Default Premium

Luis A.V. Catao Rui C. Mano∗

This Draft: December 2016

Abstract

The literature has found that sovereigns with a history of default are

charged only a small and/or short-lived premium on the interest rate war-

ranted by observable fundamentals. We re-assess this view using a metric

of such a “default premium” (DP) that nests previous metrics and apply-

ing it to a much broader dataset. We find a sizeable and persistent DP: in

1870-1938, it averaged 250 bps upon market re-entry, tapering to around

150 bps five years out; in 1970-2014 the respective estimates are about 350

and 200 bps. We also find that: (i) the DP accounts for between 30 to 60

percent of the sovereign spread within five years of market re-entry, and its

contribution to the spread remains non-negligible thereafter; (ii) The DP

is higher for countries that take longer to settle with creditors and is on

average higher for serial defaulters; (iii) our estimates are robust to many

controls including realized “haircuts”. These findings help reconnect theory

and evidence on why sovereigns default only infrequently and, when they

do, why earlier debt settlements are typically sought.

Keywords: Sovereign Debt, Country Risk, Interest Rate Spread, Haircut, Emerging

Markets

JEL Classification: F34; G15; H63; N20

∗Catao: IMF (corresponding author: lcatao@imf.org); Mano: IMF. We thank Ana Fostel andGuido Sandlieris for very helpful discussions and the research department of the IADB for support atthe onset of this project, as well as two anonymous referees for many helpful comments on earlier drafts.We are also grateful to Marcelo Abreu, Kalina Dimitrova, Ljiljana Djurdjevic, Rui Esteves, ClemensJobst, Sofia Lazaretou, Dan Li, Debin Ma, Kim Oosterlinck, Sevket Pamuk, Leandro Prados, CarolinaRoman, Solomos Solomou, Coskun Tuncer, Loredana Ureche-Rangau, and Henry Willebald for helpwith various national data sources, as well as to Olivier Blanchard, Emine Boz, Jorge de Braga Macedo,Graciela Kaminsky, Paulo Mauro, Jay Shambaugh, and seminar participants at the Banco Central doBrasil. George Washington University, IMF, Nova University, Oesterreich NationalBank, Universityof Osnabruck, and Vienna Graduate School of Economics for comments. The views expressed in thispaper are the authors’ alone and do necessarily reflect those of the IMF, its boards of directors ormanagement.

1

1 Introduction

Do countries pay an interest rate premium for having defaulted in the past? If so, how

high and for how long? These are important questions in international finance. The

prevailing view is that such a premium is small and/or short-lived on average. This view

is readily apparent from earlier studies by Eichengreen and Portes (1986), Lindert and

Morton (1989), and Jorgenson and Sachs (1989) who find that countries that honored

their debts during the 1930s Great Depression did not benefit from lower spreads in the

early post-World War II period relative to those that defaulted. This finding is only

mildly overturned in Ozler’s (1993) classic study of emerging market loan data from

1968 to 1981. Ozler found that a default memory dummy, which differentiates between

countries that defaulted in previous decades and those that did not, is statistically

significant but the implied difference in interest rates is economically small (25 basis

points at the mean). Using a similar sample but an instrumental-variables approach

to decompose the spread into a backward-looking component related to credit history

vs. a forward-looking default risk component driven by fundamentals, Benczur and

Ilut (2015) arrive at a similar estimate. For the 1880-1913 period alone, Flandreau and

Zumer (2004) estimate a slightly higher interest rate cost of default (up to 90 basis

points) in the first year after the settlement of arrears, but one that decays significantly

thereafter.

Other research finds a more sizeable premium but which dies out even more rapidly.

Using JP Morgan emerging market bond index (EMBI) spreads over 1997-2004, Boren-

sztein and Panizza (2009) estimate whether a default has an extra impact on the country

spread upon market re-entry by adding year dummies (starting from the debt settlement

date) to an otherwise standard empirical model of sovereign spreads. From the esti-

mated coefficients and standard errors on those year dummies, they find that defaults

have a very sizeable impact on the first year of market re-entry (up to around 400 basis

points), falling to 250 bps in the second year, and becoming insignificant thereafter.

Cruces and Trebesch (2013), also using JP Morgan EMBI data (extended to 2010) and

a similar set of year dummies except that interacted with the “haircut” size for each

debt settlement, obtain a more persistent effect. Yet, this is only so for defaults that

involved investors’ losses or “haircuts” above the sample average (37 percent). For the

average emerging market defaulter, they estimate an interest rate premium of under 100

2

bps four to five years after settlement, but that also ceases to be statistically significant

at 5 percent.

This evidence raises critical issues. First, the magnitude and persistence of such

interest rate premia seem puzzlingly low given that governments often go to considerable

length to avoid defaults, including through politically costly fiscal austerity and the

conditionalities of multilateral financing. This puzzle is all the more apparent if much

touted costs of default, like market exclusion and trade and military sanctions (what

Mitchener and Weidenmier, 2005 call “super-sanctions”), are frequently unimportant.

As Gennaioli et al. (2014) put it: “In reality, sanctions are rarely observed and market

exclusion is short-lived”. Costs associated with the break-down of domestic financial

intermediation could be one missing link (Bolton and Jeanne, 2011; Gennaioli et al.,

2014; Kalemli-Ozcan et al., 2015), but only so for countries and periods where and when

the share of sovereign bonds in domestic banks’ portfolios is sufficiently high. Insofar

as this deterrence mechanism is not ubiquitous, one is then back to the issue as to what

sustains non-trivial levels of sovereign debt in the first place.

Second, and as stressed in a recent survey by Wright and Tomz (2013), existing

empirical studies are unclear as to what mechanism(s) generate(s) such positive in-

terest rate premia, since none of them offers a model to guide their metrics. Indeed,

in the canonical sovereign debt model of Eaton and Gersovitz (1981) and subsequent

extensions (Aguiar and Gopinath, 2006; Arellano, 2008; and others), sovereign credit

history per se should not add a premium to country spreads. This is because those

models feature full information on fundamentals and shocks as well as investors that

break even at all times; hence default/repayment decisions per se do not add informa-

tion to bond pricing. An alternative is to do away with the assumption that investors

are risk neutral and break-even; instead, defrauded creditors could collude and impose

above-market lending rates to recoup losses and thus penalize defaulters.1 While this

could be rationalized by “cheater of the cheater” arguments (Kletzer and Wright, 2000),

this presumption seems difficult to reconcile with evidence of new lenders undercutting

older ones and of overall lenders’ surpluses being close to zero historically (Eichengreen

and Portes, 1986; Lindert and Morton, 1989 and Klingen et al., 2004).2 If so, such

1This is the interpretation favored by Benczur and Ilut (2015).2This is not to deny the importance of lenders’ market power in enforcing repayments and extracting

3

a “punishment” mechanism also seems unfit to deliver a positive default premium in

competitive bond markets. Introducing asymmetric information on the sovereign’s type

(Eaton, 1996; Alfaro and Kanczuk, 2005) or on the shocks it receives (Sandlieris, 2008;

Catao, Fostel and Kapur, 2009) can generate a positive default premium in competi-

tive bond market equilibria, but then the issue becomes how to establish which “true”

fundamentals or shocks are unobserved (or imperfectly observed) by investors. These

different modeling possibilities are notoriously hard to disentangle in the data. So,

from a purely empirical standpoint what one needs is an empirical specification that

can accommodate these distinct theoretical mechanisms - especially in a cross-country

sample (like ours) containing both sovereign bonds and loan contracts.

Third, the above studies span distinct country samples and periods and exclude

many of today’s advanced countries. These can also default and some of them have

done so in both the distant and the recent past. Before one underplays the interest rate

cost of default, a look at a more representative sample seems in order

This paper re-assesses the effect of sovereign credit history on country spreads on

the basis of a better metric and broader data. A first contribution is to build a metric

that conforms to existing theories of a positive default premium (DP henceforth) and

to relate such a metric to those used in the empirical studies reviewed above. We show

that our metric, which combines indicators of the frequency, timing and duration of

the relevant credit events (defaults and debt settlements), nests those, thereby allowing

comparability; in addition, we highlight the distinction between using our metric vs.

the unconditional difference in the spread before default and after debt settlement as a

gauge of the DP.

The paper’s second and key contribution is to measure the DP in a much broader

dataset than before.3 Our dataset spans the entire 1870-1938 period as well as 1970-

2014 for both emerging and advanced countries, rather than just for emerging markets

post-1990. Relative to classical studies of pre-WWII spreads by Bordo and Rockoff

(1997), Obstfeld and Taylor (2003), Flandreau and Zumer (2004), Mauro et al. (2006),

Tomz (2007), it adds more countries, fixes issues with the data and also comprises

surpluses from borrowers in certain circumstances, specially in the context of relationship banking andsyndicated loans (see Voth (2011) and Flandreau and Flores (2012) for eloquent historical illustrations).

3Our dataset and replication files are available at: XXXX.

4

more control variables, including for global market factors highlighted in recent work

(Borri and Verdelhan, 2011; and Longstaff et al., 2011). Regarding the post-WWII

period, we extend many series on sovereign spreads back to the 1970s and 1980s and

add advanced countries to the sample. This addition allows us to span major recent

events in sovereign debt markets as well.

By combining the first and the second contribution, the paper yields a third con-

tribution: it documents salient features of the empirical DPs hitherto uncovered. It

does so by decomposing the spread into a vector of observable fundamentals, the DP,

and an estimation error, as well as by allowing for distinct functional forms for memory

formation, and re-evaluating the importance of the actual size of the default (the final

“haircut”) in shaping the evolution of the DP.

The results are: (i) the year after a debt settlement is agreed with creditors, the

average DP was around 250 basis points in the period 1870-1938 and 350 bps in 1970-

2014. The DP then decays but slowly: by the end of the 5th year after debt settlement it

averages some 150 basis points in the pre-WWII period and 200 bps post-1970, tapering

off thereafter; (ii) while the DP is significant at conventional statistical levels, it also

displays considerable variance across countries and over time, so limited sampling (as

in previous studies) can greatly affect inference on the mean DP; (iii) the DP accounts

for between 30% to 60% of the market spread within five years after settlement with

creditors and no less than 10% of the average spread farther away from default and

debt settlement; (iv) the DP rises with the number of years a sovereign stays out of the

market - this being higher for serial defaulters implies that they pay a higher DP; (v)

The estimated DP is broadly robust to controlling for alternative sets of fundamentals,

and no smaller for alternative specifications on how investors’ memories evolve; (vi)

unlike Cruces and Trebesch (2013), direct control for the size of the actual haircut in

spread regressions is not necessary to obtain a significant and longer lasting DP, partly

because near 90% of the actual haircut can be predicted by the length of the default and

country-specific fundamentals;4(vii) reduced-form numerical illustrations suggest that a

DP of similar magnitude and persistence as that found in this paper is an economically

powerful incentive to repay debt.

4As will be discussed in Section 4.2, this result should not be construed as a dismissal that actualhaircuts are necessarily irrelevant for the evolution of the DP for individual countries, as (still) limitedhistorical data on haircuts prevent more extensive testing.

5

The remainder of the paper is structured as follows. Section 2 discusses the theoret-

ical motivations for the existence of a positive DP and lays out the empirical DP metric

used in the remainder of the paper. Section 3 introduces the new dataset. Section 4

reports our DP estimates, attendant robustness tests, and traces down the differences

with earlier estimates. Section 5 provides simple (reduced-form) numerical illustrations

on economic significance of the estimated DP as a repayment incentive. Section 6 con-

cludes. Specifics of the data and other derivations are provided in Appendices A and

B.

2 Empirical Specification

As noted in the introduction, existing theoretical models allow for the possibility of

a positive interest premium for past default if: i) lenders can “punish” and try to

recoup, in any future lending, losses due to a past default on the original terms of

the loan or the bond contract; ii) default coveys new information, over and above

observed “fundamentals” and/or observed shocks, about the borrower’s riskiness going

forward. Mechanism (i) derives inspiration from the theoretical work of Kletzer and

Wright (2000) and empirical support from Benczur and Ilut (2015). The “revelation”

mechanism (ii) can stem from imperfect information on the country’s type (Eaton, 1996;

Alfaro and Kanczuk, 2005) or on persistent shocks that affect the continuation value of

repayment decisions, as in Sandlieris (2008) and Catao, Fostel and Kapur (2009).

In all of the above models, spreads will be a function of not only the borrower’s ob-

served fundamentals and shocks (including shifts in investors’ risk aversion), but also of

the borrower’s credit history. In the particular case of asymmetric information models,

one should also expect that spreads “jump” upon default announcements and drop upon

debt settlement arrangements; away from those events, spread should also remain above

pre-default levels in both classes of models, once one properly controls for changes in

observed fundamentals. A further implication of imperfect information models is that

insofar default “discloses” an unobserved feature of the borrower that is relevant for

future default risk, but such a feature is less than fully persistent, the default premium

should decline over time. In both mechanisms (i) and (ii) the probability of default and

the expected haircut are jointly determined in the pricing of future lending and the size

of past haircuts only matters for the spread to the extent that past haircuts affect the

6

expected size of any future haircut. Benjamin and Wright (2009), Ghosal et al. (2010)

and Yue (2011) establish that haircuts tend to be highly correlated with the duration

of default as well as with the risk free interest rate and the debt-to-GDP (D/Y) ratio;

thus, the expected haircut is largely a combined function of economic fundamentals

and credit history indicators. By the same token, historical evidence (Tomz, 2007;

Sturzenegger and Zettelmeyer, 2006) suggests that the relationship between past and

future haircuts is complex, among other things because larger-than-average haircuts

may aid debt sustainability and so reduce the likelihood of large future haircuts.

Grounded on these theoretical considerations, we can now write the equation for

each country spread at time t as a function of a variety of observed macro fundamentals

(including notably D and Y) well as parameters and shocks, and a term for the default

premium. As in all the empirical literature on sovereign risk, we let this function take

a linear form. So, the determination of the sovereign spread evolves as:

si,t = αi + Fi,tβ +DPi,t + εi,t (1)

where si,t is the spread on country i’s debt at time t over the risk free interest rate, αi is

the country-specific fixed effect, Fi,t is a row vector of observable macro fundamentals

(possibly including cross-country common factors or time fixed effects), DPi,t is the

default premium, and εi,t stands for shocks to country risk.

Such a parsimonious specification can accommodate important features of existing

theories of the DP. First, to the extent that it takes time for investors to learn about

the nature of underlying shocks and parameters, and as learning is never complete,

a persistent and possibly slowing decaying DP will ensue; so αi + Fi,tβ will often

underestimate country risk. Second, insofar as some countries are more prone to such

large and persistent shocks and/or their degree of impatience can go higher than others,

they are likely to default more often; so, lenders will also learn from cross-country

repayment patterns, potentially leading to substantial cross-country variance in the

DP and hence in sovereign spreads. Third, even if the vector of commonly observable

fundamentals Fi,t evolves slowly, spreads can jump if the DPi,t jumps.

The next step in bringing theory to data is to let DPi,t take the following functional

7

form:

DPi,t = γ1MEM1i,t + γ2MEM2i,t +m∑j=1

δjDS(j)i,t (2)

where MEM1i,t is defined as the ratio of the number of years that country i is in

default to the total number of years in the sample up to year t. A similarly defined

variable appears in Reinhart et al. (2003) as proxy for credit history in their analysis of

sovereign risk ratings. We fine tune this variable by considering distinct choices for the

initial year when relevant history begins: we consider 1824 as in Reinhart et al. (2003)

but also experiment with other starting dates. One limitation of MEM1 is that it does

not differentiate between, say, two countries that have the same percentage share of

defaulted years in their history but one of which has defaulted more recently. To capture

that, we introduce the variable MEM2 (featured in previous work by Eichengreen

and Portes (1986), Reinhart et al. (2003), Flandreau and Zumer (2004), and Esteves

(2007)), defined as the number of years since the last default. Hence, while γ1 should

be positive, γ2 is expected to be negative, implying that the longer the time past since

the last default the lower the default premium. Appendix A.3 provides an illustration

of how MEM1 and MEM2 are constructed.

How do MEM1i,t and MEM2i,t relate to the theories reviewed above? If defaults

are triggered by a large and persistent negative shock that may not be anticipated and

fully unobserved by lenders, MEM1i,t is summary statistic for the unobserved nature

of both the size and the persistence of such a shock which the act of defaulting more

fully reveals.5 MEM2i,t complements MEM1i,t with further information on the timing

and duration of the relevant shock.6 In the same vein, MEM1 and MEM2 may also be

capturing variations in unobserved parameters defining a country’s “type”, about which

default/repayment decisions can be highly informative (as argued in Eaton, 1996). As

5In fact, it is reasonable to presume that such infrequent shocks are not readily identifiable evenby an econometrician armed with macroeconomic data spanning several years and with the benefit ofhindsight. See Cochrane (1994) for a pertinent discussion of the difficulties in the ex-post identificationof major macro shocks and of the role of agents’ “actions” in aiding such identification. Our MEMvariables are, essentially, summary statistics for these “actions”

6Note that upon the year of debt settlement, MEM2 is also gauging the length of the default. To theextent that longer defaults signal to investor that such countries are more subject to persistent shocksand/or that creditors’ losses in debt renegotiation will likely be higher the longer the default (consistentwith evidence to be presented in Section 4) calling for greater subsequent loss-recovery “punishment”,the sign of MEM2 could be ambiguous. Yet, because MEM1 also picks up length-of-default effects,what ultimately matters is the combined effect of MEM1 and MEM2 coefficients. Hence our focus inSection 4 on the joint significance of the two variables.

8

learning about unobserved shocks and country types can be gradual, the slow-moving

nature of MEM1i,t and MEM2i,t may be proxying for such gradual learning too.7

A limitation of MEM1i,t and MEM2i,t is that they are slow-moving variables

by construction. In a setting where discrete actions can be highly revealing of the

state of unobserved fundamentals, as well as in occasions when investors have greater

market power to reprice risk, changes in the sovereign spread can be rather discrete

around credit events. In our case, the relevant credit events are default and repayment;

accordingly, the relevant observation points for the computation of the DP are the pre-

default and post debt settlement years. The dummy sequencem∑j=1

δjDS(j)i,t in equation

(2) is meant to capture such a possible discontinuity. The dummyDS(j)i,t will equal one

if t is the jth year after which settlement on the defaulted debt is agreed with creditors;

for all other years DS(j)i,t is set zero. With 0 < j ≤ m, the number of such dummies

is m for all countries. If a country defaults more than once (i.e. is a serial defaulter),

then each of these dummies will take a value of 1 more than once per country; if instead

a country never defaulted, then MEM1i,t = MEM2i,t =m∑j=1

δjDS(j)i,t = 0, i.e., the

default premium is naturally zero. The reader is referred to the Appendix A.3 for

illustrations on how each component of the empirical default premium is constructed.

