Seok Young Hong (Lancaster University Management School)
Title:
Comparing factor models with conditioning information
Abstract:
We develop a framework to conduct asymptotically valid tests for comparing factor models with conditioning information. The tests are based on a metric analogous to the squared Sharpe ratio improvement measure that is used to gauge the extent of model mispricing in an unconditional setting. We propose an estimator for the metric and study its limiting properties in detail, establishing the asymptotic normality. An advantage of our framework is that it can be applied without an a priori knowledge of the persistence nature of the conditioning variables. We accommodate a range of dependence classes, including stationary, near stationarity, integrated, and local-to-unity. An application of our methodology to major models shows that the conditional versions of the Stambaugh and Yuan (2017) four-factor model and the Daniel, Hirshleifer, and Sun (2020) three-factor model are the best performers.