Bargaining over the Bomb: The Successes and Failures of Nuclear Negotiations

Transcription

1 Bargaining over the Bomb: The Successes and Failures of Nuclear Negotiations by William Spaniel Submitted in Partial Fulllment of the Requirements for the Degree Doctor of Philosophy Supervised by Professor Randall W. Stone Department of Political Science Arts, Sciences and Engineering School of Arts and Sciences University of Rochester Rochester, New York 2015

2 For peace. ii

3 iii Biographical Sketch The author was born in Valencia, California. He attended the University of California, San Diego, and graduated with a Bachelor of Arts degree in Political Science in He began doctoral studies in international relations at the University of Rochester in 2010 and received a Master of Arts degree from the University of Rochester in He pursued his research in international relations under the direction of Professor Randall W. Stone. The following publications were a result of work conducted during doctoral study: Preventive War and the Credibility of Arms Treaties. (Forthcoming, International Interactions.) Sanctions, Uncertainty, and Leader Tenure. With Bradley C. Smith. (Forthcoming, International Studies Quarterly.) How Uncertainty about the Legal Views of Judicial Nominees Distorts the Conrmation Process. With Maya Sen. (Forthcoming, Journal of Theoretical Politics.)

4 iv Acknowledgments I presented earlier versions of this project at the Peter D. Watson Center for Conict and Cooperation in May 2012 and November 2010 and at the University of Bualo workshop series in October I thank Phil Arena, Matthew Blackwell, Alexandre Debs, John Duggan, James Fearon, Mark Fey, Erik Gartzke, Hein Goemans, Ryan Jaimes, Bethany Lacina, Alexander Lanoszka, Nuno Monteiro, Harvey Palmer, Toby Rider, Maya Sen, Branislav Slantchev, Brad Smith, Randy Stone, Zachary Taylor, Svanhildur Thorvaldsdottir, and the participants of various talks at the American Political Science Association, Midwest Political Science Association, Southern Political Science Association, and Peace Science Society conferences. Any errors remain my own.

5 v Abstract Why do states acquire nuclear weapons? Existing theories of nuclear proliferation fail to account for the impact of bargaining on the processi.e., credible agreements exist in which rival states make sucient concessions to convince the potential rising state not to proliferate. This dissertation proves the existence of that range of settlements and the robustness of the ineciency puzzle. It then provides two main explanations as to why states proliferate anyway. First, if the would-be proliferator expects to lose the ability to construct nuclear weapons in the future, the states face a commitment problem: the rival state would like to promise to continue providing concessions into the future but will renege once proliferation is no longer an option. And second, if the proliferator's rival faces some sort of uncertaintywhether regarding the potential proliferator's ability to go nuclear or regarding its previous proliferation activitythe optimal oer can entail positive probability of nuclear investment. However, the nonproliferation regime's mission to increase the cost of building often leads to Pareto improvement.

6 vi Contributors and Funding Sources This work was supported by a dissertation committee consisting of Professors Randall Stone (advisor), Mark Fey, and Hein Goemans of the Department of Political Science, Professor Matthew Lenoe of the Department of History, and Professor Paulo Barelli of the Department of Economics. Graduate study was supported by a fellowship from the University of Rochester.

7 vii Table of Contents 1 Introduction The Central Argument Competing Explanations and Why They Are Insucient Roadmap Microfoundations: Nuclear Weapons and the Bargaining Model of War Intuition: Credible Threats in Crisis Bargaining Generalizing the Result Applying the Result Appendix The Theory of Butter-for-Bombs Agreements: How Potential Power Coerces Concessions Modeling Butter-for-Bombs Agreements Actions and Transitions Key Features

8 TABLE OF CONTENTS viii Equilibrium Robustness Interpretation Implications of Butter-for-Bombs Agreements Conclusion Appendix Proof of Lemma Proof of Proposition Proof of Proposition Proof of Proposition Does Latent Nuclear Capacity Deter Aggressive Behavior? Theory and Hypotheses Data and Statistical Model Empirical Evidence of the Eectiveness of Potential Power Robustness Discussion Conclusion The Diplomacy of Butter-for-Bombs Agreements A Theory of Formal Models and Case Studies Cold War Allies: South Korea, Australia, and Japan Egypt, the Camp David Accords, and External Subsidies Soviet Successor States: Belarus, Kazakhstan, and Ukraine.. 139

9 TABLE OF CONTENTS ix 5.5 Conclusion Institutions, Signaling, and Butter-for-Bombs Endogenous Investment Costs Contracting Nuclear Restraint Costly Signaling and the Limits of Institutional Solutions Conclusion Appendix: Proof of Theorem War Exhaustion and the Stability of Arms Treaties What Is War Exhaustion? Modeling War Exhaustion Actions and Transitions Equilibrium Illustrating the Mechanism: The Origins of the Cold War and Iranian-American Tensions The Soviet Union Iran Domestic Politics and the Robustness of Treaties Conclusion Appendix Proof of Lemma Proof of Proposition

