RDP 1999-01: The Phillips Curve in Australia 1. Introduction

In 1959, A.W. (Bill) Phillips spent a sabbatical year in Australia at the University of Melbourne. During his time there he produced what might have been the second ‘Phillips curve’ in the world. Since that time Phillips curves have appeared for many countries under many guises, have found their way into textbooks and conferences, and have become part of the standard set of tools available to macroeconomists when thinking about the effects of policy actions. At the same time, it has become apparent that there is no real agreement over exactly what the term ‘Phillips curve’ implies. The original concept of the equation related wage inflation to unemployment, but modern versions are as likely to relate price inflation to ‘output gaps’. The diverse uses to which Phillips curves have been put can be seen from the trinity of benefits that Stiglitz (1997) mentions as accruing from the possession of a Phillips curve:

1. It is useful for describing the determinants of inflation.
2. It is useful in providing a framework for policy.
3. It is useful for forecasting inflation.

As might be expected, it is unlikely that a single curve can be used for all three purposes. When forecasting or explaining history the curve might have to be augmented with a large array of variables representing both demand and supply influences and institutional factors relating to price and wage controls. Good examples of such a curve are the ‘triangle’ models of Gordon (1997). However, when it comes to establishing a policy framework, simpler versions of the curve may be desirable. Moreover, since policy has to be set within a systems context, a single equation can only be regarded as one of the building blocks of a system. Tensions created by the need to address such a range of issues have been evident both in the Australian and wider literatures and partly account for the large number of Phillips curves that have been estimated. Because of this, it is useful to have some way of organising the discussion, and the trinity of uses distinguished above is a convenient way of summarising its main conclusions. It is what we do in the present paper, although we focus only upon the first two items in the list.

The next section of this paper sets out a general framework that has been the basis of much work on the Phillips curve in Australia. It involves fairly standard descriptions of mark-up pricing and expectations-augmented wage curves, with adjustments made to reflect the fact that Australia is a small, open economy. The framework has been used within the Reserve Bank to study issues such as inflation targeting (for example, see de Brouwer and O'Regan (1997)). Furthermore, since many of the important studies of the Phillips curve in Australia emerged from the Research Department of the Reserve Bank, and since the people involved in these studies also had some input into the formulation of monetary policy, it is of interest to pay special attention to them.

Section 2 derives Phillips curves that reflect most of the historical research on the subject in Australia. Because this work has exhibited a dual focus, sometimes treating the Phillips curve as determining price inflation and sometimes labour cost movements, we derive two such curves. This duality is a feature of the paper, although, as will become clear, our preference is for the labour cost version. In conducting the research we also make some general comments about the difficulties faced by those endeavouring to estimate a Phillips curve for Australia.

Most of the research discussed in Section 2 is aimed at satisfying the first item in Stiglitz' list in that it seeks to provide a list of variables that are likely to be most influential in producing a change in inflation. While an understanding of such determinants is important, for policy there can be little doubt that it is the NAIRU and its measurement that comes to the fore, and the Phillips curve has frequently been used in this endeavour. Of course, there are well known difficulties in using a Phillips curve for this purpose and these are accentuated in the Australian case because the NAIRU has clearly not been constant over the past three decades. Consequently, in Section 3, we examine the significant recent contribution of Debelle and Vickery (1997) in which the NAIRU is allowed to change through time. Because Debelle and Vickery's model is a simpler specification than most Phillips curves recently estimated, we derive a richer specification which includes additional relevant regressors. Our results imply estimates of the NAIRU in 1997 of around 5½–7 per cent. We also examine whether the NAIRU depends, in a systematic way, on changes in the replacement ratio or in the proportion of long-term unemployed, although we are unable to find such dependence.

In Section 4 of the paper we look at the role of the Phillips curve in providing a framework for the formulation of monetary policy in Australia. We discuss the changing views about the Phillips curve within the Reserve Bank over the past three decades, and examine how these views have informed analysis of the inflationary process in Australia. Although the importance of the Phillips curve framework has fluctuated over the years, it currently has an influence comparable to that in the early 1970s when the (expectations-augmented) Phillips curve was first embraced to explain stagflation. The paper ends with a brief conclusion.