RDP 9706: Is the Phillips Curve A Curve? Some Evidence and Implications for Australia 1. Introduction

Phillips' original article indicated that he thought it likely that the Phillips curve was indeed a curve, not a straight line. However, subsequent estimates of the Phillips curve have generally been conducted in a linear framework. Using such a framework, Robert Gordon provided regular estimates of the Phillips curve in the 1970s in the United States (Gordon 1970, 1975, 1977). The Phillips curve fell into a period of neglect in academic circles during the 1980s, while remaining an important tool for policy-makers.[1] More recently, the Phillips curve has again been the subject of intensive debate (for example, the symposium in the Journal of Economic Perspectives [2]). This debate has focused on the usefulness of the Phillips curve as an analytical tool for monetary policy, given the uncertainties associated with its estimation.

Like the historical estimation of Phillips curves, the recent debate has been generally conducted in a linear framework. However, a separate stream of analysis has sought to restore the ‘curve’ in the short-run Phillips curve, and has investigated the empirical evidence for, and implications of, a non-linear model of the Phillips curve. Laxton, Meredith and Rose (1994); Turner (1995); Clark, Laxton and Rose (1996); and Debelle and Laxton (1997) all investigate the possibility that the Phillips curve is convex in various G7 countries. In contrast, Eisner (1996) finds evidence that the Phillips curve is concave. This has somewhat perverse implications for monetary policy which will be discussed below in Section 5. Akerlof, Dickens and Perry (1996) provide evidence that the long-run Phillips curve may be negatively sloped at low levels of inflation, although Groshen and Schweitzer (1997) present some counterevidence.

This paper builds on the existing literature on non-linear Phillips curves. It highlights the features of a non-linear model of the Phillips curve and examines the policy implications of non-linearities. A simple ‘horse race’ is conducted between parsimonious linear and non-linear models of the Phillips curve using Australian data. It is shown that the non-linear model appears to outperform the linear model, when plausible priors are placed on the two models.

Our purpose in this paper is not to estimate a definitive model (either linear or non-linear) of the Phillips curve. Rather our aim is simply to investigate the possibility that the Phillips curve is non-linear, and use our derived estimates of a non-linear Phillips curve as an expository device to demonstrate the implications of an asymmetrical Phillips curve.

In the next section, the basic linear and non-linear models of the short-run Phillips curve are presented and previous work estimating the Phillips curve in both Australia and the United States is summarised. Section 3 describes the technique we use to estimate the Phillips curve and the data used. The results of our estimation are presented in Section 4, and the non-linear and linear models are compared. In Section 5, we discuss the policy implications of a non-linear Phillips curve. Section 6 concludes.

Footnotes

Leeson (1997) provides a comprehensive summary of the Phillips curve debate. [1]

Journal of Economic Perspectives, 1997, 11(1). [2]