Words vs. Deeds
What Really Matters?

This paper revisits the link between the nominal exchange rate regime and inflation, based on a sample of 145 emerging market and developing countries (EMDCs) over the period 1980-2010. We contend that, just as a de jure peg that is not backed by a de facto peg will have little value, de facto pegs that lack the corresponding de jure will likewise reap few of the low inflation benefits associated with pegging the exchange rate. To test our hypothesis, we exploit a novel dataset of both de jure and de facto exchange rate regime classifications. We find that pegged exchange rates are associated with significantly lower inflation in EMDCs than flexible exchange rates, and that this effect is much stronger for de facto pegs that are matched by de jure pegs than for those that are not. When it comes to anchoring expectations and delivering low inflation, therefore, both deeds and words matter.

Abstract

This paper revisits the link between the nominal exchange rate regime and inflation, based on a sample of 145 emerging market and developing countries (EMDCs) over the period 1980-2010. We contend that, just as a de jure peg that is not backed by a de facto peg will have little value, de facto pegs that lack the corresponding de jure will likewise reap few of the low inflation benefits associated with pegging the exchange rate. To test our hypothesis, we exploit a novel dataset of both de jure and de facto exchange rate regime classifications. We find that pegged exchange rates are associated with significantly lower inflation in EMDCs than flexible exchange rates, and that this effect is much stronger for de facto pegs that are matched by de jure pegs than for those that are not. When it comes to anchoring expectations and delivering low inflation, therefore, both deeds and words matter.

I. Introduction

As the world economy recovers from the global financial crisis, inflationary pressures—driven in part by surges of capital inflows and higher food and fuel prices—are again raising concerns in emerging market and developing countries (EMDCs). Many of these countries have achieved policy credibility only in recent years, and the inflationary pressures risk reigniting inflationary expectations. While there may be many ways of dealing with such pressures, EMDCs have often relied on pegging their exchange rate to help maintaining price stability and anchor expectations. Yet this strategy is not uncontroversial, with existing studies offering very mixed evidence on whether pegging the exchange rate is associated with lower inflation, especially in emerging market (as opposed to developing) countries.

In this paper, we revisit the link between the nominal exchange rate regime and inflation performance. We argue that the varied findings of previous studies can be reconciled by the different impact on inflation that ad hoc interventions in the foreign exchange market (a de facto peg), and a formal commitment by the central bank to maintain the parity (a de jure peg), are likely to have. The early literature on this subject (Ghosh et al., 1997; Ghosh, Gulde, and Wolf, 2003) uses de jure classifications—the central bank’s declared exchange rate regime—and finds a strong statistical association between pegged exchange rate regimes and low inflation. The explanation is that pegging the exchange rate both instills monetary discipline (limits central bank credit expansion) and engenders confidence in the currency (raises money demand and thus lowers inflation for a given rate of monetary expansion). But subsequent papers (Levy-Yeyati and Sturzenegger 2001; Rogoff et al. 2003; Reinhart and Rogoff, 2004), using de facto classifications (that is, based on the actual behavior of the exchange rate) generally find a much weaker association between pegging the exchange rate and low inflation. A natural interpretation of the different findings across studies is that it is “deeds” (the de facto behavior of the exchange rate) rather than “words” (the de jure commitment) that matter.

We offer a slightly different interpretation. While acknowledging that words not backed by deeds will have little value, we contend that de facto pegging the exchange rate without a corresponding de jure commitment will likewise reap few(er) of the low inflation benefits of pegging. Our premise is based on the observation that divergences between de jure and de facto classifications reflect two distinct phenomena. The first is “soft” pegs, whereby the central bank is unable or unwilling to fulfil its commitment to maintain the parity: despite the de jure peg, the exchange rate de facto floats. The second is “closet” pegs—where the central bank intervenes in the foreign exchange market to keep the rate roughly constant, but makes no formal commitment to do so (also known as “fear of floating”; Calvo and Reinhart (2002)). Whereas soft pegs may have been relatively common in the 1970s and 1980s (including in the context of failed stabilization programs), over the past couple of decades, closet peggers have been more prevalent. That is, most de jure pegs are also de facto pegs, but many de facto pegs are not de jure pegs (and, correspondingly, many de jure floats are not de facto floats).1

If that is the case, then the different results for inflation performance under de jure versus de facto pegs cannot be attributed to deeds mattering more than words—because, when it comes to de jure pegging the exchange rate, in most cases deeds back those words. Rather, the explanation must lie in the fact that many de facto pegs are not also de jure pegs. We therefore hypothesize that it is the formal commitment to the de jure peg that instills monetary discipline and helps anchor inflationary expectations.2 Of course, if this commitment were lightly broken, then there would be few credibility benefits of the de jure peg—our point is that this seldom happens (most de jure pegs are also de facto pegs).

Testing our hypothesis using existing studies runs into several obstacles. While the source for the de jure classification is the International Monetary Fund’s Annual Report on Exchange Arrangements and Restrictions (AREAR), the source and methodology underlying the various de facto classifications are usually quite different, resulting in little agreement among them, and making it difficult to judge whether disparate findings reflect substantive differences across the classifications or simply idiosyncracies in how they have been constructed. Moreover, empirical studies differ considerably in terms of the sample considered and the methodology employed, further complicating comparisons across them.3

To get around these problems, we exploit a unique data set created by the IMF for our empirical analysis that provides both de jure and de facto classifications of the exchange rate regime of all member countries over the period 1980–2010. Since these classifications are from a common source, any differences between them should be substantive, and not just idiosyncracies in how they are constructed. Moreover, applying the two classifications to a common sample and using the same methodology allows for a direct comparison of inflation performance under de jure and de facto pegs.

Our results based on a sample of 145 EMDCs over 1980–2010 support the hypothesis that formal commitment to pegging matters.4 Specifically, we find that pegged exchange rates are associated with significantly lower inflation—through both the confidence and monetary discipline channels—than either the intermediate or floating exchange rate regimes, but that this effect is much stronger for de jure pegs than for de facto pegs. Moreover, the finding of a relatively smaller impact of de facto pegs stems from those cases where the de facto peg is not supported by the de jure commitment. Indeed, when we restrict the sample to those cases where the de facto peg is also a de jure peg, the estimated association between de facto pegs and low inflation strengthens significantly. In addition, we find that the effect of pegs evolves over time with longer duration of the regime associated with lower inflation, albeit at a diminishing rate. Our findings are generally robust, including taking account of potential endogeneity of regime choice, and to the exclusion of the extreme cases of freely falling regimes and hard pegs.

Beyond providing fresh empirical evidence on the relationship between the nominal exchange rate regime and inflation performance in a large sample of EMDCs, our main contribution to the existing literature is thus three fold. First, by employing a novel dataset of both de jure and de facto classifications from the same source (the IMF), and applying the same empirical methodology, we are able to obtain results across alternative regime classifications that are more readily comparable than has hitherto been possible. Second, we help clarify the respective roles of formal commitments versus actual behavior of the central bank in anchoring inflationary expectations. Third, as a by-product, we are able to reconcile the often starkly different findings in previous studies of how the exchange rate regime affects inflation performance.

