Journal Issue
Share
Article

The Evolution of Exchange Rate Regimes Since 1990

Author(s):
Andrea Bubula, and Inci Ötker
Published Date:
September 2002
Share
  • ShareShare
Show Summary Details

I. Introduction

The choice of an optimal exchange rate regime is one of the major unresolved questions of international macroeconomics. With “intermediate regimes” having been at the center stage in most major currency crises in recent years, there has been growing support for the so called bipolar view that hard pegs and free floats are the only regimes compatible with the current degree of financial market integration, and that countries with high capital mobility are, or should be, abandoning intermediate regimes between these two extremes.23 The proponents of this view argue that a peg cannot be viable under capital mobility unless the country makes an irrevocable commitment to the peg and is prepared to support it absolutely. The only feasible alternative to such commitment would be to float the currency by putting the country under the discipline of the markets on a continuous basis.

Judging from official notifications by country authorities to the IMF, the evolution of exchange rate regimes since the late 1970s seems to support the bipolar view. Indeed, the proportion of IMF member countries with officially pegged exchange rates has halved, whereas that of floating countries has doubled. Furthermore, within this declining group of pegged regimes, the share of hard pegs increased over the past decade or so, as a growing number of countries adopted currency boards, joined currency unions, or formally dollarized.

A country’s actual exchange rate regime, however, can differ from its official notification. Over the past decade or so, a number of countries that had officially declared to be floating were in reality targeting stable exchange rates.4 Such a divergence has in turn been used to challenge the validity of the bipolar view (for example, Calvo and Reinhart, 2000). A natural question then is whether the bipolar view would still be supported if the analysis is based on de facto exchange rate regimes. Providing an answer to this question requires establishing a comprehensive database of the exchange rate regimes actually pursued by all IMF members.

Characterizing exchange rate policies actually pursued by countries has been one of the greatest challenges to empirical analyses of exchange rate regimes. Most of the previous studies have relied on the former IMF official classification that summarized, from 1975 to 1998, members’ exchange regimes based on their official or de jure notifications to the IMF.5 This classification suffered from many shortcomings. The most important was its failure to capture differences between what the countries claimed to be doing and what they were doing in practice and also to distinguish between very rigid forms of pegged regimes and softer pegs.6 To address these shortcomings, the IMF adopted a modified system in 1999, distinguishing between various types of pegged regimes and classifying exchange regimes based on countries’ de facto policies (see IMF, 1999 for details). This system presents a marked improvement over the former classification. Unfortunately, the lack to date of a historical database has thus far limited its usefulness for empirical analysis.

The first objective of this paper is to construct a monthly database of de facto exchange rate regimes of all IMF members since 1990 using the IMF’s current nomenclature adopted in 1999. This nomenclature classifies members’ exchange rate regimes along a flexibility spectrum, ranging from regimes with no separate legal tender to currency boards, conventional fixed or crawling pegs, horizontal bands, crawling bands, managed floating with no predetermined path for the exchange rate, and to independently floating rates. The database of this paper also provides refinements on some regime categories, distinguishing backward-looking crawls from forward-looking ones and tightly managed floats from other managed floating regimes. The classifications are based primarily on information obtained through bilateral discussions with or provision of technical assistance to member countries, and from regular contacts with IMF desk economists. The views drawn from these documents are supplemented with press reports, news articles, and other relevant papers, as well as with an analysis of the observed movements in exchange rates.

The second objective of the paper is to use this database to examine the evolution of various exchange rate regimes since 1990, for all IMF membership combined as well as for particular country subgroups. A simple analysis of trends in de facto exchange rate regimes appears to support earlier claims (for example, Fischer, 2001) that, over the past decade, the proportion of countries adopting intermediate regimes has been shrinking in favor of either greater flexibility or greater fixity. However, the trend away from intermediate regimes is mostly accounted for by developed countries and by emerging markets that are already, or increasingly, integrated with the international financial markets. Intermediate regimes, despite their decline, have continued to be a common practice in many nonemerging market developing countries. The paper also finds that the extent of the decline has not been uniform across different types of intermediate regimes, with a tendency to move from less flexible toward more flexible intermediate regimes.

A more detailed statistical examination of the observed trends is then carried out by estimating Markov transition probabilities between three broad categories: hard pegs, intermediate regimes, and floating regimes. This part of the analysis is related to Masson (2001), who conducted a similar analysis using annual data from two earlier classifications of exchange rate regimes—one that relied partly on the IMF de jure classification (Ghosh, Gulde, Ostry, and Wolf (GGOW), 1997) and another based purely on observed exchange rate and reserves developments (Levy-Yeyati and Sturzenegger, 1999). The findings are consistent with Masson (2001), in that the present paper finds no strong empirical support for the bipolar view (with intermediate regimes likely to be a part of the long-run distribution of exchange rate regimes). Yet, new evidence of structural instability in the data points to the need to obtain additional evidence from the future evolution of exchange rate regimes before concluding against the “vanishing middle” proposition.

The rest of the paper is organized as follows. Section II provides a brief review of the earlier work on classifying exchange rate regimes, describes the methodology of the regime classifications adopted in this paper, and discusses issues involved in classifying exchange rate regimes. Section III discusses the observed trends in members’ exchange rate regimes since 1990 to identify if there is any tendency toward declining intermediate regimes. Section IV estimates Markov transition probabilities of shifts among alternative exchange rate regimes to examine the hollowing out hypothesis more formally. Section V concludes.

II. Classifying Countries’ Exchange Rate Regimes

A. Background

From 1975 through 1998, the official IMF exchange regime classification system categorized members’ exchange rate policies based on their official notifications to the IMF (i) within 30 days of becoming a member and (ii) promptly after any changes in their regimes occurred. Members’ regimes were grouped according to their own official statements about the degree of exchange rate flexibility. Specifically, the de jure classification distinguished between three main categories: pegged regimes, regimes with limited flexibility (those that permit the exchange rate to fluctuate within a range or within a cooperative arrangement), and more flexible arrangements, in which the exchange rate is managed or allowed to float freely.

Among the key advantages of the de jure classification were its comprehensiveness in terms of country coverage, frequent updating (on a quarterly basis), and long history, and many empirical analyses of exchange rate regimes have relied on it. However, this system had a number of well-recognized shortcomings, the most important of which was its failure to capture differences between what the countries claimed to be doing and what they were doing in reality. Some countries with pegged regimes engineered frequent devaluations, using the exchange rate as a safeguard for export competitiveness, thereby making their regime less distinct from a flexible one. Others, claiming to have floating regimes, informally pegged or managed their exchange rates along a predetermined path;7 these reflected, among other things, attempts to avoid the political costs of visible devaluations if the peg were to be announced, the desire to use the exchange rate as an inflation anchor while avoiding the risk of one-way bets against their currencies, or the “fear of floating” in the presence of large foreign exchange exposures or high degree of dollarization. The resulting divergence between the de jure and de facto regimes reduced the transparency of exchange rate policy, making effective surveillance over members’ policies difficult and clouding the policy implications derived from research and policy work using this classification.8 A second weakness of the system was that, by lumping rigid forms of pegs together with softer pegs, it failed to acknowledge the different degree of monetary autonomy afforded by each regime.

With a view to addressing these shortcomings, the IMF adopted a new classification scheme based on de facto policies, which has become official since January 1999. The new system has also distinguished between various types of pegged regimes to reflect varying degrees of monetary autonomy and commitment to a given exchange rate path. After an extensive examination of the nature of various practices followed by members and also guided by the existing literature on the variety of exchange rate regimes, the following categories were identified: (1) regimes with no separate legal tender—including regimes with another currency as legal tender (formal dollarization) and currency unions; (2) currency boards; (3) conventional fixed pegs (against a single currency or a basket of currencies); (4) pegged exchange rates within horizontal bands; (5) crawling pegs; (6) crawling bands; (7) managed floating with no predetermined path for exchange rate; and (8) independently floating.

While this new scheme is a marked improvement over the former classification, the lack of a historical database has limited its usefulness for empirical analysis. In the absence of such a database, for example, a number of researchers relied on the observation of a few data points available on the new classification in their analysis of exchange rate regimes.9 This in turn provided the main motivation for the present paper, namely to construct a historical database on de facto regimes for all member countries that extends the current classification back in time, from end-2001 to the beginning of 1990. Constructing such a database maintains the main advantage of the old classification in that it covers all member countries and is consistent with the IMF’s new official classification, which will continue to be updated as long as the new nomenclature remains in place.10 Those advantages differentiate the present paper from several other attempts at identifying differences between de facto and de jure regimes, which are discussed below.

In a comprehensive study of the macroeconomic performance of different exchange rate regimes during the period 1960–90 for about 140 countries, GGOW (1995, 1997) combine the IMF de jure classification with the actual exchange rate behavior so as to differentiate between official and actual policies. The authors supplement the stated intentions of the central bank, as suggested by the de jure classification, by categorizing the pegged regimes according to whether or not changes in parity have been frequent or infrequent, and distinguish between countries with heavy or no intervention, based on a survey of IMF desk officers for the countries in the sample. The authors propose a very fine taxonomy of regimes (about 25 categories), grouped under nine more-aggregated categories: single currency pegs, SDR pegs, other published basket pegs, secret basket pegs, cooperative systems, unclassified floats, floats within a predetermined range, floats without a predetermined range, and pure floats. The information content of the classification contributed partly to the selection of the nomenclature behind the current IMF classification back in 1997. The regime classifications specified, however, also do not distinguish between soft pegs and the hardest forms of peg commitments, and some countries are placed in unclassified categories.11 Post-1990 updates are also not automatic in that they require processing of the relevant raw data.12

In contrast to both GGOW (1995, 1997) and the present paper, Levy-Yeyati and Sturzenegger (1999) ignore the IMF’s old official classification altogether. Their analysis, based on an original framework from Holden, Holden, and Suss (1979) instead, provides an alternative classification over the period from 1990–98 that aims to capture actual exchange rate policies on the sole basis of the observed behavior of official exchange rates and international reserves, with the latter presumed to measure the extent of foreign exchange intervention.13 Exchange rate regimes are classified into four main categories (flexible, dirty float, crawling peg, and fixed) by using cluster analysis of the monthly percentage changes in the nominal exchange rate, the relative volatility of exchange rate levels, and the volatility of reserves.14 Countries without significant variability in either variable are included in the “inconclusive” group.15 While such a classification has some conceptual attractiveness for some purposes, it has some limitations as discussed below. The classification also covers only a subset of 110 IMF member countries for which the relevant exchange rate and reserve data are available and leaves a number of countries in the inconclusive category.16

In a contemporaneous paper, Reinhart and Rogoff (2002) present a monthly database of exchange rate regimes for 153 countries over the period 1946-2001. The authors recognize the merits of the new IMF classification and adopt a broadly similar nomenclature to classify the regimes.17 Exchange rate regimes are classified and verified mainly by applying a variety of descriptive statistics to either official or market-determined parallel exchange rates, depending on whether a given country has unified or multiple exchange rates. The frequency and the length of the database, as well as the application of a systematic approach to classifying regimes, are its main advantages. While the database is quite comprehensive, the lack of full availability of (official or parallel market) exchange rates tends to limit the country coverage. Furthermore, updating the database beyond 2001 requires processing of the raw data as well as classifying some exchange rate regimes based mainly on actual exchange rate movements, which, as discussed below, presents some potential limitations.

