Can Theory and Practice Justify Controls?

Controversy persists on the role of capital flows in boosting development and inducing macroeconomic instability (Cardoso and Dornbusch, 1989). Capital flows affect consumption, production, and macroeconomic management.

Capital Flows and Consumption

The consumption-smoothing advantage offered by capital inflows arises under two circumstances: in a context of cyclical fluctuations and in a context of growth where foreign savings are used to initiate growth. In the case of cyclical disturbances to the terms of trade, output, or foreign demand, optimal consumption will fluctuate less than disposable income if there is the possibility to borrow during periods of income shortfalls with subse-quent repayment when income recovers. This positive welfare effect of capital flows extends to disturbances that are domestic.

Consumption smoothing can also arise in a growth context. The case of Korea between 1960 and the mid-1980s provides a striking example of a transition toward a high saving rate, financed initially by external borrowing. Between 1960 and 1969, foreign savings equal to 9 percent of GDP financed half of investment. A growing income per capita increasingly provided the resources to finance investment, and by 1986-89 the savings-income ratio had reached 35 percent, the current account had turned toward surplus, and debt started to be retired.

Capital Flows and Production

Capital inflows add to an economy’s productive capacity and thus potentially increase welfare. Foreign investment may carry more than the traditional neoclassical benefits by adding to competition or improving technology. But because of existing distortions, these factors may also lower welfare.

The traditional analysis of foreign investment considers a barter economy where capital inflows (direct foreign investment) take the form of an increase of the economy’s stock of physical capital. The inflow of productive capital raises the economy’s output. In the case of constant returns to scale, the foreign factor earns its marginal product, but also adds to national income, that is, the income of domestic factors of production. This simple analysis facilitates consideration of the choice of optimal borrowing. A country facing a perfectly elastic supply of capital should borrow (rent) capital to the point where the marginal value of capital is equal to the world cost of capital. But if the supply of capital is upward sloping, the increasing marginal cost of capital calls for restriction of capital inflows below the competitive level. This analysis offers a first rationale for the use of taxes or quantitative restrictions on foreign borrowing.

Much of the discussion about the costs and benefits of foreign capital in developing countries involves departures from the simple neoclassical model sketched above and acknowledges that controls are welfare reducing unless they are a “second-best” policy that mitigates the effects of another market failure. Dooley (1995) offers a survey of the modern literature on market distortions and second-best arguments that justify intervention over international capital transactions. The survey reviews the analysis of a wide variety of market failures, including sticky prices in goods and labor markets, distortionary tax policies, anticipated trade reforms, and myopic private speculation. A more recent argument for government intervention in international capital markets is based on the literature on multiple equilib-ria. Special circumstances, such as a fixed exchange rate regime during transition to a monetary union, may justify capital controls to prevent self-fulfilling speculative attacks. In this situation, if multiple equilibria are possible, the “first-best” equilibrium might be achieved through government intervention in capital markets.

Capital Flows and Macroeconomic Management

It is widely recognized that capital flows pose several problems for macroeconomic policy. The most widely cited examples concern the experiences of Latin American countries during the period 1978-82 and in the mid-1990s. During these two periods, a number of countries experienced a strong real appreciation of their currencies, followed by balance of payment crises. Economists have interpreted the real appreciation in two distinct ways. Harberger (1986), for instance, highlights the sudden abundance of foreign borrowing and the resulting pressure of capital inflows on the real exchange rate. In this view, capital flows lead to real appreciation, and thus bring about an inward transfer of resources.

The alternative explanation notes that in all cases of real appreciation in Latin American countries in the early 1980s and mid-1990s, the monetary authorities followed a conscious policy of using reduced rates of exchange depreciation (or even fixing the exchange rate to achieve disinflation, as in Chile in 1979-81). The combination of expected reduced depreciation with high domestic interest rates in relation to interest rates in the United States attracted capital inflows. The real appreciation in turn led to current account deterioration. In the end, each of these experiences of the real appreciation turned out to be very costly, as illustrated by Mexico’s 1994 crisis.

The situation has also been complicated by relatively high domestic rates that have induced banks to incur open foreign exchange positions by financing local currency lending with foreign currency borrowing. Even when rules limit their foreign currency positions, banks still become indirectly exposed to the risk of devaluation. When use of the exchange rate as a nominal anchor leads to relatively high interest rates, combined with little immediate prospect of devaluation, enterprises are encouraged to take up foreign currency–denominated loans. In cases where the borrowers’ revenues are mostly denominated in the domestic currency, the quality of foreign currency loans can also deteriorate in the event of a domestic currency devaluation.

The desire to counteract the pressures to exchange rate appreciation in the face of large capital inflows and to limit inflows that are likely to be reversed has led to central bank intervention. Policies to reduce the impact of capital inflows include direct intervention through controls and taxes and a restrictive monetary policy in the form of sterilization. Sterilization can create significant fiscal costs in financing high levels of reserve holdings depending on the scale of the operation and the size of the interest differential vis-à-vis external rates in reserve centers. The instability caused by heavy inflows and the costs of sterilization seems to give governments a reason to control capital flows.

Perhaps one of the most convincing arguments in favor of the use of capital controls was advanced by Dooley (1996). He argues that large private capital inflows to developing countries have reflected a chain of official guarantees consisting of a commitment to an open capital account, the adoption of a fixed exchange rate (or limited flexibility), and the guarantee that the authorities will help stabilize the domestic financial system during a crisis. The financial system guarantees include a lender-of-last-resort provision, bank deposit insurance, and interventions in equity markets to limit price declines. Given the incentives created by these guarantees, the size of the capital inflow will be related to the country’s perceived net worth (the value of its net international reserves, the credit lines it can obtain from private markets, and the resources that are likely to be available from international financial institutions). If the guarantees lead inflows to a poorly supervised financial system, poor quality investments may occur. The solution to this problem lies in breaking the chain of guarantees offered to international investors. Dooley regards a threat to withdraw the guarantee of the bank deposits or the solvency of the banking system as not credible. This leaves either changing the exchange rate regime or imposing capital controls as the only options, if countries do not want domestic interest rates to be determined by international markets.

Costs of Capital Controls

Controls on capital flows take the form of restrictions on the assets transactions or restrictions on payments related to the acquisition of assets, Restrictions on assets transactions include direct capital controls, such as quantitative limits or prohibition of certain transactions by imposing minimum maturity limits. Price-based capital controls take the form of taxes or reserve requirements.

Recent experiments with controls on capital outflows and inflows have covered a wide variety of instruments. In response to the mid-1990s capital outflows, Venezuela introduced comprehensive exchange controls to limit current and capital account transactions. Romania responded to its balance of payments crisis of early 1996 by effectively closing foreign exchange markets. South Africa postponed the elimination of remaining exchange controls on residents’ capital outflows following a run on the rand in early 1996. In response to Mexico’s peso crisis in late 1994, Brazil prohibited prepayment of foreign loans and relaxed certain capital inflow controls.

