Why Do Countries Use Capital Controls?
Author:
Ms. Natalia T. Tamirisa
Search for other papers by Ms. Natalia T. Tamirisa in
Current site
Google Scholar
Close
and
Mr. R. B. Johnston
Search for other papers by Mr. R. B. Johnston in
Current site
Google Scholar
Close

Contributor Notes

Authors’ E-Mail Addresses: bjohnston@imf.org, ntamirisa@imf.org

Recourse to controls on capital flows among developing economies is generally quite pervasive. This paper examines the structure and determinants of capital controls based on a cross-sectional study of developing and transition economies. It identifies categories of capital transactions that can be aggregated for analytical purposes. Controls are found to be related to the balance of payments, macroeconomic management, market and institutional evolution, prudential and other factors. The relationship with the balance of payments, however, is not robust to simultaneous equation analysis.

Abstract

Recourse to controls on capital flows among developing economies is generally quite pervasive. This paper examines the structure and determinants of capital controls based on a cross-sectional study of developing and transition economies. It identifies categories of capital transactions that can be aggregated for analytical purposes. Controls are found to be related to the balance of payments, macroeconomic management, market and institutional evolution, prudential and other factors. The relationship with the balance of payments, however, is not robust to simultaneous equation analysis.

I. Introduction

This study presents initial systematic evidence on the structure of capital controls and their determinants in a cross-section of 45 developing and transition economies. To the best of our knowledge, it is the first study that examines the detailed structure of controls on capital movements. It has been facilitated by the information that is now reported in the IMF’s Annual Report on Exchange Arrangements and Exchange Restrictions (AREAER). Beginning in 1996, data in the AREAER was expanded to cover capital movements in a comprehensive manner and was presented in a new tabular format.

In the wake of the Asian currency crisis, there has been renewed interest in the use of capital controls. The nature of transactions and structure of regulation that can influence international capital movements are, however, potentially numerous and highly complex. In some countries the use of capital controls is pervasive, while in others it is selective. In some countries the structure of controls appears to be a legacy of the past, while other countries appear to use controls as active instruments of macroeconomic and structural policy. In order to design programs of liberalization of the capital account, or the regulatory frameworks for capital movements, it would be helpful to know what factors have generally led to the regulation or deregulation of components of the capital account. In designing such programs, it would also be useful to know if the regulation or deregulation of certain capital transactions has generally been associated with the regulation or deregulation of other transactions. Answering these questions requires a detailed study of the structure of individual capital controls and their relationship with economic and structural variables. This is the purpose of this study; the study does not examine in detail whether capital controls have been effective in achieving their objectives.2

By considering economic factors explaining the structure of capital controls, the study complements earlier research on determinants of the capital regime as a whole. Alesina, Grilli, and Milesi-Ferretti (1994) and Grilli and Milesi-Ferretti (1995) found that capital controls are more likely to exist in countries with fixed or managed exchange rate regimes, lower per capita incomes, larger government consumption as a ratio to GDP, less independent central banks, and larger current account deficit. Focusing on the political economy of the capital regime in OECD countries, Quinn and Inclan (1997) created measures of financial openness including controls on current and capital transactions and showed that financial openness is linked to the weakness of left-wing governments and independence of the central bank. Johnston and others (1999) constructed relatively comprehensive indices of capital controls and showed that restrictive capital regimes, as measured by the indices, are associated with low levels of economic development, high tariff barriers, large black market premia, and high volatility of the exchange rates. Unlike the previous studies, this study focuses on the structure of capital controls rather than the overall capital regime.

The paper is organized as follows. Data on capital controls and stylized facts concerning the structure of capital controls are described in Section II. Section III reviews theoretical explanations of the use of capital controls. Section IV describes a simple empirical model of capital controls that synthesizes the existing explanations, and Section V discusses findings concerning determinants of capital controls. Section VI concludes.

II. Data and Stylized Facts Concerning Capital Controls

This section describes the classification of and data on capital controls that are used in this study and presents some stylized facts concerning the structure of capital controls that are supported by the data.

A. Data on Capital Controls

The study uses disaggregated measures of capital controls based on the classification and data in the IMF’s AREAER. The framework for collecting and classifying information on controls on capital movements in the AREAER is summarized in Table 1. The classification scheme distinguishes between a number of different types of transactions, which contribute to capital movements such as securities, money market instruments, credit operations, direct investment, etc. The scheme also distinguishes between capital inflows and outflows, and between the different types of specific transactions, which can result in such flows, e.g., for inflows through securities, purchases of local securities by nonresidents and sales or issues of securities abroad by residents; and for outflows, purchases of securities abroad by residents and the sale or issue of securities locally by nonresidents. In addition, the database covers a number of provisions specific to commercial banks and institutional investors that can influence capital movements (such as reserve requirements discriminating between resident and nonresidents deposits, or open foreign position limits imposed asymmetrically with regard to short or long currency positions or vis-à-vis residents and nonresidents).

