PARSING THE RELATIONSHIP BETWEEN COUP ATTEMPTS AND CIVIL CONFLICT ONSETS By Petra Hendrickson A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of Political Science Doctor of Philoso phy 2015 ABSTRACT PARSING THE RELATIONSHIP BETWEEN COUP ATTEMPTS AND CIVIL CONFLICT ONSETS By Petra Hendrickson The separate developments of the civil war and coup literatures has meant that potential connections between the two forms of politic al violence have remained undertheorized and largely empirically unexplored. This dissertation seeks to remedy this, introducing three potential frameworks for understanding the relationship that exists between coup attempts and civil conflict onsets. The th ree proposed frameworks are the independence of events framework, the constraints framework and the trade - offs framework. The independence of events framework assumes no systematic relationship between the two phenomena, the constraints framework assumes a systematic relationship in which similar factors cause both events, and the trade - offs framework assumes a systematic relationship in which factors that make one event more likely make the other event less likely. In order to test these frameworks, I e mploy both large - N time - series cross - national bivariate probit analyses as well as qualitative analyses of Nigeria from 1966 - 1967 and Burma from 1948 - 1962. The results of the quantitative analysis provide strong support for the constraints framework. The parameter, which measures the relationship between the error terms of the two equations comprising the bivariate probit, is consistently significant and positive, suggesting that the unobserved factors that make one even more likely also make the other e vent more likely. Additionally, measures of model fit reveal that modeling the two events as part of the same underlying process provides more accurate coefficients and understanding of both phenomena than does modeling the two events as independent proces ses. The qualitative analyses provides a more nuanced approach. Indeed, the case studies reveal that the trade - offs framework is nested within the constraints framework the constellation of social, political and economic challenges facing a leader limit the range of possible policies they can implement, while the decisions they make about what policies to pursue within this range have the ability to manipulate the relative risks they face as well as the specific timing and order of coup attempts and civi l conflict onsets. Overall, this dissertation has sought to shed light on the similarities between the dynamics leading to coup attempts and civil conflict onsets. This is accomplished through new theorizing and the development of new frameworks concernin g the actors, preferences and interactions that tie the two processes together. Q uantitative and qualitative assessments of the theoretical frameworks reveal that not only are there systematic connections between the two phenomena, but that those connectio social, political and economic environment they find themselves confronted with. Copyright by PETRA HENDRICKSON 2015 v ACKNOWLEDGEME NTS I would like to thank my dissertation chair, Michael Colaresi, and my committee members, Jakana Thomas, Ben Appel and especially Ani Sarkissian, for their involvement in my dissertation journey. I would also like to thank my friends, sister and partn er for their constant encouragement and support throughout graduate school broadly and the dissertation process in particular. Melanie Bowers and Cameron Whitley have been consistent sounding boards, sympathizers and empathizers as we have made this journ ey together. They have laughed with me and let me cry, supported me and encouraged me when I doubted whether I would finish. They were some of my first true friends in graduate school, and have been there for all my highs and lows, all the peaks and valley s of the process, and their presence alone, to say nothing of their active kindness, support, love and willingness to welcome me into their family, has been the rock of my graduate school journey. Laurel Burchfield, Aaron Passman, Kate Smith - Buckler and J ohn Buckler have been welcome diversions from graduate school and the dissertation, thoroughly introducing me to the provided an outlet and a release, encouraging me to think about things other than my topic, but also to carry on and complete my studies. Anne Hunter has been my expert guide on the graduate school process, and always had encouraging and comforting words and support when I was feeling most discourage d. She has vi Final and particular gratitude is extended to Micah Holmquist, who despite an initial inadvertent poke at my thought process, has provided nothing b ut enduring love, support, encouragement and welcome distractions throughout some of my wilderness years and the final push to begin, complete, finish and defend my dissertation. He has never questioned my ability to complete my dissertation, even when I h ad doubts. He has cheered me on through my triumphs and held my hand through my setbacks, and his unwavering confidence in me helped convince me to believe in myself as I neared the end of my graduate school journey. vii TABLE OF CONTENTS LIST OF TAB LES xii LIST OF FIGURES xiv CHAPTER 1: ESTABLISHING CONNECTIONS BETWEEN CO UP ATTEMPTS AND CIVIL CONFLICT ONSETS 1 Introduction 1 Conceptual Commonality and Empirical Difference 4 Underlyi ng Characteristics of the State 4 Ethnicity 4 State Capacity 6 Legacy Effects 7 Additional International - Domestic Linkages 7 Three Models Broadly Consi dered 10 Independence of Events 10 Constraint s 11 Trade - Offs 12 Outline of the Dissertati on 14 CHAPTER 2: MODELING THE RELATIONSHIP BETWEEN C OUP ATTEMPTS AND CIVIL CONFLICT ONSETS 17 Introduction 17 The Common Framework: The Selecto rate Theory 18 Actors 18 Preferences 18 Interactions 19 Coup Attempt s and Civil Conflict Processes 23 Actors 23 Preferences 26 Interactions 29 Trade - Off s 29 Constrain ts 39 Independence of Events 49 Conclusion 50 A PPENDIX 52 CHAPTER 3 : QUANTITATIVE ASSESSMENTS OF THE RELATIONSHIP BETWEEN C O UP ATTEMPTS AND CIVIL CONFLICT ONSETS 56 Introduction 56 Primary Methodology: Bivariate Probit 58 viii Powell and Buhaug Replications 60 Variables Employed in Powe ll (2012) and Buhaug (2006) Replications 62 Variables Employed by Powell (2012) 62 Coup Attempt 62 Change in Military Expenditures 63 Soldier Quality 63 M ilitary Personnel 63 Effective Number of Military Organizations 63 Change in GDP Per Capita (t - 1) 64 GDP Per Capita (t - 1) 64 Instability 64 Dem ocracy and Autocracy 64 Military Regime 64 Peace Years and Peace Years Spline 65 Variables Employed by Buhaug (2006) 65 Civil Conflict Onset 65 Demo cracy (t - 1) and Mixed Regime (t - 1) 65 GDP Per Capita (t - 1) 65 Oil Exporter 65 Country Land Area 66 Ethnic Fractionalization 66 Conflict Onset De cay Function 66 Additional Observable Indicators for Testing the Constraints and Trade - Offs Models 66 Dependent Variables 66 Civil Conflict Onset 66 Coup Attempts 67 Independent Variables 68 Ongoing Civil Conflict 68 Previous Civil Conflict Involvement 69 Credible Rebel Group 69 International Threats and D isputes 70 Political Exclusion 71 Government Spending 72 Military Readiness 73 Regime Type 74 Capacity 75 Instability 76 Repression 77 Sample of Cases in Original Quantitative Analyses 77 Modified and Extended Replication Results 77 Extensions to Powell and Buhaug Models 81 Model Fit 89 Discussion 97 ix Conclusion 100 A PPENDIX 103 CHAPTER 4 : INTRODUCTION T O T HE CASE STUDIES 141 Introduction 141 Case Selection 142 Background/Brief Historical Lead - Up to the Period u nder Consid eration 142 Broad Expectations of the Three F ram eworks 143 Independence of Events 143 Constraints 144 Trade - Offs 145 Narrative of Period under Consid eration 146 Assessment of the Framewor ks 146 Nested Frameworks 146 Conclusion 147 CHAPTER 5 : NIGERIA, 1966 - 1967 148 Introduction 148 Co lonial Era