B. The Approach of this Paper

Classifications based solely on the behavior of exchange rates and reserves have limitations in terms of country coverage, in view of the limited availability of relevant exchange rate and official reserves data for all countries. More important, sole reliance on such data may in many cases lead to incorrect inference about the actual exchange rate regime. First, exchange rate stability could be the outcome of stable fundamentals or lack of significant shocks to the economy. Since many different aspects of economic policy have an effect on the exchange rate, it is difficult to disentangle the actions aimed at stabilizing the exchange rate from those directed at other goals (for example, direct inflation targeting). A high sample variance may also be due to a one-time large devaluation under a pegged exchange regime and thus may not indicate greater exchange rate flexibility.18 Differences in the size and depth of foreign exchange markets may also affect exchange rate volatility. Moreover, statistical analysis of official exchange rates would be misleading if the country maintains multiple exchange rates and the reported official rate affects only a fraction of the foreign exchange transactions, while remaining transactions are effected at officially recognized market-determined rates.19

Supplementing the exchange rate with official reserves data does not entirely solve the drawbacks of focusing on quantitative measures alone. Changes in reserves may, for instance, be linked to debt or reserve management strategies, and not necessarily to an exchange rate stabilization motive. Reserve data could be distorted by valuation changes and official borrowing or repayments; no allowance is made for intervention by other countries that may negate the need for exchange market action by the country under consideration; and there may not be a linear relationship between changes in exchange rates and reserves (see Holden, Holden, and Suss (1979)). In addition, the extent of intervention may be concealed if currencies are defended through intervention through swaps and forwards. Even more important, the authorities may affect the exchange rate without any direct purchase or sale of foreign exchange in the market. For example, in less-developed economies with small and shallow foreign exchange markets, indirect intervention may take the form of administrative foreign exchange controls and regulations or “moral suasion,”20 all of which attempt to limit the ability of market forces to set the exchange rate freely.

Given the limitations associated with a classification based purely on quantitative analysis, this paper uses a combination of quantitative and qualitative analysis in constructing the database on de facto regimes—an approach which coincides with that pursued by the current official IMF classification. The primary source in identifying the de facto policies is the information obtained through bilateral consultation discussions with and provision of technical assistance to member countries, and regular contacts with IMF country desk economists. These views are then supplemented with other sources of information, including press reports, news articles, and other relevant papers, and supported by an analysis of observed exchange rate and reserves behavior, to reach a final view on the de facto regime.

Assessing what countries actually do with their exchange rate policy is clearly a complicated task, except in the case of hard peg regimes, where the authorities are forced by their legal commitment to implement the de jure regime. The task is particularly difficult where countries officially announcing (managed or independently) floating regimes use systematic direct (through sale or purchase of foreign exchange by the central bank or its agents) or indirect intervention (through foreign exchange regulations or moral suasion) to target informally the exchange rate. The approach taken to identify these cases has been to supplement data on exchange rates and reserves with evidence contained in various documents to identify whether or not authorities were pursuing informal targets. The following guidelines were applied in assigning to countries the specific regime categories:

  • Exchange rate stability by itself was not sufficient to classify a given regime as a peg, unless documents from a variety of sources provided evidence that the authorities formally or informally sought to keep the exchange rate stable vis-à-vis a given currency or a basket of currencies. When available information indicated that the authorities targeted to keep the exchange rate stable and the exchange rate remained within a range less than 2 percent for at least four months vis-à-vis a given currency, the regime was classified de facto as a conventional fixed peg.
  • Basket pegs are more difficult to verify, particularly when the composition of baskets and/or weights are undisclosed. Aside from the unambiguous cases where authorities were clearly not following the announced policy (e.g., when the relevant currency stayed constant vis-a-vis another currency—e.g., the U.S. dollar for Bangladesh, or the Indian rupee for Nepal), the de jure classification of basket pegs was maintained.
  • Countries were assigned crawling peg regimes (with or without bands) when there was evidence in documents (supported by nominal and real exchange rate data) of a formal announcement or an informal pursuit of a policy of gradual depreciation of the currency by small amounts with or without a formal or unannounced band, according to a formula. Distinction was made between forward- and backward-looking crawling pegs and bands to indicate differing degrees of commitment to a given exchange rate path and role of the exchange rate as a nominal anchor. Information contained in the documents was used to make this distinction.
  • Countries were also classified as pursuing a (fixed or crawling) peg regime when they announced a band regime, but limited exchange rate movements through heavy intramarginal intervention within narrow inner bands (e.g., Belarus, Brazil, and Russia).
  • A regime was classified as a managed float rather than a peg, when the officially announced regime was a fixed peg, but the official rate was subject to frequent devaluations within very short time periods (e.g., as in Bangladesh during 1995-97), or when the country maintained multiple exchange rates and the majority of transactions were effected at a more market-based (official or officially recognized secondary) rate, rather than at the fixed official rate (e.g., Myanmar and Nigeria since 1995). As a general rule, when countries maintained multiple exchange rates, the classification of the regime was based on the determination of the exchange rate that applied to most transactions.
  • In choosing between managed and independently floating regimes, the decisive factor was the presence of some evidence that direct or indirect intervention was aimed at countering the long-term trend of the exchange rate beyond what is warranted to prevent excessive exchange rate volatility. In tightly managed floats, authorities have been considered to be aiming at a stable exchange rate path by closely monitoring and controlling the exchange rate, where such stability could not be clearly identified as distinct episodes of a peg, a crawl, or a band (e.g., Paraguay and Singapore).21

This approach of classifying exchange rate regimes offers many advantages over an approach based solely on quantitative analysis. However, its heavy reliance on the quality of available information may introduce some margin of uncertainty where the authorities may be reluctant to make actual policies publicly known, and when such reluctance influences the coverage of documents in providing a more candid description of the actual regime. There is also an element of judgment in interpreting the information contained in a particular way. Nevertheless, using the same set of criteria and a variety of information sources in forming a judgment about the nature of the exchange rate regimes for all members limits the risk of misplacing a country’s regime relative to the other countries within the membership.

The database covers the period from January 1990 to December 2001. Diagram 1 presents the 13 categories, after distinguishing between forward- and backward-looking crawling pegs and bands and between tightly and other managed floating with no predetermined path for the exchange rate. Box 1 describes each category. For the following analysis of the bipolar view of exchange rate regimes, we define categories (1)–(3) as “hard pegs,” (4)–(11) as “intermediate regimes” (i.e., consisting of tightly managed floats and “soft pegs” such as conventional fixed and crawling pegs and horizontal and crawling bands), and (12)–(13) as “floating regimes.” The classifications for all IMF members from 1990 to 2001 are provided in Appendix I on an end-year basis.22

Diagram 1.De Facto Classification of Exchange Rate Regimes

1/ Excludes tightly managed floats.

III. The Evolution of Exchange Rate Regimes Since 1990

Table 1 shows the evolution of various aggregated regime categories over 1990-2001 based on countries’ de facto policies (Appendix II provides the evolution of more disaggregated categories). Despite a marked decline in their share, various forms of pegged regimes have remained the relatively more dominant type of monetary arrangement among the IMF members, in contrast to what is suggested by the de jure system (Table 2). As of end-2001, more than half of the member countries (56 percent) were still pursuing various forms of pegged arrangements, compared with the 80 percent at the beginning of the 1990s. There has been a discernible shift in the composition of pegged regimes during this period, however, away from soft peg regimes to harder pegs, with the latter more than doubling their share (Figure 1).

Box 1.Classification of Exchange Rate Regimes

Exchange Regimes with Another Currency as Legal Tender (Formal “Dollarization”)

The country uses the currency of another country, which circulates as the sole legal tender. Dollarization is a complete surrender of the authorities’ independent control over domestic monetary policy, and as such, can be viewed as the hardest form of a pegged regime.

Exchange Regimes with No Separate Legal Tender (Currency Unions)

The member belongs to a monetary or currency union in which they share the same legal tender. By adopting such regimes the authorities also surrender control over domestic monetary policy and, hence, such regimes are also viewed as a hard exchange rate peg.

Currency Board Arrangements (CBAs)

An exchange rate regime based on an explicit legislative commitment to exchange domestic currency for a specified foreign currency at a fixed exchange rate, combined with restrictions on the issuing authority to ensure the fulfillment of its legal obligation. This implies that domestic currency remain fully backed by foreign assets, eliminating traditional central bank functions such as monetary control and lender of last resort. While leaving little scope for discretionary monetary policy, some flexibility may be afforded depending on the strictness of the board’s rules.

Conventional Fixed Peg Arrangements: Vis-à-Vis a Single Currency or a Currency Composite

The country (formally or de facto) pegs its currency at a fixed rate to another currency or a basket of currencies, where a basket is formed from the currencies of major trading or financial partners and weights reflect the geographical distribution of trade, services, or capital flows. The currency composites can also be standardized, such as that of the SDR. There is no commitment to keep the parity irrevocably. The exchange rate may fluctuate within a narrow margin of less than ±1 percent around a central rate or the maximum and minimum value of the exchange rate remains within a narrow margin of 2 percent for at least three months. The authorities stand ready to keep the fixed parity through direct (i.e., via sale/purchase of foreign exchange in the market) or indirect intervention (e.g., via aggressive use of interest rate policy, imposition of foreign exchange regulations or exercise of moral suasion that constrain foreign exchange activity, or through intervention by other public institutions). Flexibility of monetary policy, though limited, is greater than in hard pegs, as traditional central banking functions are still possible, and the authorities can adjust the level of the exchange rate, though relatively infrequently.

Crawling Pegs: Forward and Backward Looking

The currency is adjusted periodically vis-à-vis a single currency or a basket in small amounts at a fixed rate or in response to changes in selective quantitative indicators (past inflation differentials with major trading partners, differentials between the targeted or projected inflation with major trading partners, differentials between official and parallel market rates, etc.). Distinction is made between backward- and forward-looking crawls: the crawl is viewed as “backward looking” when the crawl is set to generate inflation adjusted changes in the currency (i.e., when it aims to passively accommodate past inflation differentials under a real exchange rate rule); and as “forward looking” when the exchange rate is adjusted at a preannounced fixed rate and/or set below projected inflation differentials, typically when the exchange rate is envisaged to have an anchor role. Maintaining a credible crawling peg imposes similar constraints on monetary policy as a fixed peg system, particularly in a forward-looking crawl, as the authorities are expected to intervene to ensure the targeted fixed depreciation path. The degree of intervention in a backward-looking crawl is expected to be less given the lack of commitment to a fixed depreciation path and the absence of a need to anchor expectations.

Pegged Exchange Rates Within a Horizontal Band

The currency is allowed to move within margins of fluctuation of at least ±1 percent around a formal or a de facto fixed central rate. The authorities stand ready to defend the limits of the band through direct or indirect intervention to maintain the exchange rate within these limits. Some limited degree of monetary policy discretion can be afforded, with the degree of discretion depending on the band width.

