A Journey in the History of Sovereign Defaults on Domestic-Law Public Debt

Board of Governors of the Federal Reserve System International Finance Discussion Papers ISSN 1073-2500 (Print) ISSN 2767-4509 (Online) Number 1338 March 2022 A Journey in the History of Sovereign Defaults on Domestic-Law Public Debt Aitor Erce, Enrico Mallucci and Mattia Picarelli Please cite this paper as: Erce, Aitor, Enrico Mallucci and Mattia Picarelli (2022). “A Journey in the History of Sovereign Defaults on Domestic-Law Public Debt,” International Finance Discussion Papers 1338. Washington: Board of Governors of the Federal Reserve System, https://doi.org/10.17016/IFDP.2022.1338. NOTE: International Finance Discussion Papers (IFDPs) are preliminary materials circulated to stimulate discussion and critical comment. The analysis and conclusions set forth are those of the authors and do not indicate concurrence by other members of the research staff or the Board of Governors. References in publications to the International Finance Discussion Papers Series (other than acknowledgement) should be cleared with the author(s) to protect the tentative character of these papers. Recent IFDPs are available on the Web at www.federalreserve.gov/pubs/ifdp/. This paper can be downloaded without charge from the Social Science Research Network electronic library at www.ssrn.com.

A Journey in the History of Sovereign Defaults on Domestic-Law Public Debt∗† Aitor Erce‡, Enrico Mallucci§, & Mattia Picarelli¶ Abstract We introduce a novel database on sovereign defaults that involve public debt instruments governed by domestic law. By systematically reviewing a large number of sources, we identify 134 default and restructuring events of domestic debt instruments, in 52 countries from 1980 to 2018. Domestic-law defaults are a global phenomenon. Over time, they have become larger and more frequent than foreign-law defaults. Domestic-law debt restructurings proceed faster than foreign ones, often through extensions of maturities and amendments to the coupon structure. While face valuereductionsarerare,net-present-valuelossesforcreditorsarestilllarge. Unilateral amendments and post-default restructuring are the norm, but negotiated pre-default restructurings are becoming increasingly frequent. We also document that domesticlaw defaults typically involve debt denominated in local currency and held by resident investors. We complement our analysis with a collection “sovereign histories”, which provide the fine details about each episode. JEL classification: E62, E65, F34, G01, H12, H63, K00, K41 Keywords: Public debt, sovereign default, domestic law, database. ∗WethankLauraFernandez,BrynneGodfrey,MonicaGomez,SilviaGutierrez,FloraHaberkorn,Kaede Johnson, Isaac Norwich, Kelsey Shipman and Chris Webster for research assistance, Debbie Y Briggs, Jose Cunha, Javi Diaz-Cassou, Stefano Finesi, Gabriel Gomez-Giglio, Andreja Lenarcic, Lorenzo Ricci and Lluis Sanchis for help with specific countries, and Giovanni Callegari, Ricardo Correa, Tito Cordella, Jenny Sellin and seminar participants at Bank of Spain, European Central Bank, European Stability Mechanism, and the Fourth Conference on Law and Macroeconomics for comments. These are our views and not those of the Board of Governors of the Federal Reserve or the European Stability Mechanism. †The database accompanying this paper can be accessed here and the collection of “sovereign histories” providing the fine details about each default episode here. ‡Navarra Public University & LUISS-SEP, Pamplona, Spain. E-mail: erceaitor@gmail.com §BoardofGovernorsoftheFederalReserveSystem,Washington,U.S.A.E-mail: enrico.mallucci@frb.gov ¶European Stability Mechanism, Luxembourg, Luxembourg. E-mail: m.osvaldopicarelli@esm.europa.eu 1

1 Introduction Sovereign debt markets in emerging economies have experienced radical transformations in recent decades. As many sovereigns began to tap international capital markets, bonds replacedbankloans, andincreasinglyperfectedclauseswereaddedtobondstofacilitatedebt restructuring (Buchheit et al., 2019; IMF, 2020). Another critical, yet less discussed, change istheincreasedrelevanceofdomesticdebtmarkets(GelpernandPanizza,2021;Reinhartand Rogoff, 2008). Traditionally, domestic debt markets for emerging sovereigns were either nonexisting or closed to foreigners (CGFS, 2007). Emerging sovereigns could only borrow from foreign investors in foreign currencies and international markets (Eichengreen and Panizza, 2005). Since the 90s, as a result of financial deepening and economic growth, governments are increasingly relying on domestic borrowings to fund their financing needs (Gelpern and Setser, 2004; Burger and Warnock, 2006; IMF, 2020).1 Despite this growing importance of domestic debt markets, the very definition of what constitutes a sovereign default in domestic debt remains elusive (Gelpern, 2008; Gelpern and Panizza, 2021). Some authors have focused on sovereign debt denominated in local currency (Kohlscheen,2009;JeanneretandSouissi,2016;BeersanddeLeon-Manlagnit,2019). Others (SturzeneggerandZettelmeyer,2008;ReinhartandRogoff,2011;Duggar,2013)havefocused on whether the creditors are domestic residents. In contrast, there is no systematic work focused on sovereign defaults under domestic law. This is, according to leading legal experts (Gelpern and Panizza, 2021; Gulati and Weidemaier, 2015) and key policy institutions (IMF, 2021), a major gap in the economic literature.2 Inthispaper, wepresentoureffortstofillthisgap. Weintroduceanoveldatabaseidentifying sovereign defaults on debt instruments governed by domestic law. Our database focuses on explicit defaults towards private creditors and is the first effort to systematically record domestic sovereign defaults that are identified on the basis of the legal jurisdiction. We use it to present key stylized facts that may prove informative for theoretical work.3 We also shed light on a range of related issues, including the incidence of selective sovereign defaults (Broner et al., 2010, 2014; Mendoza and D’Erasmo, 2016; Gelpern and Panizza, 2021) and the extent to which different definitions of what constitutes domestic debt coincide (Burger and Warnock, 2006; Gelpern, 2008; Bradley et al., 2016).4 1Brazil and Mexico exemplify this trend. Bank of International Settlements data reveals that domestic debt accounted for 22% of Mexico’s public debt in 1995. By 2010 that share was over 80%. In Brazil, the share increased from 54% in 1995 to over 90% in 2010. 2IMF(2021)notesthatthelackofbetterdataaboutsovereigndefaultsondomestic-lawdebtinstruments is preventing better grounded policy work on the resolution of sovereign debt crises. This is worrying considering that many sovereigns currently at risk of distress borrow domestically (IMF, 2021). 3Recenttheoreticalquantitativecontributionsstudydomesticsovereigndefaults. Bocola(2016)andSosa- Padilla(2018)studythenexusbetweendomesticdefaultsandbankcrises. Aguiaretal.(2015),Mendozaand D’Erasmo (2016) and Arellano and Kocherlakota (2014) analyze their distributional implications. Erce and Mallucci (2018) study borrowing and default decisions when sovereigns can discriminate between domestic and foreign debt. Thaler (2021) models the effect default on access to domestic sovereign debt markets. 4OurworkcomplementsAsonumaandTrebesch(2016),whocompileadatabasewithinformationon184 2