10 TABLE OF CONTENTS x 8 Resolving the Monitoring Problem: The Hidden Value of Worthless Weapons Inspections Known Unknowns, Weapons Inspections, and the Cost of Inconvenience Bargaining with Imperfect Monitoring When Monitoring Problems Are Irrelevant Why Butter-for-Bombs Agreements Still Work The Monitoring Problem Premium The Extreme: Preventive War and Proliferation The Value of Weapons Inspections The Hidden Purpose (and Apparent Futility) of Weapons Inspections Understanding the Origins Iraq War Strategic Jockeying: Costly Signaling in the Post-9/11 Era Conclusion Appendix Proof for Proposition 8.2 and Proposition Proof for Proposition Proof for Theorem Robustness Checks Preventive Strikes: When States Call the Wrong Blu 281

11 TABLE OF CONTENTS xi 9.1 Modeling the Blu Actions and Information Equilibrium Illustrating the Mechanism: Israel, Iraq, and Operation Opera Overview of the Iraqi Nuclear Program and Operation Opera The Osirak Reactor: An Open Secret Iraqi Optimism Preventive War versus Disadvantageous Peace The Future of Blung Conclusion Appendix Lessons Learned 322 Biblography 330

12 xii List of Tables 4.1 Regression results from ve logit models Timeline of Important Events in U.S./U.S.S.R Nuclear Relations Timeline of Important Events in Recent U.S./Iran Relations Timeline of the Osirak Reactor and Operation Opera

13 xiii List of Figures 1.1 Nuclear capable versus nuclear weapon states over time An example of war outcomes The central model's extensive form The pre-shift and post-shift balances of power R's equilibrium oer size as a function of p 1 R The set of Pareto settlements in a static versus dynamic bargaining game Equilibrium outcomes as a function of the cost to build and the rising state's level of future power Predicted probabilities from Model R's payo as a function of its cost k Equilibrium outcomes as a function of the extent of the power shift and the investment cost A simple arms race game

14 LIST OF FIGURES xiv 8.1 A simultaneous move game between a declining state (rows) and rising state (columns) A simple simultaneous move game Equilibrium outcomes as a function of the cost to build and the rising state's level of future power The simultaneous move decision Equilibrium payos for both actors as a function of k The utility of weapons inspections as a function of the extent of the power shift and cost of proliferation

15 1 Chapter 1 Introduction On September 25, 2009, U.S. President Barack Obama gave the following warning to Iran at a G20 summit in Pittsburgh, Pennsylvania: Iran must comply with U.N. Security Council resolutions and make clear it is willing to meet its responsibilities as a member of the community of nations. We have oered Iran a clear path toward greater international integration if it lives up to its obligations, and that oer stands. But the Iranian government must now demonstrate through deeds its peaceful intentions or be held accountable to international standards and international law. 1 For the past few years, Western powers have extended such olive branches to Iran, oering various enticements if Iran ends its (assumed) nuclear weapons program. It is still unclear whether Iran will eventually yield. However, if past oers are an indication, any eventual deal will be a quid-pro-quo, in 1 Obama 2009.

16 CHAPTER 1. INTRODUCTION 2 which Iran receives concessions, dismantles its program, and continues to receive those concessions in the future. The United States has maintained a similar strategy with North Korea. In early 2012, the U.S. and North Korea reached a food-for-nukes agreement, which called for the North to end its nuclear weapons program and suspend long-range missile tests in exchange for millions of pounds of food. Since then, North Korea has stubbornly continued to test missiles under the auspices of its edgling space program and detonated its third nuclear weapon in February As a result, whether Iran or North Korea will end their respective nuclear programs for good remains in doubt. Reasons for pessimism abound, as both sides appear to face commitment problems. The United States would only want to oer Iran and North Korea immediate concessions if those rivals end their respective nuclear programs in return. However, these states could take the those concessions, continue constructing nuclear weapons, magnify their coercive power, and coerce yet more concessions. If Washington is worried they will adopt such a strategy, it might never give the concessions necessary to induce them to end their programs. Meanwhile, without any physical power to enforce the agreement, the potential rising states must be concerned that the United States' concessions are temporary. Thus, even if settlements exist that leave both sides better o, credible commitment problems may lead to proliferation anyway. That said, it is important not to take the struggles with Iran and North Korea as conclusive evidence that bargaining is doomed to fail. Perhaps

17 CHAPTER 1. INTRODUCTION 3 those cases are salient precisely because they are so exceptional. To wit, similar agreements involving other states have lasted over the long-term. Egypt's turbulent early relationship with Israel led Cairo's to explore proliferation beginning in the 1960s. However, Egypt lost most incentive to acquire nuclear weapons when the parties signed the Camp David Accords, which returned the Sinai Peninsula to Egypt (Einhorn 2004, 48-51). Anwar Sadat formalized Egypt's commitment to nonproliferation by ratifying the Nuclear Nonproliferation Treaty in Although Egypt continued pursuing nuclear technology (Solingen 2007, ), he and his successor Hosni Mubarak never made a serious attempt to proliferate thereafter. Indeed, any nuclear experimentation would have put U.S. foreign direct aidtied to good relations with Israel the Accords outlinein jeopardy (Arena and Pechenkina 2012). Proliferation could have provided Egypt with benets, but the net benets (minus costs) could not have exceeded the value of the deal Egypt was already receiving. The end of the Cold War presented the United States a large-scale nuclear conundrum. Although the nuclear stockpiles in Belarus, Kazakhstan, and Ukraine remained under control of Moscow (Miller 1993, 71-74), each of these countries had the technological know-how to proliferate in short order. Negotiations between the United States and these Soviet successor states ultimately led to the Lisbon Protocol, which made Belarus, Kazakhstan, and Ukraine parties to the Nuclear Non-Proliferation Treaty. Contingent of the terms, the United States oered assistance with civilian nuclear energy

