In Search of Successful Inflation Targeting
Evidence From An Inflation Targeting Index
Author: Yanliang Miao

Contributor Notes

Author’s E-Mail Address: ymiao@imf.org
Publication Date:
01 Jul 2009
eISBN:
9781451872958
Language:
English
Keywords:
WP; IT index
Download PDF (1.4 MB)

In a first attempt to treat inflation targeting (IT) as a continuous variable, we construct IT subindices for 21 full-fledged ITers on three dimensions: flexibility, transparency, and explicitness. Comparing flexibility and transparency we find that (1) the impact of flexibility on both the mean and variation of inflation is more quadratic than that of transparency; (2) after adding the transparency index, the impact of flexibility is no longer significant. The significant and negative association between transparency and the level and variation of inflation is confirmed when we check for robustness by controlling for disinflation stage, subsampling, instrumental variable estimation, and principal component analysis (PCA).

Abstract

In a first attempt to treat inflation targeting (IT) as a continuous variable, we construct IT subindices for 21 full-fledged ITers on three dimensions: flexibility, transparency, and explicitness. Comparing flexibility and transparency we find that (1) the impact of flexibility on both the mean and variation of inflation is more quadratic than that of transparency; (2) after adding the transparency index, the impact of flexibility is no longer significant. The significant and negative association between transparency and the level and variation of inflation is confirmed when we check for robustness by controlling for disinflation stage, subsampling, instrumental variable estimation, and principal component analysis (PCA).

I. Introduction

Inflation targeting (IT) has shown increasing promise as the new generation of monetary policy framework ever since New Zealand first adopted it in 1990. Burgeoning research is recording the practice and reviewing the performance of IT central banks (“ITers” henceforth). The work by Neumann and Hagen (2002), Truman (2003), IMF (2005), Vega and Winkelried (2005), among others, painted more or less a success story for IT.1 In fact, no country has ever dropped IT after adoption, and an increasing number of countries are embarking on IT, despite the ongoing debate and reservations of certain academics.2

With the increasing popularity of IT, a lowest common denominator, if no consensus, of what constitutes IT is forging among central bankers and academics (c.f., section 2). But no international best practice has emerged. In fact, the “flexibility” of IT as a monetary policy framework is often cited as a primary reason for its resilience and viability (e.g., Roger and Stone, 2005; Svensson, 1999).3

It is well recognized in recent literature that IT is multi-faceted strategy and set of procedures and the practices vary across central banks. Economists have not yet agreed on how to label this diversity. Carare and Stone (2006), for example, divided IT regimes into three subcategories: full-fledged IT, implicit IT, and IT lite. They classified the U.S., the European Central Bank (ECB), and Japan as implicit ITers. Truman (2003), in contrast, dubbed the three big economies “G-3”, a special group that are not ITers and yet achieve well-anchored inflation expectations. A closer reading into central bank laws and publications reveals that divergence is the norm rather than exception. This is true even among the more homogenous group of so-called full-fledged ITers. For instance, such countries as Armenia, Colombia, Greece, Iceland, and South Korea write IT explicitly in their central bank acts; many others, like Brazil, the UK, and Israel, have published numerous government decrees regarding IT but fall short of legislating the regime. Thus, there is a de jure and de facto distinction among ITers. There are other important differences in terms of target design, reaction horizon, transparency, and accountability. These differences become more apparent with implicit ITers and central banks having gone through disinflation whilst instituting IT.

While recognizing the vast diversity of IT practices, empirical research continues to treat ITers as homogeneous. Inflation targeting is recorded as a binary variable, i.e., either one or zero, in virtually all econometric work comparing ITers with non-ITers – name just a few, the empirical work by Ball and Sheridan (2005), Mishkin and Schmidt-Hebbel (2007), and IMF (2005). Much as announcing an inflation target does not make a central bank an ITer, there are institutional arrangements and attributes that matter more for inflation performance than does the tag of IT. Yet no research has made an effort to ascertain and quantify, if possible, the impact of such institutional underpinnings.

This research attempts to fill the void by posing and answering key questions of both theoretical and operational relevance to current and potential ITers. First, do different institutional arrangements and attributes matter for the performance of an IT regime? What are the most important ones? Second, how flexible can flexible inflation targeting be in terms of target design, and, is there an optimal degree of flexibility or transparency? Our contribution is twofold. First, it contributes to the growing literature of IT evaluation by treating IT as a continuous variable on important dimensions such as institutional arrangements, target design, and transparency and accountability. It does so by first establishing a framework of constructing an IT index based on an emerging consensus on what constitutes IT. Second, and more importantly, it sheds light on vital policy debate for potential ITers eager to imitate and emulate, e.g., whether to make the regime as explicit as possible in terms of institutional arrangements; whether to make the regime as flexible as possible in deciding on a point target or a target range and choosing between a relatively short and long target horizon; and whether in principle to aim to have as much transparency as possible.

The rest of the paper is organized as follows. Section 2 first provides a brief literature review and discusses the framework of constructing IT indices on related and yet distinct dimensions of the regime. Section 3 tests empirically the validity of the newly minted indices and presents a horse race between flexibility and transparency among all IT regimes. Section 4 provides robustness checks using factor analysis, subsampling, and instrumental variable estimation. We conclude in Section 5 and offer some policy reflections. Tables, figures, and a brief theoretical review of factor analysis are provided in the appendices.

II. Literature Review and the Construction of an IT Index

A. What Constitutes Inflation Targeting?

Even after almost two decades of inflation targeting, confusion persists on exactly how to define the term. Kuttner (2004) noted that there are two alternative ways, not mutually exclusive, to think about IT: the first is in terms of the observed characteristics of the policy framework, and the second is in terms of policy rule. The second approach, spearheaded primarily by Svensson (1997, 1999), treats IT as an optimal policy rule derived from a “reasonably explicit objective function.” Woodford (2004) and Walsh (2002) also described IT in terms of optimal targeting rules.4

Bernanke et al.,(1999), in the first cross-country review of IT practices, advocated that IT is best described as a “framework” rather than a “rule.” In a similar vein, Truman (2003) observed that “inflation targeting in practice involves both more and less than a reaction function characterizing a monetary policy regime”. This research takes the practical angle and views IT as a monetary policy framework. Even from the pure practical perspective, however, prominent writers differ in what constitutes IT:

  • Mishkin (2000) posited that inflation targeting encompasses five main elements: (i) the public announcement of medium-term numerical targets for inflation; (ii) an institutional commitment to price stability as the primary goal of monetary policy, to which other goals are subordinated; (iii) an information-inclusive strategy in which many variables, and not just monetary aggregates or the exchange rate, are used for deciding the setting of policy instruments; (iv) increased transparency of the monetary policy strategy through communication with the public and the markets about the plans, objectives, and decisions of the monetary authorities; and (v) increased accountability of the central bank for attaining its inflation objectives.

