Prediction of financial failure using multivariate statistical methods and a model based on multicriteria decision analysis: An application to the Turkish banking system

Ill ABSTRACT PREDICTION OF FINANCIAL FAILURE USING MULTIVARIATE STATISTICAL METHODS AND A MODEL BASED ON MULTICRITERIA DECISION ANALYSIS: AN APPLICATION TO THE TURKISH BANKING SYSTEM Süleyman Bilgin KILIÇ P.D. Thesis, Department of Management Supervisor: Prof.Dr. Serpil CANBAŞ February 2003, 140 Pages The main objective of this study is to estimate the early warning systems, for the prediction of commercial bank failures in the Turkish banking system which is the most important component of the Turkish financial sector, based on the multivariate statistical methods and the multicriteria decision analysis. The sample set of the study contains 40 privately owned commercial banks in the Turkish banking system of which 20 banks failured during the period 1997- 2002. The Banking Association Institution in Turkey publicities 49 financial ratios for the banks in its web site. Initially, the univariate ANOVA (Analysis of Variance) test were applied to the 49 ratios, and 12 ratios determined as the early warning indicators which were best separate healthy and failed banks in the -l.year (one year before the failure). Secondly, by applying principal component factor analysis, these ratios were grouped under the meaningful factors, and the 3 common factors (capital adequacy, income and expenses, and liquidity) were determined which represent basic financial characteristics of the Turkish banking system. After then, for each of the bank, the factor scores were calculated and these scores were used as independent variables to estimate discriminant, logit and probit models. Also, as a new approach to the financial failure, the ELECTRE TRI models of the multicriteria decision analysis were estimated by using 12 early warning ratios determined previously. Banks wereIV classified according to estimated models and results compared with respect to classification achievements of the models. Finally, the importance of application of early warning systems was evaluated with respect to bank failure costs, which is the cost of restructuring financial system in Turkey. The results of the study show that, by undertaking early warning systems in the Turkish banking sector, it is possible to avoid from the restructuring costs at a significantly amount of rate in the long run. Key Words: Factor Analysis, Disciriminant, Logit, Probit, ELRCTRE TRI.