Pegged Exchange Rates Within Crawling Bands: Forward and Backward Looking

The currency is maintained within fluctuation margins of at least ±1 percent around a formal or a de facto central rate, which is adjusted periodically in small amounts at a fixed rate, or in response to changes in selective quantitative indicators. A similar distinction is made between backward- and forward-looking crawls. The degree of exchange rate flexibility is a function of the band width; bands can be chosen to be either symmetric or fixed around a crawling parity or to widen gradually with an asymmetric choice of the crawl of upper and lower bands (in the latter case, there may not be a preannounced central rate). The commitment to maintain the exchange rate within the band continues to impose constraints on monetary policy, with the degree of policy independence a function of the bandwidth.

Tightly or Other Managed Floating with No Predetermined Path for the Exchange Rate

The authorities influence exchange rate movements through active intervention to counter the long-term trend of the exchange rate, without specifying a predetermined exchange rate path, or without having a specific exchange rate target (“dirty floats”). Intervention may be direct or indirect. Indicators for managing the rate are broadly judgmental (e.g., balance of payments position, international reserves, parallel market developments), and adjustments may not be automatic. Distinction is made between “tightly managed floating” (where intervention takes the form of very tight monitoring that generally results in a stable exchange rate without having a clear exchange rate path, so as to permit the authorities an extra degree of flexibility in deciding the tactics to achieve a desired path) and “other managed floating” (where the exchange rate is influenced in a more ad hoc fashion).

Independently Floating

The exchange rate is market determined; any foreign exchange intervention aims at moderating the rate of change and preventing undue fluctuations in the exchange rate that are not justified by economic fundamentals, rather than at establishing a level for the exchange rate, in these regimes, monetary policy is in principle independent of exchange rate policy.

Figure 1.IMF Membership: The Evolution of Pegged Exchange Rate Regimes, 1990 and 2001

Source: Authors’ calculations.

Table 1.Evolution of Exchange Rate Regimes by Country Group(In percent of members in the given category)
199019911992199319941995199619971998199920002001
Whole membership
Hard pegs 1/15.716.119.415.916.216.216.218.318.324.224.725.8
Intermediate regimes 2/69.266.556.158.856.858.958.453.248.940.941.438.7
Of which: soft pegs 3/64.260.250.052.748.149.250.348.445.735.533.930.1
Floating regimes 4/15.117.424.425.327.024.925.428.532.834.933.935.5
Total100.0100.0100.0100.0100.0100.0100.0100.0100.0100.0100.0100.0
Developed countries
Hard pegs 1/0.00.04.24.24.24.24.24.24.250.050.054.2
Intermediate regimes 2/73.973.950.054.254.254.262.558.358.312.512.54.2
Of which: soft pegs 3/73.973.950.050.050.050.058.358.358.312.512.54.2
Floating regimes 4/26.126.145.841.741.741.733.337.537.537.537.541.7
Total100.0100.0100.0100.0100.0100.0100.0100.0100.0100.0100.0100.0
Developing countries
Hard pegs 1/18.418.821.817.718.018.018.020.420.420.421.021.6
Intermediate regimes 2/68.465.257.159.557.159.657.852.547.545.145.743.8
Of which; soft pegs 3/62.558.050.053.247.849.149.146.943.838.937.034.0
Floating regimes 4/13.215.921.222.824.822.424.227.232.134.633.334.6
Total100.0100.0100.0100.0100.0100.0100.0100.0100.0100.0100.0100.0
Emerging market countries5/
Hard pegs 1/6.710.09.79.49.49.49.412.512.512.515.615.6
Intermediate regimes 2/76.766.764.575.068.881.378.156.353.140.637.534.4
Of which: soft pegs 3/63.353.351.662.553.159.462.550.046.934.428.125.0
Floating regimes 4/16.723.325.815.621.99.412.531.334.446.946.950.0
Total100.0100.0100.0100.0100.0100.0100.0100.0100.0100.0100.0100.0
Developed and emerging market countries
Hard pegs 1/3.85.77.37.17.17.17.18.98.928.630.432.1
Intermediate regimes 2/75.569.858.266.162.569.671.457.155.428.626.821.4
Of which: soft pegs 3/67.962.350.957.151.855.460.753.651.825.021.416.1
Floating regimes 4/20.824.534.526.830.423.221.433.935.742.942.946.4
Total100.0100.0100.0100.0100.0100.0100.0100.0100.0100.0100.0100.0
Non-emerging market developing countries
Hard pegs 1/21.721.324.819.820.220.220.222.322.322.322.323.1
Intermediate regimes 2/66.064.855.255.654.354.352.751.546.246.247.746.2
Of which: soft pegs 3/62.359.349.650.846.546.545.746.243.140.039.236.2
Floating regimes 4/12.313.920.024.625.625.627.126.231.331.530.030.8
Total100.0100.0100.0100.0100.0100.0100.0100.0100.0100.0100.0100.0
Source: Authors’ estimates.

Comprises arrangements with another currency as legal tender (i.e., dollarization), currency unions, and currency boards.

Comprises soft pegs plus tightly managed floating regimes.

Comprises conventional fixed pegs vis-à-vis a single currency or a basket, horizontal bands, and crawling pegs and crawling bands.

Comprises independently floating regimes and managed floating with no predetermined exchange rate path, excluding tightly managed floats.

Includes 32 countries or regions: Argentina, Brazil, Bulgaria, Chile, China, Colombia, the Czech Republic, Egypt, Ecuador, Hong Kong SAR, Hungary, India, Indonesia, Israel, Jordan, Korea, Malaysia, Mexico, Morocco, Nigeria, Pakistan, Panama, Peru, the Philippines, Poland, Russia, Singapore, South Africa, Sri Lanka, Thailand, Turkey, and Venezuela.

Source: Authors’ estimates.

Comprises arrangements with another currency as legal tender (i.e., dollarization), currency unions, and currency boards.

Comprises soft pegs plus tightly managed floating regimes.

Comprises conventional fixed pegs vis-à-vis a single currency or a basket, horizontal bands, and crawling pegs and crawling bands.

Comprises independently floating regimes and managed floating with no predetermined exchange rate path, excluding tightly managed floats.

Includes 32 countries or regions: Argentina, Brazil, Bulgaria, Chile, China, Colombia, the Czech Republic, Egypt, Ecuador, Hong Kong SAR, Hungary, India, Indonesia, Israel, Jordan, Korea, Malaysia, Mexico, Morocco, Nigeria, Pakistan, Panama, Peru, the Philippines, Poland, Russia, Singapore, South Africa, Sri Lanka, Thailand, Turkey, and Venezuela.

Table 2.Evolution of Exchange Rate Regimes(In percent of IMF membership)
199019911992199319941995199619971998199920002001
De Jure Classification1/
Pegged regimes 2/65.461.358.149.147.244.445.345.144.5
Of which: limited flexibility9.29.07.87.47.97.88.88.89.3
Floating regimes34.638.741.950.952.855.654.754.955.5
Independently floating16.318.126.332.032.630.028.725.324.7
Managed floating18.320.615.618.920.225.626.029.730.8
Total100.0100.0100.0100.0100.0100.0100.0100.0100.0
De Facto Classification
Pegged regimes79.976.469.468.764.365.466.566.764.059.758.655.9
Hard pegs 3/15.716.119.415.916.216.216.218.318.324.224.725.8
Soft pegs 4/64.260.250.052.748.149.250.348.445.735.533.930.1
Floating regimes20.123.630.631.335.734.633.533.336.040.341.444.1
Independently floating5.77.512.813.713.014.112.414.516.718.319.421.5
Managed floating 5/14.516.117.817.622.720.521.118.819.422.022.022.6
Total100.0100.0100.0100.0100.0100.0100.0100.0100.0100.0100.0100.0
Memorandum item:
Intermediate regimes 6/69.266.556.158.856.858.958.453.248.940.941.438.7
Sources: IMF, Annual Report on Exchange Arrangements and Exchange Restrictions, various issues; and authors’ estimates.

Since September 1998, the de jure classification has not been updated.

Includes arrangements with no separate legal tender, currency boards, conventional fixed pegs and horizontal bands, and regimes with limited flexibility within a band and ERM.

Comprises arrangements with another currency as legal tender (i.e., dollarization), currency unions, and currency boards.

Comprises conventional fixed pegs vis-à-vis a single currency or a basket, horizontal bands, and crawling pegs and bands.

Includes tightly managed floating regimes.

Defined as to include tightly managed floating regimes and soft peg regimes.

Sources: IMF, Annual Report on Exchange Arrangements and Exchange Restrictions, various issues; and authors’ estimates.

Since September 1998, the de jure classification has not been updated.

Includes arrangements with no separate legal tender, currency boards, conventional fixed pegs and horizontal bands, and regimes with limited flexibility within a band and ERM.

Comprises arrangements with another currency as legal tender (i.e., dollarization), currency unions, and currency boards.

Comprises conventional fixed pegs vis-à-vis a single currency or a basket, horizontal bands, and crawling pegs and bands.

Includes tightly managed floating regimes.

Defined as to include tightly managed floating regimes and soft peg regimes.

Together with the significant growth in the number of floating regimes, the trend toward harder pegs implies a shift away from intermediate regimes toward the two ends of the spectrum of exchange rate regimes, providing some support for the bipolar view of exchange rate regimes (Figure 2). The share of intermediate regimes dropped by about 30 percentage points over the past decade, with a 10 percent and 20 percent increase in the share of hard pegs and floating regimes, respectively. Exits from intermediate regimes since 1990 indicate that most intermediate regimes exited to a floating regime rather than to hard pegs, although certain intermediate regimes (for example, backward- and forward-looking crawling pegs and backward-looking crawling bands) were replaced by other intermediate regimes before eventually exiting to floating regimes (Figure 3).

Figure 2.Evolution of Exchange Rate Regimes, 1990–2001

(Percent of whole membership)

Source: Authors’ calculations.

Figure 3.Exits from Intermediate Regimes, 1990–2001

(In percent of all exits within each group)

Source: Authors’ calculations.

While intermediate regimes have been shrinking, the extent of the decline has not been uniform across different types of intermediate regimes (Table 3 and Figure 4). Many countries abandoned soft peg regimes, but others have continued to hold on to exchange rate targets by adopting more flexible forms of intermediate regimes.23 The shares of crawling bands and tightly managed floats in particular have risen, while those of horizontal bands, crawling pegs, and conventional fixed pegs have declined (the latter reflecting the sharp drop in basket pegs). In many cases, the emergence of tensions between the goals of disinflation and competitiveness has been an important factor in abandoning, willingly or unwillingly, the less flexible intermediate regimes toward more flexible ones.24 Crawling pegs, which had once been very popular among high-inflation countries, were abandoned during crises or made more flexible with the introduction of a band.25 Crawling bands offered greater flexibility in coping with capital flows and in avoiding severe exchange rate misalignments, while retaining, to some degree, the anchor role of the exchange rate with a regular path of mini-devaluations. The share of tightly managed floats rose as authorities attempted to keep the anchor role of the exchange rate without precommitting to a specific target path.

