RBA Annual Conference – 1992 Discussion

1. Bill Norton

This is a very thorough paper. I will talk about what it says to policy makers, especially in Australia. Palle's messages can be put into two groups:

  1. those that are largely taken for granted, on which there is little discussion; and
  2. those on the main issues of the paper.

(a) Taken for Granted

Palle tells us that:

  • inflation fell in virtually all countries in the 1980s, triggered by a tightening of policies in the late 1970s to bring inflation rates back to about where they were in the 1960s;
  • tight monetary policy is a necessary part of a low-inflation policy, although other factors can help; and
  • there is no long-term economic cost of reducing inflation – in technical terms, the long-run Phillips curve is vertical – although there can be political costs.

These messages are, in a sense, taken for granted. This would not have been so a decade or more ago, which shows how the debate on these issues has moved on.

(b) Main Issues

Palle's main issues are the short-term economic cost of reducing inflation and what can be done to minimise that cost.

His first task is to measure the short-term economic costs. On that, he finds:

  • In terms of output, there is a short-run cost but it declines over time. There are a few surprises. For one thing, deflation costs are above average in Germany and a number of other countries that are tied to Germany by Europe's exchange rate regime. Palle's figure for Australia's deflation costs is much less than Glenn Stevens' Australia estimate. Palle's low figure for the trend growth of output seems part of the answer. Perhaps Palle and Glenn can explain why their figures differ so much.
  • Unemployment costs are generally larger. Australia has a relatively high figure; on this, Palle and Glenn agree. Other countries with high unemployment costs of deflation are Germany and those bound to it by a fixed exchange rate, such as Belgium, Ireland and the Netherlands. A tendency for people out of work today to also be out of work in the future is found for the United Kingdom, but not for other countries.

Palle's second task is to sort out what factors influence the costs of deflation. He does this by cross-country analysis and case studies of four countries. Palle's emphasis is on the deflation costs of ‘real and nominal rigidities’. (I put these terms in inverted commas as his use of them differs from common usage.)

  • First, ‘real rigidity’ – Palle argues that deflation costs increase if wages and prices respond slowly to changes in the degree of slack in product and labour markets. In keeping with this, Palle finds that Japan has ‘real flexibility’ and low deflation costs whereas Australia and Germany have ‘real rigidity’ and high deflation costs. But there are some curious results. Sweden has ‘real rigidity’ but low deflation costs, whereas Austria has ‘real flexibility’ but high deflation costs. Overall, however, Palle's estimates suggest that ‘real rigidity’ adds substantially to deflation costs.
  • Turning to ‘nominal rigidity’, Palle argues that deflation costs increase if wage growth responds slowly to a change in the pace of price rises. Australia is found to have ‘nominal flexibility’ – a surprising result in some ways – yet high deflation costs. Ireland and the United Kingdom are in the same boat. On the other hand, Japan, Sweden and the United States have ‘nominal rigidity’ but low deflation costs. Overall, Palle's results suggest that ‘nominal rigidity’ isn't a major influence on deflation costs.

Apart from these rigidities, Palle considers a number of other factors that can influence deflation costs. These factors include:

  • Policy credibility – that is, a credible anti-inflation policy will reduce deflation costs. In apparent contradiction, however, Palle finds high deflation costs for Germany and associated countries, such as the Netherlands and Ireland, that are often seen as having had a ‘credibility bonus’.
  • Expansive budgetary policy – that is, an overly expansive policy can increase deflation costs, either by diluting the anti-inflation policy credibility or reducing monetary policy's effectiveness. Palle thinks there is some evidence for this proposition, notably for Canada and Ireland. Australia, however, started its present deflation from a relatively sound budget position, yet has had high deflation costs.
  • A high effective exchange rate. Australia's experience could be consistent with that view, as its effective exchange rate was high at the start of the recent deflation and could have contributed to our relatively high deflation costs. In contrast, however, the United States is estimated to have low deflation costs, but began the 1980s with a high real exchange rate.
  • Incomes policy. Palle sees incomes policy as a help, but only for a short time, in minimising deflation costs. Australia's experience with the Accord seems consistent with that view.