  • Truman (2003) summarized four principal elements of IT: (1) price stability as the goal of monetary policy; (2) a numerical target or sequence of targets to make the framework operational; (3) a time horizon over which the target or the targets are to be met, and 4) an associated approach for evaluating whether the objective or objectives have been achieved.

  • Svensson (1997, 1999) originally emphasized the role of inflation forecasting. Today, for Svensson (2007a), the essential elements of IT are three-fold: (a) an announced numerical inflation target; (b) an implementation of monetary policy that gives a major role to an inflation forecast and has been called “inflation-forecast targeting”; (c) and a high degree of transparency and accountability.

It is worth pointing out that the point of deviation of Svensson’s definition is that a single or hierarchical mandate is no longer a necessary condition for IT. Svensson’s related work (1997, 1999, 2001) also showed in detail the consistency of inflation targeting with a dual mandate.

Despite the aforementioned divergence in defining the term IT, a lowest common denominator, if not a consensus, exists for recognized ITers. IT as a monetary policy regime must have a well-defined inflation target, be it a point or a range, with institutional arrangements to support its achievement, and a high degree of transparency and accountability. The centerpiece of IT is to establish and maintain well-anchored inflation expectations via central bank transparency and accountability.

B. A Framework for Constructing an IT Index

Our proposed framework of constructing an inflation targeting index consists of three related and yet distinct dimensions: target design, transparency and accountability, and institutional arrangements. Each of the three dimensions has a corresponding subindex, namely, flexibility, transparency, and explicitness, which in turn is built upon various components of the IT regime.

1. Target Design (IT1 – Flexibility)

The subindex of flexibility is built on four variables, each coded on a scale of 0 (least flexible) to 1 (most flexible).

  • Clarity of target: point targeter vs. range targeter, and the width of the range

A critical issue in coding numerical targets is whether the level of inflation target itself matters. The level of numerical targets for point targeters or the mid-point for range targeters often changes over time (see Table 1 for details). It is especially the case for countries going through disinflation phases. They often have a declining path of inflation targets and sometimes specify only upper but not lower bounds on inflation.5 The flexibility of inflation targeting is not so much about the level of target itself as it is about the latitude within which central banks can operate. We thus focus on the range of the target. For countries where no range is specified, we look at the range within which central banks are exempt from reporting requirements.6

Table 1 records that the target range in practice can be as wide as 5.0 percent (minus and plus 2.5 for Brazil between 2003 and 2006) or as narrow as 0 for a pure point targeter (e.g., Finland and Norway). A larger inflation target range will likely allow central banks more flexibility in taking other objectives into account. We hence code 1 for countries with the widest target range indicating most flexibility; 0 for a strict point target with least flexibility; and x/5 for countries in between with a target range of width x. The theory of IT would suggest that the clearer the target, the more concrete the commitment and the better anchored the inflation expectations. Thus, we expect to find a positive correlation between the clarity of target and the performance of inflation dynamics.

  • Horizon to achieve the inflation target

Publicly declared target horizons of ITers vary from annual to multi year/medium term and from business cycle to indefinite/long term (see Table 1). The reason for such diversity may be that the optimal choice of target horizon is economy-specific and subject to factors such as the underlying transmission mechanism of monetary policy, the magnitude and persistence of shocks, and not least the preference of central bankers.7 Batini and Nelson (2001) defined the optimal policy horizon (OPH) as the time at which it is least costly, for a given loss function, to bring inflation back to target after a shock.8 This definition is the result of standard optimization practice along the lines of Rudebusch and Svensson (1999) and Svensson (1997, 1999). More intuitively, the OPH is the horizon-analogue of the optimal speed of disinflation – the optimal time required for the dissipation of a shock. Operationally, the OPH is given by the number of periods after a shock when inflation is back on target under an optimal rule.

A longer target horizon, ceteris paribus, gives the central bank more flexibility in taking other policy objectives into account without subordinating the inflation objective. We thus code 1 for indefinite horizon, indicating maximal flexibility; 0.67 for business cycle; 0.33 for multi year; and 0 for annual horizon. An annual horizon is often adopted during disinflation phases.

  • Reporting requirements of target breach and escape clauses9

If there is no formal reporting requirement for missing the target range, central banks have more leeway to take into consideration other objectives, such as employment and financial stability. We code 1 for no target breach reporting requirement and 0 for the nine ITers (Brazil, Canada, Iceland, Israel, New Zealand, Philippines, Sweden, Thailand, and the UK) that require formal public explanations for missing the target.

The existence of escape clauses and its potential invoking may also provide central banks with extra flexibility. Escape clauses spell out in advance circumstances under which central banks are exonerated from reporting a breach of the target. Such circumstances include large adverse supply shocks and big adjustments in administered or regulated prices and often dictate a temporary accommodation of inflation disturbances. Setting the inflation target in terms of a measure of core inflation, as was pointed out by Roger and Stone (2005), often serves as an implicit escape clause. But only a few countries set their targets in terms of core inflation, and only Canada, Czech Republic, New Zealand, and South Africa have had explicit escape clauses (see Table 2 for a detailed list). We code 1 if there are escape clauses, explicit or implicit, and 0 otherwise.

The expected sign of flexibility on the mean and variance of inflation is not unequivocal. On the one hand, if inflation stability is already achieved and credibility well-established, central banks can afford more ambiguity in target clarity and slower adjustment to bring inflation back to target. On the other hand, a steadfast point target and a shorter (annual) time horizon to achieve it are often adopted during the disinflation stage. The reason might be that the central bank needs to signal its resolve and establish credibility by reducing ambiguity and increasing accountability. Put simply, there is a potential endogeneity issue between flexibility and credibility (more on this later).

As Svensson (1999) noted, the speed of monetary policy instrument adjustment depends on the degree of flexibility. Flexible IT implies that shocks that drive inflation away from the target should revert at a pace that does not harm real activity too severely. Too fast an adjustment is equivalent to strict IT, which is likely in situations whereby the central bank needs to gain or strengthen credibility. With a fast adjustment, undue real volatility might emerge, whereas in the slow adjustment case either credibility is strong enough that the central bank can reap some benefits of flexibility or the nominal anchor is lost and inflation falls to the expectation trap.

2. Transparency and Accountability (IT2 – Transparency)

There is no consensus on a definition of central bank transparency, let alone its measurement. Posen (2002), for example, presented six practical views of central bank transparency ranging from “reassurance” to “irrelevance.” Geraats (2002) and Eijffinger and Geraats (2006) provided a taxonomy to analyze the transparency of monetary policy along five distinct aspects: political, economic, procedural, policy, and operational transparency. Although very useful, the five-way categorization may not be altogether fitting here. We feel that certain aspects, especially political transparency, stray beyond what most people understand by the term “transparency.” In asking such questions as whether or not there is an explicit numerical target, it in essence applies to a target design issue. In fact, in applying the taxonomy to 37 central banks Crowe and Meade (2008) found that only economic and operational transparency are significant and bear the expected negative sign. Our proposed transparency index for ITers focuses on the forecasting aspects of an IT regime as well as central bank website coverage. Like the flexibility index, it is built on four variables, each coded on a scale of 0 (least transparent) to 1 (most transparent).

