Modeling company failure: a longitudinal study of Turkish banks |
| |
Authors: | Ozlem Ilk Didem Pekkurnaz Murat Cinko |
| |
Institution: | 1. Faculty of Arts and Sciences, Department of Statistics, Middle East Technical University, Ankara, Turkeyoilk@metu.edu.tr;3. Department of Economics, University of North Carolina at Chapel Hill, Chapel Hill, NC, USA;4. Faculty of Business Administration, Department of Business Administration, Marmara University, Istanbul, Turkey |
| |
Abstract: | Determining the factors related to the financial failure of a company is important. In this paper, we extend literature on bank failure prediction by modelling bank failures in Turkey from 1998 to 2000 using three statistical models combined with a principal component analysis on financial ratios. The three statistical models employed are a logistic regression, a logistic regression that takes serial correlation into account via generalized estimating equations and a marginalized transition model (MTM). Time and financial ratios that are related with capital adequacy and profitability, risk, non-interest income and Fx assets to Fx liabilities are found to be significant in classifying failed banks. Each of our methods achieves a correct classification rate of 93.3%. Among the three models, MTM, which is the soundest model in terms of statistical assumptions, shows slightly better model fit properties. |
| |
Keywords: | uncertainty modelling regression financial ratios hierarchical models panel data |
|
|