Examples of direct controls by countries that experienced recent surges in capital inflows include, among others, Brazil’s prohibition of some nonresident transactions (inflows to futures and options markets) in 1995, and Chile’s one-year minimum maintenance period for nonresident capital inflows. These countries also used price-based controls. For instance, Brazil raised the financial transaction tax to discourage inflows in me 1990s. Chile introduced a stamp duty in mid-1990 and extended the tax base to all foreign loans.

Financial regulatory measures and prudential measures can also affect capital movements. China, India, Korea, and Thailand differentiate their reserve requirements between resident and nonresident deposits in a way that can influence capital movements in some cases. Prudential regulations applied for the purposes of controlling banks’ open net foreign currency position may include a capital control element. Brazil, responding to capital outflow surges in 1995. raised banks’ short position limit and lowered their long position limit.

Alongside arguments that justify the use of capital controls, a strong tradition argues that government intervention does not accomplish its stated objectives. There is a question whether the costs and distortions generated by controls outweigh the potential benefits. These costs include the possibility of retaliation by other countries, evasion, administrative costs, and the inability to quantify the needed tax on capital flow. There is also the risk that controls established to mitigate a temporary distortion may generate interests of their own and outlive their purpose.

Does Practice Justify Controls?

Whether controls are welfare improving or welfare reducing is an empirical question. The empirical evidence on the effectiveness of controls has concentrated on the effect on interest differentials. Essentially, capital controls permit a breach between international and domestic interest rates even when expected devaluation and risk premium are factored in. Dooley (1995) surveys the empirical evidence on industrialized and developing economies and concludes that controls have influenced yield differentials across countries, although there is no evidence that controls have helped governments achieve policy objectives, such as avoiding real appreciation, or that controls have enhanced welfare as suggested by theory.

Data on capital controls are scarce and few empirical papers introduce them directly. Most papers use the International Monetary Fund’s Annual Report on Exchange Arrangements and Exchange Restrictions as the source of capital control data. Johnston and Ryan (1994) and Grilli and Milesi-Ferretti (1995) use panel data for industrialized and developing countries. Both papers find that the data do not support the hypothesis that control programs affect economic variables, such as the volume and composition of private flows, changes in international reserves, or the level of the exchange rate. Grilli and Milesi-Ferretti also find that capital controls are associated with higher inflation and lower real interest rates.

Chile is generally cited as an example of the effective use of capital controls, but Valdés-Prieto and Soto (1996) find mixed results. The econometric evidence between 1987 and 1995 shows that capital controls were not evaded in Chile, where substantial levels of tax revenue were levied on capital market participants. As a matter of fact, the ability to collect tax revenue on capital flows increased over time as the Chilean authorities closed loopholes and the selective capital controls have discouraged significantly particular classes of short-term credits. The results show that the taxed short-term flows were smoothly substituted by other short-term flows without measurable changes in total short-term credits. The taxes were borne by participants who were unable to substitute flows. The authors also find that selective capital controls have failed to achieve other objectives of the Chilean monetary authorities, such as delaying real exchange rate appreciation or improving the mixture of foreign financing between long- and short-term credits.

In the case of Colombia, Cárdenas and Barrera (1997) find a relative inability of controls to reduce the level of capital inflows, but suggest that nonremunerated deposits have been successful in inducing a recomposition of foreign liabilities in favor of long-term maturities.

Reinhart and Smith’s (1996) findings are consistent with the results mentioned above. After analyzing stylized facts of several recent episodes in Asia, Eastern Europe, and Latin America, they agree that capital controls had little effect on consumption, the current account, or the real exchange rate, but that in most cases the measures were capable of either reducing the overall volume of inflows, or in some cases even altering their maturity profile over the short run.

In summary, the evidence seems to be that capital controls can provide temporary breathing room for dealing with balance of payments difficulties and help to reverse capital outflows if combined with policy tightening involving higher interest rates. Controls can also serve to discourage potentially destabilizing short-term capital flows and reduce a country’s vulnerability to shifts in market sentiment. But it seems ineffective in preventing sustained outflows of savings or avoiding a crisis induced by inconsistent macroeconomic policies. Enforcing capital controls over extended periods can reveal itself as a hopeless task in a world of highly integrated international capital markets. Section II will now investigate whether these conclusions also apply to the Brazilian experience.

Capital Flows to Brazil

After the oil shock of 1973, Brazil’s reliance on commercial loans to finance both public investment and the more expensive oil led the country to the debt crisis of the early 1980s. Following a trend common to other emerging markets, private capital inflows ¹ to Brazil disappeared in the 1980s and increased dramatically after 1991. By 1993, the fall of international interest rates had eased the external debt burden and led to an agreement with creditor banks that was concluded in April 1994 with an exchange of instruments that covered over $50 billion in debt stocks and arrears.

Monthly private net capital flows averaging $39 million between 1988 and 1991 increased 25 times, and turned into an average monthly net flow of $970 million between 1992 and 1995. Since 1992, net foreign capital flows to Brazil have been sufficient to finance small current account deficits while contributing to an increase in foreign reserves (Cardoso, 1997). During this period, the capital consisted primarily of short-term resources tied to portfolio investments and other short-term investments. In 1995, for example, net capital flows amounted to more than $29 billion, of which $20 billion was short-run capital: $2.3 billion was equity and special investment funds, and approximately $18 billion consisted of short-run capital not classified under a specific category (Table 1).

Figure 1 shows the composition of capital flows. It illustrates the declining share of medium- and long-term capital flows (lines d and e in Table 1) and the growing importance of short-term capital (lines c and f) in total private capital flows. The share of net direct investment (including reinvested profits, lines a and b in Table 1) in total private capital flows oscillated between 1991 and 1995. Net direct investment, as a share of private capital flows, declined from 22 percent in 1992 to negative 5 percent in 1993, increasing to 11 percent in 1994, and falling to 9 percent in 1995.

Table 1.

Composition of Capital Flows, 1991-95

(In millions of U.S. dollars)

Sources: Central Bank of Brazil, Boletim do Banco Central; International Monetary Fund, International Financial Statistics, and Fund staff calculations. See Cardoso (1997).

Table 1.

Composition of Capital Flows, 1991-95

(In millions of U.S. dollars)

	1991	1992	1993	1994	1995
(a) Net direct investment	-408	1,268	-481	852	2,376
(b) Reinvested profits	365	175	100	83	200
(c) Equity securities and other	578	1,704	6,651	7,280	2,294
funds
(d) Debt securities and loans	2,368	5,761	5,866	3,713	9,113
(e) International organizations,	-4,131	-3,425	-2,909	-1,908	-2,227
and government agencies
(f) Short-term capital and	-2,901	1,033	-1,623	-2,054	17,554
others
(g) Financial account in the IFS=	-4,129	6,516	7,604	7,965	29,310
(a) + (b) + (c) + (d) + (e) + (f)
(h) Arrears, other short-term	-19	18.755	2,511	6,329	510
liabilities and exceptional
financing
(i) Capital account in	4,148	25,271	10,115	14,294	29,820
Boletim do Banco Central=
(g) + (h)

Sources: Central Bank of Brazil, Boletim do Banco Central; International Monetary Fund, International Financial Statistics, and Fund staff calculations. See Cardoso (1997).

View Table

Table 1.

