Author: Jim Berkovec and Peter Zorn

Start Page: 39
End Page: 56
Volume: 11
Issue Number: 1
Year: 1996
Publication: Journal of Real Estate Research

Abstract: The Home Mortgage Disclosure Act (HMDA) mandates the reporting of mortgage loan applications. Nearly all studies of mortgage lending patterns done in recent years rely on the data collected under HMDA. However, not all mortgages are reported under HMDA. Understanding the relationship between HMDA coverage and neighborhood characteristics is particularly important because neighborhood rates of applications and loan originations from HMDA are used by many analysts to measure neighborhood lending activity. If HMDA coverage rates vary systematically with neighborhood characteristics, then studies that use these neighborhood characteristics to explain lending activity will yield biased results. This paper presents the results of an analysis that attempts to estimate the fraction of mortgage activity that is reported under HMDA, and examines how HMDA coverage rates vary with the racial and income characteristics of the neighborhood. The basic idea for this analysis is simple: identify a group of loans from an independent (non-HMDA) source and count the fraction of those loans that appear in the HMDA data. Our independent source for loans is the loans purchased by Freddie Mac during 1992 and 1993; counts of those loans is compared with counts of loans reported as sold to Freddie Mac in 1992 and 193 HMDA datasets. The major finding of the analysis is that the HMDA dataset for 1992 is estimated to contain only around 70% of the total mortgage loans, and the coverage rate only improves to 75% in 1993, despite the increased reporting requirements for 1993. Both the 1992 and 1993 HMDA files exhibit substantial variability in coverage across census tracts and lenders, but the direction of bias is consistent. Measured HMDA coverage rates are higher in lower income census tracts, relative to higher income tracts.