ABSTRACT
Title of Dissertation: INTERNATIONAL MONETARY FUND.
PROGRAMS AND CAPITAL MARKET ACCESS
Diego Saravia Tamayo, Doctor of Philosophy, 2004
Dissertation directed by: Professor Carmen Reinhart
Department of Economics
This thesis studies how International Monetary Fund (IMF) loans interact
with private capital  ows and how they afiect the level of welfare of borrower
countries and private lenders.
The flrst chapter presents a model highlighting the fact that the IMF has
both de jure and de facto seniority rights over private creditors. It is shown that
IMF lending afiects borrowers and lenders in difierent ways. Ex-post, once the
initial borrowing decisions have been made, an IMF intervention always make
the borrower country better ofi. The efiects on private lenders depend on the
size of the senior intervention and on what they expect to get in case that the
IMF does not intervene. For some parameter values, IMF interventions make
existing lenders worse ofi when the liquidity situation is either good or weak and
make them better ofi when it is in an intermediate range. This is consistent with
the empirical evidence presented in Chapter 2. The expectation of a future IMF
intervention may reduce the level of borrowing and borrowers? welfare ex-ante,
because seniority allows the IMF to lend in cases where it is not socially optimal
to do so. This efiect is contrary to the moral hazard view where \too much"
rescuing leads to \too much" borrowing. Thus, the country may have incentives
to commit today not to borrow tomorrow from the IMF in the future, although
this promise is not time consistent.
The second chapter, which is a joint work with Ashoka Mody, analyzes empir-
ically if IMF programs in uence the ability of developing country issuers to tap
international bond markets and whether they improve spreads paid on the bonds
issued. It is found that Fund programs do not provide a uniformly favorable
signaling efiect. Instead, the evidence is most consistent with a positive efiect of
IMF programs when they are viewed as likely to lead to policy reform and when
undertaken before economic fundamentals have deteriorated signiflcantly. The
size of the Fund?s program matters, but the credibility of a joint commitment by
the country and the IMF appears to be critical.
INTERNATIONAL MONETARY FUND.
PROGRAMS AND CAPITAL MARKET ACCESS
by
Diego Saravia Tamayo
Dissertation submitted to the Faculty of the Graduate School of the
University of Maryland, College Park in partial fulflllment
of the requirements for the degree of
Doctor of Philosophy
2004
Advisory Committee:
Professor Carmen Reinhart, Chair/Advisor
Professor I.M. Destler
Professor Enrique Mendoza
Assistant Professor Michael Pries
Professor John Shea
c Copyright by
Diego Saravia Tamayo
2004
DEDICATION
To Rosario, for everything, and to my parents.
ii
ACKNOWLEDGEMENTS
IamextremelygratefultoFernandoBroner, EnriqueMendoza, Michael
Pries, Carmen Reinhart and John Shea for their guidance, comments
and helpful discussions over the process of writing my dissertation.
I have also beneflted from discussions and comments from Marco
Arena, Roger Betancourt, Guillermo Calvo, Mac Destler, Marcela
Eslava, Eduardo Ganapolsky, Federico Guerrero, Pedro Rodriguez,
Roberto Mu~noz and Antonio Spilimbergo. Michael Bordo, Patrick
Conway, Barry Eichengreen, Kristin Forbes, Alessandro Rebucci and
Jeronim Zettlemeyer provide useful comments on the second chapter
of my dissertation. Last but not least, I am grateful to Ashoka Mody
with whom I coauthored the second essay of this thesis.
iii
TABLE OF CONTENTS
List of Tables vi
List of Figures vii
1 On the Role and Efiects of IMF Seniority 1
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3 Equilibrium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.3.1 Period 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.3.2 Period 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.3.3 Period 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
1.4 Empirical Evidence . . . . . . . . . . . . . . . . . . . . . . . . . . 28
1.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2 Catalyzing Private Capital Flows: Do IMF-Supported Programs
Work as Commitment Devices? 32
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.2 Background and Hypothesis . . . . . . . . . . . . . . . . . . . . . 35
2.3 Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
2.4 Methodology and Data . . . . . . . . . . . . . . . . . . . . . . . . 45
iv
2.4.1 Framework for Analyzing Spreads . . . . . . . . . . . . . . 46
2.4.2 Evaluating IMF Programs: Econometric Issues . . . . . . . 50
2.4.3 Descriptive Statistics . . . . . . . . . . . . . . . . . . . . 52
2.5 The Role of Countries? Fundamentals . . . . . . . . . . . . . . . . 53
2.6 Implications Of Fund-Supported Program Design . . . . . . . . . 58
2.7 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
A Appendices 71
A.1 Proof of Lemma 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
A.2 Proof of Lemma 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
A.3 Proof that @^?@k0 < 0 . . . . . . . . . . . . . . . . . . . . . . . . . . 75
A.4 Data Sources and Construction of Variables . . . . . . . . . . . . 77
A.5 Base Regression . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
Bibliography 87
v
LIST OF TABLES
2.1 Frequency of IMF Programs. . . . . . . . . . . . . . . . . . . . . . 66
2.2 Bond Issuance, Terms, and Country Characteristics . . . . . . . . 67
2.3 Interaction of Country Characteristics with Fund Programs . . . . 68
2.4 In uence of Program Features on Bond Market Spreads and Issuance 69
vi
LIST OF FIGURES
2.1 Econometric Implications of Timing of IMF Programs . . . . . . . 70
vii
Chapter 1
On the Role and Efiects of IMF Seniority
1.1 Introduction
The role that the IMF should play in the New International Financial Architec-
ture is an important issue in the current policy and academic debate, especially
after the crisis that took place in the 1990s, beginning with Mexico in December
of 1994. It has been recognized that the IMF has some characteristics that make
it a special player in the international lending community, capable of attract-
ing capital  ows to a country and improving in this way its economic situation.
For example, it is argued that the IMF may have more information than other
lenders and that its presence may be a positive signal about countries? character-
istics that are not observed by other creditors (Rodrik 1996); a related argument
is that the IMF can be used by less informed investors as a country?s screening
device (Marchesi and Thomas 2001). Another hypothesis is that the IMF could
act as a delegated monitor through its conditionality and surveillance functions
or could serve as a country?s commitment device to behave well (for example
1
Rodrik (1996), Tirole (2002), Mody and Saravia (2003)).1
This chapter focuses on a difierent aspect of IMF lending, speciflcally its sta-
tus as senior lender. This focus is motivated by some facts about IMF lending
that have received little analytical attention. These are: (1) countries have shown
a higher aversion to defaulting on IMF loans than on loans from private credi-
tors.2 and (2) the IMF has contractual seniority on its loans. Arguably, these
two characteristics imply two other characteristics of IMF lending: (1) the IMF
lends at a lower interest rate than private creditors, and (2) the IMF lends in
circumstances where other creditors are not willing to do so.
This chapter addresses the following questions: Is IMF seniority good? For
whom? Under what circumstances? Since we are interested in the seniority issue,
we will study the IMF as a creditor of a country with the only difierence being
that it has seniority rights. The crucial distinction in the model is, therefore,
between senior and non-senior lenders. In this chapter, one can think either that
senior lending is realized by lenders acting competitively, or by a deep-pocket
investor, who can make senior loans and in addition chooses to make zero proflts
in expectation. Arguably, this is a realistic assumption about IMF behavior.
The presence of senior lending may introduce a con ict of interest between
non-senior creditors and the debtor country. Consider a country that has been
hit by shocks that prompt a need for new flnancing. It may be the case that
1Cottarelli and Gianini (2002) clasifiy the channels in which  ows are \catalyzed" in flve
categories similar to the mentioned here as an example.
2For example, Argentina, Indonesia, Ecuador, Pakistan and Ukraine have defaulted on pri-
vate debts and not on IMF loans. \...the IMF typically gets paid back (instances of arrears
being the exception to the rule..)" (Eichengreen (2003)).
2
no new lending will be provided without seniority rights; for example a highly
indebted economy would have problems attracting new non-senior funds because
of credit ceiling and debt overhang considerations. A senior lender would have
less problems lending since the probability of being repaid is higher than for non-
senior lenders. Thus, seniority may be a necessary condition to have flnancing
that allows the economy to cope with shocks. However, non-senior lenders may
be worse ofi in the presence of a senior lender since in case of bankruptcy they
have to wait until senior debts are repaid.
The chapter presents a model with three periods: a planning period, a period
when a shock hits the economy, and a flnal period where output is obtained, and
consumption and debt repayment take place. In the planning period, the country
borrows to invest in capital, which is used in the production process in order to
maximize expected utility. In the middle period, the country potentially has to
borrowmore money tocopewithaliquidityshockthathits theeconomy. The way
that the IMF adds value in this model is by lending in circumstances where non-
senior creditors are not willing to lend in equilibrium. If the shock is big enough,
non-senior lenders will expect losses on new lending and, assuming initial lenders
are atomistic and cannot coordinate efiorts to make \emergency loans", will not
be willing to ofier credit; in these cases a deep-pocket lender with seniority rights
(IMF) will be necessary to cope with the shock. Once capital is installed and the
initial lending and borrowing decisions have been made (i.e. ex-post), a senior
intervention always makes the country better ofi, since senior creditors lend at a
lower interest rate, allowing a higher consumption level. The efiects of the IMF?s
lending on non-senior lenders depend on the size of the liquidity shock and on
what non-senior lenders would get when the IMF does not intervene. On the one
3
hand, having senior lending allows the economy to cope with a higher range of
liquidity shocks, but on the other hand, a senior lender jeopardizes what private
creditors expect to get in case of bankruptcy. As a consequence of these opposing
efiects, lenders may prefer to discontinue the project, and would be ex-post worse
ofi with an IMF intervention.
Lenders take into account these efiects when making their initial lending de-
cisions (i.e. ex-ante). It may be the case that the option of a future senior
intervention makes contractual conditions more onerous in the planning period
and that, as a consequence, the country ends up borrowing (and investing) a
lower amount than in the case where the IMF is not allowed to intervene. More-
over, it may be the case that the borrower country would be ex-ante better ofi
by committing not to borrow from the IMF to cope with future shocks, because
seniority allows the IMF to lend in circumstances when it is socially optimal
not to continue with the project, making the country ex-ante worse ofi. Since
the country has incentives to borrow from the IMF once the shock occurs, this
promise is not time consistent and a commitment technology will be necessary
to maintain it.
The chapter is related to the discussion about the role of International Fi-
nancial Institutions as a Lender of Last Resort (LOLR) (for example Fischer
(1999), Zettelmeyer (2000) and Calomiris (1998)). This discussion is often based
on models where a crisis occurs as a self-fulfllling equilibrium caused by coordina-
tion problems between creditors. An important point in this debate is the trade
ofi between ex-post e?ciency and ex-ante moral hazard. Some argue that having
a LOLR institution able to flll liquidity needs reduces the probability of a crisis
and ameliorates their efiects once they occur. Others claim that having a LOLR
4
would trigger debtor and other creditors? moral hazard. Our model abstracts
from coordination and moral hazard issues and adds to this literature in two as-
pects. First, we highlight the point that IMF intervention afiects borrowers and
lenders difierently, and need not lead to ex-post e?ciency, and in fact can create
ex-post socially ine?cient outcomes. Second, contrary to the moral hazard view
that predicts that the possibility of a future bail-out will lead to excessive lending
by making lenders take riskier strategies, our model predicts that the possibility
of a future bail-out may lead to less lending, in equilibrium, as a consequence of
the con ict of interest mentioned above.
Recent theoretical work by Corsetti et al. (2003) studies the role of the IMF
in catalyzing capital  ows by providing liquidity in a model with coordination
problems between creditors having asymmetric information about the state of the
economy.3 In one of the extensions to their model, they consider the case where
the IMF is a senior lender. They conclude that since a senior lender is more
willing to intervene, the probability of a crisis would be reduced, but since the
return to junior lenders is lower they would be less willing to roll over their debts.
As noted above, in our paper, we are not concerned with coordination problems
and roll-over of short term debt issues although we recognize they are important.
Rather, our framework allows us to analyze the impact of senior interventions
on borrowers? and lenders? ex-ante and ex-post welfare, highlighting the con ict
of interest between borrowers and lenders that a senior intervention may imply.
This is something that previous work has abstracted from and it is what allows
3Morris and Shin (2003) use a similar analysis to Corsetti et.al. to analyze the IMF?s ability
to catalyze capital  ows. Penalver (2002) reaches similar conclusions to Morris and Shin?s work
with a difierent modelling strategy. None of these works analyzes the role of IMF seniority.
5
us to generate the result that the amount borrowed and the country?s welfare
may be lower when senior lending is allowed.
Section 2 describes the elements of the model. Section 3 solves the model
backwards, allowing for cases in which senior lending either is or is not allowed.
We flrst examine the efiects of senior intervention on the country?s and private
creditors? welfare ex-post, once capital is installed and the shock hits the economy.
We then study how the possibility of a senior intervention afiects the initial level
of investment and the country?s welfare ex-ante. Section 4 relates this work to
the empirical evidence presented in the second essay of this thesis. Section 5
concludes.
1.2 Model
Time. There are three periods, indexed by t=0,1,2. In period 0, agents make
real investment and borrowing decisions. In period 1, the economy can be hit
by a shock that afiects the production process. In order to cope with this shock,
agents have to borrow again. In period 2, output is realized, debt issued in period
0 and 1 is repaid and consumption takes place.
Agents and production. The economy is populated by a continuum of identical
consumer-producers with linear preferences over consumption of a single good at
date 2; i.e their utility function is U(c0;c1;c2) = c2. The production process has
a time-to-build aspect: investment is realized in period 0 and 1 and output is
realized in period 2. It is assumed that agents do not have any endowment of
goods in period 0 and 1, so they have to borrow from abroad in order to import
goods used as inputs in the production process. In period 0, agents borrow to
install capital, k0, which will be depreciated totally at the end of period 2.
6
To avoid borrower?s moral hazard considerations, we assume that investment
is veriflable, or alternatively, that there is no storage technology available, so that
the amount borrowed has to be invested in the production process.
Following Holmstrom and Tirole (1998) and Caballero and Krishnamurthy
(2001) we introduce a liquidity shock in period 1 as a production shock that
the economy has to cope with by borrowing additional funds. Let ? be the
aggregate liquidity shock that hits the economy in period 1. Agents will need a
reinvestment of ?k0 to continue the project. If they do not reinvest this amount,
then the project cannot continue and a scrap value, S(k0), is obtained in period
2. S is assumed to be quasiconcave, increasing in k0 and satisfles S ? k0.
Assume ? is a random variable distributed between [0;1] with cumulative
distribution function G(?). In order to introduce market incompleteness, we
assume that ? is observable but not veriflable, so that contracts in period 0
cannot be made contingent on realized values of the shock in period 1. We do
not consider idiosyncratic shocks since we are interested in cases in which the
economy as a whole needs liquidity, and we are not concerned with heterogeneity
between residents.
If reinvestment is made in period 1, then the project continues and output in
period 2 is ?f(k0), where ? is a random productivity shock distributed between
[0;??] with cumulative distribution F(?), and where f(k0) is a concave function.
It is assumed that E(?)f(k0) > k0; otherwise, investors will not invest in period
0.
7
period 0 period 1 period 2
k0 -?k0 -?f(k0)
@
@
@
@
@
@
@
@@R
S(k0)
reinvest
not reinvest
? ? G[0;1]
? ? F[0;??]
Financial contracts. As noted above, residents have to borrow from abroad
in order to produce. This is an ability-to-pay model with no deadweight losses
associated with bankruptcy. That is, when realized output is lower than debt
face value or when the project is discontinued, lenders can seize output or the
scrap value.
It is assumed that debt issued in period 0 and debt issued in period 1 both
mature in period 2. International lenders are risk neutral, act in a competitive
environment and have enough wealth to provide liquidity to the country when
needed. Clearly, for any amount lent they will charge a positive interest rate
since the default risk is positive (remember that the minimum value that ? can
take is zero).
Without loss of generality, it is assumed that the gross international interest
rate is equal to 1. At date 0 domestic agents borrow an amount L0 (equal to
k0) and agree to pay a total amount of D0 (i.e. initial amount borrowed plus
interest) in period 2. At date 1 they borrow an amount L1 (equal to ?k0) whose
face value in period 2 is D1.
8
1.3 Equilibrium
In what follows we will solve the model backwards beginning with period 2. In
period 1, when the shock hits, we will consider what happens when a senior
lender(s) is allowed in that period. Then we will consider period 0.