18 CHAPTER 1. INTRODUCTION 4 projects, provided substantial aid at a time of nancial crisis, and issued security assurances. The United States even oered concessions to its own Cold War allies. At various points, Japan, South Korea and Australia sought a nuclear deterrent. The United States publicly extended its nuclear umbrella over those countries in each of these cases. Rather than risk destroying their good relationships with Washington, all ultimately backed down. Of course, negotiations do not always lead to non-proliferation. Ten countries have proliferated; nine still maintain nuclear arsenals. 2 The question is what separates Egypt, Belarus, and South Korea from the Soviet Union, Pakistan, and North Korea. Why do negotiations sometimes succeed and sometimes fail? 1.1 The Central Argument This book argues that mutually preferable settlements between status quo powers and would-be proliferators always exist. No proliferation occurs, and both parties are better o as a result. However, existence does not guarantee feasibility. States may be misinformed about each other's capabilities or cannot credibly commit to such agreements as those capabilities uctuate. The presence of these bargaining frictions determines whether a state pursues 2 South Africa held deliverable nuclear weapons from 1982 to 1990 and ultimately acceded to the Nuclear Nonproliferation Treaty due to a combination of structural change to the international system and internal political strife (Albright 1994). See Gartzke and Kroenig (2009) for a list of nuclear powers and when they entered the club.

19 CHAPTER 1. INTRODUCTION 5 nuclear weapons. To be thorough, understanding nuclear proliferation is a two-step process. First, this book shows that there exist concessions-for-weapons agreements, or butter-for-bombs settlements, that leave neither side with incentive to declare war or build nuclear weapons. The book then describes the circumstances under which potential proliferators would want to reject such agreements and nuclearize anyway. More specically, this book presents a proliferation ineciency puzzle. When nuclear weapons are too expensive or the rival's threat to launch a preventive strike is credible, reaching non-proliferation settlements is trivial since the potential proliferator cannot credibly threaten to invest in nuclear weapons. Outside these cases, proliferation may seem inevitable, as the potential proliferator's temptation to build might prove too strong. However, such intuition fails to properly empathize with the potential proliferator's incentives. Nuclear weapons are exceedingly costly. The expenses start with constructing nuclear infrastructure to create weapons-grade material. Then the state must construct the physical weapons. But nuclear bombs are not eective unless the state builds the corresponding delivery mechanisms. And even once that is all complete, the proliferator must pay to maintain the weapons and delivery mechanisms over time, which can be the most burdensome part of all. As a result, perhaps it is not surprising that Schwartz (1998) estimates that the American nuclear program spent $5.5 trillion from

20 CHAPTER 1. INTRODUCTION to 1996 (in 1996 dollars). 3 Ultimately, nuclear weapons must provide some sort of benets for potential proliferators to bother with the whole operation. That being the case, why can't rival states concede those benets upfront and avoid the nuclear outcome? Indeed, a rival state could shave o a portion of these oers to keep for itself since the potential nuclear state would never have to pay the proliferation costs under such a deal. It therefore appears that the ineciency of weapons incentivizes both states to prefer an agreement. 4 The main theoretical section of this book proves the credibility of butterfor-bombs settlements. In equilibrium, for suciently high proliferation costs, the rival state makes immediate concessions to the potential proliferator. The potential proliferator accepts the concessions and never builds; although proliferating would yield further concessions, the additional concessions fail to cover the cost of construction. Proliferation only occurs when the investment cost is small. Yet, even in this case, states have mutual incentive to create non-proliferation institutions; doing so articially inates the investment cost so they may reach a settlement. Put simply, non-proliferation equilibrium outcomes generally Pareto-dominated proliferation outcomes. This has an important implication for the nonproliferation regime. By some accounts (Nye 1981; Sagan 1997; Dai 2007; Rublee 2009), the Nuclear 3 This amount represents 11% of all government spending over that time. The only programs that cost more were conventional military spending and Social Security (Schwartz 1998, 3). 4 Careful readers will note the similarity of this argument to war's ineciency puzzle (Fearon 1995), except applied to costly weapons construction instead of war.

