Duration of Capital Market Exclusion: Stylized Facts and
Determining Factors
Christine Richmondy
Daniel A. Diasz
First Draft: June 2007
This Draft: August 18, 2008
Abstract
This paper examines why some countries are able to regain access to international
capital markets immediately after resolving a default, whereas other countries appear to be
punished for longer periods. We develop a methodology to determine when market access
occurs after default settlement and establish a set of stylized facts. Our main …ndings
from examining the duration of exclusion from international capital markets between 19802005 by sovereign defaulters are: i) countries regain partial market access after 5.7 years
on average (median of 3.0 years) while it takes 8.4 years on average (median of 7.0 years)
to regain full market access; ii) partial market access depends mostly on external demand
for risk; iii) full market access depends primarily on good domestic behavior and market
expectations; iv) size matters, with large economies regaining market access twice as fast
as small countries; and v) there are regional di¤erences, with African and Middle Eastern
defaulters taking twice as long to regain market access when compared to other regions.
JEL Classi…cation: F21, F34, G15, H63
Keywords: sovereign default, market access, international capital markets, duration
This is a substantially revised version of a paper that circulated previously under the title "Regaining
Market Access: What Determines the Duration of Exclusion?". We have bene…tted from discussions with
Stijn Claessens, Sebastian Edwards, Edward Leamer, Mike Tomz, Aaron Tornell, Christoph Trebesch, Mark
Wright, and Jeromin Zettlemeyer. We are grateful for additional comments from Jonathan Eaton and an
anonymous referee. We thank seminar participants at the 2008 Royal Economic Society Annual Conference,
2007 LACEA-LAMES Meetings, Central Bank of Portugal, and UCLA International Economics Proseminar for
their comments and suggestions. All remaining errors are our own.
y
Corresponding author. Anderson Graduate School of Management, UCLA. 110 Westwood Plaza, Entrepreneurs Hall, Suite C525, Los Angeles, CA 90095-1481, United States. Tel.: +1 310.825.8207, Fax: +1
310.825.4011; E-mail: [email protected].
z
Anderson Graduate School of Management, UCLA and CEMAPRE. 110 Westwood Plaza, Entrepreneurs
Hall, Suite C525, Los Angeles, CA 90095-1481, United States. E-mail: [email protected].
1
1
Introduction
Why are some countries able to access international capital markets immediately after resolving
a default, while others seem to be punished and are forced to remain on market sidelines?
Looking at two countries in Latin America, we can contrast their market experiences, although
both are recent defaulters. Argentina has defaulted four times on foreign currency bond debt
and twice on foreign currency bank debt during the last 183 years (Beers and Chambers,
2006). The most recent default, in December 2001 on USD79.7bn in foreign debt, excluding
past due interest, took until May 2005 to be resolved when the majority of bondholders …nally
accepted the government’s terms (Dhillon et al., 2006). Despite being in default for 3.5 years
and forcing investors to realize large haircuts on their positions, once the restructured bonds
began trading in the grey market Argentina appeared to have immediately regained access to
international capital markets. Ecuador, however, has faced a very di¤erent experience with
international capital markets and has largely been cut o¤ from markets since its default in
1999 on USD6.5bn in foreign debt (which was settled in 2000).
While some recent research has begun to examine the question of when a country will
have market access (IMF, 2001, 2003, and 2005), the question of how long a country will be
excluded from international capital markets once a period of default is settled has yet to be
examined. This line of research is di¤erent from earlier work as we look explicitly at periods of
sovereign default during the modern …nancial period (1980-2005) in an e¤ort to determine the
duration of market exclusion, rather than solely identifying characteristics of market access.
We also examine whether the type of default - bank debt or bond debt - and other country
characteristics can generate di¤erent durations of market exclusion.
The question of access to capital markets is important from three points of view. First, from
the country point of view capital market access plays an important role for developing countries
economies, particularly for trade and investment activities. Domestic investment projects for
both infrastructure and capacity-building purposes can boost a country’s productivity and
growth over the long run as well as improve international competitiveness.
However, since
most countries do not have the domestic resources available to …nance such large-scale projects,
the main vehicles for funding are overseas aid, borrowing from international capital markets, or
assistance from development banks (Eichengreen, 1994). Rose (2005) presents support for the
hypothesis that the downside of a non-repayment strategy comes through the trade channel
and …nds that debt renegotiation is associated with a decline in bilateral trade of approximately
eight percent a year that persists for around …fteen years. Kohlscheen and O’Connell (2007)
show that trade credit lines, the vehicle used to …nance international trade, can disappear
during periods of default and debt renegotiation. In analyzing sovereign defaults between 1992
and 2001 the authors …nd that the volume of trade credit provided by banks typically falls
considerably following a default, with the median reduction in trade credit, relative to the year
of default, amounting to 35% after two years and 51% after four years. Thus, knowledge of the
period of exclusion and the costs associated with it can in‡uence a country’s decision whether
2
or not to default.
If the punishment of exclusion is short and economic impact is small, it
could lead to more frequent defaults.
The study of this issue may also allow borrowers to
learn what actions they can undertake, if any, in order to minimize the period of exclusion
from markets. Secondly, from the lender’s point of view, potentional lenders can understand
the behaviour of countries during default periods and assist in the evaluation of whether or
not to extend new funds post default. Finally from the institutional point of view, study of
this issue can assist in the design of country policies to allow for more continued market access
or help countries graduate to market-based borrowing.
The remainder of the paper proceeds as follows. In section 2 we present a review of relevant
literature. In section 3 we de…ne the empirical strategy of the paper, presenting our stylized
facts as well as the macroeconomic variables that we believe could drive the outcomes, and
present the results, while section 4 concludes.
2
Related Literature
Much of the literature on sovereign debt focuses on why governments wish to repay their
obligations and has largely ignored the issue of how long countries have been excluded from
international capital markets after resolving a default. This has been the case starting with
Eaton and Gersovitz’s (1981) seminal work where sovereigns repay debt because future lending
depends on reputation and there is a threat of a permanent embargo on future loans by private
sector lenders if a default should occur. However, real-world observations show that countries
do default and are able to borrow again at some point in the future.
Recent quantitative
research by Arellano (2008) and Yue (2006) extend the approach developed by Eaton and
Gersovitz (1981) and models country exclusion from borrowing as part of a stochastic general
equilibrium model with endogenous default risk. Arellano (2008) models exclusion from international …nancial markets after defaulting as a stochastic number of periods, with reaccess
occurring with an exogenous probability, or independent of both global …nancial conditions
and country-speci…c conditions. Yue (2006) considers the reaccessing of …nancial markets dependent on the country’s bargaining power, with greater bargaining power resulting in shorter
periods of exclusion, averaging around 1 year.
Preliminary work has begun to endogenize the period of exclusion from capital markets.
Pitchford and Wright (2007) model the sovereign debt renegotiation process, during which
time it is assumed that there is no market access, using an incomplete markets model with an
explicit model of the sovereign debt renegotiation process, with delays arising due to creditor
holdout for better terms and freeriding by negotiators, while Benjamin and Wright (2008)
study delays in the debt renegotiation process, where outcomes are driven by ‡uctuations in
domestic economic conditions as well as changes in creditor and debtor bargaining power. Bi
(2008) argues that delays in a debt renegotiation process may be mututally bene…cial so as to
increase the size of the "cake" and …nds that given the defaulted debt level, the expected delay
length is determined by the output process.
3
Empirical work on this area is also limited.
Lensink and Van Bergeijk (1991) present
one of the …rst papers tackling the determinants of a country’s ability to access international
capital markets during the 1985-1987 period. The authors use the observation of whether a
country has access to international capital markets or not as the dependent variable, which is
based on undertaking actual borrowing. However, by utilizing this de…nition the data sample is contaminated because it groups countries that do not need to borrow with those that
do not have access to markets, thereby in‡ating the number of countries observed to have
no market access.
Gelos et al. (2004) measure the frequency of borrowing by developing
countries during the 1980-2000 period to determine access to international capital markets and
…nd that vulnerability to shocks, perceived quality of policies, and institutions are important
determinants of access. Fostel and Kaminsky (2007) examine the question of whether volatile
international capital markets are the main reason for the boom-bust pattern of Latin America capital market participation.
