The accuracy of OECD forecasts of the international economy: Demand, output and prices

Abstract

This paper examines the accuracy of forecasts of the international economy made by the OECD. Our large data set, comprising over 7,000 pairs of forecasts and outcomes, includes one-, two-, and three-step ahead semi-annual forecasts of the main components of demand, output and prices for Canada, France, Germany, Italy, Japan, the U.K. and the U.S.A. over the twenty-year period 1968–1987. Various measures of accuracy are computed; also a comparison is made with competing naive and time-series predictions. The analysis includes a full range of diagnostic checks on forecast performance, including rationality tests for unbiasedness, efficiency and consistency. Although there is considerable variation in the accuracy of these forecasts, they are generally superior to the naive and time-series predictions. Error is predominantly non-systematic. However, our analysis exposes exceptions, particularly forecasts of government consumption, and in some of the forecasts of fixed and inventory investment, the foreign balance and inflation. Accuracy in these cases could be improved by a simple linear correction, or by incorporating information contained in recent, known forecast errors. At least half the OECD forecasts fail one or more of the rationality tests.