1 Journal of Urban Economics 49, Ž doi: juec , available online at http: on The Importance of Lender Heterogeneity in Mortgage Lending David M. Harrison* School of Business Administration, University of Vermont, Burlington, Vermont Received June 3, 1999; revised July 17, 2000; published online November 16, 2000 Previous studies of the mortgage lending process, which fail to adequately account for inter-lender differences, suffer from a potentially serious omitted-variables bias. Using a sample of 23,094 home purchase mortgage applications from Pinellas County ŽSt. Petersburg. Florida, over the interval, I demonstrate that the inclusion of lender characteristics enhances the predictive power of accept reject decision models of mortgage origination. In addition, the inclusion of lender attributes is shown to substantially influence the coefficient estimates of additional model parameters of interest, such as the applicant race and neighborhood racial composition variables. Finally, this study provides unique and supportive evidence regarding the existence and persistence of information externalities in home purchase mortgage markets Academic Press Key Words: lender heterogeneity; mortgage lending; information externalities; redlining; mortgage discrimination. I. INTRODUCTION More than 5 years after the landmark Federal Reserve Bank of Boston study 1 of discrimination in home mortgage markets, debate over the existence, persistence, and economic significance of disparate lending outcomes continues to rage. While numerous studies 2 present evidence of an unequal distribution of credit outcomes across racial, gender, and income categories, many economists * I gratefully acknowledge the financial support of the University of Florida s Real Estate Research Center, as well as the time, knowledge, and expertise of Mark Flannery, Dean Gatzlaff, Joel Houston, Christopher James, David Ling, Wayne Passmore, Stephen Ross, Mike Ryngaert, and Greg Smersh as key ingredients in enabling successful and timely completion of this investigation. Any remaining errors are, as always, my own. 1 See Munnell et al. 29 for the original version, or Munnell et al. 30 for the published version. Ross and Yinger 34 provide a comprehensive review and reanalysis of this study and the criticisms leveled against it. 2 See Schafer 35, Schafer and Ladd 36, Bradbury et al. 8, Canner and Smith 14 and 15, and Canner et al. 13 for examples of studies documenting inequalities in mortgage market outcomes across racial and income classifications $35.00 Copyright 2000 by Academic Press All rights of reproduction in any form reserved.
2 286 DAVID M. HARRISON believe mortgage markets are relatively efficient, suggesting widespread discrimination should not persist. Furthermore, they argue the observed disparities documented by previous empirical work are the by-product of an omitted-variables bias, and any disparities which do indeed exist are the result of an economically efficient and rational decision making process on the part of lenders. 3 The purpose of this paper is to examine the economic significance of one such omitted variable: the identity of the mortgage lender. Specifically, I find the inclusion of lender attributes, such as the size, capitalization, profitability, and specialization of lending institutions, significantly enhances the predictive power of accept reject decision models of home purchase mortgage applications. Furthermore, the inclusion of these characteristics is shown to substantially reduce the importance of neighborhood racial composition in such models. Finally, the results provide unique and supportive evidence on the existence and economic significance of information externalities in residential mortgage markets. II. LENDER HETEROGENEITY Literally hundreds of investigations have been undertaken to evaluate the equality and distribution of mortgage lending outcomes across racial and income categories. Over time, the sophistication of these investigations has grown to include not only the race, sex, and income of the applicant, but also much broader measures of neighborhood risk, such as housing stock attributes, detailed demographic profiles, and historic price levels and transactions volume in the subject property s vicinity. Notably absent, however, from the everexpanding information set employed by previous investigations is sufficient recognition of the role and significance lender heterogeneity plays in the mortgage market. Instead, most studies charge ahead implicitly assuming either lenders are cross-sectionally identical with respect to their risk tolerances and preferences, information collection and processing capabilities, and regulatory environment or alternatively, that any differences which do indeed exist are economically insignificant and orthogonal to those characteristics included in their model specifications. 3 For example, Becker 4 and 5 and Masulis 25 argue discrimination should not persist in economically efficient markets, such as the mortgage market, as lenders with a costly taste for discrimination will ultimately be forced out of the marketplace by more efficient, profit-maximizing entities. Similarly, Schill and Wachter 37 and 38 and Tootell 40 present evidence that a more complete set of underwriting variables, designed to control for neighborhood risk factors, significantly reduces evidence of red-lining based upon neighborhood racial composition. Finally, Phillips-Patrick and Rossi 32, LaCour-Little 21, and Rachlis and Yezer 33 argue single-equation models of the mortgage market must be interpreted with caution, as the supply and demand for mortgage financing are not independent. If supply and demand vary systematically with the racial composition or income level of neighborhoods, failure to estimate a simultaneous equations model may seriously impair the accuracy and applicability of the results.
3 LENDER HETEROGENEITY IN MORTGAGE LENDING 287 Numerous reasons abound to lend credence to the belief that the omission of lender identity may create a potentially serious omitted-variables problem. For example, financial researchers and mortgage market participants have long recognized that lenders often choose to specialize in various market niches and typically employ divergent underwriting practices to achieve institutional objectives. Stengel and Glennon 39 emphasize this point and argue the reliability of any model of disparate treatment will depend critically on the ability to accurately represent the bank s underwriting policies Ž p To substantiate this claim, the authors collected detailed information regarding the unique underwriting guidelines and policies of three lending institutions. The authors then demonstrate that the ability to control for the divergent underwriting standards across these lenders significantly enhances the predictive power of their accept reject decision models. Furthermore, Avery et al. 2 contend that the vast majority of variation in minority versus white denial rates stems from individual lenders abilities to attract applicants within the neighborhoods they serve. Specifically, they argue that 75 90% of the variation in approval rates stems from differences in the quality of the applicant pool banks are able to attract, thus highlighting the need to control for critical dimensions of an institution s market position such as institutional size and focus. Similarly, Phillips-Patrick and Rossi 32 observe that a major limitation of all aggregate credit market studies is their failure to focus on lenders. Specifically, they conclude it is the dynamics of the mortgage market and the participants, both borrower and lender, that determine lending outcomes Ž p In addition to specialization, the financial performance and competitive environment in which a lending institution operates may also influence its lending activities. For example, Calomiris et al. 12 suggest the financial strength and capitalization of an institution may influence the underwriting standards employed by the lender. While their argument was originally used to explain the notoriously poor Community Reinvestment Act Ž CRA. compliance performance of financially constrained minority-owned institutions, there is little reason to doubt the proposition s applicability to a broader spectrum of 4 institutions. Similarly, Harrison 19 argues the regulatory environment faced 4 Two empirical investigations into the lending performance of minority-owned banks also attempt to control for a limited set of lender attributes. Black et al. 6 find evidence that minority-owned institutions systematically treat minority applicants less favorably than white applicants, but also demonstrate that bank-specific characteristics used to proxy for historical underwriting standards are significant factors in explaining decisions to accept or reject applications for mortgage financing. Bostic and Canner 7, on the other hand, find no evidence that minorityowned institutions treat minorities any differently than white applicants, but do find evidence of applicant-driven cultural affinity. Specifically, blacks are more likely to apply for credit at black-owned banks, while Asians are more likely to apply for credit at Asian-owned banks. Interestingly, Bostic and Canner find the inclusion of lender characteristics in their analysis adds little to the explanatory power of the model, but does tend to attenuate observed behavioral differences.
