The Strategic Politics of Imf Conditionality

Home , Conditionality

THE STRATEGIC POLITICS OF IMF CONDITIONALITY

DISSERTATION

Presented in Partial Fulﬁllment of the Requirements for the Degree Doctor of

Philosophy in the Graduate School of the Ohio State University

Byungwon Woo, B.A./M.A.

Graduate Program in Political Science

The Ohio State University

2010

Dissertation Committee:

Daniel Verdier, Advisor

Irfan Nooruddin

Alexander Thompson c Copyright by

Byungwon Woo

2010 ABSTRACT

The dissertation theorizes how domestic politics in a borrowing country inﬂuences the design of IMF conditionality, a set of policy reform measures included in an

International Monetary Fund program. Considering the ways in which domestic politics can shape the outcome of negotiation between the IMF and a government, there are two alternative logics that can play out: the government can either tie its hands to the IMF to force reforms to domestic interests or tie its hands to domestic interests to extract a better deal from the IMF. Using a game theoretic model, I demonstrate that the eﬀect of domestic politics on the IMF program design hinges on the interaction of three parameters and suggest the following propositions: a government that is more sensitive to vote losses or less reform-minded is more likely to extract more lenient conditions from the IMF; a government free from electoral pressure receives more conditions; for those governments that are electorally less constrained, the severity of conditionality is limited only when there exist strong domestic interests that can hinder proper implementation of reform conditions.

To test the hypotheses, I constructed an original dataset of IMF conditionality by coding all 263 letters of intent agreed in between 1994 and 2006. I coded the number of conditions by conditionality type and by aﬀected economic sector. The empirical

ii analysis of public and fiscal sector conditions strongly supports the domestic politics hypotheses and yields the following findings. First, democratic countries receive fewer conditions than autocracies that are freer from electoral competition, suggesting that the IMF is strategic in limiting how much it pushes politically vulnerable negotiating partners. Furthermore, I show that among democracies, a government that won the previous election by a narrower margin receives fewer public sector conditions than a government that won the previous election with a wider margin and a government that has a more proximate election receives fewer public sector conditions than a government with a more distant election. Finally, strong domestic interests reduce the number of public sector conditions in an autocratic country but exert little influence over the number of conditions in a democratic country.

While domestic political institutions and interests exert considerable influence over the design of IMF conditionality in public and fiscal sector conditions, where each additional policy condition is politically costly, the influence of domestic and international politics vary substantially by affected economic sector. In designing financial sector conditions, where domestic financial interests are nascent or actually support

ﬁnancial reforms, domestic politics plays little role; instead, international politics plays a greater role. The empirical results justiﬁes the disaggregation approach that

I take in empirically examining the design of IMF conditionality by aﬀected economic sectors.

iii I dedicate the dissertation to those who have shaped who I am:

my late, dearly-missed grandma and my family

iv ACKNOWLEDGMENTS

I am indebted to many individuals who have helped me throughout the dissertation journey. Amazingly, they all generously allowed me to be indebted, without imposing any conditionality to me. I am grateful for their advice and support, and hope that my words can deliver at least a small fraction of my appreciation.

I am deeply grateful to my advisor, Daniel Verdier. Since my ﬁrst year in graduate school, he has taught me to think more logically as a political scientist, and from the inception of the dissertation idea to the conclusion of the ﬁnal draft, he has provided me with many rounds of insightful comments, critical advices, and much-needed encouragements. His generous criticisms and suggestions undoubtedly improved the quality of the dissertation. I admire his keen ability to recognize the core logic and his strict discipline toward his own research. I only hope that I can follow in his footsteps through my academic career.

I owe a great debt to Irfan Nooruddin. His cheerful encouragements have lifted me up and made me feel more conﬁdent, when I was frustrated with myself. During the dissertation stage, he not only provided me with crucial feedback on earlier drafts, but also presented me with many opportunities to experience the academia. He taught me how to make a better research presentation and how to prepare for the

v academic job market. Collaborating with him on a research project also allowed me to ﬁrsthand experience the best practice of doing research.

Alex Thompson has guided my academic career ever since I ﬁrst met him 10 years ago. I was only an aspiring undergraduate student when I ﬁrst met him, and he has advised me at every critical juncture in my transformation into a credible political scientist. He helped me when I applied for a graduate school, he taught me IR, he advised me to learn game theory, and he served on my dissertation committee.

I deeply appreciate his ability and willingness to read many earlier drafts of the dissertation and to give constructive and encouraging comments.

I am eternally thankful to Autumn Lockwood Payton and Daniel Blake for their camaraderie throughout the years in graduate school. Our ABD meetings were the most helpful venue to receive honest and good-hearted criticisms on earlier drafts of the dissertation, and they made our meetings something that I would look forward to.

They also provided emotional support, without which the dissertation writing could have been far more miserable. I am very fortunate to have A and D as colleagues, coauthors, and friends. I hope we continue our comradeship through our academic careers and beyond.

I would like to thank Jan Box-Steﬀensmeier and Alan Wiseman. Jan gave me many opportunities to work with her, ﬁrst as a research assistant and later as a

PRISM fellow, and provided ﬁnancial assistance that allowed me to take time oﬀ from teaching. Alan taught me game theory, allowed me to work as his teaching assistant, and encouraged me to explore further studying opportunities outside of the university. I am also grateful for their willingness to serve as a reference.

vi I am grateful to Brian Pollins for his rigorous training. He has helped me to become a better political scientist. His 747 is the best course that I have taken in graduate school and his constant push during the dissertation workshop have shaped the overall direction of the dissertation at its early stage.

I thank to all the wonderful friends that I have made in Columbus. In particular,

Quintin Beazer always makes people around him happier with his bright personality,

TongFi Kim always takes care of people around him, and Chaekwang You always makes people laugh with his own kind of humor. My years in Derby have been a lot more enjoyable thanks to all my friends, and I hope that we can have lasting friendships beyond graduate school.

On a more personal note, I am eternally grateful to my parents. I have always been their proud son, even at times when I was not very conﬁdent of myself. Their unconditional love and belief in me have sustained me throughout this long journey.

I am thankful to my brother for ﬁlling my shoes when I have been largely absent from family occasions. I also want to thank my in-laws for believing in Hee-Jung and me.

I am eternally thankful to my wife, Hee-Jung Jun. Her incredible work ethics has constantly inspired me and her emotional support has kept me going. Writing two dissertations was not easy for us, but we somehow managed to do it. We shared our joys and suﬀered together in the process, and I believe we understand and love each other better and more as a result. Thank you, Hee-Jung, for being amazing.

vii VITA

1977 ...... Born in Korea

2003 ...... B.A. in Political Science

2005 ...... M.A. in Political Science

2003-Present ...... Graduate Teaching/Research Associate, The Ohio State University

FIELDS OF STUDY

Major Field: Political Science

Specialization: International Relations

Specialization: Political Methodology & Formal Theory

viii TABLE OF CONTENTS

Abstract ...... ii

Dedication ...... iii

Acknowledgments ...... v

Vita...... viii

List of Tables ...... xii

List of Figures ...... xiv

CHAPTER PAGE

1 Introduction ...... 1

1.1 Introduction ...... 1 1.2 IMF Program Design as International Bargaining ...... 8 1.2.1 IMF Program ...... 8 1.2.2 IMF Conditionality ...... 9 1.2.3 Negotiating IMF Conditionality ...... 11 1.2.4 Variation in IMF Conditionality ...... 13 1.3 Literature Review ...... 16 1.3.1 Economic Account ...... 17 1.3.2 Political Account ...... 19 1.3.3 Two-Level Games Literature ...... 28 1.4 Criticisms and Preview of the Argument ...... 31 1.5 Conclusion and Plan of the Dissertation ...... 34

2 A Theory of IMF Program Design ...... 36

2.1 Introduction ...... 36 ix 2.2 Description of the IMF Lending Procedure ...... 39 2.3 IMF Program Negotiation ...... 41 2.4 The Model ...... 47 2.4.1 Conditionality ...... 48 2.4.2 The Players ...... 50 2.4.3 Sequence of the Game ...... 52 2.4.4 Payoﬀs ...... 55 2.4.5 Solutions ...... 65 2.5 Theoretical Implications ...... 83 2.5.1 Hypotheses ...... 85 2.6 Conclusion ...... 86

3 IMF Programs and Public Sector Reforms ...... 89

3.1 Introduction ...... 89 3.2 Public Sector Conditions ...... 91 3.3 Dataset of IMF Conditionality ...... 93 3.4 Hypotheses ...... 97 3.5 Empirical Analysis ...... 102 3.5.1 Dependent Variable ...... 102 3.5.2 Independent Variables ...... 105 3.6 Results and Discussions ...... 112 3.7 Conclusion ...... 128

4 IMF Programs and Fiscal Reforms ...... 131

4.1 Introduction ...... 131 4.2 Fiscal Reforms in IMF Programs ...... 135 4.3 Hypotheses ...... 137 4.4 Empirical Analysis ...... 142 4.4.1 Dependent Variable ...... 142 4.4.2 Independent Variables ...... 144 4.5 Results and Discussions ...... 150 4.6 Conclusion ...... 158

5 Financial Sector Reforms in IMF Programs ...... 162

5.1 Introduction ...... 162 5.2 Financial Sector Reforms in Borrowing Countries ...... 165 5.2.1 Financial Sector Conditions ...... 165

x 5.2.2 Financial Sectors in Borrowing Countries ...... 167 5.3 The Revised Model ...... 171 5.3.1 The Model ...... 171 5.3.2 Hypotheses ...... 174 5.4 Empirical Analysis ...... 175 5.4.1 Dependent Variable ...... 175 5.4.2 Independent Variables ...... 177 5.5 Results and Discussions ...... 180 5.6 Conclusion ...... 188

6 Conclusion ...... 189

6.1 Summary ...... 190 6.2 Contributions ...... 194 6.2.1 Theoretical Contribution ...... 194 6.2.2 Empirical Contribution ...... 195 6.3 Policy Implications ...... 198 6.4 Future Directions ...... 200

CHAPTER PAGE

A Proof for the Chapter 2 ...... 203

A.1 Payoffs ...... 203 A.1.1 Payoff for the IMF ...... 203 A.1.2 Payoff for the Domestic Interests ...... 204 A.1.3 Payoff for the government ...... 205 A.2 Solution ...... 208 A.2.1 Subgame below the government’s offer ...... 208

B IMF Programs 1994 — 2006 ...... 214

Bibliography ...... 228

xi LIST OF TABLES

TABLE PAGE

2.1 Parameters in the Model ...... 57

2.2 Equilibria in IMF Conditionality Design and Implementation Game . 76

3.1 Summary Statistics: Chapter 3 ...... 105

3.2 OLS Regression Model: the Natural Log of the Number of Public Sector Conditions ...... 113

3.3 Negative Binomial Model: Number of Public Sector Conditions . . . 119

3.4 Democracy and Autocracy Comparison: Public Sector Conditions . . 123

3.5 Heckman Selection Model: the Natural Log of the Number of Public Sector Conditions ...... 127

4.1 Summary Statistics: Chapter 4 ...... 145

4.2 Negative Binomial Model: Number of (Binding) Fiscal Sector Condi- tions ...... 151

4.3 Negative Binomial Model: Number of (Binding) Fiscal Sector Condi- tions (Democracy Only) ...... 159

5.1 Summary Statistics: Chapter 5 ...... 177

5.2 Negative Binomial Model: Number of (Binding) Financial Sector Conditions ...... 181

6.1 Summary of The Empirical Findings ...... 195

xii 6.2 Negative Binomial Model: Number of All (Binding) Conditions . . . 197

xiii LIST OF FIGURES

FIGURE PAGE

1.1 The Number of Conditions in IMF Programs between 1994-2006 . . . 15

2.1 IMF Negotiation Process ...... 47

2.2 Game Tree: The Main Model ...... 54

2.3 Theoretical Cases of Conditionality and the Size of Domestic Interests 70

2.4 Theoretically Predicted Outcomes in the Possible Cases ...... 73

2.5 Illustration: Reform-Minded Autocracy ...... 78

2.6 Illustration: Democracy ...... 80

3.1 Distribution of the Number of Public Sector Conditions ...... 103

3.2 Distribution of the Log of the Number of Public Sector Conditions . 104

3.3 Predicted Eﬀect of Regime Type on Public Sector Conditions . . . . 120

3.4 Predicted Eﬀect of Earned Vote % on Public Sector Conditions . . . 121

3.5 Predicted Eﬀect of Public Sector Size on Public Sector Conditions in Autocracies ...... 125

4.1 Distribution of the Number of Fiscal Sector Conditions ...... 143

4.2 Distribution of the Number of Binding Fiscal Sector Conditions . . . 144

4.3 Fiscal Freedom and Fiscal Sector Conditions ...... 146

xiv 4.4 Fiscal Freedom and Binding Fiscal Sector Conditions ...... 147

4.5 Government Expenditure and Fiscal Sector Conditions ...... 148

4.6 Government Expenditure and Binding Fiscal Sector Conditions . . . 149

4.7 Predicted Eﬀect of Regime Type on Fiscal Sector Conditions . . . . . 152

4.8 Predicted Eﬀect of Government Expenditure Score on Fiscal Sector Conditions ...... 154

4.9 Predicted Eﬀect of Loan Size on Fiscal Sector Conditions ...... 155

5.1 Game Tree: Financial Sector Conditions ...... 171

5.2 Distribution of the Number of Financial Sector Conditions ...... 175

5.3 Distribution of the Number of Binding Financial Sector Conditions . 176

5.4 Predicted Eﬀect of Loan Size on Financial Sector Conditions . . . . . 183

5.5 Predicted Eﬀect of Financial Freedom on Financial Sector Conditions 185

5.6 Predicted Eﬀect of Financial Sector Strength on Financial Sector Con- ditions ...... 186

5.7 Predicted Eﬀect of UN Voting with the U.S. on Financial Sector Con- ditions ...... 187

xv CHAPTER 1

INTRODUCTION

1.1 Introduction

Why do some IMF programs contain more policy conditions than others?

In any given year, the International Monetary Fund (IMF or Fund) lends to more than a dozen countries experiencing balance of payments problems.1 An IMF loan is typically provided under an arrangement or program between the IMF and the country in need. An IMF program includes speciﬁc policy measures that the country needs to commit to implement and program review and withdrawal schedules of the

Fund’s financial resources for an arranged program period. “IMF conditionality” refers to the specified policy measures in an IMF program. Conditionality is initially negotiated by a borrowing government and IMF staff, then later is approved by the

Executive Board of the IMF, the IMF’s decision making body. The policy measures are called conditionality or conditions, as continued access to the Fund’s ﬁnancial resources is granted conditional on implementation of them. Then what determines

1The average number of IMF programs signed in a year is around 21 between 1994 and 2006. For the 13 years of period, 275 new IMF programs had been arranged. Borrowing records of every member country are available at http://www.imf.org/external/np/ﬁn/tad/exﬁn1.aspx.

1 the conditions that a country has to promise to implement in order to receive an

IMF loan? Diﬀerently put, why do some IMF arrangements contain more extensive conditions than others?2

Existing studies in the two-level games literature provide a useful ground on which to build a theoretical account of how domestic politics aﬀects IMF program negotiations. In exploring the ways in which interactions between domestic and international actors aﬀect IMF program negotiations, studies in the two-level games literature generally suggest one of the following two competing theoretical arguments.

On the one hand, a government can tie its hands to an international negotiation partner, the IMF in the IMF program negotiation context, to force reforms on opposing domestic interests whose interests are to be compromised by the policy reforms. On the other hand, a government can tie its hands to domestic oppositions to extract a better deal from an international negotiation partner in general, or a more lenient conditionality from the IMF in the IMF negotiation context. Although the two-level games literature is rich in that it has many studies to support either of the two theoretical explanations, there is little theoretical argument over when and why one explanation may prevail over the other and vice versa.

In this dissertation project, I oﬀer a theoretical account of under what circumstances one of the two alternative arguments prevails over the other. Then I apply

2The extensiveness of conditions can be captured in various ways. The most common way is counting the number of conditions. Given the difficulty of comparing conditions in different country specific contexts, the number of conditions is one reasonable way to make comparison between programs (Dreher & Vaubel 2004, Dreher 2004b, Dreher & Jensen 2007, Gould 2006, Marchesi, Sabani & Dreher 2009). Alternatively, some studies use the scope of IMF conditionality, measuring the number of areas covered by an IMF program (Stone 2008, Marchesi, Sabani & Dreher 2009).

2 my theoretical argument to explain why some IMF programs contain more stringent conditions than others. In doing so, I ﬁrst construct a game theoretic model of IMF conditionality design and implementation in which I allow both of the two competing theoretical arguments can play out. The model captures the dynamic linkage between the deliberate negotiations over conditionality between the IMF and a government and domestic politics of conditionality implementation with emphasis given to the role of domestic interests.

The theoretical model shows that when conditions are domestically unpopular, the effect of domestic politics on IMF program design depends on the interaction of three institutional parameters: sensitivity to vote losses; reform-mindedness of the government; and the strength of affected special interest groups. Specifically, the model yields the proposition that a government that is more sensitive to vote losses and less reform-minded is more likely to extract a more lenient conditionality from the IMF. A government that is less sensitive to vote losses and more reform-minded is likely to pursue more extensive policy reforms by siding with the IMF and it is constrained by domestic politics only when there exist sufficiently strong domestic interests that can hinder proper implementation of agreed conditions. In contrast, when conditions are favored by affected domestic interests or affected domestic interests are too weak, the domestic politics exerts little influence over the design of an IMF program.

I test these propositions with an original dataset that I created by coding structural conditions included in all the letters of intent signed between 1994 and 2006.3

3I focus on structural conditions because 1) the number of quantitative conditions do not vary 3 Each structural condition is classified and counted by their types and affected economic sectors. I disaggregate IMF conditions into four affected economic sectors and analyze them separately. By doing so, I allow that factors relevant to designing IMF conditions over an economic sector vary by sectors. For instance, domestic politics may play a larger role in designing conditions over some economic sectors, while international politics behind the IMF and internal, organizational, and bureaucratic incentives of the IMF can play a greater role in designing conditions over other economic sectors.

The empirical analysis of public sector conditions in the third chapter supports the domestic politics hypotheses generated from the model and yields the following

findings. First, democratic countries that are more sensitive to vote losses receive fewer conditions than non-democratic countries that are freer from electoral competition, suggesting that the IMF is strategic in limiting how much it pushes vulnerable negotiating partners. The conclusion is bolstered by the second and the third findings. The second finding shows that within democracies, a government that won the previous election by a narrow margin has fewer conditions than a government that won the previous election with a wide margin. In addition, within democracies, a government that faces a more proximate election has fewer conditions than a government that expects a more distant next election. Finally, strong public sector interests reduce the number of conditions in autocratic countries but bear little in-

ﬂuence over the number of conditions in democratic countries. The overall ﬁndings

much across programs and 2) structural conditions are more politically salient as they pursue structural reforms.

4 suggest that democratic governments are more sensitive to popular pressures, and autocratic governments are more sensitive to strong domestic interests.

The fourth chapter examines what determines the number of fiscal sector conditions that affect the majority of ordinary citizens. I analyze the IMF conditionality dataset to examine the domestic politics hypotheses deduced from the theoretical model. The analysis of the fiscal conditions is very similar to the result from public sector conditions analysis and conforms to the domestic politics hypotheses highlighted. I show that more democratic countries are likely to have fewer fiscal sector conditions that are highly unpopular among the general public. I also show that the size of a loan is positively related to the number of fiscal sector conditions. I argue that this relationship is driven by domestic politics logic, because a government can introduce more conditions when it negotiates a larger loan to compensate domestic reform losers.

In investigating the design of ﬁnancial sector conditions in the ﬁfth chapter, I modify one important assumption in the main theoretical model presented in the second chapter, that an additional policy condition is politically unpopular. Instead,

I assume that there is little political cost involved in having more financial sector conditions. The modified assumption is justified for two reasons. First, the growing literature on the financial sector reform suggests that financial sector interests in developing countries actually favor more financial sector reforms toward financial sector liberalization. Second, in most borrowing countries, the financial sector is inchoate. Financial sectors in most developing countries are typically too weak to push their governments to extract favorable deals for them. The revised model for

5 the design of financial sector conditions yields predictions that are largely consistent with the existing studies of IMF program design: existing policies, size of a loan, and internal and international politics in and behind the IMF exert significant influence over the design of financial sector conditions, while domestic interests and institutions make little impact. Empirical investigation largely supports the predictions from the revised model. In particular, I find that there is little comparable influence of domestic special interests on the design of financial sector conditions. Instead, I show that the freer existing financial regulations in a borrowing country are, the smaller a loan is, and the closer a borrowing country votes with the United States in the

United Nations General Assembly, the fewer ﬁnancial sector conditions the program contains.

By providing a domestic political account of IMF conditionality negotiation and testing it empirically, the dissertation adds important insights to the rapidly growing

“politics of the IMF” literature. The main argument and ﬁndings of the dissertation suggest that the domestic politics of a borrowing country exerts considerable inﬂuence over the design of an IMF program.

The dissertation not only provides an alternative argument to the existing literature on the IMF, it also suggests alternative strategy to examine variation of IMF conditions. Specifically, by disaggregating IMF conditions by affected economic sectors and examine the influence of international and domestic politics on the design of IMF conditionality by each disaggregated sector, the dissertation shows that the influence of a factor is not homogenous across affected economic sectors but varies across sectors. For instance, the empirical chapters of the dissertation show that the

6 influence of domestic interests and institution is much stronger in designing public sector and fiscal policy conditions, while international politics plays a bigger role in designing financial sector conditions.

Theoretically, the argument generated from the game theoretic model provides a new insight to two-level games and the broader literature on interactions of domestic and international politics in international bargaining and international political economy. The dissertation provides an answer to under which circumstances and why one of the two competing theories of two-level games wins over the other. The model suggests that the choice between the two theories depends on characteristics of domestic institutions and conﬁguration of domestic economic interests.

There are also a number of policy implications. First of all, the dissertation provides an accurate model of how domestic politics in a borrowing country aﬀects the joint policy design between the IMF and a borrowing government. The dissertation shows that the design of an IMF program is not one-sided imposition of conditions as many observers claim but rather is jointly made by the IMF and a government.

Second, the dissertation indicates that the politics of IMF program design and the politics of IMF program implementation are interrelated; hence, the implementation of an IMF program is going to be inﬂuenced by the design of IMF conditionality.

This suggests that the evaluation of an IMF program and its subsequent effects on various social, economic, and political conditions should take the design of IMF conditionality seriously. One of the future directions for research suggested by the dissertation is to seek to better understand how the design of a program affects the implementation and its social, economic, and political effects.

7 The introduction chapter highlights the main research question, reviews existing literature, and provides a broad overview of the entire project. In the next section, I

ﬁrst illustrate the IMF lending process and conditionality in detail. I then highlight the empirical variation of IMF conditions and reemphasize the research question. In the following section, I review existing studies of IMF conditional lending and politics and two-level games literature. I summarize my argument before the concluding section in which I outline the dissertation and research strategy.

1.2 IMF Program Design as International Bargaining

1.2.1 IMF Program

The IMF lends to countries experiencing balance of payments problems, conditional on borrowing countries’ promises of reforms of economic policies and/or political institutions.4 In principle, policy conditions are required when a country needs to borrow more than 25 percent of its own IMF quota.5 While 25 percent is an arbitrary

4While the primary focus of the IMF is still on providing ﬁnancial assistance to countries with temporary balance of payment problems, the IMF has gradually extended its role into economic development. The IMF has recently engaged more in poverty reduction of highly indebted poor countries(HIPCs), which has conventionally been a domain of the World Bank lending. For instance, there have been an increasing number of poverty reduction programs through Poverty Reduction and Growth Facility(PRGF) than traditional stand-by agreements in recent years. During 2004-2006 period, about 1/3 of 37 new IMF programs are traditional Stand-by agreements while the rest of 2/3 are through PRGFs. See more details on IMF programs in the ensuring empirical chapters.

5“When a country joins the IMF, it is assigned an initial quota in the same range as the quotas of existing members that are broadly comparable in economic size and characteristics.” In October 2009, a new quota formula was adopted as a part of the IMF governance reform. “The newly agreed quota formula is a weighted average of GDP (weight of 50 percent), openness (30 percent), economic variability (15 percent), and international reserves (5 percent).”

8 cut, the reasoning is that if a country needs to borrow less than 25 percent of its own quota, the balance of payment disequilibrium is considered as “temporary;” thus, no corrective actions are considered to be required. However, if a country needs to borrow more than 25 percent of its own quota, the economic problem is deemed

“fundamental;” hence, the country is expected to install policy reforms to ﬁx unsound economic policies. In the latter case, a borrowing country and the IMF sign into an informal international agreement — an IMF program. The IMF program outlines required reform measures, commonly referred to as conditionality, and the size of the loan, along with other administrative procedures such as loan disbursement and program review schedules.

1.2.2 IMF Conditionality

IMF conditionality is at the heart of criticisms, controversies, and debates on the

IMF and its lending practices.6 According to the political left, IMF conditionality is a way for the IMF to serve selected interests of powerful sovereign stakeholders and heartless private global capital (Danaher 1994). The left point out that IMF programs advocating neoliberal economic ideology often do more harm than good to already struggling countries. They contend that austerity policies prescribed by

http://www.imf.org/external/np/exr/facts/quotas.htm. March 2010. The quota is important as it determines voting power of the country.

6Although it is widely used, the term conditionality is not precisely defined by The IMF. The IMF’s official website broadly describes conditionality as “borrowing government’s commitments on economic and financial policies and a requirement (in order to borrow from the IMF), when the country borrows from the IMF.”, IMF Conditionality by The IMF, “Factsheet — IMF Con- ditionality,” available at http://www.imf.org/external/np/exr/facts/conditio.htm, May 2008.

9 the IMF are counterproductive to economic recovery and IMF programs have little positive eﬀects on economic growth while stretching economic inequality even more by imposing costs of economic adjustment to the poor. For the political right, IMF programs are prime examples of activities of an ineﬃcient and extraneous international institution with overextended and idealistic policy agendas. According to the right, the IMF lost its “reason to exist” when the Bretton Woods exchange rate system collapsed, and the original purpose of the Fund disintegrated in the 1970s.

Since then, the IMF has morphed into a junior version of the World Bank, stepping into the unfamiliar territory of structural adjustment lending with structural policy conditions.7

Technically, the purpose of IMF conditionality is to ensure that scarce IMF resources are used in order to enhance the borrowing country’s ability to pay back the loan by helping the borrower adjust its own economic policies. In the long run, this would ensure continuing availability of IMF resources for other member countries.

The article of agreements of the IMF states:

The Fund shall adopt policies on the use of its general resources, includ-

ing policies on stand-by or similar arrangements, and may adopt special

policies for special balance of payments problems, that will assist mem-

bers to solve their balance of payments problems in a manner consistent

with the provisions of this Agreement and that will establish adequate

7For further discussion, see Killick (1995).

10 safeguards for the temporary use of the general resources of the Fund.

(article V. Section 3. (a). The International Monetary Fund (1945))

Formally, IMF conditionality is specified in a letter of intent and supplementary documents sent by a borrowing government.8 Practically, the documents are negotiated and jointly prepared by IMF staff and officials of the borrowing government, before being sent to the managing director of the IMF. Then the Letter of Intent and separate IMF staff reports are reviewed at an Executive Board meeting, where 24

Executive Directors make the final approval decision. When the Board approves the program, the country can use the first tranche of the approved loan amount. The subsequent tranches are subject to periodic reviews of conditionality implementation specified in the Letter of Intent and associated documents.

1.2.3 Negotiating IMF Conditionality

While the underlying purposes for IMF conditionality is to ﬁx economic policies by narrowing gaps between existing economic policies and the governance environment on the one hand and ideal economic policy settings that IMF economists advocate, designing IMF conditionality inevitably involves political bargaining between a government and the IMF. This is because the preferences of a borrowing government and the IMF rarely coincide with each other, as the government and the IMF serve diﬀerent constituencies. Moreover, the drafted agreement needs to be approved by

8These documents include a Letter of Intent(LOI) and/or Memorandum on Economic and Fi- nancial Policies (MEFP) that may be accompanied by a Technical Memorandum of Understand- ing(TMU) These documents will be prepared by the authorities of a borrowing country, with the cooperation and assistance of the Fund staﬀ (The International Monetary Fund 2007). 11 the Executive Board of the IMF and subsequently needs to be implemented in the borrowing country. Thus, designing IMF conditionality leaves considerable room for international and domestic politics to intervene at every stage — preparatory bargaining between the IMF and the government, approval by the Executive Board of the IMF, and implementation in the borrowing country.

The Fund itself seems to recognize the inﬂuence of politics. The most recently published “Guidelines on Conditionality,” adopted in 2002 and amended in Novem- ber 2006, acknowledges and pays due attention to domestic political and institutional circumstances. For instance, the Guidelines states that “the member has primary responsibility for the selection, design, and implementation of its economic and ﬁnan- cial policies ... The Fund will encourage members to seek to broaden and deepen the base of support for sound policies in order to enhance the likelihood of successful implementation (A. Principles 3. Ownership and capacity to implement programs. The

International Monetary Fund (2007)).” The Guidelines also emphasizes the domestic social and political circumstances of members and recommends its staﬀ be politically savvy. “In helping members to devise economic and ﬁnancial programs, the Fund will pay due regard to the domestic social and political objectives, the economic priorities, and the circumstances of members, including the causes of their balance of payments problems and their administrative capacity to implement reforms (A.

Principles 3. Ownership and capacity to implement programs. The International

Monetary Fund (2007)).”

12 1.2.4 Variation in IMF Conditionality

A quick empirical survey of IMF agreements reveals that there is much greater variation in IMF conditionality than is commonly perceived (Stone 2008, Dreher

& Vaubel 2004, Independent Evaluation Oﬃce 2007b). For example, Dreher and

Vaubel (2004) show that the number of conditions in the IMF agreements approved between April 1997 and February 2003 varies enormously. The mean number of conditions is approximately 22 and the median 18.5, though one country received only

ﬁve, while some others were asked to install more than 100 conditions. Even within the post-Soviet “transition countries” group, the variation is still considerable. One transition country had only six conditions, while a few others had 50 or more. Stone

(2008) draws a similar picture. He shows that the number of categories of conditions, measuring the scope of conditionality, also displays wide variety, ranging from 0 to

12 or more. Gould (2006) shows that bank-friendly conditions are often tailored in some IMF programs but not in others, especially when a country owes much debt to private banks. The IMF’s own “Monitoring Arrangements (MONA)” dataset also conﬁrms the overall trend. Reporting the number of structural conditions in 43 IMF programs, a recent publication by the Independent Evaluation Oﬃce of the IMF

(IEO), “Background Documents of and IEO Evaluation on Structural Conditional- ity in IMF-Supported Programs,” highlights the variation (Independent Evaluation

Oﬃce 2007b). An Analysis of data obtained from the IMF’s Policy Development and

Review Department including the MONA dataset, shows that there are signiﬁcant

13 variations in the sectoral composition, the types, and the number of IMF conditions.9

For example, among the 43 programs analyzed in the report, the programs of Cote d’Ivoire, Dominica, Estonia, The Gambia, Guatemala, Mauritania, and Mexico each contain fewer than 15 conditions with the least amount being eight conditions in the

Gambian program. In contrast, the programs of Albania, Armenia, Ghana, Lesotho,

Pakistan, and Romania each contain at least 60 conditions, with the greatest amount being 79 conditions in the Romanian program.

The original dataset of IMF conditionality that I created shows similar variation.

I constructed the dataset by coding 263 IMF Letters of Intents and Memoranda of

Economic and Financial Policies signed between 1994 and 2006. Many of the Let- ters, especially those signed in earlier years, were obtained from the IMF Archives in Washington D.C. The more recent Letters were downloaded from the oﬃcial IMF website. In analyzing the 263 programs signed between 1994 and 2006, the average number of structural conditions is 14, with a standard deviation of approximately

10.10 Figure 1.1 shows the histogram of the number of conditions in 263 IMF programs. The modal range is from eight to 12 conditions and is where more than

25 percent of IMF programs in the dataset fall. About 80 percent of all programs

9Database on Monitoring Arrangements (MONA) had been available only to IMF aﬃliated researchers until January 2009. In January 2009, MONA became available on www.imf.org as per a Board-endorsed recommendation arising from the IEO Evaluation of Structural Conditionality. But the available MONA spans relatively short period of time, from 2002 to the present. See http://www.imf.org/external/np/pdr/mona/MonaFAQ.aspx.

10Five programs out of all 275 programs signed between 1994 and 2006 do not have their Letters of Intents publicly available at the time of data collection. These programs are Mexico (1994), Guyana (2002), Guatemala (2003), Kenya (2003), and Cameroon (2005). In addition, Grenada, Sao Tome and Principe *2, Dominica *2, and Cape Verde*2 are not included in the dataset.

14 Figure 1.1: The Number of Conditions in IMF Programs between 1994-2006

15 contain 20 or fewer conditions, while the rest contain more than 20 structural conditions. Three programs contain more than 50 structural conditions, with the greatest number being 56, while more than 20 programs contain no structural conditions.11

In sum, the empirical survey of IMF conditionality reveals that the popular belief of uniform IMF programs is not close to reality. Furthermore, the signiﬁcant variation of IMF conditions raises a couple of interesting questions — how are IMF programs designed and why is there such a signiﬁcant variation among IMF arrangements?

1.3 Literature Review

Research on the IMF is burgeoning thanks to the growing availability of IMF program data and ensuing attention from academics and practitioners alike12 Instead of surveying IMF-related studies comprehensively, I focus on the more relevant empirical and theoretical studies in the following review. I divide this section into two parts: the political economy of IMF lending and theories of international bargaining.

After a brief discussion of the economic explanation based on technical aspects of the program design, I explore political accounts of IMF lending, with a number of supply and demand sides arguments. I then discuss the more theoretical literature of

11The diﬀerences between the number of conditions are due to the ways researchers code the number of conditions diﬀerently. For instance, I only collect data on the number of structural conditions in the initial Letters. In comparison, Dreher and Vaubel (2004) count the number of all conditions in initial and subsequent Letters separately by treating initial and subsequent Letters independently. I discuss the point in more detail in the third chapter.

12There are excellent comprehensive reviews on IMF research. See Vreeland(2007) and Steinwand and Stone(2008).

16 international bargaining and the two-level games literature in international political economy.

1.3.1 Economic Account

The oﬃcial intent of IMF conditionality is to ﬁx maladjusted economic policies. If one makes a reasonable assumption about the link between the degree of the economic crisis (as measured by the size of debt, foreign reserves, and balance of payments) and the degree of maladjustment of economic policies, then one can conjecture that as the economy fares worse, more conditions are needed. Using debt service, foreign reserves, and more severe balance of payments as proxies of policy maladjustment, recent empirical studies of IMF conditionality provide support for this argument.

For instance, analyzing conditions included in 206 letters of intent with 38 countries,

Dreher and Vaubel (2004) ﬁnd that a high rate of monetary expansion leads to more conditions while having a more advanced economy with a higher real GDP per capita reduces the number of conditions. Similarly, Dreher, Sturm and Vreeland (2009) ﬁnd that high foreign reserves reduce the number of conditions included in a program.

More recent empirical studies of IMF programs try to capture the degree of policy maladjustment more directly. These studies use indicators that directly measure how far oﬀ existing economic policies are from the policy standards that the IMF advocates. If the distance is wider between existing economic policies in a country and the policy standards that the IMF advocates as necessary to recover the country’s economy, most notably policies congruent with neoliberalism, the number of

17 conditions should increase. The Independent Evaluation Oﬃce of the IMF, the independent body that oversees various functions of the IMF, took this approach when it reviewed structural conditionality in IMF-supported programs in 2007. While the empirical analysis is preliminary and based on correlations between various indicators, the overall trend suggests that the above expectation is consistent with the data.

An analysis of the sector-wise disaggregated conditions included in the IMF’s MONA dataset, shows that as the distance between country’s existing economic policies and the ideal neoliberal economic policies becomes greater, the number of conditions increases. For instance, more financial sector conditions are assigned in a program when the index of financial liberalization in a country is lower. Similarly, when trade is less open, the number of trade-related structural conditions tends to increase, and as government intervention in the economy increases, it also increases privatization- related conditions (Independent Evaluation Office 2007b). Dreher (2004b) examines this argument in a more systematic way. He finds that IMF does not demand as many conditions for countries with more economic freedom.

It is worthwhile to note that the measure of economic policies should serve as the benchmark for deciding conditionality. If the IMF follows its mandate faithfully, it should be the case that a large portion of the variation in IMF conditionality would be explained by the gap between the existing economic policies and ideal liberal policies. If there is a reasonable indicator of how liberal a country’s economic policies are, direct measure of policy maladjustment should be preferred to indirectly measuring policy maladjustment with economic indicators. This is because using

18 a direct measure would likely reduce the noise introduced when one uses indirect proxies.

1.3.2 Political Account

There are two sets of actors directly involved in IMF program negotiations and the life cycle of a program. On one side, decisions are made by “suppliers” of ﬁnancial resources — the IMF and actors behind them, consisting of sovereign principals of the

IMF, including the U.S. and G-7, their domestic constituents, and private ﬁnancial institutions. Researchers focus on what inﬂuences the IMF’s decision to lend and with what conditions attached. Answers to these questions are reviewed under the

“supply” side story. On the other side, complementary decisions are rendered by

“demanders” of ﬁnancial resources or borrowers. Here, researchers focus on why or why not governments sign IMF programs. Existing answers to these questions are reviewed under the “demand” side story.

