已实现波动、偏度和峰度与股票未来收益——来自中国A股市场的证据  

Realized Volatility,Skewness and Kurtosis and Future Stock Returns-Evidence from China's A-Share Market

作　　者：潘娜 李子洋 周勇[2] PAN Na;LI Zi-yang;ZHOU Yong(Institute for Advanced Studies,Finance and Economics,Wuhan 430205,China;Academy of Statistics and Interdisciplinary Sciences,Faculty of Economics and Management,East China Normal University,Shanghai 200241,China)

机构地区：[1]湖北经济学院财经高等研究院,湖北武汉430205 [2]华东师范大学经管学部统计交叉科学研究院,上海200241

出　　处：《数理统计与管理》2023年第6期1127-1140,共14页Journal of Applied Statistics and Management

基　　金：国家自然科学重大研究计划重点项目(91546202);国家自然科学基金委重点项目(71931004);教育部人文社会科学研究规划基金项目(21YJAZH062)。

摘　　要：与美国市场中已实现偏度与截面股票未来收益显著负相关不同,本文利用中国A股市场15年的高频交易数据构建已实现高阶矩,实证研究发现已实现峰度与股票未来收益存在不能被其他变量所分散的显著负相关,日内收益中的极端值的存在会导致未来收益的负向变动,相比已实现波动与偏度,是理解A股市场中截面资产定价的显著因子。此外本文发现已实现高阶矩特别是已实现峰度对股票价格的影响是状态依存的,相对于波动趋同的牛熊市,在震荡市场行情中,基于已实现峰度构造的多空投资组合能获得更高更显著的超额收益;本文基于不同的已实现高阶矩估计方法和采样频率进行的稳健性检验也支持上述结论,并且这些组合收益并不能被CAPM的beta因子及Fama-French五因子所捕获。本文的研究还表明已实现峰度和基于低频数据的高阶矩所构造的投资组合分别抓住了未来收益中的不同的部分,使用高频数据计算出来的已实现峰度包含了更多的跳跃信息,具有显著的获取超额收益的能力。Different from the significant negative correlation between realized skewness and future returns of cross-sectional stocks in the U.S.market,this paper uses 15 years'high-frequency trading data of China's A-share market to construct realized higher moments.The empirical study finds that there is a significant negative correlation between realized kurtosis and future returns,which cannot be dispersed by other variables.Compared with realized volatility and skewness,it is a significant factor to understand cross-sectional asset pricing in A-share market.In addition,we find that the impact of realized higher moments,especially realized kurtosis,on stock prices is state dependent.Compared with the bull and bear market,in the volatile market,the long short portfolio based on realized kurtosis can obtain higher and more significant excess returns.The robustness test based on different realized higher moments estimation methods and sampling frequency also supports the above conclusion.These portfolio returns cannot be captured by CAPM's beta and Fama-French's five factors.This paper also shows that the portfolio constructed by realized kurtosis and higher moments based on low-frequency data capture different parts of future returns respectively.The realized kurtosis calculated by using high-frequency data contains more jump information and has significant ability to obtain excess returns.

关 键 词：已实现波动 已实现偏度 已实现峰度 截面资产定价 

分 类 号：C22[社会学] C23[文化科学] G11[理学—概率论与数理统计] G12[理学—数学] O212