and First Years of Independence 148 Historical Background 148 Assessment of the Frameworks 152 Independence of Events 152 Constraints 153 Trade - Offs 155 Deterioration Leading to the January 1966 Coup 156 Historical Background 156 Assessment of the Frameworks 160 Independence of Ev ents 160 Constraints 161 Trade - Offs 163 Fallout from the January 1966 Coup and the Slide to Civil War 166 Historical Background 166 The January Conspirators 169 Ironsi and the Unification Decree 170 May 1966 Anti - Ibo Riots 171 July 1966 Counter - Coup 172 September 1966 Anti - Ibo Riots 174 Assessment of the Frameworks 175 Independence of Events 175 Constraints 176 Trade - Offs 178 Final Descent to Civil War 181 Historical Background 181 x Assessment of the Frameworks 183 Independence of Events 183 Constraints 184 Trade - Offs 18 5 Over all Assessment of the Frameworks 187 Conclusion 189 CHAPTER 6 : BURMA, 1948 - 1962 192 Introduction 192 Colonial Era 193 Historical Background 193 Assessment of the Frameworks 198 Independence of Events 198 Constraints 200 Trade - Offs 204 Immediate Civil War 205 Historical Background 205 Assessment of the Frameworks 208 Independence of Events 208 Constraints 209 Trade - Offs 210 The Uneasy 1950s 213 Historical Background 213 Continued Malaise 213 1956 Elections 214 AFPFL Split 215 Caretaker Government 218 1960 Election Campaign 221 Assessment of the Frameworks 223 Independence of Events 223 Con straints 224 Trade - Offs 227 Final Decline to the 1962 Coup 232 Historical Background 232 Assessment of the Frameworks 234 Independence of Events 234 Constraints 235 Trade - Offs 237 Overall Assessment of the Frameworks 239 Conclusion 240 CHAP TER 7 : LOOKING BACK AND MOVING FORWARD 244 Introduction 244 xi Primary Contribution: Theoretical Approa ch and Nuance 245 Summary of Statistical Findi ngs 247 Comparison of Case Studies: Similarities between Nigeria and Burma 249 Limitations to the Present Study and Possible Future Research 252 Implications 254 BIBLIOGRAPHY 256 xii LIST OF TAB LES Table 2.1. Summary of Expectations of the Trade - Offs Framework 53 Table 2.2. Summary of Expectations of the Constrai nts Framework 54 Table 2.3. Summary of Expectations of the Independence of E vents Framework 55 Table 3.1. Powell (2012) Modif ied Replication 104 Table 3.2. Buhaug (2006) Modi fied Replication 105 Table 3.3. Seemingly Unrelated Bivariate Probit Combination of Powell (2012) and Buhaug (200 6) Modified Repl icati ons 106 Table 3.4. Powell Modified Replication wi th More Restricted Sample 109 Table 3.5. Buhaug Modified Replication wit h More Restricted Sample 110 Table 3.6. Seemingly Unrelated Biv ariate Probit Combination of Powell (2012) and Buhaug (2006) Modified Replications wi th More Restricted Sample 111 Table 3.7. Coup Attempt Analysis with Additional Covariates with More Restricted Sam ple 114 Table 3.8. Civil Conflict Onset Analysis with Additional Covariates with More Restricted Sam ple 115 Table 3.9. Coup Attempt and Civil Conflict Onset Bivariate Probit with Additional Covariates with More Restricted Samp l e 116 Table 3.10 . Coup Attempt Analysis with Additional Covari ates with Unrestricted Sample 119 Table 3.11. Civil Conflict Onset Analysis with Additional Covariates with Unr estricted Sample 120 Table 3.12. Coup Attempt and Civil Conflict Onset Bivariate Probit Analysis with Additional Covariates with Unr estricted Sample 1 21 Table 3.13. Bivariate Probit Log - Likeli hood and Information Criteria for Combination Powell and Buhaug Modifie d Replications 124 xiii Table 3.14. Bivariate Probit Log - Likelihood and Information Criteria for Combination Powell and Buhaug Modified Replications and Model with Additional Covariates with More Restricte d Sample 125 Table 3.15. Bivariate Probit Log - Likelihood and Information Criteria for Combination Powell and Buhaug Modified Replications and Model with Additional Covariates with Unrestricted Sample 12 6 xiv LIST OF FIGURES Figure 3.1. Plot of Coefficients and Confidence Intervals for Coup Attempts in Separate Powell Model and Simple Powell - Buhaug Combination Model 107 Figure 3.2 . Plot of Coefficients and Confidence Intervals for Conflict Onsets in Separate Buhaug Model and Simple Powell - Bu haug Combination Model 108 Figure 3.3. Plot of Coefficients and Confidence Intervals for Coup Attempts in Separate Powell Model and Powell - Buhaug Combination Model with More Restricted Sample 112 Figure 3.4. Plot of Coefficients and Confidence Intervals for Conflict Onsets in Separate Buhaug Model and Powell - Buhaug Combination Model with More Restricted Sample 113 Fi gure 3.5. Plot of Coefficients and Confidence Intervals for Coup Attempts in Separate Probit Model with Additional Covariates and Bivariate Probit Model with Additional Covariates with More Restric ted Sample 117 Figure 3.6. Pl ot of Coefficients and Confidence Intervals for Conflict Onsets in Separate Probit Model with Additional Covariates and Bivariate Probit Model with Additional Covariates with More Res tricted Sample 118 Figure 3.7. Plot of Coef ficients and Confidence Intervals for Coup Attempts in Separate Probit Model with Additional Covariates and Bivariate Probit Model with Additional Covariates with Unr estricted Sample 122 Figure 3.8. Plot of Coefficients and Confidence Intervals for Conflict Onsets in Separate Probit Model with Additional Covariates and Bivariate Probit Model with Additional Covariates with Unres tricted Sample 123 Figure 3.9. Separation Plots of Powell Coup At tempts 127 Figure 3.10. Separation Plots of Bu haug Conflict Onsets 128 Figure 3.11. Separation Plots of Coup Attempt s with More Restricted Sample 129 Figure 3.12. Fit of Each Model on Five Worst - Fitting Coup Attemp t Cases from Powell and Powell - Buhaug Models 130 Figure 3.13 . Fit of Each Model on Five Worst - Fitting Coup Attempt Cases from Separate and Bivariate Model s with Additional Covariates 131 xv Figure 3.14. S eparation Plots of Conflict Onsets with More Restricted Sample 132 Figure 3.15. Fit of Each Model on Five Worst - Fitting Conflict Onset Cases from Buhaug and Powell - Buhau g Models 133 Figure 3.16. Fit of Each Model on Fi ve Worst - Fitting Conflict Onset Cases from Separate and Bivariate Models with Additional Covariates 134 Figure 3.17. Separation Plots of Coup Attem pts with Unrestricted Sample 135 Figure 3.18. Separation Plots of Civil Conflict On sets with Unrestricted Sample 136 Figure 3.19. ls for Bivariate Prob it Models 137 Figure 3.20. Marginal Effect of Instability on Probability on Coup Attempts and Civil Conflict Onsets 138 Figure 3.21. Marginal Effect of Repression on Probability on Coup Attempts and Civil Conflict Ons ets 139 Figure 3.22. Marginal Effect of Years of Repression on Probability on Coup Atte mpts and Civil Conflict O nsets 140 1 CHAPTER 1 ESTABLISHING CONNECTIONS BETWEEN COUP ATTEMPTS AND CIVIL CONFLICT ONSETS Introduction Although some might suggest that political violence is dissipating around the globe (Goldstein 2011; Pinker 2011), notable and recent examples of political violence around the world from the Boko Haram kidnapping of school girls in Nigeria, sectarian violence in the Central African Republic, police - community confrontations in the United States, reaction to renewed Israeli campaigns in Gaza and the mounting threat of the Islamist State organization to a number of social groups in Iraq and Syria to religious minorities and Western journalists alike serve as a reminder that the study of the causes and consequences of political violence, whatever form it may take, is just as important as ever. Moreover, the variance in conflict its timing, location, targets and forms are still relative mysteries, as is starkly illustrated by are particularly useful and consistent predictors of civil war. One reason that more progress has not been made could be that the literature has tended to analyze distinct forms of conflict as separ able and discrete categories of events. For example, there are well - developed literatures on international conflict and civil conflict, and it is only recently that scholars have begun to analyze their interconnections with sustained attention (Buhaug and Gleditsch 2008; Chiozza et al. 2004; Cunningham and Lemke 2013; Gleditsch 2007; Gleditsch, Salehyan and Shultz 2008; Gleditsch et al. 2010; Thyne 2006). 