Supply Story

IMF Bureaucratic Interests Approaching from the public choice tradition, some scholars argue that bureaucratic interests of the IMF staff influence the IMF’s lending activities. Roland Vaubel argues that “international bureaucrats try to maximize their power in terms of budget size, staff and freedom of discretion and appreciate some leisure on the job (Vaubel 1986, p.52).” Empirically, Vaubel (1996) demonstrates that there is a tendency toward “hurry-up lending” as the next quota review approaches. Since the IMF’s budget comes primarily from managing its financial

19 resources contributed by member countries (IMF quotas), every ﬁve years when member countries review the quotas, the IMF hurries up its lending to showcase the need to increase the IMF’s pool of resourses. IMF historian James Boughton agrees that one of the main goals of the IMF as an organization is to maintain its resources.

He states, “the main challenge for the future is safeguarding the IMF’s identity and its resources, so that it can continue to provide adequate support to its now universal membership (IMF Survey 1994).”13 While there is no explicit argument regarding

IMF conditions from this view, one can extend the argument and come up with a reasonable hypothesis. Since the IMF hurries up its lending before a regular budget review, the IMF is more likely to engage in active lending and become more ﬂexible and accommodative when the next quota review is near. Speciﬁcally, the IMF would be willing to lend with fewer conditions when the quota review is scheduled soon.

Conversely, when the IMF has completed its regular quota review, the IMF might become tougher, demanding more conditions, since it has secured the quota for the next ﬁve years. Thus, we may see cyclical ebbs and ﬂows of the number of IMF programs and their conditions.

Sociology of the IMF Organizational sociologist Sarah Babb also emphasizes the influence of the IMF and its staff over the IMF’s lending. From a sociological orientation, she focuses on dominant ideas and interests of the IMF staff, rather than materialistic bureaucratic incentives (Babb 2003). As organizations often have am- biguous mandates, she argues, it is largely up to the professionals working in those organizations how to interpret and apply those mandates to cases. Since the IMF is

13This is cited in (Vreeland 2007). 20 populated by economists whose thinking is mostly congruent with the “Washington

Consensus” since 1990s, “it is natural for the IMF to simultaneously endorse the goals of low inflation and economic growth through government downsizing (Babb 2003, p.20).” Similarly, Barnett and Finnemore (2004) argue that the evolution of conditionality is driven by the development of knowledge expertise of IMF staff and the knowledge expertise has contributed to the expansion of the number and breadth of IMF conditions. The theory implies that IMF conditionality would reflect the shared contemporary economic ideology among IMF economists. When the neoliberal Washington consensus dominates economists’ thinking, the conditionality will also be designed based on the Washington consensus. In contrast, when Keynesian economy was more in mainstream, conditions with a greater emphasis on government’s role in the market would prevail. Chwieroth (2007b, 2007a) shows that proportion of neoclassical economists in IMF senior staff positions is positively correlated with capital account liberalization.

Principals of the IMF More political explanations come from principal-agent framework. In this line of research, scholars emphasize the inﬂuence of the principals of the

IMF — the major shareholders of the IMF who command a larger share of votes in the Executive Board of the IMF, most notably the United States and the rest of G-7 countries. The inﬂuence of the major shareholders is possible given the rules of decision making and the appointment of high-ranking oﬃcials of the IMF. 14 Numerous

14Vote share is proportional to a country’ quota, which in turn is largely determined by the size of the country’s economy. Thus, the U.S. controls about 17 percent of votes and it gives the U.S. a veto power in making important decisions, since 85 percent supermajority is required to make such decisions. In addition, the Managing Director of the IMF is appointed by the European

21 anecdotes exist ranging from Cold War-era preferential IMF lending to West-allied countries to War on Terrorism era lending to American strategical partners such as

Turkey and Pakistan. For instance, in his Washington Post article, Paul Blustein, an IMF expert, wrote on American favoritism toward Pakistan in the wake of the terrorist attack on September 11, 2001: 15

[A]ccording to a senior IMF staﬀer, the expectation among some of the

people dealing with Pakistan before Sept. 11 was that the chances for

board approval were 50-50, because although the Pakistani government

had kept many of its pledges, it had failed to deliver on others, including

tax collection. After Sept. 11, the political atmosphere was transformed

— so at Wednesday’s meeting, board members representing the IMF’s

member countries heaped praise on Islamabad’s economic performance.

(October 4, 2001, Washington Post, “A Tighter Hand in Doling Out

Global Aid?”)

There are also a number of studies backing up these newspaper articles and popular speculations. For instance, Momani (2004a, 2004b) ﬁnds that the U.S.’s geopolitical interest intervened in IMF decision making procedures. She argues that

“two IMF-Egyptian agreements were facilitated by the United States in 1987 and

governments while the World Bank President is appointed by the American government by convention.

15Blustein has written excellent accounts of global financial crises based on detailed interviews with policy makers from borrowing countries and G-7 governments, private bankers, and IMF staff members. On the Asian financial crisis, see Blustein (2001). On Argentina’s long lasting relations with the IMF and its latest financial crisis in 2001, see Blustein (2005). 22 1991, in order to protect the Egyptian regime from tough conditions and to reward the regime for its participation in the Persian Gulf War (Momani 2004a, p.881).” In the first systematic large-N study of the influence of the U.S. on the IMF, Thacker

(1999) finds that countries moving their policy positions favorably toward the U.S. foreign policy position indeed are more likely to get loans from the IMF than those moving away from the U.S.’s foreign policy position. In a more recent study, Dreher and Jensen (2007) report that the U.S.’s influence is present in IMF conditionality design. In their statistical analysis, economic variables do not explain the variation in the number of IMF conditions very well; instead, they find that closer allies of the

U.S. receive IMF loans with fewer conditions. With a diﬀerent measurement scheme,

Stone (2008) reports similar findings. Specifically, he finds that “countries receiving more U.S. foreign aid are subject to dramatically reduced degrees of conditionality,” but “only when the borrower has a pressing need for IMF support (Stone 2008, p.22).”

Others further elaborate on this line of inquiry. Recent studies broaden the scope of sovereign principals of the IMF and diversify the ways in which sovereign principals channel their preference. For instance, Dreher, Sturm and Vreeland (2009) ﬁnd that countries signing IMF arrangements during their United Nations Security Council tenure received nearly 20 percent fewer conditions. They argue that this is because

IMF loans serve as a mechanism by which the major shareholders of the Fund can win favor with voting members of the Security Council. Oatley and Yackee (2004) examine the eﬀect of the amount of U.S. bank exposure in borrowing countries on the amount of loans, and Broz and Hawes investigate the eﬀect of the total amount of

23 U.S. lending as a proportion of a developing country’s GDP on the lending practice of the IMF, emphasizing ﬁnancial interests in the U.S. (Broz & Hawes 2006a, Broz

& Hawes 2006b, Broz 2008). These studies generally agree that more favorable

IMF programs, those granting loans worth greater amounts of money with fewer conditions, are granted when American ﬁnancial and foreign policy interests at stake are deemed signiﬁcant.

Moving from state to non-state actors, Gould points out that an IMF loan is only a small fraction of the funds that a borrowing country needs at a time of ﬁnancial crisis.

Gould (2003, 2006) examines the influence of other suppliers of international financial resources — the supplementary financiers. Her argument is two-fold. Statically, the composition of the debt is a significant predictor of IMF conditionality. When supplementary financing comes from private financial institutions, the IMF program is more likely to include private bank-friendly conditions — for example, setting aside certain fiscal revenues to match international loans with fiscal revenues, using a certain percentage of the IMF loan for debt reduction payments or replenishment of foreign reserves, and making debt service payments as agreed with commercial bankers. Gould’s argument also has dynamic implications. As the primary creditors historically have evolved from sovereign states to private financial institutions, the bank-friendly conditions are more likely to be present in recent IMF programs than older IMF programs.

Gould illustrates the inﬂuence of a private ﬁnancial institution with a stark example:

As SBA (Stand-By Agreement) with Ghana in 1983 stipulated that the 24 IMF loan to Ghana be deposited directly in a Bank of Ghana account

held at the Bank of England, and that the Bank of England was to

transfer the deposit directly to the Standard Chartered Bank to repay

a short term loan it had made to Ghana. So the IMF loan never even

reached Ghana, but rather went directly to repay a commercial bank.

(Gould 2006, p.156)

Overall, the literature emphasizes actors in the chains of principal-agent relations in the command of the IMF: IMF staﬀ, principals of the IMF, such as the Executive

Directors and their governments, and principals of governments, such as voters and private financial institutions. Copelovitch (2010) further synthesizes the “supply” side story and argues that preference heterogeneity among G-5 governments provides scope conditions under which the G-5 countries exert greater influence, while under others, IMF staff have autonomous influence independent from G-5 preferences. In sum, the important factors in approving and designing IMF programs reflect the following interests: bureaucratic and ideational interests of IMF staff members; interests of the IMF’s major quota holders, most notably that of the U.S.; domestic politics of the major shareholders of the IMF; and private financial institutions who supplement IMF loans.

Demand Story

While the supply side of the IMF lending process has been under increasing attention, the demand side of the story has attracted far less attention. This may be because of the popular view of “inevitable” IMF participation and “imposed” IMF

25 conditionality. That is, more often than not, IMF participation by a sovereign county is depicted as an unavoidable, forced choice and included IMF conditionality is also seen as assigned or imposed and, unilaterally designed by IMF staﬀ.

While the suppositions may be accurate in some cases, they neglect the point that sovereign countries can choose not to participate in IMF programs. For instance, even in the middle of the Asian ﬁnancial crisis in 1997, Malaysia opted not to sign an agreement. Instead, Malaysia unilaterally imposed stricter, short-term capital control (Sundaram 2006). Similarly, India and South Africa did not participate in

IMF programs at times of looming ﬁnancial crises. Vreeland also reports a handful of countries that had extremely low hard currency reserves — average reserves as low as or less than 0.1 times the monthly imports — did not participate in IMF programs (Vreeland 2003, p.23). Kahler (1993) shows that even heavily indebted, the least developed countries often refuse to participate in IMF programs.

Moreover, some politically important countries are reported to exert strong bargaining leverage in IMF negotiation. For instance, when Russia was under a ﬁnancial crisis, it was said to have greater-than-deserved bargaining leverage, because it was too nuclear and too economically big to fail (Aslund 1999). After a series of interviews with IMF oﬃcials and brief case studies, Stiles also concludes that bargaining and compromise are the central dynamic of IMF policy making, rather than coercion

(Stiles 1987, Stiles 1990). Dreher and Jensen (2007) also acknowledge:

After all, the number and stringency of conditions are the outcome of

a bargaining process, and the Fund, eager to lend, is probably prepared

to endorse fewer conditions if it feels that this is necessary to reach an 26 agreement ... IMF participation is a joint decision between the Fund and

the borrower. (2007, p.14)

There are very few systematic studies that investigate how borrowing countries can affect the design of IMF programs, even though they certainly can in fact affect them. To study the influence of a borrowing country, one needs to fully reconsider the process of IMF program design as an interdependent decision-making process by the IMF and the borrowing country. If the government does not want to sign on to the IMF program for any reason, the government may not sign the agreement. This naturally suggests the need for studying the other side of the negotiation table — the borrowing country. Taking the politics of the borrowing country seriously involves not only accounting for political dynamics in the program negotiation stage but also accounting for domestic politics of implementation (Fearon 1998).

Lack of attention to the demand side of the story results in biased inferences based solely on the “why the IMF lends” story. It risks falsely attributing causal weight of excluded demand side factors to included supply side variables, unless there is little to gain by adding domestic dynamics of borrowers. Yet, there are enough reasons to believe that domestic dynamics do matter in IMF program negotiations.

Vreeland provides a notable exception to the trend of general inattention to domestic politics of the IMF (Vreeland 2003, Vreeland N.d.a, Vreeland N.d.b). In his ground-breaking study of domestic politics of IMF programs, Vreeland (2003) argues that the government may bring in the IMF to gain leverage against domestic opposition. He reasons that bringing the IMF in is useful for the government, because it increases the rejection cost for domestic opposition, and hence deters rejection 27 of the IMF program by domestic opposition. He contends that the rejection cost for the domestic opposition gets higher after the government successfully brings in the IMF and signs an agreement, as opposing reforms now involves not only rejecting the government’s reform initiatives but also the international commitment made with the IMF. Domestic opposition is more likely to acquiesce, since defecting from the international commitment by rejecting the IMF program may send a bad signal to foreign creditors and investors in times of economic crisis. Thus, governments with high “political will” to engage in economic reform have an incentive beside the

IMF’s ﬁnancial resources to participate in IMF programs — conditionality. In this case, IMF conditionality becomes politically desirable for the government. It allows the government to tie its hands with IMF conditionality against potential domestic opposition.

1.3.3 Two-Level Games Literature

Theoretically, the dissertation borrows insights from and builds on studies concerning interactions of international and domestic politics. There are certainly no shortage of studies in this body of literature.16 Since international economic negotiations, whether they are for free trade agreements or loan arrangements with international

ﬁnancial institutions, always bring domestic welfare and distributional consequences, scholars have sought to understand how international bargaining interacts with domestic political interests and institutions. Broadly grouped as “two-level games”

16There are a number of excellent volumes of studies in the literature. See the seminal article by Putnam (1988), the original volume on two-level games by Evans et al. (1993), a more recent studies by Milner (1997), Milner and Rosendorﬀ (1997), and by Drezner (2003). 28 literature, a number of studies have advanced our understanding of this complex phenomenon.

Yet, when it comes to applying the insights generated from the literature to a specific case, one would find two logically established and empirically supported explanations competing with each other. While equally plausible, they generate opposite predictions regarding international negotiation outcomes. Therefore, two- level games as a theory is underspecified: it provides a useful theoretical framework, but does not provides an unequivocal prediction in a specific empirical context.

On the one hand, scholars contend that governments can utilize domestic opposition to extract better deals in international bargaining (Schelling 1960). For instance,

Martin (2000) argues that the U.S. can make its bargaining proposal more credible as it has a stronger and more independent legislature. As the American Congress con- strains the President, the President can calim to his international counterpart that his hands are tied and extract a better deal for the U.S. Cowhey (1993) compares

Japanese and U.S. cases and concludes that stronger domestic checks and balances in the U.S. allowed the U.S. government extract a better deal than its Japanese counterpart. Kahler’s study of IMF stabilization programs in Jamaica and Somalia demonstrates that domestic constraints on least developed countries’ governments, over which the IMF had little control, undermined the IMF’s seemingly dominant position (Kahler 1993). In his original contribution, Putnam highlights the mechanism in the IMF context:

The actors at Level II (domestic) may represent bureaucratic agencies,

interest groups, social classes, or even “public opinion.” For example, 29 if labor unions in a debtor country withhold necessary cooperation from

an austerity program that the government has negotiated with the IMF,

Level II ratiﬁcation of the agreement may be said to have failed; ex ante

expectations about that prospect will surely inﬂuence the Level I (interna-

tional) negotiations between the government and the IMF. (Putnam 1988,

p.439)

On the other hand, there are a group of scholars who focus on how a government can use an external inﬂuence to tip the balance against domestic opposition.

Vreeland (2003) argues that governments with political will to reform would sign an

IMF agreement to tip the balance against domestic opposition. This would work, as the opposition’s rejection to the reform measures endorsed by the IMF would send a bad signal to investors; hence, bringing in the IMF would raise the cost of rejection for the opposition. Similarly, Mayer and Mourmouras (2008) contend that

“under ideal conditions, International Financial Institutions (the IMF and World

Bank) assistance can help tip the balance of the authorities’ political calculations in the direction of desiring to reduce policy distortions (which are currently in favor of domestic special interest groups).” Vaubel (1986) also argues that the IMF relieves countries of unpleasant tasks by serving as a scapegoat and imposing policy conditions on borrowing governments that want to evade the responsibility of unpleasant measures. More speciﬁc to conditionality, Drazen (2002) contends that conditionality is necessary if there are heterogeneous interests and conﬂicts within a borrowing country and claims that in these cases “conditionality can strengthen the hands of

30 the reformers who are committed to carrying out reform but face domestic opposition[p.43].” The mechanism has been applied to other topics in international political economy. For instance, Goldstein (1996) contends that international commitments provide solutions for the government to overcome diﬃcult domestic problems. In the

Canadian-U.S. Free Trade Agreement context, she shows that “the FTA (between

Canada and the US) and NAFTA reduced the autonomy (of trade bureaucracy in the

U.S.) and were therefore preferred by the President even though it reduced the ability of the United States to retaliate to unfair trade competition abroad (Goldstein 1996, p.556-557).” In a broader International Relations context, this line of argument is also well supported in democratization literature. For instance, external forces, such as membership in international organizations, increase the probability of successful democratization against potential domestic opponents (Pevehouse 2003).

1.4 Criticisms and Preview of the Argument

While there are increasing studies on the IMF, and they do provide important ﬁnd- ings, there also exist several limitations as detailed above. First, as most IMF-related studies focus on participation and eﬀects of IMF programs, the important question of

IMF program design has attracted very little attention (Drazen 2002). As the design question is closely related to participation, effects, and evaluations of IMF programs, inattention to program design is not a minor problem. Second, the existing studies do not draw linkages between design, implementation, and effects. “But program success depends on successful implementation, which in turn reflects the political constraints, raising the question of the extent to which program design should take

31 these constraints into account (Drazen 2002, p.37).” Third, as far as IMF program design is addressed, most focus on principals of the IMF without paying attention to the other side. As IMF program design is a joint decision, make by the IMF and the government, this inattention may bias our understanding of causal eﬀects of other factors. Fourth, there are few studies empirically investigating the contents of

IMF programs. This might be due to the unavailability of IMF conditionality data until recently. The dissertation seeks to address these limitations by focusing on the design of IMF programs, paying due attention to domestic politics of IMF programs, and empirically testing hypotheses with an original data set of IMF conditionality.

Theoretically, the two-level games literature provides a useful starting point to conceptualize the process of designing IMF programs. Yet when it comes to applying two-level games insights, they are indeterminate, as they provide multiple competing predictions. Given multiple plausible explanations, existing studies of the two-level games do not provide guidance to which of the competing theoretical arguments will drive the outcome of international economic bargaining in general and the design of an IMF program in the IMF lending context. One theoretical argument suggests that when you have stronger domestic opposition, a government and the IMF may want to have stronger conditions included in its IMF program to force reforms on the domestic opposition, as weaker conditions would not be enough to tip the balance against domestic opposition. The other suggests that when there is strong domestic opposition, a government would utilize strong domestic opposition to extract a lenient deal from the IMF. Thus, while the two arguments sound equally plausible, they generate opposition predictions. So when is one argument more relevant than the

32 other? Is it totally at the whim of a government to either use international inﬂuence to tip the balance against domestic opposition or use domestic opposition to extract a better deal from an international counterpart? Assuming that a government is itself a strategic player in maximizing its own utility, there must be some systematic conditions under which one argument is more relevant than the other.

In the next chapter, I construct a game theoretic model built on the basic structure of the two-level games to study the interactions between the IMF, government, and domestic interests. I allow both of the competing theories of two-level games to work and show under what circumstances one theoretical prediction prevails over the other. The theoretical model suggests that the eﬀect of domestic politics on

IMF program design depends on the interaction of three institutional parameters: sensitivity to vote losses, reform-mindedness of the government, and the strength of aﬀected domestic interests. Speciﬁcally, the model yields the proposition that a government that is more sensitive to vote losses and less reform-minded is more likely to extract a more lenient conditionality from the IMF. This is because a government that is sensitive to vote losses is forced to minimize the electorally costly reform measures. A government that is less sensitive to vote losses and more reform-minded is likely to pursue more extensive reforms, siding with the IMF, and is constrained by domestic politics only when there exist strong domestic interests that can hinder proper implementation of agreed upon conditions. For a government that is immune to electoral costs, implementation probability is the only important factor when it considers how extensive a set of reforms it can pursue. Thus, the government is only

33 constrained by strong domestic interests that can resist proper implementation of

IMF reform conditions.

1.5 Conclusion and Plan of the Dissertation

This chapter introduces the research question in both theoretical and empirical contexts and then briefly reviews the relevant bodies of literature. While existing studies on two-level games and IMF participation and effects certainly deepen our understanding of IMF and international bargaining, there are also several limitations in the literature. Two-level games literature gives valuable insights and plausible arguments regarding international bargaining, but it proves indeterminate when applied to IMF program design specifically or outcomes of international economic bargaining more generally. Existing studies of IMF lending pay little attention to the design of IMF programs, which can have major consequences. When they do, they do not recognize the designing as a bargaining process between the IMF and a borrowing government. It generally discounts the role of a borrowing government and domestic politics in a borrowing country.

The second chapter introduces the comprehensive model of international economic bargaining in the context of IMF program design and implementation. It explores strategic interactions between the IMF, government, and domestic interests in IMF program design. The model makes explicit linkages between the design and implementation stages by using domestic political considerations. By doing so, it generates a number of testable hypotheses regarding domestic and international political factors inﬂuencing design and implementation of IMF programs.

34 The remaining three chapters provide rigorous empirical analysis of theoretically generated hypotheses from the second chapter. Specifically, the third chapter examines how domestic political institutions and strength of domestic interests, the public sector, affect the number of public sector conditions in IMF programs. The fourth chapter examines fiscal conditions, covering highly unpopular austerity measures targeting tax reforms and spending cuts of government expenditure. The fifth chapter modifies the main theoretical model, then investigates determinants of financial sector conditions. The dissertation will conclude with a final chapter reviewing the findings of the study and presenting future research ideas.

35 CHAPTER 2

A THEORY OF IMF PROGRAM DESIGN

2.1 Introduction

What explains the variation in IMF conditionality? Why do some IMF programs contain numerous policy conditions while others have only a few of them? Extant political studies of IMF programs tend to focus on either organizational and bureaucratic incentives of the IMF and its staff or influences of powerful state and non-state actors such as the United States, G-7 countries, and private financial institutions.

While the recent studies of IMF programs have greatly advanced our understanding of the IMF lending process, they tend to focus on how decisions are made in and behind the IMF without paying comparable attention to the inﬂuence of domestic politics in a borrowing country. I claim that the inattention to the domestic political economy of a borrowing country is problematic as it can lead to biased inferences solely based on factors aﬀecting the IMF’s decision making.

There are many ways that domestic politics can shape the design of IMF conditionality. First, the program design is negotiated and ﬁnalized jointly by the IMF and a borrowing government and the decision making of the borrowing government is

36 often constrained by its domestic political circumstances. Second, domestic politics of reform policy implementation should be a part of the IMF’s and the government’s calculation, if they care not only about just the signing of an IMF program but also the implementation of the agreed upon program, which would substantiate economic reforms. To aid in a more comprehensive understanding of the design of IMF conditionality, I provide a domestic political account of the IMF program design in this chapter. Attempting to capture the ways in which domestic political institutions and interests of a borrowing country shape IMF conditions, I present a game theoretic model of IMF program design and implementation.

The model of IMF program design and implementation makes an explicit linkage between the bargaining or designing stage where the IMF and a government negotiate over policy conditions and the implementation stage where the government and reform-targeted economic interests interact to implement or renege on the agreed upon reform measures. With the explicit linkage between international bargaining over IMF conditions and domestic politics of implementation, there are two alternative domestic political theoretical conjectures that can play out in IMF program design: the government can either tie its hands to the IMF to force reforms on domestic interests or tie its hands to its constituents to extract a more lenient deal from the IMF. I show that the choice between the two theoretical arguments hinges on the interaction of three institutional parameters in the model: democracy and electoral competition, reform mindedness of the government, and the strength of targeted domestic interests. Speciﬁcally, the model yields the proposition that those governments that are sensitive to vote losses are pressured by domestic interests to extract

37 more lenient deals from the IMF and to stay with the minimum conditions acceptable to the IMF, regardless of their own preferences over reforms, since additional reform measures are electorally costly. In comparison, those governments that are less sensitive to vote losses can pursue their own ideological orientation (reform-mindedness) freely. These countries with few formal institutional constraints are only constrained by strong domestic interests who can intervene and disrupt the implementation of agreed upon policy reform measures. These propositions are empirically examined in the subsequent three chapters.

The main body of this chapter begins with a brief description of the IMF lending procedure. The discussion of directly relevant existing studies follows. The model of

IMF program design and implementation builds on these existing studies. The next section introduces the model of IMF program design and implementation. Actors and their payoffs and the sequence of the game are described in detail. In describing the solution to the game, only main theoretical implications and intuitions behind the implications are presented, with further technical details and proofs being re- served for the appendix A. The model yields a few clear predictions with regard to the domestic political factors influencing the design of IMF conditionality. First, governments that are more sensitive to vote losses are more likely to have less severe conditions than governments that are immune to such electoral pressures. Second, the size of the loan increases the severity of IMF conditionality. Third, more reform- minded governments are more likely to have more reform conditions. Fourth, the size of affected domestic interests reduces the severity of IMF conditions targeting the economic interests in countries where little electoral pressure exists. However, the

38 size of aﬀected domestic interests does not aﬀect the severity in IMF conditionality in countries where governments already accommodate the demand of the economic interests due to the electoral consequences of reforms. In the countries where there exist strong electoral pressure, the ability of domestic interests to resist implementation of agreed upon reform measures is essentially redundant. The concluding section recaps the model and discusses how to empirically test the predictions generated from the theoretical model.

2.2 Description of the IMF Lending Procedure

The IMF lending process begins when a country makes a request for IMF ﬁnancial assistance to the IMF and ends with the approval decision by the IMF’s Executive

Board. A country commonly makes a request when it faces a balance of payments problem with few remaining alternative financing sources.1 Once the request is made by the country, IMF staff and government officials of the requesting country start to negotiate and formulate an IMF program.2 Most importantly, IMF staff and the government negotiate over the specific policy reform measures to be included in the program. Once in agreement, the borrowing country has to implement the policy

1The IMF is often referred as the lender of last resort.

2The description of the IMF emphasizes its supporting role to provide technical expertise to the government who formulates an IMF program. In contrast, experts outside of the IMF consider the program formulation being unilaterally imposed by the IMF (e.g. Stiglitz (2002)). The IMF oﬃcial website describes the program design process as follows: “the economic program underlying an arrangement is formulated by the country in consultation with the IMF.” http://www.imf.org/external/np/exr/facts/howlend.htm.

39 reform measures in order to continue receiving scheduled tranches of the arranged loan.

The process of IMF program design is not always a quick and smooth process and can last from a few weeks to more than a few months. This is because the IMF and the government serve different constituents and thus have different preferences over policy reform measures. In particular, the government is responsible for contents and consequences of policy reforms while the IMF is relatively free from such account- ability. Instead, the IMF needs to satisfy those who control its financial resources.

Thus, the IMF and the government often disagree over what reform measures are necessary to bring the balance of payment disequilibrium back to an equilibrium and how extensive policy reforms should be. If and when a draft agreement is finally reached by IMF staff and government officials, a “Letter of Intent” stipulating the agreed upon policy reform measures is presented to the Executive Board of the IMF.

Provided that the program is approved by the Executive Board, the country can draw the loan in several phased installments or tranches. Continued access to the loan is conditional on implementation of reform measures stipulated in the IMF program. In a given year, the IMF typically approves about 21 new IMF programs, and the conditions included in each program vary signiﬁcantly across IMF programs.3

3Between 1994 and 2006, 275 new programs were signed by the IMF and member countries. Thus, on average, there are about 21 new IMF programs each year. The number of programs in a given year closely follow how the global economic condition ﬂuctuates — since the Asian Financial Crisis, there had been a steep downward trend of the number of programs signed each year to the point where experts start talking obsolescence of the IMF only to witness the surge of the number of programs and the eﬀort of G-7 to give a new life to the IMF amid the global economic downturn since 2008 (Woods 2010).

40 2.3 IMF Program Negotiation

As in any international economic negotiation, the domestic political economy in IMF program bargaining can play a critical role in shaping the ﬁnal negotiation outcome.

For one, the domestic political economy exerts significant influence in international economic negotiations because a typical international economic agreement produces winners and losers due to economic policy changes stipulated in the agreement. For instance, a free trade agreement between countries usually benefits consumers and export-oriented industries while it hurts import-competing industries. In addition, the domestic politics plays a critical role in the implementation of the agreement and the agreement remains unsubstantiated until it is domestically implemented. Signing a free trade agreement brings few consequences until the agreement is ratified and subsequent policy changes are made. The process of domestic ratification is not an automatic follow-on of the signing of the agreement, and often political battles often arise over the ratification of the agreement.

Similar to its trade counterpart, an IMF program has distributional consequences, because costs and benefits of an IMF program are distributed unevenly. On the one hand, benefits of the program, especially the benefits from the at least temporary stabilization of the economy, tend to be more evenly distributed. The government can replenish the foreign reserve account, restore confidence of foreign creditors, and bail out indebted companies and financial institutions rather than letting them go bankrupt. This will prevent the unemployment rate from rising higher, consumer prices from rising, and interest rates from going up, for instance. On the other hand, costs of the program are concentrated in certain parts of the economy. Reform

41 measures included in the program speciﬁcally target certain sectors of the economy, and the conditions are generally devised to reduce the rents that the targeted sectors have enjoyed previously. Thus, while most constituents enjoy the beneﬁts from the loan, some do so without bearing the costs that others do.

For the government, the distributional consequences of the IMF program are very important, as they can decide the political survival of the government in the next election or in other leadership decision mechanisms. In many countries, members of adversely affected economic sectors are also voters who have the right to elect their head of government in the coming election. So at the very least, the incumbent government needs to consider potential vote losses due to extensive reform conditions that it agrees to implement. In addition, especially when domestic interests are well organized, they can publicly endorse a candidate or finance the election campaign of a candidate of their choice. The overall influence of affected sectors on an electoral outcome would be especially salient in democracies, where voters (directly in presidential democracies or indirectly in parliamentary democracies) elect the head of government. Anticipating political consequences of IMF conditionality, the government may have to accommodate and represent some of the needs of affected domestic interests when it negotiates conditions with the IMF. As the electoral pressure varies across domestic political institutions and depends on the configuration of domestic interests, the effort to accommodate the needs of affected economic interests will vary depending on domestic political circumstances.

Secondly, unlike a few self-enforcing international agreements such as air traﬃc or

42 maritime navigation guidelines, agreeing on an IMF program produces few meaningful consequences, unless it is subsequently implemented and domestic politics exerts large influence over the implementation of the policy conditions. In particular, as an IMF program has distributional consequences, domestic interests who bear the costs of the IMF program have incentives and often channels by which they can intervene in the program implementation. In addition to voting an incumbent out in a formal election in most democracies, domestic interests can lobby legislators, local politicians and government officials; protest on a street; sabotage and go on strikes in work places; and in extreme cases, incite coups and revolutions. Indeed, one of the IMF’s reports cites political and social opposition along with limited implementation capacity of a government as the two main factors hampering implementation of IMF programs (The International Monetary Fund 2001, p.72).4 As more extensive policy conditions are more likely to trigger resistance from affected domestic interests, making the implementation of the reform measures more uncertain, the IMF and the government may need to make inevitable trade-offs between including more desirable, efficient conditions and increasing implementability.

Existing empirical studies of IMF programs do not pay due attention to the domestic political economy in the IMF program design process. Instead, most studies focus on economic factors (Independent Evaluation Office 2007b) or, as far as political stories are told, they center around the factors that affect the decision of the IMF — the supply side of the financial resources. For instance, scholars in the public choice

4For instance, Arpac and Bird (2009) details how special interest groups have inﬂuenced the implementation of the IMF program in Turkey.

43 tradition or in organizational sociology highlight ideational, bureaucratic, and organizational incentives of IMF staﬀ to explain the recent expansion of IMF conditions

(Babb 2003, Vaubel 1986, Vaubel 1996, Chwieroth 2007a, Chwieroth 2007b, Barnett

& Finnemore 2004). On the other hand, scholars studying the international political economy emphasize the inﬂuence of sovereign principals of the IMF, most notably the United States, and private ﬁnanciers on the decision making of the IMF via the

Executive Board (Copelovitch 2008, Dreher & Jensen 2007, Gould 2003, Gould 2006,

Momani 2004a, Oatley & Yackee 2004). The exclusive focus on the supply side of the

IMF lending — international politics around the IMF’s decision and organizational and bureaucratic incentives of the IMF — is ﬁne, if the demand side — domestic politics in a borrowing country — makes little impact in designing an IMF program.

Yet, the domestic politics in bargaining international economic agreements can exert a great deal of influence. Along with the reasons put forth above, many empirical and theoretical studies in international political economy report interesting findings on the ways in which the domestic politics influence the outcome of international negotiations. As a type of international economic agreement, therefore, it is very likely that the domestic political economy of the borrowing country exerts significant influence over the final design of an IMF program.

The theoretical framework of two-level games serves as a useful starting point to understand why there exists such a substantial variation across IMF programs and how domestic politics may inﬂuence outcomes of international negotiations. However, multiple logical conjectures exist within the two-level games framework. On the one hand, one can expect that the government may bring in the IMF to tip the balance

44 against domestic opposition (Vreeland 2003). A typical scenario would feature a reform-minded government with strong anti-reform domestic opposition. As the government is unable to carry out reform alone, the government may try to bring in the IMF to tip the balance against the domestic opposition. With the IMF on its side, the government is able to push forward the reform measures that it would not be able to implement without the IMF. According to this account, bringing in the IMF serves two purposes. First, it increases the rejection cost for the domestic opposition, as rejection to the reform is not only against the government but also against the IMF, which would have greater consequences. For instance, rejecting the

IMF program can send negative signals to foreign investors about the ability and willingness to reform. Second, the government can use the IMF as a scapegoat and blame the IMF for the economic hardship caused by implementation of the reform measures. If this is right, we can expect that the design of the IMF program would favor the position of the IMF at the expense of the domestic opposition. On the other hand, one can make the completely opposite argument that the government brings in strong domestic interests to extract more lenient conditions from the IMF.

In a typical scenario, a government can point to strong domestic opposition and tie its hands with the opposition when the government negotiates with the IMF. It will then allow the government to minimize politically costly conditions in the IMF program. As an IMF program not only needs to be signed but also implemented, and domestic opposition can exert inﬂuence over the implementation of the IMF program, the IMF could be more willing to compromise when domestic interests are

45 strong. Thus if this scenario holds, then the government would side with domestic opposition in the hope that it can minimize policy reforms that are politically costly.

In order to understand how the domestic politics in the borrowing country inﬂu- ences outcomes of international economic bargaining in the IMF program negotiation context, I constructed a formal model of IMF program design and implementation.

I build upon the model of bargaining and enforcement of an international agreement proposed by Fearon (1998) and extend his formalization by unpacking the assumption of a unitary state actor. Specifically, while I maintain that the formal bargaining happens between the IMF and a government, I allow domestic interests to play a role in the implementation or enforcement stage. This is a significant development from the Fearon’s model and more accurately reflects real world international economic negotiations. Defection in the enforcement stage, following Fearon’s terminology, or in the implementation stage in my model, is often not a matter of the government’s defecting but rather it is the resistance of domestic oppositions that frequently halts the implementation. The involuntary defection due to domestic opposition, in other words, is explicitly captured in my model.

Thematically, the process of IMF program negotiation is illustrated in ﬁgure 2.1.

The model of IMF program design and implementation is a two-staged process. In the designing phase, the IMF and the government negotiate over program conditions.

When the program is signed and approved, it is then moved to the implementation stage. In the implementation stage, the government tries to implement the program, and the aﬀected domestic interests decide to resist or to acquiesce. These two stages

46 Figure 2.1: IMF Negotiation Process

are interrelated, as the IMF and the government take the probability of implementation into consideration in the designing phase. This is because the IMF and the government not only care about making an agreement but also implementing the reform measures in the agreemeent. In turn, the decision of the aﬀected domestic interests to resist or not depends on the extent of IMF conditionality agreed upon by the IMF and the government in the designing stage. Finally, when the program is implemented, it brings political, social, and economic consequences.

2.4 The Model

In order to comprehend the complex dynamics of IMF program design and implementation, I constructed the game theoretic model and deduced a set of testable

47 implications from the model. I will begin the section by discussing conditionality, the main parameter of interest in the model.

2.4.1 Conditionality

A number of policy conditions are included in a typical IMF program, and these conditions are often categorized by targeted economic sectors. For instance, a report prepared by the Policy Development and Review Department of the IMF classi-

ﬁes all structural conditions into 14 economic sectors (The International Monetary

Fund 2001). Among the 14 economic sectors, policy conditions tend to be concentrated in a few sectors, such as public enterprise reform and restructuring, privatization, the fiscal sector, and the financial sector. Similarly, a report prepared by the Independent Evaluation Office of the IMF (2007b) uses nine reform categories to classify reform conditions. The IEO reduces the number of categories by collapsing a few minor sectors into “Other Fund Core” and “Other World Bank Core.” The main categories again cover tax policy and public expenditure management, financial sector reforms and development, state-owned economy reform, civil service reform, and privatization.

Considering the domestic eﬀect of policy conditions, IMF conditionality is multidimensional. Each set of policy conditions imposes compromises to a speciﬁc economic sector, yet bears little direct negative consequence to other economic sectors.

