Predicting US commercial bank failures via a multicriteria approach

Abstract

This paper examines the problem of bank failure and proceeds to the development of bank failure prediction models based on a multicriteria decision technique, namely UTilites Additives DIScriminantes (UTADIS). A sample data of US banks for the years 1993?2003 is used to develop a model that discriminates between the failed and non-failed banks. Although many studies have been developed examining the problem of bank failure prediction, it is the first time that a multicriteria technique is used in this specific field. The obtained results are quite efficient for the evaluation of bank failure, providing errors of 20% grade for at least four years before the failure. Moreover, the results indicate that the multicriteria approach UTADIS outperforms the traditional multivariate data analysis techniques such as discriminant analysis, supporting the fact that the multicriteria techniques could be used efficiently for the bank failure prediction problem.