Table 1.

Types of Capital Transactions Possibly Subject to Controls

article image

The information in the database is intended to be comprehensive and to include regulations that affect capital flows. The inclusion of information in the database does not imply any judgement of whether a measure is considered restrictive3 or would be justified by the relevant circumstances, nor does the database seek to distinguish the purpose of controls, e.g., whether they are motivated by prudential or balance of payments considerations.

The information used in this study refers to a sample of 45 developing and transition countries, which, first, have close to complete data in the AREAER,4 and, second, represent various geographical regions, levels of economic development, and experiences with capital account liberalization (see Appendix I for the list of countries).

B. Stylized Facts Concerning the Structure of Capital Controls

Summary statistics for aggregate measures of the extent of capital controls are presented in Table 2. These measures refer to:

  • overall controls on capital movements;

  • overall controls on capital inflows and overall controls on capital outflows;

  • controls on inflows and controls on outflows pertaining to capital and money market securities, collective investment securities and derivative instruments;

  • controls on inflows and controls on outflows pertaining to commercial and financial credits and guarantees and sureties;

  • controls on inflows and controls on outflows related to direct investment and real estate; and

  • provisions specific to commercial banks and other financial institutions.

Table 2.

Summary Statistics for Aggregate Measures of Capital Controls 1/

article image

Total number of controls in the respective category, as classified and reported in AREAER.

Includes differential reserve and liquid asset requirements, and interest rate and credit controls; investment regulations, and open foreign exchange position limits.

Table 2 shows that controls on capital outflows are more prevalent than controls on inflows on all types of transactions, except in the case of direct foreign investment and real estate purchases. The incidence of measures specific to commercial banks and other financial institutions is lower than that of other types of capital controls in the country sample.

Tables 34 present correlation matrices for aggregate and individual measures of capital controls respectively. These tables illustrate a number of regularities between uses of different capital controls. Such regularities correspond to high and medium correlations, defined as correlation coefficients of 0.8–1.0 and 0.5–0.7, respectively. The following main conclusions emerge:

Table 3.

Correlations Between Aggregate Measures of Capital Controls 1/

article image

Total number of controls in the respective category, as classified and reported in AREAER.

Includes differential reserve and liquid asset requirements, and interest rate and credit controls; investment regulations, and open foreign exchange position limits.

Table 4.

Correlations Between Individual Measures of Capital Controls 1/

article image
article image

Dummy variables based on AREAER,

Concerning correlations between aggregate measures of capital controls (Table 3):

  • There are high correlations between controls on inflows and outflows pertaining to capital and money market operations, suggesting that these are imposed together. There are also high correlations between outflows of foreign direct investment and real estate and controls on outflows (and inflows) pertaining to money and capital market transactions, suggesting fungibility between these transactions.

  • There are medium correlations between controls on inflows and outflows pertaining to credit operations; between controls on inflows of direct investment and inflows through capital and money market securities; and between controls on inflows and outflows of credits and controls on inflows and outflows through capital and money market securities.

  • There is generally limited correlation between controls on inflows and outflows related to direct investment and real estate, and between provisions specific to commercial banks and other financial institutions and other capital controls. This would suggest that these controls are normally imposed for different purposes. Inflows of direct investment, for example, are often regulated for noneconomic reasons (i.e., social, sectoral, and strategic).5

Concerning correlations between controls on individual transactions (Table 4):

  • There are high and medium correlations between controls on purchases abroad by residents of capital market securities, money market securities, collective investment securities and derivative products and the issue locally of those by nonresidents. This is the group of transactions that makes up the aggregate category “controls on outflows pertaining to capital and money market securities, collective investment securities and derivative instruments”. Similarly, high and medium correlations are found between the individual components making up the category “controls on inflows pertaining to capital and money market securities, collective investment securities and derivative instruments”. These correlations suggest that controls on inflows and outflows related to capital and money markets tend to be applied consistently regardless of the maturity of the transaction, possibly reflecting substitutability among the respective transactions. High and medium correlations also exist between the individual components comprising the category “controls on inflows and outflows pertaining to commercial and financial credits, guarantees and sureties”. Controls on outward direct investment and real estate purchases abroad are also highly correlated; there is medium correlation between controls on liquidation of direct investment and sales of real estate locally by nonresidents. All in all, high and medium correlations between controls on individual transactions provide a statistical basis for the respective aggregate measures discussed above.

