RBA Annual Conference – 1992 OECD Country Experiences with Disinflation Palle Schelde-Andersen[*]

1. Introduction

During the 1980s virtually all OECD countries have adopted policies to reduce inflation, but so far only Canada and New Zealand have introduced specific inflation targets and timetables for their central banks. To set the stage for the issues to be discussed – or left out – in this paper it is instructive to consider briefly three propositions by one of the participants in the Canadian discussion (see Fortin (1990)):[1]

  1. A policy aimed at lowering the rate of inflation permanently would, because of higher interest rates and an appreciation of the real exchange rate, be accompanied by important transitory costs, mainly in the form of lost production and higher unemployment. However, the permanent benefits of lower inflation would far exceed the transitory costs;
  2. The Bank of Canada should pursue the goal of price stability as rigorously as possible, but stop reducing inflation before full price stability is reached. Most price indices overstate the ‘true’ rate of inflation and, at a low rate of inflation, the marginal costs of reducing it further tend to exceed the marginal benefits;
  3. The approach towards the final goal should be cautious and gradual because: the anti-inflationary effectiveness of higher unemployment tends to decline as unemployment rises; large unemployment shocks risk creating severe unemployment persistence; lags in the effects of monetary policy call for a gradual approach; and the general public prefers to take the transitional pains in smaller doses over an extended period of time.

These propositions raise a number of both policy-oriented and technical issues, and more than can be dealt with in the following. The paper will not discuss the concept of price stability and the associated measurement problems. Nor will it attempt to measure the costs of inflation and the benefits of price stability. A number of the more technical issues – such as the shape of the Phillips curve (or the aggregate supply curve), the sources of unemployment persistence and the lags in the wage and price formation process – will be discussed only briefly and not analysed in depth. On the other hand, some issues not directly dealt with above will be taken up, including credibility, the initial conditions, the choice of nominal anchors and more generally the role of other policies in complementing anti-inflation monetary policies.[2] Given the subject of the paper, the approach will not be based on any particular model, but will draw on a wide range of models and theories in attempting to explain the experience of a given country or group of countries.

2. The Main Features of the 1980s Disinflation

During the 1980s the OECD area experienced a substantial fall in the rate of inflation, with the average rate – measured by the GDP-weighted private consumption deflator – falling by almost nine percentage points in the course of only six years. The principal features of the disinflation process may be summarised in four points:

  1. The origin can be traced to decisions taken in late 1979, when most OECD countries agreed not to repeat the accommodating policies adopted after the first oil shock. Led by a change in the operating procedures of the Federal Reserve Board, monetary policies were the key instrument in the anti-inflation stance, and both nominal and real interest rates rose substantially;
  2. Although more restrictive monetary policies provided the largest contribution to reducing inflation,[3] the process was also helped by the terms-of-trade gains[4] related to the steep fall in oil prices in 1985–86, and by the weakness of non-oil commodity prices during most of the decade. Regardless of whether the terms-of-trade improvements were exogenous or partly induced by the weaker demand growth in the OECD area, the resulting gains in real disposable income mitigated distributional issues, which in earlier years had been one of the main sources of inflationary pressure;
  3. Nominal and real wage restraint were also instrumental in reducing inflation and in preventing a re-acceleration of inflation during the second half of the decade, when output growth was stronger. This was especially evident in the manufacturing sectors, where the deceleration of nominal wages in almost all countries was more pronounced than that of prices (see Figure 1). In other sectors, and especially in the services sector, the degree of nominal and real wage moderation was less pronounced. When wages are measured by aggregate wages and salaries per employee, there is little evidence of lower real wage growth (see Table 1); [5]
Figure 1: Deceleration in Nominal Wage and Price Inflation, 1973/79 – 1986/89
  1. A unique feature of the disinflation process was its uniformity, despite variations in the policy mix across countries and volatile nominal and real exchange rates. Since countries entering the 1980s with a comparatively poor inflation performance tended to see a more pronounced decline in price inflation than traditional low-inflation countries, the decline in the area's average inflation rate was accompanied by a marked convergence of inflation rates across countries (see Figure 2). In addition, the year-to-year variability of inflation declined,[6] so that the ambiguous price signals typical of a high-inflation economy were significantly reduced.
Figure 2: Average and Standard Deviation of Inflation and Unemployment

However, while the process of disinflation was impressive and also unparalleled in the post-war period,[7] it was not without costs. During the first half of the 1980s real output growth for the OECD area was less than 2½ per cent per year, compared with 3½ per cent for the previous five years. Moreover, unemployment in the area rose sharply (see Figure 2) and, particularly in Europe, remained high well after output growth had returned to a stronger trend. This raises some questions about the costs of disinflation and, ultimately, whether the gains from reducing inflation justify the costs incurred.

3. The Costs of Disinflation: Theory

Economists should be circumspect when attempting to estimate the costs of reducing the inflation rate.’ Lucas (1990), p. 70.

(a) Policy Relevance of Analysing Costs

There is a general consensus that, while the benefits of lower inflation are permanent, the costs associated with reducing inflation are only transitory, so that a cost-benefit analysis would always come out in favour of adopting a policy of price stability. Yet a closer analysis of the costs is for several reasons relevant to evaluating the overall outcome:

  • while the economic costs may be transitory, the political costs incurred by a government could well be permanent, especially when the ‘pay-off’ period (i.e. the period required for the policies to produce a net gain) is longer than the electoral cycle;[8]
  • some have questioned the assumption of transitory costs on the grounds that the trade-off between inflation and unemployment is subject to hysteresis, because unemployed workers are disenfranchised from the labour market, or because investment and the capital stock decline. There is little evidence of extreme hysteresis (i.e. the case where the rate of inflation depends on changes in unemployment rather than on its level), but partial hysteresis and a lengthening of the pay-off period cannot be ruled out;
  • factors such as social costs and public acceptance of the anti-inflation policies are difficult to quantify in a cost-benefit analysis. However, they cannot be ignored as they may change the ranking of alternative policies and make certain options unfeasible. Survey evidence suggests that the relative weights assigned to the macroeconomic targets of low inflation and low unemployment change along the trade-off curve, with lower inflation given a relatively high priority when inflation is at a high level, while further ‘down’ the trade-off curve lower unemployment becomes the principal concern of the general public;
  • any government adopting anti-inflation policies would obviously want to minimise the associated costs. Hence an analysis of the sources and causes of such costs has policy relevance.

(b) Principal Determinants of Transitory Costs

The factors most often mentioned in the literature are lack of credibility and slow adjustment of prices and wages.[9] The former view is associated with the New Classical School and the latter with the New Keynesians, but in the following both will be regarded as essential to the transitory costs.[10] At the same time, a number of other factors should be added, in particular the initial conditions, terms-of-trade developments, the number of policy instruments used as ‘nominal anchors’, the speed with which the policies are implemented and the ‘mix’ of macroeconomic policies. Below we briefly discuss some features of these fundamental determinants, before turning to concrete measures of the costs of disinflation in Section 4.

(i) Credibility[11]

It goes without saying that anti-inflation policies which are not credible to economic agents will either fail to reduce inflation or will do so only at great cost. Conversely, a credible policy can, by affecting expectations directly, considerably reduce the costs of disinflation. Yet despite its potentially crucial role, the concept of credibility is not well defined in macroeconomics.[12] Moreover, the factors most likely to influence the credibility of a given policy change are not well understood, and the empirical literature in this area is still in its infancy.

According to Blackburn and Christensen (1989), the factors most likely to have an adverse influence on credibility may be classified into three broad groups: (i) technological constraints, including reliability of the data on which the policy decisions are based, controllability of the policy instruments and the acceptance and relevance of the economic theory used by the policy makers; (ii) administrative constraints, which mainly relate to the ability of the government to implement the required legislative changes; and (iii) strategic constraints, which refer to the interdependence between the behaviour of private agents and policy makers (analysed by most authors within a game-theoretic framework). One issue of particular interest in this respect is the time inconsistency dilemma, which can arise when the policy makers pursue several targets and the optimal policy strategy changes over time. As stressed by Blackburn and Christensen, the three sets of constraints are not independent of each other, and frequently the single most important determinant of credibility is the coherence or consistency of the overall macroeconomic programme, and especially whether fiscal policies support an anti-inflation monetary policy stance.

Empirical studies of credibility effects have relied on mainly two methods: (i) analysing economic relations before and after a given policy change using prediction errors or dummy variables; and (ii) deriving a credibility variable, or estimating a parameter hypothesised to capture credibility effects. The first method may be seen as a special application of the Lucas critique, and the reliability of the measures obviously depends on the stability of the underlying behavioural or reduced-form equations. Moreover, to the extent that credibility effects gradually build up (or decay) over time, models with time-varying coefficients would be required. Direct measures of credibility variables have mostly been derived from exchange rate or interest rate equations, and recent works in this area have used a Bayesian approach (see Baxter (1985) and Weber (1991)). One common problem with this method appears to be that the variables or parameters do not measure credibility alone. For instance, credibility parameters are often derived by decomposing a specific error process, and such parameters may combine policy reaction parameters with coefficients from price and wage equations (see Takeda (1992)).

Against this background – and despite their potential relevance to the topic of this paper – we have not attempted to quantify credibility effects. Instead we have relied on ‘snapshot empiricism’ (see Blackburn and Christensen) and evaluated the presence or absence of credibility on the basis of cross-country comparisons, case studies and the general consistency of the policies applied.

(ii) Wage and Price Rigidities and Unemployment Persistence

There are two major reasons why prices and wages do not immediately adjust to the inflation targets announced by the authorities. Firstly, because of built-in habits and the costs associated with obtaining new information, expectations tend to be backward-looking (or adaptive) rather than forward-looking. Secondly, even in a case of perfect credibility and forward-looking expectations, coordination failures (see Fortin (1991)) and institutional factors such as staggered contracts (see Fischer (1977 and 1986a) and Taylor (1980)) would prevent expectations from immediately affecting actual price and wage changes. Because of these lags (nominal rigidity), the adoption of anti-inflation policies will be accompanied by excess supply in labour and product markets, and the transitory costs will not only depend on the length of the lags, but also on the sensitivity of price and wage changes to a given degree of slack (real rigidity).[13]

Table 2 shows empirical estimates of the nominal and real rigidities[14] using a model (for details see the Appendix) which, to ease the exposition, can be simplified to:

with all variables measured in logs, and where w denotes nominal wages, p consumer prices (with an asterisk denoting expected values), u the rate of unemployment, and d the first difference operator. Inserting [2] into [1] then gives:

where η can be used as an indicator of real rigidity, and (1−β) as a measure of nominal rigidity. The model may be extended to include a labour demand equation with a real wage elasticity of −1, a price adjustment function where price changes equal changes in money supply (i.e. dp = dm) and constant labour supply. On these assumptions, a tightening of monetary policy aimed at lowering the long-run rate of inflation by 1 percentage point would initially reduce dw by only β points and increase real wages and the rate of unemployment by 1−β points. A convenient measure of the intermediate or historical costs (see Cozier and Wilkinson (1990)) of reducing inflation is (1−β)/η, which may be interpreted as the number of unemployment years required to reduce the rate of wage inflation by 1 percentage point, and is usually referred to as the ‘sacrifice ratio’. Nominal and real rigidities enter symmetrically, with (1−β) measuring the gap between the actual and the targeted wage inflation rate, and 1/η the rise in u required to eliminate this gap. In more realistic models two additional problems are frequently encountered. Firstly, in the case of hysteresis the unemployment term in [1] becomes – ηu + η αu−1, with α a parameter between 0 and 1. The presence of hysteresis reduces the sensitivity of dw to u from η to η(1−α), and in the extreme case of α = 1 the costs of disinflation become permanent. Secondly, the weighting pattern of the expectation formation process plays an important role as strongly backward-looking schemes would raise the numerator in the sacrifice ratio.[15]

Table 2: Sources of Labour Market Rigidities and Unemployment Persistence.
Hysteresis(α)1 Real wage
Nominal wage rigidity(1−β−1)1 Unemployment persistence
1 + σ2)2
United States 0.00 −0.23 0.73 0.75
Japan 0.68 −1.84 0.46 0.94
Germany 0.76 −0.45 0.47 0.94
France 0.00 −0.15 0.14 0.98
Italy 0.00 −0.38 0.26 0.96
United Kingdom 0.91 −0.15 0.20 0.92
Canada 0.00 −0.26 0.35 0.89
Austria 0.62 −0.94 0.78 0.91
Belgium 0.00 −0.29 0.37 0.96
Denmark 0.00 −0.19 0.31 0.97
Ireland 0.00 −0.19 0.12 0.94
Netherlands 0.36 −0.35 0.33 0.95
New Zealand 0.42 −0.09 0.38 1.07
Norway 0.00 −0.20 0.37 0.93
Sweden 0.69 −0.50 0.66 0.66
Australia 0.00 −0.18 0.00 0.95
1. Derived from a nominal wage equation specified as:
dlnW = μ + η (lnU − αlnU−1) + β1dlnPC + β2dlnPC−1 + (1−β1−β2)dlnW−1; see Appendix.
2. Derived from an autoregressive unemployment equation:
= σ1U−1 + σ2U−2; see Appendix.

(iii) Initial Conditions

Depending on their nature, these can have a direct effect on the transitory costs but may also have important indirect effects, either through the shape of the trade-off curve, or by affecting the extent to which anti-inflation policies are credible to the general public. Table 3 presents three sets of factors. Their likely influence may be summarised as follows:

Table 3: Initial Conditions1
‘Domestic’ indicators Fiscal position External indicators
CPI U PRO Bal Pbal Gdebt Sust BoP Fdebt REXCH PMR
United States 10.0 7.4 1.7 −1.2 −0.5 20.8 1.5 0.1 0.6 107.2 −2.2
Japan 6.0 2.1 −6.9 −4.1 −2.3 22.0 2.6 −0.3 2.2 97.2 −4.9
Germany 6.1 3.9 −1.6 −3.3 −1.7 18.6 1.5 −1.1 3.3 94.0 −4.3
France 13.2 6.9 −7.4 −1.0 −0.2 16.0 1.6 −0.7 4.2 97.7 −4.1
Italy 19.3 8.1 0.3 −10.1 −5.5 60.3 5.4 −2.2 2.2 88.3 −7.2
United Kingdom 13.8 7.3 −2.0 −3.0 −0.8 46.3 1.0 1.9 1.3 133.8 −2.9
Canada 10.6 7.5 −0.3 −2.1 −1.0 14.0 1.5 −1.0 −29.0 94.6 −5.0
Austria 7.0 2.2 −0.9 −1.7 −0.4 40.52 1.4 −2.1 −3.8 98.2 −5.6
Belgium 7.5 9.0 −12.5 −11.1 −4.5 85.2 10.6 −4.1 −5.8 80.9 −11.6
Denmark 11.3 8.3 −4.9 −5.1 −3.9 21.5 7.9 −3.5 −31.2 80.7 −7.2
Ireland 19.1 8.6 −4.5 −12.7 −8.0 88.02 7.7 −13.1 −49.6 96.9 −8.3
Netherlands 6.6 7.3 −0.5 −4.7 −2.5 29.4 4.0 0.1 6.4 89.5 −10.3
New Zealand 16.2 3.1 −13.3 −7.93 −3.13 48.33 1.43 −4.7 −26.7 111.0 −12.0
Norway 11.7 1.8 11.9 5.3 5.2 2.6 −4.3 2.8 −26.5 101.1 −4.4
Sweden 13.4 2.3 −3.6 −4.7 −4.7 −0.4 5.2 −3.0 −15.5 92.9 −6.5
Australia 9.9 5.8 −3.6 −0.3 −0.0 22.22 0.4 −3.8 −7.9 103.0 −3.3
1. Average 1980–81. 2. Estimated. 3. Central government only.
Notation: CPI
= consumer prices, percentage change
= unemployment as a percentage of total labour force
= change in profit share (operating surplus/GNP in factor prices), 1970/74 – 1980/81; in percentage points
= general government net lending as a percentage of GNP
= general government net lending less interest payments, as a percentage of GNP
= general government net debt, as a percentage of GNP
= indicator of sustainability, calculated as 0.01 Gdebt. (IR-Y) – Pbal. IR-Y is the long-term bond rate less change in nominal GNP (3-year moving average) and a positive sign indicates a need for raising taxes and/or reducing expenditure
= current external account as a percentage of GNP
Fdebt = net foreign assets as a percentage of GNP
= real effective exchange rate (unit labour costs) as a percentage of 1970–79 average
= change in ratio of import prices to consumer prices (1982–89), scaled by ratio of imports to GNP (1985).
  • the initial levels of inflation and unemployment[16] affect the costs when the trade-off is non-linear and/or when the weights assigned to inflation and unemployment in the welfare function of the general public change along the trade-off curve.[17] On the basis of the survey evidence referred to above and empirical estimates of the trade-off, it is frequently argued that at low rates of inflation and high rates of unemployment the costs of reducing inflation further rise disproportionately, as the trade-off curve is very flat and inflation has a low weight in the general welfare function. The profit share was also included among the initial conditions on the assumption that when profits are relatively high, enterprises may be able to lower prices without reducing labour demand or cutting investment plans;[18]
  • the fiscal position can influence the costs of disinflation in at least two ways. Firstly, an unfavourable position (measured by the overall balance, the level of public debt or the sustainability of current policies) is likely to reduce the credibility of anti-inflation monetary policies. Secondly, a large initial deficit narrows the scope for offsetting the contractionary impact of more restrictive monetary policies through a fiscal stimulus;
  • the external position can affect the costs along the same lines as the initial fiscal balance. A large external imbalance (measured by the current account or the net asset position) would tend to reduce credibility, and an unfavourable competitive position narrows the scope for offsetting the fall in domestic demand growth through net exports.

