Journal of International Money and Finance 30 (2011) 1265–1279

Contents lists available at ScienceDirect

Journal of International Moneyand Finance

journal homepage: www.elsevier .com/locate/ j imf

The relationship between investment and large exchangerate depreciations in dollarized economies

Luis Carranza a, Jose E. Galdon-Sanchez b,*, Javier Gomez-Biscarri c

aUniversidad San Martin de Porres, Facultad de Ciencias Contables, Economicas y Financieras, Ciudad Universitaria,Jr. Las Calandrias No 151-291, Santa Anita, Lima 43, PerubUniversidad Publica de Navarra, Departamento de Economia, Campus de Arrosadia, 31006 Pamplona, SpaincUniversitat Pompeu Fabra and Barcelona GSE, Departament d’Economia i Empresa, Ramon Trias Fargas, 25-27,08005 Barcelona, Spain

JEL classification:F31F33

Keywords:Large depreciationDollarizationBalance-sheet effectInvestmentCurrency mismatch

* Corresponding author. Tel.: þ34 948 169338; fE-mail addresses: luiscarranzau@gmail.com (L. C

edu (J. Gomez-Biscarri).

0261-5606/$ – see front matter � 2011 Elsevier Ltdoi:10.1016/j.jimonfin.2011.07.006

a b s t r a c t

We use a simple financial friction in an economy with high degreeof liability dollarization – and currency mismatch – to show thatthe negative balance-sheet effect of an exchange rate depreciationmay be observable only if the magnitude of the depreciation islarge enough. This result justifies the difficulty to find strongempirical evidence for balance-sheet effects and suggests theconvenience of including a “large depreciation” term in empiricalanalyses. We review some of the related empirical literature andprovide some new evidence of this large depreciation effect.

� 2011 Elsevier Ltd. All rights reserved.

1. Introduction

There is ample evidence of large real exchange rate depreciations that are accompanied by GDPcontractions, at least in the short-run. Such behavior has been observed in several countries during thelast 20 years (see Table 1 for some examples). The literature on liability dollarization and currencymismatch (Cespedes et al., 2004; Choi and Cook, 2004; Magud, 2010; Ize and Levy-Yeyati, 2005; Batiniet al., 2007; Bleakley and Cowan, 2008; Carranza et al., 2009) has suggested that a balance-sheet effectinduced by exchange rate depreciations could be an explanation for this negative impact: when firms’liabilities are denominated in a foreign currency, a depreciation may lead to a reduction in firms’ net

ax: þ34 948 169721.arranza), jose.galdon@unavarra.es (J.E. Galdon-Sanchez), javier.gomez@upf.

d. All rights reserved.

Table 1GDP Growth and Large Exchange Rate Swings.

Country Year Exchange Rate Depreciationa Domestic inflationb GDP Growth ratec

Argentina 2002 206 25.87 �10.90Brazil 1999 56 4.86 0.25Mexico 1995 90 35.00 �6.22Nicaragua 1991 2930 116.60 �0.20Paraguay 2002 39 10.51 0.00Peru 1999 15 3.52 0.91Dominican Republic 2003 66 27.45 �0.30Russia 1998 67 27.68 �5.35Thailand 1998 32 8.08 �10.50Venezuela 2002 60 22.43 �8.90

Note: a,b,cin percentages.Source: World Banka, IMFb,c.

L. Carranza et al. / Journal of International Money and Finance 30 (2011) 1265–12791266

worth which, in the presence of financial constraints, reduces access to credit and investment and,consequently, generates a contractionary effect that goes counter the traditional competitiveness effectof the depreciation. This effect is amplified when maturity mismatches exist, triggering also, in mostcases, credit crunch episodes.

Recent empirical analyses, however, have found only weak evidence for this effect (seeLuengnaruemitchai, 2003, for a review), and usually only in the context of quite large depreciations(see, among others, Burstein et al., 2005; the papers in Galindo et al., 2003; Leiderman et al., 2006).These empirical findings suggest that the aggregate investment function may present a nonlinearity inits dependence on the (real) exchange rate:

DIt ¼ HðztÞ þ ðlþ rDtÞDet ; Dt ¼ 1½Det > 4� (1)

where DIt is the change in aggregate investment, H(zt) contains the effect of relevant variables otherthan the real exchange rate, Det is the change in the real exchange rate, l is the sensitivity of investmentto “regular” real depreciations and r is the additional impact of a real depreciation that is “large” (i.e.greater than some threshold 4). Finally, 1[Det>4] is an indicator function that takes value one if thechange in the real exchange rate is larger than 4.

In Eq. (1), the coefficient l may be positive or negative, since it is a combination of a positivecompetitiveness effect (a real depreciation increases the output of firms that sell tradeables), of anincreased relative cost of imported capital (a financial cost effect) and of a negative impact of theincrease in relative worth of foreign currency liabilities (the balance-sheet effect). In this paper weargue, however, that the coefficient r is negative. We show in Section 2 how a simple financial frictionmay lead to an investment function of the form shown in (1), a result which explains the difficulty offinding robust empirical evidence for the balance-sheet effect of real depreciations. We then review inSection 3 some of the recent empirical literature on balance-sheet effects and the relationship betweeninvestment and real depreciations and show the results of an empirical analysis which support thepossible nonlinear relationship between investment and the real exchange rate.

2. Investment and large exchange rate depreciations

We use a simple model in the spirit of that of Bleakley and Cowan (2002). Assume a small countrywith a continuum of firms that produce tradeables and a continuum of firms that produce non-tradeables.1 There are two periods. Firm i enters period one with some long-term debt, which may bedenominated in foreign ðL�i Þ or local (Li) currency. Thus, the ratio L�i =Li is a measure of the degree of

1 Alternatively, we could think of firms producing a share of tradeables and a share of nontradeables. The results would notchange at all but we believe that keeping both types of firms separate facilitates the interpretation and it simplifies some of thederivations of the model’s solution.

L. Carranza et al. / Journal of International Money and Finance 30 (2011) 1265–1279 1267

currency mismatch at the firm level, and the aggregate ratio a measure of total liability dollarization.Our analysis, therefore, applies especially to developing countries, which tend to present high levels offoreign currency debt due to their inability to borrow in local currency (a phenomenon termed“original sin” by Eichengreen et al., 2003).

