Human vs. Machine: Underwriting Decisions in Finance

Abstract

Using a randomized experiment in the auto lending industry, we provide causal evidence of higher loan profitability and lower default rates with algorithmic machine underwriting, relative to human underwriting. We find that machine-underwritten loans generate 10.2% higher profit than human-underwritten loans in a sample of 140,000 randomly assigned loans. When loans to otherwise identical borrowers are compared, the loans underwritten by machines not only have higher APRs but also sustain a 6.8% lower incidence of default, relative to loans underwritten by humans. The performance gap is more pronounced with more complex loans and at discrete cutoffs. The use of a discontinuous variable space to categorize consumer credit profiles by human underwriters is associated with a 40.2% increase in default rates and 24.7% less profit. These results are consistent with findings on the human mind's limited capacity for analyzing complex problems and with agency conflicts in the underwriting process.

Keywords: Lending, Underwriting, FinTech, Automation, Household Finance

JEL Classification: O33, G21, D14