Artificial Neural Network Models to Build an Early Warning System for Turkish Commercial Banks before and after the 2001 Financial Crisis

Abstract

The objective of this paper is to measure the failure risk of Turkish commercial banks. Bank failures bring to bear high costs on economies as well as on governments and eventually on the public and the taxpayers. During the past two decades, many developed and developing economies have experienced large scale bank failures, and estimates for average bank restructuring costs range from 6% to 10% of the Gross Domestic Product. In Turkey the amount of restructuring costs is approximately 30% of the Gross Domestic Product. In this study, we use 29 selected financial ratios of banks across 1996-2012 periods and the Artificial Neural Network Models to build an early warning system. If commercial bank failure were a predictable event, bank restructuring costs could be minimized. Additionally, if early warning systems are used effectively, the regulatory actions necessary to prevent banks from failing could be taken in advance or in the least a more orderly process of bank closures could be administered. The results overall indicate that almost all commercial banks currently operating in the Turkish banking sector are quite sound and far from failure.