WELFARE POLICIES AND DOMESTIC ABUSE AMONG SINGLE MOTHERS: EXPERIMENTAL EVIDENCE FROM MINNESOTA

Transcription

1 WELFARE POLICIES AND DOMESTIC ABUSE AMONG SINGLE MOTHERS: EXPERIMENTAL EVIDENCE FROM MINNESOTA Lisa A. Gennetian MDRC 16 East 34 th Street New York, New York Tel: Fax: Acknowledgements This paper was completed as part of the Next Generation Project, a collaboration among researchers at the Manpower Demonstration Research Corporation (MDRC) and several leading research institutions, funded by the David and Lucile Packard Foundation, William T. Grant Foundation, and John D. and Catherine T. MacArthur Foundation. Many thanks to Jared Smith who provided expert data analysis; Tara Cullen who provided research assistance; Virginia Knox, Andrew London and Ellen Scott for helpful discussions; and, more generally, the Minnesota Department of Human Services for its support in the evaluation of the pilot Minnesota Family Investment Program. Key words: domestic abuse, welfare reform, poverty

2 ABSTRACT This paper examines the effects on domestic abuse of a pilot welfare program that took place in urban and rural counties of Minnesota from 1994 to Like many other random assignment evaluations of welfare programs, this pilot program was not designed to explicitly test the effects of special provisions for victims of domestic abuse. Yet, in urban counties, the Minnesota program increased employment and reduced domestic abuse among single mother welfare recipients. Similar effects were not found in rural counties. Urban/rural differences were not a result of racial/ethnic composition but likely related to differences in prior marital experiences.

3 INTRODUCTION There is growing concern among the public and policymakers about the harmful repercussions of domestic abuse among low-income women, particularly single mothers. The proportion of women on welfare who report experiencing abuse during their lifetime 50 to 60% is double the reports of abuse among women overall, and nearly one-third of welfare women disclose recently or currently experiencing abuse (Lloyd, 2000; Tolman & Raphael, 2000). Domestic abuse has been associated with mental and health problems, homelessness, employment instability and a host of serious negative effects on children (Lawrence, 2002). Furthermore, domestic abuse presents a special challenge to single mothers in a welfare policy environment where they face strict employment requirements and time limits on welfare receipt. The Family Violence Option (FVO) of the 1996 welfare reform legislation allowed states to screen welfare parents for domestic violence, offer special services, and provide temporary exemptions from program requirements. As of 1999, over 30 states opted to implement the Family Violence Option (Raphael & Haennecke, 1999). At the same time, states have significantly altered their welfare programs, instituting a number of key policies including earned income disregards, mandatory employment services and time limited welfare. Although these latter policies are not specifically targeted to nor make special exceptions for victims of domestic abuse, the resulting effects of these policies on employment and income could have important implications on reported incidences of domestic abuse among the welfare population. Fortunately, a number of pilot state welfare programs were evaluated, via a random assignment research design, as a condition for receiving a welfare waiver prior to None of these pilot programs were designed to evaluate the specific effects of the Family Violence Option, but many of them tested a package of welfare policies that had the common goal of

4 increasing employment. Because of the experimental design of these studies, any differences in outcomes observed between families who were randomly assigned to a program group and families who were randomly assigned to a control group can be attributed to the effects of the program or policies being tested. In this paper, findings are presented from one such program a pilot welfare study that took place in urban and rural counties of Minnesota from 1994 to The Minnesota Family Investment Program (MFIP) offered financial incentives to work, via an enhanced earned income disregard, and required single-parent long-term recipients of welfare to participate in employment related services. Early findings show that MFIP increased employment and income among single mothers (Miller et al., 2000). I use data collected at a three-year follow-up survey to examine the effects of MFIP on domestic abuse and to ascertain whether or not MFIP s effects on economic well-being differed for those single mothers who had a history of abuse. Because it is important not to view MFIP s findings in isolation, I place MFIP s effects on domestic abuse in the context of findings from comparable studies of welfare programs. WELFARE POLICY CONTEXT AND THE PILOT MINNESOTA FAMILY INVESTMENT PROGRAM The 1996 federal welfare reform law introduced sweeping changes to the nation s system for supporting low-income families with children. The new law eliminated AFDC, which was funded as an open-ended entitlement, replacing it with Temporary Assistance for Needy Families (TANF), which provided block grants to states, introduced time limits on cash assistance, and imposed work requirements on recipients. Arguably, the most controversial of all of these changes was the elimination of the entitlement to receive federal cash assistance.