Composition of Capital Flows, 1991-95

(In millions of U.S. dollars)

	1991	1992	1993	1994	1995
(a) Net direct investment	-408	1,268	-481	852	2,376
(b) Reinvested profits	365	175	100	83	200
(c) Equity securities and other	578	1,704	6,651	7,280	2,294
funds
(d) Debt securities and loans	2,368	5,761	5,866	3,713	9,113
(e) International organizations,	-4,131	-3,425	-2,909	-1,908	-2,227
and government agencies
(f) Short-term capital and	-2,901	1,033	-1,623	-2,054	17,554
others
(g) Financial account in the IFS=	-4,129	6,516	7,604	7,965	29,310
(a) + (b) + (c) + (d) + (e) + (f)
(h) Arrears, other short-term	-19	18.755	2,511	6,329	510
liabilities and exceptional
financing
(i) Capital account in	4,148	25,271	10,115	14,294	29,820
Boletim do Banco Central=
(g) + (h)

Sources: Central Bank of Brazil, Boletim do Banco Central; International Monetary Fund, International Financial Statistics, and Fund staff calculations. See Cardoso (1997).

At the end of 1994, Mexico’s financial crisis led to an immediate cutback in capital flows to emerging markets. During the fourth quarter of 1994 and the first quarter of 1995, the net flow of capital to Brazil was insufficient to finance the current account deficit, and the central bank lost reserves of about $9.8 billion. When the crisis erupted, the initial reaction of investors suggested that the Mexican financial crisis would compromise all emerging markets, as stock prices plunged, particularly in Argentina and Brazil; currencies weakened in developing countries from Thailand to Bulgaria; and foreign portfolio investment disappeared. The IMF joined the United States in a rescue operation under which the United States committed $20 billion from its Exchange Stabilization Fund and the IMF pledged $17,8 billion to support Mexican reforms. This infusion of capital successfully insulated financial markets from the crisis and soon capital also returned to Brazil. At the end of 1995, net capital flows were close to $30 billion, and in 1996 net flows again exceeded $29 billion. In 1996, a boom in mergers and acquisitions led to an increase in foreign direct investment, which amounted to $8 billion, while the sum of equity investment and short-term capital fell from $20 billion to approximately $17 billion.

Composition of Capital Flows

(Millions of U.S. dollars)

Citation: IMF Staff Papers 1998, 004; 10.5089/9781451974515.024.A006

Source: Banco Central do Brasil.

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Composition of Capital Flows

(Millions of U.S. dollars)

Citation: IMF Staff Papers 1998, 004; 10.5089/9781451974515.024.A006

Source: Banco Central do Brasil.

• Download Figure
• Download figure as PowerPoint slide

Composition of Capital Flows

(Millions of U.S. dollars)

Citation: IMF Staff Papers 1998, 004; 10.5089/9781451974515.024.A006

Source: Banco Central do Brasil.

• Download Figure
• Download figure as PowerPoint slide

Determinants of Capital Flows

Calvo, Leiderman, and Reinhart (1996) divide the factors that encourage or inhibit capital flows into external and internal factors. The most important external factor is found to be world interest rates. Short-term interest rates in the United States declined steadily in the early 1990s and the recessions in the United States and Japan made profit opportunities in developing countries more attractive. Agénor, Hoffmaister, and Medeiros (1997), using variance decompositions, found that world interest rate shocks explain a large component of medium-term fluctuations in capital inflows in Brazil.

On the internal side, factors that attract capital flows include sound monetary and fiscal policies and market-oriented reforms, such as trade and capital market liberalization. Inflation stabilization reduces risks and stimulates capital inflows. Yet, Fernández-Arias and Montiel (1995) conclude that formal evidence indicates that falling interest rates in advanced economies have played a dominant role in driving capital to developing countries and that flows were not restricted to countries with good reform records.

Finally, there are contagion effects. Capital flows to a couple of countries in a region generate externalities to neighboring countries and an external crisis in one country may spread to others.

We now investigate if the above conclusions apply to Brazil. The ordinary least squares (OLS) regression controlling for heteroscedasticity and serial correlation is

where nf, i, i*, and Ee are the net capital flows as a percentage of GDP, the domestic interest rate, the foreign interest rate, and expected devaluation, respectively, X is a group of variables including domestic variables such as inflation, government spending, the real exchange rate, a dummy for the Real Plan, and a variable for contagion effects, that is, a dummy for the Tequila effect, and ε is a random error. The data are described in Appendix III and results are summarized in Table 2.

As predicted by theory, the coefficient of the international interest rate is negative and significant. This result is consistent with evidence for Latin America in Calvo and others (1993), and with evidence for developing countries in Fernández-Arias and Montiel (1995). This negative relationship is illustrated in Figure 2. The result is robust across specifications and to using either returns on U.S. treasury bills or yields on 10-year treasury bonds.

Table 2.

Dependent Variable: Ratio of Monthly Total Net Private Capital flows to GDP (nf)

(January 1988-December 1995)

Notes: i* is the interest rate on U.S. treasury bills and i–Eeis the interest rate on Brazilian government bonds deflated by expected devaluation. Standard errors are corrected by Newey-West heteroscedasticity and autocorrelation consistent covariance matrix. t-statistics are in parentheses.

Table 2.

Dependent Variable: Ratio of Monthly Total Net Private Capital flows to GDP (nf)

(January 1988-December 1995)

Constant		4.97	4.78	3.87
		(8.35)	(7.75)	(3.52)
i*		-0.66	-0.66	-0.65
		(-7.14)	(-7.08)	(-6.60)
i–Ee		0.07	0.06	0.08
		(2.30)	(2.08)	(2.30)
Tequila dummy		-3.41	-4.26	-4.34
		(-2.79)	(-3.21)	(-3.28)
Real Plan dummy		—	1.04	1.41
		—	(1.84)	(1.49)
Inflation rate		—	—	0.02
		—	—	(1.18)
Ratio of government		—	—	1.56
	spending to GDP	—	—	(0.56)
Real exchange rate		—	—	-0.93
	(deviation from	—	—	(-0.85)
	equilibrium rate)
Adjusted R²		0.44	0.46	0.46

Notes: i* is the interest rate on U.S. treasury bills and i–Eeis the interest rate on Brazilian government bonds deflated by expected devaluation. Standard errors are corrected by Newey-West heteroscedasticity and autocorrelation consistent covariance matrix. t-statistics are in parentheses.

View Table

Table 2.

Dependent Variable: Ratio of Monthly Total Net Private Capital flows to GDP (nf)

(January 1988-December 1995)

Constant		4.97	4.78	3.87
		(8.35)	(7.75)	(3.52)
i*		-0.66	-0.66	-0.65
		(-7.14)	(-7.08)	(-6.60)
i–Ee		0.07	0.06	0.08
		(2.30)	(2.08)	(2.30)
Tequila dummy		-3.41	-4.26	-4.34
		(-2.79)	(-3.21)	(-3.28)
Real Plan dummy		—	1.04	1.41
		—	(1.84)	(1.49)
Inflation rate		—	—	0.02
		—	—	(1.18)
Ratio of government		—	—	1.56
	spending to GDP	—	—	(0.56)
Real exchange rate		—	—	-0.93
	(deviation from	—	—	(-0.85)
	equilibrium rate)
Adjusted R²		0.44	0.46	0.46

Notes: i* is the interest rate on U.S. treasury bills and i–Eeis the interest rate on Brazilian government bonds deflated by expected devaluation. Standard errors are corrected by Newey-West heteroscedasticity and autocorrelation consistent covariance matrix. t-statistics are in parentheses.