As will be seen in Section 4, we will set m = 5 on empirical grounds: post-default

dummies beyond a 5-year window are no longer statistically or economically significant.

Regression estimates will tell us which of the dummies up to m = 5 have significant

coefficients depending on the specification and estimation period. The coefficients on

MEM1i,t and MEM2i,t will shed light on the hypothesis that investors’ memories last

longer than this 5-year window. By combining those two “memory’ variables featuring

in previous work by Reinhart et al. (2003), Eichengreen and Portes (1986), Flandreau

and Zumer (2004), Esteves (2007) with the discrete post-settlement dummies used in

Borensztein and Panizza (2009) and Cruces and Trebesch (2013), our DP metric nests

all those previous specifications. Restricting credit history to be proxied only by the

settlement dummies reduces not only the size but also the persistence of DP (as will

be shown in section 4). Conversely, ignoring those dummies implies missing out on the

significant post-settlement dynamics of the DP, as shown below. Thus, the inclusion

7See Cogley and Sargent (2007) and Boz and Mendoza (2014) for a richer formalization of agents’learning on related deep parameters, which also boils down to using historical counting indicators akinto our MEM variables.

9

of the three components of the DP is important to account for the differences between

our estimates of the interest rate cost of default and those of previous work.

Combining equations (1) and (2) yields the following specification for the market

spread of country i for each t in a cross-country panel with fixed effects:

si,t = αi + Fi,tβ + γ1MEM1i,t + γ2MEM2i,t +m∑j=1

δjDS(j)i,j + εi,t (3)

An important feature of the DP as defined above is that it does not necessarily

move one-to-one with the (unconditional) market spread si,t. Indeed, measures of the

interest cost of defaulting based on a difference of the market spread before default and

after settlement can be misleading. To see this, suppose that a country’s (publicly-

observed) fundamentals strengthen during the spell between default and market re-

entry. Then, a fall in the market spread would be warranted by those fundamentals

and this would not imply that the DP would be lower (or higher). Yet, equating the

DP to the unconditional difference between the pre-default and post-settlement spread

would suggest that the DP is lower. This point can be readily seen by re-writing (1)

as:

∆DPi,t,T = ∆si,t,T −∆Fi,t,Tβ −∆εi,t,T (4)

The above equation shows that changes in the market spread equal changes in the

DP if, and only if, ∆Fi,t,T = 0, i.e., the observable fundamentals do not move between

the eve of the default at t and the year of market re-entry, T . In principle, the DP can

be positive even if the spread at settlement turns out to be lower than that before the

default, i.e., we can have a situation where ∆si,t,T < 0 and ∆DPi,t,T > 0. Conversely,

we can have a situation where the spread is higher and the DP is zero. In short,

looking at differences between the pre-default and the post-settlement spread may be

an inaccurate indicator of the “true” interest rate cost of default. We examine in Section

4 whether this has been the case.

3 Dataset

Following most studies, we use the Standard and Poor’s (S&P) definition of default.

The latter is either a unilateral interruption of repayment of interest and/or principal

10

on contractual debt obligations by a sovereign government, or an event where the

sovereign tenders an exchange offer to swap new for existing debt with less favorable

terms than the original issue and the offer is accepted (even if contentiously) by lenders.

Consistent with this definition, we also mostly follow the S&P dating of the end of

defaults. The timing of the latter is based on the date of the settlement between

sovereign and investors on (all or most) outstanding arrears, and on the judgment by

Standard & Poor’s that no further near-term resolution of creditors’ claims is likely

(Beers and Chambers, 2006). In a few cases, however, we depart from these strict

definition for reasons already discussed elsewhere. One reason, highlighted in Arellano,

Mateos-Planas and Rios-Rull (2013), is that the S&P dating of the end of defaults does

not always coincide the full elimination of arrears: in a few cases a non-trivial ratio of

arrears to outstanding debt obligations lingers on after the S&P settlement dating. In

our sample, this is mostly strikingly the case for Argentina which, according to Standard

& Poor’s, settled with creditors in 2005 but substantial hold-outs in such negotiations

implied that over 60 percent of the face value of outstanding debt remained in arrears

through the end of our sample period (2014). 8

Another adjustment to some of the S&P dating comes from treating defaults as

watershed-like occurrences in the sense that re-scheduling episodes that are ramifica-

tions of the initial default are not counted as new defaults. The rationale for doing

this is supported by the arguments and evidence presented in Bussiere and Fratzscher

(2006), Gourinchas and Obstfeld (2012) and Catao and Milesi-Ferretti (2014), about the

stronger (and potentially rather perverse) feedback loops between country risk, growth

and indebtedness during partial renegotiations and partial repayments – the so-called

“muddling-through” dynamics. As an illustration, take a country like Brazil which

defaulted in 1983, but then renegotiated part of the defaulted debt in 1984, 1986, and

1988 before a full-fledged renegotiation in 1992 under the so-called Brady debt deal. In

such a case, we treat 1983 as the year of default and 1993 as the first year after the final

debt settlement – rather than treating 1983, 1984, 1986, 1988 as separate default events

8Some non-trivial arrears also lingered on for Brazil after the 1992-93 Brady deals, the Ecuadordeal in 1997, and after the Indonesian settlement in 2002. However, unlike in the Argentine case, someof such delays in arrear reduction were already contemplated in the settlement deal and indeed rapidlytrended down after the S&P settlement date. No less importantly, fresh borrowing resumed with avengeance the year after those settlement dates.

11

with partial market “re-entries” in between. 9 This procedure is akin to that adopted

by other classical sources on the chronology of sovereign defaults, such as Suter (1992)

and Beim and Calomiris (2001), as well as by Reinhart and Rogoff (2009).10 Another

departure of the S&P chronology consists of not classifying as default some technical

delays in repayment amounting to no more than a few days or weeks, as by Venezuela

in July of 1998 and Peru in September 2000 11. Appendix Table A1 provides the full

list of default and settlement dates in our sample, indicating the cases in which our

dating differs from Standard & Poor’s.

Alongside with the rationale of treating defaults as major, water-shed credit events,

an important feature of our econometric estimation of the default premium is to drop

from the sample the entire period during which the country is deemed in default (as per

the chronology laid out in Appendix Tables A1 and A2). Going back to the illustration

about Brazil’s default of 1983, this means dropping the entire set of annual observations

between 1983 and to 1992 from the estimated sample. The rationale follows from the

very concept of default premium as in the theoretical literature reviewed above. As

we document below, spreads typically jump upon the default announcement, consistent

with the idea that default announcements reveal new information to investors. Fur-

thermore, spreads typically remain very high while the country is in default (at least

in the cases when defaulted bonds are at all traded so one can actually measure the

spread). There is little dispute that there is then a high interest premium associated

with default in those occasions. This is not the interest premium that this paper seeks

to measure. Indeed, not dropping the default period from the sample would impair the

interpretation of the sum of estimated coefficients on our default settlement and the

“memory” variables as a measure of the default premium consistent with the discussion

in Section 2. 12

9In the case of Brazil, a final tranche of negotiations followed in 1994, so there might arguably bea case for taking 1995 instead of 1993 as the first year after the very final round of debt settlement,but not to take Brazil 1994 as yet another credit event, as Cruces and Trebesch (2013) do

10see http://www.carmenreinhart.com/dataa/browse-by-topic/topics/7/11Another issues is how to classify defaults that were arguably averted by large multilateral inter-

vention, such during the emerging market crisis of the late 1990s and the eurozone debt crisis morerecently. Manasse and Roubini (2009) provide a rationale for potentially including these events in oursample. When we do, we find that the thrust of findings stands up (see Table 4 in the working paperversion of this paper at https://www.imf.org/external/pubs/ft/wp/2015/wp15167.pdf). But since thisinclusion also raises questions on the interpretation of the coefficients on the default premium relatedto the role of bail-out expectations and IMF seniority in those events, we do not discuss them here

12For the sake of completeness in our data collection work, we collected bond yield data during

12

By using this definition of defaults as water-shed events and dropping all the default-

year observations from our sample, we clearly loose some country-year observations for

which we have spread data. Yet, our sample still spans as many as between 1361 to

1639 annual observations for pre-WWII period and between 1421 and 1569 for the

period 1970-2014, the lower limit of those sub-sample ranges being determined by the

availability of data for the right-hand side variables that one chooses to include in

the spread regression. The gain of loosing those default-year observations is a cleaner

definition of credit events and the theoretically relevant concept of the DP, as well

as a mitigation of feedback loops between debt renegotiation rounds and fundamentals

which could affect consistency of the estimates (as discussed in Bussiere and Fratzscher,

2006).

A main contribution of our empirical analysis to the literature is our new database

on sovereign spreads. For the entire pre-WWII period, our dataset uses yields on long-

dated government bonds derived from bond prices in secondary bond markets in main

global financial centers (namely, London, New York and Paris) – with the respective

country spread computed as the arithmetic difference between those yields and the

yields in either the UK, the US, or the French government bonds (depending on which

of those financial centers the borrower’s issued) of a comparable maturity. 13 Relative

to the studies on the determination of sovereign spreads spanning pre-WWII data by

Bordo and Rockoff (1997), Obstfeld and Taylor (2003), Flandreau and Zumer (2004),

Mauro et al. (2006), Tomz (2007), and Catao, Fostel and Kapur (2009), we draw on a

variety of new sources to add more countries and correct various data inconsistencies

(as detailed in Appendix A). Regarding post-WWII data, we augment the datasets

relying only on the J.P. Morgan EMBI spread (which starts in the 1990s), by extending

the sovereign spread series backward to 1970 for both advanced and emerging markets.

the default years (when available) but, as explained above, do not include those observations in theeconometrics and DP measures presented in this paper. It is also worth noting that because defaulteddebt may not be traded at all during those default spells, secondary market observations may not beavailable. This is particularly the case for the defaults of the 1980s.

13One issue with using secondary market prices to compute the yield to maturity is that such an yieldmay not capture the effective borrowing cost to the sovereign after debt settlement if the sovereigndoes not issue new bonds, which countries may choose to do, inclusive for market “timing” reasons.Yet, this is not typically case. In our post-WWII sample, for which we have information on the timingof primary issuance, the sovereign issues new debt in the year after debt settlement in no less than 57percent of the cases. Moreover, abstaining from issuance until the spread in primary markets subsidescarries in itself an economic cost.

13

This is done by combining yield data from both primary bond issuance and secondary

market trading, obtained from the sources listed in Appendix A.1. In doing so, we

control for the potential differential effects in spreads arising from the use of primary

vs. secondary market yields, as described in Section 4. While the Reinhart and Rogoff

(2009) database does not contain spread data, we also improve on their database by

filling gaps in country series for key fundamentals, such as the ratio of external to

total debt and monetary data (see also the Appendix). The end-product is a much

more extensive annual database on sovereign spreads, totaling around 3200 observations

for econometric analysis, i.e, even after observations for default years are dropped as

explained above. This is far larger than the 131-144 observations in Ozler (1993, Table

1), 215-265 in Flandreau and Zumer (2004, Table A.7), 144-162 in Borensztein and

Panizza (2009, Table 5), 169 in Benczur and Ilut (2015), and 310-447 annual-equivalent

observations in Cruces and Trebesch’s (2013) dataset. On those grounds alone, one

might expect our estimates of the interest rate cost of default to be more globally

representative and the attendant inferences to be more robust. Figures 1 and 2 provide

a panorama of our sample, showing the evolution of the median sovereign spread, its

cross-country dispersion, and the percentage of countries in default over both the pre-

WWII and post-1970 sub-samples, respectively.

4 Estimation

4.1 Econometric Controls

Because of data discontinuity around World War II, as well as the far-reaching structural

changes in the global financial system around those years, it is natural to break down the

econometric estimation between the pre-WWII (comprising 1870-1938) and the post-

WWII sub-samples. Given the far-reaching capital controls which greatly impaired

international private lending in the 1950s and 1960s, as well as because related data

limitations before the 1970s, our post-WWII latter starts in 1970 and ends in 2014 .14

Key to a sound measurement of the default premium is the inclusion in equation (3)

14Not until the end of the Bretton Woods regime in 1971 did international capital flows becomesufficiently free, and private markets for sovereign bonds were rebuilt, so as to make our analysismeaningful. Confining analysis of post-WWII data to the post-1970 years also ensures more directcomparability with pre-WWII sub-sample.

14

of variables that span the relevant set of publicly-observable fundamentals. Accordingly,

we include in Fi,t all the variables typically featuring in theories of sovereign risk,

namely: i) the ratio of public debt to GDP ratio; ii) the ratio of external to total public

debt; iii) the real world (risk-free) interest rate; iv) a measure of external imbalance,

such as the trade or the current account balance; v) a trend-like measure of real GDP

growth (computed as a moving average of real GDP growth rates over the past three

years); vi) classic indicators of the output cost of default, such as trade openness, foreign

exchange reserves (as a ratio to imports or GDP) and a fixed exchange rate dummy.15

In addition, it is also sensible to assume that actual investors may, at times, depart

from risk neutrality. To this effect, we add to the regressions an indicator of global

risk-aversion (proxied by an index of stock price volatility in the UK for the pre-WWII

period and the US VIX index for the post-1970 period).16

We then conduct extensive tests on the robustness of the estimated DP to the

inclusion of other variables in the vector of macro fundamentals. These include the ratio

of M2/GDP – a classic indicator of financial depth and hence of the financial disruption

and related output cost of default, as well as of a country’s capacity to borrow in one’s

currency and escape “original sin” (Eichengreen, Hausmann and Panizza, 2003). We

also include an indicator of country-specific exogenous shocks, which we proxy by the

gap between the external terms of trade and its 3-year past moving average17, as well

as for country-specific investment growth, inflation (relative to that of the UK and the

US), and the general government balance to GDP. 18 In addition, in light of evidence

from Tomz (2007) that seasoned borrowers (as opposed to newly independent countries

that are new comers to world capital markets) tend to borrow at more favored terms,

we include a sovereign dummy variable, which is one for newly independent countries

in the post-WWII period. Finally, we also consider measures of institutional stability,

including the well-known Polity index and the ICRG index of political risk. Appendix

I contains further specifics of the data and the data sources.

15See, e.g., Aizenman and Marion (2004) and Borensztein and Panizza (2009)16We have also considered the high yield corporate spread for the U.S. as an alternative indicator

of global risk aversion which is sometimes used in the literature (e.g., Cruces and Trebesch, 2013 andBenczur and Ilut (2015). Yet, its significance was dwarfed by that of the U.S. VIX.

17Mendoza (1995), Acemoglu et al (2003), Blattman et al (2006), Catao, Fostel and Kapur (2009)all use the TOT as a gauge for the exogenous country-specific component of macro shocks.

18As aggregate investment data before WWII is unavailable for several countries in our sample, weonly control for investment growth in the post-1970 regressions.

15

All variables entering the regressions are lagged one year and all robust standard er-

rors are computed clustered at the country level. We also add to all pre-1939 regressions

a World War I dummy defined as equal to 1 in 1914-18 and zero otherwise; this serves

to capture the disruptive effects caused by the war on international interest spreads.

It is worth noting that some of these variables can encapsulate other (and sometimes

contradictory) effects, so the size and signs of the respective coefficients can differ and

likely to be affected by multicollinearity.19 Yet, by eliminating from the sample the

years in default as explained above, one might expect that endogeneity biases are mit-

igated for the reasons discussed in Bussiere and Fratzscher (2006) and Gourinchas and

Obstfeld (2012).

Finally, we address the issue of model uncertainty following Pesaran and Timmer-

mann (2007), who recommend averaging out the respective parameter estimates across

models and specifications, without requiring such models or specifications to be nested.

As shown below, our inferences are robust to this procedure as well.

4.2 Regression Results

4.2.1 Pre-WWII Period

The results for the 1870-1938 period are shown in Table 1. The first column shows the

estimates for our baseline specification with country fixed effects. The next six columns

add various controls for fundamentals, whereas the last column repeats the baseline

specification but including time fixed effects (hence the dropping of global factors such

as the world interest rate and global stock market volatility). Key fundamentals like

the overall public debt/GDP ratio, the ratio of external to total debt, the gold-peg

dummy yield the expected signs and are statistically significant for the most part.

The specification without time fixed effects in columns (1) shows that world financial

19One example of potentially ambiguous effect is the fixed exchange rate dummy. As in Bordoand Rockoff (1996) and Obstfeld and Taylor (2003), being pegged to the gold standard in much ofthe 1870-1939 was perceived by investors as a seal of good fiscal and monetary housekeeping, thushelping lower country spreads. But fixed exchange rates can also increase country risk by encouragingborrowing in foreign currency and constraining macroeconomic management of real shocks (see Obst-feld, Shambaugh, and Taylor, 2005). One illustration of collinearity is that the non-trivial panel-widecorrelation in the data between current account balance (as well as the trade balance) and the fiscalbalance. As will be seen in the regression results below, the explanatory power of the current accountbalance typically dominates that of the fiscal balance.

16

market conditions – proxied by the world real interest rate – matter and with the

theoretically expected sign. While UK stock price volatility also appears to play a

role in the pre-WWII sub-sample, it becomes insignificant at 5 or 10 percent once the

global interest rate indicator is included in the regression, and so was dropped from the

baseline specification. Columns (2) to (8) show that other country fundamentals do not

add significantly to the specification. In particular, foreign exchange reserves, a 3-year

moving average of real GDP growth, and the fiscal balance (as ratio to GDP) yield

the expected negative sign, while the trade balance and the terms of trade yield the

opposite sign, but none of them is significant.20 The dummy for whether the country

is a new sovereign (so to capture the “seasoned borrower effect” postulated in Tomz,

2007) is also far from being statistically significant. Finally, all columns indicate that

the fundamental most critically important in theory — the public debt to GDP ratio

— is reassuringly significant in all specifications and with similarly sized coefficients,

and regardless of whether includes or excludes time fixed effects from the specification.