The authors …nd that during the 1990s, domestic factors
(macroeconomic policies, economic activity, political risk, real exchange rate volatility, and
openness) were important for international capital market access by Argentina, Brazil, and
Chile, whereas external factors (global liquidity, world economic activity, and terms of trade)
were important for Colombia, Mexico, and Venezuela. However, during the period of 20022005, external factors were most important for market access by all of the countries analyzed.
Lastly, Trebesch (2008) studies delays in the sovereign debt renegotiation process and …nds
that the average duration from the start of debt distress until the …nal debt renegotiation deal
is about 2.5 years, which is largely driven by political instability and government actions rather
than creditor holdout.
Our paper looks to add to the literature by empirically analyzing the duration of the
exclusion from international capital markets in the aftermath of defaults. We …rst establish
a set of stylized facts regarding characteristics of market exclusion.
We then speci…cally
take into account the quantity of new borrowing and identify macroeconomic characteristics
of defaulters to measure the length of market exclusion.
3
3.1
Empirical Strategy
De…nitions
As a starting point for our analysis, we de…ne two terms: default and market access. We use
Standard & Poor’s standard de…nition of default: "...the failure to meet a principal or interest
payment on the due date (or within the speci…ed grace period) contained in the original terms
of a debt issue ... or tenders an exchange o¤er of new debt with less-favorable terms than the
original issue" (Beers and Cavanaugh, 2006). Further, Standard & Poor’s considers a country
to have emerged from default when the agency has concluded that "...no further near-term
resolution of creditors’claims is likely" (Beers and Cavanaugh, 2006).
4
Using S&P’s de…nition of default, during the period 1980-2005, we have identi…ed 128
episodes of sovereign default on foreign currency bank debt and foreign currency bonds (Beers
and Chambers, 2006), with an average default of 7.3 years.1 The average length of default on
foreign currency bank debt was 8.3 years and 2.6 years for default on foreign currency bonds.
We next consider defaults in four separate regions: Emerging Europe, Latin America and
Caribbean, Asia-Paci…c, and the Middle East and Africa and summarize the default results in
Table 1. It is worth stressing that while a country is not classi…ed by S&P as entering into
default until the expiration of the grace period, a country may in fact be in default. Thus the
reported durations of default should be recognized as having a downward bias, which could be
up to one year.
insert Table 1 around here
We de…ne market access to be the …rst of either of the following events occurring post
default exit: (i) positive net transfers in the form of bonds and commercial bank loans to the
public or publicly guaranteed sector; or (ii) positive net transfers from bonds and commerical
bank loans to the private sector.
By looking at these measures we can say whether or not
a country has access to international capital markets even if they do not borrow because we
may assume that for a private …rm to borrow from abroad, the country must be in good
…nancial standing.2 We choose to restrict our attention to net transfers, new borrowing less
debt service, to represent the net ‡ow of real resources from bank and bond creditors to the
debtor and distinguish between a country merely rolling over its debt and contracting new
debt (Eaton, 1992). This limits the problem faced during the 1980s when commercial banks
rolled over loans to developing countries rather than writing them down and prevents us from
considering these funds as a country having market access.
We consider bond and bank debt instruments to be close substitutes as sources of external
…nancing. Bonded debt is a contract with covenants and loan-granting decisions dependent
on only public information, while a bank loan uses the public information as well as additional
information gathered via costly monitoring of the borrower’s actions. A key implication of this
result is that once a country establishes a positive reputation, the need for close monitoring is
reduced.3
We also consider a country as being able to gain market access even if there is outstanding
litigation with holdout creditors.
While in earlier times it may have been more di¢ cult
1
We exclude default events in Cuba, North Korea, and the Former Yugoslavia.
Typically the ratings of private …rms are constrained by the country rating. This is especially evident in
the case of developing countries. As of June 2008 Standard & Poor’s rated only 82 corporate, counterparty,
and municipal ratings above the rating of the sovereign in the country of domicile. On a foreign currency basis,
moreover, only 31 of these entities were based in developed countries. See Cavanaugh et al. (2008).
For a discussion on emerging markets’ private sector access to international debt markets during sovereign
debt crises, see Arteta and Hale (2007, 2008).
3
See, for example, Diamond (1991).
2
5
to access markets if a country was facing legal challenges and the possibility of assets being
attached, today the global bond market has developed su¢ ciently so that debt can be issued in
di¤erent legal jurisdictions, i.e., the Eurobond market, which allows US-based …rms to purchase
government securities from countries with outstanding litigation. Further, the advent of special
purpose vehicles (SPVs) has created loopholes in existing legislation, which makes it easier for
countries to issue debt and avoid the attachment of assets.4
We next distinguish between partial and full market reaccess. We consider partial reaccess
as the …rst year in which there are positive net bond and bank transfers to the public or private
sector, whereas full market reaccess is de…ned as the …rst year of positive net bond and bank
transfers to the private or public sector greater than 1.0% of GDP. This threshold is chosen
as it is, on average, less than one-half of the annual central government borrowing requirement
over the entire sample.5 If a country exits default and regains market access in the same year,
we consider the duration of market exclusion to be one year.
The rationale for this is that
we know the duration of the event is greater than zero, but by using discrete data our …rst
observation of the change is in the following period.6
3.2
Stylized Facts
We begin by presenting some important empirical regularities that our unconditional analysis
uncovers, most previously not established.
While most facts should not be seen as a big
surprise, they are signi…cant in that they do not necessarily match the speci…cations being
used in the current sovereign debt model literature.
Fact 1 - Countries are excluded from capital markets for a long period of time.
The average length of time it takes for a country to be able to regain partial market access is
5.7 years, while regaining full market access takes 8.4 years on average. From an analysis of
the distribution of market exclusion, in Tables 2A and 2B, we see that 50% of the countries
regain partial market access within 2 years, while it takes 4 years for 50% of the countries
to regain full market access.
Regaining partial access occurs much faster than full access;
in one year 46% of countries regain partial market access, whereas only 29% of countries
regain full market access over the same time period.
These results are signi…cantly longer
than previously implied by Gelos et al. (2004) and those being modeled theoretically.
The
long duration of exclusionary periods lends support to the premise that countries are indeed
punished by markets for defaulting.
4
One can consider the case of EM Ltd. v. Russia, which was unsuccessful in its many attempts to attach
assets. In the past, government and central bank assets have been placed in the Bank of International Settlements (BIS) in Switzerland to utilize the legal protection a¤orded to the BIS against the attachment of assets.
See, for example, Sturzeneger and Zettelmeyer (2006b).
5
The idea of setting a threshold is not uncommon in the international …nance literature, particularly for
…nancial crises and currency crises. For a summary of the di¤erent currency crisis de…nitions used by researchers
see, for example, Esquivel and Larraín (1998).
6
See Appendix 2 for performance of our measure against default periods.
6
insert Table 2 around here
insert Table 2A around here
insert Table 2B around here
Fact 2 - There are regional di¤ erences to the length of market exclusion.
Defaulters from Middle Eastern & African countries are excluded from capital markets for
substantially longer periods than defaulters from other regions, in terms of both partial and
full market access. Latin America & Caribbean country defaulters regain partial market access
the fastest, with an average exclusion of 3.1 years, while Emerging European countries regain
full market access the fastest, with an average exclusion of 4.7 years.
insert Table 3 around here
Fact 3 - Larger countries regain market access faster.
We use nominal USD GDP to measure country size, and organize by quartile. We …nd that
smaller countries take signi…cantly longer to regain both partial and full market access - 7.8
years vs. 3.4 years on average and 9.4 years vs. 5.1 years on average, respectively. Countries
in the largest GDP quartile regain full market access after a median of 4 years while countries
in the smallest GDP quartile regain full market access after a median of 9 years.
We also obtain similar results when we use population as a measure of country size. Large
countries tend to have larger nominal debt stocks, which allows for more liquid debt instruments
and leads to higher weightings in asset class indices, making them more important for investors.
insert Table 4 around here
Fact 4 - Countries with larger haircuts are excluded for longer periods of time.
Utilizing Benjamin and Wright’s (2008) database of 90 defaults by 73 countries settled between
1989-2006, we sort the size of haircuts into quartiles.
As shown in Table 5, countries that
experience only a small writedown on their debt obligations (less than 14% haircut) regain
partial market access in 3.4 years on average and full market access in 4.8 years on average.
