RDP 1999-01: The Phillips Curve in Australia 4. The Phillips Curve and Monetary Policy

The Phillips curve is clearly a useful empirical device for examining the determinants of inflation in Australia. It also, however, provides an intellectual framework for the analysis of monetary policy. In this section, we discuss the intellectual development of the Phillips curve framework within the Reserve Bank of Australia, and particularly within its Research Department. This is of particular interest because many of the Australian empirical studies of the Phillips curve came from this part of the Reserve Bank. It also seems likely that the ideas formulated in this research would have had an influence, perhaps after some time, on the formulation of monetary policy in Australia.

In earlier sections of this paper we showed how conceptions of the Phillips curve and its determinants in Australia had changed over the past three decades and so it is useful to look at monetary policy developments in the light of the results that we have established. In the 1960s, the policy framework in Australia, as in most countries, almost certainly accepted the unemployment/inflation trade-off implicit in the first generation of Phillips curves. Nevertheless, strong economic growth at the time meant that the perceived trade-off did not need to be exploited. In the last year of that decade, the Reserve Bank began issuing research discussion papers, and among the crop of seven papers produced in that year was one titled ‘An Equation for Average Weekly Earnings’, by K.E. Schott. This paper was part of a project within the Reserve Bank to construct a macroeconometric model of the Australian economy, which was released in its initial form in January 1970, and became known as the RBI model (Norton 1970). The Schott paper refers to Phillips' (1958) original Economica paper, but makes no reference to either the Phelps (1968) or Friedman (1968) papers, which introduced the idea that there was no long-run trade-off between inflation and unemployment. Given this omission, it is perhaps not surprising that the econometric results presented by Schott implied the existence of a trade-off between inflation and unemployment in the long-run, although this implication is not drawn out in the paper.

With the dawning of the 1970s, and the rapid acceleration of inflation in Australia, the question of whether there was a long-run trade-off became one of greater urgency. In mid 1971, the Head of Research Department, Austin Holmes, wrote an internal paper on the problems of inflation which made a clear statement about the distinction between the short-run and long-run trade-offs between inflation and unemployment:

Inflationary expectations can help to explain the differences over time in the apparent trade-off between wage rises (and the rate of inflation) and the rate of unemployment. Empirical studies linking these two have often found that the relationship is improved if the recent rate of increase in prices (a proxy for expectations) is included. If, as seems likely, wage claims are specified in real terms, the interpretation of Phillips' curves becomes somewhat clouded. They can rule only for a given set of expectations about prices. Changes in these expectations lead to equivalent changes in the rate of increase in wages for a given level of unemployment. (Research Department 1971, p. 4)

Much of the analysis contained in this ‘Inflation’ paper remains of interest in the late 1990s. In a later section, the paper argues that inflationary expectations, once raised, might prove hard to reduce. It canvasses the possibility that expectations might not respond to announcements of the anti-inflation resolve of the authorities, and might be difficult to lower without a period of higher unemployment. Given their contemporary relevance, it is worth quoting these arguments at some length.

Suppose an economy has proceeded for some time with activity high and inflation positive but moderate. Expectations of future growth in prices have been formed.

Suppose this situation is disturbed by an upsurge in demand resulting from, say, an export or investment boom. Does this require only an application of traditional fiscal and monetary measures to reduce demand? And, after the faster price rises generated by the excess expenditure have worked their way through the economy (this could, of course, take some time), can the economy return to its previous acceptable pattern of activity and price performance? This seems to depend a good deal on what has happened to expectations about prices.

Where inflationary expectations are unaffected by these hypothetical developments (our knowledge about what actually affects these expectations is far from perfect), the economy can resume its previous pattern. However, if the episode leads to, say, an upward revision in the community's expectations about inflation, the answers seem to change somewhat. It is assumed, of course, that the revised expectations persist in the face of pronouncements of concern by the authorities and even of the announcement and effects of the fiscal and monetary measures.

In this situation, there is a rise in the economy's aggregate supply schedule. With demand given, this would lead to a lower level of activity but a faster rate of increase in prices than previously. Attempts to restore activity lead to even faster growth in prices. In other words, following a change in inflationary expectations, the fiscal and monetary policies consistent with a given level of employment have to be more expansive than before the change.

It does not require a burst of demand to induce the change in expectations. This could result from a sudden awareness that, say, previous expectations about stable prices ought to be abandoned in the face of x years of positive price increase. Perhaps the inflationary expectations can be imported. Whatever the cause, a return to price stability with high employment requires a change in price expectations. If these are firmly entrenched, as they might be if a fairly long history of price increases figures in their formation, one cannot be too optimistic about the ability of fiscal and monetary measures to do this without a period of higher unemployment. (Research Department 1971, p. 6)

It was not too long before empirical support for the arguments set out in the ‘Inflation’ paper was provided by econometric evidence from the Australian economy. In a Reserve Bank Research Discussion Paper issued in August 1973, Jonson, Mahar and Thompson estimated several equations for the annual growth in average weekly earnings. As well as variables capturing demand effects, the equations included growth in award wages and in world prices as explanators. In their preferred equation, the sum of the coefficients on the ‘price’ terms was insignificantly different from one, from which the implication was drawn that there was no long-run trade-off between the rate of inflation and the state of the labour market in Australia. This preferred equation for average weekly earnings, as well as an equation for award wages, were soon incorporated into the Bank's RBI macroeconometric model with only minor amendments.

