RBA Annual Conference – 1992 A Perspective Max Corden
1. Some Reflections on the Conference
We have had a set of excellent papers and many wise comments. Two comments sum up the principal issues. ‘Nothing in economics is as simple as we thought it was’ (Glenn Stevens). ‘Any fool, it might seem, can disinflate. The interesting thing is how to minimise the cost of doing so’ (Mike Artis).
(a) How Much Consensus?
One would like to say that there was a wide area of agreement. But in fact, while no-one questions the undesirability of high inflation and of high unemployment, even when the latter is temporary, different views were held as to the emphasis monetary policy should give to the two objectives of reducing inflation and avoiding high unemployment levels. There was by no means universal agreement that the sole target of monetary policy should be to achieve and maintain some low rate of inflation.
But it seems to me that consensus did exist on three crucial matters.
- You cannot disinflate without some cost.
- You cannot expand without some inflation risk.
- There are long and variable lags in the response of the economy to monetary policy changes.
With regard to (i), I did not detect any support for the naive rational expectations view which was briefly fashionable in the seventies that with sufficient (and attainable) credibility a more than negligible cost can be avoided. It is worth remembering that even the ultra-credible Mrs Thatcher could not avoid a significant disinflation cost. With regard to (ii), old-fashioned crude Keynesianism has clearly been killed by history, though one might argue that in current Australian conditions a mild expansion would be unlikely to re-ignite inflation: In practical terms, (iii) represents the biggest problem for policy makers, and makes fine-tuning of any kind so difficult.
Let me now make some brief remarks on three issues discussed in the papers and at the Conference.
(b) The Short-Term Phillips Curve, Hysteresis and Credibility
It is quite clear that in Australia and in other countries there is some kind of non-vertical short-term Phillips curve, and it is certainly expectations – augmented. The existence of such a relationship is supported by the Blundell-Wignall et al, Stevens, Taylor and Andersen papers. As John Taylor pointed out, the expectations are not necessarily backward-looking. Given uncertainty about the future, past experience is bound to have an effect on expectations, and to that extent they may be adaptive. But regime changes will surely be taken into account by reasonably rational agents. The object of policy should be to shift the curve in a favourable direction. Here, Australian economists would be almost unanimous that there is a big role for labour market reform. In a conference focused on monetary policy there was not much discussion of labour market issues, but many participants have been active in this field. Many of us would no doubt agree that real wage levels – not just overall but sectional – as well as various measures that effect labour costs – have a crucial influence on employment for any given rate of inflation. That is certainly my view.
The possibility of hysteresis was referred to several times at the Conference. The basic idea is that a rise in unemployment brought about by macroeconomic policy will increase long-term unemployment, in effect shifting the Phillips curve in an unfavourable direction. It is well accepted that because of effects on expectations, there is inflation-hysteresis: every rise in the rate of inflation brought about by macroeconomic expansion will raise the rate of inflation for given unemployment in the future, so that in the long-run the Phillips curve will be vertical. Once we introduce labour market hysteresis, the long-run Phillips curve becomes difficult to define. A demand contraction will lead to a long-run rise in unemployment. It appears that there is some evidence of hysteresis only in the UK case. It has not been found in the United States or in other European OECD countries. Hysteresis is not apparent in the Australian figures when one compares peaks and troughs of the 1981–83 and 1988–92 cycles. But it is true that the rate of unemployment in the 1976–81 period was bigger than that prevalent before 1974. In any case, if there is significant hysteresis (and this is still an open question), our models and policy conclusions must be changed.
The most popular word in recent macro-economic discussions and theoretical literature is ‘credibility’. It made its appearance frequently at this Conference. The view that monetary policy credibility needs to be established, and that once it is attained it should be maintained by consistent policies, was put most clearly by Charles Goodhart in his very clear paper and presentation. In his view, now that Australia has at last brought the inflation rate down, it would be foolish to throw away the gains by trying to overcome short-term unemployment with expansionary policies that would throw the credibility of the anti-inflation commitment into doubt. In his view, and that of some others at the Conference, monetary policy should make an absolute commitment to an inflation target, and if this commitment is credible, the target could actually be attained with relatively low cost. The theoretical basis for this widely accepted view can be found in articles by Barro and Gordon.
It is now also widely accepted that there are lags in price and wage adjustment that are caused not by a slowness of expectations to adjust to policy changes, but that are institutional in nature, or have other causes. Even the most credible policy change will not lead to immediate adjustments by private agents. Hence it is now widely agreed – universally at the Conference – that there is some inevitable cost of disinflation. But I wish to direct attention to another problem.
