Title: Modelling and forecasting Euro area GDP growth using a hierarchical factor model based 3PRF
Authors: Rolf Scheufele - Swiss National Bank (Switzerland) [presenting]
Alain Galli - Swiss National Bank (Switzerland)
Massimiliano Marcellino - Bocconi University (Italy)
Abstract: A multi-level factor model is set up for analyzing and forecasting GDP growth for the aggregate Euro area and its largest members (Germany, France, Italy and Spain) in one aggregate framework. This framework is used to characterize fluctuations in real activity of the aggregate and the individual countries. It allows us to decompose each country's GDP growth into a global (Euro area wide) component, a country specific-component and an idiosyncratic component. To extract the factors of the different levels, we employ a principle component based quasi-maximum likelihood using targeted predictors as well as a three-pass regression filter (3PRF) approach. The results suggest that the three-pass regression filter approach is able to explain a larger country-specific share of GDP growth than the standard quasi-maximum likelihood approach. Moreover, it turns out that our proposed 3PRF based hierarchical factor model performs similarly than an aggregate factor model for the Euro area before the outbreak of the European debt crisis, but notably better thereafter.