Our definition of domestic public debt, grounded on whether government debt is governed by the domestic law, highlights a dimension that crucially shapes the restructuring process: debt jurisdiction (Gelpern and Panizza, 2021; IMF, 2021). While the residence of investors and the currency denomination have implications for the macroeconomic consequences of sovereigndefault,thejurisdictiondirectlyaffectsgovernments’abilitytorestructuredebt. As described in Chamon et al. (2018) or IMF (2020, 2021), the terms of government debt issued in the domestic jurisdiction can be more easily restructured using legislative or executive measures, with repercussions for market access. Moreover, domestic sovereign debt markets are the backbone of domestic financial systems. According to CGFS (2007), domestic bond markets promote financial stability not only by reducing currency mismatches but also by creating a benchmark (market-determined) yield curve that reflects the costs of borrowing domestically at different maturities. In economies lacking well-functioning domestic debt markets, banks may find it hard to price and provide long-term lending. As a result, defaulting upon domestic-law debt may affect the financial standing of the private sector over and beyond what a default of debt governed under foreign laws may do (Gelpern and Panizza, 2021; IMF, 2021). In fact, as the consequences of sovereign default are increasingly borne domestically, government incentives to default have likely changed. While governments defaulting externally are concerned about being excluded from international capital markets, those defaulting domestically are more concerned with their financial stability (Sosa-Padilla, 2018; Thaler, 2021) and distributional implications (Mendoza and D’Erasmo, 2016).5 To build our database we reviewed a large number of sources, ranging from IMF official documents to local news articles, and identified 134 events of restructuring of domestic-law debt instruments in 52 countries from 1980 to 2018. We organized the information in two distinct products: a database on domestic defaults and a collection of “sovereign histories” (Erce et al., 2021). The database collects variables that measure the timing and size of domestic defaults, as well as the terms and restructuring methods used. Where information was available, the database was constructed using a bottom-up approach, collecting information at the debt instrument level and aggregating it to obtain episode-level variables. Sovereign histories keep track of the complexity and heterogeneity of domestic defaults, and report finer details of each episode. In particular, each sovereign history provides an overview of the events leading to the default and the available details of the restructuring process for each debt instrument involved.6 We draw various lessons from our database and our sovereign histories. First, domesticlaw defaults are a global phenomenon occurring in every continent. While they are more frequent in poor and middle-income countries, they also happen in advanced economies. Second, domestic-law defaults are increasingly frequent. Combining our data with the database restructuring events of sovereign debt instruments governed or held externally. 5According to IMF (2020) “While debt governed by domestic law is often easier to restructure than internationaldebtfromalegalperspective,restructuringsuchdebtcannegativelyimpactfinancialstability”. 6Domestic-law defaults receive little international coverage, and documenting them is a challenging exercise. This marks the value of our “sovereign histories”. 3

on foreign defaults of Asonuma and Trebesch (2016), we find that in the 1980s around 10 percent of default episodes involved domestic-law debt. In the 2000s, around 70 percent of defaults episodes did. Between the late 1990s and 2016, domestic defaults have actually become more frequent than foreign ones. More often than not, governments operate selective defaults. Third, defaults on bonded debt are the most common form of domestic-law defaults. Defaults on bank loans and deposits are fairly rare nowadays. Fourth, the median size of domestic-law defaults has increased over time, reaching around 11% of GDP in the last decade. The size of domestic defaults varies greatly depending on the instruments involved in each episode. Defaults that involve deposits, bonds with CACs or multiple debt instruments are typically larger than those involving bank loans or bonds without CACs. Fifth, an extension of maturities is by far the most frequent form of restructuring, featuring in almost 80% of the episodes. Amendments to the coupon structure are also fairly frequent, while face value reductions are rare. Sixth, domestic debt restructuring often proceeds much faster than external one, but it can also protract significantly. Seventh, net-present-value losses for creditors during domestic-law sovereign debt restructuring are sizable. Eighth, negotiated preemptive restructurings are gaining traction. Ninth, despite financial globalization the triple coincidence is very much alive. Domestic-law defaults typically involve debt denominated in domestic currency and held by domestic investors. We draw an additional lesson: domestic-law defaults are complex and heterogeneous. Summarizing them in a few indicators may conceal more than reveal. This is why we complement this paper with a collection of “sovereign histories” (Erce et al., 2021). Reading them, one quickly learns that shocks triggering domestic defaults are disparate ranging from regional economic crisis to fluctuations in commodity prices to wars, pandemics, extreme natural events, and political animosity. At the same time, governments’ approach to domestic default varies greatly from episode to episode. Some countries, like Jamaica in 2013, seek an early involvement of creditors and restructure government debt in a pre-default market friendly fashion. Other countries, like Peru, are still trying to reach an agreement with creditors on bonds that have been in default since the early 1990s. The rest of this paper is organized as follows. Section 2 describes the construction of our database and the data sources we consulted. Section 3 introduces the database and the sovereign histories. Section 4 compares our database to existing databases covering other aspects of domestic default. Section 5 discusses domestic-law defaults overtime. Section 6 analyzes their geographic distribution. Section 7 reports descriptive statistics about the size of domestic defaults. Section 8 discusses the duration of domestic-law defaults. Section 9 describes net-present values losses. Section 10 examines how governments choose to default and restructure debt. Section 11 documents the relevance of domestic creditors and local currency debt in domestic-law defaults. Section 12 concludes. 4

2 Data and Sources Sovereign debt is a contractual obligation referring to debt issued or guaranteed by the government of a sovereign state. Defaults can take place through a plethora of mechanisms, ranging from unilateral reduction of principal or coupons, forcible currency conversions, forcible conversions in other debt instruments, suspensions of payments, or freezes.7 Our database focuses on explicit domestic defaults towards private creditors on the basis of the legal jurisdiction. Consistent with the definitions adopted by rating agencies and the empirical literature on sovereign defaults (Reinhart and Rogoff (2008), Beers and de Leon- Manlagnit (2019)), we consider a domestic default event to have happened when one or more of the following events occur: • A government fails to meet a principal or interest payment on a debt instrument on the due date or within a specified grace period (as Brazil in 1990 or Argentina in 2001). • Debt instruments are written off the books without a proper compensation for debtholders (as Liberia in 1989). • Contractual terms of debt instruments are unilaterally amended by a government lawdecree, such as the abrogation of indexation clauses (as Brazil in 1986) or the introduction of retrospective taxes targeting debt service payments (as Turkey in 1999). • Absent an outright payment default, the government undertakes a restructuring exercise, that reduces interest rates and/or extends maturities of outstanding securities (as Greece in 2011 or Barbados in 2018). • Deposits that are either guaranteed by the government or held by public banks are frozen and/or forcibly converted from foreign to local currency or into government bonds (as Pakistan in 1999). Documenting domestic-law defaults, their timing, their size, and the details of the restructuring terms is a challenging exercise. Given the limited attention paid by the international community to these events, information about domestic-law defaults needs to be collected across a large and dispersed number of sources.8 As a result, the compilation of our database and the drafting of our sovereign “histories” was an intensive effort that required the consultation of several and diverse sources. In order to minimize errors, where feasible, we compared available details across different sources. We consulted the databases in Reinhart and Rogoff (2008) and Beers and de Leon-Manlagnit (2019) that contain information on domestic sovereign default episodes, and Asonuma and 7Other authors have considered less explicit definitions of domestic default, such as episodes of high inflation and financial repression. 8This is especially true for episodes involving smaller countries and not involving bonded debt. 5

Trebesch (2016) that contain information on external sovereign default episodes, as well as reportsfromthemajorratingagencies(S&P,Moody’sandFitch)listingepisodesofsovereign payment difficulties, both domestic and external. We complemented these resources with information collected from an extensive google search and a press review through Factiva, a news search tool that enables the consultation of more than32,000sourcesthatincludelocalandinternationalnewspapers,journals,andmagazines. Factiva proved very useful especially to identify domestic defaults in African countries that are widely publicized in the local press but receive little to no-coverage in the international press. The news search on Factiva was carried out using keywords, such as “country name” + “domestic debt” + “default” in English, French and Spanish, and analyzing the results obtained. We also conducted additional searches using the following terms: “restructuring”, “rescheduling”, “reprofiling”, “missed payments”, and “unpaid” instead of “default”; and searches using “internal” and “local” instead of “domestic’. To process the information we proceeded as follows. For each identified domestic default event, we systematically scrutinized official documents from the IMF. These often provide detailed information on the restructuring process, including volumes, terms and the type of assets involved, as well as on the economic background in which defaults and restructuring episodes unfolded. Among them, program documents and Article IVs, proved especially useful. We also searched into the websites of the countries’ Ministries of Finance, debt management agencies and Central Banks to cross-check the information and to add any additional details we could found.9 Whenever possible, we consulted the parliamentary resolutions, bills, and decrees that amended the terms of the debt instruments.10 We also browsed other resources to gain additional knowledge about each default episode. Additional sources included country reports from the World Bank and the OECD, rating agencies publications, policy reports from development banks, local and international media, debt exchange offers, academic books, and research papers.11 The information was organized and systematized in two distinct products. First, a database collectingkeyinformationaboutdefaultsandrestructurings, whichwedescribeinthispaper. Second, a collection of documents, named “sovereign histories”, that provides a detailed summary of each sovereign debt restructuring of liabilities governed under domestic law. 9Whenever possible, we contacted local authorities to confirm the accuracy of the information. 10Many of these sources are not in English. Hence, in the collection of the histories we may have run in a languagebias. WearemorefamiliarwiththeEnglish,Spanish,Italian,andFrenchterminologyforsovereign defaults than with the terminology in other languages. 11InErceetal.(2021)weprovideaepisodebyepisodelistingofallthereferenceswherewefoundrelevant information. 6

