How Do Exchange Rate Regimes Affect Firms' Incentives to Hedge Currency Risk? Micro Evidence for Latin America
Author: Herman Kamil

Contributor Notes

Author’s E-Mail Address: hkamil@imf.org

Using a unique dataset with information on the currency composition of firms' assets and liabilities in six Latin-American countries, I investigate how the choice of exchange rate regime affects firms' foreign currency borrowing decisions and the associated currency mismatches in their balance sheets. I find that after countries switch from pegged to floating exchange rate regimes, firms reduce their levels of foreign currency exposures, in two ways. First, they reduce the share of debt contracted in foreign currency. Second, firms match more systematically their foreign currency liabilities with assets denominated in foreign currency and export revenues--effectively reducing their vulnerability to exchange rate shocks. More broadly, the study provides novel evidence on the impact of exchange rate regimes on the level of un-hedged foreign currency debt in the corporate sector and thus on aggregate financial stability.

Abstract

Using a unique dataset with information on the currency composition of firms' assets and liabilities in six Latin-American countries, I investigate how the choice of exchange rate regime affects firms' foreign currency borrowing decisions and the associated currency mismatches in their balance sheets. I find that after countries switch from pegged to floating exchange rate regimes, firms reduce their levels of foreign currency exposures, in two ways. First, they reduce the share of debt contracted in foreign currency. Second, firms match more systematically their foreign currency liabilities with assets denominated in foreign currency and export revenues--effectively reducing their vulnerability to exchange rate shocks. More broadly, the study provides novel evidence on the impact of exchange rate regimes on the level of un-hedged foreign currency debt in the corporate sector and thus on aggregate financial stability.

I. Introduction

Firms’ heavy reliance on foreign-currency debt has been singled out as a key source of financial fragility in emerging market countries (Calvo (2001) and Frankel (2005)). In the late 1990s and during the last decade, sharp currency depreciations in several countries in Latin America and East Asia turned firms’ foreign currency liabilities into crippling debt burdens—especially for those companies that had a large mismatch between foreign-currency denominated debt and earnings in local currency. This, in turn, impaired many firms’ ability to service debt or drove them into bankruptcy, triggering banking crisis and economic recessions.

An intense debate on academic and policy circles in the aftermath of these crises has centered on the extent to which the exchange rate regime affects firms’ incentives to hedge their exposure to currency risk. In particular, the debate has focussed on whether fixed or pegged regimes encourage firms to take on excessive foreign currency debt that leaves them exposed to a sudden depreciation of the domestic currency. This debate has considerable relevance for economic policy, as foreign currency indebtedness has been cited as a key factor deepening output contractions during emerging market crisis, with persistent effects on economic growth (Doma ç and Martinez-Peria (2003), Calvo et al. (2004), Cerra and Saxena (2008), and Bordo, Meissner, and Stuckler (2009)). It is especially relevant today for Central and Eastern Europe, where the growing exposure of the private sector to currency risk has been highlighted as having potentially significant implications for financial stability (Rosenberg and Tirpak (2009), Ranciere, Tornell, and Vamvakidis (2010), and Zettelmeyer, Nagy, and Jeffrey (2010)).

Economists are sharply divided over the role of the exchange rate regime in contributing to currency mismatches in firms’ balance sheets. Views here fall into two camps. Proponents of flexible regimes argue that authorities’ commitment to defend a peg offer the private sector an implicit guarantee against short-term movements of the exchange rate, which leads to moral hazard and excessive foreign currency borrowing (see Burnside, Eichenbaum, and Rebelo (2001), Goldstein and Turner (2004), and Schneider and Tornell (2004), among others). A natural implication of this line of thought is that the adoption of floating exchange rate regimes would provide incentives for a more cautious management of currency exposure, thereby reducing financial vulnerabilities associated to currency mismatches in the private sector.

Other authors, however, have claimed that the problem of unhedged foreign currency liabilities in the corporate sector has deeper roots than the exchange rate regime (Eichengreen and Hausmann (1999) and Eichengreen, Hausmann, and Panizza (2005)). According to this view, currency choice is not the result of a market equilibrium but of financial market incompleteness, and thus monetary or exchange rate policies are ineffective in reducing vulnerabilities related to foreign exchange exposures. 1

Fueling this theoretical debate is the mixed empirical evidence on whether pegged regimes contribute to higher debt dollarization and currency mismatches in firms’ balance sheets.2 Martinez and Werner (2002) conclude that the exposure of Mexican firms to devaluation risk lessened after Mexico switched to a flexible regime. Using firm level data for Chile, Cowan and De Gregorio (2007) also find that currency exposure of Chilean corporates fell after 1999 when Chile moved to a floating regime. In the case of Asia, Parley and Popper (2006) find that firms in East Asia were less hedged under pegged exchange rates, while Patnaik and Shah (2010) find that low currency flexibility encouraged Indian companies to hold larger unhedged currency exposures than during periods of greater currency flexibility.

On the contrary, firm level evidence from Brazil over the period of 1996 to 2001 suggests that the collapse of the fixed exchange rate regime in Brazil, in early 1999, did not cause a major change in the currency composition of debt and the hedging behavior of the corporate sector (Berrospide, Purnanandam, and Rajan (2008)). Using bank-level data, Arteta (2005) finds that floating exchange rate regimes have been, if anything, consistently associated with greater currency mismatches in a sample of developing and transition economies from the early 1990s to 2000. Thus, whether exchange rate regimes reduce or exacerbate financial fragility by encouraging excessive foreign currency borrowing is far from settled.

Against this background, the goal of this paper is to investigate whether the choice of exchange rate regime, and the degree of exchange rate flexibility, affects firms’ foreign currency borrowing decisions and the associated currency mismatches in their balance sheets. For these purposes, I use a new hand-collected dataset with information on the foreign-currency share of firms’ assets and liabilities, and the breakdown of sales into domestic and export revenues, for a large sample of firms in six Latin American countries between 1992 and 2005. To my knowledge, this is the first cross-country dataset of emerging market firms to include information on the currency composition of both sides of companies’ balance sheet.3

Estimating the effect of exchange rate flexibility on firms’ foreign currency debt choices (and currency mismatches) is problematic due to the possible two-way causality.4 As noted, authorities’ commitment with exchange rate stability may encourage foreign currency borrowing if the central bank is perceived as effectively providing free currency risk insurance to the private sector. In turn, high levels of liability dollarization may make authorities more reluctant to allow the exchange rate to float fearing the balance sheet impacts of real exchange rate movements—giving rise to a spurious negative association between exchange rate flexibility and financial dollarization.5

To identify the causal effect of exchange rate regimes, I compare the cross-sectional changes in firms’ foreign currency debt holdings after the adoption of floating exchange rate regimes, across companies with differing abilities to generate foreign currency revenues. If floating exchange rate regimes lead to better currency risk management, firms with lower foreign currency buffers (that is, with little foreign currency earnings or assets denominated in foreign currency) should experience larger declines in dollar debt relative to those firms selling mostly to international markets or holding larger amounts of foreign currency assets.

The empirical evidence yields three key results. First, I find strong evidence of a persistent decline in firms’ foreign currency borrowing in response to the adoption of a flexible exchange rate regime. The switch to a floating regime is associated with a statistically and economically significant reduction of 7 percent on corporate debt dollarization, on average, compared with pegged regimes. Second, consistent with a causal effect of flexible regimes in reducing currency mismatches, I find that after countries switch to flexible exchange rate regimes, firms with lower natural currency buffers experienced larger declines in dollar debt relative to firms that rely principally on export revenues or have large dollar asset holding. That is, beyond reducing the share of debt denominated in dollars, flexible exchange rate regimes seem to provide incentives for firms to match more strongly the currency composition of liabilities with the currency denomination of assets and cash flows, in a way that effectively reduces the balance-sheet effects of large devaluations.

To explore the robustness of these findings, I verify that these results hold up under a wide variety of conditions and econometric specifications. Within a panel framework, I demonstrate that the results are robust to different methods for classifying exchange rate regimes and measuring exchange rate flexibility, potentially confounding macroeconomic influences, and are not driven by changes to regulations on banks’ foreign currency lending. In addition, I also use an event study approach around exchange rate regime changes, and compare cross-sectionally the changes in firms’ foreign currency debt holdings for firms with differing levels of foreign currency buffers. Again, I find that after the adoption of flexible exchange rate regimes, firms reduce their unhedged foreign currency exposures by using more systematically their assets in foreign currency to offset their dollar debt risk.

The effects of exchange rate regimes on firms’ foreign currency hedging—if present—should be particularly manifest in firms that are significantly exposed to currency risk. To test this hypothesis, I use a censored quantile regression approach to look at the effects of the exchange regime at various points of the cross-sectional distribution of firms’ dollar debt ratios. My third finding is flexible exchange rate regimes have a larger impact in reducing currency mismatches in firms in the upper tail of the conditional distribution of dollar debt ratios, i.e., those more exposed to exchange rate risk.6

The findings in the paper are consistent with, and complementary to, several strands of literature. First, results in this paper are related to previous studies of the determinants of capital structure financial decisions and risk exposures for companies in emerging markets (Allayannis, Brown, and Klapper (2003), Schmukler and Vesperoni (2006), and Desai, Foley, and Hines (2008), among others). The empirical findings are novel because I examine emerging market firms’ financing patterns simultaneously across a number of countries and years with a broad coverage of firms and detailed information of the currency and maturity composition on both sides of their balance sheets. Second, the study contributes new evidence on the impact of exchange rate regimes on the level of unhedged foreign currency debt in the corporate sector and thus on aggregate financial stability (Céspedes, Chang, and Velasco (2004)). My results support the view that fixed or pegged regimes bias corporate borrowing towards foreign currency denominated debt, and is in line with previous country-level evidence for Mexico from Martinez and Werner (2002) and for Chile from Cowan, Hansen, and Herrera (2005).7 In addition, I provide evidence of a significant change in the way firms match the currency structure of assets and liabilities after the adoption of flexible regimes, such that they reduce their level of unhedged foreign currency borrowing and become less exposed to exchange rate depreciations. Taken together, the preponderance of evidence presented in the paper supports the notion that higher exchange rate flexibility provides stronger incentives for firms to reduce their balance sheet currency exposure and take-on less risky financing structures.8

The remainder of the paper is organized as follows. Section II provides a brief overview of the literature. Section III describes the data and provides a first glance at the evolution of firms’ foreign currency borrowing and foreign currency exposures around exchange rate regime switches. Section IV discusses the empirical strategy and presents the baseline panel regression results. Section V shows that the baseline results are robust, while Section VI discusses competing or alternative explanations. Section VII uses a quantile regression approach to provide a more nuanced picture of the extent to which the exchange rate regime affects firms’ incentives to hedge their exposure to currency risk. The last section concludes.

