引入投资者关注的中国股市协方差预测——基于多元HAR类模型  

Investor Attention and Covariance Forecasting in China’s Stock Markets——A Study Based on the MHAR Type Models

作　　者：瞿慧[1] 沈微 QU Hui;SHEN Wei(School of Management and Engineering,Nanjing University,Nanjing 210093,China)

机构地区：[1]南京大学工程管理学院,江苏南京210093

出　　处：《中国管理科学》2022年第7期9-19,共11页Chinese Journal of Management Science

基　　金：国家自然科学基金资助项目(72171110,71671084)。

摘　　要：众多经济事实表明投资者并非完全理性。一方面,投资者由于有限关注,无法及时掌握市场上所有投资决策相关信息,可能导致资产价格对信息的反应不足;另一方面,某些信息会诱导投资者过度关注和过度交易,导致价格信号中包含更多噪声。这些都会造成短暂的错误定价,引起资产价格波动和资产间相关性的变化。鉴于此,本文认为投资者关注是影响资产多元波动率的一个重要外生因素,用百度指数衡量中国市场个体投资者关注,将其引入已实现协方差的多元异质自回归类模型,刻画个体投资者关注的变化对资产协方差的非对称影响,同时区分电脑端和移动端百度指数对协方差预测的不同贡献。采用2014年1月2日至2018年12月28日的50ETF成分股高频价格,和以股票简称及“50ETF”为关键词的百度指数,对上述模型进行实证。结果表明,百度指数代理的个体投资者关注蕴含对协方差预测有益的信息,将其引入协方差预测模型显著提升拟合性能和样本外预测能力;投资者关注的变化对资产波动及相关性的影响均存在非对称性;引入电脑端百度指数比引入移动端百度指数对协方差预测性能的提升更为显著。研究结果肯定了引入投资者关注对协方差预测的积极作用,对投资者的资产配置和风险管理有实际指导意义。Economic facts disclose that investors are not completely rational.On the one hand,investors can’t possess all the information in the financial markets due to limited attention,which causes the under-reaction of prices to some information.On the other hand,some information may induce investors’over attention and over trading,which results in more noises in the price signals.These may cause temporarily mispricing of stocks,which generates volatility and correlation in the market.Based on the above analysis,this paper deems investor attention as an important exogenous variable to stocks’covariance;uses Baidu index to measure individual investors’attention in China,and introduces it into the multivariate heterogeneous autoregressive(MHAR)model of realized covariance matrix.The asymmetric influence of investor attention variation on volatility and correlation are studied and characterized.Furthermore,the contributions of the PC oriented Baidu index and the mobile Baidu index to covariance forecasting are distinguished and compared.The high-frequency prices of seven 50 ETF component stocks,and the Baidu indexes with corresponding stock abbreviations and“50 ETF”being the keyword are used as empirical data.The empirical period is from January 2,2014 to December 28,2018.The results show that,investor attention contains valuable information to covariance forecasting,introducing it into the covariance forecasting model can significantly improve the in-sample fit and out-of-sample forecasting performance.The influence of investor attention variation on volatility and correlation is asymmetric.The PC oriented Baidu index contributes to covariance forecasting more significantly than the mobile Baidu index.The above results confirm the contribution of individual investor attention to covariance forecasting,which can provide practical guides for asset allocation and risk management,and enrich the empirical research on covariance modelling based on high-frequency data.

关 键 词：已实现协方差 MHAR-DRD模型 百度指数 个体投资者关注 非对称影响 

分 类 号：F830.9[经济管理—金融学]