The basic idea of credibility is that a policy once established should be firmly maintained. There should be no weakening, or a possibility of weakening, in response to political pressures or short-term problems. This is the firmness that Mrs Thatcher preached and mostly practised. ‘That lady was not for turning.’ She was a most credible decision maker. Yet the policy may turn out to be wrong. Either there are structural changes that alter the model, or new research suggests that the theory is wrong or somewhat faulty. One might argue that policies should be flexible in response to new circumstances – whether shocks coming from outside the country, changes in domestic institutions or responses, or new understanding of issues. I don't have a simple solution for the conflict between the two approaches: one approach is to establish credibility so as to induce private agents to adjust appropriately to a predictable low-inflation environment; the other approach is for policy makers to be flexible so as to allow for new, unexpected circumstances, for the necessary correction of errors, or for new theoretical understanding.
(c) What Should Be the Target of Monetary Policy?
To sort out the numerous issues discussed in the papers and the Conference, let me begin with some taxonomy.
Monetary policy might be targeted purely on a nominal variable or, alternatively, real output and employment might also be taken into account, so that the target is to reach an optimal point on a trade-off curve. Focussing now on the nominal target – whether single-valued or an element in a more complex optimisation objective – it could be an intermediate goal or it could be the ultimate target of low (or zero) price inflation. The intermediate target could be a monetary aggregate of some kind, it could be nominal income, or it could be the nominal exchange rate. Several visitors from Europe (Goodhart, King and Andersen) discussed the implications of the nominal exchange rate target to which members of the European Exchange Rate Mechanism, notably Britain, have committed themselves. The paper by Blundell-Wignall et al. is an important contribution to the practical problems involved when consumer price inflation is targeted directly, which is not far from what seems to be the case in Australia at present. A special feature of the Conference was the extensive discussion of the relevance of asset price inflation, so that the issue arose whether asset prices should be taken into account in the price inflation target, if there is to be one.
I will not try to summarise divergent views, but just give my own conclusions, though I believe they represent the predominant views on the main issues at the Conference.
There is no clear consistent relationship between any money aggregate and nominal income. Indeed, all over the world, crude monetarism that bases itself on money targeting has been discredited by the experience of unstable velocity. On the other hand, money aggregates continue to be relevant indicators, along with others, if a price inflation target is accepted, though such indicators have to be carefully interpreted. The problem about nominal income targeting – which is in principle more attractive – is that short-term changes in real income are difficult to measure and, in any case, can only be measured after the event, with a lag. It also has other problems which I shall not pursue.
Targeting the nominal exchange rate has many attractions. Above all, the exchange rate target is clear cut, very visible, and when a particular bilateral rate – e.g. the dollar or yen rate – is targeted, it can have a lot of credibility. This credibility is greatly strengthened when the target results from some international commitment, such as the European Exchange Rate Mechanism. But there are reasons, to which I come below, why the exchange rate is not a suitable target for Australia. One is thus left, almost by default, with targeting price inflation directly, using variations in nominal aggregates and in the exchange rate (and much else) as relevant indicators to be considered. Even if the minimisation of deviations of real output from trend – or unemployment from some desirable natural rate – were not targeted, or did not enter an optimisation function, they would still be indicators relevant for forecasting future price inflation in the absence of policy changes.
The nominal exchange rate is not a suitable target for Australia for two reasons, the second being the main one.
Firstly, because of Australia's trade pattern, if the Australian dollar were tied to any one of the three major currencies – the US dollar, yen or Deutschemark – there would be undesirable real exchange rate changes brought about by events outside Australia. We do not have a single predominant trade partner, even though Japan is certainly the single most important one. From this point of view our currency should be tied to a trade-weighted basket. Technically that is easy, but it would not carry as much credibility as tying to one currency. It would be too easy to change the composition of the basket. Hence, if a credible exchange rate target were to be achieved, it would be best to fix the Australian dollar either to the US dollar or to the yen, and then live with the inevitable changes in the real exchange rate that would result from US dollar-yen fluctuations.
Secondly, the equilibrium real exchange rate varies for many reasons, notably terms of trade changes and variations in productivity. These variations have been greater for Australia than for most, or all, European countries. With a fixed nominal exchange rate there would have to be large fluctuations in domestic prices and in nominal wages to bring about the required real exchange rate adjustments.
