Sovereign Risk, Currency Risk, and Corporate Balance Sheets

Abstract

During the 1980s, 1990s, and early 2000s, a number of sovereign debt crises engulfed emerging markets. While the details of each sovereign debt crisis were different, the broader story remained similar: the government borrowed from foreign investors in foreign currency (FC) during good times only to later default on their external debt as economic conditions deteriorated. Following these crises, emerging market governments curtailed their FC borrowing and moved toward borrowing in their local currency (LC). Using a newly constructed comprehensive data set on the currency composition of sovereign and corporate external debt, we find that from 2003 to 2017 major emerging market sovereigns went from having around 80|$\%$| of their external debt in FC to having more than half of their external sovereign debt in their own currency. By contrast, even as governments were dramatically changing the way they finance themselves, the private sector continued to borrow from foreigners almost entirely in FC.

Despite their shift toward LC debt, emerging market governments continue to compensate investors for sovereign default risk, paying positive credit spreads on their LC and FC debt. In this paper, we argue that the private sector’s reliance on external FC debt raises the cost of inflating away sovereign debt, leading to elevated sovereign credit spreads even in the presence of a shift toward LC sovereign debt. We find that heterogeneity in the private sector’s FC exposure helps explain sovereign default risk across countries and over time, above and beyond the debt structure of the government.

We argue that corporate borrowing helps understand sovereign risk because private sector currency mismatch constrains the government from using inflation to reduce its real fiscal burden. When the government faces fiscal difficulties, corporate mismatch therefore raises the risk of outright sovereign default by making the alternative means of reducing real repayment, inflation, less appealing. The idea that corporate balance sheet mismatch could make depreciations contractionary was studied extensively following the Asian Financial Crisis.¹ One key contribution of this paper is to demonstrate that currency mismatch on corporate balance sheets can be a source of sovereign default risk.

We begin by constructing a new data set on the sectoral composition of cross-border borrowing and use it to provide a systematic account of the currency composition of external debt issued by the sovereign and corporate sector in 14 major emerging markets between 2003 and 2017, where “external” debt is defined as borrowing from foreign investors.² We find that emerging market sovereigns increasingly borrow in LC from foreign investors, while corporate external liabilities still remain largely in FC. From 2003 to 2017, the average LC share in external sovereign debt increased from around 21.7|$\%$| to almost 59.1|$\%$|⁠. However, during this same period, the LC share in external corporate sector debt increased relatively little from 8.5|$\%$| to 13.0|$\%$|⁠.

The composition of corporate balance sheets has significant implications for sovereign credit risk. We use our aggregate cross-country data set on the currency composition of external liabilities to show that a higher reliance on external FC corporate financing is associated with a higher default risk on sovereign debt. In a panel regression, we demonstrate that increases in the LC share in the external corporate debt are robustly associated with reductions in sovereign credit risk. In addition, an increase in the LC corporate debt share also reduces the sensitivity of EM sovereign credit spreads to exchange rate fluctuations. Furthermore, we present micro-evidence showing that firms’ FC corporate bond issuance is associated with a relatively larger increase in sovereign credit default swap (CDS) spreads than LC corporate bond issuance about 1 week after the issuance date.

A high reliance on FC debt leads to currency mismatch on corporate balance sheets. While firms may borrow in FC debt to hedge their FC revenues (known as “operating hedges”) or to reduce their funding costs on the currency-hedged basis using foreign exchange derivatives (known as “financial hedges”), we show that neither operating hedges nor financial hedges can explain the full extent of FC corporate borrowing in emerging markets. Instead, we observe significant FC borrowing by firms with and without operational hedging motives. In particular, the distribution of the FC debt shares across firms is quite similar for firms with high and low export revenues, after controlling for country fixed effects and firm size. Furthermore, using data on cross-currency derivatives outstanding, we show that the ability for emerging market firms to hedge their FC debt using derivatives is generally quite limited because of the small size of the currency swap markets relative to the size of FC debt outstanding.

Motivated by these dramatic changes in emerging market borrowing and the empirical evidence on the importance of private FC debt for sovereign risk, we introduce LC sovereign debt and a corporate sector with FC external liabilities and LC revenues into the canonical quantitative sovereign debt model (Aguiar and Gopinath, 2006; Arellano, 2008). The model demonstrates quantitatively that the borrowing patterns of the private sector can have large effects on the nature of sovereign risk by affecting a government’s optimal inflation policy. When the private sector is highly mismatched, meaning private debt is overwhelmingly in FC but revenues are in LC, the sovereign is reluctant to allow an exchange rate depreciation to reduce the real value of its debt, generating a “fear of floating” as in Calvo and Reinhart (2002). In this case, when the government considers whether to default or use inflation to reduce the fiscal burden of sovereign debt repayments, it is relatively more inclined to explicitly default than to inflate away the debt because of the effect of depreciation on the private sector. A calibration of the dynamic model produces simulated moments of currency and credit risk very similar to the cross-country mean empirical moments. Further, it demonstrates that moving toward local currency sovereign debt reduces default risk much more if it is accompanied by an increase in the local currency corporate debt share.

