Colombia: Selected Issues

This Selected Issues paper uses contingent claims analysis (CCA) to assess risks to the Colombian banking sector. The CCA approach is based on the estimation of the default probability by an entity on its obligations, and is widely used by rating agencies to assess creditworthiness in the corporate sector. The paper also estimates the effects of changes in selected macroeconomic and financial variables on default probabilities for a sample of Colombian banks. The sample includes five banks for which market-based default probabilities are available.

Abstract

This Selected Issues paper uses contingent claims analysis (CCA) to assess risks to the Colombian banking sector. The CCA approach is based on the estimation of the default probability by an entity on its obligations, and is widely used by rating agencies to assess creditworthiness in the corporate sector. The paper also estimates the effects of changes in selected macroeconomic and financial variables on default probabilities for a sample of Colombian banks. The sample includes five banks for which market-based default probabilities are available.

IV. Are Capital Controls Effective in the 21ST Century? The Recent Experience of Colombia1

A. Introduction

1. The Colombian experience with capital controls has sparked a lively debate, but one which is far from settled. Confronted with rapid currency appreciation and a surge in capital inflows, controls were imposed in May 2007, and removed during the market turbulence of October 2008. Capital controls are not new in Colombia, having first been imposed in the 1990s. Research on Colombia points to a similar experience as in other countries: controls have been ineffective in altering the real exchange rate, while nevertheless enhancing the independence of monetary policy. Regarding the most recent episode, the observed decline in net private capital inflows after their imposition would appear to suggest that the controls were effective (Vargas and Varela, 2008). Cardenas (2007), however, concludes that the controls had no independent effect on total net private capital inflows, based on an econometric model that takes account of other determinants of these movements. Galindo (2007) and Concha and Galindo (2008) conclude that the controls have had a transitory effect on net private capital inflows, but without significant effects on the real exchange rate.

2. This paper attempts to add to this literature by assessing the effect of the recent controls on capital flows and the exchange rate, while addressing a number of unresolved issues. First, additional study is needed on the effects of the controls on separate categories of capital inflows. Second, most—if not all—assessments of the effects of capital controls have focused on their impact on net, rather than gross, flows, including recent studies (Cardenas, 2007; Concha and Galindo, 2008). A disaggregation of the effects on both inflows and outflows could potentially yield rich insights into how the controls affect capital markets. While the controls were imposed on gross inflows, they may also have affected outflows, owing to the effect of the new limits imposed on derivative positions and the need of some market participants to unwind their positions. Third, there is scope to more precisely quantify the effects of the controls on the exchange rate, which has received relatively less attention in recent studies.

3. The analysis in this paper focuses on an assessment of the effectiveness of capital controls, but not their costs and benefits. The paper does not analyze the microeconomic costs of controls or their adverse effects on market development. It also does not assess the potential role of the controls in helping protect financial stability by reducing the risk associated with the sudden reversal of inflows. At the same time, the analysis provides an assessment of the impact of the controls on capital flows, which could be drawn upon for a broader assessment of the effects of controls on financial stability.

4. This paper is organized as follows. First, a description of the database used for the study, and developments in exchange rates and capital flows in the periods immediately preceding and following the introduction of capital controls, is provided. This section also includes a description of the controls imposed in Colombia in May 2007. Second, the empirical methodology used to isolate and measure the effects of the controls is described. Third, the empirical results are given, followed by a summary section that also includes suggestions for further research. An appendix provides a review of the literature on capital controls, focusing on studies assessing the Colombian experience.

B. Database and Stylized Facts on Capital Flows and Exchange Rates

Database

5. The data used in this study capture all capital account transactions by the private sector that are intermediated in the foreign exchange market and recorded by the Banco de la República (BdR). Data on foreign exchange flows are derived from the BdR’s balanza cambiaria (foreign exchange balance). The categories of foreign exchange transactions included are external borrowing, portfolio investment, foreign direct investment (FDI), and other movements that involve changes in bank accounts held abroad by residents and registered at the BdR (cuentas corrientes de compensación). They exclude capital account transactions that are undertaken without an exchange of currency (e.g., certain forms of FDI).

6. The balanza cambiaria data includes foreign currency transactions between residents. Traditional balance of payments data only capture transactions between residents and nonresidents. In contrast, these operations are included in the balanza cambiaria data. For example, when a Colombian pension fund acquires foreign exchange to purchase a dollar-denominated bond owned by a Colombian investment fund, this is registered as a foreign exchange outflow. If the investment fund takes these dollars and converts them into pesos, this would be registered as an inflow. Thus, on a gross basis, the balanza cambiaria data tend to indicate a greater level of transactions than balance of payments data, although on a net basis, they present a similar picture.

7. The data is available on a weekly basis, which makes them an attractive source of information for studying the short-term behavior of capital flows. Because of its broad coverage and availability at high frequency, these data have also been employed in other studies assessing Colombia’s experience with capital controls (Concha and Galindo, 2008).

Capital flows and exchange rates in the pre-controls era

8. Following the turmoil in global markets in mid-2006, the peso appreciated significantly in the second half of the year.2 The peso recovered steadily from July through December, appreciating by about 9 percent in nominal effective terms. With the BdR beginning a period of monetary tightening (commencing in April 2006) to address overheating concerns, interest rate differentials also began to rise. Net FDI inflows were strong, while net non-FDI flows remained muted.

uA04fig01

Exchange Rates

Citation: IMF Staff Country Reports 2009, 024; 10.5089/9781451808964.002.A004

Source: Information Notice System.1/ An increase denotes an appreciation.
uA04fig02

FDI and Non-FDI Private Inflows

(US$ millions)

Citation: IMF Staff Country Reports 2009, 024; 10.5089/9781451808964.002.A004

9. Non-FDI inflows increased sharply and the exchange rate continued to appreciate in early 2007. With inflation and inflation expectations rising beyond the BdR’s inflation target, policy rates were increased by an additional 75 basis points in the first four months of the year. The exchange rate appreciated an additional 5 percent in nominal effective terms. Gross inflows rose sharply, spurred on by rising interest rate differentials and market expectations of a peso appreciation.

