Can a Simple DSGE Model Outperform Professional Forecasters?

DSGE models have recently become one of the most frequently used tools in policy analysis. Nevertheless, their forecasting proprieties are still unexplored. In this article we address this problem by examining the quality of forecasts from a small size DSGE model, a trivariate VAR model and the Philadelphia Fed Survey of Professional Forecasters. The forecast performance of these methods is analysed for the key U.S. economic variables: the three month Treasury bill yield, the GDP growth rate and the GDP price index inflation. We evaluate the ex post forecast errors on the basis of the data from the period of 1994-2006. We apply the Philadelphia Fed quot;Real-Time Data Set for Macroeconomists,quot; described by Croushore and Stark (2001a), to ensure that the information available to the SPF was exactly the same as the data used to estimate the DSGE and VAR models. Overall, the results are mixed. It appears that when comparing the root mean squared errors for some forecast horizons the DSGE model seems to outperform the SPF in forecasting the GDP growth rate. However, this characteristic turned out to be not statistically significant. In principle most forecasts of the GDP price index inflation and the short term interest rate by the SPF are significantly better than those from the DSGE model. The forecast quality of the VAR model turned out to be the worst one