This paper examines several sets of forecasts representing a variety of sources, models, and techniques. Predictions of percentage changes in GNP are relatively accurate, but partly because of negative correlations between errors in the corresponding forecasts of real growth and inflation. For the last two variables, the multiperiod quarterly forecasts for the 1970s have errors that show unusually rapid cumulations over longer spans. The accuracy and properties of the forecasts depend heavily on the economic characteristics of the periods covered but only weakly and not systematically on the differences among the forecasters.
No takes yet. Share an insight, caveat, or question.
Victor Zarnowitz (1979) studied this question.