(iv) Terms-of-Trade Changes

The last column of Table 3 gives changes in relative import prices (scaled by the ratio of imports to GNP), which for all countries show a marked decline and, to a large extent, explain the trough of consumer price inflation in 1986. In addition to this one-time effect of the fall in import prices, the associated terms-of-trade gains are likely to have had a permanent effect on the sacrifice ratios, by reducing the degree of slack required to solve the problem of inconsistent income share claims.

(v) Policies

This is potentially the most important influence on the costs of disinflation, but also the one that is most difficult to evaluate. The effects of policy changes are generally difficult to quantify and, even when reliable measures exist, the net effect of various policy changes can be uncertain. Some features of the policies adopted are summarised in Table 4, but even if the information were complete, several unresolved issues remain, among which four deserve further elaboration.

One issue frequently discussed in the literature is the choice of nominal anchors and, more specifically, whether a monetary aggregate target or a fixed nominal exchange rate is the most effective instrument for anchoring inflation expectations.[19] There is general agreement that in periods of large shifts in the demand for money function a fixed exchange rate will be a more effective anchor. But it is also recognised that fixing the exchange rate before the rate of inflation has declined entails a high risk of leading to an appreciation of the real rate.[20] Generally it is difficult to use monetary aggregate targets and a fixed exchange rate as complementary anchors, whereas several countries (Austria, Belgium, Denmark, Norway, France and, more recently, Ireland) have combined a fixed exchange rate target with permanent or transitory incomes policy measures.[21] Australia appears to be the only one of the countries included in Table 4 to rely on both monetary policy and a permanent incomes policy (the ‘Wage Accord’) as the principal means of reducing inflation.

A second issue is the policy mix. According to Mundell's (1971) assignment rule, a tight monetary policy combined with an expansionary fiscal policy provides an ideal mix for minimising the sacrifice ratio. The United States successfully applied this combination during the first half of the 1980s,[22] but it is a strategy subject to several problems:

  • a large proportion of the fall in the rate of inflation may be only temporary, as it results from an appreciation of the exchange rate, which will eventually have to be wholly or partly reversed to improve the external position. Moreover, the strategy is obviously not feasible on a worldwide basis, and a number of countries which in the past adopted Mundell's rule experienced a depreciation and not an appreciation of the exchange rate;
  • growing fiscal and external imbalances can adversely affect the credibility of anti-inflation monetary policies. Moreover, in the early 1980s, fiscal consolidation was a primary objective in most countries outside the United States, and the use of fiscal measures to influence aggregate demand was not very high on the policy agenda.

A third issue concerns the speed of implementation, or whether a shock (‘cold turkey’) approach or a gradual approach produces the lowest sacrifice ratio. According to Kahn and Weiner (1990), the gradual approach taken by the US authorities in the late 1950s led to the same sacrifice ratio as the more drastic tightening of monetary policy in late 1979. More recently, a consensus seems to have emerged that the speed of implementation should be geared to the lags in the price and wage formation process. As shown by Taylor (1983), and also in several simulations reported in Chadha et al. (1991), such a ‘matching’ can reduce inflation with virtually no output and employment costs.[23] However, a major problem is that the initial speed of disinflation is very slow,[24] which is likely to adversely affect the credibility of the announced inflation targets.

A fourth but separate issue is the interpretation of changes in monetary aggregates. Considering the importance assigned to monetary policy in reducing the rate of inflation, the development in monetary aggregates is conspicuous by its absence in Table 4. However, for mainly two reasons it was not possible to find plausible and comparable indicators for monetary policy:

  1. In virtually all countries the development of monetary aggregates has been distorted by shifts in portfolio preferences, which frequently took place in response to regulatory changes and the availability of new financial instruments;
  2. In a few countries there is clear evidence of the re-entry problem mentioned above, which obviously poses a problem of interpretation, and even more a problem of selecting an appropriate strategy. When, for instance, the rate of inflation and the nominal interest rate decline as a result of a credible anti-inflation monetary policy, the demand for money will rise,[25] creating a dilemma for the monetary authorities. On the one hand, if the rise in demand is accommodated and the aggregate target is overshot, credibility is likely to suffer. On the other hand, if the demand rise is not accommodated, the adjustment will have to come through alternative channels (higher nominal and real interest rates, lower output growth or an unsustainably steep fall in the rate of inflation), which may be equally damaging to credibility and the costs of disinflation.[26]

4. The Costs of Disinflation: Empirical Measures

(a) Methodology

As mentioned above, the transitory costs are usually presented as sacrifice ratios and three methods have been used to derive such ratios.[27] They can give quite different results[28] and each has its own particular advantages and shortcomings.

(i) Slope of Short-run Phillips Curve (or Aggregate Supply Curve)

This is the method applied by most researchers and merely requires estimates of the Phillips curve or the aggregate supply curve. The precise definition varies, as some use the real rigidity coefficient (1/η, see above), while others combine nominal and real rigidities ((1−β)/η). The ratios are easy to derive, but are only partial measures and, in particular, cannot take account of shifts induced by changes in expectations.[29] Special problems are also encountered when the trade-offs are non-linear and, as shown in Rose (1988) and Cozier and Wilkinson (1990), the sacrifice ratios are highly sensitive to different methods of estimating equilibrium unemployment and potential output. Moreover, in studies which combine trade-offs from the Phillips curve with slopes of the Okun curve in order to derive a sacrifice ratio based on foregone output, the range of estimates is particularly wide. Two examples illustrate this point and, at the same time, provide some perspective to the disinflation of the 1980s. In a well-known review of the 1970s, Tobin (1980) concluded with Figure 3, using the wage-price model generally accepted for the United States at that time. Although Tobin is careful to point out that the figure is not intended to be a forecast, but a path against which actual developments should be compared, it is instructive to note that the unemployment-based sacrifice ratio using Tobin's figures for the period 1980–81 to 1986–89 would be 2.5, or three times the value given in Table 5. A comparison with actual developments also shows that the more favourable outcome can mainly be ascribed to a much faster adjustment of inflation than predicted by the model.[30] As a second example, assume that the Phillips curve is linear with a slope equal to ‘a’ and that the Okun curve is also linear with a slope equal to ‘b’. A one point rise in the rate of unemployment would then be accompanied by an ‘a’ point decline in the rate of inflation and a ‘b’ point deceleration in real output growth, with the ratio of real to nominal output changes equal to b/(a+b) (Fortin (1990)). A consensus view would probably put ‘a’ at around 0.5 and ‘b’ at 2.5 for all OECD countries on average, producing a ratio of 0.85 and implying that a policy aimed at reducing inflation would mainly result in lower real output growth. However, as can be seen from the bottom lines of Table A3 in the Appendix, the actual change between 1973–79 and 1979–81 produced a ratio of only 0.35, and between 1979–81 and 1986–89 the ratio was negative, as real output growth increased while nominal output growth declined. It was only during 1990–91, when output growth became negative in several countries, that the ratio reached 0.85.

Figure 3: The Phillips Curve
Table 5: Sacrifice Ratios1
U Y1 Y2 Y3
United States 0.85 1.35 0.40 0.30
Japan 0.70 0.50 0.10 −0.30
Germany 6.40 2.20 1.20 1.60
France 2.15 0.50 0.50 0.20
Italy 1.40 0.65 0.45 0.20
United Kingdom 3.40 0.90 0.60 0.20
Canada 2.75 1.50 0.80 0.15
Austria 2.05 4.65 0.90 0.90
Belgium 3.00 6.10 0.90 0.50
Denmark 1.25 0.90 −0.15 0.15
Ireland 3.35 0.30 0.50 0.50
Netherlands 4.95 1.85 0.80 1.10
New Zealand 2.05 0.25 0.55 0.85
Norway 1.05 2.00 −0.00 0.50
Sweden 1.20 0.65 0.00 −0.40
Australia 6.00 1.25 0.25 −0.15
Average2 2.50 1.60 0.50 0.40
1. The first column is calculated as the cumulative rise in the rate of unemployment between 1979–81 and 1986–89 divided by the change in the rate of consumer price inflation over the same period. The next three columns are calculated as the cumulative output losses between 1979 and 1982, 1985 and 1988, respectively, divided by the change in the rate of consumer price inflation between 1979–81 and 1984, 1987 and 1990, respectively. In calculating the output losses trend or potential output was approximated by a cubic trend (see Appendix), except for New Zealand, where the 1960–79 trend was extrapolated into the 1980s.
2. Unweighted

(ii) Counter-Factual Model Simulations

Analytically, this is by far the most satisfactory method as it is comprehensive and exogenous factors are isolated. The sensitivity of costs to changes in the lag structure of the price and wage formation process can be estimated, and it is also possible to illustrate the effect of changes in credibility (see Chadha et al. (1991), Murphy (1991) and Selody (1990)). Nonetheless, the cost measures may be subject to a bias – though the size as well as the sign of the bias is unknown – as virtually all macro-models are based on the assumption that in the long run all real variables are neutral with respect to changes in monetary policy and in the rate of inflation, and thus do not include the benefits of lower inflation.

(iii) Actual Developments

This is the easiest method, as it merely requires the calculation of cumulative changes in unemployment (or lost output) over a certain period and the division of this figure by changes in the rate of inflation over the same period. It also has the advantage that credibility gains or changes in the expectation formation process are included, though not quantified. An obvious disadvantage, however, is that the influence of factors which are entirely exogenous to the anti-inflation policy is difficult to exclude[31] and the ratios are very sensitive to the period chosen. Another problem is whether to measure the cumulative output and employment losses relative to some potential (or equilibrium) level or rate of change, or to base the comparisons on the actual values observed just prior to the implementation of anti-inflation policies. Most empirical studies have applied potential levels and rates, which requires that such values have to be estimated in advance and may lead to biases if the initial year is far away from the equilibrium position. The measures discussed below rely on deviations from actual values in the initial period, but this procedure is also problematic, especially if there are trend changes in the underlying equilibrium values.

(b) The Costs of Disinflation in the 1980s

Table 5 shows sacrifice ratios based on actual developments. The first column measures cost in terms of unemployment, while the next three show output losses calculated for different time horizons. The ratios are only intended as broad indicators, but, nevertheless, may be used to highlight certain general features, provide a preliminary ranking of the sixteen countries and illustrate the influence of the factors discussed in the previous section.

In most cases the unemployment-based ratios shown in the first column give a more ‘pessimistic’ picture of the transitory costs than evaluations based on output developments. In several countries the cumulative change in unemployment between 1979–91 and 1986–89 is positive, whereas a number of countries do not seem to have suffered any output losses. Only three countries (the United States, Canada and Sweden) had managed to reduce unemployment to (or below) the initial level by 1989, and this was only temporary, as in all three cases the unemployment rates of 1991 exceeded those of the initial period. One reason for the discrepancy in the cost measures[32] may be that the equilibrium rate of unemployment has increased relative to the level of potential output. Alternatively, labour markets could have adjusted more slowly than output markets, so that unemployment had to be kept high over an extended period to obtain the required degree of wage restraint. Yet another possibility is that a cubic trend understates potential output for the 1980s, thus also understating the output losses. Both the IMF and the OECD have revised their assessments of potential growth rates and ascribe most of the revisions to the favourable impact of the supply-side policies pursued in many countries. On the other hand, if lower inflation has had a positive impact on potential output growth,[33] the use of higher rates would overstate the sacrifice ratios, since one of the expected benefits of lower inflation would be counted as a current cost.

Another feature of the measures shown is that the output losses tend to fall or to be reversed as the period is extended. This supports the assumption that the costs are only transitory, but also poses a problem regarding the choice of the time horizon for the calculation. On the one hand, if the period is too short, the lags in the wage-price adjustment process would tend to dominate and overstate the costs. On the other hand, as the time period is extended, the risk of including factors which are entirely exogenous to the policy of disinflation increases.

(c) Preliminary Ranking

Tables 7a and 7b rank the countries using the sacrifice ratios of Table 5. The position of individual countries will be discussed further below, and in the first instance we shall concentrate on the comparability of the four measures:

  • U is relatively closely correlated with Y2 and Y3 whereas the correlation with Y1 is radier low. It thus appears that the very short time horizon gives a misleading picture of the relative output losses, while the rankings tend to stabilise as the observation period is extended to include the second half of the 1980s;
  • for some individual countries the ranking also seems to be quite sensitive to the measure used. The United Kingdom and Australia show a better performance for the output measure than for the unemployment measure. In the former case this may be due to supply-side policies, which had a more pronounced influence on output markets than on the labour market, while in the case of Australia, it is more likely to reflect the unusually strong growth in the labour force. The reverse pattern is observed for Austria, which could be the result of wage flexibility, combined with a policy of letting foreign workers absorb a major part of fluctuations in labour demand;
  • in other countries the ranking is characterised by a distinct time pattern. In Canada, Sweden and Australia there is a clear improvement as the time horizon is extended, whereas in Ireland and New Zealand output performance deteriorates over time. Unfortunately these changes cannot be interpreted as indicators of relative performance, but rather underline the shortcomings of relying on one-dimensional indicators. Thus the low or negative Y3 measures observed for Canada, Australia and Sweden mainly signal excess demand conditions and rising inflation, while distortions in other areas – such as the external position – are ignored. In other words, when anti-inflation policies are relaxed, the sacrifice ratios are no longer a reliable indicator of economic performance and need to be supplemented by additional measures.
Table 6: Productivity and Inflation
% change, annual rate
1973–81 1981–89
United States 10.0 8.9 1.1   4.2 4.0 4.4
Japan 11.3 8.9 5.5   3.9 2.6 5.5
Germany 8.8 5.0 3.3   4.4 2.2 2.1
France 16.2 11.3 3.9   8.0 5.8 3.7
Italy 21.6 16.9 5.2   11.0 9.4 4.0
United Kingdom 19.0 15.1 1.5   7.5 5.3 5.3
Canada 12.2 17.0 1.6   5.8 5.1 1.7
Austria 9.4 6.5 4.6   4.8 3.1 5.0
Belgium 13.2 8.2 6.3   4.7 4.3 4.3
Denmark 13.1 10.9 4.0   5.8 5.6 0.4
Ireland 18.6 16.0 3.8   8.3 6.3 9.1
Netherlands 10.0 7.1 4.6   3.6 1.9 3.8
New Zealand 14.6 14.4 n.a.   7.8 10.7 n.a.
Norway 12.5 10.9 1.4   9.2 7.2 4.0
Sweden 13.3 10.7 2.1   8.3 7.5 2.6
Australia 13.2 11.5 3.3   7.6 8.0 1.1
Average 13.6 11.2 3.4   6.5 5.6 3.8

Notation: W = compensation per hour in manufacturing
CPI = consumer prices
Q = output per hour in manufacturing.

Sources: See Table 1 and note 5 on page 108.

Table 7a: Ranking Based on Sacrifice Ratios1
U Y1 Y2 Y3 Average2
United States 2 9 6 9 4
Japan 1 3 4 2 1
Germany 16 14 16 16 16
France 9 3 8 6 7
Italy 6 5 7 6 6
United Kingdom 13 7 11 6 12
Canada 10 11 12 4 10
Austria 8 15 14 14 13
Belgium 11 16 14 10 14
Denmark 5 7 1 4 3
Ireland 12 2 8 10 9
Netherlands 14 12 12 15 15
New Zealand 7 1 10 13 8
Norway 3 13 2 10 5
Sweden 4 5 3 1 2
Australia 15 10 5 3 11
1. For all measures countries with the lowest ranks are those with the lowest costs.
2. Calculated assigning a weight of 3 to U and 1 to each of Y1, Y2 and Y3
Table 7b: Correlations of Sacrifice Ratios*
U Y1 Y2 Y3
U 1.00 0.30 0.67 0.37
Y1 0.17 1.00 0.46 0.41
Y2 0.58 0.51 1.00 0.68
Y3 0.46 0.36 0.72 1.00
* Figures above (below) the diagonal indicate rank (simple) correlation coefficients.