The paper contributes to a number of literatures in international finance and macroeconomics. Our measurement of currency mismatch at the sectoral level provides a contribution to the literature on the causes and consequences of the currency composition of external debt (Eichengreen and Hausmann, 1999; Lane and Shambaugh, 2010). By measuring the external currency composition of emerging market borrowing at a more disaggregated sectoral level and distinguishing between types of borrowing, we are able to pinpoint the sources of change in the currency composition of external debt. Our empirical contribution is to demonstrate that the currency composition of corporate debt is a powerful explanatory variable for pricing sovereign default risk. This contributes to the literature documenting a range of determinants of sovereign risk, such as U.S. interest rates (Uribe and Yue, 2006), global financial conditions (Longstaff et al., 2011; Borri and Verdelhan, 2011), macroeconomic fundamentals (Tomz and Wright, 2007), debt maturity (Arellano and Ramanarayanan, 2012; Broner, Lorenzoni, and Schmuckler, 2013), or the existence of CDS markets (Salomao, 2017). Our modeling of local currency bonds as defaultable builds on the insights of Reinhart and Rogoff (2008, 2011), who emphasize the long history of default on domestic debt. Wu (2021) explores the role of risk premiums in driving this connection between corporate FC borrowing and sovereign credit spreads. Finally, we contribute to theoretical sovereign debt literature by introducing local currency debt, corporate balance sheet mismatch, and the choice of inflation into the canonical sovereign debt model. These additional features allow us to perform counterfactuals on the effect of changes in the currency composition of government and corporate debt on the probability of sovereign default.

1. The Changing Composition of External Portfolios

In this section, we combine various national and international data sources to construct measures of the currency composition of the external liabilities of the sovereign and corporate sectors in 14 major emerging markets. We document that emerging market sovereigns have shifted away from borrowing externally in FC to borrowing primarily in LC. However, we show that the external liabilities of the corporate sector remain largely denominated in FC.³

1.1. Measuring currency composition of external debt by sector

1.1.1 Overview

There are a number of empirical challenges when characterizing the currency composition of a country’s external liabilities. First, foreign-owned debt may not be issued in international markets and so one needs to measure foreign ownership of debt issued in the domestic markets. Second, because a large amount of lending to emerging market corporates is intermediated through offshore subsidiaries, nationality-based data provide a more economically meaningful measure of emerging market external borrowing.⁶ Third, many aggregated data sets do not provide a clean distinction between the sector of borrowing (i.e., nonfinancial corporate, financial corporate, or sovereign).

To overcome these challenges, we use a host of data sources from international organizations, national authorities, and commercial vendors, including International Debt Statistics (IDS) and Locational Banking Statistics (LBS) from the Bank of International Settlements (BIS); data on foreign ownership of domestic government debt and total government debt outstanding from national sources, Haver Analytics, Asian Bonds Online (ABO); individual bond and loan-level issuance level from Refinitiv SDC Platinum and FactSet Debt Capital Structure; and mutual fund holdings data from Morningstar.

1.1.2. Debt securities

This method will underestimate FC debt if countries borrow in foreign currencies other than the USD or EUR, such as the British pound (GBP), Swiss franc (CHF), or the Japanese yen (JPY).

Besides providing estimates for all domestically corporate bonds, we also follow the same methodology and provide separate estimates for the financial sector and the nonfinancial corporate sector.

1.1.3. Cross-border bank loans

The final component of Equation (1) we need to measure is cross-border loans. We start with the BIS LBS data (⁠|$L_{s,c}^{LBS}$|⁠) as our source of aggregate data, and we will use SDC Platinum syndicated loan data (⁠|$L_{s,c}^{SDC}$|⁠) to further separate sectors within the LBS data.⁹

The BIS LBS provides quarterly data on cross-border financial claims and liabilities of banks resident in the BIS reporting countries. The level of external loans for country |$i$| is given by the total claims of all BIS reporting countries against counterparty country |$i$|⁠. Since most developed, large developing countries, and offshore financial centers are BIS reporting countries, the aggregate lending of BIS reporting countries to country |$i$| represents the majority of private sector cross-border loans from the rest of the world to country |$i$|⁠.

The BIS LBS breaks down cross-border loans into claims denominated in the USD, EUR, JPY, GBP, CHF, and residual currencies. From an emerging market country |$i$|’s perspective, the amount of loans and deposits denominated in the residual currencies gives a very good proxy of the level of loans and deposits denominated in the LC of country |$i$|⁠.

We find support for this approximation by comparing our values for |$\widehat{CBL}_{govt,c,res}$| and |$\widehat{CBL}_{nfc,c,res}$| to |$CBL_{govt,c,res}^{LBS}$| and |$CBL_{nfc,c,res}^{LBS}$|⁠, where both are available from 2013, finding they closely align.

To estimate the tradable and nontradable sector external debt series, we combine our bond and loan series for nonbanks with the tradable and nontradable sector shares of nonbank bond and loan debt in the SDC Platinum data.¹² We scale our nonfinancial corporate series by the ratio of tradable and nontradable sector debt to estimate the amount of external loans and bonds for each sector by currency. Internet Appendix A.6 describes the procedure in more detail.

1.2. Stylized facts on the currency composition of external debt

Using our data set on the currency composition of external debt by sector, we document a major shift in the currency composition of sovereign external borrowing alongside a relatively limited change in the composition of corporate borrowing. In particular, we find that the share of external sovereign debt in LC increased from 20|$\%$| to 60|$\%$|⁠, but the LC share of corporate debt is relatively unchanged at 10|$\%$|⁠. Figure 1 plots the cross-country mean of the share of sovereign, corporate, and total debt in LC from 2003 to 2017, with the shift in the aggregate share of debt in LC sovereign debt accounting for the bulk of the increase the LC share in the aggregate external debt. Figure 2 plots the mean external LC and FC debt by sector as a percentage of GDP. Externally held LC debt is primarily issued by the sovereign, whereas the bulk of external FC debt comes from the corporate sector. Figure A.6 in the Internet Appendix presents country-level versions of Figures 1 and 2.

Figure 1

Average share of external debt in local currency

This figure plots the cross-country average of the share of external debt in local currency by sector. Each country is equally weighted. All data are grouped by nationality.

