风险管理视角下的高频波动率测度比较

Comparison of High-frequency Volatility Measures in the Context of Risk Management

鉴于波动率研究的一个重要应用是金融风险管理,提出在风险管理视角下比较各种高频波动率测度。具体考虑已实现波动、双幂次变差、中位数已实现波动、双尺度已实现波动、已实现极差和最优线性组合已实现波动,借助偏学生t分布假设下的Realized GARCH模型,预测未来一日,五日、十日和二十日的在险价值(VaR),并从统计精度、监管精度、资本运作效率和巴塞尔Ⅱ规定的市场风险资本需求四个角度,对六种不同高频波动率测度的VaR预测效果进行比较。使用沪深300指数1分钟数据的实证表明,最优线性组合已实现波动产生的VaR预测具有明显最低的市场风险资本需求,较高的监管精度和较高的资本运作效率,以及最高的统计精度,是风险管理视角下比较可靠的高频波动率测度。

In this paper,popular high-frequency volatility measures are compared in the context of risk management.Realized volatility,bipower variation,median realized volatility,two-time-scale realized volatility,realized range and optimal linear combination realized volatility in forecasting one-,five-,ten-and twenty-day ahead Value at Risk(VaR) are taken into consideration under the framework of Realized GARCH model with skewed student-t distribution.The VaR forecasting performances are evaluated in terms of statistic accuracy,regulatory accuracy,capital efficiency and Basel II regulatory market risk capital requirement.Empirical results with CSI 300 index one-minute prices indicate that the optimal linear combination realized volatility forecasted VaR has significantly lower market risk capital requirement,relatively higher regulatory accuracy and higher capital efficiency,and the highest statistical accuracy,and thus is a reliable high-frequency volatility measure from the risk management point of view.