10. Sterilized intervention failed to quell the surging peso. The authorities attempted to resist appreciation pressures with a large dose of sterilized intervention (starting in mid-January 2007), accumulating about 38 percent of base money in the first four months of the year. Nevertheless, with markets expecting even further appreciation of the peso—in part linked to markets’ belief that the authorities’ commitment to the inflation target would lead to increasing interest rate differentials—the strategy became unsustainable, and the BdR abandoned its policy of discretionary foreign exchange intervention by the end of April 2007.3

Capital flows and exchange rates in the controls era

11. On May 6, 2007, the authorities introduced capital controls as part of a new strategy to slow the appreciation of the peso. The primary measure comprised an unremunerated reserve requirement (URR) on external borrowing of 40 percent. This revived the capital controls in place in Colombia between 1993 and 2000. The URR was to be held for six months in the BdR and applied to both the private and public sectors.

12. As a prudential measure, a ceiling on currency derivative positions was also imposed. Banks were henceforth required to keep their overall gross positions in these instruments to no more than 500 percent of their capital. Market participants were allowed some time to unwind positions to meet the new ceiling, which became binding on July 4. The measure addressed prudential concerns based on the high exposures of some banks and potential counterparty risks involved with these large positions. By restricting derivative positions, this measure also potentially affected capital flows and the exchange rate, and helped prevent circumvention of the URR.

13. The capital controls were extended in late May 2007 to portfolio inflows. With portfolio inflows accelerating the first weeks of May, the URR was extended on May 23 to portfolio inflows by foreign residents. Investors were given the option of early withdrawals of funds from the URR, but with substantial penalties, ranging from 9.4 percent of the reserve requirement (for immediate withdrawals) to 1.6 percent (if held for 5 months).

14. Colombian institutional funds, including pension funds, were excluded from the capital controls. The exclusion of the pension funds is of particular importance, as they are highly active in trading in the foreign exchange market (Gómez, Jara, and Murcia, 2006). Their exclusion is also important in light of the fact that Colombian residents and firms, rather than nonresidents, accounted for about three-fourths of portfolio inflows in the pre-controls era. In addition, in June 2007, an exemption was granted for equities issued abroad. This effectively exempted from the controls the issuance of stock through American Depository Receipts (ADRs).

15. Capital controls regulations underwent a further series of changes, beginning in late 2007, and were eliminated in October 2008. Controls were relaxed in December 2007, as the URR for initial public offerings of equities was eliminated, and penalties for early withdrawal of funds were reduced by about 40 percent. Controls were subsequently tightened in May of 2008, as the URR on portfolio inflows was raised from 40 percent to 50 percent, and a minimum stay of two years was imposed on FDI.4 Also in May, the limit on gross derivatives positions was increased from 500 percent of capital to 550 percent. In June, the penalty for early withdrawal of deposits under the URR (which had previously been relaxed in December of 2007) was also raised. Controls were subsequently loosened in September as the minimum stay requirement on FDI was revoked, and purchases of equities were exempted from the URR. In October, the capital controls were eliminated, although the limit on the derivative position of banks of 550 percent of capital remained.

16. Several aspects of the nature of private capital flows in Colombia are worth noting when analyzing the potential effects of capital controls. In net terms, FDI accounts for the vast majority of inflows (Table 1). In the pre-controls era, for example, FDI accounted for 96 percent of net private inflows, as net non-FDI flows had been close to zero. FDI was not targeted by the controls—except in the later period, via a minimum stay rule. Second, portfolio flows by residents are substantially more important than nonresidents, in gross terms; in net terms, however, the opposite is the case. Third, for resident portfolio flows, the weekly net movements are dwarfed by the size of gross inflows (net are between 2 and 8 percent of gross, depending on the period). Fourth, gross flows—which are several orders of magnitude of net flows—are inflated by the high level of transactions between residents that involve offsetting transactions involving both inflows and outflows, especially for portfolio flows between residents.5 Indeed, as Figure 1 demonstrates, periods of high inflows have corresponded to periods of high outflows.

Table 1.

Colombia: Private Capital Flows Before and After Capital Controls

(Weekly average flows in US$ millions)

article image
Note: ***, **, and * indicate that average flows in the post control period are statistically significantly different from the pre-control period, at the 1%, 5%, and 10% levels, respectively.

Average flow for the period July 6th, 2006 to May 6th (for portfolio May 23rd), 2007.

Average flow for the period May 7th (for portfolio May 24th), 2007 to Jul 4th, 2008.

Average weekly flows, annualized as a share of GDP.

Figure 1.
Figure 1.

Colombia: Private Capital Flows Before and After Capital Controls

(weekly, in US$ millions)

Citation: IMF Staff Country Reports 2009, 024; 10.5089/9781451808964.002.A004

Note: The dashed lines correspond to the dates when capital controls were first imposed in 2007 (May 6th and May 23rd 2007 respectively).

