Comparative Research On Models Of Financial Distress Warning In Chinese Listed Companies

With the rapid development of China's capital market, the number of the listed company is increasing, and they are more important to the companies and social economy. The situation of the listed companies will directly affect related interests of investors and creditors, even the entire society and economy's development. So the predicting before the bankruptcy of listed company has an important theory and realistic significance.Owing to earning manipulation and time-lag of financial report disclosure, the authenticity and timeliness of financial information are limited. Thus, the models that financial ratios included only can't make an accurate predicting. The study will incorporate the corporate governance variables into the investigation of financial distress, while using several models for a comparable research. The purpose of the paper is to discover the best index system and method in forecasting financial distress, and provide stakeholders a helpful decision-making aid tool.This paper presented 92 firstly ST companies and 92 Not-ST companies in China from 2006 to 2007, used the annual report of listed companies to disclose information and data, selected 20 financial indicators and 7 corporate governance variables, used Normality tests and Equal Value Difference test to select significant variables, used factor analysis and correlation test to sieve financial indicators and corporate governance variables, and built two index systems-financial index system and index system incorporating corporate governance variables. Then, the paper used Fisher discriminant, Logistic regression and Support vector machine to create some multi-variable financial distress prediction models, and compared the three models from the perspective of theory and practice. The research finds that the correct results of these models based on aforementioned two index systems are 75.54%,89.67%, 99.46%, and 77.17%,90.22%,100%. On the whole, the effect of SVM is the best, and its 1 type error rate is the lowest. And the correct result of the models based on the index system incorporating corporate governance variables are more accurate than the models based on financial index system.