Need for speed: The lending responsiveness
of the IMF
Daniel McDowell
The Review of International
Organizations
ISSN 1559-7431
Rev Int Organ
DOI 10.1007/s11558-015-9240-x
1 23
Author's personal copy
Rev Int Organ
DOI 10.1007/s11558-015-9240-x
Need for speed: The lending responsiveness of the IMF
Daniel McDowell1
© Springer Science+Business Media New York 2016
Abstract How responsive a lender is the International Monetary Fund (IMF)? In
this paper, I introduce new data on IMF loan approval periods: The days that transpire between when a borrower submits a “Letter of Intent” to the Executive Board
requesting a loan and when the Board approves that request. The data reveal considerable variation across requests. Why are some loan requests approved swiftly while
others wait much longer for approval? I argue that the financial interests of the G-5
economies drive variation in responsiveness contingent on when a request was made.
I expect that during much of the 1980s, as G-5 commercial bank exposure increases,
borrowers will face longer waits for approval. In such cases, the G-5 should have been
more likely to press for the use of the “concerted lending” strategy. This protected
G-5 financial systems by catalyzing private financing on behalf of those countries,
but it also delayed loan approval. Into the 1990s, global capital flows grew more
complex and catalyzing private capital flows required a swift response. Thus, during these years I expect increased G-5 bank exposure to be associated with shorter
waits for approval. In such cases, the G-5 should have been more likely to press for
I would like to thank J. Lawrence Broz, Benjamin Cohen, Mark Copelovitch, Jeffry Frieden, Randall
Henning, Christopher Kilby, Steven Liao, Tom Pepinsky, David Steinberg, Michael Tierney, and
participants at the 2014 International Political Economy Society (IPES) conference for their helpful
comments on this paper. I would also like to thank the Appleby-Mosher Fund for Research for their
generous support of this project.
Electronic supplementary material The online version of this article
(doi:10.1007/s11558-015-9240-x) contains supplementary material, which is available to authorized
users.
Daniel McDowell
[email protected]
1
Department of Political Science, Maxwell School, Syracuse University, 100 Eggers Hall,
Syracuse, NY 13244, USA
Author's personal copy
D. McDowell
accelerated approval. A quick response would have the best chance of attracting back
private capital and reduce the threat posed by the crisis. Statistical analyses of 275
loan requests from 1984–2012 support these expectations.
Keywords International Monetary Fund · Catalytic financing ·
International lender of last resort
JEL Classification F3 · G2
1 Introduction
In recent years, scholars of political economy have built an impressive, cohesive literature analyzing the lending practices of the International Monetary
Fund (IMF). This research has produced important new knowledge about the
factors that influence variation in IMF loan selection, loan terms, and loan
size. Most notably, a number of studies have found consistent evidence that
the economic and geopolitical interests and ideas of the dominant countries within the institution influence IMF lending behavior (Broz and Hawes
2006; Copelovitch 2010; Dreher and Jensen 2007; Nelson 2014; Oatley and
Yackee 2004; Moser and Sturm 2011; Thacker 1999; Stone 2004, 2008, 2011;
Vreeland 2003, 2007). Yet, despite this rich body of literature, there is one component
of IMF lending we know relatively little about: the Fund’s lending responsiveness.
By responsiveness, I mean the speed with which the Executive Board (henceforth,
Board) approves government loan requests once they are formally filed with the institution. The issue of IMF lending responsiveness is important given the fact that the
Fund has assumed the role of de facto international lender of last resort for more than
three decades (Boughton 2000; Sachs 1995). Since Walter Bagehot’s (1873) essay
Lombard Street: A Description of the Money Market, it has been understood that
effective management of financial crises requires the lender of last resort to swiftly
bring substantial resources to bear.1 Indeed, in today’s global financial system, the
speed with which an economy in crisis can acquire emergency funding is arguably
as important as the size of the loan itself (Bordo-and-James 2000; Fernandez-Arias
2010; Fernandez-Arias and Levy-Yeyati 2010, 54). What good is a fire truck if it
arrives after the house has already burned to the ground?
This paper introduces and analyzes new data on what I refer to as IMF “loan
approval periods”: The number of days that transpire between the date an IMF member country formally submits a Letter of Intent to the Board requesting a loan and
the date the Board approves that request. Because the Board needs time to consider
the merits of each proposed program before putting it to a vote, I assume there is a
baseline approval period for a standard loan request. However, I argue that powerful
1 Bagehot asserted that during crises the Bank of England “should lend to all that bring good securities
quickly, freely, and readily. By that policy they allay a panic; by every other policy they intensify it”
(Bagehot 1873).
Author's personal copy
Need for speed: The lending responsiveness of the IMF
members on the Board–specifically, the G-5 economies–can invervene in ways that
speed up or slow down the approval process.2 I make two central claims about this
process. First, the decision to intervene is directly related to the exposure of G-5
commercial banks to the borrower country’s economy. The effects of foreign financial crises often spill across borders. In some cases, they can threaten the stability
of creditor country financial systems if they are sufficiently exposed to the afflicted
economy. Facing circumstances where their financial systems are threatened, the
G-5 should prefer the Fund to respond in a way that most effectively catalyzes private
financial flows on behalf of the borrower. All else equal, this should more swiftly
resolve the crisis and minimize the threat it poses. As I will argue, however, catalytic
lending strategies affect IMF responsiveness. Moreover, the IMF’s preferred catalytic
lending strategy has changed over time.
This relates to my second claim: The direction of the effect of the G-5’s
intervention–whether the approval period is lengthened or shortened–depends on the
time period in which the loan request was made. During much of the 1980s, the
IMF used the “concerted lending” strategy to forcibly catalyze private capital from a
relatively small and homogeneous group of global banks. The strategy relied on additional negotiations with these banks in between the submission of the Letter and the
Board’s approval. Only when the banks agreed to provide new loans to the borrower
in question would the Board approve the request. Consequently, efforts by the Board
to catalyze private capital flows in the 1980s should be associated with longer loan
approval periods. Around the turn of the next decade, the IMF changed its approach
to catalytic financing. Developing countries were opening up their financial markets.
Many became increasingly reliant on a more disaggregated and heterogeneous group
of international financiers that could pull out at a moment’s notice. In this context,
the Fund changed its approach to catalytic financing. Its efforts were designed to
reinstate market confidence through the provision of large loans as quickly as possible. Thus, attempts by the Board to catalyze private capital flows in the later period
should be associated with shorter loan approval periods.
To sum up, I expect that the strength of the G-5’s preference for catalytic financing increases as their commercial banks become more exposed to the borrower in
question. When the stakes are high, I expect the G-5 will use their influence within
the Board to improve the expected catalytic effect of its response. During the 1980s,
greater G-5 bank exposure should result in longer waits for approval. In such cases,
G-5 governments should have been more inclined to press for the use of concerted
lending which catalyzed private capital but delayed approval. Moreover, I expect
that the intensity of the delaying effect of bank exposure on responsiveness during
this era should increase as the borrower’s debt burden grows since in such cases the
risk of default is highest. Into the 1990s and beyond, the relationship should invert:
Waits for approval should decrease as G-5 bank exposure increases. In such cases,
the G-5 governments should have been more likely to press for an expedited vote
by the Board based on the belief that rapid disbursement would most effectively catalyze private capital. Additionally, I expect that the intensity of the hastening effect
2 The
G-5 includes the United States, United Kingdom, Germany, Japan, and France.
Author's personal copy
D. McDowell
of bank exposure on responsiveness should increase as a borrower becomes more
dependent on bond markets and short-term debt since in such cases the threat of capital flight should be highest. Empirical analysis of 275 IMF loan approval periods for
87 countries between 1984 and 2012 largely support these expectations.
This paper proceeds as follows. First, I present the puzzle this paper aims to
explain: variation in the duration of loan approval periods. Here I discuss existing
work on the topic, describe how I operationalize the dependent variable, introduce
my data on IMF lending responsiveness, and examine its distribution across requests
and time. Next, I present my argument, develop testable hypotheses and introduce
my research design. Finally, I summarize my empirical results before offering some
concluding thoughts.
2 The puzzle: Variation in IMF responsiveness
How should we measure IMF responsiveness? Answering this question is not as
straightforward as it might initially seem. In the first subsection below, I discuss how
existing studies on related topics measure this and summarize their findings. Next, I
outline how I measure responsiveness and explain how it is different from existing
measures. Finally, I briefly present my descriptive data.
2.1 Existing studies
Interest in the speed of financial disbursements by international financial institutions
(IFIs) has only recently captured the attention of scholars. In one study, Kilby (2013)
examines the speed of World Bank project preparation. He defines responsiveness as
the time that transpires between the date the World Bank begins the process of identifying a project and the date on which the Bank approves the program. However,
because the World Bank does not make project identification dates publicly available, Kilby adopts a novel approach to estimating the duration of project approval.
Using project identification numbers, he employs a stochastic frontier model to estimate the identification date and the duration of the preparation period. Preparation
time varies, Kilby finds, from less than one year to more than 10 years in duration.
Kilby expects that powerful donor countries at the Bank intervene to influence the
speed of project preparation when doing so serves their interests or the interests of an
important client state. The results largely support Kilby’s argument. Project preparation, and thus disbursement of World Bank funds, is hastened for borrowers that vote
closely with the U.S. at the United Nations (UN). Voting similarity with the Group
of Seven (G-7) countries, however, is not robustly associated with quick approvals.
Kilby also finds that countries holding temporary seats on the UN Security Council (UNSC) and those that have a representative on the World Bank Executive Board
have their projects approved more swiftly.
In another study more closely related to this one, Mody and Saravia (2013) investigate the response speed of the IMF. In their analysis, the authors conceive of
Author's personal copy
Need for speed: The lending responsiveness of the IMF
responsiveness as the time that transpires between the onset of a financial crisis and
the conclusion of the negotiation of an IMF program. To identify the onset of a speculative attack on a nation’s currency, Mody and Saravia use what is sometimes referred
to as the “Exchange Market Pressure” (EMP) index developed by Eichengreen et al.
(1995) and Kaminsky and Reinhart (1999). In some cases, this pressure may result
in a currency devaluation; in others, the government may defend the currency, resulting in a loss of reserves. The index is based on monthly changes in the exchange
rate and reserves. A higher score indicates greater speculative pressure. Mody and
Saravia identify a crisis when the index is more than one standard deviation above the
country specific mean for a given month.3 They then use the IMF’s “Date of Arrangement” for each Stand-by Arrangement (SBA) as the date on which a program was
initiated. The period of time–measured in months–from the month of arrangement
and the month of the first crisis identified in the previous two years is the dependent
variable in the study, what they call “the spell.”
Mody and Saravia use a Poisson regression to assess the determinants of response
speed against a sample of 183 SBAs ranging from 1977 through 2004. Their analysis
of the full sample finds that the Fund responds more swiftly to countries experiencing severe crises, those with larger IMF quotas, and cases where the size of the
loan requested (relative to the country’s quota) is larger. Conversely, the IMF is less
responsive to borrowers that are already under a program. They find no evidence,
based on their full sample, that increased economic ties with the U.S. influences IMF
speed. They do, however, find increased United Nations General Assembly (UNGA)
voting affinity with the U.S. is associated with faster program negotiations. However,
this result only applies to a restricted sample of more severe crises.4 Mody and Saravia also separate their full sample into two sub-samples: one which includes cases
before 1986 and a second that includes those occurring after that year. The division
is based on the assertion that the IMF changed its approach to crisis management following the 1980s Latin American debt crisis. The additional analyses find that crisis
severity and loan size are only associated with shorter “spells” in the second period.