Two other points:

  1. Palle offers opinions on a couple of issues about the deflation process. First, he argues that deflation should stop before price stability. Unfortunately, Palle offers no view on an acceptable inflation level. Measured inflation of 1 or 2 per cent a year is probably consistent with price stability after allowing for improvements in product quality. Higher inflation levels, however, seem to hamper decision-making by savers and investors. Second, Palle says that a quick deflation does not increase deflation costs. That may be so in the long run, but the time distribution of costs will differ and that may matter.
  2. Some things are missing from Palle's analysis, including: 1991 data that could alter estimates of deflation costs for some countries, including Australia; large countries from the case studies; and monetary policy differences between countries from his cross-country analysis of deflation costs.

But the main messages are correct:

  • low inflation is the only economically sensible policy;
  • tight monetary policy is vital to achieve low inflation;
  • given monetary policy's role, other things can help to minimise short-run costs of reducing inflation; and
  • now that inflation is down, the best policy is to not let it rise again.

2. Michael Artis

Any fool, it might seem, can disinflate. The interesting thing is how to minimize the costs of doing so. In order to examine this question, Palle Andersen first provides alternative estimates of the sacrifice ratio and then proposes a cross-country regression-based explanation. As a follow-up he provides a more detailed discussion of four case-histories – for Canada and for three Exchange Rate Mechanism (ERM) countries.

Controlling the price level involves using a nominal standard of some kind: internal standards include the nominal wage, a monetary aggregate, nominal GDP or prices directly; the alternative external standard involves targeting the exchange rate against a well-behaved low-inflation ‘anchor country’. Germany is supposed to have provided such an anchor for the original ERM countries in the 1980s; the United Kingdom initially chose an internal standard, the money supply, then abandoned it and, after a period of flirtation with alternatives, also decided to adopt the external standard offered by adherence to the ERM.

Palle Andersen's calculations of the sacrifice ratio confirm what others have found – that the sacrifice ratio for ERM countries (or most of them, for Italy seems to be an important exception) seems to be higher, not smaller, than the sacrifice ratio of non-ERM countries. As Palle himself points out, the drawbacks of the sacrifice ratio as a measure of the costs of disinflation are such that one has to be cautious about inferring from this evidence that the external standard is inferior to the internal one. But the evidence is provocative and the question is worth considering.

The analytical literature that is relevant here is the stochastic analysis of a decade ago (e.g., my paper with David Currie (1981)) and the considerations raised by the popular model of ‘reputational policy’ (Barro and Gordon, (1983)), which has been widely appealed to in analysis of the counter-inflationary properties of ERM participation (e.g., Giavazzi and Pagano, (1988)).

From the first strand of literature we learn that the external standard will pose problems whenever the behaviour of the anchor country becomes erratic, or when there is a structural shift which changes the equilibrium real exchange rate. In the latter event the fixity of the nominal rate will require nominal wages and prices to adjust to realize the new equilibrium real rate. This may be painful. Experience since the German unification could plausibly be categorized this way; but what of experience before? Palle's argument is that the natural rate of unemployment in Germany may have shifted upward (accounting for that country's exceptionally high observed sacrifice ratio – although of course the ‘true’ ratio should be measured as the cumulative excess of the actual over the natural rate, divided by the fall in inflation), causing ‘excess deflation’ in Germany's ERM partners. Presumably, the argument is that the rise in Germany's natural rate required a tighter monetary policy than otherwise, and the other countries were bound to follow this lead, even though it represented a movement to ‘excessively low’ rates of inflation, where – because of non-linearities in the Phillips curve or otherwise – the costs of deflation were particularly high.

The second strand of literature emphasises the issues of credibility and reputation. The usual argument, in relation to the ERM, has been that the external standard has important advantages over the internal one in this respect. The exchange rate is continuously observable, so it is easy for agents to verify whether the authorities are sticking to their commitment; the exchange rate is more ‘transparent’ than the money supply, the meaning and definition of which are matters of controversy among professionals; and, because the external standard involves a commitment by the government vis-a-vis other governments (at least in the ERM framework), the cost of reneging is increased and the authorities' commitment is the more credible. So far, so good. The analysis suggests that over time agents in Germany's partner ERM countries would begin to factor into their expectations data on German inflation – which, as Palle points out, is a possible interpretation of the findings reported by Paul Ormerod and myself on this question. This is the concrete way in which ERM participation could lead to the ‘importation of the Bundesbank's reputation’. However, Alan Walters (1986) has pointed out a possible problem with the credibility story – credibility is divisible. In a short run of non-negligible proportions, the exchange rate commitment may be highly credible to the foreign exchange market, even while it is not credible in labour and product markets. This ‘excess’ credibility has awkward implications. It implies that nominal interest rates will be more convergent than inflation rates, and therefore that real rates of interest will be ranked, across countries, perversely with respect to inflation rates. High-inflation countries will experience lower real rates of interest than low-inflation countries.

