Discriminant Analysis And Logistic Model In The Listed Companies' Financial Distress Prediction

In Chinese developing but not mature security market, whether thelisted companies are healthy or not, will affect related interests ofinvestors and creditors. The research of their financial situation is of grateimportance. The financial predicament predicts that is regarded as animportant respect of the early warning system of the financial affairs, caneffectively play a monitoring role to company's financial situation. Then,through perfect economic forecasting method to set up effective financialaffairs early warning system seems urgent day by day.This paper presented 108 ST companies and 108 Not-ST companiesin China from 2003 to 2005, used the annual report of listed companies todisclose financial information and data, selected 38 financial indicators,used stepwise regression method, selected significant predictors variables,used logistic regression methods, discriminant analysis and principalcomponent analysis and other methods, used SPSS statistical analysissoftware, created some multi-variable financial distress prediction models.Did comprehensive analysis and evaluation of the results of empiricalanalysis, got a review of some problems. Research found that the logisticand discriminant analysis can be accurately predicted financial distress oflisted companies. Discriminant analysis and multiple logistic regressionforecast are better than single variable. Discriminant analysis is closer tomultiple logistic regressions. The multiple logistic regression is slightlybetter than discriminant analysis; Through principal component analysisfound that more recent projections from two years and three years ago,Logistic regression analysis of the general prediction is effective thanPCA logistic regression analysis; As predicted from the distant years, likefour years and five years ago, the PCA logistic regression analysispredicted better than general logistic regression analysis. But PCAdiscriminant analysis is not better than general discriminant analysis.Normally, whether multivariate discriminant analysis or logisticregression analysis, the results show:Ⅰtype error rate is higher than theⅡtype error rate. But with the earlier forecast data,Ⅰtype error rate is increasing.