II. Theoretical Literature Review

How does the exchange rate regime affect firms’ incentives to hedge their exposure to currency risk? This question has been at the center of the debate over optimal exchange rate regimes in emerging markets since the financial crises of the 1990s exposed the perils of unhedged foreign currency debt. 9 Yet there is no clear consensus among economists on whether the type (or degree of flexibility) of the exchange rate regime affects the corporate sector’s incentives to take on foreign currency denominated liabilities or to insure against depreciation risk.

Two basic views exist in this respect. On the one hand, several authors have argued that pegged exchange rate regimes biases corporate borrowing towards foreign currency, due to an implicit exchange rate guarantee given by the government (see, among others, Mishkin (1996) and Goldstein and Turner (2004)). Under fixed or pegged regimes, the central bank keeps currency volatility within a pre-announced range, effectively underwriting currency risk (Dooley (2000)).10 Thus, firms borrow in dollars to benefit from the lower ex ante dollar interest rates, and expect the government to insure them from any potential loss in the event of a large devaluation.11 A second variant of this argument suggests that because of limited exchange rate volatility under fixed or tightly managed exchange rate regimes, borrowers appear to consider a steep devaluation a low-probability event, and therefore neglect or underestimate the exchange rate risk associated with borrowing in foreign currency.12 The fact that fixed/pegged exchange rates have played a role in every recent financial crisis since 1994, and that firms relied extensively on unhedged foreign currency financing in the years leading up to the crisis, is often used as strong evidence for these views.13

On the other hand, Eichengreen and Hausmann (1999) and Eichengreen, Hausmann, and Panizza (2005) dispute this view. The authors suggest that at the root of currency mismatches lies the fundamental inability of emerging markets to borrow abroad in their own currency. Inevitably, this leads to an accumulation of foreign-currency denominated debt which firms are simply unable to hedge, even if they have the foresight or prudence to match the currency structure of their assets and liabilities.14 In addition, McKinnon and Pill (1999) argue that adopting a floating rate regime will actually exacerbate currency mismatches. Because the domestic interest rate risk premium is a direct function of the stability of the currency, exchange rate volatility associated with floating rates will increase domestic interest rates (and thus the incentives to borrow in foreign currency) and make financial hedging more expensive.15

III. Data Set and Basic Stylized Facts

This section describes the data used and documents a new set of stylized facts on the evolution of firms’ share of foreign currency debt and associated balance sheet currency exposures around exchange rate regime switches.

A. The Firm-Level Dataset

The empirical analysis in this paper draws on a new database with annual accounting information for over 1,800 non-financial companies in six Latin American countries, spanning the period 1992 to 2005.16 The countries covered are: Argentina, Brazil, Chile, Colombia, Mexico, and Peru. A distinct feature of this dataset is that it contains detailed information on the currency and maturity structure of firms’ liabilities and assets, firms’ access to foreign currency revenues, and their ability to tap international capital markets. To my knowledge, this is the first cross-country dataset of emerging market firms to include information on the currency composition of both sides of companies’ balance sheet.17

The data were assembled from different sources. Financial statement data were drawn from local stock markets or regulatory agencies in each country.18 Data on foreign currency liabilities and assets were hand-collected from the financial explanatory notes of firms’ balance sheets. These include all assets or liabilities outstanding which are denominated in—or indexed to—foreign currency, issued domestically or abroad.19 Information on firms’ export revenues was obtained from income statement data. When this was not available, I used countries’ customs office records or Central Bank’s Balance of Payments trade registries. Together with information on firms’ dollar assets, the export to sales ratio captures the degree to which a company is well-positioned to capitalize on exchange-rate depreciation.20 This is a substantial improvement over previous studies in the literature that have typically used aggregate variables to proxy for firms’ access to foreign currency revenues, i.e., binary tradable/non-tradable classification or sectoral export shares. Finally, I merge the balance sheet information with firm-level issuance data on external bonds, loans and equity from Dealogic to capture firms’ access to international capital markets.

All firms in the sample are non-financial, publicly-traded companies, except for the case of Argentina, where roughly half of the firms are not publicly traded. Focusing mostly on publicly listed firms was dictated by data availability, and has the drawback that the patterns observed for publicly-traded firms might not be representative of the corporate sector as a whole. Yet it has the advantage that financial statistics are typically more accurate and comprehensive.

Moreover, relative to other available databases, such as Worldscope, the coverage of small and medium-sized publicly traded firms is significantly wider.21

Most of the variables are expressed as ratios; where this is not the case (firm size), I deflate the nominal magnitudes with 2000 values using December-to-December changes in the consumer price index and convert them to U.S. dollars using December 2000 market exchange rates. To ensure that results are not driven by outliers, I treat the data following common procedures in the literature. I dropped all firm/year observations for explanatory variables that exceeded the sample mean by more than five standard deviations.22 These exclusions leave us with complete information for an unbalanced panel of 12,223 firm-year observations, which consist of 1470 firms with an average of around 8 years each.

Table 1 shows the number of firm observations per country and year that have non-missing data on foreign-currency debt and export sales. The size of the sample changes as new firms enter and exit the sample. Attrition is mostly due to the fact that some public firms that are privatized, merged, or acquired are subsequently delisted. Few firms drop from the sample because of bankruptcy.

Table 1.

Number of Observations used in Empirical Analysis

article image
Source: Author’s calculations based on data described in Appendix 1.Note: The first panel depicts the number of firms in the sample for each country containing consistent balance sheet and income statement data, and information on the currency composition of debt and sales. The second panel breaks down the firms in the sample by economic sector.

Table 2 and Figure 1 report descriptive statistics for the variables used in the analysis. Inspection of Table 2 reveals significant cross-country variation in the currency denomination of corporate borrowing. The average share of foreign currency debt during this period ranged from 6 percent in the case of Colombia, to close to 60 percent in the cases of Argentina and Peru. Figure 1, in turn, shows the cross-sectional distribution of foreign currency debt ratios within each country for the whole sample. Again, differences across countries are striking. The data for Argentina and Peru are consistent with the fact that debt dollarization has been pervasive in all productive sectors. For several countries in the sample, however, the cross-sectional distribution of dollar debt ratios is highly skewed—clustered around zero and decidedly non-normal. In addition, and as shown in Table 2, a common pattern in firm capital structures across Latin America is the relatively low dollarization of assets (compared to liabilities). Finally, the average share of exports over sales for firms in the sample show less variation across countries.

Table 2.

Descriptive Statistics for Full Sample

article image
Source: Author’s calculations.Note: The table shows firm-level averages for each country for the variables used in the paper. The detailed information about variable definition is contained in the appendix. All variables expressed in percent, except where noted.

B. Exchange Rate Regimes

To measure the choice of exchange rate regime I rely on the de facto annual classification produced by the International Monetary Fund (IMF). As described by Bubula and Otker–Robe (2002), this classification combines market exchange rates, reserves data, and other quantitative information with the existence of formal or informal commitments to exchange rate paths assessed by IMF staff. Based on this classification, I construct a binary variable that takes the value of 0 for fixed, pegged, or crawling exchange rate regimes, and 1 for independently floating regimes.23 During the sample period, we observe six regime switches from fixed or pegged to floating regimes: Mexico in 1994, Brazil, Chile, Colombia, and Peru in 1999, and Argentina in 2002.24

Figure 1.
Figure 1.

Cross-Sectional Distribution of Dollar Debt Ratios Within Countries

Citation: IMF Working Papers 2012, 069; 10.5089/9781463939052.001.A001

Source: Author’s calculations.Note: For each country, the figure plots the histogram of the foreign currency debt ratios for the pooled sample of firm-year observations. The y-axis shows the frequency of firm-year observations for each interval of foreign currency debt share (expressed in percentage in the x-axis).

Because this binary variable may mask the heterogeneity of exchange rate policies across countries and time, I also construct a “Freedom to Float” index as in Calvo and Reinhart (2002). The index measures the variability of the rate of depreciation of the nominal exchange rate, relative to the sum of the variability of net international reserves and the variability of short term interest rates. It quantifies the extent to which the central bank chooses not to intervene to stabilize the exchange rate for a given level of pressure on its currency: a higher number indicates that the exchange rate is relatively more volatile than the policy instruments, thus indicating a more flexible exchange rate policy. Table 3 provides a description of the different exchange rate arrangements for each country during the sample period, and calculates measures of effective exchange rate flexibility as in Calvo and Reinhart (2002).

Table 3.

Exchange Rate Regimes and Measures of Exchange Rate Flexibility Within Regimes

article image
Source: Author’s calculations.

Crawling peg for Brazil starting from July 1994.

Crawling band/crawling peg for Mexico ends in November 1994. Independent floating beginning in December 1994.

Note: This table reports the de-facto exchange rate regime for each country and period based on the classification described in Bubula and Otker-Robe (2002). The Freedom to Float indicator is calculated using Calvo and Reinhart (2002) measure of fear of floating. A higher value denotes more flexibility.

C. A First Glance at the Data

As a first step in assessing the relationship between exchange rate regimes and currency mismatches, Figure 2 plots the average share of firm’s dollar liabilities in total liabilities for each country in the sample.25 In each panel, the vertical line represents the year each country switched to more flexible exchange rate regimes.

Several important trends are visible in the data. Foreign currency debt as a share of total debt of nonfinancial firms rose sharply during the 1990s and then began to fall rapidly, typically when countries introduced flexible exchange rate regimes. This general pattern is observed both in countries with relatively low foreign currency borrowing as in Brazil and Colombia (which actively discouraged financial dollarization), as well as in the highly dollarized countries in the sample, Argentina and Peru. As a result, corporations in Latin America have become significantly less dependent on foreign currency financing: the average share of foreign-currency denominated liabilities in Latin America dropped from 35 percent 1998 to 19 percent in 2005. 26 Excluding the special case of Argentina, Mexico and Peru are the countries where dollarization of corporate liabilities have decreased the fastest, compared with their peak levels during the 1990s.27

Figure 3 depicts a more precise measure of currency exposure, defined as dollar debt as a percentage of exports plus dollar assets.28 Overall, the data suggests that—accompanying the sustained decrease in dollarization levels—firms built up considerable foreign exchange buffers, by matching more strongly their dollar liabilities with export revenues and assets denominated in foreign currency, especially since the on-set of more flexible regimes. As the graphs illustrate, in the cases of Brazil, Chile, and Colombia, the sum of firm-level exports and dollarized assets is, on average, larger than foreign currency liabilities in the most recent period.29

Figure 2.
Figure 2.

