Cyclicality of uncertainty and disagreement

The empirical literature often uses disagreement (dispersion in forecasts) as a proxy for uncertainty, yet disagreement and uncertainty behave differently throughout the business cycle. The difference is especially salient in non-crisis periods, in which measures of disagreement are positively correlated with growth, while measures of uncertainty are negatively correlated with it. I explain this finding using a noisy information model with endogenous learning. In the model, agents observe noisy private information, but only when they are active. Holding uncertainty fixed, a rise in activity introduces noisy information to the market, and agents’ beliefs diverge, i.e., disagreement rises.