We assume that short-term debt in period one, Si,1, is equal to zero. For simplicity we also assumethat the level of long-term indebtedness is given and all short-term debt is contracted in foreigncurrency.2 The real exchange rate e0 at which the foreign debt L�i was contracted is equal to one and novariation is anticipated. Initial period capital for firm i, Ki,1, is also equal to zero.3 We assume thata proportion d of all capital goods is imported.4

During the initial period, after an unexpected real exchange rate depreciation has occurred (i.e.,e1 > 1), firms make their investment decisions taking into account their budget and borrowingconstraints. Firm i chooses next period capital, Ki,2, and the short-term borrowing in foreign currencycontracted at the initial period and payable at the last period, Si,2, to maximize profits. Now thedistinction arises between firms that produce nontradeables (i.e., their flow of income is denominatedin local currency) and firms that produce tradeables (i.e., their income is denominated in foreigncurrency). We assume that all firms are price-takers and they can sell their whole production F(Ki,2).We thus abstract from competitiveness effects of exchange rate changes based on increased demand(which, in any case, would favor our argument).5

2.1. Nontradeable firms

The problem for a firm i that produces nontradeables is the following:

Max�F�Ki;2

�� e2L�i � Li � e2rSi;2

�(2)

s:t:gðe2ÞKi;2 � e2Si;2 (3)

e2rSi;2 � q�F�Ki;2

�� e2L�i � Li

�(4)

where e2 is the expected real exchange rate in the second period, g(e2)¼de2 þ (1�d) is the relative priceof capital goods and r is the gross interest rate on short-term debt. Note that: 1) L�i and Li are contractedbefore the depreciation takes place and are repayable at the last period, and 2) we abstract from theinterest rate on long-term debt, so that L�i and Li can be taken to be the gross final value repayable at thelast period. Firms can borrow a fraction 0�q�1 of their financial necessities. The price of nontradeable

2 The latter assumption that short-term financing must be “hot money” can, of course, be relaxed without changing theimplications of the model. It is reasonable to think that if firms have significant levels of liability dollarization it must be thecase that domestic financial markets are not deep. This shallowness, in fact, justifies the existence of the financing constraintthat generates the balance-sheet effect. In any case, the balance-sheet effect is not, strictly speaking, a consequence of having toresort to hot money for financing, but of the fact that if the firm already has debt denominated in foreign currency, the realdepreciation generates a reduction in net worth which, in the presence of a financing constraint, prevents the firm fromobtaining additional funds.

3 This assumption implies that capital equals investment, so having no capital stock in period 1 allows us to obtain a simplesolution for investment and for the derivative of investment with respect to the real exchange rate. Many investment models(including Bleakley and Cowan, 2002), make parallel simplifying assumptions such as exogenous initial capital or capital thatfully depreciates after one period.

4 Again, this assumption is not unreasonable in the case of emerging economies, which tend to rely heavily on foreigninvestment goods (Calvo et al., 1994).

5 One could argue that, given a large enough depreciation, there is an incentive for firms to switch from non-tradable totradable sectors. This is the mechanism by which small open economies have endogenously come out of the recessionsassociated with massive real depreciations (it has been the case for some Asian economies after the 1997 crisis or Mexico afterthe 1994 crisis). Kehoe and Ruhl (2009), for example, provide a formal analysis of this transition, which they calibrate to thecase of Mexico. The transition, however, tends to be costly and slow due to asset specificity, so that during this transition thebalance-sheet effect is observed and can, in fact, be quite intense due to the necessity for fire-sale liquidation of non-tradablefirm assets. Our model is short-run in spirit and focuses on the immediate impact of a large depreciation. Therefore, weexplicitly abstract from this dynamic adjustment in the sectoral composition of output.

L. Carranza et al. / Journal of International Money and Finance 30 (2011) 1265–12791268

goods is normalized and used as a numeraire in the second period. Given that payments and output arerealized at period two, the change in e1 must influence the level of e2. Otherwise, e1 will not have anyimpact on investment decisions. Thus, we assume that the real exchange rate exhibits persistence sothat e2¼m(e1) and vm(e1)/ve1 > 0.6

Eq. (3) is a budget constraint: new capital expenditures are financed by short-term borrowing.Given that short-term debt is costly, this constraint will hold with equality. Eq. (4) is a borrowingconstraint: the maximum short-term borrowing is a fraction q of the firm’s final net worth. The ideabehind the parameter q is to make explicit that credit imperfections are due to an enforcementproblem: lenders can not force their borrowers to repay their debt, but they can seize a fraction of theborrower’s final net worth (see Kiyotaki and Moore, 1997, and Aghion et al., 2001).

For firms that are credit constrained (Eq. (4) is binding), the choice of Ki,2 depends on creditavailability rather than on optimality conditions. In that case Ki,2 is determined by replacing (4) into (3)in order to get:

Ki;2 ¼ q

gðe2Þr�F�Ki;2

�� e2L�i � Li

�(5)

The solution to (5) is a fixed point, Ki,2 ¼ K* which can be represented as in Fig. 1, Panel A, whereG(K)¼q/g(e2)r(F(K)�e2L*�L) and I(K)¼K.

The level of Ki,2 depends not only on r but also on the net worth of firm i (given by L�i and Li) and thereal exchange rate, e2. We denote it by Ki;2 ¼ KR

2 ðr; e2; L�i ; LiÞwhere the superscript Rmeans “restricted”firm. It can easily be proved that an increase in today’s real exchange rate, e1, will produce a fall in theinvestment of firm i. Taking implicit derivatives with respect to e1 we obtain:

vKi;2

ve1¼ �m0ðe1Þ

gðe2Þr � qF 0�Ki;2

��qL�i þ drKi;2�< 0 (6)

where the impact of a real depreciation on investment is negative because of higher financial costs(drKi,2) and the balance-sheet effect ðqL�i Þ.7 Note that if capital goods are not imported (d¼0),vKi;2ve1 ¼ �m0ðe1Þ

r�qF 0ðKi;2ÞðqL�i Þ < 0 so that, even though there is no financial cost effect, there is still a negative

balance-sheet effect. Alternatively, if capital goods are all imported (d¼1), vKi;2ve1

¼ �m0ðe1Þe2r�qF 0ðKi;2ÞðqL

�i þ rKi;2Þ,

the negative combined effect is more intense.If firms are not credit constrained, Eq. (4) is not binding and the solution is given by:

F 0�Ki;2

� ¼ gðe2Þr (7)

where investment depends on the interest rate and the real exchange rate, Ki;2 ¼ KU2 ðr; e2Þ, and KU

2stands for next period capital of the “unrestricted” firm i. If some of the capital goods are imported(d > 0), an increase in the real exchange rate still causes a drop in investment (a “financial cost” effect),but this effect is now smaller:

vKi;2

ve1¼ dr

F 00�Ki;2

�m0ðe1Þ < 0 (8)

Panel B of Fig. 1 shows that for a highly indebted firm, a large enough real depreciation (e > e**)could generate a strong negative balance-sheet effect that would lead to the financial collapse of thefirm: the G -curve would not intersect the I -curve, investment collapses and the firm’s liquidationfollows. Thus, a discontinuity appears in the firm’s investment function, shown in Panel C of Fig. 1.

