A New Set of Measures on Capital Account Restrictions

This paper extends the IMF’s post-1996 disaggregated capital account indices back to 1983 for a representative sample of 34 countries. All the information used to construct the indices comes from IMF’s Annual Report on Exchange Arrangements and Exchange Restrictions and is widely available. It is shown that the disaggregated indices do a better job than the pre-1996 single dummy in reflecting both global trends toward capital account liberalization and country-specific liberalization episodes that occurred during the period. Given the frequency of IMF reporting procedures, the disaggregated indices still fail to accurately track temporary control programs designed to fight off crises. Moreover, the lack of systematic information on enforcement means that the indices remain de jure. Some tentative solutions to these limitations are suggested. [JEL C82, F02, F33]

Abstract

This paper extends the IMF’s post-1996 disaggregated capital account indices back to 1983 for a representative sample of 34 countries. All the information used to construct the indices comes from IMF’s Annual Report on Exchange Arrangements and Exchange Restrictions and is widely available. It is shown that the disaggregated indices do a better job than the pre-1996 single dummy in reflecting both global trends toward capital account liberalization and country-specific liberalization episodes that occurred during the period. Given the frequency of IMF reporting procedures, the disaggregated indices still fail to accurately track temporary control programs designed to fight off crises. Moreover, the lack of systematic information on enforcement means that the indices remain de jure. Some tentative solutions to these limitations are suggested. [JEL C82, F02, F33]

A New Set of Measures on Capital Account Restrictions

JACQUES MINIANE*

The Asian crisis of late 1997 and the subsequent collapses in Russia, Brazil, and Argentina have sparked a vigorous debate about how best to improve the current global financial architecture. In this context, recent papers by De Gregorio, Edwards, and Valdés (2000), Edison and Reinhart (2001), and Miniane and Rogers (2003) have tried to assess the effectiveness of capital controls in shaping the size and maturity of capital flows.

One fundamental limitation in the capital controls literature has been the lack of a reliable measure of capital account openness. Capital controls can take many different forms, making it time-consuming to track all changes in restrictions within a single country. Moreover, the construction of any capital controls index raises the problem of aggregation. By how much should a measure drop if a country relaxes one of its many restrictions? Last but not least, the effectiveness of capital controls depends crucially on the government’s willingness and ability to enforce them. Assuming one has qualitative evidence on enforceability, how should it be weighted in the index?

As a result of the lack of a reliable index of capital controls, many studies trying to assess the ability of capital controls to affect financial flows have followed a case-study approach.1 Papers that have attempted to study the effects of capital controls cross-sectionally, such as Grilli and Milesi-Ferretti (1995), have commonly relied on the 0/1 IMF dummies. The pre-1996 editions of IMF’s Annual Report on Exchange Arrangements and Exchange Restrictions (AREAER) provide dummies for all member countries in six separate categories: bilateral payments arrangements with members and nonmembers, restrictions on payments for current account transactions, restrictions on payments for capital account transactions, import surcharges, advance import deposits, and surrender or repatriation requirements for export proceeds. Besides the obvious limitations of a dichotomic dummy, Eichengreen (2001) has pointed out that the dummy for capital account transactions accounts only for restrictions on residents, not on nonresidents.

To remedy the lack of suitable capital control measures, the IMF changed reporting procedures starting with the 1996 edition of the AREAER.2 For each of the above mentioned categories, the “new” AREAER provides dummies in not one but several different subcategories of transactions. In the case of capital account transactions, there are 13 such subcategories, some of which are in turn further disaggregated.3 Moreover, a distinction is now made between controls on inflows and outflows.

The new reporting procedures have prompted several authors, some within the IMF’s staff, to build capital control indices based on the disaggregated AREAER classification (Table 1).4 Barry Johnston and Natalia Tamirisa are rightly credited with initiating this trend (Johnston and Tamirisa, 1998; Tamirisa, 1999 and 2003). The authors build indices for 45 countries by averaging over all possible 0/1 dummies in the new AREAER. Theirs may be the most disaggregated de jure measure yet, but it is limited by the fact that it covers only the post-1996 period.5 Rossi (1999) tries to extend period coverage for a small sample of 15 countries. He builds Johnston and Tamirisa-style indices for 1997 as well as a “subjective” index for 1989. Values for intermediate years are approximated by a linear interpolation if the country experienced a gradual change in restrictions between the endpoints, or by a one-time change otherwise. Brune and others (2001) can only be praised for the coverage of their measure: 173 countries for the period 1973–1999. The index is a sum of 0/1 dummies over 5 different categories, which are in turn an aggregation of the 13 subcategories in the new AREAER.6 Four of the five categories separate controls on inflows and outflows. The main drawback of Brune and others’ data is that they are not publicly available. Also, early editions of the AREAER often lack information to code all five categories, let alone inflows and outflows separately.7 It remains an open question how the authors tackled this problem.

