New directions in predicting bank failures: The case of small banks

Abstract

This paper uses five financial accounting ratios with three alternative loan-portfolio diversification measures to classify failures among small commercial banks that occurred during 1984. Classifications for one, two and three years before failure are performed using linear probability, logit, probit, and discriminant analysis models. Validation is done through the U-Method. The results indicate that the logit and probit functional forms may offer an advantage over the more frequently used discriminant analysis. U-Method classification accuracy is approximately 86 percent for the logit and probit models.