Unemployment and Inflation: A Demand side focus

William Mitchell

Centre of Full Employment and Equity, University of Newcastle, Australia

Home Page

1. Introduction

Historical Note: This seminar was part of the PKT Discussion List Seminars and was presented from January 23, 1997. It was also presented to a European Union Workshop held at the European University Institute in Florence in November 1996.

Unemployment rates in almost all OECD economies have risen and persisted at higher levels since the first OPEC shocks in the 1970s. Several writers have argued that the persistently high unemployment is sourced in institutional arrangements in the labour market (wage setting mechanisms and trade unions) and government welfare policies (encouraging people to engage in inefficient search)(for example, Layard, Nickell and Jackman, 1991, hereafter LNJ).

In this paper, it is argued that the principle reason that OECD countries have experienced more than two decades of high unemployment lies in an unwillingness of policy makers to use fiscal and monetary policy in an appropriate manner. The rapid inflation of the mid-1970s left an indelible impression on policy makers which became captive of the resurging new labour economics and its macroeconomic counterpart, monetarism. The goal of low inflation replaced other policy targets, including low unemployment. The result has been that GDP growth in OECD countries has generally being below that required to absorb the labour force growth and the growth in labour productivity.

Consistent with this new ruling economic paradigm, the available policy instruments have narrowed with Governments reluctant to use taxation increases as a source of revenue to finance spending initiatives. This has led to an excessive reliance on reduced spending and tight monetary policy which has resulted in stifled economic activity. The experience of the OECD economies has been replicated, but with harsher consequences, in the nations that have been subjected to the disastrous IMF and World Bank structural adjustment programmes.

The paper is organised as follows. Section 2 presents the argument. Section 3 provides a statistical overview of the OECD economies to add empirical weight to the argument. Section 4 examines the Phillips curve relationships for the OECD economies. Section 5 models and calculates persistence in OECD unemployment rates and shows that the reaction to an exogenous shock is usually much longer than the average public policy cycle. In other words, activist policy can attenuate the effects of negative shocks. Section 5 addresses the issue of the inflation constraint by modelling the impact of incomes policy in Australia on the rate of wage inflation. It is shown that the impact was significantly negative and this provided the room for more expansion and lower unemployment rates that would have otherwise been the case. The use of incomes policy must be a serious consideration of governments facing high unemployment due to deficient demand. Concluding remarks follow which to some extent indicate that the conventional framework for analysing unemployment and the role of government is inadequate and has to be changed.

2. The Macroeconomics of Unemployment

Orthodox economists assert that the persistently high unemployment rates in OECD economies are a reflection that the NAIRU has risen. Several reasons, among others, have been suggested in the literature for the rise in the NAIRU (see LNJ):

(a) excessive unemployment and other social transfers distort the choice between labour and leisure,

1. a rise in the female labour force participation rate leading to higher representation in the labour force of groups who have a higher unemployment rate,
2. excessive minimum wage rates promoted by trade unions and government.

LNJ impute that rising unemployment benefits and trade union coverage were causes of the rise in unemployment. There has been very little evidence presented to substantiate these effects in any economy in the world. They are largely predictions which emerge from the orthodox competitive model which lack empirical substance.

The central idea in this paper is Keynesian because it suggests that the level of unemployment is a function of the level of aggregate demand. The major reason why policy makers have allowed the level of unemployment to rise over the last 20 years is that they have been increasingly loathe to use discretionary fiscal and monetary to stimulate the economy. The battle against unemployment has been largely abandoned in order to keep inflation at low levels. No calculus has been done to measure the relative costs of this strategy. An old Keynesian adage that it takes many Harberger triangles to fit into one Okun's Gap has been ignored.

The fight against inflation was linked to the resurgence of neoclassical macroeconomics and the new labour economics in the early 1970s. Osberg (1984, pp. 111) said:

... by 1982….inflation was down to single digits in North America, Britain, and many nations of Western Europe….the casualties….over 32 million unemployed in OECD nations, have been far higher than were anticipated and there is little current anticipation for their early recovery.

Why were the costs of the war against inflation so grossly underestimated? …

a concern with distributional issues, not with economic efficiency, that started the war against inflation…the distributional implications, and the aggregate costs,…, were substantially mis-estimated and this….was due to the reinterpretation of labour-market data, and especially of unemployment, which developed as part of the 'new labour economics' of the 1960s and 1970s.