3 The Database and the Sovereign Histories The database was constructed using a bottom up approach. We first collected information at liability level in order to build a database of default events. Then we aggregated the default events to obtain a database of default episodes, where episode-level variables measure different aspects of default. We bundled different default events into a single episode when one of these two conditions were met: (i) two or more default or debt restructuring events occurred in the same or in the following year (as in Nicaragua 1994); (ii) the government announced a comprehensive restructuring of public debt (as in Grenada 2013). Our efforts led to the identification of 134 default events on different instruments (bonds, bank loans or deposits), which we further group into 76 sovereign default and restructuring episodes. The sample spans from 1980 to 2018 and covers events in 52 countries in all five continents. Ourexercisewentbeyondthemereidentificationofeachevent. Wealsocollected finer details on the default, such as the type of instruments involved, the volumes involved, or the restructuring strategy pursued. Variables in the database measure different aspects of sovereign defaults. A first set of variables measures the timing of defaults. In particular, we collect information on the start and end dates of each default event at the monthly level. The start date is either the date in which debt instruments entered in default or were frozen, or the date of the announcement of the debt restructuring exercise. The end date is either the date in which debt payments resumed and deposits were unfrozen, or to the data in which restructuring plans were agreed and executed. Often, a comprehensive restructuring marks the end date of default episodes. Another set of variables measures the size of defaults. In particular, we collect information on the dollar value of the instruments involved and we compute aggregate default volumes for three categories of instruments: bonds, bank loans, and deposits. To ease the comparison with existing databases on domestic debt, we also collected information on the domicile (domestic or foreign) of the investors and the currency denomination (local or foreign) of the instruments involved. The database also includes variables that monitor how governments restructure each instrument. These variables keep track of the way in which governments amend the terms of its debt and of the net-present-value (NPV) losses for creditors. In particular, we collected information on whether the terms were amended through a face value reduction, a change in the maturity structure, or an amendment of the coupon structure. We also classified default events in either pre-default or post-default restructurings depending on whether the government was in default or not at the time the restructuring plan was announced. Finally, we also classified domestic defaults in unilateral or negotiated, depending on whether the government involved creditors in the design of the restructuring plan. Our “Sovereign histories” provide the full narrative for each of the 76 default episodes. In particular, eachhistoryisstructuredintwosections. Thefirstsectionprovidesanoverviewof the events leading to either the default or the restructuring. The second section provides the full details of the restructuring process for each instrument involved. Take for instance the case of Grenada’s default in 2013. The restructuring exercise involved T-Bills, government 7

bonds, arrears to domestic suppliers, loans, and guaranteed loans. Each asset received a different treatment and investors were also discriminated on the basis of their residence and identity. In the sovereign history of Grenada, we detail the terms and the timing of the restructuring process for each asset and each creditor. The detail of the information provided in each history makes the collection of our sovereign histories a unique source of examples for the ongoing policy and academic debates. Take, for instance, the debates surrounding the impact on the incidence and duration of default episodes of financial and legal innovations in sovereign financing strategies (IMF, 2020). Our histories speak to these debates as they document the seldom use of collective action clauses under domestically issued debt (as in Barbados) and provide the finer details of cases where sovereign debt was collateralized (as in St. Kitts & Nevis), or where the exit from the default required financial engineering operations, such as transforming deposits into bonds (as in Serbia) or designing payment structures that accommodate for natural disasters (as in Grenada). While digging for domestic-law defaults, we found 30 episodes of accumulation of domestic arrears, definedasoverduepaymentsbygovernments tosuppliers, civilservantsandpensioners. Given that basic information regarding these episodes, such as starting date, volumes, and clearance strategy were often impossible to accurately reconstruct (arrears are usually reported as flows and in a discontinued manner), we describe them in our stories and report the information we collected in Appendix A, but don’t include them in our database.12 4 Comparison with existing databases Our definition of domestic debt is based on the law governing the debt instruments. We consider domestic debt any government security that is issued under domestic law, regardless of the residence of the investors and the currency denomination. Other works have used different definitions of domestic debt. In particular, Reinhart and Rogoff (2008) classify debt according to the residence of investors, while Beers and de Leon-Manlagnit (2019) classify debt according to its currency denomination. Table 1 compares our database against the database of Beers and de Leon-Manlagnit (2019), that is based on the currency denomination of the instruments in default, and the database of Reinhart and Rogoff (2008), that is based on the residence of the creditors. Each cell of the table reports the number of defaults jointly reported in the databases specified by the corresponding row and column. For instance, the cell corresponding to the first row and the second column reports the number of defaults in our database that are also reported in Beers and de Leon-Manlagnit (2019). The diagonal reports the total number of default episodes reported by each database. The last column reports the number of defaults that 12Often governments accumulate arrears over multiple years, and these tend to be disputed, forcing reconciliation processes (through domestic courts or ad-hoc tribunals) to validate the claims before resolving them. Arrears are usually settled either by cash payments or through an exchange with newly issued debt to creditors. Beers et al. (2021), studies domestic arrears using data obtained from IMF documents. 8

are common to all three databases. To identify overlapping episodes that might have been reported with slightly different starting dates, we consider a time window of (−2,+2) years around our domestic debt default episodes. Table 1. Domestic Defaults: Comparing Databases Database Our Data B&LM (2019) R&R (2008) All Criterion Law Currency Residence All Our Data 75 20 17 10 B&LM (2019) 20 38 15 10 R&R (2008) 17 15 26 10 Number of defaults reported by the database specified by the row and by the database specified by the column. To identify overlapping episodes, we consider a time window of (−2,+2) years. Data for local currency defaults are from Beers and de Leon-Manlagnit (2019). Data for defaults on domestic residents are from Reinhart and Rogoff (2008). Wefindthatourdatabasecontainsabouthalfofthe38defaultepisodesreportedinBeersand de Leon-Manlagnit (2019) and 17 of the 26 default episodes reported in Reinhart and Rogoff (2008), suggesting that there is substantial overlap between the legal jurisdiction, domicile, and currency. There are two key reasons why our database does not capture some of the default episodes reported in the other databases. First, the time span is different. Second, our database does not include de facto defaults, such as those associated with hyperinflation or changes in the legal tender.13 That said, our database extends significantly the coverage of domestic defaults. It reports 30 default episodes that are not included in Beers and de Leon- Manlagnit (2019) database, and 41 default episodes that are not included in Reinhart and Rogoff (2008). Table 16 in the Appendix B reports the full list of domestic default episodes included in our database and compares them with those in Beers and de Leon-Manlagnit (2019) and in Reinhart and Rogoff (2008). 5 The Chronology of Domestic-Law Defaults Domestic-law defaults have happened in each of the four decades spanned by our database. The black line in Figure 1 reports the total number of domestic default events occurring in overlapping four-year windows from 1980 to 2018. Domestic defaults were relatively rare in the first half of the 1980s and became increasingly frequent in the 1990s and in the 2000s, peaking in 2004. The incidence of domestic defaults declined markedly in the second half of the 2000s before inching up again in the first half of the 2010s, reflecting high debt levels and 13Appendix C explains for each default episode the reason why they are included in either Beers and de Leon-Manlagnit (2019) or Reinhart and Rogoff (2008) and not in our database. 9