(d) Sacrifice Ratios and Principal Determinants

Table 8 reports the results of a very crude attempt to link the sacrifice ratios to the principal determinants discussed earlier. Thus the four sacrifice ratios were each regressed – across the sixteen countries – on the initial conditions, the relative change in import prices over 1982–89 and the parameters of the wage and unemployment equations shown in Table 2. The debt ratios had no effect, and the primary fiscal deficits – separately or combined with the debt ratio and IR-Y – were also insignificant. The same applies to the external position and the profit share, while the rate of unemployment was marginally significant in the U-equation, but not at all significant in any of the Y-equations. Among the labour market measures, the hysteresis parameter (α)was never significant when entered separately, and (1−α) η (see pages 113 and 114), which in theory provides the most satisfactory measure of the long-run real rigidity, always produced less satisfactory results than η alone.

Table 8: Sacrifice Ratios and the Principal Determinants*
U U** Y1 Y2 Y3
c 4.68 (1.0) 4.61 (1.4) −1.50 (1.0) −4.49 (2.2)
dPC −0.35 (2.5) −0.37 (3.8) −0.29 (3.3) − 0.07 (2.5) −0.06 (1.8)
GDEF −0.26 (1.5) −0.31 (2.7) −0.16 (2.0) −0.09 (2.0) −0.06 (2.0)
REXCH 0.30 (0.8) 0.35 (1.4) 0.11 (1.2) 0.12 (1.2)
dPM 0.18 (0.8) 0.25 (1.6) 0.05 (0.7)
η 2.62 (2.2) 2.68 (3.2) 1.60 (1.9) 0.48 (1.8) 0.78 (2.7)
(1 − β) 2.93(2.1) 0.72 (1.0) 1.92 (2.7)
σ1 + σ2 4.25 (3.7) 1.70(1.1) 4.11 (2.8)
R2 0.13 0.58 0.42 0.07 0.41
SE 1.7 1.1 1.2 0.4 0.4
* Notation:
dPC = percentage change in consumer prices, 1980–81
GDEF = general government net lending as a percentage of GDP, 1980–81
REXCH = real exchange rate, 1980–81 as a percentage of 1970–79 average
dPM = change in the ratio of import prices to consumer prices, 1982–89
η = indicator of real wage flexibility (see Appendix)
(1 − β) = indicator of nominal wage flexibility (see Appendix)
σ1 + σ2 = indicator of unemployment persistence (see Appendix).
t-values given in brackets.
** Also includes a dummy variable (coefficient 2.85 (3.5)) which equals −1 for the United States, 1 for Australia and 0 for other countries

Turning to Table 8, the most satisfactory results are obtained for the unemployment-based sacrifice ratio when the deviations for the United States and Australia are removed.[34] However, even for the uncorrected equation, the coefficients have rather high t-ratios and give some support to the various hypotheses discussed above:

  • a high initial rate of inflation seems to reduce the sacrifice ratio, thus suggesting that inflation is more costly to reduce when it is already very low;
  • a large fiscal deficit tends to raise the transitory costs, but, due to the crude nature of the estimates, it is not possible to say whether this effect is transmitted through lower credibility or a subsequent tightening of the fiscal stance;
  • similarly, an initially high level of the real effective exchange rate (i.e. an unfavourable competitive position) raises the sacrifice ratios, probably by narrowing the scope for increasing net real exports to offset the fall in domestic demand growth;
  • the fall in relative import prices (scaled by the ratio of imports to GNP) obtains the expected positive sign, but only affects the unemployment-based sacrifice ratio significantly. This may indicate that distributional issues are more important in the wage bargaining process than in the determination of prices, and that the terms-of-trade gains have prevented an even steeper rise in the rate of unemployment;
  • for the output-based equations, the influence of the initial conditions progressively declines as the time horizon is extended, whereas the structural labour market parameters either increase or become more significant;
  • finally, when comparing the U-equations with the Y-equations, certain differences in the structure of parameters are rather surprising and difficult to explain. Thus nominal rigidities only affect the output-based measures and (1−β1) actually obtained the wrong sign when included in the U-equation. Similarly, the indicator proposed by Alogoskoufis and Manning (1988) (see the Appendix) as an overall measure of unemployment persistence only affects the output-based measures. A possible, but very tentative, conclusion to be derived from this might be that the rise in unemployment following the adoption of a restrictive monetary policy only depends on the sensitivity of wages to a given degree of slack, whereas broader measures based on output developments depend on both real rigidities and the lags with which monetary policy affects real output.

5. The Experience of Individual Countries: An Overview

This final sub-section summarises the information presented in the preceding tables with respect to individual countries, and on this basis selects those cases which call for further analysis. Among the major countries, the United States and Japan obtain favourable rankings, especially when using the unemployment-based ratios. In the case of Japan this can probably be ascribed to the high sensitivity of nominal wages to labour market slack documented in a number of other studies, and confirmed in the Appendix. Table 7a also suggests that output markets in Japan are less flexible than the labour market, and this hypothesis is further supported by Figures 1 and 3, as Japan is the only country which has experienced virtually the same deceleration in manufacturing wages and consumer prices and a slight gain in real wage growth. For the United States, the favourable ranking is more difficult to explain, though the estimates reported in Table 8 could indicate a behavioural change which reduced the transitory costs. Moreover, several studies have emphasised the high degree of flexibility characterising the US labour market compared with those in Europe and, as a consequence, its ability to adjust smoothly to various shocks.

Germany and several of the countries which throughout the 1980s have linked their currencies closely to the Deutschemark (Austria, Belgium, France and the Netherlands) obtain a rather low ranking. For Germany the outcome is particularly unfavourable, which might be related to a relatively high degree of real rigidity, though the possibility of a trend rise in equilibrium unemployment cannot be excluded.[35] For Austria the unemployment-based measure may understate the costs, whereas for the Netherlands, France and Belgium the rankings are more or less independent of the measures used. Among the ERM countries with looser links to the Deutschemark, Italy is slightly better placed than France, even though the initial fiscal and external positions were much less favourable. According to Giavazzi and Spaventa (1989), the main reason for Italy's low sacrifice ratio lies with the expansionary policies of the 1970s and the relatively high profit share in 1980–81. Faced with large supply shocks and very rigid real wages (because of indexation), but counting on fiscal drag and some myopia on the part of employees with respect to post-tax real incomes, the authorities ‘accommodated’ the supply shocks by allowing the exchange rate to weaken, and supported profit margins further by using higher income taxes to finance a reduction in employers' social security contributions. The contrasting performances of Denmark and Ireland are also worth noting. Thus Denmark shares the lowest average rank with Japan and Sweden, whereas Ireland obtains a rather poor rank, even though since 1978 the Deutschemark has been used as a nominal anchor for anti-inflation policies. The particularly high costs, when evaluated in terms of unemployment, together with the ‘outlier’ position of Ireland in Figure 4, may suggest that the performance is partly explained by labour market rigidities. In fact, if allowance were made for the large numbers of Irish workers who emigrated during the period, the unemployment costs would be even higher.

Figure 4: Changes in Real Wage Growth and Unemployment 1973/79–1986/89

Labour market rigidities also appear to have influenced the rank obtained by the United Kingdom, as well as that country's position in Figure 4. As already noted, there is some evidence that supply-side policies have reduced rigidities in output markets, and thus helped to lower the transitory output costs. Canada obtains a poor rank when evaluated by unemployment changes and, as noted above, the output-based measures tend to get distorted towards the late 1980s.

Among the remaining countries, Sweden obtains a favourable ranking regardless of the measure used. This, however, may be more indicative of the quality and limitations of the indicators applied than of Sweden's actual performance. During much of the period Sweden continued to give high priority to the unemployment target, and the rate of inflation was still around 6 per cent in 1986–89, rising further to 11 per cent in 1990 (though mainly due to the effects of a tax reform). The recorded changes in unemployment and the output losses may also be biased, since the former was kept low through a wide range of employment measures, and the estimated potential growth rate of just over 2 per cent may understate true growth potential. Norway, too, has given high priority to the unemployment target and the deceleration in wage and price increases has been comparatively modest (see Figure 1 and Table 1). At the same time, developments in the oil-producing sector have led to changes in actual output which are not related to policies. Finally, Australia and New Zealand lag behind most other countries in terms of the timing of anti-inflation policies. In both countries the rate of consumer price inflation was still over 6 per cent in 1990, and it was only last year that an inflation performance similar to that of most other countries was achieved. Judging by Figure 4 and Table 2, real rigidities are high in New Zealand, while in the case of Australia the unemployment-based sacrifice ratio appears to be overstated (see Table 8), probably because of the rapid growth of the labour supply. In assessing the relative position of Australia, it is also important to keep in mind that the terms-of-trade gains were rather modest. In fact, a depreciating exchange rate put upward pressure on prices and provided only a minor boost to the growth of net exports.

The rather poor ranking of the countries with close currency links to the Deutschemark is one of the more surprising results, as it might have been expected that such a linkage would strengthen credibility and lower the transitory costs.[36] Relatively high sacrifice ratios for ERM countries have also been observed in other empirical analyses of inflation and labour market adjustments, and two recent studies go some way towards explaining the puzzle or at least identifying the source of the problem:

  • relying on pooled time series and cross-country data for ERM countries, van Poeck (1989) estimates price change and unemployment equations using money supply developments and measures characterising the wage formation process as explanatory variables. From his results three conclusions emerge: firstly, money supply growth only affects price changes, and has no effect on unemployment, suggesting that the convergence of inflation rates can mainly be ascribed to a convergence of monetary policies; secondly, measures of the wage formation process, which were derived from time series estimates for individual countries and include real rigidities, proxies for indexation and features of the bargaining structure, only affect the rate of unemployment, and have no influence on the rate of inflation, suggesting that the observed divergences in unemployment can mainly be ascribed to institutional differences and divergent labour market policies; and thirdly, average unemployment in the ERM area has been subject to a trend rise independent of monetary policies and labour market characteristics;
  • Artis and Ormerod (1991) first estimate price expectation equations (using an autoregressive process) and find that the German rate of inflation enters significantly into price expectations in other ERM countries after 1979.[37] At a second stage they enter price expectations as an explanatory variable in structural wage adjustment equations, while at the same time searching for parameter shifts and structural breaks. They find little evidence of changes in their definition of real rigidities, but some evidence of lower long-run real wage aspirations after 1979, except for Germany and the Netherlands, where real wage aspirations appear to have strengthened.[38] From the results reported it may be concluded that the link to the Deutschemark clearly helped to dampen inflation expectations in other ERM countries. However, it also appears that in the absence of specific changes in the wage formation process and in the natural rates of unemployment, other ERM countries were forced to follow the German rise in unemployment to validate the lower rates of inflation.

The overall conclusion which can be drawn from these two studies would thus seem to be that the exchange rate commitment was instrumental in generating a convergence of monetary policies and rates of inflation. However, because of differences in the characteristics of the wage formation process and only marginal changes in the structural parameters,[39] the exchange rate commitment did not help to reduce the transitory costs. In fact, the sacrifice ratios may have been influenced by a trend rise in the natural unemployment rate of the key currency country.

6. Case Studies

In this section countries have been selected for further analysis: France, Denmark and Ireland, which are all members of the ERM, have achieved spectacular gains in reducing inflation, but experienced very different transitory costs; and Canada, which appears to have incurred rather large costs in the past, and is one of two countries with a target of lowering inflation in accordance with a pre-announced time schedule. In discussing the four countries we shall rely on the tables and figures presented above, plus Table 9, which shows contributions to changes in the domestic demand deflator over three separate periods.

Table 9: Changes in Domestic Demand Deflator by Contributing Factor*
percentage changes and percentage point contributions
1978–81 1981–86 1986–90
dPDD   12.2 7.4 2.9
of which: dULC 6.9 3.8 1.3
dPROF 2.8 3.1 1.2
dTIND 1.5 1.0 0.5
d(PM-PX) −0.1 −0.6 0.0
Rest 1.1 0.1 −0.1
dPDD   9.5 4.8 3.9
of which: dULC 6.0 2.3 3.0
dPROF 3.8 1.7 0.9
dTIND 1.1 0.6 0.6
d(PM-PX) −1.5 0.9 0.1
Rest 0.1 −0.7 −0.7
dPDD   10.5 6.1 4.6
of which: dULC 5.2 3.7 2.2
dPROF 1.5 2.2 1.8
dTIND 1.9 0.6 0.0
d(PM-PX) 2.4 −0.5 0.2
Rest −0.5 0.1 0.4
dPDD   17.4 7.5 3.5
of which: dULC 8.5 3.6 0.3
dPROF 1.1 3.5 1.7
dTIND 2.8 0.8 0.6
d(PM-PX) 4.1 −0.5 0.5
Rest 0.9 0.1 0.4
* The contributions are calculated from national accounts data, using the equation: dPDD = a1dULC + a2dPROF + a3dTIND + a4d(PM-PX) + Rest. with:
= domestic demand deflator
= compensation per unit of output
= operating surplus (including depreciations) per unit of output
= indirect taxes less subsidies per unit of output
= import price deflator
= export price deflator
= residual items, mainly reflecting changes in composition
= first difference operator
= share (1981) of component i in domestic demand.

(a) France

Under the heading of ‘competitive disinflation’ macroeconomic policies in France since early 1983 have aimed at two targets: reducing the inflation rate to that of its major European trading partners, and making labour, output and financial markets more competitive. The policies have been pursued persistently by three successive governments and in many respects the results have been impressive:

  • from a peak of over 13 per cent in 1980, consumer price inflation has been reduced to only 3 per cent, and last year the French inflation rate fell below that of Germany, with most forecasters predicting a similar pattern for this year (1992);
  • the deceleration in nominal wage growth has been even more impressive (see Figure 1); as early as 1986 the growth of unit labour costs in French manufacturing fell below that of Germany (where it remained until 1990). Moreover, since 1983, aggregate unit labour costs growth has been below the rise in the GDP deflator, allowing the gross profit share to be restored to the level of the early 1970s;
  • abandoning a long tradition of interventionist policies, the authorities have liberalised prices in virtually all sectors without triggering any inflationary shocks.[40] In addition, though much remains to be done, the authorities have taken several steps to strengthen competition and remove unfair practices in output markets;
  • reform policies have gone even further in financial markets, including the major instruments of monetary policy. While France has relied on monetary aggregate targets since the early 1970s, the previous system based on credit and exchange controls has been replaced by a market system, where variations in short-term interest rates and reserve requirements are the principal instruments for implementing monetary policy. Furthermore, the rapid growth of money, bond and equity markets (as well as markets for derivative instruments) has reduced the role of bank intermediation and created the conditions for a more efficient allocation of resources;
  • since 1983 the nominal effective exchange rate has been virtually stable against other ERM currencies and the real exchange rate (based on unit labour costs) has actually depreciated by over 5 per cent. This is in sharp contrast to the previous eight years, when the nominal effective rate had to be depreciated by almost 20 per cent to prevent a deterioration in the competitive position.

In achieving these results France benefited from relatively favourable initial conditions: the budget deficit was small and fiscal policy almost sustainable; the external deficit was modest, net foreign assets amounted to more than 4 per cent of GDP and the real effective exchange rate was below the average for 1970–79. Yet more important to the final outcome was the fact that throughout the 1983–91 period the overall policy setting has been consistent. Monetary policy was kept restrictive[41] to support the exchange rate target, and fiscal policy was also restrained, as the general government deficit was progressively reduced from a rather small peak of 3¼ per cent of GDP, and the ratio of net public debt to GDP remained constant at around 25 per cent. Finally, during the early phase, incomes policies complemented the overall anti-inflation stance as prices and wages were frozen for four months from June 1982, private sector service prices were subject to control until end-1986,[42] and from early 1984 public sector wage adjustments were based on official inflation forecasts rather than on past price changes.[43] The last measure especially proved important, as it was also adopted in the private sector, and meant that the deceleration in price inflation was quickly reflected in wages.[44] [45]

Yet the costs of disinflation have been substantial and according to the rankings shown in Table 7a France occupies a place in the middle of the range, with no major difference between the output and unemployment based measures. The reasons for this relatively disappointing performance can be found in mainly two areas. Firstly, during the early 1980s, when other countries adopted measures to reduce inflation, France went ‘against the trend’ by implementing an expansionary fiscal policy, relaxing monetary policy and taking steps to increase the share of labour income in total GDP. As is well known, the attempt failed and the resulting devaluations of the French franc within the ERM, combined with a history of price increases well above the OECD average, probably meant that inflation expectations were only adjusted in step with moderation in the actual rate of inflation, and not in response to the policies and targets announced.[46] Developments in interest rates – though also influenced by a number of other factors – give some support to this interpretation. The difference between long and short-term rates remained in the 1.5 – 2 point range during 1980–88, and the premium vis-a-vis German long-term bond rates rose to 6.5 – 7 points in 1982–84, and only fell below 3 points in 1989. Real long-term bond rates gradually rose from 1.5 per cent in 1980 to around 7 per cent in 1987, and remained close to the 7 per cent mark for the next four years. This rise in real interest rates, which, of course, was also related to the fixed exchange rate target and the tight monetary policies, meant that the profit share gains as a result of wage moderation were more or less offset by higher net interest payments, and probably contributed to the relatively slow growth of business fixed investment and the capital stock.

