Judgment in macroeconomic output growth predictions: Efficiency, accuracy and persistence

Abstract

This study applies observations of individual predictions for the first three releases of the US output growth rate to evaluate how applied judgment affects prediction efficiency and accuracy and if judgment is persistent. While the first two issues have been assessed in other studies, there is little evidence of the formation of judgment in macroeconomic projections. Most forecasters produce unbiased predictions, but by employing the median Bloomberg projection as a baseline, it turns out that judgment generally does not improve accuracy. There seems to be persistence in the judgment applied by forecasters in that the sign of the adjustment in the first release prediction carries over to the projections of the two following revisions. One possible explanation is that forecasters use some kind of anchor-and-adjustment heuristic.