4 288 DAVID M. HARRISON by lenders varies systematically with institutional characteristics. Specifically, the paper argues regulatory effects, in addition to the above-mentioned specialization and financial strength arguments, combine to encourage lenders with differing characteristics to pursue divergent compliance policies and underwriting practices. Together, these works clearly suggest the omission of lender characteristics in traditional accept reject models of mortgage lending may lead to serious biases in identifying the existence and magnitude of lending disparities in home purchase mortgage markets. III. DATA AND METHODOLOGY To examine the impact of lender heterogeneity on the distribution of credit outcomes, a sample was constructed of conventional home purchase mortgage applications for properties located in Pinellas County, Florida, between 1993 and Pinellas County is located on the gulf coast of Florida, and comprises mainly the greater St. Petersburg metropolitan area and surrounding communities. The complete data set contains 23,094 usable observations assembled from four primary sources. First, information regarding lender decisions to accept or reject applications for home mortgage financing was obtained from loan application registers Ž LARs. of regulated lenders collected by the Federal Financial Institutions Examination Council Ž FFIEC. pursuant to the Home Mortgage Disclosure Act Ž HMDA.. These data contain basic information about the borrower and subject property, such as the requested loan amount, loan type, loan purpose, property location, and applicant s race, sex, and annual income. In addition, these records indicate whether each application was approved, denied, withdrawn, or closed for incompleteness. Second, information regarding the demographic composition and housing stock attributes of the census tract in which the subject property is located was obtained from summary files of the 1990 United States Census of Population and Housing. Third, information on historic appreciation rates and transactions volume was gleaned from the Florida Department of Revenue s property tax masterfile for Pinellas County. Finally, bank-specific characteristics were obtained from FFIEC year-end reports of condition and income Ž call reports. filed by regulated financial institutions. 5 Descriptive statistics for each of these variables, segmented by lending outcome, are presented in Table 1. A close examination of these figures reveals nearly 64% of the sample home purchase mortgage applications were approved by the prospective lender. In addition, applicant characteristics indicate the typical borrower had an annual income of slightly more than $58,000 and requested a loan of just over $86, Lender attributes were obtained for all bank holding company subsidiaries reporting at least 12 home purchase mortgage applications for properties within Pinellas county during one or more sample years. Consistent with Avery et al. 1, parent company information is used for all subsidiary banks. This criterion led to the collection of financial data for 23 unique institutions.
5 LENDER HETEROGENEITY IN MORTGAGE LENDING 289 Approximately 20% of the applications in this sample were received from single males, 20% were from single females, and 50% had a male applicant with a female co-applicant or vice versa. The remaining Ž OTHER. loan applications either failed to identify the applicant s gender or had a co-applicant whose gender matched that of the applicant. African-American and Hispanic applications account for 4.69 and 2.03% of sample applications, respectively. Lending institutions within the sample appear relatively large, with a weighted average size in excess of $14 billion; well capitalized, with risk-based capital ratios averaging 11.36%; and generally profitable, with typical return on assets Ž ROAs. in excess of 1%. Turning to neighborhood and housing stock characteristics, univariate test statistics indicate acceptance rates are inversely related to the minority population concentration, fraction of households on public assistance, and unemployment rate within a given census tract. Conversely, acceptance rates are positively related to the median income, educational attainment, and age 6 of residents within a given region. Properties located in relatively new neighborhoods, characterized by a large proportion of high-priced or owner-occupied housing units, also tend to be strong candidates for loan approval. Finally, historic transaction levels, or information effects, appear to positively influence acceptance probabilities. The data were then used to estimate a standard logit accept reject decision model of the following general form: ACCEPT f ŽApplicant Characteristics, Neighborhood Characteristics, Information Externalities, Lender Characteristics,. where ACCEPT is a discrete, binary choice variable equal to one, if the application was approved by the prospective lender, and zero if the application was denied. 7 To control for the potentially confounding influences created by differing loan purposes, only conventional home purchase applications for single-family, owner-occupied housing units are included in the sample. In addition, applications exhibiting quality or validity edit check error flags were excluded. 8 Finally, since the lender characteristics which serve as the focal 6 In the absence of detailed financial information regarding individual applicants, age is likely to proxy for accumulated wealth. 7 In the basic specification, application files which were withdrawn or closed due to incompleteness are considered to be denied by the lender. Similarly, applications which are approved by the lender, but not taken out by the applicant are still considered to have been accepted. Empirical results are not materially affected if these application categories are simply omitted from the analysis. HMDA records also include some information regarding loans purchased by an institution. These loans are not included in the analysis, as HMDA does not require institutions to report the applicant s race, sex, or income for these loans. 8 The FFIEC places edit status flags on HMDA records which appear to violate a series of logical relationships. For example, a loan record would be flagged if the census tract location reported for a subject property did not correspond with the indicated state, county, or metropolitan statistical area Ž MSA..
6 290 DAVID M. HARRISON TABLE 1 Descriptive Statistics for Mortgage Loan Applications made in Pinellas County, Florida, All applications Accepted applications Rejected applications t-test of Variable Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. difference Dependent variable Fraction of applications accepted Ž Applicant Characteristics Ž. Income $ 58,204 62,344 62,626 66,932 50,284 52, *** Loan size Ž. $ 86,161 79,046 91,285 80,806 77,063 74, *** Race Black Ž *** Ethnicity Hispanic Ž Single male Ž *** Single female Ž * M F couple Ž *** Non-conforming loan Ž *** Neighborhood Ž tract. characteristics Ž. Minority population % *** Minority population 25% Ž *** Median income Ž. $ 35, , , *** On public assistance Ž % *** High school graduates Ž % *** Residents aged 55 Ž % *** Unemployment rate Ž % ***
7 LENDER HETEROGENEITY IN MORTGAGE LENDING 291 Housing stock characteristics Median house value Ž. $ 92,662 42,109 95,172 42,170 88,203 41, *** Median rent Ž. $ *** Year built Ž yrs *** Owner occupied Ž % *** Vacant Ž % *** Information variables Historic appreciation rates Ž % Historic transaction volume Ž % *** Lender Characteristics Total assets Ž 000,000 s. Ž $. 14,200 12,000 13,100 12,100 16,700 11, *** Risk-based capital ratio Ž % *** Growth rate of assets Ž % *** Return on assets Ž % Risk-weighted assets total assets Ž % *** 1 4 family residential mortgage *** loans total assets Ž %. ***Significant at the 99% level. * Significant at the 90% level.
8 292 DAVID M. HARRISON point of this analysis should only influence the decision of a lender to accept or reject applications the institution will fund or bear credit risk upon, loans which were originated and sold into the secondary market within the same calendar year are also excluded from consideration. 9 Applicant Characteristics Clearly, one of the most fundamental determinants in a lender s decision to approve or deny a mortgage application is the potential borrower s perceived creditworthiness. Theoretically, high-income borrowers should exhibit higher debt capacities, and thus higher acceptance rates, ceteris paribus. Similarly, large loan requests would be expected to place larger burdens on the prospective borrower. Loan size should thus be negatively related to acceptance rates. In practice, applicant income and loan size are typically highly correlated, thus complicating the identification and separation of the true effects. Both the applicant s reported annual income and the ratio of the requested loan amount to the applicant s annual income Ž both measured in thousands of dollars. are included in the model specification to control for these effects. Dummy variables identifying black, Hispanic, female, and joint applicants were also included in the model specification, providing some insight into potentially disparate treatment received by these groups. 10 It is important to note, however, that a mortgage applicant s race, ethnicity, and to a lesser extent gender and marital status are often correlated with their employment history, stability, and prospects; educational opportunity and attainment; and accumulated wealth and credit history. Unfortunately, one of the major limitations of HMDA data is its well-documented lack of relevant information along these dimensions. 11 In the absence of such detailed information concerning the applicants in the present sample, race and ethnicity are likely to proxy for such omitted applicant characteristics. It is important to stress that significant coefficients on the race and ethnicity terms in this model should not be considered conclusive evidence of pervasive racial discrimination in the mortgage market. 9 Avery et al. 1 document a relationship between the origination behavior of lending institutions and their propensity to sell loans in the secondary market. Specifically, they find institutions which actively sell loans to secondary market participants process a disproportionately small number of minority loan applications, but are somewhat more likely to accept such applications than their less active counterparts. 10 For the purposes of this investigation, a joint application is considered to be any loan request with a male applicant and female co-applicant, or a female applicant with a male co-applicant. 11 Another limitation of HMDA data is the potential understatement of applicant income. Only that portion of the applicant s income necessary to satisfy the loan obligation is required to be reported. Thus, the income recorded on individual HMDA LARs may not represent the entire revenue stream available to a given applicant.