The dummy for the Tequila months (January-March 1995) is significant. as expected from the contagion effects reported during the Mexican crisis. The coefficient of the domestic interest rate adjusted for expected depreciation is positive, as predicted. The coefficients of other domestic factors do not help to explain capital flows to Brazil. The coefficients of the dummy for the Real Plan and the coefficient of the real exchange rate are insignificant. The coefficients of the inflation rate and of expenditures have a different sign than expected. We interpret the results as evidence in favor of push effects as opposed to pull effects in explaining the recent surge in capital flows.

Appendix I presents additional results using different sample periods and different data (primary flows data from Loanware and Bondware) and checks the effects of the factors above in the composition of flows. The main conclusion above holds for these samples as well. It should be noted, however, that the above results, as well as most of the analysis in the literature on capital flows, do not consider the effect of capital controls. We now introduce capital controls into the analysis.

Treasury Bill Rates and Capital Flows

Citation: IMF Staff Papers 1998, 004; 10.5089/9781451974515.024.A006

Sources: Banco Central do Brasil; and IMF, International Financial Statistics.

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Treasury Bill Rates and Capital Flows

Citation: IMF Staff Papers 1998, 004; 10.5089/9781451974515.024.A006

Sources: Banco Central do Brasil; and IMF, International Financial Statistics.

• Download Figure
• Download figure as PowerPoint slide

Treasury Bill Rates and Capital Flows

Citation: IMF Staff Papers 1998, 004; 10.5089/9781451974515.024.A006

Sources: Banco Central do Brasil; and IMF, International Financial Statistics.

• Download Figure
• Download figure as PowerPoint slide

Capital Controls

The central bank sets an adjustable band for the dollar value of the real and maintains a continuing crawling peg within it, while the National Monetary Council is responsible for formulating the overall foreign exchange policy. Regulations on capital outflows and capital inflows differ. Brazilian banks are permitted to sell foreign exchange to Brazilian investors in the Common Market of the South (MERCOSUR) countries but outward capital transfers not included in public regulations need prior authorization from the central bank.

Portfolio investment by foreign investors is restricted to two classes of fixed-income funds, and foreign investment in the Brazilian capital market may be made through one of the five alternatives established under National Monetary Council Resolution 1289. Special regulations govern borrowing abroad. Payments for current invisibles not covered by current regulations require approval from the central bank’s exchange department. Remittances abroad of income from foreign direct investment and remittances in respect of royalties and technical assistance require prior registration of the foreign capital concerned, including reinvestment, and the contracts for patents and trademarks with the department of foreign capital of the central bank.

Capital inflows in the form of financial loans require prior approval from the central bank. Proceeds of foreign borrowing are subject to a financial transaction tax with rates that range from 5 percent for loans with maturities under three years to 0 percent for loans with maturities over six years. Otherwise, inward transfers are unrestricted, although use of the proceeds for the acquisition of certain domestic assets is restricted. Remittances of interest on loans and credits and of related amortization payments are permitted freely in accordance with the terms stipulated in the respective contract and recorded in the certificate of registration. Purchasers of foreign exchange for some current invisibles are subject to the financial transaction tax of 25 percent.

Appendix II lists the monthly changes in taxes and restrictions on capital flows and on payments of invisibles from 1983 to 1995 based on the IMF’s Annual Report on Exchange Arrangements and Exchange Restrictions. The list also includes other changes in legislation that affect payments abroad used to circumvent legislation on capital flows. In the 1980s, following the debt crisis, controls on capital outflows were the norm. With the capital surge of the early 1990s, controls on outflows were lifted and controls on inflows were increased before the Mexican financial crisis and then reduced after the crisis induced an increase in capital outflows. Controls are divided into two types: restrictions on inflows and restrictions on outflows.

The changes in legislation listed in Appendix II were used to build indicators of restrictions on inflows and outflows and composite indicators as well. The first indicator is a measure of restrictions on inflows and defined as ΔRI. A change in legislation that is directed at reducing capital inflows, such as an increase in the financial tax on capital inflows, receives a value equal to 1, and a change in legislation that is directed at increasing capital inflows, such as a reduction of the financial tax on capital inflows, receives a value of –1. Each month the changes in legislation affecting capital inflows are added to obtain the total value of changes in legislation affecting capital inflows in that month, ΔRI.²

Between January 1983 and December 1995, monthly ΔRI varied between –3 and 3. The average number of changes in restrictions on capital inflows per month during the whole period was –0.051, characterizing a trend of liberalization of restrictions on capital inflows.

The second indicator measures changes in restrictions on capital outflows, ΔRO. Any change in legislation that is aimed at reducing capital outflows, such as introducing new restrictions on payments of debt amortization by public enterprises, receives a value of 1. Any change in legislation that liberalizes capital outflows, such as an agreement for the elimination of arrears, receives a value of –1. By adding up the changes in restrictions on capital outflows in a month we obtain ΔRO. Between January 1983 and December 1995, monthly ΔRO varied between 1 and –3. The average number of changes in restrictions on capital outflows per month during the whole 1983–95 period was –0.045, characterizing a trend of liberalization of restrictions on capital outflows.

This paper also uses overall measures of capital controls composed by both types of restrictions. Restrictions on inflows potentially reduce capital inflows and thus potentially reduce net capital inflows. In the calculation of an overall measure of restrictions on net flows, changes in restrictions on capital inflows, ΔRI, are thus recorded as a positive restriction on net flows. Restrictions on capital outflows, however, have two potential effects on net flows. First, they reduce officially registered outflows and thus potentially increase total net flows. But foreign investors will perceive restrictions on outflows as a threat to remitting abroad the returns of their investments. This policy, thus, can reduce inflows. The effect on net flows of restrictions on outflows will thus depend on the relative strength of the responses of outflows and inflows.

This subsection builds two composite measures of capital controls, ΔCC1 and ΔCC2, defined as linear combinations of changes in restrictions on inflows and outflows. Both composite measures allow for opposite responses of capital outflows and capital inflows to changes in ΔRO by setting the weight given to ΔRO at one-half.³ The first overall measure, ΔCC1, assumes that restrictions on outflows have a bigger impact on outflows than on inflows:

Figure 3 shows changes in the composite indicator of capital controls, ΔCC1, between January 1983 and December 1995. ΔCC1 varies between –3.5 and 6.5, averaging –0.029 per month during the whole period.

Changes in Restrictions on Capital Flows, January 1983-December 1995

(Changes in index of control, ΔCC1)

Citation: IMF Staff Papers 1998, 004; 10.5089/9781451974515.024.A006

Source: Authors’ calculation.

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Changes in Restrictions on Capital Flows, January 1983-December 1995

(Changes in index of control, ΔCC1)

Citation: IMF Staff Papers 1998, 004; 10.5089/9781451974515.024.A006

Source: Authors’ calculation.

