The objective of this article is to reflect on New Zealand's experience with inflation targeting, to help others evaluate their own policy options.

I am presuming that those interested in inflation targeting have already concluded that monetary policy's proper objective is to provide a nominal anchor, and that inflation targeting is a potential alternative to intermediate targetry for delivering on this objective. Taking these things for granted, I can direct my comments to two issues. First, inflation targets and inflation targeting can be set up in different ways. Second, the institutional arrangements within which the inflation targeting is embedded can also differ. What does the New Zealand experience have to say on both scores?

Case studies often suffer from lack of context. My starting point therefore is to describe how New Zealand got into inflation targeting—which also affords the opportunity for me to dispel some of the misleading images obtained from a reading of the academic literature. I will attempt to describe our approach in terms that allows comparison with others, and describe the evolution in that approach over the 15 plus years that monetary policy has been primarily directed at inflation control, and the 10 plus years that a formal inflation targeting structure has been used. Some evaluation of the approach and outcomes is offered, as well as a discussion of the current issues and challenges.

Before proceeding, however, I would like to draw attention to a selected bibliography attached to this article, where references to far more comprehensive treatments of the issues can be found.

I.  THE ADOPTION OF INFLATION TARGETING¹

Inflation targeting, New Zealand style, has four features—some but not all of which are found in all inflation targeting arrangements:

• A nominal variable (such as the price level or the inflation rate) is recognized as the sole achievable medium term objective for monetary policy.
  
  
• An attempt to drive policy directly at the medium-term objective via a tightly-specified inflation target, rather than indirectly via an intermediate target.
  
  
• An institutional structure that clearly articulates the respective roles and responsibilities of the key actors (the central bank and the government).
  
  
• Heavy reliance on transparency to support the arrangement, and cover the "weak points" in the institutional structure.

It is the presence of all four of these elements together that carves out the New Zealand end of the spectrum of inflation targeters—an end of the spectrum that has attracted a lot of attention amongst academics interested in the interaction of "public choice" incentives and expectations. And one could easily be forgiven for believing that the New Zealand approach was designed in a series of academic workshops.

A recent description of the origins and early development of the inflation target² makes it clear that, while strands from the relevant literature came into the relevant discussions, the adoption of inflation targeting came about more by default than by high design.

As for the adoption of low inflation as the sole medium term objective for monetary policy, New Zealand had had many years of bad experiences with multiple objectives. High average inflation/low average growth outcomes, and significant macroeconomic instability induced by switching between short term nominal and real objectives, had convinced policy makers of two things. First, the notion that short-run Phillips curves can't be exploited in a sustainable way must be right. Second, the existence of competing objectives confounds consistent decision-making, especially if the decision-making is dominated by short-term horizons (as in the politics of election cycles).

It was not a big intellectual step, therefore, to direct monetary policy consistently to inflation reduction—a step which happened in 1984. But the adoption of price stability as the end-point of the inflation reduction process, and the adoption of an explicit inflation target path, came later.

At the outset, simply getting inflation down towards OECD norms was good enough. And to do that, it was envisaged that the combination of a floating exchange rate and a form of money base target would provide a crude but effective mechanism. Rather wild exchange rate and interest rate outcomes promoted a series of re-thinks about the mechanism. Attention progressively moved from financial quantities to financial prices, in recognition of the likely instability in quantity relationships given the substantial financial sector reform. However, rather than having target paths for financial prices conditional on a target path of inflation, we were essentially caught in a no-man's land in which there was no understood and accepted policy reaction function. One could live in such a no man's land while inflation was high enough that "getting it down" was a sufficient specification of the objective; and while a crude judgment as to whether interest rates and the exchange rate were at draconian enough levels was a sufficient definition of the policy reaction function.

In the terminology of today, the situation was crying out for a well-specified policy reaction function. Unfortunately, our understanding of the state of monetary economics made us highly pessimistic that anything sensible would be achievable. When we considered the options, thinking tended towards reconsideration of intermediate targets. But targeting the exchange rate had produced poor outcomes for New Zealand—adjustments of the peg tended to validate past monetary ease—and, as already noted, quantity relationships were highly unstable.

Somewhat surprisingly, the subsequent adoption of an inflation target came not as the resolution to this problem, but instead from another direction. Public sector reform in New Zealand had been driven from the perspective of resolving principal-agent information and incentives problems. From this perspective, a simple recipe had been developed: give public sector managers a clearly specified objective; the authority to manage; and accountability for meeting the objective supported by clean information on outcomes.

Reform of the Reserve Bank had been also under consideration, motivated by the Minister of Finance's desire to provide more independence from politics for the Bank, so as to make reversion to a multiple objective function more difficult. The Minister trusted the Reserve Bank to do the right thing more than he trusted future Ministers of Finance. When reform for the Bank was set within the context of public sector reform, the task became one of finding a clear objective specification in relation to which the Bank could be given authority and be held accountable.

