Pricing of Home Mortgage Loans to Minority Borrowers: How Much of the APR Differential Can We Explain?, The

Journal of Real Estate Research, The, Oct-Dec 2007 by Courchane, Marsha J

Abstract

The public releases of the 2004 and 2005 HMDA data have engendered a lively debate over the pricing of mortgage credit and its implications regarding the treatment of minority mortgage borrowers. This research uses aggregated proprietary data provided by lenders and an endogenous switching regression model to estimate the probability of taking out a subprime mortgage, and annual percentage rate (APR) conditional on getting either a subprime or prime mortgage. The findings reveal that up to 90% of the African American APR gap, and 85% of the Hispanic APR gap, is attributable to observable differences in underwriting, costing, and market factors that appropriately explain mortgage pricing differentials. Although any potential discrimination is problematic and should be addressed, the analysis suggests that little of the aggregate differences in APRs paid by minority and non-minority borrowers are appropriately attributed to differential treatment.

Home Mortgage Disclosure Act (HMDA) reporting requirements were changed in 2004 to require lenders for the first time to provide information on the difference between the annual percentage rate (APR) and a comparable Treasury rate for all loans above a rate spread-reporting tiireshold.1 The Federal Reserve Board publicly released tiiese initial data in September 2005. Similar data for loans originated in 2005 were released in September 2006.

The 2004 and 2005 HMDA data show tiiat loans to minority borrowers have a substantially higher probability of exceeding the rate-spread reporting hurdle, suggesting that, on average, minority borrowers pay higher APRs for their mortgages than do non-minority borrowers. This observation has engendered a lively debate over its possible implications regarding differential treatment.2 Partly on the basis of these data, regulatory agencies have referred several pricing cases to the Department of Justice. In a separate action, Countrywide, one of the nation's largest lenders, reached a settlement with the New York Office of the Attorney General based on investigation of Countrywide's pricing practices.3

This paper provides a unique perspective on this debate because of the access to data not generally available to the public. Specifically, a number of lenders have allowed pooling of their records into an anonymous data set that includes detailed loan-level information on their 2004 and 2005 originations. These data allow exploration of the pricing of mortgage loans across minority and non-minority borrowers at a deeper level of detail than is possible with HMDA data alone. In particular, the data allow rough approximation of the work of regulators as they analyze apparently unexplained pricing differentials uncovered in the HMDA data.

Regulators use HMDA data to allocate efficiently their investigative activities-attention is focused typically on lenders with relatively larger pricing differentials that cannot be fully explained by HMDA data alone. However, once lenders are targeted for additional reviews, regulators significantly modify how they conduct their investigations. First, rather than restricting their attention to rate-spreads for loans above the rate-spread reporting hurdle, regulators explicitly focus on mortgage prices (generally APRs) for all loans.4 Second, regulators dramatically increase the control variables used in their statistical analyses, including variables used in underwriting loans, as well as variables used to account for market factors that can affect pricing outcomes.

The data uniquely allow approximation of this second-level analysis of regulators, albeit at a national level using pooled data rather than separately for individual lenders.5 Specifically, the data include APR, typical underwriting variables such as loan-to-value ratios (LTV), debt-to-income ratios (DTI), and FICO scores, as well as other variables that likely affect mortgage pricing, such as documentation status, the existence of prepayment penalties, and sourcing channel. Identification of whether the lenders in the sample operate primarily in the subprime or prime markets is also possible.6

The data show that there are large pricing differentials to explain. For example, the difference between the mean APR of African American borrowers and the mean APR of White non-Hispanic borrowers is 120 basis points in 2004 and 128 basis points in 2005, and the equivalent difference between Hispanic and White non-Hispanic borrowers is 56 basis points in 2004 and 74 basis points in 2005. Not surprisingly, APRs are substantially higher in the subprime market, and much of the explanation for why minority borrowers tend to have higher APRs than non-minority borrowers is because minority borrowers disproportionately take out subprime loans.

An endogenous switching regression model is used to estimate APR separately in the prime and subprime markets because the supply and, arguably, the demand functions in the two markets are quite different.7 A full set of explanatory variables is used to largely explain the differential tendency of minority borrowers to obtain subprime loans. Up to all but about 10 basis points of the difference in mean APRs between African American, Hispanic, and White non-Hispanic borrowers can also be explained.