Chapter 7. Measuring Financial Integration: A New Data Set

Christopher Crowe, Simon Johnson, Jonathan Ostry, and Jeronimo Zettelmeyer
Published Date:
August 2010
  • ShareShare
Show Summary Details
Martin Schindler

7.1. Introduction

The magnitude of cross-border financial assets holdings has grown in recent years at rising speed, from under 50 percent of world GDP in 1970 to over 300 percent in 2006, and doubling over just the last 10 years (see Figure 7.1). A more financially integrated global economy brings many opportunities, such as improved access to capital and more potential for risk diversification, but the increasing ease at which capital can flow into and out of countries may also carry risks: reversals of capital flows, for example, have contributed to financial crises, and more recently, large net capital flows into the United States may have contributed to the U.S. housing bubble. The ensuing recent subprime mortgage crisis underscored the fact that financial integration binds different parts of the world in good times and bad—in a financially integrated world, market participants in one part of the global economy are no longer sheltered from events emanating in another.

As a consequence, there is great interest in both the academic and the policy community in studying the determinants of financial globalization and its consequences for economic welfare. For example, policymakers averse to the risks of increased financial integration may consider imposing restrictions on cross-border capital flows. Assessing the optimality of such restrictions requires answering (at least) two questions: First, do the risks from increased financial integration outweigh their benefits and, therefore, should one attempt to restrict them in the first place? And second, even if the answer to the previous question is yes, are capital controls an effective tool? It is probably fair to say that economists do not yet have clear answers to these questions. Although a large and growing literature exists on the first question, less work has been done on the second.2 A key reason for this is the paucity of detailed and reliable measurement of countries’ financial globalization strategies, that is, of data on countries’ de jure policies.

Figure 7.1.De facto financial globalization, 1970–2006 (in percent of GDP). Notes: Based on the data provided in Lane and Milesi-Ferretti (2010), updated through 2006. The figure depicts the sum of countries’ total equity, foreign direct investment (FDI), debt, and other assets and liabilities relative to total GDP.

By contrast, de facto measures of financial globalization, such as those presented in the Lane and Milesi-Ferretti chapter (see Figure 7.1), are publicly available for a large number of countries and years. Which type of measure is preferable depends on the research context: for the purpose of policy analysis, de jure measures, which are under the policy maker’s direct control, are more relevant, whereas in other applications, only outcome (de facto) measures may matter. In still other situations, both may be necessary, for example, if one wants to study the extent to which de jure controls affect de facto outcomes. However, given the limited availability of detailed de jure data—available data are often too coarse, have limited time and/or country coverage, or are unavailable to the public—many authors have resorted to using de facto measures even when they were interested in studying policies. This chapter documents, and makes publicly available, a detailed panel data set on countries’ disaggregated de jure measures, in the hope that it will allow more progress to be made in answering some of the important questions in this field. 3

7.2. The Data Set

7.2.1. Methodology

The data set is a balanced panel, covering 91 countries on an annual frequency during the time period from 1995 to 2005. It provides novel detail on the various dimensions in which countries impose restrictions on financial transactions, and the sample of countries it covers is diverse in terms of regions and income levels, covering 35 high-income countries, 42 middle-income countries, and 14 lowincome countries (see Table 7.1 for the full country list by region).

Common to nearly all existing de jure capital control indices is their reliance on information contained in the IMF’s Annual Report on Exchange Arrangements and Exchange Restrictions (AREAER). Thus, although drawing on the same source, these indices differ in how, and to what extent, they extract the information provided in the AREAER. Until 1995, the AREAER summarized a country’s openness to capital flows using a binary dummy variable, where 1 represents a restricted capital account and 0 represents an unrestricted capital account. Since 1995, the AREAER has utilized a more structured approach, providing detailed information on restrictions on capital transactions in a number of subcategories.