1.3.1 Period 2
In period 2, if reinvestment has been made in period 1, output is realized, debt
is repaid, and consumption takes place. Consumption will be greater than zero
if and only if output is greater than the total face value of debt contracted in
period 0 (D0) and in period 1 (D1), which occurs when:
?f(k0)?D0 ?D1 > 0
or, equivalently:
? > D0 +D1f(k
0)
? ??: (1.1)
Thus, total debt will be repaid and consumption will be positive if and only if
the productivity shock is higher than a threshold value ??.
Assumption 1. When no senior lender is allowed, in case of default (i.e. ? <
??) the proportion of output that goes to each creditor equals the share of his loan
in total loans, i.e LiLi+L?i.
That is, absent seniority, creditors have equal footing on output in case of
bankruptcy. We have not assumed that the share of output going to each creditor
is equal to the share of his debt in total debt, i.e. DiDi+D?i, for simplicity and
because, if this were the case, second period debt could be made efiectively senior
9
by having a high enough D1. Since LiLi+L?i need not be the same as DiDi+D?i, it
is possible that the output due to a creditor in case of default is higher than his
debt face value. To rule this out, assume:
Assumption 2. In case of default, if LiLi+L?i?f(k0) is greater than Di then lender
i gets Di.
Thus, a creditor?s repayment in period 2 will be the maximum of his contrac-
tual value of debt and his share of output under the equal footing scheme.
If reinvestment has not taken place in period 1, the scrap value of the project,
S(k0), is divided between creditors, and consumption is equal to zero (remember
that by assumption S(k0) < k0 and, consequently, S(k0) < D0).
1.3.2 Period 1
Atthe beginning of this period the random variable ?is observedand the economy
inherits installed capital (k0) and a stock of debt contracted in period 0 (D0).
Agents need to borrow ?k0 in order to continue the project. Since it is assumed
that if reinvestment is not made the project ends and consumption is zero, the
borrower country will always want to reinvest as long as the highest possible
output level is higher than the total value of debt. So the demand for loans is
determined by the size of the shock.
Supply of loans under equal footing
As noted above, international capital markets are competitive and the interna-
tional gross interest rate is equal to 1. Competition between lenders will ensure
that expected proflts from lending to the country will be zero.
10
Deflne ?1 as the threshold productivity level above which period 1 lenders?
output share, computed under equal footing, is greater than their contractual
debt value,
?1 ?
?L
0 +L1
L1
? D
1
f(k0);
or equivalently, since L1 equals ?k0 and L0 equals k0:
?1 ?
?1+?
?
? D
1
f(k0): (1.2)
Similarly, deflne ?0 as the threshold value above which period 0 lenders? output
share is greater than D0:
?0 ? [1+?] D0f(k
0)
: (1.3)
This last expression follows from the fact that [1+?] is equivalent to
h
L0+L1
L0
i
.
Note that
h
?
1+?
i
?1 +
h
1
1+?
i
?0 = ??, so that the threshold productivity shock
above which all debts are repaid (??) is a weighted average of ?1 and ?0. When ?1
is lower than ??, it means that D1 is totally repaid when the productivity shock
is at least ?1; for productivity shocks between ?1 and ??, D0 holders get output
in excess of D1; and when the productivity shock is higher than ??, output is
enough to repay both D0 and D1. A comparable analysis holds when ?0 is lower
than ??. Also, note that ?0 will be higher than ?1 if and only if the interest rate
charged on period 0 loans is higher than the interest rate charged in period 1;
both interest rates are determined in equilibrium below.
Thus, period 1 lenders? zero proflt condition under equal footing satisfles:
?k0 =
? ?
1+?
?Z min(?1;?0)
0
?f(k0)dF(?)+
Z ??
min(?0;??)
[?f(k0)?D0]dF(?)+
+
Z ??
min[?1;??]
D1dF(?): (1.4)
11
The right hand side is period 1 lenders? expected repayment from investing
in the country and the left hand side is the amount lent. Alternatively, we can
express the same condition in terms of each unit lent:
1 =
? 1
1+?
?Z min(?1;?0)
0
?f(k0)
k0 dF(?)+
1
?
Z ??
min(?0;??)
??f(k
0)
k0 ?
D0
k0
?
dF(?)+
+
Z ??
min[?1;??]
r1dF(?); (1.5)
where r1 = D1?k0 is the gross interest rate charged to the country by international
lenders.
Lemma 1. The interest rate r1 is increasing in the amount lent.
Proof in Appendix A.1.
So, the higher the period 1 shock is, i.e. the higher the amount needed to
continue the project, the more expensive, per dollar, it will be for the borrower
to continue.
Proposition 1. If and only if
Z ??
0
?f(k0)2k
0
dF(?)+
Z ??
min[?0;??]
?1
2
?f(k0)
k0 ?
D0
k0
?
dF(?) < 1; (1.6)
there is a set of liquidity shocks su?ciently close to 1 for which no credit is
supplied in period 1 under equal footing.
Proof. A necessary and su?cient condition to have lending in period 1 that
satisfles the zero proflt condition under equal footing is:
?k0 ? ?1+?
Z ??
0
?f(k0)dF(?)+
Z ??
min[?0;??]
? 1
1+??f(k0)?D0
?
dF(?): (1.7)
This is because, given the loan size (?k0) and the value of debt issued in period
0 (D0), period 1 lenders? expected repayment is increasing in D1; and the right
12
hand side of (1.7) is lenders? expected repayment when the value of D1 is high
enough that total debt (D1 +D0) is greater than or equal to the highest possible
repayment (??f(k0)).4 If condition (1.7) is not satisfled then period 1 creditors
will expect losses on any loan of size ?k0. The set of values for ? satisfying (1.7)
is not empty. The right hand side is unambiguously greater than the left hand
side for values of ? near zero since R??0 ?f(k0)k0 dF(?) is greater than one.
Since the flrst term of the right hand side of (1.7) is a continuous, increasing
and concave function of ? and the second term is continuous and decreasing in
?, a necessary and su?cient condition to have a range of liquidity shocks where
expected proflts are negative is that (1.7) is not satisfled when ? is equal to one.
So, if condition (1.6) holds, there will be a threshold value of ? strictly less than
one above which expected proflts to lenders are negative. Since the expected
repayment function is increasing and continuous in D1, there will be a value of
D1 such that expected repayment equals the loan size.
In what follows we assume that condition (1.6) holds, in which case there is
a ^? less than 1 that satisfles:
^?k0 = ^?1+ ^?
Z ??
0
?f(k0)dF(?)+
Z ??
min[?0;??]
? 1
1+ ^??f(k0)?D0
?
dF(?) (1.8)
such that for ? > ^? there will be no lending under equal footing. A su?cient
condition to have ^? < 1 is that (1.6) is true even in the case where D0 is equal to
k0, which is the lowest possible interest rate on period 0 debt and thus the case
most likely to favor lending in period 1. Therefore, a su?cient condition for (1.6)
4If D1 + D0 > ??f(k0), then ?? > ?? and ?1 > ??. Thus, the left hand side of (1.7) follows
from replacing ?? by ?? in the left hand side of (1.4), taking into account that the third term
vanishes.
13
to hold is:
E(?)f(k0)2k
0
+
Z ??
min[ 2k0f(k0);??]
??f(k
0)
2k0 ?1
?
dF(?) < 1:
Note that it may be in the interest of period 0 lenders, as a group, to lend in
period 1 at an expected loss in order to protect their initial claims. However, any
individual lender will be better ofi if the other lenders provide liquidity allowing
the project to continue. That is, there is a con ict between private and collective
interests; each period 0 lender has an incentive to ?free-ride?.5 This free rider
problem has been discussed in the sovereign debt literature; see for example
Krugman (1988) and Eichengreen (2002).
Clearly, creditors that have not lent in period 0 do not have any incentive to
lend at an expected loss in period 1. In this essay we assume that lenders are
atomistic, act in a purely competitive market and cannot coordinate actions to
pursue their collective interests (i.e. the free-rider issue is severe).6
Senior Lender allowed in period 1
Consider the case where a senior lender(s) is allowed to intervene in credit markets
in period 1. The concept of seniority is relevant when contractual obligations
cannot be totally satisfled; i.e. in the case of bankruptcy. If this is not the case,
5The best way to coordinate creditors? actions in the case of a debt crisis, in order to
overcome the free-rider problem, is an important issue in current policy and academic debate
about the way to construct the New International Financial Architecture.
6In a recent speech Anne Krueger stated: \...These far-reaching developments in capital
markets over the last three decades have not been matched by the development of an orderly and
predictable frameworkfor creditor coordination. Because the creditorcommunityisincreasingly
diverse and difiuse, coordination and collective action problems result when scheduled debt
service exceeds a country?s ability to pay" (see IMF survey April 2000).
14
there is no con ict of interest between creditors and the concept of seniority is
not important.
Since senior creditors have priority on output in case of default, they do not
consider the stock of existing debt when making their own lending decisions.
Lemma 2. Senior lenders are willing to lend for any shock ?.
Proof: Senior lenders are willing to lend any amount up to E(?)f(k0), which
is greater than ?k0, for all ?, by previous assumption.
Thus, senior lenders are willing to lend in more states of nature than non-
senior creditors; seniority allows the economy to overcome more severe liquidity
shocks.
Let Ds1 be the value of debt owed to a senior creditor; the threshold produc-
tivity shock above which senior lenders are totally repaid is:
?s ? D
s
1
f(k0): (1.9)
If the productivity shock is lower than this threshold value, senior creditors will
not be totally repaid and non-senior creditors will get nothing. The interest rate
charged by a senior lender satisfles:
1
Ls1
Z ?s
0
?f(k0)dF(?)+
Z ??
?s
rs1dF(?) = 1; (1.10)
where Ls1 and rs1 are the amount lent by a senior creditor and the interest rate
charged, respectively. The interest rate charged by a senior lender will not be the
same as that charged by a non-senior one. In particular:
Lemma 3. For a given sized loan, the interest rate charged by a senior lender is
lower than that charged by a lender without seniority rights.
15
Proof in Appendix A.2.
This result implies that total expected consumption in period 2 is higher when
a senior lender intervenes and, consequently, the country is ex-post better ofi (i.e.
conditional on k0) under seniority. Obviously, borrowers prefer to pay less for a
given amount lent.
At the beginning of period 1 there is a stock of debt issued in period 0 (D0)
that matures in period 2. The period 1 value of this stock of debt will be afiected
by the size of the liquidity shock and by the nature (senior or non-senior) of
period 1 lenders.
To see the impact of a senior intervention on the period 0 lenders? position,
we have to consider whether the liquidity shock is greater or less than ^?, the
threshold value above which non-senior creditors are unwilling to lend.
Consider flrst the case when ? < ^?. In this situation non-senior lenders are
willing to lend to the borrower country and a senior intervention will make period
0 lenders worse ofi. To see why this is the case note that output is divided in
period 2 between the country, period 0 and period 1 creditors. At the beginning
of period 1, the expected value of output is given, since with ? < ^? the project
will continue whether period 1 lenders are senior or not. Meanwhile period 1
lenders, independent of their seniority rights, set the price of the new debt (r1
or rs1) so that expected repayments in period 2 are equal to the size of the loan
(?k0), by the zero proflt condition.
Since expected output and expected repayment to period 1 lenders are the
same with and without senior lending, but expected consumption is higher in
the flrst case, it must be the case that period 0 lenders? expected repayment (or,
equivalently, the period 1 value of their claims) is lower under a senior interven-
16
tion. A senior lender does not add value when the country is able to flnance the
liquidity shock using non-senior sources, but instead merely transfers resources
from period 0 debt holders to the country. So, a senior intervention when ? < ^?
reduces the period 1 price of the debt issued in period 0.
Consider now the case where ? > ^?. In this case, the only way to flnance the
liquidity shock is by issuing senior debt.
To see how senior lending afiects existing creditors in this situation, we com-
pare the period 1 value of existing debt with and without seniority. When senior
lending is not allowed, the project is cancelled and the scrap value is obtained.
Since this is an ability-to-pay model, period 0 lenders get the entire scrap value
(remember that we have assumed that the scrap value is less than k0). Let V n
be the period 1 value of D0 when there is no reflnancing, that is:
V n(k0) = S(k0)
and let V s be the period 1 value of D0 when a senior intervention is allowed,
V s =
Z ?B
?s
[?f(k0)?Ds1(?)]dF(?)+
Z ??
?B
D0dF(?)
where
?B ? D0 +D
s
1
f(k0) (1.11)
and
?s ? D
s
1(?)
f(k0):
The period 1 value of debt issued in period 0 is equal to the face value (D0)
times the probability of being fully repaid, which occurs when the productivity
shock is higher than the threshold value ?B, plus what existing creditors expect
to get when output is not enough to cover total contractual obligations. When
17
the productivity shock is between ?s and ?B output is enough to cover senior debt
in full but covers only part of non-senior debt. When the shock is less than ?s,
output is not enough to cover senior debt, and non-senior creditors get nothing.
Deflne the function ?(S;?) as the difierence between the period 1 value of
debt when a senior intervention is allowed and when it is not:
?(S;?) ? V s ?V n:
That is, positive values of ? imply that period 0 lenders are better ofi with a
senior intervention.
? is a function of the liquidity shock and of the scrap value, since both pa-
rameters afiect the present value of debt with and without senior lending. We
have 7:
@?
@? = ?
Z ?B
?s
@Ds1
@? dF(?) < 0
and
@?
@S = ?1 < 0:
Thus, ?(S;?) is a decreasing function in both arguments.
Note that when there is no scrap value (i.e. S = 0), ?(0;?) is greater than
zero for all values of ?. This is because cancellation leaves existing creditors with
zero, while continuation leaves existing creditors with strictly positive expected
returns.8 Also note that if the scrap value were equal to D0, ?(D0;?) is strictly
7The terms derived from the difierentiation of the integration limits cancel each other out.
8The only case when period 0 debt holders expect to get nothing in case of continuation is
when Ds1 is equal to ??f(k0); but in this case senior lenders? expected proflts will be strictly
positive (since k0 is lower than E(?)f(k0)) contradicting the zero proflt condition.
18
negative for all values of ? since cancellation gives period 0 debt holders the full
value of debt with certainty, while a senior intervention reduces the probability
of repayment below one.
Since ?(S;?) is a continuous and decreasing function in both arguments, and
since ?(0;?) > 0 8? and ?(D0;?) < 0 8?, there is for each ? a unique value of S,
denoted by S0(?), where ?(S;?) = 0. The higher the liquidity shock, the lower
the value of S0. We can express this in the following flgure:
-
6
aaa
aaa
aaa
aaa
aaa
aaa
a
aaa
aaa
aaa
aaa
aaa
?(? = 1)
?(? = ^?)
S0(1) S0(^?) D0 S
?(S; ??)
0
Thus, existing creditors? view of senior intervention depends on the size of the
liquidity shock and the project?s scrap value. We can distinguish three situations.
First, when the scrap value is lower than S0(1), a senior intervention will raise
the value of existing debt for all ? > ^?. In this case, the value of liquidation is so
low that even in the worst possible scenario (highest senior debt) period 0 lenders
prefer to continue the projects.
Second, when the scrap value is between S0(1) and S0(^?) there is a set of
liquidity shocks in the vicinity of 1 where a senior intervention makes period 0
debt holders worse ofi. Moreover, there is a set of liquidity shocks close enough
19
(from the right) to ^? where a senior intervention makes period 0 debt holders
better ofi. So, in this zone seniority has ambiguous efiects on existing creditors
depending on the size of the liquidity shock. In particular, there is a nonlinear
efiect of senior intervention on the price of the debt issued in period 0 that is
consistent with the empirical evidence, as will be seen in section 4 below. When
the shock is small (? < ^?) a senior intervention reduces this price (i.e. increases
spreads over the international interest rate); when the shock is not too far above
^?, a senior intervention increases this price; and when the shock is close to 1 the
price is reduced by senior intervention again.
Finally, when the scrap value is higher than S0(^?), a senior intervention always
makes period 0 debt holders worse ofi. Because the scrap value is so high, initial
lenders prefer to get that value for sure rather than continuing the project and
taking the risk of not being repaid.
We can summarize the flndings of this section in the following proposition:
Proposition 2. Conditional on k0, a senior intervention will improve debtors?
situation in all cases since it allows a higher level of consumption. The efiect on
period 0 debt holders depends on ? and S:
? If ? < ^? a senior intervention will always make existing creditors worse ofi.