5 One of the least heralded changes in state welfare policies was a set of provisions designed to make work more financially rewarding. Most states increased the earnings disregard (the amount of earnings not counted as income in calculating the amount of a family s welfare benefit) so that families could keep more of their welfare dollars when they went to work. Such financial incentives increase both income and motivation to make the transition from welfare to employment. Time limits and earned income disregards are often matched with mandatory employment services, or requirements that recipients participate in employment-related activities as a condition of receiving their welfare benefits. Such policies have been in effect since the 1970s, although they apply to parents with ever younger children with each revision of welfare policy. The primary tool used to enforce participation mandates is sanctioning, whereby a recipient s welfare grant is reduced if she or he does not comply with program requirements. Today, virtually all states are using such mandates in their attempt to reduce welfare use and increase parents selfsufficiency. Prior to 1996 a number of states received federal waivers to implement experimental welfare programs. To assess the effects of these pilot programs, they were evaluated using a random assignment research design. At the time of application or re-determination of welfare benefits, welfare applicants or recipients were randomly assigned either to a program group with new benefits and services or to a control group. Random assignment ensured that any differences in outcomes between welfare families in the program group and welfare families in the control group were due to the experimental policies implemented. Many of the pilot programs tested some package of the key policy approaches previously described enhanced earned income disregards, mandatory work, and time limits. This article analyzes MFIP, one of the pilot welfare programs.

6 MFIP is particularly interesting because it significantly increased employment and income among single mothers and because it was implemented in urban and rural counties. MFIP therefore provides empirical evidence of the possible effects of increased employment and income on domestic abuse among low-income single mothers as well as how these effects may differ based on urban/rural residential status. MFIP aimed to encourage work, decrease dependence on public assistance, and reduce poverty by implementing two complementary components: financial incentives and mandatory participation in employment-focused services for long-term welfare recipients (for more details see Miller et al., 2000). The term financial incentives is used broadly to denote the three ways in which MFIP provided an incentive to work over the old AFDC system. First, MFIP included an enhanced earnings disregard 38% of earnings were disregarded when calculating the MFIP grant, compared with a more rapid benefit reduction rate under AFDC. Second, MFIP paid child care costs directly to providers, whereas under the AFDC system, recipients paid the costs and were reimbursed later. Third, Food Stamps benefits were given in the form of cash, rather than as coupons. The second component of MFIP was its participation mandates. Single parents who had received public assistance for 24 of the past 36 months were required to work at least 30 hours per week or participate in employment and training activities in order to continue receiving their full grants. MFIP also combined the benefits of AFDC, Family General Assistance (FGA), and Food Stamps into a single program. MFIP was implemented on a field trial basis in April 1994, in the three urban counties of Hennepin (Minneapolis), Anoka, and Dakota and the four rural counties of Mille Lacs, Morrison, Sherburne, and Todd. Between April 1994 and March 1996, over 14,000 parents who were receiving or applying for welfare were randomly assigned to either the MFIP or AFDC systems.

7 Single parents in urban counties were randomly assigned to one of three groups the MFIP group, the AFDC group, or an MFIP group (referred to as the MFIP Incentives Only group) that received MFIP s incentives but did not face any participation or work requirements. This three-group design was implemented to test the separate effects of the program s two components, its incentives versus its participation mandates. PRIOR RESEARCH: EMPLOYMENT, WELFARE USE, AND DOMESTIC ABUSE Although MFIP did not have specific provisions for victims of domestic abuse, it increased employment and income among some groups of single mothers. The effects on employment and income may have influenced experiences with domestic abuse by increasing economic self-sufficiency or self-esteem, by reducing contact with abusive partners or by exacerbating abuse as abusive partners may negatively react to enhanced independence that accompanies employment. Theory does not predict a straightforward relationship between employment, income and domestic abuse. And, in fact, empirical findings on the effects of employment on domestic abuse have been mixed and inconsistent. Some studies have found no effect of work transitions on domestic abuse (Tauchen & Witte, 1995; Fox et al., 2002); others have found employment increases violence but only if a woman s partner is unemployed (Macmillan and Gartner 1999); and still others argue that employment decreases violence (Farmer & Tiefenthaler, 1997). Rodriguez and colleagues (2000), one of the few studies to consider domestic violence as a function of employment for low-income women, found that employment increased incidences of domestic violence, but only if the woman was also receiving welfare. Other studies have documented how actions by an intimate partner can hinder a woman s employment (Lloyd, 1997; Riger, 2000). These actions may include turning off an alarm clock, inflicting bruises that mar a woman s appearance, undermining her self-confidence,