subdued growth in the aftermath of the Global Financial Crisis. The incidence of domestic defaults declined sharply in the second half of the 2010s reaching a through in 2018, just before the onset of the COVID-19 crisis. Figure 1. Distribution of Domestic Default Events Thesolidblack(dashedred)lineplotsthefour-yearrollingsumofdomestic(external)default events. ExternaldefaultsareasreportedbyAsonumaandTrebesch(2016). Thedottedblue line is the cyclical component of low and middle-income countries aggregate GDP obtained applying the Hodrick-Prescott filter to the annual series. The red-dashed line in Figure 1 plots the total number of external default events, as reported by Asonuma and Trebesch (2016), using overlapping four-year windows from 1980 to 2018. Foreign-law debt restructuring peaked in the mid 1980s and declined sharply thereafter.This contrasts with domestic-law defaults, which peaked in the late 1990s. These patters underscore how the expansion of domestic debt markets in the 1990s has translated into a greater involvement of domestic debt in defaults.14 We also explore the relation between the timing of domestic defaults and economic cycles. The dotted blue line in Figure 1 plots the cyclical component of GDP in low and middleincomeeconomies, whichmakeupthebulkofoursample.15 Defaulteventsaremorefrequent 14The increase in the frequency of defaults on domestic-law debt relative to defaults on external law ones since the nineties suggests that the partial substitution of external debt for domestic one has increased governments’ inclination to operate selective defaults, as suggested by Erce and Mallucci (2018). 15Of the 134 domestic default events, 112 occurred in low or middle-income economies. 10

when economic growth falls below the trend. Domestic defaults are more frequent in bad times (Tomz and Wright, 2007). Table 2. Default Events by Instrument and Decade Full Sample 1980-1989 1990-1999 2000-2009 2010-2018 Default Events Bonds 84 6 24 32 22 Bank loans 32 2 11 9 10 Deposits 18 4 11 3 0 Total 134 12 46 44 32 Default Episodes Diverse instruments 21 2 9 7 3 Total 76 9 35 17 15 Number of default events involving bonds, bank loans or deposits. Default episodes that involve diverse (multiple) instruments are double counted. We categorize government instruments in three categories: bonds, bank loans, and deposits. Table 2 reports the evolution over time of the incidence of defaults by instrument. As shown in the first column, defaults on bonded debt are by far the most common form of domestic default: of the 134 events, 84 involve bonds, 32 involve bank loans, and 18 involve deposits. Looking at the evolution over time of the composition of defaulted debt, a clear pattern emerges. Defaults on bonded debt have become an increasingly large fraction of total domestic defaults. In the 1980s, 50% of the restructuring episodes involved bonded debt. In the last two decades, the share of domestic defaults involving bonded debt has risen to around 70%. Over the same period of time, the percentage of defaults involving deposits has dropped, declining from 33% in the 1980s to 0% in the 2010s. Finally, the incidence of defaults involving bank loans has remained fairly constant over time.16 To investigate the frequency of domestic defaults that involve multiple (diverse) instruments, we focus on default episodes. We find that almost 30% of the 76 default episodes in our sample involve diverse instruments. 6 The Geography of Domestic Sovereign Defaults Domestic-law sovereign default is a global phenomenon. Table 3 breaks down the number of defaults by continent. Looking at the number of events, America is the continent where 16These findings suggest that the shift in governments’ foreign borrowing from bank loans in the 1980s to tradable bonds in the 1990s has also involved domestic debt and is reflected in the greater involvement of bonded debt in default episodes. 11

most restructurings have occurred. All of them in Latin America: South American countries recorded 39 events; Central American countries recorded 10; and Caribbean countries recorded 27 default events. Africa, with 31 events, is next runner up. Europe also recorded a high number of default events: 16. The vast majority of the defaults occurred in emerging European countries, such as Macedonia and Ukraine. Yet, defaults have also involved two euro area countries: Greece and Cyprus. Domestic defaults are less frequent in Asia, where we only discovered 10 events. Bundling the events into episodes, as described in paragraph 3, weobserveasimilarrankingdespitevariousrestructuringsinAmericaoccurredsubsequently, triggering a large drop in the number of defaults when moving from events to episodes. Table 3. Domestic Default Events and Episodes by Continent Total Africa America Asia Europe Oceania Bonds 84 13 54 7 9 1 Bank loans 32 17 13 1 1 0 Deposits 18 1 9 2 6 0 N◦ of events 134 31 76 10 16 1 N◦ of episodes 76 25 33 5 12 1 Total number of domestic-law debt restructurings across continents from 1980 to 2018. An interesting aspect of the geography of sovereign domestic-law default episodes is that almost a third of the 52 countries in our sample are serial defaulters (Reinhart and Rogoff, 2004). They have defaulted more than once. Nicaragua is the front runner with 5 default episodes since 1980, followed by Brazil with 4 default episodes. Figure 2 plots the world map. Countries are color-coded according to the total number of default episodes in that country. Serial defaulters are found in every continent, but they are especially numerous in Latin America, where 9 countries have defaulted domestically multiple times. 12

Figure 2. Distribution of Domestic Default Episodes Total number of default episodes per country. Figure 3 shows how the geography of domestic defaults has evolved over time. The number of domestic-law defaults has been persistently high in Latin American countries. In each of the four decades from 1980 to 2018, Latin America recorded at least 6 domestic defaults. In Africa, domestic defaults have increased from the 1980s to the 1990s, likely reflecting progresses in the development of domestic financial markets. Since then, the number of defaults has remained broadly constant over the decades. Finally, domestic-law default episodes in Europe have peaked in the early 1990s, when the dissolution of the Soviet Union and the break-up of Yugoslavia have left many countries in poor financial standing. The euro-area debt crisis has also left its mark. The two European defaults in the 2010s are those of Greece and Cyprus. 13

Figure 3. Evolution of Domestic Default Episodes by Continents Total number of default episodes in each continent for each decade. 7 The Size of Domestic Sovereign Defaults Our database also tracks the volume of sovereign debt involved in each default. Figure 4 reports the median size of debt in default as a fraction of GDP in the four decades spanned by our sample. While the median size of domestic debt in default has increased from the 1980s to the 1990s, it has remained roughly constant in 2000s and then increased from 2010 to 2018, reaching around 11%. Median values mask significant heterogeneity. Peak values are more than 10 times larger than median values. For example, the Kuwait’s deposit freeze in 1990 involved deposits worth 143% of GDP and the large restructuring in St. Kitts & Nevis in 2011 involved public liabilities worth 99% of GDP. 14

Figure 4. Median Default Size by Decade Median size of default as a fraction of GDP by decade. The size of default also varies greatly depending on the instruments involved. Table 4 reports summary statistics on the size of debt in default using our sample of 134 events. The first two rows of Table 4 report summary statistics for domestic and foreign default events. Data for foreign default events is taken from Asonuma and Trebesch (2016). The comparison between default volumes in domestic and foreign default events suggests that domestic-law defaults are smaller than foreign ones. The size of domestic defaults, however, varies greatly depending on which instrument is involved in the default event. Defaults on domestic bonds are generally small in size and yet, when bonds feature collective action clauses, the average volume of restructured debt reaches 27% of GDP, which is much larger than the average default volume of any other instrument. Defaults that involve deposits are also large, with an average volume of debt in default of 23% of GDP. Defaults on domestic bank loans display the lowest average volume of debt in default: only 6.5% of GDP.17 Notably, around half of the default events that involve bank loans have occurred in Africa, and 3 of these episodes occurred in Cote d’Ivoire. Finally, we examine domestic-law default episodes that involve multiple instruments. For these, more comprehensive, restructuring episodes we find the average volume of debt involved is sizable, reaching 22 percent of GDP. 17This may reflect the authorities’ intention to protect banks in an environment in which other sources of funding are not available. 15

