A BMWTV Approach to the Conditional Arbitrage Pricing Model

Abstract

This paper develops a BMWTV approach to the estimation of factor premiums by integrating the APT model of Burmeister and McElroy (1988) with time-varying risk premiums. It provides premium estimates for macro-factors over time under a unified APT framework which allows for both observable and latent factors. We find significant negative risk premiums for the market factor and the size factor during the sample period. We discover that risk premium and sensitivity estimates for the observable factors are quite sensitive to omitted latent factors, suggesting the importance of accounting for missing latent factors in conditional multi-factor models. We also find the mispricings under the APT model and the CAPM model are relatively small, but the results are quite sensitive to omitted factors. Our study shows that the variation of the size premium appears to be related to business cycles.