Consistent with the hypothesis of a positive default premium, the post-settlement

dummies (as discussed, starting from the first year after debt renegotiation through 5-

years hence) yield sizeable and significant coefficients through the second or third year

after settlement in all four specifications. The two memory variables are economically

and statistically significant in specifications without a time effect: MEM1 (years in

default as a share of total years up to that point) is sizeable; the negative coefficient

of 0.01 for MEM2 indicates that investors’ memories fade away at a rate of 1 basis

point per year. This suggests that re-building investors’ confidence takes time. While

introducing time effects in column (9) robs the significance of MEM2 due to multi-

collinearity, MEM1 maintains significance and is also similarly sized across the various

specifications. Importantly, we show at the bottom of Table 1 that F-tests for the joint

exclusion of all MEM variables (i.e. MEM1 and MEM2) reject such an exclusion at

levels of significance well below 5 percent.

Finally, the last two rows of Table 1 report the estimated mean DP for each spec-

20In addition to the terms of trade gap and to allow for the possibility that country specific aggregatevolatility is time-varying, we have also included the standard deviation of the country’s terms of tradeover over a rolling 10-year window (as in Blattman et al, 2007) as part of our tests but that variable isnot significant either. Like the addition of stock market volatility, this variable addition test was notreported to save space but is available from the authors upon request

17

ification at the first and the fifth year after debt settlement. This is calculated by

plugging into equation (3) the estimated coefficients on the DS(1) to DS(5), as well as

those on MEM1 and MEM2. For the specification in column, the mean DP is 251 basis

points in the first year after settlement and 150 bps in the fifth year; so, it is quite

sizeable and persistent. Despite some variation across specifications (partly reflecting

also changes in sample size dictated by the availability of long pre-WWII time series

for some variables), the thrust of the results holds up very well. In particular, while

we prefer the specification without the time fixed effects because it explicitly identi-

fies the common factors that theory postulates, mitigates multicolinearity with MEM2

and post-settlement dummies, and yields a comparable adjusted R2, the inference of a

sizeable and persistent DP is clearly robust to that preference.

4.2.2 1970-2014

Table 2 reports the baseline results for 1970-2014. The first column shows that many

of the well-known determinants of country risk show up as highly significant, including

lagged real GDP growth, the U.S. VIX, foreign exchange reserves, and the current

account balance – the latter two variables being, as usual, scaled by GDP.21 As in the

pre-WWII, the stock market volatility indicator (the VIX) shows non-trivial collinearity

with changes in the global real interest rate but, unlike in the pre-WWII, dominates

the latter so we drop it from baseline specification. Newly included in these post-WWII

regressions are two dummies that equal one when the spread corresponds to that of a

primary issuance of a bond or of a loan. These two controls were unnecessary in the

pre-WWII regressions because, as explained in Section 3, all the pre-WWII spreads are

secondary-market bond spreads (see also Appendix Table A1).

Those spread observations associated with primary issuance are important to fill in

for the scarcity of secondary market observations in the period 1970-1990 (see Appendix

Table A2). Yet, they also pose a potential censoring bias to the regression estimates.

This is because the sovereign decision to issue debt or not may be strategic: the incentive

to issue is higher when spreads are lower and vice-versa. Further, this incentive will

not affect all countries equally because some unobservable country characteristic may

make it easier for some countries to abstain from entering the market during “bad”

21This result is robust also to scaling foreign reserves by imports, as is also customary.

18

times or, if they do, to obtain more favorable spreads than the “average” borrower.

Whether this censoring bias will bias our DP estimates positively or negatively will

depend on whether the net effect is one of adverse selection (when only the least credit-

worthy borrowers issue driving up the observed average spread) or positive selection

(when only the most credit-worthy borrowers issue, driving down the observed average

spread). Eichengreen and Moody (2000) and Dell’Ariccia et al (2006) deal with this

issue by applying the classic Heckman correction wherein an auxiliary selection equation

(which determines the threshold above or below which the observation is censored)

is added to the observation equation – their counter-part of our equation (3). An

important condition for the success of the bias correction is the use of one or more

instruments that capture the unobserved determinants of the decision to issue debt.

In addition to standard fundamentals already present in the observation equation, one

sensible instrument used, for instance, in Dell’Ariccia et al (2006) is per capita income.

This is because richer countries may find it either easier to abstain from issuing when

spreads are generally very adverse or being able to get the best rates during those

times due, e.g., to flight to safety.22 As in Dell’Ariccia et al (2006, table 2), we find that

the censoring bias is insignificant: the correlation coefficient ρ between the residual of

the spread observation equation and that of the selection equation is sufficiently close

to zero at -0.03, with a p-value of .10. Notwithstanding, we still need to include the

primary issuance and the loan vs. bond dummies in our regressions so as to control

for the possibility that bonds in primary markets may be priced differently from those

in secondary markets, as well as for the possibility that loan instruments may carry a

lower than bonds – as found to be the case in the pre-1990 sample used in Ozler (1993)

and Benczur and Ilut (2015). The coefficients on both dummies reported in Table 2

indicate that this is the case.23

As with the pre-WWII regressions, we subject the baseline specification of column

(1) of Table 2 to a host of variable addition tests as well as the inclusion of fixed time

22Such an instrument is akin to the concept of reservation wage in the well-known application of theHeckman model to labor market participation (see Killingsworth and Heckman, 1986).

23This raises the question of what is the correlation between primary and second bond spreads whenobservations of both spreads are available for the same country/year. In the appendix we show that thecorrelation is reasonably tight along a near-45 degree line. The significance of the dummy on primarybond issuance is largely driven by a few observations for which secondary bond spreads turn out to besignificantly higher than primary bond spreads. Results excluding all primary issuance observationsare reported in Table 5 further below

19

effects. These are reported in columns (2) through (9). One can then see that none

of the additional variables are statistically significant with the exception of inflation

in column (7); yet, further probing indicates that this significance is largely driven

by a few outlier observations associated with high and hyper-inflation bouts in some

emerging markets. Turning to the individual components of the DP, Table 2 also shows

that the post-settlement “impact” dummies are mostly significant – both statistically

and economically – through the fourth year and that their coefficients are sizeable and

higher than in the pre-WWII period. The MEM variables are individually statistically

significant at either 5 percent or very close to 5 percent (as in column (8)) and jointly

significant at 1 percent, with sizeable coefficients too.24 Memory effects are thus more

long-lasting than entailed by the post-settlement dummies alone: as with the pre-

WWII regressions, loss of investors’ memory is gradual.

Finally, as done above for the pre-WWII sample, we report in the last two rows of

Table 2 the estimated mean DP for each specification at the first and the fifth year after

debt settlement. For the baseline specification in column (1), the mean DP is 350 basis

points in the first year after settlement and 204 bps in the fifth year; so, it is even more

sizeable than in the pre-WWII era. Variation across specifications is not large (and

despite also some changes in sample size dictated by the length of some time series for

some emerging markets) through column (7), dropping only more significantly (about

90 basis points) upon the addition of fixed time effects in the last column, when the

estimated DP becomes closer in magnitude to that of pre-WWII sample. Overall, the

inference of a sizeable and persistent DP thus remains broadly robust to the alternative

specifications and sample periods.

4.2.3 Robustness to Memory Alternatives

In Tables 3 and 4 we provide two extra tests that are novel to the literature. One is

to add a variable (called “mem3”) that measures the number of defaults per country.

This extra memory indicator captures the extent to which, holding constant the average

duration of defaults to date (as captured by MEM1), a country that defaulted more

often tends to face a higher spread relative to the country that defaulted less often

24While we also experimented with 1824 as the starting point for those memory variables, the fitwas worse and suggestive that WWII was a major watershed in investors’ memories.

20

(but that stayed longer in default). The result, reported in column (2) of Tables 3

(pre-WWII) and 4 (post-1970), indicates that the difference is statistically insignificant

and, if anything, fewer but longer defaults tend to trigger higher default premia relative

to more frequent but shorter defaults.

Next we report on results of distinct specifications for the decay of investors’ mem-

ory. In addition to the linear discounting underlying MEM2, we also consider quasi

hyperbolic discounting, fully hyperbolic discounting, exponential discounting and in-

verse (geometric) discounting.25 In all cases, MEM1 is reassuringly significant, while

the mileage for MEM2 varies some (note that the change in sign is consistent with the

change in functional form, in all cases implying a reduction in the default premium as

time goes by). Importantly, F-statistics at the bottom of the table shown MEM1 and

MEM2 to be jointly significant at 5 percent for all specifications, while regression fits for

these alternative MEM2 specifications are no better than in the linear baseline specifi-

cation of column (1). So there is a case to stick to the linear functional form. Moreover,

the DP yielded by the alternative memory specifications is, if anything, higher rather

than lower in most cases. This further supports our central finding that the size of the

default premium has been typically underestimated in the literature.

4.2.4 Mapping with Previous Studies

The strength of the above results raises the question of why previous studies found a

much smaller and/or short-lived DP. How much of our results can be attributed to more

comprehensive country and time coverage and/or to our DP metric? In this subsection

we trace out such differences for the post-1970 period.

Column (1) of Table 5 reproduces our baseline specification in column (1) of Table 2.

25Let T be the time since the last default. We consider the following functional forms:a. Quasi-Hyperbolic Discounting: βδT . In this case, again, MEM2=0 when T->infinity, which

is the case of countries that never defaulted. But then, even for the mild assumption of k=1, thediscounting can be quite strong, certainly stronger than exponential discounting. And again MEM2would still be defined over (0,1).

b. Hyperbolic discounting: mem2 = γ/(1+kT ), where γ and k measure the strength of memory.We set γ = 1 and k = β defined as above. For countries that never defaulted T -> ∞, so MEM2'0for any positive k. Conversely, if T=0, i.e, the default took place an infinitesimal instant ago, thenMEM2=1. So MEM2 is bounded, being defined over [0,1).

c. Exponential discounting: mem2 = exp(−kT ), where k is a discount factor that we assume toequal the average real interest rate during each of our sub-samples.

d. Inverse (geometric) discounting: MEM2 = 1/T.

21

We then first ask whether our higher and more persistent estimates for DP coefficients

are due to the introduction of two MEM variables. Column (2) shows that the removal

of those variables greatly increases the size and the significance of the post-settlement

dummies all the way through five years after debt settlement. The regression fit is

however significantly lower than the specification with the MEM variables and the

resulting DP drops from 351 to 274 basis points in the first year after debt settlement

and then down to 135 basis points five years hence. So, dropping our MEM variables

does reduce the DP but still leaves it sizeable and more persistent than in earlier studies.

What this specification looses is what the MEM variables buy – namely, the possibility

of a default premium that persists beyond those few years and that might still account

for a non-trivial portion of the overall sovereign spread, as will be shown in section 4.4

below.

We turn next to investigating the possibility that differences in sampling are critical.

The third column (”No Adv”) starts that probe by dropping all advanced countries from

the sample so to compare our results more directly with those of the studies reviewed

in the introduction which span only emerging markets and low income countries. The

sizes of the post-settlement dummy coefficients as well as that for MEM1 marginally

increase; the main difference with the baseline specification is that MEM1 is significant

at almost but not quite at 5 percent and MEM2 is not significant. Yet, F-tests for the

joint significance of MEM1 and MEM2 cannot reject their significance at 2 percent or

higher26 Importantly, both the magnitude and the persistence of the mean DP would

be even a bit higher if we drop advanced countries from the sample. So, the inclusion

of advanced countries is not critical to our central findings.

The source of differences in results with previous estimates becomes obvious once we

narrow the sample further. As shown in the fourth column (”No prim”), when we drop

all spread observation associated with primary issuance, only the first post-settlement

dummy is statistically significant at conventional levels and both MEM variables are

estimated less precisely, with the statistical significance of MEM1 falling just below 10

percent, though its coefficient rises in magnitude. Because the primary issuance spreads

are mostly confined to the pre-1990 period, this data coverage restriction is telling

26Moreover, if one excludes the world interest rate variable which is non-significant, MEM1 recoversits significance at 5 percent.

22

us that including the 1970-1990 years to the sample is important to obtain sharper

estimates. Further truncations of the sample exacerbate this problem. As shown in

column (5), truncation to a much shorter period, so span 1997-2004 as in Borensztein

and Panizza (2008) brings the argument into sharper relief: on then obtains a rapidly

decaying DP and MEM variables that are statistically insignificant (justifying why in

Borensztein and Panizza (2008) do not include them). Conversely, dropping the post-

1990 years so to span only the 1970s and 1980s (as in Benczur and Ilut, 2015), one

finds instead a far smaller DP (64 basis points in the first year after debt settlement)

– as indeed those authors have found. The estimates in this also reveal a striking drop

in DP persistence, with the mean DP essentially vanishing (at 1 bp) by the fifth year

after settlement.

Last but not least, we use our broader sample re-examine the claim by Cruces and

Trebesch (2013) that controlling for realized haircuts are essential to obtain a more

sizeable and persistent default premium. The column labeled “Like CT” of Table

5 presents the results with the introduction of extra terms named “DS(.)xHC” that

interact the post-settlement dummies with the size of the haircut (”HC”). To allow

a more direct comparison with their results, we use the same configuration of post-

settlements dummies as theirs, i.e., we combine the dummy for the 4th and 5th year

and that for the 6th and 7th year together. In addition, we limit our sample to their

period coverage (1993-2010) and their spread series (the JP Morgan EMBI spread).

Truncation of the sample is further dictated by the availability of haircut data as per

their original series, resulting in a sharp drop in the number of annual observations to

425 from 1569 in our baseline specification. Other changes to the specification relative to

baseline include adding time effects, the institutional investors rating index (computed

as residual in a panel regression on the IIR rating on fundamentals as they do), the

ICRG political risk index, inflation, and scaling reserve by imports. As none of these

variables proved to be statistically significant they were not included in the reported

regression. In light of the evidence presented just above that sample truncation can

critically affect the DP estimates, any inference about the value added of haircuts in

accounting for the DP should therefore be treated with care. That said, an F-test

on the joint significance of interactive (DS*HC) dummies is significant at 5 percent,

thus corroborating Cruces and Trebesch’s (2013) claim that the actual haircut does

matter for the spread going forward. Note, however, that MEM1 and MEM2 are highly

23

significant – both jointly and individually – suggesting that the Cruces and Trebesch

specification (which does not feature MEM variables) underplays the persistence of

the haircut beyond the sixth or seventh year after settlement. Be that as it may, the

significance of the haircut dummies does not seem to be robust to an enlargement of

the sample and a reversal to our preferred specification without time fixed effects and

using global financial condition indicators to control for common factors. As shown

in the last column of the Table, when we use a broader spread series (that spans also

non-EMBI spread data, including those for advanced countries) and use an updated

of haircut series to 2014 (as per Reinhart and Trebesch, 2014) raising the number of

observations from 425 to 720, the haircut dummies are no longer jointly significant at

conventional levels.

There are clear rationalizations for this mixed result on haircuts. First, as noted in

Section 2 the probability of default and the expected haircut are jointly determined in

theory; so to the extent that the determinants of both the default probability and the

expected haircut are already included in the spread regression, such a regression would

then be correctly specified in theory. In other words, our point DP estimate is the

estimated mean DP associated with the panel-wide mean haircut. A related reason is

simply econometric: even if our baseline regressions omitted important determinants of

haircuts that are unrelated to default risk, and presuming that such determinants are

randomly distributed, the standard OLS normality assumptions in a large sample such

as ours would make the estimated mean DP unbiased. Third and no less importantly,

the haircut that matters in a competitive bond market with forward-looking bond pric-

ing is future haircuts. As such, past haircuts are only relevant if they provide additional

information on expected haircuts in the future. The question would then be whether

some of the right-side variables in our baseline specification do provide relevant infor-

mation on expected haircuts. Regressions reported in Table 6 provide an affirmative

answer. The dependent variable in those regressions is now the actual haircut at the

time of the respective debt settlement for the countries covered in Cruces and Trebesch

(2013) but with the updated haircut dataset of Reinhart and Trebesch (2014). The ex-

planatory variables are, in turn, those featuring in our model and previous regressions.

The first column reports results for a sample of 42 debt settlement episodes that include

a few low income countries outside our sample. This regression shows that the variables

already included in our baseline specification explain as much as 88 percent of actual

24

haircuts. If this sample is limited to only emerging markets (all of which included in our

sample), the respective R-square remains high at 0.88. In particular, the actual haircut

is highly predictable by the length of default (here captured by MEM2) and the debt to

GDP ratio. This is consistent with evidence reported in Benjamin and Wright (2009),

and Ghosal et al. (2010) of a high correlation (>0.6) between actual haircuts and the

length of default which, in turn, can vary with the severity of the triggering shock –

as per our discussion in Section 2 (see also Edwards, 2015 for supportive empirical ev-

idence).27 Overall, we interpret these results as conforming to our earlier discussion in

that the inclusion of the determinants of the expected haircut (but not necessarily the

actual haircut) in the regression is sufficient to obtain a sizeable and persistent default

premium in a broad cross-country panel.

In a nutshell, we are not asserting that actual haircuts – or indeed any other measure

of the effective size of investors’ loss from past defaults – are necessarily irrelevant for

the evolution of the DP for individual countries. We prefer instead to interpret the

evidence of this section as largely agnostic about the value added of the unexpected

component of past haircuts to understand the evolution of sovereign bond spreads.

This is not only due to the conceptual reasons mentioned above, but also due to data

limitations: more extensive historical haircut data seem needed to test that hypothesis

with greater confidence.28 Either way, the key and novel result here is that finding a

sizeable and long-lasting DP in cross-country data is not crucially dependent on adding

past realized haircuts to the attendant regression.

4.3 The Default Premium in the Aftermath of Debt Settle-

ments

We now zoom in on the decay of the DP in the first few years after debt settlement and

the cross-country dispersion thereof. Figures 3 and 4 summarize this evidence, using the

27Such a highish correlation in turn helps explain the loss of significance of the MEM variables whenactual haircuts are included in the spread regressions, since the actual haircut is highly collinear withMEM2. Both Benjamin and Wright (2009), and Ghosal et al. (2010) provide theoretical mechanismsas to why the length of default helps predict haircuts.

28Kaminsky and Vega-Garcia (2015) provide evidence that haircuts play a role in the terms of futuremarket access in pre-WWI Latin America but do not directly compute a DP as above. Incorporatingpre-WWII haircut to our DP regressions is a worthy extension of our analysis once pre-WWII hair cutdata becomes available for more than a handful of countries.

