Evaluating the Eurosystem/ECB staff macroeconomic projections: the first 20 years

This work marks a thorough analysis of a confidential real-time dataset consisting of the Eurosystem/ECB staff macroeconomic projections since their existence. By applying techniques widely employed in the literature of forecast evaluation, we examine their statistical properties with a special emphasis on optimality and rationality. Long-term GDP projections are biased (tendency to overpredict), do not fully account for available information, and are outperformed by private sector expectations. Inflation projections are optimal and rational on a full-sample analysis; however, subsample analysis reveals two distinct periods with a persistent and significant bias. Before the financial crisis inflation was persistently underpredicted, while in post-2013, the bias reverses into overprediction.