? If ? > ^? we have three possible scenarios:
1. If S < S0(1) senior lending makes existing creditors better ofi for all
values of ?.
2. If S0(1) < S < S0(^?) existing creditors? situation will improve if ? is
close enough to ^? and will be worsened if ? is close enough to 1.
3. If S0(^?) < S senior lending always makes existing creditors worse ofi.
20
That is, senior lending may afiect borrowers and lenders difierently; in some
cases, it will allow for the continuation of projects when existing creditors would
prefer to liquidate them. In these cases, there is a con ict of interest between the
borrower and the lenders since the former is always willing to flnish the project.
1.3.3 Period 0
Period 0 is the planning period. Borrowers decide how much to invest and borrow
in order to maximize their expected utility (expected consumption in period 2),
and lenders set the price of their loans in order to attain zero expected proflts.
In period 0 individuals have uncertainty about two shocks: the liquidity shock
(?) and the productivity shock (?). That is, expectations have to be taken over
tworandomvariables. Weconsiderthecasewhereallagentshaveperfectforesight
about the nature of future interventions. That is, borrowers and lenders take
their decisions knowing whether interventions in period 1 will be senior or equal
footing.
Equal footing in period 1
Agents make their decisions taking into account that if the liquidity shock in
period 1 is high enough the project will have to be discontinued and there will
be no consumption and only partial debt repayment.
In equilibrium, borrowers in period 0 decide the amount they want to borrow
inordertomaximizetheirexpectedutility, takingintoaccounthowtheirdecisions
afiect the credit conditions they face. Borrowers maximize:
V0 = max
k0
Z ^?(k0)
0
(Z ?
?
??
[?f(k0)?D0 ?D1(?k0)]dF(?)
)
dG(?) (1.12)
21
subject to
k0 =
Z ^?(k0)
0
?? 1
1+?
?Z min(?1;?0)
0
?f(k0)dF(?)+
Z ??
min(?1;??)
[?f(k0)?D1]dF(?)+
+
Z ??
min[?0;??]
D0dF(?)
 
dG(?)+
Z 1
^?(k0)
S(k0)dG(?) (1.13)
and
?k0 =
? ?
1+?
?Z min(?1;?0)
0
?f(k0)dF(?)+
Z ??
min(?0;??)
[?f(k0)?D0]dF(?)+
+
Z ??
min[?1;??]
D1dF(?): (1.14)
V0 is borrowers? expected utility, and ?? ;?1 and ?0 are as deflned above in (1.1),
(1.2) and (1.3) respectively. The outer integral of (1.12) corresponds to expecta-
tions taken over the liquidity shock, recognizing that if ? > ^?(k0) consumption is
zero under equal footing. The inner integral corresponds to expectations taken
over the productivity shock, knowing that consumption will be positive if output
is enough to cover the total value of debt contracted in period 0 and in period 1.
That is, consumption will be positive if and only if ? > ^?(k0) and ? > ??.
Equation (1.13) is the zero expected proflt condition for period 0 lenders who
face uncertainty about both the liquidity shock and the productivity shock. They
know that if ? > ^?(k0), the project will not continue and they will get the scrap
value. If ? < ^?(k0) (i.e. there is no liquidation in period 1) what they expect to
get in period 2 depends on the productivity shock. Analogously with the period
1 lenders? zero proflt condition in equation (1.4), if output is not enough to cover
either D0 or D1, period 0 lenders receive a share 11+? ?i.e. L0L0+L1?of output. If the
proportion of output that corresponds to period 1 lenders allows D1 to be repaid
for output levels lower than that required to cover total debts ?i.e. D0 + D1?,
22
then period 0 debt holders get output minus D1 until output is enough to pay
D0. When output is higher than this amount, they are repaid in full.
Equation (1.14) is lenders? zero proflt condition in period 1 for a given ?, as
analyzed above in equation (1.4).
Integrating equation (1.14) from zero to ^?(k0) and adding this expression to
equation (1.13) we get:
k0 +
Z ^?(k0)
0
?k0dG(?) =
Z ^?(k0)
0
(Z ??
0
?f(k0)dF(?)+
Z ??
??
[D0 +D1]dF(?)
)
dG(?)+
+
Z 1
^?(k0)
S(k0)dG(?): (1.15)
Adding and subtracting R??0 ?f(k0)dF(?) in equation (1.12) we get:
V0 = max
k0
Z ^?(k0)
0
(Z ??
0
?f(k0)dF(?)?
"Z ??
0
?f(k0)dF(?)+
Z ??
??
(D0 +D1)dF(?)
#)
dG(?):
(1.16)
Inserting equation (1.15) into (1.16) we can express the borrower value function
as:
V0 = max
k0
Z ^?(k0)
0
"Z ??
0
?f(k0)dF(?)
#
dG(?)?k0
?
1+
Z ^?(k0)
0
?dG(?)
!
+
Z 1
^?(k0)
S(k0)dG(?):
(1.17)
For simplicity, assume that the scrap function is linear in the investment level;
i.e. S(k0) = sk0. Then, the optimal investment (and borrowing) level under equal
footing satisfles the following flrst-order condition:
Z ^?(k0)
0
?
E(?) @f(k0)@k
0
?
dG(?)+
Z 1
^?(k0)
s dG(?) = 1+
Z ^?(k0)
0
? dG(?)?
?fE(?)f(k0)? ^?k0 ?sk0gG0(^?) @^?@k
0
; (1.18)
where @^?@k0 < 0; that is, the higher the level of investment, the lower the range of
23
liquidity shocks for which continuation in period 1 will be possible without senior
lending. See Appendix A.3 for the proof.
To set the optimal investment level borrowers balance the marginal beneflt,
given by the marginal productivity of capital and by the efiect that one more
unit invested has on the scrap value; and the marginal costs, given by the cost of
investing in period 0, the expected cost of reinvesting in period 1 and the negative
efiect that one more unit of investment has on the threshold value ^?(k0). Since
higher scrap values allow period 0 lenders to ofier better terms (see equation
(1.13)), the optimal level of investment increases in s.9
Senior lending in period 1
Assuming that senior lending is allowed in period 1, the objective function is:
V s0 = max
ks0
Z 1
0
(Z ?
?
?B
[?f(ks0)?Ds0 ?Ds1(?ks0)]dF(?)
)
dG(?) (1.19)
subject to:
ks0 =
Z 1
0
(Z
?B
?s
[?f(ks0)?D1(?ks0)]dF(?)+
Z ??
?B
D0dF(?)
)
dG(?) (1.20)
and
?ks0 =
Z ?s
0
?f(ks0)dF(?)+
Z ??
?s
Ds1dF(?); (1.21)
where the superscript \s" implies that senior lending is allowed; and ?B and
?s are as deflned in (1.11) and (1.9) above. Now individuals choose investment
knowing that the projects will continue in period 1 for all possible values of the
liquidity shock, so the expectation in (1.19) is taken over the whole range of ?.
9Analytically, this follows from applying the implicit function theorem to (1.18), taking into
account that the second order condition is satisfled.
24
Equation (1.20) and equation (1.21) are the zero proflt conditions for period
0 and 1 respectively. Period 0 lenders know that there will not be liquidation
in period 1 and, consequently, they do not consider scrap value in their zero
proflt condition. They know that senior lenders will have priority on output and
they will begin receiving repayment if and only if senior debts are totally repaid.
Equation (1.21) is the same as equation (1.10) above.
As before, integrating equation (1.21) over all possible values of ? and adding
this expression to equation (1.20) we obtain:
ks0 +
Z 1
0
?ks0dG(?) =
Z 1
0
(Z
?B
0
?f(ks0)dF(?)+
Z ??
?B
(Ds1 +Ds0)dF(?)
)
dG(?):
(1.22)
Adding and subtracting R?B0 ?f(ks0)dF(?) in equation (1.19) and plugging equa-
tion (1.22) in the resulting expression, the borrowers? value function is:
V s0 = max
ks0
Z 1
0
"Z ?
?
0
?f(ks0)dF(?)
#
dG(?)?ks0
?
1+
Z 1
0
?dG(?)
?
: (1.23)
Optimal investment satisfles the following flrst order condition:
Z 1
0
?
E(?) @f(k
s
0)
@ks0
?
dG(?) = 1+
Z 1
0
? dG(?): (1.24)
Thus, borrowers balance the expected marginal product of capital with the ex-
pected marginal cost of investing one more unit, given by the marginal cost at
date 0 plus the expected marginal cost of continuation in period 1.
25
Comparison
In this section we compare how the optimal level of investment and borrowers?
welfare is afiected by allowing senior lending in period 1.10 As noted above,
having senior lending allows the project to continue in circumstances where it
otherwise would have had to be liquidated. Although borrowers always prefer to
continue ex-post, non-senior lenders would prefer to liquidate the project if the
scrap value is high enough. In this case, the anticipation of senior lending makes
period 0 lenders ofier more onerous terms in their lending, leading to a lower
level of investment. When the scrap value is low enough, so that period 0 lenders
prefer a senior intervention in period 1, the expectation of the intervention leads
to a higher level of investment.
To see how optimal investment is afiected, compare equation (1.18) and equa-
tion (1.24). First, assume that there is no scrap value in case of liquidation (i.e.
s = 0 in (1.18)). In this case, the term in brackets that multiplies @^?@k0 in (1.18)
is positive (otherwise there will be no investment in period 0), implying that
R1
0 [E(?)f
0(ks
0)??]dG(?) <
R ^?
0 [E(?)f
0(k0)??]dG(?). This inequality can be
expressed as:
E(?)f0(ks0)
?
1? f
0(k0)
f0(ks0) Pr(? ? ^?)
?
< E(?=? > ^?)[1?Pr(? < ^?)]:
Since the flrst term on the left hand side is greater than one (by (1.18)), while
the flrst term on the right hand side is less than one by deflnition, it must be the
case that f0(k0) > f0(ks0), implying that k0 < ks0.
10Since lenders always set the price of their period 0 loans in such a way that expected proflts
are zero, allowing a senior lender in period 1 does not afiect period 0 lenders? welfare ex-ante
as long as lenders are fully informed about the nature of future interventions.
26
In this case borrowers are ex-ante better ofi with a senior intervention. The
intuition is that the only efiect of a senior intervention is to avoid ine?cient
liquidation of the project. We deflne e?ciency as the outcomes that would be
reached under a flrst-best complete contract (contingent on ?) signed in period
0. Under such a contract, liquidation would occur ex-post ifi E(?)f(k0) < ?k0 +
S(k0). If s = 0, liquidation is never e?cient, since ? ? 1.
Note that in this model the expectation of senior lending does not make
individuals take riskier actions, so the increase in borrowing and lending in period
0 is not the consequence of moral hazard but of avoiding ine?cient liquidation.
Now consider the case where the scrap value is difierent than zero. As noted
above, the scrap value makes period 0 credit conditions under equal footing less
onerous, because it represents a positive payofi in case of liquidation. From
equation (1.18) we can see that the higher is s, the higher the level of investment
under equal footing. When s is equal to one, the term in brackets on the right
hand side of (1.18) is less than or equal to zero (see equation (1.8)), and a
comparison of (1.18) and (1.24) yields
E(?)f0(ks0)
?
1? f
0(k0)
f0(ks0) Pr(? ? ^?)
?
> E(?=? > ^?)[1?Pr(? < ^?)]:
In this case we can not rule out the possibility of ks0 being lower than k0.
Note that a higher scrap value increases the ex-ante utility level when senior
intervention is not allowed in period 1. A comparison of (1.17) and (1.23) suggests
that borrowers may be ex-ante better ofi when senior lending is not allowed
in period 1, depending on the size of s. The intuition is that senior lending
guarantees continuation of the project for all sizes of the shock in period 1, even
if for some values of ? it is socially optimal to liquidate. By insuring continuation
of the project even when E(?)f(k0)??k0 < S(k0), senior lending is reducing the
27
social value of the project and making the borrowers worse ofi ex-ante.
Numerical exercise. We present a numerical example to show that for scrap
values su?ciently high it is possible to have a lower level of investment and
welfare when a senior lender is allowed. Consider the case where f(k0) = k0:8, ?
is uniformly distributed in [0,3], ? is uniformly distributed in [0,1], and s = 1. In
this case we obtain that V s0 = 0:12 < V0 = 0:15 and ks0 = 0:32 < k0 = 0:59.
As noted above, there may be circumstances where senior lending creates a
con ict of interest between lenders and borrowers in period 1. Ex-post, lenders
may want liquidation although it is always in borrowers? interest to continue.
Assume borrowers are able to set institutions in period 0 that govern the avail-
ability of senior lending in period 1. If V0 < V s0 , borrowers will allow for senior
lending in period 1, and lenders will set the price of debt, knowing that there will
be senior lending, in such a way that expected proflts are zero.
If V0 > V s0 , borrowers will maximize ex-ante expected utility by committing
not to allow senior lending in period 1. Note that this promise is not time
consistent, since ex-post, borrowers would always prefer senior lending to equal
footing lending in period one. If no commitment technology is available, then
period 0 lenders will set the price of debt anticipating senior intervention in
period 1 and the borrower country will be worse ofi.
1.4 Empirical Evidence
There are several empirical papers that study the efiects that IMF interventions
have on countries? access to capital markets, with varying conclusions among
28
them.11 The second chapter of the thesis studies the efiects of IMF loans on
spreads and on the probability of issuing bonds by emerging markets economies.
The empirical flndings that are related to this work are:
? The impact of IMF lending on spreads depends on the level of countries?
indebtedness. In particular, there is a ?U? shaped efiect on spreads; IMF
intervention raises spreads when the country?s solvency situation is at the
extremes, either solid or weak, and reduces spreads for intermediate levels.
? ?Precautionaryprograms?, inwhichthecountrydoesnotdisbursethemoney
made available by the IMF, reduce spreads and increase the probability of
issuing bonds.
The flrst flnding implies that when the countries? solvency situation is either
good or weak, an IMF intervention raises spreads, while spreads are reduced by
intervention when solvency is in an intermediate range. In our model, the higher
the period 1 (middle period) liquidity shock, the worse is the country?s solvency
situation. The model is able to show that for small liquidity shocks (when non-
senior credit is available) an IMF loan raises spreads; but when shocks are higher
than a threshold value above which non-senior lending is not available, the efiect
on spreads depends on what lenders? expect to get in the case that reinvestment
does not take place (the project?s scrap value in the model). When the scrap
value is in an intermediate range, an IMF intervention will reduce spreads when
the liquidity shock is not too far above the threshold value, and will increase
spreads when the shock is in the upper tail of the distribution. Thus, there is a
nonlinear efiect consistent with the empirical evidence.
11See Cotarelli and Giannini (2002) for a survey.
29
The second empirical flnding is related to our model?s planning period. A
precautionary program is a proxy for the possibility of future interventions, since
it is money that has already been lent to the country but is not being used
(insurance). We have seen that in equilibrium the initial borrowing level and its
cost are afiected by the possibility of a future senior intervention, and that the
model replicates the empirical flnding when the project?s scrap value is not too
high.
1.5 Conclusions
This paper presents a model that emphasizes the efiects of senior lending (such
as IMF lending) on the borrower country?s and on creditors? welfare. When the
shock that hits the economy is big and markets are incomplete, seniority allows
continuation of projects that otherwise would have to be abandoned; in this sense
the IMF completes markets by flnancing liquidity needs when existing creditors
are not willing (or cannot coordinate efiorts) to do so. Ex-post, once the shock
has occurred, an IMF loan would increase borrower welfare by providing cheaper
funds than non-senior lenders, allowing for a higher consumption level. The
efiects on non-senior creditors depend on the size of the shock and on what they
expect to get when projects are discontinued. When non-senior flnancing could
be attracted to the country, a senior intervention makes existing creditors worse
ofi, since it does not improve the country?s repayment capacity but worsens their
relative position. Even when senior lending is necessary to cope with the shock,
other creditors may be worse ofi with an IMF intervention, depending on the size
of the shock and the project?s scrap value.
In the absence of clear rules set ex-ante governing the types of permissable
30
intervention, an institution providing senior lending would have to weigh the po-
tentially con icting wishes of borrowers and lenders, and decide when to intervene
according to whose interests it more closely represents.
The anticipation of a senior lender can make the borrower country ex-ante
better ofi by avoiding ine?cient liquidation. More interestingly, however, the
anticipation of a senior intervention can make the country ex-ante worse ofi and
reduce the investment level. The reason is that the IMF may allow continuation
of the project when it is socially optimal to liquidate, reducing the social value
of the project and making borrowers worse ofi.
