基于极值统计和高维动态C藤Copula的股市行业集成风险计算  

Calculation for Industry-integrated Risk Based on EVT and High-dimensional Dynamic Canonical Vine Copula Structure

作　　者：严太华[1] 韩超[1] 

机构地区：[1]重庆大学经济与工商管理学院,重庆400030

出　　处：《数理统计与管理》2016年第6期1098-1108,共11页Journal of Applied Statistics and Management

基　　金：国家自然科学基金(71373296)

摘　　要：股市诸多行业风险之间存在着波动相依性,集成计量多维风险对投资决策意义重大。藤Copula是Copula函数高维化拓展的一个方向,其动态化是新的研究前沿。将极值理论的GPD模型和高维动态C藤Copula方法结合起来研究沪深300指数中地产、基建、银行和运输四个行业风险,能够有效描述尾部极值形态,突出关键变量的作用。再运用动态Pair-Copula分解,刻画高维行业风险变量间的动态关系,以仿真出动态集成风险变量VaR序列。VaR计算结果通过了回溯检验和稳定性测试,表明高维动态C藤Copula模型可以作为风险集成计量的一种新的有效方法。There are always industry risks in stock markets. It is of very important significance for investment decision how to integrate high-dimensional risks. Vine copula denotes the research direction of high dimension. And dynamic research is the frontier in this scholar field. This article uses GPD model of EVT and high-dimensional dynamic canonical vine copula simultaneously, and researches on four-industry risks. It is demonstrated that the models can describe the shape of extreme value of tail fraction and highlight leading variable. Furthermore, this article decomposes four-dimensional risk series by dynamic Pair-Copula, then describes the dynamic dependence, gets dynamic integrated VaR series by Monte Carlo simulations. The results gets through back testing and stability testing. Thus highdimensional dynamic canonical vine copula can be used as a new and prudent way to measure integrated risks.

关 键 词：GJR-SkewT GPD 高维动态C藤Copula VAR 风险集成 

分 类 号：F830[经济管理—金融学] O212[理学—概率论与数理统计]