The remainder of this paper is organized as follows. Section II lays out a simple theoretical framework in which the central bank’s commitment to peg helps to anchor inflationary expections and deliver low inflation, and also discusses our empirical strategy. Section III describes the exchange rate regime classification. Section IV presents our main empirical results and the robustness tests. Section V concludes.

II. Theoretical Framework and Empirical Strategy

We begin by developing a simple theoretical model to explore how the nominal exchange rate regime could affect inflation performance. The model builds on the policy credibility models of Barro and Gordon (1983a, b), as applied to the exchange rate regime (see Wolf et al., 2008). Using this setup, we first compare “pure” pegged and floating exchange rate regimes—that is, when the central bank both de jure commits to, and de facto follows, its chosen regime. We then consider the “hybrid” cases of closet pegs (a de jure non-peg, but de facto peg) and soft pegs (a de jure peg, but de facto non-peg)—modeling them as pegged exchange rate regimes with limited commitment/adherence—and compare inflation performance under these hybrid regimes to the pure pegs and floats.5

Our model suggests that, comparing various pegged exchange rate regimes, inflation will be lowest under the pure peg (where the central bank both de jure commits to maintain the parity and de facto does so) followed by closet pegs (where the central bank does not commit to maintain the peg but de facto does so) and soft pegs (where the central bank de jure commits to maintain the parity but de facto does not do so); inflation is highest under a pure float (that is, where the central bank neither de jure nor de facto pegs). This ranking of inflation rates reflects a combination of the monetary discipline that de facto maintaining the peg entails, and the confidence in the currency that a de jure commitment engenders.

A. The Setup

Output is determined by a Lucas-type “surprise inflation” supply function:

y=θ(πdπde)+η(1)

where y is the level of output (with the “natural” level of output normalized to zero), θ > 0 is the elasticity of output with respect to the inflation surprise, πd is price inflation of the domestically produced good, and πde is the private sector’s expectation of inflation, which must be formed before the realization of the shock to output, η, which has mean zero and variance ση2.6

Consumer price inflation (CPI), π, is a weighted average of domestic (πd) and imported goods’ (πi) price inflation, where β ∈ [0,1]:

π=βπd+(1β)πi(2)

Domestic price inflation is given by inverting the money demand function:

πd=Δm+νπde(3)

where Δm is broad money growth, and 0≤ v <1 is the elasticity of the growth of velocity with respect to expected inflation, reflecting forward-looking elements in household money demand, whereby higher expected inflation reduces the trend growth of money demand.

Price inflation of the imported good is given by purchasing power parity, where foreign price inflation is assumed to be zero:

πi=π*+Δe=Δe(4)

while the nominal exchange rate (defined so that an increase in e represents a depreciation of the domestic currency) is determined by a simple monetary relation:

Δe=Δm(5)

For simplicity, the banking system is not modeled, so the money supply consists of central bank domestic credit (DC) and foreign exchange reserves (R). Under a float, there are no foreign exchange reserves, and the money supply equals domestic credit. Under a peg, the central bank does not pursue an activist monetary policy but simply keeps the exchange rate constant by sterilizing any capital inflows or outflows, hence ΔDC = − ΔR ⇒Δm = 0

The central bank is assumed to have two objectives: stabilizing output around some desired level, y¯0, that may exceed the natural rate of output, and keeping inflation low:

Min L=12E{A(yy¯)2+π2}(6)

where E{.} is the central bank’s expectation and A is the relative weight placed on output.7

B. Pegged versus Floating Exchange Rate Regimes

Pure peg

Under a pure peg, the central bank both commits to maintain a constant parity and de facto does so: Δe = 0. From (5), this requires Δm = 0. In turn, from (4), πi = 0, while taking expectations of (3) yields πde=0. Substituting into (1)-(3):

π=0(7)
y=η(8)

Thus, the central bank is able to achieve its bliss level of inflation, but at the cost of not being able to react at all to the output shock.

For a given realization of the stochastic shocks, the corresponding welfare loss under a de jure pegged exchange rate regime is thus:

LPeg(η)=12{A(ηy¯)2}(9)

while the ex ante expected loss is:

LPeg=12{Aση2+Ay¯2}(10)

Free float

Under a floating exchange rate regime, the central bank is free to choose its domestic credit policy to minimize its loss function. Substituting (3) into (6) where Δm = ΔDC, and solving for the first-order condition for the optimal choice of monetary policy (DC), we get:

ΔDC=Aθ(y¯η)πde(βν+Aθ2(ν1))1+Aθ2(11)

Substituting (11) into (3) and taking expectations yields:

πde=Aθy¯1(1β)ν>0(12)

Optimal monetary policy becomes:

ΔDC=(1ν)Aθy¯1(1β)νAθη1+Aθ2(13)

From (13), the higher the central bank’s target for output, y¯>0, the more expansionary will be discretionary monetary policy under the float, and correspondingly, the higher the domestic good and overall inflation rates:

πd=Aθy¯1ν(1β)Aθη1+Aθ2(14)
π=Aθy¯Aθη1+Aθ2(15)

Using (1), however, output is independent of the central bank’s target level of output because workers perfectly anticipate the central bank’s incentive to generate surprise inflation:

y=θ(πdπde)+η=η1+Aθ2(16)

The welfare loss for a given realization of shocks under a free float is thus:

LFlt(η)=12{A(η1+Aθ2y¯)2+(Aθy¯Aθη1+Aθ2)2}(17)

and the expected loss is:

LFlt=12{Ay¯2(1+Aθ2)+Aση21+Aθ2}(18)

Comparison of Regimes

The choice between a peg and a float depends on the trade-off between the anti-inflation credibility afforded by the peg and the ability to use an activist monetary policy to offset real output shocks under a float. If the central bank has a large incentive to generate surprise inflation—or the private sector believes that it has (y¯=0,y¯e>0)—then a pegged exchange rate regime dominates by solving the time-consistency problem.8 Conversely, if the central bank does not have such an incentive—and enjoys credibility—then a flexible exchange rate that allows it to react to output shocks would be preferable. The exact condition under which ex ante expected welfare is higher under a pegged regime is given by a comparison of (10) and (18):9

LPeg<LFltAση2+Ay¯2<Ay¯2(1+Aθ2)+Aση21+Aθ2ση2Aθ2<y¯2(19)

While ex ante expected welfare (19) is the correct criterion for the optimal choice of regime, a decision to abandon an existing peg would be based on the welfare gain (for a given realization of the shock, η) from floating the exchange rate relative to the cost of exiting the peg, which we denote by c. When the private sector expects the peg to be maintained, πde=0, the central bank’s welfare gain by abandoning the peg and floating the exchange rate is given by LPeg(η)LFlt,(πde=0)(η). Therefore, the central bank exits the peg if:

LPeg(η)LFlt,πde=0(η)=A2θ2(ηy¯)21+Aθ2>c(20)

where c captures the political and economic costs of the central bank reneging on the commitment to peg the exchange rate.10