  • Number of inflation reports, quantitative inflation forecasts, and publication of fan charts

If, as was argued by Blinder (2004), Woodford (2004) and Svensson and Woodford (2005), modern central banking is about managing expectations, the importance of inflation forecasting cannot be over-emphasized. All ITers now publish inflation reports (see Table 4). But the number (x) and frequency of inflation reports varies from 0 to 4 a year, which is coded accordingly as x/4. Other forwardlooking practices such as publication of quantitative forecasts and fan charts are equally important. They are recorded and coded as taking a binary value of 0 or 1.

  • Central bank website coverage

Central bank websites play an increasingly important role in central bank communications. Heenan, Peter, and Roger (2006) documented central bank website coverage of seven items, namely, IT framework, target details, transmission mechanism, policy instruments, decision-making, policy calendar, and links to materials. We code 1 if a central bank website covers all seven items, and x/7 if it covers only x items.

  • Reporting the interest rate path

Svensson (2007b) advocated the publication of the central bank’s own policy projections – the interest rate path – as further developments to IT. “Not to publish the interest rate forecast would be to hide the most important information,” he proclaimed. But only a few ITers, e.g., the Reserve Bank of New Zealand since 1997, the Norges Bank since 2005, and the Riksbank since 2007, have adopted the practice. As such, it is not yet included as part of the transparency index.

The expected sign on transparency is negative; i.e., the more transparent ITers are, the more superior their performance on managing inflation expectations and thus reducing the mean, variance, and persistence of inflation. There is nonetheless a possibility of too much transparency, but it is doubtful that any central bank has passed that threshold yet.10

3. Institutional Arrangements (IT3 – Explicitness or Strength of Institutional Commitment)

  • Mandate: single/hierarchical vs. dual/multiple

One approach is to record 1 for a single/hierarchical mandate and 0 for a dual/multiple mandate. Another is to follow Cukierman et al., (1992) and have a finer classification on central bank objectives.11 Using the former approach and citing relevant articles from central bank laws, I found in a separate and yet related paper (Miao, 2007) that the mandate does matter for the performance and conduct of monetary policy. To the extent that a binary classification captures the essential difference among different types of central bank mandates, we adopt it in this paper as well.

Adoption of an IT regime entails subordinating exchange rate goals to the objective of price stability. We code 1 for completely free floating, 0.5 for managed floating, and 0 for a peg. But full-fledged ITers have converged in this aspect as well.

  • Central bank independence (instrument independence) and monetary policy decision-making mechanism12

A disproportionate amount of weight should be given to instrument independence given the fact that ITers are converging on aspects such as establishing monetary policy committees (MPCs), setting a single/hierarchical mandate, and intervening less and less in foreign exchange markets. To the extent that the Cukierman index (Cukierman et al., 1992, 2002) already covers important aspects of the central bank mandate and decision-making mechanism, we use the central bank independence index as a proxy for IT3. We obtain central bank independence data from Simon and Guillen (2005) for up to 2000 and from Crowe and Meade (2008) for the year of 2003 and onwards.

The expected sign of the subindex of explicitness is negative. The more explicit ITers become, the stronger the institutional commitment to it, and the better their performance in terms of reduction in the mean, variance, and persistence of inflation.

We could have added an additional subindex ranking the credibility of different IT regimes. It could potentially cover items such as whether inflation expectations are well-anchored or not, whether an ITer has a good track record of meeting targets, and whether a central bank possesses necessary IT infrastructure, e.g., research and forecast capacity. But some of the items appear to be performance criteria themselves. Credibility, even if treated separately from inflation performance, clearly hinges upon attributes that are already included in the subindices of flexibility, transparency and explicitness.

Each of the three subindices takes the form of ITi = Σ ωi * Ii, where ωi is the subjective and mostly equal weight for different attributes. The only exception is that two of the four elements in the flexibility subindex, namely reporting requirements and escape clauses, receive only half the weights assigned to the other two. The reason is based on the observation that they are seldom invoked in practice. Notice that each individual attribute Ii is coded on a scale of 0 to 1. As such, each subindex ITi is normalized to 1 and is comparable across different dimensions. In view of the uncertain impact of target design (flexibility) on inflation dynamics, it may be more meaningful to separate the index of flexibility from that of transparency and central bank independence.

III. Data and Econometric Analysis

The main question we address in this section is whether more explicit (strict) or transparent ITers differ from their more implicit (flexible) or less transparent counterparts? To this end we retrieve panel data in annual frequency for 21 full-fledged ITers spanning the full course of inflation targeting eras, including the stage of disinflation. Before conducting the panel analysis, we plot average inflation and average variation of inflation over the full sample period against our flexibility and transparency indices for selected ITers.13 Figures 1 and 2 demonstrate that both transparency and flexibility are negatively associated with average inflation. Figures 3 and 4 present the scatterplots between the variation of inflation and transparency and flexibility. While the negative association between transparency and inflation variation is salient in Figure 3, the relationship for flexibility appears to be more quadratic. In addition to cross-section data plots, we test in the next subsection the validity of the newly minted indices.

A. The Validity of the Subindices on Flexibility and Transparency

We run parsimonious fixed effects estimations of the following two equations using robust and clustering options and controlling for nonspherical errors. This in effect down-weights sample outliers and gives White-corrected standard errors.

σitπ/σity=αit+δ1*Flexibilityit+countrydummies+timedummies,σitπ/σity=αit+γ1*Transparencyit+countrydummies+timedummies,

where σitπandσity are variations in inflation and growth for country i at time t, derived from the standard deviation of the five-year rolling averages of inflation and growth, respectively.

If the IT indices are well constructed, the relative variability of inflation vis-a-vis growth should be smaller in a more transparent and more stringent IT regime, as the theory of IT would suggest. We thus expect to find a negative coefficient on transparency and a positive one on flexibility. The transparency index indeed carries the expected negative sign (-1.79) and is significant at the 1% level with a t-ratio of -2.96. Barring further refinement of the index and the data on inflation and growth variability, however, the flexibility index has a negative coefficient -0.98 and is almost significant at the 5% level with a t-ratio of -2.04, which is contrary to the prior. This points to the viability of the story that relatively credible and well-established ITers can afford more flexibility such as an indefinite horizon to bring inflation back to target.

An alternative explanation might be that 15 countries in the sample have gone through disinflation phases, in which inflation variation is often large relative to growth variation due to the reduction of inflation from double digits to single digits.14 Rerunning the regression on the subsample consisting of postdisinflation phases does lend partial support to this explanation. It yields a positive coefficient 0.27, although not significant with a t-ratio of 0.55, suggesting that high flexibility is indeed associated with large relative inflation variability.

