Applicability of generic linear scoring models in the US credit-union environment

Although credit-scoring models represent a widely used managerial aid for large financial intermediaries, the vast majority of US credit unions – relatively small cooperatively owned retail intermediaries, constrained by sample and funding limitations – have yet to adopt such techniques. Lovie and Lovie have theorized that the flat-maximum effect or curve of insensitivity associated with linear scoring models could be advantageous in areas of applied prediction such as credit scoring. In this context, we reported the relative predictive power of generic credit-scoring models versus customized models in an earlier paper. Unfortunately, these findings were not readily adaptable to the credit-union industry due to a dated sample with incomplete credit-bureau information. Consequently, from 1988 to 1991, we gathered a refined database from which to further develop and field-test generic scoring models in the credit-union environment. The results reported herein not only confirm, but amplify, the relative predictive power of such models found earlier. Relative costs and benefits of generic versus customized models are modelled for a representative credit union. Future research directions are set forth in the conclusions.