A New Pricing Factor Construction Method And Its Application In China

In China,the stock market has become one important investment channel for investors.When investors choose different portfolios,they will be faced with different risk.Risk compensation has become an important pricing factor,thus seeking effective stock returns pricing factors is of great significance.In this paper,we propose a new factor construction method in order to improve the pricing ability of the model.Specifically,we argue that factors constructed based on one single anomaly variable will contain more noise,which in turn will affect the pricing model.Based on this analysis,we utilize three indicators that can reflect firm's profitability to build a comprehensive profit factor,namely,SPF;and utilize three indicators that can reflect firm's investment size to build a comprehensive investment factor,namely,SIF.Combined with MKT and SMB,we construct one four-factor model,namely,SCM.When using this model to price HML and UMD,results indicate that they no longer contain additional pricing information when considering those four factors.After regressing SPF on MKT,SMB,HML,UMD,SIF,and removing factor common information,we obtain another comprehensive profit factor SPFN.Accordingly,we construct another four-factor model,namely,SCMN.In the empirical part,firstly,we examined the pricing power of MKT,SMB,HML,UMD,RMW,CMA,SPF and SIF.The results show that MKT outperforms the rest factors,followed by SMB and SPF.Then,we conduct random sampling method to construct the test portfolios,so that they don't contain any priori information.The result shows that only MKT is selected into the model.The above results show that when we construct a test portfolio based on some obvious variables,it will contain priori information,thus having a preference for some corresponding factors,which will lead to biases in model tests.In the model test section,the seven pricing models involved are tested based on the 18 test portfolios.The results show that the SCMN and SCM outperform other models.Finally,we carry out the robustness analysis through three methods to achieve two goals:the test portfolio contains more priori information,and we can obtain sufficient test portfolios.The results show that when comparing the optimal performance numbers,SCMN is the best,followed by SCM;when considering the proportion of non-rejections,SCMN is the highest;when comparing model stability,SCMN outperforms others,followed by SCM.Through the above robustness analysis,it is confirmed that the two models constructed in this paper have higher pricing ability.In addition,we utilize the US market data to do another robustness test,proving that our factor construction method is still useful.Based on our reseach results and the latest research literature,we believe that further research can be carried out in the following two aspects:first,testing the the difference between two non-nested factor models;second,how to contain more information in the pricing model.To propose a new method,so that the pricing model with good econometric nature can contain more information is of great significance.