Dollarization of Liabilities of the Corporate Sector in Latin America

(average across firms, in percent)

Citation: IMF Working Papers 2012, 069; 10.5089/9781463939052.001.A001

Source: Author’s calculations based on data described in Appendix I.Note: To control for changes in sample composition, we regress firm-level dollarization ratios on a complete set of firm and year fixed effects. The graphs plot the estimated time dummies from these country level regressions.
Figure 3.
Figure 3.

Dollar Debt as a Share of Natural Currency Hedges

(median across firms, in percent)

Citation: IMF Working Papers 2012, 069; 10.5089/9781463939052.001.A001

Source: Author’s calculations based on data described in Appendix I.

While these raw facts are indicative of a significant economic effect of flexible exchange rate regimes on firms’ balance sheet currency mismatches, one must be cautious in interpreting this as a causal link, due to the possible presence of omitted factors correlated both with currency exposure at the firm level and exchange rate regime switching. The rest of the paper is devoted to exploiting the panel structure of the dataset, which helps eliminate the potential confounding effects that unmeasured firm characteristics and common shocks across Latin American countries may have.

IV. Baseline Results

In this section, I describe the empirical strategy and present the baseline results from the panel data regressions.

A. Dollarization Levels in Flexible versus Pegged Regimes

The empirical strategy is based on estimating a pooled cross-section model of the main determinants of firms’ foreign currency borrowing in Argentina, Brazil, Chile, Colombia, Mexico, and Peru between 1992 and 2005. I begin the analysis by looking at whether dollar debt ratios were lower after countries switched to more flexible regimes. I estimate the following panel regression.

DOLLijct=β0+β1FLEXDatect+β2PostFLEXct+β3EXPtoSijct+β4DOLAijct1+γXijct1+Φj+χc+λt+εijct(1)

Equation (1) represents a reduced form equation which models DOLL ijct, the share of total liabilities denominated or indexed to a foreign currency (typically the dollar) of firm i in sector j, in country c in year t. Thus, DOLL ijct is between 0 and 1.FLEXDate is a binary dummy variable that takes the value of 1 the year a country switches to a flexible regime and 0 otherwise. PostFLEX is a binary variable that also varies across countries and time, and takes on the value of 1 after the year the country switched to floating exchange rates. EXPtoS is the ratio of exports (foreign currency revenues) to sales of each firm in the sample. DOLA is the share of a firm’s assets denominated in foreign currency, lagged one period. These two variables measure a firms’ ability to generate dollar revenue for repayment of their foreign currency debt obligations. The estimated equation also controls for a vector of other firm-specific, time-varying covariates lagged one period, xijt -1, as described below.30 Standard errors are clustered by country and year to take into account the level of aggregation of the exchange rate regime indicator (Bertrand, Duflo, and Mullainathan (2004)).

The specification includes an extensive set of fixed effects. First, I allow for time fixed effects, λt, to account for regional changes that affect all countries and firms equally (for example, world business cycles or other events, such as the Asian and Russian crises that have affected world financial markets). The high degree of commonality observed in the time series behavior of dollarization levels across countries described in the previous section suggests that global or regional factors may be partially driving the behavior of dollarization. Further, I also control for time-invariant country-specific factors (χc) which may drive cross-country differences in foreign currency borrowing—such as restrictions on dollar intermediation in the domestic banking systems in Brazil and Colombia.31 In addition, I group firms into eight categories based on their primary industry classification, and include industry dummies, øj, in the empirical model. Industry fixed effects control for systematic differences across economic sectors that might determine dollarization levels—like competition structure, openness, or liquidity needs.

Finally, I also include a set of additional variables that have been shown in the literature to be important determinants of debt dollarization.32 To capture firm size, I sort firms in each country and year into thirds based on total assets. Separate dummies are used for large-sized (top-third) and medium-sized (middle-third) firms (small-sized firms being the excluded category). 33 To account for firms’ ability to raise funds in international capital markets, I construct two binary dummy variables: one that takes the value of 1 starting in the year a firm issues bonds or syndicated loans in international markets (and 0 otherwise), and a second one that takes the value of 1 starting in the year a firm issues equity abroad in the form of ADRs (and 0 otherwise).

Overall, this specification captures most of the factors that are likely to affect a firm’s share of debt denominated in foreign currency and greatly attenuate problems of omitted variable bias. In particular, this differences-in-differences approach allows estimating the effect of the exchange rate regime while holding constant fixed characteristics of a country that affect dollarization and might also be correlated with the timing of flexible regimes introduction.

Table 4 presents the main results.34 Given that observations for the dependent variable are censored by 0 and 1, I use Tobit regressions to estimate the model. I find that the introduction of more flexible exchange rate regimes leads to a sharp decline in firm-level liability dollarization. Column (1) shows that the switch to a flex regime is associated with a statistically significant decrease of 7 percent on the share of debt denominated in foreign currency, on average, compared with pegged regimes. The results confirm the negative relationship illustrated in Figure 2 in section III.C.

Table 4.

The Effects of Exchange Rate Regimes on Firm-Level Dollarization

article image
Source: Author’s calculations.Note: ‘This table reports the pooled Tobit estimates of equations (1), (2) and (3) in the text, for the period 1992-2005. Coefficient estimates denote marginal effects on dependent variable, evaluated at mean values of independent variables. For dummy variables, they represent the effect of discrete changes from 0 to 1. A constant is also included but not reported. Standard errors adjusted for clustering by country-year are reported in parentheses. Asterisks denote significance of coefficients, with ***, ** and * indicating significance at the 1%, 5% and 10% level, respectively.

Estimates of the other covariates in the regression appear generally reasonable and consistent with past research. First, the results suggest that, on average during the 1992 to 2005 period, firms whose income and assets were positively correlated with the exchange rate had a higher fraction of foreign currency-denominated liabilities, given everything else constant (and are positive and statistically significant). This result confirms the findings in Bleakley and Cowan (2008), who studied firm’s investment response to exchange rate changes using a sample of 450 companies in five Latin American countries between 1991 and 1999. The authors found that the fraction of debt denominated in foreign currency was highest among firms whose earnings would typically increase in the event of a depreciation.35 These results are also consistent with Brown, Ongena, and Yesin (2011), who study the currency denomination of individual bank loans in a large cross section of firms in 26 transition countries in 2005. The authors find that firms with foreign currency income and assets are more likely to borrow in a foreign currency.

Second, the results of the model also point to the theoretically sensible finding that bigger firms and firms with access to international capital markets hold more foreign currency debt as a fraction of their liabilities. These coefficients have the predicted signs and are statistically significant at standard confidence levels. The relation between firm size and the use of dollar debt is monotonic, increasing from smaller to medium and to larger firms, consistent with evidence for East Asian firms by Allayannis, Brown, and Klapper (2003). Overall, the model explains 63 percent of the variance in the dispersion of foreign currency borrowing ratios across countries and years.36

As with any identification strategy using variation in macroeconomic policies across countries, one must be concerned with the exogeneity of the shift in exchange rate regimes. The empirical analysis thus far rests on the assumption that the cross-country timing of regime changes was not affected by the average level of financial dollarization in the corporate sector. Yet it could be the case that declining corporate debt dollarization in a country over time reduces the financial stability concerns of a steep devaluation, making authorities more inclined to let the exchange rate float. Conversely, authorities’ commitment to exchange rate stability may create perverse incentives to increase liability dollarization by artificially lowering the cost of borrowing in foreign currency. This could reinforce authorities’ tendency to stabilize the exchange rate because they fear the contractionary effects of currency depreciations (Hausmann, Panizza, and Stein (2001)).

To check whether results above are driven by reverse causality, I augment the specification in (1) by including two binary dummies that take the value of 1, one year and two years before a country adopts a flexible regime, respectively—and 0 otherwise. I thus estimate the following specification:

DOLLijct=α0+α1Pre2FLEXct+α2Pre2FLEXDatect+α4PostFLEXct+βYijct1+ϵijct(2)

where Pr e1 FLEX and Pr e2 FLEX take the value of 1 respectively in the first and second year before country c switched to a flexible regime, and 0 otherwise; FLEXDate takes the value of 1 in the year in which country c liberalized the exchange rate regime and 0 in every other year. Finally, PostFLEX takes on the value of 1 after the year the country switched to a floating exchange rate and 0 otherwise. The estimating equation also controls for the same set of country, sector, year, and firm-specific effects, Yijt- 1, as described above.

If changes in the exchange rate regime are driven by pre-existing declines in dollarization, then this “early” regime change dummies would be negative and significant, and/or would attenuate the observed negative average effect of the exchange rate regime change on the share of foreign currency debt documented above.37 Column 2 in Table 4, however, shows a significant increase in dollarization in the years immediately preceding the regime change, and still a significant decrease in the years after countries switch to floating.38 Thus, the results show that dollarization declined after floating regimes were introduced, but not before.

Finally, I estimate a specification similar as (2), but allowing for dynamic effects in the post-floating period:

DOLLijct=α0+α1Pre2FLEXα+α2Pre1FLEct+α3FLEXDatect+α4Post1FLEXct++βYijct1+ϵijctα5Post2FLEXct+α6Afterct(3)

where Post1 LEX, and Post2FLEX take the value of 1 respectively in the first and second year after the regime change and 0 otherwise; and After ct takes the value of 1 beginning the third year after the flexible regime was adopted. Yijct - 1 captures all firm-level controls and fixed effects described in (1). Results in Column 3 do not suggest a shift in the currency composition of firms’ liabilities immediately following the regime change; rather, it appears that the significant negative effect of flexible regimes on dollar debt begins to take place in the second and third year after liberalization. I next investigate whether the results are robust to excluding Argentinean firms from the sample, as the forced de-dollarization of domestic financial contracts in Argentina after the crisis may be distorting estimations. Column 4 confirms that the central results remain unaffected: a switch to a flexible regime is associated with lower corporate debt dollarization.

