Summary
This document is a report that summarizes the analysis of Upstart's personal loan underwriting model in comparison to a traditional model. The report outlines the findings from the 2023 research, which indicates that Upstart's AI model approves a significantly higher number of applicants, including Black and Hispanic borrowers, while offering lower average APRs. Specifically, the Upstart model approves 44% more applicants overall, with APRs that are 36% lower than those of the traditional model. The report details the construction of the traditional model, which is based on a Cox Proportional Hazard modeling framework that incorporates six key inputs, including FICO Score and Debt-to-Income Ratio. Additionally, it explains the methodology used to ensure a fair comparison between the two models, including the simulation of approval and pricing decisions for applicants in 2022. The report aims to enhance understanding of Upstart's efforts to improve access to affordable credit.