Secondly, a persistently high rate of unemployment combined with a deteriorating trade balance for industrial goods points to major weaknesses on the supply side. As regards the labour market, the disinflation process in France has benefited from a low degree of nominal wage rigidity (see Table 2), whereas the degree of real wage rigidity and unemployment persistence is very high.[47] In fact, a number of factors suggest that the wage moderation observed since 1983 does not reflect a change in trade union and employee behaviour, but is mainly the result of high unemployment:

  • a particular feature of the tax structure in France is the high level of social security contributions[48] and a ‘tax wedge’ which is wider than in most other OECD countries. Moreover, a sustained rise in the wedge since the early 1970s[49] has, according to Sachs and Wyplosz (1986), progressively raised the NAIRU, so that by the time of the expansionist policies of the early 1980s the actual rate of unemployment was equal to the NAIRU;[50]
  • another feature of the French economy is a relatively high minimum wage (almost 60 per cent of the average industrial wage and paid to 12 per cent of all dependent employees), which is likely to have created a wage floor for low-income and unskilled workers, and reduced employment, especially among the young (OECD (1990)).

Turning to output markets and the rising external deficit for manufactured goods, one striking development has been that, despite the improvement in the competitive position (as measured by the real effective exchange rate), gains in export market shares have been very modest, and on the domestic market the degree of import penetration has increased sharply.[51] Although changes in the composition and regional distribution of international trade may have been unfavourable to France, the poor trade performance could indicate that the capacity to take advantage of the better competitive position was absent, probably because of the slow investment growth during most of the 1980s.

Although the exchange rate commitment does not seem to have provided any direct credibility gains, its importance in terms of committing policy-makers to reducing inflation can hardly be overstated. This was particularly evident in the spring of 1983, when the Government faced the option of either leaving the ERM and going for a more expansionary course, or staying within the ERM and adjusting policies towards more restriction. The Government opted for the latter course and with that decision incurred an overall policy commitment for the rest of the decade.[52] It is more difficult to say whether the deregulation of financial markets has helped to reduce inflation and/or eased the transitory costs. On the one hand, to avoid distortions, the deregulation of financial markets took place in conjunction with the liberalisation of output markets and prices, and the absence of serious pressure on the currency following the abolition of exchange controls may have strengthened the credibility of the exchange rate commitment. On the other hand, the frequent changes in aggregate targets made necessary by the deregulation of financial markets have probably lessened the monetary authorities' ability to anchor inflation expectations. In addition, while the abolition of credit controls will have a favourable influence on resource allocation in the long run, the short-run effects were dominated by the steep rise in nominal and real interest rates and the slow growth of investment.

(b) Canada

Canada's experience with disinflation has not been particularly favourable: the sacrifice ratio based on unemployment costs is well above 2, and in the overall ranking Canada is only in tenth place. Considering that Canada is one of only two countries to have announced an explicit medium-term inflation target as the basis of monetary policy, it is natural to ask whether this performance was influenced by factors specific to the 1980s, or whether it was the result of more fundamental problems.

Initially, it may be useful to make some broad comparisons with the United States and with the other three countries discussed in this section. As can be seen from Table 3, Canada started from the same initial levels of unemployment and inflation as the United States, but the degree of disinflation during the 1980s was smaller, especially if measured by nominal wages. On the cost side the largest difference can be seen for unemployment, with the Canadian rate remaining well above that observed for the United States. The rigidity measures give a mixed picture. Hysteresis appears to be absent in both countries (though see below) and real rigidities are about the same. Nominal rigidity is much smaller in Canada than in the United States, whereas unemployment persistence is higher. On the fiscal side Canada started from a more favourable position than the United States. Nonetheless, the debt/GDP ratio rose more sharply, which can in part be explained by the tighter monetary policy as shown by the relatively steep rise in real long-term interest rates, and by the reversal of the yield curve. With respect to the external position, Canada clearly started from a less favourable situation than the United States, but the principal difference is probably the development of the exchange rate, as the United States experienced a depreciation of both the nominal and the real exchange rate, whereas for Canada a small nominal depreciation was accompanied by a large real appreciation.

Compared with the three European countries discussed in this section, Canada had higher output and employment growth, so that the reason for the higher sacrifice ratio mainly lies in the smaller reduction in price and wage inflation. Fiscal policy appears to have been less restrictive than in Denmark and Ireland, but slightly tighter than in France, and this outcome more or less matches differences in the initial positions, as Canada was better placed than Denmark and Ireland but worse off than France. Judging by the developments in nominal and real interest rates, monetary policy was more restrictive in France and Ireland than in Canada, whereas Denmark appears to have pursued a more accommodating policy, though credibility effects may explain the fall in nominal interest rates (see below). Exchange rate developments were clearly less favourable than in Ireland and France, whereas Canada did somewhat better than Denmark.

Precise conclusions would obviously require a much more rigorous analysis, but – very tentatively – these broad comparisons point to two possible sources of problems for Canada: (i) a labour market development which was clearly less favourable than in the United States, possibly because Canada did not benefit from the same degree of nominal and real wage restraint, or had a poor productivity performance; and (ii) certain problems in finding a satisfactory policy mix, as restrictive monetary policies and higher nominal and real interest rates caused a worsening of the fiscal position, even though the stance of fiscal policy may have been more restrictive than in the United States.

As a further step, the dates of some important policy events may be recalled:

  • in November 1982 Canada abandoned its monetary aggregate target (M1) because financial market innovations and large portfolio shifts had made the money demand function unstable. Since Canada also had a flexible exchange rate (since 1970) and incomes policies had been phased out in 1978, this decision essentially meant that there was no clearly-defined medium-term anchor for inflation expectations;
  • a medium-term strategy (the Agenda for Economic Renewal) was adopted in 1984 stressing two policy priorities: reducing inflation, and restoring fiscal stability by consolidating the budget and improving the productivity performance through deregulation and the removal of market distortions;
  • the 1988 Hanson Lecture by the Governor of the Bank of Canada, which emphasised the primacy of price stability as the focus of monetary policy,[53] It can also be seen as the first step towards the Bank of Canada – Finance Department agreement of February 1991, which set out a disinflation path whereby the CPI increase was to be reduced to 3 per cent by end-1992 and further to 2 per cent by end-1995.[54]

The relative inflation and output performance of Canada needs further analysis. During 1982–90 output and employment expanded at higher rates than in most other OECD countries, and by 1989 the rate of unemployment had fallen to the pre-1982 level of 7.5 per cent, a performance that was surpassed only by the United States and Sweden. Influenced by the exceptionally severe 1982–83 recession the rate of inflation also fell substantially, but then remained more or less stable at around 4 per cent during the following years, until it started to rise again in 1989–90. One reason why Canada, unlike most other countries, did not achieve a further deceleration is, of course, that as an oil producer it did not benefit from the terms-of-trade gains of 1985–86. More importantly, however, this result also appears to reflect some fundamental weaknesses on the supply side.

This brings us to another point, viz. the extent to which the microeconomic reforms foreseen in 1984 have helped to lower inflation or reduce the transitory costs. At first glance these policies seem to have met with some success, as there was a noticeable improvement in the growth of labour productivity in both manufacturing (see Table 6) and the aggregate economy. Yet on further analysis this apparent improvement looks rather like a setback:

  • because of surging business fixed investment, Canada experienced an exceptionally large rise in the capital/output ratio,[55] and the growth of total factor productivity, at only 0.5 per cent annually, was actually among the lowest in the OECD area;
  • compared with developments in other countries, the productivity performance and the related growth of unit labour costs was also unfavourable, and the real effective exchange rate appreciated by over 10 per cent during the 1980s.[56]

Supply-side problems are also evident in labour market developments, despite the impressive employment performance. Thus the unemployment gap relative to the United States which widened in 1982 and rose further during the 1980s is in sharp contrast to the previous post-War years, when the two rates had moved in parallel. There are several reasons why the Canadian rate has traditionally been higher than the US rate.[57] But despite intensive research efforts, it remains unclear why the gap should have widened during the 1980s, since most of the factors explaining the higher level became less important (see, for instance, Ashenfelter and Card (1986) and Coe (1990)).

Empirical studies (see Ford and Rose (1989), Rose (1988) and Coe (1990)) have also found that the NAIRU at 7.5 – 8 per cent in 1989 is higher than in the United States, and higher than in the 1970s. Again, however, it is not entirely clear why the NAIRU has remained so high, since most of the factors explaining the rise during the 1970s have been relatively stable in the 1980s, and there is little evidence of any real wage gaps (see Fortin (1989)).[58] One unresolved issue in this respect is the degree of hysteresis in Canada. In the wage equations discussed in the Appendix, no hysteresis effect was identified, and this was also the result obtained by Fortin (1989). However, this may be an extreme view, as alternative estimates of the trade-off relation have produced hysteresis parameters ranging from 0.68 to 100.[59] Moreover, in estimating an unemployment equation for Canada, Coe (1990) identified a significant hysteresis effect which is transmitted through the unemployment, benefit insurance system.[60]

On balance it thus appears that despite some uncertainty with respect to the precise causes, there is evidence that policy measures in the area of unemployment insurance and disappointing supply-side developments have tended to raise actual and long-run unemployment, and thus made the disinflationary policies look more costly.

These, however, are not the only effects since there are clear signs that monetary and fiscal policies interacted in such a way as to raise the transitory costs and make the announced targets less credible to the public. As already mentioned, the Bank of Canada abandoned the M1 target in 1982 and did not adopt a fixed exchange rate as an alternative nominal anchor. Instead, it preferred to maintain a flexible exchange rate as a means of protecting the domestic economy against unfavourable external shocks (notably terms-of-trade shocks) and otherwise relied on a policy of preventing exchange rate depreciations from pushing up the domestic rate of inflation.[61] At the same time, the aim of fiscal policy was to reduce the general government borrowing requirement and reverse the rise in the net public debt to GDP ratio. However, reflecting the effects of external events and a lack of firmness in implementing the necessary fiscal measures, a progressively rising share of the anti-inflation policy burden fell on monetary policy, and the overall outcome appears to have exacerbated the costs of reducing inflation:

  • during the first half of the 1980s the Bank of Canada was forced to increase short- and long-term interest rates and maintain a large differential against US rates to resist strong downward pressures on the exchange rate; [62]
  • although the net public debt to GDP ratio was relatively low by international standards in 1982, and a 1 – 2 point rise in interest rates could easily have been absorbed without adversely affecting sustainability, cyclical effects together with lax implementation increased the primary deficit. By 1986 the debt ratio at 37 per cent was well above the average for the Group of Seven countries. In addition, the high level of real interest rates worsened the already fragile financial position of the corporate sector, and probably delayed the recovery of business fixed investment;
  • later in the decade, when exchange rate pressures were reversed and interest rates declined, a relatively stable nominal effective exchange rate combined with a poor labour cost performance led to a marked appreciation of the real exchange rate. In addition, since fiscal policy was not tightened[63] during the recovery, short- and long-term interest rates were raised again, pushing up the public debt to GDP ratio as well as the nominal effective exchange rate.

As the combined result of the absence of a firm fiscal policy and high nominal and real interest rates, the debt to GDP ratio at the end of the decade had risen to over 40 per cent (or twice the level recorded in 1982), and at this level the prospects of a smooth and relatively costless reduction of inflation are not very promising. On the one hand, if a monetary tightening is to be backed up by a restrictive fiscal policy to keep the debt to GDP ratio stable and maintain credibility, the restrictive impact on the economy would be quite large, as each 1 point rise in the nominal rate requires a fiscal tightening equivalent to 0.4 per cent of GDP. On the other hand, if the fiscal authorities do not wish to reinforce the restrictive impact of monetary policy, the deficit and the debt ratio are bound to rise, which could adversely affect the credibility of monetary policy.

All in all, the sacrifice ratios found for Canada seem to have been the result of mainly two factors: (i) fundamental weaknesses on the supply side, as seen in weak productivity growth and a persistently high rate of unemployment; and (ii) inconsistent fiscal and monetary policies, with lax fiscal policies not only shifting more and more of the burden of stabilisation to the monetary authorities, but probably also damaging the credibility of monetary actions.

(c) Denmark

Denmark's inflation performance during the 1980s has in several respects been impressive:

  • the rate of consumer price inflation declined from a peak of 12 per cent in 1980 to only 2 – 2½ per cent in 1990–91, and most forecasts suggest that it has stabilised at this lower rate;
  • long-term nominal interest rates have fallen in step with the rate of inflation, and the earlier differential against German rates has virtually disappeared. At the same time, the yield curve has flattened and the ex post real rate has been stable;
  • the costs of disinflation have been low relative to those of most other countries (particularly those with currencies more closely linked to the Deutschemark) and in the overall ranking Denmark is in third place, preceded only by Japan and Sweden.

Yet the adjustment has not been entirely smooth, and the overall performance and the policy lessons for other countries can best be evaluated by distinguishing three separate phases.

Initial phase: 1982–86. A determined anti-inflation strategy and a clear break with earlier and more accommodating policies was signalled when a new minority Government came into power in the autumn of 1982.[64] Shortly after taking office, the new Government introduced a broadly based programme including: (i) a firm commitment to holding the exchange rate stable within the ERM and eliminating the current account deficit by 1987; (ii) a more restrictive fiscal policy with cuts in expenditure and a target of balancing the budget by 1990; and (iii) a wide range of incomes policy measures including suspension of wage indexation (finally abandoned in 1986), wage guidelines limiting nominal wage increases to 4 per cent a year for the next two years (further lowered to 1.5 – 2 per cent over the following two years),[65] ceilings on profit margins, a freeze on dividends and a rise in the corporate tax rate from 40 to 50 per cent to neutralise the revenue effects of lower payroll taxes.

There is little doubt that the programme was effective in reducing inflation, and that the three-pronged approach helped to strengthen credibility.[66] Short- and long-term interest rates fell by 8 percentage points in less than six months, actual wage increases were close to the guidelines and the fiscal imbalance was reduced from 9½ to 2 per cent of GDP by 1985, with sustainability already achieved in 1984. The stance of monetary policy and the developments in monetary aggregates are more difficult to interpret, as the authorities were clearly faced with a re-entry problem. Reflecting the fall in interest rates and in the rate of inflation, and reinforced by the real gains on the stock of bonds held by the household sector, money demand rose substantially. With the authorities adopting an accommodating stance, the growth of both M1 and M2 accelerated to over 25 per cent in 1983 and remained near 20 per cent over the next two years. At the same time, the rise in money demand probably reflected a fall in inflation expectations, and may thus be interpreted as a sign that the policies pursued enjoyed a considerable degree of credibility.

Intermediate phase: 1986–87. Despite the more restrictive fiscal policy and the fall in real earnings, aggregate output growth rose to an annual rate of 4 per cent during 1983–86, compared with only 2½ per cent for EC countries on average. To some extent this development was an unexpected consequence of the success of the policies pursued, since a major factor was a surge in private consumption and business fixed investment induced by real wealth gains and a restoration of business confidence. A further consequence was a marked deterioration in private sector net saving which more than offset the improvement in the fiscal balance, so that the external deficit, instead of being eliminated as planned by the Government, rose to 5½ per cent of GDP. This renewed doubts about the exchange rate commitment and, with the incomes policies becoming progressively less effective, and actual wage increases exceeding the guidelines by a widening margin,[67] the process of disinflation had obviously come to a halt. This became even more evident in early 1987, when the Government stepped back from the general wage round and the unions and the employers agreed on a settlement which raised labour costs in manufacturing by 9½ per cent, and more than eliminated the gains in international competitiveness recorded since 1982.[68]

Recent phase: 1987–91. As the authorities recognised that the stabilisation programme was moving off course, and that incomes policies which are not supported by sufficiently restrictive fiscal and monetary policies are bound to fail, the overall policy stance was progressively tightened. In the course of 1986 no less than four major fiscal packages were adopted by Parliament, including, inter alia, an income tax reform, a gradual reduction in tax deductions for mortgage interest payments, substantial increases in indirect taxes and a 20 per cent surcharge on interest payments. In addition, domestic credit growth was sharply curtailed especially for financing current consumption and residential investment. Initially the various measures had very little influence, but the cumulative effect of the successive packages eventually brought domestic demand growth to a halt. As a result, during 1987–91, when most other EC countries entered a strong recovery phase, real GDP growth in Denmark fell to an annual rate of only 1 per cent. Moreover, the earlier growth in employment (especially strong in the private sector) was sharply reversed, and by 1991 the rate of unemployment had returned to the level recorded just prior to the 1982 programme. However, the rate of inflation responded to the growing slack and in both 1990 and 1991 Denmark's inflation performance was even better than that of Japan.