Overall, I find that the adoption of floating regimes led to a statistically and economically significant decline in firms’ foreign currency borrowing. These findings are not explained by pre-existing trends in dollarization or the mandatory pesification of domestic financial contracts in Argentina after 2001. The more significant effect of flexible regimes on dollar debt begins to take place in the third year after the regime change.

B. Currency Matching in Flexible versus Pegged Regimes

The stock of foreign currency debt may not be a good indicator of the potential for exchange-rate induced financial distress of a firm. A firm may have natural hedges in the form of foreign currency cash flows or assets that buffer the dollar risk arising from its debt portfolio. Thus, what should matter for financial vulnerability is not the level of dollarization per se, but the way firms match the exchange rate exposures on both sides of their balance sheet.

Therefore, in what follows I test whether the sensitivity of dollar debt holdings to the availability of foreign currency revenues and/or dollar assets increased (or decreased) under floating exchange rate regimes. For these purposes, I use a generalized difference-in–difference-in-difference (DDD) approach, interacting a country-level measure of exchange rate flexibility with firm-level measures of export orientation and dollarization of assets, and estimate the following specification:

DOLLijct=β0+β1FLEXDatect+β2PostFLEXct+β3EXPrtoSijct+β4DOLAijct1+β5[EXPtoSijct*PostFLEXct]+β6[DOLLijct1*PostFLEXct]+γXijct1+øj+χc+λt+ϵijct(4)

The main variables of interest are the interaction terms associated with parameters β5 and β6 The estimated values measure how much (or less) tighter is the match between the currency denomination of income streams and assets with the currency composition of liabilities under floating regimes. If, following the introduction of flexible regimes, firms’ dollar debt holdings become more sensitive to the availability of foreign currency revenues and/or dollar assets (compared to countries that keep their exchange rate pegged during the same period), then β5 and/or β6 should be positive. 39 This would provide evidence that firms reduce currency mismatches in their balance sheets, thus insulating themselves more strongly from financial risks arising from exchange rate fluctuations.40

The effects of the exchange rate regime are identified from the variation in dollar debt ratios across firms with different reliance on foreign currency earnings or assets within a given industry and country.41 Thus, the estimation would suffer from reverse causality only if the relative foreign currency debt ratios of firms with high foreign currency income compared to those that mostly rely on domestic-market revenues had a causal effect on the probability that a country switched to a flexible regime. This seems very unlikely to be the case.

Table 5 reports estimates of various specifications of equation (4). Column 1 shows results for the case of export revenues as the only proxy for natural currency buffers. The interaction term on exports to sales is positive and statistically significant. That is, under flexible exchange rate regimes, there is evidence that firms significantly reduce the impact of exchange rate movements on the company’s net cash flows and corporate value by using more systematically their operating income in foreign currency to offset their dollar debt exposure. Put differently, following the adoption of floating regimes, firms with lower real hedges (foreign currency revenues) experienced larger declines in dollar debt relative to firms selling mostly to international markets. In terms of economic magnitude, the point estimates in Column 1 imply that under floating exchange rate regimes, the foreign currency debt ratio of firms that sell primarily to the domestic market (in the bottom 5th percentile by exports to sales ratio) dropped 5 percentage points relative to the highly export-oriented firms in the 95th percentile of the distribution—compared to the same relative change across firms under pegged regimes.42

Column 2 shows the results for the full specification in equation (3), which includes foreign assets as another potential source of foreign currency cash flows. The results show that under floating regimes, firms have tended to limit the impact of exchange rate movements on the company’s financial position by correlating more strongly the currency denomination of liabilities and assets. Evaluated at the sample means of the data, the point estimate in Column 2 indicates that firms on average increase the degree of currency matching between the stock of foreign currency assets and liabilities by more than 30 percent, compared with pegged regimes. When taken together, these results also indicate that the increase in currency matching is observed both in stocks and in flows.

Table 5.

The Effects of Exchange Rate Regimes on Balance Sheet Currency Mismatches

article image
Source: Author’s calculations.Note: This table reports the pooled Tobit estimates of augmented versions of equation (4) in the text, for the period 1992-2005. Coefficient estimates denote marginal effects on dependent variable, evaluated at mean values of independent variables. For dummy variables, they represent the effect of discrete changes from 0 to 1. Variables based on balance sheet stock values are measured at the beginning of the period. A constant is also included but not reported. Standard errors adjusted for clustering by firm are reported in parentheses. Asterisks denote significance of coefficients, with ***, ** and * indicating significance at the 1%, 5% and 10% level, respectively.

I next investigate whether the previous results hold when using a continuous measure of exchange rate flexibility. I estimate the following specification:

DOLLijct=β0+β1FLEXDatect+β2FtoFct+β3EXPtoSijct+β4DOLAijct1+β5[EXPtoSijct*FtoFct]+β6[DOLAijct1*FtoFct]+γXijct1+øj+χt+λt+ϵijct(5)

where FtoF is the Freedom to Float index, with higher values denoting more exchange rate flexibility FLEXDate takes the value of 1 in the year in which country c liberalized the exchange rate regime and 0 in every other year. Again, the results show that the degree of currency matching in firms’ balance sheets is—economically and statistically—higher during periods of higher exchange rate volatility.

In macroeconomic terms, the results suggest that flexible regimes lead to a reallocation of dollar liabilities towards firms with better natural currency hedges.43 One can use the estimates of Column 4 in Table 5 to quantify the effect of exchange rate flexibility in redistributing dollar debt across firms with differing abilities to bear exchange rate risk. On the basis of the estimated coefficients, I compute the difference in dollarization levels between firms with high dollar asset holdings (95th percentile of the distribution in the sample) and firms with low dollarization of assets (5th percentile) in a country with the highest average index of flexibility compared to the country with the lowest index, as follows:

β^6[(DOLA95thDOLA5th)(FtoF95thFtoF5th)](6)

The coefficient estimate on the interaction term for the asset dollarization ratio in column 4 suggests that the difference in debt dollarization levels between firms with very high and very low dollarization of assets in Chile (the country that attains the maximum level of exchange rate flexibility in the sample) is 14 percent higher than the difference in average dollar debt shares between the same type of firms in Argentina (in the bottom 5th percentile in terms of flexibility over the whole sample). As a comparison, the difference across countries in the average dollarization among firms in these two extremes of the distribution is approximately 28.6 percent. This suggests that the effect of currency regime flexibility accounts for approximately 50 percent of the mean difference.

The evolution of Chilean firms’ foreign exchange exposure over the sample period—presented in Figure 4—clearly illustrates this general result. Before 1999, Chile had an exchange rate band and an explicit commitment to exchange rate stability. Since 1999, the exchange rate was allowed to float freely. As shown in the figure, in the years following the switch to a flexible regime firms significantly reduced the negative exposure to a local currency depreciation by reducing their reliance on foreign currency borrowing, and by ramping up the accumulation of dollar assets. For the average firm in the sample, the net dollar asset position (as a percentage of total assets) jumped from -2.4 percent in 1998 (in the eve of the exchange rate regime change) to 4.2 percent in 2005.44

Figure 4.
Figure 4.

Chile: Strong Accumulation of Net Dollar Assets since Onset of Flexible Regime in 1999 1/

(Annual averages across firms, except where noted)

Citation: IMF Working Papers 2012, 069; 10.5089/9781463939052.001.A001

Source: Author’s calculations.1/ Net dollar assets are defined as the difference between dollar assets and dollar liabilities, as a fraction of total assets (median across firms).

V. Robustness Tests

In this section, I perform several robustness tests on the main results to address concerns with potential empirical biases in our estimations. I also consider competing explanations for the negative relation between the flexibility of exchange rate regimes and currency mismatches in firms’ balance sheets. Results are presented in Table 6.

A. Firms with No Foreign Currency Debt

One concern with the baseline specification is that an important fraction of the firms had zero dollar debt in every year, suggesting that the dynamics governing their financial decisions could be very different from the rest of the firms in the sample. To allay concerns about sample selectivity, we excluded firms that showed zero dollar indebtedness throughout the sample.45 As shown in the first two columns of Table 6, dropping these firms had minor qualitative impacts on the basic result of firms’ lower foreign currency exposure under floating regimes.

B. Sample Selection Bias

Another concern is that non-random exit may be biasing the results. For example, it may be the case that those firms that went bankrupt or were merged or acquired (and subsequently delisted) after regime switches were those with higher currency exposure and thus a higher level of financial vulnerability. In this case, we would tend to observe a reduction in the average foreign currency exposure post-floating due to changes in the composition of the sample. To diminish concerns about survivorship bias, I re-estimate the model by including only those firms in the sample that had at least two year-observations in both fixed (and/or pegged) and flexible regimes. Results in columns 3 and 4 in Table 6 confirm the central finding in the paper that exchange rate regimes affect firm’s incentives to hold foreign currency denominated debt and manage currency risk.

C. Importance of Interest Rate Differentials and Exchange Rate Appreciation

A firms’ decision to borrow in foreign currency may be positively related to slow moving country-level determinants, such as interest rate differentials or the strengthening of the domestic currency in real terms.46 Therefore, I allow country fixed effects to vary by year, to control for all time-varying, country-specific observed and unobserved factors that might affect the dollarization of firms’ liabilities. This specification reduces concerns that the positive interaction term might be picking up time-varying country effects that are due to factors other than the regime change, like changes in the interest advantage on foreign currency funds or the appreciation of the domestic currency. As shown in columns 5 and 6 of Table 6, the central result is not sensitive to this specification: under floating regimes, firms match more systematically the currency denomination of their liabilities with the exchange-rate sensitivity of their assets and revenues, suggesting that flexible exchange rate regimes may have encouraged firms to insulate themselves more from balance sheet risks arising from exchange rate fluctuations. If anything, this specification suggests that the increase in balance sheet currency matching was almost 50 percent compared with pegged regimes for these subset of firms.

Table 6.

The Effects of Exchange Rate Regimes on Currency Mismatches: Robustness Tests

article image
Source: Author’s calculations.Note: This table reports the pooled Tobit estimates of augmented versions of equation (4) in the text, for the period 1992-2005. Coefficient estimates denote marginal effects on dependent variable, evaluated at mean values of independent variables. For dummy variables, they represent the effect of discrete changes from 0 to 1. The key independent variable is the interaction term, and the marginal effect is calculated as in Appendix 1. Variables based on balance sheet stock values are measured at the beginning of the period. A constant, a dummy for the year of the regime switch and firm-level controls are included but not reported. Standard errors adjusted for clustering at the country-year are reported in parentheses. Asterisks denote significance of coefficients, with ***, ** and * indicating significance at the 1%, 5% and 10% level, respectively. For detailed sources and descriptions, see Section 2.