For instance, conditions targeting the financial sector, such as tighter regulations of the financial sector, tighter banking supervision, and corrective actions in problem banks, have an immediate effect on the financial sector yet have little direct effect on

48 the public sector or the agricultural sector. Likewise, state-owned enterprise reforms and privatization have a direct consequence on the welfare of the public sector, yet do not have such an eﬀect on the private sector employees.

Treating multidimensional IMF conditionality as unidimensional is potentially problematic, especially when considering how domestic politics aﬀect the design of policy conditions. Existing studies of IMF conditionality, excluding occasional IMF reports, generally treat IMF conditionality as a single dimension. The common measure of IMF conditionality in the literature is the number of all policy conditions.

This is understandable, given the main focus of the extant studies is how international politics around the IMF and internal factors within the IMF determine the degree of severity or leniency of conditions over a country’s economy (Dreher &

Jensen 2007, Copelovitch 2008). Yet, there is the possibility that certain international and organizational factors affect a particular set of policy conditions over an economic sector and explain the variation in the particular set of conditions without having significant influence over the others. Without treating IMF conditionality multidimensionally and disaggregating policy conditions by targeted sectors, it is hard to tell if those international and organizational factors equally affect policy conditions over all economic sectors or disproportionately affect a few selected sectors without affecting the others.

Therefore, in order to model how domestic interests affect the final form of the conditions negotiated by the IMF and the government, it is especially important to look at conditions by targeted sectors. Agricultural interests of a country will be interested in influencing the conditions pertaining to agricultural policies, while

49 financial interests of the country will want to exert influence over financial sector reforms. Agricultural interests would have little incentive for attempting to shape the design of financial sector conditions and vice versa. Treating IMF conditionality multidimensionally by targeted sectors allows us to identify precisely who will be the important domestic interest groups that would try to influence the design of a particular set of conditions targeting the sector.

Conditionality is the parameter of interest in the model. Being disaggregated by targeted sectors, a particular set of conditions is then negotiated by anticipating domestic political consequences of the targeted sector. For instance, public sector conditions are designed by the IMF and the government, who is anticipating response from the public sector. Similarly, ﬁnancial sector conditions are designed in consideration of the reaction from the ﬁnancial sector interest. Then, the whole set of conditions can be thought of as the additions of entire sets of conditions over all targeted sectors. The model is general in the sense that it models the interactions between the IMF, government, and generic domestic interests.

2.4.2 The Players

There are three strategic players in the IMF program negotiation model. The ﬁrst one is the IMF. The IMF needs to negotiate and decide ﬁrst whether it is willing to arrange an IMF program with a potential borrowing country and how much it would lend with what conditions attached. Note that the IMF is treated as a strategic actor in this model. The IMF, driven by its own ideological, bureaucratic, or political incentives, has its own preference over a set of policies that a potential borrower has

50 to implement. Its lending decision is based on a strategic calculation of expected beneﬁts and costs of a loan arrangement.

At the other end of the negotiation table is a government. The government negotiates with the IMF over whether it can receive ﬁnancial assistance from the

IMF and with what conditions included. When the IMF and the government ﬁnd a mutually acceptable set of policy measures, they sign an IMF program. When they fail to do so, there is no IMF program.

Finally, there is a targeted domestic interest that is going to be adversely affected by a set of conditions included in an IMF program. When public sector conditions are negotiated, the relevant domestic interests are those who are negatively affected by the conditions: for instance, public sector employees. When financial sector conditions are negotiated, the relevant domestic interests are those whose interests should be compromised when the conditions are implemented. Only those who are negatively affected get to “make a move.” Those who are not negatively affected do not make a move yet are implicitly accounted for in the payoff for the government.

In the current IMF context, a targeted domestic interest can be any faction of the population that shares a similar preference over a potential set of IMF conditions.

For instance, it may be employees in the public sector. Since IMF programs often include several speciﬁc conditions on the public sector, public sector employees have a shared interest to minimize these damaging conditions. These conditions often include restructuring of public enterprises, privatization of non-ﬁnancial state-owned enterprises, and wage and employment limits in civil service (Independent Evaluation

Oﬃce 2007a). Sometimes, public sector employees are represented by the union(s).

51 Another often targeted domestic interest is the financial sector in a country. Espe- cially during and after the Asian financial crisis in 1997, many conditions targeting the financial sector were included in IMF programs, as the Fund diagnosed the weak

financial regulations as the root of the crisis, hence started to emphasize the soundness and transparency of the financial sector (Haggard 2000). These conditions often require a government to install laws to strictly regulate strictly the financial sector and to establish firmer supervisions and corrective actions in problematic banks

(Independent Evaluation Office 2007a). A domestic interest can be comprised of workers who benefit relatively more from government spending. As studies show, there is general consensus that the effect of IMF programs on labor is not beneficial

(Vreeland 2002, Garuda 2000). These harmful effects can be delivered via many different mechanisms. For instance, as private and public firms go through restructuring, workers may be in danger of losing their jobs. In addition, as part of austerity policy measures, a government may be required to increase tax rates while cutting governmental spending on education, health care, and various subsidies, which are more sensitively received by the ones who benefit the most — the working class.

2.4.3 Sequence of the Game

The IMF and the government negotiate over a single dimensional policy space P ∈

[Pmin, Pmax] in the “design” stage. P can capture the severity of conditions over a generic economic sector. It can be positioned anywhere in between the minimum P, denoted Pmin and the maximum P, denoted Pmax. To make the model as simple as possible, I adopt take-it-or-leave-it bargaining protocol. Thus, upon receiving the

52 offer from the government, the IMF decides whether it wants to approve or to disap- prove the program. The take-it-or-leave it offer can be thought of as the final draft of the program formulated through rounds of talks held between government officials and IMF staff and sent to the Executive Board of the IMF. Then, the approval decision by the IMF can be thought of as the Executive Board approval procedure.

When the IMF approves PG, the particular P that is stipulated in the offered Letter of Intent, the designing stage ends. The approved PG becomes a part of the IMF program. When the IMF rejects the offer, the design phase ends without an IMF program. This is when there exists no mutually acceptable IMF program design. I call this outcome “No Agreement” which I denote as “NA.” When the IMF accepts the offer, then the IMF program is arranged with PG and the implementation stage begins.

Entering the “Implementation” stage, the government complies and domestic in-

5 terests get to move. Given the level of PG, the domestic interests choose whether they want to acquiesce or to resist to the agreed upon IMF program. When the domestic interests acquiesce, the IMF program is implemented without a problem. I call this outcome “Implementation” and denote it as “IMP.” Finally, when the domestic interests try to resist, the IMF program can still be implemented with probability

1 − p, but the program is aborted with probability p where p is proportional to the size of the domestic interests. Here, I posit that as the size of the domestic interests

5 I can make the game such that a government can decide whether it wants to comply to PG or not, but under complete information, this bears no analytic significance because if the government prefers non compliance, it can decide not to offer or offer such that the IMF would not want to accept. Hence, I assume when a government signs an agreement, it tries to implement the program.

53 Figure 2.2: Game Tree: The Main Model

increases, there is a linear increase in the probability of successful resistance in case the domestic interests opt to do so. I call this outcome “Resistance” and denote it as “RST.”

Note that only domestic interests whose welfare is adversely affected by a set of IMF conditions get to move and choose to resist. The rest of the population, whose interests are not harmfully affected by the IMF program, are assumed to enjoy benefits passively and are only implicitly accounted for in the government’s utility function. The model is general, in the sense that the domestic interests can still be any group of people. The precise nature of the relevant domestic interests becomes fixed only when a set of policy conditions are fixed. For instance, when one

54 looks at the policy conditions that harmfully aﬀect interests of the public sector, the relevant domestic interests becomes those who are employed in the public sector.

2.4.4 Payoﬀs

As the model tries to understand the design of an IMF program, the variable of primary interest is PG. PG represents the degree of conditions — how lenient or severe agreed upon conditions are for the domestic interests. Severe conditions may mean breadth and depth for the particular set of conditions targeting the domestic interests. Broader conditions mean that the prescribed economic reforms are not limited but target broader economic policies. Deeper conditions mean that conditions require fundamental changes, that is, a revision of laws and institutional reforms.6

PG varies from Pmin to Pmax in one-dimensional Euclidean space, where the former is the most preferred outcome for the domestic interests, meaning no speciﬁc policy conditions on the domestic interests. The latter is the most preferred outcome for the IMF. The latter can be thought of as the policy measures that the IMF wants in the ideal policy-making environment, where there is no potential resistance from a government, and there is no domestic political constraint for the government to implement the reform measures. Furthermore, I normalize the space of PG by setting

Pmin = 0 and Pmax = 1. Thus, PG varies between 0 and unity. As illustrated, the higher PG gets, the more severe the conditions are for the domestic interests, and closer they are to the most preferred policy of the IMF.

6The IMF’s Monitoring Arrangements (MONA) dataset differentiates conditions requiring change of law that has long lasting effects, conditions requiring one time institutional change, and conditions requiring no such changes. 55 I normalize the utilities of the status quo or NA for all parties involved equal to 0. Below I briefly discuss the payoffs for the IMF, the domestic interests and the government in turn. The subscript represents each player (F for the IMF, G for the government, SIG for the domestic interests), and acronyms within parentheses represent outcomes.7

Payoﬀs for the IMF

Formally, the payoﬀs for the Fund can be expressed with the probability of successful resistance (SR) deﬁned as Pr(SR) and the probability of failed resistance (FR) denoted as Pr(FR). Pr(SR) + Pr(FR) = 1.

• UF(NA) = 0

p • UF(IMP) =Beneﬁts - Costs of the IMF program = BF ∗ ( PG − Pmin) − CF

As PG varies between Pmin = 0 and Pmax = 1, this can be rewritten as p UF(IMP) = BF ∗ PG − CF

• UF(RST) = Pr(FR) ∗ UF(IMP) + Pr(SR)∗(Costs of the stopped IMF program) s p s = (1 − β ∗ )(B ∗ P − C ) + (β ∗ ) ∗ (−rC ) N F G F N F First of all, when there is no agreement the payoﬀ for the Fund is standardized to

0. Second, when the program is implemented without domestic interests resistance,

7The payoffs below are largely consistent with economic models of special interest groups. There are a few articles using the model of Grossman and Helpman (2001) to understand domestic political dynamics of IMF programs. They generally focus on special interest groups whose interests are to be compromised with a new IMF program and how conditionality reduces inefficiently dis- torted economic policies which had existed in favor of special interests. See articles by Wolfgang Mayer and Alex Mourmouras (Mayer & Mourmouras 2008, Mayer & Mourmouras 2004, Mayer & Mourmouras 2002). 56 Table 2.1: Parameters in the Model Terms Description Note BF Benefit of an IMF program to the Fund: > 0 proportional to a country’s size of economy interests of the US, other sovereign creditors and PFIs CF Cost of an IMF program for the Fund: 0 < CF < BF proportional to the size of loans Prob. of default, potentially following reputational costs BG Benefit of staying in power for the government > 0 PG Conditionality 0 6 PG 6 1 N Size of total population > 0 s Size of the domestic interests 0 < s < N wi Status quo welfare of an individual i in s > 0 β Degree of organization of the domestic interests > 0 s Pr(SR) β : Probability of successful resistance 0 6 Pr(SR) 6 1 N s Pr(FR) 1 − β : Probability of failed resistance 0 Pr(FR) 1 N 6 6 r Proportion of the cost to the IMF in case of interruption 0 < r < 1 cs Cost of resistance for each special interest group member > 0 ρ A Government’s electoral sensitivity to a vote loss 0 < ρ < 1 γ Reform-mindedness of the government 0 6 γ 6 1 σ Individual’s ideological bias σ ∈ R φ Height of the distribution of ideology > 0 α Population share of each group 0 < α < 1 δ Popularity of a challenger in an upcoming election σ ∈ R ψ Height of the distribution of popularity > 0

57 the payoﬀ for the Fund is a portion of the full potential beneﬁt when there is an ideal

IMF program for the Fund (BF) minus the cost of an IMF program for the Fund (CF).

The beneﬁt, BF, can be thought of as the beneﬁt for the IMF of agreeing on the IMF program, disbursing the loan, and subsequently implementing the economic reforms.

The model treats the source of the benefits exogenously, and BF can be the sum of benefits from financial stabilization of the borrowing country, bureaucratic benefits with better reputation and increased resources, and interests earned from the loan, which provide the operational budgets of the IMF. The cost to the Fund, CF, can be thought of as various risks involved in the loan arrangement. It includes the risk of no or delayed repayments from the borrowing country and associated financial and reputational damages to the Fund. I assume that the realized benefit for the Fund is an increasing function of the level of the conditionality PG. Thus, the realized benefit

8 decreases as PG decreases. In other words, the benefit increases when conditionality gets closer to the IMF’s ideal policy criteria and decreases with lenient conditions, that is, closer to no conditions. This is justifiable, given that the IMF conditions are made to correct deficient economic policies that have caused the balance of payments problem from the IMF’s perspective, thereby to increase the probability of timely repayment of the loan.

8Alternatively, I can divide the benefit with the distance between the ideal point of the Fund and the proposed conditionality. This makes mathematics slightly more complicated without an added insight, so here I multiply the distance between the most preferred policy point of domestic interests and the proposed conditionality. I assume that the marginal benefit decreases over the distance as an alternative specification with constant marginal benefit does not make a qualitative difference and the decreasing marginal benefit seems to capture the idea better.

58 When there is domestic interests’ resistance, the payoﬀ of implementation for the

Fund is realized with probability Pr(FR) = 1 − Pr(SR), where Pr(SR) is proportional to the size of the domestic interests. The size of the domestic interests, denoted as s, is a faction of the entire population, denoted as N. With probability Pr(SR), then resistance is successful and the program is aborted, and the Fund pays a part of the cost (r of CF) without realized beneﬁts of an IMF program. When the size of the domestic interests increases, so does the probability of successful resistance.

Finally, I let β translate the size of the domestic interests into actual influence of the special interest group, with β greater than 0. Therefore, β can be regarded as the coefficient of how well the special interest group is organized. When β is unity, the size of domestic interests directly translates into influence. When it is less than unity, the domestic interests are relatively poorly organized, and the influence is less than their sheer size. A classical example of this kind of group is consumers in trade policy formulation. Conversely, when it is greater than unity, the influence is greater than deserving influence given its size. This may be the case of a well organized economic sector that has a much larger influence in foreign economic policy than its group size, that is, the agricultural sector in advanced economies in trade policy negotiations.

Payoﬀs for the Domestic Interests

Now I turn to the payoﬀs for the domestic interests. Utility for the domestic interests can be written as follows.

• USIG(NA) = 0

59 • USIG(IMP) = Size of the special interests*(individual beneﬁts - individual p costs) = sv − sw PG

• USIG(RST) = Pr(FR) ∗ USIG(IMP) + Pr(SR) ∗ USIG(NA)−cost of resistance p s s p = sv − sw P − sc − ∗ sv ∗ β + ∗ β ∗ sw P G s N N G

When there is no IMF agreement, the payoff for the domestic interests is normalized to 0. When the program is implemented without resistance, the payoff for the domestic interests is proportional to their size and the difference between individual benefits of the program and individual cost imposed on each individual member of the domestic interests. When resisted, with probability Pr(FR) = 1 − Pr(SR), resistance fails and the program is implemented, and with probability Pr(SR), the resistance is successful and the program is aborted, giving the domestic interests the payoff of no

IMF agreement. When the domestic interests decide to resist, the individual cost of resistance, cs is imposed on all the members of the domestic interests.

Payoﬀs for the Government

The payoﬀs for the Government are assigned as follows.

• UG(NA) = 0

• UG(IMP) =Beneﬁts from the Loan + Beneﬁts from reforms + Electoral Con- p sw p sequences = Nv + γ ∗ sw ∗ P + ρ ∗ ψ(v − ∗ P )B G N G G s • U (RST) = Pr(FR) ∗ U (IMP) + Pr(SR) ∗ U (NA) = (1 − β ∗ )[Nv + γ ∗ G G G N p sw p sw ∗ P + ρ ∗ ψ(v − ∗ P )B ] G N G G

60 First of all, when there is no agreement the payoﬀ for a government is standardized to 0. This includes the expected utility of staying in power without having an

IMF agreement. When the program is implemented without domestic interests’ resistance, the utility for the government comes from three parts: one, the overall welfare increase for the population as a whole (Nv), thanks to the loan and economic stabilization, where N is the size of population and v is the benefit for an individual in the population; two, the benefit from the reform, which depends on the government’s reform-mindedness and the level of conditionality; and three, the benefit of staying in power. Reform-mindedness of the government or how much the government is biased toward the policy reforms is captured by γ. When it is 0, the government is agnostic about the magnitude of the reform. When it is positive to unity, it means the government’s utility increases when there is more reform, hence the government is reform-oriented. ρ is an electoral sensitivity parameter, or alternatively, it measures the degree of citizens’ control over government’s survival.9 When constituents have no control over the selection of their leaders, such as in authoritarian governments without meaningful elections, ρ gets closer to 0. Conversely, when constituents exercise total control over the selection of their leaders, like in democracies with highly competitive elections, ρ is close to 1. BG denotes the benefit for the incumbent government to stay in power. Lastly, when there is domestic interests’ resistance, the utility of implementation for the government is realized with probability Pr(FR) =

1 − Pr(SR). With probability Pr(SR), domestic interests’ resistance is successful, the

9The parameter is adopted from the model by Xinyuan Dai (Dai 2005).

61 IMF program is aborted, and the utility for the government is equal to that of no agreement.

I adopt the probabilistic voting model for assigning the government’s utility

(Coughlin 1992, Persson & Tabellini 2002).10 For the probabilistic voting model,

I assume the population consists of two distinct groups Js, here the domestic interests whose interests are to be compromised due to the negotiated conditions and the rest of the population.11 Thus, let:

J = SIG, Others

The domestic interests will be hurt by a particular set of IMF conditions while the rest will beneﬁt from stabilization of the economy, increased foreign investment, or subdued inﬂation brought by the IMF program. Let the population share of group

J be αJ. Hence: s αSIG = N N − s αOthers = N Others s N − s αJ = + = 1 N N J=SIG AtX the time of the elections, voters base their voting decision both on the IMF conditionality negotiated by the government and on the two candidates’ (the incumbent and the challenger) ideologies and popularity. Speciﬁcally, voter i in group J prefers the candidate of the incumbent government if:

WJ(IMF Program)> WJ(NA)+σiJ + δ

10This part is technical. While technical details are essential for the model, less enthusiastic readers can skip the discussion over the probabilistic voting model.

11The discussion here is based on Persson and Tabellini (Persson & Tabellini 2002). 62 where WJ(IMFProgram) is the utility of having an IMF program for a member i in group J and WJ(NA) is the utility of having no IMF program for a member i in group J. I assume that only the incumbent can initiate the IMF program, as it is the incumbent government’s decision to enter into an IMF agreement, and the challenger endorses the status quo, i.e., no IMF program. σiJ is an individual-specific parameter that can take on negative as well as positive values. It measures voter i’s individual ideological bias toward the challenger. A positive value of σiJ implies that voter i has a bias in favor of the challenging candidate, whereas voters with σiJ = 0 are ideologically neutral, that is, they care only about economic policy in general, or in this case, the IMF program. 1 1 The population as a whole is assumed to have a uniform distribution on [− , ]. 2φ 2φ I further assume that there is no difference between these distributions of ideological biases across different groups. In other words, voters’ ideological bias is independent of the group to which they belong.12 These distributions have density of

φSIG, φOthers, φ respectively, and each group has members inherently biased toward both candidates. As I assume that there are no group diﬀerences, these three are equal to each other, or φSIG = φOthers = φ.13

12While I assume a uniform distribution here, speciﬁc form of probability distribution does not make much diﬀerence as long as it is unimodal. Persson and Tabellini states that even if the group distributions of the parameter σ are not uniform, the result does not change much as long as the distribution remains unimodal. They suggest “the properties of the equilibrium are dictated by the group density of σ in a neighborhood of the equilibrium policy. But the interpretation remains the same (Persson & Tabellini 2002, p.57).

13No group diﬀerences substantively mean that the distribution of individual ideology of the affected interest group is same as the distribution of individual ideology of the entire population.

63 1 1 1 1 [− , ] = [− , ] 2φSIG 2φSIG 2φOthers 2φOthers The parameter δ, which measures the average popularity of the challenger in the population as a whole, can also be positive or negative. I again assume a uniform 1 1 distribution over the interval [− , ] for the parameter δ. Although the distri- 2ψ 2ψ bution assumptions regarding σiJ and δ are special, meaning a uniform distribution, they have an advantage in that they facilitate a simple closed form solution.

The government announces its position toward the IMF program by deciding whether to agree to an IMF program and if so, with what conditions. Only the incumbent can make a policy proposal while the challenger(s) holds his or her position at the status quo, that is, no IMF agreement. Then, the swing voter in group J is:

WJ(IMF Program)= WJ(No IMF Program)+σJ + δ

This can be rearranged to:

σJ = WJ(IMF Program)−WJ(No IMF Program)−δ

iJ J All voters i in group J with σ 6 σ prefer the incumbent and vote for the candidate of the incumbent government. In other words, in the uniform distribution, those who are more biased toward the incumbent than the swing voter vote for the incumbent given δ. Hence given the distributional assumptions, the incumbent’s expected vote share is: 1 1 Π = αSIG ∗ φ ∗ (σSIG + ) + αOthers ∗ φ ∗ (σOthers + ) A 2φ 2φ Now, let

For instance, the ideological distribution of ﬁnancial sector employees should have the same ideological distribution as the rest of the population.

64 1 1 ψ Others P = Prob[Π ] = + [ αJ∗σJ(WJ(IMF Program)−WJ(NA)] Incumbent A > 2 2 φ J=SIG Note that for an individual in the domesticX interests, the diﬀerence between p having the IMF program and not having the program is v − w PG. Also note that for an individual who does not belong to the domestic interests, the diﬀerence between having the program and not having it is v. The probability of reelection for the incumbent, PIncumbent is then reduced to: 1 sw p + ψ(v − ∗ P ) 2 N G

2.4.5 Solutions

As the game assumes complete information, the subgame perfect Nash equilibrium is used as the solution concept. The game can be easily solved with backward induction.

I brieﬂy discuss the results. First, I look at the subgame after the government makes the take-it-or-leave-it oﬀer.

Solution for the Subgame

In the subgame, there are three possible scenarios. First, the IMF accepts and the domestic interests resist. Second, the IMF accepts and the domestic interests acquiesce. Third, the IMF does not accept, hence there is no IMF program.

When does the IMF accept yet the domestic interests resist? There are two conditions for this scenario. First, USIG(RST) > USIG(IMP) and second, UF(RST) >

UF(NA). From the conditions, I ﬁnd:

1 Lemma 2.4.1. The domestic interests resist if and only if P (Nc + svβ)2. G > s2w2β2 s

65 ∗ 1 2 ∗ Let P ≡ (Ncs + svβ) . Thus when PG P (this reads as the cut GRST s2w2β2 > GRST point of the PG above which the domestic interests would resist), the domestic interests resist, otherwise the domestic interests acquiesce.14

Lemma 2.4.2. The IMF accepts the deal from the government when and only when 1 C2 1 C2 P F (N − sβ + rsβ)2. I let P∗ ≡ F (N − sβ + rsβ)2. G > 2 2 GIR 2 2 BF (N − sβ) BF (N − sβ) P∗ (this reads as the cut point of the P above which the IMF accepts given domestic GIR G interests resistance) is the cut point for the IMF to accept even knowing that the domestic interests are likely to resist when the program is adapted.

When do the IMF accept and the domestic interests acquiesce? Two conditions pertain. First, USIG(IMP) > USIG(RST), and second, UF(IMP) > UF(NA). The ﬁrst condition is the exact opposite of the condition for the domestic interests to resist. Thus, the solution is:

Corollary 2.4.3. The domestic interests acquiesce when and only when P P∗ . G 6 GRST

P∗ is the cut point for the domestic interests. When the conditionality is more GRST severe than the cut point, the domestic interests resist. If conditionality is less than the cut point, then the domestic interests acquiesce.

Lemma 2.4.4. The IMF accepts the proposal from the government when and only 1 1 when P C2 . I let P∗ ≡ C2 . This is the cut point for the IMF to accept G > 2 F GII 2 F BF BF when it expects that the domestic interests will acquiesce.

14Detailed proof is provided in the appendix at the end of the dissertation.

66 From the Lemma 2.4.4, I also generate two additional conditions for CF and

1 2 BF. As 0 < PG < 1, it also should be the case that 0 < 2 CF < 1. This yields BF 0 < CF < BF as both CF and BF are positive. This is the necessary condition for the conditionality to bear any meaning and the IMF program to be possible.

This condition is easy to see when we consider the null conditions. First, if CF is instead equal to or smaller than 0, the MF is always willing to lend, even without any conditions attached. Second, if CF is greater than BF, the IMF never lends, even with the most ideal conditionality attached. Theoretically, these are possible cases, yet for the purpose of the current dissertation, these fall into the trivial and uninteresting cases category. Thus, in the comparative statics session, I exclude any cases that do not meet 0 < CF < BF.

Corollary 2.4.5. The IMF rejects the oﬀer from the government below P∗ if the GIR IMF expects domestic interests’ resistance and P∗ if the IMF expect the domestic GII interests to acquiesce.

In sum, there are three cut points. If P > P∗ , then the domestic interests G GRST prefer to resist, while if P < P∗ , the domestic interests prefer to acquiesce.15 G GRST This cut point gets smaller as the size of the domestic interests increases as the

ﬁrst derivative of P∗ is negative. Substantively, this conﬁrms the common sense GRST conjecture that the domestic interests are going to resist more readily when they are stronger. P∗ is the cut point above which the IMF agrees even in anticipa- GIR tion of domestic interests’ resistance and below which the IMF does not agree in

15The probability of two are equal is statistically zero. Thus the equality case is trivial.

67 anticipation of domestic interests’ resistance. This cut point gets larger as the size of the domestic interests increases as the first derivative of P∗ is always positive. GIR Substantively, as the size of the domestic interests increases, the IMF becomes more cautious and accepts the offer from the government only if the government promises to implement more conditions. In other words, facing stronger domestic opposition, the IMF wants the government to compensate the high risk of a loan with more reform conditions. Finally, P∗ is the cut point above which the IMF agrees in GII anticipation of implementation and below which the IMF rejects the offer even expecting no resistance from the domestic interests. And this cut point does not vary as the size of the domestic interests varies. This can be thought of as the minimum reservation conditionality for the IMF to grant a loan.

Depending on how these three cut points line up, we get diﬀerent equilibria.

Theoretically, there are 3! = 6 ways to line up these cut points. Yet I prove that:

Lemma 2.4.6. P∗ P∗ for all s.16 GIR > GII

This should make sense, as the IMF presumably wants to agree on IMF program with tougher conditions when they expect domestic interests’ resistance and hence implementation is more questionable than when they expect implementation without any domestic interests’ resistance. Other conditions being equal, the IMF is more willing to approve an IMF program when it expects a smooth implementation without domestic interests’ resistance than a bumpy implementation with strong domestic

16The proof is provided in the appendix.

68 interests’ resistance. Thus, when there is little potential of domestic opposition, the

IMF is more willing to grant lenient conditions.

Given P∗ P∗ , there are now only three ways to line up these cut points. GIR > GII When there are relatively weak domestic interests, P∗ < P∗ < P . This is GII GIR GRST Case 1 in Figure 2.3. When there is are moderate domestic interests in terms of the

17 size, PGII < PGRST < PGIR obtains. This is Case 2 in Figure 2.3. Finally, when the

domestic interests are relatively strong, PGRST < PGII < PGIR ensues. This is Case 3 in Figure 2.3.

The ﬁgure describes how the three cut points behave when s and PG vary, while holding the other parameters at reasonable values. I discuss the three cases in turn.

When the domestic interests are weak, the case resembles Case 1. So, up to the cut point where the IMF accepts expecting implementation without resistance, there is no IMF program, as the IMF rejects any such offer. This first cut point can be thought of as the minimum reservation value for the IMF, the minimum conditions that the IMF wants, anticipating full implementation. Between the first cut point, where the IMF accepts expecting the domestic interests’ acquiescence, and the rightmost cut point, where the domestic interests resist even though they are weak, any proposal is accepted by the IMF, and it is implemented without any resistance. Finally, when it goes tougher over the cut point of the domestic interests’ resistance past the rightmost line, the IMF program is signed, and the domestic interests resist. In sum, when s is small, any proposal above the IMF’s reservation

17Speciﬁc cut points for the size of domestic interests as well as how cut points P∗s behave as s varies are provided in the appendix.

69 Figure 2.3: Theoretical Cases of Conditionality and the Size of Domestic Interests

70 point is accepted and implemented in between P∗ and P and is accepted and GII GRST resisted above PGRST . Case 2 is an interesting one. As happens in Case 1, up to the ﬁrst cut point from the left, where the IMF accepts expecting implementation, there is no IMF program accepted by the IMF. If the proposal from the government falls right of the cut point of the IMF accepting expecting implementation, but left of the cut point of the domestic interests’ resistance, an IMF program is approved and implemented without resistance. But if it goes over the cut point of the domestic interests’ resistance, yet falls short of the IMF accepting with expecting resistance, the IMF does not accept; hence, there is no IMF program. This is the PG interval, where the conditionality is strong enough that it would trigger domestic interests’ resistance, but at the same time, conditions are still too lenient for the IMF to accept, anticipating resistance from the domestic interests. Finally, if the proposal promises drastic conditions bigger than the rightmost cut point of the IMF expecting resistance, the IMF accepts and domestic interests’ resistance ensues.

In Case 3, when domestic interests are very large, the domestic interests are ready to resist even when the IMF and a government agree on relatively lenient policy conditions. So, there is no IMF program until the cut point of the IMF approving a program expecting domestic interests’ resistance. Having lenient conditions is not worthy of to the IMF, given resistance-ready domestic interests. But, the IMF accepts the proposal from the government when the prize gets bigger in the form of promised reforms. This is when the proposal from the government gets closer to the

71 most preferred policy of the IMF. When PG is large, the odds of successful implementation may be small, but the prize also gets bigger for the IMF. Substantively, Case

3 may be highly unlikely, because it requires extremely strong domestic interests.

The discussions are summarized in Figure 2.4. One can see what would be an expected outcome by coordinating the size of the domestic interests in the Y axis with the severity of conditionality on the X axis. For instance, when the size of the domestic interests is small at 20, any proposal from the government bigger than .5 will be accepted by the IMF and subsequently implemented without resistance from the domestic interests. When the proposal is greater than .7, however, the domestic interests are willing to resist.

Solution to the Entire Game

Until now, I have considered only the subgame after the government proposes. But the real question remains as to what the government would propose given what the government expects to happen in the following subgame illustrated in the above section. Now I turn to the government’s strategic consideration. I assume the government also maximizes its own utility given diﬀerent situations.

1 ∗ Lemma 2.4.7. When γ > ψρ ∗ BG, then the government proposes P when N GRST s falls under Case 1 or 2 described above. This is then accepted by the IMF and implemented without domestic interests’ resistance. In the case of Case 3, then the government proposes Pmax. This is subsequently accepted by the IMF, yet is resisted by the domestic interests in the implementation stage.18

18The proof is provided in the appendix. 72 Figure 2.4: Theoretically Predicted Outcomes in the Possible Cases

73 1 ∗ Corollary 2.4.8. When γ < ψρ ∗ BG, then the government proposes P when N GII s falls under Case 1 or 2 described above. The proposed agreement is then accepted by the IMF and implemented without domestic interests’ resistance. In Case 3, the government proposes P∗ , then the IMF accepts the proposal, followed by domestic GIR interests’ resistance.

There are two diﬀerent subgame perfect equilibria (SPE), depending on the values of the parameters in the government’s utility function. Speciﬁcally, when

1 ∗ γ > ψρ ∗ BG, then the government proposes P when s falls in Case 1 or 2 N GRST described above. This is then accepted by the IMF and implemented without domestic interests’ resistance. In Case 3, then the government proposes Pmax which is subsequently accepted by the IMF, yet is resisted by the domestic interests in the implementation stage. This is more likely to be obtained when γ is large, ψ is small, and ρ is small. That is, when the government is more reform-minded, when the election is less competitive and more distant, and/or the less democratic the country is, this particular equilibrium is likely. I call this equilibrium the Reform Drive Equilib- rium (RDE), as the government proposes the policy conditions closest to the IMF’s preferred policy position given the policy space in order to reform economic policies.

The Equilibrium resembles the argument that a reform-oriented government would bring in the IMF to impose more conditions on domestic interests. 1 The second equilibrium exists when γ < ψρ ∗ B . If this is the case, then N G the government proposes P∗ when s falls under Case 1 or 2 described above. The GII proposed agreement is then accepted by the IMF and implemented without domestic interests’ resistance. In Case 3, the government proposes P∗ , then the IMF GIR 74 accepts the proposal followed by domestic interests’ resistance. This is likely to happen when the conditions are exactly opposite to the previous equilibrium. Thus, when the government is less reform-minded, the electoral competition is strong and approaching soon, and the country is more democratic, this equilibrium is likely to be obtained. I call this equilibrium the Electorally Constrained Equilibrium (ECE), as the government chooses the closest policy point to the most preferred policy of the domestic interests, given the available policy space. The Equilibrium closely matches the theoretical argument that a government can tie its hands with strong domestic interests to extract more lenient conditionality from the IMF. The two Equilibria are formally summarized in the following table.

It is important to note that the Reform Drive Equilibrium is always equal to or greater than the Election Constrained Equilibrium.

Lemma 2.4.9. The ﬁrst set of SPE (Reform Drive Equilibrium) is always greater or equal to the second set of SPE (Election Constrained Equilibrium).19

To illustrate the equilibria graphically, I set some of the peripheral parameters to reasonable values and simulate how changes in the key parameters change the equilibria. Here are some representative examples.

I start with a market-oriented pro-reform authoritarian government. As the pay- p oﬀs for the government reduces to UG(NA) = 0, UG(IMP) = Nv + sw PG and

19Proof is in the appendix.

75 Table 2.2: Equilibria in IMF Conditionality Design and Implementation Game Condition Actor s < H H < s 1 RDE G P = (Nc + svβ)2 P = 1 I s2w2β2 s H IMF Accept Accept SIG Acquiesce Resist 2 1 2 1 CF 2 ECE G PB = 2 CF PA = 2 2 (N − sβ + rsβ) BF BF (N − sβ) IMF Accept Accept SIG Acquiesce Resist Condition Actor Selection 2 1 (Nv + vψρBG) RDE G or 0 if > PG s2w2 1 2 γ − N ψρBG IMF Not Accept SIG 2 1 (Nv + vψρBG) ECE G or 0 if < PG s2w2 1 2 γ − N ψρBG IMF Not Accept SIG

76 s p U (RST) = (1 − β )(Nv + sw P ), the government always prefers implementa- G N G tion over resistance over no agreement. Moreover, the government’s utility increases as PG increases. Thus, when the domestic interests are small to medium, the government chooses the maximum conditions that induce the IMF’s agreement but still falls short of invoking domestic interests’ resistance. This point is P∗ . But when GRST the domestic interests are large enough to prohibit such a window of opportunity, the government instead agrees to the maximum level of conditions, P . As P∗ max GRST decreases as s increases, the ﬁnal outcome we observe in the case of pro-reform authoritarian government is the steady decrease of the number of conditions as s increases. But above a certain point H where s is so strong that the domestic interests are almost always willing to resist, the condition jumps to the maximum number.

When this happens to an authoritarian government with no desire to reform, very different dynamics ensue. Assuming that overall benefit from the market reform still outweighs the cost imposed over the domestic interests, when the domestic interests are small to medium, the government chooses the minimum conditions that would still induce the IMF’s agreement. It is P∗ . When s is large, then the government GII still prefers to have a program but with the fewest possible conditions. But the most lenient possible conditions will still trigger domestic interests’ resistance. The equilibrium here is P∗ . Thus, the final outcome we observe in the case of an au- GIR thoritarian government with no desire to reform is the minimum, constant number of conditions for the small and medium domestic interests and steadily increasing conditions in the case of the large domestic interests. But overall the level of conditions is less reform-oriented than that of a pro-reform authoritarian government.

77 Figure 2.5: Illustration: Reform-Minded Autocracy

78 Turning to democracy, how reform-minded a government is will be less signiﬁcant in designing the conditions, as democratic governments do not show signiﬁcant differences. In general, democratic governments always want the minimum policy point for the IMF to sign on to an agreement when s is small to medium, unless there is very weak electoral competition. When s gets large, then there are two possibili- ties. As it becomes too costly, the government’s utility of having an IMF program actually becomes negative, thus a government may prefer no agreement over having an agreement. In this case, the government chooses not to participate in an IMF program by proposing unacceptable conditions to the IMF. If it is not too costly to participate in a program, then the government prefers having an agreement. This case then resembles the authoritarian government with no desire to reform. The number of conditions stays at the minimum, then steadily increases as s gets past the cut point of domestic interests’ resistance and gets larger.

