A1239
Title: Dynamic factor models with infinite-dimensional factor space: Forecasting
Authors: Marco Lippi - Universita di Roma La Sapienza (Italy) [presenting]
Mario Forni - Universita degli Studi di Modena e Reggio Emilia (Italy)
Alessandro Giovannelli - University of Tor Vergata - School of Economics (Italy)
Stefano Soccorsi - ECARES (Belgium)
Abstract: The aim is to compare the pseudo real-time forecasting performance of three different previous factor models. The standard US monthly large dataset of macroeconomic and financial time series, which includes the Great Moderation, the Great Recession and the subsequent recovery, is employed. In a ten-years rolling window framework, we find that the first two methods, based on spectral estimation, outperform the third on average. Moreover, applying a test for relative forecasting performance recently introduced, we find that the spectral-estimation methods significantly prevail in the Great Recession. We extend a previous method by introducing forecast combinations and we find that the prediction accuracy for the industrial production and inflation is greatly enhanced for some forecasting horizons.