In the next section we will see that the link between movements in the unemployment rate and capital expenditure is well defined. The restrictive policy stance adopted by OECD governments has led to a vicious circle with the parlous unemployment situation being inevitable. First, the restrictive policy held real interest rates at high levels for extended periods. The high real interest rates have resulted in lower than otherwise private capital expenditure which also reduce the potential medium term growth path of the economy and worsens the relationship between output capacity growth and labour force growth.

Capital expenditure has also been retarded directly by the public spending cuts (including privatised and to-be-privatised utilities). In the short run, it is easier for governments to cut capital expenditure instead of recurrent, "vote-attracting" programmes and services. As these cuts impact on the unemployment rate, the resulting high cyclical budget deficits have led governments to further cut public capital spending. We went from inflation mania to the balanced budget mania over the decade from 1975. For European countries trying to make the adjustments required for Maastricht the balanced budget mania remains a continued source of low activity.

For a small trading nation like Australia, the high real interest rates also pushed the exchange rate up which squeezed the export sector and allowed an advantage to imports. This has meant that Australia has pushed up against the current account constraint which has further inhibited the ability of the country to reduce unemployment.

The pursuit of balanced budgets via severe public sector contraction has also narrowed the range of available policy instruments that governments use. It is now very difficult to raise income or other taxes to provide flexibility to the budget position. The Australian government has relied excessively on monetary (interest rate) policy to pursue its inflation objectives despite the uncertainty over timing and impact that accompanies such a blunt instrument. The 1990 recession in Australia was policy-induced and the extreme severity a direct consequence of the lack of feedback that high interest rate strategies give the policy maker.

Orthodox economists have overlooked this reality and instead concentrated on the labour supply side, hypothesising that full employment now occurs at much higher unemployment rates than in the past.

Piore (1979, p.10), an antagonist to the orthodox position, argues that:

Presumably, there is an irreducible residual level of unemployment composed of people who don't want to work, who are moving between jobs, or who are unqualified. If there is in fact some such residual level of unemployment, it is not one we have encountered in the United States. Never in the post war period has the government been unsuccessful when it has made a sustained effort to reduce unemployment. (emphasis in original)

Piore's view is consistent with that espoused by Kalecki (1971) who emphasised the political aspects that arise in economies which maintain full employment.

The assumption that a government will maintain full employment in a capitalist economy if it only knows how to do it is fallacious. In this connection, the misgivings of big business about the maintenance of full employment by government spending are of paramount importance. (p.138)

The curious anomaly identified by Kalecki was that this "opposition" to full employment would seem to defy the reality that it benefits all - with both higher wages and profits being available. Kalecki listed three reasons including the dislike by business of government economic intervention, the dislike by business of the direction of government spending, and, significantly, the dislike by business of the social and political changes resulting from the maintenance of full employment.

Underlying the orthodox Phillips curve is an implicit view of the interaction between the employer and employee which is never fully articulated in political terms. Kalecki makes the view explicit and says

Under a regime of permanent full employment 'the sack' would cease to play its role as a disciplinary measure. The social position of the boss would be undermined and the self-assurance and class consciousness of the working class would grow.

In political terms it is clear that the ideas expressed by Piore and Kalecki might help us understand the rise in unemployment rates over the last 20 years. Clearly there have been moves to right-wing, rationalist, governments in most economies over this time and there has been a reluctance to use aggregate demand policies in any significant or effective manner. We have been lectured by the politicians backed up by a resurgent orthodox economics profession about twin-deficit hypotheses, outsourcing, privatisation, and the drain on incentives of welfare provision. There has also been an increase in employer power vis-a-vis the workers as unions have declined.

From an economic perspective, if there have been constrained product and labour markets over this time, then what is going on with the labour supply curve (however we might care to define this) is largely irrelevant. We simply cannot say what optimal re-arrangements of time labour suppliers are making and how it is impacting on the labour market outcomes if in fact the economy is permanently "off" the labour supply frontier. In the spirit of Keynes and then Clower, what people want to do is irrelevant if they haven't an opportunity to implement their plans.

Summers (1988, p. 19) says

In Britain today, about 60 per cent of all unemployment is due to persons in the midst of spells lasting two years or more. Can anyone seriously maintain that this outcome is the result of intertemporal substitution, misperceptions, or efficient search? For that matter, what factor or factors could have double the NAIRU over the last seven years?