The two distinct equilibria produce quite diﬀerent comparative statics. I ﬁrst examine comparative statics in the Reform Drive Equilibrium. As shown above, the

∗ 1 2 Reform Drive Equilibrium features P ≡ (Ncs + svβ) when the s falls GRST s2w2β2 in the small to medium range, and Pmax when s is large. As Pmax = 1 is a constant, meaningful comparative statics can be generated only with P∗ . GRST

Proposition 2.4.10. Reform Drive Equilibrium: As the size of the domestic interests increases, as the domestic interests are better organized, as the cost of resistance

79 Figure 2.6: Illustration: Democracy

80 decreases, and as the individual beneﬁt of the IMF program decreases, the number of conditions is expected to decrease.20

Under the Reform Drive Equilibrium, the government sides with the IMF. Only reform-oriented governments with relatively few electoral constraints can aﬀord this

Equilibrium. First of all, under this Equilibrium, the equilibrium conditionality decreases as the size of the special interest group increases. This is because the government not only cares about the design of conditionality but also about implementation of conditions. Thus, if available, the government chooses the maximum conditions that will not trigger domestic interests’ resistance. But after the threshold where the domestic interests are too large to allow the space, where it is both acceptable to the IMF and still does not trigger resistance, the government faces the decision between no IMF agreement and IMF agreement with expected resistance from the domestic interests. In this case, the government still goes for the agreement and picks the ideal reform policy identical to the IMF’s most preferred policy. The same dynamic is expected when there are more incentives for the domestic interests to resist. For instance, when the domestic interests are better organized, the number of conditions is expected to decrease until the threshold. The same is expected when the cost of resistance to the domestic interests is lower and the individual beneﬁt of the IMF program is lower.

In contrast, in the Electorally Constrained Equilibrium, quite diﬀerent comparative statics are generated. Recall that the Electorally Constrained Equilibrium

20Proof is in the appendix.

81 1 features P∗ ≡ C2 when the s falls in small to medium range, and P∗ ≡ GII 2 F GIR BF 2 1 CF 2 2 2 (N − sβ + rsβ) when s is large. BF (N − sβ) Proposition 2.4.11. Electorally Constrained Equilibrium: As the size of the domestic interests decreases, as the domestic interests are less organized, and as the beneﬁts of a program for the IMF increases, as the costs of a program for the IMF decreases, the number of conditions is expected to decrease.21

Under the Electorally Constrained Equilibrium, the government ties its hands with the domestic political constraints and tries to minimize the reform measures.

This is due to electoral pressures present in democracy as well as the tradeoﬀ between the number of conditions and the probability of successful implementation. As for the size of the domestic interests, the number of conditions stays constant at its minimum at the IMF’s reservation policy point. This point does not provide enough incentives for the domestic interests to resist until the threshold. This is because any additional policy condition to the minimum will be electorally costly. When there is no more policy space acceptable to the IMF while deterring the domestic interests from resisting, then the government proposes conditionality acceptable to the IMF. This increases as the size of the domestic interests increases. Thus, the equilibrium conditionality is a non-continuous function of the size of the domestic interests. When the size is small to medium, the conditionality stays at the minimum point. When the size is large, the conditionality increases as the size increases. As the domestic interests are better organized, the domestic interests can readily resist,

21Proof is in the appendix. 82 and this raises the minimum number of conditions that the IMF would agree on.

Thus, when the domestic interests are better organized, the number of conditions is expected to increase. In contrast to the previous Reform Drive Equilibrium, other comparative statics with regard to the factors suggested in the literature are also generated in this Equilibrium. For instance, as the beneﬁts of an IMF program for the IMF increases, as is presumably the case when the borrowing country has a bigger economy with a good record of IMF program implementation, the number of conditions decreases. The IMF is willing to reduce the number of conditions when it is likely to beneﬁt more by agreeing on a program. Conversely, when the program is deemed too costly by the IMF, that is, when the borrowing country has a bad record of implementing conditionality or repaying loans, the IMF is likely to require more conditions for the borrowing government.

Furthermore, comparing the two Equilibria generates empirically testable comparative statics. As shown in Lemma 2.4.9., the equilibrium policy is always greater in the Reform Drive Equilibrium than in the Electorally Constrained Equilibrium in the entire possible range of s. The choice between the two Equilibria is dependent on the two parameters, γ and ρ. Speciﬁcally, when the government is less reform- minded, the electoral competition is strong and approaching soon, and the country is more democratic, this equilibrium is likely to be obtained.

2.5 Theoretical Implications

As analytically and graphically demonstrated above, there are three important domestic political factors that are inﬂuential in deciding the equilibria. First of all,

83 democratic governments with stronger electoral pressure should have fewer conditions than autocratic governments with weak or no electoral pressure. While both regime types prefer to have the IMF program to not having it when the size of the domestic interests is small to medium, the electoral pressure forces democratic governments to extract the minimum PG acceptable to the IMF regardless of the ideological leaning of the government, that is, whether it is reform-minded or not.22

On a related note, the ideological commitment of a democratic government makes little diﬀerence in designing conditionality. Again, this is because of the electoral pressure that the democracy faces. In contrast, there is more room for the ideology of a government when the regime is less democratic and the election does not decide the fate of the incumbent. Speciﬁcally, a reform-minded autocratic government would choose the maximum PG available whereas an anti-reform autocratic government would choose the minimum PG, being similar to a democratic government. In the case where there exists some range of PG acceptable to the IMF and acquiesced by the domestic interest, the ideological commitment of an autocratic government plays an important role in deciding the degree of conditionality.

Finally, up to the medium size, the size of the domestic interests does not change the degree of conditionality in a democracy. After that, if the democratic government prefers having the agreement with expected resistance to not having a program, conditionality increases as the size of the domestic interests increases. This is because the minimum demand from the IMF increases as the size of the domestic interests

22I assume that when there is a conﬂict between political survival and ideological commitment of the government, political survival comes ﬁrst.

84 increases. In contrast, under a reform-oriented autocracy, as the size of the domestic interests increases to the medium size range, the equilibrium policy proposal gets smaller. This is because the maximum conditionality which can be implemented without domestic interests’ resistance decreases as the size of the domestic interests increases. Thus, the size of the domestic interests matters more in an autocracy.

When the size gets too large to have a range within which the IMF accepts and the domestic interests acquiesce, then the autocratic government proposes the IMF’s most preferred policy position.

Notice that the regime type and the ideological commitment are general predictions, while the size of the domestic interests is conditional on the regime type and the ideological commitment of the government.

2.5.1 Hypotheses

• Hypothesis 1: The more democratic a country is, the more lenient the condi-

tions will be in an IMF program.

• Hypothesis 1-1: The sooner the next scheduled election is in a democratic

country, the more lenient the conditions will be in an IMF program.

• Hypothesis 1-2: The more competitive an election is in a democratic country,

the more lenient the conditions will be in an IMF program.

• Hypothesis 2: The more reform-minded a government is, the more severe the

conditions will be in an IMF program.

The following are conditional on which equilibrium the equilibrium policy sits in. 85 When the circumstances lead to the Reform Drive Equilibrium, that is, when the government prefers reforms while facing little electoral pressure, the model suggests the following hypotheses.

Reform Drive Equilibrium:

• Hypothesis 3-1: As the size of the domestic interests increases, the severity of

conditionality decreases, given the size is not too large.

• Hypothesis 4: The less beneﬁt the IMF program brings to an individual in a

country, the more lenient the conditions will be in an IMF program.

Quite diﬀerent set of hypotheses is suggested in the Electorally Constrained Equi- librium.

Electorally Constrained Equilibrium:

• Hypothesis 3-2: As the size of the special interest group increases, the number

of conditions remains constant to a certain level, given the size is not too large.

• Hypothesis 5-1: The more beneﬁts the IMF garners from a program, the fewer

conditions an IMF program will contain.

• Hypothesis 5-2: The more costs the IMF garners from a program, the more

conditions an IMF program will contain.

2.6 Conclusion

This chapter introduces the IMF conditionality design and implementation model that captures the strategic interactions between the IMF, a government, and domestic 86 interests. The chapter first introduces the strategic players and their respective payoffs for the possible outcomes. It is followed by discussions on the solution for the game. The model highlights the importance of three domestic political variables and other variables shaping the design of IMF conditionality. Specifically, the model yields the proposition that those governments that are sensitive to vote losses are pressured to extract a more lenient deal for domestic interests from the IMF and to stay with the minimum conditions acceptable to the IMF, regardless of their own preferences over reforms, as additional reform measure are electorally costly.

In comparison, those governments that are less sensitive to vote losses can pursue their own ideological orientation (reform-mindedness) freely. The countries with few formal institutional constraints are only constrained by strong domestic interests who can intervene and disrupt the implementation of agreed upon policy reform measures.

The ensuing chapters provide rigorous empirical tests of the hypotheses generated in this chapter. Chapter 3 tests the predictions generated from the model with public sector conditions. Chapter 4 tests the predictions with austerity measures or

fiscal policy conditions. Since public sector conditions and austerity measures are generally considered as adversely affecting respective economic sectors and public and very often opposed by domestic interests, they provide sound tests of the predictions from the model. The last empirical chapter, Chapter 5, examines financial sector conditions. Given existing studies that document domestic financial sectors’ disposition toward financial liberalization, I modify the main model presented in this chapter. The simplified model provides very simple predictions with regard to the

87 relationship between the size of a loan and the severity of conditionality. The chapter examines the loan size hypothesis and other political variables suggested in the IMF literature.

88 CHAPTER 3

IMF PROGRAMS AND PUBLIC SECTOR REFORMS

3.1 Introduction

IMF conditionality includes diverse conditions over diﬀerent economic sectors. They range from tax policy reforms to civil service reforms to education, health, and pension policy reforms. While these reform measures are partly decided based on the discrepancy between IMF’s benchmark economic policies that the IMF believes are optimal for a borrowing country and the borrowing country’s actual economic policies, international and domestic political economy surrounding the IMF lending process can also inﬂuence the design of an IMF program.

The previous chapter introduced the model of IMF program design, focusing on domestic politics of IMF program design and implementation. The model provides a number of clear, testable implications. In this chapter, I empirically test the hypotheses using an original dataset of IMF conditionality, covering all IMF programs signed between 1994 and 2006.

Among diﬀerent sets of conditions, this chapter focuses on the eﬀects of the

89 domestic political factors in negotiating public sector conditions. Public sector conditions, including the restructuring of state-owned enterprises (SOEs), privatization of public enterprises, and civil service reforms are one of the main kinds of conditions.

Among all IMF conditions, 25 to 30 percent of all IMF conditions are public sector conditions. Applying the theoretical implications to the design and implementation of public sector conditions, the chapter will empirically examine how the theoretical predictions perform in explaining the design of public sector conditions.

The chapter begins with a discussion of public sector conditions. Following that, the original dataset of IMF conditionality will be introduced, with an emphasis on the public sector conditions. A set of hypotheses in the context of designing public sector conditions is introduced and the research design, including the dependent variables, the main independent and control variables, and estimation techniques, is presented. In the following section, statistical results are presented and substantively interpreted. The results highlight a number of findings. First, as the model predicts, more democratic countries are more likely to have fewer public sector conditions in their IMF programs. Second, among democracies, the ones whose previous electoral victories were won by slimmer margins are more likely to have fewer public sector conditions than those who won the previous election with wider margins. Third, among democracies, the ones whose next election is more proximate in time are more likely to have fewer conditions than those who face more distant next elections. Fourth, in autocracies, the stronger the public sector is, the fewer conditions a program contains. In contrast, the strength of the public sector does not have a statistically significant effect over the number of public sector conditions in democracies.

90 Finally, the bigger the size of a loan to a country, the more public sector conditions a program is likely to contain. Overall, the results are consistent across diﬀerent model speciﬁcations and provide strong support for the theoretically generated predictions.

3.2 Public Sector Conditions

Public sector conditions are a major part of structural conditions in IMF programs.

As IMF programs generally encourage market oriented liberalization to borrowing countries, one of the major reform areas under these programs has been the public sector. Most public sector conditions are set to reduce a government’s role in the economy. The IMF faithfully follows the neoliberal economic principles summarized as the “Washington consensus,” and with regard to the public sector, that means reducing the size of the public sector and reducing the role of the government in the economy. The original list of items of the Washington consensus includes ﬁscal discipline, re-prioritizing public expenditure priorities, privatization, and deregulation

(Williamson 2009). According to the 2001 report by the IMF’s Policy Development and Review Department, 20 to 40 percent of the structural conditions are on public sectors (The International Monetary Fund 2001). Similarly, the more recent study by the Independent Evaluation Oﬃce (IEO) of the IMF shows that about 25 to

30 percent of total structural conditions are either on civil service reforms, reforms of state-owned enterprises, or privatization of state-owned enterprises (Independent

Evaluation Oﬃce 2007a).

The 1996 Kenyan IMF program is a good example. In seeking further liberalization of the Kenyan economy, the program includes a number of public sector related

91 conditions. Speciﬁcally, the approval of the program was conditional on implementing a few prior actions, including “Cabinet approval of the restructuring of Kenya

Railways,” “Invitations to bid on a contract to commercialize the National Cereals and Produce Board (NCPB) by end-1996,” and “Cabinet approval of a framework for the liberalization of the telecommunications sector and of the privatization plan of the Kenya Posts and Telecommunications Corporation (Kenya 1996).” In addition, the program also contains a few performance criteria conditions that have to be implemented in order for the Kenyan government to keep withdrawing the loan as scheduled. The performance criteria required the government to appoint experts for the commercialization of NCPB, to privatize 20 public enterprises, and to reduce civil service positions by 6,500. Finally, there were also a few structural benchmarks including further privatization of 40 public enterprises and reduction of civil service positions by 13,000.1

While having a few public sector conditions in an IMF program is quite common, there is also variation among the number and the kind of public sector conditions.

Some programs have very few public sector conditions or even none, while others have more than 30 conditions. What determines the number of conditions included in a program? While the existing literature focuses on a number of economic and international political factors to account for the variation among IMF programs’ conditions, I provide an alternative, domestic political account for the variation in

IMF program design. The model was developed in the previous chapter, and the

1Prior actions, structural performance criteria, and structural benchmarks are all diﬀerent types of structural conditions.

92 predictions from the model will be tested with other economic and political variables highlighted in the existing studies.

One of the challenges of studying IMF programs has been data limitation. Until recently, researchers outside of the IMF had not been granted access to the rich database that the IMF Policy Development and Review Department maintain, called the Monitoring of Fund Arrangement Database (MONA). A part of the database became available to outside researchers through the oﬃcial IMF website in 2009 but provides conditionality data dating only as far back as 2002. Thus, I have constructed my own dataset of IMF conditionality, which I describe more fully in the next section.

3.3 Dataset of IMF Conditionality

The dataset that I have assembled contains all 275 IMF programs that have been signed between 1994 and 2006. Among those 275 programs, ﬁve do not have a publicly available Letter of Intent and/or Memorandum on Economic Policies. Those

ﬁve programs have missing entries in the conditionality dataset.2 I have coded the remaining 270 IMF programs by the targeted economic sectors and by the conditionality types. I have counted and recorded the number of conditions falling in each category.

Each Letter of Intent was either downloaded from the oﬃcial IMF website, where most of Letters signed after 2000 are available, or gathered through archival research

2All programs included in the data set are listed in the Appendix B. The 5 missing programs are with Mexico (1995), Guyana (2002), Guatemala (2003), Kenya (2003), and Cameroon (2005). Furthermore, 7 programs with Grenada, Sao Tome and Principe (*2), Dominica (*2), and Cape Verde (*2) are dropped from the dataset due to limited data availability.

93 at the IMF Archives located in Washington D.C. A Letter of Intent often lays out a broad scheme of intended economic reforms with supplementary tables and is often supplemented by a Memorandum of Economic Policies. IMF conditions are as often listed explicitly with their types at the end of Letter of Intent as they are embedded in the main text of the Letter.

Only initial Letters of Intent are coded, as they are the most important ones.

After scheduled reviews, revised Letters of Intent are often published, but those tend to be slight modiﬁcations of what the IMF and the country initially agreed on. Copelovitch (2008, 2010) employs a similar strategy. He relies on the conditions speciﬁed in the Letter initially approved by the Executive Board, on the ground that

“the IMF staff and Executive Board almost never alter the number of conditions from stage to stage.” As the number of Letters of Intent signed in the lifespan of an IMF program varies significantly, “counting each stage (Letter signed after each review) as a separate case would over-weight the influence of longer loans in the IMF lending dataset without actually multiplying the number of relevant observations

(Copelovitch 2008, p. 79).”

I only count structural conditions. IMF conditionality takes one of two forms: quantitative conditions that set numerical targets for various economic indicators such as, fiscal deficit level, government debt, and external arrears, and structural measures that specify how to meet the quantitative conditions. As comparing the quantitative difference between quantitative conditions is extremely difficult, the only way to compare the quantitative conditions across programs in a reasonable manner is to count the number of conditions. Yet, the number of quantitative conditions is

94 pretty constant across programs, since these conditions are standard for all programs.

Thus, the number of conditions provides little information on the degree of severity of conditionality. In contrast, structural conditions specify the speciﬁc measures that a borrowing government has to implement to meet the quantitative conditions and vary across programs widely. It is precisely the structural conditionality that is politically controversial as the structural conditionality aims to micro-manage domestic economic policies. In sum, structural conditions capture the degree of severity of conditions very well and are comparable across IMF programs. Thus, I only include structural conditions in the dataset.

Departing from the common practice of the IMF literature, I disaggregate and classify IMF conditions both by the targeted economic sectors and the condition types. In the dataset of IMF conditionality, the targeted economic sectors are divided into four. These four economic sectors are the public sector, financial sector, fiscal sector, and the remaining. Once the conditions are classified by the economic sector classification scheme, they are then further classified by the types — prior action, performance criteria, and structural benchmarks. Prior action and performance criteria are “hard” conditions, in a sense that they are binding and non-implementation of these conditions triggers program suspension in principle.3 Structural benchmarks are “soft,” and violating one or a few structural benchmarks does not automatically suspend a program.

The Independent Evaluation Oﬃce of the IMF uses nine categories to classify reform conditions in IMF programs. Among the categories, three are closely related

3Granting exemptions to the principle is a highly political decision. See Stone(2004). 95 to the public sector.4 Public sector conditions in the dataset include sectors 4,5, and

6 of the IEO’s classification, which are SOE reform, privatization, and civil service reform respectively. Thus, public sector conditions cover restructuring of public enterprises, pricing policies, regulatory reforms in utilities, all activities related to the privatization of non-financial SOEs, civil service reform wage and employment limits, and other measures affecting employment in the public sector (Independent

Evaluation Oﬃce 2007b).5

Disaggregating conditions by sectors gives an advantage over aggregating all IMF conditions into one group. Firstly, there is a possibility that determinants highlighted in the literature have disproportionate effects on IMF conditions across different targeted economic sectors. For instance, it is possible that the influence of private creditors is focused on financial sector conditions and has little influence over social spending reform measures. Similarly, it is plausible that the swing of the dominant macroeconomic perspective of the IMF staff from Keynesian to neoclassical has some bearing on designing fiscal or monetary conditions, but it might not directly influence the design of agricultural sector conditions. Secondly, it is very important to disaggregate IMF conditions by sectors if one considers their domestic consequences

4Reform categories include 1. tax policy and tax administration, 2. PEM (Public expenditure measures), 3. financial sector reforms and development, 4. state owned enterprizes reform, 5. privatization, 6. civil service reform, 7. social policies, 8. other Fund core, 9. other World Bank core (Independent Evaluation Office 2007b). These nine categories are combined into four different categories in the dataset I have created. Those are financial sector conditions, public sector conditions, fiscal sector conditions including tax, customs and budgetory reforms, and all other sectors.

5Privatization of financial SOEs are classified and coded as financial conditions.

96 and the role of domestic politics in implementing those conditions. Each economic sector has a particular interest in the set of conditions targeting itself, but it will not have such strong preference regarding conditions targeting other economic sectors.

Particular domestic interests will resist certain conditions but not others. Hence, it is critical in order to examine the domestic politics hypotheses developed in the previous chapter.

3.4 Hypotheses

There are a few domestic political factors that are highlighted by the model in the previous chapter. How sensitive a government is to vote losses, how reform-oriented a government is, and how strong are the domestic interests are all factors hypothesized to shape the design of an IMF program.6

The model produces the conjecture that the more sensitive a government is to electoral costs, speciﬁcally, to losing votes, the fewer conditions an IMF program will contain. This is because each additional reform measure brings vote losses, holding the amount of the loan constant. From the sensitivity to electoral costs parameter,

I derive three diﬀerent hypotheses that are empirically testable.

As election outcomes are uncertain in democratic countries and each reform measure is electorally costly, democratic countries should have fewer conditions than their autocratic counterparts. Assuming the government wants to stay in power,

6The hypotheses assume that there is mutually acceptable policy space where the IMF approves and the special interest group acquiesces. Formally, this would mean that s < H.

97 the government in the democratic country can little aﬀord additional reform measures. Hence, the electoral sensitivity of the government in the democracy forces the government to agree to no more than the minimum conditions acceptable to the

IMF.

Similar logic applies to explain the variation in the severity of IMF conditionality within democracies. Within democracies, the sensitivity to electoral costs can vary as some governments can afford to lose a certain number of votes. For instance, the government with a huge margin of victory in the previous election may feel more confident in pursuing electorally costly policy reforms than the government that won the previous election by a slim margin. For instance, the government that won the previous election with 30 percent margin can afford to lose up to 15 percent of the overall vote share and still remain in power in the next election. In contrast, the government that won the previous election with only a two percent margin cannot enjoy such flexibility in adopting potentially costly reform policies: pursuing policy reforms similar to the government having won the previous election with 30 percent margin would bring a loss in the next election.

Following similar logic, the government that just won an election would be better positioned to pursue policies that it desires than the government that expects a proximate next election. The government that will soon face election will discount the next election less and consider the eﬀect on the election more seriously than the government that does not face an election in the next few years. Since an additional policy reform measure is electorally costly, the government with the more temporally proximate election is less likely to pursue electorally costly reform measures

98 and more likely to bargain hard to reduce the severity of IMF conditionality than the government with a more distant election. Note that the assumption of electoral costliness of an additional reform measure does not mean that there is no one who would support the additional reform measure — there could very well be those who want more policy reforms, but as long as the number of those who oppose the measure is greater than the number of those who support the measure, the theoretical implications will hold.

The sensitivity to the electoral costs parameter thus yields the following testable hypotheses:

• Democracy Hypothesis: The more democratic a country is, the more lenient

public sector conditions will be in an IMF program.

• Election Discount Hypothesis: (Within democracy) The more proximate the

next election is, the more lenient public sector conditions will be in an IMF

program.

• Margin of Victory Hypothesis: (Within democracy) The more competitive an

election is in a country, the more lenient public sector conditions will be in an

IMF program.

The model also conjectures that the size of the loan has an eﬀect on the severity of

IMF conditionality. In both the Electorally Constrained Equilibrium and the Reform

Drive Equilibrium, the model predicts that the more economic beneﬁt brought by signing an IMF program, the more a government can aﬀord a greater number of policy conditions. Under the Electorally Constrained Equilibrium, the minimum 99 acceptable number of conditions to the IMF increases as the IMF lends a bigger loan. As the Electorally Constrained Equilibrium follows the minimum possible number of conditions, the increase of the minimum acceptable number of conditions caused by the larger amount of the IMF loan should also increase the equilibrium number of conditions. In other words, bigger loans raise the reservation point of the

IMF. This reasoning is consistent with the public choice argument with regard to the relationship between the size of a loan and the number of conditions (Dreher 2004b).

Under the Reform Drive Equilibrium, the same conjecture is derived for a diﬀer- ent reason. Under the Reform Drive Equilibrium, the equilibrium conditionality is the maximum number of conditions that do not trigger resistance by the domestic interests in the implementation stage. As the government can reduce the incentives for the domestic interests to resist by increasing the beneﬁt from a loan, the maximum possible number of conditions will increase as the size of a loan gets larger.

Thus, both the governments with and without electoral pressure will have more stringent conditions when the size of the loan is larger. One is because the IMF demands more condition,s and the other is because the domestic interests become more tolerant and less likely to resist when the size of the loan is larger. While this may sound intuitive, this contrasts with the argument based on U.S. strategic inﬂuence. The argument based on U.S. interests conjectures that closer allies of the

U.S. and other major sovereign principals of the IMF should be treated favorably with fewer conditions and larger loans. Thus, the reasoning behind the U.S. inﬂuence argument predicts a negative correlation between the number of conditions and the size of a loan.

100 The size of a loan parameter then yields the following hypothesis:

• Loan Size Hypothesis: As the size of a loan increases, the more severe public

sector conditions an IMF program will contain.

Lastly, the hypothesis related to domestic interests is contingent on a boundary condition. As governments sensitive to vote losses, that is under the Electorally

Constrained Equilibrium, are likely to stay with the minimum possible number of conditions, even without taking domestic interests’ ability to hinder policy implementation into consideration, the strength of the domestic interests, or the size of the public sector in the public sector conditions context, should have little bearing on the severity or leniency of public sector conditions. In comparison, those governments that are relatively immune to electoral pressure, that is, under the Reform

Drive Equilibrium, are likely to side with the IMF and have more conditions than countries under electoral pressure, but the number of conditions should decrease as the domestic interests (the public sector) get stronger. This is because given a certain level of severity in conditions, stronger domestic interests are more likely to resist than weaker domestic interests, and the government and the IMF would prefer proper implementation of the compromised conditionality rather than facing resistance, if possible.

In order to capture the boundary condition, I use a democracy versus non- democracy division. I assume that democracies are governments with a higher sensitivity to vote losses while non-democracies, or autocracies, are governments freer from such electoral pressure.

• Public Sector Hypothesis: As the size of the public sector increases, the severity 101 of public sector conditions decreases in nondemocracies. In comparison, as

the size of the special interest group increases, the number of public sector

conditions stays constant in democracies.

3.5 Empirical Analysis

3.5.1 Dependent Variable

The number of public sector conditions in a program is used to capture the severity or leniency of IMF conditionality. While the number of conditions does not perfectly coincide with the severity or leniency of IMF conditionality, the number of conditions is the most objective measure of IMF conditionality, given information and resources limitations. It is extremely diﬃcult to objectively measure the relative severity of conditions across diﬀerent countries, because macroeconomic and economic policy contexts vary widely. It is also widely accepted in the literature as the measure of the stringency of IMF conditions (Copelovitch 2008, Gould 2003, Dreher & Jensen

2007, Dreher & Vaubel 2004). The raw number of public sector conditions is used for the count model analysis, while the log of the number of public sector conditions is used for an OLS regression. I take the natural log to transform the number of public sector conditions, as the distribution of the number of public sector conditions is very skewed, and there is no reasonable way to eliminate a few outliers. In order to log-transform the number of public sector conditions, I add one to the number of public sector conditions, as a few observations have zero public sector conditions.

Figure 3.1 shows the distribution of the number of public sector conditions. The

102 Figure 3.1: Distribution of the Number of Public Sector Conditions

distribution of the number of public sector conditions is right skewed, with many observations falling below 10, while just a few fall between 10 and 20. There are two outliers with 26 and 29 public sector conditions, respectively.7 About 35 percent of all programs have fewer than two public sector conditions, and another 24 percent of all programs have three to four public sector conditions. In comparison, about 10 percent of IMF programs signed between 1994 and 2006 have more than 10 public

7The outlier programs are with Romania (2004) and Gabon (2000).

103 sector conditions. The average number of conditions is 3.9 with a standard deviation

4.26.

Figure 3.2: Distribution of the Log of the Number of Public Sector Conditions

Figure 3.2 shows the distribution of the natural log of the number of public sector conditions, with one added to each observation to be able to take a log when the number of public sector conditions is zero. Except for the inﬂation in between zero and 0.2 on the far left, the distribution approximates a normal distribution. About

20 percent of the observations fall in between zero and 0.2, and the average is 1.25.

The median falls in between 1.4 and 1.6. The maximum is 3.4. 104 Table 3.1: Summary Statistics: Chapter 3 Variable # of obs Mean Std. Dev. Min Max Polity Score 262 3.351 5.436 -9 10 Years Left in Current Term 227 2.185 1.485 0 6 First Round Votes 186 53.554 20.864 20.3 98.8 Public Sector Compensation 213 5.532 3.472 0.678 18.992 Government Expenditure 262 28.340 10.394 6.055 57.735 Loan to Quota Ratio 263 0.896 1.418 0.05 12.62 Loan Size (SDR Millions) 263 829 2881 4 27375 Quota (SDR Millions) 263 486 836 14 5945 GDP per Capita (US$ Thousands) 260 1.470 1.675 0.09 12.19 Trade Volume (US$ Billions) 257 15.627 35.739 0 292.563 UN Voting with U.S. 262 0.135 0.328 -0.425 1 Transition Country 263 0.293 0.460 0 1 Years Count 263 4.840 3.617 0 12 PRGF 256 0.555 0.498 0 1

3.5.2 Independent Variables

There are three main independent variables identiﬁed by the model of which two are included in the estimation. I also include a number of control variables highlighted in the literature.

The Polity IV score is used to measure democracy (Jaggers & Gurr 1995). The

Polity score is a 20-point scale measure of democracy and made by subtracting the

Autocracy score from the Democracy score, both of which are on a 10-point scale.

Thus, the theoretical minimum of the Polity score is -10, while the theoretical maximum is 10. The average Polity score in the dataset is between three and four; hence, the average country in the dataset is not democratic, judging by the conventional cut

105 point of six or above for democracy. The expected sign is negative, as the hypothesis states that democracies should have fewer conditions than non-democracies. When

I dichotomize the observations into democracy and non-democracy groups, I use the

Polity score of six as the cut point following the suggestion of the authors of the

Polity score (Jaggers & Gurr 1995).

The election data come from the World Bank Database of Political Institutions

(Beck, Clarke, Groﬀ, Keefer & Walsh 2001). The percentage of votes in the 1st or only round of presidential elections is used to proxy the competitiveness of an election. Unfortunately, this variable is only available for presidential democracies.

So, when the variable is included, the number of observations drops signiﬁcantly.

Among all countries included in the dataset, the average amount of first-round votes is 53 percent with a standard deviation of approximately 20 percent. Among all democracies, where this should be more meaningful, the average amount of the first- round vote share of the incumbent government is about 44 percent, with a standard deviation of around 16 percent. The expected sign of the coefficient is positive, as the increase of the share of votes in the previous election should increase the number of public conditions, since the incumbent government can afford to lose a few votes without losing the presidential post.

I use “Years Left in Current Term” variable in the World Bank Database of

Political institutions to capture how proximate the next scheduled election is (Beck et al. 2001). The variable records how many years a chief executive has remaining in his or her current term. As the value increases, the more distant the next scheduled election is, and the more the government will discount electoral consequences of

106 economic reform measures. On the contrary, as the value gets smaller, the sooner the next scheduled election is, and the less the government will discount electoral consequences. Among all countries, the variable varies between zero and six with a mean of 2.19 years. Among all democracies, the variable varies between zero and ﬁve with a mean of 2.05 years. I expect the increase in ”Years Left in Current

Term” should increase the number of public sector conditions, thus yielding a positive coeﬃcient.

In order to proxy the strength and the size of the public sector in a country, I use the public sector salary and compensation as a share of the Gross Domestic Product

(GDP) variable. The public sector salary and compensation as a percentage of GDP is calculated by multiplying the percentage of government spending on the public sector salary and compensation to the government expenditures as a percentage of

GDP. The data come from the World Bank World Development Indicators (World

Bank N.d.). This is a reasonable measure to proxy the strength or size of the public sector as a stronger and larger public sector would increase the share of government spending for the public sector employees salary and benefits. On average, a country spends about five percent of its GDP for public sector employees compensation and salary. The hypothesis suggests that the size of the public sector, measuring the size of the affected domestic interests, should decrease the number of conditions only in non-democracies.

In order to examine the loan size hypothesis, I use the ratio of loan to quota. The size of a loan and the quota data are gathered from the IMF’s oﬃcial website. The ratio of a loan to the quota is calculated by simply dividing the size of a loan with

107 the quota. The average loan arranged is about 90 percent of one’s quota, but there is a very large variation, since some countries have managed to arrange a loan more than ten times larger than their respective quota level. As the ratio increases with larger loans, the potential cost outweighs the beneﬁts for the IMF, or the better a government can compensate the loss to the domestic interests; thus, more conditions are expected. Conversely, as the ratio decreases with smaller loans, the potential beneﬁt outweighs the potential cost to the IMF or a government cannot compensate for the loss to the domestic interests; thus, fewer conditions are expected.

The government expenditure was recovered from 2008 Index of Economic Free- dom (Holmes, Feulner & O’Grady 2008). This comprehensive index allows a short cut for accessing the most comprehensive government expenditure data, since the In- dex of Economic Freedom augments the World Development Indicators of the World

Bank with other reputable data sources. According to the Index, “the scoring of the government size factor” is based on government expenditures as a percentage of GDP, and the following non-linear quadratic cost function is used to calculate

2 the expenditures score GEi = 100 − α ∗ (Expendituresi) where GEi represents the government expenditure score in country i. Expendituresi represents the total amount of government spending at all levels as a portion of GDP, and α is an arbitrary coeﬃcient to control for variation among scores. The minimum component score is zero. The authors of the Index used the following sources for information on the government’s role in the economy, (in order of priority): World Bank World

Development Indicators and “Country at a Glance,” oﬃcial government publications of each country, Country Report and Country Proﬁle by Economists Intelligent

108 Unit, Organisation for Economic Cooperation and Development data, African De- velopment Bank, International Monetary Fund Staﬀ Country Report and selected issues and statistical appendix, Asian Development Bank, and the U.S. Department of Commerce (Holmes, Feulner & O’Grady 2008). To recover the size of government expenditures as a percentage of GDP, I rearranged the government expenditure score with simple math. The average of the resulting Government Expenditure variable is about 28 percent with a standard deviation of 10 percent. The government expenditure is expected to be positively associated with the number of public sector conditions. Assuming the government expenditure is a reasonable proxy for government intervention in the economy, IMF conditions seeking to reduce government intervention should increase as there is more government intervention in the economy.

Indeed, using the composite measure of government consumption and government production, an IMF report ﬁnds that there is a positive relationship between government intervention in the economy and the number of privatization conditions, a part of public sector conditions (Independent Evaluation Oﬃce 2007b).

The similarity of voting patterns in the United Nations General Assembly (UNGA) between a borrowing country and the U.S. is used to control for the inﬂuence of the

U.S. in determining the number of public sector conditions (Voeten & Merdzanovic

N.d.). Dreher and Jensen (Dreher & Jensen 2007) ﬁnd that the closer a country votes with the U.S. in the UNGA, the fewer conditions the IMF program contains.

They reason that the U.S., a country that maintains signiﬁcant inﬂuence over the decision making of the IMF, rewards those countries with similar voting records. If

109 their argument holds, an increase in the similarity of UNGA voting patterns should decrease the number of public sector conditions.

I include a number of other control variables suggested in the literature. The level of economic development is captured by GDP per capita. The data are obtained from the World Bank World Development Indicators (World Bank N.d.). I expect that the richer a country is, the fewer the conditions are included in a program. This is because when a country is economically more developed, existing economic policies of the country should be more sound, all other things being equal. For those more developed countries, there should be fewer conditions that are devised to supposedly

ﬁx problems in existing economic policies.

A country may have more bargaining leverage if the country’s role in the world economy is larger. There is anecdotal evidence that the IMF has been softer toward Russia, Brazil, and Argentina, because collapses of those economies have had a greater impact on the world economy than smaller economies. In this regard, I control for the volume of trade and IMF quota of a country. The IMF quota is based primarily on the size of the economy. As the volume of trade or size of the quota increases, the number of public sector conditions should decrease. The quota data is from the IMF’s oﬃcial website, and the trade data is from the World Bank World

Development Indicators (World Bank N.d., The International Monetary Fund N.d.).

A dummy variable for the Poverty Reduction and Growth Facility (PRGF) lending is included. In general, there are two types of loan facilities in the IMF. The

ﬁrst one is a traditional stand-by agreement for a country with temporary balance of payments problems. The second one was established more recently to cope with

110 development issues of the least developed countries in the world. These loans are provided under the Poverty Reduction and Growth Facility or other development oriented lending facilities. Loans under these facilities tend to have a longer life span and lower interest rate than traditional stand-by agreements. About 55 percent of all programs were arranged under the PRGF or other development-oriented facilities during the 1994-2006 period.

To capture the distinct characteristics of those countries that have made transitions from state command economies to market oriented economies, a dummy variable for transition economies since 1990 is included. The expectation is that transitional economies, those that had been run by centralized authorities, should have more public sector conditions than non-transitional economies, as there has been more to reform than in other participating countries.

Finally, in order to control for the time trend, I include the “Years” and “Years

Squared” variables. It has been reported that there had been a steady increase in the number of structural conditions since the early 1990s, but that the trend has recently reversed, with the introduction of the Fund’s 2002 conditionality streamlining initiative (Independent Evaluation Oﬃce 2007b). The Years variable is a number count from 1994, with 1994 is set to zero. Years Squared is used to capture the decline in the number of structural conditions in more recent years.

As the nature of the main dependent variable is a count variable, I estimate the negative binomial models. I also take the natural log of the dependent variable to transform the dependent variable to the log of the number of public sector conditions

111 and run the OLS regression. The results are consistent with each other, as reported in the next section.