Open in new tabDownload slide

Figure 2

Debt-to-GDP ratios by currency and sector

This figure plots the LC (top panel) and FC (bottom panel) external-debt-to-GDP ratio for various sectors in the economy. Total refers to all debt, including corporate and sovereign. Corporate refers to the sum of financial and nonfinancial corporate debt.

Open in new tabDownload slide

Figure 3 plots the share of external sovereign debt in LC and external corporate debt in local currency for 2005 (panel a) and 2017 (panel b). The difference is stark. While the share of external sovereign debt increased significantly nearly across the board during that period, the external foreign currency corporate debt composition was nearly unchanged. Remarkably, we see little correlation between the change in a government’s reliance on LC debt externally and changes in the corporate sector’s external borrowing.¹³

Figure 3

Currency composition of external borrowing

These panels plot the sovereign and corporate LC shares of external debt for 2005 (panel A) and 2017 (panel B). Section 1 describes all variable construction.

Open in new tabDownload slide

Despite this major shift of government external borrowing toward LC, emerging market sovereigns are not issuing debt in their own currency in international markets. Instead, foreign investors are buying LC sovereign debt issued under domestic law. The rise of foreign ownership of domestically issued LC sovereign debt can be seen in Figure 4, where we plot the share of all domestic sovereign debt owned by foreign investors. For the early 2000s, we see that, on average, less than 10|$\%$| of all domestically issued sovereign debt was owned by foreigners. By the 2010s, that picture had changed, and the mean foreign ownership share of domestic sovereign debt was around 25|$\%$|⁠, with a significant dispersion across countries.

Figure 4

Foreign ownership of domestic sovereign debt

This figure plots the share of domestically issued sovereign debt owned by foreign investors. Mean reports the unweighted cross country mean.

Open in new tabDownload slide

In Figure 5, we decompose external debt into each of the components of external borrowing, international debt, foreign-owned domestic debt, and cross-border loans in FC and LC, respectively. We see clearly that the increasing external reliance of governments on LC comes almost entirely from foreign purchases of domestic debt, with the share of external debt accounted for by internationally issued LC debt actually slightly declining. For the emerging market corporate sector, the only significant change in the composition of external borrowing is the shift from FC bank loans to FC international bonds. This shift does not affect the currency composition but represents a significant change in the nature of cross-border lending.

Figure 5

Composition of external borrowing

This figure plots the cross-country mean share of external debt accounting for the components of external debt. The top panel plots the composition of sovereign borrowing, and the bottom panel plots the composition of corporate borrowing.

Open in new tabDownload slide

Taken together, these stylized facts on the heterogeneous behavior of corporate and sovereign sectors provide a changed picture on the nature of international currency exposures for emerging markets.

2. Corporate Borrowing and Sovereign Risk

Having documented a striking dichotomy between sovereigns and corporates, we now turn to its implications. First, we demonstrate that FC corporate borrowing is strongly associated with elevated levels of sovereign default risk. We then provide evidence for the mechanism behind this relationship, demonstrating that a higher reliance on corporate FC debt is associated with a higher sensitivity of sovereign CDS spreads to fluctuations in exchange rate movements. We then provide high-frequency micro-evidence that increased foreign currency corporate borrowing causes increases in sovereign credit spreads, illustrating a direct pass-through of corporate currency mismatch to sovereign default risk. We conclude the section by providing micro-evidence that corporate foreign currency borrowing is evidence of balance sheet mismatch.

2.1. From corporate mismatch to sovereign default risk

2.1.1. Aggregate evidence

Table 1 presents regression results. In columns 1 and 2, we include time fixed effects without country fixed effects to examine between country variation. In columns 3 and 4, we include country fixed effects to only examine within country variation. In column 1, we see that a higher sovereign and corporate LC share are both associated with a reduction in sovereign CDS spreads, after controlling for government and corporate debt/GDP ratios. In column 2, we see that the result is robust to excluding the peak of the Global Financial Crisis. In columns 3 and 4, we see the same qualitative patterns with country fixed effects. The effect of the corporate LC debt share remains largely similar to that in column 1, albeit with slightly reduced significance. By contrast, we see reductions in the estimates on the sovereign LC share with country fixed effects. In summary, across all specifications, we find that while holding the debt levels constant, a 10|$\%$| increase in the corporate LC debt share is associated with about a 17-basis-point (bps) reduction in the sovereign CDS spread.¹⁴

In Table 2, we ask which type of FC debt helps explain sovereign CDS spreads. We include FC sovereign debt in all columns and different types of FC corporate debt across different columns. In line with previous findings, such as Edwards (1984) and Hilscher and Nosbusch (2010), the amount of FC sovereign debt is significantly positively correlated with sovereign CDS spreads in all specifications. For the FC-corporate-debt-to-GDP ratio, column 1 shows that the total FC corporate debt is associated with higher sovereign CDS spreads. Column 2 splits the FC corporate debt into the FC debt of the financial sector and the nonfinancial sector. We see that the positive effects of the FC corporate debt on sovereign CDS spreads are similar for the financial and nonfinancial sector.eak In columns 3 and 4, we further split the FC debt of the nonfinancial corporate sector into the debt of the tradable and nontradable sectors. We consider two different definitions of the nontradable sector based on Jensen and Kletzer (2010) and Sachs and Larrain (1993) in columns 3 and 4, respectively. The effect of FC debt of the nontradable sector on sovereign CDS spreads is larger that of the tradable sector. Given that firms in the nontradable sector have less FC revenue to hedge their FC borrowing, the larger coefficient for the nontradable sector is suggestive evidence that balance sheet that FC corporate debt increases FC debt due to corporate balance mismatch. That said, firms from tradable sector are not insulated from FX fluctuations if they overborrow in FC relative to their FC revenue. In addition, a large share of the pricing of tradable goods depends on the prices of nontradable service inputs (Burstein, Eichenbaum, and Rebelo, 2005).