Professor Michael Parkin, a visitor to the Bank at the time, seems to have played an influential role within the Research Department by providing a unifying interpretation of the available econometric evidence. Jonson, Mahar and Thompson (1973) credit Parkin with pointing out that their empirical results implied the absence of a long-run inflation-unemployment trade-off. Furthermore, the Jonson, Mahar and Thompson paper was a revised version of a paper by Jonson and Mahar, issued in November 1972, which contained much the same econometric exercise as the later paper (using a slightly shorter sample), but did not draw any implications from the results about the long-run inflation-unemployment trade-off.

Parkin also produced a research paper on ‘The Short-run and Long-run Trade-offs between Inflation and Unemployment in Australia’ in the second half of 1973, in which he presented a critical analysis of the long-run inflation-unemployment trade-offs implied by several recent econometric studies of wage and price inflation in Australia. On both theoretical and empirical grounds, he argued that those studies which implied a non-zero long-run trade-off were flawed. His paper led to a series of responses and rebuttals in the pages of Australian Economic Papers that continued for several years.^[23]

Of course, the intellectual framework for analysing the inflationary process was not the only thing that was changing around this time. The economic landscape was also changing. As well as the rapid deterioration in the inflation performance (Figure 14), it is now clear that the NAIRU in Australia also rose significantly in the early 1970s.^[24]

For the Phillips-curve framework to be useful as a guide for monetary policy, it was of course necessary to have some reasonable idea of the level of the NAIRU – in order to be able to assess the inflationary implications of any given rate of unemployment.^[25] While we would now date the beginning of the significant rise in the NAIRU somewhere around 1970–1972 (based on both the price and unit labour cost Phillips curves in Figures 8 and 9), this rise was far from clear at the time. For example, in the paper discussed above, Parkin (1973) argued that the natural unemployment rate had probably fallen since the early 1960s, to be in the 1½ to 2 per cent range at the time of writing in late 1973.

By early 1976, however, the then Head of the Research Department, W.E. Norton, argued in a published paper that recent experience in several countries (including Australia) suggested that the NAIRU (which he called ‘the lowest sustainable rate of unemployment’) seemed to have increased, although he offered no quantitative estimate of the extent of the increase (Norton 1976).

The difficulties inherent in coming to a view about the level of the NAIRU in the mid 1970s probably also made some contribution towards another important change in the Reserve Bank's thinking about the inflationary process. As with the introduction of the idea that the long-run Phillips curve was vertical, this change also owed a large debt to intellectual developments overseas.

In the 1971 ‘Inflation’ paper, the causes of inflation were discussed under the subheadings: labour costs, material costs (including, importantly, the price of imported goods), taxes and profits. The paper argued that inflation was caused by both excess demand and adverse supply shocks. Excess demand led to higher inflation primarily because of rising labour costs (although firms' mark-ups on costs might also rise) as the economy travelled up a short-run Phillips curve. The result was higher inflation – rather than a once-off rise in the price level – because inflationary expectations were disturbed. In light of later developments, it is of interest to note that the paper did not discuss excess money growth as one of the causes of inflation.

Quite soon after the ‘Inflation’ paper was penned, however, monetary growth began to play a more prominent role in explanations of the inflationary process within the Research Department. Jonson, in an internal paper written in October 1973, put the argument in these terms:

…our positive knowledge of the workings of the economic system establishes a general presumption that if we desire to control inflation we should carry out any stabilization policy within the constraint of an average rate of growth of the money stock determined by the growth of ‘full employment’ demand at current inflation rates. (Jonson 1973, p. 4)

In common with developments overseas, money came to play a central role in the second half of the 1970s, both in macroeconometric models developed within the Bank, as well as in the formulation of Australian monetary policy itself. A second generation of macroeconometric models (called RBII) developed within the Research Department appeared in a series of versions during the second half of the 1970s, and well into the 1980s (the last Research Discussion Paper to use RBII was written in 1987). Money had a key role in this generation of models, with several transmission channels through which money growth had a direct and immediate influence on both real and nominal magnitudes within the economy, including, in particular, both wage and inflation outcomes.

In the formulation of Australian monetary policy, money growth became an intermediate target in 1976, when the Government began announcing an annual projection for growth in the broad monetary aggregate, M3. This practice was maintained, with only slight variations, until early 1985 when, faced with evidence of a breakdown in the empirical relationship between growth in M3, nominal income and interest rates, the Government abandoned the M3 projection.^[26]

In principle, analysis of the inflationary process based on a Phillips curve framework, with allowance made for open-economy aspects and supply shocks, could exist along side an analysis based on money growth. The two intellectual frameworks need not be seen as incompatible. To the extent, however, that excess monetary growth was seen at the time as the fundamental cause of inflation, it was natural that a framework that highlighted the importance of controlling money growth would gain pre-eminence over one that focused on the demand/supply balance in the markets for labour and goods.