I am aware of an important qualification to this view. Most of the terms of trade changes are temporary and they should be absorbed by variations in the current account – and hence reserves or net capital inflow – rather than by variations in the real exchange rate and the level of absorption. If variations in private sector savings do not bring about the necessary stabilisation through the terms of trade cycle – and usually they do not – fiscal policy would have to do so. Maintaining a fixed nominal exchange rate may be an inducement to follow such a stabilising fiscal policy. But there is still the danger that if the stabilisation policy is not followed, the fixed nominal exchange rate will simply destabilise the domestic price and output levels.
There is a particular point concerning the target rate of price and wage inflation which John Pitchford, Richard Snape and I made at the Conference. The argument leads to the conclusion that the target rate of price inflation should not be too low, and certainly not zero. There will always be a need for relative price changes, sometimes very large ones. More important, there is a need for relative wage changes. Such changes are easier to bring about if the average rate of inflation is significantly positive. I am concerned particularly with wages here. The argument is based on accepting the old-fashioned, but still true, Keynesian proposition that nominal wages tend to be rigid downwards, though flexible upwards.
If changes in relative wages are to be brought about, the average level of nominal wages will need to rise, possibly at certain times very much. Given the minimum average rate of nominal wage increase determined by this consideration, one then gets a minimum average rate of price increase for a given average level of unemployment and given average rate of growth of labour productivity. The higher is labour productivity the less the rate of price inflation needs to be. If labour productivity growth is high, it may be possible to target quite a low rate of price inflation (say 2–3 per cent), while still allowing considerable relative wage changes without any nominal wages having to fall, and without unemployment being generated by the inflation target.
A fascinating and extensive discussion took place at the Conference about the role of asset price inflation. It was generally agreed that asset prices should not be targeted by monetary policy and that they should not be included in the general price index to be targeted. It was also noted by Blundell-Wignall et al. that in Australia asset prices were not good indicators of subsequent price inflation. On the other hand, asset price inflation played an important role in recent Australian developments and – as I note below – there was, I think, an implicit view that the asset price inflation would contribute to an acceleration of price inflation later. The discussion about asset prices raised important theoretical issues. Perhaps the main point was that, ideally, a normal price index to be targeted should include the price of the services of assets, but not the value of the assets themselves.
So much for possible nominal targets. But should real output and employment also be taken into account? Some strong views were put that monetary policy can only be credible and effective if it sends out a clear message: it should be aimed solely at price stability or, at least, a low price inflation target. Yet no one really believes that short-term output effects can or should be ignored. After all, to quote Mike Artis again, ‘any fool, it might seem, can disinflate’. Some massive degree of disinflation can certainly attain the low inflation target and, if the target is 2 per cent or less, it would be difficult to overshoot it very much in a downward direction, given that prices are pretty inflexible downwards. There is a fine tuning problem because policy is inevitably constrained by the danger of a very large short-term deflation effect. Hence, the attention given to effects on output is a matter of emphasis. In my view, in the medium and long run, output and employment depend on the interaction between the average level and the structure of real wages and labour productivity, also taking into account various fiscal measures and additions to real labour costs imposed by government regulations, superannuation contributions and the rest. But in the short run monetary policy is also relevant and cannot be ignored. If monetary policy were to be firmly committed to the anti-inflationary target, a heavier burden would then rest on wage flexibility.
(d) The Costs of Inflation
The paper by McTaggart brought home to the Conference participants how difficult it is to measure the costs of inflation. This does not mean that they are necessarily low, and that there are not many distortions that result from inflation. Much clearly hinges on how the society and its institutions adjust to inflation. Above all, is the taxation system adequately adjusted to inflation (as it has not been in Australia), and is the nominal exchange rate varied appropriately to avoid real appreciation resulting from inflation above that of trading partners? With exchange rate flexibility the costs of inflation are undoubtedly less than when an attempt is made to keep the nominal exchange rate fixed.
Inflation may reduce investment, though there is hardly unambiguous evidence in the Australian case. The recent investment boom was associated with a steady rate of inflation at a level that was fairly high by recent OECD standards, but not high by past Australian standards. In any case, in the absence of the Accord one might have expected inflation to rise because of the investment boom. Eventually the investment boom came to an end – as all such booms do – and inflation declined as a result. But we would hardly attribute the decline in investment to the decline in inflation. The problem for investment and for the efficiency of the economy is uncertainty. It is uncertainty about inflation – and thus the various factors that determine it, notably monetary policy – that is the source of the problem. This uncertainty may be provoked as much by an unexpected disinflation as by an unexpected increase in inflation.