9 LENDER HETEROGENEITY IN MORTGAGE LENDING 293 Finally, Buist et al. 9 report many lenders employ secondary market standards as underwriting guidelines, even on loans they hold within the institution s portfolio. These guidelines provide the lender with additional liquidity in the event portfolio rebalancing is required. While the present data set does not allow for the inclusion of loan-to-value ratios or total obligation ratios, the classic front-end ratio can be closely approximated and controlled for. 12 Specifically, monthly payments were estimated for all borrowers assuming a 30-year, fixed rate, fully amortizing, level payment mortgage with a 10% 13 nominal interest rate. A dummy variable Ž non-conforming loan. was then constructed, taking on the value of one if the estimated monthly mortgage payment exceeded 28% of the applicant s monthly income, and zero otherwise. A strongly negative coefficient is expected for non-conforming loan. Neighborhood Characteristics To address allegations of red-lining by mortgage lenders, the percentage minority population within the census tract in which the subject property is located was included in the regression specification. 14 A significantly negative coefficient on this variable would be consistent with under-served market segments based upon the area s racial composition. Similarly, the median household income of the tract s residents is included to measure the equality of the distribution of credit outcomes across neighborhood income classifications. In this case, a significantly positive coefficient would be consistent with low-income neighborhoods being under served. In addition to racial and income measures, Schill and Wachter 38 highlight the importance of controlling for a comprehensive set of additional neighborhood characteristics. Specifically, they find race effects presented in most traditional red-lining studies disappear when a broader array of neighborhood risk factors are included in the analysis. Similarly, Tootell 40 finds a more complete set of underwriting variables reduces evidence of red-lining based upon neighborhood racial composition. Responding to these concerns, the initial regression specification is expanded to include not only the tract s racial and income characteristics, but also a detailed demographic profile containing the percentage of the tract s households receiving public assistance income Ž expected sign negative., the percentage of the tract s residents over age The absence of appraisal values for subject properties and information on non-mortgage debt obligations for potential borrowers preclude use of the aforementioned ratios. 13 Results from reestimating the model using dummy variables created from nominal interest rates of 8 and 9% are qualitatively identical to those reported. 14 Consistent with Tootell 40, an alternative model specification using a zero one dummy variable equal to one if the tract s minority population exceeded 25% of the tract s total population Ž zero otherwise. in place of the continuous percentage minority variable was also tested. The results of this alternative binary specification were qualitatively identical to those which follow below for the continuous specification.
10 294 DAVID M. HARRISON possessing a high school education or the equivalent Ž expected sign positive., the percentage of the tract s residents 55 years of age and older Žexpected sign positive., and the tract s unemployment rate Ž expected sign negative.. Housing stock attributes were also analyzed including the median house value Žexpected sign positive., median rent Ž expected sign positive., and rent-to-value ratio Ž expected sign negative. within the tract, the median year houses within the tract were constructed Ž expected sign positive., the percentage of the tract s housing units occupied by the property s owner Ž expected sign positive., and the percentage of the tract s housing units which are vacant Žexpected sign negative.. To control for potential non-linearities associated with housing values, the square of the median housing value in the neighborhood was also included. Information Externalities An information externality is created when transactions generate information which makes similar transactions in the future less risky. Lang and Nakamura s 22 and 23 seminal work provides the foundation for investigating the existence and significance of such phenomena in home mortgage lending. They argue individual home sale transactions generate information which makes future transactions less risky to lenders. Specifically, they believe a lender s decision to accept or reject an application for mortgage financing is based, in large part, on the point estimate or range of values provided by independent appraisers. Appraisal estimates, in turn, are based on the recent selling prices of comparable units within a given market area. Therefore, in areas with few historic transactions, appraisers will have a relatively difficult time providing a precise estimate of market value, and hence, accept reject decisions on the part of lenders will be plagued by more uncertainty and characterized by higher rejection rates. If historically low volume areas are correlated with socially sensitive demographic groupings, omission of information variables such as historic sales levels and appreciation rates may lead to the misidentification of lending patterns resulting from economically rational decision making processes on the part of lenders as the presence of irrational red-lining. To date, three major empirical investigations have explicitly addressed information externalities in home mortgage underwriting, all providing moderately supportive evidence for information-based theories of mortgage market activity. First, Calem 10, using a nationwide data set of major metropolitan areas, examines the Lang and Nakamura hypothesis. Interestingly, he finds evidence consistent with information externalities at work across primarily white census tracts; however, no such evidence of externalities can be detected in minority tracts. Ling and Wachter 24, on the other hand, using data for Dade county Ž Miami. Florida find strong evidence of information externalities exerting an economically significant influence across all tract groupings. Specifically, they find a one standard deviation positive shock to historic
11 LENDER HETEROGENEITY IN MORTGAGE LENDING 295 appreciation rates and sales levels increases the probability of acceptance on home mortgage applications by 3 4%, with applications in minority neighborhoods being more sensitive to such information. Finally, Avery et al. 3 attempt to identify and distinguish between internal and external information effects. Specifically, they argue that processing loan applications generates information about the neighborhood in which the property is located. Within the Lang and Nakamura context, all market participants benefit equally from the enhanced information set via more accurate appraisals. Avery et al. 3 term this societal benefit an external information effect, but also argue that the process of evaluating mortgage applications within a given neighborhood creates private information which enhances the future information set of only the evaluating underwriting lender. The authors further contend this internal information effect may be far more important than the Lang and Nakamura style external effect, and find that up to 25% of observed disparities in denial rates across neighborhood racial and income classifications can be explained by 15 these internal economies of scale. Interestingly, both Calem 10 and Ling and Wachter 24 employ data exclusively from 1990, while Avery et al. use data from 1990 and Therefore, all three studies effectively analyze data from a period of economic recession Ž particularly in real estate markets. during which Lang and Nakamura hypothesize information effects should be most pronounced. The results from the current analysis will provide the first non-recessionary evidence on the existence, persistence, and economic significance of information externalities in home mortgage lending. To control for the potential influences of external information externalities, two variables were added to the model specification. First, the department of revenue data were geocoded and cleaned using the Gatzlaff and Ling 17 methodology. Median house price indices were then constructed and 3-year historic, geometric appreciation rates were calculated and included in the model specification. Transaction volumes within each tract were also derived, with the sum of historic sales volumes over the previous 3 years, scaled by the number of owner-occupied housing units within the tract, included in all regressions. The Lang and Nakamura information externality hypothesis would predict positive signs for the coefficients on both of these variables. Lender Characteristics To evaluate the importance of lender identity on the home mortgage underwriting process, a set of institutional attributes designed to measure the financial condition and operating position of prospective lenders was also 15 This finding is consistent with Avery et al. 2 who demonstrate approval rates are directly related to the number of applications processed by an institution, but stands in direct contrast to Avery et al. 1 who find little evidence of a relationship between the number of applications processed and denial probabilities.