Figure 4.IMF Membership. The Evolution of Intermediate Regimes Between 1990 and 2001

Source: Authors’ calculations.

Table 3.The Evolution of Intermediate Exchange Rate Regimes(In percent of all intermediate regimes, end-year)
199019911992199319941995199619971998199920002001
Conventional fixed pegs63.658.960.457.053.351.447.251.552.759.257.156.9
Fixed peg to single currency35.532.733.733.635.236.734.338.440.746.144.243.1
Fixed peg to basket28.226.226.723.418.114.713.013.112.113.213.013.9
Horizontal band12.712.110.912.114.311.916.716.217.66.67.86.9
Crawling peg13.614.012.915.011.411.011.113.112.111.89.15.6
Crawling band2.75.65.05.65.79.211.110.111.09.27.88.3
Tightly managed float7.39.310.910.315.216.513.99.16.613.218.222.2
Total100.0100.0100.0100.0100.0100.0100.0100.0100.0100.0100.0100.0
Memorandum items:
Members with intermediate regimes1101071011071051091089991767772
(in percent of total membership)69.266.556.158.856.858.958.453.248.940.941.438.7
Total membership159161180182185185185186186186186186
Source: Authors’ estimates.
Source: Authors’ estimates.

The shift away from intermediate regimes has also been more pronounced among the developed and emerging market countries that are highly integrated with international capital markets (see Figure 5 and also Fischer (2001)).26 Intermediate regimes have almost disappeared in the developed countries. The substitution of the ERM bands with a monetary union accounted for most of the decline, while a part of the fall was owed to the replacement of band regimes with floating rates during the ERM turmoil in 1992 (Norway, Sweden, and the United Kingdom) or in the context of a change in the overall monetary policy framework (for example, Iceland recently). In the emerging market countries, there has been a marked shift toward floating regimes. Many moved to more flexible regimes following a sharp reversal of capital inflows during the 1990s, while others as a result of a deliberate policy choice to enhance monetary policy in coping with capital flows.27 The move to greater flexibility also reflected a desire to limit speculative capital flows typically encouraged by pegged or tightly managed exchange rate regimes. Still, several emerging markets adopted more rigid exchange rate regimes (Argentina, Bulgaria, Ecuador), sharply reducing or entirely giving up monetary autonomy, with the hope that the greater discipline imposed by such regimes would enhance policy credibility and make currencies less subject to speculative pressures.

Figure 5.Trend Toward Polarization of Exchange Rate Regimes Across Country Groups in 1990 and 2001 1/

(In percent of membership in each group)

Source: Staff estimates.

1/ Hard pegs = formal dollarization+currency unions+currency boards intermediate = conventional fixed pegs+horizontal bands+crawling pegs+crawling bands+tightly managed floats floating = independently floats+ other managed floats with no predetermined path for the exchange rate.

2/ The definitions of the developed and developing countries coincide with that of the IFS. The list of emerging market countries is based on a number of existing definitions that combine the countries included in the Emerging Markets Bond Index Plus (EMBI+) and Morgan Stanley Capital International (MSCI) index, with a few exceptions: Greece is included in the developed countries group and Singapore and Hong Kong SAR are included in the emerging countries group. This gives a list of 32 countries: Argentina, Brazil, Bulgaria, Chile, China, Colombia, the Czech Republic, Egypt, Ecuador, Hong Kong, Hungary, India, Indonesia, Israel, Jordan, Korea, Malaysia, Mexico, Morocco, Nigeria, Pakistan, Panama, Peru, the Philippines, Poland, Russia, Singapore, South Africa, Sri Lanka, Thailand, Turkey, and Venezuela.

The greater degree of regime polarization in these countries has been attributed to the principle of “impossible trinity,” that is, that only two of the three goals of exchange rate stability, capital mobility, and monetary policy independence can be attained simultaneously (see Fischer, 2001, and Frankel, 1999). The viability of intermediate regimes has been questioned as countries attempted to maintain a stable exchange rate while directing monetary policy to achieve domestic objectives in an environment with increased capital mobility. The growing integration of financial markets in recent years has in turn forced countries to choose between exchange rate stability or monetary independence—hence the apparent polarization of the regimes.28 Intermediate regimes have remained a feasible alternative for those countries that have not yet been open to capital flows.

Regardless of the underlying causes, however, the observed trend away from the intermediate regimes cannot by itself provide strong evidence that intermediate regimes will vanish eventually, as some proponents of the “hollowing middle” or “bipolar” view have argued. A more formal analysis of this hypothesis is carried out in the following section.

IV. Evolution of Exchange Rate Regimes: Evidence From a Markov Analysis

Whether the observed trend will persist until all intermediate regimes disappear can be examined more formally by analyzing if there were no exits from either hard pegs or floats (that is, if hard pegs or floats were an “absorbing state”), or if hard pegs and floats together formed a “closed set” with no transitions to intermediate regimes from the other two. A statistical test of this hypothesis was conducted in Masson (2001), who applied a Markov chain model to estimate the probabilities of transitions between three regime categories of hard pegs, intermediate regimes, and floats.29 In testing these hypotheses, Masson used two alternative regime classifications (GGOW, 1997; and Levy-Yeyati and Sturzenegger, 1999) with somewhat different definitions of hard pegs and floating regimes than those adopted in this paper.30 The paper concluded that neither data provided firm support for a substantial move away from middle regimes, and that these regimes will continue to exist in steady state, or at least, will take a very long time to disappear.

The estimated probabilities of transitions between hard pegs, intermediate regimes, and floating regimes based on the de facto regime classifications of this paper are broadly consistent with the main finding of Masson (2001), despite the differences in the frequency of the data and the use of different regime classifications. Table 4 presents these probabilities over 186 countries using monthly data over the period 1990–2001. For example, given that the current month’s regime is a hard peg, the estimated probability of remaining in a hard peg regime next month is 99.83 percent (i.e., p11 = 0.9983) and that of moving to an intermediate or a floating regime is 0.06 percent and 0.10 percent, respectively (i.e.,p12 = 0.0006 and p13 = 0.0010). Similarly, the estimated probability of moving next month from a floating regime to a hard peg or an intermediate regime are 0.03 percent and 1.07 percent, respectively.31

Table 4.Estimated Transition Matrix for All Countries, 1990–2001
Estimated probability of regime in t+1
Regime in period tHard pegsIntermediateFloatingTotal observations
Hard pegs0.99830.00060.0014,840
Intermediate0.00120.99090.007813,915
Floating0.00030.01070.9896,791
Total25,546
Memorandum items:
In percent:
Shares in Jan. 199015.2970.0614.65157 members
Shares in Dec. 200125.8138.7135.48186 members
Steady-state distribution32.9437.4329.63
Source: Authors’ estimates.
Source: Authors’ estimates.

These probabilities suggest that all three regimes are highly persistent (the most persistent being the hard pegs, followed by intermediate regimes, then floats), suggesting a high probability of remaining within a given regime once entered.32 However, the matrix is also “irreducible” (i.e., it is possible to reach any regime from any other regime), as there have been transitions from hard pegs to both intermediate and floating regimes, and from floating regimes to both intermediate regimes and hard pegs.33 Strictly speaking, then, the existence of exits from hard pegs and floats to intermediate regimes (that is, nonzero transition probabilities to intermediate regimes) provides clear evidence that the hypothesis of “hollowing out of the middle” is not supported by the data; that is, neither hard pegs or floats are “absorbing states,” nor do they form a “closed set” together.34

The transition probabilities over the whole membership therefore suggest that despite the decline in their share, intermediate regimes would be part of the long-run distribution of exchange regimes based on historical data for transitions and assuming that there are no major shocks to the system. On the one hand, there is a clear decline in the prevalence of intermediate regimes: at the beginning of the 1990s, the countries with intermediate regimes represented close to 70 percent of all membership, while at end-2001 only 39 percent maintained such regimes. On the other hand, the transition probabilities imply that intermediate regimes would not disappear in the long run, as suggested by the steady-state distribution of regimes that assign more than one-third share to intermediate regimes—a higher share than those of hard peg and floating regimes.35 If there are structural shifts over time, however (e.g., due to a sharp increase in capital mobility for the whole membership or to political shocks that may lead to more currency unions), the distribution of regimes in the future may differ from that in steady state. Such shifts could cause unpredictable changes in the direction of regimes.

The data indeed point to some evidence of structural instability over the sample period—a result that calls for a cautious interpretation of the implications suggested by the steady-state distribution. Dividing the sample into two equal subperiods, the structural stability test rejects the null hypothesis that the transition probabilities estimated from each subsample are the same as the parameters from the whole sample (the test statistic, 39.5, is significant with a p-value < 0.001). More detailed structural stability tests (discussed in Appendix III) also suggest that hard pegs and floating regimes became more persistent in the second subperiod (i.e., since 1996). The evidence of a significant increase in the retention rate of the two poles argues in favor of the bipolar view, since through time the extremes of the flexibility spectrum appear to be more persistent. The insignificant change in the persistence of intermediate regimes between the two periods, however, suggests that the trend toward polarization may not be the result of an acceleration in the exit rate from intermediate regimes, but instead of the fact that in the later period countries became less likely to change regimes once an exit occurs from intermediate regimes. Additional evidence from the future evolution of exchange rate regimes would hence be necessary to establish more concrete evidence against the vanishing middle proposition.

Table 5 provides some additional insight from the dynamics of transitions for finer categories of intermediate regimes to see how different intermediate regimes are moving toward the two poles. The diagonal elements of the transition matrix indicate that basket pegs, forward-looking crawling bands, and single currency pegs were the most persistent, and thus the most durable, regimes within the intermediate regime category, while crawling pegs, backward-looking crawling bands, and horizontal bands were the least enduring. The latter regimes, as a result, have been subject to more frequent exits compared with forward-looking crawling bands and conventional fixed pegs (Table 6 and Figure 6).

Figure 6.Exit Frequency and Regime Persistence Across Intermediate Regimes, 1990–2001

Sources: Tables 5 and 6.