Table 7.1List of Countries in the Data Set
High IncomeMiddle IncomeLow Income
East Asia and PacificEast Asia and PacificEurope and Central Asia
AustraliaChinaKyrgyz Republic
Hong Kong SARIndonesiaUzbekistan
KoreaPhilippinesMiddle East and North Africa
New ZealandThailandYemen, Republic of
Europe and Central AsiaSouth Asia
Europe and Central AsiaBulgariaBangladesh
CyprusCzech RepublicIndia
KazakhstanSub-Saharan Africa
Middle East and North AfricaLatviaBurkina Faso
BahrainMoldovaCôte d’Ivoire
Brunei DarussalamRomaniaGhana
Saudi ArabiaLatin America and CaribbeanUganda
United Arab EmiratesArgentinaZambia
North AmericaBrazil
United StatesCosta Rica
Dominican Republic
Western EuropeEcuador
AustriaEl Salvador
MaltaVenezuela, República Bolivariana de
NorwayMiddle East and North Africa
United KingdomTunisia
South Asia
Sri Lanka
Sub-Saharan Africa
South Africa

The data set documented here contains information from a subset of these subcategories in broad correspondence to the standard presentation of de facto assets and liabilities (as, for example, in Lane and Milesi-Ferretti, 2010). The asset categories covered here constitute the lion’s share of global cross-border asset holdings; thus, focusing on these categories allows for the construction of a data set that broadly reflects the structure of global de facto financial integration. 4 The main categories covered in this data set are as follows (with the names in square brackets reflecting those used in the published data set):

1. Shares or other securities of a participating nature [eq];

  • Purchase locally by nonresidents [eq_plbn];
  • Sale or issue abroad by residents [eq_siar];
  • Purchase abroad by residents [eq_pabr];
  • Sale or issue locally by nonresidents [eq_siln];

2. Bonds or other debt securities [bo]; 5

  • Purchase locally by nonresidents [bo_plbn];
  • . Sale or issue abroad by residents [bo_siar];
  • . Purchase abroad by residents [bo_pabr];
  • . Sale or issue locally by nonresidents [bo_siln];

3. Money market instruments [mm];

  • Purchase locally by nonresidents [mm_plbn];
  • Sale or issue abroad by residents [mm_siar];
  • Purchase abroad by residents [mm_pabr];
  • Sale or issue locally by nonresidents [mm_siln];

4. Collective investments [ci];

  • . By residents to nonresidents [cio];
  • By nonresidents to residents [cii];

5. Financial credits[fi];

  • . By residents to nonresidents [fco];
  • By nonresidents to residents [fci];

6. Direct investment [di];

  • Outward investment [dio];
  • . Inward direct investment [dii];
  • Liquidation of direct investment [ldi].

To allow for a flexible use of the data, the information contained in the AREAER is coded at the level of resident/nonresident restrictions, in binary form, taking a value of 0 (unrestricted) or 1 (restricted) (see below for the slightly different cases of restrictions on collective investments, financial credits, and direct investment). In each case, restrictions on capital transactions are coded as a 0 if there are none in a given year, or if they consist merely of registration or notification requirements. 6 They are also coded as 0 if a country is generally open but imposes restrictions on investments in only a small number of selected industries, for example, for national security purposes, or on financial transactions with only a small number of countries, typically for political reasons. 7

Given that capital account restrictions are coded at the level of individual transactions, the data can be aggregated in different ways, providing information along different dimensions. In particular, the coded data allow for the construction of capital control subindices byasset category, by residency, 8 and by the direction of flows9 (inflows versus outflows). The simplest way of aggregating subindices, and the one followed here, is by taking unweighted averages of the appropriate subcategories. Thus, for example, a country’s restrictiveness of individual asset categories would be captured by averaging across each asset category’s various subcomponents to obtain


for i ε { eq, bo, mm }. Given that each of the subcategories is coded as a binary variable, the resulting asset-specific aggregate can take on five different values. For collective investments and financial credits, where the AREAER provides less disaggregated information on restrictions, the aggregated index is simply the average of the two subcategories, implying that each of these can take on only three different values (0, 0.5, and 1).

For direct investment, in addition to information on inward and outward restrictions, the AREAER provides a third category regarding the “liquidation of direct investment.” To maintain symmetry to other categories in terms of the values the subindex can take on, the published data set is calculated as the average between dio and the maximum of dii and ldi. This aggregation recognizes that liquidation restrictions make reversals more costly, and thus indirectly impose costs on direct investment inflows. However, different aggregations will be appropriate in different contexts, such as a simple average of all three subcategories, and the modular structure of the data set provides researchers with the option of exploring these alternatives.

