| Due to the relatively late start of China’s capital market,the relevant research on the factor pricing model of A-share market refers more to the research results from foreign mature markets,such as FF-3 and FF-5 factor models.However,there are many differences between the A-share market and foreign mature markets,which will inevitably affect the factor pricing model of the A-share market,and one of the biggest differences between the two is the investor structure.There are many individual investors in the A-share market and there is an obvious 80/20 phenomenon,that is,the market value of individual investors’ positions only accounts for 20% of the market value of the whole market,but they contribute 80% of the trading volume.Coincidentally,Game Stop,AMC and other network meme stocks,which have been popular in the US stock market since 2021,have attracted many individual investors.The changes in their share prices reflect the differences in the pricing of stocks between the two different types of investors when the investor structure changes from institutional dominance to individual dominance.On the premise of this fact,when there are obvious differences in the types of investors,the assumption of homogeneous belief in traditional finance is obviously no longer valid,while the study of investor groups in behavioral finance is just in line with the current situation of the A-share market.Compared with the concept that the factor premium originates from systemic risk in traditional finance,Barberis and Shleifer(2003)proposed the style investment hypothesis that the factor premium originates from investors’ trading demand,and the investor structure of the A-share market is bound to bring about heterogeneous trading demand.Based on the heterogeneous belief theory of investor groups in behavioral finance,and according to the unique investor structure of the A-share market,this paper analyzes how the difference in investor structure affects the three dimensions of empirical asset pricing,namely,anomaly factor premium,factor pricing model and factor investment strategy,so as to help researchers,investors and regulators better understand the pricing system of the Ashare market,And make contributions in both theoretical and empirical aspects.Firstly,this paper discusses how to identify the investor structure.It is found that jump returns reflect the trading needs of overreacting investors,while non-jump returns reflect the trading needs of underreacting investors.In the A-share market,the former are mostly individual investors,while the latter are mostly institutional investors.Specifically,this paper discusses from four dimensions: investor behavior,herding efficiency,fundamental information and investor sentiment.The research finds the following four points:(1)Jump returns are positively correlated with the number of shareholders,while non-jump returns are positively correlated with the proportion of institutional holdings,and there is a tug-of-war phenomenon between jump and nonjump returns;(2)The jump income is significantly related to the intraday herding effect,and the jump direction is consistent with the herding effect direction;(3)The PEAD phenomenon does not exist in jump returns,but it is significant in non-jump returns;(4)Jumping returns are positively correlated with investor sentiment.Based on the above empirical results,this paper believes that jump returns reflect the trading needs of overreacting investors.In addition,this paper tests the relationship between jump returns and other overreacting proxy variables from the cross section and time series.It is found that the overreacting proxy variables are no longer significant after controlling jump returns,otherwise,jump returns can explain other overreacting proxy variables.In addition,this paper also finds that the momentum effect of A-share market originates from non-jump returns,while the reversal effect originates from jump returns.Therefore,this paper measures the investor structure at the individual stock level of the A-share market through the contribution of jump income and non-jump income to total income.The investor structure with higher jump income contribution is dominated by overreaction investors,while the investor structure with higher non-jump income contribution is dominated by underreaction investors.Secondly,this paper studies the influence of investor structure on factor premium,and finds that the source of factor premium in cross section is consistent with that in time series.If the anomaly factor exists significantly in the stocks with high jump income contribution,then its factor premium also mainly comes from jump income,and there is a similar phenomenon for non-jump income.Referring to the research on the US stock market by Hou et al.(2015)and Feng et al.