Survey-based equity premium forecasts are procyclical, less volatile, and more persistent than countercyclical statistical forecasts, a pattern the literature finds puzzling and has tried to explain with behavioral models. We show these patterns arise in disagreement models because prices aggregate beliefs differently than surveys. Prices reflect the market view, a consumption-share- and risk-tolerance-weighted average, rather than the survey consensus, which ignores gains from trade. We derive conditions for a negative correlation between consensus and statistical premium and show that disagreement about cash flows or discount rates induces contrarian and trend-chasing positions, generating this negative correlation.