Speech

International Monetary Fund
Given at the IMF Institute's High-Level Seminar on
Implementing Inflation Targets
Washington, D.C., March 20-21, 2000

Let me, first, welcome the participants in the course, many of you friends from earlier meetings in your central banks and in the IMF.

I should start by making it clear, despite any appearances to the contrary, that the IMF is not setting out to be a missionary for inflation targeting. This is intended to be a neutral conference about the merits and difficulties of inflation targeting.

However, it is the case that the inflation targeting approach, which looked very experimental 10-11 years ago, has gained increasing support in recent years and a growing number of countries have adopted it. As you listen to the proceedings and take part in them, you will want to know why. But nothing is necessarily ever the last word in monetary policy, and implementation of this approach has some difficulties, as you will discover during this conference.

I would like to thank the presenters and discussants for agreeing to share their knowledge and experience with us. You have among you an extraordinary range of experts and expertise both on the theory and practice of inflation targeting. On the theory side, Bennett McCallum and Lars Svensson will be presenting the first two papers. And then eminent central bankers from both industrial and developing countries will be discussing their country experiences-Canada, New Zealand, United Kingdom, among the industrialized countries, and Brazil, Israel, and Mexico among the emerging market countries that have moved in that direction. I would also like especially to thank Leo Leiderman for suggesting the idea for the seminar, and Mohsin Khan and the staff of the Institute for organizing it.

Interestingly, this is not the first IMF conference on inflation targeting. For example, we helped organize conferences in Brazil and Colombia as officials from both countries were examining how to introduce inflation targeting. We had a conference in Washington last fall on how to combine inflation targeting with the traditional IMF approaches to monetary policy conditionality. That is an ongoing topic that has not been fully resolved.

*** *** *** *** ***

Let me raise a set of issues and describe what I found in the excellent papers that will be presented at the seminar. The seminar starts with a discussion of alternative monetary approaches in the Ben McCallum paper. You may wonder why a seminar on inflation targeting should be discussing Taylor rules, money-base targeting, and the like. Inflation targeting, as an approach to monetary policy, does not rule out the use of particular reaction functions (which is what is in McCallum's paper) as ways of achieving those goals. It is, alternatively, possible that the approaches that McCallum discusses offer better ways of conducting monetary policy than inflation targeting. So, it is appropriate to begin with a discussion of other approaches, either as a means to keep inflation under control more or less in accordance with the inflation targeting approach, or as an alternative to inflation targeting.

In his comprehensive survey paper, Lars Svensson describes inflation targeting as constrained discretion. That is a very good description of what inflation targeting is about. There is an old debate about monetary policy rules versus discretion. That was before the problem of dynamic inconsistency had been understood. It was very hard to explain the rules versus discretion debate without that theoretical development. At that time participants in the debate—I was among them—had a nagging concern. We could see the benefits of rules. They stopped central banks from making big mistakes. But we seemed to be asking central bankers to throw away the enormous knowledge they had about the way the system worked by giving them very simple rules to implement.

The inflation targeting approach helps clarify the rules versus discretion debate. Inflation targeting works by setting out very clear goals of monetary policy and the framework for implementing it. The framework describes who has what responsibilities, and what accountability and how much transparency there should be. But it leaves to the central bank, to the experts, to actually hit the target, using the instruments the framework permits it to use.

This notion of constrained discretion—of leaving the discretion about how to deploy instruments (instrument independence) to the central bank, while withholding from it the discretion about the goals of policy and the framework in which it is to be made—resolves, in a way that could not have been resolved 30 years ago, what it is that we want central banks to do. It resolves how to combine rules (and what the rules should be about) with the central bank's discretion in the operation of policy.

Noneconomists often accuse economics of being a purely cyclical profession. Issues come and go, but they are never resolved. But that is simply not true, as shown by this example: Our understanding of how to conduct monetary policy and the framework in which it should be conducted has changed fundamentally in past decades as a result of what it is that a science should do: develop the analytic framework and draw on empirical evidence. In the case of monetary policy both of these factors have moved us along, even if we do not yet have final answers.

