Assessing Fair Lending Risks Using Race/Ethnicity Proxies

Abstract

Fair lending analysis of nonmortgage credit products often involves proxying for race/ethnicity since such information is not required to be reported. Using mortgage data, this paper evaluates a series of proxy approaches (geo, surname, geo-surname, and Bayesian Improved Surname Geocoding (BISG)) as compared with the race/ethnicity reported under the Home Mortgage Disclosure Act (HMDA). The BISG proxy predicts the reported race/ethnicity the best as judged by prediction bias, correlation coefficient, and discriminatory power. In assessing fair lending risks where classification of race/ethnicity is called for, we propose the BISG maximum classification, which produces a more accurate estimation of mortgage pricing disparities than the current practices. The above conclusions withhold various robustness tests. Additional analysis is performed to assess the proxies on nonmortgage credits by leveraging consumer credit bureau data. This paper was accepted by Amit Seru, finance.