25

coefficients of our estimated baseline specifications for pre- and post-WWII (as reported

in the first columns of Tables 1 and 2) to compute the DP (as before, using equation

(2)). Figure 3 plots the evolution of the cross-country mean DP as well as the estimated

dispersion (in dotted lines) around it. One can then appreciate not only the tapering

of the mean DP through the fifth year after debt settlement but also a considerable

dispersion around the mean. Recalling the expression for the DP in equation (2),

there is a clear and important interpretation for such a large dispersion. Since the

estimated coefficients on the default settlement dummies and memory variables (MEM1

and MEM2) are invariant across countries, this dispersion of DPs is thus entirely due

to differences in the values of MEM1it and MEM2it. Thus, observations below the

mean spread stand for the DP in countries with smaller blemishes in their credit history

and/or blemishes that occurred a long ago (i.e. have higher values of MEM2), whilst

the mean corresponds to the DP of the average defaulter, i.e. that with average values

of MEM1 and MEM2.

[Figure 3 about here.]

Corroborating the point estimates of the DP provided at the bottom of Tables 1

and 2, Figure 4 indicates that the inference of a sizeable DP for the average defaulter

is robust to various changes in model specification: instead of using our preferred

specification, we compute the median of the default premia obtained by all alternative

specifications (eight in total for each – pre- and post-WWII– sub-periods) in the spirit

of the Pesaran and Timmermann (2007) test for model uncertainty. The resulting error

bands corroborate the inference of a positive and sizeable DP.

[Figure 4 about here.]

Comparing the estimated median DP in both eras with the (unconditional) median

of market spreads, shown in Figures 5a and 5b and 2 6b, one can readily appreciate

the value added of the DP metric.29 They differ non-trivially, suggesting that the

“raw” comparison between pre-default and post-settlement spreads is not generally a

good proxy for the DP. As discussed in Section 2 and fleshed out in equation 4, the

29We base these comparisons on the median rather than the mean to mitigate the effect of largeoutlier observations for the unconditional spread

26

theoretically appropriate measure of the DP must net out changes in the publicly-

observable fundamentals from the market spread. If this is not done, Figures 5a to

6b show that one ends up underestimating the DP at market re-entry by as much as

150 bps in the pre-WWII era and overestimating it by over 50 bps the post-1970 era.

Another misleading inference if one focuses on the unconditional spread would be to

assume that the “true” interest cost of default (i.e. the DP) decayed less smoothly than

it does in the post-WWII sample.

[Figures 5a-6b about here.]

Finally, these Figures also highlight how the spread jumps around default events,

so does the DP also before and after default. Next we sharpen this association further,

by decomposing the change in the spread into changes in observed macro fundamentals

vs. changes in the DP.

4.4 Fundamentals vs. Default Premium

Combining the estimated parameters of our spread regressions with equation (4) allows

us to measure how much of the change in the sovereign spread is accounted for changes in

macro fundamentals rather than in the DP and in the estimated error term. The novelty

of this decomposition exercise relative to previous studies is not only the decomposition

of the spread during the first few years after debt settlement but also during periods

away from defaults (i.e. away that 5-year window aftermath). This allows us to gauge

the importance of the DP when credit conditions are less “distressed” and spread levels

can be a lot lower.

Figure 7 shows that the default premium accounts for between 30 to 50% of the

spread within five years after debt settlement in 1870-1938. For the post-1970 period

(Figure 8), the contribution is even higher on impact, reaching up to 60%. In com-

parison to the contribution of DP to changes in the market spread, the contribution

of macro fundamentals is strikingly stable. Moving to periods far-away from defaults,

the Figures also show that the DP contribution to the spread does fall considerably.

Yet, it is never quite trivial. For the pre-WWII it remains at about 20% of the spread

and just over 10% of the post-1970 period. In absolute terms, these percentage shares

27

correspond to default premia of about 50 bps and 30 bps respectively. While these

margins may seem low, they are not negligible (or strictly zero, as in some previous

studies). Further, their sizes need to be evaluated in due context: this includes periods

of global market bulls and/or very low global real interest rates, when investors’ search

for positive yields is typically aggressive and any difference of a few basis points in

interest spreads can make a big difference for portfolio allocation.

[Figures 7-8 about here.]

Figures 9 to 12 offer further insight. Bearing in mind the arguments in Reinhart

et al. (2003) on the distinctive features of serial defaulters, we break down this spread

decomposition exercise between countries that defaulted only once and those that de-

faulted more than once. As before, we compute this breakdown separately for pre- and

post-WWII data. Both the spread and the DP are larger for serial defaulters. This

is true for both the post-settlement period as well as for periods far-way from default

and post-settlement. The contribution of DP to the overall spread in periods of more

“normalized” credit conditions is particularly non-negligible for serial defaulters.

Finally, Figures 9 to 12 also indicate that the estimation residual (as gauged by

the gap between the spread in dotted lines and the top of the vertical bars for each

year) is mostly trivial for the pre-WWII decompositions as well as for the post-1970

decomposition for non-serial defaulters. Interestingly, however, Figure 11 indicates a

not-so-trivial residual for post-WWII serial defaulters, which is suggestive of greater

cross-country heterogeneity in this group, consistent with its comprising of advanced

countries like Cyprus and Greece and not just emerging markets. The upshot is that

serial defaulters in general face higher borrowing costs for the same set of fundamentals

vs. non-serial defaulters, and this is true in the immediate aftermath of debt settlements

as well as in periods away from default and debt renegotiation events.

[Figures 9-12 about here.]

5 Default Premium as a Repayment Incentive

As discussed in the introduction, a DP that is suficiently high can help explain why

sovereign defaults are infrequent. The question is then how high. While a thorough

28

answer to this question is best pursued in a calibrated DSGE model - an endeavor

beyond the scope of this paper – the economic significance of our DP estimates can

still be highlighted through simple numerical exercises. Benczur and Ilut (2015) use an

Euler equation on the cost-benefit trade-off of delaying debt repayment for one period to

argue that an interest rate premium in the range of about 30 to 70 basis points suffices

to discourage default for a sovereign with a discount factor no lower than 0.89, facing

a base lending rate of 5 percent per annum, and zero hair-cut. This section examines

the more general case of multi-year defaults, non-zero haircut, a slow-decaying DP, and

non-zero debt and output growth.

To focus on the role of the DP in repayment decisions, we shut off two important

costs of default featured in the literature. One is the use of external borrowing to

smooth consumption. Instead, we consider a sovereign whose utility is linear in pay-

offs, in which case foregoing consumption smoothing is not a cost. Second, our baseline

illustration rules out output costs due to default; we simply assume that output Y

is exogenous and independent from default/repayment decisions. To simplify further

the algebra and sharpen intuition, we assume also that the sovereign’s target ratio of

debt to GDP is interest rate inelastic and that all debt is financed with a one-period

bond whose coupon interest rate is re-set at each period depending on the sovereign’s

decision to repay vs. default. To aid tractability, we also rule out serial default: when

a sovereign defaults and then settles the debt, it can credibly commit not to default

again.

Under these assumptions, a sovereign stands to gain from default by pocketing

the debt service in t and buying the possibility of inflicting lenders a haircut of h on

outstanding arrears when it settles the defaulted debt, since 0 ≤ h < 1. In turn,

default entails two costs. One is a positive DP that the sovereign will pay in all future

borrowing after default ends. The other potential cost is the inability to bring forward

consumption on the back of future income growth. This cost is greater the higher

the long-term growth rate of the borrowing country, and the higher the borrower’s

subjective discount rate relative to the non-default interest rate offered by lenders r,

i.e., the smaller β relative to the inverse of gross lending rate offered by lenders, 1+r.30

30In the calibrations presented below we also mute this consumption front-loading channel, so as tofocus even more squarely on “pure” DP effects. This is done by equalizing the output growth rate andthe base lending interest rate.

29

Faced with these cost-benefit trade-offs, the sovereign will in any period t repay in full

the debt outstanding at the end of the previous period (i.e. Dt−1) only if the benefit of

repayment is no smaller than the cost of repayment.

5.1 Measurement Framework

In the setting discussed, the condition for full debt repayment can be formalized as:

Yt − (1 + r)Dt−1 +Dt + βV rt+1(.) ≥ Yt + βV d

t+1(.)

where the left hand-side of the equation represents the net benefit of repaying and

the right hand side the net benefit of defaulting. The term V r,dt+1(.), to be formally defined

shortly, denotes the future value of each “strategy” i.e., repayment (r) or default (d),

expressed in terms of its current period pay-off as per the one-period discount factor β,

where 0 < β < 1. The above equation simplifies to:

Dt + β{V rt+1(Dt, Yt, 0)− V d

t+1(Dt−1, Yt, 0)}≥ Dt−1 (1 + r) (5)

where the state variables in the argument of V r,dt+1 are assumed to be prior period

debt, output, and the default premium. The latter being zero in V rt+1 follows from our

earlier assumption of no serial default, i.e., prior to a possibility of default in t, the

sovereign is charged no default premium. The default premium argument in V dt+1 is

also zero because the default premium falls only on debt issued after the end of the

default, so will not kick in until V dt+J , where J is the duration of the default. Note that

the argument of V dt+1 in case of default is Dt−1 reflecting the fact that the debt stock is

carried forward until it is settled.

The value function under repayment is:

V rt+1(Dt, Yt, 0) = Yt+1 − (1 + r)Dt +Dt+1 + βV r

t+2(Dt+1, Yt+1, 0)

And under default:

V dt+1(Dt−1, Yt, 0) = Yt+1 + βV d

t+2(Dt−1, Yt+1, 0)

30

Generalizing for a default that lasts J periods yields:

V rt+1(Dt, Yt, 0) =

J−1∑j=0

βjYt+1+j−rJ−1∑j=0

βjDt+j +J−1∑j=0

βj∆Dt+1+j +βJV rt+J+1(Dt+J , Yt+J , 0)

V dt+1(Dt−1, Yt, 0) =

J−1∑j=0

βjYt+1+j−βJ−1 (1 + r)J (1−h)Dt−1+βJ−1Dt+J+βJV rt+J+1(Dt+J , Yt+J , s0)

where s0 is the DP on any new debt immediately after debt settlement is reached.

After default ends, i.e at J+1, and assuming the debt path is independent thereafter,

the value functions take the form:

V rt+J+1(Dt+J , Yt+J , 0) = Yt+J+1 − (1 + r)Dt+J +Dt+J+1 + βV r

t+J+2(Dt+J+1, Yt+J+1, 0)

V rt+J+1(Dt+J , Yt+J , s0) =

[Yt+J+1 − (1 + r + s0)Dt+J+

Dt+J+1 + βV rt+J+2(Dt+J+1, Yt+J+1, s0 (1− δ))

]

where s0 (1− δ) stands for the second-period-after-default DP, with s0 being the DP

at the first year after debt settlement and δ measuring the speed of decay thereafter.

A default that lasts several periods requires us to make assumptions about the

evolution of debt and GDP. Since the relevant variable for the determination of the

spread and of DP is the debt to GDP ratio in our regressions, it is useful to divide the

expressions above through by Y . Assume further that Y simply grows over time at rate

gy, and that this growth rate is invariant to the default/repayment decision. In that

case, output is no longer a state variable and we re-express equation (5) as:

1

1 + gy

(dt − dt−1

1 + r

1 + gy

)+ β

{vrt+1(dt, 0)− vdt+1(dt−1, 0)

}≥ 0 (6)

where small caps refer to variables divided by current year’s output. See Appendix

B for detailed calculations. A natural benchmark is that countries have a targeted debt

to GDP ratio (d). Coupled with the assumption that the sovereign chooses to have the

same debt to GDP ratio as it would have had if it hadn’t defaulted, and thus dt+j = d

for all j ≥ 0, one can replace the value functions expressed in terms of GDP in equation

31

(6) to get31:1 + gy − dt−1

d(1 + r) + (gy − r)

∑J−1j=0 (β (1 + gy))j+1 +

+ (β (1 + gy))J+1 s01−β(1+gy)(1−δ)

+βJ (1 + r)J (1− h)dt−1

d− (1 + gy) (β (1 + gy))

J

≥ 0 (7)

5.2 Estimates

We solve equation (7) as an equality for s0 as a function of β, h, and the remaining

parameters. While the above framework is sufficiently general to nest a variety of

scenarios, our DP estimates in the previous subsections focus on the average defaulter;

so, a natural benchmark is to fix parameters with that case in mind. We fix the

duration of the default at the post-1970 sample median which is five year (J = 5).32

We fix growth (gy) at the panel mean of 3.5 percent and assume that debt-to-GDP is

unchanged before and after default. This is again roughly consistent with the average

default case in our sample, where the ratio of debt to GDP in the five years preceding

default is roughly the same as that in the five years after the sovereign settles the

default at around 50%. We set the decay of the DP at the same level as in the baseline

specification of Table 2, which is δ = 0.12533. The base real interest rate is set to

3.5 percent, which is equal to the average nominal interest rate of the 10-year US

treasury subtracted from average US CPI inflation added of the 1 percentage point

spread that the average non-defaulting sovereign paid in the period 1970-2014. So, the

usual r=g equality is thereby incorporated in the calibrations presented below. Since

there is substantial variation regarding β in calibrated models of sovereign debt — and

empirical estimates thereof are in short supply — we let β vary between 0.8 and 1. We

then ask what is the minimum DP required to sustain repayment (i.e. for equation (7)

to hold as an equality) as we vary β and the expected haircut h.

Figure 13 shows that a DP sightly under 100 bps in the first year after debt settle-

ment is able to sustain debt for the case of no haircut and β down to 0.85. Such an

estimate is closer to the upper range of that of Benczur and Ilut (2015). But Figure

31See Appendix B for detailed calculations.32The median in the pre-WWII is very close at 4.5 years.33As it turns out, the average decay is not substantially different between and first and fifth year

from between the fifth and tenth year after settlement, although the rate of decay is not constant fromone year to the next.

32

13 also indicates that the minimum DP to ensure repayment is highly sensitive to the

expected haircut. As one might expect, the higher debt forgiveness a sovereign expects

to be able to extract from lenders, the bigger the benefit of defaulting. To illustrate

the sensitivity of the required DP to the expected haircut, consider in Figure 13 the

iso-line for a haircut not too far off the sample mean.34 At the post-1970 estimated

(first-year) DP of 350, repayment can be sustained only if the borrower is about twice

as impatient as lenders, i.e., has a discount factor of 0.93. To make more impatient

sovereigns indifferent — those with a β say in the 0.8 to 0.9 range — would require

a DP upwards of 500 bps. In an environment (like today) where the base interest is

lower and/or in the case of a country that expects to grow above 3.5 percent in the

long-term, the required minimum DP would be substantially lower. These sensitivities

can be easily computed in the above setup and deliver intuitive results. But again,

the focus here is on the average defaulter and the key message conveyed by Figure 13

is that a DP of 350 bps or thereabouts is able to sustain repayment for haircuts that

are non-trivial if the sovereign is not overly impatient. This is so even in the absence

of other well-know costs of defaults, such as output and reputational losses, as well as

foregoing of consumption smoothing through borrowing while in default. In short, the

thrust of this simple numerical exercise is that a DP within the range of our estimates

can go a long way to incentivize repayment, thus helping explain – in conjunction with

other likely default costs – why defaults are infrequent in a broad cross-country data.

[Figure 13 about here.]

6 Concluding Remarks

This paper has argued that the interest rate cost of default has been underestimated in

the literature. To appreciate further the economic significance of our estimates, consider

the average post-1970 defaulter with a mean public debt/GDP ratio of 50 percent. Our

DP point estimates imply an extra interest cost of 1.75 percent of GDP in the year after

34Using the Cruces-Trebesch data, the mean hair-cut is 0.36 over the defaulted debt. Arellano etal. (2013) note that defaulting sovereigns on average default on 60 percent of their debt. Multiplyingthe two figures yields a hair-cut of 22 percent on the total debt stock. In our sample, which comprisesmany large defaults of the 1980s and the 1990s, as well as Greece more recently, the share of defaulteddebt can be as high as 80 percent or so, bringing the average hair-cut on the total debt within a 25 to30 percent range.

33

debt settlement, tapering to about 1 percent of GDP five years hence, and vanishing

only gradually since. Countries with a worse than average credit history would face an

even higher interest bill. In a world where fiscal space is limited and where generating

primary fiscal surpluses to service debt and stabilize debt-to-GDP ratios is costly – both

economically and politically – such an extra fiscal expense due to the default premium

is hardly unimportant.

A key question is why this paper finds more economically significant and more per-

sistent estimates of the interest premium than previous studies. Since our dataset and

our measure of the DP encompass those of previous studies, we are able to identify

the sources of the difference in results. The first and most critical pertains to our use

of a much broader panel, particularly in the time-series dimension. Because sovereign

defaults are major macro/financial events that usually take years to be resolved, and

because the most relevant observations are the spread observations before default and

those in the few years after debt settlement, time-series shorter than a decade or two

can be plainly insufficient. Moreover, long time-series are also important to average out

cyclical and other circumstantial effects. For instance, confining estimation to emerg-

ing market data between the mid-1970s and early 1980s would favor very compressed

spreads and small mean DP due to the prevalence and nature of syndicated banking

lending and the high liquidity in the global inter-bank market in those years. Con-

versely, if one picks up emerging market bond data for countries re-entering the global

capital market in the early to mid-1990s, when spreads were historically very high, and

then interrupts the sample before the 2008-09 global financial crisis, when emerging

market spreads were sharply compressed, one is bound to find larger default premia

upon market re-entry but decaying rapidly due to the very bright emerging market

outlook of 2002-2007. In addition to differences in the time-series dimension, we find

that the size of the cross-section matters too. After all, defaults are infrequent events

that do not affect all countries; because estimation benefits from cross-country hetero-

geneity of credit histories, the size of cross-section can improve precision. In short, by

averaging over much longer periods and countries, a broader panel can greatly miti-

gate the pitfalls of sample selection. This explains much of the differences in our results

vis-a-vis earlier studies, as well as some of the differences in results across earlier studies.

The second source of difference stems from a broader metric of the DP, which in-

34

cludes not only post-settlement year dummies but also longer-term “memory” indica-

tors. While we show that the latter account for only a fraction of the difference in the

size of the DP in the aftermath of debt settlement, those longer-term memory indicators

do bring out the existence of a DP that is substantially more persistent than found in

previous work. As the numerical illustrations of the previous section show, such a per-

sistence can be critical as a repayment incentive. Finally, by subjecting our estimates to

a wider variety of controls (including to distinct functional forms for memory indicators

and to the actual size of haircuts), as well as by extending the estimates to pre-WWII

years with broadly similar results, we see our DP measures likely to be considerably

more robust.

