A Statistical Model for Credit Scoring

Abstract

We derive a model for consumer loan default and credit card expenditure.The default model is based on statistical models for discrete choice, incontrast to the usual procedure of linear discriminant analysis. Themodel is then extended to incorporate the default probability in a modelof expected profit. The technique is applied to a large sample ofapplications and expenditure from a major credit card company. Thenature of the data mandates the use of models of sample selection forestimation. The empirical model for expected profit produces an optimalacceptance rate for card applications which is far higher than theobserved rate used by the credit card vendor based on the discriminantanalysis. I am grateful to Terry Seaks for valuable comments on anearlier draft of this paper and to Jingbin Cao for his able researchassistance. The provider of the data and support for this project hasrequested anonymity, so I must thank them as such. Their help andsupport are gratefully acknowledged. Participants in the appliedeconometrics workshop at New York University also provided useful commentary.