Additionally, they find that democracy correlates with slowed responsiveness in the
first period. They interpret this to mean that “democracies had not yet matured and
hence had difficulty dealing speedily with crises.” Conversely, they find that democracy is linked to increased responsiveness in the second period suggesting that “as
capital markets came to place greater value on speed, democracies adapted to speed
up decisions” (Mody and Saravia 2013, 207).
In the following section, I explain how my measure of responsiveness differs
from Mody and Saravia. While neither measure is without its flaws, I argue that
3 Mody and Saravia use a cutoff of one standard deviation which is, they admit, a “softer” approach than
(Kaminsky and Reinhart 1999) which identifies speculative attacks at three standard deviations above the
mean.
4 Severe crises are identified when EMP is 1.5 standard deviations above the country mean.
Author's personal copy
D. McDowell
Fig. 1 IMF loan request and approval stages
my approach provides a more accurate measure of IMF behavior independent of the
borrower country government.
2.2 Measuring IMF lending responsiveness
The process by which a member country obtains financial assistance from the IMF
can be divided into three main stages depicted in Fig. 1. Stage 1 begins when some
event triggers the borrower country to consider approaching the IMF for assistance.
In many cases, this is perpetuated by the onset of a crisis or financial hardship. In
other cases the decision may be made under more stable conditions. The first stage
ends when the government formally contacts the IMF for assistance. The duration
of this stage is wholly dependent on the borrower country government’s sense of
urgency as the Fund cannot compel countries to seek out its services. Stage 2 begins
when the member country initially contacts the Fund, expressing interest in seeking
assistance and ends when a deal between the government and Fund staff is completed.
Before formally requesting a loan, the country must enter into negotiations with IMF
staff. These talks determine the proposed loan’s terms including its size, maturity,
and conditionality. The duration of this stage is jointly dependent on the borrower
country government and IMF staff. That is, variation in the negotiation time between
the borrower country and IMF staff does not exclusively reflect on the IMF’s responsiveness. In such bargaining situations, the length of negotiations are a function of
the preferences and strategies adopted by both parties.5 Stage 3 begins after negotiations have completed resulting in an official “Letter of Intent” and “Memorandum of
Economic and Financial Policies” (henceforth “Letter” and “Memo”) outlining the
objectives and macroeconomic and structural adjustments that the borrower government has agreed to implement in exchange for a loan. These documents are circulated
among the 24 Board members (known as Executive Directors) for review. The Managing Director (or sometimes the First Deputy Managing Director) then places the
request on the Board’s agenda for consideration at some future date.6 On that date,
the Board decides whether to approve or reject the request. The duration of the third
stage, then, is a function of the priority that the Managing Director and the Board
gives to a particular request.
Mody and Saravia’s approach lumps all three stages together in one measure. On
one hand, this approach may be viewed as the most appropriate for measuring the
Fund’s response speed since it adopts a holistic view of the concept. On the other
5 From a practical perspective, there is also no readily available record of when each country first began
negotiations with Fund staff.
6 This procedure was confirmed in a phone conversation with a high level IMF official.
Author's personal copy
Need for speed: The lending responsiveness of the IMF
hand, a definition that encompasses all three stages is measuring far more than just
the IMF’s responsiveness. It is also taking into account a number of factors that lie
outside of the IMF’s control. Namely, the borrowing country government’s sense of
urgency in approaching the IMF and its subsequent resolve in negotiations. Moreover,
since the onset of financial crises do not come with dates attached, the accuracy of
their approach depends on the way in which financial crisis onset is measured. That
is, the authors create an objective measure of when a crisis begins. However, this
may not necessarily coincide with the actual date on which policymakers subjectively
interpreted their economies as being in real trouble. Finally, Mody and Saravia’s
sample only includes cases where speculative pressure on a borrower’s currency is
identified. Neither debt crises nor banking crises are captured by this measure and,
unless they coincide with a currency crisis, they are left out of the sample. Indeed, in
a footnote, the authors acknowledge: “The focus on currency crises is determined by
the practical difficulty of dating, for example, banking and debt crises” (Mody and
Saravia 2013, 194). The authors admit that more than 100 SBAs during the period
they investigate are left out of their analysis.7 Thus, while their measure may be more
complete, it is also more complex and systematically leaves out cases that do not
fit their precise definition of a “crisis.” In short, it is a very noisy and incomplete
indicator of the IMF’s lending responsiveness. In contrast, I focus on the duration of
Stage 3. While this is admittedly a more narrow interpretation of responsiveness, it
isolates Fund behavior from a number of influences outside of its direct control. Thus,
while more narrow it is considerably less noisy. Moreover, my measure does not
omit cases as inclusion in the sample depends only on a Letter and Board approval–
something all IMF credits posses.
2.3 The dependent variable
I focused my data collection efforts on the Fund’s two primary non-concessional
lending facilities: the Stand-by Arrangement (SBA) and the Extended Fund Facility (EFF).8 For sixty years, the IMF’s workhorse emergency lending mechanism for
countries facing short-term balance-of-payments problems has been the SBA. In the
mid-1970s, the Fund introduced the EFF as a complement to the SBA for countries needing medium-term balance-of-payments assistance. Together, these facilities
account for the vast majority of the Fund’s non-concessional lending activities dating back to the institution’s creation. To calculate the loan approval period for each
request, I recorded the date on each Letter that was submitted to the IMF between
1984 and 2012. I then calculated the number of days that transpired between the date
of the Letter and the date at which the Board approved the request. Identifying the
7 “While
there were about 300 SBAs during this period, only 200 of them coincide with crises in the
previous two years” (Mody and Saravia 2013, 192).
8 The IMF’s concessional loan programs are primarily designed to promote economic development and
reduce poverty rather than respond to financial crises. Thus, from this paper’s perspective, they fall outside
of the scope of the IMF’s international lender of last resort actions. By comparison, (Mody and Saravia
2013) only look at SBAs.
Author's personal copy
D. McDowell
50
40
count
30
20
10
0
0
50
100
150
200
Loan Approval Period (days)
Fig. 2 IMF loan approval periods, 1984-2012
date on the Letters enabled me to adopt a consistent and clear method for establishing
when a loan request was formally submitted to the IMF for approval.9
Figure 2 presents variation in the duration of loan approval periods, defined as
the number of days between a formal request via a Letter and loan approval by the
Board. As the figure reports, there is a considerable amount of variation across loan
requests. The median lag time between request and approval for the sample of 275
loan requests is 30 days. 139 cases (slightly more than half) fall at or above the
median approval period. The award for longest wait in the sample goes to a 1987
SBA request from Argentina at 192 days from request to approval. Cote d’Ivoire’s
SBA request that same year tallies the second longest wait at 165 days. The award
for shortest wait goes to Peru’s 1993 SBA which was approved the same day it was
filed. South Korea’s 1997 SBA which was approved the day after the Letter was filed
is a close second. 16 loan requests in my sample were approved in fewer than ten
days. All of these rapid approvals occur since 1991.
The last point speaks to a notable and important pattern in the data: Over time,
the IMF appears to have become a more responsive lender. In Fig. 3, I plot approval
9 SBA dates were collected by the author at the IMF archives in Washington D.C.; EFF dates were collected
by a research assistant via the Fund’s digital archive.
Author's personal copy
Need for speed: The lending responsiveness of the IMF
periods by year. The longest waits are clustered in the first several years of the data.
The secular trend toward improved responsiveness over time is obvious and quite
substantial. For instance, the median lag time between request and approval from
1984 to 1987–the first four years in the sample–is 52 days. In the last four years of the
sample, the average loan approval period is a mere 12 days. Thus, over time the Board
has accelerated the pace of loan approvals. In the following section, I develop my
own argument about what drives this variation in loan approval periods and propose
several testable hypotheses.
3 Explaining variation in loan approval periods
Once a Letter is filed, what ultimately determines the duration of the loan approval
period is the date on which a loan request is scheduled for consideration–and
approved–by the Board. In general, the Fund follows a standard set of scheduling
protocol. In order to make an informed vote, Directors are afforded a baseline period
of time–known as the “circulation period”–to review the borrower’s request and program details. A minimum circulation period of four weeks was established in 1982
and was revised down to two weeks in 1996 ([EBD] Executive Board Document
1985, 1996). Thus, depending on the year a request was made, the baseline approval
Loan Approval Period (days)
200
150
100
50
0
1990
2000
Year
Fig. 3 IMF loan approval periods by year, 1984-2012
2010
Author's personal copy
D. McDowell
period is two or four weeks. Of course, borrowers are not guaranteed that their request
will be considered immediately following the expiration of this period. Indeed, many
requests languish long after this has passed. Others are approved on or shortly after
the conclusion of the circulation period. Outside of this standard scheduling protocol,
borrowers that have formally submitted their loan request can ask the Board to waive
the minimum circulation period to ensure speedier approval. So long as no member
of the Board objects to that request, a vote may be scheduled prior to the expiration
of the minimum circulation period. However, even if the Board agrees to the waiver
request, the precise date on which a vote is scheduled remains under the control of the
Managing Director or, sometimes, the First Deputy Managing Director. Thus, IMF
responsiveness is always (1) a function of the priority given to a request by the Managing (or First Deputy Managing) Director and (2) in some cases also a function of
a borrower’s willingness to ask for a favor and the Board’s unanimous acquiescence
to that request.
Based on the bureaucratic process through which requests are scheduled for a vote
and the way power is distributed within the institution, I anticipate that the interests
of major Fund shareholders influence IMF lending responsiveness. The position of
Managing Director, the top spot at the institution, has always been held by a European. The Managing Director is the primary actor in charge of scheduling requests
for a vote and, thus, has the most direct influence on the duration of loan approval
periods. During the period under investigation here, five out of six (accounting for
25 of 29 years) have hailed from G-5 countries: four from France and one from
Germany.10 The position of First Deputy Managing Director, the number two spot
at the Fund, has always been held by an American.11 Additionally, G-5 countries are
the only member states that have permanent Executive Director (henceforth, Director) seats on the Board. This guarantees them regular access to the Managing Director
and her top deputy. I assume these influential members can use this access to press
Fund leadership to take actions that will affect when the request is brought before
the Board for a vote. It also directly involves these countries in any decision to grant
a waiver of the circulation period, expediting the period in which that body can consider a request. Thus, G-5 countries should hold disproportional influence over when
the Board considers a loan request. To put it differently, they should have the capacity
to intervene in ways that speed up (or slow down) the approval of governments’ loan
requests.
I contend that the decision to intervene is directly related to the exposure of G5 commercial banks to the borrower country’s economy. Foreign financial troubles
can threaten the stability of creditor country financial systems if they are sufficiently
exposed to the afflicted economy. When G-5 financial interests are threatened, these
powerful members should push the Fund to respond in a way that most effectively
10 The only Managing Director during this period not from a G-5 country was Spain’s Rodrigo Rato who
held the position from June 2004 to November 2007.