3. Overview of the Problem

Table 1 provides a usual summary of movements in unemployment rates, inflation rates and the misery index for selected OECD countries. If we assume that the decade preceding 1973-74 was a "golden age" which can be used as a benchmark for comparing later performance, then only Japan has resisted the deterioration in misery. Most countries have adjusted to the high inflation which accompanied the first oil shock by allowing rising unemployment rates to drive the inflation rates down. The monetarist policy experience has been very successful at one thing - driving down inflation - but at the high cost of persistent excess labour.

Table 2 shows the shift in inflation misery since the 1960s. Overall misery in the 1960s was largely due to inflation, although many would argue that this was hardly misery. In 1995, most countries have achieved low inflation, but at the cost of high unemployment and its attendant output and social losses. The move from medium to high inflation-low unemployment regimes to the opposite, with stagflation in between, is stark and largely policy driven.

Table 3 shows the proportion of long term unemployment as a percentage of total unemployment. Given the persistently high unemployment rates across the OECD since the 1970s, it is little wonder that the proportion of people experiencing long duration unemployment should rise. This is a stock-flow problem. The initial rapid rise in unemployment which was followed by two decades of restricted activity has meant that many who entered unemployment in times of protracted low demand have become locked in as the next labour force entrants arrive. Eventually we see the ratio of long term unemployed rise. It is interesting to note that New Zealand has been a laboratory for extreme supply-side policies since 1985 and the long-term unemployed to total unemployed ratio has risen continuously over that time.

LNJ (1991) hint that the USA economy has been spared the worst of the rise in unemployment because they have experienced a fall in welfare benefits and trade union membership. However, if we focus on the most stable measure of activity, the employment/population ratio, several countries have experienced rising ratios. Table 4 shows the employment/population ratios and labour force participation rates for selected OECD countries from 1973 to 1995 (at selected intervals). The OECD Employment Outlook (July 1996, Chart 2.2 p. 29) shows movements in replacement ratios in OECD countries from1961 to 1995. We see no systematic relationship at all between the direction of the benefits and rise or fall in the employment/population ratio. For example, Australia has seen a rise in both, whereas Belgium has seen a fall in both.

It is also possible that the causation is exactly the opposite to that proposed by the new labour economics. With low activity moderating money wages growth, the replacement ratio will rise if in the face of persistently high unemployment rates, the governments bow to the pressure for increased benefit rates.

Moreover, LNJ (1990, p.5) argue that:

... there are very powerful mechanisms at work which have forced the number of jobs to respond to huge changes that have occurred in the numbers of people wanting work.

In general, there have been substantial rises in participation rates in most OECD countries since the early 1970s, driven by the rise in women's participation rates. LNJ must argue that the rise in participation rates have been driven by people seeking to establish a labour force status to get benefits. But once again there is nothing systematic evident which would explain the rise in unemployment. Countries with significant increases in participation since 1973 include Australia, Canada, Japan, Iceland, Norway and the United States.

Table 1 Unemployment Rates, Inflation Rates and Misery Indexes for OECD Economies, Selected Years

Table 2 The Shift in Inflation as a Percentage of Total Misery in OECD Countries 1963-1995

Table 3 Long Term Unemployment as a Percentage of Total Unemployment

Table 4 Employment/population ratios and Labour Force Participation Rates - All persons

Table 5 OECD Unemployment/Vacancy Ratios

Other countries with very moderate to insignificant rises in participation include Austria, Netherlands, New Zealand (all with rising employment/population ratio and significant rise in average unemployment rates), Belgium, Finland, France, Ireland, Sweden, and United Kingdom, (all with falling employment/population ratio and significant rise in average unemployment rates). Germany, Italy, Luxembourg, Portugal and Spain have had falling labour force participation rates and with the exception of Portugal, sharp falls in their employment/population ratios and large rises in the average unemployment rates. The data does not suggest any systematic supply side erosion.

Further, while not the topic of this paper, the countries which have experienced strong growth in employment/population ratios, have also seen substantial proportions of the new jobs growth going to part-time or casual work. So in some cases there have been large increases in available jobs but the qualitative dimension has to be questioned. For instance, in Australia, the employment/population ratio has risen by a small amount while the participation rate has risen by around 5 per cent points. Over the same period (1973-1995), part-time employment as a percentage of total employment has risen from 11.1 per cent to 24.8 per cent. The pattern is common across the OECD bloc (see OECD Employment Outlook, July 1996).