Table 5.All Countries: Estimated Transition Probabilities with a Finer Categorization of Intermediate Regimes(January 1990–December 2001)
Hard

Pegs
Single

Currency Peg
Basket PegForward-

Looking

Crawling Peg
Horizonta

1Band
Forward-

Looking

Crawling Band.
Backward-

Looking

Crawling Peg
Backward-

Looking

Crawling Band
Tightly

Managed
FloatingTotal

Observations
Hard pegs0.99830.000600000000.00104,840
Single currency peg0.00100.98550.00040.00060.00100.00020.00100.00020.00180.00845,024
Basket peg00.00080.99020.00040.001500.000800.00080.00572,653
Forward-looking crawling peg00.005100.97300.00170.01010.0034000.0067593
Horizontal band0.00570.0023000.98110.001100.00110.00110.00751,742
Forward-looking crawling band00.0013000.00260.98940000.0066757
Backward-looking crawling peg0.00090.00260.00090.00530.00180.00090.97640.00180.00260.00701.142
Backward-looking crawling band00.0037000.00370.011000.970600.0110272
Tightly managed0.00060.003500.0006000.002300.98210.01101,732
Floating0.00030.00400.00010.00040.00150.00010.00030.00010.00410.98906,791
25,546
Share on Jan. 199015.2928.0319.113.188.280.007.011.273.1814.65157
Share on Dec. 200125.8116.675.381.612.693.230.540.008.6035.49186
Steady state32.2714.221.261.474.193332.150.559.2831.29
Source: Authors’ estimates.
Source: Authors’ estimates.
Table 6.The Frequency of Regime Shifts Under Alternative Exchange Rate Regimes by Members, 1990–2001
Toward Greater FlexibilityToward Less FlexibilityTotal
Total

number 1/
Frequency

(In percent) 2/
Total

number 1/
Frequency

(In percent) 2/
Total

number 1/
Frequency

(In percent) 2/
Exit from
Hard peg regimes80.16----80.16
Formal dollarization70.8770.87
Currency unions------------
Currency boards10.12----10.12
Intermediate regimes1641.18580.422221.60
Conventional fixed peg to a single

currency
681.3550.10731.45
Conventional fixed peg to a basket240.9020.08260.98
Horizontal bands191.09140.80331.89
Crawling pegs261.50170.98432.48
Forward looking132.1930.51162.70
Backward looking131.14141.23272.37
Crawling bands80.7880.78161.55
Forward looking50.6630.4081.06
Backward looking31.1051.8482.94
Tightlymanaged floating191.10120.69311.79
Floating regimes300.44991.461291.90
Other managed floating300.90541.62842.52
Independently floating451.30451.30
Total202157359
Source: Authors’ estimates.

Indicates the total number of exits from each regime during the period January 1990-December 2001.

Defined as the total number of exits from a given regime as a ratio of the total number of observations during which the regime as in effect over the sample.

Source: Authors’ estimates.

Indicates the total number of exits from each regime during the period January 1990-December 2001.

Defined as the total number of exits from a given regime as a ratio of the total number of observations during which the regime as in effect over the sample.

Consequently, among the intermediate regimes, crawling peg, horizontal band, and backward-looking crawling band regimes are likely to lose most of their share in the long run based on past history. Basket pegs lose almost all of their share in steady state as well, although they have the highest persistence across all intermediate regimes. This is because countries seem much less likely to switch to basket pegs, conditional on starting from a different regime, as indicated by the largest number of zeros and the very small values of nonzero transition probabilities of switching to this regime (see the fourth column of Table 5). Forward-looking crawling bands and tightly managed floats, whose shares have increased over the past decade, are expected to continue this trend (provided that there are no significant shocks to the system), as suggested by the transition probabilities of switching to these regimes and their steady-state distribution. Consistent with Figure 3 in Section III, the estimated transition probabilities also show that most intermediate regimes end up with a float rather than being absorbed in a hard peg or another intermediate regime.

V. Concluding Remarks

There has been growing support in recent years for the view that increased capital mobility has pushed countries toward either greater fixity or flexibility of exchange rate regimes, and that the regimes between the two extremes will be hollowing out as countries abandon regimes in the middle ground. Some have argued that the trend toward greater flexibility is to a certain degree a fallacy, as a number of countries declaring to be officially floating maintained informal exchange rate targets for a variety of reasons. Such divergence has in turn been used to challenge the validity of the bipolar view. This paper presented a monthly database on de facto exchange rate regimes of the IMF members during 1990–2001. It then used this database to examine the evolution of various exchange rate regimes since 1990 to see whether the bipolar view would still be supported if the analysis were based on de facto, as opposed to de jure, exchange rate regimes.

The simple analyses of trends in de facto exchange rate regimes support the earlier claims (for example of Fischer, 2001) that the proportion of countries adopting intermediate regimes has been shrinking in favor of either greater flexibility or greater fixity over the past decade or so, but more so for the countries that are fully or increasingly more integrated with the international financial markets. The analysis also shows that the extent of the decline has not been uniform across different types of intermediate regimes. In particular, countries have tended to move to more flexible forms of intermediate regimes, away from less flexible ones, in part to minimize potential trade-offs between competing policy objectives in a world with growing mobility of capital.

A more detailed analysis of the exchange regime transitions using Markov chains provides mixed evidence for the validity of the bipolar view. On the one hand, despite the marked shift toward the two poles of exchange rate regimes during 1990—2001, the history of regime transitions between hard pegs, intermediate and floating regimes points to no strong evidence to suggest that intermediate regimes will disappear. Some evidence of structural instability in the sample, however, suggests that this result be taken cautiously. The structural stability tests indicate in particular that the trend toward polarization may be the result of the fact that countries became less likely in the later part of the decade to leave the two poles upon exiting from intermediate regimes. Additional evidence from the future evolution of exchange rate regimes would hence be necessary to see if this trend would continue and hence to establish more concrete evidence for or against the vanishing middle proposition.

APPENDIX I
Table 7.Database on De Facto Exchange Regime Classification(End of year)
199019911992199319941995199619971998199920002001
Afghanistan, I.S. of121213131313131313131313
Albania413131313131313131313
Algeria55551212121212121212
Angola4444124444131312
Antigua and Barbuda222222222222
Argentina1333333333333
Armenia1121111121212121213
Aruba444444444444
Australia131313131313131313131313
Austria444444444222
Azerbaijan441212121111111111
Bahamas, The444444444444
Bahrain444444444444
Bangladesh444441111114444
Barbados444444444444
Belarus1212124999978
Belgium666666666222
Belize444444444444
Benin222222222222
Bhutan444444444444
Bolivia999999999777
Bosnia44433333
Botswana555555555555
Brazil12999119997131313
Brunei Darussalam333333333333
Bulgaria513131111111333333
Burkina Faso22222222222
Burundi555555555121212
Cambodia121212121212121212121212
Cameroon222222222222
Canada121212121212121213131313
Cape Verde555555554444
Central African Republic222222222222
Chad222222222222
Chile10101010101010108131313
China4121212124444444
Colombia988888888131313
Comoros444444444444
Congo, Democratic Republic1212121212131313124413
Congo, Republic of222222222222
Costa Rica9911999779999
Côte d’Ivoire222222222222
Croatia9666666111111
Cyprus556666666666
Czechoslovakia555
Czech Republic55561212121213
Denmark666666666666
Djibuti333333333333
Dominica222222222222
Dominican Republic41111111212121212121212
Ecuador77610888881311
Egypt46666664412126
El Salvador111112444444441
Equatorial Guinea222222222222
Eritrea4444121244
Estonia3333333333
Ethiopia444121212121212121212
Fiji555555555555
Finland6613131313666222
France666666666222
Gabon222222222222
Gambia, The131313131313131313131313
Georgia1131111111113131313
Germany666666666222
Ghana121212121212121212121212
Greece777777776662
Grenada222222222222
Guatemala111111111212121212121212
Guinea12124111111121212121212
Guinea-Bissau999999922222
Guyana41211121111111112111111
Haiti41313131313131313131313
Honduras441313117888888
Hong Kong SAR333333333333
Hungary555568888886
Iceland6666666666613
India6444412111111111111
Indonesia9999910101313131312
Iran, I.R. of555444444444
Iraq444444444121212
Ireland666666666222
Israel688888888888
Italy6613131313666222
Jamaica12134121113121212121211
Japan131313131313131313131313
Jordan555554444444
Kazakhstan121313121249111111
Kenya555121212121212121212
Kiribati111111111111
Korea111111111111111313131313
Kuwait555555555555
Kyrgyz Republic11266121212121212
Lao, People’s Dem. Rep.44444441212121212
Latvia131255555555
Lebanon121212777774444
Lesotho444444444444
Liberia4444444413131313
Libya555555555555
Lithuania131133333333
Luxembourg444444444222
Macedonia1213124444444
Madagascar55551312121313131313
Malawi5554441212121313
Malaysia5121212121111124444
Maldives1212121244444444
Mali222222222222
Malta555555555555
Marshall Islands1111111111
Mauritania555555121111111111
Mauritius55551212121313121212
Mexico78881313131313131313
Micronesia111111111
Moldova1121211111113131313
Mongolia44131313131212131111
Morocco555555555555
Mozambique9912131313131313131313
Myanmar5555512121212121212
Namibia111444444444
Nepal444444444444
Netherlands444444444222
Netherlands Antilles444444444444
New Zealand131313131313131313131313
Nicaragua944777777777
Niger222221222222
Nigeria1111114411111111111111
Norway6613111111111212121213
Oman444444444444
Pakistan111111111111441241212
Palau11111
Panama111111111111
Papua New Guinea55551313131313131313
Paraguay111111111111111111111111
Peru131312121212121212131313
Philippines412121212441213121313
Poland47777888881313
Portugal886666666222
Qatar444444444444
Romania511111111111199778
Russia1261277712121212
Rwanda5555513131312121212
Samoa555555555555
San Marino1111111111
São Tóme and Principe59991212121211111111
Saudi Arabia444444444444
Senegal222222222222
Seychelles555555555555
Sierra Leone131313131313131313131313
Singapore111111111111111212121111
Slovak Republic6666612121212
Slovenia996126612121212
Solomon Islands999999999977
Somalia121313131313131313131313
South Africa121212121211121313131313
Spain666666666222
Sri Lanka9999910101010101012
St Kitts and Nevis222222222222
St Lucia222222222222
St Vincent and Grenadines22222222222
Sudan441241213121213444
Suriname44441212444464
Swaziland444444444444
Sweden6613131313131313131313
Switzerland131313131313131313131313
Syrian Arab Republic444444444444
Tajikistan12121212412121313
Tanzania111111131313131313131313
Thailand55555551212121212
Togo222222222222
Tonga455555556666
Trinidad and Tobago44413111111444411
Tunisia99999999991212
Turkey9999799977713
Turkmenistan14441144444
Uganda121213131313131313131313
Ukraine1241212666121111
United Arab Emirates444444444444
United Kingdom6613131313131313131313
United States131313131313131313131313
Uruguay10108888888888
Uzbekistan141111119991112
Vanuatu55555555555
Venezuela121212944888888
Vietnam1212121244666111111
Yemen Arab Republic444444131313121213
Yugoslavia4413444444411
Zambia9913131313121313131312
Zimbabwe55551010101212444
Note:1 = Another currency as legal tender; 2 = currency union; 3 = currency board; 4 = conventional fixed peg to single currency;5 = conventional fixed peg to basket; 6 = pegged within a horizontal band; 7 = forward-looking crawling peg;8 = forward-looking crawling band; 9 = backward-looking crawling peg; 10 = backward-looking crawling band;11 = tightly managed floating;12 = other managed floating; and 13 = independently floating.
Note:1 = Another currency as legal tender; 2 = currency union; 3 = currency board; 4 = conventional fixed peg to single currency;5 = conventional fixed peg to basket; 6 = pegged within a horizontal band; 7 = forward-looking crawling peg;8 = forward-looking crawling band; 9 = backward-looking crawling peg; 10 = backward-looking crawling band;11 = tightly managed floating;12 = other managed floating; and 13 = independently floating.
APPENDIX II
Table 8.Evolution of Exchange Rate Regimes(In percent of members in the given category)
199019911992199319941995199619971998199920002001Average