17. A comparison of mean weekly gross inflows across periods suggests that the effects of the controls was strongest for foreign borrowing and portfolio inflows from non-residents, but its overall effect on non-FDI gross inflows was insignificant. Table 1 provides mean statistics for weekly average flows across different categories. The table also reports tests for the equality of first moments between the capital control period (May 2007-July 2008) and the immediate pre-controls era (July 2006 to April 2007).6 Means tests indicate the controls have been associated with statistically significant reductions in foreign borrowing and nonresident portfolio gross inflows. Combined, these two categories accounted for about 11 percent of average total gross inflows in the pre-controls era. With respect to other flows, we find that portfolio inflows of Colombian residents have actually increased relative to the pre-controls era, while FDI inflows have also been higher.

18. On a net basis, capital inflows have not declined in the controls era. A statistically significant decline was found for just one subcategory (non-resident portfolio inflows) in net terms. This accounted, during the pre-controls period, for about 25 percent of net inflows. For foreign borrowing, resident portfolio flows, and the drawdown of bank accounts abroad monetized in local foreign exchange markets, the changes were statistically insignificant. For combined non-FDI net inflows as a whole, no statistically significant reduction was found. Because of the strong increase in gross inflows, FDI also increased, in net terms, in the controls era.

19. The capital controls have also been associated with an increase in off-shore derivative trading by the pension funds. BdR data suggest that cross-border transactions in foreign currency derivative contracts between domestic pension funds and offshore entities have increased sharply since mid-2007. The average monthly value traded rose from US$6 million in the pre-control period to an average of about US$405 million since—over a 60-fold increase.7 The average monthly value of contracts traded between off-shores and local banks, on the other hand, increased much more modestly (250 percent). The reallocation of this activity to the pension funds was probably a consequence of the ceiling of derivative positions imposed on local financial institutions, which may have led to an increase in the role of the offshore sector in providing currency hedges to the pension funds8. This may also help explain the observed increase in gross weekly inflows for Colombian residents and institutions since May 2007, while their net position has stayed roughly unchanged.

uA04fig03

Total Value Traded in Foreign Currency Derivatives between Pension Funds and Off-Shore Entities

(in millions of US dollars)

Citation: IMF Staff Country Reports 2009, 024; 10.5089/9781451808964.002.A004

20. The exchange rate continued to appreciate after the implementation of capital controls. The capital controls were not associated with an immediate effect on the exchange rate, which continued to appreciate until a period of brief global turbulence in June-August 2007. Afterwards, the peso resumed its upward climb, spurred on by the global weakness of the dollar, strong FDI flows, and improving terms of trade. From the start of the controls to end-July 2008, the bilateral rate appreciated by 16 percent.9

21. Simple comparisons of the pre-controls and controls era do not take into account changes in the domestic and external environment that may also have affected capital flows and exchange rates. On the domestic front, these included the BdR’s decision to refrain from discretionary foreign exchange intervention, and the further tightening of monetary policy, which increased interest rate differentials. Exchange rate volatility also increased, reducing the appetite of foreign investors seeking to exploit interest rate differentials. Externally, EMBI spreads increased, and the VIX—a measure of volatility in international markets—also rose. At the same time, commodity prices rose, strengthening Colombia’s trade balance, while a favorable environment for FDI and an improved security situation also helped bolster the peso. In this light, a proper assessment of the effect of the capital controls requires the development of a model that accounts for these changes in the determinants of capital flows, as attempted in Section D below.

uA04fig04

Exchange Rate Volatility, Central Bank Intervention and Capital Controls

Citation: IMF Staff Country Reports 2009, 024; 10.5089/9781451808964.002.A004

uA04fig05

Nominal and Real Exchange Rates

Citation: IMF Staff Country Reports 2009, 024; 10.5089/9781451808964.002.A004

Source: Information Notice System.1/ An increase denotes an appreciation.
uA04fig06

EMBI spread, VIX and interest rate differential 1/

Citation: IMF Staff Country Reports 2009, 024; 10.5089/9781451808964.002.A004

1/ Interest rate differential is the difference between domestic interbank rate and Fed funds rate

C. Empirical Methodology

22. The empirical analysis is based on econometric models of capital flows and the exchange rate. An OLS model is developed to assess the determinants of capital flows on a weekly basis, using the balanza cambiaria data. For the exchange rate model, daily data are used to construct a GARCH model, drawing on Kamil (2008). Capital controls are then assessed in the context of these models that control for the underlying determinants of capital flows and the exchange rate. Both the capital flows and exchange rate models are estimated using data spanning from end-July 2006 to early July 2008.

23. The focus of the analysis on both net and gross capital inflows represents a significant departure from previous work for Colombia and other emerging markets. Both flows are important to assess, as they provide different insights. Net inflows, which have traditionally been the focus of previous studies, are often seen as the clearest signal of demand pressures on the domestic currency. At the same time, gross inflow data also offer important insights for exchange rate determination, as each gross transaction provides an opportunity for the underlying asset to be repriced. The econometric results from the gross flows data—especially for the category of resident portfolio flows—should nevertheless be assessed with caution, for the reasons described earlier. The analysis also draws on more disaggregated data than previous studies. This allows us to exclude FDI flows from the analysis, which is difficult to model on a high-frequency (weekly) basis.

24. The methodology employed here overcomes many of the econometric problems of previous empirical work. First, endogeneity issues are much less severe than in many other studies. This owes to the fact that under Colombia’s inflation targeting regime, interest rates are set to achieve the inflation target, and capital inflows are fully sterilized. Thus, there is no two-way causality between capital flows and interest rate differentials. Second, in light of the use of high-frequency, weekly data over a relatively short period of time—with no discernable trend—there are no issues related to unit roots in the data. In contrast, studies using lower frequency data over a long period of time must take into account the possible nonstationarity of these series.