8 or harassing her at her place of employment (Sable et al., 1999; Riger, Ahrens & Blickenstaff, 2000; Tolman & Rosen, 2001). Evidence has been more mixed as to whether more general experiences of domestic violence affect women s employment. Browne, Solomon, and Bassuk (1999) and Moore and Selkowe (1999) present evidence suggesting that victims of domestic abuse were less likely to be employed or maintain steady employment. In contrast, Tolman and Rosen (2001), in their study of welfare mothers in Michigan, found no association between domestic violence and current or past employment. However, they did find that domestic violence negatively affects a woman s mental health, which in turn was negatively correlated with employment. Likewise, in a study of employment patterns among women residing in an impoverished neighborhood in Chicago, Lloyd and Taluc (1999) found no direct connection between experiencing domestic violence and current employment status, although they found that abused women were more likely to have been unemployed in the past. DETAILS OF DATA, SAMPLE AND EMPIRICAL ANALYSIS The key data source for this analysis is a subset of a survey that was conducted three years after families entered the MFIP study. The core part of the survey took approximately 30 minutes and collected detailed information about adult and family outcomes including participation in employment-related activities, financial hardship, and family composition. An additional component of the survey conducted in the home of the respondent, often referred to as the child section of the survey, took an additional 45 minutes to administer and contained a range of questions designed to measure child outcomes and, importantly, children s environments, including maternal experiences of domestic abuse. This particular component of the survey was targeted to a representative sample of single mothers with at least one child aged 2 to 9 who were

9 randomly assigned between April and October of 1994, representing an initial pool of 2,639 families. Eighty percent of those sample members were located and completed the follow-up survey (for a sample of 2,131 families). Analyses indicate that the survey sample is representative of the full sample and that there are no systematic differences in observed characteristics across the research groups in the survey sample (Gennetian & Miller, 2000). Of these families, 1,929 families met a few additional restrictions and constitute the final analysis sample for this paper. Problems with disclosure and disparities in measuring domestic abuse are well-noted among researchers (e.g. see Lawrence, 2002). To improve upon problems of non-response, single mothers in the MFIP study, in addition to receiving general assurances about the confidentiality of their responses, were asked to report their experiences with domestic abuse using Audio-CASI (Computer Assisted Self Interviewing) techniques. With this method, respondents use a laptop or PC, a sound card and headphones to listen to and answer questions in privacy. Subsequent analysis of responses to sensitive questions such as domestic violence using Audio-CASI as compared to other self-administered techniques show that Audio-CASI did improve upon the proportion of valid responses (Gallup-Black, 1999). The domestic abuse items in the MFIP follow-up survey were part of a larger section designed to capture life circumstances and barriers to work. Many of the original questions were used in a prior random assignment study (the National Evaluation of Welfare-to-Work Strategies, borrowed from the University of Michigan s Research on the Study of Domestic Violence, Questionnaire #3). Respondents were first asked a series of questions about whether or not those who were close to them made it difficult for them to enter or maintain employment. They were then asked more directly about their experiences with domestic abuse, i.e. did

10 anyone ever yell at you, control every move, threaten you with physical harm, force you into sexual activities or hit/slap/kick or otherwise physically harm you? If a respondent answered yes to any of these types of abuse, they were further asked for each type of abuse about who perpetrated the abuse and how long ago the most recent event happened. From these data, a number of domestic abuse outcomes were constructed for the analysis. More specifically, abused ever during the three year follow-up captures any incidence of reported abuse during the time of the MFIP study; physical abuse captures reports of being hit/slapped/kicked or otherwise physically harmed; nonphysical abuse captures reports of being yelled at or controlled or threatened with physical harm; and sexual abuse captures reports of being forced into any sexual activity. These latter three outcomes were measured the year prior to the survey interview. Alternative definitions of being physically abused that included threatened with physical harm did not alter the findings. In addition, an outcome was constructed to capture whether or not the perpetrator of abuse was an intimate partner. Finally, one measure was constructed to capture the intensity of abuse, i.e. whether or not more than one type of abuse was experienced. These outcome measures are not mutually exclusive. Note that it is not possible to identify the specific timing of past abuse with the information about type and perpetrator because of the way the survey questions were structured. In addition to the three-year follow-up survey targeted to single mother families, the MFIP evaluation also collected baseline demographic information and administrative records data covering earnings and public assistance receipt. Public assistance benefits records include monthly information on public assistance benefits (including MFIP, AFDC, Food Stamps, and Family General Assistance) provided to each member of the research sample. Unemployment Insurance (UI) earnings records give information about quarterly earnings for each sample

