We propose a measure of disagreement, which reflects differences of opinion as opposed to information asymmetry, that can be extracted from sequences of analyst forecasts. Using a Bayesian theoretical framework, we prove that, when analysts agree, a regression of an analyst's forecast on the previous forecast issued by another analyst should have a slope coefficient of one. The magnitude of the estimated regression coefficient's deviation from one is then employed as a disagreement measure. We validate the measure using tests tied to predicted relations between disagreement and trading volume and bid-ask spreads. Finally, we employ our measure to test for associations between disagreement and expected returns predicted by antecedent theoretical studies.