Measurement Of Default Risk Of Unlisted Corporate Bonds Based On Variable Selection Model

At present,China's financial industry has entered a deep-water period of re-form.With the development of China's gold in recent years,the scale of China's bond market has also accelerated expansion.But along with it,the default events of our country's credit bonds occur frequently,and there are successive default events,which have a great impact on the stable development of the credit bond market and the financial market of our country.Therefore,it is of great theo-retical and practical significance for the stable and healthy development of bond market and financial market to form a perfect system for monitoring,predicting and controlling the default risk of credit bonds.In this paper,the listed companies of China's whole market credit bonds that default for the first time from 2016 to 2018 are taken as samples,and the sam-ple default distance calculated by KMV model is taken as dependent variable.Through the analysis of default causes and the monitoring indicators of CBRC,multiple dimensional indicators including the company's asset situation,business situation,debt paying ability and leverage level are constructed as independent variables,respectively through Ling Regression,Lasso regression and adaptive regression models were used to select variables,and the results of model selection and goodness of fit test were compared to select the best.Since the results of goodness of fit are similar,but considering the compressed variable characteristics of the adaptive lasso model,useful information is lost in the calculation process of the model.Therefore,after variable selection,this paper uses lasso model to build an additive semi parametric regression model,and takes the key financial indicators selected by lasso method as parameter and other variables as nonparametric parts,and builds a PLAM model Model is used to predict the future risk of unlisted companies in China's bond market.Com-bined with China's current economic environment and policies,this paper makes a multi-dimensional analysis of the prediction results of some additive non para-metric models,and finally puts forward three feasible policy recommendations.The results show that the overall default of listed companies is on the rise,and the default risk of the market is on the rise.By market,the average default distance of bonds on the Shanghai Stock Exchange is generally higher than that on the Shenzhen Stock Exchange in 2016-2018.As far as the internal part of the exchange is concerned,the overall default distance of bonds shows a down-ward trend,especially when the default distance of the Shenzhen Stock Exchange reaches the minimum value of 0.06 in recent years in 2018,which indicates that most bonds on the Shenzhen Stock Exchange face default risk in 2018.According to the nature of enterprises,in 2018,the default distance of state-owned enterpris-es in bond market is the highest,followed by local state-owned enterprises,and the default distance of the remaining five attributes is the same,with an average of 23.97.Although the state-owned enterprises have had the event of default,they still have the advantage of less probability of default than other enterpris-es.Several suggestions are put forward as follows:the participants should make joint efforts to build a stable bond market;build an intermediary communication system to ensure the quality of practice;build a diversified risk dispersion and sharing channel.