The Relative Forecasting Performance of the Divisia and Simple Sum Monetary Aggregates

Abstract

This paper provides direct evidence on the forecasting performance of the divisia monetary aggregates relative to the traditional simple sum monetary aggregates. It is shown that forecasts of U.S. real GDP from a four variable vector autoregression (VAR) are most accurate when a divisia aggregate is used rather than a simple sum aggregate, particularly at broad levels of aggregation. Further, the two M1 aggregates, relative to the broader aggregates, are superior predictors of the GDP deflator, with a slight edge going to divisia M1 over simple sum M1.