This paper evaluates the evidence on return predictability from an economic perspective: it asks whether investors would have been able and willing to exploit dividend price signals in order to allocate capital efficiently. We use a simple model that incorporates a time varying investment opportunity set into a mean-variance portfolio maximization framework, and derive the optimal capital allocation weights for: (i) a naive strategy based on average realized returns; and (ii) a class of strategies that condition on dividend-price signals. While our data supports in-sample evidence of return predictability, the out-of-sample returns of the naive strategy are higher than those of all conditional portfolio specifications based on a certainty equivalent metric and portfolio turnover. The degree of underperformance is most dramatic in the last three decades: an investor who had used dividend-price ratios as signals for capital allocation in the period 1990-2012 would have consistently generated lower returns than by following a naive strategy. These results suggest that dividend-price predictability offers no economic value to investors.