Abstract
We document that professional forecasters adjust inflation forecasts in a lumpy way - forecasts are changed infrequently, and when adjusted, they are revised by a large amount. As the forecasting horizon shrinks, the frequency of revisions, the variance of revisions, and forecast errors decrease. Using a fixed-event forecasting framework, we assess the role of the consensus forecast and private information in shaping forecast revisions, both at the extensive and the intensive margins. A model of Bayesian belief formation with fixed revision costs and strategic concerns (i) delivers lumpy forecasts consistent with the survey evidence, (ii) rationalizes forecast efficiency tests without introducing behavioral biases, and (iii) generates the observed response to increases in inflation volatility.
Local Organizer: Jérémy Boccanfuso