Variables summarizing controls according to residency can be obtained by calculating the average of “sale or issue abroad by residents” and “purchase abroad by residents” for resident restrictions, and the average of “purchase locally by nonresidents” and “sale or issue locally by nonresidents” for nonresident restrictions. In this context, controls on direct investment inflows (as described above) can be interpreted as nonresident restrictions, and those on direct investment outflows as resident restrictions. 10

In each asset category, indicators can also be grouped according to the direction of flows. With the exception of direct investment (see footnote 11), the direction of flows is conceptually independent of the residency status of the transacting individual. For example, a capital inflow may arise from a nonresident purchasing domestic assets (increasing the country’s stock of external liabilities), or from a domestic resident’s sale of assets abroad (decreasing the country’s stock of external assets). Thus, inflow restrictions are calculated here as the average of the restriction dummies on “purchase locally by nonresidents” and “sale or issue abroad by residents,” whereas outflow restrictions are calculated as the average of the restriction dummies on “purchase abroad by residents” and “sale or issue locally by nonresidents.” 11

Further aggregation across asset categories yields broader indices of a country’s restrictiveness of capital account transactions. Again, the modular nature of the data set provides flexibility for a variety of different aggregations—researchers using these data will have to make a determination as to which aggregation is most appropriate given the research question at hand. It is also worth noting that although the basic coding at the level of individual transactions consists of a binary indicator, the cross-sectoral and time variation that results from aggregating indices along various dimensions can be interpreted as a measure of the intensity of a country’s capital controls, because such aggregations effectively “Account” how many subcategories are restricted, and within each category, how many types of transactions.12

7.2.2. Comparison with Existing Indices

A comparison of the new index with existing indices highlights the trade-offs one faces in their construction. Between a broad country coverage, long time coverage, and detailed information on the types of restrictions, typically only two can be achieved. As discussed in the previous section, the new index documented in the previous section strikes a relatively favorable balance regarding country coverage and the level of detail, but is constrained by a relatively short time series dimension because of the limited information provided by the AREAER prior to 1995. Authors of other capital account indices have made different choices.

Most closely related to the index presented here is the work done by Tamirisa (1999) who followed a similar approach. Although her index has a broad country coverage, it covers only one year. 13 By contrast, Miniane (2004) aimed to extend the time series dimension, at the cost of a more limited country coverage and less detail. His sample includes 34 countries covering the period 1983–2000. Given the more limited information available in the AREAER prior to 1995, his index cannot distinguish between inflow and outflow restrictions.

Other authors have aimed to maximize time and country coverage, at the expense of less detail at the country level. Four binary indicators that were reported in the AREAER prior to 1995 include (1) the openness of a country’s capital account; (2) the openness of the current account; (3) the stringency of requirements for the repatriation and/or surrender of export proceeds; and (4) the existence of multiple exchange rates for capital account transactions. Many authors simply use the capital account dummy under (1) as a measure of a country’s capital account openness. Given its binary nature, this is a crude approximation of a country’s capital account restrictiveness, although it has the advantage of a broad country and time coverage, being available for up to 184 countries at an annual frequency starting in 1966.

Mody and Murshid (2002) extend these dummies into the years after 1995, thus covering the years 1966–2000 and 184 countries. They calculate a “financial integration index” as the sum of the four binary variables ranging from 0 to 4, with 4 denoting the least restricted. Chinn and Ito (2007) also construct a composite measure from these four dummy variables taking a principal components approach. In an updated version of their data set, Chinn and Ito apply this procedure to 182 countries for 1970–2006. Although the Mody-Murshid and the Chinn-Ito measures provide more finely graded information than the simple IMF dummy, it is less clear to what extent these indicators are measures of capital account openness in a narrow sense, given that three of the four indices underlying these indicators represent information that is not directly related to capital account transactions.14 By contrast, some authors have chosen a narrow approach. For example, Bekaert, Harvey, and Lundblad (2005) focus on only equity controls and attempt to date equity liberalization episodes for a sample of 42 countries during 1960–2006. Edison and Warnock (2003) focus on de facto equity restrictions for a sample of 31 countries during 1989–2006 at a monthly frequency, by measuring the fraction of a country’s market capitalization that is open to foreign investment.

None of the above indices captures the intensity of controls or distinguishes between asset categories, inflows and outflows, or residents versus nonresidents. For example, regarding the intensity of restrictions, whether a financial transaction is prohibited, limited, taxed, or only requires notification/registration is likely to have different economic consequences. Quinn (1997) constructs a data set that contains information on the intensity of controls and covers 94 countries during 1950–1999. He captures the intensity of controls by ranking different control instruments by their (assumed) economic importance 15 and it is the only index doing so for a large number of countries and years.16 His index also distinguishes between restrictions on residents and nonresidents. A recently updated version extends the data coverage through 2005. Similar to Miniane, given the less structured nature of the AREAER prior to 1995, consistency and comparability requirements across country-years imply that Quinn’s (1997) index cannot distinguish between inflows and outflows (see the discussion in the previous section) or between different asset categories.

