提出一种新的金融市场波动率的测度方法——多分形波动率(multifractal volatility)测度,并以上证综指在长达8年左右时间内的高频数据样本为例,构造了多分形波动率的ARFIMA动力学模型.同时,运用最近提出的SPA(superior predictive ability)检验法,实证对比了多分形波动率模型与现有的如实现波动率(realized volatility)模型、GARCH模型以及随机波动(stochastic volatility,sv)模型对市场波动预测能力的优劣.实证结果显示,在某些损失函数标准下,文中提出的多分形波动率测度及其动力学模型具有比现有其它模型更优的波动率刻画能力和预测精度.
In this paper, a new volatility measure, muhifractal volatility, is constructed. Base on about 8 years' high-frequency data of SSEC, we choose ARFIMA model as the dynamic model of multifractal volatility and use SPA by Hansen and Lunde (2005) to test the predicting performance of the multifractal model and other popular models, such as, Realized volatility model, GARCH and Stochastic volatility model. The empirical results show that, for some kinds of loss functions, muhifractal volatility measure and its ARFIMA model have better predicting performance than other existing volatility models.