基于Skewed-t分布的FIGARCH模型与VaR的度量  

FIGARCH-Type Models and VaR Based on Skewed Student t Distribution

作　　者：黄炎龙[1] 

机构地区：[1]同济大学经济管理学院,上海200092

出　　处：《应用概率统计》2012年第2期189-202,共14页Chinese Journal of Applied Probability and Statistics

摘　　要：金融资产收益率序列的波动具有典型的尖峰厚尾和非对称性特征,描述这种特性需以合适的概率分布函数为基础.因此,寻求更好的概率分布函数对风险度量、VaR的计算有着十分重要的意义.有鉴于此引入Skewed-t分布度量VaR,并比较分析了RiskMetrics及FIGARCH类模型度量VaR值的准确程度,本文同时分析了多头头寸和空头头寸情况下的VaR.结果表明,在两种头寸情况下,Skewed-t分布在空头和多头情形对资产厚尾特性以及非对称性的拟合效果均要比正态分布好;在两种头寸中不同的置信水平下,FIAGARCH(CHUNG)模型预测的VaR值改进了使用传统模型的精确性,高估或低估风险的程度较轻.The study of volatility models exert an important influence on the calibration of Value at Risk,in that they are not only universally popular in modeling financial data,but also as basis on calculating VaR by using adaptive mathematical models.This paper studies RiskMetrics and GARCH-type models of 11 comparatively,based on the assumption of Gaussian Normal distribution and Skewed Student＇s t distribution respectively and their accuracy of computing VaR by analyzing the close price of Shanghai Composite Indice from Jan.5,1998 and Nov.6,2006 in China.The study checks the one-step-ahead forecasting VaR by invoking Failure Rate Test and Dynamic Quantile Test.The analysis on models and VaR shows the truth that Skewed Student＇s t distribution is better fitted with the feature of lepkurtosis than Gaussian Normal distribution and the models of FIGARCH（BBM）and FIEGARCH as the extension of GARCH model improve the estimation under the traditional GARCH models,which the degree of extra-high or low estimation is receivable.

关 键 词：VAR FIGARCH模型 Skewed-t分布 动态分位数测试 

分 类 号：F830[经济管理—金融学]