3.6 Results and Discussions

The first column of the tables lists the independent variables. The second column shows the coefficients of all independent variables with clustered standard errors by country in parentheses. The third column reports the coefficients and standard errors from the selection model. The fourth column limits observations to the programs with democratic countries. The fifth column reports the result of the model for presidential democracies.

The results strongly support all the main hypotheses and are consistent across different specifications. Even after controlling for the potential selection effect, the coefficients in the second and third columns are remarkably similar.8 The overall model fit is good, given there are a few outliers driving R2 lower. Exclusion of a few outliers does not significantly change the overall results.

Most importantly, the model strongly supports the democracy hypothesis that democracies are likely to have fewer conditions than autocracies. The coeﬃcient for the Polity score variable is negative and statistically signiﬁcant. The 10-point increase in the Polity score variable from zero (somewhat autocratic) to 10 (democratic) reduces the log of the number of public sector conditions by 0.28. Substantively, this will translate into almost 1.2 fewer conditions around the mean value of the number

8The Heckman selection model result is reported in detail at the end of this chapter. There is little evidence that the selection model and the number of conditions model are dependent.

112 Table 3.2: OLS Regression Model: the Natural Log of the Number of Public Sector Conditions Variable All Countries All (Selection) Democracies Presidential Dem. Polity Score -0.028 -0.029 (0.010)** (0.013)** Years Left in Current Term 0.094 0.110 (Chief Executive) (0.045)** (0.048)** Percentage of Votes 0.008 (0.004)* Government Expenditure 0.018 0.024 0.023 0.020 (% of GDP) (0.005)** (0.006)** (0.009)** (0.011)* Loan to Quota Ratio 0.449 0.443 0.402 0.329 (0.123)** (0.121)** (0.187)** (0.383) Loan Size -0.0002 -0.0002 -0.0002 -0.0006 (SDR Millions) (0.0001)** (0.0001)** (0.0001) (0.0006) Country’s Quota 0.0001 0.0001 0.0003 0.0007 (SDR Millions) (0.0001) (0.0001) (0.0002) (0.0003)** GDP per Capita -0.129 -0.117 -0.133 -0.142 (US$ Thousands) (0.030)** (0.027)** (0.039)** (0.052)** Trade Volume -0.003 -0.003 -0.003 -0.003 (US$ Billions) (0.002) (0.002) (0.002) (0.004) Similarity in UN Voting -0.136 -0.113 -0.416 -0.422 to the U.S (0.200) (0.229) (0.376) (0.481) Transition Country 0.284 0.248 0.525 0.536 (0.138)** (0.143)** (0.251)** (0.329) Poverty & Growth -0.032 -0.057 0.260 0.282 (0.130) (0.149) (0.153)* (0.209) Years 0.150 0.203 0.162 0.181 (0.046)** (0.053)** (0.068)** (0.083)** Years Squared -0.013 -0.020 -0.012 -0.014 (0.004)** (0.005)** (0.006)** (0.007)* Constant 0.506 0.194 -0.356 -0.657 (0.221)** (0.508) (0.297) (0.430) No. of obs. 247 200 123 90 No. of Clusters 89 78 49 37 R2 0.23 0.32 0.35 Notes: **p < .05, *p < 0.1 Clustered standard errors by country are reported in parentheses. 113 of raw public sector conditions. This is roughly equivalent to the eﬀect of a more than

$2,000 increase in GDP per capita of a country or 17 percent decrease in government expenditure as a percentage of GDP.

Consistent with the ﬁnding by the IEO of the IMF, the greater the government expenditure is, the more public sector conditions there are (Independent Evaluation

Oﬃce 2007b). Thus, the number of public sector conditions is partly driven by what the IMF considers desirable and what is considered as an essential part of the Washington consensus. When the government accounts for too much of the country’s economic activities, the IMF program will have more conditions on the public sector to reduce the role of the government in the economy. Holding other variables constant, a 20 percent increase in government spending would increase the number of public sector conditions by 1.5 conditions around the mean of the number of conditions.

As the ratio of the size of a loan to the quota increases, more conditions are included. With the ratio variable in the model, now the size of a loan variable turns negative and statistically signiﬁcant. Yet, the substantive eﬀect of the size of a loan variable is fairly small compared to the ratio variable. Interpreted together, the increase in the size of a loan certainly increases the number of public sector conditions, as suggested by the theoretical model presented in the second chapter.

As an example, suppose there is a country with a quota size of SDR9 500 million.

500 million SDR is about the average quota size among the countries in the dataset.

9SDR stands Special Drawing Rights and works as if it is a currency of the IMF. See more at http://www.imf.org/external/np/exr/facts/sdr.htm

114 If the country’s loan size increases from 500 million SDR to a billion SDR, the net eﬀect of the increase in the size of a loan calculated based on the loan to quota ratio variable and the loan size variable would be an increase of 0.35 in the log of the number of public sector conditions. This would translate into 1.5 more public sector conditions around the mean of the number of public sector conditions.

Transition countries have more public sector conditions in their IMF programs, and the eﬀect is statistically and substantively signiﬁcant. If a country is a transition country, it is likely to have 1.1 more public sector conditions than non-transition countries, holding other variables constant and around the mean of the dependent variable.

Contrary to the previous study by Dreher and Jensen (2007), after controlling for domestic political variables and the economic policy circumstances, “the similarity of

United Nations General Assembly voting to the U.S.” does not reduce the number of public sector conditions. If anything, the signs of the coefficients tend to be negative and statistically not significant, suggesting that the closer a country’s foreign policy position is to the U.S. does not in fact affect the number of public sector conditions.

There is still the possibility that the U.S.’s strategic interest has some inﬂuence on the overall number of IMF conditions, as suggested in Dreher and Jensen (2007), but not particularly over the number of public sector conditions.

The result highlights the advantage of investigating the number of conditions by targeted economic sectors. One cannot find that different economic and political factors have heterogeneous effects over different targeted sectors without disaggregating

115 the number of conditions by targeted sectors. To see exactly how economic and political factors influence the number of conditions across different targeted sectors, one needs to look at different sets of IMF conditions by their targeted sectors.10

There is a very strong time trend. Both the Years and Years Squared variables turn statistically signiﬁcant at the conventional level. Substantively, an average program signed in 1994 would have had 3.5 public sector conditions, which is approximately two conditions less than an average program signed in 1999, which would have had 5.4 public sector conditions. Yet, the trend is non-linear, since an average program signed in 2004 would have 4.3 public sector conditions. The ﬁnding supports the speculation that there had been an upward trend of the number of conditions until about 2002, when mounting criticisms against increasing conditionality triggered the IMF to take the streamlining initiative.

In order to examine the effects of the timing of the next election and the margin of the previous election, I estimate models with two different specifications. The

ﬁrst one includes only the Years Left in Current Term variable and the second model includes both the Years Left in Current Term and Percentage of Votes variables.

The ﬁrst model has a considerably greater number of observations, as inclusion of

Percentage of Votes variable drops all parliamentary democracies.

The election discount hypothesis is strongly supported. The Years Left in Current

Term variable is statistically signiﬁcant in both models (reported in the rightmost

10The fifth chapter investigates the number of financial sector conditions. To preview the result, the similarity of UNGA voting to the U.S. has statistically significant negative effect on the number of financial sector conditions, suggesting that the closer the voting record of a country to the U.S., the fewer financial sector conditions a program contains.

116 two columns of the table.) Substantively, a country with one more year left in the current term receives an approximately 0.2 smaller log of the number of conditions than a country with three more years left in the current term. The substantial effect of two years’ difference in years left is comparable to about a 10 percent change in the government expenditure as a percentage of GDP. I argue that this is because governments expecting more proximate elections discount vote losses less and, thus are forced to extract more lenient deals from the IMF. In comparison, governments expecting more distant elections can disregard potential vote losses due to extensive reform measures and, thus can have more room for their own reform-mindedness to influence the design of their IMF programs. This finding is consistent with the recent study by Dreher, Strum, and Vreeland (2009). The explanation they give is that the number of conditions following elections tends to be greater, because stricter IMF conditionality is needed following expansionary fiscal policies during the run-up to elections or because negotiators of new governments are less experienced.

The margin of victory in the previous election bears a statistically significant and strong substantive effect on the number of public sector conditions. If the government is elected with 60 percent of the votes, either in the only round or in the final round, the IMF program with the country will have approximately one more condition than a government elected with 40 percent of the votes. This is again because those who are elected by slimmer margins are likely to be more sensitive to vote losses caused by additional reform measures than the ones elected by a wider margin in the previous election.

These results are very robust to the inclusion and exclusion of various other

117 control variables not reported here. The results are also very consistent with respect to diﬀerent estimation techniques. The next table reports the results with negative binomial estimations.

I report the results from the negative binomial models with the same variables included in the OLS regression models listed in the leftmost column. The results are very consistent with the results from OLS regression models. Among diﬀerent count variable estimation techniques, the data analysis indicates that the dependent variable is overdispersed with α > 0; thus, I use the negative binomial model (Long

1997).

The Polity score variable is statistically signiﬁcant at the conventional level. Sub- stantively, a 10-point increase in the Polity score is equivalent to about a $2,000 increase in the GDP per capita. Further substantive eﬀect is shown in Figure 3.3.

I calculate the predicted number of public sector conditions with varying levels of democracy, while holding other variables at their medians using “CLARIFY” (Tomz,

Wittenberg & King N.d., King, Tomz & Wittenberg 2000).

The predicted number of public sector conditions follows the solid line sandwiched between two dotted lines indicating 95 percent confidence intervals. Overall, the effect of democracy is substantial. Moving from a perfect democracy (Polity score of 10) to a typical autocracy (a score of zero) increases the number of public sector conditions from 3.5 conditions to five. Moving from a typical autocracy to a complete autocracy increases the number of public sector conditions even more dramatically by raising the number to seven. Thus, holding other variables constant at their medians, changing the regime type from a perfect democracy to a complete

118 Table 3.3: Negative Binomial Model: Number of Public Sector Conditions Variable All Countries Democracies Presidential Dem. Polity Score -0.036 (0.015)** Years Left in Current Term 0.087 0.083 (Chief Executive) (0.067) (0.064) Percentage of Votes 0.008 (0.005)* Government Expenditure 0.026 0.035 0.029 (% of GDP) (0.008)** (0.013)** (0.013)** Loan to Quota Ratio 0.525 0.615 0.483 (0.183)** (0.303)** (0.500) Loan Size -0.0003 -0.0004 -0.0011 (SDR Millions) (0.0001)** (0.0002)* (0.0008) Country’s Quota 0.0002 0.0005 0.0012 (SDR Millions) (0.0002) (0.0003)* (0.0004)** GDP per Capita -0.173 -0.240 -0.224 (US$ Thousands) (0.059)** (0.074)** (0.081)** Trade Volume -0.001 -0.002 -0.002 (US$ Billions) (0.006) (0.005) (0.009) Similarity in UN Voting -0.144 -0.312 -0.232 to the U.S (0.287) (0.641) (0.727) Transition Country 0.254 0.563 0.561 (0.207) (0.402) (0.444) Poverty & Growth -0.127 0.250 0.349 (0.186) (0.225) (0.314) Years 0.213 0.210 0.194 (0.063)** (0.094)** (0.110)* Years Squared -0.017 -0.013 -0.012 (0.005)** (0.008)* (0.009) Constant 0.236 -0.954 -1.146 (0.301) (0.540)* (0.718) No. of obs. 247 123 90 No. of Clusters 89 49 37 Prob.> χ2 0.00 0.00 0.00 α 0.679 (0.113) 0.793 (0.184) 0.688 (0.164) Notes: **p < .05, *p < 0.1 Clustered standard errors by country are reported in parentheses. 119 Figure 3.3: Predicted Eﬀect of Regime Type on Public Sector Conditions

autocracy increases the number of public sector conditions twofold. Given that the mean number of public sector conditions is 3.9, the substantive eﬀect of the polity score is very signiﬁcant.

In the democracy only model, with the variables only available for democracies, the percentage of votes in the previous election for the incumbent has the expected sign and is statistically significant. The predicted number of public sector conditions is reported in Figure 3.4. As the number of observations drops from 247 to 90 when the variable Percentage of Votes is included, the confidence interval is relatively wide. Since it is generated only with the observations of democratic governments, 120 Figure 3.4: Predicted Effect of Earned Vote % on Public Sector Conditions

the number of public sector conditions stays very low. Substantively, the increase of vote share in the prior election from 30 percent to 50 percent increases the number of conditions from 1.75 to 2.4. With another 20 percent increase in the vote share, from 50 percent to 70 percent, the number of public sector conditions increases to three. Thus, the percentage of votes that an incumbent government claimed in the previous election exerts a statistically signiﬁcant and quite substantial eﬀect on the number of public sector conditions.

Other variables that are statistically significant in the OLS results are all statistically significant in the negative binomial results as well. Government expenditure 121 increases the number of public sector conditions and so does the size of a loan. As a government intervention into the economy becomes stronger, there are more public sector conditions in a program. When a larger loan is arranged, the number of public sector conditions tends to increase, holding other variables constant. More developed countries have fewer public sector conditions in their IMF programs. Lastly, there is a statistically significant and substantively strong non-linear time trend. The number of public sector conditions increases from when the data starts in 1994, peaks around the early 2000s, then starts to drop from that point onward. The similarity of a country’s UNGA voting to that of the U.S. again does not have any statistically significant effect on the number of public sector conditions.

The noticeable differences between the OLS and the negative binomial models are that the Years Left in the Current Term is not statistically significant in the negative binomial models, and that the transition country dummy variable is not statistically significant in the negative binomial models.

In order to test the domestic interests hypothesis and compare it with or without the boundary condition, I run OLS regression and negative binomial models with split samples by dichotomized regime type. Countries with a Polity score of six or above are coded as democracy and countries with a Polity score lower than six are coded as autocracy. The public sector hypothesis states that the strength of the public sector should decrease the number of public sector conditions for autocracies but should have little eﬀect for democracies.

The results are reported in Table 3.4.The second and third columns from the left are OLS and negative binomial estimation results for democracies, and the fourth

122 Table 3.4: Democracy and Autocracy Comparison: Public Sector Conditions Democracy Autocracy Variable OLS Neg. Binomial OLS Neg. Binomial Pub. Sec. Strength 0.029 0.039 -0.028 -0.032 (0.022) (0.033) (0.015)* (0.020)* Government Expenditure 0.023 0.035 0.011 0.017 (% of GDP) (0.009)** (0.015)** (0.008) (0.009)* Loan to Quota Ratio 0.283 0.365 0.310 0.431 (0.144)* (0.214)* (0.415) (0.548) Loan Size -0.0001 -0.0002 -0.0003 -0.0004 (SDR Millions) (0.0001) (0.0002) (0.0002) (0.0003) Country’s Quota 0.0003 0.0005 -0.0002 -0.0002 (SDR Millions) (0.0003) (0.0003)* (0.0005) (0.0006) GDP per Capita -0.154 -0.285 -0.128 -0.103 (US$ Thousands) (0.038)** (0.074)** (0.131) (0.130) Trade Volume -0.003 -0.003 0.010 0.015 (US$ Billions) (0.002) (0.005) (0.033) 0.036 Similarity in UN Voting -0.238 0.114 -0.158 -0.519 to the U.S (0.350) (0.554) (0.234) (0.290)* Transition Country 0.473 0.455 0.061 0.006 (0.244)* (0.360) (0.136) (0.157) Poverty & Growth 0.185 0.122 -0.061 -0.007 (0.164) (0.229) (0.238) (0.287) Years 0.181 0.259 0.172 0.196 (0.075)** (0.110)** (0.073)** (0.089)** Years Squared -0.013 -0.015 -0.022 -0.027 (0.006)** (0.008)* (0.007)** (0.010)** Constant -0.305 -0.955 1.239 1.005 (0.337) (0.633) (0.386)** (0.403)** No. of obs. 114 114 88 88 No. of Clusters 46 46 38 38 R2 0.30 0.28 Notes: **p < .05, *p < 0.1 Clustered standard errors by country are reported in parentheses.

123 and ﬁfth columns from the left are OLS and negative binomial estimation results for autocracies. Most importantly, the proposed hypothesis is statistically supported in the models. The performance of the models is generally solid, given a few outliers in the dataset. Excluding outliers from the data does not substantively change the estimation results.

In democracies, the strength of the public sector, measured by the share of the

GDP for compensation and salary for public sector employees, does not have a statistically significant effect. The coefficients are positive in both OLS and negative binomial estimations and fail to reach the conventional statistical significance level.

The coeﬃcients for other variables are very similar to the general results discussed above. Government expenditure generally increases the number of public sector conditions, and an increase in the size of a loan increases the number of public sector conditions. More developed countries have fewer public sector conditions, and there is a very strong time trend.

In autocracies, the strength of the public sector indeed reduces the number of public sector conditions. In both OLS and negative binomial models, the stronger the public sector is, the fewer public sector conditions a program contains. The results are statistically significant at the conventional significance level. A five percent increase in the share of the GDP for public sector compensation is approximately equivalent to an increase of $1,000 in the GDP per capita. The predicted number of public sector conditions is illustrated in Figure 3.5. As the public sector compensation variable varies from about one to 19, with the mean around 5.5, I vary the variable from one to 12 and calculate the predicted number of public sector conditions. The middle

124 Figure 3.5: Predicted Eﬀect of Public Sector Size on Public Sector Conditions in Autocracies

solid line indicates the predicted number of public sector conditions, and the upper and lower dotted lines indicate the 95 percent conﬁdence interval. As the public sector strength increases from one to six, the number of public sector conditions in autocracies decreases, from around seven to 5.5. If the public sector becomes very strong, with about 12 percent of the country’s GDP spent on public sector wages and compensations, then the number of public sector conditions further drops to 4.5.

Together with the null eﬀect of the strength of the public sector in democracies, the

125 strong substantive eﬀect of the strength of the public sector in autocracies strongly supports the domestic interests hypothesis generated from the model.

The result discussed here present an interesting comparison with the recent study by Anner, Caraway, and Richard (Anner, Caraway & Rickard 2010). Examining labor-related conditions in their paper, they report that countries with strong labor see fewer conditions, and the effect is more pronounced among democracies. There are a few potential explanations to why there is some difference between their analysis to mine. Among them, I suspect that the lack of control variables capturing existing policy environment in their analysis might cause the difference.

IMF programs are not randomly chosen, and some worry that the non-randomness of IMF programs may bias the overall results reported above. In order to control for the selection eﬀect, I utilize the Heckman selection model and report the results here

(Heckman 1979, Steinwand & Stone 2008). The participation equation is estimated with essential variables highlighted in the literature, and the number of public sector conditions equation is estimated with the OLS regression, with the natural log of the number of public sector conditions as the dependent variable. The results for the number of public sector conditions are very consistent with the general results reported above. In the participation equation, I ﬁnd that the similarity in the UNGA voting to that of the U.S. is a very strong predictor of the IMF program participation.

When a country votes in line with the U.S. most of time, it drastically increases the probability for the country of receiving a loan from the IMF. Among various economic indicators, only the debt service variable becomes signiﬁcant. When a country faces lots of debt burden, the country is more likely to participate in the IMF program.

126 Table 3.5: Heckman Selection Model: the Natural Log of the Number of Public Sector Conditions Variable # of Conditions Variable IMF Participation Polity Score -0.029 Polity Score 0.012 (0.013)** (0.011) Similarity in UN Voting -0.113 Similarity in UN Voting 2.178 to the U.S (0.229) to the U.S. (0.517)** GDP per Capita -0.117 GDP per Capita -0.049 (US$ Thousands) (0.027)** (US$ Thousands) (0.057) Government Expenditure 0.024 Foreign Reserve -0.021 (% of GDP) (0.006)** Months of Import 0.022 Loan to Quota Ratio 0.443 Debt Service 0.017 (0.121)** Months of Export (0.005)** Loan Size -0.0002 GDP Growth -0.007 (SDR Millions) (0.0001)** (0.008) Country’s Quota 0.0001 (SDR Millions) (0.0001) Trade Volume -0.003 (US$ Billions) (0.002) Transition Country 0.248 (0.143)** Poverty & Growth -0.057 (0.149) Years 0.203 (0.053)** Years Squared -0.020 (0.005)** Constant 0.194 Constant -1.736 (0.508) (0.233)** Number of Obs. 200 Number of Obs. 1075 Wald χ2 110.50 Prob.> χ2 0.000 ρ 0.0155 Prob.> χ2(ρ = 0) 0.725 Notes: **p < .05, *p < 0.1 Clustered standard errors by country are reported in parentheses.

127 3.7 Conclusion

This chapter asks why some IMF programs contain more public sector conditions than others. It presents the hypothesis from the model of IMF program design and implementation introduced in the second chapter and tests the hypotheses with the original dataset of IMF conditionality.

The ﬁndings from the statistical analyses suggest that domestic politics in the borrowing country do matter in the design of public sector conditions included in

IMF programs. All the hypotheses generated from the model are strongly empirically supported. Most importantly, democratic countries have fewer conditions than non-democratic countries. I argue that this is because there exists stronger electoral constraint in democracies. The electoral pressure often forces democratic governments to stay with the minimum number of conditions that the IMF would accept.

This ﬁnding is further bolstered by the other ﬁndings. Among democracies, when governments are elected with slimmer margins in previous elections or when they are facing an immediate next election, they have fewer public sector conditions than governments that are elected with wider vote margins or that were just elected, and hence face later future elections. Again, I contend that this is because of the electoral pressure being stronger in the slimmer margin or imminent election case. In the absence of electoral pressure, governments in non-democracies cannot try to reduce the number of conditions the same way as governments in democracies do. Yet, strong domestic interests that can resist and potentially interrupt the implementation of policy reform measures can reduce the number of conditions in non-democracies. The hypothesis is again empirically supported in the analyses. In comparison, strong

128 domestic interests do not have any eﬀect in democracies, as they are essentially redundant: a democratic government is already forced to agree on the minimum acceptable conditions to the IMF.

This chapter has shown that studying domestic politics of a borrowing country is important and useful in studying the political economy of IMF lending. Ignor- ing the domestic dimension might cause an omitted variable bias and, hence cause false inferences in other variables emphasized in the IMF literature. Indeed, variables highlighted by previous studies often fail to reach the conventional statistical signiﬁcance level.

In the next chapter, the same set of hypotheses are tested in a diﬀerent context

— fiscal policy conditions. As is the case with public sector conditions, fiscal policy conditions are politically unpopular, electorally costly, and often are faced by strong opposition from those who are directly affected by the conditions. In comparison to the public sector conditions case, opposition to fiscal sector conditions might be broader, and include the general public.

In the last empirical chapter of the dissertation, I will visit ﬁnancial sector conditions. Financial sector conditions have been especially highlighted since the 1997

Asian Financial Crisis and have become a very important part of a number of IMF programs. In contrast to the standard assumption that an additional condition is opposed by those who are affected by the condition, the existing literature on financial sector interests in developing countries points out that financial sector interests often support further financial liberalization targeted by financial sector conditions. Given

129 this, I modify the main theoretical model and generate a simple set of hypotheses and empirically test them with the IMF conditionality dataset.

130 CHAPTER 4

IMF PROGRAMS AND FISCAL REFORMS

4.1 Introduction

On March 3, 2010, the Greek government announced a list of austerity measures designed to help Greece close its budget gap amidst economic trouble. These austerity measures cover a wide array of taxation and government spending policy changes that will put a strain on the majority of Greeks. The list included a hike in the value-added tax from 19 to 21 percent, the introduction of an one-oﬀ corporate tax, two percent supplemental cigarette tax, supplemental electricity tax, an increase of tax rates on fuel projects which will spike the gasoline prices, and tax on luxury goods, including vacation homes and oversized properties, yachts, and luxury cars.

The measures also include a 30 percent reduction in holiday bonuses, a freeze on state pensions, a seven percent reduction in beneﬁts for public sector employees including teachers, a reduction of pension subsidies, ﬁve percent reduction of government spending on public works projects, and a 200 million Euro cut in education spending

(Wall Street Journal 2010, Foreign Policy Blog Passport 2010).

These austerity measures were a desperate attempt of the Greek government to

131 strike a bargain with the European Union and the IMF. Unless the Greek government could show it was committed to these measures, the IMF and the European Union would not extend their ﬁnancial resources to the Greek government. Two months after the announcement of the measures and rounds of negotiations for tailoring the measures, the European Union and IMF agreed to an 110 billion Euro ﬁnancing plan with Greece (The International Monetary Fund 2010). The announcement by the

IMF states:

With the budget deﬁcit at 13.6 percent of GDP and public debt at 115

percent in 2009, adjustment is a matter of extreme urgency to avoid the

debt spiraling further out of control. Accordingly, the Greek government

plans to implement rigorous ﬁscal measures, far-reaching structural poli-

cies, and ﬁnancial sector reforms. Key elements of the reform package

are:

• Fiscal policies: Fiscal consolidation — on top of adjustment already

under way — will total 11 percent of GDP over three years, with

the adjustment designed to get the general government deﬁcit under

the 3 percent level by 2014 (compared with 13.6 percent in 2009).

• Government spending: Spending measures will yield savings of 5.25

percent of GDP through 2013. Pensions and wages will be reduced

and frozen for three years, with payment of Christmas, Easter, and

summer bonuses for workers abolished, but with protection for the

lowest-paid.

132 • Government revenues. Revenues measures will yield 4 percent of

GDP through 2013 by raising value-added tax, and taxes on luxury

items, and tobacco and alcohol, among other items (The Interna-

tional Monetary Fund 2010).

While these austerity measures pass the scrutiny of the Greek parliament, the implementation of them would not be easy. Indeed, the domestic response to the announced measures was not favorable to the deal. The austerity measures were immediately faced by strong opposition from various domestic interests, including public sector workers and labor unions. On May 4, 2010, the New York Times reported:

The Socialist government of Prime Minister George Papandreou on Sun-

day announced belt-tightening measures intended to save $39 billion, over

the next three years. The plan is part of an eﬀort to clear the way for a

110 billion Euro rescue package from the European Union and the Inter-

national Monetary Fund aimed at preventing the country from defaulting

on its debt. The measures, including freezes in public sector salaries, cuts

in pensions and higher sales taxes, amount to a cultural revolution in the

social contract between state and citizen.

Hundreds of demonstrators took to the streets in Greece on Tuesday, un-

furling banners over the Acropolis to rail against new austerity measures

aimed at helping the debt-ridden country stave oﬀ economic disaster. ...

Earlier in the day, dozens of protesters from the Communist Party broke

133 the locks at the entrance to the Acropolis, the country’s most famous

tourist attraction, and hung banners saying: “Peoples of Europe - Rise

Up.” ... Public sector workers, including teachers and hospital employ-

ees, began striking Tuesday. A nationwide general strike planned for

Wednesday aims to bring services like public transport to a halt across

the country.

The latest case of IMF involvement in the economic adjustment process in Greece nicely illustrates the domestic politics of economic adjustment led by the IMF. In the middle of an economic crisis, a government often seeks external ﬁnancing resources, and often it is often the IMF that can step in as a “lender of last resort.” Yet, IMF

ﬁnancial resources do not come without strings. A government has to design an IMF program hand-in-hand with the IMF staﬀ to make structural changes to its existing economic policies believed to have caused the economic crisis. The economic program designed by a government and the IMF is often met by strong domestic opposition.

Such domestic opposition makes it hard for the government to implement the agreed policy reforms.

Fiscal policy reforms, often referred to as austerity measures, are among the most criticized policy measures in many IMF programs. These measures are highly unpopular domestically, because they often aim to reduce a government’s budget deﬁcit by simultaneously imposing more taxes and reducing government spending by cutting government subsidies, and reducing the number of salaries and beneﬁts for public sector employees, and the amount spent on healthcare, social safety, and education. These measures are often heavily criticized both by policy experts and 134 people from streets in borrowing countries, because they impose the cost of economic adjustment squarely on domestic constituents, covering a majority of the general public, at the time of economic hardship.

This chapter investigates how these fiscal sector conditions are designed when a country negotiates conditionality with the IMF. The chapter begins with a discussion of fiscal conditions included in IMF programs. The following section briefly reviews the hypotheses generated in Chapter 2. The third section presents the empirical analysis, reporting the results from statistical analyses of the dataset. I show that, among others, democratic countries have fewer fiscal sector conditions than their non-democratic counterparts. In conclusion, I discuss implications of the empirical

ﬁndings.

4.2 Fiscal Reforms in IMF Programs

Fiscal sector conditions make up a large portion of the conditions included in a program. Among all programs signed between 1995 and 2000, about 25 percent of the total number of conditions were on tax policy and public expenditure management.

From 2001 to 2004, the percentage increased to 37 percent, a substantial jump.

Poverty Reduction and Growth Facility (PRGF) programs, only available to the poorest countries in the world, have slightly more ﬁscal conditions. From 1995 to

2000, the proportion of ﬁscal sector conditions was PRGF programs is 29 percent, four percent more than the average in all programs. Between 2001 and 2004, the percentage increased to 40 percent (Independent Evaluation Oﬃce 2007b).

IMF conditionality relating to the ﬁscal sector includes broader measures that

135 affect the general public. These measures aim to increase government revenue while decreasing government spending. For instance, the 2003 Dominican Republic standby agreement included quite a number of fiscal sector conditions, such as specific measures to reduce public debt, like gradually increasing the primary surplus to 3.5 percent of the GDP facilitated by a number of fiscal structural reforms. Asset sales would contribute to achieving this objective. A number of revenue measures with an annual yield of about two percent of the GDP were also introduced, including a doubling of the airport exit tax, an adjustment in the fuel tax, and additional temporary tax increases, such as a two percent import surcharge, 0.15 percent tax on bank checks, and a temporary tax on windfall gains. The program also includes measures to restrain fiscal expenditure, and imposes strict limits on transfers and purchases of goods and services. Low priority capital projects were cut in order to contain public investment outlays at about five percent of the GDP, and the public sector minimum wage was adjusted, with a moratorium placed on additional wage increases in order to enhance fiscal revenue (Dominican Republic 2003).

Armenia’s 1995 program also included many speciﬁc measures targeting tax and spending policies. To enhance revenue performance, the enterprise ﬁxed asset tax was replaced with a comprehensive property tax on enterprises and on households beginning January 1, 1996. In 1995, the government intended to rationalize its expenditure program. Total expenditure and net lending was projected to decline from

44 percent of the GDP in 1994 to 26 percent in 1995 and then to 23 percent in 1996.

During the ﬁrst four months of 1995, the government had strictly limited budgetary expenditure, and in the second quarter intended to implement further cuts in other

136 goods and services and, to domestically ﬁnance capital expenditure, healthcare and education. By these measures, Armenia expected to save 1.5 billion dram, relative to its revised budget for 1995. Further, on June 1, 1995, most remaining domestic subsidies were to be eliminated, notably on bread, garbage collection, district heating, tap water, and hot water, saving a total of 2.2 billion dram during the remainder of 1995. In addition, the implementation of a major and wide-ranging program of expenditure reduction was planned and was expected to yield savings of

8 billion dram for the second half of 1995. Areas in which the Armenian government expected to yield the largest savings were low priority domestically ﬁnanced capital expenditure, defense spending, targeted child allowances, a limitation of the increase in state reserves, and non-essential health and education outlays (Armenia 1995).

Although their presence is close to universal across all IMF programs, fiscal sector conditions show a large variation. The average number of fiscal sector conditions is approximately 3.5, with a standard deviation of 3.4. There are few programs with zero fiscal sector conditions, while there are some program with more than

15 fiscal sector conditions. So, what explains the variation in the number of fiscal sector conditions? How does domestic politics influence the number of fiscal sector conditions?

4.3 Hypotheses

There are a few domestic political factors that are highlighted by the model in

Chapter 2. All the following factors are hypothesized to shape the design of an

137 IMF program: how sensitive a government is to vote losses, how reform-oriented a government is, and how strong domestic interests are.1

Firstly, the model produces the conjecture that the more a government is sensitive to electoral costs, that is, losing votes, the fewer conditions its IMF program will contain. This is because each additional reform measure brings vote losses, holding the amount of the loan constant. From the sensitivity to electoral costs parameter,

I derive three diﬀerent hypotheses that are empirically testable.

• Democracy Hypothesis: The more democratic a country is, the more lenient

ﬁscal sector conditions will be in an IMF program.

• Election Discount Hypothesis: (Within democracy) The more proximate the

next election is, the more lenient ﬁscal sector conditions will be in an IMF

program.

• Margin of Victory Hypothesis: (Within democracy) The more competitive an

election is in a country, the more lenient ﬁscal sector conditions will be in an

IMF program.

As election outcomes are uncertain in democratic countries and each austerity measure is electorally costly, democratic countries should have conditions that are less severe than those of their autocratic counterparts.2 Assuming the government

1The hypotheses assume that there is mutually acceptable policy space where the IMF approves and the special interest group acquiesces. Formally, this would mean that s < H.

2Note that the assumption of electoral costliness of an additional reform measure does not mean that there is no one who would support the additional reform measure. There could very well be 138 wants to stay in power, the government in a democratic country can little aﬀord additional reform measures. Hence, the electoral sensitivity of the government in the democracy forces the government to stay with no more than the minimum number of conditions acceptable to the IMF.

Similar logic applies to explain the variation of the severity of IMF conditionality within democracies. Within democracies, the sensitivity to electoral costs can vary, as some governments can afford to lose a certain number of votes. For instance, a government with a huge margin of victory in the previous election may feel more secure in pursuing electorally costly fiscal policy reforms than a government that won the previous election by a slim margin. A government that won the previous election with a 20 percent margin can afford to lose up to 10 percent of the overall vote share and still remain in power in the next election. In contrast, a government that won the previous election with only a two percent margin cannot enjoy such flexibility in adopting potentially politically costly reform policies. Pursuing austerity measures similar to the government having won the previous election with a 20 percent margin would bring a loss in the next election.

Similarly, a government that recently won an election, hence does not expect another election for a few years, would be better positioned to pursue policies that it desires than a government that expects a more proximate next election. A government that faces an election sooner will discount the next election and electoral costs of policy reforms less than a government that does not face an election in the

those who want more policy reforms, as long as the number of those who oppose the measure is greater than the number of those who support the measure.

139 next few years. Since an additional austerity measure is electorally costly, a government immediately facing an election is less likely to pursue electorally costly ﬁscal reform measures and more likely to bargain hard to reduce the severity of ﬁscal policy conditions than a government with a distant election.

The model also conjectures that the size of the loan has an eﬀect on the severity of

IMF conditionality. In both the Electorally Constrained Equilibrium and the Reform

Drive Equilibrium, the model predicts that the greater the economic beneﬁt brought by signing an IMF program, the more policy conditions a government can aﬀord.

Under the Electorally Constrained Equilibrium, the minimum number of conditions acceptable to the IMF increases as the IMF lends a bigger loan, because the IMF demands more conditions as the size of a loan increases. Since the Electorally Con- strained Equilibrium follows the minimum possible number of conditions accepted by the IMF, the increase in the minimum caused by the larger amount of the IMF loan should also increase the severity of ﬁscal policy conditions. In other words, bigger loans raise the reservation point of the IMF. This reasoning is consistent with the public choice argument with regard to the relationship between the size of a loan and the number of conditions (Dreher 2004b).

Under the Reform Drive Equilibrium, the same conjecture is derived for a diﬀer- ent reason. Under the Reform Drive Equilibrium, the equilibrium conditionality is the maximum number of conditions that do not trigger resistance by the domestic interests in the implementation stage. As the government can reduce the incentives for the domestic interests to resist by increasing the economic beneﬁt from a loan,

140 the maximum possible number of conditions will increase as the size of a loan gets larger.

Thus, both the governments with and without electoral pressure will have more stringent conditions when the size of the loan is larger. One is because the IMF demands more conditions and the other is because the domestic interests become more tolerant and less likely to resist when the size of the loan is larger. While this may sound intuitive, it contrasts to the argument based on U.S. inﬂuence. The argument based on U.S. inﬂuence conjectures that closer allies of the U.S. or other major sovereign principals of the IMF should be treated favorably with fewer conditions and larger loans. It predicts a negative correlation between the number of conditions and the size of a loan.

The size of a loan parameter then yields the following hypothesis;

• Loan Size Hypothesis: As the size of a loan increases, the more severe ﬁscal

sector conditions an IMF program will contain.

In addition to the hypotheses highlighted in the model, there are other arguments that I test in this chapter. First of all, if the IMF keeps to its mandated organizational goal of fixing existing economic policies, the gap between the existing fiscal policies and the desired fiscal policies should explain a large portion of the variation in the number of fiscal sector conditions. Thus, the farther away the existing fiscal policies are from the so-called Washington consensus, the more they should warrant more austerity conditions. Second, the influence of the major shareholders of the IMF in

IMF lending process is highlighted in the IMF literature. I test these propositions

141 along with the domestic politics hypotheses that I propose in the following empirical analysis.

4.4 Empirical Analysis

4.4.1 Dependent Variable

The number of ﬁscal sector conditions in a program is used as the main dependent variable, capturing the severity or leniency of IMF conditionality. The number of

ﬁscal sector conditions data comes from the IMF conditionality dataset introduced in

Chapter 3. While the number of conditions does not perfectly capture the severity or leniency of IMF conditionality, the number of conditions is the most objective measure of IMF conditionality, given information and resources limitations. It is extremely diﬃcult to objectively measure the relative severity of conditions across different countries, because macroeconomic and economic policy contexts vary widely.

