摘要
本文应用极值的阈值与峰值模型来度量单个资产的风险价值,用两种不同的方法度量了基于Copula函数的沪深指数收益率的相关结构,比较了不同Copula函数下基于沪深指数的二元投资组合集成风险值。结果说明:Gauss Copula函数对沪深指数收益率的相关结构拟合较好,阈值模型的极值Copula能较好地度量投资组合的集成风险值,在高置信度下(0.99以上),基于Gumble Copula函数的上尾(正收益)集成风险值、基于ClaytonCopula函数的下尾(负收益)集成风险值与真实值最为接近。直接加权的方法会高估投资组合的风险,假设沪深指数的收益率服从二元正态分布会低估风险。峰值法的集成风险值误差较大。
The paper measures the value risk of single asset with the application of threshold and peak model of extreme value theory, measures the dependence structure of the return rate of Shanghai, Shenzhen index with copula function, compared the integrated risk of the Shanghai, Shenzhen index with different copula function. And the result of empirical suggests : Gumble Copula is a better measure of the dependence structure comparing the other Copula function. Threshold of extreme copula model can measure the integrated risk of portfolio well. at high confidence level( up 0. 99), the integrated risk of upper tail ( positive return) with Gumble copula and the integrated risk of lower tail( negative return) with Clayton copula are most close to the true value. The method of straight add weight will overestimate risk, the method of supposing the Shanghai, Shenzhen index return obey duality normal distribution will underestimate the risk. Finally, the integrated risk of peak has big error.
出处
《统计研究》
CSSCI
北大核心
2011年第7期84-91,共8页
Statistical Research