Overall, we see this paper’s findings as a step toward re-connecting theory and

evidence: a positive, sizeable, and persistent default premium that rises on market

exclusion spells, helps rationalize why governments typically try hard not to default

and, when they do, to seek early renegotiation.

35

Tables and Figures

36

Table 1: PRE-WWII Sample(1) (2) (3) (4) (5) (6) (7) (8) (9)

DS(1) 1.31*** 1.30*** 1.31*** 1.45*** 1.39*** 1.05** 1.30*** 1.32*** 1.58***(0.34) (0.39) (0.39) (0.47) (0.41) (0.42) (0.34) (0.39) (0.37)

DS(2) 1.17*** 1.18*** 1.15*** 1.10*** 1.21*** 0.94** 1.17*** 1.19*** 1.37***(0.30) (0.34) (0.34) (0.39) (0.34) (0.38) (0.30) (0.34) (0.35)

DS(3) 0.71** 0.68** 0.65* 0.49 0.69** 0.65 0.71* 0.73** 0.87***(0.34) (0.32) (0.34) (0.34) (0.31) (0.39) (0.35) (0.34) (0.30)

DS(4) 0.56 0.53 0.58 0.45 0.54 0.63 0.56 0.58 0.74**(0.38) (0.36) (0.38) (0.37) (0.36) (0.42) (0.38) (0.38) (0.34)

DS(5) 0.25 0.31 0.27 0.23 0.18 0.33 0.25 0.27 0.45(0.37) (0.35) (0.36) (0.37) (0.35) (0.42) (0.37) (0.37) (0.34)

MEM 1 1824 3.72*** 3.84*** 3.56** 3.41*** 3.11 4.83*** 3.58** 3.77*** 2.99***(1.37) (1.29) (1.36) (1.06) (2.07) (1.23) (1.50) (1.34) (1.09)

MEM 2 1824 -0.01** -0.01* -0.01** -0.01 -0.01* -0.01** -0.01** -0.01** 0.00(0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01)

Debt/GDP 0.78*** 0.86*** 0.75** 0.73*** 0.88*** 0.81** 0.78*** 0.78** 0.70**(0.28) (0.30) (0.32) (0.25) (0.31) (0.38) (0.28) (0.30) (0.28)

% Ext Debt 0.27 0.48 0.41 0.39 0.46 0.44 0.30 0.29 0.51*(0.32) (0.32) (0.30) (0.31) (0.31) (0.30) (0.32) (0.32) (0.29)

Fixed -0.27** -0.25* -0.24* -0.35** -0.28* -0.27* -0.29** -0.29** -0.10(0.13) (0.13) (0.13) (0.15) (0.15) (0.13) (0.14) (0.13) (0.17)

iWorld 0.02** 0.02*** 0.02*** 0.03** 0.03*** 0.03** 0.03** 0.03**(0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01)

RGDP growth -0.02(0.02)

TB/GDP 1.19(1.37)

Reserves/GDP -3.24(3.32)

TOT gap 0.19(0.31)

M2/GDP 0.32(0.68)

New Sovereign -0.23(0.91)

Gov Bal/GDP -0.89(2.35)

R2 0.44 0.44 0.45 0.44 0.45 0.46 0.44 0.45 0.40N 1639 1499 1586 1438 1442 1361 1599 1540 1639All DS(.)=0 0.00 0.00 0.00 0.01 0.00 0.00 0.00 0.00 0.00All mem=0 0.00 0.00 0.00 0.00 0.03 0.00 0.00 0.00 0.03Time FE No No No No No No No No YesCountry FE Yes Yes Yes Yes Yes Yes Yes Yes YesImplied Default Premium in bpsYear 1 251 259 249 235 249 288 246 249 262Year 5 150 162 147 133 115 216 144 153 159

Dep. variable: sovereign spread on long-dated bonds placed in external markets. DP variables for eachsovereign: dummies DS(z)=1 if out of default z years ago, “MEM 1 1824” is % of years in default since1824, “MEM 2 1824” is years since latest default. Constant and dummy for WWI period included in theregressions but not shown. All variables defined in Appendix A.

37

Table 2: POST-WWII Sample(1) (2) (3) (4) (5) (6) (7) (8) (9)

DS(1) 1.82*** 1.82*** 1.81*** 1.90** 1.83*** 1.92** 1.74*** 1.71** 1.64***(0.66) (0.67) (0.66) (0.72) (0.64) (0.73) (0.63) (0.70) (0.59)

DS(2) 0.88* 0.86* 0.85* 0.85* 0.87* 0.86* 0.79* 0.76 0.78*(0.47) (0.50) (0.48) (0.49) (0.45) (0.50) (0.45) (0.51) (0.45)

DS(3) 0.96* 1.03 0.95* 1.04* 0.97* 1.01* 0.89 0.86 0.86*(0.55) (0.63) (0.55) (0.61) (0.55) (0.59) (0.54) (0.57) (0.46)

DS(4) 0.94** 0.93* 0.93** 0.87* 0.95** 0.90* 0.92* 0.78* 0.77**(0.47) (0.52) (0.46) (0.49) (0.46) (0.47) (0.47) (0.46) (0.38)

DS(5) 0.80 0.81 0.80 0.67 0.83 0.78 0.79 0.74 0.55(0.55) (0.57) (0.54) (0.55) (0.54) (0.54) (0.54) (0.56) (0.46)

MEM 1 1950 5.52** 5.67*** 5.49** 5.98*** 5.92** 6.16*** 5.38** 5.61** 3.88*(2.12) (1.99) (2.08) (2.19) (2.24) (2.24) (2.19) (2.20) (2.00)

MEM 2 1950 -0.02** -0.02** -0.02** -0.03** -0.02* -0.02** -0.02** -0.02** -0.04***(0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01)

Debt/GDP 1.41*** 1.40*** 1.42*** 1.28*** 1.35*** 1.38*** 1.36*** 1.34*** 1.43***(0.42) (0.43) (0.42) (0.47) (0.43) (0.46) (0.44) (0.41) (0.49)

RGDP growth -0.14*** -0.14*** -0.14*** -0.16*** -0.15*** -0.15*** -0.15*** -0.14*** -0.14***(0.03) (0.03) (0.03) (0.03) (0.03) (0.03) (0.03) (0.03) (0.03)

Reserves/GDP -2.32** -2.43** -2.54** -2.04* -2.41** -2.23* -2.36** -2.14* -2.47**(1.13) (1.11) (1.17) (1.22) (1.16) (1.20) (1.12) (1.08) (1.10)

Fixed -0.27* -0.35* -0.27 -0.32* -0.25 -0.29* -0.31* -0.21 -0.14(0.16) (0.18) (0.16) (0.17) (0.17) (0.18) (0.17) (0.16) (0.18)

Stock Mkt vol 0.05*** 0.05*** 0.05*** 0.05*** 0.05*** 0.06*** 0.05*** 0.05***(0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01)

CA/GDP -3.76*** -4.11*** -3.59*** -4.67*** -3.44*** -4.44*** -3.69*** -3.87*** -3.06**(1.27) (1.35) (1.28) (1.33) (1.22) (1.37) (1.36) (1.30) (1.23)

% Ext Debt -0.13(0.68)

TOT gap -0.42(0.71)

M2/GDP 0.22(0.42)

Gov Bal -2.80(2.87)

Polity -0.02(0.03)

Real Invest. -0.07(0.32)

Infl. diff. 0.01**(0.00)

loan -0.70*** -0.72*** -0.71*** -0.62** -0.67*** -0.78*** -0.64*** -0.73*** -0.61***(0.24) (0.24) (0.24) (0.24) (0.25) (0.27) (0.23) (0.24) (0.20)

Primary Mkt -0.70*** -0.74*** -0.71*** -0.74*** -0.75*** -0.75*** -0.70*** -0.73*** -0.53**(0.23) (0.23) (0.23) (0.24) (0.22) (0.25) (0.22) (0.23) (0.23)

R2 0.29 0.30 0.29 0.30 0.29 0.31 0.29 0.31 0.38N 1569 1504 1562 1429 1549 1421 1547 1551 1569All DS(.)=0 0.01 0.02 0.01 0.02 0.01 0.01 0.02 0.03 0.01All mem=0 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00Time FE No No No No No No No No YesCountry FE Yes Yes Yes Yes Yes Yes Yes Yes YesImplied Default Premium in bpsYear 1 350 352 348 374 365 387 338 342 262Year 5 204 205 203 193 218 216 199 199 115

Dep. variable: sovereign spread on long maturity bonds placed in external markets. DP variables for each sovereign:dummies DS(z)=1 if out of default z years ago, “MEM 1 1950” is % of years in default since 1950, “MEM 2 1950” isyears since latest default. Constant included in the regressions but not shown. All variables defined in Appendix A.

38

Table 3: PRE-WWII Memory Specification Robustness(Base) (mem3) (exp mean) (exp MA) (exp high) (qhyp) (hyp) (inv)

DS(1) 1.31*** 1.61*** 1.42*** 1.62*** 1.23* 1.07 1.08 1.16**(0.34) (0.48) (0.40) (0.34) (0.71) (0.65) (0.65) (0.57)

DS(2) 1.17*** 1.37*** 1.38*** 1.45*** 1.13* 1.07** 1.08** 1.17***(0.30) (0.40) (0.30) (0.28) (0.60) (0.51) (0.50) (0.41)

DS(3) 0.71** 0.93*** 0.94** 0.97*** 0.71 0.67 0.68 0.76(0.34) (0.32) (0.37) (0.33) (0.50) (0.48) (0.48) (0.45)

DS(4) 0.56 0.73** 0.79* 0.80** 0.59 0.57 0.58 0.65(0.38) (0.36) (0.39) (0.37) (0.47) (0.46) (0.46) (0.45)

DS(5) 0.25 0.39 0.47 0.48 0.31 0.30 0.31 0.36(0.37) (0.38) (0.37) (0.35) (0.46) (0.44) (0.44) (0.42)

MEM 1 3.72*** 7.03*** 4.38*** 4.12** 4.48*** 4.58*** 4.58*** 4.56***(1.37) (2.10) (1.29) (1.53) (1.32) (1.30) (1.30) (1.28)

MEM 2 -0.01** -0.02*** 4.45 -0.05 0.88 3.38 3.75 1.81(0.01) (0.01) (5.31) (0.33) (1.41) (3.45) (3.81) (1.81)

MEM 3 -51.91(40.05)

Debt/GDP 0.78*** 0.80*** 0.74** 0.75** 0.71** 0.70** 0.71** 0.72**(0.28) (0.28) (0.28) (0.29) (0.27) (0.27) (0.27) (0.27)

% Ext Debt 0.27 0.28 0.60* 0.62** 0.57* 0.55* 0.55* 0.57*(0.32) (0.30) (0.30) (0.30) (0.31) (0.32) (0.32) (0.31)

Fixed -0.27** -0.30** -0.17 -0.17 -0.18 -0.18 -0.18 -0.18(0.13) (0.13) (0.14) (0.14) (0.14) (0.14) (0.14) (0.14)

iWorld 0.02** 0.02** 0.02** 0.02 0.02** 0.02** 0.02** 0.02**(0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01)

R2 0.44 0.38 0.38 0.38 0.39 0.40 0.40 0.39N 1639 1639 1639 1633 1639 1639 1639 1639All DS(.)=0 0.00 0.00 0.00 0.00 0.24 0.14 0.12 0.02All mem=0 0.00 0.00 0.00 0.04 0.01 0.00 0.00 0.00Time FE No No No No No No No NoCountry FE Yes Yes Yes Yes Yes Yes Yes YesImplied Default Premium in bpsYear 1 251 187 311 300 321 335 334 327Year 5 150 82 213 200 222 236 236 229

This table compares the regression results obtained with a different specification for the MEM variables.The first column reproduces the baseline results of Table 1, column 1. The second column adds anadditional memory variable (“MEM3”) that measures the number of defaults between 1824 and thecurrent year for each country. The subsequent columns replace the linear functional form of “MEM2”with different discounting alternatives as explained in the text. Constant and dummy for WWI periodincluded in the regressions but not shown.

39

Table 4: POST-WWII Memory Specification Robustness(Base) (mem3) (exp mean) (exp MA) (exp high) (qhyp) (hyp) (inv)

DS(1) 1.82*** 1.90*** 2.20*** 2.09*** 1.71*** 1.74*** 1.77*** 1.96***(0.66) (0.64) (0.72) (0.72) (0.58) (0.62) (0.63) (0.66)

DS(2) 0.88* 0.91* 1.04** 0.99** 0.79* 0.86* 0.87* 0.98**(0.47) (0.47) (0.49) (0.48) (0.41) (0.46) (0.46) (0.47)

DS(3) 0.96* 1.00* 1.12** 1.16* 0.93 0.99* 1.01* 1.09**(0.55) (0.55) (0.56) (0.61) (0.56) (0.55) (0.55) (0.55)

DS(4) 0.94** 0.98** 1.07** 1.05** 0.91* 0.97* 0.98** 1.05**(0.47) (0.47) (0.47) (0.48) (0.51) (0.49) (0.49) (0.47)

DS(5) 0.80 0.85 0.93* 0.92* 0.80 0.85 0.86 0.91*(0.55) (0.53) (0.54) (0.54) (0.53) (0.55) (0.55) (0.55)

MEM 1 1950 5.52** 7.04** 6.79*** 7.21*** 6.72*** 6.76*** 6.78*** 6.86***(2.12) (2.88) (2.10) (2.10) (2.10) (2.08) (2.08) (2.08)

MEM 2 1950 -0.02** -0.02** -3.97 0.27 0.92 2.05 2.07 0.23(0.01) (0.01) (3.95) (0.29) (1.26) (3.12) (3.45) (1.58)

MEM 3 1950 -14.68(18.96)

Debt/GDP 1.41*** 1.48*** 1.34*** 1.35*** 1.28*** 1.29*** 1.29*** 1.31***(0.42) (0.45) (0.43) (0.46) (0.42) (0.43) (0.43) (0.43)

RGDP growth -0.14*** -0.14*** -0.14*** -0.14*** -0.14*** -0.14*** -0.14*** -0.14***(0.03) (0.03) (0.03) (0.03) (0.03) (0.03) (0.03) (0.03)

Reserves/GDP -2.32** -2.13** -2.68** -2.77** -2.75** -2.72** -2.72** -2.71**(1.13) (1.06) (1.11) (1.08) (1.13) (1.13) (1.13) (1.12)

Fixed -0.27* -0.27 -0.22 -0.26 -0.23 -0.23 -0.23 -0.22(0.16) (0.16) (0.16) (0.17) (0.16) (0.16) (0.16) (0.16)

Stock Mkt vol 0.05*** 0.05*** 0.05*** 0.05*** 0.05*** 0.05*** 0.05*** 0.05***(0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01)

CA/GDP -3.76*** -3.92*** -3.85*** -3.98*** -3.85*** -3.85*** -3.85*** -3.84***(1.27) (1.25) (1.28) (1.29) (1.30) (1.30) (1.29) (1.29)

loan -0.70*** -0.73*** -0.65** -0.71*** -0.66*** -0.66*** -0.66*** -0.66**(0.24) (0.25) (0.25) (0.23) (0.25) (0.25) (0.25) (0.25)

Primary Mkt -0.70*** -0.72*** -0.56** -0.50** -0.55** -0.56** -0.55** -0.55**(0.23) (0.23) (0.22) (0.24) (0.22) (0.22) (0.22) (0.22)

R2 0.29 0.27 0.24 0.25 0.25 0.24 0.24 0.24N 1569 1569 1569 1520 1569 1569 1569 1569All DS(.)=0 0.01 0.00 0.00 0.00 0.01 0.01 0.01 0.00All DS(.)=0 0.00 0.00 0.00 0.00 0.01 0.01 0.01 0.01Time FE No No No No No No No NoCountry FE Yes Yes Yes Yes Yes Yes Yes YesImplied Default Premium in bpsYear 1 350 351 428 483 434 439 438 432Year 5 204 196 272 304 278 282 281 275

This table compares the regression results obtained with a different specification for the MEM variables. Thefirst column reproduces the baseline results of Table 2, column 1. The second column adds an additional memoryvariable (“MEM3”) that measures the number of defaults between 1950 and the current year for each country.The subsequent columns replace the linear functional form of “MEM2” with different discounting alternativesas explained in the text. Constant included in the regressions but not shown.