6 This is a safe assumption since the empirical evidence for the high persistence of the real exchange rate is quite abundantand can be found in the literature on deviations from PPP or on structural models of the exchange rate. The literature (see Frootand Rogoff (1995) for a review, or Lothian and Taylor (1996) for a referential example) shows that these deviations are verylong-lasting, an indication that the real exchange rate has been generally accepted to be a persistent variable.

7 First and second derivatives of F($) and m($) are denoted as F0($) and m

0($), and F

0 0($).

K i2

K i2

G (K i2) I(K )

4 5 °

G (K )

k *

P A N E L AF irm ’s in v es tm ent

K i2

K i2

G (K i2 ) I(K )

4 5 °

G (K ;e *)

G (K ;e* *)

k *k * *

P A N E L B :R eal deprec ia tion in a nontradab le f irm

e1ei**

k**

k*

ei*

Ii,2=Ki,2

PANEL C:Investment function - nontradable firm

e1

Ii,2=Ki,2

ei++

k*

k**

ei+

PANEL D:Investment function - tradable firm

Fig. 1. Firm’s maximization problem and investment/exchange rate functions of non-tradable and tradable firms.

L. Carranza et al. / Journal of International Money and Finance 30 (2011) 1265–1279 1269

We assume that the only difference among nontradeable firms is their level of foreign debt L�i .8

Given the technology and institutions, there must be a critical level L*(e1), which depends on thereal exchange rate, beyondwhich firms are constrained.When a real depreciation occurs, the fraction ofunconstrained firms is reduced so dL*(e1)/de1 < 0. Letting H(L*) be the cumulative distribution of firmswith foreign debt less than L*, and h(L*) the density function, we can obtain the aggregate investmentfunction of nontradeable firms, INTt , as:

INTt ¼Z L�ðe1Þ

�NKU2 ðr; e2ÞdHðL�Þ þ

Z N

L�ðe1ÞKR2�r; e2; L

�i ; Li

�dHðL�Þ (9)

Taking the derivative of aggregate investment with respect to e1 gives the following expression:

vINTtve1

¼Z L�ðe1Þ

�N

vKU2 ðr;e2Þve1

dHðL�ÞþZ N

L�ðe1Þ

vKR2

�r;e2;L�i ;Li

�ve1

dHðL�Þ

þ�KU2 �KR

2

�hðL�ðe1ÞÞ

dL�ðe1Þde1

<0 (10)

which is negative given that all three terms are negative.

8 This assumption and the parallel one made next in the case of tradeables are merely to avoid the double integration overthe distribution of both Li and L�i , which would make the algebra unnecessarily cumbersome and would add no insight to thediscussion.

L. Carranza et al. / Journal of International Money and Finance 30 (2011) 1265–12791270

2.2. Tradable firms

For a firm producing tradeables, we assume that the revenues are given in foreign currency and sofirm i ’s problem becomes:

Max�e2F

�Ki;2

�� e2L�i � Li � e2rSi;2

�(11)

s:t:gðe2ÞKi;2 � e2Si;2 (12)

e2rSi;2 � q�e2F

�Ki;2

�� e2L�i � Li

�(13)

For credit constrained firms Ki,2 is determined by:

Ki;2 ¼ q

gðe2Þr�e2F

�Ki;2

�� e2L�i � Li

�(14)

where the solution is a fixed point Ki;2 ¼ KR2 ðr; e2; L�i ; LiÞ that depends on r, on the net worth of firm i

(given by L�i and Li) and on the real exchange rate e2. Now, however, an increase in the real exchangerate e1 produces an increase in investment, since the relative value of domestic debt falls with respect tothe value of revenues and the net worth of the company increases, as shown by the followingexpression:

vKi;2

ve1¼ m0ðe1Þ

�qF

�Ki;2

�� drKi;2 � qL�i�

gðe2Þr � qe2F 0�Ki;2

� ¼ m0ðe1Þ�ð1� dÞrKi;2 þ qLi

�e2�gðe2Þr � qe2F 0

�Ki;2

� > 0 (15)

If firms are not constrained, Eq. (13) is not binding and the solution is:

F 0�Ki;2

� ¼ gðe2Þe2

r ¼dþ 1� d

e2

�r (16)

where investment Ki;2 ¼ KU2 ðrÞ still depends on the expected exchange rate (via the relative cost of

capital) if d < 1. The derivative of investment with respect to e1 becomes:

vKi;2

ve1¼ m0ðe1Þ

�dr � F 0

�Ki;2

��e2F 00

�Ki;2

� ¼ �m0ðe1Þð1� dÞre22F

00�Ki;2� (17)

where the second equality follows from Eq. (16). Given m0(e1) > 0 and F

0 0(Ki,2) < 0, the sign of this

derivative is positive: if the proportion of capital imported is not one, then a real depreciation decreasesthe relative cost of capital in a tradable firm (“positive” financial cost effect) so investment increases.Given that the proportion of imported capital goods in highly dollarized countries tends to be large, weexpect d to be large, and therefore investment is probably not very sensitive to the real depreciation.Thus, the investment function of a tradable firm looks like that on Panel D of Fig. 1, where we allow fora massive appreciation (e < eþ) to cause a tradable firm’s bankruptcy and for a level of the real rate(eþþ) beyond which the firm is not financially constrained and the slope of the function, while stillpositive, becomes almost flat.9

9 Notice that Fig. 1 depicts more clearly the investment function of constrained firms. However, both functions (panels C andD) can apply to unconstrained firms, since the constraint depends (in both cases) on the level of the exchange rate e2 and, by thepersistence assumption, on e1. Thus, it could be the case that the exchange rate is so appreciated that a nontradeable firmbecomes unconstrained (initial range of the function in Panel C, from 0 to e�i ) or a tradable firm becomes constrained (in thefunction in Panel D, the range ð0; eþi Þ would imply a collapse of the tradable firm and the range ðeþi ; eþþ

i Þ would imply that thetradable firm is constrained). Of course, if d¼1 then the tradable firm cannot be constrained or collapse, and its investmentfunction would be flat on the exchange rate. Therefore, panels C and D apply also to unconstrained firms: one just has to keep inmind that the investment function of an unconstrained nontradeable firm corresponds to the range ð0; e�i Þ, whereas anunconstrained tradable firm corresponds to the range ðeþþ

i ;NÞ.

