Risk Correlation Based on Time-Varying Copula Function and Extreme Value Theory

Abstract

The dependence structure of financial assets in financial risk measurement is very important, the tail relations in particular. Authors of extant studies in this field tended to focus on the linear analysis of the financial assets, rarely considering nonlinear, asymmetric and thick-tail characteristics. Here, we apply the copulas connection function with time-varying factors to discuss the risk dependency relationship between financial assets. Moreover, we develop an SV-EVT model to fit variables’ marginal distribution combined with stochastic volatility and extreme value theory. Finally, we present an empirical comparative study of static and dynamic copula models applied to the sample comprising of the Chinese mainland A-shares and Hong Kong stock market. The results show that the CSJC copulas connection function describes the tail features of stock index better than the normal copulas connection function. Similarly, the time-varying model outperforms the static copulas model. Furthermore, we observe an asymmetry dependence change rule between Chinese mainland A-shares market and the Hong Kong stock market; the correlation of lower tail is significantly higher than that of the upper tail, and the bear market effect is remarkable. These findings indicate that time-varying Copulas-SV-EVT model can depict the correlation of financial asset tails exactly, and can thus be used to control investment risk and forecast abnormal fluctuations.