25. Identifying the effects of capital controls nevertheless presents several empirical challenges that need to be borne in mind when analyzing the results. First, the capital controls—in the form of the URR—was also accompanied by a limit on gross derivative positions. Thus, assessing the separate effect of the URR, and the limit on the derivative position, is particularly difficult, and is beyond the scope of the study. Second, the capital controls were imposed in stages (with the initial controls introduced on May 6 and later in May 23 of 2007). Third, the controls have undergone additional (though minor) modifications, in both December of 2007 and throughout 2008. Given the large number of changes, it was not possible to control, with dummy variables, all of these modifications.

26. Our explicit focus on assessing the effects of the controls on capital flows differs from earlier studies on the experience of Colombia in the 1990s. Villar and Rincón (2001) argue that the effect of the controls on the quantity of inflows is ambiguous, and thus of less relevance in assessing the impact of the controls. They argue that the controls impose a tax on foreign borrowing that brings about an increase in interest rates, which tends to offset the dampening effect of the controls on inflows. In the context of Colombia’s current inflation targeting regime, however, capital controls—even if they effectively place a tax on foreign borrowing—would not necessarily lead to an increase domestic interest rates, which are determined by the BdR in light of inflation objectives. Thus, the capital controls could be expected to affect the quantity of inflows, without any effect on interest rates.

The Effectiveness of Controls: Impact on Capital Flows

27. We consider five broad categories of flows: foreign borrowing, portfolio investment (bond and equity flows), movements that involve changes in bank accounts held abroad by residents (cuentas corrientes de compensación), and total non-FDI flows10. We assess the behavior of capital flows for categories not directly covered by the controls (such as the drawdown of bank accounts), given the possibility that the controls could divert carry trade towards these categories of inflows. The analysis is carried out separately for each type of capital flow, and we estimate the following OLS regression:

Flowj,t=δ0+δ0INTt+δ2(i-r*)t+δ3(EMBIspread)t+δ4VOLt+δ5APPEt+δ6DCCjt+ɛt(1)

where: Flowj corresponds to the weekly inflow for each category j of capital (either net or gross), in millions of dollars; INT is the weekly amount of the BdR’s discretionary intervention in the foreign exchange market, in millions of dollars; (ir*) is the weekly average interest rate differential between the domestic interbank rate and the U.S. Fed’s fund rate, in percent per year; EMBI spread is the weekly average yield spread on a sovereign foreign currency bond over a comparable U.S. treasury bond in percent per year; VOL is the unconditional volatility of the nominal exchange rate, over the previous three-month period; APPtE is a proxy for the expected appreciation of the exchange rate, obtained as the weekly average of the percentage difference between the spot and forward exchange rates in on-shore non-deliverable forward contracts of 15 to 35 days maturity. As a proxy for capital controls, we use a dummy variable (DCC) that takes values of 1 in the weeks after capital controls were imposed and zero otherwise.11 The main parameter of interest is δ6, the average impact of capital controls on the weekly level of capital flows. If the imposition of capital controls led to a lower level of capital inflows, then δ6 should be negative and statistically significant.

The Effectiveness of Controls: Evidence from Daily Exchange Rate Data

28. To analyze the effects of capital controls on the nominal exchange rate, we estimate a generalized autoregressive conditional heteroskedasticity (GARCH) model12 of the daily peso-dollar exchange rate return with the following general specification:

Δ(Inst)=μ+β0Δ(Inst-1)+β1INTt+β2(i-r*)t+β3Δ(EMBIspread)t+β4VIXt++β5Commt+β6DCCt+ɛt(2)
ɛt|Ωt-1N(0,σt2)(3)
σt2=α0+α1σt-12+α2ɛt-12+α3INTt+α4(i-r*)t+α5|Δ(EMBIspread)t|+α6VIXt++α7|Commt|+α8DCCt+Σi=14δiDit+vt(4)

where: Δ(ln s) is the daily percentage change in the nominal exchange rate (such that a positive change is a depreciation of the Colombian peso); VIX is the implied volatility of the U.S. stock market; Commt is the daily percentage change in a representative commodity export price index of the Colombian economy,13 and the rest of the variables are defined as above. Dit are day-of-the-week dummy variables (for example, D1t = 1 for Mondays, where Friday is the omitted category); and || denotes the absolute number operator. εt is the unexpected return, which is used to model the conditional volatility of the exchange rate in the volatility equation (4). Finally, σt2 is the conditional variance and allows for the possibility of time-varying and clustering conditional volatility. The conditional distribution of the disturbance term is normal with variance σ2.

29. Several features of the specification are worth noting. Equation (2) of the empirical model (the “mean” equation) analyzes changes in the exchange rate return (depreciation or appreciation against the dollar) as a function of the capital controls, controlling for other factors affecting exchange rates at a daily frequency.14 The main focus is on the estimate of β6, the average impact of capital controls on the level of the exchange rate. If the imposition of capital controls led to a more depreciated (less appreciated) exchange rate, then β6, the parameter of interest, should be positive and statistically significant. This would imply that the capital controls were effective in stemming the appreciation of the currency in the context of a cycle of monetary tightening.

30. The estimation controls for financial and current account developments affecting short-term exchange rate movements. The interest differential aims to capture the possible impact of monetary policy and local money market conditions on the exchange rate. Yield spreads on sovereign external debt are included as a measure of country risk and foreign investor sentiment, which are potentially key determinants of foreigners’ demand for local currency. We also include an index of the implied volatility of the U.S. stock market, the VIX index, which we take as a proxy for investors’ perception of global financial risk. We also control for daily changes in commodity prices, to capture pressures on the exchange rate associated with current account developments.

