The usual practice in economic forecasting is to report point predictions without specifying the attached probabilities. Periodic surveys of such forecasts produce group averages, which are taken to indicate the "consensus" of experts. Measures of the dispersion of individual forecasts around these averages are interpreted as indicating "uncertainty."However, consensus is best defined as the degree of agreement among the corresponding point predictions reported by different forecasters, while uncertainty is properly understood as referring to the diffuseness of the distributions of probabilities that individual forecasters attach to the different possible values of an economic variable. The NBER-ASA quarterly economic outlook surveys provide unique information on probabilistic forecast distributions reported by a large number of individuals for changes in GNP and the implicit price deflator in 1969-81. These data permit comparisons of related point and probability forecasts from the same sources.The matched mean point forecasts and mean probability forecasts are found to agree closely. Standard deviations of point forecasts are generally smaller than the mean standard deviations of the predictive probability distributions for the same targets. Thus the former tend to understate uncertainty as measured by the latter. This is so particularly for short horizons. More generally, does the dispersion of the point forecasts reflect their authors' uncertainty (i.e., their relative lack of confidence)? This paper will deal with these and other related questions, drawing on a set of data which is very rare in economics in that it includes related point and probabilistic forecasts from the same sources. ConsensusAverages from economic outlook surveys are frequently called "consensus"forecasts or treated as such. The term has entered the popular discourse without having been defined in a generally accepted way. But it is clear that the degree to which a survey average is representative of the collected individual predictions can vary greatly depending on the nature of the underlying distribution. There may be no meaningful consensus if the distribution of the point forecasts in question is highly diffuse or multimodal; on the other hand, a consensus would be strongly in evidence for any unimodal, symmetrical, and sufficiently tight distribution (cf. Schnader and Stekler, 1979). '-3.-consensus is not in doubt, though the degree of consensus varies (it is clearly much higher for I than for II). In the right-hand panel, the distributions (shown for simplicity in smoothed form only) are strongly skewed or bimodal so that no well-defined consensus may exist.Expectations or forecasts for the same aggregate variables are likely to draw upon common, publicly available information and widely recognized techniques and relationships (models)., Also, people interact and influence each other directly or indirectly, through informal exchanges, opinion polls, media dissemination of public predictions, and market arrangements for t...

No abstract

The usual practice in economic forecasting is to report point predictions without specifying the attached probabilities. Periodic surveys of such forecasts produce group averages, which are taken to indicate the "consensus" of experts. Measures of the dispersion of individual forecasts around these averages are interpreted as indicating "uncertainty."However, consensus is best defined as the degree of agreement among the corresponding point predictions reported by different forecasters, while uncertainty is properly understood as referring to the diffuseness of the distributions of probabilities that individual forecasters attach to the different possible values of an economic variable. The NBER-ASA quarterly economic outlook surveys provide unique information on probabilistic forecast distributions reported by a large number of individuals for changes in GNP and the implicit price deflator in 1969-81. These data permit comparisons of related point and probability forecasts from the same sources.The matched mean point forecasts and mean probability forecasts are found to agree closely. Standard deviations of point forecasts are generally smaller than the mean standard deviations of the predictive probability distributions for the same targets. Thus the former tend to understate uncertainty as measured by the latter. This is so particularly for short horizons. More generally, does the dispersion of the point forecasts reflect their authors' uncertainty (i.e., their relative lack of confidence)? This paper will deal with these and other related questions, drawing on a set of data which is very rare in economics in that it includes related point and probabilistic forecasts from the same sources. ConsensusAverages from economic outlook surveys are frequently called "consensus"forecasts or treated as such. The term has entered the popular discourse without having been defined in a generally accepted way. But it is clear that the degree to which a survey average is representative of the collected individual predictions can vary greatly depending on the nature of the underlying distribution. There may be no meaningful consensus if the distribution of the point forecasts in question is highly diffuse or multimodal; on the other hand, a consensus would be strongly in evidence for any unimodal, symmetrical, and sufficiently tight distribution (cf. Schnader and Stekler, 1979). '-3.-consensus is not in doubt, though the degree of consensus varies (it is clearly much higher for I than for II). In the right-hand panel, the distributions (shown for simplicity in smoothed form only) are strongly skewed or bimodal so that no well-defined consensus may exist.Expectations or forecasts for the same aggregate variables are likely to draw upon common, publicly available information and widely recognized techniques and relationships (models)., Also, people interact and influence each other directly or indirectly, through informal exchanges, opinion polls, media dissemination of public predictions, and market arrangements for t...

The research reported here is part of the NBER's research program in Economic Fluctuations. Any opinions expressed are those of the author and not those of the National Bureau of Economic Research.

The National Bureau of Economic Research, in cooperation with the American Statistical Association, conducted a regular quarterly survey of professional macroeconomic forecasters for 22 years beginning in 1968. The survey produced a mass of information about characteristics and results of the forecasting process. Many studies have already used some of this material, but this is the first comprehensive examination of all of it. This report addresses several subjects and produces findings on each, as follows: (1) The distributions of error statistics across the forecasters: the dispersion among the individual predictions is often large and it typically increases with forecast horizon, as do the mean absolute (or squared) errors. (2) The role of the time-series properties of the target data: the more volatile the time series, the larger as a rule are the errors of the forecasts. (3) The role of revisions in "actual" data: forecast errors tend to be larger the greater the extent of the revisions. (4) Differences by subpcriod: there is little evidence of an overall improvement or deterioration in forecasts between the l970s and the 1980s. (5) Combining the individual forecasts into group mean or 'consensus" forecasts: this generally results in large gains in accuracy. (6) Comparisons with a well-known macroeconometric model: the group forecasts are more accurate for most but not all variables and spanS. (7) Comparisons with state-of-the-art time series models: the group forecasts and at least half of the individual forecasts tend to outperform Bayesian vector autoregressive models in most (but not all) cases. The univariate ARIMA forecasts arc generally the weakest.