Do algorithms discriminate against African Americans in lending?

Vast evidence exists of discrimination in lending against African Americans. Does such widespread discrimination persist in peer-to-peer lending, in the absence of information about borrowers' ethnicity? To investigate this question, we gather data from a large peer-to-peer lender that uses algorithms and no face-to-face interviews to determine loan approval and conditions. With data from 3.6 million loan applications and 817,000 granted loans during 2016 and 2017, we perform regressions of loan acceptance and loan conditions on the percentage of African American residents by 3-digit zip area. The evidence shows the persistence of discrimination in peer-topeer lending against African American borrowers. In areas with higher proportions of African American residents, the results reveal higher loan rejection, higher loan spread, and shorter loan maturity. Discrimination in peer-topeer lending also affects Native American and Asian borrowers. Thus, the expansion of peer-to-peer lending cannot end discrimination in lending.

Automated summary

Research shows the persistence of discrimination against African American borrowers in peer-to-peer lending, with higher rejection rates, larger loan spreads, and shorter loan maturities in areas with higher proportions of African American residents. Discrimination also affects Native American and Asian borrowers in peer-to-peer lending. The expansion of peer-to-peer lending does not end discrimination in lending.