Evaluation of bankruptcy prediction model, using both accounting and market information, in Korea
Inro Lee
Dongcheol Kim
This paper evaluates out-of-sample prediction power of bankruptcy prediction models, accounting-based model, market-based model, and hazard model using both accounting and market information. We suggest two new accounting-based models by using multivariate discriminant analysis and logit analysis. Merton(1974)¡¯s distance-to-default(DD) model is employed as market-based-model. We also provide two hazard models. One is obtained by re-estimating coefficients of frequently used U.S. hazard model with Korean data. The other hazard model is built by modifying aforementioned hazard model for Korean firms. Therefore, we compare the predictability of these five bankruptcy prediction models. We assess the performance of these models by calculating hit-ratio, drawing ROC(receiver operating characteristic) curve, and analyzing information contents test. All three different approaches show that modified hazard model for Korean firms is superior to the others. Therefore, we recommend our modified hazard model for Korean firms as a bankruptcy prediction model.
Bankruptcy prediction model,Accounting-based model,Market-based model,Hazard model,Evaluation of the out-of-sample prediction power