LNJ (p.4) also argue that the level of unemployment has risen sharply relative to the level of vacancies and suggest this is due to a failure of the unemployed to seek work as effectively as before. Table 5 shows the unemployment-vacancy ratios for selected OECD countries since 1973. For Canada and the USA the vacancy measure is the Help Wanted Series. It is clear that the ratios have risen over the period with interspersed cycles. But if search behaviour was to explain these increases, we might expect an upward trend in unfilled vacancies. Otherwise, it is more plausible that the problem has been demand-side oriented and the rising ratios signal this.

Figure 1 shows the inverted unemployment rate and the total capital expenditure/GDP ratio for the OECD. There appears to be a strong association between the two. One might be tempted to assert that the major determining factor accounting for the changes in the level of unemployment in the OECD has been movements in the investment ratio. Figures 2 to 5 plot similar relationships for OECD Europe, Australia, United Kingdom and the USA, respectively, with similar results.

Figure 1 OECD Unemployment Rate (inverted) and Total Capital Expenditure/GDP

Figure 2 OECD Europe Unemployment Rate (inverted) and Total Capital Expenditure/GDP

Figure 3 Australia Unemployment Rate (inverted) and Total Capital Expenditure/GDP

Figure 4 UK Unemployment Rate (inverted) and Total Capital Expenditure/GDP

Figure 5 Annual Change US Non-Farm Payroll Employment and Annual Change Real Gross Fixed Capital Expenditure

4. The Phillips Curve in OECD Countries?

Figures 6 to 23 chart Phillips curve for selected OECD countries (see at end of paper). The first three concentrate on various measures of inflation and unemployment for Australia. Chart 6 is comparable to Charts 9 to 23 in that they cross plot unemployment and price inflation over 1970-1995.

Four notable features are present for most of the OECD economies. First, the instability in the trade-off in the mid-1970s after the first OPEC shock is apparent in every country except perhaps the Scandinavian economies of Finland, Norway and Sweden. At first inflation rose sharply between 1972 and 1974 and then unemployment followed after 1974. Second, there is a second wave of instability in the early to mid-1980s in almost every country except Finland, Italy, perhaps Germany, clearly Japan, Norway, and Sweden. Third, in recent years, the Scandinavian countries seem to have had very sharp declines in their inflation rates at significant cost in terms of unemployment. Fourth, the annual changes when unemployment is rising are significantly larger than changes when unemployment is falling.

An additional feature which emerges from the Australian plots (cf. Figure 6 and 7) is that the instability in the mid 1970s implicated both wage and price inflation, but although there was some large wage rises in the early 1980s, the recession of 1983 combined with the re-introduction of wage setting guidelines killed the wage inflation of quickly, but price inflation persisted for two more years.

Are there are any difference between the two periods of instability? Tables 6 and 7 provide some additional but incomplete information. Table 6 shows the real unit labour costs (RULC) in selected countries. The instability in the 1970s (shaded) is associated in every case with rising RULC which were reflecting sharp rises in money wages relative to inflation, then sharp decreases in economic activity and labour productivity. There was a large redistribution of real national income towards wages in this period. For the countries shown, except Sweden, RULC fell during the period of instability in the 1980s. There was a large redistribution of real national income over this period towards profits.

Table 7 shows real hourly earnings indexes for 1970-1995. Every country experienced rapid rises in the index in the period after the first OPEC shock. However, the experience in the early 1980s is mixed. Some countries had strong real earnings growth but different unemployment and inflation experiences (for example, Japan and the United Kingdom), while many countries had steady real earnings. Australia and Belgium (in addition to the OECD bloc as a whole) saw real earnings growth decline but still unemployment rose and price inflation persisted.

It seems that in the 1970s, some case could be made for a wages-driven inflation which led to rises in unemployment as governments finally stopped accommodating the wage-price spiral. However, in the 1980s, no such argument is possible. The high unemployment and persistent inflation does not appear to be wages-driven.