11 member, as reported by employers to the UI system and exclude earnings that are not covered by or not reported to the UI system in the state of Minnesota. Data from unemployment insurance records were used to create two measures of employment the average employment rate and average earnings during the first three years post random assignment. Data from benefit records were used to create comparably timed measures of welfare receipt. Income was computed as the combination of earnings, food stamps, and welfare benefits for the single parent as collected from unemployment insurance and public assistance benefit records. As previously described, because families were assigned at random to either the MFIP, MFIP Incentives Only or AFDC groups, there should have been no systematic difference among the groups when they entered the program. The difference in outcomes between any two groups is the effect, or impact. Program impacts were estimated by regressing, using Ordinary Least Squares, each of the measures on one (or, two when appropriate) indicator variable representing whether or not a family was randomly assigned to the MFIP group (or the MFIP Incentives Only group). In addition to the indicator variable to measure program impacts, the regressions included a number of baseline and pre-random assignment characteristics of the mother including county of residence; receipt history and current use of public assistance; number and ages of all children in the family; marital, education and employment history; race/ethnicity; and age. All of the impact estimates are regression-adjusted to increase the precision of the impact estimates and to control for any random baseline differences between the research groups. The coefficient on the experimental status of the mother captures the program impact(s). All impacts are tested for statistical significance, and only those impacts that are statistically significant using a two-tailed test at the 10 percent level are deemed program impacts. Findings in the tables include regression-adjusted means for the experimental and control groups, and the

12 differences in outcomes between these groups with an indication of whether or not the impact is statistically significant. Analyses are run separately for single mother long term welfare recipients, who had been on welfare for at least 24 of the 36 months prior to random assignment, and single mother recent applicants. The primary reason for this is because only the former group was immediately subject to mandatory participation in employment related services and thus experienced increased employment shortly after study entry. Selected baseline characteristics of the analysis sample of 784 single mother long term recipients and 689 single mother recent applicants are shown in Table 1. Table 1 about here In addition, MFIP s effects on economic outcomes and domestic abuse were estimated separately for the group of single mothers who lived in urban counties as compared to rural counties. Finally, impacts on economic outcomes were estimated separately for the group who entered the MFIP study and had a prior experience of domestic abuse. This subgroup was identified by using information at the three-year follow-up survey point about the timing of the most recent experiences with abuse and whether or not at least one of these most recent experiences occurred prior to entry into the MFIP study. These effects are compared with findings for the group of single mothers who entered the MFIP study and had no reports of a recent experience of abuse occurring prior to study entry. It was not possible to create a clean comparison group since we could not differentiate whether or not a mother who reported a very recent experience with abuse had or had not also experienced abuse over three years ago.

13 Differences in program impacts for single mothers according to their prior experience with domestic abuse were also tested for statistical significance. EFFECTS OF MFIP ON DOMESTIC ABUSE Table 2 presents MFIP s effects on economic and domestic abuse outcomes for single mother long term recipients and recent applicants. This table shows that reported rates of domestic abuse during a three-year follow-up period among control group members (under the column AFDC ), 56% for long term recipients and 50% for recent applicants, are quite consistent with what has been reported among welfare populations in prior research. More current reported rates of abuse are lower and vary, as expected, by the type of abuse. In the year prior to the survey interview, over one-third of long term recipients and recent applicants report having experienced nonphysical abuse and having experienced abuse by an intimate partner. Rates of reported sexual abuse are low at roughly 3 to 4%. Table 2 about here For single mother long term recipients, MFIP increased average quarterly employment by 15 percentage points (p<0.001), average quarterly welfare payments by $595 (p<0.001) and average quarterly income by $1,100 (p<0.001) during the three year follow-up period. MFIP increased earnings but the effect was not statistically significant. Recall that because MFIP had an enhanced earned income disregard, welfare payments will increase as employment and earnings increase up to a certain level of poverty. Although there appears to be a pattern of decreased reported incidents of abuse, MFIP had no statistically significant effect on any of the domestic abuse outcomes for single mother long term recipients. For recent applicants, MFIP

