Bayesian Analysis Of GARCH-Copula Model And Financial Application

For financial markets, most scholars agree with using GARCH model to capturethe dynamics and dependence structure between stock market returns. Also, there arecorrelations of volatilities among different markets or various assets. In order todisperse and defuse financial risks, multiple assets always been combined so that riskscan be evaded. And risk avoidances are built on the analysis of the correlatedcharacteristics among multiple markets. There are often limitations of modelparameter estimation and multivariate distribution assumptions for multivariateGARCH when dealing with multiple unknowns. Luckily, those issues can be solvedby Copula technology. The copula function not only provide the correlationstructure of multivariate distribution without considering the marginal distributionstructure, also give a convenient way to find the joint distribution function. One goodthing is，the corresponding Copula function unchanged when do the Monotoneincreasing transformation. In this paper, using Copula technology and GARCHmodels to deal with financial risk in multiple assets, and it is more convenient thanjust apply the GARCH models. In addition, most of the references, the estimation ofparameters of GARCH model as well as the copula function by using Frequencymethods. However, I choose the Bayesian method to estimate the parameters and thismethod can mine the information of parameters more sufficient. In the final empiricalpart, using the T-Copula-GARCH-T model to predicts the VaR value with0.05confidence one day in advance of the Shanghai Composite Index and ShenzhenComposite Index.