典型事实、极值理论与金融市场动态风险测度研究

本文在金融市场典型事实约束下,运用ARFIMA模型对金融市场条件收益率建模,运用GARCH、GJR、FIGARCH、APARCH、FIAPARCH等5种模型对金融波动率进行建模,进而运用极值理论(EVT)对标准收益的极端尾部风险建模来测度各股市的动态风险,并用返回测试(Back-testing)方法检验模型的适应性。实证结果表明,总的来说,FIAPARCH-EVT模型对各个市场具有较强的适应性,风险测度能力较为优越。进一步,本文在ARFIMA-FIAPARCH模型下,假定标准收益分别服从正态分布(N)、学生t分布(st)、有偏学生t分布(skst)、广义误差分布(GED)共4种分布,对各股市的动态风险测度的准确性进行检验,并和EVT方法的测度结果进行对比分析。结果表明,EVT方法风险测度能力优于其他方法,有偏学生t分布假设下的风险测度模型虽然略逊于EVT方法,但也不失为一种较好的方法;ARFIMA-FI-APARCH-EVT不仅在中国大陆沪深股市表现最为可靠,而且在其他市场也表现出同样的可靠性。

Measuring Dynamic Risk of Financial Markets Based on Stylized Facts and Extreme Value Theory

In this paper, the ARFIMA model was used to model conditional return, and GARCH, GJR, FIGARCH, APARCH and FIAPARCH models to model conditional volatility in financial markets based on some stylized facts, respectively, and then, measure extreme risk by extreme value theory (EVT), at last, we test accuracy of all models in 6 financial markets by the back-testing method. Empirical results show that the FIAPARCH-EVT model is superior to others in all market. Further, this paper uses the ARFIMA-FIAPARCH model with normal distribution (N), student t distribution(st), skew student t distribution(skst), Generalized Error Distribution (GED) and EVT to measure dynamic VaR and test accuracy of dynamic VaR measurement, and finds that EVT is better than others used in this paper,I also find that ARFIMA-FIAPARCH-skst is an excellent risk measurement model, the ARFIMA-FIAPARCHEVT model is the best risk measuring model for Chinese mainland stock markets, and it also exhibit obvious robustness in other financial markets introduced in this paper.