2 Thyne (2006) and Gleditsch (2007) both examine how interstate relationships and events can contribute involvement in a potential conflict in another, Thyne finds that costly sign als have no effect on the likelihood of civil war, while anti - government cheap signals on the part of the interstate opponent are much more disruptive, reducing the bargaining space between government and opposition and making civil war onset more likely. Gleditsch (2007), on the other hand, argues and finds support for the proposition that the regional environment states are situated within can have an impact on the likelihood of civil war, and that exclusively domestic factors cannot account for the onset of civil wars. For instance, as Buhaug and Gleditsch (2008) note, the presence of ethnic kin across a proximate border makes civil war onset more likely in a state, because ethnic rebels have an external base of support not subject to the same constraints and hardships they are. Gleditsch, Salehyan and Shultz (2008), on the other hand, develop an argument for intrastate - interstate linkages, noting that countries engaged in civil war are much more likely to also become involved in international disputes. Gi ven the important role transnational ethnic kin can play, it should be no surprise that interstate tensions may arise as a result of a civil war. Salehyan (2008) likewise finds that transnational ethnic kin participating in a civil war can encourage inters tate conflict between the target and refuge states. Similarly, even when we focus on domestic conflicts over control of a regime, the literatures on coups and civil conflicts have been bifurcated: knowledge about the causes and consequences of coup attemp ts remains largely separate and distinct from knowledge about the causes and consequences about civil war onsets. To the extent that these literatures overlap, mainly through studies examining one of the events but choosing simply to control for the other, we have 3 reason to believe there are interdependencies. However, simply grafting one event onto a model of the other for example, adding civil war as a predictor of coups is unlikely to answer questions about why these interconnections are present and what they tell us about the dynamics of political conflict and the chances at reconciliation. In addition, keeping analyses of coups and civil conflicts as separate literatures leaves scholars wondering whether results are complementary or divergent and ho w different models and scopes of analyses could contribute to a more holistic and integrated understanding of politics around the globe. To help fill this gap and explore the relationship between distinct forms of political conflict, in this chapter I th eorize about the possible connections between coup attempts and civil war onsets in the existing literature. Although both coup attempts and civil conflict onsets are relatively rare events 1 , they collectively represent the most severe repercussions rulers may face when they lose their ability and legitimacy to maintain control over the totality of their population and territory (in the case of civil conflicts) or even within the innermost circle of their base of support (in the case of coup attempts). This assessment of the existing literature helps set the stage for the remainder of the dissertation, which establishes that coup attempts and civil conflict onsets are best understood as layered processes leaders are faced with a constellation of social, po litical and economic challenges, and the policies they pursue against this backdrop have the potential to heighten or dampen the relative risks they face and the order and timing of the coup attempts and/or civil conflict onsets they ultimately experience. 1 There were 29 instances of a coup attempt and civil conflict onset occurring in the same year between 1975 and 2005, the years under consideration in the original analysis developed later in this chapter. 4 Conceptual Commonality and Empirical Difference To the extent that the separate literatures on coup attempts and civil war onsets explore the same potential causes, there are some crucial similarities. Studies of both tend to focus on underlying charac teristics of the state, and how those characteristics structure social interactions. In addition to coming from the same general theoretical position, the empirical results of many of these studies coincide with one another. Other work that links internati onal incentives and events to coups or civil conflicts is less common, but has begun to illuminate some patterns for analysis. Underlying Characteristics of the State The most obvious connection in preexisting research on coups and civil wars is that the y both commonly use similar sets of predictors. For example, studies of one or the other tend to use state capacity or ethnic diversity to explain the outbreak of an event of interest. However, the theoretical mechanisms linking these measures to violence have tended to differ across literatures. Ethnicity Concerns about the potentially destabilizing effects of ethnic diversity have loomed large in the intrastate conflict literature, sparking some theorizing in the coup literature as well. Measures of eth nic diversity have often been used in intrastate conflict studies as proxies for the degree of ethnic tension, if not outright conflict, within a country (Collier and Hoeffler 2004; Collier et al 2009; Fearon and Laitin 2003). A number of studies (see Coll ier 2001; Elbadawi and Sambanis 2002; Sambanis 2001) have identified potential problems with using a measure of ethnolinguistic fractionalization to tap ethnic tensions, an issue revisited in more detail in Chapter 3. Coinciding 5 with this controversy, ther e has been little empirical support for the idea that measures of ethnolinguistic diversity within a country are significantly related to conflict onset. The use of a new measure of ethnic dominance, however, has begun to uncover more consistent evidence on the relationship between ethnicity and conflict. Diversity in itself should not necessarily be expected to have much influence on conflict. After all, if a country is comprised of a large number of ethnic groups, none of them very large, the level of et hnolinguistic fractionalization will be high, but it is unlikely that one of those numerous small groups will be able to gain and retain repressive or violent control over the others. Rather, those states that do have such a sufficiently large ethnic group should be more expected to have political and economic tensions between groups. When one group is large enough to be able to gain and retain power and control without the participation and input of other groups (and perhaps at the expense of other groups) , violent interactions should become more likely as smaller groups lash out against their marginalization. And indeed, studies using ethnic dominance as a measure do find that it is associated with civil conflict onset (see Collier and Hoeffler 2004; Colli er at al. 2009; Montalvo and Reynal - Querol 2005); the political exclusion that ethnic dominance facilitates outbreaks (see Wimmer, Cederman and Min 2009) and in the case of minority rulers, may necessitate (see Fearon et al. 2007) is also a trait associate d with civil war. In the coup literature too, questions of the impact of ethnicity have begun to arise, with ethnic diversity assumed to reflect underlying sociopolitical instabilities (e.g., Jackman 1978; Johnson et al. 1984). Jackman (1978), for instanc forming 6 diversity might also be all the more powerful because when coupled with