C. Hybrid Cases: Closet Pegs and Soft Pegs

Closet peg

We model a closet peg as a regime where the central bank does not make a formal commitment to maintain the parity (or, put differently, the cost, c, of abandoning a de facto peg is assumed to be lower than the cost of exiting a de jure peg).11 As shown above, in the face of a sufficiently large real shock, the central bank will abandon its peg and generate surprise inflation to offset the negative shock to output. Let ρ be the probability that the peg survives, and (1-ρ) the probability that it is abandoned.12 In forming its expectations, the private sector now needs to take into the possibility that the peg may be abandoned:

πde=(ρ×0)+(1ρ)Aθy¯+πde(ν(1β)+Aθ2)1+Aθ2(21)

Solving for expected inflation as a function of the probability that the peg survives:

πde(ρ)=(1ρ)Aθy¯[1+ρAθ2(1ρ)ν(1β)]>0(22)

Expected inflation is given by (22) for some ρ < 1.13 Under the de facto peg, however, it remains the case that the central bank maintains the parity implying Δe = 0, which from (5) requires that Δm = 0. Substituting (22) into (3) shows that even if the parity is maintained (Δem=0), the expectation of inflation will lead to some positive consumer price inflation:14

π=βν(1ρ)Aθy¯[1+ρAθ2(1ρ)ν(1β)]>0(23)

Soft peg

Under a soft peg, the central bank de jure pegs the exchange rate, but de facto chooses a discretionary monetary policy that is inconsistent with maintaining the peg. Despite the de jure commitment, the central bank’s credibility is likely to be low in these circumstances, so inflation will again be given by (22) for some ρ< 1. Substituting (22) into (11) yields the optimal discretionary monetary policy, given the private sector’s expectations:

ΔDC(πde(ρ))=Aθy¯(1(1ρ)ν)(1+ρAθ2(1ρ)ν(1β)(24)

Substituting into (3) yields the domestic and overall CPI inflation rates::

πd=Aθy¯(1+ρAθ2(1ρ)ν(1β))>0(25)
π=Aθy¯(1(1ρ)ν(1β))1+ρAθ2(1ρ)ν(1β))>0(26)

both of which are again positive.15 It is noteworthy, however, that soft pegs are not likely to last long because the central bank’s monetary policy, ΔDC > 0, is inconsistent with maintaining the parity.16

D. Inflation Performance under Alternative Regimes

Abstracting from realizations of the stochastic shock,η, the results obtained above can be used to rank average inflation rates across the various exchange rate regimes. First, from (7), inflation is lowest (actually, equal to the central bank’s bliss value, here normalized to zero) under a pure peg. Under all other regimes, inflation is strictly positive.

Second, comparing (23) and (26) shows that, if the credibility of the peg is similar between closet pegs (where the central bank does not commit to maintain the peg but follows a monetary policy consistent with doing so) and soft pegs (where the central bank does commit to maintain the parity but then is not observed to do so), then inflation will be higher under the latter:

πSoftPeg>πClosetPeg1(1ρ)ν(1β)>βν(1ρ)1(1ρ)ν>0(27)

Third, comparing inflation under a soft peg to that under a free float (15) shows that inflation will be higher under the free float, that is, commitment matters:

πFlt>πSoftPeg1+ρAθ2(1ρ)ν(1β))>(1(1ρ)ν(1β))ρAθ2(1ρ)ν(1β))>(1ρ)ν(1β)ρAθ2>0(28)

Putting the chain together yields the following relative ranking of inflation performance across regimes:

πFlt>πSoftPeg>πClosetPeg>πPurePeg(29)

This ranking reflects both monetary discipline and confidence effects. To the extent that the central bank de facto maintains the parity, it must exhibit monetary discipline, Δm = 0. To the extent that the de jure commitment is credible (the private sector assigns a low probability to that the peg will be abandoned), expected inflation is low, engendering confidence in the currency and raising (the growth rate of) money demand, in turn implying that actual inflation is correspondingly lower for a given rate of monetary expansion.

E. Empirical Specification

Next, we assess whether the two key predictions of the model—that inflation is lower under pegs than under floats; and that among pegs, inflation is lowest when formal commitment is backed by deeds—hold. To do so, we first estimate the relationship between inflation and pegged exchange rate regimes, controlling for other factors that are likely to determine inflation:

πit=β0+β1Xit+βPegPegit+βIntIntit+βMonΔmit+νt+εit(30)

where πit is the inflation rate for country i at time t;17 Peg and Int are dummy variables for pegged and intermediate exchange rate regimes respectively (with the floating regime as the excluded category); Δm is the growth in broad money; X includes the other likely determinants of inflation performance (real GDP growth, trade openness, central bank independence (proxied by the central bank governor turnover rate), fiscal balance, and terms of trade shocks);18 v are year effects to capture the effect of shocks over time that are common to all countries; and ε is a random error term.19

In equation (30)—which constitutes our benchmark specification—the estimates of βPeg and βInt are the direct effects of exchange rate regimes on inflation that are obtained after controlling for all other possible determinants. However, as mentioned above, money growth itself may vary systematically by regime, so the exchange rate regime could also affect inflation indirectly through its effect on money growth:

Δmit=α0+α1Xit+αPegPegit+αIntIntit+υt+ηit(31)

Taking into account the possibility that money growth is endogenous to the exchange rate regime as in equation (31), we also estimate the total effect of pegs and intermediate regimes, which adds both the direct and indirect effects. Specifically, the total effect of pegs (γPeg) is given by βPeg+βMonαPeg, and that of intermediate regimes (γInt) is given by βInt+βMonαInt.

We then examine whether the formal commitment to pegging matters such that inflation is lower when a de facto peg is backed by a de jure peg than when it is not. We do so by estimating two alternate specifications. In the first, we estimate equation (30) but drop the observations when a de facto peg is not supported by a de jure peg (that is, restrict the sample to pure pegs and floats only). In the second case, we retain all observations but include dummy variables to capture pure pegs, closet pegs, and soft pegs respectively:20

πit=β0+β1Xit+βPPPurePegit+βCPClosetPegit+βSPSoftPegit+βMonΔmit+νt+εit(32)

which allows for a more direct comparison of the estimated effects of the pure and hybrid exchange rate regimes on inflation performance.21

III. Exchange Rate regimes: Classifications and Trends

Any empirical study of exchange rate regimes must contend with issues of regime classification. Early work used a de jure classification—the regime declared by national authorities in the IMF’s AREAR. Thereafter, de facto classifications that seek to categorize the regime according to the behavior of the exchange rate (supplemented by information on movements in foreign exchange reserves, interest rates, and parallel market exchange rates) became more common.22 The problem in comparing macroeconomic performance under de jure versus de facto pegs is that the source of the classifications is usually quite different, and even among de facto classifications there is little agreement (correlations between them are generally quite low and in the range of 0.1-0.4), making it difficult to judge whether the various findings reflect substantive differences or simply the variety of samples and methodologies employed.

To address these problems, we employ a unique dataset that provides both a de jure and a de facto classification from a common source—the International Monetary Fund. The de jure classification is based on the IMF’s AREAR database; the de facto classification draws on IMF country teams, staff reports, and staff consultations with central banks, and has been compiled by Anderson (2008).23 Appendix I establishes that our de facto classification is generally less idiosyncratic than other de facto classifications—in the sense that, observation by observation, a higher proportion of the other classifications agree with our classification than with any other classification, giving confidence that the empirical results are likely to be robust and not driven by idiosyncrasies of the classification.