Table 7.2 shows pairwise correlations of the various indices at their most aggregated levels. The correlations are reassuringly high, and particularly so between the new index and those by Miniane and Tamirisa—this is not surprising given that these indices employ similar methodologies. By contrast, the equity liberalization index by Edison-Warnock is based on a rather different methodology, and effectively is a de facto measure, but, at around 0.47, the correlation is still relatively high. 17 Overall, the high correlations with other indices at the aggregate level instill confidence that the new index also captures meaningful information at more disaggregated levels that existing indices cannot provide.

Table 7.2Pairwise Correlations of Alternative Capital Control Indices
Pairwise Correlations of Alternative Capital Control Indices
New IndexIMF DummyBekaert and OthersChinn-ItoEdison-WarnockMody-MurshidMinianeTamirisaQuinn
New Index1
IMF Dummy0.7491
Bakaert and others-0.0620.2221
Notes: The indices and their sources are described in the text. All indices are normalized to the unit interval, with higher values indicating higher restrictiveness. For each pair, the table lists the correlation coefficient, the level of significance (in parentheses), and the number of pairwise observations.
Notes: The indices and their sources are described in the text. All indices are normalized to the unit interval, with higher values indicating higher restrictiveness. For each pair, the table lists the correlation coefficient, the level of significance (in parentheses), and the number of pairwise observations.

7.3. Empirical Applications

The new index can be used to study a broad range of questions of interest that could not be examined previously. In particular, by exploiting novel features of the index, specifically the possibility of separately considering controls by asset categories, resident status, and direction of flows, new research avenues open up. This section highlights some of these features and outlines possible directions for future research.

7.3.1. Trends and Cross-Country Comparisons

Although other indices with longer time coverage are better able to present long-term trends, the new index extends into 2005 and thus can capture more recent trends. Figure 7.2 plots average trends for some of the main indices: Miniane’s, Chinn-Ito’s, the IMF dummy, Quinn’s, and the new index. The various indicators are all fairly consistent in their time series variation and all document a broad trend toward increased de jure liberalization of financial flows over most of the past decade. All of the indices also point toward a slowdown in the pace at which countries are liberalizing their capital accounts. In fact, the indices point to a possible reversal in 2005, although additional time coverage will be necessary to draw more meaningful conclusions. Although there were, in nearly all regions, both countries increasing and decreasing their average degree of restrictiveness in 2005, many European countries were among those with the highest increases in capital account restrictions, such as Austria, Kyrgyz Republic, Belgium, Czech Republic, and Uzbekistan. 18

Compared to previous indices based on the IMF’s binary capital controls dummy, the new index also allows for a more meaningful comparison of the levels of capital account restrictiveness across regions and income groups. Considering sample averages of the IMF dummy is equivalent to counting the number of countries that exceed a certain (undefined) threshold of capital account controls, without quantifying how restrictive individual countries in a group are, making cross-country rankings of capital account restrictiveness difficult to interpret.Figure 7.3 provides regional averages for 1995 (the last year the IMF dummy was officially reported) and illustrates that these differences in indices may indeed lead to different rankings. The simple dummy overstates restrictiveness in most regions, particularly in Asia, sub-Saharan Africa, and Latin America, whereas it understates average capital account restrictiveness in North America and, to a lesser extent, Europe.19 Thus, the new index arguably provides a more realistic and meaningful comparison across regions (and countries).

Figure 7.2.Trends in de jure financial openness, 1975–2005. Notes: The indices and their sources are described in the text. Where applicable, they were rescaled and normalized to [0,1], with higher values indicating higher restrictiveness.

Figure 7.3.Regional averages of de jure financial openness, 1995. Notes: The indices and their sources are described in the text. Higher values indicate higher restrictiveness.