2.1.2. Sovereign default risk and exchange rate movements

In Table 3, we consider three different versions of the LC share: sovereign, corporate, and total. The coefficient of interest is |$\beta_{3}$|⁠, and a negative |$\beta_{3}>0$| coefficient indicates that the sovereign credit risk of countries with more LC debt increase less when the LC depreciates. Unsurprisingly, given that the dollar tends to appreciate in bad times (Avdjiev et al., 2019; Kekre and Lenel, 2020), we see that |$\beta_{1}>0,$| and so emerging market sovereign CDS spreads tend to widen when the local currency depreciates against the USD. More importantly, we see that |$\beta_{3}$| is indeed negative, with us once again observing the weakest effect for the sovereign LC debt share, followed by the corporate LC debt share, and finally the total LC debt share. The magnitudes of |$\beta_{3}$| for corporate and total debt are also larger in absolute value than that of |$\beta_{1},$| indicating that countries with all of their corporate and sovereign debt in LC would be expected to witness a decline in their default risk when the LC depreciates. Columns 4–6 rerun the same specification, but with additional time fixed effects, and we find similar results.¹⁵

2.1.3. Local currency credit spreads

One way to understand the LC risk-free rate, |$\left(y_{t}^{*}+\rho_{t}\right)$|⁠, is to think of it as the hypothetical nominal interest rate that the U.S. government (assumed to be default-free) would pay if it issued a bond in an emerging market currency in the absence of market frictions. By using cross-currency swaps to convert the fixed dollar cash flows from a U.S. Treasury into fixed LC cash flows, we construct a synthetic LC instrument that is free from sovereign default risk. The LC credit spread measures how much an emerging market sovereign pays to borrow relative to this default-free benchmark in its own currency.

Figure 6 plots the mean sovereign CDS spreads and the LC credit spread between 2005 and 2021. If sovereigns are expected to simultaneously default on their LC and FC debt (as will be the case in the model)¹⁶ and financial markets are frictionless, we would expect sovereign CDS spreads and LC credit spreads to track each other closely.¹⁷ We see strong comovements between the two spreads in the early part of the sample between 2005 and 2013 and, more recently, during the COVID-19 pandemic-induced market panic of March 2020. During the period 2005–2020, the average LC credit spread was very subdued and did not track sovereign CDS spreads closely. However, as discussed in Du and Schreger (2016, 2022), besides sovereign default risk, a number of financial market frictions can affect the LC credit spreads. These frictions include capital controls, market segmentation, illiquidity in the FX swap markets for emerging markets, and balance sheet costs for financial intermediaries to arbitrage between LC and FC debt markets. In addition, the gap between the LC and FC credit spread should fluctuate over time if expected currency depreciation conditional on default varies over time (Della Corte, Jeanneret, and Patelli, 2020).¹⁸

Figure 6

Sovereign CDS and LC credit spreads

This figure plots the mean sovereign CDS and LC credit spread at the 5-year tenor. Data are plotted as 2-week rolling averages.

Open in new tabDownload slide

In Table 4, we repeat the specifications in Table 2 using the LC credit spread as the dependent variable to complement our baseline regressions using CDS. We can see that the LC credit spread is increasing with the corporate-FC-debt-to-GDP ratio, and in particular with the FC borrowing of the nonfinancial sector. In contrast to the results in Table 2, FC sovereign debt/GDP does not significantly affect the LC credit spread, and the impact of FC debt of nontradable sector is also weaker.¹⁹ The more muted effects likely reflect the impact of financial frictions, which tend to reduce the LC credit spread.

2.1.4. Sovereign CDS spreads around corporate bond issuance

In this section, we provide additional support for our argument by demonstrating that sovereign CDS spreads increase following corporate FC bond issuances. We take an event study approach and examine daily changes in the sovereign CDS around corporate bond issuance dates. We include all dollar-denominated and LC-denominated corporate bonds of issuers in our sample emerging markets with par value greater than $100 million. We calculate the changes in the sovereign CDS spread 20 business days before and after the issuance date relative to the level of the spread on the day prior to the issuance²⁰ and examine the pattern of changes in the CDS spreads around individual FC and LC corporate bond issuance.

Figure 7 reports the results. We plot the point estimate and the 95|$\%$| confidence interval for the coefficient for the date fixed effect, |$\gamma_k$| (with |$\gamma_{-1}=0$| by construction). The top panel shows changes in CDS spreads around FC corporate bond issuance. We see that the sovereign CDS spread declines prior to the FC bond issuance, which reflects the fact that issuers time debt issuance when the overall funding condition becomes more favorable. However, as soon as the FC corporate bond issuance occurs, the declining trend in the sovereign CDS spread reverses. Instead, sovereign CDS becomes more elevated and increases on average about 1 bps 1 week following issuance. The middle panel shows changes in the sovereign CDS spread around LC corporate bond issuance. While the spread also declines prior to LC bond issuance, the downward trend in the sovereign CDS spread continues after issuance.

Figure 7

Sovereign CDS spreads around corporate bond issuance

This figure shows the changes in the sovereign CDS spread around corporate bond issuance. We require the par amount of the bond to be greater than $100 million. The top panel shows changes in the sovereign CDS around FC corporate bond issuance; the middle panel show changes in the sovereign CDS spreads around LC corporate bond issuance; and the bottom panel shows the difference in the sovereign CDS spreads around FC and LC bond issuance. All changes to the sovereign CDS spreads are calculated with respect to the CDS level on the day prior to the issuance date. The dots represent point estimates, and the gray dashed lines represent 95|$\%$| confidence internal for the point estimates. The confidence intervals are constructed based on robust standard errors clustered at the issuer level.