The IMF’s de jure and de facto classifications group exchange rate regimes into eight categories: (i) exchange arrangement with no separate legal tender; (ii) currency board arrangement; (iii) conventional pegged arrangement (further divided into single and basket pegs); (iv) pegged exchange rates within horizontal bands; (v) crawling peg; (vi) crawling band; (vii) managed float with no predetermined path for the exchange rate, and (viii) independently floating arrangement. For the empirical analysis, we group the first three arrangements (excluding basket pegs) into the pegged exchange rate regime category; group the next four (including basket pegs) as intermediate regimes; and classify the last one as the floating regime (see Table 1 for the distributions).

Table 1.

Classification of Exchange Rate Regimes, 1980–2010

(in percent of observations)

article image
Source: Authors’ estimates based on IMF’s AREAER and Anderson (2008).

Figure 1 shows the evolution of exchange rate regimes for our sample of 145 EMDCs over the period 1980–2010.24 Three patterns are discernible. First, among the EMEs, floating exchange rate regimes have become more popular since the 1997/98 Asian crises—though there has not been a corresponding decrease in the proportion of pegged exchange rate regimes (including de jure pegs where the central bank formally commits to maintain the parity). Second, consistent with the bipolar hypothesis (Fischer, 2001), there has been a marked “hollowing out of the middle”—particularly in de facto terms—with the proportion of intermediate regimes in the EMEs shrinking from around 64 percent in the 1980s and 1990s to about 34 percent in the last decade. Third, there is significant divergence between the de jure and de facto classifications of pegged and floating exchange rate regimes. This divergence mainly reflects countries de facto pegging their exchange rate but being unwilling to take on the de jure commitment to the pegged exchange rate (as opposed to cases where the central bank violates its formal commitment to maintain the parity by allowing the exchange rate to move, see Table 2). Indeed, of de jure pegs, almost 95 percent are also backed up a de facto peg, but of de facto pegs, only about 70 percent are also de jure pegs; in other words, closet pegs are much more common than soft pegs.25

Figure 1.
Figure 1.

Frequency Distribution of Exchange Rate Regimes, 1980–2010

(in percent)

Citation: IMF Working Papers 2011, 112; 10.5089/9781455261413.001.A001

Source: Authors’ estimates based on IMF’s AREAR, and Anderson (2008).
Table 2.

Distribution of De Jure and De Facto Classifications, 1980–2010

(in percent of observations)

article image
Source: Authors’ estimates based on IMF’s AREAR, and Anderson (2008).

IV. Inflation and The Exchange Rate Regime

A. Some Stylized Facts

The model outlined in Section II predicts that inflation will be lower under pegged exchange rates. Table 3 indicates that this prediction is indeed borne out by the data—on average inflation is systematically lower (by about 5 percentage points per year) in countries with pegs compared to countries with intermediate or floating exchange rate regimes. What lies behind this better inflation performance? The most obvious explanation is that money growth is lower under pegged regimes (the “discipline” effect). Again, this is borne out in the data: broad money growth is, on average, 9 percentage points per year lower under pegs than under the other regimes.

Table 3.

Inflation and Money Growth under Alternative Regimes, 1980–2010*

(in percent per year)

article image
Source: Authors’ estimates.

Based on the estimated sample. Inflation is transformed as follows: inflation/(1+inflation).

The model also suggests that, in addition to the greater monetary discipline, there may be a “confidence effect” whereby inflation under pegged regimes is lower even after controlling for money growth. Figure 2 plots inflation against money growth for both pegged and floating regimes, and indicates that inflation is indeed lower under pegged regimes for a given rate of money growth with the difference initially increasing in money growth.

Figure 2.
Figure 2.

Money Growth and Inflation under Alternative De Jure and De Facto Regimes, 1980–2010* (in percent)

Citation: IMF Working Papers 2011, 112; 10.5089/9781455261413.001.A001

Source: Authors’ estimates.* Figure shows the relationship between broad money growth and inflation (in percent per year) under alternative de jure and de facto regime classifications. The inflation rate has been transformed by: inflation/(1+inflation).

While the statistics reported in Table 3 and Figure 2 provide initial support for the monetary discipline and credibility effects, they also indicate that divergence in policy announcement and actual behavior may have implications for inflation performance across countries. For example, inflation and money growth are slightly lower under de jure pegs than de facto pegs, and the inflation and money growth differential of pegs with intermediate regimes and floats is larger under the de jure classification (Table 3). Not surprisingly, the confidence effect—lower inflation for a given rate of money growth—is greater for de jure pegs than for de facto pegs (Figure 2).

Considering that a vast majority of de jure pegs are also de facto pegs (Table 2), the observed difference in inflation performance (between the de jure and de facto pegs) must stem from the de facto pegs that are in fact not de jure pegs. Figure 3 indicates that this is indeed the case—both inflation and money growth are higher in cases where the peg is not backed by formal commitment, and that the difference in inflation performance and money growth (between the de facto pegs that are not backed by words and those that are) is larger for the EMEs. Figure 3 also indicates that inflation in soft pegs is higher than in closet pegs, but that the inflation performance of soft pegs is better than that of pure floats (particularly in developing countries). The observed pattern of inflation across the pure and hybrid regimes thus corresponds with the ranking obtained from the model, as given in (29).

Figure 3.
Figure 3.

Inflation and Money Growth under Pegs and Non-Pegs, 1980–2010*

Citation: IMF Working Papers 2011, 112; 10.5089/9781455261413.001.A001

Source: Authors’ estimates.Based on the estimated sample where inflation is transformed as follows: inflation/(1+inflation).

These statistics, while consistent with the model’s predictions, are only impressionistic as they do not control for other factors that may affect money growth and inflation. In what follows, we verify that the association between the nominal exchange rate regime and inflation implied by our model still holds once we condition for other factors. In addition, we examine the importance of formal policy announcements in determining this association.

B. Estimation Results

The estimation results for the baseline equation (1) for the full sample, as well as the emerging market and developing country sub-samples, are presented in Table 4.26 The various determinants of inflation enter the regression with the expected signs and are statistically significant (with the exception of the terms of trade variable). Specifically, higher real GDP per capita growth, a strong fiscal position, and greater trade openness are associated with lower inflation, while higher money growth, and a higher central bank turnover rate (a proxy for lack of central bank independence) tend to raise it.

Table 4.

Inflation Performance: Benchmark Specification

article image
Source: Authors’ estimates.1/ Regression of inflation (decimal fraction, per year) on regime dummy variables, and other control variables. Estimates obtained from instrumental variable estimation controlling for the endogeneity of real GDP growth, fiscal balance, and money growth where lagged values are used as instruments. All specifications include time effects. t-statistics based on clustered standard errors. Negative coefficient on pegged or intermediate exchange rate regime dummies indicates lower inflation under that regime relative to inflation under floating exchange rate regimes (the omitted category). For example, the coefficient of -0.063 for pegged regimes implies 6.3 percent per year lower inflation under pegged exchange rate regimes compared to floats. *, **, and *** indicate signficance at the 10, 5 and 1 percent levels, respectively.