D. Empirical Results

Controls and Capital Flows

31. The econometric results suggest that the controls did not result in a reduction in aggregate non-FDI inflows (Table 2). The estimates yield a good fit for weekly data, especially for the equation estimated in net terms. They suggest that capital controls had no statistically significant impact on either gross or net non-FDI capital flows, once we control for other drivers of capital movements.15

Table 2.

Impact of Capital Controls on Non-FDI Private Capital Flows

(OLS model for weekly flow of capital: July 2006 - July 2008)

article image
Asterisks denote significance of coefficients, with ***, ** and * indicating significance at the 1%, 5% and 10% level, respectively.

32. The empirical evidence also suggests that the impact of the controls varied across different categories of inflows. The results indicate that the controls have contributed to a significant decline in foreign borrowing in both gross and net terms (Table 3). The point estimates imply that the controls led to a reduction in weekly gross foreign currency borrowing of approximately US$44 million—about equal to their average level in the pre-controls era. The controls also led to a reduction in foreign borrowing on net terms, although only at the 10 percent confidence level. Nevertheless, the econometric results provide some evidence that the controls were effective in reducing at least one category of capital inflows.16

Table 3.

Impact of Capital Controls on Foreign Borrowing and Drawdown of Residents’ Bank Accounts Abroad

(OLS model for weekly flow of capital: July 2006 - July 2008)

article image
Asterisks denote significance of coefficients, with ***, ** and * indicating significance at the 1%, 5% and 10% level, respectively.

33. There is no statistical evidence that the controls reduced aggregate portfolio inflows and the bank account transfers of Colombian residents, which represent the majority of total non-FDI flows (Tables 3 and 4). The estimated effect on the drawdown of bank accounts is insignificant at standard confidence levels. In addition, the controls had no impact on total portfolio inflows, in both gross and net terms. Looking at the impact on portfolio flows by residence, the results for non-resident bond and equity flows are mixed: there does appear to be an effect on gross inflows, but not on net flows.17 For resident flows, however, the effect is unambiguously insignificant in both net and gross terms. This most likely reflects the fact that the controls did not cover domestic institutional funds, and the continued high level of inflows of foreign currency from Colombian residents and firms to purchase peso-denominated assets. Taken together, this explains why total portfolio flows appear to show little response to capital account restrictions.

Table 4.

Impact of Capital Controls on Portfolio Inflows

(OLS model for weekly flow of capital: July 2006 - July 2008)

article image
Asterisks denote significance of coefficients, with ***, ** and * indicating significance at the 1%, 5% and 10% level, respectively.

34. Estimates of the other covariates in the regression appear generally reasonable and consistent with past research and economic theory. An increase in interest rate spreads, a reduction in exchange rate volatility, and an increase in expected appreciation are all typically expected to lead to higher net and gross inflows. The results are consistent with these priors.

35. In sum, the evidence on the effects of the controls on capital flows is mixed. Foreign borrowing appears to have declined in net terms, but only at the 10 percent confidence level. Given the modest share of foreign borrowing in total flows, however, and the statistically insignificant relationship found for other flows, the effect of the controls on total non-FDI flows is insignificant.

Controls and Exchange Rates

36. Results for estimates based on the entire sample period (July 2006-July 2008) are presented in Table 5. The first two columns present estimates of the baseline model described in equations (2)-(4). The first column reports results for the mean exchange rate equation, while the conditional variance equation estimates are presented in the second column.18 In addition, a slightly modified version of these equations was estimated to assess whether the sensitivity of the exchange rate to the explanatory variables differed under capital controls. Under this specification, we augment the model with a term that interacts the CC dummy with the interest rate differential, the VIX, EMBI, and commodity prices. If capital controls were effective in decreasing the sensitivity of the exchange rate to interest rate differentials, for example, then the coefficient in the interaction term should be positive and significant.19 If capital controls were effective in muting the transmission of global financial shocks to Colombia, then the EMBI and VIX interaction should be negative and significant.

Table 5.

Impact of Capital Controls on the Exchange Rate--Full Sample (July 2006-July 2008)

(GARCH (1,1) model for daily percentage changes in the US dollar-peso exchange rate)

article image
Asterisks denote significance of coefficients, with ***, ** and * indicating significance at the 1%, 5% and 10% level, respectively.

37. There is no statistical evidence that the capital controls have resulted in a more depreciated exchange rate, or reduced the sensitivity of the exchange rate to interest rate differentials. Using a standard t-test, the coefficient on the capital control dummy (column 1) and the interaction term of capital controls on the interest rates differential (column 3) are found to be statistically insignificant. Capital controls have also not reduced the sensitivity of the exchange rate to global developments, as indicated by the insignificant coefficient for the interaction terms for the VIX and EMBI spreads.

38. Capital controls appear to have increased exchange rate volatility. The results indicate that the controls have had the unintended consequence of increasing the conditional volatility of exchange rate returns (column 2). This finding is consistent with Edwards and Rigobón (2005), who show that capital controls increased the unconditional volatility of the exchange rate in Chile.20

39. Several factors could explain why exchange rates tended to be more variable following the introduction of capital controls. First, it may be that case that with controls on the quantity of inflows, more of the burden of adjustment falls on prices when demand for the currency changes. A weakness in this line of argumentation, however, is that the controls appear to have had no effect on the quantity of total flows affecting the exchange rate. Second, controls are likely to have segmented the foreign exchange market further, increasing conditional volatility. Third, and related to this segmentation, the capital account restrictions on nonresidents placed the pension funds in a privileged position, as they were excluded from the controls. This may have increased the importance of domestic pension funds in exchange rate market trading. For example, the average weekly trading in foreign exchange markets by pension funds (as proxied by the sum of gross portfolio inflows and outflows by residents in balanza cambiaria) increased 52 percent across periods. Given that minimum return rules provide a strong penalty for deviations from average performance in the industry and encourage herd behavior, increased activity of pension funds may have contributed to large and sudden movements of pension fund assets between peso- and foreign-currency denominated assets, exacerbating volatility. More work is needed, however, to shed light on the role of pension funds during capital controls and their impact on foreign exchange market conditions.