Open in new tabDownload slide

2.2. Firm-level corporate balance sheet mismatch

The previous results demonstrated a strong relationship between corporate FC borrowing and sovereign risk. This connection between corporate borrowing and sovereign risk makes economic sense if FC borrowing is evidence of corporate vulnerability to exchange rate fluctuations (a balance sheet mismatch). We now present firm-level evidence that this is indeed case. In particular, we ask whether FC use can be explained by operating hedging or financial hedging motives, or whether it is more likely to indicate a mismatch.

We examine this question using FactSet Debt Capital Structure, a commercial data set built from firm financial reporting, such as financial statements and credit agreements. The data are intended to cover not only foreign bonds and syndicated loans but also all of a firm’s debt financing. In addition, the data are aggregated to the level of the ultimate parent, consistent with our nationality-based treatment of the aggregate data in Section 1.²¹ We restrict our attention to any bonds and loans reported on the firms’ balance sheets. The set of firms included in Factset are those that report their financial information publicly and therefore are overwhelmingly likely to be among the large, publicly traded firms. For each firm-year, we measure the share of the firm’s FC debt share in total debt. Table A.2 in the Internet Appendix reports the summary statistics by country and year.

Furthermore, Figure 8 visualizes the overlap between the distribution of FC shares between tradable and nontradable sector firms and exporters and nonexporters. We residualize the FC share by running the regression in column 2 and plot the resulting distribution. Panel A reports the residuals for firms in the tradable sector versus firms in the nontradable sector. While we do see a somewhat lower residualized FC share for firms in the nontradable sector (as indicated by the statistically significant coefficient for Tradability in column 4), the overall distribution of the FC debt share in the tradable and nontradable sectors remains very similar. Panel B reports a similar finding for firms above and below the sample median in terms of foreign sales shares.²³

Figure 8

Foreign currency borrowing and currency mismatch, 2017

These figures plot kernel densities of the FC share by sector tradability (panel A) and export intensity (panel B). The FC shares are residualized to the log of total sales and country fixed effects, as in column 2 of Table 5.

Open in new tabDownload slide

While the previous analysis demonstrates that the pattern of firm-level FC borrowing is not fully explained by operating hedge motives, Table A.5 in the Internet Appendix presents evidence that the FC liabilities are unlikely to be financially hedged. In particular, we show that the outstanding amounts of FC liabilities are generally much larger than the notional outstanding of cross-currency swaps for most sample countries, the derivatives that firms would use to financially hedge their foreign currency liability exposure.

Putting this together leads us to conclude that while firms with an operating hedging motive are more likely to borrow in FC, many firms have no operational hedging motives yet still borrow significant amounts in FC. In addition, the relative explanatory power of country fixed effects motivates our decision to focus on between country heterogeneity rather than within-country differences in our theoretical framework. Furthermore, the use of financial derivatives to hedge FC debt exposure is likely to be quite limited. This evidence of corporate balance sheet FX mismatch is consistent with recent findings based on Peru (Gutierrez, Ivashina, and Salomao, 2020), Chile (Alfaro, Calani, and Varela, 2020), Hungary (Salomao and Varela, 2021), and Turkey (Di Giovanni et al., 2017).

3. Model

Having documented this relationship between the currency composition of external debt and sovereign risk, we now interpret these patterns through the lens of the canonical sovereign debt model (Arellano, 2008; Aguiar and Gopinath, 2006). The key friction in the model is that the government cannot commit and instead decides each period whether or not to repay and how much to borrow. Relative to this literature, we introduce LC sovereign debt alongside FC sovereign debt and give the sovereign another policy tool, the inflation rate, with which to reduce real repayments on the debt. In addition, we introduce a simple firm balance sheet where the firm’s debt composition generates an endogenous cost of inflation.

3.1. Setup

The sovereign maximizes this objective function by choosing how much to borrow |$\left(B'\right)$|⁠, whether to default on the outstanding debt |$\left(D\right),$| and, now, how much of the existing debt to inflate away |$\zeta$|⁠. An exogenous fraction |$\alpha_{G}$| of the face value of the borrowing the government is in LC and the remainder |$\left(1-\alpha_{G}\right)$| will be in foreign currency.

The term |$\left(1-\alpha_{G}\zeta\right)\kappa B$| gives the real value of coupon payments this period, with |$\alpha_{G}\kappa B\zeta$| capturing the reduction in real coupon value coming from inflation. The next term, |$\alpha_{G}q^{LC}\left(a,B'\right)\left(B'-\left(1-\zeta\right)\delta B\right),$| is the net revenue raised from local currency bond issuance bond issuance. The final term is the equivalent for net revenue raised from FC bond issuance.²⁸

3.2. The optimal inflation rate

This says that the government will continue to increase inflation until it maximizes period consumption, the point at which the marginal output loss from increased inflation offsets the consumption gain from reducing the real value of debt payments. The key idea of this paper is that one of the things that will determine the optimal choice of inflation is the degree of FX mismatch on corporate balance sheets. Essentially, if firms are more mismatched (with more LC revenue relative to FC liabilities), then output will fall increasingly quickly with inflation.

We do not actually need output to fall with inflation for this mechanism to be at play. The key for our mechanism is that output may simply increase relatively less for countries with a higher degree of mismatch than those of others, holding all else equal. Appendix B.2 in the Internet Appendix demonstrates how changes in the degree of currency mismatch operate very similarly to changes in a reduced form deadweight cost of inflation.

