On the consistency of Bayes estimates