21 CHAPTER 1. INTRODUCTION 7 Nonproliferation Treaty is one of the most successful international organizations ever created. 5 Only four countries have never signed (India, Israel, Pakistan, and South Sudan) and only one (North Korea) has ever withdrawn. Bargaining shows that states opt not to proliferate because of the bribes they receive from other parties; absent those bribes, many more states would leave the NPT and develop a nuclear deterrent. This helps explain the discrepancy between the number of nuclear capable states and the number of nuclear weapons states seen in Figure However, as a matter of empirical fact, states sometimes walk down the path toward proliferation. The task then is to explain these instances of proliferation within the context of the ineciency puzzle. Why do potential proliferators nuclearize when non-proliferation settlements exist that leave both states better o? This book provides three causal mechanisms. First, commitment problems preclude the rival from continuing to provide concessions once the potential proliferator can no longer credibly threaten to nuclearize. 7 In turn, this dynamic forces the potential proliferator to invest while it can to coerce concessions in the long run. Unsurprisingly, nuclearization results if the potential proliferator cares enough about the future. 5 See Mearsheimer (1993) for an opposing viewpoint. 6 Data on nuclear capable states come from Jo and Gartzke 2007, which I dene as states with latent capacity values of 6 or 7 on a 7 point scale. Data on nuclear weapons states comes from Gartzke and Kroenig Note that the potential proliferator does not face a commitment problem here. As later chapters will show, it is always willing to accept concessions in the long-term, even if it could freely proliferate after receiving a bribe.

22 CHAPTER 1. INTRODUCTION 8 50 Nuclear Trends 40 Number Legend Nuclear Capable States Nuclear Weapons States Year Figure 1.1: Nuclear capable versus nuclear weapon states over time.

23 CHAPTER 1. INTRODUCTION 9 Second, information problems lead potential proliferators to challenge rivals in the absence butter-for-bombs oer. Weak rival states have incentive to act tough and make no concessions, relying on the threat of a preventive strike to induce the potential proliferator to yield. Thus, the potential proliferator may develop nuclear weapons to test the rival's credibility. Stable butter-for-bombs agreements still exist here; it is just that the rival state has incentive to blu, causing the potential proliferator to occasionally challenge the status quo. Third, imperfect information has mixed eects. If the rival cannot eectively monitor the potential proliferator's decision to build, the potential proliferator faces great temptation to defect on butter-for-bombs agreements. If the cost of proliferation is great, the rival makes additional concessions to appease the potential proliferator. If the cost of proliferation is suciently low, however, the rival prefers preventive war to making the minimally acceptable butter-for-bombs deal. Proliferation and war result, though weapons inspections can alleviate the problem. 8 8 As in Fearon 1995, issue indivisibility provides another causal mechanism, but this is trivial and theoretically unimportant for the standard reasons.

24 CHAPTER 1. INTRODUCTION Competing Explanations and Why They Are Insucient Nuclear proliferation has received much attention from academics and policy makers alike. 9 Many theories of nuclear proliferation exist; none adequately accounts for the ineciency puzzle and therefore cannot fully explain why some states proliferate and others do not. Selecting on the dependent variable is a frequent issue. To wit, consider the standard realist account of proliferation. Thayer (1995, ) notes that every nuclear weapons state faced serious security concerns during the decision to proliferate. He then concludes that security is the only necessary and sucient cause of nuclear proliferation (486). Meanwhile, other scholars have recognized the importance of bargaining on an ad hoc basis, leading to nuclear restraint (Reiss 1988; Paul 2000; Levite 2003; Rublee 2009). Each side of this debate, however, only focuses on one side of the equation. 10 A theory for proliferation must explain nonproliferation, and a theory for nonproliferation must explain proliferation. A theory that can only do one holds no explanatory power. Moreover, even within the ad hoc bargaining literature it is unclear how and why inducements might deter proliferation or when and why negotiations may break down. 11 The lack of a comprehensive bargaining theory of 9 See Sagan 2011 for an overview of nuclear proliferation. 10 Hymans (2006) and Solingen (2007) are important exceptions that generate a theory accounting for both proliferation and non-proliferation. 11 Paul (2000), for example, implicitly treats bargaining as an important determinant of

25 CHAPTER 1. INTRODUCTION 11 nonproliferation means that no complete understanding of bargaining failure exists. Again, an ideal theory must explain both outcomes, not just the nonproliferation successes. Indeed, conventional wisdom (Sagan 1997, 57-61; Monteiro and Debs 2014) holds that the following assumptions are sucient to understand proliferation behavior: 1. The point of contention is constant sum. 2. A state's share of the point of contention increases when it has nuclear weapons. 3. The costs of proliferation are smaller than the gains made by increasing power. 4. Preventive war is not a viable option. However, as the previous section's thesis outlined and the central model will prove later, these assumptions merely outline why proliferation appears desirable; they do not imply proliferation. If these are the only constraints on bargaining, then states should negotiate their way to settlement. Nevertheless, research in this vein dates back to Thucydides and has been abundant in the recent past. 12 Neorealist theories dominate the paradigm. proliferation, but he fails to fully elucidate the mutual benet of bargaining or develop causal mechanisms for why bargaining fails. 12 A non-exhaustive list of traditional models of shifting power includes Morgenthau 1960, Organski 1968, Organski and Kugler 1980, Waltz 1979, Gilpin 1981, Howard 1983, Van Evera 1984, Van Evera 1999, Kennedy 1987, Levy 1987, Schweller 1992, Copeland 2000, Mearsheimer 2001, and Lee 2008.