The bulk of the country papers show that almost all the inflation-targeting countries started from a higher rate of inflation than the steady-state inflation rate target. For the industrialized countries, inflation rates were typically already in the single digits, frequently just a few percentage points above the long-run target. So outlining a declining path of target inflation to reach the long-run target range was not very complicated. Partly as a result of a historical accident—namely that this approach was introduced at a time when inflation was slowing worldwide with or without an inflation targeting approach—some countries got there more quickly than they had expected. However, for the developing countries, many of which had double-digit inflation, figuring out how to get rates down to the target ranges was harder. Many countries avoided setting targets until rates got down to the single-digit range; others declared target inflation ranges for a year or two ahead.

Now another interesting problem has surfaced, one not discussed in the papers: What to do with a country that has too low inflation to begin with? This is, of course, the Japanese problem. And it has been suggested that one of the ways out of Japan's current problems is to institute an inflation targeting framework with a positive inflation rate target. Some people say 4 percent, some people say 3 percent. Given that inflation in Japan has been either negative or barely positive for a few years, getting to 4 percent would be an impressive achievement. And since we need some realism in setting targets, I would say 2 percent.

In general the problem of how to set the target range, given the starting point, is one of the issues to consider in assessing whether this approach could work in a particular country.

The papers address a range of other issues. They discuss not only how to set the target path, but also how to define the target, with the interesting difference between say, the United Kingdom, which specifies a target number a couple of years out, and most other countries, which specify ranges. Thus, whether the target should be a number or a range is another issue to consider.

Either way, given that deflation is bad, you will need some symmetry in the way you approach setting targets, because just as being above target is bad, being below target, whether a range or a number, is also bad. Thus, when the inflation rate goes below the target range, policies are needed to revive the economy and bring inflation back up to target.

Thus questions about target inflation, of number versus range, of symmetry or asymmetry, need careful thought. It is a habit of thought that dies hard that less inflation is always better than more inflation. The first reaction of a central banker is always to recoil in horror at the suggestion of pursuing a higher inflation rate. That is not something we thought about in the past 40 years. But it is necessary now: inflation can be too low, as well as too high.

A related issue (discussed in the Svensson paper, the United Kingdom paper, and elsewhere) is what, precisely, should be targeted? That relates to the issue of whether volatile elements should be included in the inflation target. Should it be a CPI, minus mortgage costs, minus indirect taxes, minus prices that reflect terms of trade shocks? Should something like underlying inflation be targeted? That is one set of questions.

And then there is another, more theoretically interesting and possibly more troublesome, question for countries with less developed financial systems: whether to target an expectation of inflation. Some countries target the expected inflation rate, two years out—that is, today's expectation of what inflation will be in two year's time—at whatever the target rate is, say, 2.5 percent, in the British case. But that is difficult to do without inflation expectation surveys or econometric models that give estimates of what the inflation rate will be. Nevertheless, many analytical problems are solved by targeting the expected inflation rate, providing that the expectation is an official and not wholly a market expectation.¹

The next question, which goes to the heart of the unresolved analytic difficulties with inflation targeting, concerns what Svensson in his paper calls "flexible inflation targeting." Flexible inflation targeting says that there are issues other than inflation that concern monetary policy, and in the model that Svensson describes output is the other concern.

The question is: Should the central bank be a monomaniac (Mervyn King calls it an "inflation nutter") whose only goal is to hit the inflation target, or should it have other goals in mind? In practice, nobody, not even the toughest central banker is only an inflation targeter. Nobody tries to keep inflation on target every minute or even every quarter. If a disturbance comes, nobody says, We should be at 2.5 percent, we are at 3.75 percent, so we are going to jack interest rates up to 50 percent this week until we get inflation down to the goal in the next two months.

Central bankers think about how quickly or slowly they should get rid of inflation. Automatically, that means they are taking into account that there is a short-run, rather than a long-run tradeoff between inflation and output; therefore, they need to consider other factors than inflation in the inflation targeting approach.

Targeting inflation a couple of years out gives authorities time to eliminate the short-run tradeoff or to take it into account. The approach works well as a pragmatic matter, but it is not yet fully integrated into the theory. That theoretical dimension is one aspect that needs further thought. And it is related to the other key tradeoff that many of the countries you will be discussing have faced—the exchange rate- or current account- inflation tradeoff. This issue is discussed in the papers on New Zealand, the United Kingdom, and Canada. It has certainly been a major issue in Israel, as Leo Leiderman will explain.