Inflation targeting fell out of this exercise as the least bad of the alternatives available, rather than as the scaling of the intellectual high ground. By the time that inflation targeting was adopted³, the time was ripe for a clearer articulation of the broad objective for monetary policy. Inflation had by then fallen into single figures, and there were limits as to how far the "just get it down" approach could be sustained. But it was one thing to make a choice between a distant objective of reducing inflation to, say, trading partner levels, or to a level broadly consistent with price stability; it was another thing to embody inflation targeting as the organizing construct for the day-to-day operation of monetary policy. The choice of price stability in preference to trading partner inflation came rather quickly, more as a product of the Minister's keenness to shoot for the best outcomes rather than settle for the mediocre, than as a product of lengthy analysis. The choice of inflation targeting came because some specific period-by-period target was required by the public sector reform recipe, and neither monetary aggregate or exchange rate targets were seen as viable.⁴

II.  THE NEW ZEALAND APPROACH

Of the four features of the New Zealand framework mentioned above, three are tightly bound into the legislative structure of the Reserve Bank of New Zealand Act (1989). Interestingly, of the four features, inflation targeting is the one not tightly bound into the statutory arrangements, in a sense that will be explained in a moment.

The RBNZ Act and the associated institutional structure has the following key characteristics:

For monetary policy, the primary objective is established as maintenance of "stability in the general level of prices" (in contrast with the previous Act's objective statement which directed monetary policy to growth, full employment, balance of payments equilibrium and—last on the list—price stability).

To give greater specificity to the price stability objective, the Act requires that the Governor and the Treasurer negotiate and agree a Policy Targets Agreement (PTA)—effectively a contract providing the Governor with, to use the jargon of management consultants, his Key Performance Indicators. Once the PTA is signed, the Governor is free to implement monetary policy without further reference to or instruction from the Government. The Governor, who is the person in whom all the decision-making authority is vested, can be dismissed for inadequate performance in relation to the target set down in the PTA.

However, the Act also provides that the Government can unilaterally over-ride the primary objective (and the PTA entered into on the basis of that objective). Such an over-ride is to be effected by a formal, legal and public process. In the event of an over-ride being invoked, a new PTA must be renegotiated which specifies clearly the new target—that PTA is also a public document. And the Act provides that an over-ride can last for no more than one year before being explicitly—and publicly—renewed.

Thus the legislative framework:

• First, provides a single nominal medium-term objective for monetary policy—price stability—as a matter of law.⁵
  
  
• Second, recognizes that if any choices are to be made on the tradeoffs between medium-term price stability and other medium-term economic policy objectives, those choices are the prerogative of the elected government, consistent with our version of democracy.
  
  
• Third, forces the exercise of judgment on such tradeoffs into public view, making them transparent to the community at large.
  
  
• Fourth, gives the Governor of the Bank the effective independence to implement monetary policy as he or she judges necessary to meet the contracted target and without limitations as to technique, except that decisions "must have regard to the efficiency and soundness of the financial system." (The Bank is thus goal dependent but instrument independent.)
  
  
• Fifth, establishes the Governor's (personal and individual) accountability for decisions on the implementation of monetary policy.⁶
  
  
• Sixth, sets out key mechanisms for achieving the desired accountability, specifically:
  
  

As already foreshadowed, a description of the legislative and institutional framework does not refer to inflation targeting. There is nothing in the framework that requires inflation targeting. While there is a requirement for a Policy Targets Agreement, what that agreement contains is not specified—apart from an admonition that it should be consistent with the statutory price stability objective.

Even though adopting an intermediate targeting approach was deliberately left open as a possibility for the PTA, from the beginning the PTA has been structured around an inflation targeting approach. The main reasons for that were covered in my brief review of how we got into inflation targeting. Of course, it is inflation targeting that is the prime focus of this conference. Just as the New Zealand legislative framework does not require inflation targeting, it is also the case that inflation targeting does not require a formal institutional framework of the New Zealand type now in place—indeed many of the essential elements of inflation targeting pre-dated by some two years the formal legislative framework

But it is naturally an interesting question whether embedding inflation targeting within a well-specified institutional framework makes any difference to the results, and I will make some comments on that issue later. It is equally interesting what difference the particulars of the inflation targeting approach might make, and it is to that issue that I now want to turn.

Which inflation measure?⁸

• Being an average measure, the CPI can get buffeted by relative price movements that do not constitute ongoing, generalized, inflation. Various trend or core measures of inflation are available. For several years, we calculated and focussed our policy discussions around a measure of "underlying" inflation, but questions relating to the methodology of stripping out unusual events tended to cloud the main issues.
  
