Research On Bond Portfolio Credit Risk Modeling And Its Applications

Modern economy is a kind of credit economy, so both parties must rely on credit to cooperate. There are many problems in the financial market, such as the more and more widespread financial crisis and the prevalence of information asymmetry, leading to higher demand for credit risk management. How to establish a Portfolio credit risk model witch is more consistent with the actual situation has become one of the most important topics of research. Aiming at the deficiency of Merton model, main works in this paper are as follows:Firstly, a discrete credit risk model with the first passage time is established. On the basis of the structural model, a new breach of contract is established. The company's annual or semiannual maximum negative return can be considered as a state variable, and the distribution function of each state variable can be obtained by using the extreme value theory. Considering the companies from different industries, it is more reasonable to construct the joint distribution function of state variables using the hierarchical Gumbel Copulas function. The model which is discrete and credit losses model for first passage time improves shortcomings of the Merton model. The empirical study shows that,compared with the non- stratified Gumbel Copula model, the measured credit default loss tail is thicker, and the effect is more conservative.Next, a dynamic bond portfolio credit risk model is established.Knowing company information at and before the moment t , the company's credit losses situation at some point in the future is measured.With KPMG risk neutral model, we get the relationship between the probability and loss rate of default for the bond i at the time t under the condition of hypothesis. According to the log-return of the bond i at the time t as state variables, GARCH model is used to depict the marginal distribution of state variables; Pair-Copula-GARCH model is applied to construct the joint distribution function of state variables.Compared with other models, this model is dynamic and forward-looking.Finally, the paper gives a brief summary of the total work, and proposes directions for further research.