Did fintech loans default more during the Covid-19 pandemic? Were fintech firms “cream-skimming†the best borrowers?

A growing portion of consumer credit has been devoted to unsecured personal installment loans. Fintech firms have been active players in this market, with an increasing market share, while the market share of banks has declined. This paper uses recent loan data, including that from the Covid-19 pandemic period, and finds that a proprietary loan rating system that leverages alternative data and artificial intelligence and/or machine learning could both disaggregate the risk beyond those predicted by a traditional credit risk model and identify the “invisible prime†consumers from the nonprime pool. This loan rating system continued to perform well during the pandemic despite the unexpected adverse economic shock that affected consumers unevenly. It contained different information than the traditional measure of credit risk, and the inclusion of both measures in the regression analysis improved the fit of the model. We also show that default rates for nonprime consumers are much lower on LendingClub installment loans than on traditional consumer loans during the pandemic period. In exploring the potential cream-skimming, we find that there is an additional factor that is correlated with default but uncorrelated with credit scores, alternative ratings or economic factors. This unique factor, which is not captured by the typical credit risk model but is important in determining a loan’s default probability, is likely responsible for some believing in the occurrence of cream-skimming behavior.