How good are dynamic factor models at forecasting output and inflation?

Autori

• Sandra Eickmeier

Parametri

• Economia e affari

Maggiori informazioni sul libro

This paper conducts a meta-analysis of existing factor forecast applications for real economic activity and inflation, contributing to the debate on the determinants of forecast performance of large-scale dynamic factor models compared to other models. The findings indicate that, on average, factor forecasts are slightly superior to those from other models, particularly outperforming small-scale models, though they perform marginally worse than alternative methods that utilize large datasets. The results show that factor forecasts are more effective for US macroeconomic variables than for those in the UK and euro area, with no significant differences in performance for inflation between the US and euro area. Additionally, factor models are better at predicting output over shorter forecast horizons. The forecasting environment is a key factor, while the dataset size positively influences factor forecast performance. Quarterly data appears more suitable for factor forecasts than monthly data, and rolling forecasts are preferred over recursive ones. The choice of factor estimation technique also plays a role. Other factors, such as the type of panel used or the forecasting method, were found to be less significant, and pre-selecting variables for the factor extraction panel did not improve forecast accuracy in the past.

Lingua: Inglese

Pubblicato: 2006