I. Introduction

Over the last two decades several developing economies have liberalized their domestic financial markets and eliminated current and capital controls and restrictions, though to various extent and at different speed. Starting with the work of McKinnon (1973) and Shaw (1973), and proceeding with that, among others, of King and Levine (1993), it is broadly accepted that financial liberalization, by fostering financial development, can increase economic performance.² However, the banking crises of the 1980s and 1990s have pointed out the link between financial liberalization and financial fragility and the existence of a possible trade-off between the benefits of liberalization and the costs of increasing financial fragility, especially in developing markets.³ On these grounds, some have pointed out that, although there are clear benefits from moving away from a repressed financial system, it might not be optimal to pursue quick and extensive financial deregulation, especially at an early stage in the liberalization process when market institutions have not yet fully adjusted.⁴

The Asian crisis has, moreover, underscored the importance of the soundness of the domestic financial system in achieving the benefits and avoiding the risks of liberalization, and raised questions about the welfare-related implications of capital account liberalization and whether in some cases capital controls, in particular over short-term capital flows, could be appropriate.⁵ In this context, since capital controls may also be easily misused, some have taken the view that it may be necessary to draw up a list of all controls and restrictions that could be compatible with free participation in financial relations and orderly market conditions.

Responding to these developments, some international institutions, such as the Bank for International Settlements and the International Accounting Standards Committee, have been developing a set of guidelines to strengthen regulatory, supervisory, and accounting practices, while others, namely the International Monetary Fund and the World Bank, have stepped up surveillance of the financial sector in individual countries and urged developing economies, in particular, to adopt these guidelines. Given the banks’ central role in the payment system and in the mobilization and distribution of financial resources, effective prudential regulation and supervision of the banking sector are deemed crucial to achieving domestic financial stability and economic development, and, hence, to minimizing the odds of an abrupt turnaround in capital flows.

Earlier literature has examined the relationship between financial sector reforms, banking problems, and economic performance. Johnston and Pazarbasioglu (1995) examine the impact of financial sector variables on economic growth and efficiency using panel data for 40 countries distinguishing between those that did or did not face a banking crisis. They find evidence that countries that reform their financial systems and avoid financial crises enjoy higher average rates of economic growth; however, countries that liberalized and faced a banking crisis experienced a deterioration in economic performance following the crisis. Demirgüç-Kunt and Detragiache (1998b) study the connection between financial liberalization and financial fragility empirically on a sample of 53 countries during the period 1980–95. Their main finding is that financial liberalization increases the probability of a banking crisis, though less so if legal institutions and governance are strong. Barth, Caprio and Levine (1999) analyze the relationship between financial regulation, financial fragility, and economic performance, and find that countries with more restrictive regulatory systems have higher probability of suffering a banking crisis. Other recent studies have looked at the relationship between financial liberalization and fragility and underscored the importance of improving, in the context of capital account liberalization, the soundness of the financial system, including through stronger prudential regulation and supervision.⁶ Johnston (1998) points out that the adoption of prudential regulations based on generally accepted best practices will not normally entail restrictions on capital flows and will support the move toward capital account convertibility.

Despite the general agreement, which does not necessarily encompass policy responses, on the inherently intertwined nature of financial and capital account liberalization, prudential regulation and supervision, financial fragility and economic performance, very little empirical cross-country investigation exists that could provide evidence on these complex interrelationships and, thus, be a first step toward the formulation of a blueprint of financial and capital account liberalization. The main reason for the lack of such empirical studies seems to be the unavailability of adequate measures to describe the regulatory structure of both the capital account and the domestic financial system.

This paper is a first, albeit preliminary, effort to fill these gaps. On the one hand, it develops and uses new measures of the pervasiveness of capital account controls, prudential regulation and supervision, and the degree of depositors’ safety for each of the sample countries. On the other hand, it provides a cross-country empirical analysis of the basic correlations among capital account controls, the banking regulatory and supervisory framework, financial fragility, and economic performance.

The paper is organized as follows. Section 2 describes the data set, whereas Section 3 presents some stylized facts emerging from the panel with a focus, in particular, on capital account controls, prudential regulation, supervision, and depositors’ safety. Sections 4 and 5 report the empirical analysis; Section 6 concludes. The Appendix details the methodology used to compute the new measures of capital controls, prudential regulation, supervision, and depositors’ safety.

II. The Panel Data

Although dummies to describe the openness of the capital account have been used before in the empirical literature,⁷ they usually take a value of zero or one according to whether capital transactions in general are free or restricted, respectively. In fact, the capital account is neither open nor closed altogether, and the usual dummy variable for capital controls is too coarse a measure to reflect accurately the pervasiveness of capital controls and the structure of the capital account (controls on inflows or outflows, on short- or long-term flows, on direct or portfolio flows, and so on).⁸ With regard to prudential regulation and supervision, indices that account for the extent to which various regulatory and supervisory practices in different countries accommodate international best standards are not currently available to our knowledge.⁹

The main challenge in compiling the data set was clearly the development of new indices of capital account controls, prudential regulation, supervision, and depositors’ safety that would not simply take 0 or 1 values.¹⁰ Inevitably, both the country and time coverage had to be limited; the focus is, therefore, on 15 developing economies over the period 1990–97. Countries included in the sample are Argentina, Brazil, Chile, Colombia, India, Indonesia, Israel, Korea, Malaysia, Mexico, Peru, Philippines, South Africa, Thailand, and Venezuela.¹¹

The indices of prudential regulation, supervision, and depositors’ safety try to capture how different countries’ regulatory and supervisory frameworks fare in terms of internationally accepted guidelines.¹² These guidelines specify that a sound regulatory and supervisory framework should reflect: satisfactory capital adequacy requirements and definition of capital; adequate loan classification and provisions; suitable maximum exposure limits; effective enforcement of regulations; clear responsibilities and objectives for each agency involved in banking supervision; strong powers to review and reject specific activities that do not meet the required standards; proper disclosure and accounting requirements; and an adequate system of market incentives to complement regulatory and supervisory requirements.

