Statistical Implications of Inflation Targeting

6 Information Requirements for Inflation Targeting: Observations Based on the Australian Experience

Carol Carson, Claudia Dziobek, and Charles Enoch
Published Date:
September 2002
  • ShareShare
Show Summary Details
Glenn Stevens

Like a number of other small to medium-sized industrial economies, Australia adopted inflation targeting as its monetary policy strategy during the 1990s. Since the middle of 1993, the stated aim of policy has been to achieve an average inflation rate of between 2 and 3 percent “over the course of the cycle.” Initially, this objective was somewhat less tightly specified than those in some other countries, such as Canada, New Zealand, and the United Kingdom, though the targets in some of those countries are less tightly specified now than they were a decade ago. As it turns out, inflation in Australia since mid-1993 has ranged between 1½ percent and 3¼ percent, and has averaged 2¼ percent. This outcome is consistent with the target (see Figure 6.1).

Figure 6.1.Australia: Inflation

(End-year; percent)

Sources: Australian Bureau of Statistics; Reserve Bank of Australia. CPI, consumer price index.

1Excludes effect of general sales tax.

This chapter will not undertake a detailed description of the target or inflation performance.1 Rather, it uses aspects of Australian experience to address the question, What are the information requirements for operating an inflation target?

At one level, it could be argued that the information needs for inflation targeting are immense. Inflation targeting is a “full information” system: all the information relevant for forming a judgment about the likely path of prices over the coming year or two is to be taken into account. In contrast, a monetary target, at least in principle, requires only the information on money itself. Because in any economy a large number of factors may be affecting particular prices or groups of prices at any one time, the potential dataset to be considered is very large. Certainly, central banks that have been practicing inflation targeting for some time and publish their analysis of the macroeconomy in detail turn out a prodigious volume of material. For example, the latest Inflation Report of the Bank of England and the Statement on Monetary Policy of the Reserve Bank of Australia (RBA) both run to over 60 pages.

At the same time, the main information needs are, in fact, relatively simple. Obviously, a reliable and timely measure of prices is required. Beyond that, inflation targeting consists of forming a judgment about the relative levels, and rates of change, of aggregate demand and supply, taking some account of inflation expectations (and exchange rate changes in open economies), and adjusting policy accordingly so as to bring demand and supply into roughly the appropriate relationship over the ensuing policy-relevant horizon. Conceptually, this is not very complicated.

Inflation targeting is a fairly straightforward approach to policy whose information needs are not all that extensive, if by information we mean statistics. There are a number of issues to consider, to which the next two sections of this chapter are devoted. But it is important not to overestimate the data an economy needs to operate inflation targeting.

A separate, but equally important, issue is, once a central bank has thought about what information it needs to use in pursuing an inflation target, what information should it present to the public? This is taken up in the last section.

Statistical Issues


An inflation target, by definition, has as its centerpiece an announced numerical objective for inflation. Inflation-targeting countries have found that to give concrete expression to the target, to secure public acceptance of it, and to assist in generating a helpful response in inflation expectations, a particular measure of prices needs to be specified as the metric for judging performance. Typically, this is the consumer price index (CPI), or some variant of it. The CPI is conceptually not necessarily the best index, but it is usually the one that is best known, and hence the most credible. This is certainly the case in Australia, where the CPI is widely reported and is used in legal contracts that have indexation clauses, and where concepts such as implicit price deflators, by their very name, are not amenable to easy recognition by the general public, not to mention suffering from the problem of frequent revision.2

In cases where the CPI contains some particularly egregious conceptual flaws—such as measuring the cost of consuming housing services by the interest rate on a mortgage—some form of adjusted CPI is probably to be preferred, despite lesser recognition. This was the case with the Australian target for some years. A rise in interest rates to counter inflation actually made the CPI accelerate in the short term because of the treatment of interest charges. A 100-basis-point change in interest rates moved the CPI by 0.6 percent in the same direction with a lag of about a quarter. For this reason, a measure of core or “underlying” inflation was initially adopted as the target variable. The particular measure chosen was the so-called “Treasury measure,” which had been devised by the Treasury nearly 20 years earlier for other purposes.