Many of the countries that have targeted inflation have found themselves with appreciating exchange rates and growing current account deficits. We had a fascinating discussion in the Executive Board of the IMF last week on inflation targeting in Poland, where the current account deficit is getting very large and the Central Bank is clearly uneasy about the measures it should be taking in light of the deficit. Some Board members argued that if you are an inflation targeter you should target only inflation and let the current account deficit go where it may. But a flexible inflation targeter, like Svensson, would argue that other factors also need consideration, as long as that does not threaten the inflation target.

A purist, a constitutionalist, will say that monetary policy deals with inflation, while fiscal policy deals with the current account; so, if the fiscal authorities are not doing their job, that is not the concern of the central banker. And if the economy goes into a crisis, then we will all blame fiscal policy. That is a nice a priori approach, but it does not resolve the issue. This is one of the unresolved issues in this debate.

There is much discussion in the papers about the quality of models needed to predict inflation—particularly among countries moving to inflation targeting, such as Brazil. Many economists feel that unless a country has a good model that accurately predicts inflation it should not adopt inflation targeting because it will be trying to hit a particular target when it does not even have a model that connects the instruments to what is going to happen.

I am not quite sure where this debate stands now, because what you know and what you do not know about monetary mechanism is a fundamental problem for all monetary policy, whether it is inflation targeting, money growth targeting, or whatever. And if in the absence of a good model, a country should not adopt inflation targeting, then what should it do? The proponents of this view may propose adopting a money growth rule. But money growth rules are just another way of making forecasts about what ultimately is the concern—inflation and output growth—and so they do not resolve the issue.

How much do you actually need to know about the monetary mechanism and how good do your models have to be in order to adopt inflation targeting? I do not believe that a very accurate model or a very sophisticated set of expectations is a prerequisite to adopting inflation targeting. Obviously, having them would be preferable. I would add as a footnote that a country is more likely to develop the models once it has an inflation targeting approach that it must implement than if waits for it to arrive. So, I generally do not give as much weight to that problem as most do.

Most of you will be asking yourselves, Could this framework apply to my country? Is there something different about central banking in developing countries that makes inflation targeting, which has worked for Australia, Canada, New Zealand, and the United Kingdom, not applicable in my country? You will need to ask yourself that question. Many of the papers certainly wrestle with it.

I believe that monetary policy should have more or less the same goals in all countries. I also believe that in developing countries many of the tasks given to the central bank, like development banking, would be better put elsewhere. If the functions of a development bank are needed, one should be set up. If subsidies are going to be used, they should be financed through the budget, not the central bank. Mind you, I am not recommending the use of subsidies. I am just saying that if you are going to subsidize, do it through the budget and not through the central bank.

So, I am not convinced that fundamental differences exist among industrialized and emerging market and developing countries on what monetary policy should target. But given the less developed market structures, there are fundamental differences about the routes through which monetary policy operates.

The last issue to consider is how successful has inflation targeting been. Do we really have evidence that it works? After all we are in an era when inflation has come down everywhere. Countries that did not have inflation targeting brought inflation down just as much as countries that did have inflation targeting. The ECB is not an inflation targeter, the Bundesbank is not, and so on.

Some say that no inflation targeting country has had to deal with a full business cycle yet. But we are pretty close to that point now. For example, though New Zealand has not had to deal with a full business cycle, it has had to deal with many disturbances, including the Asian crisis upturns and big current account deficits. The United Kingdom has similarly had to deal with a wide variety of cyclical conditions, as have other inflation targeters. I believe the experience shows that this approach has done well under a variety of circumstances that 10 years ago would have raised legitimate doubts on whether the framework would hold up. But you will need to evaluate for yourself all of the evidence as you listen to these excellent papers and engage in debate with the authors and with each other.

*** *** *** *** ***

Let me again thank you for participating in this event, and hope that you enjoy yourselves as well as learn something in the next couple of days.

---

¹I am grateful to Lars Svensson for reminding me about this qualification.