• Download Figure
• Download figure as PowerPoint slide

Changes in Restrictions on Capital Flows, January 1983-December 1995

(Changes in index of control, ΔCC1)

Citation: IMF Staff Papers 1998, 004; 10.5089/9781451974515.024.A006

Source: Authors’ calculation.

• Download Figure
• Download figure as PowerPoint slide

The second overall measure assumes that the impact on capital inflows of restrictions on capital outflows dominates the effect of these restrictions on outflows:

Between 1983 and 1995, ΔCC2 varied between -3 and 6.5, averaging -0.073 per month during the whole period.

Does a reduction in controls over capital outflows and an increase in controls over capital inflows actually reduce net capital flows in a significant way? Are controls just an ineffective reaction of authorities to capital flows? That is, are restrictions on capital outflows relaxed and capital controls on inflows increased when there is a surge in inflows? Which effects are more important?

Between January 1988 and December 1995 there is a positive correlation (equal to 0.34) between total net monthly private capital flows and changes in capital controls on net flows in Brazil, ΔCC1. This is explained by a positive correlation (0.33) between RI and capital flows and a negative correlation (-0.09) between ΔRO and flows. Such correlations could be interpreted as a first intuitive result indicating that controls are ineffective in reducing capital flows.

Changes in controls are introduced in the OLS regressions from the previous subsection. Most of the previous results are confirmed; relative returns and contagion effects are important. The coefficients on capital controls are positive and significant (Table 3). This result would reject the hypothesis that capital controls reduce capital inflows and would suggest that the reaction of capital controls to flows is important.

Endogeneity of Capital Controls

We now assume that domestic and foreign economic agents compare expected relative returns on domestic and foreign assets and restrictions imposed by different types of controls. Thus, we can write that total net private capital flows, NF, respond to expected returns, ER, and to changes in net capital controls, ΔCC:

where ε is a random error.

Table 3.

Dependent Variable: Ratio of Total Net Capital Flows to GDP

(January 1988-December 1995)

Notes: ΔRI is the increase in restrictions on capital inflows, ΔRO is the increase in restrictions on capital outflows, ΔCC1 = ΔRI - 0.5 ΔRO; ΔCC2 = ΔRI + 0.5 ΔRO, i*is the interest rate on U.S. treasury bills, and i-Eeis the interest rate on Brazilian government bonds deflated by expected devaluation, t-statistics are in parentheses. Standard errors are corrected by Newey-West heteroscedasticity and an autocorrelation consistent covariance matrix.

Table 3.

Dependent Variable: Ratio of Total Net Capital Flows to GDP

(January 1988-December 1995)

Constant	4.85	5.08	4.92	4.81
	(8.28)	(8.09)	(8.48)	(8.06)
ΔRI	0.37	—	—	—
	(2.54)	—	—	—
ΔRO	—	0.31	—	—
	—	(1.21)	—	—
ΔCC1	—	—	—	0.33
				(2.26)
ΔCC2	—	—	0.39	—
			(2.86)
i*	-0.64	-0.67	-0.65	-0.36
	(-6.97)	(-6.91)	(-7.08)	(-2.91)
i–Ee	0.07	0.08	0.08	0.07
	(2.31)	(2.41)	(2.40)	(2.21)
Tequila Plan dummy	-2.64	-3.56	-2.71	-2.66
	(-2.52)	(-2.73)	(-2.48)	(-2.60)
Adjusted R²	0.47	0.43	0.48	0.46

Notes: ΔRI is the increase in restrictions on capital inflows, ΔRO is the increase in restrictions on capital outflows, ΔCC1 = ΔRI - 0.5 ΔRO; ΔCC2 = ΔRI + 0.5 ΔRO, i*is the interest rate on U.S. treasury bills, and i-Eeis the interest rate on Brazilian government bonds deflated by expected devaluation, t-statistics are in parentheses. Standard errors are corrected by Newey-West heteroscedasticity and an autocorrelation consistent covariance matrix.

View Table

Table 3.

Dependent Variable: Ratio of Total Net Capital Flows to GDP

(January 1988-December 1995)

Constant	4.85	5.08	4.92	4.81
	(8.28)	(8.09)	(8.48)	(8.06)
ΔRI	0.37	—	—	—
	(2.54)	—	—	—
ΔRO	—	0.31	—	—
	—	(1.21)	—	—
ΔCC1	—	—	—	0.33
				(2.26)
ΔCC2	—	—	0.39	—
			(2.86)
i*	-0.64	-0.67	-0.65	-0.36
	(-6.97)	(-6.91)	(-7.08)	(-2.91)
i–Ee	0.07	0.08	0.08	0.07
	(2.31)	(2.41)	(2.40)	(2.21)
Tequila Plan dummy	-2.64	-3.56	-2.71	-2.66
	(-2.52)	(-2.73)	(-2.48)	(-2.60)
Adjusted R²	0.47	0.43	0.48	0.46

Notes: ΔRI is the increase in restrictions on capital inflows, ΔRO is the increase in restrictions on capital outflows, ΔCC1 = ΔRI - 0.5 ΔRO; ΔCC2 = ΔRI + 0.5 ΔRO, i*is the interest rate on U.S. treasury bills, and i-Eeis the interest rate on Brazilian government bonds deflated by expected devaluation, t-statistics are in parentheses. Standard errors are corrected by Newey-West heteroscedasticity and an autocorrelation consistent covariance matrix.

Capital controls are endogenous. The policy reaction function reflects the effect of capital flows on the creation or withdrawal of new controls. With large capital inflows, governments tend to impose controls to avoid real exchange rate appreciation, money expansion from accumulating reserves, or increasing debt from sterilization policies, and they react to excessive inflows by taxing some forms of capital inflows and imposing other forms of restrictions. On the other hand, controls on capital outflows are imposed in moments of distress, when there is a shortage of external financing, and authorities respond to a decline in net flows by restricting outflows and creating incentives to inflows. Thus, restrictions on net inflows respond positively to net capital inflows:

where v is a random error. Substitution yields

and

where

Observe that in the reduced form (6), the effect of expected returns on net capital flows, β, would be unaffected by the introduction of controls (i.e., q = 1 and β = a) if either capital controls can be avoided (b is zero), or if capital controls are not endogenous (h is zero). But the response of capital flows to expected returns would appear to be smaller than it is— the smaller q—the bigger the combined effect of capital controls on capital flows and the response of controls imposed by the authorities if inflows increase. Assume there is a decline in foreign interest rates. The increased relative expected return on domestic assets increases capital inflows. If the authorities respond immediately by increasing controls and controls bite, the increase in capital inflows is reduced and the effect of the decline on foreign interest rates would appear smaller than it is in fact.

