Factor Model Comparisons with Conditioning Information

Abstract

We provide novel asymptotic tools for tests of asset pricing models and factor model comparisons when portfolios trade dynamically using lagged information. An Asymptotic Variance Lemma covers most of the tests in the literature that compare maximum squared Sharpe ratios. We develop finite-sample bias adjustments for Sharpe ratios of dynamic portfolios, include models with non-traded factors and validate the asymptotic results with simulations. We provide an estimator for the zero beta rate in the tests, and find values larger than the historical average risk-free rate when we use standard test portfolios. Factor models’ Sharpe ratios are improved with conditioning information, some more than others, but dynamic trading in the test asset portfolios is the larger effect.