Table 4. Default Volumes by Instrument in Default Mean Median SD Min Max N Default Events Foreign-law 13.2% 6.0% 20.9% 0.1% 183.1% 178 Domestic-law 9.8% 3.5% 18.8% 0.0% 143.2% 134 Bonds 8.1% 3.6% 14.7% 0.0% 84.1% 84 Bonds with CACs 27.1% 13.9% 30.3% 3.0% 84.1% 7 Bank loans 6.5% 1.8% 15.2% 0.1% 84.1% 32 Deposits 23.2% 15.1% 32.7% 2.1% 143.2% 18 Default Episodes Diverse instruments 22.3% 14.9% 23.9% 4.4% 99.7% 17 SummarystatisticsforvolumesofdebtinvolvedintherestructuringasapercentageofGDP. Data for foreign-law debt comes from Asonuma and Trebesch (2016). 8 The Duration of Domestic Sovereign Defaults The median duration of domestic defaults has changed over time. As shown in Figure 5, the median duration was 12 months, in the 1980s. In the 1990s the median duration tripled, almost reaching 3 years, as the number of domestic defaults also increased dramatically.18 Since the 2000s, the duration of domestic defaults has been on a downward trend. In the last decade, the median duration of debt restructuring has fallen to 7 months, reflecting the increasing role of pre-default restructurings and the adoption of collective action clauses (CACs).19 In some cases, such as Barbados 2018, CACs were introduced retroactively to accelerate the restructuring process. As a result of such strategy, the duration of Barbados’ debt restructuring was just 5 months.20 18The resolution of domestic defaults in the Balkans, Liberia and Peru was extraordinarily slow. 19As we highlight in Section 10, the negotiation process between governments and creditors has shifted from unilaterally sanctioned restructuring operations to a more frequent involvement of private creditors ahead of defaults. 20A similar strategy was pursued in Greece 2011. Also St Kitts & Nevis used CACs during its 2012 bond restructuring, but in this case, given the English law tradition of the country, the CACs were already included in domestic bonds. 16

Figure 5. Median Duration of Domestic Defaults Mediandefaultdurationbydecade. Wedefinethedurationofasovereigndebtrestructuring as the time between its start (defined by either an actual default or the announcement of a restructuring)anditsend(definedaseitherwhenarrearsareclearedorwhenadebtrestructuring is agreed or enacted). As in Asonuma and Trebesch (2016), when the information on the starting or ending month was missing we took the following approach: in case start and end years are different, we take June; in case start and end years are the same and we have no information regarding start and end months, we take June; in case start and end year are the same and we have information regarding either the start or the end month, we set the missing month to the mid-point of the remaining part of the year. Table 5 explores the duration of the default and restructuring process in our sample of 134 default events. We define the duration of a sovereign debt restructuring as the time between its start (defined by either an actual default or the announcement of a debt restructuring operation)anditsend(definedaseitherwhenarrearsareclearedorwhenadebtrestructuring is agreed or enacted). As shown in Table 5, around 40% of domestic-law debt restructurings were resolved within 6 months. Yet, a non-negligible fraction of episodes took a very long time to get resolved. Almost one third of the domestic restructuring events lasted more than 3 years, and a 6% of them lasted more than 12 years. Peru is the most extreme case: the government began to negotiate a solution in 1992 and is yet to reach an agreement with creditors. The comparison with the duration of foreign law debt restructuring is striking. While just a little over ten percent of foreign law episodes got resolved in less than six months, four of every ten domestic episodes did. 17

Table 5. Distribution of Domestic Defaults’ Duration Less than 6 Between 6 and 12 Longer than 36 Domestic-law Defaults 42% 13% 28% Foreign-law Defaults 13% 24% 29% The table reports the percentage of domestic- and foreign-law events lasting less than 6 months, from 6 to 12 months, and longer than 36 months. Data for foreign-law debt comes from Asonuma and Trebesch (2016). Table 6 provides summary statistics for the duration of our domestic-law defaults and restructurings. Domestic defaults are on average resolved faster than foreign defaults. There are, however, instances in which the resolution took more than a decade. Defaults on bank loans feature the longest duration, followed by defaults on deposits. Defaults on bonds feature, instead, the shortest average duration, especially when they encompass CACs.21 Finally, episodes that involve restructuring diverse instruments take very long to be resolved, lasting on average more than 5 years. Table 6. Duration (months) Mean Median SD Min Max N Default Events Foreign-law 38.3 17.0 50.2 1.0 271.0 184 Domestic-law 32.8 11.5 55.7 0.0 303.0 134 Bonds 20.6 2.0 45.9 0.0 303.0 84 Bonds with CACs 8.7 5.0 10.6 2.0 32.0 7 Bank loans 55.1 38.5 65.5 0.0 301.0 32 Deposits 50.4 24.5 64.9 0.0 206.0 18 Default Episodes Diverse instruments 71.6 48.0 70.9 1.0 209.1 21 Summary statistics for duration of domestic- and foreign-law defaults. Data for foreign-law debtcomesfromAsonumaandTrebesch(2016). AsinAsonumaandTrebesch(2016),when theinformationonthestartingorendingmonthwasmissingwetookthefollowingapproach: in case start and end years are different, we take June; in case start and end years are the same and we have no information regarding start and end months, we take June; in case start and end year are the same and we have information regarding either the start or the end month, we set the missing month to the mid-point of the remaining part of the year. 21AccordingtoBaiandZhang(2012),thedurationofrestructuringofforeign-lawbondsisover12months. Five times more in the case of foreign bank loans. 18

9 Investors’ Losses in Domestic Sovereign Defaults The changes in the financial terms and conditions of the debt instruments determine the size of creditors’ losses. Losses are traditionally measured in terms of their net present value. Unfortunately, our sources often fail to report all the details of the restructuring terms.22 Consequently, we decided not to try to compute investors’ NPV losses ourselves and to rely, instead, on estimates provided by our sources. Using these, we gathered information on NPV losses for 48 instruments in 28 distinct default episodes.23 Figure 6 summarizes NPV losses at episode level.24 Median NPV losses are sizable at 31%. That said, the distribution of NPV losses has a wide support. In 39% of default episodes, NPV losses were smaller than 24%. However, in the worst 11% of the episodes, NPV losses reached 70% or higher. Analyzing the evolution of creditors’ losses over time, we find that median losses have increased from 23% in the 1990s to almost 40% in the 2010s. Figure 6. Median Creditors’ Losses by Decade Median value of creditors’ losses as a fraction of the NPV of the instrument by decade. Median NPV losses in the 1980s equaled 70%, but as we have data on NPV losses for just 3 default episodes in the decade, we add them together with those of the 1990s. The dot in the fist box represents the median value computed from 1990 to 1999. 22In particular, discount rates for the different instruments are rarely available, making the comparison between the net present value of old and new instruments impossible. 23We compared the median size and duration of the 48 events with that of the full sample. The median duration is almost identical. Instead, likely reflecting that NPV data is more readily available for recent bond restructurings, the median size is somehow larger in the NPV-sample. 24Episode level NPV losses are the average of NPV losses at the instrument level weighted by debt size. 19