Table 6 Real Unit Labour Costs

Table 7 Real Hourly Earnings 1970-1995

5. OECD Output Gaps and the Required GDP growth rates

In this section we use some Okun's Law arithmetic to estimate the deficiency in GDP growth which has led to the rise in unemployment rates. We can relate the major output and labour-force aggregates to form expectations about changes in the aggregate unemployment rate based on output growth rates. A series of accounting identities underpins Okun's Law and helps us, in part, to understand why unemployment rates have risen. Take the following output accounting statement:

Y = G (1-UR)LH (1)

where Y is real Gross Domestic Product, G is labour productivity, H is the average number of hours worked per period, UR is the aggregate unemployment rate, and L is the labour-force. Equation (1) says that total output produced in a period is equal to total labour input times the amount of output each unit of labour input produces.

This identity can be converted into a dynamic equation expressing percentage growth rates, which in turn, provides a simple benchmark to estimate, for given labour-force and labour productivity growth rates, the increase in output required to achieve a desired unemployment rate.

Accordingly, assuming that the hours worked is more or less constant, we get a relationship between real output growth, labour force growth and labour productivity growth, for zero changes in the unemployment rate:

y = g + lf (2)

which shows that if the unemployment rate is to remain constant, the rate of real output growth must equal the rate of growth in the labour-force plus the growth rate in labour productivity. Cyclical movements in labour productivity and the labour-force can modify the relationships in the short-run.

The calculations are shown in Table 8 and confirm that for the selected countries except Japan, real GDP growth was insufficient to keep the unemployment rate from rising. For Australia, the cumulative GDP gaps adds nearly 5 per cent to the unemployment rate over the time period which is provides a reasonable estimate of the actual rises experienced.

Table 8 Required Real Output Growth (per annum)

7. Output Gaps and Persistence in OECD Unemployment Rates

The concepts of hysteresis and persistence have pervaded macroeconomic debate since the 1980s. LJN argued that to understand unemployment "an adequate framework requires a new combination of macroeconomics with a detailed micro analysis of the labour market". Hysteresis is a central concept in their quest. Under the heading of "Persistence", they argue that (p.18):

These disappointing experiences have raised in sharp form the issue of hysteresis in unemployment. Clearly, we have to modify our model to allow for this. If wage and price behaviour depends on the change in unemployment as well as on the level….

In terms of policy, hysteresis means that, once unemployment has risen, it cannot be brought back at once to the long-run NAIRU without a permanent increase in inflation.

The claim that we need to buttress our macroeconomic models with disaggregated explanations - the so-called micro foundations - has been used to attack Keynesian macroeconomic modelling since the early 1970s. Jackson and Pettit (1992) show that the claim is ill-conceived. But more worrying is the confusion between the concepts of persistence and hysteresis (see Mitchell, 1993 for a more complete discussion).

Coe and Gagliardi (1985) also model hysteresis in this way and say (p. 11):

The simplest test of the hysteresis hypothesis is to define U* as a distributed lag on past values of U in the estimation of equation [2]…{equation [2] is a standard Phillips curve specification with the activity variable being modelled as (U-U*)]…This also tests for a long-lived impact of unemployment rates on wage inflation if the constraint in equation [2} that U and U* have the same but opposite-signed coefficient is dropped.

In other words, Coe and Gagliardi see hysteresis as being equivalent to the inclusion of the change in the unemployment rate being included in the wage equation.

Amable, Henry, Lordon and Topol (1993) are extremely critical of LJN and their use of hysteresis. They argue that their concept of hysteresis is nothing but a unit-root system; and paradoxically, their econometrics is not unit-root econometrics. (p. 124)

One can wonder why LNJ decide to make use of the term "hysteresis" where it simply boils down to discussing the impact of changes in unemployment on the level of wage and price or, similarly, to looking for the existence of a long-run equilibrium in levels of unemployment and of real wage. This is simply a story about the persistence of shocks, or about unit roots, but not about hysteresis. (p.132)

Cross (1993, p.67) adds some light:

The problem with these linear characterisations is that hysteresis appears as a special case, arising from zero or unit roots. If the root is not exactly zero in the continuous time case, nor exactly one in the discrete time case, the property of hysteresis disappears.

A unit root process is a highly restrictive form of hysteresis where shocks have permanent effects. So how should one proceed? If we equate hysteresis in the unemployment rate with the presence of zero eigenvalue linear dynamical systems (or unit-root systems for discrete-time processes) then we can easily conclude that there cannot be hysteresis operating because the unemployment rate is a bounded-variable with a finite variance.

