Recent literature has concluded that money no longer plays a fundament
al role in determining US economic activity, especially when post-1982
data are included in the analysis. We re-examine the issue using annu
al and quarterly data sets ranging back to the Civil War. Cointegratio
n tests show that an equilibrium relationship holds between money and
income in all data samples of 35 years or longer, but frequently fails
to hold for many shorter samples. However, when the normal monetary l
ags are explicitly imposed, a cointegrating relationship is frequently
captured, even in quite short samples. Examining the issue in a stand
ard five-variable VAR model, we find that money is the most important
variable in explaining real output for the full 1874-1993 as well as 1
952-1993 periods, even allowing for interest rate effects. Finally, us
ing monthly data over the troublesome 1983-1994 sample, we show that w
hen appropriate adjustments are made to capture recent structural chan
ges, such as the growing importance of the international economy on US
output, money still Granger-causes economic activity. (C) 1997 Elsevi
er Science Ltd. All rights reserved.