Regulatory Arbitrage or Random Errors? Implications of Race Prediction Algorithms in Fair Lending Analysis
When race is not directly observed, regulators and analysts commonly predict it using algorithms based on last name and address. In small business lending—where regulators assess compliance with fair lending laws using the Bayesian Improved Surname Geocoding (BISG) algorithm—we document large prediction errors among Black Americans.
The errors bias measured racial disparities in loan approval rates downward by 40%, with greater bias for traditional vs. fintech lenders. Since errors correlate with socioeconomic characteristics, basing regulation on self-identified race would increase lending to Black borrowers, but also shift lending toward affluent areas.
Our results highlight systematic problems with policies based on race proxies.
About the Law and Economics Workshop
Michigan's Law and Economics Workshop provides an opportunity for faculty and students from across the University to engage with cutting-edge law and economics research by leading scholars on a wide range of legal and policy topics.
Professors J.J. Prescott (jprescott@umich.edu) and Ed Fox (edfox@umich.edu) organize the workshop. If you would like to receive workshop announcements, please contact Alex Wroble (arwroble@umich.edu) and ask to have your name added to the workshop’s email list.