It is also widely accepted in the literature as the measure of the stringency of IMF conditions (Copelovitch 2008, Gould 2003, Dreher & Jensen 2007, Dreher 2004a).

The raw number of fiscal sector conditions is used for the count model analysis. In order to check THE robustness of the empirical analysis, I also test the hypotheses against a slightly different measure of fiscal sector conditions — binding fiscal sector conditions. The binding sector conditions exclude structural benchmarks that are counted as soft conditions in the sense that not meeting one of these benchmarks does not automatically interrupt the program.

The distribution of the number of ﬁscal sector conditions and binding ﬁscal sector

142 Figure 4.1: Distribution of the Number of Fiscal Sector Conditions

conditions are shown in Figure 4.1 and 4.2, respectively. First thing to note is that both measures show large variations. The average number of fiscal sector conditions is 3.55, with a standard deviation of 3.4. The average number of binding fiscal sector conditions is about the half of the average of fiscal sector conditions at 1.87, with a standard deviation of 2.4. About 23 percent of the IMF programs signed between

1994 and 2006 have zero ﬁscal sector conditions, while a few have more than 10.

143 Figure 4.2: Distribution of the Number of Binding Fiscal Sector Conditions

4.4.2 Independent Variables

The Polity IV score, used to measure democracy, is used as the main independent variable. The Ratio of a Loan to the Quota variable is used to examine the loan size hypothesis. I also include a number of control variables highlighted in the literature.

The following table summarizes the independent variables used in the analysis.

The Polity score is a 20-point scale measure of democracy and made by subtracting the Autocracy score from the Democracy score, both of which are on a

10-point scale (Jaggers & Gurr 1995). The theoretical minimum score is −10, while the theoretical maximum is 10. The average of the Policy score is between three 144 Table 4.1: Summary Statistics: Chapter 4 Variable Mean Std. Dev. N Polity Score 3.351 5.436 262 Fiscal Freedom 69.709 13.589 262 Government Expenditure 72.67634 20.02103 262 Quota 486.922 836.86 263 Similarity of UN Voting 0.135 0.328 262 Years 4.84 3.617 263 PRGF 0.555 0.498 256 GNP per Capita 1.47 1.675 260 Loan to Quota Ratio 0.896 1.418 263 Transition 0.293 0.456 263 Trade Volume 15627.314 35739.295 257 Loan Size 829394.954 2881428.147 263

and four, hence the average country in the dataset is not democratic, judging by the conventional cut point of six or above for democracy. The expected sign is negative, as the hypothesis states that democracies should have fewer conditions than non-democracies. When I dichotomize the observations into democracy and non-democracy groups, I use the Polity score of six as the cut point, following the suggestion of the authors of the Polity score (Jaggers & Gurr 1995).

I control for the existing policy by including two measures of ﬁscal policy environment. The ﬁrst one is Fiscal Freedom, which is included in the Index of Economic

Freedom. Fiscal Freedom is “a direct measure of the extent to which individuals and businesses are permitted by government to keep and control their income and wealth for their own beneﬁt and use (Holmes, Feulner & O’Grady 2008).” In the

145 Index of Economic Freedom, “the burden of these taxes is captured by measuring total government revenues from all forms of taxation — including all indirect taxes such as payroll, sales, and excise taxes, tariﬀs, and the value-added tax (VAT) as a percentage of total GDP (Holmes, Feulner & O’Grady 2008).”

Figure 4.3: Fiscal Freedom and Fiscal Sector Conditions

Figure 4.3 and 4.4 show the scatter plots of the number of ﬁscal sector conditions and the Fiscal Freedom score and the number of binding ﬁscal sector conditions and the Fiscal Freedom score. One can observe a weak negative correlation in each

146 Figure 4.4: Fiscal Freedom and Binding Fiscal Sector Conditions

scatter plot, suggesting that the better the Fiscal Freedom score, the fewer ﬁscal policy conditions are included in an IMF program.

In addition to the Fiscal Freedom score that captures the soundness of a taxation system, I also control for Government Expenditure. The expectation is that the more a government intervenes in the economy, the more ﬁscal sector conditions there should be. The Government Expenditure score is also from the Index of Economic

Ereedom (Holmes, Feulner & O’Grady 2008). The Government Expenditure score is calculated such that the bigger the score is, the less government intervention there is. Figure 4.5 and 4.6 show the scatter plots of the number of ﬁscal sector

147 conditions and the Government Expenditure score and the number of binding ﬁscal sector conditions and the Government Expenditure score. As it is the case in the

Fiscal Freedom score, it seems that there exists a weak negative correlation between the two measures, suggesting that the better the Government Expenditure score, the fewer ﬁscal policy conditions are included in an IMF program.

Figure 4.5: Government Expenditure and Fiscal Sector Conditions

In order to examine the loan size hypothesis, I use The Ratio of a Loan to the

Quota. The size of a loan and the quota data are gathered from the IMF’s oﬃcial

148 Figure 4.6: Government Expenditure and Binding Fiscal Sector Conditions

website. The ratio of a loan to the quota is calculated by simply dividing the size of a loan by the quota. The average loan arranged is about 90 percent of one’s quota but there is a very large variation , because some countries manage to arrange a loan more than ten times larger than their respective quota level. As the ratio increases with larger loans, the potential cost outweighs the beneﬁts for the IMF, or the better a government can compensate the loss to the domestic interests, and thus more conditions are expected. Conversely, as the ratio decreases with smaller loans, the potential beneﬁt outweighs the potential cost to the IMF, or a government

149 cannot compensate the loss to the domestic interests, and thus the fewer conditions are expected.

In addition to the variables described above, I include a number of other control variables suggested in the literature; they are identical to those in Chapter 3. As the nature of the main dependent variable is a count variable, I estimate the Negative

Binomial models. The results are consistent with each other, as reported in the next section.

4.5 Results and Discussions

The first column of Table 4.2 lists the independent variables. The second column shows the coefficients of all independent variables, with clustered standard errors by country in parentheses and the number of fiscal sector conditions as the dependent variable. The third column reports the coefficients and standard errors of the estimation, with the number of binding fiscal sector conditions as the dependent variable.

The results strongly support the main hypotheses and are consistent across different speciﬁcations. Controlling for other variables highlighted in the literature, I

find strong support for the democracy hypothesis. The coefficient for the Polity score variable is negative and statistically significant. In order to show the substantive effect of the Polity variable, I calculate the predicted number of fiscal sector conditions using the number of fiscal sector conditions model.

The predicted number of ﬁscal sector conditions follows the solid line sandwiched between two dotted lines indicating 95 percent conﬁdence intervals. Overall, the

150 Table 4.2: Negative Binomial Model: Number of (Binding) Fiscal Sector Conditions

Variable Fiscal Conditions Binding Fiscal Conditions Democracy -0.034*** -0.036*** (0.011) (0.014) Fiscal Freedom 0.002 -0.012** (0.005) (0.006) Gov Expenditure -0.013*** -0.005 (0.003) (0.004) Loan to Quota Ratio 0.488*** 0.638*** (0.133) (0.215) Quota 0.001*** 0.001*** (0.000) (0.000) Loan Size -0.000** -0.000** (0.000) (0.000) Similarity of UN Voting 0.005 -0.435 to the United States (0.216) (0.289) Years 0.145** 0.131* (0.066) (0.072) Years Squared -0.001 0.002 (0.005) (0.006) PRGF 0.102 0.016 (0.145) (0.190) GNP per Capita -0.161** -0.212** ($1,000) (0.068) (0.088) Transition 0.243* 0.550*** (0.142) (0.191) Trade -0.000** -0.000 (0.000) (0.000) Constant 1.196*** 0.770 (0.406) (0.554) N 247 247 Notes: ***p < .01, **p < .05, *p < 0.1 Clustered standard errors in parentheses.

151 Figure 4.7: Predicted Eﬀect of Regime Type on Fiscal Sector Conditions

eﬀect of democracy is substantial. Moving from a perfect democracy (Polity score

10) to a typical autocracy (zero) increases the number of public sector conditions from fewer than 2.5 conditions to more than three. Moving from a typical autocracy to a complete autocracy (-10) increases the number of public sector conditions even more dramatically by raising the number to 4.5. Thus, holding other variables constant at their medians, changing the regime type from a perfect democracy to a complete autocracy increases the number of public sector conditions twofold. Given that the mean number of public sector conditions is 3.5, the substantive effect of the polity score is quite substantial. 152 Consistent with the finding by the Independent Evaluation Office of the IMF, the greater the government expenditure is, the more fiscal sector conditions there are (Independent Evaluation Office 2007b). However, the quality of the taxation system measured by the Fiscal Freedom score does not seem to affect the number of

fiscal sector conditions. If existing fiscal policy matters, the number of fiscal sector conditions is mostly driven by existing government expenditure policies. That is, when the government accounts for too much of the country’s economic activities, the

IMF program would have more ﬁscal conditions to reduce the role of the government in the economy.

Again, the predicted number of fiscal sector conditions follows the solid line sandwiched between two dotted lines indicating 95 percent confidence intervals. Overall, the effect of the existing Government Expenditure score is impressive. Moving from minimal government spending (score 100) to the mean Government Expenditure score (72) increases the number of fiscal sector conditions from fewer than two conditions to 2.5. When government intervention becomes more substantial, with a score of 40, the number of predicted fiscal conditions increase to more than four.

As the Ratio of the Size of a Loan to the Quota increases, the greater the number of conditions included as hypothesized. The relationship is statistically signiﬁcant.

With the ratio variable in the model, the size of a loan variable turns negative and statistically signiﬁcant. Yet, the substantive eﬀect of the size of the loan variable is fairly small compared to the ratio variable. Interpreted together, the increase of the size of a loan certainly increases the number of public sector conditions, as suggested by the theoretical model presented in Chapter 2.

153 Figure 4.8: Predicted Eﬀect of Government Expenditure Score on Fiscal Sector Con- ditions

The predicted number of public sector conditions is indicated by the solid line sandwiched by two dotted lines indicating 95 percent confidence intervals. Overall, the effect of the size of a loan is large. As the loan size increases from 25 percent of one’s quota to 100 percent, the number of fiscal sector conditions increases from two to 2.5. When the size of the loan increases to twice of the size of one’s quota, the number of fiscal sector conditions increases to around five. Thus, the larger the loan is, the greater the number of fiscal sector conditions included.

Transition countries have more ﬁscal sector conditions in their IMF programs, and

154 Figure 4.9: Predicted Eﬀect of Loan Size on Fiscal Sector Conditions

the eﬀect is statistically and substantively signiﬁcant. The result is fairly intuitive, given that all transition countries had centrally planned economies before making transitions.

More advanced economies have fewer conditions than poorer countries. Likely, this is because more advanced economies have better overall existing economic policies. Alternatively, it could be because more advanced economies have better resources and bargaining leverage against the IMF in negotiations.

Contrary to the previous study by Dreher and Jensen (2007), after controlling for domestic political variables and economic policy circumstances, “the similarity of 155 United Nations General Assembly voting to the U.S.” does not reduce the number of fiscal sector conditions. There is the possibility that while the U.S.’s strategic interest has some influence on the overall number of IMF conditions as suggested in Dreher and Jensen (2007), but not over the number of fiscal sector conditions specifically.

The result again highlights the advantage of investigating the number of conditions by targeted economic sectors. One cannot find that different economic and political factors have heterogeneous effects over different targeted sectors without disaggregating the number of conditions by targeted sectors. To see exactly how economic and political factors influence the number of conditions across different targeted sectors, one needs to look at different sets of IMF conditions by their targeted sectors.3

There is a very strong time trend. Both the Years and Years Squared variables turn statistically significant at the conventional level. The finding suggests that there is an increasing trend in the number of fiscal sector conditions. Compared to the public sector findings, the finding here suggests that the number of fiscal sector conditions does not trail off and decrease.

These results are very robust to the inclusion and exclusion of various other control variables not reported here. The results are also very consistent with respect to diﬀerent estimation techniques not reported here.

3The fifth chapter investigates the number of financial sector conditions. To preview the result, the similarity of UNGA voting to the U.S. has statistically significant negative effect on the number of financial sector conditions, suggesting that the closer the voting record of a country to the U.S., the fewer financial sector conditions a program contains.

156 In order to examine the election discount and margin of victory hypotheses, I include the Percentage of Votes in the ﬁrst or only round of presidential elections and the Years Left in Current Term variables. Table 4.3 reports the results with negative binomial estimations for only democracies, as these variables are only available and meaningful in democratic countries.

The election data come from the World Bank Database of Political Institutions

(DPI) (Beck et al. 2001). The Percentage of Votes in the first or only round of presidential elections is used to proxy the competitiveness of an election. Unfortunately, this variable is only available for presidential democracies. So, when the variable is included, the number of observations drops significantly. Among all countries included in the dataset, the average of first round votes is 53 percent, with a standard deviation of approximately 20 percent. Among all democracies, where this should be more meaningful, the average of the first round vote share of the incumbent government is about 44 percent, with a standard deviation of around 16 percent. The expected sign of the coefficient is positive, as the increase of the share of votes in the previous election should increase the number of public conditions. This is because the incumbent government can afford to lose a few votes without losing the presidential post.

I use the Years Left in Current Term variable in the World Bank DPI to capture how proximate the next scheduled election is (Beck et al. 2001). The variable records how many remaining years a chief executive has in the current term. As the value increases, the more distant the next scheduled election is and the more the government will discount electoral consequences of economic reform measures. On the contrary,

157 as the value gets smaller, the sooner the next scheduled election is and the less the government will discount electoral consequences. Among all countries, the variable varies between zero and six with mean of 2.19 years. Among all democracies, the variable varies between zero and ﬁve with mean of 2.05 years. I expect the increase in

Years Left in Current Term should increase the number of public sector conditions, thus yielding a positive coeﬃcient.

The election variables are not statistically significant. Both Percentage of Votes and Years Left in Current Term are not significant and switch signs across different model specifications. Thus, in comparison to the public sector conditions, where election related variables are statistically supported, fiscal sector conditions seem not to be influenced by election cycles and the margin of victory. The result is not definite, however, as the number of cases included in the democracy only models are fairly small.

4.6 Conclusion

This chapter asks why some IMF programs contain more ﬁscal sector conditions than others. It presents the hypothesis from the model of IMF program design and implementation introduced in the second chapter and tests the hypotheses with the original dataset of IMF conditionality.

The findings from the statistical analyses suggest that domestic politics in the borrowing country do matter in the design of fiscal sector conditions included in IMF programs. Most importantly, democratic countries have fewer fiscal sector conditions than non-democratic countries. I argue that this is because there exists stronger

158 Table 4.3: Negative Binomial Model: Number of (Binding) Fiscal Sector Conditions (Democracy Only)

Variable Fiscal Conditions Binding Fiscal Conditions Percentage of Votes -0.001 -0.001 (0.006) (0.008) Years Left in Current Term 0.007 -0.124 (0.070) (0.081) Gov. Expenditure -0.021*** -0.027*** (0.006) (0.007) Fiscal Freedom 0.013 -0.003 (0.011) (0.009) Quota 0.001* 0.002*** (0.001) (0.001) Similarity of UN Voting 0.142 -0.310 to the United States (0.349) (0.468) Years 0.548*** 0.468** (0.122) (0.229) Years Squared -0.027*** -0.016 (0.010) (0.015) PRGF 0.331 0.167 (0.263) (0.304) GNP per Capita -0.306*** -0.355*** ($1,000) (0.068) (0.061) Loan to Quota Ratio 0.495* 0.865** (0.289) (0.339) Transition 0.497* 1.038*** (0.256) (0.289) Trade -0.000* -0.000** (0.000) (0.000) Loan Size -0.000 -0.000 (0.000) (0.000) Constant -0.663 0.240 (0.795) (0.769) N 90 90 Notes: ***p < .01, **p < .05, *p < 0.1 Clustered standard errors in parentheses.

159 electoral constraint in democracies. The domestic political pressure often forces democratic governments to stay with the minimum number of conditions that the

IMF would accept. In addition, I ﬁnd that the larger the loan arranged, the more

ﬁscal conditions included in the IMF program. I contend that this is because the

IMF wants to have more conditions when the loan gets riskier. From the domestic politics perspective, this would also suggest that a larger loan helps a government to implement more conditions, as it can compensate some of those who are negatively affected by economic adjustment. Finally, I find that more government intervention in the economy increases the number of fiscal sector conditions.

This finding becomes particularly interesting when it is linked to the recent study of the effect of IMF programs on public spending. Nooruddin and Simmons (2006) show that “IMF programs do cause reductions in social spending, and, perhaps counterintuitively, that this effect is particularly pronounced in democratic countries.” Taking the finding in this chapter and the finding of Nooruddin and Simmons

(2006) together, this might suggest that democracies have fewer fiscal conditions for economic adjustment included in their IMF programs, yet the effect is greater. The missing link here is compliance. That is, when democracies have fewer fiscal sector conditions, they can implement them more successfully, resulting in the reduction of social spending by governments. In comparison, non-democracies have more fiscal conditions that they promise to implement when they negotiate IMF programs but fail to implement them, resulting in little reduction of social spending. Though this is a cautious speculation, the linkage between the design of an IMF program and its effect merits further research.

160 In the last empirical chapter of the dissertation, I visit ﬁnancial sector conditions.

Financial sector conditions have been of particular relevance since the 1997 Asian

Financial Crisis and have become a very important part of an IMF program. In contrast to the standard assumption that additional conditions are opposed by those who are affected by them, the existing literature on financial sector interests in developing countries points out that financial sector interests often support further

ﬁnancial liberalization targeted by ﬁnancial sector conditions. Given this, I modify the main theoretical model and generate a simple set of hypotheses and empirically test them with the IMF conditionality dataset.

161 CHAPTER 5

FINANCIAL SECTOR REFORMS IN IMF PROGRAMS

5.1 Introduction

Why do some countries have numerous ﬁnancial sector conditions included in their

IMF programs while others have relatively few? In this chapter, I provide an answer to the question by deriving hypotheses from the revised model that simplifies the theoretical model presented in the second chapter. The most significant modification of the model is the change of a crucial assumption with regard to the preference of the domestic interests over policy reforms. In the financial sector model, I assume that domestic financial interests, even when they are well-developed, do not oppose financial sector reforms toward financial sector liberalization. The modified assumption is based on evidence in the literature on the financial sector interests in developing countries.1 In addition, in many developing countries, very few private

financial interests exist. Given the assumption of weak or reform-favoring financial sector interests, the model predicts that the size of the loan should increase the number of financial sector conditions. The model also indicates that in the absence

1For a nice review of the literature on the ﬁnancial sector interest, see Pepinski (Pepinski 2009b).

162 of meaningful domestic interests that are adversely affected by conditions and that are willing to resist proper implementation of conditions, domestic political variables exert little influence over the design of financial sector conditions.

Empirical analyses support the hypothesis that countries receiving larger loans from the IMF are assigned more financial sector conditions than countries receiving smaller loans, holding other variables constant. In contrast to public and fiscal sector conditions cases, I find that domestic political variables do not significantly affects the decision on the number of financial sector conditions. Finally, the analysis shows that international political factors, such as the influence of the U.S., emphasized in the existing literature, are indeed important determinants of the number of financial sector conditions.

The Asian financial crisis and the recent financial crises throughout the globe underline the importance of having and maintaining a well functioning and tightly regulated financial sector. The IMF has actively promoted financial sector reforms in developing countries. Financial sector conditions first started to appear in IMF programs in the early 1990s and have increasingly been used in IMF programs since then.

The focus on financial sector reform was especially salient during the Asian Financial crisis in 1997, when weak and government-controlled financial sectors were blamed as the central cause of the crisis (Haggard 2000, Noble & Ravenhill 2000). The emphasis on financial sector reform is further justified by research studies that link financial sector development and economic growth. According to research by the World

Bank staﬀ and other development economists, there is not only a correlation between ﬁnancial sector development and economic development, but there also exists

163 a causal link between ﬁnancial development and economic growth (Levine 1997, Cull

& Eﬀron 2008).

While financial sector conditions have developed rapidly and quickly assumed a major part in economic reform packages that the IMF and borrowing countries negotiate, few studies have examined what determines the extent of financial sector conditions. On the one hand, extant studies examining determinants of IMF conditions take an aggregate approach to explain the variation of the number of all IMF conditions, and thus cannot account for what determines conditions on a specific segment of the economy. As argued in the previous chapters, this runs the risk of preventing the examination of different factors exerting influences heterogeneously over different economic sectors. On the other hand, as far as the financial sector is concerned, studies tend to focus on the effects of IMF programs without address- ing how financial sector conditions are included in IMF programs in the first place.

Since not all financial reform packages are designed the same, ignoring the design of financial sector conditions may prevent researchers from properly diagnosing the actual effects of IMF programs.

This chapter aims to present an explanation of what determines the number of

financial sector conditions and begins with a discussion of them. Following the discussion on financial sector conditions and policy preferences of those who are going to be affected by them, a revised model is introduced, and a set of hypotheses from the model is summarized. The revised model emphasizes the role of the size of a loan in determining the number of financial sector conditions. Then the research design, including the dependent variable, the main independent and control variables, and

164 estimation techniques is introduced. In the following section, statistical results are presented and substantively interpreted. Compared to the analyses of public and

fiscal sector conditions, the empirical analysis of financial sector conditions reveals that the effects of domestic political institutions and other domestic political factors are quite muted in designing financial sector conditions. I argue that this is because there exist few domestic interests that can and are willing to electorally mobilize to hurt reform-initiating governments or can and are willing to hinder proper implementation of agreed upon financial reforms. In their stead, the size of a loan and other international political variables become significant in determining the number of financial sector conditions.

5.2 Financial Sector Reforms in Borrowing Countries

5.2.1 Financial Sector Conditions

According to an IMF report produced by the Independent Evaluation Oﬃce, more than 20 percent of all structural conditions target ﬁnancial sector reforms (Independent

Evaluation Oﬃce 2007a). Breaking down the data temporally, the report shows that there is a generally increasing trend in the number of ﬁnancial sector conditions.

Among all programs in their analysis, about 18 percent of all structural conditions in IMF programs signed between 1995 and 2000 target the ﬁnancial sector. The percentage of ﬁnancial sector conditions increases to over 25 percent of all conditions in

IMF programs signed between 2001 and 2004 (Independent Evaluation Oﬃce 2007a).

Among 263 IMF programs included in the dataset that I constructed, the average

165 number of ﬁnancial sector conditions is about 2.6 conditions per program and varies from zero to 47. Financial sector conditions are also emphasized in World Bank structural lending, where about 25 percent of all structural adjustment loans disbursed between 1992 and 2003 have conditions promoting ﬁnancial sector reforms.

Broadly defined, financial sector reforms generally aim to liberalize the financial sector. Capital account liberalization reduces government control over inflows and outflows of foreign capital. Other financial sector reform measures include privatization of government-owned banking and non banking financial institutions and tightening supervision and prudential regulations over financial institutions. These measures ultimately aim increase financial openness of a country.

In the aftermath of the Asian financial crisis, Indonesia was pressed to further liberalize its financial sector. The 2000 Indonesian IMF program thus included a few financial sector conditions, such as further privatization of a state-owned share of Bank of Central Asia and the setting of a new governance framework for the

Indonesian Bank Restructuring Agency. The 2001 Romanian IMF program also contained a number of ﬁnancial sector conditions, such as an amendment of the banking laws and privatization of the governmental share of Banca Comerciala Ro- mana (Romanian Commercial Bank) and other state-owned banks. These measures were a part of the continued eﬀort to reduce the government’s role in the banking industry in Romania. Lastly, the 1995 Ghanaian IMF program also included a few conditions requiring privatization of a major share of many state-owned banks, like

Social Security Bank, National Investment Bank, and Ghana Commercial Bank. The

166 privatization of state-owned banks generally increases foreign ownership of domestic banking institutions.

5.2.2 Financial Sectors in Borrowing Countries

Financial Sectors in developing countries, especially those belonging to the least developed countries are very weak and controlled by the government. Even when a developing country has a relatively well-developed financial sector, the financial sector interests tend to support financial sector liberalization. Given the level of development of the financial sectors in the countries participating in IMF programs and the types of typical financial sector conditions in those programs, it is safe to assume that there exist few opposing domestic economic interests involved in designing financial sector conditions.

First of all, almost all of IMF program participating countries are developing countries, and a majority of them belong to the least developed countries category of the United Nations classiﬁcation. In the dataset that I assembled, 70 percent of lMF programs arranged between 1994 and 2006 were signed by countries with a per capita income of $2,000 or below. Among all IMF programs in the dataset, over 55 percent of the programs were disbursed under the Poverty Reduction and Growth

Facility or other growth-related facilities, lending facilities particularly developed to provide ﬁnancial assistance with the nominal interest rate to heavily indebted poor countries.

Financial sectors in these developing countries are far smaller than those in the middle or high income countries. Comparing the ratio of liquid liabilities to the

167 GDP, the broadest available indicator of ﬁnancial sector development, the relative size of ﬁnancial intermediaries to the size of economy of the least developed countries 1 is approximately of that of the developed economies. 2 In 1994, the average ratio 5 of liquid liabilities to the GDP among the least developed countries was about 24 percent and the average ratio among low income countries was about 28 percent.

In comparison, the average ratio of liquid liabilities to GDP among high income countries in 1994 was 106 percent. The ﬁnancial sector data are from the World

Bank World Development Indicators and the New Database on the Structure and

Development of the Financial Sector (World Bank N.d., Beck, Demirguc-Kunt &

Levine 2000).

In addition, banking sectors in developing countries had usually been under tight control by their governments until those governments decided to participate in IMF programs in the 1990s and 2000s. In most of the developing countries, the government often established and owned a major share of domestic banks, as there existed very few domestic capital owners. A majority of the ﬁnancial sector conditions in IMF programs are targeted at reducing the control of governments over their banking institutions.

In middle income developing countries and emerging market economies, ﬁnancial

2“Liquid liabilities are a general indicator of the size of financial intermediaries relative to the size of the economy or an overall measure of financial sector development (World Bank N.d., p.277).” “Liquid liabilities are the sum of currency and deposits in the central bank, plus transferable deposits and electronic currency, plus time and savings deposits, foreign currency transferable deposits, certificates of deposit, and securities repurchase agreements, plus traveller’ checks, foreign currency time deposits, commercial paper, and shares of mutual funds or market funds held by residents. The ratio of liquid liabilities to GDP indicates the relative size of these readily available forms of money , money that the owners can use to buy goods and services without incurring any cost (World Bank N.d., p.277).”

168 sector interests are at least not against capital account and ﬁnancial liberalization, if not actually in favor of it, hence the measures promoting liberalization of their

ﬁnancial sectors.

Existing studies of the financial liberalization of developing countries often report that domestic financial interests often espouse rather than oppose capital account liberalization. Working against the standard the Hecksher-Ohlin model, where relatively capital scarce developing countries’ capital factor should oppose rather than support capital account liberalization, these studies report that financial sector interests actually favor financial liberalization. While the reason for this provides an interesting question, in the context of this dissertation, it is sufficient to note that

ﬁnancial sector interests in developing countries at least do not oppose the reform measures included in IMF programs. “In developing countries and emerging markets from Argentina to Turkey and from Egypt to the Philippines, analysts have found that owners of ﬁnancial capital resist capital control, both in the form of laws that prevent them from moving funds overseas and in the form of restrictions on the in-

ﬂow of foreign capital (Pepinski 2009b, 1).” This observation challenges conventional theories of economic interest formation based on factors of production. Financial sectors in emerging market economies, since they are relatively capital poor by global standards, should be swamped by foreign capital under conditions of full openness, and thus are destined to lose in the competition for domestic lending opportunities.

Yet, domestic ﬁnancial interests do not oppose ﬁnancial liberalization.

There is abundant microeconomic evidence that foreign banks in devel-

oping economies outcompete their domestic counterparts, oﬀering lower 169 interest rates, mobilizing more funds from large depositors, and earn-

ing greater proﬁts than domestic banks. Yet still, in many parts of the

developing world domestic lenders lobby for cross-border ﬁnancial open-

ness, the very policy that standard theories predict would harm their

interests. Scholarship on Indonesia, for example, has long identiﬁed the

interests of ﬁnanciers who for decades pressured the regime to main-

tain an open capital account — one open both to outﬂows and inﬂows

(MacIntyre 1993, Winters 1996). ... Brooks (56) ﬁnds in a sample of Latin

American countries — all emerging markets or lesser developed countries

— that large ﬁnancial sectors lead to more capital account liberalization

(Pepinski 2009b).

Domestic ﬁnancial intermediaries in emerging markets are often uncertain

as to whether they stand to beneﬁt from the increased opportunities for

intermediation that can accompany liberalization or whether they stand

to be harmed from the possibility that liberalization will precipitate a

banking crisis due to the legacies of ﬁnancial repression and poor pru-

dential supervision that typically characterize emerging markets. This

uncertainty often leads interest groups to fall silent when one might ex-

pect them to be critical players (Chwieroth 2007a).

Given the circumstances in financial sectors in developing countries, it is hard to justify that there are comparable domestic interests playing an important role in designing financial sector conditions in the same way that domestic public sector interests and the general public influence the design of public and fiscal sector 170 conditions. Financial sectors in IMF participating countries are usually underdevel- oped and commonly dominated by state-owned banking institutions. Few existing private financial interests would be strong enough to put electoral pressure on the government or to threaten to hinder proper implementation of agreed upon financial sector conditions. Moreover, there are studies that report financial sector interests in developing countries often welcome more financial liberalization, and thus favor policy conditions seeking it. In sum, financial interests in developing countries may actually favor some of the financial sector conditions included in the proposed IMF programs or are at least too weak to exert significant pressure to reduce them.

5.3 The Revised Model

5.3.1 The Model

Figure 5.1: Game Tree: Financial Sector Conditions

171 The important assumption in the model presented in Chapter 2 and subsequently tested in Chapter 3 and 4 needs modification. The weak and reform-favoring financial sector presents starkly different incentives and constraints for a government negotiating financial sector conditions in IMF programs from governments negotiating public sector conditions. Lifting the assumption that an increasing number of conditions is resisted by domestic interests makes the model a simple unitary actor model where the IMF and the government negotiate over financial sector conditions. Assuming that there is little electoral consequence or possibility of domestic interests’ resisting proper implementation of agreed upon financial sector conditions, the game theoretic model presented in Chapter 2 becomes far simpler.

The game now features only two players negotiating the number of ﬁnancial sector conditions — the IMF and the government. The game starts when the government makes a proposal to the IMF. The IMF decides to reject or approve the proposal.

When the IMF accepts, the program is implemented. Compared to the main model presented in Chapter 2, the critical diﬀerence is that once a program is approved by the IMF, the program is implemented without an additional decision nod where domestic interests make a decision to resist. When the IMF rejects the oﬀer, the negotiation ends without a program.

The payoffs for the government and the IMF are fairly simple. Specifically, the payoff for having no program is set to zero for each actors. When there is a program, the government enjoys the benefit from borrowing financial resources from the IMF, denoted Nv, and may have preference over financial sector reforms. When the coef-

ﬁcient γ is positive, the government favors ﬁnancial sector reforms. When γ is zero,

172 the government is reform-neutral. When γ is negative, the government does not like

ﬁnancial sector reforms.

For the IMF, having a program provides generic benefits and costs of the program. The benefits and costs are assumed to be exogenously given. The payoffs are summarized below.

• UG(NA) = 0

p • UG(IMP) =Beneﬁts from the Loan + Beneﬁts from Reforms = Nv + γ ∗ PG

• UF(NA) = 0

p • UF(IMP) =Beneﬁts - Costs of the IMF Program = BF ∗ ( PG) − CF

As the game assumes complete information, the subgame perfect Nash equilibrium is used as the solution concept. Practically, the game can be solved with simple backward induction. The IMF approves any proposal that meets the following incentive condition:

CF 2 PG > ( ) BF Assuming that a government weakly prefers the minimum reform, the equilibrium becomes the following:3

CF 2 PG = ( ) BF

3If a government is reform-oriented, the equilibrium becomes the most preferred policy point for the IMF. Presumably, this would also be determined by the cost-beneﬁt calculation of the IMF. So, regardless of the assumption, the simple comparative statics I present below are likely to hold.

173 Spelled out, the agreed upon policy conditions will correspond with the cost- benefit calculation of the IMF. Specifically, when the cost of the program is great, the equilibrium conditionality is large. Conversely, when the benefit of the program is large, the equilibrium conditionality is small. Varying the cost and the benefit,

I generate the following hypotheses. I assume that the cost of a program increases when the size of a loan increases. I also assume that the beneﬁt of a program increases when a country is a closer friend of the major sovereign principals of the IMF.

5.3.2 Hypotheses

• Loan Size Hypothesis: As the size of a loan increases, the more severe the

conditions an IMF program will contain.

• IMF Beneﬁt Hypothesis: As a borrowing country is a closer friend of the U.S.,

the less severe the conditions an IMF program will contain.

In sum, the decision over the conditionality is purely the function of the cost- beneﬁt calculation of the IMF. When the cost looms higher with larger loans, the

IMF wants to put more conditions in the programs. When the beneﬁt of programs grows larger with U.S. support for the programs, the IMF is willing to reduce the severity of conditions included in the programs.

174 5.4 Empirical Analysis

5.4.1 Dependent Variable

The number of ﬁnancial sector conditions in a program is used to capture the severity or leniency of IMF conditionality. The number of ﬁnancial sector conditions used in the empirical analysis comes from the original IMF conditionality dataset illustrated in Chapter 3.

Figure 5.2: Distribution of the Number of Financial Sector Conditions

Figure 5.2 shows the distribution of the number of financial sector conditions. 175 The distribution of the number of financial sector conditions is right skewed, with many observations falling below 10, while only very few fall above 10. About 46 percent of all programs have fewer than two financial sector conditions, and another

27 percent of all programs have three to four. In comparison, less than 10 percent of IMF programs signed between 1994 and 2006 have more than 10 ﬁnancial sector conditions. The average number is 2.58, with a standard deviation 3.95. There are two outliers with 22 (Ecuador, 2000) and 47 (Dominican Republic, 2005) ﬁnancial sector conditions, respectively.

Figure 5.3: Distribution of the Number of Binding Financial Sector Conditions

176 Table 5.1: Summary Statistics: Chapter 5 Variable # of obs Mean Std. Dev. Min Max Polity Score 262 3.351 5.436 -9 10 Financial Sector Freedom 262 47.634 17.213 10 90 Previous IMF Program 261 0.927 0.260 0 1 Domestic Credit by Banking 253 31.598 27.560 -13.537 177.577 Loan to Quota Ratio 263 0.896 1.418 0.05 12.62 Loan Size (SDR Millions) 263 829.395 2881.428 4 27375 Quota (SDR Millions) 263 486.922 836.860 14.2 5945 GDP per Capita (US$1,000) 260 1.470 1.675 0.09 12.19 Trade Volume (US$ Billions) 257 15.627 35.739 0 292.563 UN Voting with U.S. 262 0.135 0.328 -0.425 1 Transition Country 263 0.293 0.460 0 1 Years Count 263 4.840 3.617 0 12 PRGF 256 0.555 0.498 0 1

The number of binding financial sector conditions, excluding financial sector structural benchmarks, shows a very similar distribution. About 60 percent of all programs have zero binding financial sector condition and another 20 percent of all programs have one. About eight percent of programs have two financial sector conditions and about 10 percent of programs have three to five. Two programs have more than 10. The mean number of binding financial sector conditions is 1.03, with a 2.06 standard deviation.

5.4.2 Independent Variables

In order to examine the loan size hypothesis, I use The Ratio of a Loan to the

Quota. The size of a loan and the quota data are gathered from the IMF’s oﬃcial

177 website. The ratio of a loan to the quota is calculated by simply dividing the size of a loan by the quota. The average loan arranged is about 90 percent of one’s quota but there is a very large variation , because some countries manage to arrange a loan more than ten times larger than their respective quota level. As the ratio increases with larger loans, the potential cost outweighs the beneﬁts for the IMF, or the better a government can compensate the loss to the domestic interests, and thus more conditions are expected. Conversely, as the ratio decreases with smaller loans, the potential beneﬁt outweighs the potential cost to the IMF, or a government cannot compensate the loss to the domestic interests, and thus the fewer conditions are expected.

The similarity of United Nations General Assembly (UNGA) voting of a borrowing country to the U.S. is used to control for the inﬂuence of the U.S. in determining the number of public sector conditions (Voeten & Merdzanovic N.d.). Dreher and

Jensen (Dreher & Jensen 2007) find that the closer a country votes with the U.S. in the UNGA, the fewer conditions the IMF program contains. They reason that the U.S., who maintains significant influence over the decision making of the IMF, rewards those countries voting closer to its votes. Thus, if their argument holds, an increase in the similarity of UNGA voting should decrease the number of public sector conditions.

The size of domestic banking sector is included to control for the inﬂuence of

ﬁnancial sector interests, to the extent that it exists. The domestic credit provided by domestic banking sector as a percentage of the GDP is used to measure the size of

178 domestic banking sector. The data comes from the World Bank World Development

Indicators (World Bank N.d.).

To control for the level of ﬁnancial liberalization, I use the “Financial Freedom” variable in the Index of Economic Freedom (Holmes, Feulner & O’Grady 2008).

