We forecast quarterly US stock returns using eighteen predictor variables both individually and in multivariate regressions, with the former also used in forecast combinations. Using rolling and recursive approaches, we consider a range of statistical and economic evaluation measures. We...