40

Table 5: POST-WWII: Comparison with the Literature(Base) (No Mem) (No Adv) (No prim) (Like BP) (Like BI) (Like CT) (Base+HC)

DS(1) 1.82*** 2.78*** 2.07*** 1.70** 3.85** 0.64*** 0.88 1.57(0.66) (0.65) (0.69) (0.80) (1.79) (0.17) (1.53) (1.99)

DS(2) 0.88* 1.71*** 1.00** 0.29 0.54 0.78*** -3.98*** -1.63*(0.47) (0.47) (0.45) (0.61) (1.23) (0.20) (1.39) (0.83)

DS(3) 0.96* 1.73*** 1.11* 0.50 0.23 0.31 -3.21** -1.31(0.55) (0.65) (0.56) (0.66) (1.27) (0.24) (1.56) (1.30)

DS(4) 0.94** 1.60*** 1.07** 0.76 0.06 0.01 -1.09(0.47) (0.54) (0.50) (0.61) (0.74) (0.11) (1.61)

DS(5) 0.80 1.39** 0.86 0.56 0.23 0.01 -0.69(0.55) (0.55) (0.56) (0.59) (0.44) (0.16) (1.41)

DS(3&4) -2.70**(1.31)

DS(5&6) -3.17***(1.04)

DS(1)x HC 0.04 0.01(0.04) (0.05)

DS(2)x HC 0.11** 0.06*(0.05) (0.03)

DS(3)x HC 0.09** 0.06*(0.03) (0.03)

DS(3&4)x HC 0.08**(0.03)

DS(5&6)x HC 0.08***(0.02)

DS(4)x HC 0.06*(0.03)

DS(5)x HC 0.04(0.03)

MEM 1 1950 5.52** 6.31** 7.34 -0.18 6.90** 7.67*(2.12) (2.89) (4.95) (6.96) (2.70) (3.85)

MEM 2 1950 -0.02** -0.01 -0.04* 0.02 -0.16** -0.02(0.01) (0.01) (0.02) (0.03) (0.08) (0.03)

Debt/GDP 1.41*** 1.16*** 1.42** 3.04*** 1.82* -0.08 2.85*** 3.28***(0.42) (0.39) (0.56) (0.75) (1.06) (0.22) (0.92) (0.75)

RGDP growth -0.14*** -0.13*** -0.13*** -0.09* -0.09 -0.04 -0.16*** -0.22***(0.03) (0.03) (0.03) (0.05) (0.09) (0.02) (0.05) (0.05)

Reserves/GDP -2.32** -2.67** -4.02** -4.35** -10.81*** -0.82 -1.37 -4.07***(1.13) (1.03) (1.62) (2.07) (2.40) (0.87) (1.03) (1.51)

CA/GDP -3.76*** -2.78** -4.91*** -9.72** -9.03 -0.70 -9.04*** -9.89***(1.27) (1.26) (1.70) (4.40) (5.48) (0.46) (3.20) (2.57)

fixed -0.27* -0.32* -0.15 -0.07 -0.03 0.27** 0.07(0.16) (0.18) (0.23) (0.38) (0.57) (0.12) (0.36)

Stock Mkt vol 0.05*** 0.05*** 0.07*** 0.11*** 0.35*** 0.00 0.07***(0.01) (0.01) (0.01) (0.01) (0.05) (0.01) (0.01)

iWorld 0.42*** 0.34** -0.11***(0.08) (0.16) (0.02)

Primary Mkt -0.70*** -0.55*** -1.04*** -0.50*(0.23) (0.20) (0.32) (0.29)

loan -0.70*** -1.01*** -0.79** -0.43***(0.24) (0.32) (0.33) (0.13)

R2 0.29 0.22 0.31 0.26 0.34 0.41 0.27 0.19N 1569 1569 961 460 203 225 425 769All DS(.)=0 0.01 0.00 0.00 0.02 0.20 0.00 0.03 0.20All mem=0 0.00 . 0.01 0.02 0.78 . 0.02 0.05All DS(.)xHC=0 0.05 0.16Time FE No No No No Yes No Yes NoCountry FE Yes Yes Yes Yes Yes Yes Yes YesImplied Default Premium in bpsYear 1 350 278 409 377 388 64 290 444Year 5 204 139 236 196 50 1 22 253

This table modifies the baseline specification in Table 2 col. 1 (“Base”) to ease comparisons with the literature. “No Mem”excludes mem variables; “No Adv” excludes adv markets; “No prim” excludes primary spreads; “Like BP” spans only 1997-2004; “Like BI” takes “No Adv”, excludes mem vars and t ≥ 1990; “Like CT” takes “Base” using secondary spreads fromEMBI only, changes the default premium dummies and spans 1993-2010; “Base+HC” adds settlement dummies interactedwith CT haircuts to the baseline. Constant included in the regressions but not shown. In the second to last column, reservesare measured with respect to M2 rather than to GDP.

41

Table 6: Haircut Estimation

(1) (2)DS -0.15 -7.74

(9.75) (13.35)MEM 1 (1950) 18.77 21.62

(16.15) (20.38)MEM 2 (1950) 1.91*** 2.61***

(0.56) (0.78)Debt/GDP 31.35*** 30.96**

(11.28) (12.53)RGDP growth 2.09** 1.67

(0.98) (1.25)Reserves/GDP -86.06 -94.84

(58.69) (64.12)CA/GDP -81.14 -23.97

(55.96) (74.45)R2 0.88 0.88N 42 30

This table shows cross sectional regressions of hair-cuts on the fundamentals that we use in spread re-gressions. Dep. variable: haircut from Reinhart andTrebesch (2015).

42

Figure 1: Pre-WWII: Spreads and Defaults over time. Solid line is medianspread for sub-sample in basis points including countries in default. Dashed lines areinter-quartile range. Bars (scaled on the vertical axis on the right) represent the ratioof countries in default to the total number of countries in the sample for any given yearin percent. As discussed in the main text, spread observations for default years areexcluded in the econometric analysis.

43

Figure 2: Post-WWII: Spreads and Defaults over time. Solid line is medianspread for sub-sample in basis points including countries in default. Dashed lines areinter-quartile range. Bars (scaled on the vertical axis on the right) represent the ratioof countries in default to the total number of countries in the sample for any given yearin percent. As discussed in the main text, spread observations for default years areexcluded in the econometric analysis.

44

Figure 3: Average Default Premium for (1) of Table 1 (Left Panel) and Table 2(Right Panel). Solid line is average default premium for sub-sample. Dashed lines arebands for 1.96 standard errors of the mean default premium.

Figure 4: Median Default Premium estimates across Specifications. Note:Solid line is median default premium across all specifications in Tables 1 and 2. Dashedlines are percentiles 75th and 25th.

45

(a) Market Spreads (b) Default Premium

Figure 5: Pre-WWII: Median (solid) and 25 and 75 percentiles (dashed) marketspread and default premia in the run-up to default and after default.

(a) Market Spreads (b) Default Premium

Figure 6: Post-WWII: Median (solid) and 25 and 75 percentiles (dashed) marketspread and default premia in the run-up to default and after default.

46

0.00

1.00

2.00

3.00

4.00

5.00

6.00

7.00

8.00

Away from Default Default 1 Year After 2 Years After 3 Years After 4 Years After 5 Years After

SpreadDefault PremiumFundamentals

Figure 7: Decomposing the Mean Market Spread in the Pre-WWII sample.Cross-country mean fitted values of Fundamentals, Fβ, and DP for (1) of Table 1.

0.00

1.00

2.00

3.00

4.00

5.00

6.00

7.00

8.00

Away from Default Default 1 Year After 2 Years After 3 Years After 4 Years After 5 Years After

SpreadDefault PremiumFundamentals

Figure 8: Decomposing the Mean Market Spread in the Post-WWII sample.Cross-country mean fitted values of Fundamentals, Fβ, and DP for (1) of Table 2.

47

0.00

1.00

2.00

3.00

4.00

5.00

6.00

7.00

8.00

Away from Default Default 1 Year After 2 Years After 3 Years After 4 Years After 5 Years After

SpreadDefault PremiumFundamentals

Figure 9: Decomposing the Mean Market Spread in the Pre-WWII sample,Serial Defaulters. Same as Figure 7 focusing on countries with more than one defaultsince 1824.

0.00

1.00

2.00

3.00

4.00

5.00

6.00

7.00

8.00

Away from Default Default 1 Year After 2 Years After 3 Years After 4 Years After 5 Years After

SpreadDefault PremiumFundamentals

Figure 10: Decomposing the Mean Market Spread in the Pre-WWII sample,Non-Serial Defaulters. Same as Figure 7 focusing on countries defaulting for thefirst time since 1824.

48

0.00

1.00

2.00

3.00

4.00

5.00

6.00

7.00

8.00

Away from Default Default 1 Year After 2 Years After 3 Years After 4 Years After 5 Years After

SpreadDefault PremiumFundamentals

Figure 11: Decomposing the Mean Market Spread in the Post-WWII sample,Serial Defaulters. Same as Figure 8 focusing on countries with more than one defaultsince 1950.

0.00

1.00

2.00

3.00

4.00

5.00

6.00

7.00

8.00

Away from Default Default 1 Year After 2 Years After 3 Years After 4 Years After 5 Years After

SpreadDefault PremiumFundamentals

Figure 12: Decomposing the Mean Market Spread in the Post-WWII sample,Non-Serial Defaulters. Same as Figure 8 focusing on countries defaulting for thefirst time since 1950.

49

h=0%

h=10%

h=20%

h=30%

h=40%

h=50%

Estimated DP

0

100

200

300

400

500

600

700

800

.8 .85 .9 .95 1β

Figure 13: Default premium necessary for sovereign to be indifferent betweenrepaying and defaulting in basis points. Using equation (7), J = 5, δ = 0.125,r = 3.5%, g = 3.5% and d/dt−1 = 1. Estimated DP (s0 = 350) is calibrated to firstyear DP in the baseline specification of the post-WWII sample, column (1) of Table 2.

50

A Appendix I (online material, not to be published)

This appendix reports the data sources and construction methods for the variables used

in the empirical analysis presented in the main text.

A.1 Spreads on Foreign Currency Debt

In the pre-WWII period we used secondary market yields data alone. The post-World

War series on spreads combine primary market yields (yields at issuance), mostly for

the 1970s and 1980s, and yields on secondary markets – largely from 1991. The reason

to resort to primary market yield data in the post-WWII period is due to the much

greater resort by sovereigns to syndicated bank loans and the eurodollar bond market,

for which yield data on secondary trading is hard to obtain. Further specifics are

provided next.

A.1.1 Pre-WWII

Our main data sources for sovereign bond yields up to World War II is the Global

Financial Data (GFD), downloaded in January 2016. All spreads are computed against

the British consol yield listed in GFD. Over time, GFD has improved the quality of

historical data on government bond yields, addressing of the some of the criticisms to

older vintages of the database (see,e.g., Obstfeld and Taylor, 2003 and Mauro et al.,

2006). However, we found several inconsistencies between GFD data and other sources

and a few data entries that appear to be mistaken. Thus, we complement GFD data and

replace those seemingly mistaken entries based on information available in the sources

specified next.

First, we replace GFD for Argentina, Canada, Chile, China, Greece, Japan, Mexico,

Portugal, Russia and Sweden before 1917, by those provided in Mauro et al. (2006),

where the methodology and ultimate sources are clearly specified and data is also

available at shorter frequencies if needed. In the majority of cases for these countries,

differences between the two sources were small.

Regarding inter-war bond yields for European countries, we adjusted GFD’s yield

for Bulgaria in 1933 for a reduction of coupon that GFD didn’t account for based

on information from the League of Nations yearbooks. We also used data from the

League of Nations Yearbooks for Germany, Bulgaria, Hungary, and Poland (after 1930

51

and The Economist for 1928-29). These correspond to bonds traded in the London

market so making the British consol bond the obvious risk-free reference rate counter-

part (as discussed in Obstfeld and Taylor, 2003). For Turkey, GFD data after 1919

implies an unrealistically highly negative sovereign spread, so we opted instead to use

the value provided in Obstfeld and Taylor (2003) for that year. In the case of Romania,

none of these sources provided yield data, so our spread series is the yield differential

between Romania’s 4% consol and the French 5% consol reported in Oosterlinck and

Ureche-Rangau (2012) and kindly supplied by the authors.

Regarding Latin American countries, the following adjustments were made. First,

we excluded GFD bond yields for Peru (1890-1910) and Honduras (before 1930) in

the absence of historical information corroborating their excessively elevated levels (in

excess of 3000 basis points) reported in the GFD over those years. We filled gaps

for Colombia yields in 1927-32 (except 1929) based on Marichal (1989). In the case of

Venezuela in 1915-1929, the GFD assumes the underlying bond was a perpetuity where

in fact it was not. We recompute the yield adjusting for this effect. For Uruguay, we

compute the yield by hand picking data from the Economist for 1879-83. GFD and

Mauro et al (2006) do not account for a change in the coupon rate in the settlement of

1878, which we do based on that source.

A.1.2 Post-WWII

The main source for emerging market spreads in the post-war sample was JP Morgan’s

EMBI spreads, from MorganMarkets, Datastream and Global Financial Data. We

excluded Morocco data after 2007 and also annual observations for which the respective

EMBI index was based on less than a month of daily observations.

For Euro area countries, the relevant spread was obtained by subtracting the re-

spective country yield on its 10-year

bond from that of the German 10-year Bund yield. In other cases, as specified

below, we used the US 10-year yield as a benchmark rate.

For Argentina in 1983-1993 we used the bond yield from the database underlying

Arellano (2008) – as kindly supplied to the authors – subtracted from US 10-year

bond from IFS. The 1982 yield was based on that reported in Folkerts-Landau (1985,

Table 7) on German Mark denominated bonds, having spread calculated against the

corresponding maturity German Bund yield. Also from Folkerts-Landau (1985, Table

52

8) comes the spread on Brazil in 1982-3 calculated against a 5 year maturity US bond.

For Peru in 1991-96, we applied the correlation between the EMBI Latin American

index and Peru after 1997 to back-out an estimate of the EMBI spread for Peru. The

same procedure was followed to derived the annual spread figures for Uruguay during

1991-93. For Venezuela 1982-3, we used Folkerts-Landau’s (1985, Table 7) German

Mark denominated bond yields, having spread calculated against the corresponding

matured German Bund yield.

Thailand 1996 is from IMF (2007) and Russia 1996 from GFD, which is consistent

with the Russian EMBI spread series for subsequent years.

Data from primary market issuance was sourced from the World Bank’s (WB) pub-

lication “Borrowing in International Capital Markets”, 1973 to 1981 1st half and Dia-

logic Bondware and Loanware. We excluded from the WB dataset Pakistan 1981 which

was an extreme outlier, and from Dialogic data Egypt in 1997, Tunisia, Pakistan and

Lithuania all together. Spreads were computed for each bond separately, depending on

currency and maturity and then averaged weighting by amount issued and maturity.

Currencies of denomination used were US dollar, Japanese yen, german mark, sterling,

guilden, australian dollar, canadian dollar and swiss franc in the case of the WB data

and additionally euro in the case of Dialogic.

53

05

1015

20S

econ

dary

0 2 4 6 8 10Primary

Fitted Line 45o Line

Figure 14: Spreads in the Primary and Secondary Market

A.2 Definition of Credit Events, and Default and Settlement

dates

Our chief source of information on sovereign credit events throughout 1870-2011 is

the set of Standard and Poor’s rating agency reports, as compiled by Borensztein and

Panizza (2008) for the pre-2007 years and updated in http://www.standardandpoors.com/

ratings/articles/en/us/?articleType=HTML&assetID=1245350156739. As discussed in

Section 3, partial re-scheduling and debt negotiation attempts following the original

default decision which, nevertheless, do not result in a complete or near complete set-

tlement of arrears, are not classified as new default or a final settlement of the debt. In

short, as in Sutter (1992), Beim and Calomiris (2001), and Reinhart and Rogoff (2009),

we take the years between the initial default and full (or near full) settlement of ar-

rears as per the S&P definition as the length of the default. Unlike Lindert and Morton

(1987) but in line with much of the subsequent literature, we do not distinguish between

“voluntary” and “involuntary” defaults or “market-friendly” vs. “market-unfriendly”

54

reschedulings, focusing on the common denominator of all such events – namely, a

breach in the original contractual terms of the debt agreement.

As also discussed in Section 3, we complement the definition of sovereign credit

events centered on defaults on sovereign bonds with that based on large resort to mul-

tilateral financing. As in Catao and Milesi-Ferretti (2014), this is defined as events that

imply resort to 200% or more the respective country’s quota at the IMF. We compute

the share of IMF’s credit to country members from the IMF’s International Financial

Statistics (IFS) database which is publicly available online.

Tables A1 and A2 report the number of observations and the share of secondary

market spreads per country.

55

Table A1: Pre-WWII sample: 1871-1938

Country Years in default # obs. in % sec. mkt. % bond(automatically excluded) sample spreads spreads

Argentina 1890-1893 53 100 100Australia n.a. 57 100 100Austria 1868-1870; 1915-1916; 1932-1933; 1938 54 100 100Belgium n.a. 49 100 100Brazil 1898-1901; 1914-1919; 1931-1933; 1937-1943 53 100 100Bulgaria 1916-1925*; 1932 19 100 100Canada n.a. 58 100 100Chile 1880-1883; 1931-1947 56 100 100Colombia 1900-1904; 1932-1944 19 100 100Costa Rica 1874-1885; 1895-1897; 1901-1911; 1932-1952 2 100 100Czechoslovakia 1938-1946 13 100 100Denmark n.a. 50 100 100Egypt 1876-1880 63 100 100El Salvador 1921-1922; 1932-1935; 1938-1946 17 100 100Finland n.a. 23 100 100France n.a. 41 100 100Germany 1932-1949* 40 100 100Greece 1894-1897; 1932-1964 38 100 100Hungary 1932-1937 2 100 100India n.a. 49 100 100Ireland n.a. 14 100 100Italy n.a. 68 100 100Japan n.a. 54 100 100Mexico 1866-1885; 1914-1922; 1928-1942 33 100 100Netherlands n.a. 68 100 100New Zealand n.a. 54 100 100Norway n.a. 49 100 100Peru 1931-1951 12 100 100Poland 1936-1937 11 100 100Portugal 1892-1901 58 100 100Romania 1933-1958 6 100 100Russia 1918-1995 29 100 100South Africa n.a. 68 100 100Spain 1837-1867; 1873*-1882 56 100 100Sweden n.a. 59 100 100Switzerland n.a. 40 100 100Turkey 1876-1881; 1915-1928 38 100 100United States n.a. 59 100 100Uruguay 1876-1878; 1891; 1915-1921; 1933-1938 51 100 100Venezuela 1865-1881; 1892-1895; 1898-1904 33 100 100Yugoslavia/Serbia 1895-1896; 1932*-1950 23 100 100Total 1639 100 100

Note: * departure from S&P classification.

56

Table A2: Post-WWII sample: 1973-2014

Country Years in default # obs. in % sec. mkt. % bond(automatically excluded) sample spreads spreads

Argentina 1982-1992*; 2001-2014* 17 53 53Australia n.a. 26 0 100Austria n.a. 33 45 100Belgium n.a. 31 52 100Brazil 1983-1992* 32 88 88Bulgaria 1990-1994 20 100 100Canada n.a. 34 0 100Chile 1972*-1975*; 1983-1990 27 59 63China n.a. 31 68 71Colombia n.a. 37 49 62Costa Rica 1981-1990 20 15 60Croatia 1992-1996 18 100 100Cyprus 2013 21 38 62Czech Rep. n.a. 16 31 88Denmark n.a. 33 0 100Dominican Rep. 1982-1994; 2003*-2004* 16 75 75Ecuador 1983*-1995; 1999-2000; 2008-2014 20 65 65Egypt 1984*-1992* 19 74 74El Salvador n.a. 18 72 83Estonia n.a. 5 0 80Finland n.a. 35 46 94France n.a. 35 46 100Gabon 1986-1994; 1999-2004 9 89 89Greece 2012-2013 38 42 76Guatemala n.a. 9 33 78Hungary n.a. 32 50 94Iceland n.a. 20 0 70India n.a. 31 10 13Indonesia 1998-2002 33 42 45Israel n.a. 21 0 90Italy n.a. 31 52 100Jamaica 1978-1979; 1981-1985; 1987-1993; 2010-2011; 2013 18 28 72Jordan 1989-1993 13 31 46Korea n.a. 40 28 50Latvia n.a. 11 27 73Lebanon n.a. 22 77 95Lithuania n.a. 6 100 100Malaysia n.a. 41 46 63Malta n.a. 12 75 83Mexico 1982-1990 33 88 88...