D. Additional Statistical Checks

While not reported to conserve space, I performed a battery of additional robustness checks. I confirmed that my findings are not driven by individual countries by sequentially excluding one country at a time from the sample and re-estimating the same specification. In each case, I was unable to reject the null hypothesis that the estimated coefficient on the interaction terms equaled the value estimated for the full sample at the 5 percent level. Although statistical significance varies somewhat across specifications, the magnitude of the estimated effects and their economic significance remain fairly stable.

The results are also robust to alternative sample selection rules. I considered: (1) excluding firms with negative net worth or in financial distress, (2) excluding country-years during regime transitions, and (3) winsorizing all variables in both tails at the 1 percent level to control for outliers. Dropping these firms did little to alter the main result in the paper: under floating regimes, firms have tended to limit the impact of exchange rate movements on the company’s financial position by correlating more strongly the currency denomination of assets and liabilities. That is, those firms relying more heavily on dollarized debt are also those firms that see the largest offsetting increases in current earnings and future investment opportunities following a depreciation of the domestic currency.

VI. Alternative Explanations

A. Changes in Regulations to Banks’ Foreign Currency Lending

In response to the financial crises at the end of the 1990s and early 2000s, several Latin American countries sought to limit banks’ exchange rate-induced credit risk arising from foreign currency lending to unhedged borrowers. Within our sample period, Argentina and Chile introduced prudential regulations to discourage lending in foreign currency, including tighter limits on dollar lending based on borrowers’ ability to repay in the event of unexpected exchange rate movements.47 Thus, the estimated negative effect of floating regimes on corporate currency mismatches documented above could be also capturing changes in domestic banks’ lending practices as a result of new regulations—and not only changes in the behavior of non-financial firms themselves. To check this possibility, I re-estimated equation (4) but excluding firm-year observations for Argentina and Chile in the period after the regulations were adopted.48 As shown in column 1 in Table 7, the baseline results still hold, although the degree of significance declines to the 10 percent confidence level.

Table 7.

The Effects of Exchange Rate Regimes on Currency Mismatches: Alternative Explanations

article image
Source: Author’s calculations.Note: This table reports the pooled Tobit estimates of augmented versions of equation (4) in the text, for the period 1992-2005. Coefficient estimates denote marginal effects on dependent variable, evaluated at mean values of independent variables. For dummy variables, they represent the effect of discrete changes from 0 to 1. Variables based on balance sheet stock values are measured at the beginning of the period. A constant, a dummy for the year of the regime switch and firm-level controls are included but not reported. Standard errors adjusted for clustering at the country-year are reported in parentheses. Asterisks denote significance of coefficients, with ***, ** and * indicating significance at the 1%, 5% and 10% level, respectively.

B. Differential Access to Credit and Ability to Expand Production During Crisis

A concern with the interpretation of the basic regression result is that the sharper decline in dollar debt ratios observed for firms with no foreign currency earnings (compared with firms that rely principally on export revenues or have large dollar asset holdings) could be reflecting different abilities to overcome financial constraints across these two types of firms. Large currency depreciations that typically occur during regime changes often coincide with banking crises (the “twin crises”) in which banks (even healthy ones) cut back on lending.49 Firms that generate foreign currency revenues or hold dollar assets may have better ability to raise funds in international markets, as they can pledge export receivables or hard-currency assets as collateral—allowing them to better cope with domestic credit crunches.50 Thus, if firms have also better access to international capital markets, and if large currency depreciations occur in association with banking crises, then the differential effect found may be more related to differences in the ability to access alternative sources of financing (especially foreign finance) when domestic credit is depressed, than changes in currency risk management.

Likewise, larger firms could have more available resources that enable them to take better advantage of the improved terms of trade that occur after large exchange rate depreciations.51 If exporter firms and firms with foreign currency asset holdings are, on average, larger, then scale effects could be confounding the estimated impact of the exchange rate regime on currency mismatches documented above.

To investigate the access-to-credit and the scale-effect hypothesis, I include in the specification interacted terms of the dummies for access to international capital markets and size with the exchange rate regime indicator. Results in columns 2 and 3 in Table 7 indicate that the only interaction terms that remain significant are the export intensity and dollarization of assets, respectively—providing yet additional confirmation of our basic result. Overall, the evidence of an omitted-variable bias stemming from differential access to credit or ability to expand production is not compelling.

VII. A CloseR Look at the Data: Exploiting Changes in Entire Distribution of Firms’ Dollar Debt Ratios

A. Conditional Quantile Estimates: Basic Framework

In Sections IV and V, I investigated the effects of exchange rate regimes in foreign currency borrowing solely on the conditional mean of firm-level dollarization. While the central effects are useful summary statistics of the impact of the exchange rate regime and other covariates, they may mask the heterogeneous impact at the various quantiles of the distribution of firms’ foreign currency borrowing.52

In this section, I apply Quantile Regression (QR) techniques to study the determinants of firms’ debt currency choices. Perhaps the most appealing feature of QR methods is that they allow estimation of the effect of explanatory variables on different points of the outcome’s conditional distribution, including the upper and lower tails as well as the center of the distribution.53 An additional advantage of quantile regression estimates is that the method is robust to departures from normality and homoscedasticity, thus alleviating some of the concerns regarding results obtained with Tobit models (Powell (1986)).54

In a linear setting, the quantile regression model can be written as:

yi=Xiβθ+uθiQuantθ(yi|xi)=xiβθ(7)

where Qθ(yi | xi) denotes the θ-th quantile of yi conditional on the regressor vector xi, with 0 < θ < 1. The distribution of the error term is left unspecified, so the estimation method is ut i basically semi-parametric. The QR estimator of βθ is obtained as a solution to the problem:

β^θ=argmin1N{i:yxiβθ|yiXiβθ|+:y<xiβ(1θ)|yiXiβθ|}(8)

where unique slope parameters are estimated for each θ quantile of interest. The objective function above is a weighted sum of absolute deviations, which can be interpreted as an asymmetric linear penalty function. An important special case of the quantile regression estimator is the least absolute deviation estimator (LAD), which is obtained by setting θ = 0.5 (the median regression). The first quartile is obtained by setting θ=0.25 and so on. When increasing θ from 0 to 1, it is possible to trace the effects at different points of the conditional distribution of y on x.55

When the dependent variable is censored or has corner solutions, the Censored Quantile Regression (CQR) estimator as derived by Powell (1986) is found by solving:

β^θ=argmini=1nρθ(yimax{0,xiβθ}) where ρθ(uθi) is an indicator function defined as:

ρθ(uθi)={θuθiifuθi0(θ1)uθiifuθi<0}(9)

and

Xiβθ=β0,θ+β1,θFLEXDatect+β2,θFtoFct+β3,θEXPtoSijct+β4,θDOLAijct1+β5,θ[EXPtoSijct*FtoFct]+β6,θ[DOLAijct1*FtoFct]+γθaXijct1+øj,θ+χc,θ+λt,θ+ϵijct(10)

Chernozhukov and Hong (2002) devised a tractable CQR algorithm that is robust and performs well near the censoring point. It also handles censoring non-parametrically, i.e., without any distributional assumptions. Given that in the sample 40 percent of firm-observations have zero dollar debt, this is an important econometric issue. The algorithm is a three-step procedure that predicts which observations are least likely to be censored and estimates the coefficients based on those observations using linear quantile regression methods. In Appendix 3, I provide the intuition for this CQR estimator and the practical implementation details of the programming in STATA.

B. Results

Figure 5 presents a concise visual summary of the censored quantile regression results of the model in equations (9) and (10). The solid line in each of the top two charts plots the quantile regression estimates for the interaction coefficients (β5 and β6) for each of the nine percentiles (10th, 20th, up to 90th). The shaded grey area depicts a 95 percent point-wise confidence interval. The dashed line in each figure shows the Tobit estimate of the conditional mean effect.

In the case of the interaction term on the export to sales ratio, quantile regression estimates are positive and significant only for the upper conditional quantiles. That is, a switch to flexible exchange rate regimes leads on average to higher currency matching between dollar debt ratio and export to sales ratio at the upper tail of the distribution (seventh to ninth decile). Interestingly, while I do not find a significant average effect on Tobit estimates, we do observe significant positive impacts on specific parts of the relevant distribution. For the case of the interaction on asset dollarization, all quantile estimates (except the lower decile tail) are positive and significant, and the effect increases monotonically as we move to the upper deciles, reaching the strongest impact in the eighth decile.56

Figure 5.
Figure 5.

Effect of Switch to Flexible Regimes at Different Points of the Cross-Sectional Distribution of Dollar Debt Ratios

Citation: IMF Working Papers 2012, 069; 10.5089/9781463939052.001.A001

Source: Author’s calculations.1/ Shaded area denotes 95 percent confidence interval.2/ Red bars denote changes that are statistically different from zero.

The bottom two charts in Figure 5 show the percentage increase in the degree of currency matching at each decile of the conditional distribution for the case of exports (bottom left) and dollar assets (bottom right). For each decile, I take the ratio between the estimated interaction effects and the corresponding value for the estimated parameter on the level effect (β5,θ/β3,θ for exports to sales and β4,θ/β6,θ for the dollar asset ratio). The figure suggests that the percentage increase in the degree of currency matching in firms’ balance sheets under flexible regimes increases monotonically as we move up the conditional distribution of dollarization, with a much steeper increase in the upper tails. This last result implies a very interesting observation: as countries switch to flexible regimes, the reduction in the degree of foreign exchange rate exposure in firms’ balance sheet becomes more important for firms in the highest deciles of the dollarization distribution. That is, the differential effect is stronger where the theory plausibly suggests the costs of exposure to devaluation risk are likely to be larger. These heterogeneous effects lend additional credibility to the hypothesis that following the adoption of a floating exchange rate regime firms match up more strongly the currency structure of assets and liabilities, effectively reducing currency risk.

VIII. Conclusions

This paper provides cross-country, firm-level evidence of the distinctive role of the exchange rate regime in shaping the behavior of firms towards exchange rate risk. I present robust evidence that exchange rate flexibility encourages firms to reduce currency mismatches in their balance sheets, thus insulating themselves more strongly from financial risks arising from exchange rate fluctuations. Specifically, I show that after countries switch from a pegged to a flexible regime, firms significantly cut their unhedged foreign currency exposures by reducing the share of debt contracted in foreign currency, while using more systematically export revenues and assets denominated in foreign currency to offset their dollar debt risk.