14 also increased employment and welfare payments, and, had more modest effects on income (that were not statistically significant). MFIP had no effect on domestic abuse outcomes for recent applicants. Tables 3 presents findings for single mother long term recipients by whether or not they lived in urban or rural counties, respectively, at the time of study entry. A comparison of MFIP s effects on domestic abuse outcomes across these tables suggests that the urban/rural distinction is important. More specifically, in urban counties, MFIP decreased reports of any abuse in the prior year by 8.5 percentage points or 18% (p<0.10); reports of any nonphysical abuse by 8.9 percentage points or 20% (p<0.05), and reports of any abuse by intimate partners by 8.4 percentage points or 21% (p<0.05). Furthermore, this table shows that it was primarily MFIP s financial incentives, i.e. its enhanced earned income disregard and cashing out of food stamps, that increased employment, income, and decreased domestic abuse. MFIP s effects on long term recipients in rural counties substantially differed: MFIP increased employment and income, but increased reported incidents of domestic abuse, including nonphysical abuse, having experienced more than one type of abuse, and abuse by intimate partners. It is also interesting to note that the rates of abuse reported by control group members are generally lower for long term recipients in rural counties as compared to long term recipients in urban counties. Although the rural long term recipient sample is relatively small (N=197 single mothers), the consistent pattern of statistically significant increases in domestic abuse outcomes is striking and convincing. These findings suggest that the effects of MFIP on domestic abuse over all single mother long term recipients were masking important differences according to urban/rural residential status. MFIP had no effect on domestic abuse outcomes among recent applicants who lived in urban counties or rural counties (not shown).

15 Table 3 about here Did MFIP succeed in increasing the employment of single mothers with prior histories of abuse? The first three columns of Table 4 present findings for the subgroup of single mother long term recipients who reported at least one most recent incidence of abuse occurring prior to random assignment. The findings show that MFIP generally did not increase employment among this group, regardless of their urban or rural residential status. In comparison, MFIP had large and quite consistent effects on employment, earnings and income among the group of single mother long term recipients who had no reported prior history of domestic abuse (shown in the last column of Table 4). Effects on employment for the no reported prior history of domestic abuse group were always larger and significantly different than MFIP s effects on employment for the group of single mother long term recipients who did have prior experiences with domestic abuse. MFIP s favorable effects on domestic abuse, i.e. reductions in domestic abuse, among single mother long term recipients in urban counties were also most pronounced in the group that had no prior history of abuse (not shown). These kinds of differences were not found in analyses of single mother recent applicants. Table 4 about here DISCUSSION MFIP, a pilot welfare program implemented in Minnesota in 1994, increased employment and income among single mother long term recipients and decreased reports of

16 domestic abuse for those who lived in urban counties at the start of the study. Analyses suggest that it was MFIP s financial incentives, rather than the added effects of MFIP s participation requirements, that produced the effects on decreased domestic abuse. These financial incentives contributed to increased employment as well as increased income. While several other random assignment studies of welfare programs produced significant effects on domestic abuse (Bloom et al,. 2000, Bloom et al., 2001, Beecroft et al., 2002, Fraker et al., 2002), one study of mandatory employment programs (called the National Evaluation of Welfare-to-Work Strategies) increased employment among single parents and decreased reports of domestic abuse, similar to what was found in MFIP (Hamilton et al., 2001). Nonexperimental analyses conducted with data from NEWWS and the urban single mother recipient sample in MFIP show that employment did indeed decrease abuse in both samples (Gibson et al., 2002). A similar direct relationship between income or earnings and domestic abuse was not found. Employment may have had such an effect because of enhanced self-esteem, enhanced personal resources, or because victims are physically away from their abusers. Analyses of the MFIP data also revealed that the beneficial effects of MFIP on employment and income were most pronounced for the group of single mother long-term recipients, whether urban or rural, who had no reported or identifiable prior history of abuse. This suggests that domestic abuse presents an important barrier to employment and selfsufficiency, even in the context of a program that offers relatively generous incentives to increase employment. That domestic abuse is an important barrier to employment has also been documented in prior research (e.g., see Danziger et al., 1999; Olson and Pavetti, 1996; Scott, London & Myers, 2002).

17 Although MFIP decreased reports of domestic abuse in urban counties, MFIP increased reported incidences of domestic abuse among single mother long term recipients in rural counties. Some possible reasons for these differences in effects by rural versus urban residence include less or more hindered access to social services, differences in the characteristics of the welfare population and, differences in the effects of MFIP on economic well-being. As was shown on Table 1, compared to urban recipients almost all rural recipients in the MFIP sample are White (91% versus 45%) and less than half were never married (44% as compared to 73%). Further analyses suggest that differences in racial composition were not influencing differences in effects on domestic abuse overall or by rural/urban status, but differences in prior marital status could be an important influence. MFIP s effects on decreased domestic abuse in urban counties were most pronounced among never married long term recipients, while MFIP s effects on increased abuse in rural counties were most pronounced among those long term recipients who had been previously married. In fact, among rural long term recipients rates of reported abuse are generally higher among those who had been previously married as compared to those who had never been married. Similar patterns were not found in analyses by race/ethnicity, although there is prior research that finds differences in types of abuse as reported by Black versus White welfare mothers (Brush, 2001). MFIP s effects on economic outcomes may also partially explain differences found in domestic abuse by urban/rural residential status. More specifically, while MFIP increased employment, earnings and income for urban long term recipients each of the three years of follow-up, MFIP s effects among rural long term recipients were short-lived, increasing employment, earnings and income only during the first year of follow-up (Gennetian, Redcross & Miller 2002).