This result is the opposite to the standard moral hazard story associated with
IMF interventions. The usual story is that the IMF "rescues" investors too much
ex-post and thus leads to too much investment ex-ante. Here, however, the IMF
may rescue the country too much ex-post and thus leads to too little investment
ex-ante.
It may be the case that the country would maximize expected utility by
committing itself not to borrow from a senior lender to cope with shocks that hit
the economy. Since the country has incentives to borrow from a senior institution
once the shock occurs, this promise is not time consistent.
The IMF could maximize ex-ante utility by intervening if and only if it is
socially optimal to continue, taking into account that sometimes it might be
better not to intervene even if it beneflts the country ex-post.
31
Chapter 2
Catalyzing Private Capital Flows: Do IMF-Supported
Programs Work as Commitment Devices?
2.1 Introduction
Does the International Monetary Fund (IMF or Fund) succeed in its objective of
\catalyzing" capital  ows to developing economies? A not inconsiderable litera-
ture concludes that the answer is \no"|that is, Fund programs do not enhance
countries? access to capital markets and, indeed, a program may actually make
things worse in this respect [for a recent review, see Bird and Rowlands, 2002].
Why would we expect to observe a catalytic efiect? International contracts,
more so than domestic contracts, are incomplete, and foreigners are, therefore,
often unwilling to lend. A Fund program can potentially substitute for missing
contracts and act as a commitment device that improves access to international
capital. The Fund?s role is, in Tirole?s [2002] terminology, that of a \delegated
monitor," mediating between the country and international investors.
This chapter explores the possibility that the delegated monitoring role works,
and successful catalysis occurs, when a credible joint commitment by the country
and the Fund leads to improved prospects for honoring debt contracts. In other
32
words, the catalytic efiect|or the Fund?s \seal of approval"|is not automatic
and the mere presence of a Fund program does not lead to more capital  ows.
Rather, an IMF program is efiective as a commitment device when other available
information does not negate its credibility. As such, the value of the commitment
implied by a Fund program, and its ability to catalyze capital  ows, are likely
to depend on initial country conditions, program design, and the country-Fund
relationship. Our contribution then is to move from a presumption of undif-
ferentiated efiects to identify country, program, and relationship characteristics
that create the conditions for credible commitments and, hence, contribute to
enhanced capital  ows under IMF programs.
We reach four conclusions that outline the conditions under which the market
values the Fund?s role as a commitment device:
? The presence of a Fund-supported program reduces the adverse efiect that
a country?s export volatility has on its access to international markets and
cost of funds. It is as if contracting a Fund program strengthens commit-
ment to repay when volatility is high.
? An IMF program is efiective when foreign exchange reserves and debt levels
make the country vulnerable but have not deteriorated to a point where
their restoration to normal levels within a reasonable time frame has a low
probability. Thus, the Fund catalyzes  ows when, for example, solvency is
not at stake.
? The size of the Fund-supported program matters, but large programs have
often been successful when the money committed has not actually been
used, suggesting that their precautionary deployment can be valuable.
33
? Repeated relationships between a country and the Fund can imply commit-
ment to solve structural problems, but diminishing returns set in as use of
Fund resources is prolonged, suggesting that, beyond a certain point, the
likelihood of improvement in performance begins to be called into question.
Our empirical analysis centers on the ability of Fund programs to help devel-
oping country issuers tap international bond markets and to reduce spreads paid
on the bonds issued. We use an empirical model developed by Eichengreen and
Mody [2001] to evaluate the determinants of international bond issuance and of
spreads charged at the time of issuance. The transactional data used reduce the
severity of the reverse-causality problem|that is, the possibility that observed
outcomes in uence the likelihood of Fund programs. This is so because the feed-
back from an individual bond issue to explanatory country aggregates is likely to
be less serious than when the dependent variable is, itself, a country aggregate
such as growth or capital  ows. At the same time, by allowing a more careful
consideration of timing than was possible in past studies, transactional data at
higher frequency allow us to more precisely consider the rate of issuance and
spreads paid in the period following the initiation of a Fund program and, hence,
further reduce the problem of reverse causation.
In the next section, we provide a brief background of the Fund?s objective in
stimulating capital  ows and its ability to act as a \delegated monitor." We then
review the literature on the impact of IMF programs to identify key substantive
conclusions and methodological issues. This is followed by a description of the
methodology and data. The empirical results deal flrst with the in uence of initial
country conditions and then with the implications of Fund program design. The
flnal section concludes.
34
2.2 Background and Hypothesis
Enhancing its members? access to international capital markets is widely regarded
as an important objective of the International Monetary Fund. Though the
objective is not an explicitly stated purpose in the Fund?s Articles of Agreements,
the  ow of international capital is essential to such stated purposes as the stability
of the international monetary system, e?cient trade, and productive resource use,
and to providing confldence when a member country experiences di?culties with
its balance of payments.1 The Fund?s interest in private international capital
 ows has, moreover, increased over the last decade. Re ecting this evolution, the
Fund?s Managing Director a?rmed in a recent speech:
\Because private  ows are an indispensable source of flnancing for develop-
ment, another crucial function of the IMF?s new Capital Markets Department
will be to strengthen our ability to help countries gain access to international
capital markets [K?ohler, 2001, para. 13]."2
1Article I of the International Monetary Fund?s Articles of Agreement lists a number of
objectives ("purposes") for the Fund. These include international monetary cooperation, fa-
cilitation of international trade to enable productive use of resources, exchange rate stability,
establishment of a multilateral system of payments, and giving confldence to its members by
making available the general resources of the Fund to permit "correction of maladjustments"
in their balances of payments without a high cost to the domestic or international economy.
2Each member country is required by Article IV of the Articles of Agreement to, among
other things, foster orderly growth, price stability, and orderly monetary and flnancial condi-
tions. Article IV authorizes the Fund to oversee compliance of member countries with these
obligations. The Fund is asked|and has agreed in the past|to monitor and certify a coun-
try?s policy program without any commitment of resources. A distinction may be made, in this
context, between the role of the Fund?s stafi and that of its Board. A positive stafi report can
signal to investors a professional judgment that the country has a credible adjustment program.
35
The Fund?s monitoring function is critical to the catalytic role it plays. Fund
resources do help and Fund flnancing can signal confldence in the course the
country is charting. But it is the signal that counts and hence the Fund?s knowl-
edge of, and confldence in, the country?s policies is necessary to induce private
capital  ows. The Fund?s website describes its role in these very terms:
\In most cases, the IMF, when it lends, provides only a small portion of a
country?s external flnancing requirements. But because the approval of IMF lend-
ing signals that a country?s economic policies are on the right track, it reassures
investors and the o?cial community and helps generate additional flnancing from
these sources. Thus, IMF flnancing can act as an important lever, or catalyst, for
attracting other funds. The IMF?s ability to perform this catalytic role is based
on the confldence that other lenders have in its operations and especially in the
credibility of the policy conditionality attached to its lending."3
Tirole [2002, p. 99] refers to such a role as \delegated monitoring." The
IMF, Tirole argues, acts to \substitute for the missing contracts between the
Sovereign and individual foreign investors and to thereby help the host country
to fully beneflt from its capital account liberalization." Tirole notes that missing
contracts are not just a problem when foreigners lend to the sovereign. The
problem is serious even when the lending is to private domestic borrowers. The
ability of private borrowers to repay is a function of a variety of government
actions that are unpredictable and can de facto expropriate foreign lenders.
The Board?s approval sends a signal that the international community is prepared to support
the country?s program. At least in theory, these two signals can be distinct and separable. The
Fund?s Board "expressed some degree of reservation" about unbundling policy certiflcation (or
"enhanced surveillance") from the use of its resources [Boughton, 2001, p. 413].
3http://www.imf.org/external/publs/ft/exrp/what.htm
36
In this paper, we examine the implications of Fund programs for capital mar-
ket access. A program combines resources and surveillance, with difierent pro-
grams ofiering difierent combinations of these functions. Programs are typically
contracted in periods of external economic imbalances, though the extent of the
imbalance varies and countries can, and do, enter programs for \precautionary"
reasons. For a Fund program to catalyze new private capital  ows, it must cred-
ibly convey a signiflcant likelihood of success|an improvement in the program
country?s external payments position and growth prospects. Success of a pro-
gram, in turn, depends on several factors. In particular, country and global
market conditions in uence the outcome. To deal with this heterogeneity, Fund
programs difier in design (e.g., size of resources, duration, and the nature and
extent of program conditionality). In addition, program outcomes depend on
country-Fund relationships, re ected, for example, in the frequency of programs.
Withrespecttocountryconditions, animportantconsiderationinthesovereign
debt literature has been the volatility that a country is exposed to. In an early
contribution, Eaton and Gersovitz [1981] argued that when countries are exposed
to a high degree of volatility, they are more likely to repay their external debt
since failure to do so would close them ofi from international borrowing and thus
prevent them from dampening the future efiects of continued volatility. But coun-
tries with high volatility may also flnd it more di?cult to repay debt|or may be
able to use the fact of the volatility to claim inability to repay debt. In assessing
these countervailing forces, Catao and Sutton [2002] flnd that macro volatility is
a strong predictor of sovereign debt defaults. Thus, under volatile conditions, a
commitment device should help. A Fund program is a joint commitment. From
the country, it is a commitment to good policies, and from the Fund, to provide
37
resources that serve as a substitute for a country?s reserves.
Commitment through the Fund, however, is likely to be efiective when coun-
tries are vulnerable but have not yet crossed thresholds that imply inability to
service external debts even with Fund assistance. When vulnerability is high,
the role of a \delegated monitor" may be especially valuable if a country?s com-
mitment to international contracts is more suspect than in \tranquil" or more
normal periods. Also, in periods of vulnerability, information about the country
may be fuzzy. However, when a country is past the point of vulnerability|when
reserves and external debt levels have reached levels that imply low probabil-
ity of reversing into a more normal state|the country?s ability and incentives
to achieve policy objectives are suspect and the Fund?s leverage is likely to be
limited.4 Thus, for example, a Fund program is unlikely to catalyze new capital
when solvency is at stake. Even if a country does not "gamble for resurrection,"
as some have argued [e.g., Powell, 2002], new shocks will continue to prevent
recovery. In such a situation, the Fund as a delegated monitor will add limited
value.
In recent theoretical contributions, Morris and Shin [2003] and Corsetti, Giu-
mares, and Roubini [2003] reach a similar conclusion. They show that IMF
lending is most efiective in catalyzing capital  ows when a country is an \in-
termediate" zone between bad and good fundamentals. In this intermediate, or
vulnerable zone, an IMF program elicits an adjustment efiort (IMF program and
country efiort are strategic complements). An implication of this analysis is that
4Powell [2002] suggests that a country?s response to a Fund program is likely to weaken as
its economic situation deteriorates. Supporting that notion, Ivanova, Mayer, Mourmouras, and
Anayiotos [2001] flnd that larger government flscal deflcits, which they believe re ect internal
political competition, are associated with more frequent program failure.
38
not all IMF lending is the source of moral hazard. To the contrary, lending gen-
erates a positive country response in the vulnerable region; moral hazard kicks
in when fundamentals are irretrievably bad.
These considerations are consistent with the Fund?s own preferred approach
to early intervention. For example, in discussing policy toward access to Fund re-
sources, the Fund?s Treasurer?s Department notes: Over the years, it has come to
be recognized that the e?cacy of the mixture of adjustment policies and flnancing
depends largely on the early adoption of corrective policy measures. Early resort
to an adjustment program supported by IMF resources can help to avoid more
drastic policy actions that may otherwise be required, thereby limiting the im-
pact of the adjustment on other members. [IMF 2001a, p. 29] A 1979 decision by
the Fund?s Executive Board had an almost identical wording: \Members should
be encouraged to adopt measures...at an early stage of their balance of payments
di?culties or as a precaution against the emergence of such di?culties."5
The Fund can signal strong commitment by making available a large amount
of resources. All else being equal, we would, therefore, expect programs with
larger resources (in relation to country debt obligations) to be associated with
better capital market access. However, the joint commitment is even stronger
when the country does not actually use those resources. In that situation, a coun-
try subjects itself to the discipline implied by a Fund program without drawing
on the available resources. The delegated monitoring function should be partic-
ularly valuable in such programs that are \precautionary" in nature. Finally,
the Fund can signal commitment by deeper engagement in a country. One mea-
sure of deeper engagement is the length of time over which a country contracts a
5Decision No. 6056-(79/38), March 2, 1979, in IMF [2001b], pp. 167-168.
39
Fund program. Where problems are of a structural nature, markets are likely to
value the continued presence of the Fund. However, excessive repetition of Fund
programs (\prolonged use") is likely to reduce the perception of the country?s
commitment and the Fund?s ability to resolve matters.6
2.3 Literature Review
In this review, we cover three aspects of the literature. First, we brie y describe
the main body of the literature on IMF programs, which focuses on their macro
implications. Second, we discuss the smaller set of writings on the Fund?s ability
to catalyze private capital  ows. And, flnally, we discuss two methodological
issues (the need to move away from considering programs as homogeneous and
the need to correct for sample selection bias arising from unobserved difierences
between program participants and nonparticipants).
On the implications of IMF programs for macro country performance, the
results display considerable consistency despite difierent methodologies and cov-
erage of difierent time periods. Two early studies [Edwards, 1989, and Khan,
1990] reached three conclusions that have stood the test of time. First, Fund
programs help improve the external payments position; this improvement takes
efiect within a year, and is sustained beyond the program. Second, the impact
on in ation is statistically insigniflcant. Third, growth actually sufiers during
the period of an IMF program but recovers once the program ends, though pos-
6In light of results obtained by Stone [2002], loss of credibility when there is a high incidence
of program repetition may also re ect that such repetition re ects, in part, political a?nity
with the United States, which serves to reduce the incentive to undertake demanding reform
measures.
40
sibly not to the level prior to the initiation of the program.7 The problem of the
appropriate counterfactual against which to compare IMF programs has plagued
all studies.8 However, continued econometric reflnement conflrms these flndings
[Mussa and Savastano, 1999].
That the maximum efiectiveness is achieved with respect to the external pay-
ments situation is not surprising. The Fund?s principal objective and its analyt-
ical approach both lead to that focus. Fisher [1997] notes: \Fund programs are
designed to restore balance-of-payments viability, and more generally to restore
macroeconomic stability|seen as a necessary condition for economic growth."9
Thus, though growth is an objective, especially in programs that have longer
duration and greater structural content, the immediate emphasis is on the exter-
nal payments position [see also Schadler and others, 1995]. The ambiguity with
respect to the growth efiect follows, as Krueger [2000] notes, from the remedy
7Much of the recent debate has centered on the growth efiects. Przeworski and Vreeland
[2000] flnd the most signiflcant adverse efiect on growth. Hutchison [2001] flnds a small negative
growth efiect while Barro and Lee [2001] flnd that a Fund program has no impact on growth. At
the other extreme, Dicks-Mireaux, Mecagni, and Schadler [2000], who focus only on countries
that undertake structural adjustment programs and hence are in the low-income category, flnd
a signiflcant positive growth efiect of IMF programs.
8The generalized evaluation estimator suggested by Goldstein and Monteil [1986] and em-
ployed by such in uential papers as Khan [1990] and Conway [1994] has been the preferred
approach to dealing with the problem of the counterfactual. For recent applications, see Dicks-
Mireaux, Mecagni, and Schadler [2000] and Hutchison [2001]. This estimator allows for the
possibility of "mean-reversion," that is, of a return towards normalcy from distress even in the
absence of a Fund program, possibly on account of an endogenous policy response.
9This focus leads to a Fund program being \built around three identities: the central-bank
balance sheet, the balance of payments constraints, and the government budget constraint."
41
in addressing the traditional balance-of-payments crises: devaluation of the do-
mestic currency and tightening of monetary and flscal policy to contain domestic
demand.
On the indicator of most interest to this paper, private capital  ows, strong
presumptions, anecdotal evidence, and statistical analysis lead to quite difierent
conclusions. It is often taken as axiomatic that a Fund program is necessary
for the resumption of capital  ows [Dhonte, 1997, and Fisher, 1997]. Bird and
Rowlands [2001a] say it is a \commonly held view" that the IMF helps attract
private capital to a country by endorsing the country?s economic reform plan.