B. Does IT Matter? Is There an Optimal Degree of Flexibility or Transparency?

Here we investigate inflation dynamics, i.e., the mean (annual level), variance, and persistence of inflation, among different types of ITers. We first fit simple fixed effects estimation of the mean inflation without including control variables X other than our own constructed IT indices.

πit=αi+β*X+δ1*ITit+δ2*ITit2+timedummies,

where π is the inflation rate and X denotes the vector of other control variables.

Preliminary results indicate the following pattern: (1) both flexibility and transparency indices carry negative signs and are both significant at the 1% level (see Table 5 specifications 1 and 5); and (2) the impact of flexibility is more quadratic, with the quadratic term almost significant at 10% level; but transparency does not show signs of diminishing returns (see Table 5 specifications 2 and 6). This pattern of results also holds when we estimate a New Keynesian type Phillips curve after adding output gap and inflation expectation, proxied by lagged inflation, as controls (see Table 5 specifications 3 and 6).

The significant negative sign of flexibility warrants further explanation as the theory of IT would indicate otherwise. As discussed in section 2, the negative correlation between flexibility and mean inflation may simply reflect the fact that well-established ITers, often with lower level of inflation, can afford more flexibility. We therefore regress reduction in inflation on the flexibility index:

Δπit=αi+β*X+δ1*Flexibilityit+timedummies,

where Δπit = πit − πi,t−1 is the increase (negative of reduction) in inflation. This returns a positive coefficient 2.91 and is significant at the 5% level with a t-ratio of 2.61, which confirms the a priori that more stringency is associated with increasing inflation reduction.

We now estimate the equation of inflation variability:

σitπ=αi+β*X+δ3*ITit+δ4*ITit2+timedummies,

where σπ is the variation of inflation and X denotes the vector of other control variables. The variation of growth is included as a control per the theory of inflation-output variability trade-off (Taylor, 1979). The aforementioned pattern of results again holds (see Table 6 specifications 1–3). Note that the effects of transparency become even more pronounced (significant at the 1% level) in reducing inflation variation.

An intriguing finding emerges from fitting the New Keynesian Phillips curve with added flexibility index and its squared term. The relevant coefficients (Table 5 specifications 3 and 4) suggest that the optimal degree of flexibility (Flexibility*) lies around 0.7. This number is derived from first order conditions as Flexibility* = −δ1/2δ2. It corresponds to the flexibility index of IT frontier countries such as Canada (after 1995), Chile (after 2001), New Zealand, Norway, and Sweden (see Table 3).

C. A Horse Race Between Different IT Components

We now run a horse race between the two important components of the IT index: flexibility and transparency. Both indices are included in the same equations for estimating the mean inflation and variation of inflation.

πit=αi+β*X+γ1*Flexibilityit+γ2*Transparencyit+timedummies,σitπ=αi+β*X+γ3*Flexibilityit+γ4*Transparencyit+timedummies,

In the mean inflation equation, flexibility is no longer significant, but transparency remains significant at the 1% level (Table 5 specification 7). In the inflation variation equation (Table 6 specification 3), transparency continues to be significant at the 1% level and with expected negative sign; but flexibility turns barely significant at the 10% level. The sharp contrast demonstrates that all that matters in the aggregate IT index is transparency and accountability.15 The ultimate question is not so much the label of IT as the substance on these important dimensions.

IV. Robustness Checks

A. Principal Component Analysis (PCA)

We have been assigning subjective and mostly equal weights to different components of the flexibility and transparency indices of inflation targeting. For example, the weight somewhat arbitrarily given to target horizon is equal to that of target range and to the combined weight of reporting requirements and escape clauses. There is nonetheless a possibility that certain attributes may be more important than others. Target horizon, for instance, may carry more weight than is assigned given that escape clauses and reporting requirements are seldom invoked and target range is not always binding.

One alternative to adjust for the subjectivity of constructing an index is to let the data decide relative weights. Factor analysis (e.g., Stock and Watson 1989, 2002; Bai and Ng, 2002) enables us to detect the most important components of IT while automatically correcting for potential multicollinearity among the different attributes of the indices. The resulting factors, the principal components, are latent variables that are linear combinations of different underlying attributes. The factor loadings associated with the linear combinations will be alternative weights for the newly constructed indices. Banaian, Burdekin, and Willett (1998) used PCA to examine the role played by 15 of the attributes of central bank independence in the Cukierman index. They found that most appear to have an insignificant or a positive rather than a negative relationship with mean inflation rates.

To conduct the factor analysis, we refine the indices and include annual observations on the following nine attributes: target range, target horizon, reporting requirements, escape clauses, number of inflation reports, forecasts, fan charts, central bank website coverage, and central bank independence. Three eigenvalues of the correlation matrix are greater than one (see the screeplot in Figure 5). The scree test associated with the plot also suggests that the optimal number of common factors is three. We thus retain three principal components, with each one uncorrelated to any other.

The first principal component (Z1) is the one that matters most and is usually referred to as the “general” factor. The factor loadings associated with Z1 are the weights. It is noticeable that roughly equal weights are assigned to different underlying attributes except for reporting requirements and escape clauses (see Table 7). The sum of the loadings (in absolute terms) of these two items, however, is almost equal to that of other attributes, especially those of target horizon and target range. This coincides with our a priori assumption in assigning only half weights to these two attributes in constructing the flexibility subindex.

Regressing both the level and the variation of inflation on Z1, Z2, Z3, and a full set of country dummies, we find that only Z1 is significant at the 5% level and carries the expected negative sign (see Table 8). Not surprisingly, all the explanatory power resides with the first principal component. The drawback with the PCA, however, is that components other than the first, e.g., Z2, Z3, are usually hard to interpret. Thus it might be preferable to work with some related and yet more interpretable variables, F′s, that are linear combinations of the Z′s. F′s become more interpretable in the sense that each F is highly correlated to a specific subset of the underlying attributes. We can therefore employ the “orthogonal rotation” technique and construct three new factors, each significantly associated with certain underlying attributes and yet not correlated with one another.

The factor loadings associated with the three rotated components F1, F2, and F3, and the correlation coefficients between these components and the nine underlying attributes of IT are reported in Tables 9 and 10. It is noticeable that the No. 1 factor is mostly significantly correlated with the following three attributes in the transparency subindex, e.g., number of inflation reports, quantitative forecasts, and publication of fan charts. Using the rotated components F1, F2, and F3 as regressors and including country fixed effects and other controls, we find that F1 is significantly and negatively associated with both the level and the variation of inflation across different specifications (see Table 11). The rotated factor F3, which is predominantly correlated with central bank website coverage, is also significantly and negatively associated with the variation of inflation (Table 11 specifications 3 and 4).

