Study On Credit Risk Of China's Credit Debt Market Based On Logit Model

Since the reform and opening up,the bond market in our country has been continuously developed and improved.With the support of various government policies,the size of stock and the speed of development in various types of bond markets are at the forefront of the world.With the continuous expansion of the stock of bonds,The problem existed in various types of bonds is gradually emerging,the brunt of which is the bond default problem.Before 2014,no default had been found in the development of China's bond market.Since 2014,China's economy has enjoyed sustained and rapid growth.However,a bond market with no default on bonds is definitely not a sustainable bond market.In reviewing the possible existing default bonds in the bond market,the government often transports blood to the issuer through implicit administrative measures or the government intervents the constructing between issuers to ensure the stability of the entire bond market.Since China's economy entered “new normal”,chinese economic growth rate has dropped continuously.Supply-side reform has been continuously promoted.The implicit "rigid payment",previously maintained by administrative means,has been gradually broken.From the beginning of 2015 to the end of 2016,a series of instances of bond defaults,the voice between conmunity is becoming louder and louder on sound bound market,Therefore,the research on the credit risk of the entire bond market is particularly important and of great significance.This paper first analyzes the process of bond default,and then combines credit risk related theory at home and abroad,and establishes a credit risk model to predict bond credit risk.In analyzing the process of bond default,it is found that the bond defaulting mainly goes through three stages,the sprouting,accumulation,and appearance of the risk are finally reflected in the change of credit rating.After combing the relevant credit risk models at home and abroad,a risk model suitable for the actual situation in China is established by combining the risk changes in the process of the bond default in China.Taking the defaulted bond as the default sample,according to the scale of the defaulting entity and the industry in which the entity is selected,the normal existence of the issuer's bond is selected as the normal sample.Refer to domestic and foreign research and bond default process.A total of 12 financial indicators at the end of the first two periods of the default time are selected.Two non-financial indicators were established respectively.The firstphase time interval was one year before the default and the t-year model was the first period.The second period was the t-1 year model.The t-test and nonparametric rank sum test were used to screen the two models.The indicators with significant differences between normal and default samples are used as modeling variables.From the screening results,the closer to the time of default,the more variables with significant differences are mainly reflected in cash flow indicators,indicating that the closer they are to breach of contract.The bigger the cash flow gap.The logit model was established by the selected variables.By analyzing the models established at different time points,the rating changes,and the size of the net assets has a good discriminative effect on early warning at different modeling time points.In the t-1 year model,The capital structure has discriminative effects on credit risk.In the t year model,it pays more attention to the industry factors caused by cash flow and policy factors;the t year model and the t-1 year model are analyzed,and the t-1 year model is the second.The probability of class errors is high,and the overall discriminant effect is not ideal.The second type of error probability in the t year model is significantly reduced to 18.7%,and the overall false positive rate is 5.5%.Finally,the t and t-1 models are predicted through the ROC curve.According to force analysis,the area under the ROC curve rose from 0.84 in year t-1 to 0.98 in year t,indicating that the predictive power increased gradually;at the same time,as the timing of the breach was approaching,the overall false positive rate and probability of committing the second type of error were substantially higher.Decreased,discriminative efficiency increased,and predictive power increased.