Also observe that the covariance between net capital flows and capital controls depends on the relative size of b, the coefficient of capital controls in the equation that determines net flows, and h, the coefficient of net flows on the equation that determines changes in capital controls:

Equation (8) implies that the relationship between capital flows and capital controls can appear as positive or negative depending on the strength of different effects. For the given variance of the shocks v and ε. if capital controls are not endogenous and effective (b is big relative to h), the correlation between net flows and changes in capital controls will tend to be negative. But the correlation would tend to be positive if controls increase in response to an increase in capital inflows but are not effective (h is big in relation to b).⁴

The positive correlations between capital flows and capital controls reported in the previous subsection were interpreted as a first intuitive signal that evidence may support the hypothesis that one direction of causality runs from capital flows to capital controls. It also warns against using capital controls as exogenous variables in OLS regressions. The positive coefficients found in the regressions reported in Table 3 are also consistent with the hypothesis that the positive response of controls to flows dominates the negative effect controls may have on flows. Furthermore, it also implies that the coefficients in the OLS regressions are biased downward.

The government reaction function to capital flows can be estimated through a reverse regression using expected returns as instruments to capital flows. Table 4 shows the results of two-stage least squares regressions that use foreign interest rates and domestic returns in dollars as instruments for net foreign flows in regressions relating net flows to capital controls. The regressions include the dummy variables, ANNEX and REALPLAN, to control for the change in regime with the introduction of Annex IV legislation in 1991 and the Real Plan in 1994, respectively. The regressions show that an increase in net flows increases controls. The coefficient of net foreign capital flows is always positive and significant (consistently, the coefficient in the ΔRO regression is negative).

The regressions in Table 4 show that the authorities react immediately to an increase in net flows. But it is possible that when they react they look not only at the last observed monthly inflow but also at the behavior of capital flows during the previous three to six months. Table 5 shows the results of regressions in which accumulated lagged capital flows appear as the independent variable determining capital controls. The coefficient of these accumulated lagged flows is positive and significant in all regressions. The independent variables are the lagged average net flows during the previous three, four, five, and six months. Observe that the response of controls increases as the period of observation increases, indicating that sustained increases in net flows will meet with a stronger policy response than would a single event. This subsection thus finds strong evidence that net capital flows strongly influence policy decisions on implementing or reducing restrictions on capital flows.

Table 4.

Dependent Variable: Change in Capital Controls Instruments for Net Capital Flows Are i* i–Ee, and the Tequila Effect

(January 1988-December 1995)

Notes: ΔRI is the increase in restrictions on capital inflows, ΔROis the increase in restrictions on capital outflows, ΔCC1 = ΔRI -0.5ΔRO; ΔCC2 = ΔRI + 0.5ΔRO, i* is the interest rate on U.S. treasury bills, i–Ee is the interest rate on Brazilian government bonds deflated by expected devaluation, and nf are the net private capital flows; f-statistics are in parentheses.

Table 4.

Dependent Variable: Change in Capital Controls Instruments for Net Capital Flows Are i* i–Ee, and the Tequila Effect

(January 1988-December 1995)

Dependent variable	ΔCC1	ΔCC2	ΔRI	ΔRO
Constant	0.20	0.16	0.18	-0.04
	(1.78)	(1.21)	(1.52)	(-0.78)
nf	0.56	0.42	0.49	-0.14
	(5.85)	(5.35)	(6.04)	(-2.23)
ANNEX	-1.48	-1.33	-1.4	0.15
	(-5.49)	(-5.31)	(-5.85)	(0.78)
REALPLAN	0.32	0.35	0.33	0.03
	(0.52)	(0.88)	(0.68)	(0.10)

Notes: ΔRI is the increase in restrictions on capital inflows, ΔROis the increase in restrictions on capital outflows, ΔCC1 = ΔRI -0.5ΔRO; ΔCC2 = ΔRI + 0.5ΔRO, i* is the interest rate on U.S. treasury bills, i–Ee is the interest rate on Brazilian government bonds deflated by expected devaluation, and nf are the net private capital flows; f-statistics are in parentheses.

View Table

Table 4.

Dependent Variable: Change in Capital Controls Instruments for Net Capital Flows Are i* i–Ee, and the Tequila Effect

(January 1988-December 1995)

Dependent variable	ΔCC1	ΔCC2	ΔRI	ΔRO
Constant	0.20	0.16	0.18	-0.04
	(1.78)	(1.21)	(1.52)	(-0.78)
nf	0.56	0.42	0.49	-0.14
	(5.85)	(5.35)	(6.04)	(-2.23)
ANNEX	-1.48	-1.33	-1.4	0.15
	(-5.49)	(-5.31)	(-5.85)	(0.78)
REALPLAN	0.32	0.35	0.33	0.03
	(0.52)	(0.88)	(0.68)	(0.10)

Notes: ΔRI is the increase in restrictions on capital inflows, ΔROis the increase in restrictions on capital outflows, ΔCC1 = ΔRI -0.5ΔRO; ΔCC2 = ΔRI + 0.5ΔRO, i* is the interest rate on U.S. treasury bills, i–Ee is the interest rate on Brazilian government bonds deflated by expected devaluation, and nf are the net private capital flows; f-statistics are in parentheses.

Table 5.

Dependent Variable: Controls

(1988-95; constant not reported)

Note: t-statistics are in parentheses.

Table 5.

Dependent Variable: Controls

(1988-95; constant not reported)

Dependent variable	ΔCC1	ΔCC2
Average of nf(-1) to nf(-3)	0.17	0.13
	(2.32)	(1.84)
Average of nf(-1) to nf(-4)	0.19	0.15
	(2.51)	(1.92)
Average of nf(-1) lo nf(–5)	0.20	0.14
	(2.45)	(1.74)
Average of nf(-1) to nf(-6)	0.22	0.16
	(2.46)	(1.80)

Note: t-statistics are in parentheses.

View Table

Table 5.

Dependent Variable: Controls

(1988-95; constant not reported)

Dependent variable	ΔCC1	ΔCC2
Average of nf(-1) to nf(-3)	0.17	0.13
	(2.32)	(1.84)
Average of nf(-1) to nf(-4)	0.19	0.15
	(2.51)	(1.92)
Average of nf(-1) lo nf(–5)	0.20	0.14
	(2.45)	(1.74)
Average of nf(-1) to nf(-6)	0.22	0.16
	(2.46)	(1.80)

Note: t-statistics are in parentheses.

Controls and Total Net Flows

We now analyze the effects of capital controls on capital flows in a VAR framework. This framework permits inference of the dynamic response of flows to controls over time now that the previous sections have shown that there is an inverse causality running from flows to controls through a policy reaction function.

The estimated VAR takes into account the dynamic interactions between capital flows, capital controls, and interest differentials. Formally, we can express the system in a reduced form format:

where X is the set of endogenous variables that includes capital flows (nf), changes in capital controls (ΔCC1), and interest differentials (i – E(e) – i*), B(L) is a lag operator of the order L, and ε is a vector of reduced-form residuals.⁵

We are particularly interested in the effect of innovations of changes in capital controls on capital flows over time once the reverse effect is taken into account. The system above can be inverted and represented as a moving average of past shocks:

The objective is to graph the impulse response of structural shocks to our endogenous variables. In general, the reduced-form residuals are a linear combination of the structural innovations that can only be obtained once sufficient identifying assumptions are made. In our case, it is crucial to disentangle the simultaneous correlation of capital controls and flows.