Note: * departure from S&P classification.

57

Table A2 (continuation): Post-WWII sample: 1973-2014

Country Years in default # obs. in % sec. mkt. % bond(automatically excluded) sample spreads spreads

...Morocco 1983; 1986-1990 27 59 70Netherlands n.a. 23 70 100New Zealand n.a. 19 0 100Norway n.a. 27 0 100Pakistan 1998-2000 17 82 82Panama 1983-1996 27 67 67Peru 1976; 1978-1992* 25 88 88Philippines 1983-1992 32 69 69Poland 1981-1994 20 100 100Portugal n.a. 37 43 76Romania 1982*-1994* 18 78 94Russia 1917*-1995*; 1998-2000 16 100 100Slovak Rep. n.a. 18 89 94Slovenia 1992-1996 17 59 100South Africa 1985-1993 32 63 81Spain n.a. 41 39 71Sweden n.a. 34 0 100Thailand n.a. 33 33 64Tunisia n.a. 23 43 83Turkey 1978-1982 33 58 85Ukraine 1998-2000 14 100 100Uruguay 1983-1991*; 2003 30 73 73Venezuela 1983-1997 32 75 75Yugoslavia/Serbia 1983-2004 10 100 100total 1569 50 80

Note: * departure from S&P classification.

58

A.3 Construction of Credit History Indicators

Illustration Case 1: Single Default

Suppose that the relevant sovereign history starts in 1850 for country i.35 In 1898

the country defaults and it settles its debt with creditors in 1902. There is no default

thereafter. First, MEM1i,t = 0 for all t < 1898, MEM1i,t = t−1898t−1850

for 1898 ≤t ≤ 1902, and MEM1i,t = 1902−1898

t−1850for all t > 1902. Second, MEM2i,t = 0 for all

t<1898 and MEM2i,t = t−1898 for all t > 1898.Let’s now turn to the debt settlement

dummies, DS. Consider a window of m = 2 . The two DS(j)i,t dummies are defined as:

DS(1)i,t = 1 for t = 1903 and DS(1)i,t = 0 for t 6= 1903, and DS(2)i,t = 1or t = 1904

and DS(2)i,t = 0 for t 6= 1904.

Illustration Case 2: Serial default

Consider the same information as in Case 1 but now suppose that the country

defaults again 1928 and re-enters the markets after settlement in 1935. So in this

example country i defaults twice. In this case the variables are defined as follows.

First, MEM1i,t = 0 for all t < 1898, MEM1i,t = t−1898t−1850

for 1898 ≤ t ≤ 1902, and

MEM1i,t = 1902−1898t−1850

for 1902 < t ≤ 1928; then MEM1i,t = 1902−1898+t−1928t−1850

for all

1928 < t ≤ 1935 and MEM1i,t = 1902−1898+1935−1928t−1850

for all t > 1935.

Second, MEM2i,t = 0 for all t < 1898 and MEM2i,t = t−1898 for 1898 ≤ t ≤ 1902;

then, MEM2i,t = t− 1928 for t > 1928. Thus, serial default implies that MEM2i,t is

reset: upon the year country i defaults for the second time, MEM2i,t is reset to zero

(in 1928 in this example) and keeps growing from t > 1928 as per MEM2i,t = t−1928.

Finally, turning to how the debt settlement dummies, DS, are defined in this ex-

ample, they become: DS(1)i,t = 1 for t = 1903, 1936 and DS(1)i,t = 0, otherwise.

DS(2)i,t = 1 for t = 1904, 1937 and DS(2)i,t = 0 otherwise.

35We thank Guido Sandlieris for suggestions on these illustrations.

59

A.4 GDP data

A.4.1 1870-39

Nominal and Real GDP data are Mitchell (2013) and Maddison (2003), complemented

with the dataset from Barro and Urzua (2008), with exception of the following cases:

Argentina, Brazil, Chile and Mexico: from Aiolfi, Catao, and Timmermann (2011),

available at: http://rady.ucsd.edu/faculty/directory/timmermann/LAC-4 database Aiolfi-

Catao-Timmermann 1870-2004. The authors present a variety of robustness tests to

show that their estimates are superior to those provided by Maddison.

Greece: new estimates kindly provided by George Kostelenos, based on his earlier

research (Money and Output in Modern Greece, 1858-1938, Athens, 1995).

Russia: the net national product estimate from Paul Gregory, Russian National

Income, 1885-1913 (Cambridge: Cambridge University Press, 1983), Table 3.1, pp.

56-7, (”variant 1”).

Spain: Prados de la Escosura, Leandro 2003, El Progreso Economico de Espana,

1850-2000 (Madrid: Fundacion BBVA), Table A. 9.1 and A.13.5, pp. 517-22 and 681-82.

Venezuela: Baptista, Asdrœbal, 1997, Bases Cuantitativas de la Economia Vene-

zolana, 1830-1995, Caracas.

Nominal GDP from Obstfeld and Taylor (2003) except for the above countries (which

are provided in the respective sources), as well as for New Zealand (which is from

Rankin, Keith, 1992, “New Zealand’s Gross National Product”, Review of Income and

Wealth, 38(1), pp.49-6, Table 4, p.60/61), and for Hungary and Yugoslavia which are

taken from Mitchell, Brian, 2003, International Historical Statistics: Europe, New York.

A.4.2 1970-2014

IMF’s International Financial Statistics database.

A.5 Total Government Debt

A.5.1 1870-39

Our chief source was Obstfeld and Taylor (2003), except for:

Argentina, Brazil, Chile and Mexico: from Aiolfi, Catao, and Timmermann (2011),

see above;

60

Bulgaria: Dimitrova and Ivanov (2014);

Greece: Lazaretou, Sophia, 1993, “Monetary and Fiscal Policies in Greece: 1833-

1914”, Journal of European Economic History, vol.22, no.2, as kindly supplied by the

author.

Peru: Kelly (1998) and League of Nations, op. cit., several issues;

Romania: Stoenescu et al. (2014);

Serbia/Yugoslavia: Hinic, Durdevic, and Sojic (2014);

Venezuela: Baptista, Asdrœbal, 1997, Bases Cuantitativas de la Economia Vene-

zolana, 1830-1995, Caracas; Kelly, Trish,1998, Ability and Willingness to Pay in the

Age of Pax Britannica, Explorations in Economic History 35, pp.31-58, as kindly pro-

vided by the author; and League of Nations, op. cit., several issues.

For the remaining countries, the sources were the League of Nations Yearbooks,

several issues, and Reinhart and Rogoff (2009).

A.5.2 1970-2014

IMF’s World Economic Outlook and World Bank’s World Development Finance databases.

In cases of conflict or missing information in either databases, we used Abbas et al

(2010), “A Historical Public Debt database”, IMF working paper 10/245.

A.6 External Government Debt

A.6.1 1870-39

The chief sources were Flandreau, M. and F. Zumer, 2004, The Making of Global

Finance, 1880-1913, Paris, Table DB3, for the period 1880-1913 and the League of

Nations Yearbooks for 1919-1939. In addition we used the following country-specific

sources:

Argentina: della Paolera and Taylor (2003), Straining the Anchor, Chicago;

Austria: From Jobst and Scheiber (2014), with pre-1880 separately and kindly

communicated by Clemens Jobst;

Brazil: IBGE, 1987, Estatisticas Historicas do Brasil, Rio de Janeiro;

Chile: Braun et al., 1997, Economia Chilena: Estatisticas Historicas, 1810-1995,

Santiago;

61

Colombia, Mexico, India, and Venezuela: Kelly Trish, op cit., kindly provided by

the author.

Egypt and Turkey: Tuncer and Pamuk (2014), kindly provided by Ali C. Tuncer;

Greece: Lazaretou, Sophia, 1993, “Monetary and Fiscal Policies in Greece: 1833-

1914”, Journal of European Economic History, vol.22, no.2, as kindly supplied by the

author;

Uruguay: Nahum, B., 2007, Estatisticas Historicas de Uruguay, 1900-1950, Vol III,

Montevideo

A.6.2 1970-2014

World Bank’s Global Development Finance database. For advanced countries, it was

assumed that externally issued debt was zero except for countries/periods under which

some of those countries were under IMF program, in which case, debt to the IMF was

deemed to represented all external debt.

A.7 Fixed Exchange Rate Regime (Fixed)

A.7.1 1870-1939

Defined as dummy variable, taking the value of 1 when the country was on the gold

standard. The information source for gold standard membership is Catao and Solo-

mou (2005) for the sub-period 1870-1913 and Obstfeld and Taylor (2003) and Officer,

Lawrence, “Gold Standard”, available at: http://eh.net/encyclopedia/gold-standard/.

A.7.2 1970-2014

The fixed exchange rate dummy was constructed based on the IMF classification (cat-

egories “1” and “2”), as compiled in Reinhart,C., K. Rogoff and Ethan O. Ilzet-

zki, 2009 “Exchange Rate Arrangements Entering the 21st Century: Which Anchor

Will Hold?” (available at: http://www.carmenreinhart.com/research/publications-by-

topic/exchange-rates-and-dollarization/, to 2007 and updated by the authors, based on

IMF information, for the remaining years.

62

A.8 Nominal interest rate in the world (iWORLDt−1 )

A.8.1 1870-1939

GDP weighted average of the (short-term) money market interest in the UK and the

US (the latter from 1920) deflated by CPI inflation. The source for the nominal interest

is Holmer, Sidney and Richard Silla, “A History of Interest Rates,” New Jersey, 1996.

The source for CPI inflation data in 1870-1913 is Catao and Solomou (2005). For post-

1914 data is Obstfeld and Taylor (2003) and the Global Financial Data database. GDP

data is from Maddison (1995).

A.8.2 1970-2014

Computed the same was as for 1870-1939, but including all G-7 economies plus Aus-

tralia. Money market rates, CPI and inflation and US dollar GDP data for all countries

are from the IMF’s International Financial Statistics.

A.9 Stock Market Volatility

A.9.1 1870-1939

Computed as the standard deviation of the month to month change over the 12-month

annual window for the London stock market index from Global Financial Data for

the period 1870-1918. For 1919-1939, the US stock market index provided in Global

Financial database was used.

A.9.2 1970-2014

Stock market volatility is measured by the VIX/VOX index compiled in Philips, Steve

and al. “External Balance Assessment Methodologies”, IMF working paper 13/296

from 1986. For the pre-1986 the index was spliced based on the S & P 500 index actual

volatility, calculated standard deviations of monthly changes as for 1870-1939.

A.10 Exports and Imports

A.10.1 1870-1939

63

Mitchell, B.R. International Historical Statistics, all four volumes, from different edi-

tions.

A.10.2 1970-2014

IMF’s International Financial Statistics.

A.11 Terms of Trade

A.11.1 1870-1939

For most countries, the source was Blattman, C. J. Hwang and J. Williamson (2007),

kindly supplied by the authors, and Catao and Solomou (2005), online appendix,

available at https://www.aeaweb.org/aer/data/sept05 app catao.pdf. For some Latin

American countries that are not covered by those studies, data was taken from the Base

de Datos de Historia Economica de America Latina Montevideo-Oxford, available at

http://moxlad.fcs.edu.uy/es/basededatos.html.

A.11.2 1970-2014

IMF’s International Financial Statistics and World Economic Outlook databases, vari-

ous vintages, as compiled in Catao and Milesi-Ferretti (2014).

A.12 Foreign Exchange Reserves

A.12.1 1870-1939

Data for advanced countries was mostly from Flandreau and Zumer (2004), comple-

mented with data from della Paolera and Taylor (2004), Dimitrova and Ivanov (2014),

Jobst and Scheiber (2014), Lazaretou (2014), Mata and Valerio (1993), Prados (1988),

Tuncer and Pamuk (2014), Nahum (2009), and the database underlying Aiolfi et al.

(2011), cited above.

A.12.2 1970-2014

IMF’s International Financial Statistics.

64

A.13 Broad Money (M2)

A.13.1 1870-1939

Mitchell (2013) and Flandreau and Zumer (2004), complemented with data from della

Paolera and Taylor (2004), Dimitrova and Ivanov (2014), Jobst and Scheiber (2014),

Lazaretou (2014), Mata and Valerio (1993), Prados (1988), Tuncer and Pamuk (2014),

Roman and Willebald (2011), and the database underlying Aiolfi et al. (2011).

A.13.2 1970-2014

IMF’s International Financial Statistics, World Economic Outlook, and the World

Bank’s Global Development Finance databases, as revised and combined by Catao

and Milesi-Ferretti (2014).

A.14 Government Expenditures and Revenues

A.14.1 1870-1939

Mitchell (2013) and Flandreau and Zumer (2004), complemented with data from della

Paolera and Taylor (2004), Dimitrova and Ivanov (2014), Jobst and Scheiber (2014),

Lazaretou (2014), Mata and Valerio (1993), Prados (1988), Tuncer and Pamuk (2014),

Nahum (2009), and the database underlying Aiolfi et al. (2011).

A.14.2 1970-2014

IMF’s International Financial Statistics, World Economic Outlook, and the World

Bank’s Global Development Finance databases, as revised and combined by Catao

and Milesi-Ferretti (2014).

A.15 Polity index (Polity)

Available at: http://www.icpsr.umich.edu/icpsrweb/ICPSR/studies/6695 and

http://www.systemicpeace.org/polity/polity4.htm.

65

A.16 haircuts on Foreign Debt

Cruces and Trebesch (2013), updated to 2014 as available from

https://sites.google.com/site/christophtrebesch/, and Reinhart and Trebesch (2014).

B Appendix II (online material, not to be pub-

lished)

This appendix presents the calculations in section 5 in detail.

B.1 Equations in levels

Assuming default lasts J > 1 periods, the value function under repayment is:

V rt+1(Dt, Yt, 0) = Yt+1 − (1 + r)Dt +Dt+1 + βV r

t+2(Dt+1, Yt+1, 0)

and the value function under (multi-period) default:

V dt+1(Dt−1, Yt, 0) = Yt+1 + βV d

t+2(Dt−1, Yt+1, 0)

Iterating forward until default settlement at which point arrears are paid and new

debt is issued:

V rt+1(Dt, Yt, 0) =

J−1∑j=0

βjYt+1+j−rJ−1∑j=0

βjDt+j +J−1∑j=0

βj∆Dt+1+j +βJV rt+J+1(Dt+J , Yt+J , 0)

V dt+1(Dt−1, Yt, 0) =

J−1∑j=0

βjYt+1+j−βJ−1 (1 + r)J (1−h)Dt−1+βJ−1Dt+J+βJV rt+J+1(Dt+J , Yt+J , s0)

After default ends, i.e at J+1, the value function is:

V rt+J+1(Dt+J , Yt+J , 0) = Yt+J+1 − (1 + r)Dt+J +Dt+J+1 + βV r

t+J+2(Dt+J+1, Yt+J+1, 0)

V rt+J+1(Dt+J , Yt+J , s0) = Yt+J+1 − (1 + r + s0)Dt+J +Dt+J+1 + βV r

t+J+2(Dt+J+1, Yt+J+1, s0 (1− δ))

where s0 (1− δ) stands for the evolution of the default premium.

66

B.2 Equations expressed as a ratio to GDP

Note that the level of output is no longer a state variable, as we assume constant

exogenous GDP growth. To get to equation (6), we first divide equation (5) by Yt+1:

Dt

Yt+1

+ β

{V rt+1(Dt, 0)

Yt+1

−V dt+1(Dt−1, 0)

Yt+1

}≥ Dt−1

Yt+1

(1 + r)

Denoting variables in percent of GDP with small caps, e.g. V rt+1 (.) /Yt+1 = vrt+1 (.):

Dt

Yt+1

+ β{vrt+1 (dt, 0)− vdt+1(dt−1, 0)

}≥ Dt−1

Yt+1

(1 + r)

Multiplying and dividing by the same period’s output:

Dt

Yt

YtYt+1

+ β{vrt+1 (dt, 0)− vdt+1(dt−1, 0)

}≥ Dt−1

Yt−1

Yt−1

Yt+1

(1 + r)

using the fact that output grows at exogenous rate gy:

dt1 + gy

+ β{vrt+1 (dt, 0)− vdt+1(dt−1, 0)

}≥ dt−1

(1 + gy)2 (1 + r)

Re-arranging terms yields equation (6).