3.2.1. Currency mismatch and inflation

We assume that entrepreneurs have access to a linear production technology that allows them to invest to produce intermediate goods |$X=\xi I$|⁠, where |$\xi$| is the productivity of the intermediate good production technology, and |$I$| denotes the units of tradable goods invested. The key financial friction is that we assume entrepreneurs cannot access external finance to invest in intermediate good provision, so we must have that investment is less than net worth |$\left(I\leq\Pi\right).$|

This captures the trade-offs the sovereign faces in choosing the optimal inflation rate. First, inflation is countercyclical, as a lower aggregate productivity makes it more tempting to inflate away the debt. Second, the larger today’s debt service, |$\kappa B,$| the higher the optimal inflation rate. Third, the term |$\delta Bq^{LC}\left(A,B'\right)$| captures the present value of outstanding long-term debt that can be inflated away. Because the expression for inflation is for a fixed amount of debt to be issued, |$B',$| net revenue raised is increasing with the amount of debt inflated away. Therefore, the higher the price a sovereign will receive for new bond issuances, the more tempting it is to inflate away the existing debt. Of course, this temptation will be captured by the bond price schedule in equilibrium.

3.3. Sovereign borrowing and default

A primary benefit of our introduction of LC debt and the entrepreneurial sector into the canonical model is that our model is a generalization of the benchmark model used for modeling FC debt. If we were to restrict inflation to always be zero, then this setup collapses exactly to a model with FC debt, particularly the version with long-term debt studied by Hatchondo and Martinez (2009) and Chatterjee and Eyigungor (2012). If we also were to restrict the sovereign to borrowing with one-period debt (⁠|$\delta=0$|⁠), then this would be equivalent to the model studied by Arellano (2008) and Aguiar and Gopinath (2006).³⁵

3.3.1. Equilibrium definition

We study the recursive Markov equilibrium for this economy, where all decision rules are functions only of the state variables |$A$| and |$B$|⁠. An equilibrium is a set of policy functions for consumption |$\tilde{c}\left(A,B\right)$|⁠, debt issuance |$\tilde{B}\left(A,B\right)$|⁠, default |$\tilde{D}\left(A,B\right)$|⁠, and inflation |$\tilde{\zeta}\left(A,B\right)$|⁠, and a price function for debt |$q\left(A,B'\right)$| such that (1) taking as given the government policy functions, household consumption satisfies the resource constraint; (2) taking the bond price function |$q\left(A,B'\right)$| as given, the government’s policy functions satisfy the sovereign’s optimization problem; and (3) the bond price function satisfies the foreign lenders’ pricing condition. The government’s lack of commitment is captured by the fact that equilibrium policy functions are restricted to be functions of today’s state variables |$A$| and |$B$| and cannot be history dependent. Instead, the government policy functions must satisfy the government’s optimization problem period by period.

3.4. Understanding the trade-off between inflation and default

The model rationalizes the empirical findings documented in the empirical section through three model assumptions. The first is that the cost of inflation varies across countries only through differences in corporate sector mismatch. Second, we assume that the depreciation rate and inflation rate coincide, meaning PPP holds. While PPP is generally a poor description of the data in low inflation times, it holds as a better approximation during high inflation periods (Froot and Rogoff, 1995), which we believe is closer to our concept of fiscal inflation. The third assumption, closely related to PPP, is that fiscal inflation to reduce the real debt burden has a direct effect on the exchange rate, but sovereign default does not. Therefore, governments with less mismatched corporate sectors will choose higher levels of fiscal inflation before turning to outright default, and in equilibrium this will lead to a lower risk of explicit sovereign default.

The assumption that default does not have a direct effect on the exchange rate is standard in the sovereign debt literature with inflation and nominal debt. For instance, much like the current paper, Aguiar et al. (2013, 2015) consider a government’s decision to inflate way or default on nominal debt. Conditional on default, the government chooses zero inflation because there is no longer a need to reduce the fiscal burden via inflation, as in the baseline case of our model. The same is true for Engel and Park (2016). While our model with full default delivers this sharp prediction, it is not important for us that inflation upon default is actually zero, but rather that optimal inflation in default states is lower than optimal inflation conditional on repayment for the same level of aggregate productivity |$A$| and debt |$B$|. This feature should hold in any model where the driver of inflation is the desire to reduce the real debt burden, as the defining feature of default is a reduction in the face value of debt.

In the data, however, there is usually significant inflation upon default. In our previous work (Du and Schreger, 2016), we found that CDS contracts quoted in multiple currencies imply a depreciation upon default of 36|$\%$|⁠. Looking at the Reinhart and Rogoff (2008) default dates, we estimate the historical average to be between 27|$\%$| and 34|$\%$|⁠. The high comovement between inflation and default that we observe in the data is generally explained by the fact that in emerging markets, both tend to occur in bad times (a low aggregate state |$A$| in the model). While the zero inflation upon default in the baseline case is therefore counterfactual, the relevant but unobserved measure for us is how much higher inflation would have been if a government had chosen to repay rather than default, for a given aggregate state |$A$| and debt level |$B$|⁠.

To generate inflation upon default, the sovereign would need to continue have an incentive to inflate after the default-induced debt reduction. One way to accomplish this in the present framework would be to allow for partial default, as then additional inflation would still reduce the fiscal burden. An extension along these lines is considered in Appendix B.1 in the Internet Appendix. More generally, any nominal liabilities that exist after default would produce this same incentive. For instance, Galli (2020) considers how the existence of money can also generate positive inflation upon default via the desire to generate seigniorage. To be clear, however, these types of incentives still mean that inflation and default are substitutes as optimal inflation policy conditional on default would be lower than optimal inflation conditional on repayment, because nominal liabilities are reduced by default. Of course, there may be other reasons for a government to inflate or devalue in default states than reducing nominal liabilities, which we do not model in our paper. For instance, Na et al. (2018) generate inflation upon default because of the assumption of downward nominal wages. Because default occurs in bad times, the government finds it optimal to devalue in the same states that it defaults, even though inflation has no direct effect on the government’s liabilities.