40. We also analyze how the results differ if the model is estimated over a shorter sub-sample (January 2007–July 2008) that starts during the period of most intense capital inflows. During this period, strong appreciation pressures coincided with a marked increase in capital inflows (see Figure 1). From this angle, this period provides a useful benchmark for comparing the environment immediately before the implementation of the controls. This also coincided with a period of heavy BdR intervention, and market expectations of continued intervention. During this period, there were also unusually high levels of currency derivative activity by the off-shore sector with local banks (Kamil and Reveiz, 2008). Using a sub sample of the available time period to estimate the model nevertheless has some drawbacks, as valuable information on the pre-controls regime is ignored in the estimation, and the fit of the model is inferior to that obtained from the larger sample.

41. Estimates from the reduced sample yield largely similar results, except with respect to the effect of the controls on the autonomy of monetary policy (table 6). The results again suggest that the controls had no effect on average exchange rate returns, as the capital controls dummy variable (DCC) in the mean equation is not statistically different from zero. At the same time, the results for the variance equation continue to indicate that the imposition of controls led to a substantial increase in volatility. We also find, consistent with the results in the baseline sample, that the imposition of capital controls did not help to insulate the foreign exchange market from changes in global risk and EMBI spreads. However, the results for the augmented model indicate that the interactive term for the interest rate spread and the capital controls dummy is positive.21 Based on the estimates reported in Table 6, the controls enhanced the independence of monetary policy once interest spreads exceeded approximately 7 percentage points, at a 95 percent confidence level.

Table 6.

Impact of Capital Controls on the Exchange Rate--Sub-Sample (January 2007-July 2008)

(GARCH (1,1) model for daily percentage changes in the US dollar-peso exchange rate)

article image
Asterisks denote significance of coefficients, with ***, ** and * indicating significance at the 1%, 5% and 10% level, respectively.

E. Conclusions and Suggestions for Future Research

42. Capital controls have been successful in reducing foreign borrowing, but other, more significant categories of inflows remain high. Foreign borrowing accounted for a modest share of total capital flows in the pre-control era. For non-FDI net flows as a whole, we find the controls have had no statistically significant impact.

43. Capital controls have not affected the level of the exchange rate, but have been associated with an increase in volatility. Exchange rate levels do not appear to have been affected by the capital controls, regardless of the time period used for estimating the model. In addition, capital controls have not reduced the sensitivity of the exchange rate to global financial developments, suggesting that controls on capital inflows did not reduce Colombia’s vulnerability to external shocks. We also consistently find that controls have actually increased the volatility of the exchange rate, which over the past two years has well exceeded the Latin American average.

44. There is no robust evidence that the capital controls contributed to enhancing monetary policy autonomy. The econometric estimates of the effect of capital controls on the sensitivity of the exchange rate to interest rate differentials are sensitive to the estimation period. For the entire sample period, from end-July 2006 to early July 2008, capital controls do not appear to have had an impact on the effect of interest rate differentials on exchange rate determination. If the estimation period begins in January 2007, however, capital controls are found to have a significant impact on the interface between interest rate differentials and the exchange rate.

45. Further research is needed to fully understand the effects of the controls. Additional study of the channels through which carry trades are executed could provide important insights into the effectiveness of capital controls. In particular, the role of the offshore sector and pension funds in the exchange rate market could be more fully explored. Further empirical work on the effects of capital controls on the volatility of the exchange rate is also warranted. Given the sensitivity of some of the econometric results to the choice of sample, it would be also useful to extend the analysis to a longer period, especially covering the experience of the 1990s. In addition, there is room to use different measures of capital controls in future empirical work. In particular, a “tax equivalent” measure—which quantifies the implicit premium that must be paid on local returns to compensate investors for the financial costs associated with capital controls—could be utilized (Villar and Rincón, 2001; Concha and Galindo, 2008). A measure of the effect of capital controls could also be derived from information on banks’ gross derivatives positions, in particular how closely these have come to the ceiling of 550 percent of gross capital.

Appendix I: The Effectiveness of Capital Controls in Colombia: Insights from Previous Research

Empirical studies find that capital controls have had mixed success in emerging market countries.22 As suggested by Magud and Reinhart (2007), the effectiveness of capital controls can be assessed from the standpoint of their ability to (1) limit capital inflows; (2) change the composition of flows (in particular, by shifting flows to long-term assets); (3) alleviate real exchange rate pressures; and (4) provide more independence for monetary policy by severing the link between domestic and international interest rates. Studies in this area have confronted a minefield of methodological issues, and not always with great success (Magud and Reinhart, 2007; International Monetary Fund, 2007). Among these challenges is the difficulty of isolating the effects of capital controls per se on capital flows and exchange rates, as well as endogeneity issues. The area where capital controls have been most successful is in providing more autonomy for monetary policy and altering the composition of capital inflows, while success on reducing the volume of inflows and reducing exchange rate pressures has been mixed.

