FinTech Lending to Borrowers with No Credit History

Abstract

Despite the promise of FinTech lending to expand access to credit to populations without a formal credit history, FinTech lenders primarily lend to applicants with a formal credit history and rely on conventional credit bureau scores as an input to their algorithms. Using data from a large FinTech lender in Mexico, we show that alternative data from digital transactions through a delivery app are effective at predicting creditworthiness for borrowers with no credit history. We also show that segmenting our machine learning model by gender can improve credit allocation fairness without a substantive effect on the model’s predictive performance.

Presented by Paolina Medina.