Revised article published in American Economic Review 90, no. 3 (June 2000): 367-90.In earlier work (Fuhrer 1996), I document what I view as the failure of standard models of representative consumer and firm behavior to replicate the dynamics that we observe in the aggregate data. In essence, these models fail because they imply that both inflation and real variables must “jump” in response to monetary policy (and other) shocks, in contrast to identified VAR evidence that shows a gradual, “hump-shaped” response. This paper discusses a rigorous empirical standard for monetary policy models. The motivation for this discussion is that, if one wishes to conduct welfare analysis, one must be reasonably confident that the model provides a good approximation to underlying consumer and firm behavior over the monetary policy horizon, i.e., in the short-run. The paper examines a specific alternative to the standard consumption model in which consumers' utility depends in part on current consumption relative to past consumption. This formulation of habit formation allows one to nest habit formation, life-cycle consumption, and Campbell and Mankiw's “rule-of-thumb” consumers within a more general model. The empirical tests developed in the paper show that one can reject the hypothesis of no habit formation with tremendous confidence. This result suggests that models that are unable to produce a hump-shaped response will be strongly rejected empirically.