B. Subsampling

As a further robustness check, we divide the full sample of 21 ITers into two groups. The first consists of 13 emerging market economies; the second is composed of 8 industrial countries. The same pattern of results as highlighted in section 3 holds in the subsample of emerging market economies, but not in the group of industrial countries (Table 12). This contrast of results should not be too surprising since the differences of IT practices among emerging economies are much larger than those among industrial countries. Noticeable intragroup heterogeneity withstanding, it is worth emphasizing that there is an even larger intergroup difference in terms of flexibility and transparency. The intergroup heterogeneity also explains the significant deterioration of the goodness-of-fit in the two subsamples compared to the full sample. Precisely because of both the intergroup and the intragroup heterogeneity, however, the search for attributes of successful inflation targeting becomes even more meaningful.

Another caveat is that some ITers (15 of 21) have gone through the process of disinflation while moving toward de jure or more full-fledged IT. To control for potential confounding effects due to disinflation, we add a dummy variable indicating the phase of disinflation. It aims to test the robustness of our story of the irrelevance of flexibility and the importance of transparency in a subsample closer to a natural experiment. The results (Table 13) indicate that, after controlling for disinflation, transparency remains significant in reducing inflation variation across both subsamples and the full sample.

C. Endogeneity

The estimates of the impact of transparency might be subject to endogeneity bias. So might the results for flexibility, although the sources of endogeneity could differ. For flexibility, potential endogeneity might be due to simultaneity or reverse causality. A significant and positive association between inflation dynamics and flexibility can be observed among stable ITers. It can be interpreted either as that flexibility leads to superior performance of inflation dynamics; or, the other way around, that superior track record compensates for the potential credibility loss from increased flexibility. For transparency, the sources are more likely to be omitted variables or unobservables relegated to the error term. To account for the possibility of an endogeneity problem due to either reverse causality or omitted variables, we run two-stage least-squares (2SLS) regressions with heteroskedasticity-consistent standard errors of the following form:

Second Stage:Inflation or Variation of Inflation=δ2SLS[Flexibility]+βX+u; andInflation or Variation of Inflation=δ2SLS[Transparency]+βX+uFirst Stage:Flexibility=α[Instruments]+γX+υ; andTransparency=α[Instruments]+γX+υ

The error terms in the first and second-stage regressions are v and u respectively. X is a set of included exogenous variables, meaning they are exogenous variables that are included in second-stage regressions. Instruments, often dubbed excluded exogenous variables, are adopted to extract the exogenous components of the flexibility and transparency index. Notice that included exogenous variables X can be part of the instruments, but the instruments should have at least one variable that is exogenous and excluded from the second-stage regression.

Potential instruments for flexibility include inflation history (10-year lagged inflation), neighbors’ flexibility, and political economy aspects of central bank operation. Promising instruments for transparency include education (tertiary school enrollment rate), development of stock markets (10-year lagged market capitalization as a percent of GDP), and neighbors’ transparency. Current education level and lagged stock market development are chosen because of the plausible thesis that the more educated the public and the more advanced the financial markets, the more pressure there will be on central banks to be transparent.16 Neighbors’ transparency and flexibility are also potential instruments because they are unlikely to be correlated with omitted variables of a foreign country and yet very likely influenced by a neighboring country’s inflation targeting practice due to peer pressure or spillover effects. In fact, we do observe clustering of IT and similar degrees of transparency and flexibility among “neighbors” (see Tables 3 and 4). Notice that one’s economic neighbor is likely to be, but not necessarily, its geographic neighbor. Factors such as historical and linguistic links and development stage also matter. Taking all these factors into account, we could divide the 21 ITers into five neighborhoods, e.g., developed Anglo-Saxon countries (Australia, Canada, New Zealand, and the UK); Nordics (Finland, Iceland, Norway, and Sweden); emerging Europe (Czech Republic, Hungary, Israel, and Poland); Latin America (Brazil, Chile, Colombia, Mexico, and Peru); and other emerging markets (Korea, Philippines, South Africa, and Thailand).

We can test the validity of the proposed instruments using the over-identifying restriction (OIR) test. The OIR-test has as its null hypothesis that the instruments can be excluded from the second-stage regression. The test yields a Lagrange multiplier test statistic that under the null hypothesis is distributed Chi-squared (m), where m is the number of OIRs. The degree-of-freedom m equals the number of excluded exogenous variables minus the number of endogenous variables included as regressors in the second-stage regression. The results of the second-stage regressions and OIR-tests are reported in Table 14. The proposed instruments for transparency appear to be valid as the null hypothesis can’t be rejected at the 5% significance level.17 The extracted exogenous component of transparency remains significant in reducing both the variation and the level of inflation in virtually all specifications. It still retains the expected negative sign in the one specification that it is not significant. Comparing the coefficients of transparency in Table 14 with those from corresponding specifications in Tables 5, 6, and 13, it appears that the instrumental variable estimation actually strengthens the magnitude of the impact of transparency. Corresponding Hausman tests, however, fall short of proving that the differences are statistically significant.18

V. Avenues for Future Research and Reflections

A. The Credibility and Flexibility Trade-off

A potential trade-off with inflation targeting, as previously mentioned, is that with increased flexibility there is more risk of losing credibility. Less flexibility (more stringency) is often desirable during disinflation phases for the sake of establishing credibility and thus well-anchored inflation. However, the other side of the same coin is that with credibility well-established, central banks can afford and often desire more flexibility in taking other objectives into account. As is observable in the data, target horizons tend to lengthen when inflation stabilizes; whereas annual target horizons are commonly used in disinflation. To put it differently, central banks with well-established IT regimes may face different credibility and flexibility trade-off curves. Follow-up research might plot the track record of meeting targets, a proxy of credibility, against our constructed flexibility index and ascertain whether there is indeed a trade-off, and whether ITers at different stages are facing different trade-off curves.

B. Final Reflections

What is IT? And why move to IT? For the U.S. and some others, the question is whether moving to full-fledged inflation targeting would matter much for the conduct and performance of monetary policy. Put differently, is announcing a target and having a single/hierarchical mandate really necessary for well-anchored inflation expectation? The Fed’s current regime appears to have successfully obtained one of the most important benefits ascribed to a regime based on explicit guidelines. U.S. financial markets and the public do not seem to be overly bothered by the lack of an explicit number, and inflationary expectations are well anchored.19

For emerging ITers, the question is different, but similar policy implications can be drawn from the findings of this study. If, as we have demonstrated, the ultimate hallmark that differentiates a more successful ITer from its less successful counterpart is the subindex of transparency and accountability, then the ultimate question is not so much the label of IT as the substance of increased central bank transparency and accountability.