Countries with large haircuts, greater than 46%, regain partial market access in 6.2 years on
average and full market access in 8.3 years on average.
These results are in line with Benjamin and Wright’s (2008) …nding that longer defaults
are associated with larger haircuts and show that even though countries experiencing larger
haircuts, which should improve their debt and debt servicing position post default, are punished
by investors for having to realise a writedown.
insert Table 5 around here
7
Fact 5 - Defaulters in the 1980s take the longest to regain market access.
Looking at the decade in which the default episode begins, we …nd that defaulters in the 1980s
are excluded from capital markets for longer periods of time that defaulters from more recent
decades. This …nding is in line with Trebesch (2008) who …nds that debt renegotiations before
1999 took substantially more time to be completed.
insert Table 6 around here
Fact 6 - Favorable external market conditions generate shorter exclusion periods.
We measure external market conditions by the spread between US Treasury yields and US
high yield bonds, which serves as a proxy for emerging market assets and risk levels. A small,
tight, spread between US Treasuries and High Yield assets would indicate more liquidity in
the market as well as greater investor risk appetite. We …nd that in periods of tight spreads
(less than 1.28%), countries regain full market access more quickly, with an average exclusion
of 5 years, while periods of wide spreads (greater than 2.22%) result in regaining market access
very slowly, with an average exclusion of 11 years.
insert Table 7 around here
Fact 7 - The existence of an IMF program does not signi…cantly shorten market
exclusion periods.
Countries with an IMF program, either Standy-By Agreement (SBA), Extended Fund Facility
(EFF), Poverty Reduction and Growth Facility (PRGF), or Structural Adjustment Facility,
regain full market access after 7.4 years on average, while countries without an IMF program
regain full market access after 7.7 years on average. We also check speci…cally whether an SBA
can shorten the period of market exclusion but we do not …nd substantially di¤erent exclusion
periods between countries with and without SBAs. This type of program is typically over a
1-2 year period with loan conditions focused on macroeconomic targets.
insert Table 8 around here
Fact 8 - Higher credit ratings result in shorter exclusion periods.
Countries with the highest credit ratings (above 36.4) regain full market access in 5.9 years on
average, while countries with credit ratings below 19.9 regain full market access in 9.5 years
on average.
The Institutional Investor credit rating of 36.4 corresponds to roughly a BB-
rating by S&P, which is the minimum rating observed for countries with continued market
8
access.7 These results support Reinhart’s (2002) view that credit ratings play a "crucial role
in determining the terms and the extent to which countries have access to international capital
markets", with the lowest ratings unable to borrow from international capital markets at all.
insert Table 9 around here
Fact 9 - Political risk does not signi…cantly impact exclusion periods.
Countries with the highest political risk (less than 55 on a 100 point scale, where 100 is least
risky) do not appear to have a signi…cant impact on the length of exclusion from capital
markets, particularly for partial market access. The political risk rating is designed to assess
the political stability of a country, including the ability of the government to carry out its
declared program and ability to stay in o¢ ce.
And as long as the political risk does not
contribute to repayment risk, creditors should be indi¤erent.8
insert Table 10 around here
3.3
Determining Factors of Market Access
In this section we consider the important macroeconomic variables that may in‡uence the
duration of capital market exclusion by distinguishing between the forces of (i) better domestic
behavior, (ii) greater demand for external risk, and (iii) market expectations.
Better domestic behavior may be captured by a number of macroeconomic variables including growth, in‡ation, and as Uribe (2006) notes, "certain monetary-…scal regimes [where]
the risk of default and thus the emergence of sovereign risk premiums are inevitable". The
capacity to pay problem is tied to the issue of debt sustainability, where debt ratios should be
stable or decreasing over time. Speci…cally, the primary surplus, the …scal de…cit excluding
interest payments, necessary to stablize debt ratios is considered by this literature.9
While
the use of the primary surplus measure faces criticism over its simplistic nature and its inability to say what is a "safe" debt ratio for a country (see Goldstein (2003) for an overview of
the criticisms), more advanced extensions of the debt sustainability models still look at the
primary surplus a country needs to achieve. Using the central government primary balance
as measures of good domestic behavior we would predict that larger primary surpluses, or at
least positive primary balances, would lead to faster market reaccess.
7
See Bloomberg for sovereign credit ratings at the time of debt issuance. We map the existing ratings from
S&P into the ratings by Institutional Investor, which exist for many countries starting in 1979.
8
One may consider, for instance, the two-month strike by the state-owned oil company, PDVSA, and attempted coup to overthrow President Chavez, which resulted in a signi…cant decline in GDP between 2002 and
2003. Despite the political upheaval, Venezuela continued servicing its external debt throughout this period
and the Venezuela EMBI spread declined.
The required primary surplus to keep debt ratios stable can be calculated using the formula: P S = r1+gg dt ,
where r is the real interest rate, g is the real growth rate, and dt is the debt to GDP ratio being targeted.
9
9
Better domestic behavior may also be captured by the existence of an IMF program, which
typically focuses on macroeconomic policies and reforms in public …nance, the …nancial sector,
and statistics collection as well as building capacity through technical assistance (Mumssen,
2008).
Thus, we would expect that the existence of an IMF program would show market
participants that the country was moving in a positive direction and result in shorter exclusion
periods.
To measure external …nancial market conditions and investor demand for sovereign debt
we focus our attention on the interest rate spread between risk free assets (US Treasury yields)
and riskier assets such as developing country sovereign debt (proxied by high yield returns).
We believe that this can capture global credit concerns given that international investors are
largely US-based, or at least use US assets as benchmarkets in pricing risks and returns in
international …nancial markets (Hartelius et al., 2008). Following the …ndings of Hartelius et
al. (2008) that US interest rates have an e¤ect on emerging market debt spreads, implying that
the US can reduce the risk of disruption to emerging debt markets, we anticipate that tighter
spreads indicate greater investor demand and overall market liquidity would lead to shorter
periods of market exclusion.
We also look at the level of US interest rates, which sets an
implicit ‡oor on the interest rate that defaulters face when accessing markets. We anticipate
that higher US interest rates, either short-term or long-term rates, will lead to longer exclusion
periods.
Market expectations can be captured by credit ratings, which aim to measure the forwardlooking estimate of default probability of a national government on its obligations (Beers and
Cavanaugh, 2006).
The economic rationale for credit ratings are twofold.
First, they pro-
vide information economies of scale - it is e¢ cient for creditors and investors in initiating and
monitoring transactions because of the economies of scale achieved in gathering and analysing
information.10
Secondly, ratings help formulate a simple and veri…able rule with low trans-
action costs so as to be able to monitor and constrain the actions of agents (Gonzalez et al.
(2004). Further, ratings can contribute towards determining the …nancial cost of issuing debt
and the quality of the investor base. Speci…cally the Institutional Investor country credit rating surveys senior economists and sovereign risk analysts at global banks, money management,
and securities …rms, rating countries in terms of likelihood of default. Thus we would expect
that countries with higher credit ratings to regain market access faster.11
Finally, we consider the size of the country, in terms of nominal GDP.12 We would predict
that large countries regain market access more quickly than small countries due to their relative
importance in providing signi…cant investment opportunities for investors.
Large countries
10
One can think of this as helping creditors or investors to minimize the "lemon problem". See Akerlof (1970)
for a discuss of this problem. For a good discussion of the determinants and impact of country credit ratings,
see Cantor and Packer, 1996.
11
We use the Institutional Investor country credit rating due to data limitations imposed by S&P and Moody’s
country coverage. Prior to the mid-1990s both rating agencies did not provide ratings to countries that were
below investment grade.
12
We de…ne the largest 10% of countries, in terms of USD nominal GDP, as "big".
10
tend to have larger nominal debt stocks, which allows for more liquid debt instruments and
typically leads to higher weightings in asset class indices.
3.4
Measuring the Duration of Market Exclusion
In this section we start by looking at the unconditional survival and hazard functions for the
duration of market exclusion and then proceed with our analysis by estimating a discrete time
duration model with time varying regressors in order to analyze the impact of some of the
variables previously identi…ed in section 3.3.
Our analysis going forward is based on 106
default episodes (out of 128 events originally) due to data limitations.