sociopolitical mobilization, ethnicity serves to form the foundation for societal identity, thus highlighting and emphasizing such differences among groups seeking power. The coup and civil war literatures diverge from one another in their conclusions about the role and impact of ethnicity, however. Auvinen (1997), for example, finds that coups and civil wars are more likely under different conditions of ethnic distribution in a country: ethnic dominance makes civil war more likely (a finding confirmed by others, like Collier and Hoeffler 2004; Collier et al. 2009; Fearon and Laitin 2003), following the logic outlined above. However, more likely to foster coups (a finding confirmed by Jenkins and Kposowa 1990, 1992; Kposowa and Jenkins 1993; Roessler 2011). Here, it is expected that the more control a leader has over her winning coalition, as through ethnic stacking, the less likely that leader is to experience an uprising from that winning coalition. State Capacity In the intrastate conflict literature, s tate capacity is a ubiquitous concept. No examination of either civil war or coups is complete without it (Collier and Hoeffler [1998] is one of the first studies to consider economic causes of civil war; in the civil war literature, see also Blattman and Miguel 2010; Chassang and Padró i Miquel; Collier 2007; Collier et al. 2003; Fearon 2008. In the coup literature, see for example Belkin and Schofer 2003, 2005; Londregan and Poole 1990), and there is general consensus that states with higher GDP are less vulnerable to coup attempts and civil war onsets. States with weak capacity are more vulnerable to any challenges because they lack the 7 resources to effectively bolster themselves. These challenges may arise from the military or rebel groups, each seeking to maximize their access to and control of the power structures of the state. Legacy Effects Although state capacity and ethnic diversity have been the most consistent recurring themes in the coup and civil war literatures, research has also considered t he potential legacy effects of each event: are coups/coup attempts and civil war onsets one - time affairs, or do they have some broader impact on society? The finding that states that have recently experienced a civil war are more vulnerable to a recurrence is widely accepted (see, for instance, Collier 2007; Collier et al. 2003; Quinn et al. 2007; Walter 2004), though there is debate over the exact reasons for that vulnerability. The relationship between coup attempts and future coup attempts is a bit more complex and raises questions about the role of various forms of instability on coup attempt occurrence. For example, there is strong evidence that once a leader falls victim to a successful coup attempt, successful coup attempts are more likely to occur in the future (Belkin and Schofer 2003, 2005; Jackman who are able to withstand a coup attempt (that is, the coup attempt fails) are less likely to be unseated by a coup attempt in the future. Additional International - Domestic Linkages bridge a gap between phenomena traditionally considered international in nature and it s domestic consequences. Involvement in such an international dispute can in itself be interpreted as a policy 8 failure, given that all wars are costly, albeit to varying degrees, and that there would have been some range of compromises that would have left both parties satisfied enough to not pursue a revisionist conflict (Fearon 1995). Accurate inventories of military strength and capability are private information, which means that actors are unable to make fully - informed decisions about their chances of victory should they decide to pick a fight or the costs they may face if they opt to retreat. Absent this information about capacity and capability (and thus their prospects for success if such an attack breaks out), it is much more difficult, if not impos sible, for states to reach mutually acceptable agreements and avoid spiraling into an otherwise - preventable conflict. 2 Nonetheless, the outbreak of war is a policy failure, and leaders who find their states engaged in war should expect a greater chance of punishment. One shortcoming of this strand of literature is that much of it is focused on the impact of international conflict on an incumbent (Bueno de Mesquita and Siverson 1995; Bueno de Mesquita, Siverson and Woller 1992; Blimes 2009; Chiozza and Go emans 2004; Colaresi 2004a, 2004b; Gelpi and Grieco 2001; Goemans 2000, 2008; Wolford 2007), with no real similarly well - developed literature looking civil war and other domestic factors. 3 An exception is Desch (1999), who takes on the question of civil - mi litary relations in times of both internal and external threat and finds that civilian control of the military tends to be stronger when there is low or no internal threat, but high external threat. Given this, leaders are less likely to face removal in ti mes of international conflict, but face a higher risk when there are other forms of domestic turmoil. 4 2 Other potential causes of bargaining breakdowns are credible commitment problems and issue indivisibility. 3 But Bienen and van de Walle (1989, 1992) buck this trend, and Bueno de Mesquita and Smith (2009, 2010) explore the consequences of rents on revolutionary threats within states. 4 This logic is supported by Colaresi (2004a), who finds that leaders who face instability as a res ult of war costs and external threats are less likely to lose power than leaders who face only war costs. Likewise, Blimes (2009) finds that leaders who reciprocate challenges are less likely to be removed from power and Goemans (2008) finds that initiatin g conflict (as opposed to being the target) decreases the risk of turnover. These findings all suggest that leaders can use external threats to their survival advantage. 9 A second shortcoming is that even to the extent that there is consistent evidence that civil wars make leaders more vulnerable to removal (Arriola 2009; Chiozza and Goemans 2004; Debs and Goemans 2010), the findings for international conflict are more varied, and are conditional on regime type. To the extent that democratic regimes have been investigated, the (mis)fortunes of war seem to affect elected lea ders less than they do nondemocratic leaders defeat makes regular removal of democratic leaders from office more likely, but democratic incumbents do not benefit much in terms of job security as a result of a victory. 5 For nondemocratic leaders, conflict outcomes strongly affect the likelihood of removal from office, with losses making turnover particularly irregular turnover (i.e., coups) more likely (Debs and Goemans 2010). Most vulnerable, though, are leaders of mixed regimes, who face the highest risk of post - tenure punishment after even a moderate war loss. It is worth noting, however, that leaders do not face additional costs from wars as compared to militarized crises (Chiozza and Goemans 2004). Whatever the costs faced by incumbents following an interstate conflict, they are likely to be amplified following an intrastate conflict, given that any war costs (in terms of battle deaths, increasing financial burden, destruction of infrastructure and civilian casualties) will be more visible to all s egments of the population, whereas they might be more concentrated to a few localized areas in the case of interstate war, such as areas directly adjacent to border skirmishes. The relationship between war performance and the status of the incumbent vis - à - vis the military is not an automatic one. As in the case of interstate conflicts, not all conflict losses hurt incumbents uniformly, nor do all conflicts victories help them uniformly. A sweeping victory will not likely result in special protection of demo cratic incumbents by the military as it might for autocratic 5 Nondemocratic leaders consistently suffer as a result of war losses, though militar y, civilian and monarchical dictatorships, like democratic regimes, do not receive much benefit from a war victory (Chiozza and Goemans 2004; Debs and Goemans 2010). 