They cite, for example, a U.K. Treasury Committee report on the IMF that refers
to \an all pervasive conventional wisdom" that an IMF program buys a \good
housekeeping seal of approval." Marchesi and Thomas [1999] state: \Overall,
there is evidence to suggest that those who accept the intervention of the Fund
can more easily obtain better conditions on their loans, consistent with our thesis
that program adoption plays an information role." However, with the exception
of Marchesi [2001], which is a follow-up to Marchesi and Thomas [1999], the
statistical evidence to date goes the other way.
Killick, Malik, and Manuel [1992] do a before-after comparison of net capital
 ows and flnd that these  ows decline after an IMF program is put in place. Much
of the decline is due to an increase in repayments rather than to a decline in gross
in ows. Bird and Rowlands [1997 and 2001a] are especially skeptical of the Fund?s
\catalytic efiect." They flnd no empirical evidence for such an efiect, consistent
with their priors. IMF programs are a sign of economic distress and they are
not persuaded that the country?s macroeconomic performance improves following
the start of a program. Similarly, in a regression to explain spreads charged on
42
commercial bank loans, Ozler [1993] flnds a positive sign on the dummy variable
for an IMF program, suggesting that the program is an indicator of \repayment
di?culties."10 These studies, however, have their limitations. Ozler?s results are
quite sensitive to the inclusion of other explanatory variables. Once variables
are added to characterize the loan and whether a country achieved sovereign
status only recently, the coe?cient falls sharply and is no longer signiflcant at
the 5 percent level. The Bird and Rowlands [1997 and 2001a] and Ozler [1993]
analyses also do not formally address the possibility that a drop in capital  ows
may trigger IMF programs, the reverse causality or selection problem.
Edwards [2000], in reexamining the catalytic efiect of Fund programs, con-
siders the possibility that self-selection into Fund programs may bias the results,
but flnds that correction for self-selection makes no difierence|there is still no
evidence of a catalytic efiect. This is not surprising since probit estimates of pro-
gram participation fare poorly in their predictive ability [Hutchinson, 2001, and
Garuda, 2000], and tend, moreover, to be highly sensitive to choice of sample [see,
especially Bird and Rowlands, 2001b, for an extensive discussion of the history
and weaknesses of these estimates]. Edwards does flnd, however, that program
countries that have a recent history of lack of compliance with the agreed re-
form agenda are penalized in terms of access to capital markets. Thus, he flnds
evidence for an asymmetric efiect: Fund programs do not necessarily help, but
programs with noncompliance appear to hurt. The important point the paper
makes is that all Fund programs cannot be taken to have the same efiect, since
10Hajivassiliou [1986] reaches the same conclusion as Ozler [1993] in his estimate of a supply
function for capital, where he flnds that a dummy variable representing IMF programs (and
also instances of debt rescheduling) is associated with reduced capital  ows.
43
the nature of country and Fund involvement is likely to vary considerably across
programs.
The one study that flnds an indirect impact of IMF programs on capital
market access is Marchesi [2001]. She examines a country?s ability to reschedule
its private debt obligations and flnds that the presence of a Fund program helps
in this respect. She interprets her flnding as evidence that participation in a
Fund program signals a commitment to policy reform that is a precondition to
debt rescheduling and continued market access.
The bulk of the literature described above treats IMF programs as undif-
ferentiated. Thus, a single dummy variable represents the presence or absence
of a Fund program. However, difierences between Fund programs have recently
received some attention. An advance, in this respect, is distinguishing between
types of Fund programs (for example, Stand-By Arrangement and Extended Fund
Facility) as in Eichengreen and Mody [2001] and Bird and Rowlands [2002]. In
addition, program efiectiveness is likely to vary with country conditions. Edwards
[2000], as noted, flnds difierential efiects for countries in and out of compliance
with the program. Ivanova, Mayer, Mourmouras, and Anayiotos [2001], in ex-
plaining the success or failure of Fund programs, distinguish between countries
on the basis of internal political competition for resources. Stone [2002] focuses
on a number of difierences, important among which is a measure of the country?s
political a?nity to the United States as a proxy for the inability of the Fund to
discipline domestic policymakers in that country. Garuda [2000] also difierenti-
ates across country characteristics.11 Speciflcally, within the group of countries
11He classifles countries by a "propensity" score, that is, by a measure of the likelihood
that the country is in a Fund program. A country?s propensity is derived from a probit as
44
with a high propensity to enter IMF programs, an IMF program is associated with
a worsening income distribution. In the medium- and low-propensity groups, an
IMF program is associated with an improvement in the income distribution.
In summary, this review of the literature on Fund programs highlights both
substantive conclusions and methodological issues. On substance, Fund programs
help with respect to the current account and the balance of payments. Thus, net
capital  ows should decrease following the start of a Fund program. With respect
to gross  ows also, the literature has generally concluded that no IMF catalytic
efiect exists. Methodologically, the literature points to concerns with regard to
counterfactuals, reverse causality, and omitted variables that afiect both program
participation and capital market access.
2.4 Methodology and Data
In this paper, we move away from using volumes of gross capital  ows and fo-
cus instead on the probability of bond issuance in international markets and the
spreads charged on individual bonds. In thus limiting our focus, we do not con-
sider other forms of capital  ows, such as syndicated loans and foreign direct
investment. However,  ows through bond issuance were a major source of inter-
national capital to emerging markets in the 1990s. The spotlight on the available
transactional bond data improves, we believe, the prospects of addressing both
substantive and methodological issues. In this section, we flrst present our basic
framework for analyzing the determinants of bonds issuance and spreads. We
then discuss our approach to dealing with the econometric concerns highlighted
the probability of IMF program participation and is a function of such variables as growth,
in ation, reserves, and current account balances of current and past periods.
45
by the literature review. Finally, we present some descriptive statistics.
2.4.1 Framework for Analyzing Spreads
We adopt an estimation approach developed in earlier papers [see Eichengreen
and Mody, 2001]. We estimate a two-equation model: the \spreads" equation,
which specifles the determinants of spreads charged on a particular bond, and
the \selection" equation, which is a probit for the decision to issue the bond.
Throughout, the spread we use is the so-called primary or launch spread and
is deflned as the premium paid at the time of bond issuance over the risk-free
rate for a bond of similar maturity and currency denomination. Because we
use primary spreads, we do not \follow" a particular bond \over time." Bond
frequency issuance varies over time, resulting in varying numbers of bonds for a
given country in any given time period.
The spreads equation is a linear relationship:
(1) log(spread) = X +u1
where the dependent variable is the logarithm of the spread; X is a vector of issue,
issuer 12, and period characteristics; and u1 is a random error. The X vector
contains a dummy variable for an IMF program, other program characteristics,
and also interactions between the program and country characteristics, as we
discuss below in detail. Since the spread will be observed only when the decision
to borrow and lend is made, we correct for this sample selection problem. Assume
12We believe that the reverse causality problem that may be argued to be present in some of
the country characteristics control variables is not serious in our case. Our dependent variable
is the individual bond issue, not an aggregate variable, and it is less likely that an individual
spread observation determines the aggregate variables that we use as controls.
46
that spreads are observed when a latent variable B crosses a threshold B0 deflned
by:
(2) B0 =  Z +u2
where Z is the vector of variables that determines the desire of borrowers to
borrow and the willingness of lenders to lend (and will also contain the IMF
program variables and their interactions), and u2 is a second error term. We
further assume that:
u1 ? N(0;s)
u2 ? N(0;1)
corr(u1;u2) = ?
This is a sample selection model  a la Heckman [1979] and equations (1) and (2)
can be estimated simultaneously by a maximum likelihood procedure. Estimating
the determinants of market access requires information on those who did not issue
bonds. For each country we consider three categories of issuers: sovereign, (other)
public, and private. For each quarter and country where one of these issuers did
not come to the market, we record a zero, and where they did we record a one.
Leung and Yu [1996] note that the estimation does not require the variables
in the selection equation and the spread equation to be difierent. What is critical
instead is to avoid multicollinearity between the variables in the spreads equation
and the "inverse-Mills ratio" constructed from the selection equation. That, in
turn, requires the value of the variables not be concentrated in a small range and
that the truncated observations (no bond issuance) should not dominate the set
of observations. In our case, most variables have a large range and about a third
of the observations have a bond issued. We do include in the probit selection
47
equation, the ratio of debt service to exports, which appears to in uence the
issuance decision but not the determination of spreads.13
The data sources for the dependent and explanatory variables are documented
in Appendix A.4. Details on bonds issued and their characteristics are obtained
from Bondware, a commercial data source. Bond characteristics included in the
spreads equation are: the dollar value of the bond issued, its maturity, whether
the issuer was in the public or private sectors, the industrial sector of the issuer,
the currency of issue, and whether the bond had a flxed or  oating rate.
The global variables included in both the spreads and selection equation are:
U.S. industrial growth rate during the quarter in which the bond was issued; the
daily swap rate (as a measure of liquidity risk); and, as a measure of market
uncertainty, the standard deviation of daily Emerging Market Bond Index (a
commonly followed index of emerging market spreads) over the relevant quarter.
In the spreads equation, we use the following country characteristics as control
variables: country credit ratings provided by Institutional Investor, external debt
relative to GNP, a dummy variable for whether the sovereign has restructured
debt within the previous year, the growth rate of real GDP, the variance of export
growth, the ratio of short-term debt to total debt, the ratio of reserves to imports,
13Dell?Ariccia, Godde, and Zettelmeyer [2000] follow a similar research strategy but also add
as instruments in the probit equation the bonded debt issued in the previous year, the number
of bonds issued in the previous year, the natural logarithm of per capita GDP in 1993, and
a dummy variable for countries directly afiected by the Asian crisis. Ideally, the instruments
should in uence the issuance decision but not the spreads. It is not obvious that these four
variables fulflll that objective and, in practice, it is hard to flnd such variables. For example,
Asian-crisis countries were rationed during speciflc years but also paid higher spreads in those
years. Hence, it is not su?cient to rely on exclusion conditions.
48
and the ratio of domestic private credit to GDP. Note that the debt-restructuring
variable we use is not the same as debt rescheduling: restructuring re ects a
positive efiort at debt management and typically involves exchange new debt for
old more expensive or in exible debt. Also, while it is common to use the ratio
of reserves to short-term debt as a measure of country liquidity, we use short-
term to total debt and reserves to imports since we want to examine separately
the in uence of short-term debt and reserves. The IMF variables we use in
alternative speciflcations in the spreads and selection equation are: IMF program
dummy, a measure of repeated Fund programs, the size of the program relative
to the country?s external debt, and whether a program was \precautionary,"
that is, if in practice there was no, or limited, drawing down of Fund resources.
The onset of the Fund program was dated by the month in which it originated,
which contrasts with the typical practice of using an annual dummy variable if a
program was initiated at any time during the year. Dicks-Mireaux, Mecagni, and
Schadler [2000] note that the timing of IMF programs makes a difierence to the
empirical results. In their analysis, they code the IMF dummy variable to take
the value one if a program was in efiect for six or more months during the year.
Kaminsky and Schmukler [1999] use daily announcements to track movements in
stock markets and flnd, on average, that stock markets respond positively on the
days agreements are reached with international organizations such as the IMF.
We believe that the more precise timing of programs in this paper helps with
reducing the reverse-causality problem.
49
2.4.2 Evaluating IMF Programs: Econometric Issues
The literature review has highlighted the need to: (1) identify the direction of
causality; and (2) consider the possibility of omitted variables bias (of which, as
we discuss below, selection bias is a special case).14
Consider flrst the identiflcation issue. In a recent paper, Barro and Lee [2001]
use as instruments for participation in Fund programs, such variables as the
political a?nity of the country to the United States, the national composition of
the Fund?s stafi, and past participation in Fund programs.15
The reverse-causality problem, we believe, is less serious in our case than for
those who have addressed this issue in the past. All previous studies use data
at frequencies of at least one year [Barro and Lee, 2001, use flve-year averages].
Moreover, the outcomes they test (such as growth, current account balances,
in ation) are national outcomes just as the IMF program is a national decision.
It is quite likely that over these time spans, and especially as the time span gets
longer, national economic outcomes will in uence the decision to participate in
Fund programs.
In our case, there are two key difierences relative to the past literature. First,
the outcome we observe is an individual bond issue. While a bond issue may be
large and re ect broad market sentiment towards the country, a single bond issue
14. The mean-reversion problem does not apply in our case. That problem arises when the
change in the indicator of interest is the dependent variable. The extent of the change, it is
argued, depends among other things on the initial level of the indicator. In our case, we are
not examining changes in spreads but rather the level of spreads in any period.
15It is not clear that past participation in Fund programs is a good instrument since it could
re ect unobserved country characteristics that constrain the country?s economic growth. In
that case, some part of the country?s low economic growth will be attributed to the Fund.
50
is unlikely to trigger an IMF program. And this leads to our second point. Our
observations are at a much higher frequency than is the case with past studies.
Figure 2.1 shows that if the IMF program is initiated at the time shown by the
solid vertical line, but the dotted line is the starting date that the econometrician
uses, then we are likely to flnd a positive correlation between IMF programs and
spreads, re ecting reverse causation. However, if we record the actual starting
date, then we are more likely to observe whether a Fund program was associated
with a reduction in spreads. A bond issuance is recorded on the day it occurs
and the start of an IMF program is recorded in our data in the month in which it
occurs. Since the actual start of a program re ects many considerations, including
negotiations between a country and the Fund and internal Fund procedures, this
further reduces the likelihood that there is signiflcant feedback from an individual
transaction to an IMF program. As it turns out, the sign on the coe?cient of the
IMF program in the spreads equation is typically negative, implying that a Fund
program is, all else equal, associated with lower spreads. Thus, if poor market
sentiment towards a country leads to a Fund program, then our result suggesting
that a Fund program lowers spreads would only be strengthened.
Figure 2.1 also points to the importance of controlling for variables that move
the level of spreads: thus the rise in spreads before the onset of a Fund program
could re ect worsening of country characteristics, which could be misattributed
to the Fund program. This further concern with respect to omitted variables is
often stated as a "selection bias" problem and the Heckman selection correction is
sometimes applied. However, as the literature review has shown, Fund programs
are in place in a variety of circumstances that are not easily captured through a
probit equation that forms the flrst step of the selection bias correction. Edwards
51
[2000] flnds that correction for selection bias does little to change his results. With
our higher frequency data, predicting Fund programs is likely to be even more
di?cult (not least because the right-hand-side variables are often measured at
much lower frequencies). In addition, a variable that consistently works well in
predicting participation in Fund programs is past participation [see Barro and
Lee, 2001, and Bagci and Perraudin,1997]. If this is a key omitted variable, then
it appears to us that the best approach is to include it directly in the outcome
equation. Indeed, our results below show that the history of past participation in
Fund programs has a signiflcant bearing on capital market access. In addition, we
include some nonlinear terms that also could proxy for some omitted variables.
2.4.3 Descriptive Statistics
Between1990and2000, over250IMFprogramswerenegotiated, withthenumber
ofprograms varying between20 and 35 a year, except in 1990 and 2000 when there
were less than 20 programs. There is no trend in either the number of programs or
the amount of flnancial support committed by the Fund. In particular, flnancial
support has been large at times of crises: the big jump in 1995 re ects the large
package to Mexico and the large commitments in 1997 and 1998 followed the
East Asian and Russian crises.
Table 2.2 shows that between 1991 and 2000, the period covered by this
paper, about one-third of all developing country and emerging market bonds were
issued by borrowers from countries with IMF programs. The spreads charged
(yield to maturity minus the risk-free rate) on the bonds were typically higher
for program countries (406 basis points) compared with countries that did not
have such programs at the time the bonds were issued (223 basis points). Also,
52
bonds issued by program countries had shorter maturities (5.44 years versus 6.67
years).
It was thus the case that IMF programs were associated with poorer access
terms. This is not surprising, since Fund programs were also associated with
worse fundamentals: higher debt/GDP ratios, lower recent growth, and greater
volatility. Countries with Fund programs appear to have better credit quality
in one dimension: among those that issued bonds, those with Fund programs
have higher reserves. Also, issuers with Fund programs have had lower ratios of
short-term debt to total debt; however, that may re ect their lack of access to
short-term credit. In the next section, we examine the relationship between Fund
programs and capital market access after controlling for country fundamentals-
and also for bond characteristics and global fundamentals.
