Consumer Lending Discrimination in the FinTech Era
50 Pages Posted: 6 Nov 2017 Last revised: 8 Dec 2017
Date Written: December 7, 2017
Lending discrimination can stem from loan officer facial biases or algorithmic scoring, especially with big-data use in FinTech. Using never-before-linked mortgage data covering loan-level ethnicity, scoring variables, contract terms, and lender identifiers, we implement a treatment-based Oaxaca-Blinder discrimination estimation, based on the unique default-risk setting of the GSEs. We find that African-American and Hispanic borrowers have a 5% higher loan rejection rate, especially among low-credit-score applicants. Consistent with facial biases, rejection-rate differences are less pronounced for FinTech lenders, for whom the minority rejection rates are 1% lower. Ethnic-minority borrowers pay a 0.08% higher interest rate for purchase mortgages and a 0.03% higher rate for refinance mortgages, which translates into almost half a billion dollars per year in extra interest payments. FinTech lenders charge only a 0.01% differential for refi loans to African-American and Hispanic borrowers. These latter results are consistent with profit-taking opportunities in weaker competitive environments.
Keywords: Discrimination, FinTech, Mortgages, Credit Scoring, Algorithmic Underwriting, Big Data Lending, Platform Loans, Disparate Impact
JEL Classification: G21, G28, G23, J14, K22, K23, R30
Suggested Citation: Suggested Citation