3.4.1
Preliminary Analysis
As a preliminary analysis of the data on market exclusion we start by presenting the empirical
survival and hazard functions for the duration of market exclusion.13
In order to estimate
the survival functions we use a non-parametric estimator that is very popular in the duration
literature, the Kaplan-Meier estimator. This estimator is de…ned as follows:
t
Q
SbKM (t) =
j=1
1
dj
nj
;
(1)
where dj denotes the number of exits in the j-th period and nj denotes the total number of
possible exits in the j-th period. The estimator for the hazard function follows immediately
from the survival function estimator and it simply uses a fundamental relationship between
the hazard and the survival functions
b
b
bKM (t) = SKM (t 1) SKM (t) = dt ;
nt
SbKM (t 1)
(2)
where dt and nt have the same interpretations as before. The results for the survival function
can be seen in Figure 1.
Figure 1 around here
Figure 2 around here
From the initial analysis of the empirical survival functions, we see that 50% of defaulters
regain partial market access within 3 years, while it takes 7 years for 50% of defaulters to
13
The survival function is de…ned as S (t) = 1 F (t), where F (t) is the cumulative distribution function.
This function tells us what percentage of the population is still in the state after t periods, in our case, it tells
us the percentage of countries that have not regained market access after t periods. The hazard function is
f (t)
de…ned as (t) = S(t)
, where f (t) is the density function. This function tells us the instantaneous probability
of exiting a state at time t conditional on not having exited after t periods. In this case it states that the
probability of a country regaining market access after t years conditional on not having got access until then.
11
regain full market access. Regaining partial access occurs relatively quickly compared to full
market access; in less than 7 years 75% of defaulters were able to borrow from abroad again.
In order for 75% of defaulters to regain full market access it takes 11 years, which suggests
that creditors remember defaults and make obtaining large quantities of external funds quite
di¢ cult. The question that this raises, but we are not able to answer with the data that we
are analyzing, is to know whether the fact of a country not being able to borrow large amounts
of money from abroad for fairly long periods of time is a su¢ cient punishment for default or
whether the country does not feel the cost of being …nancially constrained.
In Figure 2 we present the empirical hazard functions.
What we learn from this graph
is that the speed at which countries that have not yet regained market access regain market
access is non-decreasing over time.14 This means that, over time for those countries that have
not regained market access, the probability of being able to access the market again does not
decrease (in some cases it actually increases) from period to period.
In the case of partial
market access, in each of the …rst 8 years, close to 20% of the countries that are excluded from
the market are able to access it again.
In the case of full market access, we see a similar
pattern, with the speed of access fairly constant during the initial 8 years, around 10% per
year, and then it increases somewhat.
From the analysis of the hazard functions, the result that should be seen as the most
concerning one from an incentives point of view is the fact that a country can simply wait and
the odds of being able to regain market access are in its favor.
That is, as time passes, it
becomes more likely that a country will be able to again have access to international capital
markets.
The analysis thus far has been unconditional and this does not tell us anything
about the sources of di¤erences between the di¤erent events.
3.4.2
Econometric Analysis
We proceed with our analysis by specifying and estimating a discrete time duration model with
time varying covariates, for both partial and full market access, in order to understand the
quantitative impact of di¤erent factors a¤ecting the duration of market exclusion. The bene…ts
to this approach are the fact that it allows us to incorporate episodes in which market access
has not yet occurred (censored observations) and for the interaction between the duration of
exclusion and the evolution of the variables previously identi…ed that can potentially impact the
length of market exclusion.15 The parametric speci…cation we consider here is a proportional
hazard model with time varying covariates.
This means that, in our model, the hazard
function is the product of two elements: the baseline hazard function,
14
0 (t),
and some factor
Besides the economic interpretation inherent to this feature, there is also an important statistical interpretation that this result signals, which is the fact that there aren’t any signals of major sources of unobserved
heterogeneity which is something we use in the next subsection of our econometric approach. For a detailed
discussion of this topic see Lancaster (1990).
15
Note that the approach that we are following here is, to some extent, similar to the one adopted in Gelos
et al. (2004). These authors use a probit model to analyze the same phenomena, but they do not give it an
hazard interpretation nor use direct measures for the length of market exclusion.
12
of proportionality that varies with the covariates, g (xit ; ).16 We assume
0 (t)
to be a piece-
wise constant, which leads to the model proposed by Prentice and Gloeckler (1978) for grouped
data.17
insert Table 11 around here
Based on the results presented in Table 11 several conclusions can be drawn: i) the factors
that explain partial market access are not the same that explain full market access. In the
case of partial market access, external risk seeking behavior matters the most, captured by
the spread between US Treasuries and high yield assets as well as the level of short-term US
interest rates.
For full market access, the most important factors are market expectations
and good domestic behavior, captured by the Institutional Investor country credit rating as
well as the central government primary balance and existence of an IMF program; ii) good
domestic behavior is rewarded while poor domestic behavior is not punished. To capture this
asymmetric e¤ect we separately consider central government primary surpluses and de…cits.
Achieving a primary surplus is signi…cant in reducing the full market exclusion period, while
running a primary de…cit is insigni…cant in lengthening the period of full market exclusion;
iii) size matters for regaining market access. We see that bigger countries can regian market
access almost twice as quickly, certeris paribus, as a small country. Our interpretation is that
investors face a …nite number of countries to invest in and cannot excluse themselves from
large markets for long periods of time, particularly those that are heavily weighted in asset
class indices;18 and iv) the Africa and Middle Eastern region take signi…cantly longer to regain
market access than other regions.
This holds for both partial and full market access and
highlights the fact that there are regional di¤erences between borrowers. However, we do not
…nd any statistical di¤erence between defaulters in the Emerging Europe, Latin America and
the Caribbean, and Asia regions.
We next consider how the type of debt instrument defaulted on - bank or bond debt can impact the duration of market exclusion by reestimating our previously de…ned models
to separately consider the type of defult.
By separating events by the type of default we
raise the issue of creditor knowledge of debtors: bank debt creditors typically monitor debtors
more closely, at a cost, than bonded debt creditors do, which may contribute to the relative
importance placed on the di¤erent variables.
insert Table 12 around here
16
Notice that in the speci…cation of this model, t denotes elapsed time and not historical time, that is, it
represents the amount of time during which a country did not have market access after settling its default and
not the chronological time of market exclusion.
17
See Appendix 4 for a derivation of the likelihood function.
18
This makes sense when one considers that most institutional investors’performance is compared to benchmark index returns.
13
For bank debt defaults, size matters for both partial and full market access, with big
countries regaining market access faster. Market expectations matter for full access but good
domestic behavior is not rewarded.
External risk seeking behavior is signi…cant.
Finally,
we …nd that the Africa and Middle Eastern region regains market partial market access more
slowly, but …nd that the duration of full market exclusion is not signi…cantly longer than other
regions.
When looking at bond defaults, we …rst acknowledge that our results are limited by the
small sample size. Interestingly, we do not …nd that size matters for regaining market access
after bond defaults.
This may be due to the fact that bond investors tend to view smaller
countries as enhancing portfolio diversi…cation.
Again, good domestic behavior does not
appear to be a signi…cant driver for market reaccess, however, the existence of an IMF program
is signi…cant for partial market reaccess. Market expectations prove to be signi…cant for both
partial and full market access, which we anticipated due to its monitoring e¢ ciency. Finally,
there does not appear to be any regional di¤erence for regaining market access after defaulting
on bond debt, which may follow for the same reason of portfolio diversi…cation.
The fact that a certain variable has a positive or a negative impact on the hazard function
(or on the average duration) is not very informative because it is not easy to translate the value
of an estimated coe¢ cient into more tangible information, like an average or an elasticity. For
this reason we decided to construct the following experiment. Suppose that during the periods
of market exclusion, instead of having an average Institutional Investor country credit rating
of 23.6, the average credit rating was 34.7 or 12.4, what would be the implied average market
exclusion duration? Based on this idea we estimated the average duration of market exclusion
that is implied by the econometric model assuming that the observed average of the variable
being analyzed is higher/lower by a fraction of its standard deviation, x
% stdev (x). The
results of this experiment are summarized in Figure 3 for partial market access, and in Figure
4 for full market access.19
Note that our base case scenario is a small country and African
countries we consider are small.