Appendices

Appendix 1 – Factor Analysis

Let Xit be the observed data for the ith cross-section unit at time t, for i = 1,…, N, and t = 1,…, T. Consider the following factor representation of data:

Xit=∑j=1kλijFjt+εit, Fit=Fjt−1+ut

where λij is a factor loading coefficient associated with factor Fjt.

The number of common factors k is estimated by solving the following optimization problem:

V(k)=minΛ,Fk(NT)−1∑i=1N∑t=1T(Xit−λikFtk)2s.t.F′/T2=Ik

where F is the vector of common factors, Ft = (F1t, F2t,…, Fkt), Λ = (λ1, λ2,…, λN)′, and λ′i=(λi1,λi2,…,λik).

Λ′=FX/T2

In deciding the optimal number of common factors, we can employ criteria involving penalty functions proposed by Bai and Ng (2002). The column components of the Λ′ matrix are the estimated eigenvectors corresponding to the k largest eigenvalues of the T × N matrix XX′.

Once the number of common factors and the eigenvalues are identified, we restrict ourselves to those components that are associated with an eigenvalue greater than one. We then regress this subset of factors against the mean and variance of inflation in the sample. As a next step, we back out the attributes of IT indices from the common factors and re-estimate the equations of mean and variance of inflation.

Appendix 2 – Tables and Figures

Figure 1.
Figure 1.

Inflation and Transparency for Selected ITers

Citation: IMF Working Papers 2009, 148; 10.5089/9781451872958.001.A001

Figure 2.
Figure 2.

Inflation and Flexibility for Selected ITers

Citation: IMF Working Papers 2009, 148; 10.5089/9781451872958.001.A001

Figure 3.
Figure 3.

Inflation Variation and Transparency for Selected ITers

Citation: IMF Working Papers 2009, 148; 10.5089/9781451872958.001.A001

Figure 4.
Figure 4.

Inflation Variation and Flexibility for Selected ITers

Citation: IMF Working Papers 2009, 148; 10.5089/9781451872958.001.A001

Figure 5.
Figure 5.

Scree Plot of Eigenvalues after PCA

Citation: IMF Working Papers 2009, 148; 10.5089/9781451872958.001.A001

Table 1.

Flexibility of IT and Target Design

article image
Sources:Roger and Stone (2005), Mishkin and Schmidt Hebbel (2001), and Truman (2003).

Officially, there is not a range, but deviations of more than 1 percent from target requires an official explanation.

Table 2.

Formal Procedures for Target Breach and Escape Clauses

article image
Sources:Tuladhar (2005), Roger and Stone (2005), and central bank websites
Table 3.

The Flexibility Index of Full-fledged Iters

article image

The overall flexibility index is the average over three items, target horizon, target range, escape clauses and target breach. The code for the last item is the average of the numerical code for escape clause sand target breach from Table 2.

Table 4.

The Transparency Index of Full-fledged Iters

article image
Sources:Roger and Stone, 2005; Heenan, Peter and Roger, 2006; Wyplosz and others, 2003; central bank websites
Table 5.

Mean Inflation and Inflation Targeting Indices

(21 ITers, 1990-2006, t values in italics)

article image

reject rail at 1% significance level

reject null at 5% significance level

reject null at 10% significance level but not. 5%

Table 6.

Inflation Variation and Inflation Targeting Indices

(21 ITers, 1990-2006, t values in italics)

article image

reject null at 1% significance level

reject null at 5% significance level

reject null at 10% significance level but not 5%

Table 7.

Principal Components without Orthogonal Rotation

article image
Table 8.

Princrpal Components and the Level and Variation, of Inflation

(21 ITers, 1990-2006, t values in italics)

article image

reject null at 1% significance level

reject null at 5% significance level

reject null at 10% significance level but not 5%

Table 9.

Principal Components with Orthogonal Rotation

article image
Table 10.

Correlations between the Rotated Factors and the Underlying Attributes

article image
This table presents the correlation coefficients between the three rotated principal components and the nine underlying attributes of IT regimes.
Table 11.

Rotated Principal Components and the Level and Variation of Inflaton

(21 ITers, 1990-2006, t values in italics)

article image

reject null at 1% significance leve1

reject null at 5% significance level

reject null at 10% significance level but not 5%

Table 12.

Mean and Variation of Inflation_Subsanp ling

(21 ITers, 1990-2006, t values in italics)

article image

reject null at 1% significance level

reject null at 5% significance level

reject null at 10% significance level but not 5%

Table 13.

Mean and Variation of Inflation_Controlling for Disinflation

(21 ITers, 1990-2006, t values in italics)

article image

reject null at 1% significance level

reject null at 5% significance level

reject null at 10% significance level but not 5%

Table 14.

Instruments for Transparency and 2SLS Estimation

(21 ITers, 1990-2006, t values in italics)

article image

reject null at 1% significance level

reject null at 5 % significance level

reject null at 10 % significance level but not 5%

References

  • Bai, Jushan, and Serena Ng (2002), “Determining the number of factors in approximate factor models”, Econometrica 70:191–221.

  • Ball, Laurence, and Niamh Sheridan. 2005. “Does Inflation Targeting Matter?” in Bernanke, Ben S., and Michael Woodford, eds., The Inflation Targeting Debate, University of Chicago Press, Chicago.

    • Search Google Scholar
    • Export Citation

  • Banaian, King & Burdekin, Richard C K & Willett, Thomas D, 1998. “Reconsidering the Principal Components of Central Bank Independence: The More the Merrier?,” Public Choice, Springer, vol. 97(1–2), pages 1–12, October.

    • Search Google Scholar
    • Export Citation

  • Batini, N. and E. Nelson. 2001. Optimal horizons for inflation targeting. Journal of Economic Dynamics & Control 25: 891–910.

  • Bernanke, Ben S., Thomas Laubach, Frederic S. Mishkin, and Adam S. Posen. 1999. Inflation Targeting: Lessons from the International Experience. Princeton, NJ: Princeton University Press.

    • Search Google Scholar
    • Export Citation

  • Blinder, Alan S. 2004. The Quiet Revolution: Central Banking Goes Modern. Yale University Press.

  • Blinder, Alan S. and Ricardo Reis, 2005. “Understanding the Greenspan standard,” Proceedings, Federal Reserve Bank of Kansas City, issue August, pages 11–96.

    • Search Google Scholar
    • Export Citation

  • Carare, Alina and Mark R. Stone. 2006. “Inflation targeting regimes,” European Economic Review, Elsevier, vol. 50(5), pages 1297–73, July.

    • Search Google Scholar
    • Export Citation

  • Crowe, Christopher and Meade, Ellen E., 2008. “Central bank independence and transparency: Evolution and effectiveness,” European Journal of Political Economy, Elsevier, vol. 24(4), pages 763–777, December.

    • Search Google Scholar
    • Export Citation

  • Cukierman, A., S. Webb and B. Neyapti. 1992. Measuring the independence of central banks and its effect on policy outcomes. World Bank Economic Review 6, 353–398.