From late 1998, the Australian Bureau of Statistics changed its treatment of housing costs, dropping the interest charge component in favor of an acquisitions measure of the cost of house purchase. Because this removed the principal conceptual problem with the CPI, and because the CPI was still better known among the general population than alternative measures, the RBA shifted to targeting the “headline” CPI at that time. The CPI is still decidedly more volatile than core inflation measures, but the bank was comfortable making this change, for two reasons. First, the target is a medium-term one that does not require policy to maintain inflation within a narrow corridor at all times; and, second, by 1998 the policy framework had achieved considerable credibility, which allows a measure of comfort in coping with temporary price fluctuations.

The main point here, however, is that in deciding what index to target, a choice has to be made weighing both conceptual appropriateness and community recognition.

Credibility of the Data

Credibility is likely to be maximized by having the statistics compiled by an independent statistical authority. Enhancing credibility is important, and is probably more important the more unhappy a country’s history of inflation has been. In Australia, it was helpful to be able to use a measure of underlying inflation that had been invented by the Treasury and was calculated and published by the Australian Bureau of Statistics, rather than by the RBA (though the Bank does compute and use a number of core measures; see below).3

Another question under this heading is the frequency of the statistics. Most industrialized countries have monthly CPI data, which is regarded as desirable under the IMF’s Special Data Dissemination Standard (SDDS). Australia and New Zealand are the two industrialized economies whose CPI is issued only quarterly. Both are successful inflation targeters. Hence, high-frequency price data are desirable, though apparently not essential for a country to have. In fact, if a choice has to be made, having better quality is probably preferable to data that are more frequent.

A related observation is that it is useful to have a strong professional relationship with the statistical authority. There is information beyond the purely statistical that a close relationship with the statisticians can provide. This sort of knowledge can help to ensure that the central bank’s evaluation of the data is well informed. In addition, the central bank’s data needs can be taken into account when the statistical bureau makes its own resource allocation decisions.

Measuring Persistent Inflationary Pressures

It is highly desirable for the central bank to be able to abstract, as far as possible, from transitory factors, so as to gauge the pace of persistent price increases. Usually doing so involves decomposing the CPI in some way. In many countries, computation of a CPI excluding food and energy is a quick way of achieving this. The assumption is that food prices are subject to supply-side disturbances that are driven by the weather and so on, and are temporary. This assumption seems defensible, though the issue may be more difficult for developing countries, where food is a large share of the CPI basket. In those cases, care with target design—not pursuing too close a degree of short-run price control—will also be important. The assumption that energy prices are driven by supply shocks rather than demand shocks is probably rather less defensible in recent years: it is fairly clear that swings in global demand have driven the price of crude oil, with the Organization of Oil Exporting Countries’ (OPEC’s) supply changes accommodating the demand swings in an effort to stabilize the price. The other reason for abstracting from energy price changes has been that petroleum has been subject to rather substantial tax imposts in a number of countries, and tax increases do not, in themselves, necessarily signal excess demand pressures.

In Australia, as noted above, a measure of core inflation was used as the target variable. This measure abstracted not only from interest charges, but also from petrol (gasoline) price changes and a host of other price movements deemed to be driven mainly by nonmarket forces (such as, tax increases or administrative pricing decisions). The set of excluded prices amounted to about half the CPI by weight. Looking back, it is remarkable that such a concept was accepted by informed observers as a reasonable gauge of price pressures, but as noted above, the average rate of increase of this series was about the same as that of the CPI over a couple of decades, and the measure had been in existence for a long time and was well recognized by the professional economics community. These factors weighed in favor of adopting the underlying series as the target variable even though other core measures, which retained much more of the CPI regimen, could be constructed.

With the changes to the CPI in 1998, the RBA moved to using the CPI as the target variable. There is, however, extensive use of “core” measures of inflation in analytical and forecasting processes. In fact, without the requirement that inflation-targeting performance be judged by a particular core measure, the Bank has been inclined to expand its use of a suite of core measures, each somewhat different in nature. A particular feature is the way in which exclusion-based core measures of inflation—which take out of the CPI a defined set of items each and every quarter—have been supplemented by measures that exclude or down-weight extreme observations on purely statistical criteria, with the use of trimmed mean and median CPI measures.4 These measures seek to remove inordinately large impacts on the average price level of extreme movements in any price, in both directions. In so doing, they are trying to measure where the bulk of the distribution of prices is moving, on the assumption that this is likely to provide a reasonable gauge of persistent inflationary pressures. They are relatively simple to compute, are symmetric (not only the large price rises are removed, but also the falls), and do not rely heavily on ad hoc judgmental adjustments.

In recent times, the RBA has examined up to four measures of core inflation and then made a general assessment of the current state of core inflation. Doing so prevents the assessment of inflation being unduly affected by the idiosyncratic nature of any particular measure (see Figure 6.2).

