Transcript edwards_F330_1_pres

Implementation Uncertainty
and the Design of IMF
Conditionality
Martin S. Edwards
Whitehead School of Diplomacy
Seton Hall University
The Limits of Conditionality



40% of International Monetary Fund lending
programs are suspended for noncompliance
Program suspension has real costs for both IMF
and the borrowing state
To what extent is the IMF to blame for the
failings of conditionality?
Question

How does the IMF design loan programs if
it does not know ex ante whether the
borrowing state will honor its promises?
The Argument:

IMF devises conditionality both
prospectively and retrospectively
 Conditionality
is front-loaded in states with
approaching elections and federal states
 Conditionality is front-loaded in states with a
history of poor program implementation.
Does Implementation Uncertainty
Matter for the IMF?


Joseph Stiglitz (2002:52): “Sometimes conditionality was
even counterproductive, either because the policies were
not well suited to the country or because the way they
were imposed engendered hostility to the reform
process.”
Stanley Fischer (1998): “We don't need to form very
sophisticated judgments about the political forces in
(those) countries. We basically have to form a judgment
on whether the government will do what it says it will do
in an overall satisfactory way.”
Forming international agreements

Variations in the form of international
agreements are responses to uncertainty
 Escape
clauses / renegotiation are ways to adjust
agreements

Agreements made under high transaction costs
look different than those in which parties face
low transaction costs
 Ex
ante commitments can be used to make promises
more credible
Assumptions
Fund faces uncertainty about borrower
commitment
 Program causes domestic dislocation
 Renegotiation costly for IMF
 Renegotiation costly for borrower

If I’m right…..

Expect more conditions in states with a
higher level of implementation uncertainty.

But these conditions should be imposed
on states ex ante.
 IMF
imposes conditions (prior actions) in
order for Executive Board to approve program
What shapes implementation
uncertainty?

Approaching elections
 We
know that electoral cycles are common in
developing countries, and that elections often
coincide with program interruptions

Federalism
 Economic

adjustment more difficult in these states
Past performance
 Failed
programs lead to changes
Expectations

Prospective
 Programs
with approaching elections are more likely
to have prior actions than those in which elections are
not approaching
 Federal states more likely to have prior actions

Retrospective
 Programs
negotiated in the wake of a suspended
program are more likely to have prior actions than
those negotiated following a successfully completed
program.
Cases

Random Sample of 38 States under
Programs from 3rd qtr 1997 to 2nd qtr 2003.
 Asia
(7 states), Latin America (8), Africa (11),
Eastern Europe (12)
 183 agreements initiated in this time period
 States required to submit new memoranda as
part of program review process
Dependent Variable

Count of fiscal prior actions for each
agreement
 Range:

0-12
Prior actions are those policy measures
that are prerequisites for review by
Executive Board
Examples of Fiscal Prior Actions

Measures adopted to reduce wage bill


Measures to increase revenue



Russia 99q3: Delay law on VAT reduction
Pakistan 01q1: Mandated increases in rates for electricity and
gasoline
Measures to reduce expenditure


Cambodia 99q3: freeze on new hiring for civil service
Armenia 98q4: Publish decree detailing govt plan to reduce
expenditures by 7 billion dram.
Passage of Fund-compliant budget by legislature
Why Fiscal Prior Actions?

Have to look at prior actions to ascertain whether Fund
incorporates implementation uncertainty


Looking at performance criteria makes causal chain unclear
Fiscal criteria serves as “most likely” case



Electoral cycles manifest themselves through higher levels of
spending
Federal states esp. prone to fiscal problems
Noncompliance stems most frequently from fiscal shortfalls
Independent Variables

Policy Stance
 Lagged
budget deficit / GDP (-)
 Lagged growth of expenditures / GDP (+)
 Lagged growth of reserves (-)
 Lagged GDP (-)

Type of Agreement
 Dummy

for PRGF/ESAF (+)
US influence
 Lagged
US foreign aid / GDP (-)
Independent Variables

Approaching elections (Dummy)
 Is
an executive election six months away?

Federalism (Dummy)

Status of past program (Dummy)
 Did
the Fund interrupt the program for
noncompliance?
Research Design

Confront substantial missing data
problems
 Missingness
on fiscal variables approaches
40% of sample

Used multiple imputation (King et al 2000)
to address missing data
Research Design

Empirical Test is a negative binomial
model
 Appropriate
because data are counts
 Distribution of count dictates model
specification

Population averaged to address panel
heterogeneity
Change in Total Reserves t-1
-0.00025**
(9.102 E-05)
GDP t-1
-2.43 E-06
(1.008 E-05)
Budget Deficit / GDP t-1
0.0114
(0.0159)
Change in Expenditure / GDP t-1
0.0022
(0.0013)
PRGF Dummy
-0.1651
(0.2652)
US Foreign Aid / GDP t-1
0.3860
(0.5012)
Approaching Elections
0.7242*
(0.3591)
Prior Program Noncompliance
0.8669***
(0.2581)
Federalism
1.061*
(0.5360)
Number of Performance Criteria
0.0879
(0.0504)
Dummy for Ukraine Program
2.6957***
(0.2711)
Constant
-0.6510
(0.3481)
Substantive – Discrete Change
Change in Total Reserves t-1
-8.510
GDP t-1
-1.685
Budget Deficit / GDP t-1
1.509
Change in Expenditure / GDP t-1
1.199
PRGF Dummy
-.403
US Foreign Aid / GDP t-1
1.288
Approaching Elections
2.572
Prior Program Noncompliance
3.338
Federalism
4.571
Number of Performance Criteria
2.816
Dummy for Ukraine Program
2.256
Robustness Checks

Results hold in the presence of the
following
 Controls
for Quota in IMF
 Controls for Democracy
 Trade Openness
 Changes in Exchange Rate level

Results unchanged with nine-month
window.
Summary

IMF does not lend “in the dark”
 Expectations
about electoral horizons affect
program design
 Programs look differently in federal states
 Expectations about program compliance
affect program design
A Further Question to Consider

Do prior actions make a difference ex
post?
 In
fiscal performance?
 In the probability of program suspension?