AN EMPIRICAL STUDY TO MODEL CORPORATE FAILURES IN TURKEY: A MODEL PROPOSAL USING MULTIVARIATE ADAPTIVE REGRESSION SPLINES (MARS)

Authors : Mehmet Sabri TOPAK

Pages : 1-21

View : 22 | Download : 5

Publication Date : 2011-12-01

Article Type : Research Paper

Abstract :This paper is prepared to model financial distress cases in Turkey using a non-parametric technique, Multivariate Adaptive Regression Splines insert ignore into journalissuearticles values(MARS);. For this purpose, a sample of 114 firms with 665 annual observations between the years 1994 and 2003 was used to predict financial distress for one year prior failure. Our modeling study on 41 independent variables, 39 financial data-based and 2 non-financial, has resulted in a condensed model including 10 basis functions based on 8 original variables. The final model has an overall rate of correct classification of 81,8 % and is proved to be significantly superior to a naïve model. Its Type I and Type II performances are respectively 91,5 % and 80,9 %. Furthermore, profitability performance, capital structure decisions, and macroeconomic conditions are found to be the major determinants that influence Turkish firms’ risk profiles.
Keywords : Financial Distress Prediction, Non Parametric Modeling, Multivariate Adaptive Regression Splines, Turkey