  • Correlations between the individual controls that pertain to commercial banks and institutional investors and other capital controls are generally low. For instance, there is little correlation between the use of differential reserve requirements on resident and nonresident liabilities and controls on other transactions or measures pertaining to other activities of banks and institutional investors. However, correlations exist between controls on banks’ open position limits, on their lending locally in foreign exchange and controls on derivative transactions more generally; and between differential interest rate controls and controls on various underlying capital transactions. This suggests that certain bank regulations complement other types of capital controls, or vice versa.

III. Theoretical Explanation of Capital Controls

A number of different motivations have been suggested for maintaining controls on capital movements. While there is some overlap between the motivations, they can generally be classified into those related to: (1) balance of payments and macroeconomic management; (2) underdeveloped financial markets and regulatory systems (referred to as market and institutional evolution reasons); (3) prudential; and (4) other reasons. The latter may include international and national security, structural considerations related to economic size, openness to trade, etc.; and social, sectoral, and strategic considerations. These reasons are discussed in more detail below.

A. Balance of Payments and Macroeconomic Management

Historically, capital controls have most often been justified as one of the instruments for balance of payments and macroeconomic management. Thus, the classic rationale for the use of exchange controls has referred to countries with weak balance of payments which sought to prevent an outflow of capital by imposing capital controls. The evidence that controls have protected the balance of payments in developing countries against outflows is, however, generally weak (see Johnston and Ryan (1994)).

This balance of payments argument is also variously presented in terms of preserving a shortage of domestic savings for domestic uses, or of preserving a capacity to finance the domestic fiscal deficit through an inflation tax. If a country faces problems in financing the fiscal and balance of payments deficits, capital controls may have an the objective of reducing the domestic debt-servicing costs and preserving the domestic inflation base by keeping domestic interest rates low (Drazen (1989)). The role of capital controls in debt financing, however, is likely to be limited, partly because, by depressing domestic interest rates, capital controls penalize investors in domestic assets and discourage domestic saving.

A somewhat more sophisticated version of the rationale for capital controls concerns their use to help maintain or achieve a degree of monetary and exchange rate policy autonomy. In essence, capital controls are used in an attempt to reconcile the use of interest rates and the exchange rates to pursue, simultaneously, at least partially inconsistent internal and external balance objectives. Generally, outflow controls seek to avoid nominal currency depreciation pressures without the tightening of monetary conditions or other difficult policy measures otherwise needed. Inflow controls, conversely, are used as a way to minimize nominal exchange rate appreciation pressures in the face of substantial capital inflows, without sacrificing control over domestic monetary conditions. Policy autonomy, however, often comes at the expense of policy discipline.

Capital controls are sometimes justified for macroeconomic reasons by asymmetric information problems and herding behavior in capital markets (e.g., see Tobin (1978), Dornbusch (1986), Tornell (1990), Mishkin (1996), and Eichengreen and Mussa (1998)), and the need to reduce the volatility of short-term capital flows and the associated volatility of exchange rates.

Capital controls have also been linked to the objective of protecting a fixed exchange rate regime. If exchange rates are pegged, free short-term movements of capital could lead to large fluctuations in international reserves, interest rates, and even a collapse of a fixed exchange rate regime. Capital controls may serve as a short-term line of defense (see for example, Krugman (1979), Flood and Garber (1984), Grilli (1986), and Obstfeld (1986)).

Capital controls may also be motivated by the overall framework for macroeconomic and balance of payments management in financially repressed economies (Johnston and Sundararajan 1999). Such economies are generally characterized by widespread restrictions on interest rates, high reserve requirements, credit ceilings, government directed credit, weak financial sector competition, and capital controls.

B. Market and Institutional Evolution

A second rationale for capital controls is linked to the insufficient level of development of domestic financial markets and institutions. Controls on inflows are sometimes justified by the need to protect infant industries and less developed financial markets. Two different versions of the infant industry argument can be found in the trade literature. One version states that, when protected by a tariff, infant industries will gain economies of scale and thus achieve a higher level of output. The second version notes that domestic industry will be able to lower its production costs thanks to learning by doing. Both versions try to justify temporary tariffs. However, the existence of economies of scale or learning effects is not sufficient to justify protection, since the initial high cost of production could be borne by the firm as a kind of investment financed via the capital market. A case for infant industry protection could be made when the capital market does not work efficiently. Direct subsidization, development of the capital market or government loans to the industry would be superior because such measures would deal directly with the original distortion and avoid the adverse effects of trade protection on consumption and other domestic industries. Besides, there is a risk that temporary protection would become permanent. Similar issues arise in the application of the infant industry argument to financial markets. By restricting the introduction of new financial products and limiting competition in the domestic financial system, capital controls may hinder rather than facilitate the development of financial markets and institutions.