This might help explain how it is that Germany's partners seem to have experienced high sacrifice ratios. Palle's Phillips curves are presented as structural equations. However, the ‘efficient Phillips curve’ literature (see, e.g. Henry, Karakitsos and Savage, (1982) reminds us that observed Phillips curves (hence, sacrifice ratios) will reflect the set of instruments available to the authorities. If the authorities were to be deprived of their first-best policy instruments, they could only use cruder tools to bring down inflation and the costs of disinflation would be greater. In inflation-unemployment space the efficient frontier would move out. The ‘Walters effect’ suggests that something like this might have happened in Europe. Excess credibility immobilizes the interest rate, which is forced into convergence with the German rate. Getting inflation down implies resort to less preferred instruments, fiscal policy or ‘bleeding to death’ through a loss of competitiveness. The loss of the first-best counter-inflationary instrument raises the cost of disinflation. Is there any evidence that the ‘excess credibility’ problem actually exists? The answer is ‘Yes’. There is some evidence of systemic Walters effects for the first half of the EMS period, and there is anecdotal evidence of such effects in the later part of the period. For example, the data in Table 1 give the rank correlation (R) between inflation and real interest rates across the core ERM countries (Ireland excluded for data reasons) for the periods shown:

Table 1: Rank Correlation Between Inflation and Real Interest Rates
Period 79:4–
R −0.60** 0.73** −0.60** 0.73* −0.47+ 0.47+
** indicates a 1% level;
* a 5% level; and
+ a 10% significance level.

Clearly, the correlation is significantly inverse over the first part of the EMS period, though the significance and strength of the correlation weakens as the period is extended. In the second half, the correlation appears to be positive (but less significant for more recent periods). Despite the lack of systemic inverse correlation in the latter part of the period, however, it is well known that Spain and, for a time, Italy, encountered the ‘Walters problem’.

However, the good news is that the Walters effects must be ‘transitory’ – even if, as asserted above, the period of transition is seriously embarrassing for policy. Expectations can hardly go on being inconsistent for ever: experience, we should expect, must bring into line the expectations held in the foreign exchange market with those held in the labour market. Then, presumably, the ERM effect will be wholly ‘benign’, at least in the absence of assymetric shocks which call for changes in real exchange rates. De Grauwe (1988) has made a related relevant point. This is, that a country following an internal standard may be able to make some ‘cheap’ (or at least quick) gains against inflation by letting its currency appreciate; this course will not be open to the country pursuing the external standard – but, by the same token, those quick gains can also be dissipated by depreciation. (The United Kingdom provides a good example.)

The De Grauwe point and the Walters effect point both suggest that the time horizon over which the relative merits of the alternative methods of controlling inflation should be compared may be quite along one – and one thing we do know about the sacrifice ratio is that it is sensitive to choice of period. There may, however, be other explanations we should be looking at. Palle's cross-country regression-based explanation of the sacrifice ratio successfully relates it to initial conditions (it is easier to reduce high inflation than low inflation), and to labour market conditions as reflected in his parameter estimates of the Phillips curve. This part of Palle's exercise is very similar to, though not identical with, attempts that have been made to relate economic performance indicators to measures of wage bargaining structure (e.g., Bean, (1992)). It would have been interesting to see what results Palle would have obtained if he had used, as Bean does, the Calmfors-Driffill ranking of wage bargaining structures. Calmfors and Driffrll (1988) themselves, it may be recalled, find that the performance/centralization relationship is U- or hump-shaped (depending which way round you look at it): that is, performance is good for both highly centralized and for highly decentralized bargaining structures. The poor performers are on the whole those with bargaining structures which are in the middle of the spectrum. In these studies it is unemployment alone, or a measure of performance closely related to unemployment, which is under examination. But much the same findings would probably emerge with respect to the sacrifice ratio. At least we can see from Palle's data that the low sacrifice ratio countries include the highly decentralized United States and Japan as well as the highly centralized Sweden, Denmark and Norway (Palle's Table 5).

