Assessing the Lucas critique in monetary policy models.

LUCAS (1976) ARGUES that the parameters of traditional macroeconometric models depend implicitly on agents' expectations of the policy process and are unlikely to remain stable as policymakers change their behavior. Historically, this critique was influential in two respects. First, it helped

reorient macroeconomic research toward models with explicit expectations and "deep" parameters of taste and technology. These models, which many hoped would be invariant to policy shifts, included estimated first-order conditions or Euler equations, calibrated general equilibrium models with explicit optimization, and, most recently, "New Keynesian" models. Second, the Lucas critique helped change the focus of policy evaluation from consideration of alternative paths of the policy instrument to consideration of alternative policy rules, which allowed individual agents to formulate forward-looking dynamic optimization problems.

A heightened interest in policy rules in general has spurred many recent estimates of the monetary policy rule using postwar U.S. data on a short-term interest rate, output, and inflation. From the standpoint of the Lucas critique, the most interesting result from these studies is that they typically reject the stability of historical monetary policy rules. Notably, Clarida, Gali, and Gertler (2000), Estrella and Fuhrer (2000), Taylor (1999), and Judd and Rudebusch (1998) estimate Taylor rules that exhibit discrete shifts in the response of the Federal Reserve to inflation and output during the past few decades. (1) This evidence suggests that the Lucas critique should be particularly relevant for monetary policy analysis. Specifically, given the apparent policy shifts over time, the Lucas critique suggests that lagged autoregressive models will be plagued by parameter instability and will make a poor choice for analyzing monetary policy.

In fact, however, empirical autoregressive macroeconomic models without explicit expectations are still widely used for monetary policy analysis. The most prominent examples are the ubiquitous monetary VARs, in which lagged representations of the economy are treated as invariant structural models (e.g., Leeper and Zha 2001). But researchers have also used other nonexpectational autoregressive macroeconomic models for monetary policy analysis, including most recently Rudebusch and Svensson (1999, 2002), Onatski and Stock (2002), Smets (1999), Dennis (2001), Laubach and Williams (2003), and Fagan, Henry, and Mestre (2001). Even more surprising in view of the Lucas critique is that these VAR and non-VAR macroeconomic models without explicit expectations often appear to be fairly stable empirically. For example, Rudebusch and Svensson (1999), Bernanke and Mihov (1998), Estrella and Fuhrer (2000), Dennis (2001), and Leeper and Zha (2001) all provide evidence of the stability of various autoregressive reduced forms. (2)