A new test of factor model for asset returns: based on pleiotropy model
提出一种基于多效性模型的检验方法，通过判断时间序列回归中非零截距的数量来比较不同资产定价模型的有效性，模拟和实证表明该方法优于传统GRS检验。
The validity of factor model for asset returns relies on testing the null hypothesis of whether the intercepts are jointly indistinguishable from zero, which, however, is often rejected, resulting in imperfect comparison among competing asset pricing models. In this paper, we consider a new test based on the pleiotropy model, through which we could determine how many intercepts in the time-series regressions in portfolio-level are statistically non-zero and thus provide a more direct way to compare the validity of competing factor models. A sequential testing procedure regarding the estimated parameters under unconstrained and constrained conditions is proposed, based on some likelihood ratio test statistics. Simulation studies suggest that our proposed test statistics have better performance than the influential GRS test. Finally, we apply the proposed method to real data of portfolio returns with Fama-French five factor models, and demonstrate that it provides important and irreplaceable information for determining the efficacy of different asset pricing models.