7.3.2. Compositional Changes

A key strength of the new index is its ability to provide information on a country’s composition of capital account restrictions in addition to simply measuring the country’s overall restrictiveness. Figure 7.4 shows a decomposition by asset categories, by the direction of flows, and by residency. The figure exhibits substantial changes in the relative importance of controls across asset categories, with an overall trend of convergence across asset groups. This convergence may be a response to growing market sophistication which increasingly enables market participants to circumvent differential treatment of different asset categories—equal restrictions across asset categories may thus facilitate their enforcement. Although the relative levels of resident/ nonresident and inflow/outflow restrictions have been fairly stable during most of the decade, the 2005 data points to a divergence in relative inflow and outflow controls, with countries imposing more restrictions on outflows than on inflows.

Recent research has started to take advantage of the information on the composition of capital controls. For example, Prati, Schindler, and Valenzuela (2009), among other things, exploit the inflow/outflow distinction in combination with firm-level credit ratings data to identify the channel through which capital account liberalization affects an economy. Dell’Ariccia and others (2010) investigate the link between de facto financial integration and de jure capital account restrictiveness using a gravity approach. To do this, they combine a country’s outflow controls with its partner country’s inflow controls to construct a measure of bilateral capital controls.

7.3.3. Event Studies

Another important feature of the index is its relatively fine gradation, allowing researchers to identify large changes in de jure regimes, thus being able to date reform events. This type of approach has been advocated by Henry (2007). For example, in assessing the existing literature on effects of capital account liberalization on economic growth, he argues that attempting to find long-term growth benefits is ill-conceived, as simple growth theory would predict only temporary growth effects during the transition to a new steady state. Event studies focusing on the immediate period around liberalization episodes may, therefore, be a more appropriate framework for testing for growth effects.

An application of this approach to the effects of de jure liberalization on de facto financial integration is illustrated in Figure 7.5 where large capital account reforms and reversals are identified both in the aggregate and by asset category.

Figure 7.4.The composition of capital controls, 1995–2005. (A) Asset categories. (B) Inflows vs. outflows. (C) Residents vs. nonresidents. Notes: The debt category in panel (A) is defined as the average of the bond and money market restriction indices.

Figure 7.5.Changes in de facto financial integration following large de jure reforms/reversals (in percent). Notes: Based on Lane and Milesi-Ferretti (2010) and the new de jure index. The figure plots the percent difference between countries’ average financial integration, defined as the ratio of external assets and liabilities to GDP, three years before and after reforms (reversals). Reforms (reversals) are defined as annual changes in the new de jure index exceeding the median positive (falling below the median negative) annual change in the index.

The dating of such events can, for example, help answer questions regarding the effectiveness of controls in enabling or reducing de facto capital movements. The figure suggests that there may indeed be an association. The strength of this association varies between asset categories and also between reform and reversal episodes. Broadly speaking, countries’ de facto integration jumps up substantially around the time of reform episodes. De facto integration also increases following reversal episodes, but to a much lesser extent. One interpretation is that liberalizing the capital account can substantially help a country attract foreign capital, but that the reverse is not true: that is, countries may not be able to completely insulate themselves from financial flows by imposing restrictions.

Figure 7.5 is also suggestive of another result, namely, that capital controls for some asset categories are more effective than for others. For example, lifting equity controls (and, to a lesser extent, debt controls) coincides with dramatic increases in de facto integration, whereas there is virtually no such association for FDI. Although such correlations do not establish a causal relationship, they are suggestive of a link and warrant more careful investigation.

7.4. Conclusion

This chapter has presented and documented a new data set of countries’ de jure restrictions on cross-border financial transactions. As any measure of capital account restrictions, its construction required striking a balance between various features of the data, such as breadth (information by assets, direction of flows, and residency), depth (the intensity of controls), and country and time coverage. Besides a fairly broad country coverage, the distinguishing feature of the data set presented here is its level of disaggregation, not found in other indices. By coding the data at the level of individual types of transactions, the data set has a modular setup which allows researchers to “mix and match” by averaging across the various subcategories in ways that best fit their research objectives. The chapter also outlined several research avenues that the new index makes possible and that could help make progress in better understanding the many facets of financial globalization.


    BekaertGeert Campbell R.Harvey and ChristianLundblad2005“Does Financial Liberalization Spur Growth?,”Journal of Financial EconomicsVol. 77No. 1pp. 3–55.

    • Search Google Scholar
    • Export Citation

    BruneNancy2006“Financial Liberalization and Governance in the Developing World,”Ph.D. Dissertation Yale University.