    • Search Google Scholar
    • Export Citation

  • Cukierman, A., G. P. Miller and B. Neyapti 2002, “Central Bank Reform, Liberalization and Inflation in Transition Economies - An International Perspective”, Journal of Monetary Economics, 49,237–264.

    • Search Google Scholar
    • Export Citation

  • Eijffinger, Sylvester C.W. & Geraats, Petra M., 2006,.“How transparent are central banks?,” European Journal of Political Economy, Elsevier, vol. 22(1), pages 1–21, March.

    • Search Google Scholar
    • Export Citation

  • Faust, Jon, and Dale W. Henderson. 2004. “Is Inflation Targeting Best-Practice Monetary Policy?” International Finance Discussion Papers 807, Board of Governors of the Federal Reserve System.

    • Search Google Scholar
    • Export Citation

  • Geraats, Petra M., 2002. “Central Bank Transparency,” Economic Journal, Royal Economic Society, vol. 112(483), pages 532–565, November.

    • Search Google Scholar
    • Export Citation

  • Heenan, Geoffrey, Marcel Peter and Scott Roger. 2006. Implementing Inflation Targeting: Institutional Arrangements, Target Design, and Communications. IMF Working Paper No. 06/278.

    • Search Google Scholar
    • Export Citation

  • IMF (International Monetary Fund). 2005. World Economic Outlook Chapter IV. “Does Inflation Targeting Work in Emerging Markets?”, pp. 161–186. Washington.

    • Search Google Scholar
    • Export Citation

  • IMF (International Monetary Fund). 2006. De Facto Classification of Exchange Rate Regimes and Monetary Policy Framework. Washington.

  • Kuttner, Kenneth N, 2004. “A Snapshot of Inflation Targeting in its Adolescence,” RBA Annual Conference Volume, in: Christopher Kent & Simon Guttmann (ed.), The Future of Inflation Targeting. Reserve Bank of Australia.

    • Search Google Scholar
    • Export Citation

  • Lybek, Tonny and Morris, JoAnne. 2004 “Central Bank Governance: A Survey of Boards and Management.” IMF Working Paper No. 04/226.

  • Miao, Yanliang. 2007. Central Banks’ Law and Its Impact on the Conduct and Performance of Monetary Policy. Typescript. Princeton University.

    • Search Google Scholar
    • Export Citation

  • Miao, Yanliang. 2008. Monetary Policy Committee Decision-making Dynamics—A Political Economy Model. Typescript. Princeton University.

  • Mishkin, Frederic S. 2000. “Inflation Targeting in Emerging-Market Countries.”.American Economic Review. American Economic Association. vol. 90(2). pages 105–109, May.

    • Search Google Scholar
    • Export Citation

  • Mishkin, Frederic S. 2004. “Can Central Bank Transparency Go Too Far?,” NBER Working Papers 10829, National Bureau of Economic Research, Inc.

    • Search Google Scholar
    • Export Citation

  • Mishkin, F.S. and K. Schmidt-Hebbel. 2007. “Does Inflation Targeting Make a Difference?” NBER Working Papers No 12876.

  • Neumann, M.J.M. and J. von Hagen. 2002. “Does Inflation Targeting Matter?” Federal Reserve Bank of St. Louis Review (July/August):127–48.

    • Search Google Scholar
    • Export Citation

  • Posen, Adam S., 2002 “Six Practical Views of Central Bank Transparency,” Institute for International Economics Working Paper.

  • Reinhart, Carmen M. & Kenneth S. Rogoff. 2004. “The Modern History of Exchange Rate Arrangements: A Reinterpretation.” The Quarterly Journal of Economics, MIT Press, vol. 119(1), pages 1–48, February.

    • Search Google Scholar
    • Export Citation

  • Roger, S. and M. Stone. 2005. “On Target? The International Experience with Achieving Inflation Targets.” Working paper 05–163. Washington: International Monetary Fund.

    • Search Google Scholar
    • Export Citation

  • Rudebusch, Glenn, and Lars E.O. Svensson. 1999. “Policy Rules for Inflation Targeting,” in John B. Taylor (ed.), Monetary Policy Rules, University of Chicago Press, 1999.

    • Search Google Scholar
    • Export Citation

  • Simone Polillo and Mauro F. Guillén, 2005. “Globalization Pressures and the State: The Global Spread of Central Bank Independence.” American Journal of Sociology 110(6):1764–1802.

    • Search Google Scholar
    • Export Citation

  • Smets, F. 2003. Maintaining price stability: How long is the medium term? Journal of Monetary Economics 50: 1293–1309.

  • Stock, James H., and Mark W. Watson (1989), “New indexes of coincident and leading economic indicators”, NBER Macroeconomics Annual, 351–393.

    • Search Google Scholar
    • Export Citation

  • Stock, James H., and Mark W. Watson (2002), “Forecasting using principal components from a large number of predictors”, Journal of the American Statistical Association 97:1167–1179.

    • Search Google Scholar
    • Export Citation

  • Svensson, Lars E.O. 1997, “Inflation Forecast Targeting: Implementing and Monitoring Inflation Targets,” European Economic Review 41 (1997), 1111–1146.

    • Search Google Scholar
    • Export Citation

  • Svensson, Lars E.O. 1999, “Inflation Targeting as a Monetary Policy Rule,” Journal of Monetary Economics 43, 607–654.

  • Svensson, Lars E.O. 2001 “Inflation Targeting: Should It Be Modeled as an Instrument Rule or a Targeting Rule?” European Economic Review 46 (2002) 771–780.

    • Search Google Scholar
    • Export Citation

  • Svensson, Lars E.O. 2007a. “Inflation Targeting.” CEPS Working Paper No. 144. Princeton University.

  • Svensson, Lars E.O. 2007b. Optimal inflation targeting: further developments of inflation targeting. In Monetary Policy Under Inflation Targeting, ed. F. Mishkin and K. Schmidt-Hebbel. Santiago, Chile: Banco Central de Chile.

    • Search Google Scholar
    • Export Citation

  • Svensson, Lars E.O., and Michael Woodford. 2003. “Indicator Variables for Optimal Policy,” Journal of Monetary Economics 50, 691–720.

    • Search Google Scholar
    • Export Citation

  • Svensson, Lars E.O., and Michael Woodford. 2005. “Implementing Optimal Policy through Inflation-Forecast Targeting,” in Bernanke, Ben S., and Michael Woodford, eds., The Inflation Targeting Debate, University of Chicago Press, Chicago.

    • Search Google Scholar
    • Export Citation

  • Truman, Edwin M. 2003. Inflation Targeting in the World Economy. Washington, DC: Institute for International Economics.

  • Taylor, J.B. 1979. “Estimation and Control of a Macroeconomic Model with Rational Expectations.” Econometrica 47(5): 1267–86.

