UC Berkeley Library

In this paper, we suggest that financial performance of loans and loan portfolios, rather than statistical performance of scores, should be the critical focus for lenders and lending institutions using credit scoring to fund and acquire new borrowers. We propose two measures for measuring relative financial performance of loan accounts when portfolio acquisition decisions are based on risk assessments. These compare with numerous statistical performance measures for response and risk scores that assess borrower response to offers, default, bankruptcy, late payment and fraud. Popular statistical measures often used in scorecard development and validation testing include the K-S statistic, Gini coefficient, AUROC, ROC curves, divergence, and others. The measures that we suggest focus on relative return on equity (ROE) and market share; in our opinion, such financial performance measures are more meaningful to lenders and businesses than statistical measures of scorecard performance.