sovereign debt and default - university of...

Sovereign Debt and Default

Sewon Hur

University of Pittsburgh

February 10, 2015

International Finance (Sewon Hur) Lecture 6 February 10, 2015 1 / 22

Introduction

Sovereign debt di�ers from corporate debt because of the lack of

a legal mechanism to enforcement repayment

In this lecture, we consider the Panizza, Sturzenegger,

Zettelmeyer (2009) survey on the existing theories of sovereign

debt/default and the empirical literature


Theories of Sovereign Debt and Default

Can a sovereign debt market exist if repayment cannot be

enforced?

Eaton and Gersovitz (1981) - yes, with the threat of permanent

exclusion from credit

However, Bulow and Rogo� (1989) show that even with the

threat of exclusion from credit, if there exist other methods to

smooth consumption such as savings or insurance, then

borrowing is impossible.



Sachs and Cohen (1982), Bulow and Rogo� (1989), Fernandez

and Rosenthal (1990) focus on direct punishment for reasons to

repay (interference with a country's current transactions such as

seizure of trade and payments)

Cole and Kehoe (1998) - if default can damage government

reputation, for instance, with the government's domestic

partners, then debt can be sustained

Sandleris (2005), Catao and Kapur (2006) focus on information

revealed by default. If default signals, for instance, that

government �nancial position is weaker than previously thought,

then future output may be reduced (perhaps due to increase in

expected taxation)



Mendoza and Yue (2008) assume defaults limit the ability of

private agents to obtain private capital - model is consistent

with rapid output and TFP collapses

Broner et al. (2006) highlight the role of secondary markets in

limiting sovereign risk.

if foreigners can sell debt to domestic residents in secondary

markets, then debt will always be repaid back, even in the

absence of punishments.

this result is ex-post ine�cient for the borrower (because if

domestic agents could coordinate and not buy back debt, the

government could default and be better o�)

but ex-ante e�cient (because it allows the country to borrow, by

solving the sovereign risk problem)International Finance (Sewon Hur) Lecture 6 February 10, 2015 5 / 22


A strand of the literature takes the existence of sovereign debt as

given, and explore the e�ect of investor behavior or expectations

run on debt (Sachs 1984, Alesina 1990, Cole and Kehoe 1996,

2000)

run on currency (Aghion 2001, 2004, Krugman 1999, Burnside

et al. 2004)

sudden stops (Mendoza 2012, Calvo 1998, Hur and Kondo 2013)



Many recent papers have gone back to Eaton and Gersovitz's

(1981) implicit assumption that countries do not have a savings

opportunity after defaulting

Aguiar and Gopinath (2006), Arellano (2008), Benjamin and

Wright (2008), Yue (2006) to name a few...


Empirical Literature

When do countries borrow?

When do countries default?

What are the default costs?


When do Countries Borrow?

Levy-Yeyati (2009) �nds that private lending to sovereigns is

procylical, while o�cial lending is coutercyclical, with a net

procyclical e�ect.

This is in contrast to standard theory where sovereign borrowing

is countercyclical (to smooth consumption)


Why is Borrowing Procylical?

market failure: lack of access to international lending credit

during recessions (Gavin and Perotti 1997), incomplete markets

(Caballero and Krishnamurthy 2004), limited enforcement

(Kehoe and Perri 2002), moral hazard (Atkeson 1991), etc

political failure: con�ict of interest groups (Tornell and Lane

1999), political pressure for wasteful spending (Talvi and Vegh

2005), corrupt politicians (Alesina et al. 2008)

nature of output shocks: Aguiar and Gopinath (2006) and

Rochet (2006) show that a model with persistent shocks can

generate procyclical borrowing even in the absence of political or

market imperfections


When do Countries Default?

In standard sovereign debt models, countries borrow during bad

times and repay during good times.

countries might be tempted to default, but anticipating this,

creditors will not lend beyond a threshold level of debt at which

defaulting is preferable to repaying.

as a result, in the simplest models, defaults never happen.

Defaults can arise in equilibrium in sovereign debt models with

output uncertainty and incomplete contracts.

countries default in bad states


Evidence

The evidence is broadly consistent with theory

Levy-Yeyati (2006) �nds that defaults tend to follow output

contractions (1982-2003)

Tomz and Wright (2007) �nd a negative correlation between

output and defaults (1820-2004)


Puzzles Remain..