  
• Our CPI contained, up until recently, mortgage interest rates. A tightening of monetary policy would thus result in an increase in the measured rate of inflation, and vice versa. This was one of the motivations for calculating an "underlying" rate of inflation, a motivation that has disappeared with the removal of interest rates from the index in late 1999.
  
  
• The CPI is an economy-wide measure of prices, and some simulation experiments suggest that a "domestic" or "home" inflation measure might produce superior outcomes (reduced variance in inflation, output and policy instruments). But this result depends significantly on how the exchange rate behaves in response to shocks, the typical nature of those shocks, the way in which prices respond, and may be model specific.

Despite these issues, we have retained consumer price inflation as the focus of the PTA, and indeed have backed away from attempts to "formularize" treatment of the problem events. Familiarity is a significant benefit, and problem events can be handled in other ways, primarily by aiming policy at trend factors and accepting the resulting potential for variations in the quarter to quarter inflation track — variations that need to be explained convincingly of course.

What numerical value should the target have?

There are two components to this question—the location of the central aiming point, and whether a band should be specified (and if so how wide). The latter component is treated later.

The initial target for inflation was set at 0 to 2 percent. At one percent, the central aiming point was understood to be approximately consistent with genuine price stability, after adjustment for biases in the measurement of consumer prices. With the current range of 0 to 3 percent, most observers would still consider that genuine price stability still lies within the target range.

Why aim at genuine price stability, and not something noticeably higher? The initial choice was not made on the basis of large volumes of careful research and reflection. It was instead set, in the spirit of the uncompromising reforms of the time, as simply the "best" aiming point available. There are of course arguments as to why a small positive level of inflation might be more efficient than genuine price stability, and some limited empirical evidence to support that case. On the other hand:

• empirical work suggesting that a small positive level of inflation is beneficial has necessarily drawn on periods where average inflation rates were high, and expectations and behavior probably had adapted to that context;
  
  
• the world has seen long periods of genuine price stability coinciding with growth (inflation is a recent phenomenon);
  
  
• the nominal rigidities that form the basis of many arguments for tolerating a small amount of inflation are, to a significant extent, endogenous; and
  
  
• in New Zealand at least, the issue of a nominal floor on interest rates does not bite (given consistently high real interest rates and the presence of a channel for monetary policy running through the exchange rate).

Re-examining this issue from today's perspective, the increasing number of countries where strong trend growth rates have been happily co-existing with very low rates of inflation suggests that the uncompromising choice was probably not a bad one. Indeed, the issue seems to have virtually disappeared into trivial debates between, for example, those who prefer 0 to 3 percent inflation target bands and those who prefer 1 to 4 percent inflation target bands.

A point target or a range?

We preferred a range, because it provided clear outer bounds to the extent to which inflation could vary without demanding a reaction—or at least a convincing explanation—from the Reserve Bank. Without such bounds, it was felt more likely that the Bank would prevaricate when action was needed, and more likely that expectations would embed such a probability (with resultant costs associated with a higher sacrifice ratio).

Of course, there are problems with ranges. Ranges can be too narrow, both in the sense that frequent inflation outcomes outside the range are likely, and in the sense that monetary policy is forced to become highly aggressive to minimize the frequency of breaches of the range. The former type of problem risks raising doubts about the power of monetary policy, and undermines the credibility of the range boundaries as an accountability device. The latter type of problem raises the important issue of a potential trade-off between inflation variance on the one hand and output and instrument variance on the other. (More on this in a moment.)

It is worth noting, before moving on, that policy-makers have some options in how to react to the prospect of going outside the target range. Having experienced remarkable (and unexpected) success in getting inflation within the 0 to 2 percent range and keeping it there, we came to treat the possibility of a departure from the range as a signal of failure. Recognizing—perhaps belatedly—that it was not sensible to treat the difference between a 1.9 percent inflation rate and a 2.1 percent inflation rate as highly significant, we have shifted our emphasis towards being able to explain adverse outcomes in a way that retains our credibility.

In sum, we are comfortable with the continued use of a range. It has to be acknowledged, though, that the combination of the widening in the target to 0 to 3 percent in 1996, the shift in our treatment of the edge of the range, and more anchored inflation expectations, makes the choice between a point and a range rather less sharp. In this respect, as in others, the distinction between New Zealand's and Australia's approaches to inflation targeting has narrowed.

How forward looking?

• Although monetary policy affects inflation with a (substantial) lag, it is possible to choose to react to current inflation or forecast inflation—in both cases, relative to the target. Both approaches can be shown to be effective in controlling inflation, albeit potentially with implications for the variances of output and the policy instruments (as well as inflation itself).
  
