Machine Learning-Based Profit Modeling for Credit Card Underwriting - Implications for Credit Risk

54 Pages Posted: 1 Jul 2020 Last revised: 1 Feb 2023

See all articles by George Krivorotov

George Krivorotov

Government of the United States of America - Office of the Comptroller of the Currency (OCC)

Date Written: June 6, 2020

Abstract

Retail credit issuers traditionally assign credit based on cutoffs from risk-based models. However, in recent years, advances in technology such as AI/ML have given rise to more models that predict more complicated facets of customer behavior, such as projected NPV. These can be used to precisely target profitable but risky customers. Using a unique regulatory panel dataset of credit cards combining data from many major banks, I construct both traditional risk and ML-based profit models and find that profit score cutoffs generally target wealthy, high-spending, "revolving" customers, while risk score cutoffs target low-activity "transacting" customers. Conducting simulations using both types of cutoffs, I find that, absent risk guardrails, profit-based underwriting could potentially cause an increase in riskiness in card portfolios. However, this is highly portfolio dependent and may only occur in those that concentrate on "revolvers" in the lower end of the credit spectrum.

Keywords: Machine Learning, Credit Risk, Credit Cards, Consumer Finance, Profit Models

JEL Classification: C55, C53, G21, G17, D12

Suggested Citation

Krivorotov, George, Machine Learning-Based Profit Modeling for Credit Card Underwriting - Implications for Credit Risk (June 6, 2020). Journal of Banking and Finance, Vol. 149, No. 106785, 2023, Available at SSRN: https://ssrn.com/abstract=3620895 or http://dx.doi.org/10.2139/ssrn.3620895

George Krivorotov (Contact Author)

Government of the United States of America - Office of the Comptroller of the Currency (OCC) ( email )

400 7th Street SW
Washington, DC 20219
United States

Do you have negative results from your research you’d like to share?

Paper statistics

Downloads
599
Abstract Views
1,575
Rank
83,515
PlumX Metrics