With the end of money-growth targeting in Australia, there was a transition period for monetary policy, in which policies became more pragmatic and there was a search for alternative guiding principles. For a time, there was a policy ‘checklist’, which was a range of variables, which were to be consulted in assessing economic conditions and making policy decisions. An early description of the checklist approach by Governor Johnston makes clear the very wide range of variables that were considered relevant. They included:

…all the monetary aggregates; interest rates; the exchange rate; the external accounts; the current performance and outlook for the economy, including movements in asset prices, inflation, the outlook for inflation and market expectations for inflation. (Johnston 1985, p. 812)

The checklist was essentially an amalgam of instruments, intermediate and final policy objectives, and general macroeconomic indicators. Although the demand/supply balance in the labour and goods markets would undoubtedly have been considered relevant elements of the checklist, the Phillips curve certainly did not play a central role in this view of the inflationary process.

In the late 1980s and into the 1990s, the framework for monetary policy evolved gradually. A medium-term inflation target formed the centrepiece of the framework from around 1993, although many of its essential elements were in place several years earlier.^[27] A monetary policy framework with a medium-term focus on inflation as the policy objective, no intermediate objective, the short-term interest rate as the instrument, and a transmission process that works via the effect of interest rates on private demand, had been analysed in several pieces published by the Bank in 1989, including its conference volume.^[28]

In many ways, the intellectual framework for analysing the inflationary process within the Reserve Bank has come full circle. The framework of the 1990s has much in common with the one enunciated in the ‘Inflation’ paper written in 1971, although the modern version would perhaps contain a few elements not present in the earlier version. These are the main elements of the modern version.

In the short-run, output above potential (or, equivalently, unemployment below the NAIRU) generates rising wage growth and, perhaps, increases in firms' mark-ups, which in turn, feed into inflation. Speed-limit effects are also relevant, so that strong output growth (or rapidly declining unemployment) also generates inflationary pressures.^[29]

Inflationary expectations are central to the inflationary process; the Phillips curve is indeed of the expectations-augmented variety, so that there is no trade-off between inflation and unemployment in the long-run. Inflationary expectations seem relatively immune to announcements of the authorities' anti-inflation resolve. To achieve a sustained reduction in inflation and inflationary expectations, it appears to be necessary to run the economy for a period with output below potential and unemployment above the NAIRU. Furthermore, inflationary expectations take a long time to fully adjust to a fall in the trend rate of inflation – especially after an extended period of high inflation. Thus, for example, the transition to low inflation in Australia was complete by 1992. Nevertheless, while inflationary expectations both in the bond market and among consumers fell to some extent at that time, the fall in inflationary expectations in the bond market did not fully reflect the lower trend rate of inflation until 1997, and even then, consumers' inflationary expectations appeared not to have fully adjusted to the lower trend rate of inflation (Figure 14).

Open-economy aspects are also important to the inflationary process. A fall in the exchange rate, triggered, for example, by a fall in the world price of Australian commodity exports, leads to a rise in import prices, which feeds into consumer prices with a lag. Whether a once-off fall in the real exchange rate translates into a rise in the rate of consumer inflation (rather than simply a once-off rise in the level of consumer prices) depends on whether inflationary expectations are disturbed. This is an empirical issue; there should not, however, be an automatic presumption that inflationary expectations are immune to such an exchange-rate-induced rise in consumer prices.

The rate of money (or credit) growth is an important indicator of the pace of financial intermediation in the economy. Money growth does not, however, have an independent effect on either activity or inflationary expectations in the economy, once the effects of the level of real interest rates and asset prices, including the exchange rate, have been allowed for.

Footnotes

See, for example, Challen and Hagger (1975), McDonald (1975), Nevile (1975), Parkin (1976) and Rao (1977). Hagger (1978) later reviewed the debate at length. [23]

The rapid deterioration in inflation is consistent with our earlier empirical finding that the unemployment rate was below the NAIRU for much of the decade 1966–75 (see the proportional NAIRU gap in Figure 12). The significant rise in the NAIRU is clear from both the price and unit labour cost Phillips curves (Figures 8 and 9). [24]

Wieland (1998) provides a modern discussion of the optimal interplay between policy gradualism and experimentation when there is uncertainty about the level of the NAIRU. [25]

See Argy, Brennan and Stevens (1990) for a comparison of the monetary targeting experience of Australia and other countries. [26]

The numerical objective of 2–3 per cent underlying inflation began appearing in public statements by Governor Fraser in 1992 and 1993. International organisations (for example, the Bank for International Settlements) date the adoption of the Australian inflation target from 1993. Grenville (1997) discusses the history of the inflation target in more detail. [27]

See Macfarlane and Stevens (1989), Macfarlane (1989) and Grenville (1989). [28]

Recall that speed-limit effects are present in both the price and unit labour cost Phillips curves presented earlier. The Bank's 1995 Annual Report also drew attention to them: ‘…unemployment remains well above the point at which serious inflationary pressures are likely to be experienced. The relatively rapid speed of its fall over the past two years, however, has been such as to prompt a pick-up in labour costs’ (p. 15). [29]