Is Broader Better? A Monetary Approach to Forecasting Economic Activity

Abstract

This paper investigates whether the use of broader Divisia monetary aggregates improves money’s performance in forecasting economic activity within a time-varying parameter vector autoregressive (TVP-VAR) framework. We evaluate entire predictive densities from several alternative models of US output growth and inflation, each using eight different Divisia monetary aggregates. Using the broadest, M4 aggregate produces out-of-sample forecasts which consistently outperform those based on narrower measures of money, pooling of forecasts from several models, and a large-scale, 143-variable model. Our results show that TVP-VARs with Divisia M4 forecast economic activity more accurately than constant-parameter models with alternative or no measures of money.