In this paper, we examine the relation between stock returns and analysts' heterogeneous expectations. We find that stock returns are positively associated with divergence of opinion. Our evidence provides no support for Miller's (1977) overvaluation hypothesis, which predicts lower (higher) future returns for high (low) divergence of opinion stocks in the presence of short-selling constraints. Our findings are based on the use of the diversity measure, which is free from the confounding effects of uncertainty in analysts' forecasts and is therefore a more accurate measure of divergence of opinion than dispersion. Our results refute the view that dispersion in analysts' forecasts reflects divergence of opinion. Our evidence is robust to the use of alternative measures of short-selling constraints, time intervals, optimism in analysts' forecasts, and herding in analysts' behavior.