How Professional Forecasters Compress Macroeconomic Tail Risks

Professional forecasters report full predictive distributions, yet most research on expectation formation focuses on point forecasts. We use density forecasts from the Survey of Professional Forecasters to study how subjective beliefs differ from benchmark predictive distributions constructed in real time from statistical forecasting models. Comparing survey densities with these benchmark distributions , we construct empirical weighting functions that characterize how forecasters allocate probability across future outcomes. The results reveal substantial heterogeneity across macroeconomic variables. Forecasts for real GDP growth and core inflation exhibit pronounced S-shaped weighting functions, indicating that forecasters assign relatively less probability to extreme outcomes and more probability to outcomes near the center of the distribution. In contrast, forecasts for GDP deflator inflation and the unemployment rate remain close to the linear benchmark. The results are robust across alternative benchmark models, sample periods, and forecast horizons. Overall, the evidence suggests that professional forecasters systematically compress tail risks for some macroeconomic variables but not for others, highlighting an important dimension of expectation formation that is not captured by point forecasts alone.