Forecasting and nowcasting economic growth in the euro area using factor models

Many empirical studies have provided evidence that the use of factor models, which use large data sets of economic variables, can contribute to the computation of more accurate forecasts. In this study, we examine the performances of four different factor models in a pseudo real-time forecasting competition for the euro area and five of its largest countries. Our aim is to identify empirically the best factor model approach for the forecasting and nowcasting of the quarterly gross domestic product growth rate. We also propose some modifications of existing factor model specifications, with the aim of improving their forecast performances empirically. We conclude that factor models consistently outperform the benchmark autoregressive model, both before and during the crisis. Moreover, we find that the best forecast accuracy is generally produced by the collapsed dynamic factor model.