基于极值理论和多元时变copula模型的我国外汇储备汇率风险度量  

Measuring Currency Risk of Foreign Exchange Reserve in China with Extreme Theory and Multivariate Time-varying Copula Model

作　　者：崔百胜[1] 陈浪南[2] 

机构地区：[1]上海师范大学金融学院 [2]中山大学经济研究所

出　　处：《国际贸易问题》2011年第12期158-168,共11页Journal of International Trade

基　　金：国家自然科学基金项目(70673116);北京大学汇丰金融研究院2009年课题;国家社科基金重点课题(08ATL007);广东省自然科学基金(9151027501000032);社科基金课题;广东省普通高校人文社会科学重点研究基地经费资助;上海师范大学原创与前瞻性课题;上海市哲学社会科学规划课题(2009BJB022);上海市教委科研创新重点项目(09ZS142)资助成果之一

摘　　要：本文利用极值理论和ARMA-AGARCH-t模型,得到人民币对美元、欧元、港元、日元和英镑五种货币汇率日收益率的残差序列,利用多元静态和时变copula-t模型,分别求出各自VaR与CVAR值,并计算不同目标日收益率下最优外汇储备持有结构。研究结果表明,根据极值理论得到的广义帕累托分布能够较好拟合汇率日收益率序列的尾部特征,与多元静态copula-t模型相比,时变coupla-t模型能够更好度量外汇储备汇率风险。在给定目标收益率区间,美元最优持有比率随目标收益率提高而下降,日元则同方向增加。This paper first uses extreme value theory and the ARMA-AGACH-t model to derive the residual series of daily returns of the five exchange rates of RMB against the U.S. dollar, euro, yen, Hang Kong dollar, and pound. And then it uses multivariate static and time-varying copula-t model respectively to esti- mate its VaR and CVAR value. Finally it calculates the optimal structure of foreign exchange reserves under the target of different return ratios. The result shows that generalized Pareto distribution can well fit the tail of exchange rate re- turns series. Compared with multivariate static copula-t model, time-varying copula-t model can better measure the currency risk of foreign reserves. Given the target of return ratio, the optimal holding ratio of dollar declines when the target of return ratio increases while that of yen increases in the same direction.

关 键 词：外汇储备 汇率风险 度量 极值理论 多元时变copula模型 

分 类 号：F830.92[经济管理—金融学]