(1990-

2001)
Whole membership
Dollarization1.91.96.12.72.72.72.73.23.23.23.84.33.2
Currency union11.911.810.610.410.310.310.310.810.816.716.717.212.3
Currency board1.92.52.82.73.23.23.24.34.34.34.34.33.4
Fixed peg to a currency24.521.718.919.820.021.620.020.419.918.818.316.720.1
Fixed peg to a basket19.517.415.013.710.38.67.67.05.95.45.45.410.1
Horizontal baud8.88.16.17.18.17.09.78.68.62.73.22.76.7
Crawling peg9.49.37.28.86.56.56.57.05.94.83.82.26.5
Crawling band1.93.72.83.33.25.46.55.45.43.83.23.24.0
Tightly managed float5.06.26.16.08.69.78.14.83.25.47.58.66.6
Other managed float9.49.911.711.514.110.813.014.016.116.714.514.013.0
Independently floating5.77.512.813.713.014.112.414.516.718.319.421.514.1
Developed countries
Dollarization0.00.04.24.24.24.24.24.24.24.24.24.23.5
Currency union0.00.00.00.00.00.00.00.00.045.845.850.011.8
Currency board0.00.00.00.00.00.00.00.00.00.00.00.00.0
Fixed peg to a currency13.013.012.512.512.512.512.512.512.50.00.00.09.5
Fixed peg to a basket0.00.00.00.00.00.00.00.00.00.00.00.00.0
Horizontal band52.252.233.333.333.333.341.741.745.812.512.54.233.0
Crawling peg4.34.34.24.24.24.24.24.20.00.00.00.02.8
Crawling band4.34.30.00.00.00.00.00.00.00.00.00.00.7
Tightly managed float0.00.00.04.24.24.24.20.00.00.00.00.01.4
Other managed float4.34.34.24.24.24.24.28.34.24.24.20.04.2
Independently floating21.721.741.737.537.537.529.229.233.333.333.341.733.1
Developing countries
Dollarization2.22.26.42.52.52.52.53.13.13.13.74.33.2
Currency union14.013.812.212.011.811.811.8¡2.312.312.312.312.312.4
Currency board2.22.93.23.23.73.73.74.94.94.94.94.93.9
Fixed peg to a currency26.523.219.920.921.123.021.121.621.021.621.019.121.7
Fixed peg to a basket22.820.317.315.811.89.98.78.06.86.26.26.211.7
Horizontal band1.50.71.93.24.33.15.03.73.11.21.92.52.7
Crawling peg10.310.17.79.56.86.86.87.46.85.64.32.57.1
Crawling band1.53.63.23.83.76.27.56.26.24.33.73.74.5
Tightly managed float5.97.27.16.39.310.68.75.63.76.28.69.97.4
Other managed float10.310.912.812.715.511.814.314.817.918.516.016.014.3
Independently floating2.95.18.310.19.310.69.912.314.216.017.318.511.2
Emerging markets
Dollarization3.33.33.23.13.13.13.13.13.13.16.36.33.7
Currency union0.00.00.00.00.00.00.00.00.00.00.00.00.0
Currency board3.36.76.56.36.36.36.39.49.49.49.49.47.4
Fixed peg to a currency13.33.33.26.39.412.512.512.512.512.59.49.49.7
Fixed peg to a basket20.013.312.915.612.59.46.33.13.13.13.13.18.8
Horizontal band6.73.36.56.36.33.16.30.00.00.00.06.33.7
Crawling peg20.020.016.118.812.59.49.49.46.33.13.10.010.7
Crawling band3.313.312.915.612.525.028.125.025.015.612.56.316.3
Tightly managed float13.313.312.912.515.621.915.66.36.36.39.49.411.9
Other managed float10.016.722.615.618.86.36.318.818.818.815.615.615.3
Independently floating6.76.73.20.03.13.16.312.515.628.131.334.412.6
APPENDIX III

Markov Chain Analysis

Some Basic Definitions and Terminology

Assuming that the stochastic process for the choice of exchange rate regimes can be represented by a Markov chain, the probability of a country being in a given exchange rate regime depends only on its regime in the most recent previous period.36 A great advantage of the Markov representation is that it compresses the history of exchange rate regimes into a single transition matrix that can be used to characterize the dynamics of shifts between various classes of regimes. Letting the row vector xt represent the (1xn) distribution of n regimes at time t, the probability of moving from regime i to j can be written as pij= Pr (Xt+1 = j | Xt = i), where i,j= 1, 2, …, n, and collected in an (n x n) transition matrix. Focusing on regime transitions among the three broad regime categories of hard pegs, intermediate, and floating regimes, the transition probability matrix, P, is given by:

where, p11 is the probability of staying in a hard peg in period t+1 given that period t regime is a hard peg, p12 is the probability of moving from a hard peg to an intermediate regime, etc.

Given the history of exchange rate regimes and given that the observations are drawn from a multinomial distribution, the maximum likelihood estimates for the transition probability pij is simply obtained from the sample frequencies of transitions between regimes, that is, the ratio between the number of transitions from i to j (nij) and the number of times the time t regime was i(ni):p^ij=nijni.37 The larger the number of observations the more precise point estimates are, since ni(p^ijpij) is distributed approximately as a normal with mean 0, variance pij (1 - pij).

Given the transition matrix, the time t forecast of the distribution at time t+1 is given by: Xt+1|xt = xt * P. The exchange regime process is Markovian if Pr (Xs = j|Xt = i, Xt-1 = k, Xt-2 = l,..) = Pr (Xs =j| Xn = i), where i, j, k, l = 1,2,…, n, and t<s; that is, the most recent information on the regime is the only relevant information to make predictions about the next period’s regime. By repeatedly iterating the transition matrix into the future, one can obtain the expected distribution of Xt+s, conditional on the information available at time t as: Xt+s|xt = xt * Ps.

The invariant (alternatively, long-run or steady state) distribution, xss, characterizes the long run proportion of regimes, given by: xss= xt,* P, where xt is the initial distribution. The invariant distribution can be expressed as the unique distribution of exchange rate regimes that is mapped by P into itself: xss=xss * P.38 A state is said to be intransient if it is not destined to be empty in steady state.

The transition matrix of regimes is said to have an absorbing state, if it is not possible to reach other states from that state, that is, once entered, the probability of leaving that state is zero. For example, hard pegs are an absorbing state if and only if p11= 1 (hence, p12 = p13 = 0). Similarly, floating is an absorbing state if and only if p33 = 1 (hence, p31 = p32 = 0). A transition matrix can have more than one absorbing state. If there exists a unique long-run distribution, the non-absorbing states are transient, that is, they are expected to be empty in the long run.

The transition matrix has a closed set of states if no state outside this set can be reached from any state in this set. Transition between states within this set could take place but no transition can occur to the states outside this set. For example, if hard pegs and float form a closed set together, the only transitions from hard pegs or floats can be to each other, and no transition from either hard pegs or floats can occur to intermediate regimes (i.e., p12 = p32 = 0), while transitions can occur from intermediate regimes to either hard peg or floating regimes. If a transition matrix has no closed set, it is called irreducible, which means that any state can be reached from any other state, and that all states would prevail in the long run.

Testing Hypothesis of Absorbing States and Closed Sets

Letting p^ij be the component in row i and column j in the estimated Markov chain, in large samples the vector ni(p^ijpij) is distributed approximately as a normal with mean 0, variance pij (1 − pij) and covariance − δigpijpgh, where ni is the number of observations starting in regime i, pij is the true population parameter and δig is an indicator function equal to one when i = g and zero otherwise (i.e., the estimates in each row are independent of the estimates in other rows). Therefore, hypotheses tests on the Markov chain can be conducted using a goodness of fit statistic, since under the null H0:pi1=pi10,pi2=pi20,pi3=pi30,j=13ni(p^ijpij0)2pij0 is distributed as a χ2 with 2 degrees of freedom.

This approach cannot be adopted for testing whether regime i is absorbing, however, because, under the null, the asymptotic distribution is not normal but collapses to a degenerate distribution. Considering, for instance, the hypothesis of hard pegs being absorbing, the null hypothesis is H0: p11 = 1, p12 = 0, p13 = 0. Under this null, the asymptotic variance and covariance of ni(p^ijpij) are equal to zero and the asymptotic distribution is not normal. For these reasons, it is sufficient to find in the sample at least one exit from hard pegs in order to reject with certainty the null hypothesis of hard pegs being an absorbing state. Intuitively, since a sample should represent independent draws from the population, only one exit in the sample is inconsistent with a population with no exits. For similar reasons, the evidence in the sample of shifts from hard pegs or from floats to intermediate regimes rules out the possibility that the two polar regimes form a closed set.

Test for Structural Stability

To test for structural stability, the sample was divided into two sub-periods of equal length (January 1990–December 1995 and January 1996–December 2001), estimating the transition matrices for each sub-period and comparing them with the matrix obtained in text Table 3. The estimated transition probabilities for the two sub-periods are given in Table 9.

Table 9.Estimated Transition Matrix for All Countries
Estimated probability of regime in t+1
Regime in period tHard pegsIntermediateFloatingTotal

observations
(1990–95)
Hard pegs0.99660.00140.00192,074
Intermediate0.00030.99160.00817,637
Floating0.00000.01730.98272,711
12,422
(1996–2001)
Hard pegs0.99960.00000.00042,766
Intermediate0.00240.99010.00756,278
Floating0.00050.00640.99314,080
13,124
Source: Authors’ estimates.
Source: Authors’ estimates.

The null hypothesis of structural stability is conducted with a χ2 test on the statistic:

where nit and p^ijt are, respectively, for each sub period, the number of observations that started in the ith regime and the estimated transition probabilities; p^ij0 stands for the transition probability under the null hypothesis of structural stability using the whole sample. Under the null, the statistic has a χ2 distribution with 6 degrees of freedom.39

Calculations based on this statistic significantly reject the null hypothesis of structural stability: S = 39.5,P2 ≥ 39.5)< 0.001. Since each row in the estimated transition matrix is independent of the others, the stability of each row can be tested separately, using the statistic, Si=t=12j=13nit(p^ijtp^ij0)2p^ij0, which is distributed as a χ2 with 2 degrees of freedom under the null.

Table 10 shows that time homogeneity for each row is rejected; among the three exchange rate regimes, hard peg regimes appear to be the least heterogeneous through time.