Financial freedom is “a measure of banking security as well as a measure of inde- pendence from government control. State ownership of banks and other financial institutions such as insurers and capital markets is an inefficient burden that reduces competition and generally lowers the level of available services. The authors score this component by determining the extent of government regulation of financial services; the extent of state intervention in banks and other financial services; the difficulty of opening and operating financial services firms (for both domestic and foreign individuals); and government influence on the allocation of credit. The authors use this analysis to develop a description of the countrys financial climate and assign it an overall score on a scale of 0 to 100 (Holmes, Feulner & O’Grady 2008, p.447).”

The score 100 is given to those countries with negligible government inﬂuence over the ﬁnancial sector. The score 50 is given when there is considerable government in-

fluence, and the score 10 is given when most financial activities are strictly controlled by the government. As the purpose of financial sector conditions is to promote financial liberalization, the Financial Freedom score should strongly predict the number of financial sector conditions. In countries where financial freedom is high, there should be very few financial sector conditions. In countries where financial freedom is low, there should be many financial sector conditions to improve the freedom of

ﬁnancial sector activities. The average Financial Freedom in the dataset is around

179 48, with a standard deviation of 17. The minimum score is 10, and the maximum

90.

I include a number of other control variables suggested in the literature. The set of the control variables is identical to the set used in Chapter 3 and 4. The detailed discussion of the control variables can be found in Chapter 3. As the nature of the main dependent variable is a count variable, I estimate the negative binomial models. I also take the natural log of the dependent variable to transform the dependent variable to the log of the number of public sector conditions and run the

OLS regression. The results are consistent with each other as reported in the next section. Only the results from the negative binomial models are reported.

5.5 Results and Discussions

I report the results from the negative binomial models in Table 5.2. The first column of the tables lists the independent variables. The second column shows the coefficients of all independent variables with clustered standard errors by country in parentheses, with the number of financial sector conditions as the dependent variable. The third column shows the coefficients of all independent variables, with the number of binding

ﬁnancial sector conditions as the dependent variable. The results are very consistent with the results from OLS regression models, which are not reported here. Among diﬀerent count variable estimation techniques, the data analysis indicates that the dependent variable is overdispersed with α > 0. Thus, I use the negative binomial model (Long 1997).

180 Table 5.2: Negative Binomial Model: Number of (Binding) Financial Sector Condi- tions Variable Financial Conditions Binding Financial Conditions Polity Score 0.010 0.026 (0.016) (0.031) Financial Freedom -0.010 -0.020 (0.005)** (0.008)** Previous IMF Program -0.708 -0.074 (0.232)** 0.300 Domestic Financial Sector 0.007 0.009 (0.002)** (0.004)** Loan to Quota Ratio 0.764 1.053 (0.224)** (0.263)** Loan Size -0.0003 -0.0005 (SDR Millions) (0.0001)** (0.0001)** Country’s Quota 0.0002 0.0005 (SDR Millions) (0.0002) (0.0002)** GDP per Capita -0.104 -0.359 (US$ Thousands) (0.076) (0.121)** Similarity in UN Voting -0.640 -0.697 to the U.S (0.198)** (0.280)** Transition Country 0.567 0.775 (0.184)** (0.302)** Poverty & Growth -0.438 -0.633 (0.270)* (0.326)* Years 0.295 0.326 (0.102)** (0.116)** Years Squared -0.021 -0.021 (0.008)** (0.010)** Constant 0.874 -0.397 (0.313)** (0.517) No. of obs. 242 242 No. of Clusters 88 88 Prob.> χ2 0.00 0.00 α 0.725 (0.185) 1.351 (0.340) Notes: **p < .05, *p < 0.1 Clustered standard errors by countryreported in parentheses.

181 The analysis reports very interesting results. Compared to the previous two Chap- ters, in which domestic political variables showed significant effects on the number of conditions in the respective sectors, the analysis of financial sector conditions shows that domestic political variables do not affect the design of the conditions. Instead, the existing policy environment and the influence of the U.S. become significant. As predicted from the modified model, the size of a loan exerts much influence that is statistically significant on the number of financial sector conditions.

The more Financial Freedom a country enjoys, the fewer the financial sector conditions. Previous IMF program experience reduces the number of financial sector condtions. The larger the domestic financial sector is, the more financial sector conditions a program contains. Larger loans increase the number of financial sector conditions. The more developed a country is, the fewer conditions the program contains. The closer a country votes with the U.S. in the United Nations, the fewer conditions the country’s IMF program contains. Transition economies tend to have more conditions. Programs arranged under the Poverty Reduction and Growth Fa- cility tend to have fewer conditions. And there is strong time trend.

Highlighting the differences between public and fiscal sector interests and financial sector interests, the model also shows that regime type does not influence the number of financial sector conditions. If at all, the coefficient is positive, suggesting that the more democratic a country is, the more financial sector conditions a program would contain. But the relationship is not statistically significant. Moreover, the larger the existing banking sector is, the more financial sector conditions are included in a program. This confirms the consensus in the literature on the financial liberalization

182 of developing countries that states existing ﬁnancial interests tend to prefer more

ﬁnancial liberalization.

As hypothesized from the model, an increase of the size of loan increases the number of financial sector conditions. Further substantive effect is shown in the following figures. I calculate the predicted number of public sector conditions with varying levels of democracy while holding other variables at their medians using

“CLARIFY (Tomz, Wittenberg & King N.d., King, Tomz & Wittenberg 2000).”

Figure 5.4: Predicted Eﬀect of Loan Size on Financial Sector Conditions

183 The predicted number of financial sector conditions follows the solid line located between the two dotted lines indicating 95 percent confidence intervals. Overall, the effect of the size of a loan is substantial. Moving from a usual loan size at .9 times the quota to 1.5 times the quota, the number of financial sector conditions increases from around two to four. If the size gets to double that of the quota, the number of financial sector conditions further increases to 5.5. When a country gets a smaller loan from the IMF, at 25 percent of its quota, the number of financial sector conditions decreases to 1.5, holding other variables constant at their medians. Thus, the effect of the size of a loan is substantial. This is because the IMF wants more conditions as the potential cost of the program, that is, the cost of default, increases with the size of a loan.

The current Financial Freedom score is a strong predictor of the number of financial sector conditions. When a country’s Financial Freedom is low, the country is more likely to have more financial sector conditions. Conversely, if a country has already liberalized its financial sector, the country has fewer financial sector conditions.

Additionally, previous IMF program participation decreases the number of financial sector conditions. This may be because of the improved financial sector liberalization due to the previous IMF programs. Figure 5.6 illustrates the substantive effect of the Financial Freedom score. The predicted number of financial sector conditions follows the solid line. Overall, the effect of financial freedom score is strong. Moving from a financial sector tightly regulated by a government (Financial Freedom score of 20) to a relatively free financial sector (financial freedom score of 80) halves the predicted number of financial sector conditions from 2.6 to 1.3.

184 Figure 5.5: Predicted Eﬀect of Financial Freedom on Financial Sector Conditions

The predicted number of financial sector conditions follows the solid line between the two dotted lines indicating 95 percent confidence intervals. Overall, the effect of the size of the domestic banking sector is substantial and highlights the differences between the design of financial sector conditions and the design of conditionality over other sectors of economy. Basically, the larger the domestic banking sector is, the more financial conditions are included in an IMF program. Substantively, when a country’s domestic credit provided by the domestic banking sector increases from 20 percent of the GDP to 90 percent, the number of financial sector conditions increases from less than two to close to three conditions. 185 Figure 5.6: Predicted Effect of Financial Sector Strength on Financial Sector Con- ditions

Consistent with the previous study by Dreher and Jensen (2007), after controlling for domestic political variables and the economic policy circumstances, “the similarity of United Nations General Assembly voting to the U.S.” reduces the number of

financial sector conditions. When a country votes with the U.S. more often, the country is likely to have fewer financial sector conditions in its IMF program. Con- versely, when a country votes against the U.S.’s positions more often, the country is likely to have more financial conditions in its IMF program. Specifically, when a country votes along with the U.S. half of the time (score of zero), the country would

186 Figure 5.7: Predicted Eﬀect of UN Voting with the U.S. on Financial Sector Condi- tions

have little less than two conditions. In comparison, when it votes with the U.S. all the time, the number of ﬁnancial sector conditions further decreases to one. When a country votes against the U.S. most of the time and earns -.5 for its voting score

(close to the minimum in the dataset), the number of ﬁnancial sector conditions increases to 2.5, holding other variables at their medians.

187 5.6 Conclusion

This chapter asks why some IMF programs contain more financial sector conditions than others. It presents the hypothesis from the revised, simplified model of IMF program design and tests the hypotheses with the original dataset of IMF conditionality. The model is based on the modified assumption that in financial liberalization, the domestic financial interests tend to prefer more liberalization; hence, adding additional financial sector conditions is not electorally costly nor is it more readily resisted. Under the revised assumption, the model becomes a simple, two-player game, where the IMF and a government, as a unitary actor, negotiate over the design of the financial sector conditionality.

The findings from the statistical analyses suggests that domestic politics in the borrowing country do not matter much in the design of financial sector conditions included in IMF programs. Instead, the size of a loan, as predicted from the model, is an important determinant of the number of financial sector conditions. In addition, the existing policy environment, measured by the financial freedom score, and the size of domestic banking sector significantly influence the number of financial sector conditions. The more liberalized a country’s financial sector is, and the smaller the country’s domestic banking sector is, the fewer the financial sector conditions. Lastly, the influence of the U.S. is significant in the absence of domestic political influence.

I ﬁnd that the closer a country votes with the U.S. in the UNGA, the fewer ﬁnancial sector conditions the country’s program contains.

188 CHAPTER 6

CONCLUSION

International political economy has a bearing on program selection, con-

ditionality, and implementation. Similarly, domestic political economy

shapes attitudes toward the IMF and, hence, the ability of the authori-

ties to enter into constructive IMF-supported programs. A richer political

economy analysis could make for more eﬀective engagement.

Raghuram G. Rajan, Economic Counselor and Director, Research De-

partment, International Monetary Fund, In the Forward of the IMF-

Supported Programs (Mody & Rebucci 2006, v)

The dissertation provides a richer political economy analysis of IMF program design and implementation. The explicit theoretical linkage between the design and implementation serves as the bridge to connect the international political economy of IMF program negotiation and the domestic political economy of economic reforms. By exploring the theoretical linkage between the design and implementation in the IMF program context, the dissertation contributes to both empirical studies of IMF lending and theoretical literature of two-level games in international political economy. 189 This short concluding chapter briefly summarizes the main findings of each Chap- ter. Taking findings of the empirical chapters together, the following section highlights the advantage of the disaggregating approach. I also discuss the policy implications of the dissertation with regard to the evaluation of effects of IMF programs and its relation to the ownership of the program. At the end, I suggest future research directions.

6.1 Summary

Why do some IMF programs contain more policy conditions than others?

In this dissertation, I oﬀer a theoretical account of under what circumstances one of the two alternative theoretical arguments prevails over the other. I then apply the theoretical argument to explain why some IMF programs contain more stringent conditions than others. In doing so, I ﬁrst construct a game theoretic model of IMF conditionality design and implementation in which I allow both of the two competing theoretical arguments to play out. The model captures the dynamic linkage between the deliberate negotiations over conditionality between the IMF and a government and the domestic politics of conditionality implementation, with emphasis given to the role of domestic interests.

190 reform-minded is more likely to extract a more lenient conditionality from the IMF.

A government that is less sensitive to vote losses and more reform-minded is likely to pursue more extensive policy reforms by siding with the IMF, and it is constrained by domestic politics only when there exist sufficiently strong domestic interests that can hinder proper implementation of agreed upon conditions. In contrast, when conditions are favored by affected domestic interests, or affected domestic interests are too weak, the domestic institutions and interests exert little influence over the design of an IMF program.

I test these propositions with an original dataset that I created by coding structural conditions included in all the letters of intent signed between 1994 and 2006.1

Each structural condition is classified and counted by its types and affected economic sector. I disaggregate IMF conditions into four affected economic sectors and analyze them separately. By doing so, I allow that factors relevant to designing IMF conditions over an economic sector vary by sectors. For instance, domestic politics may play a larger role in designing conditions for some economic sectors, while international politics behind the IMF and ideological, organizational, and bureaucratic incentives of the IMF can play a greater role in designing conditions for other sectors.

Among diﬀerent sets of conditions, the Chapter 3 focuses on the eﬀects of the domestic political factors in negotiating public sector conditions. Public sector conditions, including restructuring of state-owned enterprises (SOEs), privatization of public enterprises, and civil service reforms, are one of the main kinds of conditions.

1I focus on structural conditions because 1) the number of quantitative conditions do not vary much across programs and 2) structural conditions are more politically salient as they pursue structural reforms.

191 Among all IMF conditions, 25 to 30 percent of all IMF conditions are public sector conditions. Applying the theoretical implications to the design and implementation of public sector conditions, the Chapter empirically examines how the theoretical predictions perform in explaining their design.

The ﬁndings from the statistical analyses in Chapter 3 suggest that domestic politics in the borrowing country do matter in the design of public sector conditions included in IMF programs. All the hypotheses generated from the model are strongly empirically supported. Most importantly, democratic countries have fewer conditions than non-democratic countries. I argue that this is because there exists stronger electoral constraint in democracies. The electoral pressure often forces democratic governments to stay with the minimum number of conditions that the

IMF would accept. This finding is further bolstered by the other findings. Among democracies, when governments were elected with a slimmer margin in the previous election, or when they are facing an immediate next election, they have fewer public sector conditions than governments that were elected with a wider vote margin or that were just elected, and hence face a later next election. Again, I contend that this is because of the electoral pressure being stronger in the slimmer margin or immediate election cases. In the absence of electoral pressure, governments in non-democracies cannot try to reduce the number of conditions the same way as governments in democracies do. Yet, strong domestic interests that can resist and potentially interrupt implementation of policy reform measures can reduce the number of conditions in non-democracies. The hypothesis is again empirically supported by the analyses. In comparison, strong domestic interests do not have any effect in

192 democracies, as they are essentially redundant: a democratic government is already forced to agree on the minimum acceptable conditions to the IMF.

Chapter 4 looks at fiscal sector conditions, the highly criticized austerity measures. The findings from the statistical analyses confirm the findings in Chapter 3 and further support the argument that domestic politics in the borrowing country do matter in the design of conditionality included in IMF programs. Most importantly, democratic countries have fewer fiscal sector conditions than non-democratic countries. I argue that this is because there exists stronger political constraint in democracies. The domestic political pressure often forces democratic governments to stay with the minimum number of conditions that the IMF would accept. In addition, I find that the larger the loan arranged, the more fiscal conditions are included in the IMF program. I contend that this is because the IMF wants to have more conditions when the loan gets riskier. From the domestic politics perspective, this would also suggest that a larger loan helps a government to implement more conditions, as it can compensate some of those who are negatively affected by economic adjustment. Finally, I find that more government intervention in the economy increases the number of fiscal sector conditions.

In Chapter 5, I provide an answer to how financial sector conditions are designed by deriving hypotheses from a revised model that simplifies the theoretical model presented in the second chapter. The most significant modification of the model is the change of a crucial assumption with regard to the preference of the domestic interests over policy reforms. In the financial sector model, I assume that domestic

ﬁnancial interests, to the extent that they are developed, do not oppose ﬁnancial

193 sector reforms toward capital account liberalization. The modified assumption is based on the evidence in the literature on financial sector interests in developing countries. In other developing countries, very few private financial interests exist.

Given the assumption of weak or reform-favoring financial sector interests, the model predicts that the size of the loan should increase the number of financial sector conditions. The model also indicates that in the absence of meaningful domestic interests that are adversely affected by conditions and that are willing to resist proper implementation of conditions, domestic political variables exert little influence over the design of financial sector conditions.

Empirical analysis supports the hypothesis that countries getting larger loans from the IMF receive more financial sector conditions than countries receiving smaller loans, holding other variables constant. In sharp contrast to public and fiscal sector conditions cases, I find that domestic political variables do not have significant effects in deciding the number of financial sector conditions. Finally, the analysis shows that international political factors, such as the influence of the United States emphasized in the existing literature, are indeed important determinants of the number of

ﬁnancial sector conditions.

6.2 Contributions

6.2.1 Theoretical Contribution

Theoretically, the dissertation provides boundary conditions under which one of the two competing theoretical arguments of the two-level games prevails over the other.

194 Table 6.1: Summary of The Empirical Findings

Variable Public Sector Fiscal Sector Financial Sector Polity Score - - 0 Years Left in Current Term + 0 0 Similarity of UN Voting 0 0 - Years + + + Loan to Quota Ratio + + +

The theoretical model shows that the ability of a government to tie its hands with the IMF to force more reforms on domestic opposition is limited when there are large domestic political consequences. On the other hand, greater domestic political constraints allow the government tie its hands with opposing domestic interests to extract more lenient deals from the IMF in international negotiation. The theoretical argument presented in the dissertation is widely applicable to other international negotiations scenarios where domestic and international politics are intertwined.

6.2.2 Empirical Contribution

On the empirical side, the dissertation clearly shows the advantage of disaggregating

IMF conditions by aﬀected sectors and analyzing them separately. The following table highlights the advantage.

Table 6.1 summarizes the signs of the coefficients for the main independent variables in the dissertation. As one can see, the effect of a variable often varies across different economic sectors. For instance, democracy and election related variables

195 are significant in public sector and fiscal sector conditions cases. More democratic countries have fewer public sector conditions and fiscal sector conditions. In comparison, the regime type of a country does not affect the number of financial sector conditions. Similarly, the influence of the United States is salient and statistically significant in determining financial sector conditions. Closer friends of the United

States receive fewer ﬁnancial sector conditions than those who often disagree with the United States. This relationship is only for ﬁnancial sector conditions, however.

In designing public and ﬁscal sector conditions, the inﬂuence of the United States seems to be muted.

If one takes into account only the number of aggregate IMF conditions, one can misstate the inﬂuence of a particular variable. The following table uses the aggregate number of IMF conditions and binding IMF conditions as dependent variables and estimate with all the independent variables used in the previous empirical analyses.

When one analyze by taking the aggregate number of IMF conditions as the dependent variable, a different conclusion emerges. For instance, one could mistakenly argue that the similarity of a country’s UNGA voting to that of the U.S. reduces the number of overall IMF conditions, when in fact it is only the financial sector conditions that are affected. Similarly, one could also mistakenly argue that democratic countries receive fewer conditions than non-democratic countries, when it is not true for the financial sector conditions case.

In sum, by taking the disaggregating approach, the dissertation enhances our understanding of the IMF program design. When those conditions with large, adverse domestic distributional consequences are negotiated, domestic political institutions

196 Table 6.2: Negative Binomial Model: Number of All (Binding) Conditions Variable All Conditions All Binding Conditions Economic Freedom -0.024*** -0.045*** (0.008) (0.013) Polity Score -0.019* -0.022 (0.011) (0.016) Quota 0.000*** 0.001*** (0.000) (0.000) Similarity of UN Voting -0.244 -0.462** to the United States (0.173) (0.236) Years 0.177*** 0.181*** (0.054) (0.067) Years Squared -0.011*** -0.008 (0.004) (0.005) PRGF -0.190 -0.207 (0.123) (0.155) GNP per Capita -0.084* -0.210*** ($ 1000) (0.047) (0.077) Loan to Quota Ratio 0.616*** 0.815*** (0.117) (0.173) Transition 0.406*** 0.384* (0.126) (0.202) Trade -0.000* -0.000 (0.000) (0.000) Loan Size -0.000*** -0.000*** (0.000) (0.000) Constant 3.240*** 3.358*** (0.411) (0.658) N 247 247 Notes: ***p < .01, **p < .05, *p < 0.1 Clustered standard errors by countryreported in parentheses.

197 and interests weigh in the designing process. When those conditions with little comparable domestic political consequences are negotiated, international politics and organizational incentives of the IMF weigh in more heavily, while domestic political circumstances do not make much diﬀerences.

6.3 Policy Implications

There are a number of policy implications. First of all, the dissertation provides an accurate model of how domestic politics in a borrowing country aﬀects the joint policy design between the IMF and a borrowing government. The dissertation shows that the design of an IMF program is not a one-sided imposition of conditions as many observers claim, but rather is jointly created by the IMF and a government.

Second, the empirical analyses show that the IMF is indeed strategic in negotiating policy conditions by taking domestic political constraints of a borrowing country into account. The findings demonstrate that the IMF pushes a government with few political constraints harder, yet limits its pressure on a government with more vulnerable domestic political circumstances. The IMF seems to take into account the trade-off between the optimal conditions and the implementation possibility, and the strategic behavior of the IMF should be considered when one evaluates effectiveness of IMF programs.

Third, the dissertation indicates that the politics of IMF program design and the politics of IMF program implementation are interrelated; hence the implementation of an IMF program is going to be inﬂuenced by the design of IMF conditionality. This suggests that an evaluation of an IMF program and its subsequent eﬀects on various

198 social, economic, and political conditions should take the design of IMF conditionality seriously. Many recent IMF-related studies explore various effects of IMF programs on borrowing countries. For instance, there are studies examining the effect of IMF programs on the political survival of leaders of borrowing countries or political regime change (Smith & Vreeland 2006, Dreher 2004a, Pepinski 2009a, Remmer 1986), on foreign investment (Edwards 2006, Jensen 2004, Biglaiser & DeRouen Jr 2010), on economic growth (Przeworski & Vreeland 2000, Vreeland 2003, Dreher 2006), and on respect for human rights (Abouharb & Cingranelli 2009). One of the weaknesses of these studies is that they treat all IMF programs homogenously without differenti- ating them. Yet, I have demonstrated, along with others, that not all IMF programs are the same; in fact, they are very different from each other, and domestic and international politics heavily intervenes in every step of IMF lending process. Thus, it is very likely that the design of IMF programs will be closely linked to the implementation of them, which also influences the various effects examined in the current

IMF-related studies. Taking the design of IMF conditions into account will enhance our understanding of various consequences of IMF programs.

Fourth, related to the third point, the lessons from the dissertation related to the recent discussion of IMF program ownership (Bird & Willett 2004, Arpac &

Bird 2009). While program ownership is often considered as politically desirable and may enhance the implementation, the literature recognizes that the concept of ownership is hard to empirically proxy (Bird & Willett 2004). I suggest that the number of structural conditions can actually be a very nice measure of program ownership, as it captures how much policy measures are mandated in an IMF program

199 as opposed to be left to a participating government. This will provide an opportunity for researchers to examine the eﬀect of increased program ownership on various eﬀects of IMF programs.

6.4 Future Directions

A possible future direction for research stemming from this dissertation is to seek to better understand how the design of a program affects the implementation and its social, economic, and political effects.The follow-up research should attempt to make linkage between the design of IMF programs and their effectiveness. This would allow researchers and practitioners to better understand and evaluate the effects of IMF programs, with regard to both participation and potential noncompliance.

Building on my existing research projects on the politics of the IMF conditionality design, I aim to examine how the design of IMF conditionality is linked to the implementation of policy conditions and how design shapes subsequent economic, social, and political consequences, such as economic growth, foreign capital ﬂows, social unrest, and political leadership changes.

As part of the larger research agenda of evaluating effects of IMF programs and their design, I intend to revisit the so-called “catalytic effects of IMF lending” question (Bird & Rowlands 2004, Edwards 2006). In the IMF literature, there is an unre- solved question regarding the effect of IMF program agreement in catalyzing foreign capital flows. Simply put, the question asks whether an IMF program triggers foreign capital flows in a program participating country or not. Generally conceptualizing

200 the signing of an IMF program as a signal to the financial market, theoretical arguments suggest that when a country signs into an IMF program, it should increase the volume of capital flows into the country. Unfortunately, there are few empirical studies that examine the catalytic effect hypothesis, and those few existing empirical studies do not agree with each other.

In the proposed research, I intend to reformulate the catalytic effect hypothesis and empirically investigate the theoretically reformulated hypothesis. I will point out that the signal that the financial market receives is not homogenous across IMF programs, as it is commonly conceptualized in the literature. Rather, the signal that the financial market receives when a country signs into an IMF program varies and the international financial market may respond accordingly. Specifically, financial investors not only look at whether a country signs a program with the IMF or not but also look at with what conditions the country does so. This is a plausible scenario, as policy conditions attached to an IMF program are widely public during and after program negotiations. When financial investors observe an IMF program and the design of IMF conditionality, they can weigh both the scope and depth of proposed policy reforms and the probability of proper implementation of the conditions. Thus, by paying attention to different signals that IMF programs can send, I will be able to identify under what conditions the catalytic effect will be more or less likely.

Another research will look at the effects of sector specific conditions in sector specific reforms. The specific policy reform outcomes should only be affected by the policy measures targeting the specific policy areas. Thus, when one is interested in

fiscal policy reform outcome, one should focus on specific fiscal policy conditions,

201 rather than looking at the whole set of conditions. I have a research project along this line, examining the eﬀect of public sector conditions on the public sector reform in post-Soviet, transition economies.

The dissertation on the IMF program design and the follow-up projects investigating the eﬀect of program design on various political economic indicators, with taking into account the program design explicitly, should be able to make important contributions to the growing literature on the IMF lending and on other international ﬁnancial institutions. The theoretical arguments in the dissertation could be also applied to other empirical contexts in international political economy.

202 Appendix A

PROOF FOR THE CHAPTER 2

A.1 Payoﬀs

A.1.1 Payoﬀ for the IMF

Utility for the IMF not to have an agreement is normalized to 0. When there is an agreement and is implemented without resistance from the special interest group,the

IMF’s beneﬁt is proportional to the conditionality. PG varies from 0 to 1, and 1 is the ideal policy point for the IMF when there is no obstacles for implementation.

On the other hand, 0 is the ideal point for the special interest group as it means no conditions attached. When PG is 1, I further assume that the IMF always prefers IMP over N/A. This should be a reasonable assumption, as it only assumes that the

IMF would prefer to have a program implemented than no agreement when it can assign the ideal conditions for the borrowing country. Thus, the utility of IMP for the Fund increases as the conditions gets closer to the IMF’s ideal point (increases).

A few words on BF and CF. The beneﬁt is proportional to the probability of successful implementation by the expected reward. I assume the expected reward is proportional to the size of the borrowing country’s economy. Whether it comes from

203 organizational interests or backpadding of the principals of the IMF, the beneﬁt for the IMF increases as the IMF lends to and rescues larger economy. On the other hand, the cost is proportional to the probability of failed implementation and the size of the default loan. Thus as the size of the loan increases, holding probability of default IMF loans constant, the cost also increases.

When there is no agreement, the status quo payoff is 0. When there is an agreement and implemented, the payoff is BF by PG minus CF. Finally, when there is resistance of the special interest group(SIGR), the utility for the Fund have two terms; the resistance fails with probability PF, and when the resistance fails the Fund enjoys the implementation payoff. When the resistance is successful with probability

PS, the Fund gets the utility of N/A minus the cost of partial lending.

UIMF(N/A) ≡ U1 = 0 p UIMF(IMP) ≡ U2 = BF PG − CF 1 p 1 U (SIGR) ≡ U = P U +P (−rC ) = sβ − 1 C − B P − rsβC = IMF 3 F 2 S F N F F G N F 1 p 1 sβ − 1 C − B P − rsβC N F F G N F

U1 = 0

p U2 = BF PG − CF

1 p 1 U = sβ − 1 C − B P − rsβC 3 N F F G N F

A.1.2 Payoﬀ for the Domestic Interests

Before normalization

s B1 =i=1 wi 204 s W =i=1 wi = sw W given the average of W, = w, thus W = sw s s p B2 =i=1 (wi(1 − PG) + v)

B3 = PFB2 + PSB1 − scs

After normalization

USIG(N/A) ≡ S1 = B1 − B1 = 0 s s p USIG(IMP) ≡ S2 = B2 − B1 = v − wi PG − 1 − wi i=1 i=1 s X X p p = v − wi PG = sv − sw PG i=1 X s s 1 1 p U (SIGR) ≡ S = B −B = sβ w − sβ − 1 v − w P − 1 − SIG 3 3 1 N i N i G i=1 i=1 s s X Xs p 1 1 p sc − w = sv − P w − sc − s2vβ + sβ P w s i G i s N N G i i=1 i=1 i=1 X X X

S1 = 0

p S2 = sv − sw PG

p 1 1 p S = sv − sw P − sc − s2vβ + s2βw P 3 G s N N G

A.1.3 Payoﬀ for the government

Before normalization probability of reelection (Average for the special interest group: the mean voter, the mean voter when the interest is normally distributed, should be pretty close to the median voter)

205 Probabilistic voting model

There are two groups, J = SIG, Others or simply,

J = S, O

The population share of group J = αJ s N − s αS = , αO = N N voter i in J prefers the incumbent if

J J iJ w (IMF) > w (N/P) + σ + δ 1 1 σS∼ [− , ] 2φS 2φS 1 1 σO∼ [− , ] 2φO 2φO 1 1 σ∼ [− , ] 2φ 2φ 1 1 δ∼ [− , ] 2ψ 2ψ

given the swing voter in group J, σJ

σJ = wJ(IMF) − wJ(N/P) − δ

wJ(IMF) = σJ + wJ(N/P) + δ

Actual vote share of the incumbent is 1 1 Π = αSφS(σS + ) + αOφO(σO + ) incumbent 2φS 2φO the probability of reelection is 1 P = Pr[Π > ] E incumbent 2 1 1 1 = Pr[αSφS(σS + ) + αOφO(σO + ) > ] 2φS 2φO 2 1 = Pr[αSφS(wS(IMF) − wS(N/P) − δ + ) + αOφO(wO(IMF) − wO(N/P) − 2φS 1 1 δ + ) > ] 2φO 2 206 = Pr[αSφS(wS(IMF) − wS(N/P)) + αOφO(wO(IMF) − wO(N/P)) − δ(αSφS + 1 1 1 αOφO) + αSφS + αOφO > ] 2φS 2φO 2 = Pr[αSφS(wS(IMF) − wS(N/P)) + αOφO(wO(IMF) − wO(N/P)) > δ(αSφS +

αOφO)] αSφS(wS(IMF) − wS(N/P)) + αOφO(wO(IMF) − wO(N/P)) = Pr[ > δ] (αSφS + αOφO) 1 ψ = + [αSφS(wS(IMF) − wS(N/P)) + αOφO(wO(IMF) − wO(N/P))] 2 φ 1 ψ s p N − s 1 ψ s p = + [ φ(v − PGw − 0) + φ(v − 0)] = + [ (v − PGw) + 2 φ N N √ 2 1 N N − s 1 s p N 1 −s PGw (v)] = + ψ[ (− PGw) + (v)] = + ψ( + v) N 2 N √ N 2 N 1 −s P w P = P = + ψ( G + v) E 2 N √ 1 −s P w P = + ψ( G + v) 2 N

Utility before normalization 1 A = ( ρ + (1 − ρ))B 1 2 G p A2 = Nv + γ PGW + (ρP + (1 − ρ))BG

A3 = PFA2 + PSA1

After normalization

UG(N/A) ≡ G1 = A1 − A1 = 0 1 p 1 U (IMP) ≡ G = A − A = B ρ ψ v − sw P + − ρ + 1 + G 2 2 1 G N G 2 1 p p 1 p B ρ − 1 + Nv + swγ P = Nv + vψρB + swγ P − swψρB P G 2 G G G N G G

UG(SIGR) ≡ G3 = A3 − A1 1 1 1 p 1 = B ρ − 1 − sβ − 1 B ρ ψ v − sw P + − ρ + 1 G 2 N G N G 2

207 p 1 1 +Nv + swγ PG − sβBG ρ − 1 N 2 √ √ N2v + NvψρB + Nswγ P − swψρB P = (N − sβ) G G G G N2

G1 = 0 √ p PGsw G = Nv + γ P sw + B ρψ(v − ) 2 G G N √ √ N2v + NvψρB + Nswγ P − swψρB P G = (N − sβ) G G G G 3 N2

A.2 Solution

A.2.1 Subgame below the government’s oﬀer

The Game is solved in two steps. First, I identify conditions under which the special interest group resists or acquiesces and/or the IMF accepts or rejects the oﬀer from the government. Then I calculate the optimal oﬀer from the government that will maximize the government’s utility.

A)special interest group resists iﬀ

S3 > S2 p 1 1 p p sv − sc − sw P − s2vβ + s2wβ P > sv − sw P s G N N G G 1 1 p −sc − s2vβ + s2βw P = 0 s N N G 1 Solution is: (Nc + svβ)2 < P s2w2β2 s G 1 P = (Nc + svβ)2 ;P is the cutpoint for SIG above which it resists and I s2w2β2 s I below which it acquiesces. ∂P N I = −2 c (Nc + svβ) < 0 ∂s s3w2β2 s s 208 if PG > PI then special interest group prefers to resist

if PG < PI then special interest group prefers to acquiesce The probability of two are equal is statistically zero.

And as the size of interest group increases, the cutpoint gets smaller. Substan- tively, this conﬁrms the common sensual conjecture that the special interest group is going to resists more readily when it is stronger.

B)IMF acceptance expecting special interest group resist iﬀ

U3 > U1 1 p 1 sβ − 1 C − B P − rsβC > 0 N F F G N F p 1 1 p 1 B P − C + sβC − sβB P − rsβC = 0 F G F N F N F G N F 1 2 2 2 2 2 2 N BF + s β BF − 2NsβBF 2 2 2 2 2 2 2 2 2 2 2 2 2 2 ∗ N CF + s β CF + r s β CF − 2rs β CF − 2NsβCF + 2NrsβCF 2 1 CF 2 = 2 2 (N − sβ + rsβ) BF (N − sβ) 2 1 CF 2 2 2 (N − sβ + rsβ) < PG BF (N − sβ) 2 1 CF 2 PA = 2 2 (N − sβ + rsβ) ; PA is the cutpoint for F above which the BF (N − sβ) IMF agrees in anticipation of SIG resistance while below which the IMF rejects the oﬀer from the government in anticipation of SIG resistance. 2 ∂PA β CF = 2Nr 2 3 (N − sβ + rsβ) > 0 assuming that N − sβ > 0 ∂s BF (N − sβ) If PG > PA, then IMF agrees even with anticipation of SIG resistance

If PG < PA, then IMF does not agree with anticiaption of SIG resistance As the size of the special interest group increases, the IMF becomes more cautious and accepts the oﬀer from the government only if the government promises to

209 implement more conditions. In other words, facing stronger domestic opposition, the

IMF wants the government to compensate the high risk with more conditions.

C)IMF acceptance expecting special interest group acquiesce iﬀ p U2 > U1 = 0 < BF PG − CF p 0 = BF PG − CF

1 2 Solution is: PG > 2 CF BF 1 2 PB = 2 CF; PB is the cutpoint for F above which the IMF agrees in anticipation BF of implementation without SIG resistance while below which the IMF rejects the oﬀer. ∂P B = 0 ∂s This does not vary as the size of the special interest group varies. 1 P = (Nc + svβ)2 I s2w2β2 s 2 1 CF 2 PA = 2 2 (N − sβ + rsβ) BF (N − sβ) 1 2 PB = 2 CF BF When PI meetsPA : s 1 1 C2 (Nc + svβ)2 = F (N − sβ + rsβ)2 2 2 s 2 2 s β BF (N − sβ) 1 L = − ∗ (NCF −2βCF + 2rβCF + 2vβBF q 2 2 2 2 2 2 −N v BF + CF + BFcs + 2vBFcs − 2vBFCF − 2BFCFcs + 4rBFCFcs−NvBF+NBFcs

L = sL

When PI meetsPB : s 1 1 (Nc + svβ)2 = C2 2 2 s 2 F s β BF cs H = NBF βCF − vβBF 210 H = sH

2 2 1 CF 2 1 2 1 CF 2 PA−PB = 2 2 (N − sβ + rsβ) − 2 CF = 2 2 (N − sβ + rsβ) − BF (N − sβ) BF BF (N − sβ) 2 1 2 β CF 2 CF = rs 2 2 (2N − 2sβ + rsβ) > 0 BF BF (N − sβ) There are three theoretically potential cases

1)When N > sH > sL

PIand PA,PIandPBmeet

2) When sH > N > sL

PIand PAmeet

3)When sH > sL > N

No crossing (Not possible) (PAincreases exponentially as sgets very close to N)

Let the ﬁrst cutpoint sLand the second cutpointsH Then there are three intervals.

The ﬁrst interval covers 0 < s < sL

The second interval covers sL < s < sH

The third interval covers sH < s < N

As always PI > PB and PH > PA, two comparative statics are possible here. First of all, when γ increases, all other things equal, the slope increases and is more likely to lead the slope positive, hence more likely to lead to tougher conditionality.

Conversely, when it decreases, all other things equal, the slope decreases and is more likely to lead the slope negative, especially when it itself is negative, hence more lenient conditionality. Secondly, when ρ increases, the slope decreases and is more likely to lead the slope negative, hence more lenient conditionality. When it

211 approaches to 0, the slope increases and is more likely to lead the slope positive, hence tougher conditionality. 1 P = (Nc + svβ)2 I s2w2β2 s ∂P N I = −2 c (Nc + svβ) < 0 ∂s s3w2β2 s s ∂P N I = −2 c (Nc + svβ) < 0 ∂β s2w2β3 s s ∂P N I = (Nc + svβ) > 2 2 2 2 s 0 ∂cs s w β ∂P 2 I = (Nc + svβ) > 0 ∂v sw2β s

PH does not change. These comparative statics holds only when the slope is negative.