12 296 DAVID M. HARRISON included in the accept reject model specification. 16 First, to address the issue of lender specialization and internal information effects, the ratio of 1 4 family residential mortgage loans to total assets was added. 17 Banks specializing in mortgage lending will exhibit high ratios and may exhibit high acceptance rates if experience or economies of scale in data processing and collection facilitate the profitable funding of mortgage loans. Alternatively, if specialization is achieved through marketing efforts designed exclusively to increase the size of the applicant pool, this ratio may be unrelated to acceptance rates. 18 Second, Calomiris et al. Ž hereafter, CKL. 12 argue undercapitalized institutions may fund fewer loans. While Avery et al. 1 provide only weak empirical support for this contention, to account for the possibility of such a relationship, each institution s risk-based capital ratio is included in the analysis. If the CKL hypothesis is correct, a positive relationship between capitalization and acceptance rates should be observed. Differing risk tolerances and preferences across mortgage lenders may also influence underwriting practices. To investigate this possibility, the ratio of the lender s risk-weighted assets to total assets is included in the model specification. Mortgage lenders who are willing and able to assume higher risk levels may be expected to exhibit higher acceptance rates, as they should be more willing to fund marginal loans Žsuch as those not conforming to secondary market standards. in attempts to generate additional fee income. Alternatively, lenders already possessing weak or risky asset portfolios may be hesitant to reach deeply into the applicant pool in efforts to avoid overextending the institution s financial position. Size effects may also alter the dynamics of mortgage underwriting via both internal economies of scale and the regulatory process. For example, Neuberger and Schmidt 31 suggest information should be cheaper in areas where a bank 16 Note, the standard errors reported in the following section assume there are 23,094 independent observations for each lender characteristics, when in reality there are only 23. Moulton 27 argues that regression errors are often correlated within groups, and demonstrates that group effects may well lead to a downward bias in estimates of unadjusted standard errors. This implies that the statistical significance of the lender characteristic estimates to follow may be overstated. Moulton 28 further explores these issues and provides a correction within the traditional OLS framework. Unfortunately, at this time I am unaware of an equivalent correction applicable to the logistic analysis. 17 Avery et al. 1 examine lender specialization using the ratios of mortgages to assets and core deposits to assets. They find institutions which focus primarily on lending activities originate a disproportionately high number of minority loans. Bostic and Canner 7 employ real estate loans to total assets, non-real estate loans to total assets, and core deposits to assets to investigate lender focus. They find minority-owned institutions tend to hold fewer mortgages per dollar of assets than their white-owned peers. 18 In the extreme, it is possible that efforts to increase the size of the applicant pool will lead to a dilution in the creditworthiness of prospective borrowers, and thus, a negative relationship between acceptance rates and specialization.
13 LENDER HETEROGENEITY IN MORTGAGE LENDING 297 has a large presence, while Avery et al. 3 suggest large lenders may well have access to more and better information than their smaller competitors. This enhanced information set should reduce the uncertainty Ž i.e., risk. faced by the bank, and thus lead to higher acceptance rates at larger lenders. Similarly, Flannery and Houston 16 argue bank size may affect regulatory treatment on a variety of issues, while Harrison 19 finds institutional size is positively related to an institution s CRA compliance performance. Again, all of these findings suggest acceptance rates may be positively related to bank size. The natural log of the lending institution s total assets, as well as the growth rate in total assets, are thus included in the model to investigate the potential for size effects. 19 Size is only one dimension of an institution s regulatory environment. Avery et al. 2 show that lending activities may also vary by type of institution Ži.e., thrift, commercial bank, mortgage bank.. Specifically, they show that acceptance rates may vary by more than 10% across different types of institutions. Similarly, Moore 26 documents differential CRA enforcement practices, even after controlling for lending activities, across regulatory agencies. To control for any additional regulatory influences, a set of dummy variables is added to the regression, identifying which agency has primary regulatory oversight for the institution. Finally, to analyze the importance of institutional efficiency and profitability on acceptance rates, the institution s return on assets for the previous year is included in the analysis. 20 In the absence of discrimination or other market imperfections, as lenders reach more deeply into the applicant pool, the expected risk adjusted return should monotonically decrease. 21 Therefore, in an efficient market we would expect profitability to be inversely related to acceptance rates. Alternatively, if bank examiners employ a pecking order theory of regulation in which institutions are first required to meet safety and soundness goals, and then forced to invest in community reinvestment initiatives, profitability may be directly related to acceptance rates. Given the conflicting expectations of these alternative hypotheses, the expected sign on institutional profitability is left as an empirical question. In the next section, the economic significance of each of these lender characteristics will be analyzed, as well as their joint effect on the other model parameters. 19 The ability of large, multibank holding companies to pool loans from various subsidiary banks, and thus geographically diversify their loan pools also suggests a positive relationship between bank size and acceptance rates. 20 Avery et al. 1 use an institution s ROAs to control for profitability, and find banks with higher than expected minority application rates also exhibit higher than expected profits. Bostic and Canner 7 also use ROAs as a measure of financial performance, and find minority-owned banks exhibit lower profitability Ž ROA. than their white-owned peers. 21 Note, rational investors will undertake their most profitable opportunities first, moving on to less profitable alternatives as time and resources permit.
14 298 DAVID M. HARRISON IV. EMPIRICAL RESULTS Table 2 presents results from the accept reject model of home purchase mortgage applications under four alternative model specifications. The first column presents results from a naive HMDA model which incorporates only the applicant characteristics, neighborhood minority population percentage, and information variables. As expected, higher applicant incomes are associated with higher acceptance rates. Similarly, the measure of non-conforming loans is highly significant and negatively related to acceptance probabilities. The loanto-income ratio also exhibits a negative coefficient, but is surprisingly insignificant under this model specification. Interestingly, single-male applicants are consistently less likely to be approved for home mortgage financing, under this basic model specification, than their single-female or joint counterparts. In addition, the race of both the applicant and of the census tract appear highly significant, with black applicants and applications for properties in minority neighborhoods being accepted less often. Again, recall that the applicant race variable likely proxies for a variety of missing applicant characteristics, such as wealth, employment prospects, and credit history, and should not be taken as conclusive evidence of pervasive racial discrimination in the mortgage market. Finally, turning to the information variables, previous sales levels, but not appreciation rates, exhibit a significantly positive relationship with acceptance probabilities. This result provides further evidence in support of Lang and Nakamura s information externalities hypothesis. The second column of Table 2 expands the initial specification to include a broader array of neighborhood Ž tract level. demographic characteristics as well as information regarding the area s housing stock attributes. Similar to the naive specification, applicant income and front-end ratios are typically important determinants of the lender s underwriting decision. Single-female and joint applicants continue to receive more credit than their single-male counterparts, while black applicants continue to receive less. The neighborhood racial composition variable remains negative across this model specification. However, with the inclusion of the expanded information set, the magnitude is reduced by over 40%. The expanded neighborhood characteristics suggest both the percentage of tract residents who are high school graduates and the percentage of tract residents who are 55 years of age or older tend to positively influence acceptance rates. This result is consistent with expectations, as the two variables are designed to capture educational attainment and accumulated wealth, respectively. Surprisingly, the tract s unemployment rate is marginally positively related to the probability of acceptance. Only two housing stock attributes appear significant in this specification, with both exhibiting the expected signs. Specifically, applications for houses located in newer neighborhoods appear more likely to be approved, while a high proportion of vacant structures within the area detracts from the probability of acceptance. Turning again to the information variables, appreciation rates
15 LENDER HETEROGENEITY IN MORTGAGE LENDING 299 TABLE 2 Logit Regressions Explaining Probability of Acceptance for Home Purchase Mortgage Applications Made in Pinellas County, Florida, during Variable Model 1 Model 2 Model 3 Model 4 Constant Ž 2.14.** Ž 2.16.** Ž Ž Applicant characteristics Applicant income Ž 8.07.*** Ž 7.09.*** Ž 5.88.*** Ž 6.21.*** Loan-to-income ratio Ž Ž Ž Ž Race Black Ž 3.47.*** Ž 3.59.*** Ž 4.41.*** Ž 4.11.*** Ethnicity Hispanic Ž Ž Ž Ž Single female Ž 8.51.*** Ž 8.36.*** Ž 7.47.*** Ž 7.17.*** M F couple Ž 9.37.*** Ž 9.07.*** Ž 8.13.*** Ž 8.13.*** Other Ž 6.00.*** Ž 5.91.*** Ž 6.67.*** Ž 5.19.*** Non-conforming loan Ž 7.77.*** Ž 7.87.*** Ž 8.74.*** Ž 8.78.*** Neighborhood Ž census tract. characteristics Minority population percentage Ž 8.05.*** Ž 3.29.*** Ž 2.46.** Ž 2.52.** Median income Ž Ž Ž On public assistance Ž Ž Ž High school graduates Ž 1.91.* Ž 2.01.** Ž Residents aged Ž 3.44.*** Ž 3.50.*** Ž 3.38.*** Unemployment rate Ž 1.89.* Ž 0.81.* Ž 2.01.** Information variables Historic appreciation rates Ž Ž Ž Ž Historic transaction volume Ž 9.19.*** Ž 2.79.*** Ž 3.05.*** Ž 3.10.*** Housing stock characteristics Median house value Ž Ž Ž Median house value squared Ž Ž Ž Median rent Ž Ž Ž Rental index Ž Ž Ž 1.18.
