Is money's role relevant to describing the post-WWII U.S. macroeconomic dynamics? Has this relevance changed over time? These questions are answered using both fixed-coefficient and rolling-window Bayesian estimations of a structural model of the business cycle with money. Our empirical evidence favors a speci cation with dri ng parameters for money-consumption nonseparability and the Federal Reserve's reaction to nominal money growth. The role of money is estimated to have been important during the 1970s and declined afterwards. The omission of money produces severely distorted impulse response functions (relative to the model with money). Money is found to be important in replicating the U.S. output volatility during the Great Inflation. These results are shown to depend on the de nition of the monetary aggregate employed in our analysis.