Temporal consistency of forecasts and data releases

We provide key insights on expectation formation based on the Bloomberg economic survey: around two thirds of professional forecasters provide GDP forecasts that are temporally consistent, meaning that quarterly forecasts add up to the annual. Temporally consistent forecasts are not more accurate than the inconsistent ones, but inconsistent ones might drive estimates of information frictions in some cases. For the overwhelming majority of consistent forecasts, annual GDP predictions almost immediately reflect the monthly GDP releases. These findings suggest that most forecasters make at least minor forecast updates after each data release. Indeed, the inattention rate is found to be between 3% and 6% at the quarterly frequency.