Exploring Multifactor Models: Horse Races, Forecasts, and Bootstrap

Abstract

Multifactor asset pricing models play an important role in evaluation of anomalies and managed portfolios. In empirical studies, however, the researchers' prior often includes a rich set of competing models which all seem plausible, but none is conclusively dominant. In this paper we propose a unified approach to model selection and inference for application and evaluation of factor pricing models. We run horse races among a set of high profile models to select winners according to predictive ability and then utilize White's (2000) reality check methodology to test the top performers. We provide empirical results on testing multifactor explanations of industry momentum, and present a test of the three moment CAPM.