It is on the whole the countries ranked by Calmfors and Driffill as ‘intermediate’ – Germany, Netherlands, Belgium, New Zealand, Australia that do worst. Interestingly, Italy which has the lowest sacrifice ratio among the ERM countries is ranked by Calmfors-Driffill as ‘decentralized’. But there are some awkward cases – most notably perhaps Austria (centralized but with a high sacrifice ratio) and the United Kingdom (decentralized but with a particularly high sacrifice ratio) – and some of the judgments change when the output-based measures are used rather than the unemployment-based measure.

In sum, Palle Andersen's paper performs a useful function in providing us with measures of the disinflation cost, and in making an ingenious and thought-provoking attempt to explain differences in cross-country experience. That the jury is still out on some of the answers is a comment on the difficulty of the problems uncovered.


Artis, M.J. and D.A. Currie (1981), ‘Monetary Targets and the Exchange Rate’, Oxford Economic Papers, 33, pp. 176–200.

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

Barro, R. and D. Gordon (1983), ‘Rules, Discretion and Reputation in a Model of Monetary Policy’, Journal of Monetary Economics, 12, pp. 101–21.

Bean, C. (1992), ‘European Unemployment: A Survey’, Centre for Economic Performance Discussion Paper No. 71.

Bruno, M. and J.D. Sachs (1985), The Economics of Worldwide Stagflation, Harvard University Press, Cambridge, MA.

Calmfors, L. and J. Driffill (1988), ‘Bargaining Structure, Corporatism and Macro Performance’, Economic Policy, 6, pp. 14–47.

De Grauwe, P. (1988), ‘The Cost of Disinflation and the European Monetary System’, CEPR Discussion Paper No. 326.

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

Henry, S.G.B., E. Karakitsos, and D. Savage (1982), ‘On the Derivation of the “Efficient” Phillips Curve’, Manchester School of Economic and Social Studies, 50, pp. 151–77.

Walters, A. (1986), ‘Britain's Economic Renaissance, Margaret Thatcher's Reforms 1979–1984’, Oxford University Press, New York.

3. General Discussion

The discussion of Andersen's paper comparing the disinflationary experiences of OECD countries focused on two main areas:

  1. the measurement and interpretation of sacrifice ratios; and
  2. identifying those factors which might reduce the activity and unemployment costs of disinflation.

With regard to sacrifice ratios, several speakers noted that the various measures are highly sensitive to the sample periods for which they are calculated. If too long a period was selected, the measure would risk counting the next recovery, and hence under-estimating the cost of the earlier deflation. There was also considerable discussion of how to separate the influence of longer-run trends, and the problem of structural change. For example, the activity costs of deflation need to allow for changes in potential output. Here the reliability of any linear time trend measure of potential output was questioned. Similarly, sacrifice ratios defined in terms of the degree of slack in the labour market need to measure the natural rate of unemployment, and allow for its influence on the current unemployment rate. The natural rate should not necessarily be assumed to be stationary, particularly if hysteresis effects are important. Sacrifice ratios are increased by such effects, since a temporary recession augments the natural rate of unemployment.

With regard to minimising the costs of disinflation, there was some support for Andersen's observation that the credibility bonus was probably small. Policy maker's credibility was acknowledged as a necessary but not sufficient condition for achieving a better trade-off between activity and inflation. A number of speakers emphasised the need for labour market flexibility if low sacrifice ratios were to be achieved. There was also broad agreement that the sacrifice ratio is non-linear. That is, at high rates of inflation it is possible to reduce inflation by a given amount with little cost in terms of activity and unemployment. But at low rates of inflation the cost of each additional percentage point reduction becomes much higher.

Some participants expressed surprise at Andersen's observation that the speed of deflation did not seem to affect the sacrifice ratio and that, if anything, rapid deflations were less costly. One explanation here is that rapid deflations might cause inflation expectations to adjust more rapidly. If there are to be credibility gains, there may be some benefit in convincing agents with a ‘shock’ effect. In this respect, there was some support for Andersen's view that the more policy levers that could be bought to bear on inflation in addition to monetary policy – such as trade liberalisation, incomes policy and exchange rates – the more ‘successful’ a deflation episode was likely to be.