11 It should be noted that this position was created in 1994.
Author's personal copy
Need for speed: The lending responsiveness of the IMF
catalyzes private financial flows on behalf of the borrower. All else equal, this should
more swiftly resolve the crisis and reduce the threat it poses. Yet, I also argue that
the IMF’s catalytic lending strategies affect the Board’s responsiveness. During the
1980s, the Fund relied on the concerted lending approach to forcibly catalyze new
loans from banks on behalf of select borrowers. However, this approach required
additional negotiations that lengthened the approval period. Around the turn of the
next decade, the Fund altered its approach to catalytic financing. As developing
countries became increasingly dependent on disaggregated, heterogeneous global
financiers, the Fund worked to catalyze private capital flows by rapidly approving large loans. Below I further develop my argument and propose four testable
hypotheses.
3.1 The debt crisis, concerted lending, and IMF responsiveness
The Latin American debt crisis is widely recognized as the defining moment in the
IMF’s emergence as de facto international lender of last resort. The experience necessitated a dramatic revision of the Fund’s lending strategy (Boughton 2000; Sachs
1995). Prior to 1982, when a country approached the IMF for a loan, it was standard practice for the Fund to first determine how much financing the borrower could
expect to acquire from private as well as other official creditors before calculating the
amount of financing the borrower needed from the institution. As Boughton (2001)
explains, “That strategy collapsed, at least for the most heavily indebted countries,
with the Mexican crisis of August 1982” (406). Boughton goes on to explain how,
after the Mexican default, commercial banks actively sought ways to reduce their
exposure to the most heavily indebted countries. Therefore, as the Fund was increasing its lending to the economies in crisis, commercial banks were planning to do just
the opposite.
This was a problem because the Fund did not have sufficient lendable resources
to keep debtor countries solvent on its own. Without continued financing from the
banks, the probability of disorderly default would increase significantly. Were this
allowed to happen, one or more highly exposed, systemically important Western
commercial banks may have been forced into bankruptcy (Sachs 1988, 253). The fundamental problem was the collective action problem facing the commercial banks.
Despite the fact that it was in the banks’ collective interest to continue to roll over
the principal on existing loans, individually banks faced incentives to “cut and run.”
As Boughton notes, again discussing the Mexican crisis,
The riskiness of loans to Mexico had become great enough that each individual bank had an interest in demanding repayment now, as long as it could do
so ahead of any general stampede. The conflict between the individual and collective interest was by this time a quietly but rapidly burning wick (Boughton
2001, 289).
In response to this unfolding dynamic, the IMF altered its traditional approach and
adopted a new strategy which became known as concerted lending. This approach
relied on issuing an ultimatum: The Fund would not approve loan requests until the
Author's personal copy
D. McDowell
group of commercial banks (referred to as a syndicate) agreed to increase their exposures to the afflicted economies.12 The banks wanted the IMF in the lending game
since the institution could provide additional credit to keep debtors current on their
debt payments. Moreover, the Fund was the one actor that could impose economic
reforms on the heavily indebted countries. The Fund used this leverage to help the
banks overcome their collective action problem and secured additional financing for
countries in crisis.
In practice, concerted lending worked as follows. First, IMF staff and government
officials would negotiate program details. Once agreement was reached, a Letter
would be formally submitted to the IMF, as usual. Then, rather than scheduling the
formal request for a vote, the Fund would take the program as proposed in the Letter
to the banks. The syndicate was informed that a vote would not be scheduled until it
agreed to provide additional loans to the borrower in question. Only when a deal with
the banks was reached would the Managing Director put the program to a vote.13
Of course, rounding up new lending commitments from banks takes time. The process behind Argentina’s 1984 SBA is illustrative of this. On September 25, Argentina
signed and filed a Letter with the Fund requesting $1.2 billion in assistance. Despite
the large size of the request, the Fund calculated that Argentina needed an additional
$8 billion in financing to repay arrears to banks, official creditors, and to replenish its
dwindling foreign exchange reserves. So, Managing Director Jacques de Larosiere,
set up a series of bilateral meetings with official creditor countries and bank syndicates in order to round up additional money. These negotiations took months to
complete. Once all parties had signed on, the Board approved the loan on December
28–94 days after Argentina first filed its request for assistance (Boughton 2001, 393394). In effect, in exchange for larger overall financing packages, the Fund sacrificed
speed.
Given the delays associated with the strategy, concerns among some Board members that it undermined the Fund’s impartial position in negotiations, and fears that
its overuse would render it ineffective, concerted lending was not used on behalf of
every borrower during the debt crisis (EBM 1983a, 13, 17, 26-28, 37; EBM 1983b,
29). The Fund’s key shareholders had reason to support an approach that would limit
the use of concerted lending to preserve its effectiveness and protect Fund impartiality between debtors and creditors except when it came to borrowers where their own
financial systems were significantly at risk. For instance, in one Board meeting where
the strategy was discussed at length, Directors from the U.S., Germany, and the U.K.
each suggested that concerted lending should only be used in “exceptional cases.”
Specifically, situations where the stability of the international financial system was
threatened (EBM 1983a, 21; EBM 1983b, 22-23; Erb 1983). A disorderly default on
12 Boughton
(2001) describes the moment this became the Fund’s new approach: “The turning point came
at the November 1982 meeting in New York...at which the Managing Director informed the banks that
the Fund would not approve Mexico’s requests for an extended arrangement until the banks provided him
with written assurances that they would increase their exposure by enough to cover a substantial fraction
($5 billion) of Mexico’s scheduled interest payments for 1983” (406).
13 For a detailed description of the concerted lending approach, see Boughton (2001), 405-409.
Author's personal copy
Need for speed: The lending responsiveness of the IMF
the part of such systemically important borrowers could have had catastrophic consequences for financial stability within the G-5 domestic economies. In such cases,
G-5 governments should have been more inclined to press the Managing Director to
implement the catalytic, ultimatum technique. Doing so would provide them ex ante
assurances that the banks would overcome their collective action problem and the
borrower in question would receive sufficient financing and remain solvent. Because
concerted lending had the effect of delaying the approval of loan requests due to the
additional negotiations with banks, I expect that,
H1: All else equal, an increase in G-5 bank exposure will be associated with longer
loan approval periods during the concerted lending years.
A linear relationship between G-5 bank exposure and the duration of loan approval
periods may be overly simplistic, however. It is likely that along with the vulnerability
of their banks, major shareholders also weighed the imminence of default when considering the use of concerted lending. For example, if G-5 banks were highly exposed
to a borrower country that was not in danger of defaulting, then implementing the
concerted lending strategy would be unnecessary. Thus, I expect that,
H2: All else equal, an increase in G-5 bank exposure will be associated with longer
loan approval periods during the concerted lending years. This effect will strengthen
as the borrower’s debt burden increases.
3.2 Financial liberalization and IMF lending responsiveness
Concerted lending was the Fund’s primary strategy for managing the Latin American
debt crisis until 1987.14 By that year, the IMF recognized that the ultimatum approach
was no longer as effective at catalyzing bank lending and so it abandoned the practice
(Boughton 2001, 418). Over the decades that followed, IMF lending responsiveness
improved considerably and consistently. The Fund’s improving responsiveness coincided with another important trend that occurred over the same period of time: the
globalization of financial liberalization. As developing countries began to remove
barriers to international capital flows in the 1990s, the volume and complexity
of those flows grew quickly. Portfolio capital flows–investments in debt securities
(bonds) and equities (stocks)–from industrial to developing countries surged (Prasad
et al. 2003).15 Moreover, these investments were driven by institutional and retail
investors alike. Copelovitch (2010, 30-35) explains in great detail how, beginning in
the 1990s, there was a significant shift in the composition of international lenders.
What was once a global financial system dominated by a relatively small number of
Western banks had become far more complicated. A growing number of countries
found their economies indebted to a “disaggregated, heterogeneous group of private
international lenders” (Copelovitch 2010, 9). As Copelovitch argues, private creditors’ collective action problems should become more severe as a greater share of an
14 Bird and Rowlands (2004) date the strategy from 1982 through 1986 while Caskey (1989) notes that the
strategy was adopted during the Mexican debt adjustment program through 1987.
15 For example, between 1990 and 1994, roughly $670 billion of foreign capital flowed into countries in
Asia and Latin America as investors around the world began putting their money into emerging stock
markets and securities (Truman 1996, 201; Calvo, Leiderman and Reinhart 1996, 123).
Author's personal copy
D. McDowell
IMF borrower’s debt is owed to bondholders. Thus, greater dependence on bond debt
should increase the likelihood of capital flight during crises. Additionally, the 1990s
also saw a rapid increase in the amount of short-term debt accumulated by emerging
markets. Increased dependence on loans with maturities under one year have been
linked to the occurrence of more severe financial crises and capital flight (Rodrik
and Velasco 1999, 3). Thus, as the 1990s began, the debt structure of many countries
shifted from long- and medium-term loans from banks to increased dependence on
short-term capital flows from a herd of global banks and bondholders.
The combination of financial openness and the complexity of capital flows had
significant implications for economic stability in these countries. Just as foreign capital can easily flow in when the capital account is open, it can just as easily flow out
as conditions become less appealing to investors. Moreover, improvements in technology now meant that these investments could be pulled out of a country “with
little more than the flick of a computer key” (Calvo, Leiderman and Reinhart 1996,
127). Increasingly, the IMF was called upon to help countries dealing with crises that
developed quickly in the capital account as a result of such short-term capital movements.16 Moreover, the IMF was precluded from implementing concerted lending in
most cases. Efforts to forcibly catalyze lending from commercial banks in the 1980s
was possible due, in part, to the relatively small number of institutions involved.
Such a strategy would not work when dealing with a disaggregated, heterogeneous
pool of investors hellbent on pulling their money out of an economy in duress.17 The
need for speed on the part of the IMF as international lender of last resort increased
considerably as a result of these changes. For instance, Guitian (1995, 817) noted
that “capital account problems typically require a rapidly agreed and relatively large
financial support package, in which a substantial share of the funds is made available up front” (emphasis added). Similarly, Boughton (2000, 275) explained “What
characterizes a twenty-first century crisis is simply the speed and magnitude of the
resulting flows...A financial crisis calls for a similar response from the Fund as any
other balance of payments problem except that the response must be quicker and
possibly much larger than in a more traditional case” (emphasis added).18
This was not lost on the IMF. The Board’s handling of Mexico’s 1995 SBA
request is illustrative of this new approach. Mexico’s request for assistance was
16 Calvo
(1998) referred to the crises that developed in the 1990s as capital account crises to distinguish
them from current account crises which were the dominant variety in previous decades.
17 One notable exception to this point is the case of Korea during the Asian Financial Crisis. In January,
1998, the IMF implemented a concerted lending-like approach to forcibly catalyze additional lending from
major banks. However, this approach was only used after the Fund’s initial, massive loan to Korea had
already been approved in December 1997. Indeed, Korea’s SBA was approved one day after the Letter was
filed making it the second shortest loan approval period in the sample. Thus, in the one known case where
concerted lending was employed after 1987, it did not have the effect of delaying Board approval. Rather,
it came following an incredibly swift vote. For more on the Korean crisis, see Boughton (2012, 557-565.)
18 This view of catalytic financing has been compared to as “Powell Doctrine” in military affairs: Once a
decision is made to intervene, it must be done with “overwhelming force” if it is to succeed (Zettelmeyer
2000; Jeanne and Wyplosz 2001). See Morris and Shin (2006) for an alternative view on the necessity
of large credits for catalytic financing. Empirically, Mody and Saravia (2006) find limited evidence that
larger loans from the IMF are more effective at catalyzing private financial flows on behalf of borrowers.