10 leaders, and especially severe losses by autocratic leaders may overwhelm the benefits the military accrues from the provision of private goods and encourage a military revolt against the leader, while simply making the public policy offerings of a democratic challenger more appealing than the continued public policy provisions of a demonstrably incompetent incumbent. Three Models Broadly Considered Based on the theories and evidence from the li terature, we can categorize the potential relationship, or absence of such a connection, between coup attempts and civil conflict onsets into three types or models. These frameworks have not been explicitly discussed in the literature previously, highligh ting the lack of theorizing about these connections. The three types of connections, then, are models that contribute to the study of the coup attempts and civil conflict onsets. Independence of Events The first model encapsulates the argument that ther e exists no relationship between the two phenomena: coup attempts and civil conflict onsets are independent events, unaffected by one another. The preponderance of the literature on these phenomena, for all intents and purposes, is a reflection of this mod el. 6 However, even cursory theorizing suggests that this assumption of independence between the occurrence of coup attempts and civil conflict onsets is problematic, if not implausible. Both are a sign of, and the product of, a breakdown in the legitimacy of the regime, albeit vis - à - vis two different sectors of the population. Rebellions (a necessary predecessor to civil wars) result from a loss of popular legitimacy, while coup attempts signal that the incumbent no 6 Most research makes no comment on any connection between the two. A much smaller body o f literature suggests such a relationship might exist, but does so only anecdotally, with neither a clear articulation of the specific nature of the relationship nor an empirical analysis that demonstrates that such a relationship does indeed exist 11 longer has the confidence of his 7 own win ning coalition. The latter is particularly telling, given that members of the current winning coalition have no way to guarantee that a new leader will a) be any more competent or b) include members of the current winning coalition in the new winning coali tion. Although significant, the latter concern really only applies to civilian elites 8 , while the competence of the incumbent, or potential replacements, will always be of concern to the military. The competence, or perceived competence, of the incumbent i s particularly relevant to the military frequency of challenges to the leader (Wolford 2007) and, by extension as representatives of the state, the military. Indee d, scholars like Posen (1984) argue that competent civilian leadership is required for military effectiveness. However, because this model has not been specifically tested against the alternatives (discussed below), the implausibility of this lack - of - conne ction remains a conjecture. Constraints The second model highlights the role of political and social constraints. Here, coup attempts and civil conflict onsets share the same general causes, which primarily take the form of limitations on the state, be it low income and economic growth or extreme sociopolitical divisions. According to this model, using the resources at their disposal, leaders trying to avert crises that threaten their rule, may use those resources successfully and not experience either a coup attempt 7 The d ecision to use the masculine pronoun is deliberate: the vast majority of leaders, historically speaking, have been male. 8 et al. 2003); th winning coalition, their continued inclusion is contingent on their support for the incumbent. These members are inclined to remain loyal, as doing so vi rtually guarantees that they will continue to receive private goods, a promise that any potential challenger is unable to credibly make; once in office, the new incumbent can renege on all such promises, with the now - excluded civilian elites all but powerl ess to do anything about it. 12 or a civil conflict onset. On the other hand, if their states are especially weak and they are particularly limited in terms of the policies they can enact, leaders may face the risk of both coup attempts and civil conflict onsets, but be ab le to address neither adequately. In this case, we would expect to see both civil conflict onsets and coup attempts occur in similar circumstances. Further, knowing when a coup was likely could be informative as to when a civil war might also be likely. Ye t importantly, this perspective suggests that there is little a leader can do to mitigate one risk over another, as both are assumed to be caused by similar social and political forces. Trade - Offs The third model brings to the forefront the potential trad e - offs that leaders may face in response to risks of coups versus civil wars. Leaders confronted with the risk of a coup attempt or civil conflict are likely to respond with policies that will increase their chances of survival. According to the framework of this model, however, in attempting to prevent one attack on their regime, they may actually increase the chances they will face the other threat as well. In these cases, leaders are successfully able to prevent the first phenomenon, but in doing so actu ally foster the second; in preventing a civil conflict onset, leaders fall victim to a coup, for instance, or they are able to avoid a coup, but in so doing provoke a civil conflict. If this third model is operable, we might observe policies designed to de crease coup risk instead increase the probability of civil conflict. It is this form of connection that Pilster and Böhmelt (2011) attempt to test when analyzing the consequences of military counterbalancing creating additional and alternative paramilita ry organizations - offs model: leaders may be able to put off the occurrence of a coup, but in doing so (and with varying degrees of 13 consciousness) i ncrease the risk of civil war. In coming to power, especially if they do so through extraconstitutional means, leaders may use the assistance of other ethnic groups, who then become members of the ruling coalition. Leaders may soon feel threatened by these additional groups, who groups from rising to power, he expels all ethnic groups but his own from the ruling coalition. This essentially eliminates the risk of a coup, but carries with it an increased risk of civil war instigated by the expelled groups. These groups, no longer able to extract concessions as part of the government, are forced to resort to more drastic measures to obtain favorable police outcomes. tampering with the make - up of the ruling coalition, leaders are able to manipulate the likelihood of a coup, which has a direct bearing on civil war risk, as i t is the method of coup - risk mitigation consciously exchange risk. By expelling other ethnic groups from their winning coalitions, leaders know they are increasin g the chances of broader discontent within those groups. While Roessler (2011) addresses exchanging coup risk for civil war risk, he does not address the possibility of a relationship in the opposite direction, where leaders, in effect, trade civil war r isk for coup risk. That is, leaders may be able to reduce the risk of a civil war but increase quite important, since it lays out for the first time an expl icit explanation for why and how these were centered on sub - Saharan Africa, it is possible both to expand his logic and explore the scope of his empirics. 