基于Copula函数的股指尾部相关性研究——以道琼斯工业指数与恒生指数为例

Study of Tail Dependence Analysis of DJI&HIS Based on Copula Method

鉴于两步参数估计法在应用中存在误差大、计算复杂等缺陷,采用基于经验分布的半参数估计与非参数估计法确定相应边缘分布与Copula参数,对突发事件下的道琼斯工业指数与恒生指数之间的尾部相关性进行量化.研究发现ClaytonCopula,Gumbel Copula能够较好地刻画股指收益率序列间的尾部相关关系;道指与恒生指数存在着正的尾部相关且这种相关是非对称性的;在各个置信水平上,下尾损失均较上尾收益高,且下尾相关系数的增长幅度远大于上尾相关系数的增长幅度;极端事件造成的道指收益的剧烈下跌引发了恒生指数收益更强烈的相关反应,其造成的影响远超过两个市场同时上涨时的作用.

The semiparametric method metric method for Copula estimation are for marginal distribution estimation and nonpara- employed due to relatively big deviation produced by the fully parametric methods, namely inference function and maximum likelihood methods. This paper investigates the tail dependence structure between DowJone Index（DJI） and Hangseng Index（HIS） under the emergency. We find that Gumbel copula and Clayton copula can depict the tail dependence between indexes very well. The correlation is positive and asymmetric. Under each confidence level, the loss of lower tail is more than the return of upper tail, and the growth of lower dependence is larger than upper denpendence. The loss of DJI caused by extreme events causes more serious loss of HIS, and its influence has exceeded the effect of two-market increase at the same time.