Abstract: Bayesian dynamic correlation Wishart volatility model is established in this paper to address estimation problem of time-varying coefficient matrix in multivariate stochastic volatility. In order to make the correlation coefficient matrix incorporated time-varying characteristics, the precision matrices in CC-MSV models are set to following the Wishart distribution. Based on the analysis of statistic structure of model and the selection of parameters prior, the Gibss-MTM-ARMS sampling algorithm method is utilized to estimate model parameters. The empirical research applies the data of Shanghai Composite Index, S&P500 and crude oil future price. The research results show that the correlation between crude oil market and stock market is strong during the financial crisis, but it is difficult to identify the direction. Further, after the financial crisis, the U.S. stock market is obviously positive correlated with crude oil market, while the correlation between China stock market and crude oil market is very weak. All those indicate that models used in this paper can effectively depict the dynamic dependency between crude oil market and stock market.

Key words: dynamic dependence, stochastic volatility, Bayesian analysis, Wishart distribution, Gibbs-MTM-ARMS algorithm