14 Outline of the Dissertation In this dissertation, I will be examining which of these three models is most useful for thinking about the relationship between civil wars and coups. In order to accomplish this, employ a mixed - methods approach to testing the theory I develop. Chapter 2 contains the theoretical model, focused on modifying the a selectorate model developed by Bueno de Mesquita et al. (2003) , which provides a common framework and language to discuss the new theoretical models I develop as well as reconcile the seemingly contradictory findings that emerge between the quantitative and qualitative analyses. For example, the types of coalition , as I illustrate below . I then focus on modifying the selectorate theory to apply more directly to cases involving the outbreak of political violence. After those modifications have been discussed and incorporated into a broader theoretical base, I explicitly consider the three potential forms of the relationship between coup attempts and civil conflict onsets. First, and as implicitly argued by the vast majority of literature on the two topics, there could be no relationship all -- what I term the independence of events f ramework. To the extent that both phenomena occur (or do not occur) in a state, it is purely coincidental: there are no underlying similarities or differences in the risks or occurrences of the two events. The second framework is the constrain framework, w hich argues that there is indeed a systematic relationship between coup attempts and civil conflict onsets, and it stems from similarities in the underlying causes of the two events: the same types of factors and conditions that put a state at higher risk of one also put it at higher risk of the other. The final framework is the tradeoffs framework, which also conceives of a systematic relationship between coup attempts and civil conflict onsets, but one in which the underlying causes are in contrast to one another. 15 That is, factors and conditions that put states at higher risk of one event decrease the risk of the other. Chapter 3 contains the statistical assessment of the theoretical frameworks. I conduct several bivariate probit analyses to systematically test the relationship between coup attempts and civil conflict onsets. I first construct models that combine past research on coup attempts and civil conflict onsets to establish that even without the possibility of a connection being held in mind by the researchers, there is evidence that the two phenomena are related and more exploration is merited. I then construct my own models with additional covariates that bear both events in mind. The models both those constructed from previous research and those developed for the purposes of this study suggesting that the constraints model is most useful: there is an underlying relationship between the two phenomena, with the caus es of one similar to the causes of the other. An additional significant contribution of this modeling approach is that my statistical models fit the data more accurately. That is, including information typically associated with coup attempts in civil confl ict onset equations and vice versa enhance the explanatory power of the models. Chapter 4 provides a very brief introduction to the two case studies. Key to the implementation of the case studies is being able to utilize finer - grained data that are consis tent within the case, but may be idiosyncratic across cases and thus missed in the statistical analyses. Using this more micro - oriented approach, the case studies also assess the fit and usefulness of the three theoretical frameworks. Chapter 5 and Chapter 6 contain the case studies of Nigeria and Burma, respectively. In each case study I track economic and sociopolitical factors beginning in the colonial era. I assess if and how these factors evolved over time and what impact they had on leaders' decision making. 16 I subsequently follow the consequences of these decisions and identify how the states under consideration transitioned from the risk of one or both events to the actual occurrence of one and the eventual occurrence of the other. In both Nigeria and Burma, I find that both the constraints and tradeoffs frameworks are useful in understanding the connections between events. The decisions leaders made certainly produced tradeoffs in the relative risks of each event, but these scenarios were unlikely to have unfolded as they did without the overarching environment of constraints leaders were situated within. Chapter 7 provides a conclusion to the study. I briefly review the main theoretical propositions and the most notable findings from the qualitative a nd quantitative analyses. I then consider the limitations of the present study as well as possible future avenues of research to further parse the relationship between forms of political violence. I end with a discussion of some of the practical and academ ic implications of the theory and findings of this study. 17 CHAPTER 2 MODELING THE RELATIONSHIP BETWEEN COUP ATTEMPTS AND CIVIL CONFLICT ONSETS Introduction In this chapter, I develop a model of the relationship between coup attempts and civil conflict o nsets based on the selectorate theory developed by Bueno de Mesquita et al. (2003). The institutional environment in which they are situated. This institut ional environment helps shape the incentives facing leaders in their quest to retain power. The selectorate theory offers the foundation for a common framework to understand the coup and civil war literatures in a more comprehensive and holistic manner tha n is currently possible. I describe the basic logic of this selectorate framework, and make three additional contributions to the framework. The first is to discuss additional actors that are unique to considerations of civil wars and coups, but common acr oss the three models. I then explore the preferences of those actors, which are in line with the selectorate theory but again, more specific to the particular phenomena being examined in this dissertation. The final contribution is an exploration of the in teractions of those actors based on their unique preferences in each of the three conceptual models. In the course of discussing those interactions, I derive expectations about the observable implications of each of the three models, which will be tested q uantitatively in Chapter 3 and qualitatively in Chapters 5 and 6. 18 The Common Framework: The Selectorate Theory Actors groups, as well as a national leadership 9 and prospective substitutable leaders 10 identify residents as the largest of the three groups. Residents, in turn, are divided into two groups: the selectorate and the winning coalition 11 for mal role in expressing a preference over the selection of the leadership that rules them, though coalition is nested within the selectorate, and is the group w ithout whose support the incumbent is unable to retain power (38). The key link between the selectorate and the winning coalition is that an important support coalition relative to the selectorate, which in turn drives the relationship between the winning coalition and the leader, the policy outputs of the leader and the kinds of threats to her tenure the incumbent is most likely to face. Preferences The preferences of each actor are relatively straightforward, and shape the interactions 9 Although the leadership is assumed in the selectorate theory and here to be a single, unitary incumbent, in reality nt (executive, legislature, judiciary, bureaucracy, etc.) carrying out a vast array of actions. 10 These challengers, like the incumbent, are assumed to be unitary actors. 11 Bueno de Mesquita et al. (2003) also discuss the disenfranchised, those excluded f rom both the selectorate and the winning coalition, at length, but do not identify them as a distinct subset of residents, despite their potential Mesquita et al. 2003, 39). 