I suspect that steady inflation at a modest rate does not impose big costs, but only a certain amount of inconvenience (aside from the distortions in the tax system), but that there are growing costs if the inflation rate accelerates, and especially if there is some loss of policy control. This is the story of various Latin American countries. As was pointed out at the Conference, the big cost is the cost of disinflation, and inevitably accelerating inflation eventually leads to disinflation. If accelerating inflation is not ended fairly quickly it eventually leads to a loss of control – as evidenced, above all, by Latin American experience. There is a strong case for not setting out on that road: the reversal is too painful, as is the outcome of accelerating inflation.
2. Recent Events in Australia
Finally, I should like to reflect on the current disinflation and recession in Australia.
The episode has similarities with the Volcker disinflation in the United States of 1979 to 1982. In both cases there was a sharp monetary contraction involving a steep rise in nominal and real interest rates. In both cases it led quite quickly to a sharp decline in the rate of inflation and a similar sharp rise in unemployment and decline in real output relative to trend. In other words, in both cases there was some kind of short-term non-vertical Phillips curve. Thus in both cases disinflation was achieved, but at considerable short-term cost. Furthermore, in both cases, the movement down the ‘curve’ was much greater than expected: people were surprised both by the extent of the output and employment declines and by the declines in the rate of inflation. Also, in both cases, there was some modification of the policy once the extent of the recession was appreciated. In the US case this was set off by the developing country debt crisis that suddenly broke with the Mexican debt moratorium in 1982.
We know what happened in the United States once the recession came to an end in November 1982: a steady and quite impressive rate of growth was maintained to the middle of 1990, while the inflation rate stayed down. There were actually two recessions – one in 1980 and another in 1981–82 – but by the end of 1982 it was all over. A highly relevant question is whether the same sort of thing might happen here.
But, now let me note some differences between these two cases. First I shall note differences in the disinflationary process and its motivation, and then in the actual and possible subsequent developments.
In the US case there was a sharp acceleration of the rate of inflation – from about 5 per cent in 1976 to nearly 15 per cent in early 1980 – during the Carter administration. The inflation rate was already high before the second oil shock, on the eve of which it was about 11 per cent. This was associated with depreciation of the dollar. It was highly unpopular in the United States and played a role in President Carter's election defeat in 1980. Both the acceleration of inflation and the public unpopularity of inflation are worth noting, the latter no doubt being caused to a great extent by the former. Some argument can be made that reasonably steady inflation is not a problem, but acceleration is a danger signal that calls for drastic action. Even though the recession was certainly not welcome, it can also be argued that there was an implicit public endorsement in the United States for severe anti-inflationary action.
The Australian case is considerably more complex. Inflation had been fairly steady for some years in the 6–8 per cent range (with a brief higher interlude in 1986 following exchange rate depreciation). At the time, when the deflationary measures were instituted – 1988 and 1989 – there were no clear signs of accelerating inflation, or so it seemed. But I shall come back to this issue shortly. If one takes the view that the principal problem is accelerating inflation, rather than a steady rate of inflation at a fairly modest level, (though higher than the OECD average) one would conclude that costly disinflation was less justified than in the US case. Furthermore, it seems to me clear that the fairly steady rate of inflation of 6–8 per cent or so in Australia was not particularly unpopular in the community. The community would not have chosen a policy that yielded a significant increase in unemployment for the sake of reducing the rate of inflation.
Yet the Australian case is more complex for two reasons, namely the Accord and the asset price inflation, the two features of recent events that have received so much attention at the Conference. There was a demand expansion – an increase in consumption spending and, above all, investment spending – supported by the money and credit expansion, but driven by an ‘animal spirits’ motivated investment boom – which raised output and markedly deteriorated the current account. But it failed to accelerate the rate of inflation because of the Accord. This, indeed, was the great success of the Accord: compared with a likely non-Accord outcome, more of the extra demand went into output and employment growth rather than into wage and price inflation. Of course, it also went into imports. In addition, the monetary expansion must have been an element in the famous asset price inflation. I leave aside the current account issue for the moment, but will put forward two alternative points of view.
One view would be highly critical of the decision to proceed to contractionary policies in 1988 and 1989. Inflation was steady and growth was high. Thanks to the demand expansion combined with the Accord, the unemployment rate was relatively low by the end of 1989. The public was perfectly happy with the situation. Why stir things up, and worry about a perfectly tolerable, if marginally, costly inflation and put an end to this happy equilibrium? There was no public endorsement, explicit or implicit, for even modest deflation, let alone the unexpectedly sharp one that actually resulted. There had been some decline in real wages since 1983 and, in any case, few signs of a wages break-out of the kind that had done so much harm during two earlier episodes.