Author's personal copy
Need for speed: The lending responsiveness of the IMF
placed on the Board’s agenda and approved a mere 6 days after the country
filed its Letter. Thus, the standard four week circulation period was scrapped
in the interest of speed. Expediting Mexico’s request was in line with American preferences. The Clinton Administration believed the crisis required a “massive and fast” response (Greenspan 2007, 156-160). When the Board met to
discuss the SBA, the American Director explained that skirting the standard circulation period was necessary because “the size of the crisis is...unprecedented
and the potential for spillover–an extremely quick spillover–is enormous”
([EBM] Executive Board Minutes 1995, 53). The Japanese Director noted the
Mexican crisis was both “exceptional and urgent” (ibid, 61).19 In the aftermath of the
Mexican Peso crisis of 1995 the IMF introduced the Rapid Financing Instrument
(RFI). This was created to give the IMF a mechanism through which it could bypass
standard procedures and accelerate the approval of requests when the Board felt this
was necessary.20 While the IMF had established the RFI expecting that it would be
used in very limited fashion, “it turned out to be an essential tool” (Boughton 2012,
495).21 Then, in 1996 the Board implemented new guidelines that reduced the minimum circulation period for loan requests from four weeks down to two weeks ([EBD]
Executive Board Document 1996). In the years since, the Fund has continued to
enhance its ability to respond to crises with greater speed.22
Though the shifting global financial landscape led the Fund to respond more
quickly to all loan requests, the institution did not apply this sense of urgency evenly
across borrowers. Even with a reduced baseline loan approval period, the Fund continued to prioritize some requests over others. Some requests, like Mexico’s 1995
SBA, Korea’s 1997 SBA, and Greece’s SBAs in 2010 and 2012, were expedited
once the Letters were filed. Most, however, were not. Once again, I expect that
G-5 financial interests should impact IMF lending responsiveness, though in different ways than they did during the early-1980s. In the more recent era, I anticipate
that the G-5 will intervene to shorten the baseline loan approval period when their
commercial banks are highly exposed–the opposite of my expectations during the
earlier period. In such cases, they should have been inclined to press the Managing
19 For
a excellent account of the Board’s debate surrounding the Mexican request, see Copelovitch (2010,
213-227).
20 The RFI is sometimes referred to as the Emergency Financing Mechanism. It is not a separate lending
arrangement per se rather it is a mechanism which can be activated when a member country is “facing
an urgent balance of payments need” yet does not need a “full-fledged program” (International Monetary
Fund 2012d).
21 The RFI has been used on behalf of Korea and Indonesia in 1997, Turkey in 2001, Armenia, Georgia,
Hungary, Iceland, Latvia, Pakistan, and Ukraine from 2008-2009, and Greece (twice), Ireland, Portugal,
and Cyprus from 2010-2013.
22 After the Asian financial crisis in 1997-1998, the Fund debuted the Contingent Credit Line (CCL) which
was designed to prevent contagion by providing precautionary lines of credit to countries at risk (Bird
and Rajan 2002). More recently, in the aftermath of the Global Financial Crisis of 2008, the IMF has
introduced three new facilities designed to provide speedier financing to member states facing balance of
payments crises: the Flexible Credit Line (FCL), the Precautionary and Liquidity Line (PLL), and the and
Rapid Credit Facility (RCL) (IMF 2012a; 2012b). Even the SBA was recently overhauled in order to make
the workhorse arrangement more effective for members who may not qualify for an FCL arrangement “by
providing increased flexibility to front-load access” intended to improve its “crisis prevention and crisis
resolution” performance (IMF 2009, 3).
Author's personal copy
D. McDowell
Director to expedite the placement of the request on the Board’s agenda, even if this
resulted in a reduced circulation period. The G-5’s desire for speed in such cases
relates to the IMF’s changing approach to catalytic financing. The growing speed and
complexity of financial markets now required large loans to be disbursed quickly in
order to reinstate market confidence, prevent dangerous spillovers, and attract back
private capital. Foreign financial crises put the profitability and, in extreme cases, the
solvency of G-5 commercial banks in jeopardy when they are significantly exposed
to a foreign economy in crisis. Because of these risks, G-5 countries have a clear
interest in stabilizing, as quickly as possible, economies where their banks are most
vulnerable. Thus,
H3: All else equal, an increase in G-5 bank exposure will be associated with
shorter loan approval periods during the post-concerted lending years.
A linear relationship between bank exposure and reduced loan approval periods
may be overly simplistic, however. Alongside the vulnerability of their banks, major
Fund shareholders likely weigh other factors when determining which requests the
Board should prioritize. In particular, the hastening effect of bank exposure on Board
approval should increase as a borrower becomes increasingly exposed to the vagaries
of international financial markets via dependence on bond markets and short-term
debt. In such cases, the prospects for capital fight should be higher and, thus, a large
and speedy IMF loan becomes necessary in order to catalyze financial flows. Thus,
H4: All else equal, an increase in G-5 bank exposure will be associated with
shorter loan approval periods during the post-concerted lending years. This effect
will strengthen as the borrower’s dependence on bond markets and short-term debt
increases.
4 Analyzing variation in IMF lending responsiveness
To test my argument, I compile a data set consisting of 275 SBA and EFF loan
requests from 1984 through 2012. As I have already discussed the dependent variable
in detail above, I turn to the remainder of my research design below. Here, I review
the explanatory and control variables employed as well as my model specifications.
4.1 Independent variables
To account for G-5 financial interests, I construct the covariate Bank Exposure.
International banking data were gathered from the Bank for International Settlement’s (BIS) Consolidated Banking Statistics database.23 Specifically, I use the
stated foreign claims by nationality of reporting bank, immediate borrower basis.24
I measure G-5 bank exposure as the sum total of G-5 banks’ consolidated foreign
23 The BIS consolidated bank claims data is the highest level aggregate data type that includes both private
and public (sovereign) foreign debts.
24 Data available at http://www.bis.org/statistics/consstats.htm
Author's personal copy
Need for speed: The lending responsiveness of the IMF
claims to country i in year t-1 divided by the collective GDP of the G-5 countries
in year t-1. To account for the time period in which a loan request was made, I create a dummy variable Concerted Lending equal to 1 if a request was made
from 1984 through 1987, 0 otherwise. Using my full sample, I interact Concerted
Lending with Bank Exposure to test for the effect of G-5 financial interests on
IMF responsiveness contingent on the time period in which a request was made.
To account for the intensity of a borrower’s debt burden during the concerted lending years, I rely on the covariate Debt Service. This accounts for a borrower
country’s public and private sector debt service expenses (principle repayments and
interest paid) expressed as a percentage of its exports in year t-1. As this increases,
so too does the country’s debt burden and, ostensibly, the imminence of default. Debt
service data were gathered from the World Bank’s Word Development Indicators
(WDI) database. Using a sub-sample of requests made while the concerted lending strategy was actively used (1984-1987) I interact Debt Service with Bank
Exposure to assess whether the anticipated delaying effect of G-5 financial interests on IMF responsiveness during these years intensifies as the imminence of default
increases.
To capture the dependence of a borrower on bond markets, I construct the measure
Bond Debt consistent with Copelovitch (2010, 76). The variable accounts for the
share of public and private debt held by bondholders relative to banks.25 To account
for a borrower’s dependence on short-term capital flows, I employ Short-term
debt. This measures a country’s public and private debts having an original maturity of less than one year as a percentage of its total debt. As each of these covariates
increase, so too should the vulnerability of the borrower country to rapid capital
account reversals. Data for both measures were gathered from WDI and relevant IMF
documents. Using a sub-sample of requests made in the post-concerted lending era
(1988-2012) I separately interact these covariates with Bank Exposure to assess
whether the anticipated accelerating effect of G-5 financial interests on IMF responsiveness during these years intensifies as the exposure to the vagaries of international
financial markets increases.
4.2 Control variables
It is possible that the geopolitical interests of the G-5 countries also influence IMF
responsiveness. For example, it may be the case that the G-5 countries use their
influence with the Fund to speed up loan approval periods on behalf of allies or
strategically important borrowers. To account for a borrower’s status as an important “client” of the major Fund shareholders, I use WDI data to construct Foreign
Aid: the sum total of G-5 net official development assistance to (ODA) to country i in year t-1 divided by the collective GDP of the G-5 countries in year t-1.
Consistent with Stone (2004), I assume that the distribution of aid across recipient
25 This is calculated as follows: publicly guaranteed bond debt + private non-guaranteed bond debt/total
public and publicly guaraneed debt owed to private creditors + total private non-guaranteed debt. All data
are from WDI.
Author's personal copy
D. McDowell
countries is reflective of the relative priority donors attach to them. That is, foreign
aid can be viewed as an “investment in a valued regime, representing a direct monetary measure of the importance of a particular recipient to a particular donor” (ibid,
579). To account for a borrower’s status as a geopolitical friend of the G-5 countries, I rely on Bailey et al.’s (2015) measure of ideal point distance within the United
Nations General Assembly (UNGA).26 Specifically, I construct UN Ideal Point
Distance which is the mean ideal point distance of the G-5 countries vis-a-vis
country i in year t-1. Lower scores indicate more similar voting profiles (and, hence,
voting preferences) in that body. Recent studies have found that temporary membership on the U.N. Security Council (UNSC) is associated with improved access to
IMF and World Bank resources as well as increased foreign aid flows from major
donors (Dreher et al. 2009a, b; Kuziemko and Werker 2006). Thus, I include UNSC
member which is equal to 1 if the borrower held a temporary seat on that body in
year t, 0 otherwise. Data are from Dreher (2009b).27
I control for two additional political variables. First, because variation in borrower
country political institutions may also influence IMF responsiveness, I account for
Regime Type based on Polity scores (Marshall et al. 2012). Theoretically, regime
type could influence loan approval periods if the Board systematically scrutinizes
requests from non-democracies more than those from democracies. Second, I also
control for each borrower’s IMF Quota Share in year t. This reflects each country’s IMF “subscription” and can be thought of as a proxy for a member’s influence
on the Fund’s Board (Barro and Lee 2005). Countries with larger quotas may be able
to parlay that influence into speedier loan approvals. Data were acquired from the
IMF’s International Financial Statistics (IFS) database.
The business and procedures of the Fund may affect the institution’s responsiveness. In some cases, countries file their Letter while they are still under a Fund
supported program. To control for this, I include a dummy variable Under IMF
Program equal to 1 if a country was under a program at the time a Letter was
filed, 0 if otherwise. In such cases, I expect approval periods will be longer as the
Board should be more likely to delay approval of the new request until after the old
program has expired. Variation in the Board’s workload could influence its responsiveness. Thus, I control for the total number of Requests Per Month since
the Board may be unable to respond as quickly when it has more business to conduct.28 Because the IMF reduced the minimum circulation period from four to two
26 Bailey et al. (2015, 1) estimate dynamic national ideal points using voting data in the UNGA from 1946-
2012. They show that ideal point estimates improve upon conventional dyadic similarity indicators such
as Affinity or S-scores (Gartzke 1998; Signorino and Ritter 1999) by distinguishing UN agenda changes
from changes in state preferences.