Still, the theory is inconsistent with several features of default

theories underpredict default - Aguiar and Gopinath (2006)

generate higher default rates through persistent shocks;

Hatchondo and Martinez (2008) with long-duration bonds. Still

default probabilities are lower than observed.

empirical relationship between bad output and defaults is not as

tight - Tomz and Wright show that only 62 percent of defaults

occur when output is below trend. This could be due to other

shocks such as political shocks, credit shocks, interest rate

shocks


Defaults Happen in Clusters

Defaults tend to happen in clusters, suggesting that defaults are

in�uenced by the behavior of creditors and international capital

markets (interest rate shocks, sudden stops), in addition to

debtor country shocks (output or political shocks)669Panizza, Sturzenegger, and Zettelmeyer: Sovereign Debt and Default

A third perspective on debt crises, in addi-tion to output shocks and sudden reversals in international capital markets, focuses on the idea that some countries may in fact “over-borrow,” that is, accumulate debt that is too high from a welfare perspective, and at—or perhaps just below—the debt level that competitive creditors will accept.26 Given this high debt, small shocks (of whatever kind) could trigger a default. Debt accumulation

26 Evidence for overborrowing is provided in International Monetary Fund (2003), chapter 3, and Mendoza and Jonathan D. Ostry (2002). Reinhart and Rogoff (2008a) examine domestic and external debt and defaults jointly and show that external defaults are often driven by the accumulation of unsustainable domestic debt.

might be excessive because the parties that contract the debt do not bear the full costs of repayment or crises. For example, excessive borrowing could benefit specific groups at the expense of the average domestic taxpayer (Perotti 1996). Moral hazard could also occur at the expense of the foreign taxpayer, if countries in crisis are “bailed out” by institu-tions such as IMF or World Bank, or through bilateral lending.27 Finally, overborrowing

Num

ber

of s

over

eign

def

ault

epis

odes

Year in which the country declared default

18

16

14

12

10

8

6

4

2

01820 1828 1836 1844 1852 1860 1868 1876 1884 1892 1900 1908 1916 1924 1932 1940 1948 1956 1964 1972 1980 1988 1996 2004

Figure 1. Default Clusters, 1820–2005

Source: Sturzenegger and Zettelmeyer (2007) and Borensztein and Panizza (2008).

27 For this argument to make sense, official loans must contain a subsidy, either by carrying an interest rate that does not reflect the riskiness of the loan for the official lender or because the debtor country expects part of the loan to be forgiven. If this is not the case, the safety net would be operated at no one’s expense and, hence, could not be a source of moral hazard by definition (Jeanne and Zettelmeyer 2001, 2005a).


Costs of Default

Capital market exclusion

Sandleris et al. (2004) �nd that countries were excluded for an

average of four years after defaults ended (1980s) and 0-2 years

since then

Richmond and Dias (2008) using a stronger de�nition of

exclusion (positive net transfers) �nd exclusions of 5.5 years

(1980s), 4.1 years (1990s) and 2.5 since

bottom line: calibrated models with exclusion alone cannot

generate debt levels and default frequencies, many assume

additional exogenous output costs (Alfaro and Kanzcuk 2005,

Arellano 2008, Aguiar and Gopinath 2006, Benjamin and Wright

2008)


Costs of Default

Higher borrowing costs

Borendztein and Panizza (2010) �nd that spreads are 400 basis

points higher in the year after default, 250 higher in second year,

and loses statistical signi�cance thereafter (1997-2004)

Flandreau and Zumer (2004) �nd that spreads are 90 basis

points higher in the year after, but the e�ect dies out rapidly

(1880-1914)

bottom line: calibrated models with borrowing costs alone

cannot generate debt levels and default frequencies, unless for

instance output costs are assumed (Alfaro and Kanzcuk 2005)


Costs of Default

Domestic costs

models predict defaults happen in bad states. Do defaults cause

output drops, or make already bad states worse? evidence is

mixed

Sturzenegger (2004): defaults are associated with a reduction in

growth of 0.6 percent (2.2 percent if default comes with banking

crisis)

De Paoli et al. (2006): output losses are correlated with defaults

and increase with duration of default

Levy-Yeyati and Panizza (2006): defaults tend to happen in

trough, and often mark beginning of recovery

su�er from endogeneity biases


Recent Default Episodes683Panizza, Sturzenegger, and Zettelmeyer: Sovereign Debt and Default

restructuring with basically no haircut (the Dominican Republic). Finally, about half of the restructurings listed in table 4 were preemptive (i.e., they took place before the country missed any payment on its existing debt) while the other half were postdefault restructurings.49

Consider first whether there is any evi-dence of capital market exclusion. Figure 4, based on a regression that includes all developing countries with an income per

