Forecasting economic activity with higher frequency targeted predictors

In this paper we explore the performance of bridge and factor models in forecasting quarterly aggregates in the very short-term subject to a pre-selection of monthly indicators. Starting from a large information set, we select a subset of targeted predictors using data reduction techniques as in Bai and Ng (2008). We then compare a Diffusion Index forecasting model as in Stock and Watson (2002), with a Bridge model specified with an automated General-To-Specific routine. We apply these techniques to forecasting Italian GDP growth and its main components from the demand side and find that Bridge models outperform naive forecasts and compare favorably against factor models. Results for France, Germany, Spain and the euro area confirm these findings.