STTES DEPRTMENT OF GRICULTURE Wshington, DC 20250 Federl Grin Inspection Service FGIS Directive 4711.2 6/16/80 STRIKE PREVENTION ND STRIKE CONTINGENCY PLNS I PURPOSE This Instruction: Estlishes

More information

Data quality issues for accounting information systems implementation: Systems, stakeholders, and organizational factors

Data quality issues for accounting information systems implementation: Systems, stakeholders, and organizational factors Journl of Technology Reserch Dt qulity issues for ccounting informtion systems implementtion: Systems, stkeholders, nd orgniztionl fctors ABSTRACT Hongjing Xu Butler University Qulity informtion is one

More information

Numeracy across the Curriculum in Key Stages 3 and 4. Helpful advice and suggested resources from the Leicestershire Secondary Mathematics Team

Numeracy across the Curriculum in Key Stages 3 and 4. Helpful advice and suggested resources from the Leicestershire Secondary Mathematics Team Numercy cross the Curriculum in Key Stges 3 nd 4 Helpful dvice nd suggested resources from the Leicestershire Secondry Mthemtics Tem 1 Contents pge The development of whole school policy 3 A definition

More information

Impact of the National School Lunch Program on Fruit and Vegetable Selection in Northeastern Elementary Schoolchildren, 2012 2013

Impact of the National School Lunch Program on Fruit and Vegetable Selection in Northeastern Elementary Schoolchildren, 2012 2013 Brief Reports Impct of the Ntionl School Lunch Progrm on Fruit nd Vegetble Selection in Northestern Elementry Schoolchildren, 2012 2013 Srh A. Amin, MPH Bethny A. Yon, PhD Jennifer C. Tylor, MS b Rchel

More information

Factors Affecting Electronic Medical Record System Adoption in Small Korean Hospitals

Factors Affecting Electronic Medical Record System Adoption in Small Korean Hospitals Originl Article Helthc Inform Res. 2014 July;20(3):183-190. pissn 2093-3681 eissn 2093-369X Fctors Affecting Electronic Medicl Record System Adoption in Smll Koren Hospitls Young-Tek Prk, PhD 1, Jinhyung

More information

Performance Evaluation of Academic Libraries Implementation Model

Performance Evaluation of Academic Libraries Implementation Model Performnce Evlution of Acdemic Librries Implementtion Model Luiz Bptist Melo CIDEHUS UE nd Librries of the Fculty of Science, University of Porto (Applied Mthemtics nd Botnicl Deprtments) Ru do Cmpo Alegre,

More information

2. Transaction Cost Economics

2. Transaction Cost Economics 3 2. Trnsction Cost Economics Trnsctions Trnsctions Cn Cn Be Be Internl Internl or or Externl Externl n n Orgniztion Orgniztion Trnsctions Trnsctions occur occur whenever whenever good good or or service

More information

Pre-Approval Application

Pre-Approval Application Pre-Approvl Appliction In tody s rel estte mrket, Pre-Approved mortgge provides you the buyer with powerful tool in the home purchse process! Once you hve received your Pre-Approvl, you cn shop for your

More information

INJURY MANAGEMENT & REHABILITATION

INJURY MANAGEMENT & REHABILITATION OBJECTIVES To estblish systemtic pproch to injury mngement nd rehbilittion throughout the City immeditely following work relted illness, injury nd disbility. To estblish tht it is the responsibility of

More information

GAO DISASTER ASSISTANCE. USDA and SBA Could Do More to Help Aquaculture and Nursery Producers

GAO DISASTER ASSISTANCE. USDA and SBA Could Do More to Help Aquaculture and Nursery Producers GAO United Sttes Government Accountbility Office Report to the Chir, Committee on Smll Business nd Entrepreneurship, U.S. Sente September 2012 DISASTER ASSISTANCE USDA nd SBA Could Do More to Help Aquculture

More information

Econ 4721 Money and Banking Problem Set 2 Answer Key

Econ 4721 Money and Banking Problem Set 2 Answer Key Econ 472 Money nd Bnking Problem Set 2 Answer Key Problem (35 points) Consider n overlpping genertions model in which consumers live for two periods. The number of people born in ech genertion grows in

More information

A COMPARISON OF ALCOHOL SCREENING INSTRUMENTS AMONG UNDER-AGED DRINKERS TREATED IN EMERGENCY DEPARTMENTS

A COMPARISON OF ALCOHOL SCREENING INSTRUMENTS AMONG UNDER-AGED DRINKERS TREATED IN EMERGENCY DEPARTMENTS Alcohol & Alcoholism Vol. 37, No. 5, pp. 444 450, 2002 A COMPARISON OF ALCOHOL SCREENING INSTRUMENTS AMONG UNDER-AGED DRINKERS TREATED IN EMERGENCY DEPARTMENTS THOMAS M. KELLY 1 *, JOHN E. DONOVAN 1, JANET

More information

2011 Statistics on Social Work Education in the United States

2011 Statistics on Social Work Education in the United States 2011 Sttistics on Socil Work Eduction in the United Sttes Council on Socil Work Eduction 2011 Annul Sttistics on Socil Work Eduction in the United Sttes Council on Socil Work Eduction The Annul Survey

More information

S a l e s Ta x, U s e Ta x, I n c o m e Ta x W i t h h o l d i n g a n d M i c h i g a n B u s i n e s s Ta x E s t i m a t e s

S a l e s Ta x, U s e Ta x, I n c o m e Ta x W i t h h o l d i n g a n d M i c h i g a n B u s i n e s s Ta x E s t i m a t e s 78 (Rev. 8-08) S l e s T x, U s e T x, I n c o m e T x W i t h h o l d i n g n d M i c h i g n B u s i n e s s T x E s t i m t e s Forms nd Instructions Cll (517) 636-4730 or visit Tresury s Web site t

More information