Previous research on the experience of Colombia in the 1990s with capital controls is broadly consistent with these findings. Most studies, but not all, have concluded that capital controls have been ineffective in reducing inflows or influencing the real exchange rate, but did affect the composition of inflows and monetary policy autonomy.23 Cárdenas and Barrera (1997) find that capital controls between 1985 and 1995 had no impact on total capital flows but changed their composition. A similar finding is reported in Rocha and Mesa (1998) who argue that capital controls reduced the short-term interest rate differential, increasing medium- and long-term foreign debt. Using a data set that covers the period between 1993 and 1998, Ocampo and Tovar (1999, 2003) argue that capital controls diminished not only short-term but also long-term capital flows, while Rincón (2000) finds that capital controls reduced short-term flows. Using a similar data set, Villar and Rincón (2001) find that capital controls allowed the authorities to increase domestic real interest rates without creating pressures for real appreciation of the peso.

Three recent studies on the 2007 controls arrive at mixed results. Cardenas (2007) examines the effects of the 2007 controls, using weekly data on total capital inflows on assets of more than one year of maturity. Using ordinary least squares, and controlling for other determinants of inflows such as interest rate differentials, he finds that controls had no statistically significant effect on capital inflows. This approach, however, may be biased towards finding that capital controls are ineffective, as the dependent variable—total capital inflows of more than one year—misses the effect of controls on inflows seeking to purchase short-term assets. In addition, the inclusion of foreign direct investment (FDI) is problematic, as FDI flows are unlikely to be influenced by the same factors that drive portfolio inflows, such as interest rate differentials. As such, the model is unlikely to find that the controls affect total inflows, even if they have a tangible effect on portfolio inflows. Galindo (2007) and Concha and Galindo (2008), using a vector error correction model (VECM) with monthly data, find that capital controls had a significant effect on short-term capital inflows, but not the real exchange rate. They find that capital controls do, however, reduce the volatility of the real exchange rate.

References

  • Bollerslev, Tim, and Jeffrey Wooldridge, 1992, “Quasi-Maximum Likelihood Estimation and Inference in Dynamic Models with Time-Varying Convariances,” Econometric Reviews, Taylor and Francis Journals, Vol. 11, pp. 143172.

    • Search Google Scholar
    • Export Citation
  • Braumoeller, Bear, 2004, “Hypothesis Testing and Multiplicative Interaction Terms,” International Organization, Vol. 58, pp. 807820.

    • Search Google Scholar
    • Export Citation
  • Cardenas, Mauricio, and Felipe Barrera, 1997, “On the Effectiveness of Capital Controls: The Experience of Colombia during the 1990s,” Journal of Development Economics, Vol. 54, pp. 2757.

    • Search Google Scholar
    • Export Citation
  • ——— 2007, “Controles de Capitales en Colombia: ¿Funcionan o No?” Debate de Coyuntura Económica, December (Bogotá, Colombia, Fedesarrollo).

    • Search Google Scholar
    • Export Citation
  • Concha, Alvaro, and Arturo Galindo, 2008, “An Assessment of Another Decade of Capital Controls in Colombia: 1998–2008.” Paper presented at the XIII LACEA Meeting (Rio de Janeiro: Brazil).

    • Search Google Scholar
    • Export Citation
  • David, Antonio, 2007, “Are Price-Based Capital Account Regulations Effective in Developing Countries?,” World Bank Policy Research Paper Working Paper No. 4175 (March).

    • Search Google Scholar
    • Export Citation
  • Edwards, Sebastián, ed., 2007, Capital Controls and Capital Flows in Emerging Economies: Policies, Practices, and Consequences (Chicago: The University of Chicago Press).

    • Search Google Scholar
    • Export Citation
  • Edwards, Sebastián, and Roberto Rigobón, 2005, “Capital Controls, Exchange Rate Volatility and External Vulnerability,” NBER Working Paper No. W11434 (June).

    • Search Google Scholar
    • Export Citation
  • Galindo, Arturo 2007, “Controles de Capitales en Colombia: ¿Funcionan o No?,” Debate de Coyuntura Económico, December (Bogotá, Colombia: Fedesarrollo).

    • Search Google Scholar
    • Export Citation
  • Gómez, Carlos, Diego Jara, and Andrés Murcia, 2006, “Impacto de Las Operaciones de Los Fondos de Pensiones Obligatorias en Los Mercados Financieros Colombianos,” Borradores de Economía No. 406 (Bogotá, Colombia: Banco de la República de Colombia).

    • Search Google Scholar
    • Export Citation
  • International Monetary Fund, 2007, World Economic Outlook, October, (Washington: International Monetary Fund).

  • Kamil, Herman, 2008, “Is Central Bank Intervention Effective Under Inflation Targeting Regimes? The Case of Colombia,” IMF Working Paper 08/88 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Kamil, Herman, and A. Reveiz, 2008, “Carry Trades, Special Vehicle Purposes and Derivatives Markets in Colombia: Mechanisms and Policy Implications,” (Unpublished, Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Magud, Nicolas, and Carmen Reinhart, 2007, “Capital Controls: An Evaluation,” in S. Edwards eds., Capital Controls and Capital Flows in Emerging Economies: Policies, Practices, and Consequences (Chicago: The University of Chicago Press).

    • Search Google Scholar
    • Export Citation
  • Ocampo, José Antonio, and Camilo Ernesto Tovar, 1999, “Price-Based Capital Account Regulations: The Colombian Experience,” CEPAL, Serie Financiamento del Desarrollo (Santiago, Chile).

    • Search Google Scholar
    • Export Citation
  • Ocampo, José Antonio, and Camilo Ernesto Tovar, 2003, “Colombia’s Experience with Reserve Requirements on Capital Inflows,” CEPAL Review, No. 81 (December), pp. 731.