The size of NPV losses varies greatly depending on the instrument involved. As shown in Table 7, creditors’ losses are smaller when government defaults on deposits, and are larger when bank loans are involved. Losses on bonds fall in the middle. When we compare investors’ losses in domestic and external restructurings, we find that investors’ losses on domestic debt are larger than those reported by Asonuma and Trebesch (2016) for foreign law debt. Table 7. NPVs of Creditors’ Losses by Instrument in Default Mean Median SD Min Max N Foreign law 38% 33% 27% -10% 97% 178 Domestic law 41% 48% 23% -5% 100% 48 Bonds 40% 47% 23% -5% 100% 32 Bonds with CACs 44% 49% 16% 23% 65% 7 Bank loans 50% 54% 19% 5% 65% 8 Deposits 37% 31% 29% 0% 74% 8 Summary statistics for creditors’ losses. Losses are expressed as a percentage of the NPV of the liability in default. 10 The Mechanics of Debt Restructuring Inthissectionweshedlightontheapproachfollowedbygovernmentstorestructuredomesticlaw debt. On the footsteps of the external sovereign default literature (Enderlein et al., 2012; Asonuma and Trebesch, 2016; Reinhart and Trebesch, 2016), three dimensions are especially relevant for us. First, we classify restructuring episodes in either pre-default or post-default. Second, we check whether restructurings were either unilateral or negotiated. Finally, we collect information on how the original terms were amended distinguishing between cases featuring a modification of maturities, a change in coupons, or a reduction of face values. Pre-default restructurings happen when the government is able to reach a restructuring agreement with creditors before default. Post-default restructurings, instead, happen when the government defaults on its debt before a restructuring process is started. Table 8 reports the split between pre-default and post-default restructurings. About 39% of the domestic debt restructurings are pre-default.25 The incidence of pre-default restructuring is in line with the one found by Asonuma and Trebesch (2016) for external defaults: 39%. Table 8 also compares the size, the duration, and investors’ losses for pre-default and post-default restructurings. Pre-defaultrestructuringsarequicker, theyinvolvebroadlythesamevolumes of debt, and deliver smaller losses than post-default ones. 25Pre-default restructurings have become more frequent over time. While only 30% of the restructurings were pre-default before 2000, their incidence increased to 41% after 2000. 20

Table 8. Pre-default versus Post-default: Main Features % Size Duration NPV (all events) (% of GDP) (months) Losses Pre-default 39% 10.4% 2.2 31.8% Post-default 61% 10.7% 50.9 40.6% The table reports average values for pre-default and post-default domestic-law debt restructurings. Turning to the restructuring procedure, we identify two: unilateral and negotiated conversions. As in Enderlein et al. (2012), we determine that a unilateral conversion occurs when contractual terms are modified unilaterally by the debtor government. This occurred, for instance, in Turkey in 1999. Negotiated conversions, instead, are characterized by the involvement of creditors. Governments may approach investors informally or propose an exchange offer. Governments often choose an informal approach when they seek to restructure selectively a portion of the debt (as in the case of Antigua and Barbuda in 1998) or when they have a direct relation with the creditor (as it is generally the case with bank loans). Exchange offers are more often used in larger restructurings that involve multiple instruments and investors. Table 9 reports the breakdown of restructuring events by instrument and procedure. While defaults on bonds and bank loans are usually resolved through negotiated restructurings, governments always took a unilateral approach when deposits were involved. Table 9 also shows that pre-default defaults are usually negotiated with investors, suggesting that governments take a more market-friendly approach for pre-default restructurings. On the contrary, post-default restructurings are often unilateral. Table 9. Restructuring procedure by instrument Unilateral conversion Negotiation Bonds 38% 62% Bank loans 29% 71% Deposits 100% 0% Pre-default 33% 67% Post-default 63% 37% Domestic-law restructuring events by restructuring procedure and instrument. Figure 7 plots the evolution over time of unilateral and negotiated defaults for pre-default and post-default restructurings. A clear pattern emerges. Along calls for increased trans- 21

parency and a more open dialogue between creditors and debtors, negotiated restructurings have become more frequent over time. This pattern is especially visible for pre-default restructurings. Since Slovenia’s debt restructuring in 2002, there has been only one unilateral restructuring (Argentina in 2012). Figure 7. Unilateral and Negotiated Restructuring Overtime: Pre- vs Post- Default Cases Breakdownofrestructuringeventsinpre-defaultcases(lefthandsidechart)andpost-default cases (right hand side chart) according to the procedure used. Finally, we review how terms are amended when government debt is restructured.26 Table 10 summarizes our findings. Maturity extension is by far the most common amendment. In 75% of the default events in our sample, the government amended the original maturity of instruments, typically extending it.27 Maturity extensions vary greatly from case to case ranging from just a few months (as in Venezuela 2003) to 50 years (as in Bosnia 1992). Amendments to the coupon structure are also frequent and are featured in 62% of the restructuring events in our sample. Coupon amendments often involve a reduction of coupon and the exchange of variable-rate coupons for fixed-rate ones. There are, however, instances (such as Argentina’s 2001 Megaswap) in which coupon payments were increased, at least on a fraction of the instruments in default, to partly compensate investors for the losses inflicted by the restructuring. 26The terms of restructuring are a key determinant of the macroeconomic consequences of external sovereign defaults (Reinhart and Trebesch, 2016). 27There are, however, cases, such as Nicaragua 1994, in which maturities were actually reduced. 22

Table 10. Restructurings by Amendment of the Terms Maturity Change Coupon Change Face Value Reduction Bonds 65 49 12 Bank loans 21 20 7 Deposits 14 14 5 N◦ of events 100 (117) 83 (100) 24 (100) Number of restructurings by type of amendments of the original terms. Events featuring morethanonetypeofamendmentsoftermsaredoublecounted. Thenumberinparenthesis corresponds to the number of events for which information on the corresponding change is available. The numbers reported in parenthesis are bigger than the number of the events, as for several default episodes our sources did not report information about amendments of the terms at the security level. Face value reductions are far less common. Only around 18% of the restructuring events in our database feature face value reductions. Pre-default restructuring almost never feature them.28 The experiences of Uruguay in 2002, Jamaica in 2010, or Cyprus in 2013 shed a light onthereasonwhyfacevaluereductionsarenotwelcomedbyinvestorsandareuncommon. In all these cases, amid a twin sovereign and bank crisis, the authorities approached investors to discuss the terms of a pre-default restructuring of sovereign debt. Creditors expressed their preference for a maturity extension over face value reductions, motivating their choice with the more negative impact that a face value reduction would have on their balance sheets.29 11 Triple Coincidence? Our definition of domestic debt is based on the law governing the debt instruments, but we also collected information regarding other aspects of domestic. Namely, we gathered information regarding whether the domestic-law debt in default was held domestically and whether it was denominated in local currency. From our sources, data on the currency denomination and investors’ residence of restructuring events were available for 87 of our 134 default events. We use this information to asses the extent to which domestic-law defaults involve debt denominated in domestic currency and held by domestic investors. Table 11 reports the mean share of domestic-law debt that involve local currency debt and the mean share of debt that is held by resident investors by decades. 28Theonly3caseswhereapre-defaultrestructuringinvolvedfacevaluereductionsareBrazil1996,Ukraine 1998 and Greece 2011. 29Restructuringsthatinvolvedepositsareaspecialcase. Whiletherestructuringofbondsandbankloans normally only affect the terms of the securities, the restructuring of deposits typically involve transforming themintoadifferentinstrument: bonds. In13ofthe16restructuringepisodesthatinvolvedeposits,investors were either given the option or were forced to convert deposits into bonds. 23

Table 11. Triple Coincidence Local currency Domestic residents No. events 1980-1990 63% 88% 8 1990-2000 66% 86% 29 2000-2010 44% 86% 21 2010-2018 79% 75% 29 Average share of restructured domestic-law debt denominated in local-currency debt and held by domestic residents. Data by decade. In each of the four decades spanned by our database, the share of domestic-law debt restructuring that involve assets held by domestic investors exceeds 75%, indicating a strong overlap between residence and jurisdiction. The share of domestic-law restructurings involving local-currency debt hovered around 65% in the 1980s and 1990s before declining to 44% in the 2010s. In the last decade, however, the trend has reversed. The share of local-law restructurings involving debt instruments denominated in local currency has increased to 79%. All told, results suggest that triple coincidence is very much alive, despite financial globalization. The domestic nature of the law governing public debt in default often translates into the residence of creditors, who are mostly domestic residents, and, to a lesser extent, to the currency denomination of the debt, which is the local currency. 12 Conclusion This paper introduces a novel database on domestic sovereign defaults based on the jurisdiction governing public debt: Domestic debt is defined as public debt issued under domestic law. The database contains 134 domestic-law default events in 52 countries from 1980 to 2018, and systematically reports information on the timing and outcome of each event. The stylized facts we present in this paper provide interesting insights that can inform the growing theoretical work on this area. In particular we draw the following lessons: 1. Domestic defaults are a global phenomenon occurring in every continent. 2. Domestic defaults are frequent and, from the mid-1990s to 2016, have become more frequent than foreign ones. Moreover, governments operate selective defaults. 3. Domestic defaults on bonded debt are by far the most common form of domestic defaults. Defaults on bank loans and deposits, used to be more frequent, but are fairly rare nowadays. 24

