Firm Characteristics and Empirical Factor Models: A Model Mining Experiment
通过数据挖掘数百个随机构建的三因子模型，发现许多模型优于文献中知名的四因子和五因子模型，表明因子模型成功的门槛需要提高。
Abstract In a novel model mining experiment, we data mine hundreds of randomly constructed three-factor models and find that many outperform well-known models from the literature, including those with four and five factors. The results provide compelling evidence that the threshold of factor model success needs to be raised. Confidence intervals for model rankings, derived from a bootstrap simulation, offer new insights into the consistency of a model’s pricing ability. Rankings for some well-known models are unusually volatile, which have wider confidence intervals than that of most of the random factor models.