The instrumental variables regressions in Table 4 indicate that capital controls react contemporaneously to capital flows. We thus assume in this subsection that shocks to capital flows can potentially affect capital controls contemporaneously. In addition, we now assume that capital flows do not react contemporaneously to capital controls, at least on a monthly basis. This is justified by the results obtained in the OLS regressions in Table 3. which show a positive relationship between controls and net flows, indicating that the reverse causality of flows to controls dominates any immediate effect that controls can have on flows. The reasons for no contemporaneous response of capital flows to capital controls are as follows. Investment projects may take more than one month to react to new legislation and part of the reaction of the capital market will be reflected in price movements. These assumptions make the system recursive in the following order: interest differentials, capital flows, and capital controls. We can thus use the standard Cholesky decomposition to orthogonalize the reduced-form residuals.

One could argue that the ordering should be the inverse, with capital flows reacting to controls contemporaneously but not vice versa. This will be the case if one believes that markets react very fast to controls (including foreign direct investment flows), but the authorities take time to implement a response to a change in capital flows. All the results described below are robust to reordering the variable shocks in this manner when using the Cholesky decomposition.

The impulse responses using the components of capital controls are shown in Figures 4 and 5. The effects of ARO and ARI on capital flows have opposite signs. While shocks to ΔRI reduce capital flows, shocks to ΔRO increase them. These results indicate that ΔCC1 is the appropriate controls measure where restrictions on outflows enter with an opposite sign to restrictions on inflows.

The impulse responses using ΔCC1 are shown in Figure 6. The evidence shows that a permanent increase in controls (a shock to ΔCC1) has a temporary negative effect on capital flows, with a peak five months after the increase and the effect fading out fast after the sixth month. This result implies that policymakers were able to change temporarily, but not permanently, the amount of capital flows to Brazil.

The impulse responses also show that capital flows have a positive effect on changes in capital controls. A boom in capital flows induces a permanent effect on controls (a temporary effect on the changes of controls in the first two months). This confirms the instrumental variable results from the previous section. In addition, as expected, an increase in interest differentials boosts capital flows to Brazil (not shown in Figure 6).

VAR Using Restrictions on Inflows: Response of Capital Flows to Restrictions on Inflows

(Response to 1 standard deviation innovation ± 2 standard errors)

Citation: IMF Staff Papers 1998, 004; 10.5089/9781451974515.024.A006

• Download Figure
• Download figure as PowerPoint slide

View Full Size

VAR Using Restrictions on Inflows: Response of Capital Flows to Restrictions on Inflows

(Response to 1 standard deviation innovation ± 2 standard errors)

Citation: IMF Staff Papers 1998, 004; 10.5089/9781451974515.024.A006

• Download Figure
• Download figure as PowerPoint slide

VAR Using Restrictions on Inflows: Response of Capital Flows to Restrictions on Inflows

(Response to 1 standard deviation innovation ± 2 standard errors)

Citation: IMF Staff Papers 1998, 004; 10.5089/9781451974515.024.A006

• Download Figure
• Download figure as PowerPoint slide

VAR Using Restrictions on Outflows: Response of Capital Flows to Outflows Controls

(Response to 1 standard deviation innovation ± 2 standard errors)

Citation: IMF Staff Papers 1998, 004; 10.5089/9781451974515.024.A006

• Download Figure
• Download figure as PowerPoint slide

View Full Size

VAR Using Restrictions on Outflows: Response of Capital Flows to Outflows Controls

(Response to 1 standard deviation innovation ± 2 standard errors)

Citation: IMF Staff Papers 1998, 004; 10.5089/9781451974515.024.A006

• Download Figure
• Download figure as PowerPoint slide

VAR Using Restrictions on Outflows: Response of Capital Flows to Outflows Controls

(Response to 1 standard deviation innovation ± 2 standard errors)

Citation: IMF Staff Papers 1998, 004; 10.5089/9781451974515.024.A006

• Download Figure
• Download figure as PowerPoint slide

VAR Using Controls

(Response to 1 standard deviation innovation ±2 standard errors)

Citation: IMF Staff Papers 1998, 004; 10.5089/9781451974515.024.A006

• Download Figure
• Download figure as PowerPoint slide

View Full Size

VAR Using Controls

(Response to 1 standard deviation innovation ±2 standard errors)

Citation: IMF Staff Papers 1998, 004; 10.5089/9781451974515.024.A006

• Download Figure
• Download figure as PowerPoint slide

VAR Using Controls

(Response to 1 standard deviation innovation ±2 standard errors)

Citation: IMF Staff Papers 1998, 004; 10.5089/9781451974515.024.A006

• Download Figure
• Download figure as PowerPoint slide

Table 6.

Variance Decomposition of Capital Flows

Table 6.

Variance Decomposition of Capital Flows

		Percentage	Percentage	Percentage
	Forecasting	attributed to	attributed to	attributed to
Period	error o fnf	ΔCC1	nf	i–E(e)–i*
1	1.203730	1.563098	98.05191	0.384990
2	1.423215	1.301870	94.99661	3.701523
3	1.530930	1.650398	93.72625	4.623349
4	1.639868	11.82882	84.12475	4.046434
5	1.766686	21.02420	74.62018	4.355625
6	1.835479	26.38333	69.53941	4.077260
7	1.875571	25.32275	70.22430	4.452946
8	1.907860	24.57291	70.06804	5.359048
9	1.933946	24.26400	69.39995	6.336045
10	1.961063	26.00287	67.75109	6.246041
11	1.986576	27.13729	66.54943	6.313274
12	1.998234	27.84834	65.87634	6.275316
13	2.000508	27.81296	65.78667	6.400378
14	2.001842	27.82971	65.69959	6.470701
15	2.004149	27.88063	65.55388	6.565488

View Table

Table 6.

Variance Decomposition of Capital Flows

		Percentage	Percentage	Percentage
	Forecasting	attributed to	attributed to	attributed to
Period	error o fnf	ΔCC1	nf	i–E(e)–i*
1	1.203730	1.563098	98.05191	0.384990
2	1.423215	1.301870	94.99661	3.701523
3	1.530930	1.650398	93.72625	4.623349
4	1.639868	11.82882	84.12475	4.046434
5	1.766686	21.02420	74.62018	4.355625
6	1.835479	26.38333	69.53941	4.077260
7	1.875571	25.32275	70.22430	4.452946
8	1.907860	24.57291	70.06804	5.359048
9	1.933946	24.26400	69.39995	6.336045
10	1.961063	26.00287	67.75109	6.246041
11	1.986576	27.13729	66.54943	6.313274
12	1.998234	27.84834	65.87634	6.275316
13	2.000508	27.81296	65.78667	6.400378
14	2.001842	27.82971	65.69959	6.470701
15	2.004149	27.88063	65.55388	6.565488

The effectiveness of controls can also be gauged by analyzing the variance decomposition of the forecasted errors of capital flows. Capital controls explain almost 28 percent of total variance of capital flows after 15 periods, while interest differentials explain only 6.5 percent (Table 6).

The variance decomposition of capital controls shows that almost 20 percent of the variance of capital controls can be explained by reactions to shocks to capital flows. As expected, little is explained directly by interest differentials (Table 7).

