摘要
基于金融高频数据的已实现波动相关理论的提出,使得对跳跃波动与连续波动的识别及区分建模成为可能。不同规模的跳跃可能对应于不同的风险源并存在不同的时间序列特征,但现有文献尚未研究过对大、小跳跃的区分建模。鉴于此,在已实现波动HAR-RV-CJ模型基础上,提出选择合适的阈值,将跳跃波动进一步细分为大跳跃与小跳跃,构建基于跳跃规模细分的HAR-RV-CJ-BS模型。使用沪深300指数5分钟高频价格的移动窗样本外预测表明,HAR-RV-CJ-BS模型对短期、中期、长期已实现波动的预测能力均较HAR-RV-CJ模型有所提升,且对长期波动预测精度的改进最为显著。
The realized volatility theory which is based on financial high frequency data has made it possible to identify and model the jump volatility and the continuous-time volatility separately.The jumps of different sizes might correspond to different risk sources and have different time series characteristics.However,Modeling big jumps and small jumps separately is not studied in current researches.Thus,based on the HAR-RV-CJ model of realized volatility,selecting the appropriate threshold which separates big jumps and small jumps,new HAR-RV-CJ-BS model is built.Empirical tests using CSI300 index 5-minute high-frequency prices and moving windows indicate that HAR-RV-CJ-BS model outperforms HAR-RV-CJ model' s out-of-sample forecasting accuracy for short-term,middle-term and long-term realized volatilities,and the performance improvement for long-term volatility is the most significant.
出处
《中国管理科学》
CSSCI
北大核心
2013年第S1期310-314,共5页
Chinese Journal of Management Science
基金
国家自然科学基金资助项目(71201075)
江苏省自然科学基金面上项目(BK2011561)
高等学校博士学科点专项科研基金资助项目(20120091120003)
教育部留学回国人员科研启动基金资助项目