18 Increased reported incidences of domestic abuse also emerged in other pilot welfare programs, although these effects were not isolated to rural areas (e.g., see Bloom et al., 2002; Fraker et al., 2002). Both of these programs time limited welfare benefits and this may have pressured women to increase employment because a publicly provided financial support system would not be available if they were not able to enter or maintain employment. Using longitudinal ethnographic data collected from welfare families living in high poverty neighborhoods in Cleveland, Scott, London and Myers (2002) found that welfare mothers rely on abusive partners for instrumental or financial support as they transition from welfare to work in the context of time limited welfare. The combined evidence from MFIP and other random assignment evaluations of welfare or work programs studies support in two ways the current structure of the Family Violence Option, allowing states to offer exemptions, extensions and special services as appropriate for victims of domestic abuse. First, some women who have had or who are currently experiencing abuse may benefit from requirements or financial incentives that lead to increased employment and economic independence and thus, universal exemptions or extensions of requirements or services for victims of abuse may not be appropriate. It is particularly reassuring that findings from MFIP (and NEWWS) on domestic abuse provide little support for the hypothesis that abuse will necessarily increase as a backlash to women s economic independence. Second, however, is that policies such as FVO may be particularly important for protecting the financial support network and providing targeted services for some welfare and low-income women who are domestic abuse victims to keep them from relying on abusive partners and perpetuating the cycle of abuse. Rates of reported current abuse ranging from 20 to 40% among the welfare samples in these random assignment studies are at levels that cannot be ignored by the policy and

19 research community. Future evaluations of similar kinds of welfare or employment programs should pay attention to this population, collecting information from women about their experiences with domestic abuse as well as from communities about services available to assist victims of abuse.

20 REFERENCES Beecroft, E., K. Cahill, & B. D. Goodson. (2002). The impacts of welfare reform on children: The indiana welfare reform evaluation. Boston, MA: Abt Associates, Inc. Bloom, D., S. Scrivener, C. Michalopoulos, P. Morris, R. Hendra, D. Adams-Ciardullo, & J. Walter. (2002). Jobs First: Final report on Connecticut s welfare reform initiative. New York: Manpower Demonstration Research Corporation. Bloom, D., J.J. Kemple, P. Morris, S. Scrivener, N. Verma & R. Hendra (2000). The Family Transition Program: Final report on Florida s initial time-limited welfare program. New York: Manpower Demonstration Research Corporation. Browne, A., A. Salomon & S. S. Bassuk (1999). The impact of recent partner violence on women's capacity to maintain work. Violence Against Women 5(4): Brush, L. (2001). Poverty, battering, race and welfare reform: Black-white differences in women's welfare-to-work transitions. Journal of Poverty 5(1): Danziger, S.K., M. Corcoran, S. Danziger, C. Heflin, A. Kalil, J. Levine, D. Rosen, K.S. Seefeldt, K. Siefer, & R.M. Tolman (1999). Barriers to the employment of welfare recipients. Joint Center for Research on Poverty Working Paper 90. Northwestern University.

21 Farmer, A. & J. Tiefenthaler (1997). An economic analysis of domestic violence. Review of Social Economy 55: Fox, G. L., M. L. Benson, A. DeMaris & J. Van Wyk (2002). Economic distress and intimate violence: Testing family stress and resources theories." Journal of Marriage and the Family 64: Fraker, T., C. Ross, R. Stapulonis, R. Olsen, R. M. Kovac, M. Dion, & A. Rangarajam (2002). The evaluation of welfare reform in Iowa: Final impact report. Washington, DC: Mathematica Policy Research, Inc. Gallup-Black, A. (1999). Using audio-casi to collect data on domestic violence. Mimeo. New York: Manpower Demonstration Research Corporation. Gennetian, L, C. Redcross & C. Miller (2002). Welfare reform in rural Minnesota: Experimental findings from the Minnesota Family Investment Program. In Weber, Duncan and Whitener (eds.) Rural Dimensions of Welfare Reform. Kalamazoo, MI: W.E. Upjohn Institute. Gennetian, L. & C. Miller. (2001). Reforming welfare and rewarding work: Final report on the Minnesota Family Investment Program, volume 2: Effects on children. New York: Manpower Demonstration Research Corporation. Gibson, C., K. Magnuson, L. Gennetian, G. Duncan, & P. England (2002). Does employment lead to decreased domestic abuse? Presented at Association of Public Policy and Management annual meetings, November Hamilton, G., Freedman, S., Gennetian, L., Michalopoulos, C., Walter, J., Adams-Ciardullo, D., Gassman-Pines, A., McGroder, S., Zaslow, M., Brooks, J., & Ahluwalia, S. (2001).