16 300 DAVID M. HARRISON TABLE 2 Continued Variable Model 1 Model 2 Model 3 Model 4 Housing stock characteristics Year built Ž 2.03.** Ž Ž Owner occupied Ž Ž Ž Vacant Ž 3.23.*** Ž 3.22.*** Ž 2.97.*** Lender characteristics Log of total assets Ž Risk-based capital ratio Ž 2.61.*** Growth rate of totals assets Ž 2.83.*** Return on assets Ž Risk-weighted assets total assets Ž 8.31.*** 1 4 family residential mortgage Ž 5.47.*** loans total assets Time Ž year. dummies Yes Yes Yes Yes Lender fixed effects No No Yes No Regulator dummies No No No Yes Number of observations 23,094 23,094 23,094 23,094 2 Ž *** *** 1,506.99*** 1,779.06*** 2 Pseudo-R Log likelihood 14,566 14,537 14,117 13,919 *** Significant at the 99% level. ** Significant at the 95% level. * Significant at the 90% level. remain insignificant, while the inclusion of the expanded information set substantially reduces the magnitude of the transaction volume coefficient. Historic sales levels do, however, remain highly significant even after the inclusion of these neighborhood attributes. Next, column three expands the analysis in model two by incorporating fixed effects for each lender. Once again, few qualitative changes are readily noticeable between model two and three. Applicant income is still a positive and strongly significant determinant of loan approval, while non-comforming loans remain less likely to be accepted. Single-female and joint applications continue to receive more favorable treatment than applications submitted by single men, while black applicants continue to receive less favorable treatment. Turning to neighborhood profiles, acceptance probabilities remain negatively related to
17 LENDER HETEROGENEITY IN MORTGAGE LENDING 301 both the proportion of minority residents within the tract and the proportion of vacant structures. On the other hand, acceptance rates are again positively related to measures of educational attainment, accumulated wealth, and surprisingly, the tract s unemployment rate. Finally, I once again find historic transactions levels to be strongly related to acceptance probabilities. While the qualitative changes between models two and three are very subtle, the quantitative differences are far more striking. For example, the magnitude of both information externality coefficients increases more than 10% across models. This suggests previous studies which ignore lender heterogeneity may be understating the importance of such information effects. Similarly, the coefficient for the tract s minority population percentage declines by nearly 25%. This finding suggests that controlling for inter-lender differences may dramatically reduce evidence of red-lining based upon neighborhood racial composition if controlled for in traditional studies. Conversely, the coefficient on applicant Race Black also changes by nearly 25%. This change suggests that controlling for lender differences enhances the importance of applicantspecific characteristics such as accumulated wealth, credit history, employment prospects and stability, and perhaps even race ethnicity. This result highlights the need for future researchers to find alternative data sources which include more comprehensive measures of individual applicant creditworthiness such as credit scores. Finally, the inclusion of lender fixed-effects is shown to enhance the predictive power of the model, as both chi-square and pseudo-r 2 goodnessof-fit test statistics exhibit marked increases. Clearly, the inclusion of lender fixed-effects has striking implications for the magnitudes of additional model parameters. While the lender fixed-effects model in column three of Table 2 may adequately control for heterogeneity across institutions, it fails to provide any insight regarding which dimensions of heterogeneity matter to the underwriting decision. Therefore, to investigate which aspects of lender heterogeneity are most important, column four replaces the lender fixed effects with the set of institutional attributes described above. As this set of lender characteristics is only available for a subset of the data, a modified zero-order regression was used to control for the missing values. This technique replaces missing values for each lender attribute with the value of zero, while simultaneously creating a 0 1 indicator variable identifying those observations with missing data. During the estimation procedure, the influence of each lender attribute is thus effectively estimated over only that subset of data with non-zero values, while the corresponding dummy variable provides an indication of whether the observations without information regarding lender attributes are systematically different from those with complete information. 22 Interestingly, empirical results show this zero-order coefficient to be insignificant in the analysis which follows. 22 For a more complete description of modified zero-order regression, see Greene 18.
18 302 DAVID M. HARRISON The results from this final specification indicate that financial institutions specializing in mortgage lending accept a significantly higher percentage of conventional home purchase mortgage applications than their peers, as evidenced by the significant positive coefficient on the ratio of 1 4 family residential mortgage loans to total assets. This suggests economies of specialization Ž or economies of scale in data collection and processing. may accrue to mortgage lenders. This finding is also consistent with internal economies of scale, such as those previously described by Avery et al. 23 Turning to the coefficient estimate for the risk-based capital ratio, I find evidence supporting the CKL contention that undercapitalized institutions may fund fewer loans, as higher capital ratios appear to be associated with enhanced acceptance probabilities. Interestingly, institutional risk exhibits a significantly negative relationship with acceptance rates. This is consistent with the notion that banks which currently possess a relatively risky portfolio of assets may be hesitant to reach deeply into the applicant pool. Surprisingly, institution size is insignificant, and the growth rate in total assets is negatively related to application acceptance probabilities. This finding is inconsistent with the previous literature that suggests large banks may have an informational and cost advantage with respect to lending decisions. Rather, this result suggests small community banks with fewer growth ambitions may be more focused on meeting the unique credit needs of individual applicants than larger, regional organizations. Finally, institutional profitability does not appear to influence underwriting decisions within this sample. Clearly, the evidence discussed above provides ample justification to warrant the inclusion of lender characteristics in mortgage market models. Additional evidence supporting their relevance can be found by examining the effect of their inclusion on other model parameters of interest. As with the lender fixed-effects, inclusion of the bank-specific characteristics is shown to markedly enhance the explanatory power of the model, as evidenced by improvements in both chi-square and pseudo-r 2 values. Furthermore, including these lender attributes again dramatically decreases Ž by over 20%. the magnitude of the tract s minority population percentage. Thus, previous studies which fail to account for differing underwriting practices and policies across mortgage lenders may well be overstating potential disparities in credit market outcomes. Specifically, previous studies may have erroneously identified red-lined areas in some situations where they do not indeed exist. Similarly, the inclusion of lender characteristics also enhances the magnitude and significance of the 23 Indeed, when the ratio of 1 4 family residential mortgage loans to total assets was replaced in the regression specification with the number of applications processed by the lender, I again find strong evidence of internal scale economies. A high degree of collinearity between these variables Ž suggests the simultaneous inclusion of both specialization metrics within the same model specification is inappropriate.
