Substantial resources are devoted to forecasting macroeconomic series by the International Monetary Fund (IMF) and the Organization for Economic Cooperation and Development (OECD). Forecasts for these series are also available from many private sources. For example, Consensus Economics, an English firm, combines forecasts from 10 to 30 private sector forecasts for each country. Batchelor asks whether the IMF and OECD, with their statistics and highly qualified analysts, might contribute to forecast accuracy beyond what is offered by the private firms. Batchelor used a two-tailed significance test on whether these forecasts do better than the consensus. He examined forecasts for the G7 countries for seven years (1990-1996) using three starting points for each year for six variables. In all, there were comparisons for 1,200 forecasts. Using the Mean Absolute Errors, the consensus forecasts were superior to the OECD and IMF forecasts on 23 of 27 comparisons (summarized by variable by starting month). On average, the combined forecast error was 9.6% less than the IMF forecast and 8.1% less than the OECD forecast. These gains are in line with the estimate that combining will reduce the error by about 12% versus a single component forecast (Armstrong, 2001). Batchelor then examined whether the Consensus Forecasts would be improved through the addition of the OECD and IMF forecasts. It was not, which is not surprising given the large number of forecasts included in the combined forecasts; the results for combining suggest that the gains in accuracy from adding forecasts become small after one has combined five forecasts (Armstrong, 2001).
Batchelor has updated the study to cover the ten years for 1990 through 1999. As might be expected, the results are a bit `more consistent (e.g., the consensus is now better on 25 of 27 comparisons), The results are available in "The IMF and OECD versus the Consensus" at www.staff.city.ac.uk/r.a.batchelor
Armstrong, J. S. (2001), "Combining forecasts" in Armstrong, J. S. (Ed.), Principles of Forecasting. Kluwer Academic Publishers, 2001.