Table 1.

Summary of Alternative Capital Control Measures

article image
article image

Parallel to these de jure, AREAER-based efforts, researchers have been constructing de facto indices of capital account openness.8 Edison and Warnock (2003) compute the ratio of total market capitalization of equities that are available for purchase by foreign investors over total market capitalization.9 While their index has the great advantage of being both monthly and readily interpretable, restrictions on capital inflows for equity purchases are a small subset of possible capital controls. Another de facto index is based on the work of Lane and Milesi-Ferretti (2001) on countries’ net external wealth and consists of the ratio of a country’s portfolio and direct investment assets and liabilities over GDP. This is, in a sense, a capital account counterpart to the trade openness measures commonly used in the literature.

This paper builds on the tradition of disaggregated AREAER-based measures initiated by Johnston and Tamirisa. Specifically, I use text information in the AREAER to extend this methodology back to 1983 for a sample of 34 countries, thus extending period coverage significantly. As will be shown in the paper, the disaggregated indices do a better job than the standard pre-1996 dummy at tracking both global trends towards capital account liberalization as well as country-specific liberalization episodes. Unlike the Johnston and Tamirisa data, my indices do not disaggregate beyond the 13 main subcategories of capital account transactions in the post-1996 classification. Moreover, they do not separate between controls of inflows and outflows despite a priori advantages of doing so, since it is questionable whether the pre-1996 editions of the AREAER contain enough text information for such a disaggregation. The indices are very transparent in this regard, as I have systematically indicated when the coding of a dummy was done through explicit information in the text and when it was the result of logical induction as explained further in the paper. To see why this is relevant information, note that a full 25 percent of the dummies for the early 1980s were coded through logical induction because of a lack of explicit indications.

The paper is organized as follows. Section I presents and justifies the choice of sample countries and explains in detail the sources of information and the methodology used in the construction of the indices. Section II compares the index with the pre-1996 IMF single dummy, as well as with the de facto measure by Lane and Milesi-Ferretti (LMF).10 Section III addresses the limitations of the post-1996 methodology. In particular, the indices miss temporary capital control programs designed to fight off external crises, and their purely de jure nature says nothing about countries’ enforcement of controls. Some possible solutions are suggested. Section IV contains concluding remarks.

I. Sample and Methodology

Countries and Period Covered

The choice of countries in the data set is somewhat arbitrary. The original impulse behind this project was to improve on the existing IMF single dummies to better study whether capital controls insulate countries from foreign monetary shocks (see Miniane and Rogers, 2003). Country selection for the indices was thus restricted by the availability of data necessary to pursue Miniane and Rogers’s research. The list of countries includes Argentina, Australia, Austria, Belgium, Brazil, Canada, Chile, Colombia, Denmark, Ecuador, Finland, France, Germany, Greece, Hong Kong, India, Italy, Japan, Korea, Luxembourg (pooled with Belgium before 1996), Malaysia, Mexico, the Netherlands, Norway, the Philippines, Portugal, Singapore, South Africa, Spain, Sweden, Switzerland, Turkey, the United Kingdom, and the United States. The data are annual and cover the period 1983 to 2000.11

Despite the arbitrariness of the list, it is worth noting that the sample is diverse along several lines. The five main continents are represented, with particular emphasis on Europe, Asia, and the Americas. Countries range from poor (India, Ecuador) to very rich (Germany, Japan, etc.), with intermediate cases like Chile or Mexico. The whole spectrum of exchange rate arrangements is covered, from a full-blown currency board in Hong Kong, to a managed float in countries like Chile, to a free float in several OECD countries. It is hoped that this diversity somewhat compensates for the relative shortness of the list.12

Information Sources

There are many sources of information on capital controls, including country bulletins on laws enacted. A notable source is the AREAER, which not only contains the previously mentioned dummies but also contains very detailed reports on each country’s exchange arrangement, administration of control, prescription of currency, regulations on import and import payments, payments for invisibles, exports and export proceeds, proceeds from invisibles, capital account transactions, and gold. For each country there is also a section called “Changes,” where the date and details of any change in regulations in any of these categories is registered. As we shall see, the “Changes” section played a crucial role for the construction of my data.

I have decided to use the AREAER as the sole source of information. The reasons for this decision are twofold. First, the AREAER is the only publication that records and classifies the information in a systematic way, both throughout the years and, more importantly, throughout the countries. The same cannot be said about local publications, and this becomes an important consideration when constructing an index. Second, the AREAER is easy to access to verify the indices or their extension to a wider sample of countries or years.

