One-year-ahead forecasts by OECD and by national institutes of annual growth of GDP of 13 European countries are analysed for accuracy 1971.1995. Average errors were large: 1.9 % in RMSE and 1.4 MAE. Wilcoxon signed-rank tests showed that only four (five) OECD and two (four) institute forecast records were significantly better than an average growth forecast (the last observed growth) as measured by RMSE. Still, both OECD and institute forecast could jointly beat the naive variants, but there was no significant difference in accuracy between OECD and the institutes. Few one-year-ahead forecasts were biased or had autocorrelated errors. Directional forecasts were also jointly informative. For some years, e.g. 1988, the direction was perceived wrongly by many forecasters. Only the OECD forecasts of Italy and Sweden improve significantly over time. OECD´s two-year-ahead country forecasts were significantly correlated with the outcome only for half of the countries, and they were positively biased.
