基于GAS模型的动态VaR预测效果分析  

Analysis of Time-varying VaR Forecast Based on GAS Model

作　　者：刘赛可 何晓群[2] 夏利宇 LIU Sai-ke;HE Xiao-qun;XIA li-yu(School of Computer and Data Engineering,Zhejiang University Ningbo Institute of Technology,Ningbo 315199,China;Center for Applied Statistics,Renmin University of China,Beijing 100872,China;State Grid Energy Research Institute,Beijing 102209,China)

机构地区：[1]浙大宁波理工学院计算机与数据工程学院,浙江宁波315199 [2]中国人民大学应用统计科学研究中心,北京100872 [3]国网能源研究院有限公司,北京102209

出　　处：《数理统计与管理》2022年第1期179-189,共11页Journal of Applied Statistics and Management

基　　金：教育部人文社会科学重点研究基地重大项目(15JJD910002).

摘　　要：GAS模型是一种基于观测的动态模型,理论简单且应用灵活,可以直接估计VaR.将GAS模型和GARCH类模型应用于不同条件下生成的模拟数据和三个时间段的沪深300指数的日对数收益率数据,并比较模型关于VaR的预测效果。结果表明:在对称的条件分布下,GAS模型容易高估风险且不稳健,其表现不如GARCH类模型;但在条件分布为有偏的时,GAS模型与GARCH类模型的表现相当,部分情况下会优于GARCH类模型,尤其在实证分析中关于序列2和序列3的VaR的估计,GAS模型的预测效果较好。因此,实际应用中,对于具有较明显偏态分布或尖峰分布的数据可以考虑使用GAS模型预测动态VaR.GAS(Generalized Autoregressive Score)model is a dynamic model which is based on observation.Due to its simplicity and broad applicability,GAS model can be directly used to estimate VaR.GAS model and GARCHs model are applied to both simulated data and returns of CSI300 data during three periods to compare their prediction accuracy.The results show that,when the conditional distribution is symmetrical,GAS model is inferior to GARCHs model since the former tends to overestimate the risk and is not robust.But when the conditional distribution is asymmetrical,GAS model and GARCHs model show similar performance except for some cases where GAS model has better results.For example,the GAS model achieves better prediction effect about the estimation of VaR of time series 2 and time series 3 in empirical analysis.Therefore,GAS model is capable of the application to VaR forecast for the data with obvious skewed or kurtosis distribution.

关 键 词：动态VAR GAS模型 GARCH类模型 

分 类 号：O212[理学—概率论与数理统计]