19 LENDER HETEROGENEITY IN MORTGAGE LENDING 303 information effect. Specifically, the magnitude of the relationship between historic transactions volume and acceptance probabilities jumps over 15%. This coefficient remains highly significant after the inclusion of institutional attributes, suggesting previous empirical investigations into the existence of information externalities may have Ž significantly. understated their true effects. The results presented in column four of Table 2 provide clear evidence supporting the inclusion of lender characteristics in models of the home mortgage underwriting process. 24 Simulations The importance of including lender characteristics in accept reject models is further evidenced through the use of logistic simulations. The values reported in Table 3 represent the estimated marginal effects of each independent variable on the probability of acceptance. Rather than reporting simple odds ratios, to estimate the marginal impact of each independent variable, I first estimated the acceptance probability for a standard application. I then calculated the change in acceptance rates associated with altering individual applicant characteristics. In the base case, a prospective borrower is assumed to be a white applicant with an annual income of $50,000. The applicant, whose gender was not specified on the loan application register, is not liquidity constrained, and is assumed to be applying for an $87,500 loan. The property is assumed to be located in a neighborhood where 10% of the resident population is composed of racial or ethnic minorities. All remaining continuous application variables are set at their means, while all remaining dichotomous application characteristics are set to zero. Factor sensitivities were then constructed by independently shocking each continuous variable by the stated amount. Factor sensitivities for dichotomous independent variables were constructed by comparing acceptance rates in the presence and absence of the characteristic. This approach is consistent with an odds ratio analysis, but is less prone to distortion caused when continuous covariates are modeled linearly in the logit One potential explanation for the observed significance of lender characteristics is that the presence or absence of each attribute may facilitate the purchase of discrimination. The present data do not support such an explanation, as dividing the observations into quartiles for each attribute fails to produce any evidence of race effects within classifications. Similarly, estimating the model individually for each sample institution fails to identify a pattern of race effects correlated with institutional characteristics. Alternatively, some researchers may argue bank characteristics are highly collinear with socially sensitive variables, and their inclusion simply masks the truly discriminatory behavior occurring in the mortgage market. Variance inflation factors Ž VIF. do not support this hypothesis, however, as the maximum VIF for any institutional attribute in model three is only 2.00, while the average VIF across all six characteristics is only Both values are well below generally accepted problem indicator levels, suggesting collinearity among lender characteristics is not materially influencing the results. 25 See Hosmer and Lemeshow 20.
20 304 DAVID M. HARRISON TABLE 3 Estimated Marginal Effect of Selected Independent Variables on the Probability of Acceptance Estimated Change in Borrower characteristic probž default.ž%. probž default.ž%. Applicant RACE Black Hispanic Applicant income Ž. $ 25, , , , Applicant status Single male Single female M F couple Liquidity constrained Yes Neighborhood % minority Historic sales volume Ž % Risk-based capital ratio Ž % family residential mortgages to total assets Ž %
A Response to Comments Response to Critiques of Mortgage Discrimination and FHA Loan Performance James A. Berkovec Glenn B. Canner Stuart A. Gabriel Timothy H. Hannan Abstract This response discusses the
Flaws in the Use of Loan Defaults To Test for Mortgage Lending Discrimination Flaws in the Use of Loan Defaults To Test for Mortgage Lending Discrimination Stephen L. Ross University of Connecticut Abstract
The Social Science Journal 42 (2005) 479 486 Community socioeconomic status and disparities in mortgage lending: An analysis of Metropolitan Detroit Robert Mark Silverman Department of Urban and Regional
Comparing Loan Performance Between Races as a Test for Discrimination George Galster The Urban Institute Abstract This critique argues that analysis of loan performance data cannot provide a clear indicator
working p a p e r 9 6 2 0 Neighborhood Information and Home Mortgage Lending by Robert B. Avery, Patricia E. Beeson, and Mark S. Sniderman FEDERAL RESERVE BANK OF CLEVELAND Neighborhood Information and
PUBLIC DISCLOSURE JANUARY 8, 2015 MORTGAGE LENDER COMMUNITY INVESTMENT PERFORMANCE EVALUATION NEW FED MORTGAGE CORP. MC1881 98 HIGH STREET DANVERS MA. 01923 DIVISION OF BANKS 1000 WASHINGTON STREET BOSTON
Wodshg Paoer 9219 CROSS-LENDER VARIATION IN HOME MORTGAGE LENDING by Robert B. Avery, Patricia E. Beeson, and Mark S. Snideman Robert B. Avery is an associate professor in the Department of Consumer Economics
Mark W Olson: Home Mortgage Disclosure Act Testimony by Mr Mark W Olson, Member of the Board of Governors of the US Federal Reserve System, before the Subcommittee on Financial Institutions and Consumer
How Banks Can Self-Monitor Their Lending To Comply with the Equal Credit Opportunity Act James H. Gilkeson, Drew B. Winters, and Peggy D. Dwyer T he Equal Credit Opportunity Act (ECOA) prohibits lenders
Have the GSE Affordable Housing Goals Increased the Supply of Mortgage Credit? Brent W. Ambrose * Professor of Finance and Director Center for Real Estate Studies Gatton College of Business and Economics
The Home Mortgage Disclosure Act History, Evolution, and Limitations Joseph M. Kolar Buckley Kolar LLP March 14, 2005 Washington, DC History of HMDA Enacted 30 years ago, HMDA s purposes and requirements
Running head: DISCRIMINATION IN ACCESS TO FINANCE 1 Discrimination in Access to Finance: Evidence from the United States Small Business Credit Market Ethan Yiqi Pan Advisor: Shannon Mudd Haverford College
Federal Reserve Bank of Dallas A Banker s Quick Reference Guide to CRA As amended effective September 1, 2005 This publication is a guide to the CRA regulation and examination procedures. It is intended
March 31, 2005 FREQUENTLY ASKED QUESTIONS ABOUT THE NEW HMDA DATA General Background 1. What is the Home Mortgage Disclosure Act (HMDA)? HMDA, enacted by Congress in 1975, requires most mortgage lenders
Mortgage Lending Discrimination and Racial Differences in Loan Default 117 Journal of Housing Research Volume 7, Issue 1 117 Fannie Mae Foundation 1996. All Rights Reserved. Mortgage Lending Discrimination