Taking the value function under repayment up to period J+1 and dividing through

by Yt+1:

vrt+1(dt, 0) =J−1∑j=0

βjYt+1+j

Yt+1

−rJ−1∑j=0

βjDt+j

Yt+1

+J−1∑j=0

βj∆Dt+1+j

Yt+1

+βJV rt+J+1(Dt+J , Yt+J , 0)

Yt+J+1

Yt+J+1

Yt+1

67

Solving for some terms in the RHS of the above equation:

J−1∑j=0

βjYt+1+j

Yt+1

=J−1∑j=0

βj (1 + gy)j

rJ−1∑j=0

βjDt+j

Yt+1

= r

J−1∑j=0

βjYt+1+j

Yt+1

Yt+jYt+1+j

Dt+j

Yt+j=

r

1 + gy

J−1∑j=0

βj (1 + gy)j dt+j

J−1∑j=0

βj∆Dt+1+j

Yt+1

=J−1∑j=0

βjYt+jYt+1

Dt+1+j −Dt+j

Yt+j=

J−1∑j=0

βjYt+jYt+1

[Yt+j+1

Yt+j

Dt+1+j

Yt+j+1

− Dt+j

Yt+j

]=

=J−1∑j=0

βjYt+1+j

Yt+1

Yt+jYt+1+j

[(1 + gy) dt+1+j − dt+j] =

=1

1 + gy

J−1∑j=0

βj (1 + gy)j [(1 + gy) dt+1+j − dt+j]

replacing these three terms in the equation above, and simplifying:

vrt+1(dt, 0) =

∑J−1

j=0 βj (1 + gy)j − rf

1+gy

∑J−1j=0 β

j (1 + gy)j dt+j+1

1+gy

∑J−1j=0 β

j (1 + gy)j [(1 + gy) dt+1+j − dt+j] +

βJ (1 + gy)J vrt+J+1(dt+J , 0)

Assume further that ∀k≥0dt+k = d, the above reduces to:

vrt+1(d, 0) =

∑J−1

j=0 βj (1 + gy)j − rf

1+gy

∑J−1j=0 β

j (1 + gy)j d+1

1+gy

∑J−1j=0 β

j (1 + gy)j[(1 + gy) d− d

]+

βJ (1 + gy)J vrt+J+1(d, 0)

Collecting terms and simplifying leads to:

vrt+1(d, 0) =

[1 +

gy − r1 + gy

d

] J−1∑j=0

βj (1 + gy)j + βJ (1 + gy)J vrt+J+1(d, 0) (8)

Taking the value function under default up to period J + 1 and dividing through by

68

Yt+1:

vdt+1(dt−1, 0) =

∑J−1

j=0 βj Yt+1+j

Yt+1−

βJ−1(1 + rf

)J(1− h)Dt−1

Yt+1+ βJ−1Dt+J

Yt+1+

βJ (1 + gy)J vrt+J+1(dt+J , s0)

simplifying and using an expression from above for the first term on the RHS:

vdt+1(dt−1, 0) =

∑J−1

j=0 βj (1 + gy)j −

βJ−1(1 + rf

)J(1− h)Dt−1

Yt−1

Yt−1

Yt+1+ βJ−1Dt+J

Yt+J

Yt+J

Yt+1+

βJ (1 + gy)J vrt+J+1(dt+J , s0)

Collecting terms leads to:

vdt+1(dt−1, 0) =

∑J−1

j=0 βj (1 + gy)j −

βJ−1 (1+rf)J

(1+gy)2(1− h)dt−1 + (β (1 + gy))J−1 dt+J+

βJ (1 + gy)J vrt+J+1(dt+J , s0)

Finally, assuming the debt ratio is constant after t = 0, leads to:

vdt+1(d, 0) =

∑J−1

j=0 βj (1 + gy)j −

βJ−1 (1+rf)J

(1+gy)2(1− h)d+ (β (1 + gy))J−1 d+

βJ (1 + gy)J vrt+J+1(d, s0)

(9)

Now we need to replace the continuation values at time t+J + 1. For what follows,

the difference between the two is all what matters. Under no default premium and the

condition that β (1 + gy) < 1, we have that:

vrt+J+1(d, 0) =

(1 +

gy − r1 + gy

d

)1

1− β (1 + gy)

Under a positive default premium, that takes the initial value of s0 and decays

69

geometrically at rate (1− δ), where 0 < δ < 1, we have:

vrt+J+1(d, s0) =

(1 +

gy − r − s0

1 + gyd

)+ β (1 + gy)

(1 +

gy − r − s0 (1− δ)1 + gy

d

)+ ...

=

(1 +

gy − r1 + gy

d

)+−s0

1 + gyd+ β (1 + gy)

(1 +

gy − r1 + gy

d

)+ β (1 + gy)

−s0 (1− δ)1 + gy

d+ ...

= vrt+J+1(d, 0)− s0/ (1 + gy)

1− β (1 + gy) (1− δ)d

and thus the difference between value functions is:

vrt+J+1(d, 0)− vrt+J+1(d, s0) =s0/ (1 + gy)

1− β (1 + gy) (1− δ)d (10)

Plugging equations (9) and (8) into equation (6) yields:

1

1 + gy

(d− dt−1

1 + r

1 + gy

)+β

(

1 + gy−r1+gy

d)∑J−1

j=0 (β (1 + gy))j + (β (1 + gy))J vrt+J+1(d, 0)

−

( ∑J−1j=0 (β (1 + gy))j − βJ−1 (1+r)J

(1+gy)2(1− h)dt−1+

(β (1 + gy))J−1 d+ (β (1 + gy))J vrt+J+1(d, s0)

) ≥ 0

simplifying and multiplying through by (1 + gy):

d− dt−11 + r

1 + gy+ β (1 + gy)

d(gy−r1+gy

)∑J−1j=0 (β (1 + gy))j +

βJ−1 (1+r)J

(1+gy)2(1− h)dt−1 − (β (1 + gy))

J−1 d+

(β (1 + gy))J{vrt+J+1(d, 0)− vrt+J+1(d, s0)

} ≥ 0

Using equation (10) yields:

d−dt−11 + r

1 + gy+

d(gy−r1+gy

)∑J−1j=0 (β (1 + gy))j+1 + (β (1 + gy))J+1

s01+gy

d

1−β(1+gy)(1−δ)

+βJ (1+r)J

1+gy(1− h)dt−1 − (β (1 + gy))

J d

≥ 0

which we divide by d and multiply by 1+gy to obtain our final expression in equation

(7).

70

References

Aguiar, M. and G. Gopinath, 2006, “Defaultable Debt, Interest Rates and the Current

Account,” Journal of International Economics, 26, pp. 64-83.

Aiolfi, M., L.A.V. Catao, and A. Timmermann, 2011, “Common Factors in Latin

America’s Business Cycles”, Journal of Development Economics, 95, pp. 212-228.

Alfaro, Laura and Fabio Kanczuk, 2005,“Sovereign Debt as a Contingent Claim: A

Quantitative Approach,” Journal of International Economics, 65(2), pp. 297-314.

Arellano, C., 2008, “Default Risk and Income Fluctuations in Emerging Economies,”

American Economic Review, 98(3), pp. 690-712

Arellano, C., X. Mateos-Planas, and V. Rios-Rull, 2013, Partial Default, unpub-

lished manuscript

Aizeman, J and N. Marion, 2004, “International Reserve Holdings with Sovereign

Risk and Costly Tax Collection,” Economic Journal, 114, pp 569-591.

Barro, R and C. Urzua, 2008, “Macroeconomic Crises since 1870”, Brooking Papers

in Economic Activity, 39(1), pp. 255-350.

Beers, D. and J. Chambers, 2006, “Sovereign Defaults at 26 Year Low, to Show

Little Change in 2007”, Standard and Poors.

Beim, D. and C. Calomiris, 2001, Emerging Financial Markets, New York.

Benczur, P and C. Ilut, 2015, “Evidence for Relational Contracts in Sovereign Bank

Lending,” Journal of European Economic Association, 14, pp. 375-404.

Benjamin, D and M.L.J Wright, 2009, “Recovery Before Redemption: A Theory of

Delays in Sovereign Debt Renegotiations”, mimeo, UCLA

Blattman, C., J. Hwang, and J. Williamson, 2006, “Winners and losers in the com-

modity lottery:The impact of terms of trade growth an volatility in the Periphery

1870–1939” Journal of Development Economics, 82, pp. 156-179.

Bolton, P. & O. Jeanne, 2011. “Sovereign Default Risk and Bank Fragility in Fi-

nancially Integrated Economies,” IMF Economic Review, 59(2), pp. 162-194

Bordo, M.D., and H. Rockoff, H., 1996, “ The Gold Standard as a good housekeeping

seal of approval,” Journal of Economic History, Vol. 56, pp. 389-428.

Borensztein, E. & U. Panizza, 2009. “The Costs of Sovereign Default,” IMF Staff

Papers, 56(4), pp. 683-741.

Borri, N., and A. Verdelhan, 2011, “Sovereign Risk Premia”, MIT, mimeo.

71

Boz, E. and E Mendoza, 2014, “Financial Innovation, the Discovery of Risk, and

the U.S. Credit Crisis”, Journal of Monetary Economics, 62, pp. 1–22

Bussiere, M. and M. Fratzscher, 2006. “Towards a New Early Warning System of

Financial Crises,” Journal of International Money and Finance 25(6), pp. 953-973.

Cantor, R. and F. Packer, 1996, “Determinants and Impact of Sovereign Credit

Ratings”, Economic Policy Review, Federal Reserve Bank of New York, issue Oct,

pages 37-53.

Catao, L.A.V., A. Fostel, and S. Kapur, 2009, “Persistent gaps and default traps,”

Journal of Development Economics, 89(2), pp. 271-284.

Catao, L.A.V and G.M. Milesi-Ferretti, 2014, “External Liabilities and Crises”,

Journal of International Economics, 94(1), pp. 18-32.

Cochrane, J., 1994, “Shocks”, Carnegie-Rochester Conference Series on Public Pol-

icy, 41, pp.296-364.

Cogley, T and T. Sargent, 2008, “The Market Price of Risk and the Equity Premium:

A Legacy of the Great Depression?”, Journal of Monetary Economics, 55 (3), pp. 454–

476

Cohen, D, and J. Sachs, 1986, “Growth and External Debt Under the Risk of Debt

Repudiation”, European Economics Review, 30, pp.437-468.

Cruces, J and C. Trebesch, 2013, “The Price of Haircuts”, American Economic

Journal: Macroeconomics, 2013, 5(3), pp. 85–117.

D’Erasmo, P., 2011, “Government Reputation and Debt repayment.” University of

Maryland.

Dell’Ariccia, G., Schnabel, Isabel and Zettelmeyer, Jeromin, 2006. “How Do Official

Bailouts Affect the Risk of Investing in Emerging Markets?,” Journal of Money, Credit

and Banking, vol. 38(7), pp.1689-1714, October.

Dias, D., C. Richmond and M.L. Wright, 2014, “The stock of external sovereign

debt: Can we take the data at ‘face value’?” Journal of International Economics, Vol.

94 (1), pp.1-17.

Dimitrova, K. and M. Ivanov, 2014, “Bulgaria: from 1879 to 1947” in South-Eastern

European Monetary and Economic Statistics from the Nineteenth Century to World

War II, Bank of Greece, Bulgarian National, National Bank of Romania, OeNB.

Eaton, J., and M. Gersovitz, 1981, “Debt with Potential Repudiation: Theoretical

and Empirical Analysis,” Review of Economics and Statistics, Vol. 48, pp. 284-309.

72

Eaton, J., 1996, “Sovereign Debt, Repudiation, and Credit Terms”, International

Journal of Finance and Economics, 1(1), pp.25-35

Edwards, S., 2015, “Sovereign Default, Debt Restructuring, and Recovery Rates:

Was the Argentinean “Haircut” Excessive?”, NBER Working Paper 20964.

Eichengreen, B., and R. Portes, 1986, “Debt and Default in the 1930s: Causes and

Consequences”, European Economic Review, Vol. 30, pp. 599-640.

Eichengreen, B. and Mody, Ashoka. 2000, “What Explains Changing Spreads of

Emerging Market Debt? ” in Capital Flows and the Emerging Economies: Theory,

Evidence, and Controversies, ed, by Sebastian Edwards, Chicago, pp. 107-34.

Eichengreen, B., R. Haussmann, and U. Panizza, 2003, “Currency Mismatches, Debt

Intolerance, and Original Sin: Why they are not the same and why it matters”, NBER

Working Paper 10036.

Esteves, Rui, 2007, “Quis custodiet quem? Sovereign Debt and Bondholders’ Pro-

tection Before 1914”, Discussion paper 323, Oxford University.

Flandreau, M. and F. Zumer, 2004, The Making of Global Finance 1880-1913, Paris:

OECD.

Flandreau, M. & J. H. Flores, 2012. “Bondholders versus bond-sellers? Investment

banks and conditionality lending in the London market for foreign government debt,

1815-1913,” European Review of Economic History,16(4), pp. 356-383.

Folkerts-Landau, David, 1985, “The Changing Role of Banking Lending in Devel-

opment Finance”, IMF Staff Papers, vol. 32 (2), pp. 317-63.

Gennaioli, N., A. Martin, S. Rossi, 2014, “Sovereign Default, Domestic Banks, and

Financial Institutions”, Journal of Finance, 69(2), pp. 819-866.

Ghosal, S, M. Miller, and K. Thampanishvong, 2010, “Delay and Haircuts in Sovereign

Debt: Recovery and Sustainability”, University of St. Andrews, WP 10/15

Gourinchas, Pierre-Olivier, and Maurice Obstfeld. 2012. “Stories of the Twentieth

Century for the Twenty-First”, American Economic Journal: Macroeconomics 4(1):

226-65

Hinic, B., L. Durdevic, M. Sojic, 2014, “Serbia/Yugoslavia: from 1884 to 1940” in

South-Eastern European Monetary and Economic Statistics from the Nineteenth Cen-

tury to World War II, Bank of Greece, Bulgarian National, National Bank of Romania,

OeNB.

International Monetary Fund, 1999, International Capital Markets. Developments,

73

Prospects, and Key Policy Issues, September, Washington, DC.

Jobst, C. and T Scheiber, 2014, “Austria-Hungary: from 1863 to 2014” in South-

Eastern European Monetary and Economic Statistics from the Nineteenth Century to

World War II, Bank of Greece, Bulgarian National, National Bank of Romania, and

OeNB.

Jorgensen, E., and J. Sachs, 1989, “Default and Renegotiation of Latin American

Foreign Bonds in the Interwar Period,” The International Debt Crisis in Historical

Perspective, ed. by B.Eichengreen and P. H. Lindert, Cambridge, MA, pp.48-85.

Kalemli-Ozcan, S., L. Laeven, and D. Moreno, 2015, “Debt Overhang in Europe:

Evidence from Firm-Bank-Sovereign Linkages”, University of Maryland, mimeo.

Kaminsky, G. and Vega-Garcıa, P, 2014, “Varieties of Sovereign Crises: Latin Amer-

ica, 1820-1931”, NBER Working Paper 20042.

Kletzer, K. and B. Wright, 2000, “Sovereign Debt as Intertemporal Barter,” Amer-

ican Economic Association, vol. 90(3), pp. 621-639.

Killingsworth, Mark and James Heckman, 1986, “Female Labor Supply: A Survey”,

in O. Ashenfelter and R. Layard (eds), Handbook of Labor Economics, Volume I, pp.

103-204.

Klingen, C., B. Weder, and J. Zettelmeyer, 2004, “How Private Creditors Fared in

Emerging Debt Markets”, 1970-2000, IMF working paper WP/04/13.

Lazaretou, S., 2014, “Greece: from 1833 to 1949”, in South-Eastern European Mon-

etary and Economic Statistics from the Nineteenth Century to World War II, Bank of

Greece, Bulgarian National, National Bank of Romania, OeNB.

Lindert, P. H., and P.J. Morton, 1989, “How Sovereign Debt Has Worked,” in J.

Sachs (ed.) Developing Country Debt and Economic Performance, Cambridge, MA.

Longstaff, F.A., J. Pan, L. Pedersen, and K.J. Singleton, 2011, “How Sovereign is

Sovereign Credit Risk”, American Economic Journal 3(2), pp. 75-103

Maddison, Angus (2003), The World Economy: Historical Statistics, Paris

Mauro, P., N. Sussman and Y. Yafeh, 2002, “Emerging Markets Spreads: Then

versus Now”, Quarterly Journal of Economics, Vol. 117, No. 2, pp. 695-733.

Mauro, P., N. Sussman and Y. Yafeh, 2006, Emerging Markets and Financial Glob-

alization: Sovereign Bond Spreads in 1870–1913 and Today, Oxford.

Mitchell, Brian, 2013, International Historical Statistics 1750-2005, 3 volumes, Lon-

don. Available online: http://www.palgraveconnect.com/pc/doifinder/10.1057/9781137305688

74

Mitchener, K.J., and M. D. Weidenmier, 2005. “Supersanctions and Sovereign Debt

Repayment,” NBER Working Papers 11472.

Obstfeld, M., and K. Rogoff, 1996, Foundations of International Macroeconomics,

Cambridge, MA.

Obstfeld, M. and A. Taylor, 2003, “Sovereign Risk, Credibility and the Gold Stan-

dard: 1870-1913 Versus 1925-31,” Economic Journal, Vol. 113, No. 487, pp.241-75

Obstfeld, M., J. Shambaugh and A. Taylor, 2005, “The Trilemma in History: Trade-

offs Among Exchange Rates, Monetary Policies, and Capital Mobility, Review of Eco-

nomics and Statistics, 87 (3), pp. 423-438

Oosterlinck, K, and L. Ureche-Rangau, 2012, “Inter-war Romanian sovereign bonds:

the impact of diplomacy, politics, and the economy”, Financial History Review 19, pp.

219-44

Ozler, S., 1993, “Have Commercial Banks Ignored History,” American Economic

Review, Vol. 83, pp. 608-20.

Pamuk, S. and Tuncer, C, 2014, “Ottoman Empire: from 1830 to 1914” in South-

Eastern European Monetary and Economic Statistics from the Nineteenth Century to

World War II, Bank of Greece, Bulgarian National, National Bank of Romania, OeNB.

Panizza, U., F. Sturzenegger and J. Zettelmeyer, 2009. “The Economics and Law

of Sovereign Debt and Default,” Journal of Economic Literature 47(3), pp 651-98.

Pesaran, M.H. and A. Timmermann, 2007, “Selection of Estimation Windows in

the Presence of Breaks”, Journal of Econometrics, 137(1), pp. 134-161.

Prados, L., 1988, De imperio a nacion: Crecimiento y atraso economico en Espana

(1780-1930), Madrid

Reinhart, C.M., K. Rogoff, and M. Savastano, 2003, “Debt Intolerance,” NBER

Working Paper 9908.

Reinhart, C. and K. Rogoff, 2009, This Time is Different: Eight Centuries of Finan-

cial Folly, Princeton.

Reinhart, C. and C. Trebesch, 2014, “A Distant Mirror of Debt, Default and Relief”,

NBER Working Paper 2057.

Roman, C., Willebald, H., 2011, “Apuntes metodologicos para la construccion de

indicadores de inversion y de calidad institucional en el largo plazo: una propuesta para

el caso uruguayo”, II Jornadas Academicas, Montevideo.

Roubini, A. and P. Manasse, 2009, “Rules of thumb” for sovereign debt crises,”

75

Journal of International Economics, vol. 78(2), pages 192-205.

Sandlieris, G., 2008, “Sovereign Defaults: Information, Investment, and Growth”,

Journal of International Economics 76(2), pp. 267-75.

Sutter, C., 1992, “Debt Cycles in the World Economy,” Westview.

Sturzenegger F., and J. Zettelmeyer, 2006, “Debt Defaults and Lessons from a

Decade of Crises,” Cambridge, MA.

Tomz, M., 2007, “Reputation and International Cooperation: Sovereign Debt across

Three Centuries,” Princeton, NJ: Princeton University Press.

Tomz, M. and M. L.J. Wright, 2013. “Empirical Research on Sovereign Debt and

Default,”Annual Review of Economics, Annual Reviews, vol. 5(1), pp. 247-272.

Voth, H-J, 2011, “Lending to the Borrower from Hell: Debt and Default in the Age

of Phillips II, 1556-1598”, Economic Journal, 21, pp.1205-1227.

Yue, V., 2010, “Sovereign Default and Debt Renegotiation”, Journal of International

Economics, 80 (2), pp. 176-187.

76