Direct effect of exchange rate regime on inflation, plus indirect effect through money growth.

Turning to the impact of the exchange rate regime, inflation under (de jure) pegged regimes is about 6 percentage points per year lower than under floating exchange rates. Of this differential, about 3 percentage points are associated with lower money growth (the discipline effect), and the remaining 3 percentage points represent the confidence effect (lower inflation for a given rate of money growth). Inflation under pegs is also lower than under the intermediate regimes, and at 5.5 percentage points, the differential is only slightly smaller than that of pegs with floats. For emerging markets and developing countries sub-samples, de jure pegs are associated with about 9 and 5 percentage points lower inflation than floating regimes, respectively, while intermediate regimes, on average, have about a 1 percentage point differential with floats.

Consistent with the findings of earlier studies, the inflation differential (through both the confidence and discipline effects) associated with de facto pegs is smaller than that associated with de jure pegs. De facto pegs are associated with about 3 percentage points lower inflation that floats, of which 2 percentage points are due to the confidence effect. The relatively smaller effect of de facto pegs on inflation for the full sample seems to come from the EMEs—where the inflation differential between pegs and floats is about 2 percentage points, less than half of that for the developing countries (about 6 percentage points).27

These results are in line with policy credibility models, which suggest that pegs should be associated with lower inflation both because they instill monetary discipline and engender confidence in the currency. But our findings also raise the question of why the estimated effects are weaker for de facto pegs than for de jure pegs. If the difference between de jure and de facto classifications is a matter of “deeds versus words,” as argued by some earlier studies (for example, Levy-Yeyati and Sturzenegger, 2005), then this result is puzzling, since in most cases deeds are backed by words (Table 2). Rather, the effect must be coming from the de facto pegs that are not also de jure pegs, in other words, pegs that lack the central bank’s commitment to maintain the parity. In the theoretical model, it is precisely this commitment, which is costly to break, that leads the central bank to maintain monetary discipline and imparts confidence in the currency, leading to lower inflation.28 Thus, de facto pegs that are not supported by a formal commitment may not deliver the full disinflationary benefits of pegs.

Does this account for the difference in results between the de jure and de facto classifications? It does. This is readily verified by dropping those de facto pegged observations that are not also classified as de jure pegs (Table 5).29 The inflation differential of a de facto peg relative to floats increases to 4 percentage points for the full consensus sample (4 and 8 percentage points for the EME and developing country samples, respectively).30

Table 5.

Inflation Performance: Consensus between De Jure and De Facto Pegs

article image
Source: Authors’ estimates.1/ Regression of inflation (decimal fraction, per year) on regime dummy variables, and other control variables. Estimates obtained from instrumental variable estimation controlling for the endogeneity of real GDP growth, fiscal balance, and money growth where lagged values are used as instruments. All specifications include time effects. t-statistics based on clustered standard errors. Negative coefficient on pegged or intermediate exchange rate regime dummies indicates lower inflation under that regime relative to inflation under floating exchange rate regimes (the omitted category). For example, the coefficient of -0.064 for pegged regimes implies 6.4 percent per year lower inflation under pegged exchange rate regimes compared to floats. *, **, and *** indicate signficance at the 10, 5 and 1 percent levels, respectively.

Direct effect of exchange rate regime on inflation, plus indirect effect through money growth.

These results are also verified by the estimated alternate specification, equation (32), where we directly compare the estimated effect of de facto pegs matched by de jure pegs, with those that are not. The obtained results presented in Table 6 correspond to the pattern observed in Figure 3—inflation relative to non-pegs is lower by about 6 and 3 percentage points in the EMEs and developing country samples, respectively, when a de facto peg is supported by a de jure peg than when it is not; words matter. The results also show, however, that de jure pegs that are not matched by de facto pegs (that is, soft pegs) imply somewhat lower inflation than pure floats. While, as an empirical matter, there are very few of these cases (soft pegs are much rarer than closet pegs), this finding implies that deeds matter as well.31

Table 6.

Inflation Performance: Consensus between De Jure and De Facto Classifications

article image
Source: Authors’ estimates.1/ Estimates obtained from instrumental variable estimation controlling for the endogeneity of real GDP growth, fiscal balance, and money growth where lagged values are used as instruments. Pure peg is a binary variable equal to one if both classifications indicate a peg (and zero otherwise). De facto peg is a binary variable if the de jure classification indicates a nonpeg (intermediate or float) but the de facto classification indicates a peg (and zero otherwise). Soft peg is a binary variable if the de jure classification indicates a peg but the de facto classification indicates a nonpeg (and zero otherwise). Negative coefficient on pure, closet, or soft pegs indicates lower inflation under that regime relative to inflation under (pure) nonpegs. For example, the coefficient of -0.066 for pure pegs implies 6.6 percent per year lower inflation under pegged exchange rate regimes compared to floats. All specifications include time effects. t-statistics based on clustered standard errors. *, **, and *** indicate signficance at the 10, 5 and 1 percent levels, respectively.

Direct effect of exchange rate regime and regime duration on inflation, plus indirect effect through money growth.

C. Alternative Specifications and Robustness

The empirical analysis presented thus far indicates that the predictions of the policy credibility models hold—pegs are indeed associated with lower inflation as compared to more flexible exchange rate regimes—and that (de facto) pegs in which the central bank also makes a formal commitment imply lower inflation than pegs without such a commitment. We now establish the robustness of these findings through a range of sensitivity tests.

Dynamic effects

The benchmark specification looks only at the “static” relationship between the exchange rate regime and inflation, yet the impact of regimes may phase in over time. To examine any dynamic effects of pegs, we alter the baseline specification to include a variable equal to the (inverse of) the years into the peg (in addition to the peg dummy). Thus, for the first year of a peg, this variable is equal to 1; for the second year, it takes the value of 0.5, and so forth; asymptotically it approaches zero as the duration of the peg increases.32 If the effect of pegs is dynamic, then the estimated coefficient of this (duration) variable would be significantly different from zero.

The results presented in Table 7 show that the duration of pegs is indeed important; inflation is reduced gradually over time as the peg is maintained, albeit at a diminishing rate. For example, for both de jure and de facto pegs, there is on average about a 10 percentage points reduction in inflation during the first ten years of pegging, but only 0.1 percentage point reduction in the following ten years. The full effect of pegging—given by the sum of the estimated coefficients of the peg dummy and the duration variable—is larger for de jure pegs than for de facto pegs. For example, five years under a de jure peg is associated with 5 percentage points lower inflation relative to a float, whereas this differential under a de facto peg is 3 percentage points. The larger effect of de jure pegs is consistent with the results reported in Table 4. Further, as before, dropping the observations where the de facto pegs are not classified as de jure pegs, strengthens the estimated effect of de facto pegs. For both EMEs and developing countries, the inflation differential of de facto pegs relative to floats (five years into the regime) is about 1 percentage point greater in the consensus sample (Figure 4).

Figure 4.
Figure 4.