26 CHAPTER 1. INTRODUCTION 12 This tradition emphasizes that maximizing security is a state's central priority. Thus, the rst assumption is sucient to understand proliferation, at least in the absence of a competing state's credible threat to initiate preventive war. As a result, the construction of weapons is usually taken as a given, and the focus instead turns to how states mitigate power transitions through military means. In eect, proliferation should not be surprising since it is merely a form of internal balancing (Lavoy 1993, 196). But these theories fail to address the ineciency puzzle and thus provide insucient causal explanations for proliferation. To explain the variation in proliferation outcomes, more recent scholarship has moved to a two-prong attack with quantitative and formal research. Quantitative studies fall into three approaches. First, supply-side scholars seek to explain proliferation behavior as a matter of capability. States that lack the physical means to create a nuclear weapon at a reasonable price are unlikely proliferators (Jo and Gartzke 2007; Meyer 1984). This is especially true if no status quo nuclear power is oering assistance (Kroenig 2009a; Kroenig 2009b; Kroenig 2010; Fuhrmann 2008; Fuhrmann 2009). But these studies critically underestimate the role of bargaining in the proliferation decision. While technological know-how is a prerequisite for nuclear weapons, states must have some coercive bargaining relationship to make a return on their investments. And since security is zero-sum, this opens up bargaining as an important determinant of proliferation, which the supply-side literature fails to address.

27 CHAPTER 1. INTRODUCTION 13 Second, demand-side scholars quantitatively investigate the traditional security side of nuclear proliferation. Intuitively, states with more intense security relationships ought to proliferate with greater frequency; Singh and Way (2004) verify this, showing that militarized interstate disputes and enduring rivalries are positively correlated with proliferation. Yet proliferation does not beget proliferation; Bleek (2010) nds a null result on how one rival acquiring nuclear weapons aects the other rival's proliferation decision. But again, desire for nuclear weapons is a necessary but not sucient condition for proliferation. A third group of quantitative scholars looks to leaders and regime type as the unit of analysis. Leaders with experience as rebels (Fuhrmann and Horowitz 2012) or personalist dictators (Way and Weeks 2014) may nd nuclear weapons more attractive and therefore are more willing to invest the costs necessary to proliferate. 13 Meanwhile, democracies are more likely to abstain from the NPT if they wish to explore nuclear options whereas nondemocracies are more likely to sign but then cheat (Fuhrmann and Berejikian 2010). However, leader- and regime-specic analysis still fails to explain proliferation behavior in the presence of a bargaining problem. 14 Alternatively, principal-agent problems and domestic political constraints may induce leaders to build nuclear weapons even when it goes against the 13 Relatedly, Hymans (2006) argues that nationalistic leaders are more likely to proliferate. 14 Given mechanisms for bargaining failure, though, these traits could explain when the mechanisms are more likely to are up.

28 CHAPTER 1. INTRODUCTION 14 national interest (Sagan 1996; Solingen 2007). This is sensible; principalagent problems ought to sabotage negotiations over nuclear weapons just as they sabotage other security issues. 15 Nevertheless, the unitary actor approach remains useful for two reasons. First, if leaders must weigh personal gains versus state interest, unitary actors results still apply (to varying degrees) when competing political forces within a state return to the model (Slantchev 2011, 9-10). Second, and perhaps more compelling, principleagent problems may be the source of commitment problems or incomplete information, which later chapters integrate into the bargaining environment. Finally, formal models of shifting power have neglected negotiating over potential power. 16 Instead, realized proliferation is a successful outcome for the rising state in the existing models. 17 In contrast, this book's main theoretical model shows such an outcome is Pareto inferior to equilibrium nonproliferation outcomes. Perhaps this is because key works (Fearon 1995; Powell 2006; Chadefaux 2011; Debs and Monteiro 2013; Reed, Wolford, and Arena 2013; Dittmeier 2013) wish to explain the outbreak of war. While that is a noble task, this book explains how states divide goods peacefully in the 15 Bueno de Mesquita et. al. 2003; Goemans 2000; Chiozza and Goemans 2011; Downs and Rocke 1994; Goemans and Fey See Bueno de Mesquita and Smith 2012 for an overview. 16 Fearon (1996) provides the closest exception. However, in his model, power is inherently tied to the bargaining good and not produced as a part of an inecient process. The focus here is on why states might produce power given that eciency. 17 This book adopts the rising state/declining state terminology from previous work on shifting power. Rising and declining are minimalist in this sense. They do not exclusively refer to major power transitions between world powers. Rather, they refer to any coercive bargaining relationships in which one side's outside option improves over time while the other's diminishes.

