중소기업의 부실예측모형에 관한 연구

Abstract

The objective of this study is to build a model to predict small business bankruptcy using logit analysis. The initial sample was composed of 554 firms with 277 firms in each of the two groups. The companies in the bankrupt group are (1) the small manufacturing firms that went bankrupt in a single year, 2003, (2) the firms which had received credit guarantee by S-Fund. The size, in terms of assets, of these firms fell in between 1,000 million won and 7,000 million won. And the size, in sales, of these firms was over 1,000 million won. Non-bankrupt firms were selected through a paired sampling method stratified by industry and size. The study selected companies which were still in existence in 2004. All the financial data were obtained from the electronic data center of S-Fund. Because the large number of financial variables in the previous bankruptcy studies were found significant, the indicators used in the past studies - 32 variables - were involved in evaluation of the model proposed in this study. The variables were classified into six standard categories including liquidity, profitability, activity, productivity, growth, and cash flows. In order to test the efficiency of the model, secondary sample of 410 firms with 205 firms in each for bankrupt and non-bankrupt groups had been chosen, which were collected separately to test the model's efficiency. To derive the final model, the following procedures were taken: (1) observation of statistical significance of each variable; (2) evaluation of correlations among the relevant variables; (3) observation of predicting accuracy of various profiles; and (4) subjective judgment by the authors. From the original 32 financial variables used in the study, four variables below were chosen to be included in the final prediction model for the corporate bankruptcy. The four variables were Liquidation Value/Total Sales, Interest Expenses/Total Sales, Value Added/Total Sales, and a Growth Rate in Total Assets. These variables were found to be statistically significant at 1% or 5% level of significance. The directions of the three variables were come out as we expected. In addition, a dummy variable, indicating proprietorship or corporate, was included in the model and found to be statistically significant at 5% level of significance. All the five variables have been significant at 1% or 5% level of significance. The predictability of the model for the estimation sample of 2003 was a little over of 66% and that for the validation sample of 2004 was 68%. This model can be used for financial institutions to evaluate small firms which apply for credit rating or letter of credit guarantee. The model developed in this study is very valid to predict, because the model's predictability derived from the estimation sample was almost the same as that from the validation sample.