Figure 6.2.Australia: Underlying Inflation1

(End-year; percent)

Sources: Australian Bureau of Statistics; and Reserve Bank of Australia. CPI, consumer price index.

1 Excluding general sales tax.

Trade Prices

The bulk of economies pursuing inflation targets have flexible exchange rates, which means that price movements for traded goods and services can on occasion be superimposed onto the price trajectory being driven by domestic forces. Hence, such economies routinely monitor exchange rate developments closely.

Among the group of industrial country inflation targeters from the 1990s, the Australian economy is, in fact, one of the least exposed to foreign trade. Nonetheless, exports and imports are each about 20 percent of GDP. Given that, over the past 15 years, the Australian dollar has moved in a range extending perhaps 15 percent either side of its mean, the effects of exchange rate movements on consumer prices can be significant. Hence for Australia, as for Canada, New Zealand, Sweden, the United Kingdom, and others, gauging the extent of these forces has been important.

Of course, the effects of any given exchange rate change on prices, and its implication for monetary policy, depend heavily on why the change occurred. Exchange rate changes driven by real factors—such as shifts in the terms of trade—may have less impact on prices than changes driven by monetary factors. Whether or not the change is persistent or only temporary also matters. It can further be argued that even a permanent exchange rate change, which permanently alters the price level for trade prices, affects inflation only temporarily, and so may not require any response from monetary policy.

In practice, it is often impossible to identify the originating shock driving an exchange rate change, let alone decide whether or not the change is permanent. But in any event, it is useful to have some way of assessing the impact of such movements on inflation at any given point in time, even if their implications for policy are not clear. Measures of import prices “at the docks,” which are required anyway to compile the national income accounts, are helpful here. In addition, it is desirable to be able to divide final consumer prices between those that could generally be regarded as for tradables and those that could not. The Australian statistical collections allow both, though the division into tradables versus nontradables is inherently somewhat arbitrary. Still, even at a very broad level, we have found this distinction to be helpful.

Broader Information Issues

The above discussion has focused on statistical requirements. At a bare minimum, a country contemplating inflation targeting needs a credible CPI series of reasonable quality, and some basic capacity to dissect it so as to try to assess persistent versus temporary inflation pressures. Desirably, this measure would be compiled by an independent statistical authority. For open economies—that is, most economies—some basic statistical capacity to evaluate the effects of exchange rate changes is also desirable.

Most countries probably already meet these minimum requirements, or could do so without too much difficulty. There is a little more to inflation targeting than just price statistics, however, and this section briefly deals with some of these considerations.

Forecasting Inflation

Ideally, inflation targeting is forward looking. A careful assessment of current inflation should be the basis for forming a view about the likely course of inflation over the policy-relevant horizon, say one to two years ahead. In other words, a forecast of inflation, conditional on various assumptions, should be compiled.

For this purpose, analytical rather than statistical resources are required. Reservations sometimes expressed about the suitability of inflation targeting for developing countries feature concerns about lack of forecasting capacity, perhaps because no fully articulated macro model is available.

It is, of course, true that the better an economy’s structure and dynamics are understood by policymakers, the better policy outcomes are likely to be. Many countries will struggle to develop the degree of sophistication in forecasting inflation that is seen in, say, the Bank of England’s Inflation Report. (Though one could note, contrary to this assertion, that the Bank of Thailand and the Banco Central do Brasil both publish fan charts.)

But one could ask, What do economists know about the behavior of inflation? In a great many instances, the answer is that, much of the time, inflation has considerable inertia—that is, a slow-moving mean—and a cyclical component related to the business cycle, at least insofar as the cycle is driven predominantly by demand-side fluctuations. Open economies experience an effect of exchange rate changes as well, over the policy horizon.5

In Australia, the process of forming a forecast draws heavily on these rather simple ideas. Inflation a year from now is quite likely to be something like today’s inflation rate, plus or minus some components depending on whether the economy seems to be operating with excess demand or excess supply, and some allowance for the exchange rate. Australia’s models of the inflation process are fairly simple; we do not, for example, maintain a large-scale macro model (although we have access to two or three large-scale models maintained by others). The largest system we have has five equations. Two or three different models for inflation are routinely run, each of which has a very simple structure with only a few variables, and the results are compared to get some sense of the range of plausible outcomes. The final forecast presented to the Board of the Reserve Bank of Australia (RBA) and that forms the basis of the bank’s public comments on the inflation outlook is a judgmental one; it is informed by, but is not a slave to, the various model-produced projections.