27 Data available at http://www.uni-heidelberg.de/fakultaeten/wiso/awi/professuren/intwipol/datasets.html
28 There is no systematic variation in the Board’s meeting schedule that should affect IMF responsiveness.
Since its creation, the Board has functioned in “continuous session at the principal office of the Fund”
and meets “as often as the business of the Fund may require (Articles of Agreement, Article XII, Section
3(g)). Van Houtven (2002, 14-15) explained that, at the time of his writing, “total Board meeting time
averages more than 12 hours a week and over 600 hours per year, which demonstrates the intense oversight
exercised by the Board on activities of the IMF. Nearly one-third of the Board meeting time is devoted to
policy issues, about 60 percent to surveillance, and the remainder to administrative and budgetary matters.”
Author's personal copy
Need for speed: The lending responsiveness of the IMF
weeks in 1996, I include a dummy variable Circulation Period. This takes on
a value of 1 for years when the reduced period was in effect as this should increase
Fund responsiveness, all else equal. All of these variables were constructed by the
author.
The Fund might respond differently to requests from countries experiencing a
severe financial crisis. Moreover, the IMF may respond differently depending on
the crisis variety that led a government to seek assistance. To account for the presence of pressure on a country’s currency, I include Speculative Attack. Using
monthly IMF exchange rate and foreign exchange reserves data, I code this measure
as 1 if the EMP index score is more than two standard deviations above the country
mean, 0 otherwise. I then annualized the data; the variable is equal to 1 if a speculative attack is present in any month of year t, 0 otherwise. This approach is consistent
with Leblang (2003).29 Similarly, I include Debt Crisis which is equal to 1 if a
country defaulted on or rescheduled its external debts in year t, 0 otherwise. Debt crisis data through 2007 are drawn from Laeven and Valencia (2008); 2008-2012 data
from FOC Project (2015).
I also control for GDP and Population. I rely on WDI data for both measures. In
line with Copelovitch (2010) I include regional dummy variables.30 Given the secular
improvement in IMF lending responsiveness over time, I include year dummies in all
models.
In keeping with previous studies, a number of explanatory and control variables
are lagged one year, reflecting the fact that most IMF programs are designed and
approved based on information and data that lags behind the date of approval (Knight
and Santaella 1997, 413). Variables with heavily skewed distributions are logged.31
Transformed variables are identified in the tables below.
29 The
variable Speculative Attack was created by the author based on the EMP index. It is
calculated as follows:
EMPi,t =
ei,t
ri,t
−
σei
σri
e is the end-of-the-month U.S. dollar exchange rate of country i and r is the end-of-the-month non-gold
reserves. A higher score on the index indicates increased pressure on a country’s currency. I identify a
speculative attack as:
SpeculativeAttacki,t
= 1 if EMPi,t > 2σEMPi + μEMP1
= 0 otherwise
σEMP and μEMP are the country-specific mean and standard deviation of EMP, respectively.
30 Regional variables were created in Stata using the “Kountry” command based on the “Middle
East
broad” classification. Additionally, countries listed as “Oceana” were recoded as Asia resulting in five
regional dummies: Africa, Americas, Asia, Middle East/North Africa (MENA) and Europe. In my
statistical models, Europe is the baseline category to which other regions are compared.
31 In each of these cases, I add 1 to the raw data prior to transformation in order to prevent zero values
from becoming missing observations after transformation as the log of zero is undefined.
Author's personal copy
D. McDowell
4.3 Model specifications
I fit four Cox proportional hazards models of IMF loan approval duration periods to
estimate the effect of G-5 financial interests on IMF lending responsiveness.32 Model
1 includes the full sample of cases. Model 2 only includes requests made during the
concerted lending years while Models 3 and 4 only include requests made in the
post-concerted lending era. Models are fitted using the R package Zelig (Imai et
al. 2007, 2008). A request becomes “at risk” for approval when a country submits a
Letter of Intent to the Board and subsequently “fails” when the request is approved.
I model the expected duration in days between request and approval as a function of
a baseline hazard rate and list of covariates that change this baseline rate.33 Since all
loan requests in the sample were approved, there are no censored observations.34 I
compute robust standard errors clustered by country since residuals may be correlated
across observations. I also test that the covariates in each model do not violate the
proportional hazard assumption.35 Finally, I exclude the regional dummy variables
from Model 2 (concerted lending era) as their inclusion causes a number of other
covariates to violate the proportional hazards assumption.36
5 Results
Hazard estimates for models 1 through 4 are displayed in Table 1. The estimates are
presented in the log hazard (non-exponentiated) metric. Coefficient estimates greater
than 0 indicate the covariate is associated with a higher hazard rate, hence, swifter
approval. Conversely, coefficient estimates less than 0 indicate the covariate is associated with a lower hazard rate, hence, a longer wait for approval. In other words,
factors correlated with faster approval times display positive coefficients while those
correlated with slower approval times are negative. To save space, year dummy
coefficients are not reported.
I find that the G-5 financial interests do in fact influence IMF responsiveness contingent on the time period in which a loan request was made. Model 1 presents the
results for the full sample of requests. The interaction between Bank Exposure
and Concerted Lending is negative and highly significant. This result is consistent with Hypotheses 1 and 3, indicating that during the concerted (post-concerted)
32 For
handling ties, I use the Efron method.
advantage of the Cox model is that it does not make parametric assumptions about the baseline
hazard rate. This contrasts with parametric models, including Poisson and Negative Binomial models,
which assume the hazard rate takes on a specific shape. Box-Steffensmeier and Zorn (2001, 21, 46) explain
that unless “there exists strong theoretical expectation regarding the ‘shape’ of the hazard rate (or by
extension, survival times), conditional on covariates in the model” then non-parametric models, like Cox,
are preferable to parametric models.
34 In my data collection efforts at the IMF archives, there were a total of three Letters of Intent filed
for SBAs that I was unable to locate a corresponding approval date. In chronological order, these are:
Malawi (8/11/1986), Brazil (9/13/1990), and Paraguay (1/9/1991). I was unable to determine whether these
requests were rejected by the Board or withdrawn by the borrower countries. Because of this uncertainty
and their rarity, I opted to exclude these cases from the analysis.
35 Specifically, I calculate Shoenfeld residuals as described in Box-Steffensmeier and Zorn (2001).
36 Nonetheless, the results are substantively unchanged when these are included.
33 An
Author's personal copy
Need for speed: The lending responsiveness of the IMF
Table 1 Cox Proportional hazards estimates: Robust standard errors clustered by country in parentheses;
* p < 0.10, ** p < 0.05, *** p < 0.01
Model 1
Model 2
Model 3
−0.398∗∗
0.977∗∗∗
−0.421∗
−0.406
(0.177)
(0.432)
(0.249)
(0.253)
Population t-1 (log)
0.234∗
−0.261
0.260
0.266∗
(0.129)
(0.378)
(0.179)
(0.173)
Under IMF program
−0.077
−0.245
0.083
−0.047
(0.175)
(0.360)
(0.234)
(0.236)
Requests per month
0.014
0.073
−0.126∗
−0.087
(0.045)
(0.074)
(0.068)
(0.068)
Speculative attack
0.206
0.416
0.092
0.106
(0.174)
(0.381)
(0.224)
(0.225)
GDP t-1 (log)
Model 4
−0.179
−0.229
−0.211
−0.039
(0.303)
(0.909)
(0.379)
(0.367)
0.030
−0.172
0.054
−0.039
(0.184)
(0.451)
(0.272)
(0.259)
Regime type t-1
−0.001
0.044
0.005
−0.003
(0.017)
(0.035)
(0.023)
(0.023)
UNSC member
−0.003
0.289
−0.244
−0.133
(0.252)
(0.541)
(0.340)
(0.326)
−0.413∗∗
−0.924
−0.403∗
−0.324
(0.200)
(0.720)
(0.238)
(0.238)
−0.032
0.326∗
−0.029
−0.018
(0.024)
(0.212)
(0.032)
(0.029)
Debt crisis
IMF quota share (log)
UN Ideal point distance t-1
Foreign aid t-1 (log)
Africa
−0.463
−0.925∗∗
−1.298∗∗∗
(0.407)
(0.525)
(0.528)
Asia
0.848∗∗∗
0.366
−0.127
(0.358)
(0.427)
(0.430)
0.351
0.096
−0.271
(0.290)
(0.382)
(0.762)
MENA
−0.266
−0.248
−0.647
(0.429)
(0.511)
(0.518)
Circulation period
4.318∗∗∗
0.541
4.011∗∗∗
(1.275)
(1.285)
(1.332)
Concerted lending
−6.590∗∗∗
Americas
(1.810)
Debt service t-1 (log)
−16.166∗∗∗
(4.329)
Bond debt t-1 (log)
−7.525∗
(5.857)
Short-term debt t-1 (log)
2.942∗∗∗
(1.129)
Author's personal copy
D. McDowell
Table 1
(continued)
Model 1
Model 2
Model 3
Model 4
Bank exposure t-1 (log)
0.138∗
1.531∗∗∗
0.092
−0.219∗∗
(0.078)
(0.518)
(0.100)
(0.136)
Concerted lending*bank exposure
−0.226∗∗∗
(0.075)
−0.625∗∗∗
Debt service*bank exposure
(0.183)
−0.314
Bond debt*bank exposure
(0.253)
0.132∗∗∗
Short-term debt*bank exposure
(0.046)
Observations
275
63
175
Number of countries
87
39
59
180
63
Years
1984-2012
1984-1987
1988-2012
1988-2012
Logrank test
360.9∗∗∗
36.91∗∗∗
180.7∗∗
227.2∗∗∗
lending era, increased G-5 bank exposure is associated with longer (shorter) waits
for loan approval. To illuminate the magnitude of bank exposure’s effect on loan
approval periods, I graph survival curve and first difference estimates based on
Model 1.
Figure 4 presents these estimates as Bank Exposure changes from low (1
standard deviation below its mean) to high (1 standard deviation above its mean) setting Concerted Lending equal to 1 and then 0. All other covariates are held
at their means.37 The first difference curve in Fig. 4 displays the substantive effect
of increased G-5 bank exposure on loan approval during the concerted and postconcerted lending eras. From 1984 through 1987, I estimate that after 30 days, a
request from a country where G-5 bank exposure is high is associated with a 8.4 percent increase in the probability of non-approval (survival) relative to a country where
exposure is low. This difference increases to a maximum of 14.9 percent at 45 days.
In other words, during the concerted lending years, requests from countries where
G-5 banks were more heavily exposed were about 15 percent more likely to still be
waiting on loan approval after one and a half months compared to countries where
G-5 banks were less exposed. This result is consistent with the expectation that the
G-5 countries use their role in Fund leadership, their informal influence with that
leadership, and their power within the Executive Board to implement the concerted
lending strategy as a method to catalyze private financing when their commercial
banks were highly exposed to the borrower in question. In doing so, they could
37 Based on Model 1, I generated simulations using the R package Zelig to replicate estimates of survival
rates at levels specified for covariates of interest while holding all other covariates at their means. Point
estimates displayed are the median observation from the simulation at each day after the loan request based
on 1000 sample replications. The same process was used to generate all other post-estimation figures.
