Research On The Factor Structure Of Stock Return In China

The Arbitrage Pricing Theory of Ross(1976)posits that When there is no arbitrage opportunity in the market,only the loadings on the pricing factor and the pricing factor affect the return.However,the arbitrage pricing model does not specify the quantity and type of risk factors,nor does it indicate the sign and size of each factor's risk premium.In the practical application of the factor model,scholars often first observe the anomalies existing in the market,find the common factor between the assets,and then test whether the factor can be used as the pricing factor to affect the expected rate of return.For a long time,according to the CAPM,only a single market factor was widely applied to the prediction model.Fama-French(1993)added SMB,HML to the model to construct a three-factor model.Since then,a large number of studies have found a series of anomalies that violate the three-factor model.Among the anomalies that these three-factor models cannot explain,some are regarded as pricing factors by scholars.Hou,Xue,and Zhang(2015a)add investment and profitability based on market factors and scale factors to construct a four-factor model.Fama-French(2015)also added investment and profitability factors to the model based on the three-factor model,and constructed a five-factor model.The number of anomalies discovered in this process is increasing,and the number of factors affecting asset common factors is also increasing,and the number of factors added to the pricing model is also increasing.Cochrane(2011)call for a reorganization of the factor structure of stock returns to decide which factors are the most important.Which factors should we be writing deeper macroeconomic models to explain?This paper takes all the A-share listings in the market from January 1996 to December 2017 as the research object,and selects the monthly transaction data and financial data of all listed companies as the original samples.The first step,the Fama-MacBeth cross-sectional regression is used to estimate the expected return rate of each stock based on firm characteristics.The second step is to construct the stock portfolio with the expected return rate as the only indicator.The last step uses principal components analysis to extract factors from the portfolio returns.Finally,the paper extracts the first principal component factor that exhibits the level structure and the second principal component factor that exhibits the slope structure.Although the number of words is only two,these two factors can effectively summarizing the key features of the cross-section of stock returns.This procedure enhances the factor structure associated with the expected rate of return,offsets the interference of information that is not related to the expected rate of return,and ensures that the factors that can be extracted are the risk factors involved in the formation of the expected return on the stock to find the most economically important factor,narrow the existing factor space,and simplify the factor structure.I perform for time series test and cross-sectional test.Using the level factor and slope factor models proposed in this paper compared to several leading Fama-French as three-factor model,four-factor model,five-factor model and finally calculate the number of portfolios for each model Hansen,Jagannathan(1997)distance,and the intercept is not equal to zero to measure the ability of each model to explain the rate of return.Both the time series test and the cross-sectional test results show that the two-factor model of the level factor and the slope factor is better than the other models in explaining the yield cross section.Finally,with reference to Cochrane's(2005)Hourse Race test,a cross-sectional test is performed again after adding a single factor to the two-factor model.Finally,I follow the procedure in Cochrane(2005)to conduct factor horse races.I run ordinary least squares regressions with returns on each individual asset pricing factor.When the estimated coefficient is added to a cross-sectional asset pricing test with other factors,the resulting coefficient estimate yields the marginal significance of the factor.If a factor is insignificant,it adds little explanatory power to the model.The level factor and slope factor models showed good explanatory power in all tests.The regression coefficients of the level factor and the slope factor are significantly different from zero,both of which are pricing factors.At the same time,the increase factor does not improve the adjustment R2.Overall,the horse races suggest that the level factor and slope factor model largely succeeds at the goal of summarizing the key features of the cross-section of stock returns.