The task involved gathering, analyzing and integrating relevant and homogenous information from many sources, and, then, translating it into synthetic indices. Grading scales were developed to include the main features of internationally accepted regulatory and supervisory guidelines, and, then, used to compute the indices for each individual country. The three indices of prudential regulation, supervision, and depositors’ safety vary between 1 and 4; the first two increase in tightness, whereas the third in safety.

The indices of capital controls on inflows and outflows are calculated using, as a starting point, those developed in Johnston and others (1999a). They range between 0 and 1 increasing in the number of controls, and are computed as the actual number of restrictions applying to a particular capital flow (inflow or outflow) divided by the total number of possible restrictions on each category of capital flows. After being slightly modified to account for an alternative classification of some of the items in the capital account, the 1997 indices are, then, back casted to 1990 by developing an algorithm that mimics the main episodes in the process of capital account liberalization in our sample countries. The information about the process of capital account liberalization is contained in Johnston and others (1999b). As necessary, the relevant issues of the Annual Report on Exchange Arrangements and Exchange Restrictions are also used.

The entire data base comprises 27 qualitative and quantitative variables. The latter are mainly control factors that have been shown to provide consistent and significant information about financial fragility [Kaminsky and Reinhart (1996), Demirgüç-Kunt and Detragiache (1998a)], currency vulnerability [Kaminsky and Reinhart (1996), Frankel and Rose (1996)], and economic development [King and Levine (1992), and (1993)]. Table 1 lists all the variables by name, definition, and source.

Table 1.

Data Base

^1/Such a measure of the real interest rate can be interpreted as a proxy for the process of financial liberalization since real rates are usually lower, or negative, in repressed financial systems. Johnston and Pazarbasioglu (1995), and Fry (1997) use the real interest rate as a proxy for financial liberalization. Demirgüç-Kunt and Detragiache (1998b) create a financial liberalization dummy that takes 1 on the first year in which some interest rates are liberalized and on any subsequent year if no policy reversal occurred; 0 otherwise. RINT seems, however, a preferable indicator of the process of financial liberalization. First, we might argue that RINT is a measure of the de facto liberalization process, whereas the financial liberalization dummy is rather a de jure indication of whether the process has already started, and, thus, measures whether a financial system is “liberalized” rather than the process of liberalization itself. Second, the financial liberalization dummy seems, by construction, to be too rough a measure to pin down effectively the effect of the liberalization process. Regression results using a financial liberalization dummy based mainly on the information contained in Table 1 in Demirgüç-Kunt and Detragiache (1998b) are not reported as this dummy appears to have negligible explanatory power.

^2/King and Levine (1992) suggest several measures of financial development. In this paper we consider two of them: gross claims on private sector as a share of GDP (DOMCR) and the deposit money bank domestic credit as a share of GDP (DMBCRGDP). These measures are somehow similar in that they try to assess whether the recipients of loans are principally private or public institutions. Data to compute additional measures of financial development were not available for all countries over the sample period.

Table 1.

Data Base

Name	Definition	Source
Qualitative Variables
CURR	(0, 1) currency crises dummy: 1 indicates a crisis. For the methodology used to select crises dates, please refer to the source.	Aziz, Caramazza, and Salgado (1998).
BANK	(0, 1) banking crises dummy: 1 indicates a crisis. Banking crises are defined as “cases where there were runs or other substantial portfolio shifts, collapses of financial firms, or massive government intervention. Extensive unsoundness short of crisis is termed significant”. Both definitions are considered in compiling the dummy variable.	Lindgren, Garcia, and Saal (1996).
ARTVIII	(0, 1) current account restrictions dummy: 0 if a country has accepted the obligations of Article VIII of the Fund’s Articles of Agreement, 1 otherwise.	Annual Report on Exchange Arrangements and Exchange Restrictions.
KINF	Index of controls on capital inflows ranging between 0 and 1.	Developed in Johnston and others (1999a), and extended in this paper.
KOUT	Index of controls on capital inflows ranging between 0 and 1.	Developed in Johnston and others (1999a), and extended in this paper.
PRUD	Index of prudential regulation ranging between 1 and 4.	Developed in this paper.
SUP	Index of supervision ranging between 1 and 4.	Developed in this paper.
DEPSAF	Index of depositors’ safety ranging between 1 and 4.	Developed in this paper.
LAW	Index of law and order ranging between 0 and 6, and increasing in the quality of the institution.	International Country Risk Guide, published by Political Risk Service, Syracuse, NY.
CORR	Index of corruption ranging between 0 and 6, and increasing in the quality of the institution.	International Country Risk Guide, published by Political Risk Service, Syracuse, NY.
CONENF	Index of contract enforcement ranging between 0 and 4, and increasing in the quality of the institution.	Business Environmental Risk Intelligence, Washington, DC.
BURQUA	Index of bureaucratic quality ranging between 0 and 4, and increasing in the quality of the institution.	Business Environmental Risk Intelligence, Washington, DC.
Quantitative Variables
GDPCAP	Real GDP per capita in U.S. dollars. National currency- denominated real GDP (1990 prices) was converted into common international units (US$) by multiplying it by the 1990 US$ per local currency exchange rate.	International Financial Statistics (IFS).
GDPCAPGR	Real GDP per capita growth.	IFS.
REALGR	Real GDP growth as annual change of 1990 GDP.	IFS.
RINT 1/	Real interest rate, calculated as r=[(1+i)/(1+π)]-1. Because of the unavailability for all sample countries of interest rates on short-term government paper, we use nominal interest rates on banks’ deposits.	IFS.
INF	Inflation rate as annual change of the GDP deflator (1990 base).	IFS.
TOT	Change in the terms of trade.	World Economic Outlook (WEO).
INVGDP	Investment-to-GDP ratio	WEO.
GOVGDP	Government consumption-to-GDP ratio	WEO.
CAGDP	Current account balance-to-GDP ratio	IFS and WEO.
OPENNESS	Imports plus export-to-GDP ratio	WEO.
CREDITPRI	Private sector real credit growth	IFS.
M2RAT	M2-to-reserves ratio	IFS.
CASHRAT	Banks’ liquid reserves-to-assets ratio	IFS.
DOMCR 2/	Domestic private sector credit-to-GDP ratio	IFS and WEO.
DMBCRGDP 2/	Deposit money bank domestic credit-to-GDP ratio	IFS and WEO.