19 In her efforts to do so, the incumbent will set tax rates and spending levels that maximize the welfare of those inside her winning coalition. With beneficial tax rates and spending levels in place, members of the winning coalition are more likely to remai n loyal, providing the incumbent an opportunity to siphon off some resources for her own, discretionary use. Challengers, in turn, seek to gain office, and will try to encourage residents particularly members of the current to support him in his quest to depose the incumbent. In order to will make promises regarding tax and spending rates and levels. All residents, from the dise nfranchised to those in the selectorate and those in the winning coalition, also seek to maximize their welfare. Based on the tax rate set by the leader, residents determine how much energy they devote to labor and how much energy they devote to leisure. B ased on their specific group membership (i.e., whether or not they are in the winning coalition), residents will be provided with a particular assortment of public and private goods 12 . Those not in the winning coalition will receive only the public goods pr oduced by the government. Those who have membership in the winning coalition, in addition to those public goods that benefit all members of society, are privileged with the provision of private goods that benefit only those inside the winning coalition. In Interactions Bueno de Mesquita, et al. (2003) establish the selectorate theory as an infinitely repeated game in which incumbents and challengers compete for the support of residents. Under 12 A more detailed discussion of public and private goods, their provision by the incumbent and the relationship between them is included below. 20 coalition while also holding back some resources for herself (Bueno de Mesquita et al. 2003, 87). unseat the incumbent. At the beginning of the selectorate game, the incumbent chooses a winning coalition from the selectorate 13 and offers them a mixed portfolio of public and private goods. For reaso ns discussed in greater detail below, if the winning coalition is small, the leader will provide a portfolio weighted more heavily toward private goods (and if the selectorate is very small, so too will the portfolio be weighted very heavily toward private goods), while the leader will emphasize public goods in her policy portfolio when the winning coalition is larger. Because members of the winning coalition are drawn from the selectorate, members of the selectorate have some chance of being in a future winning coalition. The size of this chance depends on the size of the winning coalition relative to the size of the selectorate. When the selectorate and winning coalition are both large, the chances a member of the selectorate will make it into the winnin g coalition is fairly good; we can think here of elections in a presidential democracy. All voters are members of the selectorate, and those who vote for the victorious repeats. Assuming for a moment that the incumbent wins re - election, although there is likely to be 13 Although in order for the incumbent to gain office in the first place she must be chosen by a sufficient number of members of the selectorate in order to form a winning coalition that contains at least some members of the old is vital to her ability to retain office and provide enough goods to maintain the loyalty of only the minimally - necessary number of coalition members. That is, in order to gain office, a winning coalition chooses the leader; after the leader is in office, she chooses who will remain in her winning coalition (of course, if she dips below a minimum threshold of support, she will lose office). 21 considerable overlap between the two, her winning coalitions across the two elections are unlikely to be identical some previous supporters will have becom e disillusioned and voted for the challenger and she will have gained some new supporters who had not previously voted for her allowing more members of the selectorate with opportunities to be in the winning coalition. If she loses, of course, the two wi nning coalitions the second are likely to have significantly less overlap, providing an even wider variety of individuals in the selectorate with the chance to be in the winning coalition. A simi lar dynamic exists when the winning coalition is small but the selectorate is also small there are fewer positions in the winning coalition, but not such a large pool of potential members, either. A key dynamic unfolds when the selectorate is large but the winning coalition is small: those in the selectorate have a very slim chance of making it into any future winning coalition. This serves to make those in the winning coalition quite loyal, since they receive private goods only because they belong to t he winning coalition (Bueno de Mesquita et al. 2003, 59). However, the high risk of exclusion from future coalitions in such systems means that members of the winning coalition are more loyal to the leader than the leader is to the winning coalition; the r educed risk of defection means that leaders do not have to spend quite as much on such coalitions as they would on larger coalitions or on small coalitions pulled from a small selectorate. Also central to the selectorate theory are the expectations about what mix of public and private goods the incumbent produces; these expectations derive from the size of the winning coalition. In order to keep her winning coalition and herself in power the incumbent provides the members of her coalition with policy outputs, or goods. These goods can be public or private; the latter solely benefit the winning coalition, while the benefits associated with the former spill over to all members of the selectorate and even the disenfranchised. Private goods benefit specifi c 22 other hand, are much broader in their effect and include things coalition is small, the leade r will emphasize the provision of private goods over public goods, ensuring that the winning coalition is amply rewarded for their support while relatively neglecting those outside the winning coalition. When the winning coalition is larger, it becomes inc reasingly and prohibitively expensive to provide a steady supply of private goods to the winning coalition, and the incumbent will shift to a policy portfolio that emphasizes the provision of public goods. These public goods benefit those in the winning co alition in a more cost effective manner, with the side effect that everyone else in society benefits, too. Although in equilibrium the leader always stays in power, on occasion the challenger is able to muster enough support to unseat the incumbent. Beca use those currently excluded from the winning coalitions, the challenger should have little difficulty in wooing them over to his side with the promise of private goods, even if for a single time period. However, the support of those against the incumbent. Although he is unable to credibly commit to including the supporte rs who helped him defeat the incumbent in his post - transition winning coalition 14 , those in the 14 Like the incumbent, the challenger is going to build the smallest possible winning coalition to keep him in power. In order to unseat the leader, though, the challenger is going to accept and encourage help from all quarters of society. In order to convert this oversized transition coalition to a minimally - sized post - transition coalition, some members of the transitio n coalition will have to be jettisoned. Those offering their support to the incumbent know and understand this process, so the challenger cannot realistically promise to retain the entire transition coalition in his post - transition coalition. 