  • Tuladhar, Anita. 2005. “Governance Structures and Decision-Making Roles in Inflation-Targeting Central Banks.” IMF Working Papers 05/183. International Monetary Fund.

    • Search Google Scholar
    • Export Citation

  • Vega, M. and D. Winkelried. 2005. “Inflation Targeting and Inflation Behavior: A Successful Story?” Macroeconomics working paper 0502026.

    • Search Google Scholar
    • Export Citation

  • Walsh, C.E. 2002. “Teaching inflation targeting: an analysis for intermediate macro”, Journal of Economic Education, 33(4), pp 333–346.

    • Search Google Scholar
    • Export Citation

  • Woodford, M. 2004. “Inflation Targeting and Optimal Monetary Policy.” Federal Reserve Bank of Saint Louis Review 86 (4):15–41.

1

I am grateful to Alan Blinder, Stanley Fischer, Peter Kenen, and John Londregan for their continued support and valuable inputs. I wish to thank Peter Allum, Karnit Flug, Norbert Funke, Patrick Imam, Weicheng Lian, Jean Noah Ndela, Zhuan Pei, Rodney Ramcharan, Scott Roger, Mark Stone, and especially Andy Berg for helpful comments and suggestions.

1

A notable exception is Ball and Sheridan (2005), in which the superior performance of ITers was attributed to “regression to mean”. However, the authors stated clearly that “our results do not provide any argument against (italics original) inflation targeting, for we have not found that it does any harm”.

2

Sims (2003), for example, suspected that adoption of inflation targeting simply reflected a lack of viable alternatives.

3

The meaning of “flexibility” depends on the context. It can refer to the fact that inflation targeting even in its “strictest” form, with a zero weight on output gap in a quadratic loss function, still takes into account output in its reaction function (Svensson, 1997). In this sense, inflation targeting is always “flexible”; it is never strict inflation only. In a more practice-oriented context, flexibility also refers to the latitude that IT regimes enjoy in bringing inflation on target. The latter interpretation applies in this paper.

4

The word “rule” can be a source of confusion itself, especially when used in the context of inflation targeting. Kuttner (2004) made two useful distinctions. The first is between optimal and ad hoc rules, with the former referring to rules derived from explicit optimization problems and the latter often linked with the Taylor rule or the inflation forecast-based (IFB) rule. The second distinction is between targeting and instrument rules. Targeting rules are specified entirely in terms of the targets of monetary policy (inflation and output); instrument rules are defined in terms of the optimal setting of the monetary policy instrument, typically the short-term interest rate under the central bank’s control.

5

For the case of specifying only the upper bound, it is debatable whether this gives the central bank more or less flexibility. On the one hand, any point below the upper bound is “on target” and thus appears to give ITers more latitude. On the other hand, a clear-cut upper bound implies that central banks must by all means bring inflation under control, indicating a stringent condition. I tend to follow the latter interpretation when coding numerical targets.

6

Roger and Stone (2005) reported that the Bank of England since 2004 has not had an explicit target range but deviations of more than 1 percent from target requires an o¢cial explanation.

7

Using a small estimated forward-looking model of the euro area economy, Smets (2003) found that the optimal policy horizon became longer the greater the weight on other objectives such as minimizing the output gap and interest rate variability; it became shorter the higher the degree of “forward-lookingness” in the economy and the greater the slope of the Phillips curve.

8

Batini and Nelson (2001) also used an alternative definition dubbed “optimal feedback horizon (OFH)”. In contrast to “optimal policy horizon”, which is a result of standard optimization, OFH views targeting expected future inflation simply as setting the policy instrument in response to deviations of future inflation from target. In other words, under OFH the inflation forecast at some specific horizon is a key input into policy-makers’ decisions.

9

These two items are also important in holding central banks accountable.

10

For whether in principle there can be too much central bank transparency, see, among others, Mishkin (2004), “Can central bank transparency go too far?”

11

They assigned a numerical value of 1 if price stability is the major or only objective in the charter, and the central bank has the final word in case of conflict with other government objectives; 0.8 if price stability is the only objective; 0.6 when price stability is one goal, with other compatible objectives, such as a stable banking system; 0.4 if price stability is one goal, with potentially conflicting objectives, such as full employment; 0.20 if no objectives stated in the bank charter; and 0 if stated objectives do not indude price stability.

12

The only two ITers that do not have a full-fledged committee are Israel and New Zealand; but they operate under a shadow committee (Miao, 2008).

13

Brazil, Colombia, and Hungary are treated as outliers and excluded from graphical studies because they had financial crises and run-up inflations in the middle of the sample period.

14

Going through disinflation often entails giving relatively less weight to growth variation. But the variation of inflation is still much larger in magnitude given the double digit level of inflation to start with.

15

A single-minded mandate and central bank independence could also potentially explain the performance and conduct differences between ITers and non-ITers, the focus of my previous paper (Miao, 2007). This paper focuses on attributes that distinguish ITers from each other. Given that ITers converge on mandate (only a few still have dual/multiple mandates), on committee decision-making (all except two), and on other important independence parameters, the independence index is not entered in the horse race.

16

Lagged financial market development is adopted to avoid contemporary correlation between error term and current financial depth because the latter is often cited as an important monetary policy transmission mechanism. A relatively large lag, 10 year, is chosen because of the observed high persistency of financial depth within time spans of a few years.

17

Proposed instruments for flexibility, e.g., 10-year lagged inflation and neighbors’ flexibility, also can’t be rejected under the OIR-tests. However, they appear to be weak instruments and fail to reject the null in both the weak identification and the under identification tests across different regression specifications. The results are thus not reported here but are available upon request.

18

The Hausman test has as its null hypothesis p lim(δOLS − δ2SLS) = 0 and the test statistic H is calculated as H=(δOLS−δ2SLS)2(S.E.(δ2SLS))2−(S.E.(δOLS))2. Under the null hypothesis, H follows a Chi-square distribution with degree of freedom one. The test results should be taken with great caution for two reasons. First, the Hausman test often has a low power as there is no explicit alternative hypothesis. Second, it may be subject to the “small sample” problem here due to limited number of observations.

19

Blinder and Reis (2005), among others, hinted that inflation targeting might institutionalize monetary policy decision-making and thus overcome reliance on personality. Other potential benefits of moving to full-fledged IT include that IT changes the dynamics of inflation and this change of dynamics can’t be otherwise achieved in a non-IT or implicit IT regime.

Same Series

Other IMF Content

Other Publishers

Cover IMF Working Papers
In Search of Successful Inflation Targeting: Evidence From An Inflation Targeting Index
Author: Yanliang Miao
Volume/Issue:
Volume 2009: Issue 148
Publisher:
International Monetary Fund
ISBN:
9781451872958
ISSN:
1018-5941
Pages:
35
DOI:
https://doi.org/10.5089/9781451872958.001