29 CHAPTER 1. INTRODUCTION 15 shadow of war. To accomplish that goal, I build on previous works by analyzing a model with an endogenous and costly power shift. To address concerns that anarchy implies unenforceable contracts, it allows the rising state to freely proliferate no matter the attractiveness of the declining state's oers. And to fully address commitment issues that may occur over the long term, the interaction repeats indenitely. Despite the apparent obstacles, equilibrium agreements exist in which the declining state oers immediate concessions, prompting the rising state to not build. All told, these previous attempts to explain proliferation behavior do not adequately address the ineciency puzzle. Building on the main theoretical model, the explanations in later chapters will. 1.3 Roadmap Going forward, this book relies on game theoretical models to understand the origins of nuclear ambition. Formalization is particularly useful for this subject for a couple reasons. First, bargaining is a key determinant of nuclear proliferation. The logic of bargaining behavior can grow convoluted quickly; formal modeling provides accounting standards (Powell 1999, 29-34) that ensure claims maintain internal validity. Second, few states have ever made serious attempts to proliferate; even fewer have actually successfully tested a nuclear weapon. Worse, nuclear

30 CHAPTER 1. INTRODUCTION 16 programs are closely guarded state secrets (Levite 2003, 64-66). Political scientists lack the classied knowledge of behemoth intelligence organizations. Formal theory resolves some of these issues by directly elucidating state incentives given a commonly understood set of assumptions. Game theory is not a perfect solution to all of the challenges facing a scholar of nuclear proliferation, but modeling provides a useful starting point for the analysis. To that end, this book's explanation of the successes and failures of butter-for-bombs deals consists of nine additional chapters. The next chapter analyzes the role of deterrent and compellent threats so that the models presented in future chapters have external validity. Despite substantive dierences, both deterrence and compellence share common formal representations. Consequently, the models of nuclear bargaining this book develops apply to both uses of atomic weapons. The third chapter introduces the core model of the book. Without bargaining frictions, only three outcomes are possible. First, if the rising state cannot credibly threaten to proliferateeither because the declining state would launch a preventive strike or the cost of nuclear weapons is too great the rising state receives no concessions. This helps explain why there have been so few preventive wars over nuclear weapons; if they would occur, the rising state internalizes the threat and terminates its program. If proliferation is a reasonable option but the costs are still prohibitive, the declining state oers immediate concessions and convinces the rising state not to seek nuclear weapons. Proliferation only occurs when the threat to obtain nuclear

31 CHAPTER 1. INTRODUCTION 17 weapons is credible and the costs are low. The fourth chapter searches for empirical evidence that the potential for proliferation is sucient to coerce concessions. To do so, it looks at target behavior during militarized interstate disputes. If a target is worried that resistance today might cause proliferation tomorrow and a dispute under worse terms, it might not reciprocate the initiator's belligerence. Indeed, this holds empirically: states with higher levels of latent nuclear capacity are signicantly more likely to induce their opponents to back down. While the fourth chapter investigates latent nuclear capacity on the macroscale, it lacks the ne-grained diplomacy of individual case studies. Thus, the fth chapter reviews the success of butter-for-bombs negotiations in practice. Specically, the chapter applies the theory to Egypt, a set of American Cold War allies (Japan, South Korea, and Australia), and the nuclear procient Soviet successor states (Belarus, Kazakhstan, and Ukraine). At various times, each of these states seriously considered developing a nuclear deterrent. But other actors ultimately oered concessions up front and convinced them to give up their aspirations, just as the butter-for-bombs model predicts. Yet butter-for-bombs agreements only work when the cost of proliferation is relatively large. What happens when it is small? Chapter 6 shows that declining states and rising states mutually benet from articially inating the costs of proliferation so as to encourage bargaining and avoid the ineciency. To that end, the chapter analyzes the successful negotiations between Argentina and Brazil in the 1980s.

32 CHAPTER 1. INTRODUCTION 18 Combined, the rst half of the book creates a proliferation puzzle that the remainder of the book seeks to answer. The seventh chapter investigates the stability of non-proliferation treaties in the shadow of war exhaustion. Declining states sometimes uctuate in their resolve to handle various international crises, which opens up windows of opportunity for rising states. Specically, rising states may choose to proliferate today if they expect that the declining state will launch preventive war if they attempt to proliferate at a later date. The declining state's inability to credibly promise to keep the concessions owing to the rising state in the future sabotages arms deals in the present. Proliferation results. Chapter 7 then applies these ndings to the Soviet Union's decision to proliferate in 1949 and Iran's push toward proliferation in the mid 2000s. The eighth chapter transitions to information-based explanations, beginning with when the declining state cannot eectively monitor the rising state's decision to build. Debs and Monteiro (2013) take a similar approach while disallowing butter-for-bombs bargaining. They nd that imperfect monitoring leads the rising state to secretly attempt proliferation, while the declining state disincentivizes investment by sometimes launching preventive war. However, incorporating butter-for-bombs bargaining allows the declining state to sidestep the monitoring problem by making generous oers to the rising state. Negotiations only fail when the extent of the power shift is great. But even then the states mutually benet from raising the costs of proliferation so as to encourage butter-for-bombs agreements. This re-