Mitchell (1993) outlines one way of approaching the difference between persistence and hysteresis (defined narrowly as a unit root process) by developing the distinction between trend-stationary (TS) and difference-stationary processes introduced by Nelson and Plosser (1982). While it is difficult to distinguish the two processes in practice, we know that they behave in quite different ways following a shock. An innovation to a TS process will not have permanent effects, although if the process is a near unit root, the memory of the process with respect to the shock will be long. Long memory in a time-series is referred to as persistence. While in analytical terms, persistence is a sub-category of a TS process, and is thus, clearly distinct from hysteresis; in practical terms, it may be virtually equivalent, given that the shock may last long enough to make policy effective.

This distinction relates to the debate in macroeconomics between natural rate theorists and economists who advocate aggregate demand intervention to reduce unemployment. Mitchell argues that the hysteresis hypothesis suggests that expansionary policy could permanently reduce unemployment at the cost of some inflation. Conversely, the natural rate hypothesis, distinguishes between the long-term secular trend and the short-term (transitory) fluctuations in the economy. At best, aggregate demand management can only stabilise the short-term variations, but it inhibits the 'natural' tendencies of an economy (if shocked) to equilibrate, and ultimately only influences nominal magnitudes (that is, causes inflation).

Table 9 Summary Unemployment Data for OECD countries

Mitchell (1993) also identifies an apparent tension between the theoretical and the empirical literature on unit roots and hysteresis where the unemployment rate was a commonly-used hysteretic variable (for example, Blanchard and Summers, 1988; Franz, 1990). In reality, the unemployment rate, being a bounded variable, cannot be a process with an infinite variance (see Nelson and Plosser, 1982; Mitchell, 1993; Mitchell and Wu, 1995).

So if there can be no hysteresis in the unemployment rate, does this mean that the conclusions of the natural rate are valid? Mitchell (1993) argues that once we clear the discussion from a comparison of hysteresis and persistence, then the real question is whether the degree of persistence is sufficiently large to go beyond the temporal boundaries of the typical political cycle. If the answer is yes, then activist government policy can attentuate the effects of shocks on the economy which would persist for long periods of time. In addition, some appropriate intervention can also reduce the compound effect of multiple shocks.

Further, Mitchell (1987) argues that cyclical changes impact on the structural characteristics of the labour market (affecting trainability and hiring standards) which blur the traditional distinction, on which natural rate theory is based, between structural and cyclical parameters.

Table 9 provides summary statistics for OECD unemployment rates. There is a strong rise in the mean of all unemployment rates over the sample and between the two periods shown. Also there is evidence that the variance in the rates has narrowed and indication of persistence at a level once it is established.

On a practical level, the degree of persistence in the unemployment data can be approached via output gap calculations. For each country a tested down autoregression was estimated ensuring that the dynamics remaining were free of serial correlation. The polynomials were solved for the "steady-state" unemployment rates in each country and an output gap, defined as the difference between the current and the steady-state unemployment rate, was created by introducing a 3 per cent negative shock.

The time paths back to equilibrium were simulated and the results are shown in Table 10. All the non-reported calculations and the simulation models are available on request.

While the results do not discriminate between the nature of the shock (demand or supply side), it is clear that there are considerable time elapses in all countries (Sweden) even before half of the output gap is eliminated (other things equal). So even if we fail to formally establish hysteresis (unit root hypothesis), in practical terms the policy implications are equivalent given the high persistence evident in the data.

The results suggest significant persistence in the unemployment rates across the OECD block and may indicate that our failure to resolve the exact statistical nature of the DGP driving unemployment is less important at a practical level. In practice, the persistence is of such a degree that policy-activism appears to still be a valid tool for reducing the costs of output gaps.

Clearly, macroeconomic policy can be designed to minimize the costs of each shock (that is, reduce the output gaps) before the next shock impacts. A non-interventionist policy would see the impacts of the previous shocks still substantially in the system as the next shock arrives. Thus, the Okun losses would be magnified.

