The Empirical Analysis Of The Probability Of Default Of China's Listed Company On The Basis Of Logistic Model And KMV Model

In order to improve the vulnerability of banking system concealed in the financial cysis and to raise the quality and liquidity of the bank's capital,'the Basel Capital AccordⅢ' was officially released in September 2010. National banks have increased their capital regulatory standards. As the main business performance of the bank is consisted of the principal and interest of the corporate lending and the ability to recover the loan, the credit risk is the greatest risk of the commercial banks. When calculating the risk-weighted assets of credit risk by means of Internal Ratings Based Approaches (IRB), banks are needed to consider the four main indicators, that is, the company's default probability, loss given default, exposure at default and the exposure of maturity, which is required in the Basel Accord. Therefore, it's necessary to research the company's default probability.There are three models to measure the default rates of listed companies in this paper. The samples are composed of 1020 Shanghai and Shenzhen A-share listed companies whose financial data from 2007 to 2010 is needed to analysis. First, in the process of two-classified Logistic model, we indicate whether a company defaults judging from whether it is labeled as ST or *ST. and use 26 financial indicators obtained from time -weighted average from 2007 to 2009 to get eight main components by means of factor analysis. Then through stepwise regression, we come to a conclusion that shareholder profitability factor, operating capacity factor, cash flow factor and industry sectors have a significant impact on default rates. At the same time, the model has the strongest predictive power when the cut ratio of default is 0.1552 through ROC curve. And then it's proved by using of financial indicators for 2010 to predict default rates.Secondly, we use the overall investment ratings marked by various research institutions as a measurement of default classification of corporate. Then through regression by SAS software, we draw a conclusion that profitability factor, shareholder profitability factor, long-term solvency factor, operating capacity factor, and industry sectors have a significant impact on default rates.Finally, in the KMV model, using part of the data to calculate the theoretical default rates, we can conclude that there is showing a same cycle and trend of the three sectors from the fourth quarter 2000 to third quarter of 2010. Meanwhile, with the default distance increasing, the theoretical default rates are reduced accordingly. When considering the same default distance, it's satisfied, whether in manufacturing, wholesale or retail real estate, that default rates of t-distribution are greater than those of the general extreme-value distribution which are greater than the value of normal distribution. Therefore, normal distribution underestimates the default rates of sample. Because there is a negative correlation between default distance and default rates and we couldn't obtain the data of actual default rates, we use the weighted default distance from 2007 to 2009 to make stepwise regression and generalized least squares regression to eight principal components. It's to be found that shareholder profitability factor, long-term solvency factor, and cash flow factor have a significant impact on default rates. Nevertheless the sector is not relevant with default distance. When adding the random effects of industry category to the model, we find that there is no industry-assemble between the variation of default distance.By comparing the three models above, we can see that corporate probability of default is closely related to their own financial ratios. Although the significance indicators of each model are different, default rate is mainly affected by profitability, shareholder accrual, long-term solvency, operating capacity and cash flow. Logistic model is to better reflect the impact of industry sectors on default rates, while KMV model can not. Obviously, Logistic model have a stronger sensitivity of macroeconomic and industry sectors than KMV model. Meanwhile, how to map the default distance to the actual default rates is open to question in research of KMV model.Considering the comparative advantages of each software, SPSS 18 is selected to make the two-classified Logistic regression and the random effect analysis of KMV model. The ordered multinomial Logistic regression analysis is mainly by means of SAS9.2. We use Matlab2010 to solve the nonlinear equations group in the KMV model. And finally, Eviews5 is used for analyzing regression of default distance.