Figure 3 around here
Figure 4 around here
From our experiments, we reach the following conclusions regarding partial market access:
i) small countries and African countries bene…t more from variation in spreads and interest rate
levels than big countries do. A one standard deviation decline in the average spread between
US interest rates and high yield assets can lower the average duration of partial exclusion by
1.4 years for small countries and 3.2 years for African countries, while big countries only see a
19
See Table A.1 in Appendix 1 for the summary statistics of the di¤erent variables during the periods of
market exclusion.
14
decrease in the market exclusion period of 0.3 years; ii) the level of the risk free interest rate
and the interest rate spread have a very similar impact on the length of exclusion periods; iii)
big countries regain partial market access almost twice as fast as small countries (1.4 years
versus 3.6 years); and iv) African countries take more than twice as long as small countries in
other regions to regain partial market access, averaging 7.9 years.
Regarding full market access we conclude the following: i) small countries and African
countries have greater incentive to improve their domestic behavior. A 1 standard deviation
improvement in the primary surplus (+2.9% of GDP) reduces market exclusion by 3.6 years
for African countries and 1.7 years for small countries.
Big countries only see a reduction
of market exclusion by 0.9 years; ii) market expectations matter more for small and African
countries. A big country with a credit rating 1 standard deviation below the average rating
takes only 2.8 more years than a country that is 1 standard deviation above the average rating
(7.9 years versus 5.1 years). In the case of small country, this di¤erence becomes 5.2 years
(14.1 years versus 8.9 years) and for African countries the di¤erence is 11.7 years (29.3 years
versus 17.6 years).
These results tell us that big countries have less incentive to undertake
domestic reforms or work to improve market expectations because the potential upside is much
less than it is for small countries.
4
Concluding Remarks
This paper examines why some countries are able to regain access to international capital
markets immediately after resolving a default, whereas other countries appear to be punished
for long periods.
Our main …ndings from examining the duration of exclusion from inter-
national capital markets between 1980-2005 by sovereign defaulters are: i) countries regain
partial market access after 5.7 years on average (median of 3.0 years) while it takes 8.4 years
on average (median of 7.0 years) to regain full market access; ii) partial market access depends
mostly on external demand for risk; iii) full market access depends primarily on good domestic
behavior and market expectations; iv) size matters, with large economies regaining market
access twice as fast as small countries; and v) there are regional di¤erences, with African and
Middle Eastern defaulters taking twice as long to regain market access when compared to other
regions.
Our results complement earlier …ndings by Gelos et al. (2004) in that we …nd that the
quality of policies perceived by the market matter substantially as captured by the Institutional
Investor country credit rating and central government primary balance.
However, we …nd
signi…cantly longer periods of capital market exclusion. Our …ndings also support the recent
work of Fostel and Kaminsky (2007) in that global liquidity and country size are important
drivers of market access.
There are many interesting extensions of this work that can be undertaken. First, we would
like to utilize our resulting stylized facts to establish a theoretical model of market access.
Second, we would extend our analysis of post default market reaccess to consider periods of
15
market exclusion resulting from …nancial crises and …nancial contagion.
By understanding
the loss of market access under various circumstances we would like to be able to formulate
speci…c policies to assist countries in the market reaccess process.
Third, we believe that
our …ndings can be incorporated into recent work on sovereign debt, which is beginning to
endogenize the international capital market reaccess process.
The addition of this friction
may lead to di¤erent results, particularly with regards to the emerging market business cycles
literature and explanations of interest rate spreads.
Finally, it may be useful to study the
composition of debt ‡ows during a default cycle to see how a country is able to get around
the closure of capital markets, which may explain the low associated costs of capital market
exclusion.
16
5
References
1. Akerlof, G. 1970. The market for "lemons": quality uncertainty and the market mechanism. Quarterly Journal of Economics 84(3): 488-500.
2. Arellano, C. 2008. Default risk and income ‡uctuations in emerging economics. American Economic Review, forthcoming.
3. Arteta, C. and G. Hale. 2008. Sovereign debt crises and credit to the private sector.
Journal of International Economics 74:53-69.
4. Arteta, C. and G. Hale. 2007. Currency crises and foreign credit in emerging markets:
credit crunch of demand e¤ect? Federal Reserve Bank of San Francisco Working Paper
2007-02.
5. Beers, D., Cavanaugh, M., 2006.
Sovereign Credit Ratings: A Primer.
Standard &
Poor’s RatingsDirect, October 19.
6. Beers, D., Chambers, J., 2006. Default Study: Sovereign Defaults At 26-Year Low, To
Show Little Change In 2007. Standard & Poor’s CreditWeek, September 18.
7. Benjamin, D. and Wright, M. L. J. 2008.
Recovery before redemption? A theory of
delays in sovereign debt renegotiations. University of California, Los Angeles, mimeo.
8. Bi, Ran. 2008. "Bene…cial" delays in debt restructuring negotiations. International
Monetary Fund Working Paper No. 38.
9. Cantor, R., Packer, F., 1996.
Determinants and Impact of Sovereign Credit Ratings.
Federal Reserve Board of New York Policy Review, October 1996.
10. Cavanaugh, M., S. Bugie, and E. Dubois-Pelerin. 2008. Corporate and counterparty
credit ratings that exceed the sovereign’s rating. Standard & Poor’s RatingsDirect, June
5.
11. Chuhan, P., Sturzenegger, F., 2004.
Default Episodes in the 1980s and 1990s: What
have we Learned? in: World Bank (Ed.), Managing Volatility and Crisis: A Practitioners
Guide. Mimeo.
12. Dhillon, A., García-Fronti, J., Ghosal, S., Miller, M., 2006.
Debt Restructuring and
Economic Recovery: Analyzing the Argentine Swap. World Economy 29(4): 377-398.
13. Diamond, D. W., 1991. Monitoring and Reputation: The Choice Between Bank Loans
and Directly Placed Debt. Journal of Political Economy 99(4): 689-721.
14. Eaton, J. 1992. Sovereign debt: a primer. World Bank Policy Research Working Paper
No. 855.
17
15. Eaton, J., Fernandez, R., 1995. Sovereign Debt in: Grossman, G. M., Rogo¤, K. (Eds.)
Handbook of International Economics (3). Amsterdam: North-Holland.
16. Eaton, J., Gersovitz, M., 1981.
Debt with Potential Repudiation: Theoretical and
Empirical Analysis. Review of Economic Studies 48(2): 289-309.
17. Eichengreen, B. 1994. Financing infrastructure in developing countries: lessons from the
railway age.
University of California, Berkley Economics Department Working Paper
No. 230.
18. Esquivel, G. and F. Larraín.
1998.
Explaining currency crises.
John F. Kennedy
School of Government Faculty Research Working Paper No. R98-07.
19. Feinland Katz, L. 2005. Ratings above the sovereign: foreign currency rating criteria
update. Standard & Poor’s RatingsDirect, November 3.
20. Fostel, A., Kaminsky, G., 2007. Latin America: Access to International Capital Markets.
Good Behavior or Global Liquidity? NBER Working Paper 13194.
21. Gelos, R. G., Sahay, R., Sandleris, G., 2004. Sovereign Borrowing by Developing Countries: What Determines Market Access? IMF Working Paper 04/221.
22. Goldstein, M. 2003. Debt sustainability, Brazil, and the IMF. Institute for International
Economics Working Paper No. 03-1.
23. Gonzalez, F. F. Haas, R. Johannes, M. Persson, L. Toldeo, R. Violi, M Wieland, and
C. Zins.
2004.
Market dynamics associated with credit ratings, a literature review.
European Central Bank Occassional Paper Series No. 16.
24. Hartelius, K., K. Kashiwase, and L. E. Kodres. 2008. Emerging market spread compression: is it real or is it liquidity? IMF Working Paper 08-10.
25. International Monetary Fund, 2005.
Assessing the Determinants and Prospects for
the Pace of Market Access by Countries Emerging from Crises: Further Considerations.
Prepared by the International Capital Markets Department.
26. International Monetary Fund, 2003. Access to International Capital Markets for FirstTime Sovereign Issuers. Prepared by the International Capital Markets Department.
27. International Monetary Fund, 2001. Assessing the Determinants and Prospects for the
Pace of Market Access by Countries Emerging from Crises.
Prepared by the Policy
Development and Review Department.