49 For detail on these restructurings, see Sturzenegger and Zettelmeyer (2007b, 2008). Harald Finger and Mauro Mecagni (2007) examine the question of whether the restructurings were successful in restoring debt sustainability.

capita of at least US$500 and controls for both country and year fixed effects, sug-gests not. In the year of the default episode, private capital flows to the defaulting coun-tries were slightly below trend but started to recover immediately thereafter and, within three years of the episode, they were already above trend. Argentina, Russia, and Ecuador observed a collapse of capital flows one or two years before the default. In these countries, capital inflows reached a trough in the year after the default but then recov-ered quickly. The case of Argentina is par-ticularly interesting. This country had by far the largest and least creditor friendly default in this group. Nonetheless, two years after the default, private inflows were so high that

TABLE 4 Characteristics of Recent Debt Restructurings

Country Year

Total amount restructured1

(bill US$)Haircut

(%) Type of restructuring

Russia 1998–2000 38.7 52.6 Postdefault

Ukraine 1998–2000 7.8 28.9 Predefault

Pakistan 1999 0.61 31 Predefault

Ecuador 1999–2000 6.5 28.6 Postdefault

Argentina 2001–2005 145 75 Pre- and postdefault

Uruguay 2003 5.4 13.3 Predefault

Moldova 2002 0.08 37 Pre- and postdefault

Dominican Republic 2005 1.5 2 Predefault

1 Domestic and external debt with private creditors.

Source: Sturzenegger and Zettelmeyer (2007, 2008).


Some Evidence of Temporary ExclusionJournal of Economic Literature, Vol. XLVII (September 2009)684

the country had to impose capital controls on inflows. Profit opportunities—perhaps linked to the behavior of the real exchange rate—seemed to have dominated any rep-utational considerations. This is not to say that the restructuring strategy may not have had an impact on the behavior of capital flows around the time of the restructuring. Countries that opted for a preemptive strat-egy (Dominican Republic and Uruguay) seemed to enjoy a recovery of private inflows even before the restructuring. However, the evidence suggests that effects on the volume of capital flows were at best transitory.

Figure 5 examines whether there is any evidence for “punishment” via borrowing

costs, based again on a large panel regres-sion that controls for country and time fixed effects, and excludes the months in which a country is in default. The main result is that—controlling for changes in global financial conditions (via time dummies)—postdefault spreads return to predefault levels within twenty-four months or less. However, figure 4 does not control for changes in the fundamentals of defaulting countries. If these improve as a result of the restructuring, the rapid convergence shown in the figure could still be consistent with the idea that defaulters pay higher spreads than nondefaulters with similar fundamentals.

Ecuador Moldova Pakistan

.1

0

−.1

−.2

All countries Argentina Dominican Republic

Russia Ukraine Uruguay

Event time Event time Event time



.05

0

−.05

−.1

.020

−.02−.04−.06

.020

−.02−.04−.06−.08

.2.15

.1.05

0−.05

.1

.05

0

−.05

.05

0

−.05

−.1

.05

0

−.05

−.1

.10

−.1−.2−.3

−5 −4−3 −2−1 0 1 2 3 4 5 −5 −4 −3 −2 −1 0 1 2 3 4 5 −5 −4−3 −2 −1 0 1 2 3 4 5

−5 −4−3 −2 −1 0 1 2 3 4 5 −5 −4 −3 −2 −1 0 1 2 3 4 5 −5 −4−3 −2 −1 0 1 2 3 4 5

−5 −4−3 −2 −1 0 1 2 3 4 5 −5 −4 −3 −2 −1 0 1 2 3 4 5 −5 −4−3 −2 −1 0 1 2 3 4 5

Figure 4. Recent Defaults and Private Capital Flows

Notes: The figure plots the residuals of a regression that controls for country and year fixed effects and that includes all developing countries that had an income per capita greater than US$500 in the year 2000.

private capital �ows (residuals of regression with country and year

�xed e�ects)International Finance (Sewon Hur) Lecture 6 February 10, 2015 19 / 22

Limited Evidence of High Spreads Post-Default685Panizza, Sturzenegger, and Zettelmeyer: Sovereign Debt and Default

To deal with this objection, we regress average spreads in year t on economic fun-damentals measured in year t − 1, in addi-tion to year fixed effects, in a 1994–2008 sample (that is, centered on the recent default episodes).50 Our variables of interest are three dummies that take value 1 in each of the three years after the resolution of the default episode (DEFt+1, DEFt+2, and DEFt+3) and a dummy that takes value