More generally, the infant industry argument for trade protection belongs to the category of the second-best solutions in welfare economics. The basic idea of the second best is simply that if an economy is assumed to suffer from one distortion and it is difficult to tackle this distortion directly, it might be possible to increase welfare by adding another, off-setting distortion (see Dooley (1996)). The first-best solution, of course, would be to address the underlying distortion. In emerging countries sources of distortions may include information problems, weak or insolvent banking systems, moral hazard problems related to official guarantees and the absence of developed financial markets.6

Capital mobility reduces the effectiveness of direct credit and interest rate controls maintained for monetary policy reasons because of the increased scope for the circumvention of such controls through capital flows (see Johnston (1998)). Thus, a country might seek to limit capital movements when its indirect monetary control instruments are nascent and time is needed to develop the instruments, markets, and institutional capacity to rely on indirect measures. Such instruments can, nevertheless, often be introduced quickly through “open market-type” operations (see Johnston and Sundararajan (1999)).

C. Prudential Reasons

A third general motivation for the use of capital controls is their role as prudential measures. It is generally recognized that international transactions may involve different types of risk from those affecting comparable domestic transactions. These different types of risk include, inter alia, transfer risk, sovereign risk, and country risk. Differential requirements for the listing and trading of foreign securities in domestic markets could also be justified because of different supervisory and accounting standards, or greater difficulties in enforcement in the context of different national jurisdictions (see Johnston (1998)).

The use of capital controls is also justified by the need to preserve systemic stability. To the extent that capital controls help limit excessive foreign exchange exposure of domestic institutions, or help lengthen the maturity of liabilities of financial institutions, they could help protect the stability of the financial system. However, the prudential role of capital controls may be limited: by preventing portfolio diversification, capital controls tend to increase investment risk, and, by slowing down the development of financial markets, they reduce liquidity and hence the quality of financial assets. Capital controls are also ill-designed for addressing specific financial risks, and the circumvention of the controls may result in the channeling of transactions through the instruments or institutions that are more risky and less well regulated, thereby increasing systemic risks.

Imposing capital controls is often a less efficient and less effective way of controlling financial risks than an oversight of the internal capacity of supervised institutions to manage risk and greater public disclosure of information; such prudential measures would have little, if any, restrictive impact on capital movements. Nevertheless, in the countries that are in the process of developing their prudential and supervisory arrangements, and where there is a limited capacity to adequately design, implement, or enforce prudential measures, the authorities may need to rely on capital controls for prudential reasons. An example would be prohibitions imposed by countries on the issue of securities by nonresidents on their local financial markets because of their limited supervisory capacity or an inadequate regulatory framework for such issues. Restrictions aimed at limiting the volume of capital inflows or changing the composition of such inflows might also be justified where the financial system is unsound and the member needs time to restructure the banking system and to introduce, implement, and strengthen the enforcement of minimal prudential standards.

D. Other Reasons

In addition to the reasons listed above, the use of capital controls could be motivated by a variety of other factors. These include economic size, and openness; the general features of the regulatory system; and social, sectoral, and strategic concerns, particularly as regards controls on inward direct investment. Larger countries, it is argued, have more opportunities for the diversification of investment, and hence have less incentives to open their capital account than smaller countries. The overall openness of the economy may affect the intensity of capital controls: on one hand, more open economies tend to be more prone to external shocks and may introduce exchange and capital controls to mitigate such shocks; on the other hand, there are more opportunities for circumventing capital controls in a more open economy, and, more generally, the liberalization of certain components of the capital account, such as trade finance, is complementary to trade liberalization. Widespread capital controls may also reflect the government’s philosophy about the optimal extent of intervention in the economy in general, weakening public and corporate governance and reducing transparency. Inflows of direct foreign investment to important or sensitive sectors could also be controlled for security, national sovereignty and cultural reasons.

IV. A Simple Empirical Model of Capital Controls

The synthesis of the theoretical explanations of capital controls points to a possible conceptual framework for modeling capital controls. Consistent with the motivations discussed above, capital controls are likely to depend on factors related to balance of payments and macroeconomic management, market and institutional evolution, prudential, and other reasons. Thus, a general single-equation model of capital control of type k could be constructed as follows:

Y k = α k + β BOP X BOP + β MACRO X MACRO + β INST X INST + β PRUD X PRUD + β OTHER X OTHER , ( 1 )

where Yk represents a measure of the intensity of capital control of type k, XBOP is a vector of regressors representing balance of payments factors, XMACRO represents macroeconomic management factors, XINST is market and institutional evolution factors, XPRUD is prudential factors, and XOTHER captures other factors.

