Cluster Analysis-based Anomaly Factors In The Chinese A-Share Market

Existing asset pricing studies witness too much anomalies factors such that there is no way to analyze simultaneously.There are common components for different factors to affect the return.Using a single anomaly to construct a factor will be disturbed by data noise,and the significance of a single factor varies in different historical periods.It is intuitive to combine the information in multiple anomalies to construct a factor.Cluster analysis,an unsupervised learning technique,is used to group a set of objects in such a way that objects in the same group(called cluster)are more similar to each other to those in other groups(clusters)under the condition that the labels are unknown.This paper selects a set of 15 widely used factors in the Chinese A-share market,applies cluster analysis to exploit their common components,and creates two factors based on the obtained two clusters,i.e.,F1 and F2.The first type factor F1 contains 7 basic factors,including indicators related to the business operation status;the second type factor F2 has 8 basic factors,including relative valuation indicators and investment growth indicators.Fama-Macbeth regression of clustering factors F1 and F2 shows that F1 and F2 can significantly explain stock returns.After the controlling the other variables,the regression coefficients of F1 and F2 on stock returns are still significant,and the signs are unchanged.In order to further verify the above findings,this paper adds F1 and F2 factors to the pricing model,and constructs a four-factor model including MTK,SMB,F1 and F2,and take regression on the original 15 factors.The empirical results show that the F1 and F2 factors get the information that the original pricing model cannot explained,and have a better performance than the other models.This paper use different algorithm to cluster the 15 factors in the robustness test.The results are not fundamentally different,and the empirical results are robust.The empirical results of this paper question and supplement the original factor pricing model in Chinese stock market.The existing model cannot explain the source of income of the F1 and F2,and the empirical results confirm that the higher ? coefficients of F1 and F2 are lower.Compared with the original classical model,the clustering factor model perform well in explaining the anomanies.In addition,this paper argues that in the real market,due to the irrational behavior of investors and financing constraints,the high ? stock alpha value is lower.