  Both approaches have strengths and weaknesses that depend on the degree and type of uncertainty that policy-makers face in practice. For instance, several central banks react pre-emptively to forestall forecast inflation disturbances, partly on the grounds of uncertainty over the power of monetary policy actions—playing catch-up can be costly. On the other hand, one could interpret the Fed's recent success as being partly related to a tendency-given uncertainty over the evolution of the economy's supply potential—to react more to current indicators of inflation developments.
  
  
• The degree to which policy looks forward can have implications for the aggressiveness of policy.
  
  To illustrate, for a considerable period we focussed on inflation 12 months ahead, which is a time frame over which monetary policy could be expected to have substantial influence, primarily through the direct price effects of exchange rate movements. Because the full effect of monetary policy does not come until later, policy had to act somewhat more aggressively to offset inflation deviations 12 months out. And because the 12 month horizon contains inflation deviations that are temporary, policy can end up reacting to temporary events.
  
  At the same time, the narrow 0 to 2 percent inflation target range—interpreted as applying to the annual inflation rate at all times—motivated a focus on likely inflation developments 12 months ahead. That gave the best chance of nipping deviations in the bud, and remaining within the range.
  
  More recently, we have attempted to look further out, to between one and two years. In essence, that implies focusing more on trend inflation, and not reacting so much to short-term surprises that will cause wider fluctuations in realized inflation outcomes. The widening of the inflation target range to 0 to 3 percent is consistent with the willingness to accept wider fluctuations in inflation outcomes in order to focus more on trend developments.
  
  
• The extent to which policy looks ahead also influences which of the various policy transmission channels are emphasized. As already intimated, when policy in New Zealand focused on inflation developments within the 12 month time frame, most emphasis was placed on the direct price effects of exchange rate movements. Thus influencing the exchange rate's path through actual or threatened adjustments of our policy instruments became a key part of monetary policy implementation. With a longer time frame now in mind, direct exchange rate price effects are de-emphasized relative to real exchange rate and real interest rate channels.

Transparency

Great weight is placed on transparency throughout the design of the framework, and even more so within the practical implementation of monetary policy. Transparency is used as a device to moderate incentive problems that might arise from areas where discretion is allowed; as a key part of the accountability mechanism; and to assist in aligning expectations with policy intent.

Within the formal structure, there are important areas where the exercise of discretion can substantially alter the direction and intent of policy. The Government can over-ride the Reserve Bank, even to the point of directing monetary policy at other than the statutory objective of price stability. And in terms of the Policy Targets Agreement, the Bank can choose not to offset inflation disturbances that will take inflation outside the target range. While there is some guidance as to the nature of the disturbances that the Bank might choose to accommodate, the guidance is very general.

How can one be sure that these "loopholes" do not re-open incentive problems for both politicians and central bankers? At each point where a loophole is opened, there is also the statutory or contractual requirement to advertise publicly using a prescribed procedure what is being done and why. While not foolproof, this approach at least makes it more difficult for the relevant actors to claim consistency of their actions with medium term price stability when in reality there is no such thing.

In relation to accountability, the problem is that performance is hard to assess from inflation outcomes alone. That is especially the case when the effect of today's policy actions are not seen for some considerable time. In order to reinforce the accountability structure, therefore, the Governor—the person being held to account—is required to lay out and explain his intentions and actions quite fully and openly. It is for others, including the financial markets, to assess the quality and validity of those explanations, and from that interpret the underlying intent.

In the context of forward-looking inflation targeting, laying out one's intentions quite naturally resolves into publishing inflation forecasts. We have for some time been publishing detailed inflation forecasts in the same document in which policy decisions are described. Until 1997 those forecasts were of the standard form: projected inflation based on an assumption of no policy change. Observers could easily relate policy decisions to the forecasts in terms of direction, but not in terms of magnitude.⁹ Since 1997, we have been publishing projected tracks for interest rates and exchange rates consistent with achieving our inflation target. The extent of inflationary or disinflationary pressure that we believe will be present over the period ahead can be read directly from the extent to which the policy instruments are projected to adjust. Through this device, we provide a numeric representation of the extent of our policy bias, not only over the period until the next decision point but also over the next 2 years or so.

The obvious advantage of being so explicit in our forecasting, and in the connection of policy actions to those forecasts, is that the relevant actors can learn to anticipate our policy interests. To a significant extent, market prices adjust automatically on the arrival of new information that is relevant for inflation pressures. There are two other advantages worth noting. First, the reasoning behind policy adjustments is easier for observers to see, helping remove residual doubts as to the motivations underlying policy actions. Second, without inflation projections, pre-emptive policy actions would be harder to justify.

