Font Size: a A A

The Factors That Provide Independent Information About China's Average Monthly Stock Returns

Posted on:2021-05-20Degree:MasterType:Thesis
Country:ChinaCandidate:Y X WeiFull Text:PDF
GTID:2439330602489342Subject:Management Science and Engineering
Abstract/Summary:PDF Full Text Request
Since the CAPM model was proposed by Sharpe,Linter,and Mossin in 1964,the capital asset pricing model has continued to develop in practice.In order to improve the interpretation of the capital asset pricing model and better explain the stock return anomaly,Fama and French proposed a three-factor model in 1992 and a five-factor model in 2015,and Carhart proposed a four-factor model based on the momentum effect.Hou,Xue,and Zhang also proposed their four-factor models based on the q theory.These models explained the stock return from different perspectives.In addition to the above classic models,a large number of researchers have proposed their multi-factor models that are statistically significant to explain stock returns through empirical tests based on the world's stock markets.These research has made outstanding contributions to improving the efficiency of capital market pricing,enhancing the efficiency of the capital market,and promoting the development of capital asset pricing theory.Summarizing the results of previous studies from 1964 to the present,there are at least 430 factors that can statistically explain the changes in stock returns.It is certain that not all of these 430+factors can provide independent information to explain stock returns.There is a strong correlation between these factors,and the influence of some factors has also been confirmed being explained by other factors in some literatures.So far researchers have not concluded which factors can provide independent information to explain stock returns in this large factors.So John H.Cochranechallenged researchers to identify the firm characteristic that provide independent information about average U.S.stock returns in his 2011 American Finance Association Presidential address.The purpose of the John H.Cochrane challenge is to summarize the research results of the past 50 years and find out which of these 430+factors can accurately and efficiently explain the average stock returns.In this paper,38 feature factors are selected from 330 features listed in Green,Hand,and Zhang(2013)that have been proven to significantly explain stock returns in previous literatures.Through the WIND financial database,the monthly returns and data of 38 factors of 841 stocks in the Shanghai A-share market from January 1,2007 to January 1,2019 were obtained.In order to keep a sufficient sample size as much as possible while maintaining the consistency of the dimensionality of the coefficients of each factor,this paper first performs a Winsorize process on the stock data set at 10%and 90%quantiles,and then transform the data of each factor to standard normal distribution.At last,the missing values of the factor data are filled with 0(the mean value of each factor).The overall method in this paper follows the two-step regression method of Fama-Macbeth(Fama,Macbeth 1973).This paper innovates in the following two aspects.First,all 38 characteristic factors are used as explanatory variables simultaneously in Fama-Macbeth regression in order to find factors providing independent information about average stock returns.Second,in the cross-section data regression(the second step of Fama-Macbeth regression),the LASSO method and the Elastic Net method were used to screen factors,respectively.These two method aimed to solve the multicollinearity problem by adding a penalty function in the OLS regression,which are more efficient methods when facing large number of explanatory variables.The factor models consturcted by the LASSO method and the Elastic Net method are modified by deleting the factors that are not significant in the t test.Comparing the two models,this paper conclude that the multifactor model consturcted by the Elastic Net method is better than the LASSO method.The model constructed by the Elastic Net method is composed of four factors:momentum in the past 1 month,maximum daily return in the past 1 month,market excess return,and working capital.In order to test the efficiency and accuracy of the four-factor model proposed in this paper,based on the same data,this paper validates the Cahart's four-factor model,the five-factor model of Fama and French,and the four-factor model of Hou,Xue and Zhang.There are factors in these three classic models that cannot statistically explain stock returns.Comparedas a whole,the factor model consturcted by the Elastic Net method in this paper performs better than the three classic models based on China's Shanghai A-share market.In order to test the interpretation ability of the four-factor model selected in this paper on the generalized data set,this paper selects 579 stocks from the Shenzhen A-share market to construct a generalized data set within the same time period,and adopts the same data processing method.Through Fama-Macbeth regression,wefound that the four-factor model proposed in this paper based on the Shanghai A-share market data also has a good explanatory ability on the Shenzhen A-share data set.Analysing the multifactor model proposed in this paper,the momentum of the past 1 month and the daily maximum return of the past 1 month can be categorized into momentum factors,and the market excess return is regarded as systematic risk factors.Therefore,this paper proposed that average stock returns of China's A-share market are mainly affected by momentum and systematic factors.So the stocks of companies that have performed well in the past tend to perform well in the future resulted by the momentum effect.
Keywords/Search Tags:Capital asset pricing model, Fama-Macbeth Regression, Large factors set, LASSO, Elastic Net
PDF Full Text Request
Related items