^1/Such a measure of the real interest rate can be interpreted as a proxy for the process of financial liberalization since real rates are usually lower, or negative, in repressed financial systems. Johnston and Pazarbasioglu (1995), and Fry (1997) use the real interest rate as a proxy for financial liberalization. Demirgüç-Kunt and Detragiache (1998b) create a financial liberalization dummy that takes 1 on the first year in which some interest rates are liberalized and on any subsequent year if no policy reversal occurred; 0 otherwise. RINT seems, however, a preferable indicator of the process of financial liberalization. First, we might argue that RINT is a measure of the de facto liberalization process, whereas the financial liberalization dummy is rather a de jure indication of whether the process has already started, and, thus, measures whether a financial system is “liberalized” rather than the process of liberalization itself. Second, the financial liberalization dummy seems, by construction, to be too rough a measure to pin down effectively the effect of the liberalization process. Regression results using a financial liberalization dummy based mainly on the information contained in Table 1 in Demirgüç-Kunt and Detragiache (1998b) are not reported as this dummy appears to have negligible explanatory power.

^2/King and Levine (1992) suggest several measures of financial development. In this paper we consider two of them: gross claims on private sector as a share of GDP (DOMCR) and the deposit money bank domestic credit as a share of GDP (DMBCRGDP). These measures are somehow similar in that they try to assess whether the recipients of loans are principally private or public institutions. Data to compute additional measures of financial development were not available for all countries over the sample period.

View Table

Table 1.

Data Base

Name	Definition	Source
Qualitative Variables
CURR	(0, 1) currency crises dummy: 1 indicates a crisis. For the methodology used to select crises dates, please refer to the source.	Aziz, Caramazza, and Salgado (1998).
BANK	(0, 1) banking crises dummy: 1 indicates a crisis. Banking crises are defined as “cases where there were runs or other substantial portfolio shifts, collapses of financial firms, or massive government intervention. Extensive unsoundness short of crisis is termed significant”. Both definitions are considered in compiling the dummy variable.	Lindgren, Garcia, and Saal (1996).
ARTVIII	(0, 1) current account restrictions dummy: 0 if a country has accepted the obligations of Article VIII of the Fund’s Articles of Agreement, 1 otherwise.	Annual Report on Exchange Arrangements and Exchange Restrictions.
KINF	Index of controls on capital inflows ranging between 0 and 1.	Developed in Johnston and others (1999a), and extended in this paper.
KOUT	Index of controls on capital inflows ranging between 0 and 1.	Developed in Johnston and others (1999a), and extended in this paper.
PRUD	Index of prudential regulation ranging between 1 and 4.	Developed in this paper.
SUP	Index of supervision ranging between 1 and 4.	Developed in this paper.
DEPSAF	Index of depositors’ safety ranging between 1 and 4.	Developed in this paper.
LAW	Index of law and order ranging between 0 and 6, and increasing in the quality of the institution.	International Country Risk Guide, published by Political Risk Service, Syracuse, NY.
CORR	Index of corruption ranging between 0 and 6, and increasing in the quality of the institution.	International Country Risk Guide, published by Political Risk Service, Syracuse, NY.
CONENF	Index of contract enforcement ranging between 0 and 4, and increasing in the quality of the institution.	Business Environmental Risk Intelligence, Washington, DC.
BURQUA	Index of bureaucratic quality ranging between 0 and 4, and increasing in the quality of the institution.	Business Environmental Risk Intelligence, Washington, DC.
Quantitative Variables
GDPCAP	Real GDP per capita in U.S. dollars. National currency- denominated real GDP (1990 prices) was converted into common international units (US$) by multiplying it by the 1990 US$ per local currency exchange rate.	International Financial Statistics (IFS).
GDPCAPGR	Real GDP per capita growth.	IFS.
REALGR	Real GDP growth as annual change of 1990 GDP.	IFS.
RINT 1/	Real interest rate, calculated as r=[(1+i)/(1+π)]-1. Because of the unavailability for all sample countries of interest rates on short-term government paper, we use nominal interest rates on banks’ deposits.	IFS.
INF	Inflation rate as annual change of the GDP deflator (1990 base).	IFS.
TOT	Change in the terms of trade.	World Economic Outlook (WEO).
INVGDP	Investment-to-GDP ratio	WEO.
GOVGDP	Government consumption-to-GDP ratio	WEO.
CAGDP	Current account balance-to-GDP ratio	IFS and WEO.
OPENNESS	Imports plus export-to-GDP ratio	WEO.
CREDITPRI	Private sector real credit growth	IFS.
M2RAT	M2-to-reserves ratio	IFS.
CASHRAT	Banks’ liquid reserves-to-assets ratio	IFS.
DOMCR 2/	Domestic private sector credit-to-GDP ratio	IFS and WEO.
DMBCRGDP 2/	Deposit money bank domestic credit-to-GDP ratio	IFS and WEO.

^1/Such a measure of the real interest rate can be interpreted as a proxy for the process of financial liberalization since real rates are usually lower, or negative, in repressed financial systems. Johnston and Pazarbasioglu (1995), and Fry (1997) use the real interest rate as a proxy for financial liberalization. Demirgüç-Kunt and Detragiache (1998b) create a financial liberalization dummy that takes 1 on the first year in which some interest rates are liberalized and on any subsequent year if no policy reversal occurred; 0 otherwise. RINT seems, however, a preferable indicator of the process of financial liberalization. First, we might argue that RINT is a measure of the de facto liberalization process, whereas the financial liberalization dummy is rather a de jure indication of whether the process has already started, and, thus, measures whether a financial system is “liberalized” rather than the process of liberalization itself. Second, the financial liberalization dummy seems, by construction, to be too rough a measure to pin down effectively the effect of the liberalization process. Regression results using a financial liberalization dummy based mainly on the information contained in Table 1 in Demirgüç-Kunt and Detragiache (1998b) are not reported as this dummy appears to have negligible explanatory power.