28. International Monetary Fund and World Bank. 2006. Applying the debt sustainability
framework for low-income countries post debt relief.
18
29. Jenkins, S. P. 1995. Easy estimation methods for discrete-time duration models. Oxford
Bulletin of Economics and Statistics 57(1): 129-138.
30. Kohlscheen, E. and S. A. O’Connell. 2007. Trade credit, international reserves and sovereign debt. The Warwick Economics Research Paper Series (TWERPS) 833, University
of Warwick, Department of Economics.
31. Lancaster, T, 1990. The Econometric Analysis of Transition Data. Econometric Society
Monographs 17, London: Cambridge University Press.
32. Lensink, R., Van Bergeijk, P., 1991. The Determinants of Developing Countries’Access
to the International Capital Market. Journal of Development Studies 28(1): 86-103.
33. Levy-Yeyati, E. and U. Panizza. 2006. The elusive costs of sovereign defaults. InterAmerican Development Bank Working Paper 581.
34. Lindert, P. H., Morton, P. J., 1998.
How Sovereign Debt Has Worked in: Sachs, J.
(Ed.) Developing Country Debt and Economic Performance (1). Chicago: University
of Chicago Press.
35. Mumssen, C. 2008. IMF conditionality: theory and practice. International Monetary
Fund, African Department.
36. Pitchford, R. and Wright, M. L. J. 2007. Restructuring the sovereign debt restructuring
mechanism. University of California, Los Angeles, mimeo.
37. Prentice, R. L., Gloeckler, L. A., 1978. Regression Analysis of Grouped Survival Data
with Application to Breast Cancer Data. Biometrics, 34, 57-68.
38. Reinhart, C. M., Rogo¤, K. S., Savastano, M. A., 2003.
Debt Intolerance.
NBER
Working Paper No. 9908.
39. Reinhart, C. 2002. Sovereign credit ratings before and after …nancial crises. In Levich,
R., C. M. Reinhart, and G. Majnoni, (eds.) Ratings, Rating Agencies and the Global
Financial System. New York: Kluwer Academic Press: 251-268.
40. Rose, A. 2005. One reason countries pay their debts: renegotiation and international
trade. Journal of Development Economics 77(1): 189-206.
41. Sturzenegger, F., Zettelmeyer, J., 2007.
Haircuts: Estimating Investor Losses in Sov-
ereign Debt Restructurings, 1998-2005. Forthcoming in Journal of International Money
and Finance.
42. Sturzenegger, F., Zettelmeyer, J., 2006a.
Creditor Losses versus Debt Relief: Results
from a Decade of Sovereign Debt Crises. IMF preliminary draft.
19
43. Sturzenegger, F., Zettelmeyer, J. 2006b.
Has the Legal Threat to Sovereign Debt
Restructuring Become Real? Universidad Torcuato Di Tella, Centro de Investigación de
Finanzas. Mimeo.
44. Tomz, M., Wright, M. L. J., 2007. Do Countries Default in "Bad Times"? Journal of
the European Economic Association 5(2-3): 352-360.
45. Trebesch, C. 2008. Delays in debt restructurings. Should we really blame the creditors?
Free University of Berlin, mimeo.
46. Uribe, M. 2006. A …scal theory of sovereign risk. Journal of Monetary Economics 53:
1857-1875.
47. Yue, V. Z. 2006.
Sovereign default and debt renegotiation.
mimeo.
20
New York University,
Tables
in yrs
FC Bank Debt
FC Bond Debt
Total Debt
Average
5.9
2.0
4.9
Median
5.0
2.0
5.0
N
12
4
16
Average
8.5
3.0
6.9
Median
7.0
2.0
5.0
N
30
12
42
Average
9.2
2.5
8.5
Median
6.5
1.5
5.5
N
54
6
60
Average
5.9
1.0
5.3
Median
5.0
1.0
2.0
9
1
10
Average
8.3
2.6
7.3
Median
6.0
2.0
5.0
N
105
23
128
Emerging Europe
Latin America & Caribbean
Middle East & Africa
Asia-Paci…c
N
Full Sample
Table 1 - Summary default statistics
in yrs
Partial
Full
Average
5.7
8.4
Median
3.0
7.0
N
128
128
Table 2 - International capital market exclusion periods
21
Duration
Frequency
Percent
Cumulative
1
59
46.09
46.09
2
13
10.16
56.25
3
11
8.59
64.84
4
8
6.25
71.09
5
4
3.13
74.22
6
6
4.69
78.91
7
5
3.91
82.81
8
6
4.69
87.50
9
3
2.34
89.84
10
4
3.13
92.97
11
2
1.56
94.53
12
2
1.56
96.09
13
1
0.78
96.88
14
1
0.78
97.66
15
1
0.78
98.44
16
1
0.78
99.22
17
0
0.00
99.22
18
0
0.00
99.22
19
0
0.00
99.22
20
0
0.00
99.22
21
0
0.00
99.22
22
0
0.00
99.22
23
0
0.00
99.22
24
1
0.78
100
Total
128
100
Table 2A - Duration of market exclusion, partial access
22
Duration
Frequency
Percent
Cumulative
1
37
28.91
28.91
2
13
10.16
39.06
3
8
6.25
45.31
4
11
8.59
53.91
5
4
3.13
57.03
6
7
5.47
62.50
7
8
6.25
68.75
8
7
5.47
74.22
9
8
6.25
80.47
10
7
5.47
85.94
11
5
3.91
89.84
12
2
1.56
91.41
13
2
1.56
92.97
14
2
1.56
94.53
15
2
1.56
96.09
16
2
1.56
97.66
17
0
0.00
97.66
18
1
0.78
98.44
19
0
0.00
98.44
20
0
0.00
98.44
21
0
0.00
98.44
22
0
0.00
98.44
23
1
0.78
99.22
24
1
0.78
100
Total
128
100
Table 2B - Duration of market exclusion, full access
Emerging Europe
Latin America & Caribbean
Middle East & Africa
Partial
Full
Partial
Full
Partial
Full
Partial
Full
Average
3.4
4.7
3.1
6.0
8.2
11.8
4.1
5.4
Median
2.0
3.0
2.0
5.0
5.0
9.0
4.0
8.02 0
17
17
41
41
61
61
9
9
in yrs
N
Asia-Paci…c
Table 3 - International capital market exclusion periods, by region
20
This result is driven by one censored event which takes at least 8 years to regain full market access. However,
after 2 years 60% of countries have regained full market access.