Confirming the logic of the baseline results, I find that a switch to more flexible regimes reduces the degree of currency exposure by significantly more for those firms at the higher conditional quantiles of the dollar debt distribution. These results are consistent with theories proposing that switching to flexible regimes will reduce the corporate sector’s vulnerability to exchange rate fluctuations, as high-frequency volatility discourages foreign-exchange-denominated borrowing and increases incentives for firms to hedge currency risk.

From a macroeconomic point of view, the results imply that flexible regimes lead to a reallocation of dollar debt towards firms that are in a stronger position to absorb the impact of a large exchange-rate change on their balance sheets. This reduces the risks associated with the increasing domestic burden of debts of a country’s non financial sectors (and those of their bank creditors) after exchange rate devaluations, which were devastating during past crises.

Appendix I. Data Construction

In this appendix, I describe the firm-level data used in more detail, and the sources employed to construct them.

I use data reported on a calendar year basis, rather than fiscal year. While firms in many cases report both consolidated and unconsolidated financial statements, I use unconsolidated figures to the extent possible, to reduce variations arising from changes in subsidiaries’ ownership and to work with comparable accounting data. For the purpose of the empirical implementation, I modified the original accounting data in three ways:

(i) Convert all data to real 2000 U.S. dollars using December-to-December changes in the country’s consumer price index and the exchange rate for December 31, 1999.

(ii) Drop all firm/year observations if the accounting data are not self-consistent. In particular, we drop observations if dollar liabilities (assets) exceed total liabilities (assets) or if accounting variables do not accord with sign conventions.

(iii) Compute the change in total assets and construct a Z-score using the sample mean and standard deviation for each country/year. We drop firm/year observations that have absolute value of Z>5. These controls for outliers (either because of inadequate accounting, typing errors, or extreme values).

Definition of Variables

Total assets. Sum of total current assets, long-term receivables, investment in unconsolidated subsidiaries, other investments, net property, plant and equipment, and other assets (Balance Sheet).

Total liabilities. Book value of total liabilities (Balance Sheet).

Foreign currency liabilities. Liabilities denominated or indexed to a foreign currency (in dollars or in other non-domestic currencies), issued domestically or abroad. These include bank loans, borrowing from non-bank financial institutions, securities issuance, commercial debt, and trade credit. Consistent with accounting standards in each country, items that are in foreign currency at the end of the year are converted to domestic currency at the contemporaneous exchange rate. (Balance Sheet Notes).

Short-term foreign currency liabilities. Foreign currency liabilities coming due in the upcoming fiscal year. Includes foreign currency denominated debt issued at short maturities as well as long term issues whose terminal date falls in the next year (Balance Sheet Notes).

Foreign currency assets. Assets denominated or indexed to a foreign currency. These include cash, government issues indexed to the dollar, bank deposits abroad, and overseas client credits. Converted into local currency using end of period exchange rate (Balance Sheet Notes).

Exports. Total sales in foreign markets. (Income Statement, when available or Customs data in each country). Dollar export values were converted into domestic currency using the year’s average exchange rate.

Sales. Gross sales and other operating revenues from main activities (Income Statement).

Leverage. Total liabilities as a share of total assets in the balance sheet.

Industry Dummies. Is the industry in which the firm has its main operations. I code firms according to the one-digit ISIC 2 classification.

International Access. A dummy variable that takes on a value of one starting the year the firms accessed international equity markets (by cross-listing shares in foreign stock markets) and/or tapped foreign credit markets (by issuing bonds or taking loans abroad).

The third major source of firm-level data captures firms’ access to international capital markets. I obtained firm-level issuance data on private bonds and syndicated loans from Dealogic Bondware and Loanware. For access to equity markets, I used Bank of New York data to identify those firms whose shares listed in a foreign stock exchange in the form of American Depositary Receipts (ADRs).

Sources

Balance sheet information was mostly collected from annual reports and corporate filings obtained from local stock markets, regulatory agencies, and/or trade chambers in each country.57 Where appropriate, I complemented and cross-checked these sources with data obtained from Economática and Bloomberg. For access to equity markets, we used Bank of New York data to identify those firms whose shares listed in a foreign stock exchange in the form of American Depositary Receipts (ADRs). Firm-level issuance data on private bonds and syndicated loans were extracted from Dealogic Bondware and Loanware. Firms’ main sector of operations was identified using Economatica and Lexis Nexis.

The country-specific sources are as follows:

Argentina. Balance sheet information up to 2001 comes from Galiani, Levy-Yeyati, and Schargrodsky (2003), and from financial statements compiled from the Buenos Aires Stock Exchange. From 2002 onwards, data is from Economática. Data on exports are matched using customs data from ExiNet (NOSIS).

Brazil. Data are compiled from corporate filings submitted to the Securities and Exchanges Commission of Brazil (CVM), and complemented with data from Economática and Bloomberg for 2003 to 2005. Export data come from Notes to Financial Statements and LAFIS.

Chile. Balance sheet information is obtained from the Ficha Estadística Codificada Uniforme (FECUS) database and notes to financial statements obtained from the SuperValores of Chile. Data on exports come from Pro-Chile.

Colombia. Balance sheet information and export data obtained from Super Intendencia Financiera de Colombia.

Mexico. Balance sheet information and export data obtained from Mexican Stock Exchange.

Peru. Balance sheet information come from Comisión Nacional de Valores (CONASEV). Data on exports comes from COMEXPERU.

Appendix II. Results Based on Event Study Techniques

Specification

In this section, I try a different approach to identifying the effect of exchange rate regimes on firm’s balance sheet currency exposures: examining the variation in firm-level dollarization immediately before and after a country’s switch to a flexible exchange rate regime. To do so, I estimate the following specification:

DOLLijct=β0+ηi+β1FLEXDatect+β2PostFLEXct+β3EXPtoSijct+β4DOLAijct1+β5[EXPtoSijct*PostFLEXct]+β6[DOLAijct1*PostFLEXct]+γXijct1+øj+χc+ϵijct(B1)

This specification is more flexible than (4) in that it accommodates a firm-specific term ηi which also captures any variation in initial conditions at the firm-level at the time of the switch to a floating regime.58 Letting T=0 (T=1) denotes before (after) the exchange rate switch, then we can first difference (B1) and estimate the following specification:

ΔDOLLijct=DOLLijctDOLLijct0=β2ΔPostFLEXct+β3ΔEXPtoSijct+β4ΔDOLAijct1+β5Δ[EXPtoSijct*PostFLEXCT]+β6Δ[DOLAijct1*PostFLEXct]+γΔXijct1+Δϵijct(B2)

This event study approach isolates the independent effect of a move to flexible exchange rate regimes purely from the within-firm changes in dollarization.59 Note that the β0 term has dropped out of the regression, as well as country and sector fixed effects. First differencing also removes unobserved heterogeneity across firms, such as differences in technologies, market power, and/or managerial behavior.60 By eliminating these important sources of omitted variable bias, one may obtain coefficients that come closer to representing a causal impact.

I estimate the specification in (B2) with the post-flexible regime dummy, so that:

ΔPostFLEXct=PostFLEXctPostFLEXc0=10=1 for all countries and (B2) reduces to:

ΔDOLLijct=β2+β3ΔEXPtoSijct+β4ΔDOLAijct+β5EXPtoSijcT1+β6DOLAijcT1+γΔXijct+uijct(B3)

This way, the differential impact of exchange rate regimes across firms with different export to sales and asset dollarization ratios is estimated by the coefficients β5 and β6, respectively.

Unlike the panel analysis, the event study approach uses only one observation per country-firm: the change in foreign currency debt (and its determinants) around the regime change event. It thus requires defining the horizon over which the effects are expected to be realized. Based on results reported in the previous section, I measure ΔDOLLijct as the difference in average dollarization between (t–3, t–1) and (t+ 1, t+ 3) around the exchange rate regime switch in year t. Likewise, ΔEXPtoSDOLA) is computed in each case as the difference between the average export ratio (average dollar asset ratio) between t ‒3 and t ‒1, and the corresponding average ratio between t + 1 and t + 3. Averaging ensures that the results are not influence by temporary movements in dollar debt, dollar asset, and export intensity ratios.

Results

The first column in Table A-1 reports results from the estimation of equation (B3). The effects of moving to flexible exchange rate regimes appear even in this econometrically stringent model.61 Cross-sectional changes in firms’ debt dollarization ratios after countries switch to flexible regimes are significantly correlated with the firms’ holdings of foreign currency assets. That is, firms alter their balance sheet structures to provide a more effective buffer against exchange rate shocks and reduce their insolvency risk. These differences are statistically significant, or too large to have easily occurred by chance. The point estimates from the event study specifications are approximately half of the magnitudes from the earlier panel results.

One potential concern with the analysis is that export intensity can be itself endogenous to the exchange rate regime, as currency reforms are typically accompanied by steep increases in the nominal exchange rate which lead to gains in competitiveness. In this case, we would be violating the identification restriction that the exchange rate regimes can only have an independent effect on the dependent variable.62 However, changes in export status from non-exporter to exporter at the time of the crisis where relatively limited in our sample (5 percent of the observations at the time of the regime changes for the whole sample) and therefore, limit the potential sample selection concerns (see Table A-2).

Table A-1.

The Effects of Exchange Rate Regimes on Currency Matching: Event Study

article image
Source: Author’s calculations.Note: This table reports the pooled OLS estimates of equation (B3) in the text. The dependent variable is the difference in a foreign currency debt ratios between (t-1 to t-3) and (t+1 to t+3) where t is the year of the change in exchange rate regime. Robust standard errors are reported in parentheses. Asterisks denote significance of coefficients, with ***, ** and * indicating significance at the 1%, 5% and 10% level, respectively.

In countries where domestic sales are conducted exclusively in local currency, exports are a good indicator of whether a firm has foreign currency income. But several countries in our sample display real dollarization, i.e., many domestic transactions are conducted in—or indexed to—foreign currency (typically the dollar). Thus, export earnings could be a noisy measure of the exchange rate-sensitivity of a firm’s earnings. To address this concern, Columns 2 and 3 show the estimated results when using dollar assets as the only proxy for natural currency buffers. Again, results confirm the central finding in the paper that under flexible regimes, firms match more systematically the currency denomination of their liabilities and assets, that is, a higher share of foreign currency debt is backed by foreign currency assets. This would tend to cushion the effect of depreciation shocks on firms’ balance sheets.63

Table A-2.