1 2 PB = 2 CF BF ∂PB 2 2 = − 3 CF < 0 ∂BF BF ∂PB 2 = 2 CF > 0 ∂CF BF These comparative statics holds only when the slope is positive.

In addition, 2 1 CF 2 PA = 2 2 (N − sβ + rsβ) BF (N − sβ) 2 ∂PA 2 CF 2 = − 3 2 (N − sβ + rsβ) < 0 ∂BF BF (N − sβ) ∂PA 2 CF 2 = 2 2 (N − sβ + rsβ) > 0 ∂CF BF (N − sβ) 2 ∂PA β CF = 2Nr 2 3 (N − sβ + rsβ) > 0 ∂s BF (N − sβ) 2 ∂PA s CF = 2Nr 2 3 (N − sβ + rsβ) > 0 ∂β BF (N − sβ) 2 ∂PA β CF = 2s 2 2 (N − sβ + rsβ) > 0 ∂r BF (N − sβ) Finally, the screen/the selection (not to sign in) occurs when G1 > G2. 212 1 p 0 > G , 0 > sw γ − ψρB P + (Nv + vψρB ) 2 N G G G 2 1 1 (Nv + vψρBG) if γ − ψρBG > 0, Solution is: > PG N s2w2 1 2 γ − N ψρBG ∂P N2 (Nv + vψρB )2 s = − G < 2 3 2 2 0 ∂s s w (Nγ − ψρBG) ∂P N3 (Nv + vψρB )2 s = − G < 2 2 2 3 0 ∂γ s w (Nγ − ψρBG) ∂P N3 v2 γ + 1 s = ψB (N + ψρB ) > 2 2 2 G G 3 0 ∂ρ s w (Nγ − ψρBG)

Thus, as the size increases, the selection is less likely. The more reform minded, the selection is less likely. The more democratic, the selection is more likely. 2 1 1 (Nv + vψρBG) if γ − ψρBG < 0, Solution is: < PG N s2w2 1 2 γ − N ψρBG 2 1 (Nv + vψρBG) Ps = s2w2 1 2 γ − N ψρBG ∂P N2 (Nv + vψρB )2 s = − G < 2 3 2 2 0 ∂s s w (Nγ − ψρBG) ∂P N3 (Nv + vψρB )2 s = − G > 2 2 2 3 0 ∂γ s w (Nγ − ψρBG) ∂P N3 v2 γ + 1 s = ψB (N + ψρB ) < 2 2 2 G G 3 0 ∂ρ s w (Nγ − ψρBG) Thus, as the size increases, the selection is more likely. The more reform minded, the selection is less likely. The more democratic, the selection is more likely.

213 Appendix B

IMF PROGRAMS 1994 — 2006

Country Loan Type Date Year Loan Size Quota

Central African Republic PRGF 12/22/06 2006 36,200 55.70 Mauritania PRGF 12/18/06 2006 16,100 64.40 Haiti PRGF 11/20/06 2006 73,710 81.90 Madagascar PRGF 07/21/06 2006 54,990 122.20 Afghanistan PRGF 06/28/06 2006 81,000 161.90 Rwanda PRGF 06/12/06 2006 8,010 80.10 Paraguay Standby 05/31/06 2006 65,000 99.90 Sierra Leone PRGF 05/10/06 2006 31,110 103.70 Moldova PRGF 05/05/06 2006 110,880 123.20 Grenada PRGF 04/17/06 2006 10,530 11.70 Albania PRGF 02/01/06 2006 8,523 48.70 Iraq Standby 12/23/05 2005 475,360 1,188.40 Cameroon PRGF 10/24/05 2005 18,570 185.70

214 Country Loan Type Date Year Loan Size Quota

Macedonia Standby 08/31/05 2005 51,675 68.90 Benin PRGF 08/05/05 2005 6,190 61.90 Malawi PRGF 08/05/05 2005 38,170 69.40 Sao Tome & Principe PRGF 08/01/05 2005 2,960 7.40 Uruguay Standby 06/08/05 2005 766,250 306.50 Armenia PRGF 05/25/05 2005 23,000 92.00 Turkey Standby 05/11/05 2005 6,662,040 1,191.30 Colombia Standby 05/02/05 2005 405,000 774.00 Kyrgyz Republic PRGF 03/15/05 2005 8,880 88.80 Chad PRGF 02/16/05 2005 25,200 56.00 Dominican Republic Standby 01/31/05 2005 437,800 218.90 Niger PRGF 01/31/05 2005 26,320 65.80 Congo, Republic PRGF 12/06/04 2004 54,990 84.60 Bulgaria Standby 08/06/04 2004 100,000 640.20 Croatia Standby 08/04/04 2004 99,000 365.10 Romania Standby 07/07/04 2004 250,000 1,030.20 Mozambique PRGF 07/06/04 2004 11,360 113.60 Mali PRGF 06/23/04 2004 9,330 93.30 Zambia PRGF 06/16/04 2004 220,095 489.10 Peru Standby 06/09/04 2004 287,279 638.40 Georgia PRGF 06/04/04 2004 98,000 150.30

215 Country Loan Type Date Year Loan Size Quota

Gabon Standby 05/28/04 2004 69,440 154.30 Ukraine Standby 03/29/04 2004 411,600 1,372.00 Honduras PRGF 02/27/04 2004 71,200 129.50 Burundi PRGF 01/23/04 2004 69,300 77.00 Dominica PRGF 12/29/03 2003 7,688 8.20 Paraguay Standby 12/15/03 2003 50,000 99.90 Kenya PRGF 11/21/03 2003 225,000 271.40 Nepal PRGF 11/19/03 2003 49,900 71.30 Argentina Standby 09/20/03 2003 8,981,000 2,117.10 Dominican Republic Standby 08/29/03 2003 437,800 218.90 Tanzania PRGF 08/16/03 2003 19,600 198.90 Mauritania PRGF 07/18/03 2003 6,440 64.40 Bangladesh PRGF 06/20/03 2003 400,330 533.30 Guatemala Standby 06/18/03 2003 84,000 210.20 Burkina Faso PRGF 06/11/03 2003 30,100 60.20 Ghana PRGF 05/09/03 2003 184,500 369.00 Macedonia Standby 04/30/03 2003 20,000 68.90 Senegal PRGF 04/28/03 2003 24,270 161.80 Sri Lanka PRGF 04/18/03 2003 269,000 413.40 Bolivia Standby 04/02/03 2003 145,780 171.50 Equador Standby 03/21/03 2003 151,000 302.30

216 Country Loan Type Date Year Loan Size Quota

Croatia Standby 02/03/03 2003 105,880 365.10 Argentina Standby 01/24/03 2003 2,174,500 2,117.10 Colombia Standby 01/15/03 2003 1,548,000 774.00 Nicaragua PRGF 12/13/02 2002 97,500 130.00 Tajikistan PRGF 12/11/02 2002 65,000 87.00 Guyana PRGF 09/20/02 2002 54,550 90.90 Uganda PRGF 09/13/02 2002 13,500 180.50 Brazil Standby (w SRF) 09/06/02 2002 27,375,120 3,036.10 Dominica Standby 08/28/02 2002 2,973 8.20 Rwanda PRGF 08/12/02 2002 4,000 80.10 Bosnia and Herzegovina Standby 08/02/02 2002 67,600 169.10 Gambia PRGF 07/18/02 2002 20,220 31.10 Jordan Standby 07/03/02 2002 85,280 170.50 Albania PRGF 06/21/02 2002 28,000 48.70 Congo, Democratic R PRGF 06/12/02 2002 580,000 533.00 Serbia EFF 05/14/02 2002 650,000 467.70 Cape Verde PRGF 04/10/02 2002 8,640 9.60 Guatemala Standby 04/01/02 2002 84,000 210.20 Uruguay Standby (w SRF) 04/01/02 2002 1,988,500 306.50 Cote d’Ivoire PRGF 03/29/02 2002 292,680 325.20 Bulgaria Standby 02/27/02 2002 240,000 640.20

217 Country Loan Type Date Year Loan Size Quota

Turkey Standby 02/04/02 2002 12,821,200 1,191.30 Peru Standby 02/01/02 2002 255,000 638.40 Kyrgyz Republic PRGF 12/06/01 2001 73,400 88.80 Pakistan PRGF 12/06/01 2001 1,033,700 1,033.70 Romania Standby 10/31/01 2001 300,000 1,030.20 Mongolia PRGF 09/28/01 2001 28,490 51.10 Sierra Leone PRGF 09/26/01 2001 130,840 103.70 Brazil Standby (w SRF) 09/14/01 2001 12,144,400 3,036.10 Lithuania Standby 08/30/01 2001 86,520 144.20 Azerbaijan PRGF 07/06/01 2001 67,580 160.90 Serbia Standby 06/11/01 2001 200,000 467.70 Armenia PRGF 05/23/01 2001 69,000 92.00 Guinea PRGF 05/02/01 2001 64,260 107.10 Lao People’s Democratic R PRGF 04/25/01 2001 31,700 52.90 Latvia Standby 04/20/01 2001 33,000 126.80 Sri Lanka Standby 04/20/01 2001 200,000 413.40 Vietnam PRGF 04/13/01 2001 290,000 329.10 Ethiopia PRGF 03/22/01 2001 100,277 133.70 Croatia Standby 03/19/01 2001 200,000 365.10 Peru Standby 03/12/01 2001 128,000 638.40 Lesotho PRGF 03/09/01 2001 24,500 34.90

218 Country Loan Type Date Year Loan Size Quota

Madagascar PRGF 03/01/01 2001 91,650 122.20 Georgia PRGF 01/12/01 2001 108,000 150.30 Niger PRGF 12/22/00 2000 59,200 65.80 Cameroon PRGF 12/21/00 2000 111,420 185.70 Malawi PRGF 12/21/00 2000 45,110 69.40 Moldova PRGF 12/21/00 2000 110,880 123.20 Macedonia PRGF 12/18/00 2000 10,335 68.90 Guinea-Bissau PRGF 12/15/00 2000 14,200 14.20 Pakistan Standby 11/29/00 2000 465,000 1,033.70 Gabon Standby 10/23/00 2000 92,580 154.30 Kenya PRGF 08/04/00 2000 190,000 271.40 Nigeria Standby 08/04/00 2000 788,940 1,753.20 Benin PRGF 07/17/00 2000 27,000 61.90 Panama Standby 06/30/00 2000 64,000 206.60 Uruguay Standby 05/31/00 2000 150,000 306.50 Sao Tome & Principe PRGF 04/28/00 2000 6,657 7.40 Equador Standby 04/19/00 2000 226,730 302.30 Tanzania PRGF 04/04/00 2000 135,000 198.90 Papua New Guinea Standby 03/29/00 2000 85,540 131.60 Argentina Standby (w SRF) 03/10/00 2000 16,936,800 2,117.10 Lithuania Standby 03/08/00 2000 61,800 144.20

219 Country Loan Type Date Year Loan Size Quota

Estonia Standby 03/01/00 2000 29,340 65.20 Indonesia EFF 02/04/00 2000 3,638,000 2,079.30 Chad PRGF 01/07/00 2000 47,600 56.00 Turkey Standby (w SRF) 12/22/99 1999 15,038,400 1,191.30 Colombia EFF 12/20/99 1999 1,957,000 774.00 Kazakhstan EFF 12/13/99 1999 329,100 365.70 Latvia Standby 12/10/99 1999 33,000 126.80 Cambodia PRGF 10/22/99 1999 58,500 87.50 Djibouti PRGF 10/18/99 1999 19,082 15.90 Burkina Faso PRGF 09/10/99 1999 39,120 60.20 Mali PRGF 08/06/99 1999 51,315 93.30 Romania Standby 08/05/99 1999 400,000 1,030.20 Zimbabwe Standby 08/02/99 1999 141,360 353.40 Russian Federation Standby 07/28/99 1999 3,300,000 5,945.40 Mauritania PRGF 07/21/99 1999 42,490 64.40 Mexico Standby 07/07/99 1999 3,103,000 2,585.80 Mozambique PRGF 06/28/99 1999 87,200 113.60 Peru EFF 06/24/99 1999 383,000 638.40 Ghana PRGF 05/03/99 1999 228,800 369.00 Jordan EFF 04/15/99 1999 127,880 170.50 Uruguay Standby 03/29/99 1999 70,000 306.50

220 Country Loan Type Date Year Loan Size Quota

Honduras PRGF 03/26/99 1999 156,750 129.50 Zambia PRGF 03/25/99 1999 278,900 489.10 Brazil Standby (w SRF) 12/02/98 1998 13,024,800 3,036.10 Bulgaria EFF 09/25/98 1998 627,620 640.20 El Salvador Standby 09/23/98 1998 37,680 171.30 Bolivia PRGF 09/18/98 1998 100,960 171.50 Ukraine EFF 09/04/98 1998 1,919,950 1,372.00 Indonesia EFF 08/25/98 1998 5,383,100 2,079.30 Central African Republic PRGF 07/20/98 1998 49,440 55.70 Guyana PRGF 07/15/98 1998 53,760 90.90 Gambia PRGF 06/29/98 1998 20,610 31.10 Kyrgyz Republic PRGF 06/26/98 1998 73,380 88.80 Rwanda PRGF 06/24/98 1998 71,400 80.10 Tajikistan PRGF 06/24/98 1998 100,300 87.00 Zimbabwe Standby 06/01/98 1998 130,650 353.40 Bosnia and Herzegovina Standby 05/29/98 1998 94,420 169.10 Albania PRGF 05/13/98 1998 45,040 48.70 Senegal PRGF 04/20/98 1998 107,010 161.80 Philippines Standby 04/01/98 1998 1,020,790 879.90 Nicaragua PRGF 03/18/98 1998 148,955 130.00 Cote d’Ivoire PRGF 03/17/98 1998 285,840 325.20

221 Country Loan Type Date Year Loan Size Quota

Cape Verde Standby 02/20/98 1998 2,496 9.60 Argentina EFF 02/04/98 1998 2,080,000 2,117.10 Estonia Standby 12/17/97 1997 16,100 65.20 Panama EFF 12/10/97 1997 120,000 206.60 Korea Standby (w SRF) 12/04/97 1997 15,500,000 2,927.30 Uganda PRGF 11/10/97 1997 100,425 180.50 Indonesia Standby 11/05/97 1997 8,338,240 2,079.30 Yemen PRGF 10/29/97 1997 264,750 243.50 Pakistan PRGF 10/20/97 1997 682,380 1,033.70 Latvia Standby 10/10/97 1997 33,000 126.80 Ukraine Standby 08/25/97 1997 398,920 1,372.00 Cameroon PRGF 08/20/97 1997 162,120 185.70 Thailand Standby 08/20/97 1997 2,900,000 1,081.90 Mongolia PRGF 07/30/97 1997 33,390 51.10 Uruguay Standby 06/20/97 1997 125,000 306.50 Romania Standby 04/22/97 1997 301,500 1,030.20 Bulgaria Standby 04/11/97 1997 371,900 640.20 Macedonia PRGF 04/11/97 1997 54,560 68.90 Croatia EFF 03/12/97 1997 353,160 365.10 El Salvador Standby 02/28/97 1997 37,680 171.30 Guinea PRGF 01/13/97 1997 70,800 107.10

222 Country Loan Type Date Year Loan Size Quota

Azerbaijan PRGF 12/20/96 1996 93,600 160.90 Madagascar PRGF 11/27/96 1996 81,360 122.20 Tanzania PRGF 11/08/96 1996 181,590 198.90 Haiti PRGF 10/18/96 1996 91,050 81.90 Egypt Standby 10/11/96 1996 271,400 943.70 Ethiopia PRGF 10/11/96 1996 88,470 133.70 Lesotho Standby 09/23/96 1996 7,170 34.90 Benin PRGF 08/28/96 1996 27,180 61.90 Estonia Standby 07/29/96 1996 13,950 65.20 Bulgaria Standby 07/19/96 1996 400,000 640.20 Kazakhstan EFF 07/17/96 1996 309,400 365.70 Venezuela Standby 07/12/96 1996 975,650 2,659.10 Peru EFF 07/01/96 1996 300,200 638.40 Congo, Republic PRGF 06/28/96 1996 69,480 84.60 Mozambique PRGF 06/21/96 1996 75,600 113.60 Burkina Faso PRGF 06/14/96 1996 39,780 60.20 Niger PRGF 06/12/96 1996 57,960 65.80 Latvia Standby 05/24/96 1996 30,000 126.80 Moldova EFF 05/20/96 1996 135,000 123.20 Ukraine Standby 05/10/96 1996 598,200 1,372.00 Tajikistan Standby 05/08/96 1996 15,000 87.00

223 Country Loan Type Date Year Loan Size Quota

Kenya PRGF 04/26/96 1996 149,550 271.40 Djibouti Standby 04/15/96 1996 8,250 15.90 Argentina Standby 04/12/96 1996 720,000 2,117.10 Mali PRGF 04/10/96 1996 62,010 93.30 Russian Federation EFF 03/26/96 1996 6,901,000 5,945.40 Yemen Standby 03/20/96 1996 132,375 243.50 Hungary Standby 03/15/96 1996 264,180 1,038.40 Uruguay Standby 03/01/96 1996 100,000 306.50 Georgia PRGF 02/28/96 1996 172,050 150.30 Armenia PRGF 02/14/96 1996 109,350 92.00 Jordan EFF 02/09/96 1996 238,040 170.50 Uzbekistan Standby 12/18/95 1995 124,700 275.60 Pakistan Standby 12/13/95 1995 562,590 1,033.70 Zambia PRGF 12/06/95 1995 701,682 489.10 Costa Rica Standby 11/29/95 1995 52,000 164.10 Panama Standby 11/29/95 1995 84,300 206.60 Azerbaijan Standby 11/17/95 1995 58,500 160.90 Gabon EFF 11/08/95 1995 110,300 154.30 Malawi PRGF 10/18/95 1995 50,960 69.40 Cameroon Standby 09/27/95 1995 67,600 185.70 Belarus Standby 09/12/95 1995 196,280 386.40

224 Country Loan Type Date Year Loan Size Quota

Chad PRGF 09/01/95 1995 49,560 56.00 Lesotho Standby 07/31/95 1995 7,170 34.90 El Salvador Standby 07/21/95 1995 37,680 171.30 Papua New Guinea Standby 07/14/95 1995 71,480 131.60 Ghana PRGF 06/30/95 1995 164,400 369.00 Armenia Standby 06/28/95 1995 43,875 92.00 Georgia Standby 06/28/95 1995 72,150 150.30 Kazakhstan Standby 06/05/95 1995 185,600 365.70 Algeria EFF 05/22/95 1995 1,169,280 1,254.70 Macedonia Standby 05/05/95 1995 22,300 68.90 Latvia Standby 04/21/95 1995 27,450 126.80 Estonia Standby 04/11/95 1995 13,950 65.20 Russian Federation Standby 04/11/95 1995 4,313,100 5,945.40 Ukraine Standby 04/07/95 1995 997,300 1,372.00 Moldova Standby 03/22/95 1995 58,500 123.20 Haiti Standby 03/08/95 1995 20,000 81.90 Mexico Standby 02/01/95 1995 12,070,200 2,585.80 Mauritania PRGF 01/25/95 1995 42,750 64.40 Guinea-Bissau PRGF 01/18/95 1995 10,500 14.20 Bolivia PRGF 12/19/94 1994 100,960 171.50 Malawi Standby 11/16/94 1994 15,000 69.40

225 Country Loan Type Date Year Loan Size Quota

Vietnam PRGF 11/11/94 1994 362,400 329.10 Lithuania EFF 10/24/94 1994 134,550 144.20 Croatia Standby 10/14/94 1994 65,400 365.10 Lesotho Standby 09/23/94 1994 8,365 34.90 Togo PRGF 09/16/94 1994 65,160 73.40 Uganda PRGF 09/06/94 1994 120,510 180.50 Senegal PRGF 08/29/94 1994 130,790 161.80 Poland Standby 08/05/94 1994 333,300 1,369.00 Slovak Republic Standby 07/22/94 1994 115,800 357.50 Guyana PRGF 07/20/94 1994 53,760 90.90 Kyrgyz Republic PRGF 07/20/94 1994 88,150 88.80 Turkey Standby 07/08/94 1994 610,500 1,191.30 Nicaragua PRGF 06/24/94 1994 120,120 130.00 Philippines EFF 06/24/94 1994 791,200 879.90 Algeria Standby 05/27/94 1994 457,200 1,254.70 Congo, Republic Standby 05/27/94 1994 23,160 84.60 Jordan EFF 05/25/94 1994 189,300 170.50 Equador Standby 05/11/94 1994 173,900 302.30 Romania Standby 05/11/94 1994 320,495 1,030.20 Cambodia PRGF 05/06/94 1994 84,000 87.50 Bulgaria Standby 04/11/94 1994 139,480 640.20

226 Country Loan Type Date Year Loan Size Quota

Gabon Standby 03/30/94 1994 38,600 154.30 Central African Republic Standby 03/28/94 1994 16,480 55.70 Sierra Leone PRGF 03/28/94 1994 101,904 103.70 Chad Standby 03/23/94 1994 16,520 56.00 Cameroon Standby 03/14/94 1994 81,060 185.70 Cote d’Ivoire PRGF 03/11/94 1994 333,480 325.20 Niger Standby 03/04/94 1994 18,596 65.80 Senegal Standby 03/02/94 1994 47,560 161.80 Pakistan PRGF 02/22/94 1994 606,600 1,033.70 Kazakhstan Standby 01/26/94 1994 123,750 365.70

227 BIBLIOGRAPHY

Abouharb, Rodwan M. & David L. Cingranelli. 2009. “IMF Programs and Human Rights, 1981 - 2003.” The Review of International Organizations 4(1):47 – 72. Anner, Mark, Teri L. Caraway & Stephanie J. Rickard. 2010. “International Negoti- ations and Domestic Politics: The Case of IMF Labor Market Conditionality.” Paper Presented at the Third Political Economy of International Organizations Conference. Armenia. 1995. “The Letter of Intent.”. Arpac, Ozlem & Graham Bird. 2009. “Turkey and the IMF: A Case Study in the Political Economy of Policy Implementation.” The Review of International Organizations 4(2):135 – 157. Aslund, Anders. 1999. “Russia’s Collapse.” Foreign Affairs 78(5):64. Babb, Sarah. 2003. “The IMF in Sociological Perspective: A Tale of Organizational Slippage.” Studies in Comparative International Development 38:3–27. Barnett, Michael & Martha Finnemore. 2004. Rules for the World. Cornell University Press. Beck, Thorsten, Asli Demirguc-Kunt & Ross Levine. 2000. “A New Database on the Strucutre and Development of the Financial Sector.” The World Bank Economic Review 14(3):597 – 605. Beck, Thorsten, George Clarke, Alberto Groff, Philip Keefer & Patrick Walsh. 2001. “New Tools in Comparative Political Economy: The Database of Political In- stitutions.” The World Bank Economic Review 15(1):165–176. Biglaiser, Glen & Karl DeRouen Jr. 2010. “The Effects of IMF Programs on U.S. Foreign Direct Investment in the Developing World.” The Review of Interna- tional Organizations 5(1):73 – 95. 228 Bird, Graham & Dane Rowlands. 2004. “Financing Balance of Payments Adjust- ment: Options in the Light of the Elusive Catalytic Effect of IMF-Supported Programmes.” Comparative Economic Studies 46:468–486.

Bird, Graham & Thomas D. Willett. 2004. “IMF Conditionality, Implementation and the New Political Economy of Ownership.” Comparative Economic Studies 46:423–450.

Blustein, Paul. 2001. The Chastening: Inside the Crisis that Rocked the Global Financial System and Humbled the IMF. PublicAﬀairs.

Blustein, Paul. 2005. And the Money Kept Rolling In (And Out). PublicAﬀairs.

Brooks, Sarah M. 56. “Explaining Capital Account Liberalization in Latin America: A Transitional Cost Approach.” World Politics 3(389-403).

Broz, J. Lawrence & Michael Brewster Hawes. 2006a. “Congressional Politics of Financing the International Monetary Fund.” International Organization 60:367–399.

Broz, J. Lawrence & Michael Brewster Hawes. 2006b. US Domestic Politics and International Monetary Fund Policy. In Delegation and Agency in International Organizations, ed. Darren G. Hawkins, David A. Lake, Daniel L. Nielson & Michael J. Tierney. Cambridge University Press pp. 77–106.

Broz, Lawrence J. 2008. “Congressional Voting on Funding the International Finan- cial Institutions.” The Review of International Organizations 3(4):351–374.

Chwieroth, Jeﬀrey. 2007a. “Neoliberal Economists and Capital Account Liberaliza- tion in Emerging Markets.” International Organization 61:443 – 463.

Chwieroth, Jeﬀrey. 2007b. “Testing and Measuring the Role of Ideas: The Case of Neoliberalism in the International Monetary Fund.” International Studies Quarterly 51:5–30.

Copelovitch, Mark. 2010. “Master or Servant? Agency Slack and the Politics of IMF Lending.” International Studies Quarterly 54(1):49 – 77.

Copelovitch, Mark S. 2008. “Banks, Bonds, and Bailouts: Global Finance and the International Monetary Fund.”.

Coughlin, Peter J. 1992. Probabilistic Voting Theory. Cambridge University Press.

229 Cowhey, Peter F. 1993. “Domestic Institutions and the Credibility of Interna- tional Commitments: Japan and the United States.” International Organiza- tion 47:299–326.

Cull, Robert & Laurie Eﬀron. 2008. “World Bank Lending and Financial Sector Development.” The World Bank Economic Review 22(2):315–343.

Dai, Xinyuan. 2005. “Why Comply? The Domestic Constituency Mechanism.” International Organization 59:363–398.

Danaher, Kevin, ed. 1994. 50 Years Is Enough: The Case against the World Bank and the International Monetary Fund. South End Press.

Dominican Republic. 2003. “The Letter of Intent.”.

Drazen, Allan. 2002. Conditionality and Ownership in IMF Lending: A Political Economy Approach. In IMF Staﬀ Papers. International Monetary Fund.

Dreher, Axel. 2004a. “The Inﬂuence of IMF Programs on the Re-election of Debtor Governments.” Economics & Politics 16:53–75.

Dreher, Axel. 2004b. “A Public Choice Perspective of IMF and World Bank Lending and Conditionality.” Public Choice 119:445–464.

Dreher, Axel. 2006. “IMF and Economic Growth: The Eﬀects of Programs, Loans, and Compliances with Conditionality.” World Development 34:769–788.

Dreher, Axel, Jan-Egbert Sturm & James Raymond Vreeland. 2009. “Global Horse Trading: IMF Loans for Votes in the United Nations Security Council.” Euro- pean Economic Review 53:742 – 757.

Dreher, Axel & Nathan Jensen. 2007. “Independent Actor or Agent? An Empirical Analysis of the Impact of US Interests on IMF Conditions.” The Journal of Law and Economics 50:105–124.

Dreher, Axel & Roland Vaubel. 2004. “The Causes and Cosequences of IMF Conditionality.” Emerging Markets Finance and Trade 3:26–54.

Drezner, Daniel W., ed. 2003. Locating the Proper Authorities: The Interaction of Domestic and International Institutions. University of Michigan Press.

Edwards, Martin S. 2006. “Signalling Credibility? The IMF and Catalystic Fi- nance.” Journal of International Relations and Development 9:27–52. 230 Evans, Peter B., Harold K. Jacobson & Robert D. Putnam, eds. 1993. Double- Edged Diplomacy: International Bargaining and Domestic Politics. University of California Press.

Fearon, James D. 1998. “Bargaining, Enforcement, and International Cooperation.” International Organization 52:269–306.

Foreign Policy Blog Passport. 2010. “New Greek Austerity Measures.”.

Garuda, Gopal. 2000. “The Distributional Eﬀects of IMF Programs: A Cross- Country Analysis.” World Development 28:1031–1051.

Goldstein, Judith. 1996. “International Law and Domestic Institutions: Reconciling North American ”Unfair” Trade Laws.” International Organization 50(4):541– 564.

Gould, Erica. 2006. Money Talks: The International Monetary Fund, Conditionality and Supplementary Funanciers. Stanford Univeristy Press.

Gould, Erica R. 2003. “Money Talks: Supplementary Financiers and International Monetary Fund Conditionality.” International Organization 57:551–586.

Grossman, Gene M. & Elhanan Helpman. 2001. Special Interest Politics. The MIT Press.

Haggard, Stephan. 2000. The Political Economy of the Asian Financial Crisis. Institute for International Economics.

Heckman, James J. 1979. “Sample Selection Bias as a Speciﬁcation Error.” Econo- metrica 47:153 – 161.

Holmes, Kim R., Edwin J. Feulner & Mary Anastasia O’Grady. 2008. 2008 Index of Economic Freedom. Heritage Foundation.

Independent Evaluation Oﬃce. 2007a. “An IEO Evaluation of Structural Condi- tionality in IMF-Supported Programs.”.

Independent Evaluation Oﬃce. 2007b. “Structural Conditionality in IMF-Supported Programs: Background Documents.”.

Jaggers, Keith & Ted Robert Gurr. 1995. “Tracking Democracy’s Third Wave with the Polity III Data.” Journal of Peace Research 32(4):469–482.

231 Jensen, Nathan. 2004. “Crisis, Conditions, and Capital: The Effect of International Monetary Fund Agreements on Foreign Direct Investment Inflows.” Journal of Conflict Resolution 48(2):194–210.

Kahler, Miles. 1993. Bargaining with The IMF: Two-Level Strategies and Devel- oping Countries. In Double-Edged Diplomacy: International Bargaining and Domestic Politics, ed. Harold K. Jacobson Evans, Peter B. & Robert D. Put- nam. University of California Press pp. 363–394.

Kenya. 1996. “The Letter of Intent.”.

Killick, Tony. 1995. IMF Programmes in Developing Countries: Design and Impact. Routledge.

King, Gary, Michael Tomz & Jason Wittenberg. 2000. “Making the Most of Statis- tical Analyses: Improving Interpretation and Presentation.” American Journal of Political Science 44(2):347 – 361.

Levine, Ross. 1997. “Financial Development and Economic Growth: Views and Agenda.” Journal of Economic Literature 35(2):688 – 726.

Long, J. Scott. 1997. Regression Models for Categorical and Limited Dependent Variables. Sage Publications.

MacIntyre, Andrew J. 1993. The Politics of Finance in Indonesia: Command, Con- fusion, and Competition. In The Politics of Finance in Developing Countries, ed. Stephan Haggard, C. H. Lee & Sylvia Maxﬁeld. Cornell University Press.

Marchesi, Sylvia, Laura Sabani & Axel Dreher. 2009. “Agency and Communication in IMF Conditional Lending: Theory and Emprical Evidence.” Working Paper Series, Department of Economics, University of Milan - Bicocca.

Martin, Lisa L. 2000. Democratic Commitments: Legislatures and International Cooperation. Princeton University Press.

Mayer, Wolfgang & Alex Mourmouras. 2002. Vested Interests in a Positive Theory of IFI Conditionality. In IMF Working Paper. The International Monetary Fund.

Mayer, Wolfgang & Alex Mourmouras. 2004. “IMF Conditionality and the Theory of Special Interest Politics.” Comparative Economic Studies 46:400–422.

232 Mayer, Wolfgang & Alex Mourmouras. 2008. “IMF Conditionality: An Approach Based on the Theory of Special Interest Politics.” The Review of International Organizations 3(2):105 – 121.

Milner, Helen V. 1997. Interests, Institutions, and Information: Domestic Politics and International Relations. Princeton University Press.

Milner, Helen V. & B. Peter Rosendorﬀ. 1997. “Democratic Politics and Interna- tional Trade Negotiations.” Journal of Conﬂict Resolution 41(1):117 – 146.

Mody, Ashoka & Alessandro Rebucci, eds. 2006. IMF-Supported Programs: Recent Staﬀ Research. International Monetary Fund.

Momani, Bessma. 2004a. “American Politicization of the International Monetary Fund.” Review of International Political Economy 11:880–904.

Momani, Bessma. 2004b. “The IMF, US War on Terrorism, and Pakistan: A Lesson in Economic Statecraft.” Asian Aﬀairs 31:41–50.

Noble, Gregory W. & John Ravenhill, eds. 2000. The Asian Financial Crisis and the Architecture of Global Finance. Cambridge University Press.

Nooruddin, Irfan & Joel W. Simmons. 2006. “The Politics of Hard Choices: IMF Programs and Government Spending.” International Organization 60:1001– 1033.

Oatley, Thomas & Jason Yackee. 2004. “American Interests and IMF Lending.” International Politics 41:415–429.

Pepinski, Thomas. 2009a. Economic Crises and the Breakdown of Authoritarian Regimes: Indonesia and Malaysia in Comparative Perspective. Cambridge Uni- versity Press.

Pepinski, Thomas. 2009b. “Ownership and Opportunity: Why Bankers in Emerg- ing Markets Support Financial Internationalization.” Paper Presented at the International Political Economy Meeting.

Persson, Torsten & Guido Tabellini. 2002. Political Economics: Explaining Eco- nomic Policy. The MIT Press.

233 Pevehouse, Jon C. 2003. Democratization, Credible Commitments, and Joining In- ternational Organizations. In Locating the Proper Authorities: The Interaction of Domestic and International Institutions, ed. Daniel Drezner. University of Michigan Press pp. 25–48.

Przeworski, Adam & James Raymond Vreeland. 2000. “The Eﬀect of IMF Programs on Economic Growth.” Journal of Development Economics 62:385–421.

Putnam, Robert D. 1988. “Diplomacy and Domestic Politics: The Logic of Two- Level Games.” International Organization 42:427–460.

Remmer, Karen. 1986. “The Politics of Economic Stabilization: IMF Standby Programs in Latin America, 1954-1984.” Comparative Politics 19:1–24.

Schelling, Thomas C. 1960. The Strategy of Conﬂict. Harvard University Press.

Smith, Alastair & James Raymond Vreeland. 2006. The Survival of Political Leaders and IMF Programs. In Globalization and the Nation State: The Impact of the IMF and the World Bank, ed. Gustav Ranis, James Vreeland & Stephen Kosack. Routledge pp. 263–289.

Steinwand, Martin & Randall W. Stone. 2008. “The International Monetary Fund: A Review of the Recent Evidence.” The Review of International Organizations 3(2):123 – 149.

Stiglitz, Joseph E. 2002. Globalization and Its Discontents. W. W. Norton & Company.

Stiles, Kendall W. 1987. “Argentina’s Bargaining with the IMF.” Journal of Inter- American Studies and World Aﬀairs 29:55–85.

Stiles, Kendall W. 1990. “IMF Conditionality: Coercion or Compromise?” World Development 18:959–974.

Stone, Randall W. 2004. “The Political Economy of IMF Lending in Africa.” Amer- ican Political Science Review 98:577–592.

Stone, Randall W. 2008. “The Scope of IMF Conditionality.” International Orga- nization 62:589–620.

Sundaram, Jomo Kwame. 2006. “Pathways Through Financial Crisis: Malaysia.” Global Governance 12:489–505.

234 Thacker, Strom Cronan. 1999. “The High Politics of IMF Lending.” World Politics 52:38–75.

The International Monetary Fund. 1945. “Articles of Agreement of the International Monetary Fund.”.

The International Monetary Fund. 2001. “Structural Conditionality in Fund- Supported Programs.”.

The International Monetary Fund. 2007. “Selected Decisions and Selected Docu- ments of the International Monetary Fund.”.

The International Monetary Fund. 2010. “Europe and IMF Agree 110 Billion Financing Plan With Greece.”.

The International Monetary Fund. N.d. “International Monetary Fund Oﬃcial Website.”. URL: www.imf.org

Tomz, Michael, Jason Wittenberg & Gary King. N.d. “CLARIFY: Software for Interpreting and Presenting Statistical Results. Version 2.0.”.

Vaubel, Roland. 1986. “A Public Choice Approach to International Organization.” Public Choice 51:39–57.

Vaubel, Roland. 1996. “Bureaucracy at the IMF and the World Bank: A Compari- son of the Evidence.” World Economy 19:195–210.

Voeten, Erik & Adis Merdzanovic. N.d. “United Nations General Assembly Voting Data.”.

Vreeland, James Raymond. 2002. “The Eﬀect of IMF Programs on Labor.” World Development 30:121–139.

Vreeland, James Raymond. 2003. The IMF and Economic Development. Cambridge University Press.

Vreeland, James Raymond. 2007. The International Monetary Fund: Politics of Conditional Lending. Routledge.

Vreeland, James Raymond. N.d.a. “Institutional Determinants of IMF Agree- ments.”.

235 Vreeland, James Raymond. N.d.b. “The International and Domestic Politics of IMF Programs.”.

Wall Street Journal. 2010. “Greece’s Cuts and Tax Increases: How Austerity Looks.”.

Williamson, John. 2009. “A Short History of the Washington Consensus.” Law and Business Review of the Americas 15:7 – 23.

Winters, Jeﬀrey A. 1996. Power in Motion: Capital Mobility and the Indonesian State. Cornell University Press.

Woods, Ngaire. 2010. “Global Governance after the Financial Crisis: A New Multilateralism or the Last Gasp of the Great Powers?” Global Policy 1(1):51 – 63.

World Bank. N.d. “World Development Indicators.”. URL: http://go.worldbank.org/3SGLDH5V10

236