^2/King and Levine (1992) suggest several measures of financial development. In this paper we consider two of them: gross claims on private sector as a share of GDP (DOMCR) and the deposit money bank domestic credit as a share of GDP (DMBCRGDP). These measures are somehow similar in that they try to assess whether the recipients of loans are principally private or public institutions. Data to compute additional measures of financial development were not available for all countries over the sample period.

III. Some Stylized Facts about Capital Controls

Over time, countries have resorted to controls on capital movements for different reasons, which stem from distributional and tax efficiency to balance of payment and macroeconomic management considerations.¹³ The 15 countries in the panel are no exception; they also have followed different policies in terms of liberalization of the capital account. Some have liberalized far and fast; others have pursued a gradualist approach; and some others have temporarily reversed the liberalization process when facing external shocks. Most of these countries have simultaneously reinforced their banking regulatory and supervisory framework, though again at an idiosyncratic pace.

A formal empirical analysis of the determinants of capital controls is beyond the scope of this paper.¹⁴ The purpose of this section is to identify some stylized facts about the relationship among capital controls, macroeconomic developments, banking regulation and supervision, depositors’ safety, financial fragility, and economic activity. Table 2 presents some basic correlations. In the following sections, the robustness of the relationship among capital account convertibility, prudential regulation and supervision, depositors’ safety, banking and currency crises, and economic activity is checked by controlling for economic variables that have been used in previous empirical studies of financial crises and economic performance.

Table 2.

Correlations (in descending order, absolute values)

Table 2.

Correlations (in descending order, absolute values)

	KINF		KOUT
KOUT	.70	KINF	.70
SUP	-.50	DOMCR	.42
PRUD	-.40	PRUD	-.39
ARTVIII	.30	SUP	-.36
DEPSAF	.29	GOVGDP	.34
GDPCAP	-.21	OPENNESS	.33
DOMCR	.18	CORR	.28
BANK	.16	ARTVIII	.20
LAW	-.15	BANK	.17
TOT	-.14	DEPSAF	.17
OPENNESS	.14	GDPCAPGR	-.11
CAGDP	.12	TOT	-.10
CORR	.12	CAGDP	.09
GOVGDP	.11	INF	-.08
CURR	-.07	CURR	.04
RINT	.05	GDPCAP	-.03
GDPCAPGR	.02	RINT	-.03
INF	.00	LAW	.01

View Table

Table 2.

Correlations (in descending order, absolute values)

	KINF		KOUT
KOUT	.70	KINF	.70
SUP	-.50	DOMCR	.42
PRUD	-.40	PRUD	-.39
ARTVIII	.30	SUP	-.36
DEPSAF	.29	GOVGDP	.34
GDPCAP	-.21	OPENNESS	.33
DOMCR	.18	CORR	.28
BANK	.16	ARTVIII	.20
LAW	-.15	BANK	.17
TOT	-.14	DEPSAF	.17
OPENNESS	.14	GDPCAPGR	-.11
CAGDP	.12	TOT	-.10
CORR	.12	CAGDP	.09
GOVGDP	.11	INF	-.08
CURR	-.07	CURR	.04
RINT	.05	GDPCAP	-.03
GDPCAPGR	.02	RINT	-.03
INF	.00	LAW	.01

The high positive correlation between KINF and KOUT suggests that controls on both capital inflows and outflows have been eliminated or imposed together. Stronger prudential regulation (PRUD) and more effective supervision (SUP) look relatively highly correlated with fewer controls on capital flows, providing some evidence that some countries have, in the past, used capital controls in pursuit of prudential goals, despite more advanced regulatory and supervisory frameworks being better designed to address the specific risks inherent in both domestic and cross-border financial transactions. In this context, moving toward stronger capital requirements and loan provisioning guidelines, clearer responsibilities and objectives for each agency involved in banking supervision, stronger powers to review and reject specific banking activities that do not meet the required standards, and better disclosure and accounting requirements have each contributed to make controls on capital flows less necessary. By contrast, countries with a larger depositors’ safety net, in the form of higher deposit insurance and/or bail-out expectations, appear more likely to control, in particular, the inflow of funds from abroad.

Higher financial development (DOMCR), a greater need to raise revenue (GOVGDP), and a less corrupt environment (CORR) seem positively associated with controls on capital outflows. Freer current account transactions, either de facto (OPENNESS) or de jure (ARTVIII), having accepted the obligations of Article VIII of the Fund’s Articles of Agreement, look moderately correlated with fewer controls on both inflows and outflows. Finally, sample countries at a lower level of economic development (GDPCAP), and, arguably, with a less efficient tax system, appear to have maintained higher controls on capital inflows.

IV. Financial Fragility

Over the last two decades, many countries have been liberalizing their capital account and domestic financial system. Simultaneously, many countries, irrespective of whether industrialized, developing, or in transition, have experienced significant banking sector problems (Lindgren, Garcia, and Saal (1996)). In addition many countries, from Europe to Latin America and East Asia, have endured turmoil in the foreign exchange market, taking the form of a sharp devaluation (or depreciation), a large reduction in foreign exchange reserves, a brisk hike in interest rates, or a moratorium on servicing their foreign debt (World Economic Outlook (1998)).

Various theoretical linkages have been identified between capital account liberalization and financial fragility. It is argued that financial and capital account liberalization tend to diminish banks’ franchise value by reducing banks’ operating margins and lowering barriers to entry in the banking sector.¹⁵ Liberalization also allows banks to diversify further their investment portfolio into riskier assets and to engage in new types of transactions with both residents and nonresidents. Additional liquidity created by larger capital inflows can result in an expansion of bank loan portfolios; in the absence of strong risk assessment and management procedures, banks could easily increase their financial exposure beyond a sustainable level.