33 CHAPTER 1. INTRODUCTION 19 sult explains how International Atomic Energy Agency (IAEA) inspections succeed even when inspectors nd no weapons and know ex ante that no weapons exist. Consequently, imperfect information explanations are unconvincing. The model proves useful in understanding the Iraq War. Had imperfect information been the root cause, the United States would have stopped short of invasion in The ninth chapter introduces uncertainty regarding the eectiveness of preventive war. Even when preventive war is unattractive, the declining state might oer no concessions, hoping that the possibility of a strike will induce the rising state to not build. However, due to incentives to misrepresent, the rising state sometimes challenges the declining state by investing in nuclear weapons. Thus, investment occurs even though the declining state would retroactively prefer a butter-for-bombs deal. The chapter then traces the logic in Operation Opera, Israel's 1981 raid on an Iraqi nuclear reactor. The tenth chapter concludes the book, highlighting overall lessons, investigating policy implications, and exploring the generality of the models. Throughout the chapters, case studies accompany the formal models. These case studies serve a number of purposes. Primarily, they motivate the formal theory; without them, the numerical excursions are cumbersome, opaque mathematical expositions. With them, game theory answers critical questions about foreign policy. Understanding why Iran seeks nuclear weapons today, for example, rst requires recognition of its strategic constraints. Ultimately, this leads to the conclusion that American credible

34 CHAPTER 1. INTRODUCTION 20 commitment problemsnot Iranian commitment problemsis prompting Iranian belligerence. Second, the case studies provide new explanations for historical proliferation and preventive war behavior. Previously, scholars have argued that Soviet nuclear proliferation was an obvious byproduct of rising Cold War tensions and that the Iraq War began due to imperfect monitoring of Saddam Hussein's weapons of mass destruction programs. Applying the logic of the formal models raises strong objections to the validity of each of these theories and simultaneously produces compelling alternative mechanisms. Finally, the case studies illustrate the underlying logic of formal models. Game theory provides wonderful accounting standards but often lacks straightforward application. Consequently, each model has a corresponding case study that operationalizes critical explanatory variables. These case studies make a plausible argument that states behaved as the models predicted. In essence, they serve as an existence test of the models by plausibly demonstrating that the logical constructs have real world applications.

35 21 Chapter 2 Microfoundations: Nuclear Weapons and the Bargaining Model of War Looking at the history of nuclear proliferation, two stylized facts are clear. First, rising states benet from developing a nuclear arsenal, otherwise they would not spend billions of dollars on such projects. And second, declining states actively want to avoid proliferation outcomes and are thus willing to pay large costs to limit the number of nuclear states in the world. Combined, these two factors indicate that a complete theory of nuclear proliferation and nonproliferation explains why states reach and do not reach a bargain. As such, this book turns to bargaining theory to provide answers. However, before the book can begin the modeling process, it rst needs to establish why rising states want nuclear weapons and why declining states want to quash those programs; otherwise, the book risks modeling topics unrelated to the nuclear problem. Fortunately, a large literature explores these issues; unfortunately, researchers reach vastly dierent conclusions, potentially complicating the modeling process.

36 CHAPTER 2. MICROFOUNDATIONS 22 In brief, one side of the literature suggests that nuclear weapons coerce concessions from rivals, forcing those opponents to take actions that they otherwise would not. This can result from two separate mechanisms. First and most straightforward, nuclear weapons have incredible destructive power and thus are militarily threatening (Betts 1987; Trachtenberg 1985; Pape 1996, 35-38). Recognizing that nuclear weapons alter the balance of power, rivals shift concessions to nuclear states to avoid war. Second, nuclear weapons states can engage in brinkmanship, escalating crises and increasing the likelihood of an unavoidable nuclear strike against military or civilian targets. While such threats are dicult to make credible, the threat that leaves something to chance nevertheless encourages rivals to concede issues and end the possibility of disaster. Moreover, because nuclear weapons are so destructive, compellent coercion can succeed even if the chance of nuclear disaster is small at any given time. 1 And outside of the theoretical literature, policy makers have believed that nuclear compellance is an important bargaining strategy (Gaddis 1987, ; Truman 1955, 87). Indeed, Beardsley and Asal (2009) nd empirical support for nuclear weapons states prevailing more often in conict. 2 1 This literature is extensive and nuanced. Schelling (1960, ) gave the rst discussion of brinkmanship and threats left to chance. Other theorists have rened the concept in various ways (Jervis 1989; Powell 1987; Powell 1988; Powell 1989; Powell 1990), but a common theme remains: nuclear weapons force concessions. 2 Kroenig (2013) also nds empirical support for the theory that the size of the arsenal mattersstates with nuclear superiority over other nuclear-armed states fare better in crises. See Brodie 1959 ( ) for diculties in keeping war limited with atomic weapons, especially against other nuclear states.