In retrospect it appears that inflation was not brought effectively under control until the authorities adopted more traditional and restrictive fiscal and monetary policies, and that the broadly-based approach adopted initially provided only temporary relief and no permanent gain. The gradual worsening of the output-based sacrifice ratio as the time period is extended confirms this view and an unemployment-based measure calculated for the 1987–91 period produces a ratio which is about twice that shown in Table 5. On the other hand, the initial package was accompanied by structural changes that have probably helped to make the economy less prone to inflation in the longer run. Thus the share of public sector employment was reduced, the growth of the real capital stock accelerated (though admittedly from a low level), the gross profit share rose substantially, which strengthened business confidence, and a shift of resources from the non-tradable to the tradable sectors got slowly under way. Moreover, a major element of earlier real wage resistance – the indexation scheme – was removed, without major disturbances in the labour market. In this context it should also be noted that the Government's target of eliminating the external imbalance together with mounting net interest payments to abroad, reduced disposable income and domestic demand relative to output, and thus made a non-inflationary solution to the distributional issue more difficult.[69]

To what extent membership of the ERM helped to reduce inflation expectations and provided a net welfare gain is more difficult to say. Econometric studies of the inflation performance of countries inside and outside the ERM are inconclusive (see Giavazzi and Pagano (1988)), and the real appreciation of the Danish krone (see Table 4) increased the transitory costs of lowering inflation. On the other hand (see Andersen and Risager (1988) and Christensen (1988)), the firmer exchange rate commitment was clearly instrumental in reducing long-term interest rates, which in turn stimulated business fixed investment and faster growth of the capital stock.

On one point the Danish experience provides a particularly clear policy message. As is well known, a policy based on a fixed nominal exchange rate commitment places the burden of relative price adjustments on wages. One condition for a successful policy is therefore that real and relative wages are sufficiently flexible. However, a further condition for achieving a sustainable reduction in inflation and for keeping the adjustment costs to an acceptable level is that national saving increases in step with investment. In other words, if relative price and wage adjustments are not backed up by a corresponding change in the saving-investment balance, the disinflation process will come to a halt. This was the case in Denmark during the intermediate phase, when the external deficit rose to an unsustainably high level and the Government was forced to tighten fiscal and monetary policies. During the most recent phase Denmark has finally achieved the target of balancing the external current account (after twenty-six consecutive years of deficit) and the inflation rate is one of the lowest worldwide. It remains, however, to be seen whether the improvement is sustainable. The aggregate private saving rate has increased impressively during the last three to four years, but the surplus on the external account was also to a large extent the result of low investment and the boom in exports to Germany following unification. Moreover, there is as yet no firm evidence that the position of the Phillips curve has shifted, and that a lower rate of unemployment will not lead to stronger wage claims.

(d) Ireland

Ireland's disinflation process has been spectacular. In the course of just ten years the rate of consumer price inflation has fallen by over 16 percentage points, making Ireland a member of the group of low-inflation countries. However, the costs of disinflation have also been exceptionally high, with total domestic demand falling throughout 1981–88 and the unemployment rate doubling. Hence, judged by the sacrifice ratios shown in Table 5, Ireland obtains a very poor rating, well below the ranking for Denmark and only marginally better than Canada.

Considering that Ireland abandoned the currency link with sterling to join the ERM at its inception in early 1979, this performance raises two key questions: (i) why does Ireland not appear to have obtained any credibility gains by linking its currency to that of a low-inflation country? and (ii) were credibility gains present but overshadowed by the effects of other factors, in particular the initial conditions and the wage and price adjustment process?

The credibility issue has been discussed in three recent articles. Dornbusch (1989) finds no evidence of any gain, mainly because of the unsustainability of fiscal policy; Kremers (1990), relying on econometric estimates, finds evidence of some gain; and Dornbusch and Fischer (1991) come to the conclusion that ‘there was no obvious credibility bonus for the government’ (p. 58).

In trying to find an answer to the second question it is useful to look at the disinflation process over two phases, as the policy instruments applied and the sources of adverse influences changed during the 1980s.

Initial phase: 1981–86. It seems fairly evident that Ireland did not immediately benefit from joining the ERM, as inflation continued to accelerate and nominal interest rates remained very high. Moreover, during the first two years every realignment of ERM currencies involved a devaluation of the Irish pound. However, with the coming into power of a new coalition Government in 1982, a more determined anti-inflation policy was launched, including a fiscal policy aimed at reducing a large imbalance (see Table 3) and a monetary policy with the principal target of maintaining a stable central rate for the Irish pound within the ERM. This new and more restrictive policy stance was certainly successful in lowering the rate of inflation, which had fallen to only 4½ per cent by 1986, and also in improving the external position, the current account deficit falling from 15 to only 3 per cent of GDP between 1981 and 1986.[70] This outcome, however, was mainly the result of domestic demand restraint, and there were several signs that some of the fundamental problems had not been addressed – let alone solved – by the new policies.

Firstly, the degree of real wage rigidity is higher than for most of the other countries included in the sample,[71] and for mainly four reasons:

  1. young and skilled workers prefer to emigrate in periods of labour market slack, so that a large proportion of the unemployed are unskilled workers who have been without a job for more than a year, and exert relatively little downward pressure on wages;
  2. although the share of public sector employment in total employment at 16 per cent is below the OECD average, public sector wages have been a major source of real wage rigidity, as the public sector has tended to be a wage leader, and public sector wage trends are rather insensitive to labour market conditions; [72]
  3. unemployment benefits have – until recently – been tax-exempt and rather generous, with replacement ratios of nearly 80 per cent providing little incentive for intensive job search;
  4. the ratio of total labour costs to net real earnings (the ‘wedge’) is high and, like in France, widened during the first half of the 1980s as labour costs per employee rose by 20 per cent while real earnings declined by 10 per cent.

Secondly, despite the restrictive stance of fiscal policy, the general government deficit fell by less than 2 per cent of GDP and the gross public debt to GDP ratio rose to 135 per cent, which was larger than in any other industrialised country at that time. Moreover, monetary financing of the public sector deficit remained high, constituting two-thirds of domestic credit expansion in 1986, and increasing M3 by over 13 per cent. It thus appears that despite the restrictive measures taken, fiscal policy was not sufficiently tight to lessen concern about the sustainability of public finances and the exchange rate commitment.[73] The sharp rise in the ex post real long-term interest rate (from −1½ to over 6 per cent) may be taken as a further sign that disinflation was mainly the result of demand restraint and did not benefit from credibility gains.

Thirdly, a further setback to hopes of reducing inflation and inflation expectations through a shift in the exchange rate regime was the actual development of the nominal effective exchange rate. While the Irish pound was kept more or less stable around its central ERM rate, the effective nominal rate depreciated by over 16 per cent between 1980 and 1985 mainly because of the appreciations of sterling and the US dollar.[74] Consequently, while the Danish authorities could underline their exchange rate commitment by resisting devaluation pressures in 1982, the Irish authorities were faced with an exogenous event of the opposite sign. Indeed, this episode, together with developments later in the 1980s when the authorities were forced to devalue against ERM currencies because of a sharp depreciation of sterling, raises certain doubts about the usefulness of an exchange rate commitment which excludes the currencies of two of the country's major trading partners.

Recent phase: 1986–91. A decisive change in the policy priorities and the range of instruments applied occurred when a new Government took office in February 1987, and shortly afterwards presented a revised and much more restrictive budget and a three-year Programme for National Recovery (PNR). While the exchange rate commitment remained the cornerstone of the anti-inflation approach, it was now complemented by other measures which helped to remove some of the fundamental weaknesses of the earlier period:

  • fiscal policy was firmly aimed at reducing the deficit and reversing the growth of the public debt to GDP ratio, and the results were quick and impressive: the primary balance was in surplus by 1988, by 1990 the general government deficit had been reduced to only 2½ per cent of GDP, and last year the debt ratio was more than 20 percentage points below the earlier peak;
  • a principal element of the PNR was an agreement with the trade unions to limit wage increases in return for lower taxes. Thus, following an unsuccessful attempt in 1982, incomes policies became an important instrument in the approach to reducing inflation.[75] As a result, despite some improvement in labour market conditions, the rate of nominal wage increases fell to 4 per cent in 1990, compared with 7 per cent in 1986.

The new policies were also successful in other respects: the external balance moved into surplus, real output growth accelerated sharply and the rate of unemployment started to fall, though remaining well above the European average. Some of the achievements were the result of favourable external developments (including booming exports to the United Kingdom) rather than of the policies pursued, but there were also signs that the credibility benefits which had eluded the authorities during the earlier phase were finally starting to appear. For instance, long-term bond rates declined precipitously as the new programme took effect, and the differential against German rates almost disappeared.

When evaluating the entire disinflation process in Ireland it is natural to compare it with the Danish experience. Both countries joined the ERM at its inception and relied on exchange rate commitments rather than monetary aggregate targets as nominal anchors. This meant that neither country was able to use expenditure-switching policies to ease the transitory costs and the competitive position could only be improved through wage restraint.[76] In Denmark the process of disinflation started with a broadly-based approach that produced quick results at rather low costs, but which progressively lost its effectiveness and was replaced in the end by measures with demand restraint as their principal element. Ireland, by contrast, initially relied almost exclusively on demand restraint, and its success in quickly reducing the rate of inflation was equal to that of Denmark, but the costs were extremely high. During the more recent phase Ireland has adopted a broader approach and, in contrast to Denmark, where the sacrifice ratios have tended to rise over time, the costs have been substantially lower than during the earlier phase. In evaluating the Irish experience it should, of course, also be noted that the initial conditions were much less favourable than in Denmark, and that doubts about the consistency and sustainability of the programme prevented credibility benefits. In particular, the high levels of public and external debt were critical factors in the initial phase, and it was only as the primary public sector balance and the current external account moved into surplus that the programme appears to have been accepted with confidence by the general public.

7. Summary and Conclusions

Although some partial and tentative conclusions have already appeared earlier in this paper, it is not an easy task to draw the various pieces together and derive the relevant policy implications. A short review of the experience of each country could easily develop into another paper, or uncover shortcomings in the earlier sections. Instead, I have attempted to derive a few conditions for successful anti-inflation policies using success in reducing inflation, as well as in avoiding output and employment losses, as the main criteria.

Starting with the propositions of the Canadian debate, there is some, albeit weak, support for the view that a central bank should stop short of full price stability. A high initial rate of inflation does seem to lower the unemployment-based sacrifice ratio and the non-linear trade-off worked well for the estimates given in the Appendix. However, the initial inflation rate has little or no impact on the output-based sacrifice ratios, and the non-linear form was chosen in the light of recent international practice and not dictated by econometric tests.

On the other hand, there appear to be no convincing empirical reasons for adopting a gradualist approach. Although unemployment rates seem to follow a random walk, this feature is not – except for the United Kingdom – the result of hysteresis in the wage formation process. Moreover, even though the review does not include many episodes of shock treatment, those that are included appear to have reduced inflation at rather low costs: the Danish broad-based approach in the early 1980s was clearly successful, and the surprisingly low sacrifice ratio found for the United States could be the result of the 1979 change in the conduct of monetary policy, and the subsequent surge of the US dollar. Finally, going back to the 1970s, the dramatic tightening of fiscal and monetary policies in Japan following the first oil price shock produced quick inflation gains at low costs, and left a permanent impact on the behaviour of trade unions and enterprises. The Thatcher stabilisation programme of 1979 appears to be an exception, as the costs in terms of unemployment were quite high. However, this merely underlines the fact that an essential condition for a successful shock treatment is that labour and product markets respond flexibly.[77] In the United Kingdom this was the case for output markets, but not for the labour market.

There is also evidence to suggest that countries without a medium-term nominal anchor for inflation have done less well than countries using either monetary aggregate or exchange rate targets. The rather poor performance of Canada has already been discussed, and the high sacrifice ratios experienced by Australia, Norway and (until recently) New Zealand could also in part reflect the absence of a nominal anchor. On the other hand, the country experiences reviewed do not provide a clear answer to the question of whether monetary aggregate or exchange rate targets are more effective as nominal anchors. The United States and Japan, which have relied on monetary aggregate targets, have incurred rather low transitory costs while Germany, which has experienced fewer distortions in the targeted aggregate (M3) than most other countries, obtains a very poor ranking.

In this context it is important to recall the rather high sacrifice ratios found for countries that have linked their currencies closely to that of a country with a well-established reputation for persistently pursuing anti-inflation policies.[78] The most convincing explanation might be that such a linkage leads to a convergence of inflation rates towards that of the key currency country, through a convergence of monetary policies. However, the costs will not be lower than for countries outside the currency arrangement, unless the structure and institutions of the wage and price formation process also converge. In fact, the costs could even rise if the natural unemployment rate of the key currency country is subject to a trend rise.[79]

Incomes policies could provide some temporary relief in this respect, but cannot be relied upon as the only anti-inflation policy instrument, and they also tend to create distortions if maintained for too long. This risk is even greater for wage and price controls, though some countries have achieved some success in using such controls as a short cut to reducing expectations – or the degree of inertia – during the initial phase of stabilisation programmes. The Austrian experience is the most puzzling with respect to the usefulness of long-term incomes policies, as it is hard to find a policy which is better adapted to the exchange rate target. Moreover, it has generated a wage adjustment process which is fully compatible with this target. And yet Austria's sacrifice ratio is one of the highest among the sixteen countries reviewed.

The precise role of fiscal policies is also difficult to evaluate. Large budget deficits do seem to have damaged the credibility of monetary policies in Canada and Ireland, and the initial budget positions were also found to increase the sacrifice ratios. On the other hand, Italy has done comparatively well, and while the US authorities have attracted much national and international criticism for failing to reduce the federal deficit, there is little evidence that this has made reducing inflation more costly.

A major uncertainty concerns the reliability of the sacrifice ratios presented in this paper, and the extent to which policy implications can be drawn from the ranking suggested by these ratios. It is well known that ratios derived from actual developments are influenced by factors unrelated to the policies pursued, and a particularly disturbing result in this respect is that sacrifice ratios based on unemployment give a more pessimistic picture than ratios based on output losses. The random walk nature of unemployment could suggest that equilibrium unemployment has been subject to a trend rise, which biases the estimated transitory costs. It is also possible that rigidities in output markets are much smaller than in labour markets, thus forcing up unemployment relative to potential output. However, further exploration of this possibility requires a much more rigorous analysis, since this paper has only highlighted large deviations between output and labour market rigidities in individual countries.

Another open question is the influence of credibility, as we have not used any indicators nor attempted to quantify references to the presence or absence of credibility effects. There is no need to stress that this is an area in need of further research, because an impressive amount of theoretical and empirical work has already been done. Consequently, we shall merely add a few observations on the type of work that might be particularly important for analyses of countries' experience with disinflation:

  • given the role of monetary policies, empirical measures derived from the relationship between short- and long-term interest rates could provide useful additional information. Takeda' s (1992) preliminary estimates look promising in this respect, though one problem, which is also found in other empirical studies of this nature, is that the parameter capturing the anti-inflation stance of the monetary authorities cannot be separated from structural parameters of the wage and price mechanism;
  • although risking a further blurring of the distinction between reputation and credibility, it is important to recognise that economic agents have long memories with respect to lax policies, and do not easily forgive slippages in an otherwise impeccable anti-inflationary stance;
  • as stressed by Dornbusch (1991), evaluating the presence (or absence) of credibility on the basis of past experience easily leads to circular reasoning. In other words, future attempts to incorporate the credibility issue should pay more attention to the ex ante conditions, and attempt to assess the probability that a given programme will be credible;
  • only few credibility models consider the timing of anti-inflation measures. There is some evidence that stabilisation programmes undertaken in periods of severe economic stress enjoy a higher degree of credibility than programmes proposed when conditions are less gloomy. Hence delaying anti-inflation measures may reduce the transitory costs but the optimum delay is as yet not well known;
  • one short cut to directly influencing inflation expectations and creating favourable ex ante conditions for a credible programme might be to pre-announce a time path for the inflation target. Canada and New Zealand have recently embarked on such a course, but it is still too early to evaluate its effects on the actual rate of inflation and the costs of reaching the target.

It goes without saying that inflation cannot be reduced without an appropriately tight monetary policy. What is perhaps less obvious is that monetary policy alone is not enough if the transitory costs are to be kept low. The best that can be achieved through monetary policies is a reduction in nominal income growth, but how the slowdown is ‘split’ between lower inflation and lower real growth (and thus higher costs) depends on other factors and policies. It is mainly because of these other factors that the experience with disinflation has differed so widely between the sixteen countries reviewed in this paper, as I have found only a few instances where monetary policies have not been sufficiently restrictive. This also suggests that future attempts to reduce the transitory costs of disinflation should probably focus on these other factors (notably fiscal policy and rigidities in labour and product markets) rather than on the role of monetary policies.


