Choosing Factors in a Multifactor Asset Pricing Model When Returns are Nonnormal

We use Bayesian techniques to select the factors in a multifactor asset pricing model when the assumption of normally distributed returns is relaxed. More precisely, we assume that asset returns are multivariate t-distributed. This setup allows us to capture the well known fat tail property of asset returns. Interest rates, premiums, returns on broadbased portfolios and macroeconomic variables are included in the set of factors. Data from both the US market and the Swedish market are investigated