基于MCMC算法的时变Copula-GARCH-M-t模型及组合风险预测  

Time-varying Copula-GARCH-M-t Model Based on the MCMC Procedure and It's Application to Portfolio Risk Prediction

作　　者：彭选华[1] 傅强[2] 

机构地区：[1]西南政法大学经济学院,重庆401120 [2]重庆大学经济与工商管理学院,重庆400044

出　　处：《数理统计与管理》2013年第1期180-190,共11页Journal of Applied Statistics and Management

基　　金：重庆市自然科学基金项目(cstc2012jjA00023);教育部博士点专项基金(20100191110033);国家自然科学基金项目(70501015)等资助

摘　　要：为了准确地量化资产之间的时变相依结构和预测组合风险,本文考虑到投资者对资产风险偏好的差异,假设资产收益率序列的新息服从标准t分布,提出时变Copula-GARCH-M-t模型,推导了模型参数的两步MCMC估计方法,还得到了组合风险(VaR和CVaR)的一步预测方法。最后选取上证综合指数和标准普尔500指数,验证了所提模型及方法的可行性和优越性,同时该模型较为准确地量化了两指数在次贷危机后的时变相依结构特征。Considering the risk preference of investor and assuming the innovations come from a standard t-distribution, this paper proposes a time-varying Copula-GARCH-M-t model in order to accurately quantify the time-varying dependency structure and forecast the portfolio risk. We design a two-steps MCMC procedure to estimate the model parameters, and obtain one-step predictor of the VaR and CVaR. Finally, the example of Shanghai composite index and the s＆=p 500 index confirms the modelling methods are effective and feasible. Our models accurately quantify the time-varying characteristics of the two-index dependency structure after the sub prime crisis.

关 键 词：MCMC算法 COPULA-GARCH 风险管理 预测 

分 类 号：O212.1[理学—概率论与数理统计] F830.91[理学—数学]