Case Study

Statistical Support to the Fair Housing and Equal Opportunity’s Office of Systemic Investigations at HUD

Challenge: The Fair Housing and Equal Opportunity’s Office of Systemic Investigations at the U.S. Department of Housing and Urban Development (HUD) is charged with identifying lenders, servicers, and/or brokers in residential real estate transactions that potentially violate the Fair Housing Act by discriminating in pricing and/or underwriting residential mortgages. Such types of behavior include incidences of high-cost loans, differences in borrowers' interest rates, and disparities in denial rates among different classes of borrowers. When cases of potential discrimination are found, HUD is tasked with investigating individual instances of discriminatory practices.

Solution: Summit utilized advanced statistical techniques to screen residential-mortgage data (HMDA) provided to HUD for disparities in denial rates among protected classes of individuals. For investigations of particular lenders, Summit analyzed lender-provided loan-application data and specified a set of econometric models meant to mimic the underwriting guidelines for particular loan programs. These models indicated whether race has a systematic impact on the underwriting decision, when taking into consideration borrower characteristics such as creditworthiness, income, and other factors. In addition, for lenders currently under investigation, Summit used statistical-matching techniques including Mahalanobis Distance Matching (MDM) and Coarsened Exact Matching (CEM) to identify particular loan applicants for further review by HUD investigators.