Forecasting growth of U.S. aggregate and household-sector M2 after 2000 using economic uncertainty measures

This paper evaluates the predictive out-of-sample forecasting properties of six different economic uncertainty variables for both growth in aggregate M2 and growth in household-sector M2 in the U.S. using data between 1971m1 and 2014m12. The core contention is that economic uncertainty improves both forecast accuracy as well as direction-of-change forecasts of real money stock growth. We estimate linear ARDL models using the iterated rolling-window forecasting scheme combined with two different indicator selection procedures. Forecast accuracy is evaluated by RMSE and the Diebold-Mariano test. Direction-of-change forecasts are assessed by means of the Kuipers Score and the Pesaran-Timmermann test. The results indicate an increased relevance of certain economic uncertainty measures for forecasting growth in both real aggregate as well as real household-sector M2 since 2000.