22 National Evaluation of Welfare-to-Work Strategies: How effective are different welfareto-work approaches? Five-year adult and child impacts for eleven programs. Executive Summary. Washington, D.C.: U.S. Department of Health and Human Services, Administration for Children and Families and Office of the Assistant Secretary for Planning and Evaluation, and U.S. Department of Education. Hetling, A. H. (2000). Addressing domestic violence as a barrier to self-sufficiency: The relationship of welfare receipt and spousal abuse. Journal of Public and International Affairs 11: Kalil, A. & S. Danziger (2000). How teen mothers are raring under welfare reform. Journal of Social Issues 56(4): Lawrence, S. (2002). Domestic violence and welfare policy: Research findings that can inform policies on marriage and child well-being. Research Forum on Children, Families and the New Federalism. New York: National Center for Children in Poverty. Lloyd, S. (1997). The effects of domestic violence on women's employment." Law and Policy 19: Lloyd, S. & N. Taluc (1999). The effects of male violence on female employment. Violence Against Women 5(5): Macmillan, R. & R. Gartner (1999). When she brings home the bacon: Labor-force participation and the risk of spousal violence against women. Journal of Marriage and the Family 61: Miller, C., V. Knox, L. Gennetian, M. Dodoo, J. Hunter-Manns & C. Redcross. (2001). Reforming welfare and rewarding work: Final report on the Minnesota Family

23 Investment Program, volume 1: Effects on adults. New York: Manpower Demonstration Research Corporation. Moore, T. & V. Selkowe (1999). Domestic violence victims in transition from welfare to work: Barriers to self-sufficiency and the W-2 response. Madison, WI, The Institute for Wisconsin's Future. Olson, K. & L. Pavetti (1996). Personal and family challenges to the successful transition from welfare to work. Washington, DC: The Urban Institute. Raphael, J. & S. Haennicke (1999). Keeping battered women safe through the welfare-to-work journey: How are we doing? Family Violence Option Report. Ann Arbor, MI, Poverty for Research and Training Center. Raphael, J. & S. Haennicke (1998). The family violence option: An early assessment. A Taylor Institute Report Submitted to the Federal Department of Health and Human Services, Washington, DC. Riger, S., C. Ahrens & A. Blickenstaff (2000). Measuring interferences with employment and education reported by women with abusive partners: Preliminary data. Violence Against Victims 15(2): Rodriguez, E., K. E. Lasch, P. Chandra & J. Lee (2001). Family violence, employment status, welfare benefits, and alcohol drinking in the United States: What is the relation? Journal of Epidemiological Community Health 55: Sable, M. R., M. K. Libbus, D. Huncke & K. Anger (1999). Domestic violence among AFDC recipients: Implications for welfare-to-work programs." Affilia 14: Scott, E., A. S. London & N. A. Myers (2002). Dangerous dependencies: The intersection of welfare reform and domestic violence. Gender & Society 16(6):

24 Tauchen, H. & A. D. Witte (1995). The dynamics of domestic violence. The American Economic Review 85(2): Tjaden, P. & N. Thoennes (1998). Prevalence, incidences, and consequences of violence against women: Findings from the national violence against women survey. Washington, D.C., National Institute of Justice and Centers for Disease and Prevention. Tolman, R. & J. Raphael (2000). A review of research on welfare and domestic violence. Journal of Social Issues 56(4): Tolman, R.& D. Rosen (2001). Domestic violence in the lives of women receiving welfare. Violence Against Women 7(2):

25 Table 1 Selected Characteristics of the MFIP Sample by Welfare Status at Random Assignment Long Term Recipients Recent Applicants Characteristic All Urban Rural All Urban Rural Average age (years) Race/Ethnicity (%) White, non-hispanic Black, non-hispanic Hispanic Native American/Alaskan Native Asian/Pacific Islander Marital status (%) Never married Married, living with spouse Married, living apart Separated Divorced Widowed Highest credential earned (%) GED certificate High school diploma Technical/2-year college degree year college degree or higher None of the above Any earnings in past 12 months (%) Total prior AFDC receipt a (%) None or more years Sample size SOURCES: Calculations using Minnesota Family Investment Program administrative and 36-month survey data. NOTES: Long term recipients were on welfare for at least 24 of the 36 month time period prior to random assignment. Sample is comprised of mothers with children aged 2 to 9 at study entry. a This refers to the total number of months an individual or her spouse has spent on AFDC at one or more periods of time as an adult. It does not include AFDC receipt under a parent's name.

