Resumen de Economic forecasting with multivariate models along the business cycle

Carlos Cuerpo Caballero, Pilar Poncela Blanco

• The absence of an agreed model to forecast the main economic aggregates at different time horizons remains an important challenge for econometric analysis, especially in light of the recent subprime crisis. We conduct an out-of-sample forecasting performance exercise with structural (Dynamic Stochastic General Equilibrium) and non-structural (Dynamic Factor) models. We implement it over different stages of the business cycle to Spanish, euro area as well as US data over 1980Q1-2010Q4. Our results suggest accuracy gains from non-structural models in the short-run. Structural models perform better in the medium-run and their benefits increase at all time horizons when considering disruptive times.