Identification of problem banks and binary choice models

Abstract

This paper develops and tests logit and discriminant models that could aid regulatory agencies, as well as bank examiners, investors, analysts and others in identifying the potential failures in the banking industry. As a secondary objective, the paper assesses the comparative abilities of logit and discriminant analyses in distinguishing failed from non-failed banks. The models are validated by two alternate means. Classification accuracies and validation tests indicate that the logit and discriminant models developed are useful in predicting potential failures, and that their weighted efficiencies compare favorably to models developed in the past.