On the limits of rational expectations for policy analysis

In this lecture, I review work addressing three questions. First, are predictions about macro stabilization policies robust to reasonable departures from rational expectations? Second, do people's expectations converge to a particular set of rational expectations? Third, if they do converge, how quickly? I discuss examples from the literature where the answer to the first question is no. The answer to the second question is that learning equilibria converge to the “standard” rational equilibria analyzed in new Keynesian models. Finally, I discuss circumstances under which the answer to the third question is very slowly. In the examples, learning is slowest and policy analysis based on rational expectations is least robust in the face of shocks that render the stakes of getting policy “right” the highest.