On the other hand, there are downsides. Explicit forecasting reveals forecast errors quite directly. Given that forecast errors are commonplace, it can be embarrassing, and there is a danger that confidence in the capacity of the central bank is eroded. We have certainly found it difficult to convey the messages that projections are surrounded by uncertainty, and that the connection between the forecasts and the policy decisions is not mechanical.

III.  EVALUATION

To get use out of an evaluation of the framework, it is worthwhile first to establish which features stand out as different. At the risk of oversimplifying, in the New Zealand framework:

• Inflation targeting is embedded in a more formal institutional structure than is typical, with considerable emphasis on formal mechanisms for accountability.
  
  
• The inflation target is comparatively tightly specified and appears to demand an aggressive policy reaction function.
• Policy is set in a forward-looking framework, with policy actions keying off inflation forecasts.
  
  
• Significant emphasis is placed on transparency, with policy announcements accompanied by detailed inflation forecasts containing an endogenous monetary policy track.

As background to an examination of each of these points, it is worth laying out the numbers. The figure and table below set out the record with respect to inflation, real growth, real exchange rate and the real interest rates, over recent years. The period since 1985 contains the years during which monetary policy has been solely directed at inflation control; the period since 1990 contains the years during which a formal inflation targeting structure has been in use;¹⁰ and the period since 1993 contains the years during which inflation targeting has involved maintenance of low and stable inflation, rather than achievement thereof.¹¹

The most striking feature presented by the figure and table is the success in getting inflation under control, with New Zealand moving from the worst record to the best record in the OECD group (which is true in respect of the average level of inflation as well as its variance). At the same time, the variability of real growth has not much changed in New Zealand, although because a reduction in real variance has been seen elsewhere in the OECD group, New Zealand's relative ranking has worsened.

The picture in relation to interest rates and the exchange rate is a little harder to describe.

Real short-term interest rates in New Zealand have become markedly more stable over time, dropping in variance by around two-thirds, and moving from nearly most volatile to middle-of-the-pack in relative terms. It should be noted, in this context, that absolute and relative interest rate variance was unnecessarily boosted between mid-1997 and early 1999. Over that period we were using a Monetary Conditions Index (MCI) intensively to guide markets as to our likely policy response to exchange rate developments in the period between the quarterly reconsiderations of policy settings. This approach caused a sharp jump in interest rate volatility as high frequency exchange rate volatility was transferred into high frequency interest rate volatility. Thus some of the relative increase in real interest rate variance is attributable to things other than inflation targeting per se.

In respect of the real exchange rate, variance has not changed by much over this period, but in relative terms the real exchange rate has become less stable in the most recent period where it now ranks 4th highest in the OECD group for this measure of variance. This, however, has more to do with the dramatic drop in real exchange rate volatility of European countries en route to EMU—such as Italy, Finland, Spain and Sweden—as it has to do with events in New Zealand.¹⁴

With these pieces of evidence in mind, it is now time to offer some evaluative comments on the New Zealand framework for inflation targeting.

As noted, one of the features of that framework is that inflation targeting is embedded in a more formal institutional structure than is typical, with considerable emphasis on formal mechanisms for accountability. To what extent is the remarkable transformation of New Zealand's inflation record—a more comprehensive transformation that any other OECD country—attributable to these institutional features?

The consensus view amongst my colleagues is that the institutional framework has played a role in shaping the behavior of the monetary policy decision-makers, by making the objective very clear and by making inconsistent policy actions difficult to support. But while the institutional structure had a clear impact on keen observers of monetary policy—including some who would be expected to be relevant to price setting, such as key businessmen and the head of the union movement—judging by measured expectations it played almost no role, independent from the influence on expectations of the outcomes of policy decisions. Inflation expectations in New Zealand seemed to follow actual inflation down with the usual lag, notwithstanding the clear shift amongst leading opinion-shapers in their understanding of likely monetary policy actions. And slow adaptation of the rules-of-thumb that New Zealanders use to assess the real effect of given nominal interest rates certainly made life harder through the 1990s as we struggled to bring a strong demand cycle under control.

How significant the influence of the institutional structure was on decision-making by the central bank is hard to tell. Other countries also succeeded in reducing inflation and inflation expectations over this period, without the buttress of a formal change in institutional arrangements. As for my own judgment, I find it difficult to imagine that we would have been as disciplined were it not for the framework, especially given the background of our history and our own expectations. Our history was of worse inflation than elsewhere, which in turn promised a more difficult and costlier disinflation phase. Put that together with the skepticism that we all shared—though privately—as to whether the inflation target was actually achievable, and one has a recipe for prevarication at the point where decisive action was needed.

This is not to say, however, that in each and every case a tightly structured institutional arrangement, with the full gamut of transparent contracting and accountability, is required for sustained success. In some countries it might be the case that financial discipline and the long-view comes more naturally, and it is sufficient to rely on central bankers to simply "do the right thing."

