摘要
本文以上证A股指数为例对GARCH类模型在估计Value-at-Risk(VaR)值时所存在的模型风险进行了分析。我们分别考虑了基于EWMA,GARCH,EGARCH和FIGARCH模型的VaR估计方法。模型风险的存在意味着使用不同的估计方法得出的VaR值可能迥然不同。为了对这四种估计方法进行评判,我们在似然率和Kullback-Leibler信息准则的基础上运用四种统计检验方法对不同置信度水平下的VaR估计值进行了返回检验。实证结果表明EGARCH和FIGARCH方法的表现明显比其它两种优越。
This paper studies the model risk of GARCH-type models of volatility in modeling the daily Value-at-Risk (VaR) of Shanghai Stock Exchange index. Four estimates based on EWMA, GARCH, EGARCH and FIGARCH models to VaR are examined in this paper. As alternative VaR implementations may yield very different VaR estimates, we evaluate the performances of these four volatility models in different confidence levels by four statistical tests based on likelihood ratio or Kullback-Leibler information criterion. Our results show that the obtained VaR estimates by EGARCH method and FIGARCH method are much better than the ones provided by the other two methods.
出处
《管理评论》
2005年第10期3-7,54,共6页
Management Review
基金
高等学校全国优秀博士学位论文作者专项资金(200267)
新世纪优秀人才支持计划(NCET-04-0798)
国家自然科学基金项目(70471018)