Author's personal copy
Need for speed: The lending responsiveness of the IMF
Fig. 4 Survival and first difference estimates as G-5 bank exposure varies from low to high (Model 2)
acquire ex ante assurances that the banks had overcome their collective action problem and the borrower in question would remain solvent. While this served to protect
domestic financial stability in the G-5 countries, it also led to increased loan approval
periods for these borrowers.
As Fig. 4 shows, the direction of the effect of between bank exposure on approval
duration periods inverts in the post-concerted lending era. I estimate that after 30
days, a request from a country where G-5 bank exposure is high is associated with a
20.9 percent decrease in the probability of non-approval relative to a country where
exposure is low. In other words, in the post-concerted lending era, requests from borrower countries where G-5 banks were more heavily exposed were about 20 percent
more likely to have their request approved after one month when compared to countries where G-5 banks were exposed at lower levels. This result is consistent with
Hypothesis 3: Throughout the 1990s and 2000s the G-5 countries used their influence
in the Fund to prioritize requests from borrowers where their commercial banks were
significantly invested. In an era where financial flows originate from a disaggregated,
heterogeneous group of investors and capital can flow in and out of an economy at a
moment’s notice, large credits provided swiftly became the preferred method of catalyzing private financial flows. These findings suggest that when the financial stakes
are high for the G-5, the Board systematically responded with a greater sense of
urgency.
Of course, assuming a linear relationship between G-5 bank exposure and
the duration of loan approval periods–in either period–may be overly simplistic.
Hypothesis 2 predicts that during the concerted lending years, the delaying effect
of bank exposure should grow stronger as the imminence of default increases. In
these cases, major Fund shareholders should have been most likely to implement
concerted lending. The negative and statistically significant coefficient for the interaction term between Debt Service and Bank Exposure in Model 2 supports
the hypothesis. Figure 5 presents survival curve and first difference estimates as
Bank Exposure changes from low to high setting Debt Service at a standard
deviation below and then a standard deviation above its mean. All other covariates
are held at their means. When the debt burden is low, the delaying effect of increased
bank exposure on Board approval is quite substantial. However, when a borrower’s
Author's personal copy
D. McDowell
Fig. 5 Survival and first difference estimates as G-5 bank exposure varies from low to high (Model 3)
debt service costs are high–indicating default is increasingly imminent–the effect of
increased G-5 bank exposure is considerably stronger. Indeed, I estimate that 55 days
after a loan request was filed, high bank exposure is associated with a sizable 87 percent increase in the probability of non-approval (survival) when debt service costs
are high.
Hypothesis 4 expects that in the post-concerted lending years, the hastening effect
of bank exposure should be stronger as the borrower becomes increasingly exposed
to the vagaries of international financial markets. I rely on two measures of a nation’s
debt structure to account for this: the share of debt owed to bondholders and the share
of debt with short-term maturities. The results offer mixed support of this hypothesis. The interaction term between Bond Debt and Bank Exposure in Model
3 is not consistent with my expectations. The coefficient is not statistically significant and, surprisingly, is negative.38 On the other hand, the interaction term between
Short-term Debt and Bank Exposure is positive and highly significant.
This indicates that the accelerating effect of G-5 bank exposure on loan approvals
was intensified for borrowers more dependent on short-term capital flows. Figure 6
presents survival curve and first difference estimates as Bank Exposure changes
from low to high, setting Short-term Debt at a standard deviation below and
then a standard deviation above its mean. All other covariates are held at their means.
When the borrower is not dependent on short-term debt, bank exposure has virtually
no impact on the priority the Board gives to a request. However, when the country is
exposed to short-term capital–indicating increased borrower vulnerability and a need
to catalyze private capital flows–borrowers where G-5 banks are highly invested are
far more likely to have their requests expedited by the Board. In such cases, after 28
days, I estimate that requests are 32 percent more likely to have been approved by
the Board when compared to cases where short-term debt exposure is high but G-5
financial systems are far less at risk.
38 The inclusion of Bond Debt leads to the omission of more than 30 observations which may bias
these results against Hypothesis 4. Among these missing observations are most IMF credits to European
countries since 2008. This includes credits to Greece (2010, 2012), Iceland (2008), Ireland (2010), Latvia
(2008), and Portugal (2011). In each of these cases, G-5 bank exposure is above the mean and approval
periods are quite swift.
Author's personal copy
Need for speed: The lending responsiveness of the IMF
Fig. 6 Survival and First Difference Estimates as G-5 Bank Exposure Varies from Low to High (Model 4)
Turning now to the controls, the coefficient for UN Ideal Point Distance
is negative in all models and statistically significant in two of the four. This indicates,
albeit with mixed support, that as voting preferences between a borrower and the
G-5 within the UNGA grow apart, loan approval periods increase. Inversely, as voting profiles grow more similar, requests are approved more swiftly. Thus, geopolitical
allies of the G-5 countries may have their requests prioritized by the Board. However,
this result does not hold during the concerted lending years when, perhaps unsurprisingly, financial factors appear to be the only variables affecting Fund responsiveness.
Additionally, the results do not hold in Model 4. Borrowers with temporary membership on the UNSC do not appear to benefit from speedier approvals. During the
concerted lending years, important “client states” as measured by foreign aid flows
appear to have had their loans approved more swiftly. However, this result does not
hold in other models. A borrower’s influence within the Fund as measured by its
quota share does not seem to influence the speed of the Board’s response. Unlike
(Mody and Saravia 2013), Regime Type does not influence IMF responsiveness
as measured here. Moreover, the Board does accelerate its response to requests from
borrowers facing a speculative attack or those that default on debt obligations.
6 Robustness checks
I estimate negative binomial models and substitute (Mody and Saravia 2013) measure
of IMF responsiveness to check to robustness of my findings.
6.1 Alternative estimating Approach
While my dependent variable measures time, it may also be conceived of as “count”
data since the outcome I am measuring takes on integer values greater than zero.
Therefore, as a robustness check, I fit four negative binomial models.39 As before,
39 This is a parametric model, in contrast to the Cox model. Thus, it assumes a particular “shape” of the
hazard rate. However, as Box-Steffensmeier and Zorn (2001, 21) explain, “if the distribution of failure
times is parameterized incorrectly” then the findings may be inaccurate.
Author's personal copy
D. McDowell
models are fitted using the R package Zelig. Table 2 reports these results. Interpretation of the coefficients is the opposite of the Cox models presented above: negative
coefficients indicate swifter approvals (fewer days) while positive coefficients indicate longer waits (more days). Also, regional dummy variables are now included in
the concerted lending model (Model 7) since the proportional hazards assumption no
longer applies.40
Looking first to Model 5, the results again support Hypotheses 1 and 3: during the
concerted lending years, increased G-5 bank exposure is associated with a longer wait
for Board approval while the inverse is true for the post-concerted lending years.41
Indeed, the mean estimated loan approval period where G-5 bank exposure is high
is 62 days compared to 48 days where banks are less exposed for a difference of 14
additional days. In the post-concerted lending years, I estimate a wait time of 26 days
for borrowers where G-5 bank exposure is high compared to a wait of 32 days when
major shareholder banks are less at risk–a difference of 6 fewer days. Model 6 only
includes years when concerted lending was the Fund’s primary catalytic financing
strategy. Here the results still support Hypothesis 2: bank exposure slows the IMF
approval process down the most when the imminence of default is high. For example,
when debt service costs are low (one standard deviation below its mean), increasing
G-5 bank exposure from low to high adds an estimated 18 days to a borrower’s wait
time. However, when debt service costs are high (one standard deviation above its
mean) increasing bank exposure along the same lines adds an estimated 91 days
between request and approval. Models 7 and 8 only include requests from the postconcerted lending era. Again, support for Hypothesis 4 is mixed. As the Model 7
results indicate, a borrower’s exposure to bond markets does not appear to interact
with bank exposure to hasten Board approval. However, Model 8 results again show
in the post-concerted lending era, increased exposure of G-5 banks has its strongest
hastening effect when borrowers are exposed to short term capital flows as measured
by Short-term Debt. For example, when a borrower is less dependent on shortterm debt (one standard deviation below its mean), increasing bank exposure from
low to high reduces the borrower’s expected wait time just about a day. However,
when a borrower is highly dependent on short-term debt (one standard deviation
above its mean), increasing bank exposure along the same lines reduces the estimated
wait time by 12 days.
6.2 Alternative measure of the dependent variable
Above, I discuss a number of advantages my own measure of IMF responsiveness
has over the measure used by Mody and Saravia (2013). Succinctly, my measure is
40 The MENA (Middle East, North Africa) variable is absent because none of the borrowers in this subsample are from that region. The results are substantively unchanged if regional controls are removed.
41 I generated simulations using the R package Zelig to replicate estimates of counted days between
request and approval at levels specified for covariates of interest while holding all other covariates at their
means. Point estimates used are means from simulations based on 1000 sample replications. The same
process was used to generate all other post-estimation estimates discussed here.
Author's personal copy
Need for speed: The lending responsiveness of the IMF
Table 2 Negative Binomial Regression Estimates: Robust standard errors clustered by country in
parentheses; * p < 0.10, ** p < 0.05, *** p < 0.01
Model 5
Model 6
0.135
−0.390∗
0.143
0.149
(0.088)
(0.222)
(0.122)
(0.125)
−0.070
0.292
−0.104
−0.098
(0.077)
(0.229)
(0.107)
(0.103)
0.049
0.043
−0.013
0.032
(0.076)
(0.152)
(0.104)
(0.102)
Requests Per Month
−0.018
−0.033
0.029
0.012
(0.026)
(0.043)
(0.033)
(0.032)
Speculative Attack
−0.058
−0.056
−0.040
−0.029
(0.087)
(0.157)
(0.111)
(0.114)
0.052
0.367
0.063
−0.039
(0.164)
(0.457)
(0.189)
(0.180)
0.002
−0.003
0.006
0.008
(0.009)
(0.027)
(0.013)
(0.012)
GDP t-1 (log)
Population t-1 (log)
Under IMF Program
Debt Crisis
Regime Type t-1
IMF Quota Share (log)
UNSC Member
UN Ideal Point Distance t-1
Foreign Aid t-1 (log)
Africa
Asia
Americas
Model 7
Model 8
−0.022
−0.024
−0.014
−0.005
(0.087)
(0.149)
(0.152)
(0.139)
−0.008
−0.061
0.111
0.068
(0.129)
(0.241)
(0.153)
(0.160)
0.189∗
0.404
0.143
0.107
(0.105)
(0.333)
(0.119)
(0.126)
0.012
−0.086
0.009
0.006
(0.010)
(0.102)
(0.010)
(0.010)
0.125
−0.082
0.315
0.476
(0.241)
(0.211)
(0.314)
(0.289)
−0.510∗∗
−0.931∗∗∗
−0.237
−0.041
(0.201)
(0.285)
(0.252)
(0.242)
−0.228
−0.157
−0.109
0.031
(0.174)
(0.435)
(0.241)
(0.233)
MENA
0.031
0.143
0.300
(0.241)
(0.305)
(0.292)
Circulation Period
−1.617∗∗∗
−0.228
−0.957∗∗
(0.333)
(0.351)
(0.390)
Concerted Lending
3.152∗∗∗
(1.135)
Debt Service t-1 (log)
5.341∗∗∗
(1.483)
Bond Debt t-1 (log)
2.280
(2.918)
Author's personal copy
D. McDowell
Table 2
(continued)
Model 5
Model 6
Model 7
Model 8
−1.258∗∗
Short-term Debt t-1 (log)
(0.0170)
Bank Exposure t-1 (log)
−0.052
−0.500∗∗
−0.034
0.089
(0.034)
(0.187)
(0.047)
(0.073)
Concerted Lending*Bank Exposure
0.109∗∗
(0.044)
0.213∗∗∗
Debt Service*Bank Exposure
(0.063)
Bond Debt*Bank Exposure
0.096
(0.122)
−0.056∗∗
Short-term Debt*Bank Exposure
(0.010)
−6.706
Constant
0.005
0.634
3.167
(2.228)
(5.694)
(3.312)
(3.392)
Observations
275
63
175
180
Number of countries
87
39
59
63
Years
1984-2012
1984-1987
1988-2012
1988-2012
AIC
2364.6
605.04
1461.5
1486.5
less noisy and does not systematically omit cases. Yet, it is also quite narrow. Thus, it
cannot tell us anything about the factors that affect the duration of the first two stages
(see Fig. 1). What if the same factors that I argue influence the duration of Stage 3
also affect the duration of Stages 1 and 2? If this were the case, then it might suggest
the stages are not as neatly separable as I have argued. With this in mind, I fit four
additional Cox models substituting Mody and Saravia’s measure of IMF responsiveness for my own. In keeping with their approach, I first identify the month in which
a speculative attack on a country’s currency occurred using the approach developed
by Eichengreen et al. (1995) and Kaminsky and Reinhart (1999). This is the same
approach I used to create the covariate Speculative Attack with one subtle
difference. In that case, an attack was identified when the index was two standard
deviations above the mean. In this case, I follow Mody and Saravia’s lead and identify
a “crisis” as occurring when the indicator is only one standard deviation above the
mean.