26 Table 2 MFIP's Effects on Economic and Domestic Abuse Outcomes for Single Mother Long Term Recipients and Recent Applicants Long Term Recipients Recent Applicants MFIP AFDC Impact MFIP AFDC Impact Over a three-year follow-up period Average quarterly employment (%) *** *** Average quarterly earnings ($) 4,563 4, ,156 7, Average quarterly welfare payment ($) 7,223 6, *** 5,025 4, *** Average quarterly income ($) 11,786 10,658 1,128 *** 12,181 11, Any abuse (%) In the past year Any abuse (%) Any abuse by intimate partner (%) Any nonphysical abuse (%) Any physical abuse (%) Any sexual abuse (%) More than one type of abuse reported (%) Sample size SOURCES: Calculations using Minnesota Family Investment Program administrative and 36-month survey data. NOTES: Long term recipients were on welfare for at least 24 of the 36 month time period prior to random assignment. Data on domestic abuse were collected for mothers of children aged 2 to 9 at study entry and are based on maternal reports of her own experiences with domestic abuse. Sample sizes may slightly vary by outcome. Impact estimates are regression-adjusted to control for a number of baseline and pre-random assignment characteristics. A two-tailed t-test was used to assess the statistical significance of each difference in characteristics between the program and control groups. Statistical significance levels are indicated as * = p<0.10; ** = p<0.05; *** = p<0.01.

27 Table 3 Effects of MFIP and MFIP Incentives Only on Economics Outcomes and Domestic Abuse for Single Mother Long Term Recipients in Urban and Rural Counties Urban Counties Rural Counties Outcome MFIP vs. AFDC MFIP Incentives Only vs. AFDC Outcome MFIP vs. AFDC AFCD Impact Impact AFCD Impact Over a three-year follow-up period Average quarterly employment (%) *** 8.2 *** ** Average quarterly earnings ($) 4, , Average quarterly welfare payment ($) 6, ** 1,080 *** 5, Average quarterly income ($) 10,791 1,192 *** 1,103 *** 9,652 1,396 ** Any abuse (%) ** -9.9 ** In the past year Any abuse (%) * ** Any nonphysical abuse (%) ** -7.8 * ** Any physical abuse (%) Any sexual abuse (%) More than one type of abuse reported (%) ** Any abuse by intimate partner (%) ** -8.2 * ** Sample Size SOURCES: Calculations using Minnesota Family Investment Program administrative and 36-month survey data. See Notes in Table 2.

28 Table 4 MFIP's Effects on Economic Outcomes for Single Parent Long Term Recipients by Prior Experiences with Domestic Abuse At least one most recent experience No reported experience with with domestic abuse prior to domestic abuse prior to random assignment to random assignment Variation in Subgroup MFIP AFCD Impact MFIP AFCD Impact Impacts All Long Term Recipients Over a three-year follow-up period Average quarterly employment (%) *** Average quarterly earnings ($) 4,471 4, ,514 3, * Average quarterly welfare payment ($) 7,094 6, ** 7,278 6, Average quarterly income ($) 11,564 10, ,792 10,651 1,141 *** N = 276 N = 428 Long Term Recipients in Urban Counties Over a three-year follow-up period Average quarterly employment (%) *** Average quarterly earnings ($) 4,910 4, ,543 3, * Average quarterly welfare payment ($) 7,072 6, * 7,346 6, Average quarterly income ($) 11,983 10,850 1,133 * 11,888 10,697 1,192 *** N = 198 N = 326 Long Term Recipients in Rural Counties Over a three-year follow-up period Average quarterly employment (%) ** Average quarterly earnings ($) 3,164 5,339-2,175 * 4,889 3,170 1,718 Average quarterly welfare payment ($) 7,398 5,581 1,817 *** 6,256 5, Average quarterly income ($) 10,562 10, ,145 8,923 2,222 ** N = 78 N = 102 SOURCES: Calculations using Minnesota Family Investment Program administrative and 36-month survey data. See Notes in Table 2.

29 Lisa A. Gennetian, Ph.D. in Economics, is a Senior Research Associate at Manpower Demonstration Research Corporation (MDRC). At MDRC, her research focuses on the effects of welfare and employment policies on various aspects of family and child well-being.