Dynamic Effect of De Jure and De Facto Pegs on Inflation*

Citation: IMF Working Papers 2011, 112; 10.5089/9781455261413.001.A001

Source: Authors’ estimates.*Figure shows estimated inflation dynamics following adoption of a de jure or de facto peg.
Table 7.

Inflation Performance: Regime Duration

article image
Source: Authors’ estimates.1/ Regression of inflation (decimal fraction, per year) on regime dummy variables, and other control variables. Estimates obtained from instrumental variable estimation controlling for the endogeneity of real GDP growth, fiscal balance, and money growth where lagged values are used as instruments. All specifications include time effects. t-statistics based on clustered standard errors. Negative coefficient on pegged or intermediate exchange rate regime dummies indicates lower inflation under that regime relative to inflation under floating exchange rate regimes (the omitted category). The overall effect of pegs on inflation is the sum of the main coefficient for pegs and the duration of pegs. For example, the coefficient of -0.074 for pegged regimes and 0.107 for the duration of pegged regimes implies that for peg duration equal to 5 years, inflation is 5.2 percent lower under pegged exchange rate regimes compared to floats. *, **, and *** indicate signficance at the 10, 5 and 1 percent levels, respectively.

Direct effect of exchange rate regime and regime duration on inflation, plus indirect effect through money growth.

Capital inflows and current account surpluses

The results of the baseline specification (presented in Table 4) pertain to the average performance of the exchange rate regimes. There may however be particular circumstances in which they do not hold. For example, while pegging the exchange rate is supposed to import the credibility of the anchor currency, it can also import inflation—either because the anchor currency itself is subject to inflationary pressures or because of excessive money growth through unsterilized reserves accumulation in the face of positive balance of payments pressures (capital inflows or current account surpluses).33

To examine these possibilities, we restrict the sample to observations where capital inflows are large—defined as inflows above 2.5 percent of GDP (the full sample’s 30th percentile for positive net flows)—or the current account surplus is large (defined as 2 percent of GDP, or about the full sample’s 30th percentile for positive balances). The results presented in Table 8 show that in the face of large capital inflows, money growth is indeed higher for pegged regimes; however, the inflation differential between floating and pegged regimes remains positive for the full, and individual EME and developing country samples. By contrast, in the face of large current account surpluses, money growth is sufficiently higher under pegged regimes that the differential moves in favor of floating exchange rates. Specifically, for the full sample, de jure and de facto pegs are associated with 4-6 percentage points higher inflation than floating exchange rate regimes—with the effect stemming entirely from the money growth rather channel. For developing countries, even under large current account surpluses, pegs do not have significantly higher inflation than floating regimes.

Table 8.

Inflation Performance under Balance of Payments Surpluses

article image
Source: Authors’ estimates.1/ Regression of inflation (decimal fraction, per year) on regime dummy variables, and other control variables. Estimates obtained from instrumental variable estimation controlling for the endogeneity of real GDP growth, fiscal balance, and money growth where lagged values are used as instruments. All specifications include time effects. t-statistics based on clustered standard errors. Negative coefficient on pegged or intermediate exchange rate regime dummies indicates lower inflation under that regime relative to inflation under floating exchange rate regimes (the omitted category). For example, the coefficient of -0.063 for pegged regimes implies 6.3 percent per year lower inflation under pegged exchange rate regimes compared to floats. *, **, and *** indicate signficance at the 10, 5 and 1 percent levels, respectively.

Direct effect of exchange rate regime on inflation, plus indirect effect through money growth.

Corresponds to the 30th percentile of positive net capital flow and positive current account balances, respectively.

In addition to the effect of large current and capital account surpluses, we also assess the effect of inflation in the anchor country (to which the currency is pegged) by including it in the regression.34 The results show that the estimated effect of anchor country inflation is strongly positive, while the negative effect of pegs increases—indicating that higher inflation of the anchor currency is indeed imported by pegged countries. That said, at least for the EMDCs, inflation of the anchor currencies was not sufficiently high over the sample period for the countries that pegged to them to experience higher inflation than countries with intermediate regimes or floats.

Regime endogeneity

One difficulty in interpreting the results pertains to the potential endogeneity of the exchange rate regime itself. For example, if countries with low inflation—perhaps because of strong national consensus on the need for price stability—also have a greater proclivity, or ability, to peg their exchange rate, then the estimated effect on inflation of pegging the exchange rate regimes may be biased upward. To address this issue, we estimate a simultaneous equation system that allows explicitly for regime endogeneity.35

Specifically, we estimate a probit model where the decision to adopt a pegged regime (rather than an intermediate or floating exchange rate regime) is modeled as a function of the inflation rate, the control variables included in the baseline specification, and several variables deemed exogenous that may affect the choice of exchange rate regime but otherwise are not expected to have an effect on the inflation rate (such as the country’s population size and the geographic concentration ratio of its exports). The predicted value of the exchange rate regime variable from the probit model is then used in the second stage regression to estimate the inflation rate.36

The results presented in Table 9 indicate that taking account of regime endogeneity does not make a large difference to the estimated effect of pegs on inflation performance: for the full sample, de jure and de facto pegs are associated with about 7 and 4 percentage points lower inflation, respectively, than the intermediate and floating regimes through both the credibility and monetary discipline channels.

Table 9.

Inflation Performance: Regime Endogeneity, Freely Falling, and Hard Pegs

article image
Source: Authors’ estimates.1/ Estimates obtained from instrumental variable estimation controlling for the endogeneity of real GDP growth, fiscal balance, and money growth where lagged values are used as instruments. All specifications include time effects. t-statistics based on clustered standard errors. *, **, and *** indicate signficance at the 10, 5 and 1 percent levels, respectively.

Direct effect of exchange rate regime on inflation, plus indirect effect through money growth.

Instrumental variable regression (with population size and geographic export concentration ratio as instruments) estimating the effect of pegged regimes relative to intermediate and floating regimes.

Excludes floating regimes with rate of nominal exchange rate depreciation exceeding 25 percent per year.

Excludes floating regimes with inflation over 40 percent per year.

Excludes hard pegs (currency boards and monetary unions) from the pegs category.

Freely falling regimes and hard pegs

Another concern with both de jure and de facto classifications is that the floating category conflates countries that choose to float their exchange rate (“freely floating”) with those that must float because they are in a state of economic and financial collapse (“freely falling” in the terminology Reinhart and Rogoff (2004)). To address the concern that the float may not be a country’s natural choice, we exclude all observations from the sample that correspond to floaters but where either the nominal exchange rate depreciation is greater than 25 percent per year or the inflation rate exceeds 40 percent per year.

Table 9 shows that excluding the freely falling regimes (using either definition) from the sample does not change the finding of lower inflation under pegged regimes—the inflation differential of pegged regimes vis-à-vis the floaters remains about 1-3 percentage points for the full sample, while intermediate regimes are associated with about 1-3 percentage points higher inflation than floats.37

Likewise, although hard pegs (currency boards and monetary unions) tend to have the lowest inflation rates among pegged exchange rate regimes, reassuringly the results are not entirely driven by them. Excluding all observations of hard pegs from the pegged regime category leads to a statistically significant inflation differential in favor of pegs vis-à-vis floats.38

V. Conclusion

This paper revisits the link between exchange rate regimes and inflation performance in 145 emerging market and developing countries over the period 1980–2010. We employ a novel dataset on de jure and de facto exchange rate regime classifications, which affords a more meaningful comparison of the results obtained from the two classifications, and allows us to reconcile differences in the findings of previous studies.