37 CHAPTER 2. MICROFOUNDATIONS 23 The other side counters that compellent threats are not credible and that nuclear weapons only deter rivals; in other words, proliferation is merely useful in preserving the status quo. 3 From this perspective, successful compellence is challenging because nuclear weapons are perhaps tactically and strategically useless (Sechser and Fuhrmann 2013, ). To wit, because nuclear explosions poison the surrounding land with radiation, a state cannot credibly threaten to launch a nuclear attack on enemy forces unless those troops vacate a territorial possession. This forces the aggressor to instead utilize countervalue strategies; namely, a nuclear weapons state could threaten to bomb its rival's capital unless the rival surrenders outlying territories. But given the taboo against nuclear weapons (Tannenwald 1999), such countervalue strikes strain credulity due to the risk of international backlash. These nuclear skeptics also have some support among policymakers (Trachtenberg 1985, 137). From a modeling perspective, this debate is problematic because it seemingly forces the modeler to make a decision. Are nuclear weapons useful for compellent threats? Or are they only useful for deterrent threats? A model that speaks to one side might not speak to the other, forcing the modeler to develop two separate theories or risk alienating one side of the debate. Luckily, this chapter shows that the compellence versus deterrence debate is a false dilemma from a bargaining perspective. A single model can encapsulate both types of threats. Even better, that model is the standard 3 Compellence and deterrence date back to Schelling (1960, ; 1966, 69-72).

38 CHAPTER 2. MICROFOUNDATIONS 24 bargaining model of war (Fearon 1995). The logic is as follows. Credible compellent threats increase the value of p, the parameter normally used to represent one side's probability of winning a war. In contrast, credible deterrent threats do not eect the probability of winning but rather increase the value of a post-war settlement in the event of a loss since a conquering state must worry that a complete invasion will trigger a nuclear response. Rivals thus lower their war aims, in turn increasing the war payo for the state with the deterrent threat. However, this chapter shows that a better bargaining position is formally equivalent to having a larger value of p. Consequently, from a bargaining perspective, nuclear proliferation increases a proliferator's p value regardless of whether the mechanism is through compellent or deterrent threats (or both). As such, researchers ought to drop the interpretation of p as the probability of victory; rather, it is the average outcome of war. Using this modeling trick has two major benets. First, this chapter bridges the gap between nuclear threats and crisis bargaining so that the remaining chapters can stand on the shoulders of giants and incorporate decades of research into a theory of bargaining over proliferation. Second, the new theories this book develops are easily comparable to past work, making it easier to spot where conventional assumptions leads researchers astray. As such, the remainder of the book will adopt the formulation. This chapter proceeds as follows. The next section provides a simple motivating example to illustrate why the bargaining model of war implicitly

39 CHAPTER 2. MICROFOUNDATIONS 25 covers both compellent and deterrent threats. The section after generalizes the results from the motivating example. The chapter concludes with a brief discussion of how to apply the result. 2.1 Intuition: Credible Threats in Crisis Bargaining The conventional framework on the bargaining model of war is as follows. Two states, R (a rising state) and D (a declining state), bargain over a good standardized to value 1. If the states cannot agree on a negotiated settlement, they ght a war. R wins with probability p R and D wins with complementary probability; the victor obtains the entire good and the loser receives nothing. War is costly; states pay respective costs c D, c R 0 if they ght. Let the payo pair px, 1 xq be R's and D's respective shares of an ecient negotiated settlement. Working through their expected utility calculations, a division x p R c R satises R while 1 x 1 p D c D satises D. Thus, x is mutually satisfactory if and only if x P rp R c R, p R c D s. This interval is called the bargaining range. These war payos aect the credibility of threats. Let q be the status quo distribution of the good. If q falls within the bargaining range, compellent threats from both sides are inherently incredible. Although both would prefer shifting the status quo in their favor without escalating to war, suppose R demanded more of the good from D. When D decides whether to acquiesce, it

40 CHAPTER 2. MICROFOUNDATIONS 26 must anticipate R's reaction to noncompliance. Yet if D stands rm and R's only alternative is war, following through on the threat yields p R c R. This is less than what R receives from preserving the status quo. Consequently, D safely ignores the threat. Now imagine that R successfully developed nuclear weapons and could credibly threaten to use them oensively, as the compellence literature argues. Then the destructive power of nuclear weapons shifts p R to a larger value, say p 1 R. Here, a division x is mutually satisfactory if and only if x P rp 1 R c R, p 1 R c Ds. 4 Consequently, if the status quo division q is in rp R c R, p R c D s but not in rp 1 R c R, p 1 R c Ds, D must give R concessions to avoid war after R has proliferated. In turn, compellent threats follow the general framing of the bargaining model of war, namely that p R represents R's probability of victory in combat. Alternatively, scholars who argue that nuclear weapons are only useful for deterrent threats might take issue with that interpretation. Under their framework, nuclear weapons cannot shift the probability of victory and thus seemingly have no eect on the bargaining range. However, such an inference relies on too strict an interpretation of crisis bargaining's mechanics. While p R normally represents R's probability of victory, it is equivalent to the average outcome of war. Although this distinction is normally inconsequential, the dynamics of nuclear deterrence require 4 The values for the range assume that the costs of war remain static in this example. Relaxing this is inconsequential to the core argument.