2.5 The Role of Countries? Fundamentals
We begin with the conventional approach representing an IMF program as a
dummy variable signifying whether an IMF program was ongoing or not. Then
we explore the in uence of the country?s external vulnerability by interacting the
IMF program dummy with a variety of country characteristics. As noted above,
we jointly estimate the decision to issue a bond and the determination of the
spread on the bond. A complete set of results for the base equation is reported
in Appendix A.5. In the rest of this paper, we continue to use the controls in
this base equation but, to conserve space, we report only the coe?cients on the
relevant IMF variables and their interactions with other determinants of bond
issuance and spreads.
Column 1 in Table 2.3 shows the simple efiect of the Fund?s presence at
53
the time of bond issuance. Fund presence is seen in the selection equation to
signiflcantly improve market access, raising the frequency of bond issuance. Fund
programs are also associated with reduced spread. The point estimate suggests
that the presence of a Fund program reduces spreads by about 10 percent. If there
were mainly \reverse" causation, with periods of market aversion to a country
causingaFundprogram, wewouldhavefoundthecoe?cientontheFundprogram
dummy to be negative in the selection equation and to be positive in the spread
equation. If we repeat this regression omitting country characteristics from the
control variables ( keeping bond features and global variables), then we do flnd
that the IMF dummy is negative and highly signiflcant in the selection equation
and positive and signiflcant in the spreads equation. Clearly, the failure to control
for the country variables results in this misattribution of the country weakness to
IMF programs. The omission of relevant country controls in some of the studies
cited above could be the reason for their reaching a bleaker conclusion on IMF
programs than is warranted.
We next examine how IMF programs interact with country characteristics.
The flrst question we ask is whether a Fund program is helpful in dampening
the efiect of external volatility (Table 2.3, Column 2). We use a measure of
the volatility of the country?s exports (the standard deviation of the monthly
growth of exports). When entered independently, this measure is associated with
higher spreads and lower probability of bond issuance suggesting that volatility
shifts the supply of funds to the left (see Appendix A.5). This is consistent with
the Cat~ao and Sutton [2002] flnding that under conditions of macroeconomic
volatility, sovereign debt defaults are more likely and hence will result in reduced
access and higher spreads. When we interact export volatility with the IMF
54
program dummy, the interaction term enters with a negative and signiflcant sign.
Thus, absent a Fund program, an increase in volatility from the median to the
75th percentile raises spreads by 6.5 percent; with a Fund program, that increase
is only 2 percent. The evidence supports the possibility, discussed above, that
the Fund program acts a commitment mechanism that counteracts the efiect
of volatility. Viewed alternatively, the result indicates that an IMF program
is particularly beneflcial as country volatility increases, reducing spreads and
increasing the probability of bond issuance. The efiects are not small. At the
median volatility, an IMF program is associated with a 7.4 percent reduction in
spreads but at the 75th percentile of volatility, spreads are lower by 12.0 percent.
Thus, once again, where volatility is high, the presence of the IMF acts to reduce
investor aversion to the country.
But the Fund is not able to counteract all types of volatility. The interaction,
for example, of Fund programs with the volatility of the Emerging Market Bond
Index (the EMBI) is statistically insigniflcant in the selection and spreads equa-
tion, suggesting that a country with a Fund program is not insulated from high
volatility in international capital markets.
We next consider the possibility that IMF programs are most efiective when
countries are vulnerable but not without hope of return to normalcy. This could
be the case, as discussed above, if contracting and information problems are
especially severe in periods of vulnerability. The goal is to determine if the
credibility of joint commitment is eroded if country fundamentals are past the
point of early remedial action. We examine the efiectiveness of IMF programs
with respect to the availability of reserves (proxied by the ratio of reserves to
imports) and the country?s external debt-to-GDP ratio. A simple interaction
55
of the IMF program dummy and reserves/imports showed no statistical efiect.
Thus, we were led to consider the possibility that IMF programs may interact
with reserves availability (and other domestic conditions) in a nonlinear manner.
In other words, could it be that countries with very high or very low reserves do
not beneflt from IMF programs but those in the middle do?
To examine the nonlinearity, we specifled a piece-wise linear function.16 We
split the reserves-to-imports ratio at the median, creating two variables: the
reserves-to-imports in the low range and in the high range. Column 3 of Table 2.3
reports the coe?cients for the IMF dummy intercepts and the interaction terms
for low and high ranges of country reserves-to-imports.17 For countries with low
reserves, the results suggest that spreads are higher with a program rather than
without a program. The IMF efiect improves with reserve availability and a Fund
program turns beneflcial when the reserves cover at least 3 months of imports.
At the median value of reserves to imports (about 4.5 months of imports), the
efiective coe?cient on the Fund program is -0.19, that is, a Fund program lowers
spreads by about 19 percent. However, past the median value of reserves to
imports, the Fund efiect worsens again, and turns to a small positive efiect on
spreads when reserves are larger than about a year?s worth of imports. The efiects
on probability of issuance are also nonlinear and we flnd, in particular, that the
IMF?s assistance in improving the probability of issuance declines rapidly after
the median value of reserves-to-imports, though the magnitude of the efiects are
16While this approach imposes considerable structure, it allows a simple test of the possibility
that country conditions matter in a nonlinear manner. Adding a quadratic interaction term
was not feasible because of a multicollinearity problem.
17The full equation, which is not reported here, now has two variables representing reserves
to imports in the low and high ranges.
56
not large in this case.
The evidence, therefore, suggests that if a country?s reserves are very low,
Fund programs are unable to compensate for the economic di?culties faced by
the country. In contrast, when reserves are low|but have not yet fallen to the
extremely low levels that signify deeper structural problems|Fund programs
can be very efiective. The results further suggest that as reserves increase Fund
efiectiveness falls ofi, as may be expected. But a point may also be reached
where Fund programs may come in the way of market access if undertaken when
reserves are high|as if the presence of an ambulance is a sign of trouble.
We repeated the same methodology with debt-to-GNP ratio, with similar
results.18 Thus, once more we created two variables, one with the debt-to-GNP
ratio in the low range (below its median value) and another in the high range. We
interacted these two variables with the IMF dummy to test if these interactions
vary with the range in which the debt-to-GNP ratio falls. The results support
the analysis above. The estimates presented in Column 4 of Table 2.3 imply that
IMF programs are efiective in reducing spreads when the debt-to-GNP ratios are
between 34 and 61 percent. For debt-to-GNP, interactions with IMF programs
are also strongly nonlinear in in uencing the probability of issuance, with the
favorable efiects on issuance lying in the range of 25 to 63 percent. Interestingly,
Pattillo, Poirson, and Ricci [2002] flnd that an external-debt-to-GDP ratio of
about 35 percent marks the threshold beyond which additional debt accumulation
hasanegativeefiectongrowth. Thisthresholdisatthelowerendofourestimated
range in which the Fund has a catalytic efiect. Thus, once a country has crossed
the threshold, the Fund can counteract the negative impact of the high level of
18Results were similar also when considering the ratio of short-term to total debt.
57
debt, but at a diminishing rate.
In summary, the results clearly support the idea that country fundamentals
matter in determining the efiects of Fund programs. These programs help when
thecountry?sexportcompositionmakesitispronetoexternalvolatility. However,
along other dimensions|reserves and debt|poor country fundamentals can hurt.
It is as if the credibility of Fund programs is weakened when the country has
already placed itself in a highly vulnerable external position. Instead, the Fund
is efiective when countries are in the early stages of external payment di?culties
and the restoration of balance is reasonable likely.
2.6 Implications Of Fund-Supported Program
Design
In this section, we explore three dimensions of IMF programs: (1) size of lending
(normalized by country debt); (2) whether a program was \precautionary" or
not; and (3) \prolonged" use of Fund resources.
The Fund can signal the credibility of a stabilization process and its intention
to support that process not just through its presence but also through the size
of the program. In Column 1 of Table 2.4, we replace the IMF program dummy
with the amount committed (as a percentage of the country?s long-term debt).
The results show that program size is important. Larger programs both increase
the probability of bond issuance and lower spreads. The results imply that an
additional program size equal to 10 percent of the country?s long-term debt lowers
spreads by about 13 percent.
Why would a larger program size have a greater catalytic efiect? It could be
58
that investors view the country?s repayment capacity to have improved when IMF
funds become available. However, while repayment di?culties may be relieved in
the short term, over a more medium term, the Fund also has to be repaid and so,
over that longer time horizon, the country?s repayment capacity is not improved
by the mere fact of an IMF loan. Moreover, to the extent that the IMF is a
preferred creditor, it is possible that some private creditors may take the view
that their repayment prospects have in fact become worse. Thus, if it were mainly
the case that the amounts received from the Fund were helping repay existing
debt, access to new debt should not improve and spreads on that new debt should
not decline. An alternative interpretation of the better market access is that the
size of the Fund program signals greater commitment to economic reforms that, in
turn, improves the medium-term capacity of the country to honor new contractual
obligations.
The amount committed in a Fund program is not necessarily disbursed|
programs may be \precautionary." Programs may be precautionary in two senses.
First, at the time the program is agreed upon, the borrowing country may declare
its intention to not draw on the resources made available. While this is not a
contractually binding restriction, and the country can change its mind with no
penalty, declaration of the intent to not borrow implies that Fund resources are
not critical. Rather the country is volunteering to subject itself to the discipline
of the Fund?s program. Second, the country may negotiate a Fund program and
draw on Fund resources initially but thereafter voluntarily halt disbursements
while keeping the IMF program in place. Yet, by its later action of not draw-
ing on the resources, the country may treat the program as precautionary. Such
59
programs may be referred to as \turned precautionary."19 Programs that \turn"
precautionary are larger in size than the \outset" precautionary programs and,
presumably, have more demanding policy conditionality. Of the 245 programs
between 1991 and 2000 covered in this analysis, only 38 were precautionary at
the outset and 13 \turned" precautionary. However, of the bonds issued while
a country was in a Fund program, over 45 percent were during precautionary
programs (18 percent were precautionary at outset and 33 percent \turned" pre-
cautionary).
How do precautionary programs fare? Column 2 of Table 2.4 shows that pre-
cautionary programs of the two varieties (\outset" and \turned" precautionary)
have difiering implications. The results suggest that \declared" precautionary
programs do not have a signiflcant efiect over and above that already implied by
the presence of the program and the program size (as re ected in the variable
representing the IMF committed amount divided by the country?s debt level).
In contrast, turned precautionary programs add signiflcantly to the value of the
IMF?s presence both in terms of spreads and access.
Recall, that \turned" precautionary programs are signiflcantly larger than
the \outset" precautionary programs.20 Thus, the highly signiflcant sign on the
\turned" precautionary dummy re ects the beneflts deriving both from the pre-
cautionary nature of the program and the program?s large size. Thus, the size
coe?cient falls from 1.28 to 0.82. The evidence suggests, therefore, that those
19The country chooses to continue the arrangement and pay the commitment fee rather than
simply cancelling it.
20The median size of \outset" precautionary programs is about 40 million SDRs whereas
that for \turned" precautionary programs is almost 10 times larger at 330 million SDRs. When
normalized by country debt, the \turned" precautionary programs are still much larger.
60
subjecting themselves to the Fund?s discipline can beneflt even when the resources
are not drawn.
There is yet another Fund facility that combines large size and precautionary
intent. This is the \Supplemental Reserve Facility (SRF)," that has been used
for large-sized programs at times of crises. The premise is that a country in the
midst of a crisis could be subject to a loss of investor confldence even though its
fundamentals are relatively sound. The goal thus is to restore investor confldence
to the country, and at the same time to prevent \contagion", or the loss of
confldence from spreading to other countries. The SRF overlaps to a considerable
extent with programs that turned precautionary.21 As such, in Column 3 of Table
2.4, we flnd that while the coe?cients on program size, the dummy variable for
the SRF, and the dummy variable for \turned precautionary" programs are all
negative, their signiflcance is marginal. When we repeat the regression without
the IMF amount (Column 4), both the SRF and the "turned precautionary"
programs come in with signiflcantly negative signs in the spreads equation and
signiflcantly positive signs in the selection equations.
These flndings can be linked back to our discussion of country fundamentals.
Though the fundamentals in countries with \turned precautionary" programs are
worse than in non-program countries, they are superior to those in countries with
other forms of IMF programs. In particular, \turned precautionary" countries
do better than other program countries with respect to lower debt/GDP ratios,
higher growth, and lower volatility. In contrast, their reserves/import ratios
are not very difierent from those in countries with other programs. Thus, a
21The analysis here includes those loans that were made under "exceptional circumstances"
and were thus similar in intent to the SRF.
61
possible interpretation of our results is that \turned precautionary" countries are
vulnerable to external pressures and that their vulnerability has further led to
a liquidity problem. However, because the underlying fundamentals are not yet
beyond a point of no return in the short-run, a reform program with IMF support
carries credibility.
Finally, we ask if there may be diminishing returns to a country?s repeated
interactions with the Fund. Two opposing forces may be at work here. Repeated
Fund involvement may be warranted in light of medium-term problems that the
country faces and may re ect a joint commitment on the part of the country and
the Fund to resolve the problems. However, it may be the case that \prolonged"
use of Fund resources implies an inability to resolve the problems at hand and is
an indicator, therefore, of more deep rooted problems. The term \prolonged" use
has many difierent indicators but one of them recently proposed by the IMF?s
new Independent Evaluation O?ce is the existence of a program for more than
70 percent of the time over a given period.
Thus, we examine if repeated Fund presence in a country makes a difierence
to program efiectiveness in a non-linear manner. The measure of repeated Fund
presence we use is the number of months that a Fund program was in place in
the country during the four-year (48-month) period from 1987 to 1990. Use of a
prior time period ensures that we do not pick up a reverse causation from poor
market access to a high frequency of Fund programs. It turns out that there is
a high correlation between the number of months the country had a program in
the late 1980s and the number of months a program was in place in the 1990s.
This persistence suggests that our measure of the Fund?s ongoing involvement
in a country re ects a combination of continuing economic di?culties and the
62
inability of the Fund and the country to work together to achieve the necessary
reforms. Since we have already controlled for external indicators such as debt
and reserves, the number of months that a Fund program was in place in the
late 1980s proxies for these other (unobserved) country characteristics and the
Fund-country relationship. With this interpretation in mind, the results support
the speculation above (see Column 5). Continued Fund presence helps up to a
point. These results suggest that the \turning point" is at about 18 months out
of the 48-month window over which our measure of repeated interaction is taken.
Beyond that length of time, continuing Fund efiectiveness in helping with market
access begins to decline and at about 32 months, or about 75 percent of the time
window, continued presence raises spreads. At that point, investors apparently
believe that the problems are either deep-rooted or that the Fund is unable to
exercise the necessary in uence to resolve them. These results and interpretation
are consistent with Conway?s [2001] conclusion that a continuing Fund-country
relationship reaches diminishing returns.
To summarize the flndings in this section, the evidence suggests that construc-
tive engagement between the Fund and the member country can be demonstrated
in difierent ways. Credibility is established by the size of the program, and thus
resources made available do matter. However, the results also show that large
Fund resources and voluntary country commitment under so-called "turned" pre-
cautionary programs go together in signaling both country intentions and Fund
discipline. The programs under the Supplemental Reserve Facility (SRF) have
also been of this nature, but the SRF has not been the only vehicle to establish
confldence by committing signiflcant resources that are ultimately not used. Fi-
nally, where Fund programs are frequently repeated, the credibility of efiective
63
reforms seems to be called into question by the market.
2.7 Conclusions
Except for some recent efiorts to distinguish between programs in terms of their
degree of compliance with agreed policy initiatives, the vast bulk of the empirical
literature does not distinguish between one program and another|each program
takes an identical value of 1 in the program dummy variable. This chapter takes
seriously the diversity in Fund programs and demonstrates that they do vary sig-
niflcantly in their efiects. on the impact of IMF programs Country fundamentals
and program design difier widely across interventions and, not surprisingly, these
do have a bearing on the outcomes.
Thus, a Fund program is not an automatic or standardized \good housekeep-
ing seal of approval."22 Investors appear to value the Fund?s participation in
resolving a country?s external payment di?culties only when they view it is as
likely that the efiort will be successful.23 Our further contribution, we believe, is
22The gold standard, which apparently did provide a "good housekeeping seal," was associ-
ated with a narrow range of prudential macroeconomic policies [Bordo and Rockofi, 1996]. In
contrast, the current range of IMF member countries|and the variety of economic challenges
facing them|leads to a much larger variation in appropriate economic policy measures. Ob-
stfeld and Taylor [2002] conclude even in the case of the gold standard that its credibility was
diminished in the interwar period [1925-31] and unlike Bordo, Edelstein, and Rockofi [1999],
they flnd that only those who devalued before reentering the gold standard beneflted in the
form of lower spreads. The implications of the Obstfeld and Taylor analysis are thus, similar
to ours: country conditions matter in determining the credibility of policy actions.
