par Barot, Bharat
Référence Brussels economic review, 47, 2, page (249-278)
Publication Publié, 2004
Article révisé par les pairs
Résumé : This study evaluates the performance of the eight most important Swedish domestic forecasters of real GDP-growth, CPI-inflation and unemployment for the sample period 1993-2001. The evaluation is based on the following measures: mean absolute error, the root mean square error, bias and finally directional accuracy. The forecasts are even compared to naive random walk and random walk with drift models. The results indicate that the current forecasts compared to the year ahead forecasts decline over the forecasting horizons as more information becomes available. The results with respect to the directional accuracy indicate that we are equally good/bad in predicting the directional accuracy for all three macro aggregates. According to the comparisons with the naive random walk model six out of seven Swedish CPI-inflation forecasters were outperformed by the naive random walk model. Tests of bias indicate that the Swedish forecasters underestimate GDP-growth and overestimate CPI-inflation and the unemployment rate for the sample period. All the Swedish forecasters have been successful in predicting the downward trend in CPI-inflation and the unemployment rate. The performance of the Swedish domestic forecasters is better using preliminary GDP-growth outcomes than final. The performance for the current year forecasts is better than the year ahead forecasts for all three macro economic variables. Revisions are positively biased.