par Vuchelen, Jozef;Gutierrez, Maria
Référence Brussels economic review, 46, 2, page (31-64)
Publication Publié, 2003
Article révisé par les pairs
Résumé : We propose an easy direct regression test to evaluate forecasts with respect to the naïve most recent observation alternative. The test is applied to thirteen macroeconomic forecasts published by the OECD. Four time horizons are considered: starting with eighteen months and declining to zero for the December current year forecast. Our results show that the longest horizon forecasts are essentially worthless. Unfortunately, the naïve forecasts do not have any value either. In general, naïve forecasts are inferior to OECD forecasts, but frequently the current year forecasts compare in quality to the one-year-ahead forecast.