23It is not straightforward to distinguish a "good" catalytic efiect from a "bad" one in which
moral hazard predominates. We flnd that Fund efiectiveness in catalyzing  ows declines as the
64
to suggest the conditions under which programs are likely to succeed. A success-
ful outcome, measured in this paper as improved access to international markets,
depends on the market?s perception of credible reform measures.
The interplay of country fundamentals with IMF programs also points to
the importance of the credibility of reform measures. Here our flnding is that the
Fund can help mitigate the market?s aversion to volatility of export growth, acting
as if to bolster a country?s reserves. The market apparently discounts stated
efiorts to undertake reform and, indeed, countries with weak external payments
positions could adopt risky strategies to overcome their problems, hence deviating
from the course of action agreed on with the Fund.
A large program size can help signal stronger commitment on the part of
the country and the Fund, but it appears the program-size efiect weakens when
the efiect of \precautionary" programs is considered. Precautionary programs
help boost the frequency of market access and reduce spreads, especially for pro-
grams that turn precautionary, which are much larger in size than those declared
precautionary at the outset. Thus, both the voluntary nature of inviting Fund
discipline and the potential for drawing on resources, if needed, help improve
market access. Repeated use of Fund programs sends a bad signal in this re-
spect. The implication of our results is that where programs are repeated often,
markets infer additional problems that are not re ected in the most commonly
observed indicators of country solvency and liquidity.
country?s own repayment capacity becomes less credible, which suggests that moral hazard is
not dominant. Moreover, moral hazard is likely to re ect itself in cross-country efiects rather
than a renewed burst of imprudent lending to a country that enters into a Fund program [Lane
and Philips, 2000].
65
Table 2.1: Frequency of IMF Programs.
Number of Programs per year
(Size of Programs in billions of SDRs )
SBA EFF ESAF Total
1990 12 0 3 15
(1.70) (0) (0.56) (2.25)
1991 19 2 8 29
(5.30) (2.34) (0.69) (8.33)
1992 15 4 6 25
(2.74) (4.59) (0.44) (7.78)
1993 13 2 7 22
(1.64) (1.42) (0.28) (3.34)
1994 18 4 13 35
(2.61) (1.49) (2.19) (6.29)
1995 21 2 7 30
(19.09) (1.28) (1.20) (21.57)
1996 12 6 14 32
(3.52) (14.25) (1.30) (19.07)
1997 10 4 7 21
(28.02) (1.03) (1.37) (30.42)
1998 6 4 11 21
(11.71) (10.01) (1.06) (22.77)
1999 7 4 9 20
(9.94) (2.80) (0.86) (13.60)
2000 10 2 0 12
(7.36) (3.66) (0) (11.02)
1990-2000 143 34 85 262
(93.62) (42.87) (9.95) (146.45)
Note: SBA is for Stand-By Arrangement; EFF is for Extended Fund Facility;
and ESAF is for Enhanced Structural Adjustment Facility (includes Structural
Adjustment Facility and the now renamed Poverty Reduction Growth Facility.)
66
Table 2.2: Bond Issuance, Terms, and Country Characteristics
No Fund Total
Program Program
A: Bond Issued
Number of Bonds 2156 1139 3295
Spread (basis points) 223 406 282
Maturity(years) 6.67 5.44 6.25
Amount ($ millions) 154 177 162
Debt/GDP 0.27 0.43 0.32
Annual GDP growth (percent) 5.04 3.29 4.40
Short-term/total debt 0.66 0.50 0.56
Reserves/imports (months of imports) 5.91 6.78 6.21
Volatility of exports 0.08 0.11 0.09
B: No Bond Issued
Debt/GDP 0.39 0.56 0.46
Annual GDP growth (percent) 4.00 2.65 3.30
Short-term/total debt 0.55 0.49 0.52
Reserves/imports (months of imports) 4.89 4.68 4.83
Volatility of exports 0.14 0.18 0.16
67
Table 2.3: Interaction of Country Characteristics with Fund Programs
Log of Spreads at Time of Issue
(1) (2) (3) (4)
IMF program, low range -0.089 0.012 0.371 0.680
(-3.05) (0.18) (2.11) (2.00)
IMF program, high range -0.176 -0.287
(-1.67) (-2.28)
IMF program interacted with:
EMBI volatility -0.396
(-0.13)
Export growth volatility -0.902 -0.974 -1.151
(-2.84) (-3.10) (-3.58)
Low range of Debt/GNP -1.958
(-1.82)
Debt/GNP 0.470
(1.96)
Low range of Reserves/Imports -0.332
(-2.01)
High range of Reserves/Imports 0.056
(1.82)
Probability of Issuance
(1) (2) (3) (4)
IMF program, low range 0.337 0.181 -0.100 -1.782
(10.60) (1.81) (-0.59) (-5.13)
IMF program, high range 1.009 1.327
(10.20) (8.28)
IMF program interacted with:
EMBI volatility -2.326
(-0.70)
Export growth volatility 1.725 1.773 2.390
(2.39) (2.40) (3.35)
Low range of Debt/GNP 6.526
(6.07)
High range of Debt/GNP -2.448
(-8.96)
Low range of Reserves/Imports -0.155
(-1.03)
High range of Reserves/Imports -0.223
(-8.28)
Note: Regressions have the controls specifled in the full regression presented in Ap-
pendix V. z-statistics, based on robust standard errors, are in parentheses.
68
Table 2.4: In uence of Program Features on Bond Market Spreads and Issuance
Log of Spreads at Time of Issue
(1) (2) (3) (4) (5)
IMF amount/debt -1.290 -0.825 -0.554 -1.328
(-3.62) (-2.38) (-1.55) (-3.61)
Precautionary program:
Outset 0.021 0.001 -0.019
(0.38) (0.01) (-0.34)
Turned -0.139 -0.091 -0.101
(-3.58) (-2.13) (-2.34)
Supplemental Reserve Facility -0.119 -0.161
(-1.98) (-2.73)
Number of months in -0.029
IMF program, 1987-1990 (-5.92)
Square of Number of months 0.001
in IMF program, 1987-1990 (5.92)
Probability of Issuance
(1) (2) (3) (4) (5)
IMF amount/debt 5.659 4.940 3.623 5.945
(16.57) (13.02) (8.56) (16.48)
Precautionary program:
Outset -0.091 -0.053 0.002
(-1.40) (-0.83) (0.04)
Turned 0.253 0.145 0.211
(5.24) (2.44) (3.51)
Supplemental Reserve Facility 0.464 0.818
(5.57) (10.55)
Number of months in 0.065
IMF program, 1987-1990 (16.94)
Square of Number of months -0.002
in IMF program, 1987-1990 (-15.40)
Note: Regressions have the controls specifled in the full regression presented in
Appendix V. z-statistics, based on robust standard errors, are in parentheses.
69
Figure 2.1: Econometric Implications of Timing of IMF Programs
 
 
Time
Spreads 
on 
Bonds
Good Fundamentals Weak Fundamentals Good Fundamentals
No IMF Program IMF Program No IMF Program
Recorded
IMF Program
 
70
Chapter A
Appendices
A.1 Proof of Lemma 1
From zero expected proflt condition we write the implicit function
Q(?;r1) ? 1? 11+?
Z Min(?1;?0)
0
?f(k0)
k0 dF(?)?
1
?
Z ??
Min(?0;??)
??f(k
0)
k0 ?
D0
k0
?
dF(?)?
?
Z ??
Min[?1;??]
r1dF(?) = 0
First consider the case where ?0 < ??; applying the implicit function theorem
we have that @r1@? = ?
@Q(:)
@?
@Q(:)
@r1
@Q(:)
@? =
1
(1+?)2
Z ?0
0
?f(k0)k
0
dF(?)+ 1?2
Z ??
?0
?
?f(k0)k
0
? D0k
0
?
dF(?)+
+
??0f(k
0)
?k0 ?
D0
?k0 ?
1
1+?
?0f(k0)
k0
?
F0(?0)@?
0
@? +
?
r1 ? ?
?f(k0)
?k0 +
D0
?k0
?
F0(??)@?
?
@?
Taking into account that ?0 = (1+?)D0f(k0) and that ?? = D0+D1f(k0) we have that the
last two terms are both equal to zero. Thus, @Q(:)@? > 0. Moreover,
71
@Q(:)
@r1 = ?
Z ??
??
r1dF(?) < 0:
Thus, @r1@? > 0.
Proceeding in the same way we can show that this is also the case when
?1 < ??.
72
A.2 Proof of Lemma 3
To simplify the exposition of this proof consider the special case when ?0 =
?1 = ?? . Without seniority, the interest rate is pinned down by:
Z ??
??
r1dF(?)+
? 1
L0 +L1
?Z ??
0
?f(k0)dF(?) = 1
and with seniority by
Z ??
^?
rs1dF(?)+
? 1
Ls1
?Z ^?
0
?f(k0)dF(?) = 1
The proof proceeds by contradiction. Assume that r1=rs1. This implies that
Rs1=R1 since Ls1 = L1, and this implies that ^? < ?? for sure. Splitting the integral
limits and equating both expressions:
Z ??
??
r1dF(?)+
? 1
L0 +L1
?"Z ^?
0
?f(k0)dF(?)+
Z ??
^?
?f(k0)dF(?)
#
=
=
Z ??
^?
rs1dF(?)+
Z ??
??
rs1dF(?)+
? 1
Ls1
?Z ^?
0
?f(k0)dF(?)
Rearranging we get:
Z ??
??
(r1 ?rs1)dF(?) =
Z ??
^?
rs1dF(?)+
Z ^?
0
?f(k0)
? 1
Ls1 ?
1
L0 +L1
?
dF(?)?
?
? 1
L0 +L1
?Z ??
^?
?f(k0)dF(?)
The second term of the right hand side is positive and the flrst term is greater
than the third one under the assumption that rs1 = r1. So the right hand side is
unambiguously positive. So, the left hand side should be positive and not zero
as it is under our original assumption.
73
There is a contradiction.
Now we have to show that rs1 cannot be greater than r1. Again we proceed
by contradiction. Assume rs1 > r1, which implies that Rs1 > R1. There are
two possible cases: ^? < ?? and ^? > ??. In the flrst case the proof is the same
as before. In the second case, split the integral limits as above, but now with
^? > ??.We get
Z ??
^?
(r1 ?rs1)dF(?) =
? 1
L1 ?
1
L0 +L1
?Z ??
0
?f(k0)dF(?)+
Z ^?
??
[?f(k0)?r1]dF(?)
The second term of the right hand side is positive under our assumption that
^? > ??. Conditional on ? being greater than ?? and lower than ^? output is
greater than r1. This is because output is higher than the necessary to totally
repay the contractual interest rate r1 (i.e. ? > ??). So, the left hand side is
unambiguously positive and so should be the left hand side. But this contradicts
our initial assumption.We conclude that rs1 must be lower than r1.
74
A.3 Proof that @^?@k0 < 0
From equation (1.8), deflne the function F(k0; ^?):
F(k0; ^?) ? ^?? ^?1+ ^?
Z ??
0
? f(k0)k
0| {z }
A
dF(?)?
?
Z ??
Min[?0;??]
2
66
4
 1
1+ ^?
?
? f(k0)k
0| {z }
A
? D0k
0|{z}
B
3
77
5dF(?) = 0
Applying the implicit function theorem to this expression:
@^?
@k0 = ?
@F(:)
@k0
@F(:)
@^?
@F(:)
@k0 = ?
^?
1+ ^?E(?)
@A
@k0 ?
Z ??
Min[?0;??]
? 1
1+ ^?
?
?@A@k
0
? @B@k
0
?
dF(?) > 0
Since A is a concave function and B is a convex function (analogous to Lemma
1), this expression is greater than zero.
@F(:)
@^? = 1?
1
(1+ ^?)2E(?)
f(k0)
k0 +
Z ??
Min[?0;??]
1
(1+ ^?)2?
f(k0)
k0 dF(?)
This expression will have the same sign as:
(1+ ^?)? 1(1+ ^?)E(?)f(k0)k
0
+
Z ??
Min[?0;??]
1
(1+ ^?)?
f(k0)
k0 dF(?);
from the deflnition of ^? (equation (1.8)) we have that:
1
1+ ^?E(?)
f(k0)
k0 < 1
75
so that,
@F(:)
@^? > 0
These imply that @^?@k0 < 0.
76
A.4 Data Sources and Construction of Vari-
ables
Bond characteristics
The bond dataset, obtained from Bondware, supplemented by the former
Emerging Markets Division of the International Monetary Fund for the early
1990s, covers the period 1991 to 1999 and includes: (1) launch spreads over risk
free rates (in basis points, where one basis point is one-hundredth of a percentage
point); (2) the amount of the issue (millions of dollars); (3) the maturity in years;
(4) whether the borrower was a sovereign, other public sector entity, or private
debtor; (5) currency of issue; (6) whether the bond had a flxed or  oating rate;
and (7)the borrower?s industrial sector: manufacturing, flnancial services, utility
or infrastructure, other services, or government (where government, in this case,
refers to subsovereign entities and central banks, which could not be classifled
in the other four industrial sectors). Global variables included the United States
industrial production growth rate, constructed as average month-month growth
rate over a quarter; the United States ten-year swap spread; and the quarterly
standard deviation of log difierences of daily spreads of the Emerging Market
Bond Index.
Global variables
United States industrial production growth rate: average of month-month
growth rate over a quarter.
United States ten-year swap spread.
Emerging Market Bond Index: standard deviation of difierence in log of daily
spreads.
77
Table 5: Country Characteristics
Variable (Billions) Periodicity Source
Total external debt US$ Annual WEO
(EDT)
Gross national product US$ Annual WEO
(GNP, current prices)
Gross domestic product National Annual WEO
(GDPNC, current prices)
Gross domestic product National Annual WEO
(GDP90, 1990 prices)
Total debt service US$ Annual WEO
(TDS)
Exports (XGS) US$ Annual WEO
Exports (X) US$ Monthly IFS
Reserves US$ Quarterly IFS
(RESIMF)
Imports (IMP) US$ Quarterly IFS
Domestic bank credit National Quarterly IFS
(CLM PVT)1
Short term bank debt (BISSHT)2 US$ semi-annual BIS
Total bank debt (BISTOT)3 US$ semi-annual BIS
Credit rating (CRTG) Scale semi-annual Institutional
Investor
Debt rescheduling (DRES)4 Indicator Annual WDT/GDF
78
Table 5 (Continued). Country Characteristics
Constructed Variables
Debt/GNP EDT/GNP
Debt service/exports TDS/XGS
GDP/growth 0.25 * ln[GDP90 t/GDP90 t-1]
Reserves/imports RESIMF/IMP
Reserves/GNP RESIMF/GNP
Reserves/short-term debt RESIMF/BISSHT
Short-term debt/total debt BISSHT/BISTOT
Domestic credit/GDP CLM PVT/(GDPNC/4)
Sources: International Monetary Fund?s World Economic Outlook (WEO) and In-
ternational Financial Statistics (IFS); IMF program data from the IMF?s Executive
Board Documents and Stafi Estimates; World Bank?s World Debt Tables (WDT) and
Global Development Finance (GDF); Bank of International Settlements, The Maturity,
Sectoral, and Nationality Distribution of International Bank Lending. Credit ratings
were obtained from Institutional Investor?s Country Credit Ratings. Missing data for
some countries was completed using the US State Department?s Annual Country re-
ports on Economic Policy and Trade Practices (which are available on the internet
from http:www.state.gov/www/issues/economic/trade reports/). U.S. industrial pro-
duction, Federal Reserve Swap rates and EMBI data are taken from Bloomberg.
1 Credit to private sector.
2 Cross-border bank claims in all currencies and local claims in nonlocal currencies of
maturity up to and including one year.
3 Total consolidated cross-border claims in all currencies and local claims in nonlocal
currencies.
4 Indicator variable, which is equal to one if a debt rescheduling took place in the
previous year and zero otherwise.
