基于AR(m)-QAR-GARCH模型的沪深指数VaR测度研究

VaR Measurement of Shanghai and Shenzhen Composite Index Based on AR(m)-QAR-GARCH Model

传统的GARCH模型在测度我国沪深指数日对数收益率VaR时,由于不能兼顾其尖峰、厚尾、有偏性和自相关性的特征往往效果不佳。针对沪深指数日对数收益率的上述特点,提出利用AR(m)-QAR-GARCH模型测度我国三大股指的VaR。基于Kupiec似然比和DQ检验表明:AR(m)-QAR-GARCH模型测度沪深指数VaR预测准确性要好于自相关性不明显的恒生指数和日经指数;AR(m)-QAR-GARCH模型对沪深指数VaR测度效果要好于几种没有考虑自相关性的GARCH模型;对于我国沪深指数收益率自相关性可能存在的阶段性特点,AR(m)-QAR-GARCH模型也适合。

Traditional GARCH-based model usually performs not so good when measuring VaR about log return rate of Shanghai and Shenzhen composite index without consideration of their property of leptokurtic, skewed and fat-tailed distribution and autocorrelation at the same time. This study proposes AR（m）-QAR-GARCH to measure three security exchange indices in our country with consideration of their distribution property and autocorrelation. Three main results of this study are attained by the use of Kupiec likelihood ratio test and DQ test. Firstly, we find that the model proposed by this study predicts more exactly than in the use of measuring VaR of Shanghai composite index and Shenzhen composite index than it has been used of HIS and Nikkei 225 index. Secondly, the model performs better than other GARCH-based model ignorance of volatility autocorrelation when measuring VaR of Shanghai composite and Shenzhen composite index. Thirdly, we also find that AR（m）-QAR-GARCH model also has good performance for measuring the VaR of Shanghai composite and Shenzhen composite index when the autocorrelation of them appear in a phase and disappear in another phase.