Searching for an Optimal Strategy for Identifying Files to Review for Fair Lending Exams (WP 2005-3)

Abstract

A manual review of applications is an important component of statistically modeled fair lending exams. How files to review are identified affect both resource allocation and reliability of conclusions. This study uses Monte Carlo simulation to compare how six outlier identification strategies perform at identifying disadvantaged applicants. The results show that the optimal strategy for minimizing cost and maximizing reliability of conclusions depends on the likelihood and severity of disadvantage. Further, none of the strategies are highly successful at identifying disadvantaged applicants or minimizing the number of non-disadvantaged applicants reviewed.

Author

Jason Dietrich