L. Carranza et al. / Journal of International Money and Finance 30 (2011) 1265–1279 1271

Assuming that the only difference among tradable firms is their level of domestic debt Li, there isa critical level L(e1), a function of the real exchange rate, beyond which tradable firms are constrained.Now, however, when a real appreciation occurs, the fraction of unconstrained firms is reduced andtherefore dL(e1)/de1 > 0. Let F(L) be the cumulative distribution of tradable firms with domestic debtless than L, and let f(L) be the density function, we obtain the aggregate investment function fortradable firms, ITt , such that:

ITt ¼Z Lðe1Þ

�NKU2 ðrÞdFðLÞ þ

Z N

Lðe1ÞKR2�r; e2; L

�i ; Li

�dFðLÞ (18)

and the first derivative of Eq. (18) with respect to e1 is given by:

vITtve1

¼Z N

L�ðe1Þ

vKR2

�r; e2; L�i ; Li

�ve1

dFðLÞ þ�KU2 � KR

2

�f ðLðe1ÞÞ

dLðe1Þde1

> 0 (19)

which is positive given that both terms in (19) are positive.

2.3. Aggregate investment

If we now calculate aggregate investment, It ¼ ITt þ INTt , we obtain a function of the real exchangerate such as that in Fig. 2, which can be linearized around e

0and e

0 0to obtain a linear aggregate

investment function with a kink which can be represented by Eq. (1). Notice that the slope of the firstsegment (l in Eq. (1)) may be negative or positive: the negative balance-sheet effect for smalldepreciations in the nontradeable sector may not be enough to compensate the competitiveness pluspositive balance-sheet effects in the tradable sector. This does not affect our main result that, when thedepreciation is large, there will appear a stronger negative effect so that the slope of the secondsegment will certainly be lower than that of the first segment (r < 0 in Eq. (1)). In other words,depending on both H(L*) and F(L) the first section of It may be increasing or decreasing on the realexchange rate (l > 0 or l < 0, respectively) but eventually, for a large enough e, the function will bedecreasing or, at least, flatter because of r < 0 (almost flat ITt combined with decreasing INTt ).

Hence, the coefficient r measures the magnitude of the “large depreciation” effect. From the modelit can be seen that this magnitude depends:

Fig. 2. Linearized aggregate investment function.

L. Carranza et al. / Journal of International Money and Finance 30 (2011) 1265–12791272

- Positively on the degree of currency mismatch (liability dollarization) of the economy. In fact, bothl and r are functions of the level of liability dollarization: the negative balance-sheet in non-tradeables is more intense the larger L�i is -regardless of Li - and the positive balance-sheet effect intradeables is less intense the larger L�i compared to Li.

- Positively on the proportion of imported capital goods, denoted by d.- Negatively on the proportion of tradeables in the composition of output (or, as formalized in ourmodel, on the number of firms selling tradeables).

- Positively on the level of indebtedness of the firms, denoted here by the distributions H(L*) andF(L).

- Positively on the extent of the financial friction, denoted by q. This friction, in turn, depends onfactors such as the country’s legal framework -the extent to which repayment of credit contractscan be enforced- and the strength of the banking sector: banks with stronger balance-sheets orwith less currency mismatch in their balance-sheets will tend to lend more.

The above yields implications that can be the object of empirical analyses. We identify severalvariables that should affect the extent of the balance-sheet effect but, also, we stress the role played bythe size of the real depreciation that induces that effect. The following Section deals with the empiricalevidence on these effects: we first review the literature and then present an empirical analysis.

3. Empirical evidence on the effect of a large depreciation on investment

3.1. A brief review of the literature

Empirical evidence on the balance-sheet effects of large depreciations is, so far, scarce and, indeed,very few contributions have linked those effects to nonlinearities in the response of investment.

Interest in the analysis of the relationship between exchange rates and firms’ investment hasrecently focused on the impact of firm’s net worth and balance-sheets. Tornell et al. (2004) analyzedthe Mexican 1994 crisis and found that liquidity effects induced by the real depreciation led to a largedecline in investment for nontradeable firms. Bleakley and Cowan (2010) analyzed the maturitymismatch in firms’ balance-sheets and found a small impact of major exchange rate swings oninvestment. Campa and Goldberg (1999) and Nucci and Pozzolo (2001) examined the effect ofexchange rate fluctuations on investment decisions at the industry and firm level, respectively, andfound that the positive effects of a depreciation are associated to larger shares of exports, and thenegative effects are associated to the share of imported inputs. These studies do not consider balance-sheet effects directly but the results are very much in line with our predictions. Pratap and Urrutia(2004) show that a model where a real depreciation leads to an increase in debt burden and a largedecline in investment is quite consistent with the effects of the Mexican crisis.

Regarding balance-sheet effects and the currency mismatch stemming from dollarization of firms’liabilities, Céspedes (2004) provided evidence that exchange rate depreciations have a less expan-sionary effect when the level of dollar debt is higher. Eichengreen et al. (2003) showed that countrieswith a higher proportion of dollar-denominated foreign debt have more volatile growth of output andcapital flows, and their sovereign debt obtains lower ratings from rating agencies. These papers focusedon aggregate evidence, but the results were not strong. For analyses with a more micro focus theevidence has been, at best, mixed. Bleakley and Cowan (2008) analyzed a sample of firms from fiveLatin American countries and found no negative significant effect of dollar debt on investmentfollowing a depreciation. They argued that this result was due to firms matching the composition oftheir liabilities with their assets and revenue. This paper spawned a series of country-level replications.Bonomo et al. (2003) focused on Brazilian firms and found no significant balance-sheet effect oninvestment. Benavente et al. (2003) analyzed a sample of Chilean firms in the aftermath of the Asiancrisis of the late 1990s but could not conclude that balance-sheet effects on investment were signifi-cant. Echeverry et al. (2003) used a large sample of Colombian firms and found a negative balance-sheet effect on profitability but not on investment. They attributed the lack of prior balance-sheetevidence to the endogeneity of debt choice. Luengnaruemitchai (2003) studied the impact of

L. Carranza et al. / Journal of International Money and Finance 30 (2011) 1265–1279 1273

depreciations on investment in nonfinancial firms in Asia and found no negative significant effect ofthe interaction between dollar debt and the rate of currency depreciation.

Currency matching has been given as the main explanation for the lack of strong evidence: firmswith more dollar debt are affected by a contractionary balance-sheet effect but this could be offset bya competitiveness effect derived from dollar-denominated assets or from income being positivelycorrelated with the currency depreciations. Once the competitiveness effect is controlled for, thebalance-sheet effect appears, in fact, important and statistically significant (Cowan et al., 2005). Aguiar(2005) finds that after the 1994 crisis, therewas a contraction on investment byMexican firms -even inthe tradable sector- driven by their weak balance-sheet position. Pratap et al. (2003) also analyzedMexican firms and found that holding dollar debt in a devaluation adversely affected firms’ earningsand investment, and in the aggregate there was some evidence of balance-sheet effects resulting fromimperfect currency matching by firms. Galiani et al. (2003) found evidence of a negative impact of theinteraction between real depreciations and the level of dollar debt on firm’s investment in Argentina.Finally, Carranza et al. (2003) analyzed Peruvian firms and found that, for firms holding dollar-denominated debt, investment decisions were negatively affected by real exchange rate deprecia-tions: in periods of real exchange rate stability (prior to 1997), the high level of foreign currency couldnot explain investment behavior, whereas in periods of large real exchange rate depreciation the effectbecame negative and significant.

