Title: Problems in applying discriminant analysis in credit scoring models
Abstract: Since the mid-1960s financial institutions and other creditors with increasing frequency have applied credit scoring and related loan review procedures to appraise the creditworthiness of loan applicants. The passage of the Equal Credit Opportunity Act and promulgation of the Federal Reserve's Regulation B to implement this act place an important burden on institutions that are subject to the regulation and that employ screening models to ensure that their procedures are statistically and methodologically sound. This paper reviews the types of credit scoring models that have been described in various journals. It gives particular attention to the methodological approaches that have been employed and the statistical problems associated with those models using discriminant analysis techniques. The paper points out that the statistical scoring models discussed in the literature have focused primarily on the minimization of default rates, which is in fact only one dimension of the more general problem of granting credit. To the extent that for the lender profit maximization or cost minimization is, or should be, the objective of a scoring model, then most of the applied literature seems incomplete. The paper also shows that, even ignoring these shortcomings, the models used typically suffer from statistical deficiencies. And it finds that some of the problems of these models seem to be inherent in the discriminant analysis techniques employed or seem to be hard to remedy, given the state of the art concerning estimation and sampling procedures.
Publication Year: 1978
Publication Date: 1978-10-01
Language: en
Type: article
Indexed In: ['crossref']
Access and Citation
Cited By Count: 103
AI Researcher Chatbot
Get quick answers to your questions about the article from our AI researcher chatbot