基于藤copula–已实现GARCH的组合收益分位数预测  

Portfolio quantile forecasts based on vine copula and realized GARCH

作　　者：黄友珀[1] 唐振鹏[1] 唐勇[1] 

机构地区：[1]福州大学经济与管理学院,福建福州350116

出　　处：《系统工程学报》2016年第1期45-54,共10页Journal of Systems Engineering

基　　金：国家自然科学基金资助项目(71171056);福建省社科规划办重点资助项目(2013A017);福建省高等学校新世纪优秀人才支持计划资助项目(JA11025S)

摘　　要：为准确预测分位数,利用已实现GARCH模型在边缘分布建模中纳入高频信息,通过藤copula刻画资产收益两两之间相异的相依结构,构建了资产组合收益分位数预测的藤copula–已实现GARCH模型.选取中国股市风格指数组合展开实证分析,回测检验结果表明高频信息和异质相依结构是准确预测分位数的关键环节,藤copula–已实现GARCH模型能够提供更准确的资产组合收益分位数预测.Aiming at enhancing quantile forecasting performance, this article incorporates high-frequency information using realized GARCH model, then captures the variant dependency structure of pairs of asset returns by vine copula, and constructs a quantile forecasting approach based on vine copula-realized GARCH model. Selecting Chinese style index portfolio to carry out an empirical analysis, the backtesting results show that both high-frequency information and heterogenous dependency structure are key links for precise quantile forecasts, and the proposed method provides more accurate quantile forecasts of portfolio returns.

关 键 词：分位数预测 藤copula–已实现GARCH 高频信息 相依结构 回测检验 

分 类 号：F830[经济管理—金融学]