Next, I calculated the number of days from the first day of the crisis month until
the date on which an IMF program is approved.42 This span serves as the dependent variable. Of course, there is no direct way to link a specific crisis to a particular
IMF program. Thus, I follow Mody and Saravia in assuming that any such program
that was negotiated within two years of the crisis month was related to that crisis.
42 Mody
and Saravia (2013) calculate the number of months between crisis and IMF program. However,
measuring in days allows me to avoid losing information due to rounding up or down to the nearest month.
Author's personal copy
Need for speed: The lending responsiveness of the IMF
Table 3 Cox Proportional Hazards Estimates: Robust standard errors clustered by country in parentheses;
* p < 0.10, ** p < 0.05, *** p < 0.01
Model 9
Model 10
Model 11
Model 12
GDP t-1 (log)
−0.644∗
−1.566∗
−0.627
−1.027∗∗
(0.291)
(0.860)
(0.464)
(0.423)
Population t-1 (log)
0.062
0.658
0.119
−0.035
(0.185)
(0.671)
(0.254)
(0.227)
1.013∗∗
1.355∗∗∗
0.574
0.593
(0.362)
(0.605)
(0.490)
(0.496)
Requests Per Month
0.058
−0.240∗∗
0.142
181∗∗
(0.064)
(0.114)
(0.101)
(0.098)
Speculative Attack
−0.302
−0.442
−0.090
−0.179
(0.235)
(0.493)
(0.442)
(0.433)
−0.330
−0.487
0.656
0.372
(0.486)
(1.620)
(0.612)
(0.610)
IMF Quota Share (log)
0.668∗∗
1.023
0.514
1.150∗∗
(0.279)
(0.642)
(0.458)
(0.487)
Regime Type t-1
−0.026
0.081∗
0.005
−0.011
(0.025)
(0.053)
(0.041)
(0.040)
UNSC Member
0.619
−1.952
1.055∗∗
1.622∗∗∗
(0.379)
(1.101)
(0.512)
(0.487)
UN Ideal Point Distance t-1
−0.203
2.635∗∗∗
−0.139
−0.205
(0.299)
(1.087)
(0.398)
(0.383)
Under IMF Program
Debt Crisis
−0.070∗∗
0.345
0.024
−0.051
(0.036)
(0.353)
(0.077)
(0.080)
−0.225
0.166
0.056
(0.670)
(1.025)
(0.987)
Asia
0.325
0.948
1.140
(0.556)
(0.819)
(0.802)
Americas
1.377∗∗
0.897
1.005
(0.598)
(0.856)
(0.862)
MENA
0.773
0.980
1.540∗
(0.660)
(0.856)
(0.885)
Circulation Period
2.963∗∗∗
2.819∗∗∗
3.485∗∗∗
(0.795)
(0.913)
(0.957)
Concerted Lending
−0.897
Foreign Aid t-1 (log)
Africa
(2.889)
Debt Service t-1 (log)
−3.079
(7.583)
Bond Debt t-1 (log)
3.341
(8.857)
Short-term Debt t-1 (log)
7.068∗∗
(2.779)
Author's personal copy
D. McDowell
Table 3
(continued)
Model 9
Model 10
Model 11
Model 12
Bank Exposure t-1 (log)
0.010
0.732
0.049
−0.541∗
(0.160)
(1.002)
(0.235)
(0.321)
Concerted Lending*Bank Exposure
−0.071
(0.116)
−0.180
Debt Service*Bank Exposure
(0.307)
Bond Debt*Bank Exposure
0.064
(0.372)
0.286∗∗
Short-term Debt*Bank Exposure
(0.115)
Observations
141
39
86
Number of countries
67
29
42
90
43
Years
1984-2008
1984-1987
1988-2008
1988-2008
Logrank Test
82.271∗∗∗
36.305∗∗∗
72.777∗∗∗
78.142∗∗∗
Moreover, I also use the earliest crisis within that window as a starting point.43
Unfortunately, quite a large number of IMF programs were negotiated without an
identified crisis having occurred in the previous two years. Consequently, these cases
are dropped from the analysis. Additionally, a number of IMF programs were negotiated several months prior to the occurrence of a crisis, suggesting authorities reached
a deal with the Fund in anticipation of trouble. However, again, such cases are
dropped from the analysis using this approach. Ultimately, using Mody and Saravia’s
measure means that more than 130 cases are omitted from the analysis.
Despite these challenges, I fit four additional Cox models using Mody and Saravia’s measure of IMF responsiveness. Table 3 reports the results. Looking first to
Model 9, the interaction between Concerted Lending and Bank Exposure
is negatively signed yet it is not statistically significant. Thus, based on this measure,
we cannot say that G-5 bank exposure is associated with slower IMF responsiveness
during the concerted lending years and speedier responsiveness in the more recent
era. This may signify that the additional “noise” introduced into the dependent variable has weakened the relationship between G-5 financial interests and speed. Or,
it may reflect the loss of so many observations. The story is similar for Model 10
which is restricted to the concerted lending era. Here, the interaction between Debt
Service and Bank Exposure remains negatively signed, yet again, it is no
longer statistically significant. This suggests that G-5 bank exposure and the imminence of a borrower’s default do not interact to explain the duration of all three stages
collectively during that period. Models 11 and 12 include only the post-concerted
43 There is one caveat here: In a number of cases, a borrower negotiated a new IMF program less than two
years after its previous program. In such cases, I use the earliest crisis since the previous IMF program as
the starting point.
Author's personal copy
Need for speed: The lending responsiveness of the IMF
lending years. As was the case in the previous two estimations, model 11 results suggest that borrower country exposure to bond markets does not interact with G-5 bank
exposure to affect IMF responsiveness.
However, the interaction between Short-term Debt and Bank Exposure
in Model 12 is positive and highly significant. This is the only case where the results
using Mody and Saravia’s measure are consistent with the results using my own
measure. Above, I showed that in the post-concerted lending era, exposure to shortterm capital flows and G-5 bank exposure interact to accelerate the duration of Stage
3 in isolation. These results suggest that they may also hasten the duration of stages 1
and 2. Several things may explain this results. First, borrowers with greater exposure
to short-term capital flows may recognize their vulnerability and approach the Fund
more swiftly in the first place. This would shorten the duration of Stage 1. It may
also be the case that when the borrower government and IMF Staff are in negotiations
(Stage 2) both parties may seek a speedier deal given the clear need for speed. Lastly,
it may be the case that G-5 may intervene to reduce the duration of Stage 2 in some
cases. If these key Fund shareholders believe their financial systems are seriously
threatened by an unfolding crisis in a borrower’s economy, they may exert pressure
on either the borrower or on IMF staff to speed up the negotiation process so that the
crisis can be resolved as quickly as possible. Of course, this is speculation. The best
way to answer these questions is to craft measures that isolate the duration of each
stage separately. Ultimately, the results using Mody and Saravia’s measure support
my assertion that focusing on Stage 3 in isolation paints a less noisy picture of IMF
responsiveness and makes it easier to identify the factors that affect speed during that
specific part of the lending process.
7 Discussion and conclusions
As financial crises unfold in today’s global financial system, the speed of the emergency response is often as important as the loan itself. In this paper, I have presented
new data on the responsiveness of the IMF. Specifically, I calculate the number of
days that transpire between when a country formally requests IMF assistance and
when the Executive Board approves that request–what I refer to as the loan approval
period. Analysis of these data reveal two important findings.
First, my data reveal that the IMF has become a more responsive lender over time.
Indeed, loan approval periods decreased year-by-year–almost monotonically–across
my sample. Coupled with numerous steps taken by the Fund to improve the speed
of its crisis response abilities, this trend provides clear evidence that the IMF has
adapted its bureaucratic routines to a changing external environment. Thus, critiques
of the Fund as an unresponsive lender may no longer be as credible as they once were.
For an institution that is often criticized for its inability to reform itself to changing
conditions, this is a notable accomplishment. On its own, this trend reflects positively on the Fund’s ability to learn and adapt to changing global conditions within
which it operates. This paper provides clear evidence that the IMF is better able to
rapidly respond to countries’ requests for help today than it was in the past. All else
equal, the Fund’s speed should make it a more effective crisis manager which is a
Author's personal copy
D. McDowell
positive development for global financial stability. Yet, even as the IMF’s responsiveness improved broadly, the increased sense of urgency over time has not been applied
evenly across borrowers.
This points to the second–and key–finding of the study: the financial interests of
the G-5 countries influence IMF responsiveness in systematic ways contingent on the
time period in which a request was made. IMF responsiveness is directly related to
strategies used by the Fund to catalyze private lending on behalf of a borrower. From
1984 through 1987, increased G-5 bank exposure led to longer waits for approval,
especially as borrowers edged closer to default. The results clearly suggest that during
this era, the Fund was more likely to implement its concerted lending strategy on
behalf of such borrowers in order to secure ex ante guarantees of additional bank
financing needed to prevent default. The result of this was delayed Board approval.
In the post-concerted lending era, increased exposure of G-5 banks led to swifter
approvals as the Fund shifted its catalytic financing strategy. During these years, the
IMF was more likely to expedite requests from such borrowers in order to attract
back global creditors–especially when the borrower was also highly dependent on
short-term debt markets. In short, when G-5 financial interests were at stake, these
major shareholders tended to get what they want. However, what they want changed
over time as the structure of the international financial system evolved and the need
for speed increased. These findings contrast with Mody and Saravia’s (2013) study
on the same topic, indicating that how we measure responsiveness influences how we
explain variation in responsiveness. Broadly speaking, however, my results are quite
consistent with numerous studies that have shown the financial interests of major
IMF shareholders affect the institution’s lending behavior.