50 We measure fundamentals with the log of GDP per capita (GDP_PC), the current account balance divided by GDP (CA/GDP), log inflation (INF), total public debt over GDP (TPuD/GDP), and the share of public external debt over total public debt (EPuD/TPuD).

one in each year after the third year of the default episode (D > t+3).51 Column 1 of table 5 reports the results of the model esti-mated controlling for both credit ratings and economic fundamentals. We find a positive but statistically insignificant effect of default on spreads in the first two years, while the coefficients on (Dt+3) and (D > t+3) are statistically significant and negative. When we repeat the regression without controlling

51 Thus, if a country defaulted in 1998 and concluded the episode in 2000, Dt+1 takes value one in the year 2001, Dt+2 takes value one in 2002, Dt+3 takes value one in 2004, and D > t+3 takes value one in 2005, 2006, 2007, and 2008. In countries that never defaulted, the four dummies always take value zero.

bps

bps

bps

bps

bps

bps

bps

3000

−1000


Event time, months Event time, months Event time, months−36 −24 −12 0 12 24 36 −36 −24 −12 0 12 24 36 −36 −24 −12 0 12 24 36

Ecuador Pakistan Russia

Ukraine Uruguay

3000

0

2000

0

2000

0

5000

0

2000

0

2000

0

2000

0

Event time, months Event time, months Event time, months

Event time, months Event time, months

−36 −24 −12 0 12 24 36 −36 −24 −12 0 12 24 36 −36 −24 −12 0 12 24 36

−36 −24 −12 0 12 24 36 −36 −24 −12 0 12 24 36

Figure 5. Spreads Before and After Defaults

Notes: The figure plots the residuals of a regression that controls for country and year fixed effects and that includes all countries that are included in the JP Morgan EMBI+ Global. spreads (residuals of regression with country and year �xed e�ects)


Mixed Evidence of Output Costs689Panizza, Sturzenegger, and Zettelmeyer: Sovereign Debt and Default

output declines, financial sector disruptions and generally a period of capital market exclusion while debt is renegotiated. This fact motivated a set of proposals, beginning in the 1980s and particularly following the 1995 Mexican and 1998 Russian crises, to reform the institutions and/or contracts gov-erning debt flows and debt renegotiation in order to reduce the cost of crises. 55 For the most part, these proposals—which climaxed in the proposal by management and staff of the IMF, in 2001, to create a bankruptcy-type “sovereign debt restructuring mechanism”

55 For an early proposal, see UNCTAD (1986).

(SDRM) for countries—focused on making the debt renegotiation process smoother and faster, in particular, by mitigating credi-tor coordination failures (see Rogoff and Zettelmeyer 2002 for a history).

As argued in the previous section, pro-posals in this class can perhaps be criticized (with the benefit of hindsight) for having barked up the wrong tree—creditor coordi-nation failures did not, in the end, turn out to be a significant impediment to the debt rene-gotiations of the 1998–2005 period. Beyond questions of empirical relevance, however, proposals that aim to reduce the costs of debt crises raise a deeper issue (Michael P. Dooley 2000; Andrei Shleifer 2003). If sovereign

Ecuador Moldova Pakistan


Russia Ukraine Uruguay

.1.05

0−.05−.1

−.15

.050

−.05−.1

−.15

.05

0

−.05

−.1


.020

−.02−.04−.06

.050

−.05−.1

−.15

.05

0

−.05

−.1

.1.05

0−.05−.1

−.15

.1

0

−.1

−.2

.050

−.05−.1

−.15



−5 −4−3 −2 −1 0 1 2 3 4 5 −5 −4−3 −2 −1 0 1 2 3 4 5 −5 −4−3 −2 −1 0 1 2 3 4 5

−5 −4−3 −2 −1 0 1 2 3 4 5 −5 −4−3 −2 −1 0 1 2 3 4 5 −5 −4−3 −2 −1 0 1 2 3 4 5

−5 −4−3 −2 −1 0 1 2 3 4 5 −5 −4−3 −2 −1 0 1 2 3 4 5 −5 −4−3 −2 −1 0 1 2 3 4 5

Figure 7. Recent Defaults and GDP Growth

Notes: The figure plots the residuals of a regression that controls for country and year fixed effects and that includes all developing countries that had an income per capita greater than US$500 in the year 2000.

GDP growth (residuals of regression with country and year �xed

e�ects)


Costs of Default

Sovereign defaults are typically associated with output declines,

�nancial sector declines, and a temporary period of capital

market exclusion

Evidence for persistent e�ects of default is limited


sovereign debt and default - university of...

Documents