This latter paper suggested that the size of the real depreciation could be a factor in finding evidenceof balance-sheet effects, and it therefore links directly with our paper. Analyses of nonlinearities in theeffect of exchange rate changes have uncovered significant effects mostly in the context of inflationpass-through. Burstein et al. (2005) focused on large devaluations and showed their impact on infla-tion, given slow adjustment of prices in the nontradeable sector. Pollard and Coughlin (2003) for theUS, Bussiere (2007) for developed economies, Khundrakpam (2007) for India, and Bigio and Salas(2006) for Peru also found empirical evidence consistent with a nonlinear pass-through, althoughno clear cross-country patterns arise. Forbes (2002) went beyond inflation and examined how largedepreciation events during the period 1997–2000 affected firms’ performance: firms with higherindebtedness tended to have lower income growth, but firms with a higher share of foreign salesperformed better after depreciations. Darby et al. (1999) focused on investment in OECD countries andfound some evidence of nonlinearities in the reaction to real exchange rate changes. Carranza et al.(2009) brought together the nonlinear effects of exchange rates and balance-sheet considerations.These authors found, in a cross-country panel, that inflation pass-through was significantly affected bythe level of dollarization and a negative balance-sheet effect appeared for large depreciations, whereasfor small depreciations this effect was overwhelmed by the competitiveness effect. These authors alsosuggested that the nonlinearity was due to the impact of exchange rate changes on aggregateinvestment.

3.2. Some new empirical evidence of nonlinearities in the investment-real exchange rate relationship

Most of the empirical evidence described comes from firm level studies that focus on one country ora small number of countries considered separately. Our model, however, generates testable predictionsalso at the aggregate level. A full analysis that controls for all the factors mentioned in Section 2 isbeyond the scope of this paper but we believe it is important to provide some evidence that may serveas a guide for future analyses.

We carry out a country-level analysis by collecting data on real aggregate investment, real exchangerates and openness measures (exports plus imports over GDP) from the IFS database of the IMF. Inorder to control for the degree of dollarization, we use the measure of dollarization developed byReinhart et al. (2003). This is a time invariant measure so, even though dollarization tends to be quiteconstant -a fact noted by Reinhart et al. (2003)- we are aware that it may be difficult to identifycorrectly the effect of the dollarization level on the nonlinearity.10 A final set of control variables

10 Alternative measures of dollarization are available for a very limited set of countries or years. We leave a deeper investi-gation of the impact of this variable for future research.

Table 2Large real depreciations and investment growth: baseline equations.

Value of 4 4¼5% 4¼10% 4¼15% 4¼20%

DIit�1 �0.162 (�0.56) �0.16 (�0.55) �0.149 (�0.55) �0.201 (�0.56)Deit 0.021 (0.26) 0.015 (0.2) 0.003 (0.04) �0.026 (�0.40)Dit 0.028 (1.32) 0.045 (1.14) 0.031 (0.5) 0.125 (1.34)Dit$Deit �0.441 (�2.51) �0.499 (�2.47) �0.438 (�1.86) �0.602 (�2.19)NC 73 73 73 73N4 383 170 101 66N 1665 1665 1665 1665R2 0.053 0.054 0.057 0.039

Note: 2SLS-dynamic panel estimates of investment growth equations for fixed 4:DIit ¼ d0i þ d0t þ d1DIit�1 þd2Deit þ d3Dit þ d4DitDeit þ uit. Dit ¼ 1(Deit > 4) is a dummy variable that takes value one if the real depreciation exceeds somelevel 4. NC: number of countries included in the sample; N4: number of large depreciation observations; N : total number ofcountry-year observations. Robust t-stats are shown in parenthesis. Three lags of the real depreciation rate are used asinstruments. Time and country effects are included in all specifications.

L. Carranza et al. / Journal of International Money and Finance 30 (2011) 1265–12791274

accounts for the exchange rate regimes to which the countries are subject. Theoretical analyses haveshown that balance-sheet effects ought to lead to greater falls in output and investment underintermediate regimes than under flexible rates (see Cespedes et al., 2004; Magud, 2010). We includetwo dummy variables which control for intermediate (dirty floats or crawling pegs) and floatingregimes. These data come from Levy-Yeyati and Sturzenegger (2003, updated online).

Given the data availability for emerging economies, we use annual data from 1970 to 2007. Our finalpanel contains 73 countries �71 in some specifications- and a maximum of 38 time periods percountry. Countries with missing observations at the beginning or at the end of the sample wereincluded so our panel is unbalanced. In every panel estimated we have around 1600 country-yearobservations available (see Tables 2, 3a and 3b).11 We estimate reduced-form dynamic panels forinvestment of the form:

DIit ¼ d0i þ d0t þ d1DIit�1 þ d2Deit þ d3Dit þ d4DitDeit þ d0Xðopitintit floitbsiÞ þ uit (20)

where DIit is the growth rate in aggregate real investment of country i during year t, Deit is the growthrate in the real exchange rate (an increase meaning a real depreciation), Dit ¼ 1(Deit>4) is an indicatorfunction that takes value one if the real depreciation exceeds some level 4 and zero otherwise, andX(opit intit floit bsi) is a vector of controls which includes openness ratios opit; the exchange rate regimedummies intit for intermediate regimes (dirty floats or crawling pegs) and floit for floating regimes, andthe level of dollarization bsi as a measure of balance-sheet effects. These control variables are includedin levels and interacted with Deit and/or Dit in the different specifications. Note that the above reduced-form equation mimics Eq. (1) and can be rationalized as the dynamic version of the investmentequation which is the outcome of our model. Our main interest in this specification lies on theparameters attached to the variable Dit.

