Estimating Models with Rational Expectations

Abstract

IN THIS PAPER we wish to examine the assumptions needed to estimate models in which expectations of future variables are formed rationally. Starting with Robert Lucas [6], there has been much criticism of traditional econometric techniques based on estimating aggregate behavioral relationships. This criticism has been primarily focused on whether the traditional estimates of macroeconomic, behavioral relationships were structural. Lucas showed that many of the coefficients in traditional macroeconometric models would vary with changes in the policy regime. Thus these estimates were of little use for policy simulations. Many economists have called for a revolution in macroeconometric methods in light of Lucas's insight. One group of critics (e.g., Sims [14], Lucas and Sargent [7], Sargent and Sims [13]) proposes abandoning the use of a priori behavioral information in developing regression equations. Another group (e.g., Sargent [12] and Wallis [21]) attempts to estimate behavioral equations, but abandons the traditional Cowles Commission type of identifying restrictions. Instead, identification is achieved by the use of nonlinear cross-equation restrictions implied by rational expectations. In this paper we do not take issue with Lucas's [6] basic insight; we concur with the critics who argue that macroeconometric methods need to be changed to take account of it. However, just as one must be concerned about structural instability when performing policy simulations, one must also be concerned about structural