Mortgage default rates and borrower race

Journal of Real Estate Research, The, Sep/Oct 1999 by Anderson, Richard, VanderHoff, James

Abstract. We estimate a mortgage default model with national data on conventional mortgages that were current from 1986 to 1992. Our analysis confirms the results of previous analyses of Federal Housing Authority mortgages: Black households have higher marginal default rates, controlling for differences in borrower and property characteristics. Further, we do not find that Black borrowers have significantly more home equity. These results do not provide evidence of racial discrimination in mortgage lending and suggest that differences in default costs or transaction costs may explain differences in default rates.

Introduction

This article presents evidence that Black borrowers have higher default rates on conventional residential mortgages than other borrowers. Our results do not support racial discrimination in mortgage lending, consistent with analyzes of Federal Housing Authority (FHA) mortgages. Our confirmation of previous findings is important because of the prominence of the discrimination issue in public policy initiatives and the questions surrounding previous studies of racial differences in mortgage default rates. For decades public policy goals have included increasing minority access to owner occupied housing by eliminating racial discrimination in mortgage lending. Congress passed several laws aimed at eliminating discrimination (e.g., Fair Housing Act (1968); Equal Credit Opportunity Act (1974); Community Reinvestment Act (1977, amended 1989); Home Mortgage Disclosure Act (1975)). Various federal agencies monitor lenders for compliance with lending laws and provide information to the Department of Justice for prosecution. Evanoff and Segal (1997) provide a detailed discussion of policy initiatives.

Recent studies provide not only conflicting evidence of the existence of racial discrimination but also fundamental disagreements on what constitutes evidence of discrimination. Munnel, Browne, McEneaney and Tootell (MBMT) (1996) analyze mortgage application data collected for a study by the Federal Reserve Bank in Boston. They find lower mortgage loan approval rates for minority borrowers and conclude that racial discrimination exists. Home (1997) analyzes a subset of the Boston Fed data and concludes that there is no difference in approval rates among racial groups. Critics of the MBMT approach argue that the lower minority borrower loan approval rates may be the result of higher credit risks and that evidence of discrimination' is manifest in differences in mortgage default rates' and mortgage loan profitability. Using data on FHA mortgages, Berkovec, Canner, Gabriel and Hannan (BCGH) (1994a, b) estimate a default probability model that accounts for characteristics of the property, the neighborhood and the borrowers, including race. Their analysis indicates that minority borrower default rates are higher and minority borrower loans are less profitable, results not suggestive of racial discrimination in the FHA mortgage market.

Both the interpretation and the reliability of the BCGH studies are questioned. Brueckner (1995) and Ferguson and Peters (1995) demonstrate that taken together lower minority borrower default rates and lower minority borrower approval rates indicate discrimination, but that higher minority borrower default rates cannot be interpreted as evidence of the absence of discrimination. Higher minority default rates may result from lower default costs, regardless of whether lenders discriminate or not. Questions about the reliability of the BCGH default rate estimates stem from three sources of bias: (1) omitted variables due to the absence of variables identified in theoretical default models-, (2) simultaneous equations due to the correlation between equity and (unmeasurable) components of default costs; and (3) measurement errors due to the use of variables measured at the time of loan origination instead of at the time of default.'

Our study extends knowledge on default rates in two important ways. First, we analyze conventional mortgage defaults. Some researchers' assume that discrimination in the conventional mortgage markets forces minority borrowers into the government mortgage market although conventional mortgage default rates by racial groups have not been analyzed.' Second, we estimate a default rate model with procedures and data to correct the above biases. The model derives from theories of the optimal exercise of the default option to lessen omitted variable bias. In addition, this model is estimated with an instrument for equity to correct simultaneity bias and with data values contemporaneous to the default decision, not loan origination, to lessen measurement error bias.

Our default probability model indicates that the marginal default rate for Black households' is significantly higher and the difference in default rates is not reduced when controlling for equity, borrower age, borrower education and the number of dependents. Further analysis of house equity does not suggest mortgages issued to minority borrowers are more profitable. These results do not provide evidence of racial discrimination in the conventional mortgage market.