The views expressed in this paper are strictly those of the author and are not necessarily shared by the BIS. I am indebted to H. Bernard, J. Bisignano, B. Cozier, O. Risager and S. Roger for comments on an earlier draft of this paper. I also wish to thank my two discussants at the Reserve Bank Conference, as well as those commenting on the paper during the general discussion. [*]

These propositions are not shared by all participants, including the Bank of Canada (see Freedman (1991)). [1]

A comprehensive theoretical review of the various issues can be found in Buiter and Miller (1986). [2]

See Coe et al. (1988), who estimate the contribution of monetary policy by a counter-factual simulation where the nominal rate of interest is kept at the pre-1980 level. [3]

This can also be seen from a comparison of the GDP deflator and the private consumption deflator. While the latter fell on average by 8.8 percentage points between 1980 and 1986, the former fell by only 6.1 points, and the trough occurred one year later. [4]

In Figures 1 and 3 nominal wage growth has been measured by hourly compensation in manufacturing as published by the US Department of Labor, whereas in Table 1 and Figure 2 and in the equations reported in the Appendix, nominal wages are derived from the OECD National Accounts. The former indicator is frequently used for international comparisons, but since manufacturing employment on average accounts for only 20 percent of total employment, it is not representative of aggregate trends. [5]

For the OECD private consumption deflator the year-to-year variability fell from 3.6 to 1.0 between 1980–85 and 1985–1990, or, measured by the coefficient of variation, from 0.6 to 0.3. [6]

Inflation also declined steeply in the early 1950s. However, this disinflation process started from a lower initial rate (8.5–9 per cent in 1951) and was completed in just two years. [7]

Another, though related, problem is that the discount factor applied by policy makers to future gains from lower inflation is likely to be very high. [8]

Some countries have managed to reduce inflation without major output losses, but most of these cases refer to episodes of hyperinflation, which are typically characterised by a virtual absence of backward-looking expectations, and a high degree of credibility of the policy measures taken (see Vegh (1991) and Sargent (1982)). When the initial rates of inflation are similar to those shown in Table 1 for 1979–81, inflation has rarely been reduced without major output losses (Gordon (1982)). [9]

On this point see also Ball (1990, 1991 and 1992), who argues that both lack of credibility and lags in price and wage adjustment are necessary conditions for the transitory costs. Ball in fact goes one step further, by proving that a credible anti-inflation policy will be accompanied by excess demand pressures because of a substantial rise in the real money stock. A related problem has been discussed by other analysts (see for instance Freedman (1989), Scarth (1990) and Chadha et al. (1991)), but is usually referred to as the ‘re-entry problem’. [10]

For a comprehensive review of the theoretical and empirical literature see Agenor and Taylor (1991) and Blackburn and Christensen (1989). [11]

In theory a distinction should be made between reputation and credibility, with the former referring to the behaviour of the authorities and the latter to the perceived effects of the policies announced. However, in practice it is difficult to maintain this distinction. [12]

The precise definitions and measurements of real and nominal rigidities differ between authors. Taylor (1989), for instance, relates differences in output fluctuations between the United States and Japan to ‘nominal’ rigidities, but we interpret his results as reflecting ‘real’ rigidities as defined above. [13]

Table 2 also includes an overall measure of real wage or unemployment persistence, which is further explained in the Appendix. [14]

See the Appendix and the various simulations reported in Chadha et al. (1991). Note that while a high degree of real and nominal flexibility is helpful in reducing the transitory costs of anti-inflation policies, their general usefulness very much depends on the kinds of shocks to which a country is exposed. For instance, a low degree of nominal flexibility helps to reduce the effects of adverse supply shocks, and there are also cases where a low sensitivity to economic slack is desirable. For further discussion see Andersen (1989). [15]

Ideally unemployment should have been measured as deviations from the natural rate and not as actual rates. Moreover, recent changes in the rate of inflation may provide additional information about the initial conditions. [16]

Large regional differences in inflation and unemployment may be another source of non-linearities. [17]

See Giavazzi and Spaventa (1989), who argue that a favourable profit situation – partly the result of tax measures – was one reason for the relatively low sacrifice ratio found for Italy (see below). [18]

See in particular Fischer (1986b) and Bruno (1991). [19]

A real appreciation will, of course, strengthen an anti-inflation policy stance, but if the nominal anchor is not reflected in the wage and price adjustment process it will eventually become unsustainable. [20]

On the other hand, except for France in 1982, Belgium in 1982–83 and New Zealand in 1983, countries have generally not resorted to price and wage controls as a means of reducing inflation expectations. [21]

For further discussion see Sachs (1985). [22]

Large regional differences may be another argument for adopting a gradual approach; see Freedman (1991). [23]

According to Taylor (1983), the rate of inflation would only fall from 10 to 8¾ per cent during the first two years of a programme of disinflation geared to a wage and price adjustment process with staggered three-year contracts and forward-looking expectations. [24]

As discussed in Section 5, the growth of Ml and M2 in Denmark jumped to over 25 per cent shortly after the successful implementation of disinflation policies and remained exceptionally high over the next two years. [25]

See Chadha et al. (1991), especially p. 19. [26]

By convention, sacrifice ratios for periods of falling inflation are given as positive figures when the cumulative change in unemployment is positive or the cumulative deviation between actual and potential output growth is negative. This rule has also been followed throughout this paper. [27]

Selody (1990) reports sacrifice ratios for Canada ranging from 4.7 for calculations based on actual developments during the 1981–82 recession (Howitt (1990)), to 2.0 for estimates derived from a short-run aggregate supply curve (Cozier and Wilkinson (1990)), with simulations on the RDXF macro-model yielding an intermediate ratio of 3.4. [28]

To overcome this shortcoming, the slope measures may be supplemented with parameter stability tests, but usually parameter shifts are difficult to identify (see for instance Blanchard (1984) and Englander and Los (1983)). [29]

The sacrifice ratio implied by Tobin's path is by no means an outlier relative to other assessments. It is very close to the ‘rule-of-thumb’ proposed by Kahn and Weiner (1990) and, applying an Okun-coefficient of 2.5, it is also close to the ratios reported by Gordon and King (1982), but at the lower end of Okun's (1978) range of 6–18. Our estimate of 0.85 is close to Sachs' (1985) output-based ratio of 2.5, while Dornbusch (1989), relying on 1981–88 changes in US unemployment, arrives at a negative figure. [30]

In particular, sacrifice ratios calculated in this way may be severely distorted by supply shocks. [31]

This discrepancy becomes particularly evident when calculating the Okun coefficients corresponding to the sacrifice ratios. [32]

See Jarrett and Selody (1982), who, on the basis of data for Canada, find that lower inflation tends to increase productivity gains, and thus potential output growth. In a more recent study Cozier and Selody (1992) conclude that lower inflation mainly affects the long-run level of potential output, while the effect on output growth is much smaller and only transitory. Developments in inflation and productivity growth in manufacturing are shown in Table 6. [33]

The coefficient on the dummy variable would suggest that the sacrifice ratio for the United States is understated by 2.85 percentage points, and that for Australia overstated to a similar extent. The use of a dummy variable is, of course, very ad hoc, but it leaves the coefficients of other variables virtually unchanged. [34]

The high sacrifice ratio does not result from using a price deflator rather than a wage deflator. In fact, when the sacrifice ratio is calculated for the wage changes shown in Table 1, it rises to more than 10, compared with 1 for the United States and 3.3 for EC countries on average. [35]

When a dummy variable with 1 for ERM members (excluding Germany but including Austria) and 0 for non-members was added to the equations in Table 8, a negative coefficient was obtained, but it was never significant. [36]

While van Poeck includes all ERM countries, Artis and Ormerod confine their sample to Germany, France, Italy, Belgium and the Netherlands. [37]

Artis and Ormerod also attempt to evaluate the length of the period required for transition to the new regime and find that for France it took about four years (1982–86), whereas for the other countries the transition was much faster. [38]

The absence of converging wage formation processes is also evident in Artis and Nachane (1990), who find that price and wage changes in other ERM countries are not cointegrated with those for Germany, either before or after 1979. [39]

In this respect France no doubt benefited from the fall in import prices which, according to Cling and Meunier (1986), may account for more than half of the deceleration between 1981 and 1985, and largely offset the effect of higher profit margins due to deregulation of industrial prices (see Table 9). [40]

Because of the deregulation of financial markets and shifts in portfolio preferences in response to new instruments, actual changes in monetary aggregates have to be interpreted with caution. However, the reduction in the growth target for M2 from 10 per cent in 1983 to 5½–6½ per cent in 1984 points to a marked tightening of monetary policy which was more or less maintained throughout the period. [41]

Private sector service prices were decontrolled in early 1987 and this led to a one-time price jump in 1987–88. However, the restoration of severely distorted profit margins was largely complete by 1989. [42]

Except for the minimum wage (the SMIC) formal indexation did not exist in France, but de facto indexation was widespread and usually involved price compensation of nearly 100 per cent. [43]

The new indexation scheme included a catch-up clause providing an ex post adjustment of wages when actual price inflation exceeded the official forecast. However, because the new scheme also increased the speed of price adjustments, the effect of the catch-up clause has been only marginal (see Cling and Meunier (1986)). [44]

According to estimates reported in OECD (1990), the elasticity of nominal wage changes with respect to current price inflation increased significantly after 1982, whereas other parameters of the wage equation were largely stable. [45]

The rather long transitional period found by Artis and Ormerod should be recalled in this respect (see note 38, above). [46]

Except for 1990 (when it fell to just below 9 per cent), the rate of unemployment has been near 10 per cent since 1984, and at the current level a further 1 point rise would – with a one-year lag – lower wage inflation by only 0.15 percentage points. [47]

In 1989 employers' social security contributions equalled 40 per cent of total wages, and contributions by employees (including the self-employed and a new ‘contribution sociale de solidarité’) added another 25 percent, pushing the share of social security contributions in total tax revenue to 46 per cent, higher than in any other OECD country. [48]

Almost 140 per cent between 1970 and 1989 according to OECD. Since 1985 the average social security contribution rate has been more or less stable, but alternative revenue sources had to be sought to finance growing expenditure and a widening deficit in the social security sector. [49]

Cotis and Loufir (1990), on the other hand, find that changes in social security contributions are fully absorbed by wage earners. However, their results look rather implausible, since they find the same tax shift parameter for both gross and net wages and employees appear to lower wage claims in response to increases in both their own contributions and those imposed on employers. Our own estimates (Andersen (1992)) support the results obtained by Sachs and Wyplosz. [50]

Between 1980 and 1990 import penetration for manufactured goods has increased from 25½ to 36½ per cent. [51]

For a more detailed discussion of this episode see Sachs and Wyplosz. [52]

The Bank of Canada left no doubt that it was serious about the price stability target, as only two years later it had raised short-term interest rates by more than 5 percentage points, let the nominal effective exchange rate appreciate by over 7 per cent and allowed a real appreciation of more than 15 per cent. However, because the economy was probably close to or even above potential in 1989, the Bank alone could not enforce the stability target, and the rate of inflation, instead of decelerating, increased by 1½ percentage points. [53]

The timing of the announcement of the inflation target should be seen in relation to a major change in the structure of indirect taxes introduced in early 1991. The resulting one-time price jump makes it difficult to evaluate the underlying inflation trend. [54]

According to OECD (1991), the capital/output ratio increased by 20 percent between 1979 and 1989, compared to 10 per cent for other OECD countries on average. [55]

The crude estimates reported in Table 8 may also be indicative of a poor productivity and growth performance. In the Y3-equation, which only includes structural labour market parameters, Canada deviates by almost two standard errors of estimate, suggesting that the high output costs cannot be explained by labour market rigidities alone. [56]

Institutional factors such as higher minimum wages and unemployment compensation ratios, as well as stronger unionisation of the labour force are frequently mentioned. In addition, demographic factors (including a sharp rise in the female participation rate) and wide differences in regional labour market conditions combined with relatively low mobility have played a role. [57]

In his analysis of changes in the NAIRU in Canada, Coe (1990) identifies three policy-related variables: the minimum wage, the unemployment benefit insurance system and social security contributions by employers. The minimum wage has been largely neutral since 1985, and some tightening of the insurance system has reduced the overall compensation rate, while higher social security contributions have raised the NAIRU, leaving an overall net effect of around 0.5 percentage points. On the other hand, considering the NAIRUs reported in Setterfield et al. (1992), which cover virtually the whole range of feasible actual rates, one is well advised not to draw too firm a conclusion from estimated changes in the NAIRU. [58]

See Cozier and Wilkinson (1990) who find a parameter of 0.68 by regressing quarterly price changes on current and lagged values of the ratio of actual to potential output, and Fortin (1991) who, on the basis of annual price changes regressed on current and lagged rates of unemployment, finds a hysteresis parameter of 1 after 1973. [59]

According to Coe, the 1982–83 recession combined with hysteresis in the unemployment benefit insurance system left the NAIRU 0.5 percentage points higher (in 1988) than it otherwise would have been. [60]

The reaction function of the Bank of Canada is rather more complicated than described above, since in the event of unfavourable external shocks the Bank attempts to absorb the adverse effects through changes in both interest rates and the exchange rate, depending on where in the Bank's judgment the damage will be least. [61]

During 1983–84 short- and long-term interest rates were raised by 1.5–2 percentage points, and the differential against US short-term rates had increased to 2.5 points by 1986. Nonetheless, the nominal effective exchange rate declined by over 10 per cent between 1982 and 1985. [62]

The primary balance was in surplus by 1987, but much of the reduction in the primary deficit and in general government net borrowing during 1985–89 was the result of cyclical factors and last year the general government deficit/GDP ratio was almost back at the 1982 level. [63]

Virtually throughout the post-War period Denmark has been governed by minority governments and all major policy changes have been preceded by difficult negotiations with non-government parties. [64]

The Government also tried – unsuccessfully – to advance wage negotiations in the public sector, hoping that an early and low settlement would give a clear signal to the private sector. In evaluating wage developments in the 1980s two additional features should be kept in mind. Firstly, on several occasions the Government sought – again unsuccessfully – to reinforce the guidelines in tripartite negotiations in which tax reductions were proposed in return for lower wage claims. Secondly, in the event of a breakdown in the general wage bargaining round, Parliament is empowered to set wage increases for the next two years on the basis of the latest proposal by the public arbitrator, and most of the major settlements in the 1980s were actually imposed by the legislature. [65]

See Andersen and Risager (1988). A more fortuitous factor was the 16 per cent devaluation by Sweden in October 1982, as it provided the Government with a unique opportunity to confirm its exchange rate commitment by resisting strong devaluation pressures. [66]

The guidelines only applied to contractual wage increases, which typically account for one-half to two-thirds of total wage growth. [67]

Several Danish commentators explained the passive role of the government by the general election to be held in 1987. [68]

On this point see Casella and Eichengreen (1991), who argue that Marshall Aid and a willingness to incur external deficits helped to solve the distributional issues and reduce inflation in Italy and France during the immediate post-War period. [69]

Measured against GNP the ratios would be somewhat higher, reflecting very high net interest payments to abroad, and more recently also large dividend payments. GNP is less than 90 per cent of GDP (in 1990). [70]

Although the unemployment coefficient reported for Ireland in the Appendix is not much lower than that found for Denmark, and actually higher than those for Germany, France and the United Kingdom, the semi-logarithmic specification implies that while in 1981 a 1 point rise in the rate of unemployment would have reduced wage inflation by almost 0.25 percentage points, the fall in wage inflation would have been only 0.1 percentage points in 1986. [71]

Most public sector wages fall under the jurisdiction of various Conciliation and Arbitration Systems. [72]

On the other hand, since 75 percent of the public debt is held abroad, the exposure to exchange rate fluctuations is unusually large and this should serve to strengthen the credibility of the exchange rate commitment. [73]

The weights of sterling and the US dollar in the effective exchange rate index are 36 and 16 per cent respectively, compared with 15 per cent for the currency of Ireland's most important ERM trading partner (Germany). [74]

The national wage agreements of the 1970s had many features of a centralised incomes policy, but gradually developed an inflationary bias (mainly because of excessive pay gains for public employees) and were abandoned by the employers in 1980. [75]

Judging by the real exchange rate changes shown in Table 4, Ireland was more successful than Denmark in this respect. However, average productivity gains in Irish manufacturing probably overstate the change in the competitive position because of exceptionally large gains in the foreign-owned office and data-processing sector, and the closing-down of some of the traditional but inefficient manufacturing firms. In Denmark, by contrast, alternative indicators suggest that productivity growth may be understated. [76]

As discussed in Tsuru (1992), the shift in macroeconomic policies in Japan led to important institutional changes in the labour market and to a significant reduction of wage spillovers. [77]

One argument frequently advanced against exchange rate targets – viz. the loss of an instrument of adjustment to meet external shocks – does not seem valid in this case, as the external shocks of the 1980s were mostly favourable for the ERM countries. On the other hand, the high sacrifice ratios should not be taken as proof that these countries would have been better off with monetary aggregate targets and flexible exchange rates. [78]

A similar problem may occur if the key currency country, for reasons not present in other countries, is forced to tighten monetary policy and increase nominal and real interest rates. This situation could be observed in 1991–92, and has also been a problem during part of the period reviewed in this paper. [79]

Appendix: Sources of Rigidities and Persistence

1. Wage and Labour Market Rigidities

One important paradigm of New Keynesian economics is that changes in monetary policy are accompanied by short-run real effects due to rigidities in the wage and price formation process. When price and wage expectations partly depend on past developments (nominal rigidities) a more restrictive monetary policy will tend to reduce real demand as well as the rate of inflation. Moreover, since prices and wages do not immediately fall to clear output and labour markets (real rigidities), real demand will remain below the initial level until the rates of price and wage changes have fully adjusted to the change in monetary policy, with the size of the decline and the length of the period during which real demand remains depressed depending on both nominal and real rigidities.