The literature on exchange crises also points out that free capital mobility can exacerbate the likelihood of a currency crisis by limiting the scope for national policy makers to pursue an inconsistent policy mix (first generation models of balance of payment crises), or by reducing the authorities’ willingness (ability), as perceived by the markets, to fend off a speculative attack because of the limited amount of foreign exchanges at their disposal and/or the unattractiveness of very high short-term interest rates (second generation models).¹⁶ Free capital mobility can also contribute to the spreading of currency crises by enhancing the interdependencies among countries and shortening geographical distances, especially if markets are subject to herd behavior.¹⁷

Although specific operational advice on the optimal sequencing of capital account liberalization that could be applied generally to a wide range of countries is hard to formulate, if only because different countries are not necessarily at the same stage of economic and financial development, and do not necessarily share the same institutional and legal structures, it is usually argued that, to achieve its benefits, capital account liberalization should be preceded, or at least accompanied, by the strengthening of prudential regulations and supervision and supported by strong and consistent micro and macroeconomic policies.

The question addressed in this section is whether the increasing freedom of action, in particular with regard to capital account transactions, has significantly increased financial fragility based on the evidence provided by the 15 developing economies in our sample. Also the extent to which the prudential and supervisory framework, together with the perceived depositors’ safety, have contributed to crises is explicitly considered.

A. Banking Crises

We estimate a logit model for banking crises with country fixed effects in which the probability of a banking crisis is a function of a set of control variables and the indices of capital controls, prudential regulation, supervision, and depositors’ safety.¹⁸ Country fixed-effects are included to allow for the likelihood of a banking crisis to vary across sample countries independently of the explanatory variables as, for instance, the structure of the domestic banking system may well intensify or reduce the consequences of exogenous factors. Table 3 shows the results for various specifications.¹⁹

Table 3.

Banking Crises

Note: p-values in parenthesis. Country dummies are included among the explanatory variables. Column (2) includes indices of institutional quality; column (3) includes the lagged value of the currency crisis dummy.

^1/The LR statistic tests the joint null hypothesis that all slope coefficients, but the constant, are zero. It is asymptotically distributed as a χ² variable.

^2/The smallest information criterium indicates the model that best provides a balance between goodness of fit and a parsimonious specification.

Table 3.

Banking Crises

	(1)	(2)	(3)
Control Variables
REALGR	-.95 (.02)	-1.21 (.02)	-1.25 (.02)
RINT	-.42 (.20)	-.70 (.08)	-.66 (.12)
INF	.02 (.37)	.05 (.17)	.05 (.17)
M2RAT	-.07 (.18)	-.10 (.19)	-.11 (.19)
TOT	.04 (.72)	-.01 (.94)	-.02 (.92)
DOMCR	.102 (.21)	.12 (.14)	.13 (.15)
CREDITPRI (t-2)	.06 (.11)	.08 (.04)	.08 (.07)
CASHRAT	.35 (.01)	.48 (.01)	.48 (.01)
GDPCAP	35.9 (.03)	33.74 (.1 1)	34.76 (.11)
CURR (t-1)			.75 (.79)
LAW		1.58 (.31)	1.54 (.31)
CORR		2.79 (.21)	2.73 (.21)
Indices of Capital Controls, Prudential Regulation, Supervision, and Depositors’ Safely
KINF	11.47 (.31)	7.13 (.64)	7.20 (.64)
KOUT	23.81 (.02)	33.84 (.02)	33.14 (.02)
PRUD	-11.72 (.08)	-18.74 (.03)	-19.27 (.03)
SUP	10.89 (.49)	29.35 (.12)	30.16 (.12)
DEPSAF	7.93 (.09)	15.64 (.11)	16.37 (.11)
LR statistic 1/	94.59 (.00)	99.90 (.00)	99.97 (.00)
Hannan-Quinn criterion 2/	1.055	1.063	1.089

Note: p-values in parenthesis. Country dummies are included among the explanatory variables. Column (2) includes indices of institutional quality; column (3) includes the lagged value of the currency crisis dummy.

^1/The LR statistic tests the joint null hypothesis that all slope coefficients, but the constant, are zero. It is asymptotically distributed as a χ² variable.

^2/The smallest information criterium indicates the model that best provides a balance between goodness of fit and a parsimonious specification.

View Table

Table 3.

Banking Crises

	(1)	(2)	(3)
Control Variables
REALGR	-.95 (.02)	-1.21 (.02)	-1.25 (.02)
RINT	-.42 (.20)	-.70 (.08)	-.66 (.12)
INF	.02 (.37)	.05 (.17)	.05 (.17)
M2RAT	-.07 (.18)	-.10 (.19)	-.11 (.19)
TOT	.04 (.72)	-.01 (.94)	-.02 (.92)
DOMCR	.102 (.21)	.12 (.14)	.13 (.15)
CREDITPRI (t-2)	.06 (.11)	.08 (.04)	.08 (.07)
CASHRAT	.35 (.01)	.48 (.01)	.48 (.01)
GDPCAP	35.9 (.03)	33.74 (.1 1)	34.76 (.11)
CURR (t-1)			.75 (.79)
LAW		1.58 (.31)	1.54 (.31)
CORR		2.79 (.21)	2.73 (.21)
Indices of Capital Controls, Prudential Regulation, Supervision, and Depositors’ Safely
KINF	11.47 (.31)	7.13 (.64)	7.20 (.64)
KOUT	23.81 (.02)	33.84 (.02)	33.14 (.02)
PRUD	-11.72 (.08)	-18.74 (.03)	-19.27 (.03)
SUP	10.89 (.49)	29.35 (.12)	30.16 (.12)
DEPSAF	7.93 (.09)	15.64 (.11)	16.37 (.11)
LR statistic 1/	94.59 (.00)	99.90 (.00)	99.97 (.00)
Hannan-Quinn criterion 2/	1.055	1.063	1.089

Note: p-values in parenthesis. Country dummies are included among the explanatory variables. Column (2) includes indices of institutional quality; column (3) includes the lagged value of the currency crisis dummy.

