Research On The Effectiveness Of Pricing Factors In China's Stock Marke

Sharpe proposed the Capital Asset Pricing Model(CAPM)in 1964 to explain the relationship between individual asset returns and market portfolio returns.Ross proposed the Arbitrage Pricing Theory(APT)in 1976 to extend the impact of asset prices from a single factor to multiple factors.After that,a large number of scholars did research on asset pricing.In the research on stock cross-sectional return prediction,the anomaly factor is the factor that can predict the cross-sectional return of the stock market and cannot be explained by the existing factor pricing model.At the same time,scholars are constantly improving or rebuilding new factor pricing models to explain these anomalies.Such as the three-factor and five-factor models proposed by Fama &French(1992,2015),and the four-factor model proposed by Carhart(1997),the Qfactor model proposed by Hou et al.(2015)and the CN three-factor model proposed by Liu et al.(2019).The chairman of the American Finance Association pointed out that academic research has put forward hundreds of factors that can predict the stocks cross-sectional return since 1970,which can form a "Factor Zoo"(Cochrane,2011).How do these pricing factors and models perform in Chinese stock market? Why these anomalies existed? Studying these issues is of great significance for understanding the asset pricing,forecasting stock return and measuring market effectiveness.This paper uses China’s stock market data from April 2004 to April 2020 to study the effectiveness of pricing factors in China’s stock market,discuss the effectiveness of factors,the reasons for the effectiveness,and the actual value of these factors.In Chapter 3 of this paper,referring to the research of Harvey et al.(2015)and Green et al.(2017),we constructed hundreds of pricing factors,and selected 40 factors from hundreds of pricing factors(Anomaly factor),analysis the effectiveness of multifactor models and the existence of the anomalies.The factor pricing models we tested include the Fama French three-factor model,the Fama French five-factor model,the Q-factor model(Hou et al.2015)and the CN4 factor model(Liu et al.2019).The results show that nearly half of the factors’ t values greater than 3.0 in single-factor Fama-Mac Beth regression and 5 factors’ t values greater than 3.0 in market valueweighted least squares regression(WLS).In the Fama-Mac Beth regression after considering the pricing model,most factors especially calculate by using price and trading volume data are still significant.When putting all the factors together and using WLS regression,only 6 of the 40 factors are significant,which are maximum daily return(12 month),idiosyncratic volatility(1 month),max price(1 month),standard deviation of turnover rate(1 month),cash to size ratio and size.The results of this paper show that in the Fama-Mac Beth regression,none of the pricing models examined in this paper can explain the anomaly well.When using the long-short portfolio for regression testing,the China four-factor model(CN4)proposed by Liu et al.(2019)has the best ability to explain anomalies,the worst is the Fama French three-factor model.In order to understand the source of anomalies,chapters 4 and 5 of this paper study the effectiveness of factors from two aspects: financial information release and retail investor attention.Scholars such as Engelberg et al.(2018)and Bowles et al.(2020)found that anomaly return is related to the release of financial information.Therefore,we choose 30 factor from three categories to study the relationship between financial information release and anomaly return.First of all,this chapter studies the compare the anomaly return at different update frequencies.The results show that after increasing the data update frequency,the returns of most factors have increased significantly.These results are similar to the results of Asness & Frazzini’s(2013)research.After that,this chapter uses the cross-sectional regression and the portfolio analysis to study the time-varying anomaly return and information release.All results show that profit,growth and some valuation factors has significant time-varying characteristics related to the release of financial information.In the first month after financial statement released(May,September,and November),the profit and growth factors return are insignificant or significantly negative,and the factor returns mainly exist in the second month.The results also show that these factors’ anomaly return in the second month after financial statement released is significantly higher than in other months.However,the anomaly calculated by price and volume did not find similar time-varying characteristics.The anomalous returns of idiosyncratic volatility and volume volatility were significantly negative in each month,and the continuity of the anomaly returns was often also only 1 month.In addition,there are certain differences between the performance of the anomaly of profit growth in the month of information release and other months,the anomaly returns in these month is not significant or significantly negative.Further research found that after May 1 and November 1,the changes in earnings and growth factor returns have a common trend with the one-month reversal factor.These results show that the fluctuation of anomaly returns may be related to price overreaction.When new information is released,the market has overreacted to profit or loss information,and prices often fluctuate sharply,In the following months,there will be cyclical fluctuations in prices,resulting S-shape change of anomaly returns,and then gradually weaken until it stabilizes or the next new information is released.The results of this paper are consistent with the over-reaction and under-reaction theories proposed by Hong & Stein(1999),which indicates that the growth and profit anomalies resulted from market mispricing.Da et al.(2011),Yu Qingjin,Zhang Bing(2012),Song Shuangjie and other scholars(2011)’s research all show that investor attention has an impact on stock prices,and retail investor behavior,especially investors attention can help us analyze the source of anomalies’ return.Therefore,the chapter 5 of this paper uses the data posted by Eastmoney Stock online stick from 2010 to 2019 to construct indicators of retail attention and attention changes,and uses Fama-Mac Beth regression and group analysis to study the relationship between investor attention and the cross-sectional anomaly of the stock market.The research results show that most anomalies based on price and volume can be absorbed by attention in the Fama-Mac Beth regression except return skewness,idiosyncratic volatility,turnover rate fluctuations and volume fluctuations.Financial anomalies and analyst coverage can’t be absorbed by investor attention.From the perspective of behavioral finance,investor attention is scarce,and the information processing capabilities of retail investors are weaker than professional investors.The excessive attention of retail investors often increases the noise part of the price and reduces market efficiency.Therefore,comparing the performance of anomalies in different groups can helps us understand the source of anomalies’ return.This chapter analyzes the heterogeneity of anomaly return,and examines whether there is a significant difference in anomaly return between different attention groups.Except skewness and liquidity,most price and volume anomalies have obvious heterogeneity,which is more significant in high-attenttion group.The heterogeneity in financial valuation anomalies is more complicated.Some indicators,such as analyst coverage、cash ratio and change of quick ratio,show no significant difference in factor return in different groups.Indicators such as ROE or ROA have obvious heterogeneity.These results are consistent with Chapter 4,which means some financial anomalies may be due to mispricing.Doing research on pricing factors is helpful to understand the operation of the stock market,find mispricing in the market.Market participants can also eliminate mispricing through trading on these anomalies.With the popularity of big data methods in recent years,how to apply big data technology to the financial market is also one of the hot research topic.Chapter 6 of this paper uses rolling Fama-Mac Beth regression,rolling linear regression,elastic network regression,partial least squares regression,random forest and gradient boosting tree regression to predict stock cross-sectional returns,the results show that all method has a strong ability to predict cross sectional stock return.Different from existing studies,the Fama-Mac Beth regression and random forest performs well in prediction of long-short portfolio returns.In terms of long-term earnings forecasts,OLS performed best,the nest is random forest.The portfolio analysis result show that excess return in long portfolio dropped significantly in recent years,which may be related to the popularity of machine learning algorithms and increasing in arbitrage traders.The research results of this article provide evidence for financial technology to improve market pricing efficiency and enhance market effectiveness.This paper studies the importance of features in forecasting and timevarying returns and finds that the price-volume features have a higher weight in the random forest model.This may be related to the machine learning model using more market volume and price information.Due to the time-varying feature of financial anomalies,the stability of financial factors is low,and their feature weights in machine learning models are also lower.The time-varying feature of factor returns is a big challenge for using machine learning to predict stock returns.