Changes in Export Status After Exchange Rate Regime Changes Brazil Chile

article image
Source: Author’s calculations.Note: This table reports transition matrices for changes in export and non-export status after switches to flexible exchange rate regimes.

Appendix III. Censored Quantile Regression

In this section, I provide more intuition for the Censored Quantile Regression (CQR) estimates, and provide practical implementation details.

Basic Model and Interpretation of the Quantile Regression

A least square estimator of the mean regression model would be concerned with the dependence of the conditional mean of Y on the covariates X. The quantile regression estimator tackles this issue at each quantile of the conditional distribution, providing thus a more complete description of how the conditional distribution of Y given X=x depends on x. Instead of assuming that the covariates shift only the location and scale of the conditional distribution, quantile regression looks at the potential effects on the shape of the distribution as well.

More formally, for any variable Y whose distribution is continuous and monotone and for any θ E (0,1), the θth unconditional quantile of Y is a number QY (θ) such that:

F(QY(θ))=θ(C1)

where F(.) is the distribution function of Y. Equivalently:

Q(θ)=inf{y:F(y)θ}(C2)

Let (yi, xi = 1,…,n be a sample from some population, where xi is a Kx1 vector of regressors. Assuming that the θth quantile of the conditional distribution of yi is linear in xi, we can write the conditional quantile regression model as:

yi=xiβθ+uθi,
Qθ(yi|xi)inf{y:F(yi|xi)θ}=xiβθ(C3)

Where βθ is the unknown vector of parameters to be estimated for different values of θ ∈ (0,1). If Fθ(.) was known then various techniques could be used to estimate βθ. However, here the distribution of the error term uθi is left unspecified. The only constraint is i that the conditional quantile of the error term is zero:

Qθ(uθi|xi)=0(C4)

The QR estimator of βθ is obtained by minimizing its sample counterpart, i.e., the average of the asymmetrically weighted sum of the absolute errors, with weight θ on positive errors and weight on negative errors:

β^θ=argmin1N{i:yxiβ|θ|yixiβθ|+:yxiβ(1θ)|yixiβθ}=argmini=1nρθ(uθi)(C5)

For example, when estimating the 75th percentile, negative residuals are weighted by 0.25 and positive residuals by 0.75. The criterion is minimized when 75 percent of the residuals are negative. In the specific case of the median regression, all residuals receive equal weight. This minimization problem can be solved by linear programming techniques for the different quantiles of the dependent variable as described in, for example, Koenker and Hallock (2001). It can be shown that the estimator for βθ is consistent and asymptotically normal (Koenker and Bassett (1978)).

When the conditional quantile of the error term is zero, a CQR model can be written as:

Qθ(yi|xi)=max{0,Qθ(xiβθ+uθi|xi)}=max{0,xiβθ}(C6)

Censoring induces attenuation bias in quantile regression much in the same way it induces bias in mean regression: when a zero is observed in the place of a value that should be much smaller, a line that fits the observed values will be biased toward zero. Since quantile regression uses information from the entire sample to generate the estimate at each quantile, if some observations on the dependent variable are censored, the quantile regression lines can be biased toward zero at all quantiles. The Powell (1986) estimator overcomes this difficulty by incorporating censoring directly into the estimator as follows:

β^θ=argmini=1nρθ(yimax{0,xiβθ})

where

ρθ(uθi)={(1θ)I(uθi<0)+(θ)I(uθi>0)}|uθi|(C7)

and I is an indicator function taking the value of unity when the expression holds and zero otherwise.

Chernozhukov and Hong (2002) devised a tractable computational censored quantile regression (CQR) algorithm for Powell’s estimator based on the idea that Powell’s censored regression model estimates the coefficients using observations that are not likely to be censored. The algorithm is a three-step procedure that predicts which observations are least likely to be censored and estimates the coefficients based on those observations. The algorithm is simple, robust, and performs well near the censoring point.

This procedure selects a sub sample by a separation restriction that is put on the censoring probability (first step), and estimates the model twice by quantile regression. The goal of the first estimation (second step) is to find an appropriate sub sample, and the purpose of the second estimation (third step) is to make the estimator efficient. In the first step, a set fraction of observations that are unlikely to be censored are retained for estimation via quantile regression in the second step. After the second step, a larger set of observations is retained based on the predicted values of the dependent variable. This sample gets asymptotically close to the ideal sample of non-censored observations, and consistent estimates are obtained through a third step of quantile regression on this sample.

Below I describe in more detail the algorithm that was programmed in STATA:

Step 1: The first step involves a parametric prediction of the probability of censoring based on a Probit or Logit model. I estimate a probability model on the sample:

Pr(yi>0|xi)=F(xiγ)+ϵi Use the probability model to select the subsample J0={i:xiγ^>1θ+c} where c is a trimming constant between 0 and 1. To choose the trimming constant, c, Chernozhukov and Hong (2002) suggest using the minimum value of the Powell objective function in (C7). The goal of Step 1 is to choose a subset of the observations where Pr(yi>0|xi)>1θ, that is, where the quantile xiβθ line is above the censoring point.

Step 2: Obtain the initial estimator βθ0, by ordinary QR on the sample J0. It is shown by Chernozhukov and Hong (2002) that this step gives a consistent but inefficient estimator. Use the initial estimator to select the sample J1={i:xiβθ0>0} to be used in Step 3.

Step 3: Estimate the model by ordinary QR on the sample J1. Chernozhukov and Hong (2002) show that this step gives a consistent and efficient estimate of β^θ.

References

  • Aghion, Philippe, P. Bacchetta, and A. Banerjee, 2004, A corporate balance-sheet approach to currency crises, Journal of Economic Theory 119 (1), 630.

    • Search Google Scholar
    • Export Citation
  • Aguiar, Mark A., 2005, Investment, devaluation, and foreign currency exposure: The case of Mexico, Journal of Development Economics 78, 95113.

    • Search Google Scholar
    • Export Citation
  • Allayannis, George, G. Brown, and L. F. Klapper, 2003, Capital structure and financial risk: Evidence from foreign debt use in East Asia, Journal of Finance 58, 26672709.

    • Search Google Scholar
    • Export Citation
  • Arteta, Carlos O., 2005, Exchange rate regimes and financial dollarization: Does flexibility reduce currency mismatches in bank intermediation?, The BE Journals in Macroeconomics, Topics in Macroeconomics 5, 12261246.

    • Search Google Scholar
    • Export Citation
  • Berkmen, Pelin, and E. Cavallo, 2010, Exchange rate policy and liability dollarization: What do the data reveal about causality?, Review of International Economics 18 (5), 781795.

    • Search Google Scholar
    • Export Citation
  • Berrospide, Jose, A. Purnanandam, and U. Rajan, 2008, Corporate hedging, investment and value, Finance and Economics Discussion Series 2008-16 (Board of Governors of the Federal Reserve System, Washington).

    • Search Google Scholar
    • Export Citation
  • Bertrand, Marianne, E. Duflo, and S. Mullainathan, 2004, How much should we trust differences-in-differences estimates?, Quarterly Journal of Economics 119, 249275.

    • Search Google Scholar
    • Export Citation
  • Bleakley, Hoyt, and K. Cowan, 2008, Corporate dollar debt and depreciations: Much ado about nothing?, The Review of Economics and Statistics 90 (4), 612626.

    • Search Google Scholar
    • Export Citation
  • Bordo, Michael, C. M. Meissner, and D. Stuckler, 2009. Foreign Currency Debt, Financial Crisis and Economic Growth: A Long Run View?, NBER Working Paper No. 15534 (National Bureau of Economic Research, Cambridge, Massachusetts).

    • Search Google Scholar
    • Export Citation
  • Brown, Martin, S. Ongena, and P. Yesin, 2011, Foreign currency borrowing by small firms in the transition economies, Journal of Financial Intermediation, 20, pp. 285302.

    • Search Google Scholar
    • Export Citation
  • Bubula, Andrea, and I. Otker-Robe, 2002. The Evolution of Exchange Rate Regimes Since 1990: Evidence from de Facto Policies, IMF Working Paper No. 02/155 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Burnside, Craig, M. Eichenbaum, and S. T. Rebelo, 2001, Hedging and financial fragility in fixed exchange rate regimes, European Economic Review 45, 11511193.

    • Search Google Scholar
    • Export Citation
  • Caballero, Ricardo, and A. Krishnamurthy, 2003, Excessive dollar debt: Financial development and underinsurance0000, The Journal of Finance 58, 867893.

    • Search Google Scholar
    • Export Citation
  • Calvo, Guillermo A., 2001, Capital markets and the exchange rate with special reference to the dollarization debate in Latin America, Journal of Money, Credit, and Banking 33 (2), 312334.

    • Search Google Scholar
    • Export Citation
  • Calvo, Guillermo A., and C. M. Reinhart, 2002, Fear of floating, The Quarterly Journal of Economics 67, 379408.

  • Calvo, Guillermo A., and C. M. Reinhart, A. Izquierdo, and L. F. Mejίa, 2004. On the Empirics of Sudden Stops: The Relevance of Balance Sheet Effects, NBER Working Paper No. 10520 (National Bureau of Economic Research, Cambridge, Massachusetts).

    • Search Google Scholar
    • Export Citation
  • Carranza, L., Cayo, J.M., and J.E. Galdon-Sanchez, 2003, Exchange rate volatility and economic performance in Peru: a firm level analysis, Emerging Markets Review 4, 472496.

    • Search Google Scholar
    • Export Citation
  • Cayazzo, Jorge, A. Garcia-Pascual, E. Gutierrez, and S. Heysen, 2006. Toward an Effective Supervision of Partially Dollarized Banking Systems, IMF Working Paper 06/32 (International Monetary Fund, Washington).

    • Search Google Scholar
    • Export Citation
  • Cerra, Valerie, and S.C. Saxena, 2008, Growth dynamics: The myth of economic recovery, American Economic Review 98 (1), 439457.

  • Céspedes, Luis Felipe, R. Chang, and A. Velasco, 2004, Balance sheets and exchange rate policy, American Economic Review 94 (4), 11831193.

    • Search Google Scholar
    • Export Citation
  • Chamon, Marcos, and R. Haussmann, 2005, Why do economies borrow the way they borrow?, in Barry Eichengreen and Ricardo Haussmann, eds.: Other People’s Money: Debt Denomination and Financial Instability in Emerging Market Economics (University of Chicago Press).