To estimate these panels, we use the 2SLS procedure in Arellano (2003) for dynamic panels withlagged dependent and predetermined variables in a large T setting. We take three lags of the depre-ciation rate as instruments, along with all the control variables.12 Year and country dummies are

11 The final list of countries included in the smallest analysis (Table 2b) is: Australia, Austria, The Bahamas, Bahrain, Belgium,Belize, Bolivia, Bulgaria, Burundi, Cameroon, Canada, Central African Republic, Chile, Colombia, Costa Rica, Cote d’Ivoire, Croatia,Cyprus, Czech Republic, Denmark, Dominica, Dominican Republic, Ecuador, Fiji, Finland, France, Germany, Ghana, Greece,Grenada, Guyana, Hungary, Iceland, Iran, Ireland, Israel, Italy, Japan, Lesotho, Luxembourg, Malawi, Malaysia, Malta, Morocco,The Netherlands, New Zealand, Nicaragua, Nigeria, Norway, Pakistan, Papua New Guinea, Paraguay, Philippines, Poland,Portugal, Romania, Saudi Arabia, Sierra Leone, Slovak Republic, South Africa, Spain, St. Kitts and Nevis, St. Lucia, St. Vincent &Grens., Sweden, Switzerland, Togo, Trinidad and Tobago, Tunisia, Uganda, United Kingdom, United States, Uruguay, Venezuela,Zambia. We believe this set to be quite representative.12 This procedure is consistent, for large T, even if the instruments are only predetermined. Becker et al. (1994), for example,use this same approach for a panel shorter than ours (T ¼ 31).

Table 3aLarge real depreciations and investment growth.

Controls Openness ratios Value of 4 Exchange rate regimes

Value of 4 10% 15% 20% 10% 15% 20%

DIit�1 �0.226 (�0.75) �0.214 (�0.71) �0.25 (�0.85) DIit�1 �0.134 (�0.47) �0.135 (�0.47) �0.202 (�0.7)Deit �0.192 (�1.44) �0.199 (�1.57) �0.264 (�2.17) Deit 0.017 (0.16) 0.02 (0.21) �0.006 (�0.07)Dit 0.049 (1.19) 0.038 (0.6) 0.134 (1.42) Dit 0.059 (1.47) 0.063(0.98) 0.169 (1.76)Dit$Deit �0.215 (�0.79) �0.148 (�0.50) �0.127 (�0.36) Dit$Deit �0.45 (�1.99) �0.464 (�1.83) �0.668 (�2.22)op it 0.158 (2.96) 0.157 (2.94) 0.165 (3.11) floit$Deit �0.011 (�0.06) �0.02 (�0.11) �0.028 (�0.16)opit$Deit 0.329 (1.69) 0.329 (1.73) 0.388 (2.11) Dintit$Deit �0.003 (�0.02) �0.043 (�0.31) �0.051 (�0.38)opit$Dit$Deit �0.433 (�1.09) �0.481 (�1.16) �0.824 (�1.86) floit$Dit$Deit �0.035 (�0.12) �0.015 (�0.05) 0.032 (0.11)

intit$Dit$Deit �0.904 (�2.58) �0.772 (�1.97) �0.856 (�1.86)NC 73 73 73 NC 73 73 73N4 170 101 66 N4 170 101 66N 1663 1663 1663 N 1665 1665 1665R2 0.042 0.047 0.033 R2 0.054 0.057 0.039

Note: 2SLS-dynamic panel estimates of investment growth equations for fixed 4: DIit ¼ d0i þ d0t þ d1DIit�1 þ d2Deit þ d3Dit þ d4DitDeit þ d0Xt þ uit. Dit ¼ 1(Deit > 4) is a dummy variable that

takes value one if the real depreciation exceeds some level 4. NC: number of countries included in the sample; N4: number of large depreciation observations; N : total number of country-year observations. Robust t-stats are shown in parenthesis. Three lags of the real depreciation rate are used as instruments. Time and country effects are included in all specifications. Xt

contains the different controls (see main text for the definition of the variables). The “Exchange rate regimes” equation also includes intit and floit in levels, but the coefficient estimates arenot shown.

L.Carranzaet

al./Journal

ofInternational

Money

andFinance

30(2011)

1265–1279

1275

Table 3bLarge real depreciations and investment growth

Controls Dollarizationindex Value of 4 Allcontrols

Value of 4 10% 15% 20% 10% 15% 20%

DIit�1 �0.124 (�0.44) �0.117 (�0.41) �0.18 (�0.64) DIit�1 �0.142 (�0.50) �0.142 (�0.50) �0.231 (�0.79)Deit 0.081 (0.86) 0.065 (0.74) 0.036 (0.41) Deit �0.126 (�0.81) �0.131 (�0.89) �0.209 (�1.46)Dit 0.073 (1.46) 0.076 (0.98) 0.215 (1.92) Dit �0.059 (�0.63) �0.150 (�1.05) 0.030 (0.09)Dit$Deit �0.653 (�2.62) �0.621 (�2.15) �0.859 (�2.63) Dit$Deit 0.129 (0.32) 0.384 (0.80) 0.065 (0.08)bsi �0.000 (�0.02) �0.000 (�0.02) �0.000 (�0.04) opit$Deit 0.290 (1.49) 0.296 (1.56) 0.372(2.01)bsi$Deit �0.012 (�1.28) �0.011 (�1.26) �0.011 (�1.3) opit$Dit$Deit �1.097 (�1.64) �1.606 (�1.91) �1.423 (�0.83)bsi$Dit �0.006 (�0.93) �0.008 (�1.00) �0.025 (�1.41) intit$Dit$Deit �1.013 (�2.72) �0.855 (�2.05) �0.889(�1.65)bsi$Dit$Deit 0.032 (1.11) 0.036 (1.12) 0.072 (1.45) bsi$Deit �0.011 (�1.12) �0.009 (�1.00) �0.009 (�1.00)

bsi$Dit$Deit 0.023 (0.43) 0.030 (0.52) 0.061 (0.82)opit$bsi$Dit$Deit �0.017 (�0.20) �0.035 (�0.39) �0.092 (�0.76)

NC 71 71 71 NC 71 71 71N4 170 101 66 N4 170 101 66N 1637 1637 1637 N 1635 1635 1635R2 0.066 0.067 0.049 R2 0.078 0.076 0.046

Note: 2SLS-dynamic panel estimates of investment growth equations for fixed : 4:DIit ¼ d0i þ d0t þ d1DIit�1 þ d2Deit þ d3Dit þ d4DitDeit þ d0Xt þ uit. Dit ¼ 1(Deit > 4) is a dummy variable that

takes value one if the real depreciation exceeds some level 4. NC : number of countries included in the sample; N4: number of large depreciation observations; N : total number of country-year observations. Robust t-stats are shown in parenthesis. Three lags of the real depreciation rate are used as instruments. Time and country effects included in all specifications. The “Allcontrols” panel includes Xt ¼ (opit, opit$Deit, opit$Dit, opit$Dit$Deit, intit, intit$Deit, intit$Dit$Deit, bsi, bsi$Deit, bsi$Dit$Deit, opit$bsi$Dit$Deit) but not all coefficients are shown.

L.Carranzaet

al./Journal

ofInternational

Money

andFinance

30(2011)

1265–1279

1276

L. Carranza et al. / Journal of International Money and Finance 30 (2011) 1265–1279 1277

included in the reduced-form equation. We are not directly interested in estimating 4 but rather ingiving evidence that large real depreciations have a negative impact on investment growth, and thatthis negative impact may be linked to openness and to liability dollarization.We therefore show resultsconditional on specific values of 4.

