Comparative Analysis Of Stock Pricing Model Based On Bayesian Method

The Chinese stock market has only a short history of more than 20 years,but it has almost completed the development of the US stock market for 200 years.With the continuous expansion of Chinese opening up to the outside world,Chinese stock market will play an increasingly important role in the financial field of the whole world,and the theoretical research of the stock market can promote the development of Chinese stock market.Throughout the domestic research on Chinese stock market,the focus is on the applicability of multi-factor models in the Chinese market.Few people specifically compare the factors between models,and compare the multi-factor models from an empirical perspective.There are few,but the comparison of different construction methods of similar factors has not been involved.In recent years,foreign scholars have proposed a multi-factor model comparison idea based on all asset portfolios and a Bayesian comparison method for multi-factor models.In the empirical study of the stock multi-factor model,if a factor model can explain the excess returns of all assets,then the factor model is perfect.However,in the existing research results,there is no such factor model,which leads to different choices of factor models for different asset portfolios.Therefore,from the perspective of the explanatory variables,the existing factor model research is one-sided,and the research results can only show that the interpretation ability of a certain factor model based on certain asset portfolios is strong.In order to make the results more objective,when the two factor models compare pricing power,they should compare their pricing power to all asset portfolios,and this portfolio also includes the model and another factor model.The Bayesian comparison method can offset the non-factorial component portfolios in the calculation process,leaving only the factors that are not in the model,and then using the regression results of the factors to estimate the likelihood.The Bayesian comparison method compares multiple factor models and similar factors by Bayesian posterior estimation.Based on theoretical analysis,this paper uses Python to construct a 10-factor set of Chinese A-share market:MKT,SMB,HML,RMW,CMA,ME,ROE,IA,HMLm and UMD.These 10 factors can be grouped into six categories:market(MKT),market capitalization(SMB,ME),book-to-market ratio(HML,HMLm),profitability(RMW,ROE),investment style(CMA,IA)and momentum.(UMD).Then,the Bayesian model is constructed by python,and the Bayesian posterior probability of 162 models is calculated.The top 10 models,the cumulative probability of each factor and the cumulative probability of each major factor are obtained.The validity of the momentum factor,the reason of the difference between the optimal model at home and abroad,and the influence of the factor construction method of the market value classification on the factor validity are further analyzedChinese A-share market has the characteristics of short-selling restrictions,numerous retail investors and large liquidity.It is likely to have a liquidity premium and a speculative bubble.In this paper,the liquidity and speculative proxy factor turnover rate is included in the 10-factor set to form the 11-factor set of the A-share market:MKT,SMB,HML,RMW,CMA,ME,ROE,IA,HMLm,UMD and FMS.This set of factors can be grouped into seven categories of factors:market,market capitalization,book-to-market ratio,profitability,investment style,momentum,and turnover.Based on this,a Bayesian model is constructed,and the Bayesian posterior probability of a total of 324 multi-factor models is calculated.The top 10 models,the cumulative probability of each factor,and the cumulative probability of each major factor are obtained.The main conclusions of this paper are as follows:First,with all asset portfolios as the interpretation target,the optimal model of the A-share market under the 10-factor set is a 6-factor model consisting of CMA,HMLm,ME,MKT,RMW and UMD.In the optimal 6-factor model,although the cumulative probability of the momentum factor is small relative to other types of factors,the momentum factor can be tested by the redundancy test and the maximum Sharpe ratio increase test.Second,after adding the turnover rate factor,the optimal multi-factor model of the A-share market is a 7-factor model consisting of MKT,FMS,HML,UMD,RMW,CMA and SMB.In the increase of the maximum Sharpe ratio of the multi-factor model,the turnover rate factor FMS is much larger than the momentum factor.Third,in terms of building profitability factors,Chinese A-share market prefers operating profit rather than net profit.Fourth,the market value classification is a relatively good factor construction method,which can greatly improve the effectiveness of the factor.By comparing the momentum factor with market value classification and the momentum factor with no market value classification,it is found that the momentum factor constructed by market value classification is far superior to the no market value classification factor.The empirical results of Bayesian method also show that the optimal 5-factor model is better than the 6-factor model with no market-valued classification momentum factor,and the market-free classification momentum factor performs poorly.