    • Search Google Scholar
    • Export Citation
  • Chang, Roberto, and A. Velasco, 2006, Monetary policy and the currency denomination of debt: A tale of two equilibria, Journal of International Economics 69, 150175.

    • Search Google Scholar
    • Export Citation
  • Chernozhukov, Victor, and H. Hong, 2002, Simple 3-step censored quantile regression and extramarital affairs, Journal of the American Statistical Association 97 (459), 872882.

    • Search Google Scholar
    • Export Citation
  • Christiano, Lawrence J., C. Gust, and J. Roldos, 2004, Monetary policy in a financial crisis, Journal of Economic Theory 119, 64103.

  • Corden, W. Max, 2002. Too Sensational: On the Choice of Exchange Rate Regimes (MIT Press, Cambridge).

  • Cowan, Kevin, E. Hansen, and L. O. Herrera, 2005, Currency mismatches, balance sheet effects and hedging in Chilean non-financial corporations, in Ricardo J. Caballero, Cesar C., and L.F. Cespedes, eds.: External Vulnerability and Preventive Policies (Central Bank of Chile).

    • Search Google Scholar
    • Export Citation
  • Cowan, Kevin, E. Hansen, and L. O. Herrera, and J. De Gregorio, 2007, International borrowing, capital controls, and the exchange rate: Lessons from Chile, in Sebastian Edwards, ed.: Capital Controls and Capital Flows in Emerging Economies: Policies, Practices and Consequences (University of Chicago Press).

    • Search Google Scholar
    • Export Citation
  • Cowan, Kevin, E. Hansen, and L. O. Herrera, and J. De Gregorio, A. Micco, and A. Yáñez, 2007, Evolución y efectos de los descalces cambiarios: 1995-2004, Super Intendencia de Bancos e Instituciones Financieras, Chile.

    • Search Google Scholar
    • Export Citation
  • Desai, Mihir, Foley, C. Fritz and K. J. Forbes, 2008, Financial constraints and growth: Multinational and local firm responses to currency depreciations, Review of Financial Studies 21 (6), 28572888.

    • Search Google Scholar
    • Export Citation
  • Desai, Mihir, Foley, C. Fritz and K. J. Forbes, C. Fritz Foley, and J. Hines, 2008, Capital structure with risky foreign investment, Journal of Financial Economics 88 (3), 534553.

    • Search Google Scholar
    • Export Citation
  • Domaç, Ilker, and M. S. Martinez-Peria, 2003, Banking crises and exchange rate regimes: Is there a link?, Journal of International Economics 61 (1), 4172.

    • Search Google Scholar
    • Export Citation
  • Dooley, Michael, 2000, A model of crises in emerging markets, Economic Journal 110, 256272.

  • Eichengreen, Barry, and R. Hausmann, 1999, Exchange rates and financial fragility, in New Challenges for Monetary Policy (Kansas City: Federal Reserve Bank of Kansas City).

    • Search Google Scholar
    • Export Citation
  • Eichengreen, Barry, and R. Hausmann, R. Hausmann, and Ugo Panizza, 2005, The pain of original sin, in Barry Eichengreen and Ricardo Hausmann, eds.: Other People’s Money: Debt Denomination and Financial Instability in Emerging Market Economics (The University of Chicago Press).

    • Search Google Scholar
    • Export Citation
  • Eichengreen, Barry, and R. Hausmann, R. Hausmann, and Ugo Panizza, and R. Razo-Garcia, 2006, The international monetary system in the last and next 20 years, Economic Policy 21 (47), 393442.

    • Search Google Scholar
    • Export Citation
  • Frankel, Jeffrey A., 2005, Contractionary currency crashes in developing countries, IMF Staff Papers 52 (2), 149192.

  • Galiani, Sebastian, E. Levy-Yeyati, and Ernesto Schargrodsky, 2003, Financial dollarization and debt deflation under a currency board, Emerging Markets Review 4 (4), 340367.

    • Search Google Scholar
    • Export Citation
  • Galindo, Arturo, U. Panizza, and Fabio Schiantarelli, 2003, Debt composition and balance sheet effects of currency depreciation: A summary of the micro evidence, Emerging Markets Review 4 (4), 330339.

    • Search Google Scholar
    • Export Citation
  • Gelos, R. Gaston, 2003, Foreign currency debt in emerging markets: firm-level evidence from Mexico, Economics Letters 78, 323327.

  • Goldstein, Morris, 1998. The Asian Financial Crisis: Causes, Cures, and Systemic Implications, Policy Analyses in International Economics No. 55 (Institute for International Economics, Washington).

    • Search Google Scholar
    • Export Citation
  • Goldstein, Morris, and P. Turner, 2004. Controlling Currency Mismatches in Emerging Markets (Institute for International Economics, Washington).

    • Search Google Scholar
    • Export Citation
  • Harvey, Campbell R., and A. H. Roper, 1999, The Asian bet, in Alison Harwood, Robert E. Litan, and Michael Pomerleano, eds.: The Crisis in Emerging Financial Markets (Brookings Institution, Washington).

    • Search Google Scholar
    • Export Citation
  • Hausmann, Ricardo, U. Panizza, and E. Stein, 2001, Why do countries float the way they float?, Journal of Development Economics 66 (2), 387414.

    • Search Google Scholar
    • Export Citation
  • Iannariello, Maria Pia, 2005, Financial dollarization level and response to shocks: an in-depth look at the relationship between financial dollarization and regime flexibility unpublished, The George Washington University, Department of Economics.

    • Search Google Scholar
    • Export Citation
  • Inter-American Development Bank, 2002. Debt Composition and Balance Sheet Effects of Exchange and Interest Rate Volatility: A Firm Level Analysis, Terms of Reference Research Proposals, Washington. http://www.iadb.org/res/laresnetwork/studies/st74tor.pdf

    • Search Google Scholar
    • Export Citation
  • Ize, Alain, and E. Levy-Yeyati, 2003, Financial dollarization, Journal of International Economics 59, 323347.

  • Ize, Alain, and E. Levy-Yeyati, 2010, Financial dollarization, Unpublished manuscript, World Bank, Washington.

  • Jeanne, Olivier, 2005, Why do emerging economies borrow in foreign currency?, in Barry Eichengreen and Ricardo Hausmann, eds.: Other People’s Money: Debt Denomination and Financial Instability in Emerging Market Economics (University of Chicago Press).

    • Search Google Scholar
    • Export Citation
  • Kalemli-Ozcan, Sebnem, H. Kamil, and C. Villegas-Sanchez, 2010. What Hinders Investment in the Aftermath of Financial Crises: Insolvent Firms or Illiquid Banks?, NBER Working Paper No. 16528 (National Bureau of Economic Research, Cambridge, Massachusetts).

    • Search Google Scholar
    • Export Citation
  • Kamil, Herman, 2007, A new database on currency and maturity composition of firms’ balance sheets in Latin America: 1992-2005, Unpublished manuscript. (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Kaminsky, Graciela L., and C. M. Reinhart, 1999, The twin crises: The causes of banking and balance-of-payments problems, The American Economic Review 89 (3), 473500.

    • Search Google Scholar
    • Export Citation
  • Koenker, Roger, and G. Bassett Jr., 1978, Regression quantiles, Econometrica 46, 3350.

  • Koenker, Roger, and G. Bassett Jr., and K. F. Hallock, 2001, Quantile regression, Journal of Economic Perspectives 15 (4), 143156.

  • Krugman, Paul, 1999, Balance sheets, the transfer problem, and financial crises, in Peter Isard, Assaf Razin, and Andrew K. Rose, eds.: International Finance and International Crises (International Monetary Fund, Washington).

    • Search Google Scholar
    • Export Citation
  • Levy-Yeyati, Eduardo, 2006, Financial dollarization: Evaluating the consequences, Economic Policy 21(45), 61–118.

  • Magud, Nicolas, 2010, Currency mismatch, openness and exchange rate regime choice, Journal of Macroeconomics 32, 6889.

  • Martinez, Lorenza, and A. Werner, 2002, The exchange rate regime and the currency composition of corporate debt: The Mexican experience, Journal of Development Economics 69 (2), 315334.

    • Search Google Scholar
    • Export Citation
  • McKinnon, Roland, and H. Pill, 1999, Exchange-rate regimes for emerging markets: Moral hazard and international over-borrowing, Oxford Review of Economic Policy 15, 1938.

    • Search Google Scholar
    • Export Citation
  • Mendoza, Enrique, 2002, Credit, prices and crashes: business cycles with a sudden stop, in J. Frankel and S. Edwards, eds.: Preventing Currency Crises in Emerging Markets (University of Chicago Press).

    • Search Google Scholar
    • Export Citation
  • Mishkin, Frederic S., 1996, Understanding financial crises: A developing country perspective, in Bruno Michael and Boris Pleskovic, eds.: Annual World Bank Conference on Development Economics (World Bank, Washington).

    • Search Google Scholar
    • Export Citation
  • Parsley, David C., and H. A. Popper, 2006, Exchange rate pegs and foreign exchange exposure in East and South East Asia, Journal of International Money and Finance 25, 9921009.

    • Search Google Scholar
    • Export Citation
  • Patnaik, Ila, and A. Shah, 2010, Does the currency regime shape un-hedged currency exposure?, Journal of International Money and Finance 760769.

    • Search Google Scholar
    • Export Citation
  • Powell, James L., 1986, Censored regression quantiles, Journal of Econometrics 32 (1), 143155.

  • Ranciere, Romain, A. Tornell, and A. Vamvakidis, 2010, Currency mismatch and systemic risk in emerging Europe, Economic Policy 25 (64), 597658.

    • Search Google Scholar
    • Export Citation
  • Rennhack, Robert, and M. Nozaki, 2006, Financial dollarization in Latin America, in Adrián Armas, Alain Ize, and Eduardo Levy-Yeyati, eds.: Financial Dollarization (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Rosenberg, Christopher, and M. Tirpak, 2009, Determinants of foreign currency borrowing in the new member states of the EU, Czech Journal of Economics and Finance 59 (3), 216228.

    • Search Google Scholar
    • Export Citation
  • Rossi, Jose Luis, 2009, Corporate financial policies and the exchange rate regime: Evidence from Brazil, Emerging Markets Review 10 (4), 279295.

    • Search Google Scholar
    • Export Citation
  • Ro