Table 2 shows the baseline specification where we do not include controls. Results are quiteenlightening. First, investment growth is not persistent: the coefficient of lagged growth is neversignificant in any of the specifications (see, e.g., Benavente et al., 2003). The change in the real exchangerate is also not significant: i.e., there is little aggregate evidence that real depreciations affect invest-ment positively. However, d4, the parameter attached to the interaction between the real depreciationand Dit is significant and negative, as expected from our conclusions. This result is quite robust to thethreshold 4. Indeed, Panel A of Fig. 3 shows how the t-statistic of d4 in the investment growth equationchanges with 4 for a grid of values in the range [0,0.25]. We use this range since we are interested inlarge depreciations but few countries experienced real depreciations larger than 25%. The estimatedcoefficient is significant for most relevant values of 4 (note that for 4 > 0.2 the number of depreciationepisodes that can be considered large is already small).

This negative impact of large depreciations on investment growth is robust to the inclusion ofcontrols in the investment equation, though admittedly some of these are quite crude measures. Table3 shows parameter estimates for empirical specifications that include openness measures or exchangerate regime dummies (Table 3a) or the dollarization index (Table 3b). Interestingly enough, in the caseof openness, d4 is still negative but insignificant. The nonlinear effect is then captured by two terms.First, the interaction (opit$Deit) has a positive and significant effect, so openness is positively related tothe effect of real depreciations on investment, as implied by our model. However, when the

Panel A: Baseline specification

-0.5

0

0.5

0 0.05 0.1 0.15 0.2 0.25

Panel B: Interaction with Openness

0

0.5

-3

-2.5

-2

-1.5

-1

-2.5

-2

-1.5

-1

-0.5

dit· Δ eit openit·dit· Δeit

Panel D: Dollarization index

1

-0.5

0

0.5

Panel C: Differences across Exchange Rate Regimes

-0.5

0

0.5

1

0 0.05 0.1 0.15 0.2 0.25

0 0.05 0.1 0.15 0.2 0.25

0 0.05 0.1 0.15 0.2 0.25

-3

-2.5

-2

-1.5

-

dit· Δ eit bsi· Δeit

-3.5

-3

-2.5

-2

-1.5

-1

dit· Δ eit: Floating dit· Δ eit: Intermediate dit· Δ eit: Baseline (Fixed)

Fig. 3. Robust t - statistic of estimated values of parameter d4 (Panels A, C and D) and interactions with controls (Panels B, C and D) asa function of the threshold 4 for a real depreciation to be considered large.

L. Carranza et al. / Journal of International Money and Finance 30 (2011) 1265–12791278

depreciation rate is large the effect becomes negative, so large depreciations are again associated withlower investment growth. Panel B of Fig. 3 shows how this effect is also robust to the value of 4. Theexchange rate regime matters in an interesting way. In the case of floating and fixed regimes thenegative relationship postulated appears: note the negative and significant value of the estimate of d4,which measures the effect for fixed regimes, and the fact that the additional effect measured by theinteraction (floit$Dit$Deit) is not significant. However, for intermediate regimes the effect is still larger:note the negative and significant parameter of the interaction (intit$Dit$Deit). Thus, intermediateregimes seem to lead to more negative impacts of real depreciations, a finding that deserves furtheranalysis. Again, Panel C of Fig. 3 shows the robustness to the choice of 4. Finally, the dollarization indexdoes not provide us with strong evidence: parameters attached to bsit are not in general significant,although the relationships go in the direction expected. Panel D of Fig. 3 shows the evolution of d4-which stays significant throughout- and of the parameter attached to the interaction (bsit$Deit).Results with a fuller set of controls do not add much to what we have presented so far, so we do notcomment on them.

The analysis shows aggregate evidence consistent with the implications of the model, even thoughit is not meant to be a comprehensive test. Admittedly, some of the factors we control for deservefurther attention -for example, more complete measures of liability dollarization should help uncoverother significant results- but the general evidence supports the negative effect of large depreciationson investment and the fact that this negative effect is related to openness and to liability dollarizationor currency mismatch.

4. Conclusion

We have shown that in a small open economy with currency mismatch (liability dollarization) thepresence of a simple financial friction not only generates the traditional balance-sheet of a realdepreciation, but also a possible “large depreciation” effect. This effect may lead to a kink in theinvestment/real exchange rate function so that it becomes downward sloping or, at least, its positiveslope is significantly reduced when the real depreciation is large. Contractionary balance-sheet effectscould, therefore, be empirically noteworthy only in the presence of large enough depreciations.Furthermore, these effects should be more noticeable in countries that are highly indebted, that havepoorly developed financial markets, low proportion of tradeables in the composition of output or highproportion of imported capital, and that show a high level of currency mismatch in firms’ liabilities.

Our results are, therefore, quite relevant in order to extend the empirical literature on the effects ofdepreciations for emerging markets, which tend to be open countries with both high degrees of dol-larization and large exchange rate swings (Bigio and Salas, 2006; Leiderman et al., 2006; Ca’Zorzi et al.,2007). We have presented new empirical evidence supportive of the model’s conclusions, althoughmore empirical work is needed in order to understand, especially, the implications of institutionalfactors such as the development of the financial system or the impact of the extent of the currencymismatch and the availability of hedging instruments. We believe, in any case, that our emphasis onthe necessary large size of the real depreciation may help uncover the, so far, quite elusive balance-sheet effects.

Acknowledgments

Jose E. Galdon-Sanchez and Javier Gomez-Biscarri acknowledge financial support from the SpanishMinisterio de Ciencia e Innovacion from projects ECO2008-02641 and ECO2008-05155, respectively.Javier Gomez-Biscarri also thanks the Barcelona GSE. The usual disclaimers apply.

References

Aghion, P., Bacchetta, P., Banerjee, A., 2001. Currency crises and Monetary Policy in an economy with credit constraints.European Economic Review 45, 1121–1150.

Aguiar, M., 2005. Investment, devaluation, and foreign currency exposure: the case of Mexico. Journal of DevelopmentEconomics 78, 95–113.

Arellano, M., 2003. Panel Data Econometrics. Oxford University Press, Oxford, U.K.

L. Carranza et al. / Journal of International Money and Finance 30 (2011) 1265–1279 1279

Batini, N., Levine, P., Pearlman, J., 2007. Optimal Exchange Rate Stabilization in a Dollarized Economy with Inflation Targets.Mimeo, London Metropolitan University.

