Propensity score matching: a tool for consumer risk modeling and portfolio underwriting

Area of Science:

• Financial Risk Modeling
• Statistical Methodology
• Consumer Asset Valuation

Background:

• Financial services rely on empirical techniques for risk and value assessment.
• Uncontrolled samples in predicting future performance from past data introduce bias.
• Propensity score matching (PSM) is a statistical method used in medical and social sciences to address non-randomized experimental designs.

Purpose of the Study:

• To demonstrate the practical application of propensity score matching in financial risk modeling and portfolio underwriting.
• To adapt a methodology from social sciences for use in financial contexts.
• To assess the utility of PSM for predicting risk in consumer assets.

Main Methods:

• Application of propensity score matching (PSM) to a portfolio of distressed consumer credit accounts.
• Utilizing PSM to address issues of bias in financial risk assessment.
• Evaluating PSM's effectiveness in scenarios with limited observations and imbalanced covariates.

Main Results:

• Propensity score matching effectively predicts both account-level and portfolio-level risk.
• The study validates PSM as a viable tool for financial risk modeling.
• PSM proves useful in contexts with limited data and numerous potential features.

Conclusions:

• Propensity score matching should be integrated into the methodological toolkit for financial risk modeling and valuation.
• PSM offers a practical solution for underwriting consumer assets, especially with data limitations.
• This approach enhances the reliability of risk assessment in financial portfolios.