^1/The LR statistic tests the joint null hypothesis that all slope coefficients, but the constant, are zero. It is asymptotically distributed as a χ² variable.

^2/The smallest information criterium indicates the model that best provides a balance between goodness of fit and a parsimonious specification.

Among the macroeconomic variables (REALGR, INF, and TOT) that are likely to affect adversely banking performance via a deteriorating loan portfolio, only real GDP growth is significant across the three different specifications; its positive sign is consistent with the view that banking crises are less likely when economic activity is growing faster.

RINT is the real interest rate on banks’ deposits, and, as such, may well be considered a proxy for the process of financial liberalization rather than a measure of adverse macroeconomic shocks to the banking sector.²⁰ Besides, the average value of RINT over the sample is around one percent, which clearly is not high.²¹ RINT is marginally significant only in one regression, and its negative coefficient is an indication that moving from a repressed to a more liberalized, or restrained, banking system decreases the likelihood of a banking crisis.

Of the variables controlling for specific features of the banking system, that is, liquidity in terms of foreign reserves (M2RAT), exposure to the private sector (DOMCR), liquidity in terms of banks’ assets (CASHRAT), and lagged real credit growth to the private sector (CREDITPRI), only two (CASHRAT, and CREDITPRI) turn out to be statistically significant. The positive sign on the lagged real credit growth to the private sector is consistent with the view that buoyant credit expansion can well contribute to an unsustainable increase in asset prices and poorer-quality loans, hence, aggravating conditions in the banking sector.²² The positive sign on the coefficient for liquidity (CASHRAT) could be explained by the observation that cash ratios tend to be higher in more repressed financial systems. A higher liquid reserves-to-assets ratio may also reflect the fact that banks, if experiencing some difficulties over a certain period of time, may have already reduced their loan portfolios willingly, or been prevented from lending as part of a restructuring agreement, when a banking crisis eventually occurs.

Finally, the remaining control variable (GDPCAP) turns out to be statistically significant only in one specification. The sign of its coefficient suggests that banking crises are more likely in more developed countries, as proxied by the (log of the) level of real GDP per capita. We believe there are at least two good reasons to support this result, though it is apparently in stark contrast with that of other studies, for instance Kunt-Detragiache (1998b). First, the evidence provided in Lindgren, Garcia, and Saal (1996) shows that banking crises have occurred at some point or another across a wide spectrum of countries at different stages in economic development, suggesting that economic development is not a robust criterium to identify countries that are relatively more prone to experiencing disruptions in the banking system. Second, and more important, because of the structure of our sample, which comprises only developing economies, the positive sign of GDPCAP simply indicates that developing countries that are at a higher level of development, and whose financial markets are, hence, larger, are relatively more likely to incur banking crises. Interestingly, our findings, especially if read together with those about the indices of prudential regulation and depositors’ safety, hint at the existence of a threshold (critical mass) beyond which the size of the banking system, directly related to the level of economic development, becomes large enough to enhance financial fragility, should prudential regulation and measures to limit depositors’ safety not be buttressed accordingly.

The statistical insignificance of CURR indicates, as in Kaminsky and Reinhart (1996), that balance-of-payment crises do not help predict banking crises. Also the quality of institutions (LAW and CORR) and the degree of liberalization of capital inflows (KTNF) turn out to be statistically insignificant.

The indices of capital controls on outflows (KOUT), prudential regulation (PRUD), and depositors’ safety (DEPSAF) are all significant irrespective of the specification. The sign of the coefficients suggests that banking crises are more likely in the presence of controls on outflows, of laxer prudential regulation, and higher depositors’ safety. With regard to PRUD and DEPSAF, the results conform fully with the conventional wisdom about the importance of setting not only prudent minimum regulatory requirements ensuring that banks could support the risks that arise in their business, but also appropriate market incentives to foster clients’ monitoring of their banks’ risk-taking behavior. A possible interpretation of the positive sign of KOUT is that banks may try to evade controls by setting up offshore or other types of operations that can eventually expose them to worse risks.²³

Based on these preliminary results, there seems to be a higher probability of a banking crisis in the context of slower economic growth, and more rapid bank expansion associated with weaker prudential regulation, less market discipline and less developed financial systems. Capital account liberalization does not seem to contribute to these crises.

B. Currency Crises

We estimate a logit model of currency crises with time-specific effects in which the probability of a crisis depends on a series of control variables that were found to be relevant in previous empirical analysis,²⁴ and the indices of capital controls, prudential regulation, supervision, and depositors’ safety. Time-fixed effects are included to account for the fact that currency crises tend to be clustered in time. A remarkable feature of recent currency crises, in particular, has been the extent to which instability in foreign exchange markets has been transmitted across countries; an attack on one currency has spilled over or spread contagiously to the currencies of other countries, even, with apparently sound fundamentals. Table 4 reports the results for various specifications.²⁵

Table 4.

Currency Crises

Note: p-values in parenthesis. Time dummies are included among the explanatory variables. Columns indicate different specifications: column (2) includes the one-year lagged banking crisis dummy, and column (3) includes the two-year lagged banking crisis dummy.

Table 4.