23
Smallest 25th
in yrs
Second 25th
Third 25th
Largest 25th
Partial
Full
Partial
Full
Partial
Full
Partial
Full
Average
7.8
9.4
3.9
8.9
4.6
6.0
3.4
5.1
Median
6.0
9.0
3.0
9.0
3.0
6.0
2.0
4.0
N
36
38
35
35
27
23
30
32
Table 4 - International capital market exclusion periods, by country size
Smallest 25th
in yrs
Second 25th
Third 25th
Largest 25th
Partial
Full
Partial
Full
Partial
Full
Partial
Full
Average
3.4
4.8
2.9
6.0
3.3
7.9
6.2
8.3
Median
3.0
4.0
1.0
4.0
3.0
7.0
4.0
7.0
N
17
13
12
16
21
17
78
82
Table 5 - International capital market exclusion periods, by haircut size
1980s
in yrs
1990s
2000s
Partial
Full
Partial
Full
Partial
Full
Average
5.5
8.2
4.1
5.8
2.5
3.7
Median
4.0
8.0
2.0
6.0
3.0
3.0
N
79
79
32
32
17
17
Table 6 - International capital market exclusion periods, by decade of default
Smallest 25th
in yrs
Second 25th
Third 25th
Largest 25th
Partial
Full
Partial
Full
Partial
Full
Partial
Full
Average
2.6
5.0
3.2
5.1
5.0
6.1
7.5
11.0
Median
1.0
5.0
2.0
4.0
5.0
6.0
7.0
11.0
N
24
23
30
20
21
23
53
62
Table 7 - International capital market exclusion periods, by external market conditions
IMF Program
No IMF Program
Partial
Full
Partial
Full
Average
4.5
7.4
5.4
7.7
Median
2.0
6.0
4.0
8.0
N
73
64
55
64
in yrs
Table 8 - International capital market exclusion periods, by IMF program
Lowest 25th
in yrs
Second 25th
Third 25th
Highest 25th
Partial
Full
Partial
Full
Partial
Full
Partial
Full
Average
7.7
9.5
5.6
6.5
2.4
6.7
3.9
5.9
Median
7.0
9.0
4.0
5.0
1
7.0
3.0
5.0
N
35
39
33
35
29
27
31
27
24
Table 9 - International capital market exclusion periods, by credit rating
Lowest 25th
in yrs
Second 25th
Third 25th
Highest 25th
Partial
Full
Partial
Full
Partial
Full
Partial
Full
Average
5.4
9.1
5.4
5.9
3.7
7.5
4.5
6.9
Median
4.0
9.0
6.0
4.0
2.0
8.0
3.0
6.0
N
28
35
24
26
33
27
43
40
Table 10 - International capital market exclusion periods, by political risk
25
Partial Market Access
Full Market Access
1:0893
Constant
1:9800
(1:027)
( 1:237)
1:0380
Big, Dummy
0:7180
(3:582)
(2:025)
0:0153
CG Primary Balance, if >0
CG Primary Balance, if <0
(1:974)
0:0004
0:0189
(0:093)
Institutional Investor rating
(0:254)
0:0117
0:0262
(1:033)
IMF program, dummy
(1:974)
0:3682
0:1696
(1:502)
Spread
(0:622)
0:6287
0:2117
( 2:466)
3mo UST
0:0685
( 0:405)
( 0:895)
0:1887
( 2:175)
0:0567
10yr UST
( 0:450)
Africa, Dummy
0:7758
0:8385
( 1:9117)
Latin America, Dummy
( 1:649)
0:2596
( 0:675)
Asia, Dummy
0:7132
( 1:373)
T2
0:5800
1:4358
0:6152
0:9884
0:4091
1:1244
(2:418)
(2:638)
T4
(2:071)
(2:006)
T5
0:4549
( 0:660)
(2:881)
T3
0:2641
( 0:605)
(2:024)
0:5396
0:5727
(3:239)
(1:483)
T6
1:2519
T7
0:7828
(1:9366)
(1:241)
T8
1:9908
T9
2:5125
(1:790)
(1:647)
1:7996
T10
(2:085)
Schwarz B.I.C.
Log Likelihood
# Observations
210:570
249:617
166:941
336
186:757
537
1) T-statistics in parenthesis; 2) *, **, *** denote signi…cance at 10%, 5%, and 1% respectively.
Table 11 - Duration model results
26
Partial Market Access
Bank default
Bond default
1:8039
Constant
(1:602)
6:2868
( 1:758)
1:3276
Big, Dummy
0:2059
(3:295)
CG Primary Balance, if >0
CG Primary Balance, if <0
(0:302)
0:0292
0:0549
(0:831)
(0:582)
0:0269
0:3148
1:4840
0:2094
0:5075
0:4678
0:2898
(2:022)
(2:434)
0:1675
( 2:524)
(0:251)
0:2328
0:1573
(1:825)
(0:682)
( 1:727)
0:0425
0:8325
0:8517
(0:834)
0:1385
( 0:310)
0:374
( 0:384)
0:7454
(0:313)
0:7473
( 1:324)
0:1312
( 0:266)
0:0291
( 0:084)
0:6489
(0:378)
0:0181
(0:010)
0:5427
1:0283
1:2980
2:9325
0:5254
1:3583
0:6243
4:1355
0:4384
0:8095
0:7039
(1:618)
(0:833)
0:5105
1:2079
0:4233
2:9611
(1:282)
(2:227)
(1:546)
(1:871)
T5
(0:751)
(0:133)
(2:588)
T4
(2:050)
0:1002
( 1:125)
T3
0:1004
(0:733)
( 1:757)
T2
0:1799
( 0:836)
0:1337
( 0:336)
Asia, Dummy
(1:392)
0:0063
10yr UST
Latin America, Dummy
0:2648
0:1418
0:0172
( 2:494)
Africa, Dummy
6:5771
( 1:012)
( 0:207)
0:0253
(0:081)
Bond default
(2:045)
0:0341
0:7994
3mo UST
0:8242
(0:185)
(1:104)
Spread
1:6726
( 1:009)
0:0146
(0:491)
IMF program, dummy
Bank default
( 0:689)
(0:500)
Institutional Investor rating
Full Market Access
(0:804)
(2:654)
(1:468)
(2:197)
(1:687)
(1:202)
(0:820)
(0:891)
5:776
(0:661)
1:0908
T6
(1:713)
T7
0:8722
T8
2:3924
(1:230)
(1:682)
2:6132
T9
(1:452)
1:9997
T10
(1:753)
Schwarz B.I.C.
173.036
58.325
207.018
64.209
Log-likelihood
-130.829
-29.902
-145.925
-32.947
278
58
450
87
# Observations
1) T-statistics in parenthesis; 2) *, **, *** denote signi…cance at 10%, 5%, and 1% respectively.
Table 12 - Duration model results by default type
27
Figures
1.0
Full Access
0.9
Partial Access
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0
1
2
3
4
5
6
7
8
9
10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Analysis Time
Figure 1 - Empirical survival function estimates
1.0
Full Access
0.9
Partial Access
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
1
2
3
4
5
6
7
8
9
10
11
12 13
14
15
16
17
18
19
20
Figure 2 - Empirical hazard function estimates
28
21
22
23
24
Market liquidity - Spread
Risk free interest rate
17
Average market exclusion
(in years)
Average market exclusion
(in years)
17
15
13
11
9
7
5
15
13
11
9
7
5
3
3
1
-1
-0.8
-0.6
-0.4
-0.2
1
0
0.2
0.4
0.6
0.8
1
-1
-0.8
-0.6
Deviation from average
(number of standard deviations)
Small country
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
Deviation from average
(number of standard deviations)
Big country
African country
Small country
Big country
African country
Figure 3 - Impact of statistically signi…cant variables on the average duration of market
exclusion, partial
II Rating
31
31
28
28
Average market exclusion
(in years)
Average market exclusion
(in years)
Primary surplus
25
22
19
16
13
10
7
25
22
19
16
13
10
7
4
4
1
-1
-0.8
-0.6
-0.4
-0.2
1
0
0.2
0.4
0.6
0.8
1
Deviation from average
(number of standard deviations)
Small country
Big country
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
Deviation from average
(number of standard deviations)
African country
Small country
Big country
African country
Figure 4 - Impact of statistically signi…cant variables on the average duration of market
exclusion, full
29
1
Appendix 1 - Summary statistics
In this appendix we present the average and standard deviation for all the variables used in
our models during the periods between default settlement and market access for both partial
and full market access.
Partial Market Access
Variables
Full Market Access
Average
Std. Deviation
Average
Std. Deviation
CG Primary Balance
-0.01
5.18
-0.03
4.70
CG Primary Balance, >0
3.37
3.51
3.42
3.44
CG Primary Balance, <0
-3.65
3.80
-3.14
3.31
Rating
23.55
11.18
24.46
11.03
IMF_all
0.54
0.50
0.54
0.50
Spread
1.93
0.80
1.83
0.76
3mo UST
4.69
2.39
-
-
10yr UST
-
-
6.40
1.88
dummy_afr
0.50
0.50
0.49
0.50
dummy_lac
0.29
0.46
0.34
0.47
dummy_asia
0.08
0.28
-
-
big
0.09
0.29
0.09
0.29
# Observations
336
Table A.1 - Summary statistics
30
537
Appendix 2 - Discussion of Market Access Measure
We believe that using net bank and bond debt transfers to public and private creditors is a
good way to measure access to international capital markets. To test this, we check whether
during years of a sovereign default, net transfers are negative, which implies no market access.
We exclude the starting year of default because due to annual data we cannot tell whether
the transfers occurred before the default episode began. We also exclude the year of reaccess
if it is in the same year of exiting default, because, again, the transfers could occur after the
default episode ended.
We …nd that approximately 10% of the sample shows positive net transfers during bank
default episodes, and substantially less for bond default episodes.
These measures are not
weighted by the number of events, but are weighted by the length of the event because potentially a country experiencing a longer period of default can have more years of positive net
transfers than those countries experiencing short periods of default.
