法玛—弗兰奇五因子模型的动态分析*
DYNAMIC ANALYSIS OF FAMAFRENCH FIVEFACTOR MODEL

摘要 本文采用动态模型平均（dynamic model averaging，DMA）算法对法玛—弗兰奇五因子模型（FamaFrench fivefactor model，FF5模型）进行了系统性研究。基于法玛和弗兰奇（Fama and French）的数据，笔者所进行的实证分析表明：资产定价模型因子对投资组合收益的预测能力和系数是随时间动态变化的；不存在固定的因子模型能够同时解释和预测多种投资组合的收益；在对投资组合收益率进行预测时，DMA算法的预测均方误差（mean square error，MSE）显著低于固定系数的FF5模型。
Abstract： In this study, we employ the dynamic model averaging (DMA) method to analyze dynamic behaviors of FamaFrench fivefactor (FF5) model. With the same datasets of Fama and French, our results show that the predictive power and effectiveness of FF5 vary all the time. No fixed factormodel can explain and predict the returns of different portfolios simultaneously. The prediction mean square error (MSE) of DMA method is significantly lower than FF5.