In order to get an approximate idea of the rigidities and their possible role in explaining the variation in sacrifice ratios discussed in the text, nominal wage equations were estimated for sixteen countries, using the following general specification:

with: W = aggregate nominal wages per employee
  U = rate of unemployment
  PC = private consumption deflator
  ln = natural logarithm and
  d = first difference operator.

Assuming that the equations satisfy the long-run neutrality condition (β3 = 1−β1−β2) and that a change in monetary policy has an immediate effect on the current rate of price inflation, the degree of nominal rigidity may be measured by 1−β1. If the lagged term in U is disregarded for the moment, the degree of real rigidity is captured by η or 1/η) (i.e. the rise in InU required to generate a one percentage point fall in the rate of wage inflation). However, since equilibrium unemployment may be partly dependent on the historical path of actual unemployment (hysteresis), equation [A1] also includes a hysteresis parameter (α). In the case of extreme hysteresis (α = 1), wage inflation only depends on changes in InU and the costs of disinflation will be permanent rather than transitory. On the other hand, for 0 <α < 1, the deceleration of wage inflation depends on the level of InU, as well as its change, and, ceteris paribus, the degree of real wage rigidity will be higher than in the absence of hysteresis.[1]

Estimates of [A1] are presented in Table A1, and in most cases the assumption of a long-ran vertical Phillips curve holds. For some countries, though, there is a need for further experiments with the lag structure, or for introducing country-specific variables. Thus, for the United States and Japan, the neutrality condition is doubtful, and the condition is even more clearly rejected for Italy and Sweden. The result for Italy could be due to a major change in the indexation system in 1985,[2] while for Sweden all alternative formulations implied money illusion and a high degree of nominal rigidity.[3] For the United Kingdom and Australia the neutrality condition is easily satisfied, but it was necessary to include dummy variables and – in the case of Australia – constrain (β1 to unity to obtain a plausible adjustment pattern.

Table A1: Nominal Wage Equations1
(1960–90, annual data)
lnU lnU−1 d ln PC d ln PC−1 d ln W−1 SE DW F-test
United States −0.23 (3.3)   0.27 (2.7) 0.19 (1.3) 0.54c 0.76 2.36 4.5*
Japan −1.84 (5.9) 1.26 (3.5) 0.54 (3.9) 0.46c 1.52 1.19 3.7*
Germany −0.45 (5.3) 0.34 (3.8) 0.53 (3.1) 0.47c 1.43 2.09 0.7
France   −0.15 (4.6) 0.86 (7.0) 0.14c 1.23 1.72 0.7
Italy2 −0.38 (3.4)   0.74 (6.7) 0.26c 1.70 1.32 10.4**
United Kingdom3 −0.15 (1.2) 0.14 (1.1) 1.00 (c) 0.20 (1.5) −0.20c 1.48 2.39 0.2
Canada −0.26 (3.0)   0.64 (4.4) 0.36c 1.49 1.80 0.1
Austria4 −0.94 (4.4) 0.58 (2.5) 0.28 (1.1) 0.50 (1.9) 0.22c 1.82 1.67 0.4
Belgium −0.28 (5.7)   0.63 (3.4) 0.37c 2.03 2.11 0.8
Denmark5   −0.19 (3.9) 0.69 (2.8) 0.31c 2.13 2.65 0.9
Ireland4   −0.19 (1.7) 0.88 (4.6) 0.12c 2.88 2.04 0.6
Netherlands −0.35 (3.4) 0.13 (1.2) 0.67 (5.8) 0.13 (1.0) 0.20c 1.35 1.71 0.1
New Zealand4 −0.09 (1.4) 0.04 (0.6) 0.62 (5.0) 0.38c 3.17 1.67 0.1
Norway5 −0.20 (3.0)   0.63 (5.2) 0.37c 1.87 1.80 0.2
Sweden −0.50 (2.5) 0.34 (2.2) 0.35 (2.0) 0.65c 1.56 2.36 13.1**
Australia5, 6 −0.18 (4.6)   1.00 (c) 1.53 1.58 2.3
1. At a first stage all equations were estimated as:
dlnW = μ + ηlnUηαlnU−1 + β1dlnPC + β2dlnPC−1 + β3dlnW−1 with
W = wages per employee, aggregate economy
PC = consumer prices
U = rate of unemployment.
At a second stage the constraint β3 = 1 − β1 − β2 was imposed and tested and variables with insignificant coefficients were dropped. μ was significant in all countries, but estimated values are not shown.
c = constrained parameter
* = parameter restriction rejected at 95% level of significance.
** = parameter restriction rejected at 99% level of significance.
2. Also includes a dummy variable for 1963: 0.046 (2.6).
3. Also includes a dummy variable for 1976 (0.5) and 1977 (1): −7.81 (5.6).
4. Weekly or monthly earnings in manufacturing, instead of aggregate wages.
5. Compensation per employee, aggregate economy, instead of aggregate wages.
6. Also includes a dummy variable for 1974(1) and 1986(−1): 4.1 (3.7)

Leaving aside these less satisfactory aspects, which partly arise from using a uniform specification for all countries, the following principal results are worth noting:

  • a rather high measure of real wage flexibility is found for three of the major countries and for Austria and Sweden. However, when the hysteresis parameter is taken into account, only Japan and Austria remain in the high-flexibility group, as in Sweden and Germany the unemployment coefficient falls to or below 0.15;
  • a comparatively high degree of nominal flexibility is obtained for France, Italy, the United Kingdom, Ireland and Australia, which are all countries with above-average inflation during most of the period. By contrast, and in line with earlier studies, nominal wages adjust slowly to price changes in the United States and Austria and, as noted above, Sweden is also characterised by high nominal rigidity.

2. Real Wage and Unemployment Persistence

As an alternative to evaluating labour market rigidities from the adjustment equations for nominal wages, the constrained version of [A1] may be rewritten as a real wage adjustment equation:

with β3 indicating the degree of real wage persistence and (1−β1) the short-ran impact of inflationary surprises. This is the approach adopted by Alogoskoufis and Manning (1988),[4] who also include persistence in the labour demand equation and estimate overall persistence from a compact expression based on an autoregressive equation for unemployment:

On certain assumptions σ12 measures overall persistence, with the particular case of σ1 = 1 and σ2 = 0 occurring when η = ∞ and α= 1. σ1 and σ2 will sum to unity when η = 0 and β3 = 1, or when α and β3 are both equal to 1. Table A2 presents our own estimates of [A3], with σ1 + σ2 included in Table 2 of the text. The estimates are generally satisfactory,[5] and for thirteen of the countries both σ1, and σ2 are significant (90 per cent level of significance or, in the case of σ1 much higher) and of the expected signs. The estimated values for a, range from 1.6 to 1, with most of the ERM countries located towards the higher end of the range, while Austria comes close to the special case of σ1 = 1 and σ2 = 0. σ2 is negative for all countries and the range – from 0.7 to 0.1[6] – is such that for ten of the countries σ1 + σ2 ≥ 0.94. Only the United States and Sweden show an overall degree of persistence which is significantly below 1, while for the remaining countries σ1 + σ2 is around 0.9.

Table A2: Unemployment Equations*
(1960–90, annual data)
σ0 u−1 u−2 SE h
United States 1.49 (2.0) 1.01 (5.5) −0.26 (1.4) 0.97 7.6
Japan 0.13 (1.3) 1.27 (7.1) −0.33 (1.9) 0.17 0.1
Germany 0.33 (1.5) 1.41 (7.9) −0.47 (2.6) 0.70 1.4
France 0.27 (1.6) 1.35 (7.2) −0.37 (2.0) 0.49 −0.7
Italy 0.36 (1.1) 1.32 (7.6) −0.36 (1.9) 0.56 1.5
United Kingdom 0.44 (1.6) 1.52 (9.8) −0.59 (3.9) 0.86 1.6
Canada 0.84 (1.5) 1.24 (6.7) −0.35 (1.9) 0.93 3.0
Austria 0.24 (1.1) 1.00 (5.4) −0.09 (0.5) 0.39 0.6
Belgium 0.33 (1.5) 1.63 (11.3) −0.67 (4.7) 0.66 0.5
Denmark 0.35 (1.2) 1.21 (6.3) −0.24 (1.2) 0.93 0.8
Ireland 0.62 (1.6) 1.51 (8.9) −0.57 (3.3) 0.96 −0.0
Netherlands 0.36 (1.5) 1.52 (9.3) −0.57 (3.5) 0.80 1.3
New Zealand 0.13 (0.9) 1.18 (6.0) −0.11 (0.5) 0.61 0.7
Norway 0.23 (1.0) 1.27 (6.8) −0.34 (1.4) 0.53 1.5
Sweden 0.72 (2.8) 1.11 (6.4) −0.45 (2.5) 0.35 1.1
Australia 0.37 (1.1) 0.96 (5.0) −0.01 (0.1) 0.96 16.1
* Estimated from: U = σ0 + σ1U −1 + σ2U−2.

When combining the results of Table A1, which reject the hypothesis of extreme hysteresis in all cases except for the United Kingdom, with those of Table A2, where the hypothesis that U follows a random walk process cannot be rejected for any of the sixteen countries, two tentative conclusions can be derived:

  1. on the one hand, there is a risk that anti-inflation policies can result in a permanent rise in the rate of unemployment, though the transmission channels of this effect remain unknown until the process making U a random walk is further analysed;[7]
  2. on the other hand, even though the forces generating the random walk process do not seem to be embedded in the wage formation process, the results of Table A2 imply that the estimated unemployment coefficients in the wage equations may be biased and that sacrifice ratios calculated from the trade-off coefficients should be interpreted with some caution.

3. Price and Output Market Rigidities

Focusing on wage equations or evaluating rigidities from unemployment functions clearly involves a bias, since it is generally recognised that rigidities are equally important in product markets and in the adjustment of output prices. It may also be argued (see Gordon (1988) and (1990)) that focusing on the wage equation is misleading and irrelevant because wage changes only affect the share of wages and have no influence on the rate of price inflation. Mehra (1991) also disputes the validity of the traditional mark-up model for the United States, and a recent study by Cozier (1991) finds a similar result for Canada. On the other hand, experience shows that it is difficult to estimate real and nominal rigidities from price equations, as the sensitivity of price changes to output slack is low and the homogeneity condition is not easily satisfied. Moreover, in the case of Australia, Fahrer and Myaat (1991) show that the wage equation is the key relation for anti-inflation policies.[8]

Nonetheless, as a supplement to the estimates discussed above we have included Table A3, showing developments in the ‘split’ between nominal and real changes in GDP. Corresponding to the disinflation documented in Table 1 of the text, there is a marked improvement between the 1970s and the 1980s. Even for 1990, when inflation accelerated in several countries, it was still the case that for the sixteen countries on average a rise in nominal GDP of 5¾ per cent was split between a real change of 2½ per cent and a price change of just over 3 per cent. On the other hand, as nominal GDP growth fell to 3¾ per cent last year (partly because of restrictive monetary policies) most of the deceleration appeared as lower real growth.

Table A3: Developments in the ‘Split’
Real Output Growth Relative to Changes in Output Deflator
1973–79 1979–81 1981–86 1986–89 1990 19911
United States 0.30 0.23 0.83 1.06 0.24 −0.19
Japan 0.45 1.18 2.57 8.33 2.48 2.37
Germany 0.48 0.18 0.53 1.38 1.32 0.70
France 0.25 0.12 0.22 0.94 1.00 0.39
Italy 0.22 0.13 0.16 0.57 0.27 0.15
United Kingdom 0.09 −0.11 0.57 0.61 0.16 −0.31
Canada 0.45 0.24 0.65 0.79 0.17 −0.55
Austria 0.47 0.22 0.36 1.37 1.59 0.88
Belgium 0.27 0.39 0.24 1.24 1.23 0.45
Denmark 0.19 −0.07 0.55 0.15 0.81 0.50
Ireland 0.35 0.20 0.20 1.00 (2) 0.48
Netherlands 0.36 0.03 0.62 2.78 1.34 0.61
New Zealand 0.00 0.16 0.21 0.03 0.37 −1.54
Norway 0.60 0.18 0.77 0.10 0.40 1.40
Sweden 0.17 0.11 0.29 0.40 0.05 −0.17
Australia 0.18 0.27 0.36 0.48 0.33 −1.54
Average 0.30 0.22 0.57 1.33 0.80 0.25
GDP growth, real 2.8 1.5 2.4 3.0 2.5 0.8
GDP growth, nominal 12.4 8.6 6.8 5.3 5.7 3.8
1. Negative figures indicate a combination of falling real output and higher prices.
2. Excluded from the calculation since real output was rising faster than nominal output.

This recent episode might be used to illustrate the nature of price rigidities and their relevance to the disinflation process.[9] Denoting annual changes in log nominal GDP by dy, real growth by dq and changes in the output deflator by dp, the following identities hold:

In this set-up the crucial question is why ‘a’ is not close to unity. A preliminary answer can be obtained by assuming that price changes are determined by two factors: the degree of excess supply in the product market and inertia. Using Inline Equation to indicate the ratio of actual to potential output and letting inertia enter via adaptive price expectations, an equation for the changes in output prices can be specified as:

As a second step the national accounting identity (dy ≡ dq + dp) may be rewritten as:

Inserting [A6], with dq* denoting potential output growth, into [A5], and letting e = b/(1+b) then gives the following price adjustment equation:

where nominal rigidity can be measured by (1−e) and real rigidity by e(1−c).

As an illustration, assume that the growth of monetary aggregates is permanently reduced by 1 percentage point. When velocity is constant dy also falls by 1 percentage point and, if there is no excess supply initially, price inflation will decline by e points and output growth by (1−e) points. In other words, the initial response to the anti-inflation policy is entirely determined by the degree of nominal rigidity. In the following year, however, real rigidities will also enter as the fall in output growth will lower Inline Equation by (1−e) per cent, so that the cumulative decline in price inflation amounts to e + e(1−c)(1−e) + e(1−e). Continuing this example, it can be shown that if inflation is to be permanently reduced by 1 percentage point with no permanent loss of output, dp will have to ‘undershoot’ to allow dq to exceed dq* and bring Inline Equation back to the initial level.[10] Moreover, it is in the second and subsequent years that the degree of real rigidity becomes important to the transitory output costs, since it is the sensitivity of dp to output slack which determines how far Inline Equation has to fall to bring dp into line with the change in monetary policy.

Finally, as a tentative evaluation of output market and price rigidities, Table A4 shows estimates of the parameters of equation [A7] using national accounts data for the period 1960–90. As discussed in Andersen (1989), applying [A7] as a ‘short cut’ to estimating aggregate supply curves and the degree of real and nominal price rigidities can be problematic. This may be one reason why the ranking of real and nominal rigidities differs between labour and output markets (see Table A5). On the other hand, the differences may reflect ‘real’ factors which were also apparent in the sacrifice ratios.

Table A4: Output Price Equations(1)
C e(1 − c) (1 − e) R2 DW SE F(2) dPM(3) t1982–
United States −0.95 (4.2) 0.43 (5.7) 0.68 (6.0) 0.67 1.91 0.78 0.7 3.3
Japan −2.16 (3.8) 0.57 (3.6) 0.27 (7.7) 0.68 2.18 1.66 1.8 0.03 (3.7) 4.0
Germany −0.56 (1.8) 0.19 (2.4) 0.62 (5.8) 0.53 1.83 0.83 3.5 0.01 (2.0) 1.9
France −1.37 (4.8) 0.34 (3.2) 0.60 (6.1) 0.63 1.58 0.88 0.1 2.0
Italy −1.93 (9.7) 0.40 (5.5) 0.45 (15.7) 0.90 1.00 0.79 2.3 2.4
United Kingdom −1.80 (1.9) 0.56 (3.3) 0.23 (6.5) 0.89 1.74 1.67 0.1 2.8
Canada −1.99 (6.5) 0.30 (3.3) 0.51 (8.8) 0.74 1.64 1.02 0.3 3.1
Austria −1.25 (3.9) 0.11 (1.4) 0.64 (4.6) 0.41 2.62 1.05 5.6 1.6
Belgium −1.33 (3.8) 0.21 (2.1) 0.57 (5.5) 0.53 2.13 1.23 2.5 1.6
Denmark −0.45 (2.9) 0.45 (4.3) 0.71 (3.6) 0.46 1.66 1.06 0.3 2.1
Ireland −3.15 (6.6) 0.22 (1.5) 0.22 (10.0) 0.78 1.59 1.94 0.9 2.8
Netherlands −1.39 (4.1) 0.06 (0.6) 0.52 (6.1) 0.59 1.80 1.23 7.2 0.11 (3.0) 1.4
New Zealand −2.01 (1.3) 0.51 (2.9) 0.20 (5.3) 0.66 1.78 3.02 0.2 1.5
Norway −3.00 (2.9) 0.50 (2.6) 0.20 (8.3) 0.76 1.65 1.56 0.1 2.5
Sweden −1.29 (4.8) 0.47 (3.9) 0.48 (7.6) 0.71 2.00 0.88 1.8 2.1
Australia −1.94 (5.6) 0.39 (3.6) 0.48 (7.6) 0.69 1.80 1.19 0.1 3.3
(1) The estimated equation is [A7] with Q* determined by a cubic trend; annual data 1960–90.
(2) F-test for homogeneity condition; critical values are 4.5 (95%) and 7.8 (99%).
(3) For Japan and Germany the homogeneity condition was only met when the lagged change in import prices was included in [A7]. For the Netherlands the lagged import price change is highly significant, but the homogeneity condition is still violated.
(4) Estimated trend output growth, 1982–90; annual rate.
Table A5: Ranks Based on Labour and Product Market Parameters*
  η(1 − α) e(1 − c) (1 − β) (1 − e)
United States 6 7 15 15
Japan 1 1 12 5
Germany 14 14 13 13
France 13 10 3 12
Italy 2 8 5 6
United Kingdom 16 2 4 4
Canada 5 11 8 9
Austria 3 15 16 14
Belgium 4 13 9 11
Denmark 9 6 6 16
Ireland 9 12 2 3
Netherlands 7 16 7 10
New Zealand 15 3 11 1
Norway 9 4 9 1
Sweden 12 5 14 7
Australia 11 9 1 7
Rank correlations: −0.15 0.42
* A high degree of either real or nominal fiexibility is indicated with a low rank.