As the speed and complexity of financial markets continues to grow, the need for
speed on the part of the Fund as a financial crisis manager is only going to increase.
Consequently, more work needs to be done to understand all of the stages that comprise the time between crisis onset and the approval of an IMF program. Here, I focus
on the third and final stage in this process. However, my analysis tells us little about
why some countries approach the IMF quickly while others wait much longer to ask
for help. Nor do I seriously consider the factors that speed up or slow down negotiations between IMF staff and borrower governments. My results employing Mody and
Saravia’s measure of responsiveness are suggestive that–at least in the post-concerted
lending era–short-term debt dependence and G-5 bank exposure may also affect the
duration of the earlier stages as well.
References
Bagehot, W. (1873). Lombard street: a description of the money market. London: Henry S. King and
Company.
Bailey, M.A., Strezhnev, A., & Voeten, E. (2015). Estimating dynamic state preferences from United
Nations voting. Journal of Conflict Resolution. Forthcoming.
Bird, G., & Rajan, R.S. (2002). The evolving asian financial architecture. Essays in international
economics. No. 226. Princeton University: International Economics Section.
Bird, G., & Rowlands, D. (2004). Does the IMF perform a catalytic role? And What if it Doesn’t? World
Economics, 5, 117–132.
Author's personal copy
Need for speed: The lending responsiveness of the IMF
Barro, R.J., & Lee, J. (2005). IMF programs: who is chosen and what are the effects? Journal of Monetary
Economics, 52, 1245–1269.
Bordo, M.D., & James, H. (2000). The international monetary fund: its present role in historical
perspective. Greek Economic Review, 20, 43–76.
Boughton, J.M. (2000). From suez to Tequilla: the IMF as crisis manager. The Economic Journal, 110,
273–291.
Boughton, J.M. (2001). Silent revolution: the international monetary fund, 1979–1989. Washington D.C.:
International Monetary Fund.
Boughton, J.M. (2012). Tearing down walls: the international monetary fund 1990-1999. Washington
D.C.: International Monetary Fund.
Box-Steffensmeier, J.M., & Zorn, C.J.W. (2001). Duration models and proportional hazards in political
science. American Journal of Political Science, 45, 972–988.
Broz, J.L., & Hawes, M.B. (2006). Congressional politics of financing the international monetary fund.
International Organization, 60, 367–399.
Calvo, G.A. (1998). Varieties of capital-market crises. In G.A. Calvo & M. King (Eds.), The debt burden
and its consequences for monetary policy (pp. 181–202). New York: St. Martin’s Press.
Calvo, G.A., Leiderman, L., & Reinhart, C.M. (1996). Inflows of capital to developing countries in the
1990s. Journal of Economic Perspectives, 10, 123–139.
Caskey, J.P. (1989). The IMF and concerted lending in latin American debt restructurings: A formal
analysis. Journal of International Money and Finance, 8, 105–120.
Copelovitch, M.S. (2010). The international monetary fund in the global economy: banks, bonds, and
bailouts. Cambridge: Cambridge University Press.
Dreher, A., & Jensen, N. (2007). Independent actor or agent? An empirical analysis of the impact of US
interests on IMF conditions. Journal of Law and Economics, 50, 105–124.
Dreher, A., Sturm, J.E., & Vreeland, J. (2009a). Global horse trading: IMF loans for votes in the United
Nations Security Council. European Economic Review, 53, 742–757.
Dreher, A., Sturm, J.E., & Vreeland, J. (2009b). Development aid and international politics: does membership on the UN Security Council influence World Bank decisions? Journal of Development
Economics, 88, 1–18.
[EBD] Executive Board Document (1985). requests for waiver of circulation periods of papers on the
executive board agenda. EBD/85/5. International Monteary Fund. 4 January.
[EBD] Executive Board Document (1996). Minimum circulation periods for board documents.
EBD/96/33. International Monteary Fund. 13 March.
[EBM] Executive Board Minutes (1983a). Minutes of the executive board meeting 83/57. International
Monetary Fund. 6 April.
[EBM] Executive Board Minutes (1983b). Minutes of the executive board meeting 83/65. International
Monetary Fund. 22 April.
[EBM] Executive Board Minutes (1995). Minutes of the executive board meeting 95/11. International Mon
teary Fund. 19 November.
Eichengreen, B., Rose, A.K., Wyplosz, C., Dumas, B., & Weber, A. (1995). Exchange market mayhem:
The antecedents and aftermath of speculative attacks. Economic Policy, 10, 249–312.
Erb, R.D. (1983). Statement by Mr. Erb on fund policies and external debt servicing problems, executive
board meeting 83/57. International Monetary Fund. 6 April.
Fernandez-Arias, E. (2010). International lending of last resort and sovereign debt restructuring. In C.A.P.
Braga & G.A. Vinclette (Eds.), Sovereign debt and the financial crisis: Will this time be different? (pp.
331–353). Washington, DC: World Bank.
Fernandez-Arias, E., & Levy-Yeyati, E. (2010). Global financial safety nets: where Do we go from here?
Inter-american development bank, working paper No. IDB-WP-231. Available at http://www.iadb.org/
en/research-and-data/publication-details,3169.html?pub id=idb-wp-231. Accessed 30 Nov 2015.
Gartzke, E. (1998). Kant we all just get along? Opportunity, willingness, and the origins of the democratic
peace. American Journal of Political Science, 41, 1–27.
Greenspan, A. (2007). The age of turbulence: adventures in a new world. New York: Penguin Press.
Guitian, M. (1995). Conditionality: past, present, future. Staff Papers - International Monetary Fund, 42,
792–835.
Imai, K., King, G., & Lau, O. (2007). Zelig: everyone’s statistical software. Available at http://GKing.
harvard.edu/zelig. Accessed 25 Oct 2015.
Imai, K., King, G., & Lau, O. (2008). Toward a common framework for statistical analysis and
development. Journal of Computational and Graphical Statistics, 17(4), 892–913.
Author's personal copy
D. McDowell
International Monetary Fund (2009). GRA lending toolkit and conditionality: Reform Proposals. Available
at http://www.imf.org/external/np/pp/eng/2009/031309A.pdf. Accessed 30 Nov 2015.
International Monetary Fund (2012a). The IMF’s rapid credit facility: Factsheet. Available at http://www.
imf.org/external/np/exr/facts/rcf.htm. Accessed 13 Sept 2012.
International Monetary Fund (2012b). The IMF’s flexible credit line: Factsheet. Available at http://www.
imf.org/external/np/exr/facts/fcl.htm. Accessed 13 Sept 2012.
International Monetary Fund (2012c). IMF Conditionality: Factsheet. Available at http://www.imf.org/
external/np/exr/facts/conditio.htm. Accessed 28 Sept 2012.
International Monetary Fund (2012d). The IMF’s rapid financing instrument (RFI). Available at http://
www.imf.org/external/np/exr/facts/rfi.htm. Accessed 26 Oct 2012.
Jeanne, O., & Wyplosz, C. (2001). The international lender of last resort: how large is large enough?
NBER Working Paper, 8381.
Kaminsky, G.L., & Reinhart, C.M. (1999). The twin crises: the causes of banking and balance-of-payments
problems. The American Economic Review, 89, 473–500.
Kilby, C. (2013). The political economy of project preparation: an empirical analysis of world bank
projects. Journal of Development Economics, 105, 211–225.
Knight, M., & Santaella, J.A. (1997). Economic determinants of IMF financial arrangements. Journal of
Development Economics, 54, 405–436.
Kuziemko, I., & Werker, E.D. (2006). How much is a seat on the security council worth? Foreign aid and
bribery at the United Nations. Journal of Political Economy, 114, 905–930.
Laeven, L., & Fabian V. (2008). Systemic banking crises: a new database. IMF Working Paper,
WP/08/224. https://www.imf.org/external/pubs/ft/wp/2008/wp08224.pdf. Accessed 23 Dec 2015.
Leblang, D. (2003). To devalue or defend? The political economy of exchange rate policy. International
Studies Quarterly, 47, 533–559.
Marshall, M.G., Jaggers, K., & Gurr, T.R. (2012). Polity IV.
Mody, A., & Saravia, D. (2006). Catalysing private capital flows: do IMF programmes work as
commitment devices? The Economic Journal, 116, 843–867.
Mody, A., & Saravia, D. (2013). The response speed of the international monetary fund. International
Finance, 15, 189–211.
Morris, S., & Shin, H.S. (2006). Catalytic finance: when does it work? Journal of International Economics,
70, 161–177.
Moser, C., & Sturm, J.E. (2011). Review of international organizations. Explaining IMF lending decisions
after the cold war, 6, 307–340.
Munk, R. (2010). Exchange stabilization fund loans to sovereign borrowers: 1982-2010. Law and
Contemporary Problems, 73, 215–240.
Nelson, S. (2014). Playing favorites: how shared beliefs shape the IMFs lending decisions. International
Organization, 68(2), 297–328.
Oatley, T., & Yackee, J. (2004). American interests and IMF lending. International Politics, 41, 415–429.
Prasad, E., Rogoff, K., Wei, S.J., & Kose, M.A. (2003). Effects of financial globalization on developing
countries. IMF Working Paper.
Rodrik, D., & Velasco, A. (1999). Short-term capital flows. NBER Working Paper, 7364.
Sachs, J. (1988). International policy coordination: the case of the developing country debt crisis. In M.
Feldstein (Ed.), International economic cooperation (pp. 233–320). Chicago: University of Chicago
Press.
Sachs, J. (1995). Do we need an international lender of last resort? Frank D. Graham Lecture, Princeton
University, April 20.
Schwartz, A.J. (2002). Earmarks of a lender of last resort. In C.A.E. Goodhart & G. Illing (Eds.), Financial
crises, contagion, and the lender of last resort (pp. 449–460). Oxford: Oxford University Press.
Signorino, C.S., & Ritter, J.M. (1999). Tau-b or Not Tau-b: measuring the similarity of foreign policy
positions. International Studies Quarterly, 43, 115–144.
Smuc, T., Gamberger, D., Antulov-Fantulin, N., Lucanin, D., & Piskorec, M. (2015). FOC Project. http://
lis.irb.hr/foc/data/data.html. Accessed 4 Aug 2015.
Stone, R. (2004). The political economy of IMF lending in Africa. American Political Science Review, 98,
577–591.
Stone, R. (2008). The scope of IMF conditionality. International Organization, 62, 589–620.
Stone, R. (2011). Controlling institutions: international organizations and the global economy. Cambridge: Cambridge University Press.
Author's personal copy
Need for speed: The lending responsiveness of the IMF
Thacker, S.C. (1999). The high politics of IMF lending. World Politics, 52, 38–75.
Truman, E.M. (1996). The mexican peso crisis: implications for international finance (pp. 200–209).
Federal Reserve Bulletin. March.
Van Houtven, L. (2002). Governance of the IMF: Decision making, institutional oversight, transparency,
and accountability. Washington D.C.: International Monetary Fund.
Vreeland, J. (2003). The IMF and economic development. New York: Cambridge University Press.
Vreeland, J. (2007). The international monetary fund: politics of conditional lending. New York:
Routeledge.
Zettelmeyer, J. (2000). Can official crisis lending be counterproductive in the short run? Economic Notes,
29, 12–29.