초록 열기/닫기 버튼
The study of a failure prediction has been studying for a long time. But, practical models which can be used in actual affairs have not been developed. After the nineties, corporate failures which are radically increased have resulted in serious losses in society. The efforts to predict corporate failure are very important because they help decrease the social losses. The main purpose of this study is to develop a failure prediction model which use the ROE(return on equity)ratio and the stability variable of the ROE ratio. The secondary purpose of this study is to examine the choice based sample bias by comparing the result of a 1:3 model with the result of a 1:1 model. The sample of this study consists of 135 failed firms which had been delisted from The Korea Stock Exchanges during the year of 1991 and 1998, and 254 nonfailed firms matched on the basis of a industry classification and fiscal year. To examine predictive ability, this study splits a samples into two groups ; an estimate sample and a holdout sample. This study uses logistic regression analysis on the five year ROE ratio data of sample firms. The findings of the empirical study can be summarized as follows : First, because predictive ability is not only more than 70% in all periods but also stable in the holdout sample, ROE ratio may be used to develop useful prediction models. Second, predictive ability of expanded model which uses stability variables of the standard deviations over the past 4 years is about 10% higher than the reduced model. Therefore, stability variable is useful variable. Third, as compared a 1:3 model with a 1:1 model, there is considerable upward-bias in 1:1 model. Fourth, net worth to total assets is the most confident variable.
키워드열기/닫기 버튼
ROE(return on equity), Corporate Failure, Logistic Regression
