Uncertainty is a key concept in both economic theory and economic practice. Yet, economic forecasts are usually stated as single numbers, or "point estimates," that convey no information about the full array of possible outcomes. The dispersion of individual forecasters' point estimates is often used as an approximation of forecast uncertainty, even though it is neither logically nor empirically related. In fact, the diversity of point estimates is a poor guide to the accuracy of a point estimate forecast.

This article examines explicit estimates of forecast uncertainty, taken from the Survey of Professional Forecasters. It concludes that most individuals' estimates of inflation and real GNP uncertainty are well calibrated at both the 50 and 90 percent, though not at the 100 percent, confidence intervals. In contrast, the mean probability distribution of all respondents is well calibrated at all three intervals. Despite their overall reliability, the uncertainty estimates are not correlated with the accuracy of point estimate forecasts. This lack of correlation should not be construed as evidence that uncertainty cannot be reliably anticipated, however.