A Conditional Asset Pricing Model with the Optimal Orthogonal Portfolio

Abstract

This paper develops a conditional asset pricing model with latent factors, based on the optimal orthogonal portfolio approach. We construct a factor portfolio that embodies all the latent factors important for pricing a given set of test assets. The out-of-sample performance of this portfolio is evaluated against several conventional factors, using both cross-sectional and time-series regression approaches. Our results show that our suggested factor outperforms all the predefined factors regarding the contribution in return variations as well as the pricing ability. Our suggested methodology may have important applications in risk management, portfolio selection and performance evaluation.