Study On The Credit Risk Of Listed Companies In China Which Has Based On The Bound Logistic

With the rapid development of China’s market economy, many listed companies emerge in the market, and many companies try their best to be listed companies. However, company’s owners will bring incalculable damage to the enterprise even lead to bankruptcy, once they judge mistakely. Bankruptcy, to some degree,could cause serious economic losses for investors and creditors, even affect social stability. Therefore,how to establish credit risk assessment model for national condition by learning the advanced credit risk assessment methods becomes a serious problem.In this paper, the researcher research the credit risk of listed companies in China based on theoretical and empirical aspects.Firstly, the listed company’s credit risk is not only affected by the micro factors also affected by macroeconomic factors. At present, the establishment of a credit risk model is based on the indicators of financial companies.In order to explore the macroeconomic factors also affect listed company’s credit risk,this paper will consider the macro factors to establish the credit model.Secondly, because the micro and macro indicators selected indicators are correlated each other. The Logistic model don’t require collinearity among the variables.so we must process the macro and micro indicators,then apply to the Logistic model.Through testing the relevance and independence,we found a nonlinear relationship between the index,so choose kernel principal component analysis to deal with these indicators. This paper describes the definitions, theorems and calculation steps of kernel principal component from theoretical angle and testify the cumulative contribution rate and the effect of dimensionality reduction of kernel principal component is better than the main ingredients.Thirdly, the kernel principal component analysis applied to Logistic model, and then study the credit risk of listed companies in China. However, on the one hand we found the Logistic model require the large number of sample capacity, on the other hand if we establish the model based on the data which don’t obey the Logistic Distribution, it is difficult to pass the goodness of fit test. To solve these problems, we introduce boundary Logistic model and describe boundary effect if it obey the uniform distribution and normal distribution. Also the paper testify the superiority of boundary Logistic model than tradional model. When the boundary effect obey uniform distribution, the default rate of boundary Logistic Model obey the generalized Logistic distribution, while the default rate obey Logistic Distribution, so the boundary Logistic model is applicable to more data; If the boundary effect obey the normal distribution, the shape of boundary Logistic model becomes obvious "compact", and the gray area is compressed significantly, greatly reduce the probability of false judge. Then this paper also describes the parameter estimation boundary effects test, goodness of fit test and so on.,when boundary effect obey different distributions.The paper selected eight different ST companies and 22 non-ST companies in 2011 and 2012 to test the theory, through parametric estimation, significance tests, goodness of fit test and predicting the probability of default, the paper compare the boundary logistic model with general model. Empirical findings show that if the default rate bound obey the uniform distribution, the default rate of boundary logistic models are significantly dispersed in 0 and 1, the bound logistic model could eliminate the gray area, which could predict the default rate accurely.