摘要
针对金融数据的重尾、波动聚集、非对称性等特征,提出了基数据驱动的GAS模型的两种新模型:E-GAS-AST模型和E-GAS-AST-GPD模型,并利用新模型对实际数据进行了风险度量和回测.基于GAS模型,结合具有重尾特征的非对称学生t-分布(AST),参照EGARCH模型提出了E-GAS-AST模型,并使用GPD分布对尾部极值特征进行进一步的描述,重新得到E-GAS-AST-GPD模型.通过研究两个模型各自的残差分布计算出VaR值和ES值,并分别进行回测检验.引入参数驱动模型比如半参数GARCH模型、EGARCH-t模型和GJR-GARCH-t模型进行风险度量的估计,并与本文提出的两个模型进行比较.对道琼斯指数和上证指数在考虑收益率序列可能存在变点的情况下进行的实证研究表明,该数据驱动的E-GAS-AST模型是一个较好可行的模型,可用于对金融市场进行风险度量的模型.
Concerning financial data's fat-tail,volatility clustering and asymmetry,we raise two data-driven models:E-GAS-AST model and E-GAS-AST-GPD model,and proceed risk measuring and backtesting with real data.Based on generalized autoregressive score(GAS)model,combining asymmetric student-t(AST)distribution with heavy tail,we propose an new model denoted by E-GAS-AST referring to EGARCH model.Considering describing more of tail features,we propose another E-GAS-AST-GPD model with generalized pareto distribution(GPD).Afterwards,the paper computes VaR and ES by studying distributions of residuals,and backtests them separately.Introducing parameter-driven models,such as semi-parameter generalized autoregressive conditional heteroskedasticity(GARCH)model,EGARCH-t model and GJR-GARCH-t model to produce risk measurement we compare them with two above models proposed.Empirical analysis using Dow Jones Index and Shanghai Stock Exchange Composite Index concerning change point reveals E-GAS-AST model is proper to model financial market and measure risk.
作者
夏艺萌
陈昱
XIA Yimeng;CHEN Yu(School of Management, University of Science and Technology of China, Hefei 230026, China)
基金
国家重点研发计划项目(2016YFC0800104)
国家自然科学基金(11671374,71771203,71631006)资助.
