Forecasting under Model Uncertainty

This paper investigates the accuracy of point and density forecasts of four dynamic stochastic general equilibrium (DSGE) models for output growth, inflation and the interest rate. The model parameters are estimated and forecasts are derived successively from historical U.S. data vintages synchronized with the Fed’s Greenbook projections. In addition, I compute weighted forecasts using simple combination schemes as well as likelihood based methods. While forecasts from structuralmodels fail to forecast large recessions and booms, they are quite accurate during normal times. Model forecasts compare particularly well to nonstructural forecasts and to Greenbook projections for horizons of three quarters ahead and higher. Weighted forecasts are more precise than forecasts from single models. A simple average of forecasts yields an accuracy comparable to the one obtained with state of the art time series methods that can incorporate large datasets. Comparing density forecasts of DSGE models with the actual distribution of observations shows that the models overestimate uncertainty around point forecasts.