No 129. Estimating New-Keynesian Phillips Curves: A Full Information Maximum Likelihood Approach

by Jesper Lindé

December 2001, Revised March 2005

Abstract: The New-Keynesian Phillips curve has recently become an important ingredient in monetary policy models. However, the empirical support for the New-Keynesian Phillips curve is very mixed using limited information methods. In this paper, I argue by means of Monte Carlo simulations that limited information methods, e.g. GMM, are likely to produce imprecise and biased estimates using a simple New-Keynesian sticky price model. I then show that estimating the model with Full Information Maximum Likelihood (FIML) is a useful way of obtaining good estimates when the limited information methods fails. Finally, I estimate a version of the model on U.S. data with FIML and find that the pure forward-looking New-Keynesian Phillips curve is strongly rejected, whereas a version with both forward- and backward-looking components provides a reasonable approximation of U.S. inflation dynamics.

Keywords: Monetary policy rule; New-Keynesian Phillips curve; Rational expectations IS-curve; Backward-looking Phillips curve; Measurement errors; Full Information Maximum Likelihood estimation.

JEL Classification: E52, C52, C22.