基于赋权修正测度的时变参数Realized HAR GARCH模型及其实证研究  

Time-Varying Parameters Realized HAR GARCH with the Weighted Modified Information Volatility and Its Empirical Analysis

作　　者：项琳 陈宇峰 胡昊 XIANG Lin;CHEN Yufeng;HU Hao(College of Economics and Management,Zhejiang Normal University,Jinhua 321000;School of Mathematics Sciences,Zhejiang Normal University,Jinhua 321000)

机构地区：[1]浙江师范大学经济与管理学院,金华321000 [2]浙江师范大学数学科学学院,金华321000

出　　处：《系统科学与数学》2024年第7期2060-2087,共28页Journal of Systems Science and Mathematical Sciences

基　　金：国家自然科学基金项目(72174180);教育部人文社科重点研究基地重大项目(22JJD790080);浙江省杰出青年科学基金项目(LR18G030001);浙江省哲学社会科学基金项目(22QNYC13ZD,21NDYD097Z)资助课题。

摘　　要：基于高频数据视角,文章提出时变参数(TV)Realized HAR GARCH混合预测模型,同时考虑资产波动率的聚集性、长记忆以及由测量误差引起的参数衰减偏差效应.进一步,为充分利用价格信息并提升估计效率,本文基于日内“OHLC”数据构建赋权修正已实现信息波动率(WRIV),并将其用于驱动条件方差的动态变化.在偏t分布假设下,以沪深300指数为样本探究中国股票市场的波动性规律,并在实证中评估所提模型在收益率拟合、波动率预测以及风险度量上的能力.结果显示:中国股票市场的收益波动存在明显的异质性与长记忆特征,TV-Realized HAR GARCH能够充分捕捉指数波动率的动态变化,在样本内拟合效果和样本外波动率与风险预测准确性上均能显现出优势,且WRIV测度的引入能显著提升模型的预测精度,凸显出日内高频数据信息的充分利用对于波动率刻画与风险测度的重要性,综合而言,TV-Realized HAR GARCH(WRIV)模型具有最优的整体实证表现.Based on the high-frequency data,a hybrid volatility forecasting model named time-varying parameters,(TV)Realized HAR GARCH is constructed in this paper.This model accommodates the volatility clustering,long memory,and attenuation bias effect of parameters caused by measurement error.In order to make full use of price information and improve estimation efficiency,this paper proposes a weighted modified realized information volatility(WRIV)based on intra-day'OHLC'data.Then the WRIV is introduced into the Realized GARCH-type models to drive conditional variance dynamics.Under the skewed-assumption,the volatility pattern of the Chinese stock market is explored with the CSI 300 index as the empirical sample.The performance of the extended model combining different realized measures is empirically evaluated in terms of in-sample fitting,volatility forecasting,and risk measurement.The results show that the return volatility in the Chinese stock market is characterized by significant heterogeneity and long memory.Moreover,the TV-Realized HAR GARCH-type models can fully capture the volatility dynamics for the CSI 300,showing superiority in both in-sample fitting and out-of-sample forecasting accuracy.Our results also suggest that the introduction of the WRIV measure can significantly improve the model's forecasting accuracy,highlighting the importance of leveraging high-frequency data information for volatility characterization and risk measurement.The TV-Realized HAR GARCH,(WRIV)model exhibits the best overall empirical performance.

关 键 词：高频数据 时变参数 Realized HAR GARCH 波动率 风险管理 

分 类 号：F832.51[经济管理—金融学] F224