Empirical Study Of RRA-Based Multi-Factor Model In China’s Stock Market

The hotness of asset pricing in finance has been a constant.The A-share market in China,having commenced at a later stage,is still in a state of low efficiency,with intricate market alterations.He,Huang,Li,and Zhou’s(2022)Reduced-Rank Approach(RRA)is employed in this paper to reduce the factor dimension,extract a few from a multitude of original factors associated with company characteristics,and construct a multi-factor model to explain and forecast the return of assets in China’s A-share market.This paper uses the monthly data of A-share listed companies from January 2002 to December 2021 as a research sample,and constructs four sets of test assets consisting of two sets of investment portfolios and two sets of individual stocks.Using RRA,PCA,and PLS,a few factors were taken from the 99 original ones to assess the pricing performance of the various factor models produced by different dimensionality reduction techniques,and to contrast them with the Fama-French multi-factor model.Exploring the effect of RRA factors’ pricing performance with pricing error restrictions,this paper is secondly presented.Then,this paper gives pre-specified five Fama-French factors,and then extracts other pricing factors from the remaining original factors through RRA,PCA and PLS methods to study whether there are other pricing factors that can explain the cross-section of expected stock returns and provide incremental information.Finally,this paper conducts a robustness test by replacing basic assets and testing assets,and further studies whether the pricing performance of the RRA method is effective.This paper’s empirical analysis reveals that,although the 99 original factors’ factors extracted by RRA enhance pricing performance at the portfolio level,they do not yield much new information in comparison to CAPM at the individual stock level.At the same time,the RRA method has a better performance in explaining the return of the investment portfolio,and is better than PCA,PLS and other methods.However,compared with the Fama-French five-factor model,the multi-factor model composed of factors extracted from a large number of original factors through RRA can hardly provide more useful incremental information for asset pricing.