In addition, in my view the presence of the institutional framework did affect the expectations formation process in a material way—not at the point of initial implementation, but later. To be sure, expectations lagged reality. But inflation expectations continued to follow actual inflation down into territory unheard-of for decades with no increase in the lag, whereas I would have anticipated a lengthening in that lag as price stability was approached. Lest I be misunderstood, expectations still lagged and we have still not fully shaken off the old rules of thumb people use to distinguish real from nominal returns, but there was some moderation at the margin.

The idea that New Zealand's monetary policy became very aggressive with the introduction of formal inflation targeting is widespread. In one sense, the idea is correct. Since the introduction of formal inflation targeting, monetary policy has responded to much smaller perturbations of inflation than had previously been the case. In other important senses, however, it is far from true.

Prior to formal inflation targeting, strongly disinflationary monetary policy generated significantly higher, and more variable, real interest rates than has been the case subsequently. With the move to formal inflation targeting, a dramatic reduction in inflation variability has been achieved with a large reduction in real interest rate variability. I do not mean to imply a wholly causal relationship, but I do mean to imply some causality running from formal inflation targeting to superior variance outcomes. Real exchange rate variance did not fall, but nor did it rise materially following the advent of inflation targeting.

Moreover, the impression that one obtains from looking at an inflation target range of 0 to 2 percent or even 0 to 3 percent, without considering all aspects of the targeting approach, can be misleading. As Lars Svensson has observed—having examined the application of inflation targeting in New Zealand—all inflation targeters including the New Zealanders are better described as "flexible" than "strict" in their approach.¹⁵ The gap between appearances and reality is associated with a number of things including:

• the extent to which the departures of the target variable from the target range can legitimately be accommodated without a monetary policy response (to allow, for example, for the price level effect of supply shocks);
  
  
• the speed with which the authorities attempt to bring inflation back to target after shocks; and, more generally
  
  
• the extent to which forward-looking targeters "look through" short term noise in the inflation rate to medium term trend developments.

Each of these factors is relevant in our context (though in respect of the third factor, with more force now than was the case in the mid-1990s).

What about comparisons of the New Zealand experience through the 1990s with the experience in Australia-comparisons that indicate a larger swing in real interest rates, the real exchange rate, and growth in New Zealand, with similar inflation outcomes? (In respect of real interest rates and growth at least, the same can be said of a comparison between New Zealand and US experience.) And what about research that points to a potential trade-off between inflation, output, and instrument variance?

Clearly, the move to inflation targeting has either reduced variances or left them essentially unchanged in absolute terms. In this important respect, there is no variance trade-off. The move to inflation targeting has been undeniably beneficial. But one has to accept the conceptual case that within the class of inflation targets there may be such a trade-off. The comparison with Australia suggests on the face of it that we could have done still better with a different specification of the inflation target. And the reasons cited earlier for favoring a widening in the inflation target range from 0 to 2 percent to 0 to 3 percent go in the same direction. However, again the New Zealand case offers the potential for a substantial gap between first appearances and the more considered view:

• While New Zealand has experienced increasing variance in the real exchange rate in relative terms, and while that has been associated with a relative (and absolute) reduction in inflation variance, it has not been associated with an increase in the relative (or absolute) variance of real interest rates or output. (Table 1 refers.)
  
  
• When one compares the 1990s in New Zealand with the same period in Australia, our neighbors have clearly had the smoother ride, in real interest rate, real exchange rate, and output terms. And their inflation targeting structure is undoubtedly less tightly specified than ours. But in such evaluations one must use conditional variances. The exogenous shocks to the New Zealand economy were, over this period, clearly larger than was the case in Australia. In such circumstances, had both central banks been running identical inflation targeting approaches, we would expect to have observed precisely the relative ranking of variances that we do in fact observe.
  
  
• On the other hand, monetary policy in New Zealand has progressively attempted to look further forward, triggering monetary policy responses off inflation developments 1 to 2 years ahead rather than the 2 to 6 quarter ahead horizon used in the mid-1990s. As already noted, there is a relationship between the policy horizon and the effective bandwidth of the inflation target range. One might expect on revealed preference grounds, therefore, that some of the observed relative variance ranking vis-à-vis Australia would be associated with too short a targeting horizon in New Zealand during the period.
  
  
• So it looks like a duck, and walks like a duck, but is it a duck? One last bit of confounding argumentation. An examination of the actual conduct of monetary policy through the 1990s shows that there are periods where monetary policy exacerbated the subsequent demand cycle, by not being quick enough to recognize a turning point in the economic cycle. In most cases, these policy problems were attributable in large part to forecast errors that would probably not, in my view, have been corrected by looking further forward. The information available at the time, and our interpretation of it, was captured in the forecasts we published. Playing the what-if game, using these published forecasts, points to the likelihood of bigger policy mistakes as often as it points to the likelihood of smaller mistakes.
  