Currency Crises

	(1)	(2)	(3)
Control Variables
REALGR	-.32 (.00)	-.33 (.00)	-.44 (.00)
TOT (t-1)	-.08 (.09)	-.11 (.03)	.14 (.01)
CREDITPRI (t-2)	.02 (.38)	.03 (.18)	.07 (.06)
GDPCAP	-.95 (.18)	-.80 (.27)	-.48 (.54)
BANK (t-1)		2.0 (.04)
BANK (t-2)			3.44 (.00)
LAW	.48 (.19)	.71 (.09)	.92 (.06)
CORR	1.10 (.12)	.87 (.24)	-.57 (.47)
Indices of Capital Controls, Prudential Regulation, Supervision, and Depositors’ Safety
KINF	-6.36 (.02)	-7.44 (.01)	-6.81 (.03)
KOUT	1.15 (.58)	.88 (.66)	1.76 (.43)
PRUD	1.19 (.31)	1.66 (.22)	2.13 (.18)
SUP	-1.51 (.17)	-2.10 (.11)	-2.66 (.096)
DEPSAF	1.12 (.08)	1.17 (.07)	1.37 (.06)
LR statistic	39.01 (.00)	43.40 (.00)	49.56 (.00)
Hannan-Quinn criterion	1.045	1.034	.983

Note: p-values in parenthesis. Time dummies are included among the explanatory variables. Columns indicate different specifications: column (2) includes the one-year lagged banking crisis dummy, and column (3) includes the two-year lagged banking crisis dummy.

View Table

Table 4.

Currency Crises

	(1)	(2)	(3)
Control Variables
REALGR	-.32 (.00)	-.33 (.00)	-.44 (.00)
TOT (t-1)	-.08 (.09)	-.11 (.03)	.14 (.01)
CREDITPRI (t-2)	.02 (.38)	.03 (.18)	.07 (.06)
GDPCAP	-.95 (.18)	-.80 (.27)	-.48 (.54)
BANK (t-1)		2.0 (.04)
BANK (t-2)			3.44 (.00)
LAW	.48 (.19)	.71 (.09)	.92 (.06)
CORR	1.10 (.12)	.87 (.24)	-.57 (.47)
Indices of Capital Controls, Prudential Regulation, Supervision, and Depositors’ Safety
KINF	-6.36 (.02)	-7.44 (.01)	-6.81 (.03)
KOUT	1.15 (.58)	.88 (.66)	1.76 (.43)
PRUD	1.19 (.31)	1.66 (.22)	2.13 (.18)
SUP	-1.51 (.17)	-2.10 (.11)	-2.66 (.096)
DEPSAF	1.12 (.08)	1.17 (.07)	1.37 (.06)
LR statistic	39.01 (.00)	43.40 (.00)	49.56 (.00)
Hannan-Quinn criterion	1.045	1.034	.983

Note: p-values in parenthesis. Time dummies are included among the explanatory variables. Columns indicate different specifications: column (2) includes the one-year lagged banking crisis dummy, and column (3) includes the two-year lagged banking crisis dummy.

In conformity with conventional wisdom, higher real growth (REALGR) and improvements in the terms of trade (TOT) tend to reduce the likelihood of a crisis. Also, as in Kaminsky and Reinhart (1996), the emergence of banking problems or the occurrence of a banking crisis (BANK) help predict a balance-of-payment crisis, especially within a two-year period. The quality of institutions (LAW and CORR) and the degree of liberalization of capital outflows (KOUT) are statistically insignificant.²⁶

Countries with a larger number of controls on capital inflows (KINF) tend to be less prone to currency crashes, perhaps, by limiting the amount of funds that could leave a particular country should foreign investors’ sentiments change.²⁷ Finally, there is also some evidence that a stronger banking supervisory framework (SUP) and lower degree of depositors’ safety (DEPSAF) are positive factors influencing the prospect of sharp movements in the exchange rate.

V. Financial Development and Economic Performance

King and Levine (1983) concludes that financial development may well be an important determinant of economic performance. Financial intermediation fosters growth not only by increasing the rate of capital accumulation, but also by enhancing the efficiency with which resources are allocated in the economy. Accordingly, countries that can rely on a well-developed financial system, which effectively allocate credit to the private sector, tend to enjoy more rapid economic growth.

The empirical literature on economic growth is vast.²⁸ It usually focuses on the analysis of a large number of countries (cross-sections) over many years (panels) using ten- or five-year averages of all the variables. The available data set in this paper is clearly too limited to allow an empirical analysis of long-term growth dynamics. It is, nonetheless, adequate to explore the impact of the regulatory structure of the capital account, and of banking regulation and supervision, on the economic performance of sample countries over the business cycle.

The equation specification is that of a standard neoclassical growth model in which the growth rate of real GDP per capita (GDPCAPGR) is regressed on a time trend, the (log) of lagged real GDP per capita (GDPCAP), and other determinants of the level of output per capita. These latter are the investment-to-GDP ratio (INVGDP), the government consumption-to-GDP ratio (GOVGDP), the inflation rate (INF), the alternative measures of financial development (DOMCR and DMBCRGDP), the measure of financial restraint (RINT), the measure of the current account convertibility (ARTVIII), and the index of corruption (CORR). Besides these variables, our indices of capital controls, prudential regulation, supervision, and depositors’ safety are considered.

Table 5 contains the results. The two specifications include both a country-specific (heterogenous) intercept and a country-specific (heterogenous) trend; the difference between them is the consideration of the two different measures of financial development. There seems to be a strong justification for a specification that, ceteris paribus, takes into account the possibility not only that countries may be experiencing a different rate of growth in any particular year (heterogenous intercepts), but also that they may be at a different phase in the business cycle (heterogenous trends).

Table 5.

Determinants of Economic Performance

Note: p-values in parenthesis. Country-specific intercepts and trends are included among the explanatory variables. Columns indicate different specifications whereby two alternative measures of financial development are considered.

Table 5.

Determinants of Economic Performance

	(1)	(2)
Control Variables
GDPCAP (t-1)	-78.68 (.00)	-78.12 (.00)
INVGDP	.68 (.00)	.67 (.00)
GOVGDP	.63 (.00)	.62 (.00)
INF	-.01 (.10)	-.01 (.06)
DOMCR	-.02 (.40)
DMBCRGDP		-.01 (.72)
RINT	.18 (.02)	.20 (.01)
ARTVIII	-2.84 (.00)	-2.81 (.00)
CORR	.18 (.27)	.20 (.24)
Indices of Capital Controls, Prudential Regulation, Supervision, and Depositors’ Safety
KINF	-8.53 (.00)	-8.32 (.00)
KOUT	1.06 (.06)	1.19 (.03)
PRUD	2.29 (.00)	2.27 (.00)
SUP	-2.85 (.12)	-2.68 (.14)
DEPSAF	-3.10 (.00)	-3.16 (.00)
Adjusted R2	.932	.932
No. Of Panel Observations	120	120

Note: p-values in parenthesis. Country-specific intercepts and trends are included among the explanatory variables. Columns indicate different specifications whereby two alternative measures of financial development are considered.