Controls and the Composition of Flows

Most of the control measures implemented during the 1990s were not intended to reduce the overall flow of capital to Brazil but rather to change its composition. Controls were directed against equity and debt flows, which were suspected of higher volatility, while the authorities tried to reinforce incentives to net direct investments.⁶

This subsection investigates the effect of controls on the composition of flows and discusses the evidence of separate VARs for each component of the capital flows. The results are shown in Figures 7, 8, and 9. A shock to controls reduces debt securities flows between the fourth and the seventh months after introduction but the effect slowly fades away. The same is true with portfolio investment. In contrast, net direct investment flows remain stable and are not affected by a shock to ΔCC. In other words, a permanent increase in controls reduces temporarily the proportion of debt and equity relative to net direct investment.

Table 7.

Variance Decomposition of Capital Controls

Table 7.

Variance Decomposition of Capital Controls

		Percentage		Percentage
	Forecasting	altri baled to	Percentage	attributed to
Period	error of ΔCC1	i–E(e)–i*	attributed to nf	ΔCC1
1	1.079158	0.026848	2.56346	97.40969
2	1.165686	1.716712	14.10603	84.17726
3	1.181754	1.871907	16.14342	81.98467
4	1.212708	1.777683	15.53215	82.69017
5	1.246686	4.560536	17.17321	78.26626
6	1.255428	4.741208	18.03469	77.22410
7	1.263243	4.853373	18.76243	76.38420
8	1.275088	5.467750	18.54992	75.98233
9	1.283901	5.520693	18.42165	76.05766
10	1.293778	5.627142	18.75923	75.61363
11	1.298784	5.680307	19.27598	75.04371
12	1.300532	5.861304	19.26517	74.87353
13	1.302401	5.844983	19.24474	74.91028
14	1.307080	5.898993	19.12518	74.97583
15	1.308675	5.922094	19.08979	74.98811

View Table

Table 7.

Variance Decomposition of Capital Controls

		Percentage		Percentage
	Forecasting	altri baled to	Percentage	attributed to
Period	error of ΔCC1	i–E(e)–i*	attributed to nf	ΔCC1
1	1.079158	0.026848	2.56346	97.40969
2	1.165686	1.716712	14.10603	84.17726
3	1.181754	1.871907	16.14342	81.98467
4	1.212708	1.777683	15.53215	82.69017
5	1.246686	4.560536	17.17321	78.26626
6	1.255428	4.741208	18.03469	77.22410
7	1.263243	4.853373	18.76243	76.38420
8	1.275088	5.467750	18.54992	75.98233
9	1.283901	5.520693	18.42165	76.05766
10	1.293778	5.627142	18.75923	75.61363
11	1.298784	5.680307	19.27598	75.04371
12	1.300532	5.861304	19.26517	74.87353
13	1.302401	5.844983	19.24474	74.91028
14	1.307080	5.898993	19.12518	74.97583
15	1.308675	5.922094	19.08979	74.98811

In contrast, the government reaction function seems to be driven solely by equity securities flows (the figures are not shown in the paper). Higher debt or net direct investment flows have no effect on controls. The combination of the results above shows an asymmetry. While debt flows are largely affected by capital controls, the government control function does not react to debt flows. One hypothesis is that government mainly targets equity flows but the effect of controls spill over to debt flows.

Robustness of the Results

One of the main shortcomings of the results is that monthly data flows to Brazil do not report very short-term flows. This may bias our results to downplay the effectiveness of controls (although we have already found a temporary effect).

VAR Using Debt Flows: Response of Debt Flows to Controls

(Response to 1 standard deviation innovation ± 2 standard errors)

Citation: IMF Staff Papers 1998, 004; 10.5089/9781451974515.024.A006

• Download Figure
• Download figure as PowerPoint slide

View Full Size

VAR Using Debt Flows: Response of Debt Flows to Controls

(Response to 1 standard deviation innovation ± 2 standard errors)

Citation: IMF Staff Papers 1998, 004; 10.5089/9781451974515.024.A006

• Download Figure
• Download figure as PowerPoint slide

VAR Using Debt Flows: Response of Debt Flows to Controls

(Response to 1 standard deviation innovation ± 2 standard errors)

Citation: IMF Staff Papers 1998, 004; 10.5089/9781451974515.024.A006

• Download Figure
• Download figure as PowerPoint slide

VAR Using Portfolio Flows: Response of Portfolio Flows to Controls

(Response to 1 standard deviation innovation ± 2 standard errors)

Citation: IMF Staff Papers 1998, 004; 10.5089/9781451974515.024.A006

• Download Figure
• Download figure as PowerPoint slide

View Full Size

VAR Using Portfolio Flows: Response of Portfolio Flows to Controls

(Response to 1 standard deviation innovation ± 2 standard errors)

Citation: IMF Staff Papers 1998, 004; 10.5089/9781451974515.024.A006

• Download Figure
• Download figure as PowerPoint slide

VAR Using Portfolio Flows: Response of Portfolio Flows to Controls

(Response to 1 standard deviation innovation ± 2 standard errors)

Citation: IMF Staff Papers 1998, 004; 10.5089/9781451974515.024.A006

• Download Figure
• Download figure as PowerPoint slide

VAR Using FDI Flows: Response of FDI Flows to Controls

(Response to 1 standard deviation innovation ± 2 standard errors)

Citation: IMF Staff Papers 1998, 004; 10.5089/9781451974515.024.A006

• Download Figure
• Download figure as PowerPoint slide

View Full Size

VAR Using FDI Flows: Response of FDI Flows to Controls

(Response to 1 standard deviation innovation ± 2 standard errors)

Citation: IMF Staff Papers 1998, 004; 10.5089/9781451974515.024.A006

• Download Figure
• Download figure as PowerPoint slide

VAR Using FDI Flows: Response of FDI Flows to Controls

(Response to 1 standard deviation innovation ± 2 standard errors)

Citation: IMF Staff Papers 1998, 004; 10.5089/9781451974515.024.A006

• Download Figure
• Download figure as PowerPoint slide

As a first approach to this limitation, this subsection checks how robust the results are by substituting the information on net capital flows with information on gross primary flows. These data include bonds, equities, and syndicated loans issued by Brazilian entities in international markets (mainly New York and London, but increasingly global). The data are compiled by Dataware and published by Loanware and Bondware, Primary flows are useful because they are primary (not including transactions in the secondary market), gross (they do not account for nonresident purchases of Brazilian securities in Brazil and amortization and purchases of securities by Brazilian residents from nonresidents), and do include short-term flows.

The impulse response for primary flows confirms our previous results. Capital flows are negatively affected by controls within the first six months but the effect slowly disappears thereafter. Similarly, capital controls react immediately (two-four months) to a shock in capital flows.

Another approach is to replicate the exercise using quarterly data that include short-term flows.⁷ The capital flows data are obtained from the financial account of the balance of payments published by the IMF and the interest differential is borrowed from estimates in García and Barcinski (1996). The measure of capital controls is the quarterly average of the monthly indicator. The results confirm that the effect of controls on capital flows is strongest at two quarters and slowly fades after that. In contrast to monthly results, capital controls react within the first three quarters, instead of the first two months.