  

In summary, the conceptual case for the existence of a variance trade-off should be accepted, but one should be careful about jumping to the conclusions that seem immediately to beckon. Adopting inflation targeting is likely to reduce variances across the board, unless the inflation targeting structure promotes extremely aggressive policy responses. Whereas a couple of decades ago we would have thought that inflation target ranges as narrow as 0 to 3 percent were bound to promote overly aggressive responses (even allowing for the "exceptions" clauses noted above), that does not seem to have been the case in practice. We are comfortable with the combination of the 0 to 3 percent range and the longer policy horizon, rather than the previous 0 to 2 percent range and shorter horizon, but that comfort is aided by a sense that inflation expectations are now better anchored.¹⁶ And, to the extent that "unnecessary" variance in output, real interest rates and the real exchange rate has been added by the Bank's policy decisions, there is a plausible case that the main problem was not altering the policy stance early and/or quickly enough.

In essence, looking forward means—or is supposed to mean—reacting to trend inflation disturbances and ignoring transitory ones. The outcome ought to be a smoother path for the instruments and a more wobbly track for inflation outcomes.

There is another obvious connection between forward-looking inflation targeting and the quality of policy outcomes in terms of real economic stability. Monetary policy impacts with a lag. If one does not allow for that lag, and policy reacts instead to revealed inflation outcomes, inflation disturbances are given more time to become persistent. This is especially an issue where expectations are not well anchored and the expectations process is sensitive to perceived policy intentions (in a Barro-Gordon type construct). Stochastic simulations we have done seem to confirm these priors. Forward-looking policy produces a superior inflation-output variance trade-off than backward-looking policy.

However, what about the quality of inflation forecasts? Economic forecasts are notoriously poor, especially over the medium to longer horizons that are most relevant to forward-looking policy. Ours are no exception—as I have already indicated, the policy mistakes that we made over the 1990s were usually associated with forecast errors (not helped by lags in data collection). What impact does the quality of forecasting have on the supposed superiority of forward-looking inflation targeting over backward-looking inflation targeting?

To my mind, uncertainties arising from imperfect forecasts narrow the gap between the two approaches considerably. I would make the following arguments:

• Extracting the gains from looking forward involves making a distinction between events that will have a transitory impact on inflation and events that will have persistence. That is a rather difficult task at the best of times.
  
  
• Moreover, once expectations become well anchored, inflation persistence becomes less of an issue. So long as agents are convinced that the central bank will take the necessary steps to prevent an inflationary or deflationary spiral, forecasting may not add much value.
  
  
• Given the prevalence of errors in reading the current state of the economy, it turns out that the updating of the forward view is driven more by new information on the starting point than it is by the dynamics of adjustment paths within one's model of the economy. That tendency is compounded by innate forecaster conservatism. Forecasters are rarely willing to produce sharp adjustment paths. In other words, the practical difference between forward- and backward-looking approaches is not nearly as large as the conceptual difference.
  
  
• Finally, one might add the point that forecasts almost always only contain information that is already to hand. How often do forecasters assume a future shock (which is almost a contradiction in terms)? Looked at from another perspective, given persistence, current information often contains its own forecast of the future. Thus, even at that conceptual level, the distinction between forward- and backward-looking approaches may not be as large as first seems.

What does one make of all this? First, while the gap between the two approaches might not be as large in practice as in theory, I would still argue a preference for the forward-looking construct. Given the existence of policy lags, so long as forecasts improve the policy response on average, it is better to forecast than to wait for events to reveal themselves. Forecasts are usually wrong in terms of magnitude, but relatively rarely wrong in terms of direction.¹⁷

But I would also conclude that it would be misguided to eschew inflation targeting because one's forecasting capability is ill-developed. New Zealand's early venture into inflation targeting happened with pretty rudimentary forecasting technology, and in circumstances of considerable structural change, which made forecasting even more difficult. And as suggested already, the main issue is to show by one's deeds that inflationary and deflationary impulses will not be allowed to persist.

New Zealand is further down the transparency path than most other inflation targeters, especially as regards the publication of inflation forecasts based on endogenous monetary conditions. Earlier in this article, I made the point that publication of forecasts is a two-edged sword. Publication of projections significantly aids observers' understanding of the policy reaction function but also reveals quite sharply the weakness of forecasting. What advice would I offer others?

The pros and cons of publishing inflation projections are often debated within the Bank. Apart from the arguments already mentioned, the following issues are relevant.

• It is harder for the central bank to "cheat" on its mandate when it is forced to lay out an internally consistent basis for the decisions being made. To be sure, a good publicist can make almost any position sound reasonable, but when it matters financial markets seem to have good noses for spin-doctoring.
  