View Table

Table 5.

Determinants of Economic Performance

	(1)	(2)
Control Variables
GDPCAP (t-1)	-78.68 (.00)	-78.12 (.00)
INVGDP	.68 (.00)	.67 (.00)
GOVGDP	.63 (.00)	.62 (.00)
INF	-.01 (.10)	-.01 (.06)
DOMCR	-.02 (.40)
DMBCRGDP		-.01 (.72)
RINT	.18 (.02)	.20 (.01)
ARTVIII	-2.84 (.00)	-2.81 (.00)
CORR	.18 (.27)	.20 (.24)
Indices of Capital Controls, Prudential Regulation, Supervision, and Depositors’ Safety
KINF	-8.53 (.00)	-8.32 (.00)
KOUT	1.06 (.06)	1.19 (.03)
PRUD	2.29 (.00)	2.27 (.00)
SUP	-2.85 (.12)	-2.68 (.14)
DEPSAF	-3.10 (.00)	-3.16 (.00)
Adjusted R2	.932	.932
No. Of Panel Observations	120	120

Note: p-values in parenthesis. Country-specific intercepts and trends are included among the explanatory variables. Columns indicate different specifications whereby two alternative measures of financial development are considered.

The initial level of income per capita (GDPCAP) enters significantly in the regression, supporting the conditional convergence idea. In this case, however, we cannot interpret this (very high) coefficient multiplied by (-1) as the speed of convergence toward the steady state since, as mentioned before, the span of the data set can only allow the analysis of the business cycle. The investment-to-GDP ratio (INVGDP) and the public consumption-to-GDP ratio (GOVGDP) have a positive effect on economic performance, whereas the opposite seems true for inflation (INF).

There is not much evidence of a positive effect of financial development (DOMCR and DMBCRGDP) on economic outcomes over this period. By contrast, moving toward a less repressed financial system (RINT), freer current account transactions (ARTVIII), and higher standards of governance (CORR), enhances economic activity. Finally, countries with banking sector regulations that strengthen prudential guidelines (PRUD) and reduce depositors’ safety (DEPSAF) are more likely to achieve better economic results.

With regard to the impact of capital controls, results are partly—those on inflow controls—consistent with the “classic” view that free capital mobility enhances a more efficient allocation of resources raising welfare in the process.²⁹ Controls on inflows (KINF) seem to hamper economic performance. Fewer controls on inflows would allow capital-scarce developing economies to benefit from the possibility to tap the international capital markets and lower the cost of borrowing, improving investment perspectives and smoothing intertemporal consumption. Controls on outflows (KOUT) appear to contribute, although to a lesser extent, to better economic outcomes even if they might hinder the capability of economic agents, especially financial institutions, to share and spread their risks. It would seem, therefore, that the ability to augment the size of the ‘pot’ of available financial resources within a country, through both increasing the inflows by eliminating controls on them and reducing ‘leakages’ by stepping up controls on outflows, could positively affect the economic performance of a particular country.

VI. Concluding Remarks

Over the last two decades, many countries have actively liberalized both their domestic financial systems and their capital account. At the same time, they have also experienced severe disruptions and outright crises in their banking systems, and foreign exchange markets. As a response to these developments, many have argued in favor of strengthening prudential regulation and supervision of financial systems, and, more recently, in favor of controlling those capital flows that are deemed to be of a more unstable nature. These actions are believed to improve the soundness of existing financial systems and to shield, to a greater extent than recently experienced, countries from a turnaround in (global) market sentiments, thereby reducing financial fragility and enhancing economic performance.

Despite the importance of these issues, very little empirical investigation exists of the links among capital account liberalization, prudential regulation and supervision, financial crises, and economic development, mainly because of the lack of comparable measures to describe regulatory practices for a range of countries. The purpose of this paper was to examine empirically, albeit in a preliminary manner, these links by developing and using new measures of capital controls, prudential regulation, supervision, and depositors’ safety for a sample of 15 developing economies over the period 1990–97.

There are several correlations that emerge from these initial results. Financial fragility seems exacerbated by lenient prudential practices, higher depositors’ safety, and more controls on capital outflows. Banks’ risk-taking behavior could be affected by the imposition of controls on capital outflows as such restrictions may drive banks’ operations offshore or create the incentives for some sort of ‘creative compliance’ that could eventually undermine banks’ own prospects. By contrast, controls on capital inflows appear to reduce the likelihood of a currency crisis, perhaps, by limiting the amount of funds that could leave a particular country should foreign investors’ sentiments change.

Moreover, it seems likely that the size and sophistication of the banking system in developing economies at higher level of economic development may reach a ‘critical mass’ that requires a stronger regulatory and supervisory framework in order to take care of the increasing risks involved in the business. Interesting in this context is the finding that moving away from financial repression may well strengthen financial stability.

The degree of capital account convertibility and the regulatory and supervisory framework look likely to influence economic performance, as well. Again, stronger prudential practices and incentives to limit depositors’ safety, together with a less repressed financial system, seem to allow countries to achieve higher economic activity over the business cycle. With regard to controls on capital account transactions, there is evidence that restrictions on capital movements, especially inflows, might have an impact on economic performance.

Finally, it should be said that all of these findings ought to be interpreted cautiously because of the small size of the panel, the well-known limitations of the estimation procedures, and the endogeneity problems that are likely to exist given the inherently dynamic nature of the issues examined. Further empirical analysis using larger data sets is necessary to check the robustness of the basic correlations reported in this paper.