VAR modelling of the euro area GDP on the basis of principal component analysis - Produkte Statistical working papers

This study outlines how Principal Component technique can be useful in the short-run economic analysis and forecasting of euro area GDP growth. With reference to a previous work, we examined a restricted vector autoregressive (VAR) model based on selected components from industry and consumer confidence indicators of Business and Consumer surveys (BCS) to forecast the quarterly year-on-year growth of GDP in the euro area. One of the main conclusions was that this restricted VAR model outperforms a single autoregressive model in the short term forecasting, since we excluded all noise variables, which do not help to explain GDP growth . In this paper we focus on VAR and autoregressive estimations based on a more deep approach. Their combination with principal component analysis is considered, which will remove the noise factor from the various shocks among variables (i.e. extracting only the common trend). The derived forecasts of these new models are compared with those of the old VAR. The predictive performance of VAR and autoregressive models has been improved by including the first principal component based in all questions from the four main domains (Industry, Consumption, Construction, Retail Trade).