    BruneNancy and AlexandraGuisinger2007“Myth or Reality? The Diffusion of Financial Liberalization in Developing Countries,”Yale University MacMillan Center Working Paper(New Haven, CT: Yale University).

    • Search Google Scholar
    • Export Citation

    ChinnMenzie D. and HiroIto2007“A New Measure of Financial Openness,”Journal of Comparative Policy AnalysisVol. 10 No. 3pp. 307–20.

    • Search Google Scholar
    • Export Citation

    Dell’AricciaGiovanni Julian diGiovanni AndréFaria M. AyhanKose PaoloMauro Jonathan D.Ostry MartinSchindler and MarcoTerrones2010“Reaping the Benefits of Financial Globalization,” Chapter 9 in this volume.

    • Search Google Scholar
    • Export Citation

    EdisonHali and FrankWarnock2003“A Simple Measure of the Intensity of Capital Controls,”Journal of Empirical FinanceVol. 10 No. 1pp. 81–103.

    • Search Google Scholar
    • Export Citation

    HenryPeter Blair2007“Capital Account Liberalization: Theory, Evidence, and Speculation,”Journal of Economic LiteratureVol. 45 pp. 887–935.

    • Search Google Scholar
    • Export Citation

    International Monetary Fund1993Balance of Payments Manual 5th ed.(Washington,International Monetary Fund)

    International Monetary Fund various years1996–2006Annual Report on Exchange Arrangements and Exchange Restrictions(Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

    Johnston R. Barry and Natalia T. Tamirisa 1998“Why Do Countries Use Capital Controls?,”IMF Working Paper 98/181(Washington:International Monetary Fund). pp.

    • Search Google Scholar
    • Export Citation

    Kose M. AyhanEswar Prasad Kenneth Rogoff and Shang-Jin Wei 2010“Financial Globalization: A Reappraisal,”Chapter 10 in this volume.

    • Search Google Scholar
    • Export Citation

    Lane Philip R. and Gian Maria Milesi-Ferretti 2010“The External Wealth of Nations Mark II: Revised and Extended Estimates of Foreign Assets and Liabilities, 1970–2004,”Chapter 8 in this volume.

    • Search Google Scholar
    • Export Citation

    Miniane Jacques 2004“A New Set of Measures on Capital Account Restrictions,”IMF Staff PapersVol. 51 No. 2pp. 276–308.

    ModyAshoka and Antu Panini Murshid.2002“Growing Up With Capital Flows,”Journal of International EconomicsVol. 65 No. 1 pp. 249–66.

    • Search Google Scholar
    • Export Citation

    Prati Alessandro Martin Schindler and Patricio Valenzuela 2009“Who Benefits from Capital Account Liberalization? Evidence from Firm-Level Credit Ratings Data,”IMF Working Paper No. 09/210((Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

    Quinn Dennis 1997“The Correlates of Change in International Financial Regulation,”American Political Science ReviewVol. 91(September) pp. 531–51.

    • Search Google Scholar
    • Export Citation

    Tamirisa Natalia T. 1999“Exchange and Capital Controls as Barriers to Trade,”IMF Staff PapersVol. 46 No. 1 pp. 69–88.

This chapter is a slightly revised version of an article that appeared in IMF Staff Papers, Vol. 56, No. 1 (2009): 222-38.


The author gratefully acknowledges the contributions by Lore Aguilar during the initial stages of this project. An early version of the data was used in Dell’Ariccia and others (2010)—all collaborators on that project also contributed in some form to the data effort presented here. The author also benefited from discussions with Enrica Detragiache, Peter Blair Henry, Ayhan Kose, Gian Maria Milesi-Ferretti, Jacques Miniane, Eswar Prasad, Dennis Quinn, and Frank Warnock. Gian Maria Milesi-Ferretti and Dennis Quinn kindly provided their updated data sets. Patricio Valenzuela and Ermal Hitaj provided outstanding research assistance. The data set described in this chapter can be downloaded from the IMF Staff Papers website.


Even so, however, no clear consensus has emerged on the effects of financial cross-border flows on economic growth and other outcome variables, unlike, for example, the literature on the cross-border trade of goods and services. For recent reviews of the state of the financial globalization literature, see, for example, the chapters by Kose and others (2010) and Dell’Ariccia and others (2010).