Table 10.Test of Structural Stability of the Parameters
RegimeSi=t=12j=13nit(p^ijtp^ij0)2p^ij0P-value
Hard pegsS1 = 6.83P2 ≥ 6.83) = 0.033
IntermediateS1 =12.94P2 ≥ 12.94)< 0.001
FloatingS1 =19.72P(χ2 ≥ 19.72)<0.001

To better understand the nature of the instability, the structural stability of the diagonal and the off-diagonal components of the Markov chain can be studied separately, following Henry (1971). Markov chain can be regarded as a combination of two distinct processes: a process establishing when to change regime and a process determining the result of the shift. Instability could occur not because of time variation in all the components of the matrix, but because of heterogeneity in the diagonal components only (countries shift regimes due to changes in the feasibility of the current regime) or because of the instability in off-diagonal components exclusively (due to a time-varying attractiveness of the other two regimes). The following two hypotheses can be tested: (i) whether the transition probabilities of the diagonal components are stable, and whether the instability comes from changes in the off-diagonal probabilities; and (ii) whether the time heterogeneity derives from changes in the diagonal components accompanied with constant conditional probabilities to shift to the other two regimes. The hypotheses are tested with respect to the whole matrix and each row. The test relies on a χ2 statistic distributed with 3 degrees of freedom (1 degree of freedom when each row is tested separately).

Table 11 shows that the partial stability hypothesis is rejected when testing the entire transition matrix. The analysis of each row separately decomposes the causes for the overall rejection: First, the null hypothesis of stability in the diagonal component of intermediate regimes cannot be rejected. In other words, the retention rate of (or the exit rate from) intermediate regimes does not appear to have changed significantly through time. The instability of the intermediate regime row, as indicated in the table, comes from the heterogeneity in the off-diagonal components: while a switch to floating is the most likely outcome in both sub-samples, the attractiveness of hard pegs has significantly increased in the more recent period.40 Conversely, the null of homogeneity in the off-diagonal components in both hard pegs and floating regimes cannot be rejected. In other words, there is no significant difference between the direction of an exit from the two polar regimes in the two sub periods. What seems to have changed is the retention rate of these two regimes. In particular, as can be seen from Table 9, both regimes appears to have become more persistent in the second subperiod.

Table 11.Test of Partial Stability
H0: stable diagonal, but unstable

off-diagonal
H0: unstable diagonal, but stable off-

diagonal (conditional on leaving the

regime)
Whole matrixS = 23.28 P2 ≥ 23.28)<0.001S1 = 16.11 P2 ≥ 16.11) = 0.001
Hard pegsS1 = 6.52

P2 ≥ 6.52)= 0.01
S1 = 0.685

P2 ≥.685) = 0.407
IntermediateS1 =0.858

P2 ≥ .858) = 0.354
S1 = 11.97

P2 ≥ 11.97)<0.001
FloatingS1 = 15.89

P2 ≥ 15.89)< 0.001
S1 =3.45

P2 ≥ 3.45) = 0.063

The evidence of a significant increase in the retention rate of the two poles points in favor of the bipolar view, since through time the extremes of the flexibility continuum appear to be more persistent. On the other hand, the insignificant change in the persistence of intermediate regimes indicates that the attraction to the two poles proceeds slowly. This suggests that the tendency of polarization does not come from an acceleration in the exit rate from intermediate regimes, but from the fact that it is now less likely to exit from the two poles.

References

    AndersonT.W. and L.Goodman1957Statistical Inference about Markov ChainsAnnals of Mathematical StatisticsVol. 28 pp. 89109.

    BaliñoT. B.A.Bennett and E.Borensztein1999Monetary Policy in Dollarized Economies IMF Occasional Paper No. 171 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

    BhatU. N.1972Elements of Applied Stochastic Processes (New York: Wiley).

    BillingsleyP.1961Statistical Methods in Markov ChainsAnnals of Mathematical StatisticsVol. 32No. l pp. 1240.

    BofingerP. and T.Wollmershäuser2001Managed Floating: Understanding the New International Monetary OrderWürrzburg Economic Papers No. 30 (Würrzburg: Universität Würzburg).

    • Search Google Scholar
    • Export Citation

    BubulaA. and I.Otker-Robe2002Are Pegged Regimes More Crisis Prone?” (unpublished; Washington: International Monetary Fund).

    CalvoG. and C.Reinhart2000aFear of FloatingNBER Working Paper 7993 (Cambridge, Massachusetts: National Bureau of Economic Research).

    • Search Google Scholar
    • Export Citation

    CalvoG. and C.Reinhart2000bFixing For Your LifeNBER Working Paper 8006 (Cambridge, Massachusetts: National Bureau of Economic Research).

    • Search Google Scholar
    • Export Citation

    CollinsS. M.1996On Becoming More Flexible: Exchange Rate Regimes in Latin American and the CaribbeanJournal of Development EconomicsVol. 51 (October) pp. 117138.

    • Crossref
    • Search Google Scholar
    • Export Citation

    EdwardsS. M. and A.Savastano1998Exchange Rates in Emerging Economies: What do we Know? What do We Need to Know?Paper presented at the Stanford University Conference on Economic Policy Reform: What We Know and What We NeedSeptember 17–19, 1998.

    • Search Google Scholar
    • Export Citation

    EichengreenB.1994International Monetary Arrangements for the 21st Century (Washington: Brookings Institution).

    EichengreenB.2001From Bretton Woods to Bipolarity; Singapore and the WorldPaper Prepared for the 30th Anniversary Conference of the Monetary Authority of SingaporeJuly 20, 2001.

    • Search Google Scholar
    • Export Citation

    EichengreenB.FischerS.2001Exchange Rate Regimes: Is the Bipolar View Correct?Distinguished Lecture on Economics in GovernmentJournal of Economic PerspectivesVol. 15No. 2 (Spring) pp. 3-24.

    • Search Google Scholar
    • Export Citation

    FellerW.1967An Introduction to Probability Theory and its ApplicationsVol. 2 (New York: Wiley3rd ed.).

    FrankelJ.1999No Single Currency Regime is Right For All Countries or at All TimesEssays in International Finance No. 215 (Princeton: Princeton University Press).

    • Crossref
    • Search Google Scholar
    • Export Citation

    FrankelJ.E.FajnzylberS. LSchmukler and L.Servén2001Verifying Exchange Rate RegimesJournal of Development economicsVol. 66 (October) pp. 35186.

    • Search Google Scholar
    • Export Citation

    GlickR.2000Fixed or Floating: Is it Still Possible to Manage in the Middle?Pacific Basin Working Paper Series Center for Pacific Basin Monetary and Economic Studies Economic Research Department Federal Reserve Bank of San Francisco Working Paper No. PB00-02.

    • Search Google Scholar
    • Export Citation

    GoldsteinM.1999Safeguarding Prosperity in a Global Financial System: The Future International Financial Architecture (Washington: Institute for International Economics).

    • Search Google Scholar
    • Export Citation

    GhoshA.A.GuideJ.Ostry and H.Wolf1995Does the Nominal Exchange Rate Regime Matter?IMF Working Paper 95/121 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

    GhoshA.A.GuideJ.Ostry and H.Wolf1997Does the Nominal Exchange Rate Regime Matter?NBER Working Paper No. 5874 (Cambridge, Massacusetts: National Bureau of Economic Research).

    • Search Google Scholar
    • Export Citation

    GhoshA.A.Gulde and H.Wolf2000Currency Boards: More than a Quick Fix?Economic PolicyVol. 31 (October) pp. 269335.

    GhoshA.A.Gulde and H.WolfExchange Rate Regimes: Choices and Consequences (Cambridge, Massachusetts: MIT Pressforthcoming).

    GuilbaudO.1977Estimating the Probability Eigenvector and Related Characteristics of an Ergodic Transition MatrixScandinavian Journal of StatisticsVol. 4No. 3 pp. 97-104.

    • Search Google Scholar
    • Export Citation

    HausmannR.U.Panizza and E.H.Stein2000Why Do Countries Float the Way they Float?Inter-American Development Bank Working Paper No. 418 (Washington: Inter-American Development Bank).

    • Search Google Scholar
    • Export Citation

    HenryN.W.1971Retention Model—Markov Chain with Variable Transition ProbabilitiesJournal of the American Statistical AssociationVol. 66 (334) pp. 264-267

    • Crossref
    • Search Google Scholar
    • Export Citation

    HoldenP.M.Holden and E. C.Suss1979The Determinants of Exchange Rate Flexibility: An Empirical InvestigationReview of Economics and StatisticsVol. 61 (August) 3 pp. 327333.

    • Crossref
    • Search Google Scholar
    • Export Citation

    IMF1999Exchange Rate Arrangements and Currency Convertibility: Developments and IssuesWorld Economic and Financial Surveys (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

    IMFAnnual Report on Exchange Arrangements and Exchange Restrictionsvarious issues (Washington: International Monetary Fund).

    IMFStaff Country Reportsvarious issues (Washington: International Monetary Fund).

    KemenyJ. G. and J. LSnell1960Finite Markov Chains (PrincetonN.J.).

    Levy-YeyatiE. and F.Sturzenegger1999Classifying Exchange Rate Regimes: Deeds vs. Words” (unpublished; Universidad Torcauto Di Tella).

    • Search Google Scholar
    • Export Citation

    MAS2000Implementation of Monetary PolicyMonetary Authority of Singapore Economics Explorer Series No. 5February.

    MAS2001Singapore’s Exchange Rate PolicyMonetary Authority of SingaporeFebruary.

    MassonP. R.2001Exchange Rate Regime TransitionsJournal of Development EconomicsVol. 64 (April) pp. 57186.

    ObstfeldM. and K.Rogoff1995 (Fall) “The Mirage of Fixed Exchange RatesJournal of Economic PerspectivesVol. 9No. 4 pp. 7396.

    PoirsonH.2001How Do Countries Choose Their Exchange Rate Regime?IMF Working Paper 01/46 (Washington:International Monetary Fund).

    RajanR. S. and R.Siregar2000The Vanishing Intermediate Regime and the Tale of Two Cities: Hong Kong Versus SingaporeCentre for International Economic StudiesAdelaide University, Adelaide, Australia, Policy Discussions Paper No. 0031.

    • Search Google Scholar
    • Export Citation

    ReinhartC. M. and K.Rogoff2002The Modern History of Exchange Rate Arrangements: A ReinterpretationNBER Working Paper No. 8963 (Cambridge, Massachusetts: National Bureau of Economic Research).

    • Search Google Scholar
    • Export Citation

    SummersL.1999Testimony Before the Senate Foreign Relations Subcommittee on International Economic Policy and Export/Trade Promotion U.S. Senate 106th Congress 1st SessionJanuary 27.

    • Search Google Scholar
    • Export Citation

    WilliamsonJ.2000aAre Intermediate Regimes Vanishing?Speech given at the International Conference on Exchange Rate Regimes in Emerging Market EconomiesTokyoDecember, 17-18, 1999.

    • Search Google Scholar
    • Export Citation

    WilliamsonJ.2000bCrawling Bands or Monitoring Bands: How to Manage Exchange Rates in a World of Capital Mobility International FinanceVol. 1 (October) pp. 5979October.

    • Search Google Scholar
    • Export Citation

    WilliamsonJ.2000cExchange Rate Regimes for Emerging Markets: Reviving the Intermediate Options Policy Analyses in International Economics 60 (Washington: Institute for International Economics).