Small Institutions Examination Procedures and Sample Format for Public Disclosure of Examination Results The Examination Procedures for Small Institutions (which include the CRA Ratings Matrix for Small
Fair Lending Update 2012 Banker Outreach Program Fair Lending Discussion Topics: Fair Lending High Risks Areas Pricing and Underwriting Examination Focus and Procedures How to Conduct a Comparative Analysis
SMALL BANK Comptroller of the Currency Administrator of National Banks Washington, DC 20219 PUBLIC DISCLOSURE May 29, 2012 COMMUNITY REINVESTMENT ACT PERFORMANCE EVALUATION Midland FS&LA Charter Number
LEGISLATIVE REPORT ANALYSIS OF HOME MORTGAGE DISCLOSURE ACT (HMDA) DATA FOR TEXAS, 1999-2001 REPORT PREPARED FOR THE FINANCE COMMISSION OF TEXAS AND THE OFFICE OF CONSUMER CREDIT COMMISSIONER BY THE TEXAS
Neighborhood Diversity Characteristics in Iowa and their Implications for Home Loans and Business Investment Liesl Eathington Dave Swenson Regional Capacity Analysis Program ReCAP Department of Economics,
O SMALL BANK Comptroller of the Currency Administrator of National Banks Washington, DC 20219 PUBLIC DISCLOSURE March 02, 2009 COMMUNITY REINVESTMENT ACT PERFORMANCE EVALUATION First National Bank of Michigan
Exploratory Study of the Accuracy of Home Mortgage Disclosure Act (HMDA) Data Final Report Cambridge, MA Lexington, MA Hadley, MA Bethesda, MD Washington, DC Chicago, IL Cairo, Egypt Johannesburg, South
Issue Brief HUD s AFFORDABLE LENDING GOALS FOR FANNIE MAE AND FREDDIE MAC Fannie Mae and Freddie Mac, government-sponsored enterprises (GSEs) in the secondary mortgage market, are the two largest sources
Characteristics of Home Mortgage Lending to Racial or Ethnic Groups in Iowa Liesl Eathington Dave Swenson Regional Capacity Analysis Program ReCAP Department of Economics, Iowa State University September
Federal Reserve Bank of Dallas A Banker s Quick Reference Guide to CRA This publication is a guide to the CRA regulation and examination procedures. It is intended for bank CEOs, presidents, and CRA and
Banker s Guide To Risk-Based Fair Lending Examinations RESERVE BANK OF FEDERAL CHICAGO INCORPORATED 1914 MAY 18, Introduction Banker s Guide To Risk-Based Fair Lending This publication is a guide to the
Fact Sheet: Unequal Opportunity Disparate Mortgage Origination Patterns for Women in the Chicago Area Introduction This fact sheet examines women s access to mortgages following the collapse of the housing
Bank Consolidation and Small Business Lending within Local Markets April 2003 Working Paper 2003-02 Katherine Samolyk Division of Insurance and Research Federal Deposit Insurance Corporation Washington,
April 3, 2006 FREQUENTLY ASKED QUESTIONS ABOUT THE NEW HMDA DATA General Background 1. What is the Home Mortgage Disclosure Act (HMDA)? HMDA, enacted by Congress in 1975, requires most mortgage lenders
II. ACCESS TO CREDIT FOR MINORITY-OWNED BUSINESSES Racial Differences in Patterns of Small Business Finance: The Importance of Local Geography Raphael W. Bostic and K. Patrick Lampani Competition, Small
I Cross-Lender Variation in Home Mortgage Lending by Robert B. Avery, Patricia E. Beeson, and Mark S. Sniderman Robert B. Avery is an associate professor in the Department of Consumer Economics and Housing,
CRA Special Lending Programs Robert B. Avery, Raphael W. Bostic, and Glenn B. Canner, of the Board s Division of Research and Statistics, prepared this article. Increasing the flow of credit to lower-income
January 212 No. 212-1 Home-Mortgage Lending Trends in New England in 21 Ana Patricia Muñoz The views expressed in this paper are solely those of the author and do not necessarily represent those of the
SMALL BANK Comptroller of the Currency Administrator of National Banks PUBLIC DISCLOSURE July 12, 2004 COMMUNITY REINVESTMENT ACT PERFORMANCE EVALUATION First National Bank of LaGrange Charter Number 13941
Income and Racial Disparities in Subprime Lending May 2000 U.S. Department of Housing and Urban Development Office of Policy Development and Research UNEQUAL BURDEN IN CHICAGO: INCOME AND RACIAL DISPARITIES
Information Externalities and Residential Mortgage Lending in the Hardest Hit Housing Market: The Case of Detroit Lei Ding Federal Reserve Bank of Philadelphia 1 Abstract The flow of credit to the residential
TABLE OF CONTENTS EXAMINATION PROCEDURES FOR LARGE INSTITUTIONS... 2 EXAMINATION SCOPE... 2 PERFORMANCE CONTEXT... 3 ASSESSMENT AREA... 4 LENDING, INVESTMENT, AND SERVICE TESTS FOR LARGE RETAIL INSTITUTIONS...
PUBLIC DISCLOSURE FEBRUARY 12, 2015 MORTGAGE LENDER COMMUNITY INVESTMENT PERFORMANCE EVALUATION NE MOVES MORTGAGE, LLC MC2584 52 SECOND AVENUE, 3RD FLOOR WALTHAM MA. 02451 DIVISION OF BANKS 1000 WASHINGTON
COMMUNITY REINVESTMENT ACT EXAMINATION PROCEDURES FOR LARGE INSTITUTIONS Examination Scope For institutions (interstate and intrastate) with more than one assessment area, identify assessment areas for
UCTC Final Report Why Do Inner City Residents Pay Higher Premiums? The Determinants of Automobile Insurance Premiums Paul M. Ong and Michael A. Stoll School of Public Affairs UCLA 3250 Public Policy Bldg.,
PUBLIC DISCLOSURE January 07, 2014 COMMUNITY REINVESTMENT ACT PERFORMANCE EVALUATION Lake Federal, FSB Charter Number 706128 7048 Kennedy Ave Hammond, IN 46323-2212 Office of the Comptroller of the Currency
HMDA DATA ON DEMAND FREQUENTLY ASKED QUESTIONS About the Home Mortgage Disclosure Act (HMDA) One of the most comprehensive sources of publically available application-level information on the single-family
CRA and fair lending regulations: Resulting trends in mortgage lending Douglas D. Evanoff and Lewis M. Segal In response to concerns that banks were not adequately serving the credit needs of their local
Small Bank Performance Evaluation PUBLIC DISCLOSURE November 30, 1998 COMMUNITY REINVESTMENT ACT PERFORMANCE EVALUATION Guaranty National Bank of Tallahassee Charter Number 21162 111 South Monroe Street
Literature Review Introduction An extensive literature on small business lending has examined the impacts of restructuring in the financial industry on banks capacities to satisfy credit needs of small
Tips for Reporting Accurate HMDA and CRA Data Outlook Live Webinar - November 17, 2010 Cindy J. Anderson, Senior Examiner Karin Modjeski Bearss, Senior Examiner Visit us at www.consumercomplianceoutlook.org
The Pricing of Home Mortgage Loans to Minority Borrowers: How Much of the APR Differential Can We Explain? Author Marsha J. Courchane Abstract The public releases of the 2004 and 2005 HMDA data have engendered
DISCRIMINATION IN THE SMALL BUSINESS CREDIT MARKET David G. Blanchflower Dartmouth College and National Bureau of Economic Research Phillip B. Levine Wellesley College and National Bureau of Economic Research
Joint Center for Housing Studies Harvard University Subprime Lending and the Community Reinvestment Act Kevin Park November 2008 N08-2 by Kevin Park. All rights reserved. Short sections of text, not to