Table 10 Persistence of Output Gaps following a 3 per cent negative shock

8. Removing the inflationary constraint on growth

It has been argued that the principle reason for the persistently high unemployment rates in OECD economies has been the lack of adequate GDP growth largely due to long periods of restricted policy. While the political analysis is the topic of another paper there has been good coverage of the proposition in Cornwall (1984). The question that arises is whether the inflationary constraint is perceived or real. In this section, we show that Australia's recent history of wage setting guidelines, called the Accord, which ended after the election of a pro-market Government in March 1996, was extremely effective in controlling the growth of wages. While other problems, principally a structural balance of payments constraint, prevented GDP from growing fast enough to reduce unemployment to very low levels, the fact remains that the period of the Accord saw relatively fast output and employment growth and falling real wages. Appendix C contains a brief history of the Accord.

In 1972, Australia's inflation rate was 6.2 per cent, but following the first OPEC oil shock in 1974, aided by some large wage increases, the inflation rate reached 17 per cent in 1975. By the end of the 1970s, despite a period of subdued activity and rising unemployment, the inflation rate was still high in relation to our trading partners at 9.2 per cent. The wage increases that followed the breakdown of the period of wage indexation in the early 1980s pushed the inflation rate, once again above 10.4 per cent, and provided the background to the introduction of the Accord in 1983. At that time, the unemployment rate and the inflation rate were at around 10 per cent due to the sluggish economy.

The Accord period in Australia was associated with strong employment and GDP growth from 1983-84 to 1989-90 (with the help of a growth oriented Labour Government), negative growth during the recession, and then a strengthening recovery after 1993-94. For the period 1984-85 to 1994-95, Australia's total employment growth per annum averaged 2.19 per cent, while the corresponding growth per annum for the OECD countries in total was 1.05 per cent. For the 1984-85 to 1989-90 period of expansion, the Australian figure was 3.43 per cent compared to 1.65 per cent for the OECD. Over the recession of 1990-91 to 1994-95, Australia's employment growth was 0.70 per cent per annum compared to the OECD outcome of 0.33 per cent per annum.

Mitchell (1987) found that there were constraining effects on wages growth in Australia as a result of imposing wage fixing guidelines. Watts and Mitchell (1990) updated and extended this study to estimate the effects of the first three stages of the Accord (up until the third quarter of 1988). They found (1990, p.160) "that the different eras of wage-fixing guidelines can be statistically differentiated and are robust across different specifications. Except for the third and fourth phases of the guidelines…which signalled the end of centralised wage fixation in 1981, incomes policy successfully imposed a negative trend on the growth of real earnings, which was consistent with prevailing policy objectives."

They also found no evidence of the "existence of a conventional Phillips Curve relating inflation to unemployment….the annual growth of real weekly earnings is largely independent of conventional excess demand proxies and is strongly influenced by the prevailing institutional arrangements for wage fixing." (p.161).

Chapman and Gruen (1990) compare all the empirical work to that time which estimated the impacts of the Accord on wage inflation. They concluded that on balance the Accord had reduced the growth of nominal wage inflation.

With the Accord now history, this section of the paper updates the econometric modelling to assess the extent to which it influenced the path of wage and price inflation. A model is estimated to test for cointegration as the first stage in modelling an error-correction representation of the wage-setting dynamics. This is an advance on the work of Watts and Mitchell (1990) and Mitchell (1987) in that the modelling explicitly considers the possibility of integrated data.

In Australian wage setting , the period 1968(3) to 1996(1) has been dominated by incomes policy with several distinct phases of wage fixation. Table 11 describes the phases and the specification of the econometric variables.

Table 11 Wage Setting Phases in Australia, 1968(3)-1996(1)

8.1 Time Series Properties

The data is quarterly and is filtered for deterministic seasonality. All analysis is in terms of the logarithm. Appendix B describes the data.

Table 12 displays the sample autocorrelations for all the data in levels, seasonal differences, and the first-difference of the seasonal difference. They are a preliminary guide to assist our interpretation of the more formal unit root tests.

There is considerable variation in the sample correlations shown. The price variables (LAWE and LP) reveal similar patterns, with the level of each showing very pronounced inertia. The ACF of a random walk exhibits behaviour similar to this (see Nelson and Plosser, 1982, p.147). The seasonal difference for both variables also decay slowly and it is not until this difference is first-differenced do the lags drop off rapidly and resemble a stationary series. All the levels of the other variables appear to be non-stationarity. However, it seems that seasonal differencing results in ACFs which decay fairly quickly.

Table 12 Sample Autocorrelation Functions for 1966(3)-1996(1)

Table 13 Unit Root Statistics