Credit Risk Metric Based On Improved Creditmetrics Framework

This article mainly uses creditmetrics framework to study the credit risk calculation problem of multi-bond portfolio.The model is divided into third parts.The first part is mainly based on historical corporate bond rating data to estimate credit transfer matrix.The second part is to establish a correlation model between each bond.The third part uses Monte Carlo simulation method to estimate the credit risk of bond portfolio based on correlation and transfer matrix.In the first part,since the bond credit risk is often closely related to the macroeconomic cycle,the article adds an offset in the time homogeneous Markov process to capture the difference between different years.In the second part,the stock value is used to replace the company’s asset value,and the bond correlation is modeled using the t-copula function.Since many of the suspension data in the actual portfolio will affect the estimation effect,this paper uses the EM algorithm to estimate the parameters containing missing data.In the third part,the Monte Carlo simulation method is used to simulate the change of the next asset value according to the copula function.And get the credit loss of the portfolio according to the transfer matrix,and use var and cvar to assess the credit risk.