In line with the predictions of policy credibility models, our empirical results suggest that pegs are a useful commitment device and are associated with lower inflation in emerging market and developing countries. This effect is achieved through both monetary discipline (implying a lower rate of money growth) and greater credibility (whereby lower inflationary expectations lead to higher money demand and therefore lower inflation for a given rate of money growth).

Naturally, if the central bank does not follow through on its de jure commitment, then its credibility will soon be lost, as will the low-inflation advantage of pegging. Empirically, however, such “soft pegs” are rare. More common are closet pegs (de facto peg without the de jure commitment) where our analysis suggests that inflation will be significantly higher than if the central bank had made the corresponding de jure commitment as well. As such, deeds matter, but so do words.

Since most de jure pegs are also de facto pegs, but many de facto pegs are not de jure pegs, our results help reconcile the findings of studies based on the de jure classification (which generally conclude that pegging the exchange rate is associated with lower inflation) and studies based on de facto classifications (which tend to find much weaker effects of pegs). Also, while the literature in recent years has tended to shift away from the use the de jure classification of exchange rate regimes, our results underscore the importance of both de jure and de facto classifications.

Of course, low inflation is just one macroeconomic objective—albeit an important one for a central bank, and one that is likely to be closely related to the choice of exchange rate regime. What our results show is that promising to maintain the parity and delivering on that promise yields the greatest benefit in terms of price stability. A fuller analysis would take account of the impact of de jure versus de facto pegs on other variables—for instance, susceptibility to currency crises. This is left to future research, which could usefully explore the overall cost-benefit calculus of pegged exchange rate regimes, including when words do—or do not—match deeds.

References

  • Alesina, A., and Wagner, A. F., 2006, “Choosing (and Reneging on) Exchange Rate Regimes,” Journal of the European Economic Association, Vol. 4, pp. 770799.

    • Search Google Scholar
    • Export Citation
  • Anderson, H., 2008, “Exchange Policies before Widespread Floating (1945–89),” mimeo (Washington DC: International Monetary Fund).

  • Barro, R., and D. Gordon, 1983a, “A Positive Theory of Monetary Policy in a Natural-Rate Model,” Journal of Political Economy, Vol. 91, pp. 589-610.

    • Search Google Scholar
    • Export Citation
  • Barro, R. and D. Gordon, 1983b, “Rules, Discretion, and Reputation in a Model of Monetary Policy,” Journal of Monetary Policy, Vol. 12, pp. 101121.

    • Search Google Scholar
    • Export Citation
  • Baxter, M. and A. Stockman, 1989, “Business Cycles and the Exchange Rate Regime: Some International Evidence,” Journal of Monetary Economics, Vol. 23, pp 377400.

    • Search Google Scholar
    • Export Citation
  • Bleaney, M., and M. Francisco, 2005, “Exchange Rate Regimes and Inflation: Only Hard Pegs Make a Difference,” Canadian Journal of Economics, Vol. 38, pp 14531471.

    • Search Google Scholar
    • Export Citation
  • Calvo, Guillermo and Carmen Reinhart, 2002, “Fear of Floating,” Quarterly Journal of Economics, Vol. 117, pp. 379408.

  • Cooper. R, 1971, “Currency Devaluation in Developing Countries,” Essays in International Finance No. 86 (Princeton, NJ: Princeton University).

    • Search Google Scholar
    • Export Citation
  • Fischer, Stanley, 2001, “Exchange Rate Regimes: Is the Bipolar View Correct?Journal of Economic Perspectives, Vol. 15, pp. 3-24.

  • Frankel, J., 2005, “Contractionary Currency Crashes in Developing Countries,” IMF Staff Papers, Vol. 52, pp. 149-192.

  • Genberg, H., and A. K. Swoboda, 2005, “Exchange Rate Regimes: Does What Countries Say Matter?IMF Staff Papers, Vol. 52, pp. 129-141.

    • Search Google Scholar
    • Export Citation
  • Ghosh, A., A. Gulde, J. Ostry, and H. Wolf, 1997, “Does the Nominal Exchange Rate Regime Matter?NBER Working Paper No. 5874 (Cambridge, MA: National Bureau of Economic Research).

    • Search Google Scholar
    • Export Citation
  • Ghosh, A., A. Gulde, and H. Wolf, 2003, Exchange Rate Regime: Choices and Consequences, (Cambridge and London: MIT Press).

  • Ghosh, A., J. Ostry, and C. Tsangarides, 2010, Exchange Rate Regimes and the Stability of the International Monetary System, IMF Occasional Paper 270 (Washington: IMF).

    • Search Google Scholar
    • Export Citation
  • Guisinger, A. G., and D.A. Singer, 2010, “Exchange Rate Proclamations and Inflation-Fighting Credibility,” International Organization, Vol. 64, pp. 313-337.

    • Search Google Scholar
    • Export Citation
  • Levy-Yeyati, E., and F. Sturzenegger, 2001, “Exchange Rate Regimes and Economic Performance,” IMF Staff Papers, Vol. 47, pp. 62-98.

  • Levy-Yeyati, E., and F. Sturzenegger, 2003, “To Float or to Trail: Evidence on the Impact of Exchange Rate Regimes,” American Economic Review, Vol. 93, pp. 117393.

    • Search Google Scholar
    • Export Citation
  • Levy-Yeyati, E., and F. Sturzenegger, 2005, “Classifying Exchange Rate Regimes: Deeds versus Words,” European Economic Review, Vol. 49, pp. 160335.

    • Search Google Scholar
    • Export Citation
  • Reinhart, C. and K. Rogoff, 2004, “The Modern History of Exchange Rate Arrangements: A Reinterpretation,” Quarterly Journal of Economics, Vol. 119, pp. 1-48.

    • Search Google Scholar
    • Export Citation
  • Rogoff, K., A. Husain, A. Mody, R. Brooks, and N. Oomes, 2003, Evolution and Performance of Exchange Rate Regimes, IMF Working Paper No. 03/243 (Washington D.C.: IMF).

    • Search Google Scholar
    • Export Citation
  • Romer, D., 1993, “Openness and Inflation: Theory and Evidence,” Quarterly Journal of Economics, Vol. 108, pp. 869-903.

  • Shambaugh, J., 2004, “The Effects of Fixed Exchange Rates on Monetary Policy,” Quarterly Journal of Economics, Vol. 119, No. 1, pp. 301-352.

    • Search Google Scholar
    • Export Citation
  • Tornell, A., and A. Velasco, 2000, “Fixed versus Flexible Exchange Rates: Which Provides More Fiscal Discipline?Journal of Monetary Economics, Vol. 45, pp. 399-436l

    • Search Google Scholar
    • Export Citation
  • Wolf, H., A. R. Ghosh, H. Berger, and A. Gulde, 2008, Currency Boards in Retrospect and Prospect (Cambridge, MA: MIT Press).