Subprime Foreclosures: The Smoking Gun of Predatory Lending? Harold L. Bunce, Debbie Gruenstein, Christopher E. Herbert, and Randall M. Scheessele One of the most striking features of home finance in the
Chapter 9 Home Mortgage Disclosure Act Data Analysis The Home Mortgage Disclosure Act (HMDA) was passed by Congress in 1977. The goal of this act is to make information available to community groups, government
Issue Brief Advocacy: the voice of small business in government Issue Brief Number 3 Access to Capital for Women- and Minority-owned Businesses: Revisiting Key Variables By Christine Kymn At A Glance Small
Risk or Race: an assessment of subprime lending patterns in nine metropolitan areas U.S. Department of Housing and Urban Development Office of Policy Development and Research Visit PD&R s website www.hud.gov/policy
Chapter 5: Analysis of The National Education Longitudinal Study (NELS:88) Introduction The National Educational Longitudinal Survey (NELS:88) followed students from 8 th grade in 1988 to 10 th grade in
REGULATORY ALERT NATIONAL CREDIT UNION ADMINISTRATION 1775 DUKE STREET, ALEXANDRIA, VA 22314 DATE: June 5, 1997 NO.: 97-RA-9 TO: All Credit Unions Compliance Officer or Chief Executive Officer SUBJECT:
MINORITY-OWNED SMALL BUSINESSES, SBA LENDING, AND STATE-LEVEL ECONOMIC PERFORMANCE Cortes, Bienvenido S. Pittsburg State University McKinnis, Michael R. Pittsburg State University ABSTRACT Aside from poor
An Empirical Analysis of Insider Rates vs. Outsider Rates in Bank Lending Lamont Black* Indiana University Federal Reserve Board of Governors November 2006 ABSTRACT: This paper analyzes empirically the
PUBLIC DISCLOSURE June 2, 2014 CRA FOR MORTGAGE LENDERS PERFORMANCE EVALUATION RADIUS FINANCIAL GROUP INC. ML1846 600 LONGWATER DRIVE SUITE 107 NORWELL, MA 02061 DIVISION OF BANKS 1000 WASHINGTON STREET
A Companion to Urban Economics Edited by Richard J. Arnott, Daniel P. McMillen Copyright 2006 by Blackwell Publishing Ltd C H A P T E R T W E L V E Discrimination in Mortgage Lending Anthony M. Yezer 12.1
MORTGAGE LOAN PREQUALIFICATIONS: APPLICATIONS OR NOT A GUIDE FOR COMPLYING WITH REGULATIONS B AND C Federal Deposit Insurance Corporation Division of Compliance & Consumer Affairs Federal Deposit Insurance
New and Re-emerging Fair Lending Risks Article by Austin Brown & Loretta Kirkwood BY AUSTIN BROWN & LORETTA KIRKWOOD Austin Brown Loretta Kirkwood Regulators have been focused recently on several new and
Why Default Rates Cannot Shed Light on Mortgage Discrimination John Yinger 1 Syracuse University Abstract Some scholars argue that racial discrimination in mortgage lending can be observed in default rates.
Journal of Economic Perspectives Volume 12, Number 2 Spring 1998 Pages 41 62 Evidence on Discrimination in Mortgage Lending Helen F. Ladd The ability to get a mortgage is often the key to an individual
Mortgage Lending I. Definitions A lending institution is a bank, savings institution, credit union or mortgage company. A bank, savings institution, or credit union is required to report HMDA data if within
DEMOGRAPHICS OF PAYDAY LENDING IN OKLAHOMA Haydar Kurban, PhD Adji Fatou Diagne HOWARD UNIVERSITY CENTER ON RACE AND WEALTH 1840 7th street NW Washington DC, 20001 TABLE OF CONTENTS 1. Executive Summary
Fannie Mae and Freddie Mac Mortgage Purchases in Low-Income and High-Minority Neighborhoods: 1994 96 Fannie Mae and Freddie Mac Mortgage Purchases in Low-Income and High- Minority Neighborhoods: 1994 96
Demystifying the Refi-Share Mystery Authors Yan Chang and Frank E. Nothaft Abstract The refinance shares reported by Freddie Mac s Primary Mortgage Market Survey (PMMS), the Mortgage Bankers Association
PUBLIC DISCLOSURE April 4, 2012 MORTGAGE LENDER COMMUNITY INVESTMENT PERFORMANCE EVALUATION CRESCENT MORTGAGE COMPANY ML4247 5901 Peachtree Dunwoody Road Suite 250, Bldg C Atlanta, GA 30328 DIVISION OF
Ten Ways to Avoid Common HMDA Reporting Errors Home Mortgage Disclosure Act (HMDA) data must be submitted to the regulators by March 1 st. During this environment of enhanced regulatory and public scrutiny,
Kauffman Dissertation Executive Summary Part of the Ewing Marion Kauffman Foundation s Emerging Scholars initiative, the Kauffman Dissertation Fellowship Program recognizes exceptional doctoral students
Using Big Data to Improve the Mortgage Industry Operating Model Overview of the HMDA S B A Score By David K. Moffat Mortgage TrueView December 2014 Copyright 2014 Mortgage TrueView Inc. All Rights Reserved
DISCRIMINATION IN THE SMALL BUSINESS CREDIT MARKET David G. Blanchflower Dartmouth College and National Bureau of Economic Research Phillip B. Levine Wellesley College and National Bureau of Economic Research
OCC 97-24 OCC BULLETIN Comptroller of the Currency Administrator of National Banks Subject: Credit Scoring Models Description: Examination Guidance TO: Chief Executive Officers of all National Banks, Department
Access Denied Low-Income and Minority Families Face More Credit Constraints and Higher Borrowing Costs Christian E. Weller Senior Fellow, Center for American Progress August 2007 w w w. a m e r i c a n
Bank Consolidation and the Provision of Banking Services: The Case of Small Commercial Loans by Robert B. Avery and Katherine Samolyk* December 2000 Robert B. Avery Federal Reserve Board Washington, DC
THE ATTORNEY GENERAL S 2006 ANNUAL REPORT TO CONGRESS PURSUANT TO THE EQUAL CREDIT OPPORTUNITY ACT AMENDMENTS OF 1976 SUBMITTED BY WAN J. KIM ASSISTANT ATTORNEY GENERAL MARCH 13, 2007 This report is submitted
O SMALL BANK Comptroller the Currency Administrator National Banks Washington, DC 20219 PUBLIC DISCLOSURE March 17, 2008 COMMUNITY REINVESTMENT ACT PERFORMANCE EVALUATION Pioneer Trust Bank, National Association
Portfolio Media. Inc. 860 Broadway, 6th Floor New York, NY 10003 www.law360.com Phone: +1 646 783 7100 Fax: +1 646 783 7161 firstname.lastname@example.org CFPB Proposal Would Make 'HMDites' Of Us All Law360,
109 Rates and Race: An Analysis of Racial Disparities in Mortgage Rates Thomas P. Boehm University of Tennessee Knoxville Paul D. Thistle University of Nevada Las Vegas Alan Schlottmann University of Nevada
SMALL BANK Comptroller of the Currency Administrator of National Banks PUBLIC DISCLOSURE November 6, 2002 COMMUNITY REINVESTMENT ACT PERFORMANCE EVALUATION City National Bank Of New Jersey Charter 16142
Mortgage Lending in New England: Key Trends and Observations in 2012 by Kseniya Benderskaya* 1 Community Development Issue Brief 2, 2014 May 2014 Summary Using the annual Home Mortgage Disclosure Act (HMDA)
Fair Lending Laws and Regulations Introduction This overview provides a basic and abbreviated discussion of federal fair lending laws and regulations. It is adapted from the Interagency Policy Statement