Quantitative information on the current state of the economy is crucial to economic policy-making, but the quarterly national accounts data for GDP in the euro area are released with a significant delay. This paper presents alternative models for the real-time forecasting of euro area GDP and assesses their performance. We estimate univariate/multivariate statistical models, bridge models (systems of autoregressive distributed lags equations with indicators) and a small structural model. The models are estimated for aggregate GDP and components both area-wide and for the three main countries. They are estimated and tested for the period 1980-1999. Data from 1999 to 2001 are used to compare the forecasting ability, gauged by rolling-origin one-step-ahead errors.
Published in 2004 in: International Journal of Forecasting, v. 20, 3, pp. 447-460