3.5. Quantitative results

3.5.1. Calibration and numerical solution

We calibrate the model to a quarterly frequency. Table 6 reports the parameter values. We split the parameters between those we set using external data and those we estimate via simulated method of moments. For the parameters calibrated externally, we set the intermediate good share |$\gamma$| to one-third so that the labor share is two-thirds. To calibrate the productivity process, set the autocorrelation |$\rho_{z}=0.8$| and |$\sigma_{z}=0.034$| (Aguiar and Gopinath, 2006). We follow Tauchen (1986) to discretize the productivity process. We let |$\delta=0.9595$| to set the risk-free duration of the LC bonds to 5 years when the quarterly risk-free rate is 1|$\%$|⁠.³⁷ For the probability of reentry into credit markets, we follow Cruces and Trebesch (2013) and set the probability of reentry |$\lambda$| to |$4.9\%$|⁠. The share of sticky price goods |$\mu$| is set to |$0.75,$| a common calibration parameter for Calvo pricing. We set the amount of FC external corporate financing |$Z$| to 0.68, so that in the absence of inflation the mean debt/output ratio is equal to 17|$\%$|⁠. Our benchmark calibration sets the corporate LC debt share, |$\alpha_{P}$|⁠, to |$10\%$| and the sovereign LC share, |$\alpha_{G}$|⁠, to |$60\%$| to match the evidence in Section 1. In the counterfactuals, we will vary both of these debt share parameters for our key comparative statics.

That leaves a number of parameters we need to estimate. To do so, we jointly target a number of empirical moments. In particular, we need to estimate the default cost parameters, |$d_{0}$| and |$d_{1}$|⁠, investor risk aversion, |$\psi,$| the government discount factor, |$\beta$|⁠, and entrepreneurial relative productivity, |$\xi.$| To do so, we use the method of simulated moments, targeting the default rate, the external debt/GDP ratio, the share of the risk premium in the LC spread, the average inflation difference, and the standard deviation of LC spreads.³⁸ While we estimate all parameters jointly, default costs are roughly targeted with the default rate and standard deviation of the spread, investor risk aversion is targeted with the difference between historical spreads and default rates, the sovereign discount factor for the level of debt, and entrepreneurial productivity with mean inflation. We estimate moderately convex default costs, risk-averse investor |$\left(\psi=0.889\right),$| and impatient sovereign with a quarterly |$\beta=0.91,$| and |$\xi=1.2195$|⁠.

To solve the model, we employ value function iteration over a discretized state space. Because our recursive representation is identical to the model studied in Hatchondo and Martinez (2009) and Chatterjee and Eyigungor (2012), with one additional constraint on the policy maker (Equation (28)), we can simply follow the solution methods used in the FC sovereign debt literature. The state space for productivity shocks is discretized to a 31-state grid. The state space for bonds is discretized into 301 grid points. A finer grid is used for the endogenous state variable to keep the discretization from affecting the sovereign’s choices. Following the recommendations in Hatchondo, Martinez, and Sapriza (2010), we iterate backward from the solution of the final period of the finite-horizon model so that we select the equilibrium bond price of the finite-horizon model.³⁹ With our policy functions and bond price schedules in hand, we can calculate the model-implied moments by simulating the model 20 times for 3,000 quarters per simulation. We discard the first 500 periods of each simulation.

3.5.2. Quantitative results and key mechanisms

Table 7 reports the empirical and model-generated moments for five different calibrations of the corporate (⁠|$\alpha_{P})$| and sovereign |$\left(\alpha_{G}\right)$| local currency debt share. In the first column, we report the empirical moments. In our baseline calibration, we come quite close to matching the average cross-country empirical moments, with an average FC and LC spread of 1.06|$\%$| and 2.95|$\%$|⁠, compared to 1.59|$\%$| and 4.32|$\%$| in the data. Countries have a default rate of 0.9|$\%$| in the model as compared to 1.5|$\%$| in the data.

In the bottom-five rows of the table, we compare the business cycle moments between the data and the various model calibrations. We find that the baseline calibration come fairly close to matching the excess volatility of consumption to output |$\left(\sigma\left(y\right)/\sigma\left(c\right)\right)$| and the correlation between output and consumption (0.92 in model vs. 0.72 in data). In addition, the model generates a countercyclical trade balance. More closely related to the innovation of the model, out baseline model generates strongly countercyclical inflation, with a correlation between inflation and output of |$-$|0.65 as compared to |$-$|0.28 in the data.⁴⁰ In addition, the mean inflation differential of 1.73 is quite close to the data, where emerging markets have had an average inflation of around 1.5|$\%$|–2|$\%$| higher than the United States since 2005.

Next, we perform counterfactual analyses by varying the sovereign and corporate local currency debt shares. In column 6, we see that by moving from our baseline of 60|$\%$| local currency sovereign debt and 10|$\%$| local currency corporate debt to 80|$\%$| sovereign debt and 30|$\%$| corporate debt, the probability of default is cut by two-thirds. In columns 4 and 5, we see that when we vary only the corporate and sovereign LC shares individually, and we see smaller declines in default risk. For completeness, in column 3, we report the model-simulated moments for a model with only foreign currency debt.