(2020),as well as other research on the anomaly factors of the A-share market,this paper has constructed 188 cross-section factors,including 15 momentum factors,20 value growth factors,37 investment factors,30 profitibility factors,36 intangiblefactors,24 trading friction factors and 26 behavior factors.This paper uses CAPM model,CH-3 model and FF-5 model to test whether these factors have significant market anomalies in cross section and time series.Only when the three models are significant can the anomalies be considered significant.The results show that 41 of the 188 cross section factors are significant in cross section and28 in time series in the whole market.By decomposing the cross sectional sub samples of different investor structures and factor premiums in the jump and non jump parts,this paper divides the 188 cross sectional factors into the following four categories based on the CAPM model:(Ⅰ)The pricing factors with consensus are significant and have the same direction within the sub samples and after the income decomposition;(Ⅱ)The pricing factors with divergence are significant and opposite in the sub sample and after income decomposition;(Ⅲ)The "alpha" factor,which cannot be used as a pricing factor,is only significant in a certain part within the sub sample and after revenue decomposition;(Ⅳ)Ineffective factors are not significant within the sub sample and after income decomposition.There are 20 type Ⅰ factors,33 type Ⅱ factors,66 type Ⅲ factors and 69 type Ⅳ factors.In addition,this paper also examines the use of market value and institutional investors’ shareholding ratio to cut the sub sample,and finds that the return contribution can better distinguish the factor premium of anomaly factor s in different sub samples.Then,this paper analyzes the impact of investor structure on the construction of factor model.It is found that adding share turnover and nominal share price to the CH-3 model can significantly increase the explanatory power of the model in cross section and time series.Firstly,we use stepwise regression and lasso model to screen out the factors that have significant explanatory power to jump returns and non-jump returns in cross section and time series from all the anomaly factors.Then,this paper constructs cross section and time series factor pricing models respectively.The cross section model uses the factor value directly,while the time series model uses the factor premium calculated from grouping by the factor value.It is found that,compared with the original CH-3model,the explanatory power of the improved model for the change in cross section returns is increased from 5.43% to 8.32%,and the explanatory power of the change in time series returns is increased from 40% to 42.5%,and the ability to explain anomaly factor s in cross section and time series is improved,and the above improvements are statistically significant.Finally,this paper discusses the impact of investor structure on factor investment strategies,and finds that stock selection and timing based on investor structure can effectively improve factor long short returns.Specifically,the difference in investor structure of different stocks on the cross section can improve the anomaly factor long short returns through the cross-section factor stock selection strategy,while the change in the whole market investor structure on the time series can improve the anomaly factor long short returns through the factor timing strategy.For all anomaly factors,the average monthly CH-3 model alpha and FF-5 model alpha of long short portfolio bought and held in the whole market are 0.34% and 0.32% respectively.After considering the investor structure,the average monthly CH-3 model alpha and FF-5model alpha of the cross-sectional factor stock selection strategy significantly increased to 0.53% and 0.51%,both of which increased by more than 50%;The timing strategy of time series factor was significantly increased to 0.42% and 0.38%,with an increase of nearly 20%.Among the factors of all types,the factors of type Ⅱ and type Ⅲ defined in this paper have the greatest improvement in cross section,while the factors of type Ⅱ have the greatest improvement in time sequence,while the factors of type I and Ⅳ have no significant improvement.Compared with the existing literature,the innovation and contribution of this paper are mainly reflected in the following four aspects:(1)in the study of investor heterogeneity,this paper uses the method of income decomposition to identify the types of investors.In the existing literature,the discussion on investors’ heterogeneous beliefs is usually based on the construction of indicators representing the degree of divergence of investors’ opinions.This paper creatively distinguishes the types of investors and provides a method to judge whether a certain type of investor has the pricing power on a certain stock in a certain period.(2)In the research of market anomaly factors,this paper combs and reappears the anomaly factors at home and abroad,and classifies them into agreed pricing factors,divergent pricing factors,"alpha" factors with limited scope of use and invalid factors.On the basis of testing the effectiveness of factors,this paper discusses the impact of investor structure on factor premium from the two dimensions of cross section and time series,and finds out those pricing factors that are not significant in the whole market but are actually important due to investors’ heterogeneous beliefs(3)In the study of factor model construction,this paper constructs a factor pricing model for the returns reflecting the trading needs of different types of investors based on the recurring anomaly factors,What’s more,we can find the pricing factors that can explain the concerns of both types of investors,and build a stock factor pricing model in the A-share market,so as to better explain the anomaly factor premium in the A-share market.(4)In the research of factor investment,this paper provides another perspective for factor stock selection and factor timing.It judges how to practice factor investment strategies according to the investor structure,that is,when and how to use which factors on what stocks.This paper improves the application of factor investment in practice and provides a reference for how to continue to carry out factor strategy research. |
PDF Full Text Request