79
A.5 Base Regression
In this appendix we present the full details of the base regression, which corre-
sponds to Column 1 of Table 2.3. As noted, in Tables 2.3 and 2.4 of the main
text we present only the variables of direct interest to this paper. The signs and
signiflcance of the controls variables presented here remain very similar across the
various variations in Tables 2.3 and 2.4. The flrst two columns of the table in this
appendix present the coe?cient and z-statistic for the variables in the selection
equation; and the next two columns refer to the spreads equation.
While much of the table is self-explanatory, a few comments are in order. In
earlier work [e.g., Eichengreen and Mody, 2001], we used the United States? 10-
year treasury rate as one of the \global" variables. That variable gave ambiguous
signs. In ongoing work, we flnd that the U.S. industrial growth rate gives a
consistent sign and also has an intuitive explanation in terms of U.S. higher
growth improving credit quality for emerging market borrowers. Thus, higher
U.S. growth is associated with lower spreads and more frequent bond issuance,
as if the demand for emerging market bonds shifts to the right when the United
States grows more rapidly. Another new variable used in this analysis is the
quarterly standard deviation of the daily log change of the EMBI index. A higher
standard deviation implies greater market uncertainty with respect to pricing of
bonds. We flnd that such uncertainty reduces bond issuance signiflcantly and
raises spreads (that the efiect on spreads is not always signiflcant at the 5 percent
level).
80
Table 6: Base Regression Results
Probability of Bond Issunace Log of Spread at Time of Issue
Coe?cient z-statistic Coe?cient z-statistic
Bond Characteristics
Log amount -0.031 (-2.14)
Maturity 0.010 (4.98)
Yen -0.321 (-6.97)
Deutsche Mark -0.091 (-2.09)
Euro -0.058 (-1.24)
Other currencies -0.190 (-4.39)
Fixed rate 0.366 (11.04)
Global Variables
U.S. growth rate 52.908 (10.90) -25.052 (-5.25)
Log swap rate -0.319 (-8.28) 0.460 (11.61)
EMBI volatility -17.359 (-11.15) 6.059 (4.27)
Country Characteristics
Credit rating 0.033 (29.79) -0.044 (-26.70)
Debt/GNP -1.264 (-15.61) 0.970 (10.77)
Debt service/exports 1.281 (24.87)
Debt restructured dummy 1.058 (15.15) -0.450 (-9.72)
GDP growth 0.994 (0.93) -9.372 (-6.58)
Short-term debt/total debt -0.674 (-8.91) 0.841 (7.42)
Export growth volatility -2.118 (-5.71) 0.666 (3.10)
Reserves/imports 0.073 (8.22) -0.006 (-0.52)
Bank credit stock/GDP -0.000 (-0.51) 0.000 (1.37)
Sector
Public 0.024 (0.61) 0.033 (0.54)
Finance -0.127 (-1.96)
Services 0.506 (3.23)
Utilities -0.085 (-1.26)
Private 0.639 (25.19) 0.083 (1.75)
Finance -0.199 (-6.18)
Services 0.129 (2.32)
Utilities 0.021 (0.63)
Latin America dummy 0.021 (0.63)
IMF program dummy 0.337 (10.60) -0.089 (-3.06)
Constant -0.249 (-1.41) 5.238 (28.02)
Lambda -0.520 (-11.55)
Number of observations 7882
Number of Bonds 2990
Note: z-statistics, based on robust standard errors, are presented in parentheses.
81
Bibliography
Bagci, P., and W. Perraudin, \The Impact of IMF Programs," The Institute
for Financial Research, Working Paper IFR35 (London: Birkbeck College,
University of London, 1997).
Barro, R. J., and J. W. Lee, \IMF Programs: Who Is Chosen and What Are
the Efiects?" paper presented at the Second Annual IMF Research Confer-
ence, Washington, November 7-8, 2001. Also available on the internet at:
www.imf.org/external/pubs/ft/stafip/2001/00-00/arc.htm.
Bird, G., and D. Rowlands, \The Catalytic Efiect of Lending by the Interna-
tional Financial Institutions," World Economy, Vol. XX, No. 7 (November
1997), 81-97.
|, and | \Catalysis or Direct Borrowing: The Role of the IMF in Mobilising
Private Capital," World Economy, Vol. XXIV, No.1 (January 2001a), 81-
97.
|, and | \IMF Lending: How is it Afiected by Economic, Political, and
Institutional Factors?" Policy Reform, Vol. IV, (2001b), 243-270.
|, and | \Do IMF Programmes Have a Catalytic Efiect on Other Interna-
tional Capital Flows?" Oxford Development Studies, Vol. XXX, No. 3,
(2002), 229-49.
Bordo, M. and H. Rockofi, \The Gold Standard as a Good Housekeeping Seal
of Approval," Journal of Economic History, Vol. LVI, No. 2 (June, 1996),
389-428.
Bordo, M., M. Edelstein, and H. Rockofi, \Was Adherence to the Gold Stan-
dard a ?Good Housekeeping Seal of Approval? During the Interwar Period?"
NBER Working Paper No. 7186 (1999). Cambridge, Massachusetts: Na-
tional Bureau of Economic Research.
Boughton, J.M., \The Silent Revolution: The International Monetary Fund:
1979-1989," (Washington: International Monetary Fund, 2001).
82
Broner, F., G., Lorenzoni, and S. Schmuckler, (2003), \Why Do Emerging
Markets Borrow Short Term?". Mimeo. University of Maryland.
Caballero, R.J. and A., Krishnamurthy, (2001), \International and Domestic
Collateral Constraints in a Model of Emerging Markets Crisis", Journal of
Monetary Economics, Vol. 48, pp. 513-548.
Calomiris, C., (1998), \The IMF?s Imprudent Role as Lender of Last Resort",
The Cato Journal, Vol. 17, No.3.
Cat~ao, Luis A., and Bennett W. Sutton, \Sovereign Defaults: The Role of
Volatility," IMF Working Paper No. 02/149 (Washington: International
Monetary Fund, 2002).
Clementi, G.L., and H.A., Hopenhayn, (2002), \A Theory of Financing Con-
straints and Firm Dynamics", RCER Working Papers 492, University of
Rochester-Center for Economic Research.
Conway, P., \IMF lending programs: Participation and Impact," Journal of
Development Economics, Vol. XLV, No 2 (December 1994), 365-91.
|, \IMF Programs and External Balance: the Crisis-Participation Dynamic"
(University of North Carolina, mimeo, 2000).
|, \Evaluating Fund Programs" (unpublished; Chapel Hill: University of
North Carolina, 2001).
Corsetti, G., B., Guimar~aes , and N. Roubini, (2003), "The Tradeofi Between an
International Lender of Last Resort to Deal with Liquidity Crisis and Moral
Hazard Distortions: A Model of the IMF?s Catalytic Finance Approach.",
www.stern.nyu.edu/globalmacro/.
Cottarelli, C., and C. Giannini, (2002), \ Bedfellows, Hostages, or Perfect
Strangers?: Global Capital Markets and the Catalytic Efiect of IMF Crisis
Lending", IMF working paper: wp/02/193, November.
Dell?Ariccia, Giovanni, Isabel Schnabel, and Jeromin Zettelmeyer, \Moral Haz-
ard and International Crisis Lending: A Test", IMF Working Paper 02/181
(Washington: International Monetary Fund, 2002).
Dhonte, P., \Conditionality as an Instrument of Borrower Credibility." IMF
Paper on Policy Analysis and Assessment, No. 2 (1997).
Diamond, D.W., (1993), \Seniority and Maturity of Debt Contracts", Journal
of Financial Economics, 33, pp. 341-368.
83
Dicks-Mireaux, L., M. Mecagni, and S. Schadler, \Evaluating the Efiect of IMF
Lending to Low-Income Countries," Journal of Development Economics,
Vol. LXI (2000), 495-526.
Eaton, J., and R., Fernandez, \Sovereign Debt", Handbook of International
Economics, Vol. 3, Eds: Grossman, G. and K. Rogofi, Chapter 39.
Eaton, J., and M. Gersovitz, (1981),  ?Debt with Potential Repudiation: The-
oretical and Empirical Analysis",Review of Economic Studies, Vol.48, pp.
289-309.
Edwards, S., \The International Monetary Fund and the Developing Coun-
tries: A Critical Evaluation," NBER Working Paper 2909 (Cambridge,
Massachusetts: National Bureau of Economic Research, 1989).
Edwards, M. S., \Reevaluating the Catalytic Efiect of IMF Programs," pre-
sented at the Annual Meeting of the American Political Science Association
(2000). Available via the Internet www.rci.rutgers.edu n mse.
Eichengreen, B. (2002), \Financial Crisis and what to do about them", Oxford
University Press, Oxford.
Eichengreen, B. and A. Mody, \Bail-Ins, Bail-Outs, and Borrowing Costs,"
IMF Stafi Papers, Vol. XLVII (2001), 155-87.
Fisher, S., \Applied Economics in Action: IMF Programs," The American
Economic Review, Vol. LXXXVII (1997), 23-27.
Fischer, S., (1999), \On the Need of for an International Lender of Last Resort",
The Journal of Economic Perspectives, Vol. 13, pp. 85-104.
Freixas, X., and J.C., Rochet, \Microeconomic of Banking", The MIT Press,
Chapters 4 and 5.
Garuda, G., \The Distributional Efiects of IMF Programs: A Cross-Country
Analysis." World Development, Vol. XXVIII, No. 6 (2000), 1031-51.
Goldstein, M., and P. Montiel, \Evaluating Fund Stabilization Programs with
Multicountry Data: Some Methodological Pitfalls," IMF Stafi Papers, Vol.
XXXIII (1986), 304-44.
Hajivassiliou, V., \The External Debt Repayment Problems of LDC?s: An
Econometric Model Based on Panel Data," Journal of Econometrics, Vol.
XXXVI (1987), 205-30.
Hale, D., \The IMF, Now More than Ever: The Case for Financial Peacekeep-
ing," Foreign Afiairs, Vol. LXXVII, No. 6 (1998), 7-13.
84
Heckman, James, \Sample Selection Bias as a Speciflcation Error," Economet-
rica, Vol. XLVII (1979), 153-161.
Holmstrom, B., and J., Tirole , (1998) ?Private and Public Supply of Liquidity",
Journal of Political Ecomomy, Vol. 106, pp. 1-36.
Hutchison, M., \A Cure Worse Than The Disease? Currency Crises and the
Output Costs of IMF-Supported Stabilization Programs," NBER Working
Paper No. 8305 (Cambridge, Massachusetts: National Bureau of Economic
Research, 2001).
International Monetary Fund, \Financial Organization and Operations of the
IMF," Treasurer?s Department (Washington: International Monetary Fund,
2001a).
International Monetary Fund Survey Vol. 31 No. 7 April 8 2000 \Krueger
Modifles Sovereign Debt Plan to Lighten IMF Role, After Wide Consulta-
tion".
|, \Selected Decisions and Selected Documents of the International Monetary
Fund" (Washington: International Monetary Fund, 2001b).
|, \Evaluation of the Prolonged Use of Fund Resources," Volumes I and II,
(Washington: International Monetary Fund, 2002).
Ivanova, A., W. Mayer, A. Mourmouras, and G. Anayiotos, \What Determines
the Implementation of IMF-Supported Programs?" IMF Working Paper
No. 03/8 (Washington: International Monetary Fund, 2001).
Kaminsky, G.L. and S.L. Schmukler, \What Triggers Market Jitters: A Chron-
icle of the Asian Crisis," Journal of International Money and Finance, Vol.
XVIII (1999), 537-60.
Khan, M., \The Macroeconomic Efiects of Fund-Supported Adjustment Pro-
grams," IMF Stafi Papers, Vol. XXXVII (1990), 195-231.
Killick, T., M. Malik, and M. Manuel, \What Can We Know About the Efiects
of IMF Programmes?" World Economy (U.K.), Vol. XV (1992), 575-97.
Kletzer, K.M., (1984), \Asymmetries Of Information And LDC Borrowing
With Sovereign Risk", The Economic Journal, 94, June, pp. 287-307.
K?ohler, Horst, \Promoting Stability and Prosperity in a Globalized World."
Remarks by IMF Managing Director, Council of the Americas, Washington,
DC, May 7, 2001.
http://www.imf.org/external/np/speeches/2001/050701.htm
85
Krueger, A., \Con icting Demands on the International Monetary Fund,"
American Economic Association Papers and Proceedings, Vol. XC, No.
2 (2000), 38-42.
Krugman, P.R., (1988), \Financing Versus Forgiving a Debt Overhang", Jour-
nal of Development Economics, 29, 1988.
Krugman, P.R., (1989), ?Market Based Debt Reduction Schemes", Eds: J.
Frankel, M. Dooley, and P. Wickhan, Analytical Issues in Debt, Washing-
ton: International Monetary Fund, pp. 258-278.
Lamont, O., (1995), \Corporate-Debt Overhang and Macroeconomic Expecta-
tions", American Economic Review, Vol. 85, December, pp. 1106-1117.
Lane, T., and S. Phillips, \Does IMF Financing Result in Moral Hazard," IMF
Working Paper No. 00/168 (Washington: International Monetary Fund,
2000).
Leung, S. F., and S. Yu, \On the Choice Between Sample Selection and Two-
Part Models," Journal of Econometrics, Vol. LXXII (1996), 197-229.
Marchesi, S., \Adoption of an IMF Programme and Debt Rescheduling. An
Empirical Analysis," Centro Studi Luca D? Agliano, Development Studies
Working Papers, No. 152 (2001).
|, and J. P. Thomas, \IMF Conditionality as a Screening Device," The Eco-
nomic Journal, Vol. CIX (1999), 111-125.
Morris, Stephen, and Hyun Shong Shin, \Catalytic Finance: When Does it
Work?" Cowles Foundation Discussion Paper No. 1400 (February 2003).
Mussa, M., and M. Savastano, \The IMF Approach to Economic Stabilization,"
IMF Working Paper No. 99/104 (Washington: International Monetary
Fund, 1999).
Obstfeld, M., and K. Rogofi, \Foundations of International Economics", The
MIT Press, Chapter 6.
Obstfeld, M., and A. M. Taylor, \Sovereign Risk, Credibility and Gold Stan-
dard: 1870-1913 Versus 1925-31," NBER Working Paper 9345 (Cambridge,
Massachusetts: National Bureau of Economic Research, 2002).
Ozler, S., \Have Commercial Banks Ignored History?" The American Eco-
nomic Review, Vol. LXXXIII (1993), 608-20.
Parlour, C.A., and U. Rajan, (2001), \Competition in Loan Contracts", Amer-
ican Economic Review, Vol. 91, December, pp. 1311-1328.
86
Pattillo, Catherine, H?el enePoirson, andLucaRicci, "ExternalDebtandGrowth,"
IMF Working Paper No. 02/69 (Washington: International Monetary
Fund, 2002).
Penalver, A. , (2002), \How can IMF catalyze private capital  ows? A model",
Bank of England, forthcoming working paper.
Powell, A, \Countries in Payments? Di?culties: What Can the IMF Do?"
International Monetary Fund Seminar Series (International); No. 2002-
66:1-[36], May 2, 2002.
Prezeworski, Adam and James Raymond Vreeland, \The Efiect of IMF Pro-
grams on Economic Growth," Journal of Development Economics, Vol.
LXII (2000), 385-421.
Rodrik, D., (1995), \Why is there Multilateral Lending?", NBER working pa-
per No. 5160, June 1995.
Rodrik, D., and A. Velasco, (1999), \Short Term Capital Flows", NBER work-
ing paper No. 7364, September 1999.
Schadler, S., and others, IMF Conditionality: Experience Under Stand-By and
Extended Arrangements, Part I: Key Issues and Findings, IMF Occasional
Paper No. 128 (Washington: International Monetary Fund, 1995a).
|, and | IMF Conditionality: Experience Under Stand-By and Extended
Arrangements, Part II: Background Papers, IMF Occasional Paper No. 129
(Washington: International Monetary Fund, 1995b). Stone, R., "Lending
Credibility" (Princeton, New Jersey: Princeton University Press, 2002).
Tirole, J., Financial Crises, Liquidity, and the International Monetary System
(Princeton, New Jersey: Princeton University Press, 2002).
Winton, A., (1995), \ Costly State Veriflcation and Multiple Investors: The
Role of Seniority"., The Review of Financial Studies, Vol. 8, Spring, pp.
91-123.
Zettelmeyer, J., (2000), \Can O?cial Crisis Lending be Counterproductive in
the Short Run?", Economic Notes by Banca Monte dei Paschi di Siena SpA,
vol. 29, no. 1-2000, pp.13-29.
87