Study Overview
This research explores the welfare consequences of data-based targeted pricing in the consumer loan market. We develop a general model of optimal loan pricing. Intuitively, loans should be more expensive for consumers with high demands, high default risk, and with demand and default that do not significantly vary with interest rates. We first apply our model to the pricing of credit scores. We use experimental data from a large European bank to estimate how loan take-up and repayment vary across borrowers with different credit scores. This analysis allows us to better understand the shortcomings of risk-based pricing models used in the industry. We then combine machine learning and experimental data to determine the important consumer characteristics that lenders should consider when pricing loans. Our algorithm partition consumers based on characteristics estimated to maximize either firm profits or consumer welfare under a zero-profit condition. This analysis allows us to better understand which consumer characteristics matter most for predicting both demand elasticities (how demand responds to variations in interest rates) and cost elasticities (how expected repayments respond to these shocks – a measure of adverse selection). Our quantitative analysis reveals how moving from a standard risk-based pricing scheme to personalized pricing can affect lender’s profit and consumer welfare.
Study Results
Pending
Intervention: Randomized variations in interest rates for consumer loans
Populations: Middle income households
IBSI Funding Acknowledgement: Lab for Inclusive FinTech (LIFT)