생존분석을 이용한 기업부실예측

Abstract

A survival analysis method that have been extensively used in biology and medical science have advantages, such as no requirement of assumptions from the binary taxonomy model, the ability to provide more information, and better solutions regarding the limitations of existing methodologies by predicting survival period or hazard probability rather than simple probability taxonomy (i.e., bankrupt vs. healthy companies). And this method led us to apply survival analysis in analyzing the manufacturing industry from stock-listed corporations. Nine variables were selected based on the test of t-test: (1) Stockholders' Equity Growth rate; (2) Return on Net Sales; (3) Return on Equity; (4) Financial Expenses to Sales; (5) Owner's Equity Ratio; (6) Debt Ratio; (7) Current Liabilities Ratio; (8) Loan Ratio; and (9) the ratio of operating income of equity. The value of fitness for the model using nine variables is higher than the one using five variables, which indicates that the model with nine variables is better suited to this study. Coefficient regression values of Stockholders' Equity Growth rate, Return on Net Sales, Debt Ratio, Owner's Equity Ratio, the ratio of operating income of equity, Financial Expenses to Sales, and Loan Ratio were all found to be negative values, which indicates a reduced risk of bankrupt in those values, while Return on Equity and Current Liabilities Ratio showed positive values, which indicated a increase risk of bankrupt in those values. Based on Wald values of variables, values of Owner's Equity Ratio and Stockholders' Equity Growth rate are found to be more significant compared with values of other variables. Also, Exp(B) from the equation , which typically represents the relative risk of variables, showed that, in the case of Return on Equity, Exp(B) is 1.015(1.015-1=0.015), indicating that the rate of risk increases by 1.5% when one unit increases. According to the results of the Cox proportional hazard model, the shape of survival function based on survival probability was downward sloped: when ‘0’ month was set at the time of January in 2003, the function showed a rapid drop in October, 2004 ('20' months), and there was another rapid drop between '60' months (December, 2007) and '80' months (August, 2009). The survival probability of companies was found to be about 75% within 80 weeks.