Unwarranted Gender Disparity in Online P2P Lending: Evidence of Affirmative Action

Closing the gender gap in financial access is important. Most research tends to empirically uncover the direct effect of gender on decisions. Yet, this view overlooks other indirect channels of gender discrimination, leading to systemic bias in identifying the overall discrimination effect. In this work, by collaborating with one of the largest online P2P lending platforms in China, we estimate a broadened discrimination notion called unwarranted gender disparity (UGD). UGD recognizes any disparate lending decisions that do not commensurate with the loan's return rate, encompassing direct, indirect, and proxy discrimination. We develop a two-stage predictor substitution (2SPS) approach to estimate UGD. Somewhat surprisingly, we find significant female favoritism at almost all return rate levels. On average, female borrowers are 3.97% more likely to be funded than male borrowers with identical return rates. We further decompose and find at least 37.1% of UGD is indeed indirect or proxy discrimination. However, we also identify the observed UGD favoring female can be completely attributed to \emph{accurate statistical distribution}, which is rationalized by women being less likely to default on their P2P loans. Our results suggest that online P2P lending can complement traditional bank lending in closing the gender gap, by providing an alternative credit market where the affirmative action to support women can arise naturally from the rational crowd.