Economic forecasts and fluctuations: statistical and economic analyses.

Summary The first objective of this paper is to obtain, from an original database, robust results on the average quality and properties of economic forecasts. We analyze the statistical means available to judge this quality and show the characteristics of past errors. The "vertical" analysis of forecasts allows a comparison of variables within each organization and the "horizontal" analysis of forecasts allows a comparison of organizations among themselves. We review the literature on forecast combinations. Finally, we analyze the predictions by keeping the temporal dimension of the error chronicles. We show that forecasters face a major problem: their inability to predict turning points. However, being wrong at these particular dates means that the cyclical dynamics of the economy that we are trying to forecast are poorly understood. The second objective of this paper is therefore to show that the recent tools for identifying the trend and cyclical components of growth are part of the techniques necessary for the elaboration of a business cycle diagnosis and the construction of a forecast scenario. We describe all the statistical and economic tools available and their implications through their critical analysis. We show that the various methods described above, although of very different inspiration, lead to relatively comparable profiles of GDP deviation in the past.