The RBA’s forecasts are far from perfect, but there is some evidence that the mean square error on the forecasts has been smaller since inflation targeting was adopted than it was before. Of course, this may be either because the Bank works harder at forecasting or because Australian inflation has just been inherently more stable in this period, as it has been worldwide. In reality, both factors are probably at work. It may even be that the inflation-targeting framework itself, by producing a policy response to incipient trends in prices, has helped inflation to remain more stable; that is, after all, the intention. Regardless of the reason, there has not been an occasion since inflation targeting began, when a significant forecast error has produced a serious policy error.

This is partly because the forecasts, while far from perfect, have been acceptable. But it is also because policymakers do not respond only to the particular point forecast for inflation, but also to their informed judgment about the balance of risks to the outlook and to the likely costs associated with the various policy errors they might make.

Responding to the Balance of Risks

It would be a mistake to focus only on the point forecast; it makes much more sense to think of the central forecast as simply the modal point on the distribution of possible outcomes, with a sequence of progressively less likely outcomes on either side. Nor is the distribution necessarily symmetric; it may be skewed one way or the other.

A useful forecast, then, is one that contains not just a central number, but also some sense of the balance of risks. Is it more likely, for example, that inflation will be above the central forecast than below it? The policy implications of such a distribution of risks might be quite different from those stemming from an identical central forecast that had the distribution of risks skewed in the other direction. This is why some central banks have presented their forecasts in terms of fan charts, which are an attempt to quantify both the degree of (im)precision in the forecast and the extent of skewness in the distribution of possible outcomes. Many central banks will not feel that they have the resources to produce fan charts per se but will want nonetheless to assess and articulate the balance of risks, even if only in a qualitative way.

And it is to that balance of risks that policymakers will want, quite properly, to respond. The future is, by definition, uncertain and so policy is never a matter of simply making a point forecast and solving for the “correct” interest rate that goes with that forecast. It is also about thinking through what might go wrong with the forecast, and formulating some notion of what response is appropriate in the event that things do go off track, or even to the risk of such an event.

The policymaker will be seeking an answer to the question, Which mistake, of those I could possibly make, would I regret most? The policymaker is then likely to try to make reasonably sure that that particular mistake is not being made, perhaps at some risk of making another, but less costly, mistake.

Another way of putting this, in slightly more technical language, is that policymaking involves consideration both of the shape of the distribution of forecast outcomes and of the “loss function” that the policymaker is carrying around in his or her head.6

Presentational Issues

Publishing Analysis

The issues of what sort of information needs to be used by a central bank in targeting inflation have been discussed above. An additional question worth asking is what sort of information should be published by the central bank. A notable feature of the public discussion of monetary policy in countries with well-established inflation targets is the “inflation reports” of the central banks.

In fact, these documents are about much more than prices: they typically amount to a comprehensive treatment of the macroeconomy, including real economic activity, the labor market, balance of payments, financial markets, and international events, as well as price developments. Of course, the analysis is usually structured so that the various pieces lead up to an assessment of the outlook for inflation. But the real aim of such documents is not just to make a forecast; it is to show that the central bank has a strong understanding of what is going on in the economy. Ideally, the central bank is trying to demonstrate an unrivalled capacity for macro analysis, and especially analysis of prices, so as to strengthen the credibility of its policy actions.

A greater emphasis on building this credibility has accompanied the adoption of inflation targeting, because the approach necessarily involves being explicit about the objective, and usually is accompanied by sufficient independence for the central bank to move its instrument in pursuit of that objective. Furthermore, in a number of countries that began targeting inflation in the 1990s, the credibility of monetary policy was not high because of a long period of poor performance on inflation and/or the obvious failure of the preexisting rule-like regime. This was certainly so in Australia, where CPI inflation had averaged about 8 percent between 1970 and 1990. New Zealand and the United Kingdom also both had credibility problems as a legacy of the past conduct of monetary policy.

Hence presentation of a detailed, public treatment of the macroeconomy and inflation in particular was a necessary part of the move to inflation targeting. The central bank has to account for how things are traveling relative to the target. When the target is not being achieved, the central bank has to explain why not, and what it is doing about that. Even when the target is being achieved, the central bank has to have a credible story for why its actions will lead to continued good performance in the future.

