How to Dominate the Historical Average?

We provide a new methodology to forecast the equity premium. We show, theoretically and empirically, that our forecast first-order stochastically dominates the historical average and incurs a better mean squared forecast error than the historical average. Our forecast combines the low variance advantage of the historical average with the bias reduction afforded by a predictor, thus outperforming the historical average. We distinguish the existence from the magnitude of return predictability. Using our methodology, we establish the existence of predictability for valuation ratios based on dividends or earnings, consumption-wealth ratio, investment, or bond returns