The following types of capital control are examined:

  • overall controls on capital movements (denoted by CC);

  • overall controls on capital inflows and outflows (IN and OUT, respectively);

  • controls on inflows and outflows pertaining to capital market, money market and collective investment securities and derivative instruments (ICM and OCM, respectively);

  • controls on inflows and outflows pertaining to commercial and financial credits, and guarantees and sureties (ICR and OCR, respectively);

  • controls on inflows and outflows pertaining to direct foreign investment and real estate transactions (IDFI and ODFI, respectively); and

  • provisions specific to commercial banks and other financial institutions (FIN)

The regressors that are candidates for explaining capital controls include:

  • XBOP: overall balance as a ratio to GDP (denoted by BOP), current account deficit as a ratio to GDP (CURDEF), external borrowing as a ratio to GDP (EXTBOR), and gross international reserves in months of imports (RESIM);

  • XMACRO: government deficit as a ratio to GDP (denoted by GOVDEF), inflation (INF), Eurodollar rate spread (INTEURO), real interest rate (REALINT), an index representing the de facto exchange regime (ERDF), an index representing official exchange regime (EROFF), a dummy variable indicating an exchange rate peg (PEG), average annual change in nominal effective exchange rate (NEER), absolute value of average annual change in nominal effective exchange rate (NEERABS), and average annual change in real effective exchange rate (REER);

  • XINST: bank deposits as a ratio to GDP (denoted by DEPBK), intermediation spread (INT), value of stocks traded as a ratio to GDP (STOCKVAL), dummy variable indicating the existence of a forward market (FORWARD), and a dummy variable indicating the existence of a treasury bill market (TB);

  • XPRUD: a dummy variable indicating the recent problems in the banking sector (denoted by BKPROB); and

  • XOTHER: trade as ratio to GDP (EXIM), an index of economic freedom in banking (FRBK), an index of economic freedom in tax policy (FRTAX), an index of economic freedom in domestic regulation (FRREG), gross domestic product (GDP), GDP per capita (GDPPC), and a dummy variable indicating oil-producing countries (OIL).

For each type of capital control, a parsimonious model is developed through general-to-specific modeling. The modeling of a given type of capital control starts from a general, unrestricted multivariate regression model incorporating a multitude of proxies for each group of factors. The general model is estimated by the ordinary-least-squares method (OLS)7 and subjected to an array of diagnostic tests. The model is sequentially respecified until diagnostic tests point to a satisfactory, parsimonious model. The diagnostic tests include F and t tests for omitted and irrelevant variables, Chow tests for the constancy of parameters and the model as a whole (including 1-step, break-point, and forecast F-tests), and Durbin-Watson test. If a diagnostic is significant, it is interpreted as indicating misspecification.

Next we model capital controls as systems of equations to reflect the simultaneity of decision-making about various types of capital control. Three systems are considered:

  • System I, with equations for overall capital controls (CC and financial controls (FIN);

  • System II, with equations for overall controls on inflows and controls on outflows (IN and OUT, respectively); and

  • System III, with equations for controls on inflows and controls on outflows related individually to capital and money markets, credit operations, and direct foreign investment (ICM, ICR, IDFI, OCM, OCR, ODFI respectively).

The systems of equations are estimated using full-information maximum-likelihood methods (FIML). Starting with a general system, we sequentially impose zero restrictions on coefficients. Testing the validity of restrictions points to a parsimonious, encompassing model of capital controls.

The estimation of the above models requires data on various proxies of balance of payments, macroeconomic, institutional and other factors that could explain the use of capital controls. Data and notation for the explanatory variables are described in Appendix II. Summary statistics are presented in Table 5. The explanatory variables characterized by relatively large short-term fluctuations (mainly macroeconomic variables) are averaged. For the dependent and explanatory variables related to institutions and regulations that tend to change over medium to long term, averaging is not critical. However, as discussed below, the results need to be interpreted cautiously because of the potential endogenous relationship between the use of controls and some of the explanatory variables. A simple check for such feedback effects is performed by examining capital controls as part of a simple simultaneous equation framework, including equations for the balance of payments, macroeconomic variables and capital controls.

Table 5a.

Descriptive Statistics

article image
Table 5b.

Correlations

article image