Overall, the range of real rigidities in the output market is much narrower than for the labour market and, except for the Netherlands, all coefficients are statistically well determined. Japan again appears to have the lowest real rigidity, but otherwise the correspondence between the two measures of real rigidities is rather poor and the rank correlation is negative. The largest discrepancies can be observed for the United Kingdom and New Zealand, where the ranking based on the labour market parameter is very high relative to the output market measure, and for Austria and Belgium, where the reverse pattern is seen.

The nominal rigidities are somewhat closer, though again the rank correlation is not very high. In the United States and Germany both output and labour markets have a relatively high degree of nominal rigidity, while in Japan prices appear to respond more quickly than wages. This is somewhat unexpected given the discussion of sacrifice ratios in the text, but the lag structure for Japan is uncertain since it was difficult to satisfy the homogeneity condition. For France and Denmark wages respond more quickly than prices, while New Zealand and Norway have a response pattern similar to that of Japan.

Appendix Footnotes

Rewriting the terms in InU as – η(1−α)lnU − η αdlnU, the degree of long-run real wage rigidity is easily seen to be η(1−α) or 1/η(1−α). [1]

For further discussion and evidence see Barrell et al. (1990). [2]

In general, failure to satisfy the neutrality condition may also indicate that the market power of employees and their unions is low, and that output prices should have been included in the wage equation. [3]

[A2] is equivalent to their wage adjustment equation, except that they enter W and PC in levels. We also tested the level version and found virtually the same values for η and α as shown in Table A1. However, for most countries β3 was close to unity and in all cases the DW-statistics were very low, implying that the dynamic structure was misspecified. [4]

For the United States, Canada and Australia the inclusion of only two lagged terms in U is clearly insufficient. In all three cases the addition of U−3 reduced the h-statistics to below 1, while the sum of the coefficients remained close to those shown in Table 2. [5]

For New Zealand and Australia σ2 is not significantly different from 0, but for Australia this was no longer the case when U−3 was added. [6]

In a study of the Okun's law relationship for Australia, Mitchell and Watts (1991) find that a lower level of capacity growth and a rise in the sectoral variability of employment changes have been the major factors behind the increase in the equilibrium unemployment rate. For Canada Coe (1990) finds that the unemployment benefit insurance system has been an important source of hysteresis. [7]

See also Boehm and Martin (1989), Carmichael (1990) and EPAC (1991 and 1991b). [8]

The following is the approach preferred by Gordon (see Gordon (1981) and (1990)). [9]

See Coe and Holtham (1983) and Bispham (1984). [10]


Agenor, P.R. and M. Taylor (1991), ‘Testing For Credibility Effects’, IMF Working Paper No. 91/110.

Alogoskoufis, G. and A. Manning (1988), ‘On the Persistence of Unemployment’, Economic Policy, 7, pp. 427–69.

Andersen, P. (1989), ‘Inflation and Output: A Review of the Wage-Price Mechanism’, BIS Economic Papers No. 24.

Andersen, P. (1992), ‘Taxes and Labour Market Developments’, mimeo., BIS.

Andersen, T. and O. Risager (1988), ‘Stabilisation Policies, Credibility and Interest Rate Determination’, European Economic Review, 32, pp. 669–79.

Artis, M. and D. Nachane (1990), ‘Wages and Prices in Europe: A Test of the German Leadership Thesis’, Weltwirtschaftliches Archiv, 126, pp. 59–77.

Artis, M. and P. Ormerod (1991), ‘Is There an “EMS” Effect in European Labour Markets?’, CEPR Discussion Paper No. 598.

Ashenfelter, O. and D. Card (1986), ‘Why Have Unemployment Rates in Canada and the United States Diverged?’, NBER Working Paper No. 1840.

Ball, L. (1990), ‘Credible Disinflation with Staggered Price Setting’, NBER Working Paper No. 3555.

Ball, L. (1991), ‘The Genesis of Inflation and the Costs of Disinflation’, Journal of Money, Credit and Banking, 23, pp. 439–61.

Ball, L. (1992), ‘Disinflation with Imperfect Credibility’, NBER Working Paper No. 3983.

Barrell, R., J. Darby and C. Donaldson (1990), ‘Structural Stability in European Wage and Price Systems’, NIESR Discussion Paper No. 188.

Baxter, M. (1985), ‘The Role of Expectations in Stabilisation Policy’, Journal of Monetary Economics, 15, pp. 343–62.

Bispham, J. (1984), ‘Inflation, Recession and Recovery, A Nominal Income Analysis of the Process of Global Disinflation’, BIS Economic Papers No. 11.

Blackburn, K. and M. Christensen (1989), ‘Monetary Policy and Policy Credibility’, Journal of Economic Literature, 27, pp. 1–45.

Blanchard, O. (1984), ‘The Lucas Critique and the Volcker Deflation’, The American Economic Review, Papers and Proceedings, 74, pp. 211–15.

Boehm, E. and V. Martin (1989), ‘An Investigation into the Major Causes of Australia's Recent Inflation and Some Policy Implications’, The Economic Record, 65, pp. 1–15.

Bruno, M. (1991), ‘High Inflation and the Nominal Anchors of an Open Economy’, Essays in International Finance, No. 183, Princeton University.

Buiter, W. and U. Miller (1986), ‘Costs and Benefits of an Anti-inflationary Policy: Questions and Issues’ in V. Argy and F. Neville (eds.), Inflation and Unemployment, Theory, Experience and Policy Making, Allen and Unwin, London.

Carmichael, J. (1990), ‘Inflation: Performance and Policy’, in S. Grenville (ed.), The Australian Macro-economy in the 1980s, Reserve Bank of Australia, Sydney.

Casella, A. and B. Eichengreen (1991), ‘Halting Inflation in Italy and France After World War II’, NBER Working Paper No. 3852.

Chadha, B., P. Masson and G. Meredith (1991), ‘Models of Inflation and the Costs of Disinflation’, IMF Working Paper No. 91/97.

Chouraqui, J-C, R. Hagemann and N. Sartor (1990), ‘Indicators of Fiscal Policy: A Reassessment’, OECD Working Paper No. 78.

Christensen, M. (1988), Theoretical and Empirical Evaluation of Credibility and Reputation in Macroeconomics, Institute of Economics, University of Aarhus, Denmark.

Cling, J-P. and F. Meunier (1986), ‘La désinflation en France: Le point de vue de l'économètre’, Revué Economique, 6, pp. 1093–1126.

Coe, D. (1990), ‘Structural Determinants of the Natural Rate of Unemployment in Canada’, IMF Staff Papers, 37, pp. 94–115.

Coe, D., M. Durand and U. Stiehler (1988), ‘The Disinflation of the 1980s’, OECD Economic Studies, No. 11, pp. 89–121.

Coe, D. and G. Holtham (1983), ‘Output Responsiveness and Inflation’, OECD Economic Studies, No.1 pp. 93–146.

Cotis, J-P. and A. Loufir (1990), ‘Formation des Salaires, Chomage ‘d'equilibre’ et Incidencedes Cìtisations sur le cìut du Travail’, Economie et Prevision, pp. 97–110.

Cozier, B. (1991), ‘Wage and Price Dynamics in Canada’, Bank of Canada Technical Report No. 56.

Cozier, B. and J. Selody (1992), ‘Inflation and Macroeconomic Performance’, Bank of Canada Working Paper No. 92-3.

Cozier, B. and G. Wilkinson (1990), ‘How Large are the Costs of Disinflation in Canada ?’, Bank of Canada Working Paper No. 90-6.

Dornbusch, R. (1989), ‘Credibility, Debt and Unemployment: Ireland's Failed Stabilisation’, Economic Policy, 8, pp. 173–210.

Dornbusch, R. (1991), ‘Credibility and Stabilisation’, Quarterly Journal of Economics, 106, pp. 837–49.

Dornbusch, R. and S. Fischer (1991), ‘Moderate Inflation’, NBER Working Paper No. 3896.

Englander, S. and S. Los (1983), ‘The Stability of the Phillips Curve and the Implications for the 1980s’, Federal Reserve Bank of New York Research Paper No. 8303.

EPAC (1991a), ‘Australia's Inflation Problem’, Background Paper No. 11.

EPAC (1991b), ‘Improving Australia's Inflation Performance’, Council Paper No. 48.

Fahrer, J. and J. Myatt (1991), ‘Inflation in Australia: Causes, Inertia and Policy’, Reserve Bank of Australia Research Discussion Paper 9105.

Fischer, S. (1977), ‘Long-Term Contracts, Rational Expectations and the Money Supply Rule’, Journal of Political Economy, 85, pp. 191–205.

Fischer, S. (1986a), ‘Contracts, Credibility and Disinflation’ in S. Fischer (ed.), Indexing, Inflation and Economic Policy, MIT Press, Cambridge, MA.

Fischer, S. (1986b), ‘Exchange Rates Versus Money Targets in Disinflation’ in S. Fischer (ed.), Indexing, Inflation and Economic Policy, MIT Press, Cambridge, MA.

Ford, R. and D. Rose (1989), ‘Estimates of the NAIRU Using an Extended Okun's Law’, Bank of Canada Working Paper No. 89-3.

Fortin, P. (1989), ‘How “Natural” is Canada's High Unemployment Rate?’, European Economic Review, 33, pp. 89–110.

Fortin, P. (1990), ‘Can the Costs of an Anti-Inflation Policy be Reduced?’ in R.C. York (ed.), Taking Aim: The Debate on Zero Inflation, CD. Howe Institute, Ottawa.

Fortin, P. (1991), ‘The Phillips Curve, Macroeconomic Policy and the Welfare of Canadians’, Canadian Journal of Economics, 24, pp. 774–803.

Freedman, C. (1989), ‘Monetary Policies in the 1990s: Lessons and Challenges’, in Monetary Policy Issues in the 1990s, Symposium sponsored by Federal Reserve Bank of Kansas City, Jackson Hole, Wyoming.

Freedman, C. (1991), ‘The Goal of Price Stability: The Debate in Canada’, Journal of Money, Credit and Banking, 23, pp. 613–18.

Giavazzi, F. and M. Pagano (1988), ‘The Advantage of Tying One's Hands’, European Economic Review, 32, pp. 1055–82.

Giavazzi, F. and L. Spaventa (1989), ‘Italy, the Real Effects of Inflation and Disinflation’, Economic Policy, 8, pp. 133–72.

Gordon, R. (1981), ‘Output Fluctuations and Gradual Price Adjustment’, Journal of Economic Literature, 19, pp. 493–532.

Gordon, R. (1982), ‘Why Stopping Inflation can be Costly: Evidence from Fourteen Different Episodes’, in R.E. Hall (ed.), Inflation: Causes and Effects, NBER Project Report Series, University of Chicago Press, Chicago.

Gordon, R. (1988), ‘The Role of Wages in the Inflation Process’, The American Economic Review, 78, pp. 276–83.

Gordon, R. (1990), ‘The Phillips Curve Now and Then’, NBER Working Paper No. 3393.

Gordon, R. and S. King (1982), ‘The Output Cost of Disinflation in Traditional and Vector Autoregressive Models’, Brookings Papers on Economic Activity, No. 1, pp. 205–44.

Howitt, P. (1990), ‘Zero Inflation as a Long-Term Target for Monetary Policy’, in R. Lipsey (ed.) Zero-Inflation, CD. Howe Institute, Ottawa.

Jarrett, P. and J. Selody (1982), ‘The Productivity-Inflation Nexus in Canada: 1963–79’, Review of Economics and Statistics, 64, pp. 361–67.

Kahn, G. and S. Weiner (1990), ‘Has the Cost of Disinflation Declined?’, Economic Review, Federal Reserve Bank of Kansas City, 15, pp. 5–24.

Kremers, J. (1990), ‘Gaining Policy Credibility for a Disinflation’, IMF Staff Papers, 37, pp. 116–44.

Lucas, R. (1990), ‘The Case for Stable but not Zero Inflation’, in R.C. York (ed.), Taking Aim: The Debate on Zero Inflation, CD. Howe Institute, Ottawa.

Mehra, Y. (1991), ‘Wage Growth and the Inflation Process: Empirical Note’, The American Economic Review, 81, pp. 931–39.

Mitchell, R. and M. Watts (1991), ‘Alleged Instability of the Okun's Law Relationship in Australia: An Empirical Analysis’, Applied Economics, 23, pp. 1829–38.

Mundell, R.(1971), ‘The Dollar and the Policy Mix: 1971’, Essays in International Finance, No. 85, Princeton University.

Murphy, C. (1991), ‘The Transitional Costs of Reducing Inflation Using Monetary Policy’, in EPAC, Australia's Inflation Problem, Background Paper No. 11.

OECD (1990), OECD Economic Survey of France, OECD, Paris.

OECD (1991), OECD Economic Survey of Canada, OECD, Paris.

Okun, A. (1978), ‘Efficient Disinflationary Rules’, The American Economic Review, 68, pp. 348–52.

Rose, D. (1988), ‘The NAIRU in Canada: Concepts, Determinants and Estimates’ Bank of Canada Technical Report No. 50.

Sachs, J. (1985), ‘The Dollar and the Policy Mix: 1985’, Brookings Papers on Economic Activity, No. 1, pp. 117–97.

Sachs, J. and C. Wyplosz (1986), ‘The Economic Consequences of President Mitterrand’, Economic Policy, 9, pp. 261–322.

Sargent, T. (1982), ‘The Ends of Four Big Inflations’, in R.E. Hall (ed.), Inflation: Causes and Effects, NBER Project Report Series, University of Chicago Press, Chicago.

Scarth, W. (1990), ‘Are the Costs of Getting to Zero Too High?’, in R.C. York (ed.), Taking Aim: The Debate on Zero Inflation, CD. Howe Institute, Ottawa.

Selody, J. (1990), ‘The Goal of Price Stability: A Review of the Issues’, Bank of Canada Technical Report No. 54.

Setterfield, M, D. Gordon and L. Osberg (1992), ‘Searching For a Will 'o the Wisp’, European Economic Review, 36, pp. 119–36.

Takeda, U. (1992), ‘The Term Structure of Interest Rates as an Indicator of the Central Bank's Policy Stance’, mimeo., BIS.

Taylor, J.B. (1980), ‘Aggregate Dynamics and Staggered Contracts’, Journal of Political Economy, 88, pp. 1–23.

Taylor, J.B. (1983), ‘Union Wage Settlements During a Disinflation’, The American Economic Review, 73, pp. 981–93.

Taylor, J.B. (1989), ‘Differences in Economic Fluctuations in Japan and the United States: The Role of Nominal Rigidities’, Journal of the Japanese and International Economies, 3, pp. 127–44.

Tobin, J. (1980), ‘Stabilisation Policy Ten Years After’, Brookings Papers on Economic Activity, No. 1, pp. 19–90.

Tsuru, T. (1992), ‘The Spring Offensive: The Spillover Effect and the Wage-Setting Institution in Japan’, The Institute of Economic Research, Discussion Paper No. 247, Hitotsubashi University, Tokyo.

van Poeck, A. (1989), ‘Wage Formation, Labour Market Characteristics and the EMS’, Report No. 891235, University of Antwerp.

Vegh, C. (1991), ‘Stopping High Inflation: An Analytical Overview’, IMF Working Paper No. 91/107.

Weber, A. (1991), ‘Reputation and Credibility in the European Monetary System’, Economic Policy, 12, pp. 57–102.