Appendix A. Data Description

Comparison of Exchange Rate Regime Classifications

The commonly used de facto exchange rate classifications differ considerably in their methodologies, and the ultimate coding of the exchange rate regimes in individual countries over the years. While a detailed review of the other available classifications is beyond the scope of this paper, a comparison of IMF’s de facto classification with the other classifications reveals that it is less idiosyncratic than the other commonly used de facto classifications.39 This means that—on average—for each (country-year) observation, the other commonly used de facto classifications agree more with the Fund’s classification than with each other. Figure A1 compares the IMF’s de facto (DF) and de jure (DJ) classifications with the classifications of Levy-Yeyati and Sturzenegger (LYS, 2003) and Reinhart and Rogoff (RR, 2004) using a composite measure of similarity when regimes are grouped as fixed, intermediate and floating. The constructed “similarity” index—which is a weighted average of the consensus between the classifications across the three regimes—takes a value between 0 and 1 indicating perfect dissimilarity and perfect similarity of the classifications, respectively.40

Figure A1.
Figure A1.

Similarity Index across different exchange rate regime classifications*

Citation: IMF Working Papers 2011, 112; 10.5089/9781455261413.001.A001

Source: Authors’ calculations based on Anderson (2008).*Sample period is restricted to 1980-2004 corresponding to data availability for the other classifications. Country coverage is expanded to include all countries for which data is available (advanced, EMEs, and developing countries).

A comparison of the indices reveals that the IMF’s classification has overall greater similarity with the other two classifications. On average, it receives a score of about 0.75 if the de jure classification is included in the comparator category (and of about 0.74 if it is not), while LYS and RR receive overall scores of 0.67, and 0.58, respectively. A low score means that the other classification methods would classify that observation differently—which does not necessarily imply that the classification is incorrect, but that any empirical results obtained using such a classification are unlikely to be robust.

The Shambaugh (JS, 2004) classification is not included in the similarity indices as it is available as a binary variable (pegs versus nonpegs) only. To include it in the comparison, we group the other exchange rate regimes (IMF, LYS, and RR) into binary variables, and compute the correlation matrix. The IMF’s de facto classification is found to be the closest to the JS classification and the least similar to LYS (Table A1). About 87 percent of the observations in the IMF de facto classification are coded (as pegs or nonpegs) in the same way as in the JS classification, and the overall correlation between the two series is 0.76.41

Table A1.

Agreement of Different Exchange Rate Regime Classifications, 1980–2004

article image
Source: Authors’ calculations.*JS, LYS and RR refer to the de facto classifications of Shambaugh (2004), Levy-Yeyati and Struzenegger (2003), and Reinhart and Rogoff (2004), respectively.
Table A2.

List of Countries in the Sample

article image
Table A3.

Data Description and Sources

article image

We are grateful to Marcos Chamon and Jonathan D. Ostry for helpful comments and suggestions, and to Dorothy Kornick for research assistance. The usual disclaimer applies.

1

“Soft” pegs are generally associated with weak institutional quality and macroeconomic management (Alesina and Wagner, 2006). Several hypotheses have been put forward regarding closet peggers. Calvo and Reinhart (2002) argue that countries de facto peg because exchange rate volatility could affect their risk premium on borrowing, and may also give rise to dollarization. Rogoff et al. (2003) suggest that de facto pegs may arise because of currency mismatch and balance sheet concerns, and/or fear of Dutch Disease (in the face of potentially large appreciations). Levy-Yeyati and Sturzenegger (2005) contend that de facto pegs not backed by words reflect a “fear of pegging,” as official pegs are more likely to face speculative attacks. Alesina and Wagner (2006) put forward a signaling argument—that is, countries de facto peg to avoid wide exchange rate fluctuations, which may be taken as an indicator of poor economic management, while retaining the flexibility to respond to idiosyncratic shocks. By contrast, Genberg and Swoboda (2005) suggest that fear of floating may not indicate the breaking of any commitment at all, but rather reflect exchange rate stability achieved through optimally chosen monetary policies.

2

By analogy, the de facto interest rate policy of a central bank that does not inflation target (IT) may be very similar to that of an inflation-targeting central bank—but proponents of IT would argue that it is the commitment to the formal IT regime that helps anchor inflationary expectations.

3

For example, using the de jure and several de facto classifications, Bleaney and Fransisco (2005) find that hard pegs reduce inflation and money growth in developing countries. The estimated coefficients of hard pegs are, however, not strictly comparable across their estimations since the sample composition varies considerably. Guisinger and Singer (2010) use a common sample but different sources for their de jure and de facto classifications. Moreover, they pool developed and developing countries in the same sample, which may explain why they find that pegs are not associated with lower inflation under either de jure or de facto classifications alone (previous studies using both de jure (Ghosh et al. 2003) and de facto (Rogoff et al., 2003) classifications find no association between the regime and inflation performance in developed countries).

4

We focus on EMDCs because they are more likely to benefit from importing credibility (by pegging to a strong anchor currency) than advanced economies, which tend to have credible policy institutions of their own. Within the sample of EMDCs, the results for emerging market economies (EMEs) are of particular interest as the differences in findings between de jure and de facto classifications is greatest for this group (see, for example, Rogoff et al., 2003).

5

Under a peg, the central bank subordinates its monetary policy (domestic credit expansion) to maintaining the nominal exchange rate constant; under a float, the central bank is free to respond to shocks, but cannot commit ex ante to the monetary policy that it will follow. For closet pegs, monetary policy is consistent with maintaining the peg, whereas for soft pegs it is not.

6

Alternative formulations, for example where workers are assumed to care about overall consumer price inflation rather than just domestic price inflation, are possible. While some of the specific results of the model depend on the precise formulation chosen, the model’s key results about the ranking of inflation rates under the various regimes is unchanged.

7

More generally, A can be interpreted as the marginal benefit of generating surprise inflation for any reason (for example, erosion of public debt).

8

When the central bank does not have the incentive to create surprise inflation(y¯=0), but the private sector believes that it has,y¯e>0, then πde=Aθy¯e/(1(1β)ν)>0,y=Aθ2y¯e/(1+Aθ2)<0 . Therefore, even though the central bank does not actually have the incentive to generate surprise inflation, its lack of credibility implies that the economy has an inflationary bias and output is lower than its natural rate, yielding a welfare costLFlt=1/2{(A(Aθ2y¯e)2+ση2)/(1+Aθ2)}>0 . This explains why disinflation under imperfect credibility will be costly (and why central banks with a history of high inflation often peg the currency in exchange-rate based stabilization programs).

9

When the central bank does not have the incentive to generate surprise inflation but the private sector believes that it hasy¯=0,y¯e>0 , then the condition becomes:ση2<y¯e2 .

10

Politically, as documented by Cooper (1971) and Frankel (2005), the incumbent government faces a significantly greater probability of losing office following devaluation from a de jure peg than otherwise. In terms of economic costs, the greater the private sector’s confidence that the peg will be maintained, the larger its foreign currency debt exposure is likely to be, making any eventual devaluation that much more costly.

11

This corresponds to Calvo and Reinhart’s (2002) “fear of floating” case.

12

That is,1ρ=Pr(A2θ2(ηy¯