摘要
结合国内外研究新进展对金融高频数据波动率的特征进行综述,对比分析了已实现波动、已实现极差波动、已实现双幂次变差、偏差校正已实现双幂次变差、已实现极差四次幂变差等波动测算方法以及考虑日历效应对上述方法进行的赋权修正。从稳健性、有效性、无偏性等角度对已实现类波动的特征进行分析,并对上述成果建立HAR模型完成滚动预测和评价,最后指出赋权已实现极差四次幂变差为这十种高频数据波动率中最优的估计量。
The paper summarizes the characteristics of the volatility of high frequency data of the domestic and foreign research progress,and conducts a comparative analysis of fluctuation measuring methods,such as realized volatility,realized range-based volatility,realized bi-power variation,bias-corrected realized bi-power variation,realized range-based quad-power variation,considering the weighted correction given by calendar effect to the above methods.From the perspective of robustness,effectiveness and non bias,the characteristics of the realized fluctuation have been analyzed,and the HAR model is established to complete the rolling prediction and evaluation.Finally,it is pointed out that realized range-based qua-dpower variation is the best estimate of the ten kinds of high frequency data volatility.
出处
《山东科技大学学报（社会科学版）》
2017年第5期78-87,共10页
Journal of Shandong University of Science and Technology(Social Sciences)
基金
山东科技大学科研创新团队支持计划资助项目(2015TDJH103)
青岛市知识产权软科学研究计划项目(QDISSP-1305)
关键词
已实现波动
日历效应
HAR模型
滚动预测
realized volatility
calendar effect
HAR model
rolling prediction
