Revised article published in American Economic Review vol. 92, no. 4 (September 2002): 1013-1028.

A number of recent papers have developed dynamic macroeconomic models that incorporate rational expectations and optimizing foundations. While the theoretical motivation behind these models is sound, the dynamic implications of many of the specifications that assume rational expectations and optimizing behavior can be seriously at odds with the data, for both inflation and real-side variables. In a nutshell, the models imply that inflation or real spending "jump" in response to shocks, in contradiction to a host of empirical evidence that shows that both price and real-side variables exhibit gradual and "hump-shaped" responses to real and monetary shocks. For models that are intended for monetary policy analysis, these dynamic shortcomings should be considered quite serious. When monetary policy has only short-run effects on real variables, the inability to approximately capture the short-run responses of inflation or real variables to policy shocks makes a model unsuitable for policy analysis. This paper identifies a simple feature common to many dynamic specifications for prices and real variables that causes the problem. The paper also discusses several potential solutions to the problem, including alterations to the expectations assumption, to the order of differencing implicit in the model, and to the underlying behavioral assumptions. Revised October 1999.