4. The median size of domestic defaults has increased over time reaching around 11% of GDP in the last decade. 5. Maturity extension is the most frequent form of restructuring. Amendments to the coupon structure are also frequent, while face value reductions are relatively rare. 6. Investor losses are sizable, as large as those faced during foreign law defaults. 7. Domestic debt restructurings often proceed much faster than external ones, but they can also protract significantly. 8. Post-default restructurings were the norm, but governments show an increasing preference for negotiated, pre-default restructurings. 9. Despite financial globalization, the triple coincidence is very much alive. Domestic defaults are complex and highly heterogeneous, and summarizing them using a limited number of variables may conceal more than reveal. Mindful of this risk, we complement this paper and our database with a collection of “sovereign histories” (Erce et al., 2021) that provide both an overview of the events leading to the default and the full details of the restructuring process for each debt instrument involved. This paper is just a first step in the direction of exploiting our database to foster the understanding of public debt dynamics and sovereign defaults. Topics that are especially interesting, in our view, include: the co-existence domestic and external defaults, the role played by disasters in shaping default risk, the interplay between domestic defaults and financial stability, and the interaction between political instability, inequality, and sovereign risk. We plan to address these topics in future research. 25

References Aguiar, Mark, Gita Gopinath, Manuel Amador, and Emmanuel Farhi, “Coordination and Crisis in Monetary Unions,” Quarterly Journal of Economics, 2015, 130 (4). Arellano, Cristina and Narayana Kocherlakota, “Internal debt crises and sovereign defaults,” Journal of Monetary Economics, 2014, 68 (Supplement), S68–S80. Asonuma, Tamon and Christoph Trebesch, “Sovereign Debt Restructurings: Preemptive or Post-Default,” Journal of the European Economic Association, 2016, 14 (1), 175–214. Bai, Yan and Jing Zhang, “Duration of sovereign debt renegotiation,” Journal of International Economics, 2012, 86, 252–268. Beers, David and Patrisha de Leon-Manlagnit, “The BoC-BoE sovereign default database: what’s new in 2019?,” Bank of England working papers 829, Bank of England September 2019. , Elliot Jones, and Walsh John, “The BoC-BoE sovereign default database: what’s new in 2020?,” Bank of Canada analytical notes, 2021. Bocola, Luigi, “The Pass-Through of Sovereign Risk,” Journal of Political Economy, 2016, 124 (4), 879–926. Bradley, Michael, Irving De Lira, and Mitu Gulati, “A Sovereign’s Cost of Capital: Go Foreign or Stay Local,” Duke Law Working Papers, 2016, (3620). Broner, Fernando, Aitor Erce, Alberto Martin, and Jaume Ventura, “Sovereign Debt Markets in Turbulent Times: Creditor Discrimination and Crowding-Out Effects,” Journal of Monetary Economics, 2014, 61 (C), 114–142. , Alberto Martin, and Jaume Ventura, “Sovereign Risk and Secondary Markets,” American Economic Review, 2010, 100 (4), 1523–1555. Buchheit, Lee, Guillaume Chabert, Chanda DeLong, and Jeromin Zettelmeyer, “The Restructuring Process,” In “Sovereign Debt: A Guide for Economists and Practitioners” (edited by S. A. Abbas, A. Pienkowski, and K. Rogoff), Oxford University Press Scholarship Online, 2019. Burger, John and Francis Warnock, “Local Currency Bond Markets,” NBER Working Paper, 2006, (12552). CGFS, “Financial stability and local currency bond markets,” Committee on the Global Financial System Papers 28, Bank for International Settlements September 2007. Chamon, Marcos, Julian Schumacher, and Christoph Trebesch, “ Foreign-Law Bonds: Can They Reduce Sovereign Borrowing Costs?,” Journal of International Economics, 2018, 114, 164–179. 26

Duggar, Elena, “The Role of Holdout Creditors and CACs in Sovereign Debt Restructurings,” Moody’s Sovereign Default Series Compendium, 2013. Eichengreen, Barry and Ugo Panizza, “Currency Mismatches, Debt Intolerance, and the Original Sin: Why They Are Not the Same and Why It Matters,” In “Capital Controls and Capital Flows in Emerging Economies: Policies, Practices and Consequences” (edited by S. Edwards), University of Chicago Press, 2005. Enderlein, Henrik, Christoph Trebesch, and Laura von Daniels, “Sovereign debt disputes: A database on government coerciveness during debt crises,” Journal of International Money ad Finance, , keywords=, abstract=, url= 2012, 31 (2), 250–266. Erce, Aitor and Enrico Mallucci, “Selective Sovereign Defaults,” International Finance Discussion Papers 1239, Board of Governors of the Federal Reserve System (U.S.) November 2018. , , and Mattia Picarelli, “A Journey in the History of Sovereign Defaults on Domestic Law Public Debt – Sovereign Histories,” Universidad Publica de Navarra, Economics Department Working Paper Series DT2108, 2021. Gelpern, Anna, “Domestic Bonds, Credit Derivatives, and the Next Transformation of Sovereign Debt,” Kent Law Review, 2008, 147. and Brad Setser, “Domestic and External Debt: The Doommed Quest for Equal Treatment,” Georgetown Journal of International Law, 2004, 35 (4). and Ugo Panizza, “Emerging Market Debt Crises: Economic and Legal Aspects,” Mimeo, 2021. Gulati, Mitu and Marck Weidemaier, “The Relevance of Law to Sovereign Debt,” Annual Review of Law and Social Science, 2015. IMF, “The International Architecture for Resolving Sovereign Debt Involving Private- Sector Creditors - Recent Developments, Challenges, and Reform Options,” Policy Paper 12020/043, International Monetary Fund 2020. , “Issues in Restructuring of Sovereign Domestic Debt,” Policy Paper 2021/071, International Monetary Fund 2021. Jeanneret, Alexandre and Slim Souissi, “Sovereign Defaults by Currency Denomination,” Journal of International Money and Finance, 2016, 60 (C), 197–222. Kohlscheen, Emanuel, “Domestic vs. External Sovereign Debt Servicing: an Empirical Analysis,” International Journal of Finance and Economics, 2009, 15 (1), 93–103. Mendoza, Enrique and Pablo D’Erasmo, “Distributional Incentives in an Equilibrium Model of Domestic Sovereign Default,” Journal of the European Economic Association, 2016, 14 (1), 7–44. 27

Reinhart, Carmen and Kenneth Rogoff, “Serial Default and the ”Paradox” of Rich-to- Poor Capital Flows,” American Economic Review, 2004, 94(2), 53–58. and , “The Forgotten History of Domestic Debt,” Working Paper 13946, National Bureau of Economic Research April 2008. and , “The Forgotten History of Domestic Debt,” Economic Journal, 2011, 121 (552), 319–350. Reinhart, Carmen M. and Christoph Trebesch, “SOVEREIGN DEBT RELIEF AND ITS AFTERMATH,” Journal of the European Economic Association, 2016, 14, 215–251. Sosa-Padilla, Cesar, “Sovereign Defaults and Banking Crises,” Journal of Monetary Economics, 2018, 99, 88–105. Sturzenegger, Federico and Jeromin Zettelmeyer, “Haircuts: Estimating investor losses in sovereign debt restructurings, 1998-2005,” Journal of International Money and Finance, September 2008, 27 (5), 780–805. Thaler, Dominik, “Sovereign Default, Domestic Banks and Exclusion from International Capital Markets,” The Economic Journal, 2021, 131, 1401–1427. Tomz, Michael and Mark Wright, “Do Countries Default in ”Bad Times”,” Journal of the European Economic Association, 2007, 5 (2-3), 352–360. 28