Taking a disaggregated approach appears to be promising. As Henry (2007) notes, existing evidence suggests that opening equity markets to foreign investors may avoid some of the problems associated with the liberalization of debt flows, and so “[a]t a minimum, the distinction between debt and equity is critical” (p. 889). The data set documented here allows researchers to investigate such differences


Not all categories reported in the AREAER are coded here, given their limited importance in the composition of de facto flows and resource constraints in the data collection process. These categories include:derivatives and other instruments, credit operations (except for the subcategory financial credits, see main text), real estate transactions, and personal capital transactions.


Restrictions on bonds transactions were not recorded in the AREAER in 1995 and 1996


For example, in 1999, Bulgarian residents’ purchases abroad of capital market securities (shares and bonds) only required “[p]rior registration with the BNB” (IMF, 2000, p. 150) and were therefore coded as 0 (unrestricted).


For example, in 2005, in the United States local purchases by nonresidents of shares or other securities of a participating nature were free from restrictions except for investments “in the nuclear energy, maritime, communications, air and land transport, and shipping industries” (IMF, 2006, p. 1258); transactions were also prohibited with “Cuba and Cuban nationals; the Islamic Republic of Iran; Myanmar; Sudan” (IMF, 2006, p. 1259). These transactions were coded as 0 (unrestricted).


The definition of residence in the Balance of Payments is based on the “transactor’s center of economic interest” (IMF, 1993, p. 20) and may therefore differ from other definitions based on nationality or (other) legal criteria, such as tax laws.


The interpretation of directionality is a complex issue–see below for a more detailed discussion


This view follows the Balance of Payments Manual which notes that “[d ]irect investment is classified primarily on a directional basis–resident direct investment abroad and nonresident investment in the reporting economy” (IMF, 1993, p. 81). Thus, unlike the other categories, direct investment inflows and outflows can be equated with nonresident and resident transactions, respectively. For symmetry, the analogous approach is taken for the collective investment and financial credit categories.


Matching these inflow/outflow aggregates with their de facto counterparts is nontrivial: capital flows data are typically reported as the net changes in external assets (outflows) and liabilities (inflows), which mixes different types of transactions. For example, a reduction in liabilities because of nonresidents selling domestic bonds is effectively counted as a negative inflow, whereas it would be considered a (positive) outflow in the de jure aggregation discussed here. Transforming the de facto data by defining Outflows = max(ΔAssets,0) – min(ΔLiabilities,0) and Inflows = –min(ΔAssets,0) + max(ΔLiabilities,0) is a possible solution.


This is only one aspect of intensity. A broader intensity measure would reflect the different types of restrictions (such as approval vs. taxation vs. prohibition) as well as the degree to which de jure restrictions are actually enforced in practice. Quinn (1997) attempts to tackle the former aspect, described in more detail in the next section.


Johnston and Tamirisa (1998) analyze in more detail the various subcomponents of Tamirisa’s (1999) index.


This is not to say, however, that these other three variables have no bearing on capital account restrictions; for example, multiple exchange rate practices may make capital account transactions more costly even in the absence of other, more direct restrictions on capital account transactions.


Such a ranking is difficult as the relative importance of restrictions likely depends on the specific context. For example, whether “approval required but frequently granted” is equally restrictive as “approval not required, but heavily taxed” (as assumed in Quinn, 1997) will depend on the level of the tax rate and the precise definition of “frequently granted.”


A possible exception is Brune (2006) who, as described in Brune and Guisinger (2007), constructed a data set covering 187 countries during 1965–2004 and containing separate information on inflow and outflow restrictions in five categories. She reports high correlations with the IMF dummy and the indices by Tamirisa (1999), Miniane (2004), and Quinn (1997); however, her data set has not been available to the author.


The correlation with the binary equity liberalization index by Bekaert, Harvey, and Lundblad (2002) (switching from 1 to 0 when equity markets are liberalized) is statistically insignificant as their data set reports only six liberalization episodes that fall in the sample of the new index: Cöte d’Ivoire, Kenya, and Tunisia in 1995, South Africa in 1996, and Oman and Saudi Arabia in 1999.


The 2005 reversal is also reflected in individual asset categories except for FDI where the trend toward fewer restrictions continues even into 2005; inflow restrictions on average also continued to decrease (see Figure 7.4).


The capital controls index for the United States, for example, is coded as non-zero in the new index because of restrictions on foreign mutual funds (“sale or issue locally by nonresidents”) under the Investment Company Act (see IMF, 1996).

    Other Resources Citing This Publication