But there is another view, and I suspect this was the rationale for the policies actually followed. The above view would be acceptable if one could rely on the continuance of the Accord – not just any Accord but one involving continued nominal and real wage moderation without a difficult fiscal trade-off being required. Yet the pressures on the Accord were increasing, it showed signs of weakening and it was highly likely that the increase in profits resulting from the demand expansion would indeed lead to some kind of wages break-out. Furthermore – and here I touch on matters extensively discussed at the Conference – the asset price inflation represented an increase in perceived wealth and this would be bound to lead to further increases in spending shortly, and thus put wage moderation under even greater pressure. In other words, while there was not actually any acceleration of inflation in 1988, all the signs were that it would accelerate shortly. In this view, asset price inflation should not in itself be a target of monetary policy, but it is an indicator of future goods and services inflation, which should indeed be a major, or the main, target. This is not to say that the extent of the disinflation turned out to be correct, even though it did reduce the rate of inflation to a highly desirable 2–3 per cent. But the direction of policy was justified.
Comparing this Australian case with the US case, the clear difference is that in the US case inflation had actually accelerated, so that there was great public concern, while in the Australian case it had not accelerated, so that there was little or no public concern. Yet inflation was very likely to accelerate. This leads to a question which I shall not try to answer: should a central bank disinflate when the public does not think it necessary and is not ready for the costs, even though the bank can see that the current situation is not stable?
By way of an extended footnote, something should be said about the current account issue. The view developed that the larger current account deficit caused by the spending boom was undesirable, and that high real interest rates would reduce it through the obvious mechanism of discouraging investment spending and household borrowing for consumer durables. The problem about this view is that, given floating exchange rates and high capital mobility, a monetary contraction would appreciate the exchange rate and hence have also an opposite effect on the current account. The more internationally mobile is capital the more powerful would be the latter effect. The best instrument for improving the current account (if this were desired) is a fiscal contraction, so that increased public savings or reduced public investment offset reduced private savings and higher private investment. As is well known, fiscal contraction combined with monetary expansion can (allowing for the inevitable lags) improve the current account while maintaining aggregate demand for domestically produced goods and services. On the other hand, it has to be noted that the current account improved between January 1989 and March 1992, moving from 6.6 per cent of GDP to 3.7 per cent of GDP, and this was clearly attributable to the decline in private investment, which in turn was at least partially induced by the higher real interest rates.
Let me come back to the comparison of the Australian case with the Volcker deflation and what followed it. Why was the Volcker deflation followed by a period of low inflation and fairly high growth? It would be nice to know, so that we could reproduce the situation in Australia. Inevitably, I must oversimplify, but it seems to me that there are two elements in the answer. The first has to do with the US fiscal expansion and the second with credibility. We will not wish to reproduce the first, though we would surely like to reproduce the second.
In the US case, there was a notorious fiscal expansion which provided the short-term demand expansion – a Keynesian expansion – that compensated for the contractionary effect of the monetary tightness. In addition, it led to a real appreciation of the dollar that helped to moderate inflation, in effect shifting the short-term Phillips curve in a favourable direction. The appreciation itself had a contractionary effect, but I would guess that the direct expansionary effect of the fiscal expansion was more powerful. In addition, the budget deficit produced a large current account deficit – indeed a complete transformation of the US current account situation. I am sufficient of a neo-classical economist to believe that, over a longer period, wage moderation was a more important element in explaining the growth recovery in the United States than the fiscal expansion, but it would be widely agreed that the real appreciation (which reduced the competitiveness of US industry and generated strong protectionist pressures) contributed to moderating inflation.
In any case, aside from some short-term measures in the recession, we would not want Australia's public sector to get more into debt. Furthermore, some people would also regard a continued or even higher current account deficit as undesirable, even when caused by a private spending boom, let alone a public one. Thus we would not wish to reproduce this aspect of US policy (or non-policy).
The aspect of US experience that Australians might hope to achieve is the credibility of a low-inflation policy – a credibility that was achieved by Mr Volcker backed by the conviction of public opinion, especially as represented in Congress, that the reduction in inflation was a great achievement and should be maintained. Whether this will be achieved in Australia – and without excessive cost – time will tell!