Appendix A Payment Arrears We found fragmented information on 30 events involving payment arrears with domestic suppliers. As the information about these events is often incomplete, we chose to report them in our sovereign histories but to exclude them from our database. In this appendix we review some empirical facts about domestic payment arrears, using the information from our admittedly incomplete sample. Chronology As reported in Table 12, the number of default events involving domestic arrears peaked in the 1990s at the same time in which domestic defaults also peaked, suggesting that payment arrears often accumulate when governments default on domestic debt. As a consequence, the number default events involving domestic arrears as a fraction of the total number of defaults has remained fairly constant over time. Table 12. Events of Domestic Payment Arrears by Decade Full Sample 1980-1989 1990-1999 2000-2009 2010-2018 Payment Arrears 30 3 15 9 3 Number of default events involving domestic non-financial creditors by decade. Default Volumes Domestic payment arrears tend to be large. Table 13 reports summary statistics for default volumes. The average volume of payment arrears as a fraction of GDP is close to 19%. That said, the distribution is skewed toward zero. The median value is 6%, suggesting that the volume of domestic arrears is moderate in most domestic default events. Table 13. Default Volumes as a Percentage of GDP Mean Median SD Min Max N Payment Arrears 18.9 6.18 43.45 0.74 229.64 30 Summary statistics for domestic payment arrears as percentage of GDP. 29

Duration Payment arrears to employees and suppliers feature very long duration (Table 14). Their resolution is significantly slower than the resolution of defaults that involve any other debt instruments. Delaysintheresolutionofdomesticarrearshappeninpartbecausetheplanning and execution of clearance operations are complex and normally involve a reconciliation process between the government and creditors to assess and validate the arrears. Once the validation process is completed, payment arrears are usually settled either by cash payments or through an exchange with newly issued debt. Table 14. Duration (months) of Domestic Payment Arrears Mean Median SD Min Max N Payment Arrears 88.6 72.0 74.8 2.0 305.0 30 Summary statistics for the duration of events of domestic payment arrears. Investors’ Losses We only managed to find information for five events. For these events, investors’ losses associated with domestic arrears are significant. On average 54% of the NPV of the unpaid liability is never repaid. For our very limited sample, average losses on payment arrears are larger than those on any other instrument. Table 15. NPVs of Creditors’ Losses for Domestic Payment Arrears Mean Min Max N Losses on payment arrears 54% 30% 74% 5 Summary statistics for creditor’s losses for domestic arrears. Losses are expressed as a percentage of the NPV of the liability in default. 30

31

Appendix B List of Defaults Table 16. Default and restructuring episodes Database OurDataset ReinhartandRogoff(2008) BeersanddeLeon-Manlagnit(2019) Criterion Law Residence Currency Angola 2010 1992 1990 Antigua-Barbuda 1998,2008 Argentina 1989,2001 1982,1989,2001 1982,1989,2002 Barbados 2018 2018 Bolivia 1982 1982,1984 Bosnia 1992 Brazil 1986,1990,1993,1996 1986,1990 1983,1986,1990 Cameroon 1993,2001 CaboVerde 1998,2018 1999 CentralAfricanRep. 1992 CongoDem. Rep. 1997 CongoRep. 1992 Coted’Ivoire 1989,2001,2011 Croatia 1992 Cyprus 2013 Dominica 2003 DominicanRep. 1996 1981 Ecuador 1997 1999,2008 ElSalvador 2017 1981 Gabon 1997,2001 Gambia 2017 Ghana 1982 1982 Greece 2011 2012 Grenada 2004,2013 Iraq 1990,1993 Jamaica 2010,2013 2010,2013 Korea,North 1992,2009 Kuwait 1990 Liberia 1989,2016 Macedonia 1991 Madagascar 2002 2002 Mali 2011 Mexico 1982 1982 Mongolia 1997 1997 Montenegro 1991 Mozambique 1980 Myanmar 1984,1987 1985,1987 Nicaragua 1994,1996,1999,2003,2008 2003,2008 1988,2005 Nigeria 1995 1984 Pakistan 1998 Panama 1988,1998 1988 Paraguay 2002 Peru 1985,1992 1985 1980 Russia 1998 1998 1991,1993,1998 Rwanda 1989,1994 1995 Serbia 1991 Seychelles 2010 Slovenia 1991,1995,2002 SolomonIslands 1996 1995 SriLanka 1996 1996 1996 St. KittsandNevis 2011 Sudan 2007 1991 Suriname 2001 Turkey 1999 1999 Ukraine 1998 1998 Uruguay 2002 Venezuela 1998,2002 1995,1998 1998,2016 Vietnam 32 1985 Zimbabwe 2001,2006 2000,2006 2001,2006 Domesticdebtdefaultsandrestructuringsfrom1980to2018. Thefirstcolumnlistsepisodes included in our database that follows the law criterion. The second column lists episodes includedinReinhartandRogoff(2008)whoclassifydomesticdebtaccordingtotheresidence ofinvestors. ThethirdcolumnlistsepisodesincludedinBeersanddeLeon-Manlagnit(2019) who classify domestic debt according to the currency denomination.

Appendix C Comparison with existing datasets After cross-checking our list of episodes with those reported by Reinhart and Rogoff (2008), we decided not to include various episodes they report. There were five main reasons: (i) the episode refers to arrears accumulation with suppliers (which we cover separately), (ii) the episode refers to a period of hyperinflation, (iii) the episode relates to foreign law debt (it is included because some debt-holders were residents), (iv) default corresponds to currency reforms, and (v) we found no information on the episode other than its presence in Reinhart and Rogoff (2008) database. Specifically, our database does not cover the following episodes: • Bolivia (1984) and Angola (1992) are hyperinflationary episodes. For Angola, we found information regarding the accumulation of domestic arrears to suppliers since 1992. • Argentina (1982) and Ecuador (2008) relate to external debt restructuring events reported in Asonuma and Trebesch (2016), which affected both foreigners and residents. • Ghana (1982) and Myanmar (1984, 1987) correspond to defaults due to currency changes. • Panama (1988) and Venezuela (1995) relate to an accumulation of domestic arrears toward suppliers.30 • For El Salvador (1981) and Dominican Republic (1981), we did not find any available information regarding the episodes.31 In our search, we identified 49 domestic-law default episodes involving residents, which were not covered in Reinhart and Rogoff (2008). Of these, 41 correspond to the period starting from 1980 until the last episode reported in that database. After cross-checking our list of episodes with those reported by Beers and de Leon-Manlagnit (2019), we did not take into account various episodes they report. There were three main reasons: (i) the episode relates to foreign law debt (denominated in domestic currency), (ii) corresponds to a currency reform, and (iii) we found no information on the episode other than its presence in Beers and de Leon-Manlagnit (2019). Specifically, we did not include the following episodes: • Angola(1990),Ghana(1982),Iraq(1990),(1993),Korea,North(1992),(2009),Mozambique(1980),Myanmar(1985),(1987),Nicaragua(1988),Nigeria(1984),Russia(1991), (1993),Sudan(1991),Venezuela(2016),Vietnam(1985)correspondtocurrencychanges. 30We report Panama (1988) in our database but as a deposit freeze episode. 31El Salvador (1981) is classified as a local currency default by Standard & Poor (2004), which reports governmentandcentralbanksecurities,bankloans,andcentralbankcurrencyaslocalcurrencyobligations. 33

• Argentina (1982) and Brazil (1983) relate to the external debt restructuring interventions reported in Asonuma and Trebesch (2016), which affected both foreign and local currency-denominated debt. • Suriname (2001) relates to an accumulation of external arrears, amounting to U.S. $36 million, on bank loans and not reported in Asonuma and Trebesch (2016). • For Nicaragua (2005) we did not find any available information regarding the episode. In our search, we identified 30 domestic-law default episodes involving local currency debt, which were not covered in Beers and de Leon-Manlagnit (2019). 34