Rounding bias in forecast uncertainty

The European Survey of Professional Forecasters (SPF) is a dataset that is widely used to derive measures of forecast uncertainty. Participants in the SPF provide not only point estimates but also density forecasts for key macroeconomic variables. The mean individual variance, defined as the average of the variances of individual forecasts, shifted up during the Great Recession and has remained elevated since the crisis. The paper seeks to explain this puzzling lack of countercyclicality by applying a smooth transition analysis on data from the European SPF. The analysis indicates that the mean individual variance is a function of the modelling preferences of forecasters and consequently shifts in individual variance are likely to be misleading for the actual changes in the perceived uncertainty. The results remain robust after potential endogeneity has been accounted for. (C) 2020 University of Venice. Published by Elsevier Ltd. All rights reserved.