Ultimately, performance will do the most to enhance (or diminish) credibility. But insofar as the central bank’s words can make a contribution, it should already be clear that a full treatment of all aspects of the macroeconomy in the published material is the way to best contribute to building the kind of credibility the central bank needs. There will be target misses; that is inevitable. The best defense a central bank can have on those occasions, apart from ensuring the episode is shortlived, is having a track record of prior published views that, based on the information available at the time, reasonable opinion had found credible.

Publishing Forecasts

It is hard to avoid publishing some kind of inflation forecast, be it in a tabular form with quarterly projections to the first decimal place, in more qualitative form in words though still reasonably explicit (as is the RBA approach), or in purely graphical form, such as the Bank of England’s fan charts, which do not give an exact central forecast but put the emphasis on the general shape of the profile for forecast inflation and on the extent of uncertainty. Exactly how to promulgate an outlook is a matter for judgment in each individual country, and there are trade-offs to be made. A forecast band wide enough to allow a high chance that the actual outcome will fall within the band can be so wide that, to the ordinary person, not much of value is conveyed. But a precise forecast will almost certainly be wrong, with possible risk of damage to the central bank’s reputation. Careful judgment is all that can be prescribed in dealing with these issues.

One intriguing issue in releasing forecasts of any kind is the extent to which a deviation from the target can be forecast over the medium term. When the forecaster is not the policy advisor, such a forecast can be, and often is, made in an effort to persuade the decision maker to act. If the forecaster is the decision maker, a forecast of a persistent deviation from target seems a logical contradiction. A forecast that inflation will be away from target persistently will prompt the question, Why has policy not already been adjusted to address the problem? Most likely, inflation-targeting central banks that are truly independent will rarely make such a forecast. This of course raises a new risk: that the decision maker will manipulate the forecast to fit the setting of the instrument, rather than the other way around. What guards against this is the openness of the process: the decision maker needs to have a convincing story to back up the forecast.

Another issue is whether to publish extensive forecasts of variables other than inflation. Practice varies, with the RBA not usually making forecasts for, say, GDP growth or unemployment, whereas the Reserve Bank of New Zealand regularly releases model-based forecasts for most variables, including interest rates and exchange rates. Opinions differ here, and the outcome may need to be tailored to individual countries’ circumstances. The objective should be to do what maximizes the effectiveness and credibility of the policy process, rather than to pursue an extreme form of transparency for its own sake.


Inflation targeting does place some information demands on the central bank. A CPI of reasonable quality is essential, as is some capacity to do analysis aimed at separating temporary price shocks from persistent ones. Ideally, the central bank has access to an economy-wide dataset of reasonable quality, so that it can form a judgment about the course of aggregate demand and supply, and hence price pressures. A capacity to form a forecast is also important.

But the information requirements should not be overemphasized. Alternative policy regimes, especially of a discretionary nature, will in most cases have similar requirements. So upgrading the quality of information is a worthwhile goal, irrespective of the framework in use. Many of the concerns over potential information or analytical shortfalls could have been voiced at one stage in the countries that have now been successfully practicing inflation targeting for about a decade. The design of the target can also, to some extent, take account of the information set that is available. If information is less detailed, it may make sense not to promise close control over inflation trends over short periods. In fact, this approach may make sense anyway, regardless of the quality of information.


See Macfarlane (1999) and Stevens (1999).


Even the CPI is not regarded as entirely credible by some sectors of the population. It is not uncommon to hear members of the public say that the prices they themselves face have risen by much more than the quoted CPI rise.


The term “underlying inflation,” incidentally, is probably not a helpful one, as it can leave the impression that the inflation rate the authorities are targeting always “lies under” what is really happening. This was not true in Australia—the underlying CPI had a long-run average rate of increase about the same as the CPI, which helped to deflect this criticism—but on reflection, “core inflation” or “ongoing inflation” are probably superior labels.


For countries with very high inflation, or hyperinflation, the dynamics are of course different: inflation can rise (or fall) much more quickly, and is much more volatile. But inflation targeting is probably not well suited to these cases.


As Alan Greenspan said: “The center of the forecast distribution, of necessity, is still important to our deliberations but, more than many people realize, policymaking is to a substantial extent focused on the potential deviations from the central forecast and the costs should those outcomes prevail. In short, our policy behavior is the result of examining the implications of the interaction of probability distributions and loss functions. We do not engage in the formal mathematics of such a model, of course, but we do follow its underlying philosophy” (Greenspan, 2001).

    Other Resources Citing This Publication