The Unassuming Economist

From a paper by Pedro M. Esperança, and Paulo Júlio:

“The literature has focused primarily on the quality of forecasts for real Gross Domestic Product (GDP) and for other macroeconomic variables (e.g. inflation or unemployment), but has not satisfactorily addressed the forecast accuracy of the major expenditure components of GDP – private consumption (C), government consumption (G), investment (I), exports (X) and imports (M)4. In a recent article, Júlio et al. (2011) have analyzed, for the first time, the quality of forecasts for GDP expenditure components. The authors showed that overpredictions in investment and exports explain most of Portuguese GDP overpredictions at 1-year horizons. GDP forecast bias diminishes significantly for same-year predictions, a fact that is mostly explained by canceling out effects in component prediction errors rather than by accurate component predictions. The authors have also proposed two new statistics – Mean of Total Weighted Absolute Error (MTWAE) and Mean of Total Weighted Squared Error (MTWSE) – to objectively evaluate the overall accuracy of component predictions.

This article uses similar techniques to analyze the forecast quality of GDP expenditure components for G7 countries. Three dimensions of forecast quality are addressed here: bias, accuracy, and efficiency. We use forecast data issued by the Organization for Economic Cooperation and Development (OECD) and by the International Monetary Fund (IMF) for the 1993-2010 period, and evaluate both 1-year ahead and same-year predictions. Our focus lies on the overall quality of institutions’ forecasts, and thus we pool evaluation statistics across countries in order to obtain an aggregate overview of the main features driving these forecasts. In addition, we propose panel versions of two types of efficiency tests presented in the literature, and analyze the effects of the 2008 crisis on the quality of forecasts.”