Transcript copelovitch_F1000_1_pres

Master or Servant?
Agency Slack and the Politics of IMF Lending
Mark Copelovitch
Department of Political Science
University of Wisconsin-Madison
IPES Conference
November 17, 2006
Who “Controls” the International Monetary Fund?
•
“The IMF…is a set of ‘silk-suited dilettantes’ given to ‘champagne
and caviar at the expense of the American taxpayer.’”
– Sen. Lauch Faircloth (R-NC), The Wall Street Journal, 3/27/98
•
“Everywhere else in the world…politicians and businessmen insist
that one of the biggest problems with the IMF is that…it acts as the
United States Treasury's lap dog.”
– David Sanger, New York Times, 10/2/98
The Politics of IMF Lending: Two Main Perspectives
Principal(s)
Executive Board?
US government
Agent
What explains
lending variation?
IMF staff
Technocratic economic
criteria or staff
rent-seeking
(Knight/Santaella,
Dreher/Vaubel)
IMF staff
US geopolitical or
financial interests
(Thacker, Oatley,
Broz, Stone)
Existing Explanations: Problems and Questions
A mixed empirical record
• Similar economic circumstances, different loans
• Some countries with strong ties to US get better deals than others
Conceptual gaps
• US has strongest voice, but not veto, over IMF decisions
– “G-5” countries all exercise significant authority
• Bureaucratic rent-seeking: when does the staff “get away” with it?
Argument in Brief
A principal-agent model of IMF policymaking
• G-5 governments as the “collective principal”
• IMF staff as agent
Main findings
• Both states and IMF staff exercise partial but incomplete
control over Fund lending
• Agency slack is case-specific: staff autonomy is conditional on
the intensity and heterogeneity of principal (G-5) interests
– Must account for other large shareholders’ interests, not just US
– IMF staff are not “runaway” bureaucrats
A Collective Principal Model of IMF Lending
IMF Executive Board (“G-5”)
US
UK
GE
R
JP
N
IMF staff
FR
A
Preferences influenced
by domestic interests
(geopolitical, financial)
Preferences influenced
by economic criteria and
bureaucratic incentives
Borrower country
Key question: how much “agency slack” in a given lending case?
G-5 Bank Exposure in Recent IMF Lending Cases
Aggregate
G-5 exposure
($billions)
100%
68.5
39.5
50.3
60.6
34.6
0.3
6.9
90%
80%
70%
France
60%
Germany
50%
Japan
40%
UK
30%
US
20%
10%
0%
Korea 1997
Mexico 1995
Brazil 2002
Thailand 1997 Russia 1999
Bosnia 2002
SOURCE: Bank for International Settlements. Consolidated International Banking Statistics
Croatia 2003
Measuring Agency Slack:
G-5 Interest Intensity and Heterogeneity
Heterogeneity of G-5 interests
Intensity of
G-5 interests
High
Low
Low
High
• G-5 consensus
• G-5 conflict
• Largest loans
• Large loans, but
“logrolling” cost
• G-5 consensus
• G-5 conflict
• Smallest loans
• Small loans, but
“rent-seeking” premium
Mean Loan Size (Amount/Quota) by G-5 Bank Exposure
Short-term IMF loans, 47 countries, 1984-2003
G-5 bank exposure, coefficient of variation
Aggregate G-5
bank exposure
High*
Low*
Low*
High*
AMTQTA=2.04
(N=84)
AMTQTA=1.12
(N=34)
AMTQTA=0.63
(N=26)
AMTQTA=0.60
(N=64)
*Above or below sample mean in a given year
Empirical Analysis
Dataset
• 197 short-term IMF loans to 47 countries, 1984-2003
Sources
• IMF archival documents
• IMF, World Bank, and BIS databases
Dependent variables
• Loan sizei,t – new short-term IMF lending/quota (log)
• Robust to alternative specifications (raw amount, amount/GDP)
Models
• OLS, panel-corrected standard errors, “modified” lagged DV,
country fixed effects
• Robust to alternative specifications
Variables
Explanatory variables
• Measures of aggregate G-5 interests
• Bank exposure, foreign aid commitments, UN voting affinity
• Weighted by relative voting power of G-5 countries
• Measures of G-5 interest heterogeneity
• Coefficients of variation of bank exposure, foreign aid, and UN affinity
• 100*(std/mean) = measures dispersion as % of mean
Control variables
• Borrower macroeconomic/political characteristics
• Temporal trends/global conditions
• “Modified” lagged DV (dummy for outstanding previous loans)
First Differences - IMF Loan Size (Model 3)
Predicted change in
loan size (AMTQTA)
Interpretation
Length of loan
45.77%
20 to 30 months
GDP / quota
27.61%
67.4 to 127.74 times quota
External debt / GDP
24.30%
58.49 to 94.3
Debt service / exports
34.35%
23.55 to 42.26
Short-term debt / reserves
14.39%
0.79 to 3.09
Quota review
-12.30%
0 to 1
Variable
Coefficient of variation, G5BANK
G5BANK=3.55
G5BANK=7.36
G5BANK=9.9
Predicted change in loan size (AMTQTA)
11.16%
-10.18%
-12.29%
Effect of G-5 Interest Heterogeneity at Different
Levels of G-5 Interest Intensity - Bank Exposure
Effect of G-5 Interest Heterogeneity at Different
Levels of G-5 Interest Intensity - Foreign Aid
Effect of G-5 Interest Heterogeneity at Different
Levels of G-5 Interest Intensity - UN Voting Affinity
Main Findings
G-5 governments’ interests heavily influence IMF lending
• Amount and distribution of bank exposure and foreign aid
significantly influence loan characteristics
• UN voting affinity has less clear effects
Agency slack depends on G-5 interest intensity & heterogeneity
• Staff autonomy increases when G-5 interests are weak and divided
IMF lending is highly political
• Evidence for both common political explanations, but each is
conditional on the other
Implications and Conclusions
Understanding IO behavior
• Beyond questions of cooperation and institutional design
– How do IOs make decisions once the rules/institutions are
established?
• Focusing on delegation/agency and internal decision-making
rules is critical
Reforming the IMF
• Abolishing the IMF/curtailing lending
– Would not eliminate G-5 interests, but would simply shift the focus
to bilateral/ad hoc official lending
• Executive Board voting reform
– Replacing G-5 domestic interests with other countries’ is unlikely to
remove politics from the process