Methodology

The “new” AREAER introduced in 1996 subdivides capital account transactions into 13 subcategories, to which I added a fourteenth as explained below:

  • Capital market securities: shares or other securities of a participating nature, and bonds and other securities with an original maturity of more than one year.

  • Money market instruments: securities with an original maturity of one year or less, such as certificates of deposit, Treasury bills, and so forth.

  • Collective investment securities: share certificates or any evidence of investor interest in an institution for collective investment, such as mutual funds.

  • Derivatives and other instruments: refers to operations in other negotiable instruments and nonsecuritized claims not covered under the previous three items.

  • Commercial credits: covers operations directly linked to international trade transactions.

  • Financial credits: credits other than commercial credits.

  • Guarantees, sureties, and financial backup facilities: securities pledged for payment of a contract, such as warrants, letters of credit, and so on.

  • Direct investment.

  • Repatriation of profits or liquidation of direct investment.

  • Real estate transactions.

  • Personal capital movements: not considered in this paper because of a lack of consistent information in past editions of the AREAER.

  • Provisions specific to commercial banks and other credit institutions: regulations that are specific to these institutions, such as monetary and prudential controls.

  • Provisions specific to institutional investors: one common example is a limit on the share of the institution’s portfolio that may be held in foreign assets.

  • Multiple exchange rate arrangements. These are not part of the capital account subdivision of the AREAER, but I felt it was an important form of capital control. The AREAER systematically provides information on multiple exchange rate regimes.

The rules to construct dummies in each category are as follows:

  • The starting point is the disaggregated dummies provided by the 1996 to 2001 editions of the AREAER. Note that these dummies correspond to the period 1995 to 2000, since each edition reports on restrictions existing as of December 31 of the previous year.

  • Using text information in the 1995 edition, I fill as many of the 13 subcategories as possible for the year 1994. The rule is always the following: one if at least one restriction for that item, zero otherwise.

  • The next step is the “Changes” sections of both the 1995 and 1996 editions. For instance, the text in the 1995 edition might not mention anything about subcategory X in 1994, but the “Changes” section in the 1996 edition might refer to the elimination of some restriction in X during 1995. If one subcategory remains blank after completing this process, I assign it the same value it had in 1995. I call this rule “filling by default.”

  • In the rare case that a subcategory has an NA (not available) value for 1995, I keep the NA unless the text has some explicit information for 1994.

  • Once the 1994 indices have been completed, the process is repeated for 1993, 1992, …, back until 1983.

Other elements in the construction of the indices are worth mentioning. First, whereas the post-1996 editions clearly distinguish between controls on inflows and outflows, the text information in pre-1996 editions may not always contain information on both types of flows. My indices then account for restrictions on inflows and outflows without systematically discriminating between the two. This is in contrast with the pre-1996 IMF dummies, which, as noted earlier, are limited to restrictions on outflows. Second, some countries have restrictions on foreign equity participation in some sectors. The new AREAER will compute this as a double restriction in both capital market securities and in foreign direct investment (FDI). I believe this to be double counting and thus compute the measure as a single restriction on FDI. Since no approach is unambiguously correct, I indicate these sensitive cases with a note in the relevant cell on the spreadsheet. Third, many countries have restrictions on foreign investment in sectors related to defense and public order. Contrary to the AREAER, I chose not to consider these as capital controls. I again indicate these sensitive cases with a note.13

Finally, given that the AREAER fails to provide consistent information on whether countries enforce their restrictions, 1 attribute a value of 1 whenever a control exists, regardless of whether it is enforced. These are fully de jure measures. An exception happens whenever the AREAER indicates explicitly that a given restriction has never been used or enforced. In this case I consider it as nonexistent.14 Once again, I report these rare situations with a note in the relevant cell.

II. Comparison with Alternative Measures

Space considerations prevent a display of the 13 dummies for all countries and all years in the paper. An average over the 13 dummies for each country and each year is presented in Table 2. Tables 3 and 4 present the IMF single dummy and the LMF measure, respectively.15

Table 2.

Disaggregated Indices

article image
article image
Note: For a given country and year, the index represents the average over the 13 different dummies.Source: Author’s own data.
Table 3.

IMF Single-Dummy Indices

article image
article image
Table 4.

