Choosing Factors in a Multifactor Asset Pricing Model: A Bayesian Approach

We use Bayesian techniques to select factors in a general multifactor asset pricing model. From a given set of 15 factors we evaluate all possible pricing models by the extent to which they describe the data as given by the posterior model probabilities. Interest rates, premiums, returns on broadbased portfolios and macroeconomic variables are included in the set of considered factors. Using different portfolios as the investment universe we find strong evidence that a general multifactor pricing model should include market excess return, size premium, value premium and the momentum factor. There is some evidence that yearly growth rate in industrial production and term spread also are important factors.