Becker, G., Grossman,M.,Murphy, K.,1994. Anempirical analysis of Cigarette Addiction.American Economic Review84 (3), 396–418.Benavente, J.M., Johnson, C.A., Morande, F.G., 2003. Debt composition and balance sheet effects of exchange rate depreciations:

a firm level analysis for Chile. Emerging Markets Review 4, 397–416.Bigio, S., Salas, J., 2006. Non-linear Effects of Monetary Policy and Real Exchange Rate Shocks in Partially Dollarized Economies:

An Empirical Study for Peru Working Paper of the Banco Central de Reserva del Peru 2006–008.Bleakley, H., Cowan, K., 2002. Corporate Dollar debt and Depreciations: Much Ado About Nothing?. Working Paper University of

Chicago Business School.Bleakley, H., Cowan, K., 2008. Corporate dollar debt and depreciations: much Ado about Nothing? Review of Economics and

Statistics 90 (4), 612–626.Bleakley, H., Cowan, K., 2010. Maturity mismatch and financial crises: evidence from emerging market corporations. Journal of

Development Economics 93, 189–205.Bonomo, M., Martins, B., Pinto, R., 2003. Debt composition and exchange rate balance sheet effect in Brazil: a firm level analysis.

Emerging Markets Review 4, 368–396.Burstein, A., Eichenbaum, M., Rebelo, S., 2005. Large devaluations and the real exchange rate. Journal of Political Economy 113,

742–784.Bussiere,M., 2007. ExchangeRatePass-Through toTradePrices: theRoleofNon-Linearities andAsymmetries ECBWorkingPaper822.Ca’Zorzi, M., Hahn, E., Sanchez, M., 2007. Exchange Rate Pass-Through in Emerging Markets ECB Working Paper 739.Céspedes, L.F., 2004. Financial Frictions and Real Devaluations. Central Bank of Chile. Mimeographed.Calvo, G.A., Leiderman, L., Reinhart, C.M., 1994. Inflows of Capital to Developing Countries in the 1990s: Causes and Effects Inter-

American Development Bank Office of the Chief Economist Working Paper 302.Campa, J.M., Goldberg, L.S., 1999. Investment, pass-through, and the exchange rates: a cross-country comparison. International

Economic Review 40, 287–314.Carranza, L., Cayo, J.M., Galdon-Sanchez, J.E., 2003. Exchange rate volatility and Economic performance in Peru: a firm level

analysis. Emerging Markets Review 4, 472–496.Carranza, L., Galdon-Sanchez, J.E., Gomez Biscarri, J., 2009. Exchange rate and inflation dynamics in dollarized economies.

Journal of Development Economics 89 (1), 98–108.Cespedes, L., Chang, R., Velasco, A., 2004. Balance-sheet and exchange rate Policy. American Economic Review 94 (4), 1183–1193.Choi, W.G., Cook, D., 2004. Liability dollarization and the bank balance sheet Channel. Journal of International Economics 64,

247–275.Cowan, K., Hansen, E., Herrera, L.O., 2005. Currency Mismatches, Balance Sheet Effects and Hedging in Chilean Nonfinancial

Corporations Central Bank of Chile Working Paper 346.Darby, J., Hughes Hallet, A., Ireland, J., Piscitelli, L., 1999. The impact of exchange rate variability on the level of investment.

Economic Journal, 55–67.Echeverry, J.C., Fergusson, L., Steiner, R., Aguilar, C., 2003. `Dollar’ debt in Colombian firms: are sinners punished during

devaluations? Emerging Markets Review 4, 417–449.Eichengreen, B., Hausmann, R., Panizza, U., 2003. The Pain of Original Sin. University of California at Berkeley, Harvard

University, and Inter-American Development Bank. Mimeographed.Forbes, K., 2002. How do large depreciations affect firm performance? IMF Staff Papers 49, 214–238.Froot, K.A., Rogoff, K., 1995. Perspectives on PPP and long-run real exchange rates. In: Gene Grossman, M., Rogoff, K. (Eds.),

Handbook of International Economics, vol. 3. North-Holland, Amsterdam, pp. 1647–1688.Galiani, S., Levy-Yeyati, E., Schargrodsky, E., 2003. Financial dollarization and debt deflation under a currency board. Emerging

Markets Review 4, 340–367.Galindo, A., Panizza, U., Schiantarelli, F., 2003. Debt composition and balance sheet effects of currency depreciation: a Summary

of the micro evidence. Emerging Markets Review 4, 330–339.Ize, A., Levy-Yeyati, E., 2005. Financial De-Dollarization: Is It for Real? IMF Working Paper 05/187.Kehoe, T., Ruhl, K., 2009. Sudden Stops, sectoral Reallocations, and the real exchange rate. Journal of Development Economics 89

(2), 235–249.Khundrakpam, J.K., 2007. Economic Reforms and Exchange Rate Pass-Through to Domestic Prices in India BISWorking Paper 225.Kiyotaki, N., Moore, J.H., 1997. Credit Cycles. Journal of Political Economy 105, 211–248.Leiderman, L., Maino, R., Parrado, E., 2006. Inflation Targeting in Dollarized Economies IMF Working Paper 06/157.Levy-Yeyati, E., Sturzenegger, F., 2003. A de Facto Classification of Exchange Rate Regimes: A Methodological Note. Data

Appendix for the paper: To Float or to Fix: Evidence on the Impact of Exchange Rate Regimes on Growth. AmericanEconomic Review 93 (4), 1173–1193. Available at: http://www.aeaweb.org/aer/contents/.

Lothian, J.R., Taylor, M.P., 1996. Real exchange rate behavior: the recent float from the Perspective of the Past two Centuries.Journal of Political Economy 104 (3), 488–589.

Luengnaruemitchai, P., 2003. The Asian Crisis and the Mystery of the Missing Balance Sheet Effect UC Berkeley working paper.Magud, N., 2010. Currency mismatch, openness and exchange rate regime choice. Journal of Macroeconomics 32, 68–89.Nucci, F., Pozzolo, A.F., 2001. Investment and the exchange rate: an analysis with firm-level panel data. European Economic

Review 45, 259–283.Pollard, P.S., Coughlin, C.C., 2003. Size Matters: Asymmetric Exchange Rate Pass-Through at the Industry Level Federal Reserve

Bank of St. Louis Working Paper 2003-029C.Pratap, S., Urrutia, C., 2004. Firm dynamics, investment and debt portfolio: balance sheet effects of the Mexican crisis of 1994.

Journal of Development Economics 75, 535–563.Pratap, S., Lobato, I., Somuano, A., 2003. Debt composition and balance sheet effects of exchange rate volatility in Mexico: a firm

level analysis. Emerging Markets Review 4, 450–471.Reinhart, C., Rogoff, K., Savastano, M., 2003. Addicted to Dollars NBER Working Paper 10015.Tornell, A., Westerman, F., Martinez, L., 2004. NAFTA and Mexico’s Less than Stellar Performance NBER Working Paper 10289.