  
• The central bank's sense of the future path of monetary policy is on display when the inflation target is published. As the Fed and others have found, there is more interest in the likely policy path to follow any given action than there is in the action itself. Projections indicate policy bias quite directly.
  
  
• It is relatively easy to deal with the forecast uncertainty/forecast error problem through careful articulation of uncertainties and the associated judgment s that have gone into policy decisions. I would, however, acknowledge that we have not done a particularly good job in this regard. With its "fan" charts, the Bank of England offers a better model of how to present forecasts with the inherent uncertainties fully on display. And we have, I believe, tended to present the link between policy decisions and inflation projections in too mechanistic a fashion, understating the role of judgment.

IV.  CONCLUDING REMARKS

Our understanding of inflation targeting has come a long way since it was adopted in New Zealand a little over 10 years ago. Inflation targeting is now seen as a conventional choice that is, in many respects, quite similar to other monetary policy arrangements that were previously viewed as stark alternatives. It turns out that most of the alternative monetary policy arrangements place nominal variables at the heart of the medium term objectives; most consider the impact of policy actions on output—variance rather than trend—as an important policy design criterion; and most struggle to find some way of anchoring behavior and expectations while simultaneously allowing an appropriate degree of flexibility for policy to respond to different shocks in a cost-minimizing fashion.

It turns out that inflation targeting is different by degree, rather than by quantum leap. The differences involve the way in which information is considered and processed—inflation targeting being described by some as an information-inclusive approach. Further, it is interesting that one of the places where inflation targeting was seen to be quite different—the explicit structuring of an institutional framework in order to align incentives—is not strictly speaking part of inflation targeting per se. And, at least on the evidence we have available to date, nor is the institutional framework an important determinant of success in inflation targeting.

However, it seems to me that the organizing framework provided by inflation targeting has the wonderful advantage of clarity-of-purpose, with a close alignment of that purpose with what monetary policy can actually be expected to achieve over the medium term. Embedding inflation targeting within a suitable institutional framework, especially one that demands transparency, must be rated as a useful check against slippage—when circumstances are adverse, right-headed decisions are difficult. Inflation targeting appears to be robust to the inflationary pressures that come with strong growth cycles. And, although we haven't been fully tested on this score, inflation targeting appears likely to be more robust than other monetary policy arrangements to serious deflationary pressures.

Selected Bibliography

The following articles contain further information about the monetary policy framework and inflation targeting in New Zealand, and our experience with that framework. Most can all be found in the Reserve Bank of New Zealand's quarterly Bulletin; all are available on the Bank's website at www.rbnz.govt.nz.

Brash, Donald, 1999, "Inflation targeting: An alternative way of achieving price stability". A speech delivered on the occasion of the 50th anniversary of central banking in the Philippines. Reserve Bank of New Zealand Bulletin, Vol. 62, No 1.

Brook, Anne-Marie, Sean Collins and Christie Smith, 1998, "The 1991—97 business cycle in review," Reserve Bank of New Zealand Bulletin, Vol. 61, No 4.

Drew, Aaron and Adrian Orr, 1999, "The Reserve Bank's role in the recent business cycle: actions and evolution," Reserve Bank of New Zealand Bulletin, Vol. 62, No. 1.

Nicholl, Peter and David Archer, 1992, "An announced downward path for inflation," Reserve Bank of New Zealand Bulletin, Vol. 55, No. 4.

Reddell, Michael, 1999, "Origins and early development of the inflation target," Reserve Bank of New Zealand Bulletin, Vol. 62, No 3.

Roger, Scott, 1998, "Core inflation: concepts, uses and measurement," Reserve Bank of New Zealand Discussion Paper G98/9. Sherwin, Murray, 1999, "Strategic choices in inflation targeting: the New Zealand experience," A speech to a seminar on inflation targeting hosted by the IMF and the Central Bank of Brazil, 3-5 May 1999, Rio de Janiero. Reserve Bank of New Zealand Bulletin, Vol. 62, No. 2.

Sherwin, Murray, 1999, "Inflation targeting 10 years on." A speech to the New Zealand Association of Economists conference, July 1999. Reserve Bank of New Zealand Bulletin, Vol. 62, No 3.

Svensson, Lars, 1997, "Inflation Targeting in an Open Economy: Strict or Flexible Inflation Targeting?," Reserve Bank of New Zealand Discussion Paper G97/8.

Other references of interest (also on the website) include:

Briefing on the Reserve Bank of New Zealand, November 1999

Monetary policy under uncertainty. The proceedings of a workshop held at the Reserve Bank of New Zealand, 29—30 June 1998.

Monetary Policy Statement. Various issues.