Lane and MiIesi-Ferretti Indices

article image
article image
Source: Author’s own computations using data provided by Gian-Maria Milesi-Ferretti.Note: The indices are computed as portfolio and direct investment assets and liabilities as a share of GDP.

global Trend Toward Capital Account Liberalization

Most economists agree that in the past two decades the world has slowly but steadily moved toward greater capital account openness. While this trend may have been deeper in developed countries, it was not necessarily restricted to them.16 To see whether the disaggregated indices capture this trend, I computed a global index for each year by simply averaging over the individual country indices. I also computed a global index from the single dummies and the LMF measure.17 As can be seen in Figure 1, the disaggregated global index (labeled “Miniane” index in the figure) exhibits a pronounced and continued downward trend, from a peak near 0.7 in 1983 to a low around 0.41 in 1999–2000. It is also interesting to note that the downward trend seems to accelerate from the late 1980s to early 1990s. As I show in the next subsection, this is precisely the period during which many European and some Latin American countries were opening their capital accounts.

Figure 1.
Figure 1.

Evolution of Global Indices—All Countries

Citation: IMF Staff Papers 2004, 002; 10.5089/9781589063235.024.A004

The single-dummy and LMF measures exhibit a similar downward trend despite a reversal in the mid- to late 1990s for the former.18 An important difference between the three processes is that, in the case of the IMF dummy, the trend is driven solely by developed countries. To illustrate this, Figures 2 and 3 repeat the exercise in Figure 1 by constructing developing and developed countries subindices for each of the three measures.19 All of them fall considerably throughout the period for the group of developed countries, but only the disaggregated and LMF measures fall for developing nations. The IMF dummy is flat (or increases) for much of the period and falls only in the last year of the sample.

Figure 2.
Figure 2.

Evolution of Global Indices—Developed Nations

Citation: IMF Staff Papers 2004, 002; 10.5089/9781589063235.024.A004

Figure 3.
Figure 3.

Evolution of Global Indices—Developing Nations

Citation: IMF Staff Papers 2004, 002; 10.5089/9781589063235.024.A004

Showing a more detailed perspective, Table 5 computes pairwise time-series correlations between global indices. While these are generally high (above 75 percent), they fall substantially to around 50 percent in the case of the developing countries’ single-dummy subindex. Table 5 also computes for each of the three measures the share of countries for which the country-specific index was higher on average in the first half than in the second half of the period. That share is 100 percent, 82 percent, and 36 percent for the LMF, disaggregated, and single-dummy measures, respectively. In the case of developing countries, the IMF single dummy falls from the first to the second half in just 17 percent of the countries. In short, the single dummies belie the fact that liberalization has also occurred in developing nations, albeit at a slower rate than in developed ones.

Table 5.

Summary Statistics for Global Trends

article image
Source: Author’s own computations.Note: The liberalization measure computes the percentage of countries in which the value for the second half of the period was lower than for the first half.

Cross-Sectional Comparisons

To compare the indices not through time but cross-sectionally, Table 6 ranks the countries from most to least open according to the three measures. For each measure, a country’s index is computed as the average value across all years in the sample. Note that, in the case of the single dummies, all countries up to the United States rank as equally open, and all countries below Brazil as equally closed.20 Reassuringly, the rankings are pretty similar across measures. If one decomposes the sample in three groups according to the IMF single dummy (above the United States, between the United States and Brazil, and below Brazil), then five out of nine countries in the top group also rank among the top nine in the two other measures.21

Table 6.

Cross-Sectional Comparison

article image
Source: Author’s own computations.Notes: The ranking is from most to least open. In the IMF dummy ranking, all countries above the United States are tied with the United States, and all countries below Brazil are tied with Brazil. An asterisk denotes that the country appears in the same group in all three measures (see text for details).

For the middle and lower groups, this number is 6 out of 13 and 6 out of 10 respectively. In other words, the measures agree pretty closely about which countries are very open or very closed and less so about the middle ground.

If one computes cross-country correlations (also in Table 6), the correlation is highest between the two AREAER-based measures (73 percent). This is not surprising given that they share a common source of information. Correlations remain relatively high (around 60 percent) with respect to the LMF measure. Finally, note that many countries in the single-dummy classification are shown as having no restrictions in any year (see Table 3). This is a fallacy that is refuted by the text in the AREAER itself. Indeed, no country in the disaggregated sample ever shows an index of zero for any year. This can be partly explained by the fact that the single dummies account only for controls on outflows as noted by Eichengreen (2001).

Country-Specific Liberalization Episodes

So far, comparisons between measures have dealt with general time-series or cross-sectional trends. The purpose of this section is to study specific liberalization episodes and show how these episodes are tracked by the disaggregated index but not by the single dummy. The LMF measure cannot be used to date liberalization episodes, as it shows a rather smooth opening throughout the period for all countries in the sample. Table 7 summarizes the measures undertaken during 16 liberalization/ tightening episodes captured by the disaggregated index.22

Table 7.

Summary of Main Liberalization/ Tightening Episodes Captured in the Data

article image
article image
article image
article image
article image
article image
article image