    • Search Google Scholar
    • Export Citation
  • Rocha, Ricardo, and Ricardo Mesa, 1998, “Flujos Internacionales de Capital en Colombia: Un Enfoque de Portafolio,” Archivos de Macroeconomia, No. 78 (Bogotá, Colombia: Departamento Nacional de Planeación)

    • Search Google Scholar
    • Export Citation
  • Rincón, Hernán, 2000, “Efectividad del Control a Los Flujos de Capital: Un Reexamen Empírico de la Experiencia Reciente en Colombia,” Revista de Economía de la Universidad del Rosario (February), pp. 1536.

    • Search Google Scholar
    • Export Citation
  • Uribe, José Dario, 2005, “Capital Controls and Foreign Exchange Market Intervention in Colombia,” BIS Paper No. 23 (May), pp. 12840.

    • Search Google Scholar
    • Export Citation
  • Vargas, Hernando, and Carlos Varela, 2008, “Capital Flows and Financial Assets in Colombia: Recent Behavior, Consequences and Challenges for the Central Bank,” Borradores de Economía No. 502 (Bogotá, Colombia: Banco de la República de Colombia).

    • Search Google Scholar
    • Export Citation
  • Villar, Leonardo, and Hernán Rincón, 2001, “Flujos de capital y regímenes cambiarios en la década de los 90,” Ensayos sobre Política Económica, No. 39 (Bogotá, Colombia: Banco de la República de Colombia).

    • Search Google Scholar
    • Export Citation
1

Prepared by Benedict Clements and Herman Kamil.

2

From mid-December 2004 through mid-June 2006, capital controls were also imposed in the form of a minimum one-year stay for nonresident portfolio inflows.

3

See Kamil (2008) for an analysis of the effectiveness of intervention policies in Colombia during this period. Kamil and Reveiz (2008) provide further elaboration on the relationship between sterilized intervention and carry trade in Colombia.

4

This followed less important extensions of the controls in end-April that broadened the coverage of transactions classified as foreign borrowing (and hence subject to the URR). These included (1) import financing of greater than six months (financing of capital goods and loans under US$10,000 were excluded); and (2) credit obtained from abroad by Colombian firms for the purpose of financing outward FDI.

5

This can be seen be the high correlation (about 90 percent) between Colombian resident portfolio inflows and outflows.

6

This pre-controls era captures a period when international markets were relatively calm and during which interest rate differentials in Colombia, vis-à-vis the rest of the world, started to rise. Thus, it represents a good period for establishing a baseline before the start of the capital controls.

7

This reached US$800 million in July 2008.

8

Another reason may have to do with the implementation of an unremunerated reserve requirement on new deposits that was also implemented in May 2007. While not part of the capital controls per se, this encaje marginal constrained the ability of local banks to offer forwards to pension funds, as it reduced their liquidity and thus their ability to hedge their forward currency transactions.

9

Following the weakening of commodity prices and global financial turbulence, the peso, like many other emerging market currencies, depreciated sharply in the second half of 2008.

10

For total non-FDI, we also analyze the determinants of gross outflows.

11

For total non-FDI, foreign borrowing, and the net drawdown of bank accounts abroad, the dummy takes on a value of one from May 6, 2007 onward. For portfolio flows, the starting date is May 23.

12

In all cases a GARCH (1,1) model was estimated as it is typical in the literature. Given the presence of heteroskedastic disturbances in our sample, we use the methods described in Bollersev and Wooldridge (1992) to compute standard errors.

13

The commodity price index was constructed as a weighted average of daily commodity prices for the main commodities exported by Colombia, using weights from 2006. The variable EMBI enters the equation in first differences as this increases the explanatory power of the model significantly.

14

Given the reduced-form nature of the estimation, the framework can only identify the average response of exchange rate returns to the imposition of capital controls. It does not identify a structural relationship, or the channels through which these restrictions affect exchange rates.

15

We also estimated the model with all regressors lagged one period, in an attempt to reduce simultaneity bias. The main results were unaltered. The results were also similar when the estimated over the period January 2007–July 2008. Results are available upon request.

16

During the peak period of appreciation pressures—during the first four months of 2007—these inflows, in net terms, equaled ¾ percentage point of annualized GDP.

17

Non-resident gross portfolio inflows averaged about 5 percent of total gross capital inflows in the pre-controls era.

18

In the results presented below, central bank intervention is expressed per US$100 million.

19

Given that an increase in interest rates typically leads to an appreciation of the currency (i.e., it has a negative coefficient), then a decrease in the sensitivity under capital controls would imply that the sign of the coefficient in the interaction term is positive. In this case, the sum of the lower-order coefficient on the interest spread and the corresponding interaction term (which measures the effect of interest rates on the exchange rate during the capital controls period) would be reduced in absolute value.

20

We cannot reject the null hypothesis that all interaction terms in the variance equation in the augmented model (column 2) are jointly equal to zero. As such, the model suggests that the effect of the key explanatory variables on the variance of the exchange rate did not change with the imposition of controls.

21

Caution must be exercised when interpreting the coefficients of an interaction model. In particular, the sign, magnitude, and significance of the effect of capital controls are given by the estimate on the dummy DCC and the other coefficients with interactive terms. The total effect depends on the assumed values of the explanatory variables that are part of the multiplicative terms. See Braumoeller (2004).

22

For cross country reviews of the evidence on capital controls, see also IMF (2007), Magud and Reinhart (2007), and Edwards (2007).

23

See also Uribe (2005) for a review of the Colombian literature. For a minority view—arguing that capital controls were effective in influencing both short and long-term capital flows—see David (2007).

Colombia: Selected Issues
Author: International Monetary Fund