Comparing Asset Pricing Models: The (Neglected) Impact of "Tradability"

Abstract

I use pairwise differences in cross-sectional R^2 to compare equity factor models (where all factors are traded) with ICAPM/macro models (where some factors are not traded) in cross-sectional asset pricing tests. Critically, I impose the theoretical restriction that the risk price estimates of traded factors are equal to the corresponding factor means. The results indicate that the traditional OLS R^2 fails in discriminating models. The reason is that such metric often relies on largely incorrect risk price estimates. One can improve substantially the power of statistical tests (when evaluating and comparing factor models) by imposing restricted factor risk premia.

Keywords: asset pricing models; traded risk factors; equity risk factors; macro factors; cross-sectional $R^{2}$; stock market anomalies; cross-section of stock returns; linear multifactor models; factor risk premia; ICAPM

JEL Classification: G10, G12