An empirical study on early warning systems for banking sector

Boyraz, Mustafa Fatih
Early Warning Systems (EWSs) for banking sectors are used to measure occurrence risks of banking crises, generally observed with a rundown of bank deposits and widespread failures of financial institutions. In countries with a small number of banks, for example Turkey with 48 banks (BDDK, 2011), every bank may be considered to have a systematic importance since the failure of any individual bank may carry a potential threat to lead to a banking crisis. Taking into account this fact the present study focuses on EWSs in Turkey. Since there is no single correct EWS to apply to all cases, in this study, 300 models were constructed and tested to find models as accurate as possible by using a trial-and-error process and by searching optimal feature subset or classifier methods. Empirical results indicate that prediction accuracy did not increase significantly while we got closer to the actual occurrence of bankruptcy. An important finding of the study was that trends of financial ratios were very useful in the prediction of bank failures. Instead of failures as a result of instant shocks, the banks' failures followed through a path: first a downward movement affected the efficiency of the banks' officers and the quality of management structure measured with "Activity Ratios", then the profitability of the banks measured with "Profit Ratios" declined. At last, the performance and the stability of banks' earnings stream measured with "Income-Expenditure Structure Ratios" and the level and quality of the banks' capital base, the end line of defense, measured with "Capital Ratios". At the end of study, we proposed an ensemble model which produced probability ratios for the success rates of the banks. The proposed model achieved a very high success rate for the banks we considered.
Citation Formats
M. F. Boyraz, “An empirical study on early warning systems for banking sector,” M.S. - Master of Science, Middle East Technical University, 2012.