A Bayesian solution to the equity premium puzzle