To better understand how shifting the currency composition of debt generates this different default behavior, we will begin by looking at the sovereign’s policy for different degrees of corporate and sovereign mismatch. In Figure 9, we plot the sovereign’s inflation policy functions and default threshold for different currency compositions of external debt for a single productivity level. As the country shifts toward more local currency corporate and sovereign debt, the government finds it optimal to begin to engage in fiscal inflation at lower levels of debt. In addition, the government chooses higher levels of inflation to reduce the value of the debt before turning to explicit sovereign default.

Figure 9

Inflation policy and default

This figure plots the inflation and default policy functions for different sovereign (⁠|$\alpha_G$|⁠) and corporate (⁠|$\alpha_P$|⁠) LC debt shares for a single realization of the aggregate TFP state. The x-axis plots the amount of sovereign debt entering the period (B). The y-axis plots the amount of inflation chosen by the government for this combination of productivity and debt. The circle represents the maximum debt level for which the sovereign repays, and the government therefore defaults for any higher debt levels.

Open in new tabDownload slide

Because lenders recognize the incentives facing the sovereign, these policy functions are mirrored in the bond price schedules to ensure that foreign lenders break even in expectation. Figure 10 plots the local currency and foreign currency bond price schedule the sovereign faces for five different combinations of corporate and sovereign local currency exposure. Panel A plots the local currency bond price schedule, and panel B plots the price schedule for foreign currency bonds. Turning to local currency bond prices first, we see a straightforward relationship: as the government moves from more foreign currency corporate and sovereign debt to more local currency debt, we see lower local currency bond prices for any level of borrowing. As seen in Figure 9, this is expected as sovereigns with more local currency debt are relatively more inclined to reduce the value of the debt using inflation. Turning to panel A, however, we see a starkly different pattern. Here, we see for relatively low levels of debt, governments with relatively more local currency debt (solid red lines) actually receive the highest foreign currency bond prices. The reason for this is that the government’s ability to partially reduce the value of their debt stock via inflation, and the incentive of the government when borrowing to avoid the inflation region to maintain a higher bond price, keeps the government away from the default threshold in equilibrium. Because the only risk faced by investors is foreign currency debt is that of outright default, this figure speaks to the main finding of the paper, that a larger reliance on local currency debt can reduce default. Importantly, however, as seen in the dashed pink line, increasing the sovereign local currency share is not sufficient to do this. Instead, it is the joint movement of sovereign and corporate debt toward local currency that reduces equilibrium default risk.

Figure 10

Bond prices

This figure plots the bond price functions for different sovereign (⁠|$\alpha_G$|⁠) and corporate (⁠|$\alpha_P$|⁠) LC debt shares for a single realization of the aggregate TFP state. The upper panel is for LC bond prices, and the lower panel is for FC bond prices. The x-axis plots the amount of sovereign debt issued in the period (B’). The y-axis plots the bond price.

Open in new tabDownload slide

3.5.3. Model and data

In the second column, we additionally control for the debt/GDP ratio. The third column reports the model counterparts. We see that our model-generated estimates for the effect of an increase in the LC debt share on market-implied default risk is roughly in line with data. For corporate debt, while the model generates a significant effect of moving toward local currency, it is less so than in the data.⁴¹

4. Conclusion

This paper examines why FC corporate borrowing increases sovereign default risk, even when the sovereign’s external borrowing has been largely denominated in LC. We argue that a government is more inclined to default than inflate when the currency mismatch of the corporate sector implies large adverse balance sheet effects from a currency depreciation. In making this argument, we construct a new data set on the currency composition of emerging market external borrowing by sector and show that the corporate sector remains reliant on external FC debt even as sovereigns have swiftly moved toward borrowing in their own currency. We show that a higher level of external FC corporate debt is associated with more sovereign credit risk.

Motivated by these empirical findings that a higher level of external FC corporate debt is associated with more sovereign credit risk, we present a model in which currency-mismatched corporate balance sheets increase the cost of inflating away sovereign debt and make default relatively more appealing. We embed a corporate balance sheet channel in the canonical Eaton and Gersovitz (1981) sovereign debt model and demonstrate how higher shares of LC private debt can reduce the default risk on sovereign debt in equilibrium by affecting the cost of inflation relative to default.

More generally, this paper underscores the importance of integrating private sector vulnerabilities into analyses of government borrowing and default decisions. This paper demonstrates how aggregate corporate mismatch can tilt the balance between outright default and fiscal inflation. But, of course, corporate exposures may affect the costs of default through many channels and thereby affect the government’s decisions.

Authors have furnished an Internet Appendix, which is available on the Oxford University Press Web site next to the link to the final published paper online.

Acknowledgement

We thank David Baqaee, Luigi Bocola, Carol Bertaut, Laura Blattner, John Campbell, Max Eber, Charles Engel, Emmanuel Farhi, Jeff Frankel, Jeff Frieden, Marcio Garcia, Gita Gopinath, Pierre-Olivier Gourinchas, Galina Hale, Herman Kamil, Ricardo Hausmann, Sebnem Kalemli-Ozcan, Illenin Kondo, Hanno Lustig, Matteo Maggiori, Andrea Raffo, Romain Raniciere, Carmen Reinhart, John Rogers, Ken Rogoff, Jeremy Stein, Alexandra Tabova, Adrien Verdelhan, Steve Pak Wu, and Vivian Yue and various seminar and conference participants for helpful comments. We thank Lina Beatriz Gomez Castillo, Joao Henrique Freitas, Emmanuel Kohlscheen, and Renzo Jimenez Sotelo for assistance in acquiring and interpreting various data sources. Angus Lewis, Conor Howells, and Christine Rivera provided excellent research assistance. The views in this paper are solely the responsibility of the authors and should not be interpreted as reflecting the views of the Federal Reserve Bank of New York or any other person associated with the Federal Reserve System. Supplementary data can be found on The Review of Financial Studies web site.

Footnotes

References