A Study On The Construction And Application Of Credit Risk Correlation Model

Credit is the cornerstone of market economy, and the harm of credit risk has been widely paid attention by financial industry. Modern credit risk is featured of contagion, and the credit risk burst of one financial institution or industrial company may lead to a wide range of credit default, thus the credit risk may spread nonlinearly among the enterprises and financial institutions. Since the21st century, the credit risk contagion events among the real economy and the financial market have emerged enomorously, and under this background, the correlation among credit risks especially extreme risk events should be taken into consideration in the process of credit risk management. Thus, the implementation of credit portfolio management has become a development trend for credit risk management.A large amount of researches have been conducted to detect the credit risk, but there is still not paid enough attention on the measurement of the correlation in the credit portfolio management. Therefore, this paper chooses the credit relationship as the study object, and a theoretical analysis, modeling analysis and application study are respectively conducted to explore the mechanism of credit risk correlation, evaluate the correlation of credit risk, and apply the correlation in credit portfolio management.In the section of mechanism study of credit risk correlation, some related concepts including credit risk, credit risk correlation are defined, and then the formation mechanism of credit risk correlation. And the factors influencing the credit risk correlation contain common factors such as macroscopically economy environment, macro economic variables, political and policy events, public safety events, technical factors and social factor, contagion factors from market and equity market, and coupling factors of contagious and common factors.In the section of credit risk and correlation modeling, credit risk and correlation evaluation models are constructed by using Chinese Listed Company data from1990to2010, and the effectiveness of the models are examined, then the characteristics of credit risk correlation are evaluated through the model analysis. This section is organized as follows. Firstly, industry credit risk indexes are established based on credit risk evaluation models, and a hierarchical cluster analysis is conducted for the industry credit risk. The Hybrid model integrating MDA model, SVM model and KMV model can better synthesize financial information and capital market information and effectively evaluate enterprises’ credit risk. In the SU space, the minimum spamming tree is applied to conduct hierarchical cluster on the industry credit risk for the purpose of dimension reduction. The empirical results show that electronically power industry, food industry, petrochemical industry and information technology industry can represent the strong cyclical industry, defensive industry, weak cyclical industry and growing industry. The Johnasen cointegration test shows that there is cointegration relationship among the above industries’credit risk. Secondly, the overall features of credit risk coorelation are described based on the static Copula models. On the whole, industry credit risk correlation has an asymmetric characteristic, and the credit risk correlation is more sensitive to disadvantergous external environment changes. Thirdly, considering the characteristics of dynamic change, jump and regime switching, the bivariate dynamic Copula models, Jump Copula model and Markov Regime Switching Copula models are respectively built. The KS test and AD test shows that the Jump Clayton copula is superior to other Copula models by judging the goodness fit of the models. And the dynamic correlation coefficient indicates that the credit correlation of sample industries is high and is featured with asymmetric, low tail sensibility, and susceptible to systematic risks. Fourthly, a multivariate copula model is constructed to measure the credit risk correlation. Under the pair copula framework, the multivariate copula is decomposed following the Canonical vine and D vine structure. A multivariate copula model that can accurately describe the multivariate credit risk correlation is acquired through the empirical study. Additionally, the multivariate correlation is relatively lower than the bivariate correlation, indicating that multiple credit assets jointly change with a relatively lower possibility.In the section of application study of the credit risk correlation models, the Copula VaR model is applied in credit portfolio management. On the basis of Pair Copula model, the critical steps of credit portfolio management based on the Copula VaR model is designed. By analyzing the credit portfolio of the sample commercial bank, the optimization directs of the credit portfolio management are pointed out. And finally, based on the theoretical and empirical study, some available suggestions for the implementation of commercial banks’credit portfolio management are proposed.