    • Search Google Scholar
    • Export Citation
1The authors are grateful to Stanley Fischer for his encouragement to contruct a historical database on de facto exchange rate regimes. The paper has greatly benefited from comments and suggestions from Jorge Canales-Kriljenko, Rupa Duttagupta, Karl Habermeier, Shogo Ishii, and Michel Robe. The authors also thank Phoebus Dhrymes, Jeffrey Frankel, Hassanali Mehran, Alexei Onatski, and Carmen Reinhart for helpful comments and discussions. The authors are responsible for all errors and omissions.
2Intermediate regimes include soft peg regimes (including conventional fixed and crawling pegs and horizontal and crawling bands) and tightly managed floating regimes (where authorities heavily monitor and control exchange rate movements without any commitment to a predetermined target path, but attempt to keep the exchange rate path stable). Hard pegs include regimes with another currency as legal tender, currency unions, and currency boards.
3The bipolar view is also referred to as “vanishing or missing middle,” “hollowing out of intermediate regimes,” or “corner solutions.” See Eichengreen (1994) for initial discussions and also Obstfeld and Rogoff (1995), Goldstein (1999), Summers (1999), and Fischer (2001) for related arguments.
4Divergences between stated and actual policies could reflect, among other things, concerns about the political cost of undertaking visible devaluations under a formally announced peg, or the “fear of floating” to limit the potential impact of exchange rate depreciation on inflation and balance sheets when there is high degree of dollarization or large exchange rate exposure in the domestic economy (see, for example, Collins, 1996; Baliño, Bennett, and Borensztein, 1999; Calvo and Reinhart, 2000; and Hausmann, Panizza, and Stein, 2000).
5Countries must notify the IMF within 30 days of becoming a member and also promptly following any changes in their regimes (see IMF, 1999).
6Holden, Holden, and Suss (1979) were among the first to point out deficiencies in the IMF classification. See also Collins (1996), IMF (1999), Levy-Yeyati and Sturzenegger (1999), Masson (2001), and Poirson (2001) for recent discussions.
7Close to 60 percent of the countries classified as managed floaters at the end of 1997 maintained some form of de facto peg regimes (15 percent conventional fixed pegs, 13 percent crawling pegs, 20 percent crawling bands, and 11 percent horizontal bands). See IMF (1999).
8Accounting for members’ de facto policies indicated, for example, that exchange rate anchors remained as the predominant monetary arrangement at end-1997.
9See, for example, Fischer (2001), Glick (2000), and Poirson (2001).
10Note, however, that in the overlapping years from 1999 (the year in which the new classification system became official), the classifications in this paper’s database diverge in a number of instances from the official classifications published in the IMF’s Annual Report on Exchange Arrangements and Exchange Restrictions, owing to the additional information assembled during this research. The paper also adds to the IMF official classification more details on some regimes (e.g., introducing backward- and forward-looking crawling pegs and bands, and tightly managed floats).
11In further research on the topic, Ghosh, Guide, and Wolf (2000) and Ghosh, Guide and Wolf (forthcoming), address the distinction between de jure pegs and hard pegs, and perform robustness tests explicitly addressing the issue of “soft pegs” and “hard floats.”
12The data have been updated up to 1999. Masson (2001) uses an earlier update to 1997 for an analysis of regime transitions.
13Several other studies used this approach to study the choice or behavior of exchange regimes. These include: Baig (2001), Bofinger and Wollmerhäuser (2001), Calvo and Reinhart (2000a and 2000b), Hausmann, Panizza, and Stein (2001), and Poirson (2001).
14Countries with high exchange rate volatility and low reserve volatility are mapped into the flexible category. Those with a low variation in the exchange rate data but high volatility in reserves are classified as fixers, presuming that monetary authorities were intervening to offset market forces. Countries with relatively high volatility across all variables are classified as dirty floaters, while those with significant but relatively stable exchange rate variations and active foreign exchange intervention are identified with a crawling peg.
15For example, Belgium and France with horizontal bands in IMF classification were included in the “inconclusive” group together with Argentina with a currency board, Nepal with a conventional fixed peg, and several countries with currency unions. Other countries with similar regimes (for example, Hong Kong SAR with a currency board, other members of currency unions, and several countries participating in the European Union’s exchange rate mechanism (ERM)) were classified in the “fixed” category.
16An additional feature of the paper’s methodology is its reliance on cluster analysis, which sorts countries into a given number of groups according to the characteristics of the three variables of reference. The regime classifications generated by this approach are not robust to the choice of the number of countries in the sample and the number of clusters specified ex-ante by the user. The lack of such robustness is an important limitation for a methodology to classify exchange regimes for the whole membership that may change over time.
17The main exception is the introduction of an additional category of “freely falling” rates when monthly inflation exceeds 50 percent and when there is no official announcement of the regime by the authorities.
18Reinhart and Rogoff (2002) use mean absolute deviations to minimize the impact of such outliers.
19Using movements in the exchange rates prevailing in the parallel markets that are not officially recognized or tolerated would be misleading in inferring about the actual exchange rate regime pursued by the authorities. The use of such rates would also be misleading when the parallel market is shallow and is primarily used for illicit transactions, with the discount between the official and the parallel market rates simply reflecting transaction costs involved in conducting such activities.
20With the threat of a more severe supervision or loss of licenses, banks may refrain from transacting at rates different from those perceived as preferred by the authorities. Foreign exchange may be allocated through auctions in which authorities may control the price process by requesting the bids before deciding the amount to sell or by putting limits on the bids that could be submitted. State-owned banks and corporations may be forced to comply with implicit directives to stabilize the rate.
21In Paraguay, the overall policy throughout the 1990s was to maintain an “adjustable peg” vis-à-vis the U.S. dollar with monetary policy geared toward keeping the exchange rate under control. The authorities intervened broadly to maintain the real effective value of the currency, achieved mainly through relatively short periods of frequent adjustment of the nominal rate, followed by more stable exchange rates, and occasional step adjustments in the exchange rate. Singapore has operated an undisclosed “monitoring band” system—a variant of the “basket, band, and crawl” regime referred to by Williamson (2000b). The authorities manage the Singapore dollar against a basket of currencies of the main trading partners, allowing it to float within a target band around a central parity; both the central parity and the band width are undisclosed, reviewed and adjusted periodically to ensure consistency with economic fundamentals and market conditions, with no obligation to defend the edges of the band (see Eichengreen, 2001; Rajan and Siregar, 2000; and MAS, 2000 and 2001).
22Regime classifications at monthly frequency are available from the authors upon request.
23We base the ordering of regimes on the degree of flexibility of exchange rate policy rather than the flexibility of the exchange rate. While the two concepts often overlap, sometimes there may not be a clear cut ordering without knowing the precise parameters of each regime.
24These regimes have broadly coincided with those that came under more frequent exchange market pressure (see the companion paper, Bubula and Otker-Robe (2002), for an analysis of the frequency of currency crises across alternative exchange rate regimes). However, the sharp decline in the share of horizontal bands also reflected the substitution of ERM bands with the membership in the European Monetary Union as part of a planned political and economic integration.
25For example, Russia (1998), Brazil and Kazakhstan (1999), Turkey (1994, 2001), and Venezuela (1994) abandoned crawling pegs during currency crises, while others, including Portugal (1990), Mexico (1991), Indonesia, Poland, and Sri Lanka (1995), Honduras (1996), and Belarus and Romania (2001) introduced a band around the crawling parities.
26The developed country definition used here coincides with that of the IMF’s International Financial Statistics (IFS). There is no broadly agreed definition for emerging markets. The list used here is based on a number of existing definitions that combine the countries included in the Emerging Markets Bond Index Plus (EMBI+) and Morgan Stanley Capital International (MSCI) index, except Greece, which is included in the developed countries group, and Singapore and Hong Kong SAR in the emerging markets group. This gives a list of 32 emerging markets.
27For example, Mexico (1994), the Czech Republic, Korea, Indonesia, the Philippines, and Thailand (all in 1997), Russia (1998), Brazil and Colombia (1999), and Turkey (2001) floated during crises. Chile (1999) and Poland (2000) floated in a relatively tranquil period to make the direct inflation targeting the overriding target of monetary policy.
28Note, however, that for many emerging market countries, the extent of monetary independence has remained constrained by the reluctance to accept wide fluctuations in exchange rates, given the highly volatile nature of capital flows to these countries.
29Using the Markov chain model for exchange regime transitions assumes that the probability of being in one or another regime next period depends only the current regime, supposing that “the typical country will face the same likelihood that some shock will push it from its current regime to one of the others—independent of past history” (see Masson, 2001, p. 573). Some basic definitions and terminology on Markov chains are given in Appendix III.
30Using data from GGOW, 1997, over 1979-97, “floating regimes” were associated with free floats and “hard pegs” with announced pegs (soft or hard) with virtually no changes in parities. All other regimes (including managed floats, target zones, crawling pegs, and hard or soft fixed pegs with parity adjustments) were included in the “intermediate category.” The definition of hard pegs was hence based on expost exchange rate rigidity, and not necessarily the commitment implied by the formal arrangement. With Levy-Yeyati and Sturzenegger data over 1991-98, the inconclusive group was dropped, dirty floats and crawling pegs made up the intermediate group, and fixed and flexible categories formed the hard peg and floating categories, respectively.
31The last column of the table reports the number of observations each regime was initially in effect over the sample; more specifically, 4,840, 13,915, and 6,791 are the total number of initial hard pegs, intermediate regimes, and floating regimes, respectively. Not all countries’ regimes were available for all dates given the varying membership over time.
32The very high degree of persistence is partly due to the monthly frequency of the regimes data, as there is a higher probability of staying within the same regime from month to month than from year to year. The probabilities using annual (end-year) regimes data over the 1990-2001 period are p11 = 0.9780, p22 = 0.9031 p33 = 0.8876.
33The few shifts that occurred from hard pegs are accounted for by the exits from using as sole legal tender the ruble and the rand by several former Soviet Union states and Namibia, respectively, and by Argentina’s float in end-December 2001 from a currency board. There are many examples of transitions from floats to intermediate regimes (including, for example, Finland and Italy from free float to ERM bands, Venezuela, Jamaica, Malawi from free floats to fixed pegs, Malaysia and Pakistan from managed float to fixed pegs, etc.).
34See Masson (2001, p. 576) for a similar conclusion and Appendix III for a discussion.
35The analysis of the previous section had indicated that the trend away from intermediate regimes was more pronounced for the developed and emerging market countries that are fully or increasingly more integrated with international capital markets. Partial tests not reported here suggest that the result against the bipolar view is robust to restricting the sample to these groups of countries, although for this group, intermediate regimes are expected to hold a much lower share in steady state than hard pegs and floating regimes.
38The steady-state probabilities are estimated by imposing the following three conditions: xss = xss * P, xss * J = 1 xss0, where J is an (n x 1) vector of ones. The latter two conditions entail that xss is a probability distribution.
39The degrees of freedom are given by (T- 1)*m *(m-1), where T is the number of subperiods and m the number of states.
40The probability of switching to hard pegs or floating, conditional on leaving intermediate regimes is given by p21/(1-p22) and p23/(1-p22) respectively. Using the estimates from the two sub-matrices in Table 9, the conditional probability of switching to hard pegs has increased from 3.1 percent to 24.2 percent; the conditional probability of switching to floating regimes, has decreased from 96.9 percent to 75.8 percent.

Other Resources Citing This Publication