# Observations
Percentage
Bank default&public>0% of GDP
87
10.3
Bank default&private>0% of GDP
56
6.6
Bank default&sum>0% of GDP
88
10.4
Bank default
848
Bond default&public>0% of GDP
4
6.8
Bond default&private>0% of GDP
3
5.1
Bond default&sum>0% of GDP
4
6.8
Bond default
59
Table A.2 - Positive net transfers during default years
We also include some …gures for a graphical representation of our measures ability to match
default years.
31
% of GDP
8
Argentina
% of GDP
Cote D'Ivoire
6
4
6
4
2
Default
2
Default
private borrow
public borrow
0
0
private borrow
public borrow
-2
sum_borrow
sum_borrow
-2
-4
-4
-6
-6
-8
1980 1983 1986 1989 1992 1995 1998 2001 2004
% of GDP
1980 1983 1986 1989 1992 1995 1998 2001 2004
Pakistan
% of GDP
2.5
Moldova
8
1
2
0.9
6
1.5
0.8
0.7
4
1
Default
0.5
private borrow
0
2
public borrow
0
sum_borrow
-0.5
-1
0.6
Default
0.5
private borrow
0.4
public borrow
0.3
sum_borrow
0.2
-2
0.1
-1.5
-4
-2
0
1980 1983 1986 1989 1992 1995 1998 2001 2004
1980 1983 1986 1989 1992 1995 1998 2001 2004
Figure A.2 - Net public and private transfers by private creditors
32
Appendix 3 - Derivation of the Likelihood Function
The econometric model that we use in the paper is the piece-wise constant proportional hazard
model. This model is de…ned as follows: let
hazard function, where
(t) exp (X 0 ) be the conditional
(t jX; ) =
(t) is the baseline hazard and exp (X 0 ) is proportional factor that
control the e¤ect of the regressors X on the hazard. The main characteristic of the piece-wise
constant hazard model is that its baseline hazard is de…ned by pieces, that is, the hazard varies
with time but it is constant within certain ranges:
(t) =
8
>
>
>
>
<
>
>
>
>
:
1
if 0 < t
1
2
if 1 < t
2
(:::)
J
if t > J
Using the de…nition of the baseline hazard we can write the conditional survival function:
S (t jX; ) = exp
(
Z
Z
(u) du =
0
(u) du exp X 0
0
t
with
t
PJ
j=1
Pt
j=1
j
+ (t
j
if t
J)
J
J
if t > J
The model described above assumes constant covariates, and therefore, in order to allow
for time-varying covariates it is necessary to describe the method by which we are able to
use time varying covariates. The way we incorporate time varying covariates in our model is
through "event split". The idea of event split was proposed by Jenkins (1995) and the best
way to describe is through an example. Let’s suppose that we are interested in analyzing the
time between …rst employment and a house purchase. Let’s assume further that one of the
explanatory variables is a dummy variable indicating whether the individual is married or not.
As it is obvious, this variable can change over time. The way to incorporate a change in the
marital status (from singe to married) is by considering 2 events instead of only one. The …rst
event is the time between …rst job and marriage and the second is the time between marriage
and purchase of house conditional on not having purchased the house until that moment. The
…rst event, when the individual is still single, is going to be considered a censored event, as
we do not observe the purchase of a house while single. The second event is also a censored
event, but in this case a left censor event for which we know the time the event started. If we
assumed that it took t1 periods until the purchase of the house but at t0 < t1 the individual
got married, we can write the individual contribution to the likelihood as follows:
f (t jmarital status ) = S (t0 jsingle )
33
f (t jmarried )
:
S (t0 jmarried )
In our particular case the variables we use are not always discrete variables, but for those that
are continuous, we assume that the value of the variable did not change during the year. Using
the previous information, we can write the individual contribution to the likelihood as follows:
8 h
< Qt 1
j=1
Li ( jX (t) ; t ) =
: Qt
ih
S(jjX(j); )
f (tjX(t); )
S(j 1jX(j); )
S(t 1jX(t); )
S(jjX(j); )
j=1 S(j 1jX(j); ) if censored
i
if non
censored
This expression can be simpli…ed further:
S (j jX (j) ; )
S (j 1 jX (j) ; )
Rj
exp
(u) du exp X (j)0
0
=
Rj
exp
1
(u) du exp X (j)0
0
Z
= exp
j
(u) du exp X (j)0
j 1
= exp
= 1
j
exp X (j)0
h (j jX (j) ; )
Also, since we are working with yearly data, we can assume that the data is discrete, and
therefore, f (t jX (t) ; ) = S (t
1 jX (t) ; )
f (t jX (t) ; )
S (t 1 jX (t) ; )
=
S (t jX (t) ; ). This implies:
S (t
= 1
= 1
1 jX (t) ; ) S (t jX (t) ; )
S (t 1 jX (t) ; )
S (t jX (t) ; )
S (t 1 jX (t) ; )
0
exp
t exp X (t)
= h (t jX (t) ; )
Now, if we de…ne d = 1 if the event is right censored, that is, the end is not observed, we
can re-write the individual contribution to the likelihood as follows:
Li ( jX (t) ; t ) =
nYt
1
j=1
[1
o
h (j jX (j) ; )] [1
h (j jX (j) ; )]d [h (j jX (j) ; )]1
And the individual contribution to the log-likelihood function is:
li ( jX (t) ; t ) =
Xti
1
j=1
(ln [1
di ln ([1
(1
h (j jX (j) ; )]) +
h (j jX (j) ; )]) +
di ) ln h (j jX (j) ; ) :
34
d
Appendix 4 - Data Description
Default events. Standard and Poor’s de…nition of start and end date.
Political risk rating.
Provided by the PRS Group, available from 1985-2005.
Scale of
1-100 with 100 being least risky and results below 50 considered a very high level of risk to
the political stability in the country.
Institutional Investor Country Credit Rating.
Based on information provide by senior
economists and sovereign risk analysts at global banks, money management, and securities
…rms. Each country is rated on a scale of 0 to 100, with 100 representing those countries that
have the least chance of default.
IMF program.
Taken from the IMF webpage for History of Lending Arrangements.
A
country must have a program for at least 5 months out of 12 to be considered as having a
program in a given year. We consider just the existence of a Stand-By Agreement as well as
any type of program, including SBA, EFF, Structural Adjustment Facility, and PRGF.
Macroeconomic variables. Taken from GDF and IFS. Missing data is …lled in as much as
possible using Inter American Development Bank, African Development Bank, Asian Development Bank, and European Bank for Reconstruction and Development resources.
Partial market access. The …rst year of net positive bank or bond transfers from private
creditors to either the public and publicly guaranteed sector or the private sector.
This
de…nition excludes arrears from being considered as a positive transfer to the debtor. Data
is available from GDF using the following series: DT.NTR.PNGB.CD; DT.NTR.PNGC.CD;
DT.NTR.PBND.CD; DT.NTR.PCBK.CD.
Full market access. The …rst year of 1% of GNP net positive bank or bond transfers from
private creditors to either the public and publicly guaranteed sector or the private sector. This
de…nition excludes arrears from being considered as a positive transfer to the debtor. Data
is available from GDF using the following series: DT.NTR.PNGB.CD; DT.NTR.PNGC.CD;
DT.NTR.PBND.CD; DT.NTR.PCBK.CD.
Haircut. We use Benjamin and Wright’s (2008) database of private creditor haircuts on
90 defaults and renegotiations.
HIPC status. Determined by the World Bank HIPC Initiative. Only Completion Point
and Decision Point countries are considered.
Region.
Determined
by
the
CIA
World
Factbook.
https://www.cia.gov/library/publications/the-world-factbook/
Country size. We use nominal US$ GNP from the GDF, series: NY.GNP.MKTP.CD.
Population. United Nations Population Division, World Population Prospects: 2006 Revision.
Data is available in 5-year increments and we use linear interpolation to …ll in the
missing years.
US Treasury rates. Global Financial Data - CBOE 30-year US Government Bond Yield
Index.
35
High Yield rates.
Global Financial Data - Barron’s Intermediate-Grade Bond Yield, an
index of 10 medium-grade corporate bonds.
Spread. Di¤erence between high yield and US Treasury rates. Used to measure liquidity
and investor risk appetite. A smaller spread indicates higher liquidity and demand for riskier
assets.
36