The role of money in DSGE models: a forecasting perspective

• Forecasting performance of DSGE models with and without money.
• Money tends to improve the forecasts, either point forecasts or density forecasts.
• The absence of portfolio adjustment costs generally improves the forecasts.

This paper studies the importance of money in a New Keynesian model by considering the forecasting performance of DSGE models both without and with money. While the estimation results are in line with previous studies, favoring the inclusion of money mostly in the form of portfolio adjustment and policy effects, a few interesting results emerge with respect to the accuracy of forecasts. Along with the various filtering methods and forecasting methods used (both recursive and rolling), in many cases money tends to improve the forecasts, either for point forecasts, density forecasts or both. However, the detrending method seems to influence the particular findings. Paradoxically, the absence of a portfolio adjustment costs channel tends to generally improve the accuracy of forecasts.