23 period in office as part of his transition winning coalition and the possibility of continued provision of private goods in the future are sufficient that supporting the challenger will increase their overall welfare more than will their continued support of the incumbent. If the incumbent is defeated and the challenger takes power, he will pursue the same policy strategies as the incumbent: reward the winning coalition as cheaply as possible while siphoning off as many resources as possible for his own di scretionary use. With the foundation laid by the selectorate theory in place, it becomes possible to explore the dynamics between actors involved in coup and conflict processes more specifically. This in turn allows us to explore whether any connections exist between the two processes as well as a fuller understanding of the nature of those connections. Coup Attempt and Civil Conflict Processes Actors Building on the framework of the selectorate theory, I organize the politically - relevant actors in a society into four sets: the leader, the population (which includes the disenfranchised, the selectorate and part of the winning coalition), the military (which is an integral component of any 15 and a rebel group 16 (which may be included in the selectorate, but is excluded from the winning coalition). As the two actors common to and active in all societies, the population and the incumbent are often considered the primary political actors in a 15 While ther e are a few states, like Iceland, that do not have a military, the vast majority of countries do have some 16 In practice, there may exist a number of rebel groups. For ease of theory development, however, assuming the existence of just one rebel group is sufficient. 24 society, with politics unfolding a s the interaction between these two groups in isolation of inputs of or reactions by other, potentially smaller, societal groups. Although this conception is erroneous, most political interactions do take place within the broad population - incumbent princip al - agent relationship 17 . Because of the focus on politics at this level, examining how the military and rebel groups alter political interactions within the state can be especially useful in explaining the (in)stability of the state. The military is a key component of a leader's winning coalition. 18 Because it and more importantly, the incumbent knows it is a kingmaker, 19 the military is the most likely to receive private goods, but also the most likely to press for more. While ethnic group members have an incentive to remain loyal (Amegashie 2008), knowing they will likely be excluded from the winning coalition if an individual from a different group takes power, the military does not have this incentive. Because the military is such an important part of the winning coalition, and one that cannot be removed without a dire threat to the leader's ability to stay in power, it has no real reason not to request more funding, autonomy and perquisites. As a permanent and guaranteed member of the winning coalitio n, the military is not bound by the loyalty norm in the same way other members of the winning coalition are. 20 17 Although there is variation in the extent to which this relationship exists, in general, the population (the principal), in the interest of efficiency and prac ticality, delegates the responsibility of making and enforcing policy to the incumbent (the agent). The level of delegation is quite high, and the agent operates with a large degree of autonomy from the principal. Thus, the principal has only a very limite d role in the daily operations of the government, but also in having any say over who the agent is. Indeed, only in elections, which are relatively infrequent even in democracies, and perhaps nonexistent in nondemocracies, is the population given any oppor tunity to choose the incumbent and reward (re - elect) or punish (decline to re - elect) him in subsequent elections. 18 As with this other primary actors discussed above, the military is assumed to be a unitary actor with a single preference. 19 A number of mi litaries have been described as historically playing this role, including the Egyptian (Gotowicki 20 Here the differentiation between the military as a unitary actor and the military as a collection of actors becomes pertinent. Individuals within the military can be replaced just like any other individual in the winning coalition. Idi Amin, for instance, was known for replacing particular members of the military brass he viewed as thr eats to his power (Roessler 2011). However, for the reasons articulated above, the military as an organization cannot be excluded from the winning coalition in the way individual members of the military or the public as a whole can be. The importance 25 The military is always included in a leader's winning coalition because the military has the power and choice to protect or depose the leader; wh ile a leader might be able to remain in power in the face of military indifference or ambivalence, it cannot survive if the military has an active interest in its removal. The Turkish military provides an example of the various kinds of relationships betwe en incumbents and the military. The military has tended to view itself as the protector of secularism within Turkey, carrying on the political order put in place by Atatürk as the Ottoman Empire collapsed. Over time, Turkey has fluctuated in terms of the d egree of Islamism supported and proposed by incumbents. While it has remained neutral in some cases of Islamist - leaning politicians, it has, in the past, overthrown leaders whom it views as a threat to that secular order, most recently in 1997. Not only do es the military choose to be viewed as the guardian of secularism within Turkey, it has the power to enforce that choice (Heper 2005; Narli 2000). A rebel group may be in the early stages of the organization process and not yet fully mobilized, or it may be actively engaged in anti - regime activities. The rebel group is comprised of dissatisfied members of the public. These individuals can be defectors from the winning coalition or people who were excluded from the outset. The rebel group challenges the aut hority of the leader, perhaps going so far as to initiate a civil war. These individuals have nothing to lose: they are not receiving any private benefits from the leader they could be, and cannot, in broad strokes, be excluded from using the public goods that are being provided. The portion of the public excluded from the winning coalition is nonetheless aware of the leadership's spending decisions. More precisely, they are aware of what, if any, public goods they of me addition to having the ability to stage coups, members of the military may also be able to credibly threaten the incumbent with rebellion against the state, as in the case of M23 in the Democratic Republic of Congo. 26 are receiving. 21 That domestic factors hav e been overlooked in the literature obscures the fact that economic performance, a domestic - level factor, has a direct bearing on a leader's ability to provide goods. Because they do not expect to receive sufficient private goods from the incumbent in the future, those outside the winning coalition will look elsewhere for opportunities to improve their economic well - being. Therefore, the amount of public goods provided by the incumbent is compared to the level of private and public goods promised, or perhap s directly provided, by rebel groups. This increase in expected welfare between what is currently being provided by the incumbent and what is being offered by rebel groups may be an important component of decisions to rebel. The portion of the public that is included in the winning coalition is also acutely aware of leaders' spending decisions and the overall state of the economy. Producing additional public s may not make up for the lost private spoils. Additionally, a poor economy means less to go around, again decreasing the level of private goods the winning coalition may receive. Those outside the winning coalition may try to recruit members from the incu coalition. Preferences The exact dynamics of the interplay between the coup and civil war actors identified above and their preferen ces depend on whether the actors are in a world characterized more by trade - offs or characterized more by constraints. As will be discussed in much more detail below, in a world of 21 While Colaresi (2012), for instance, has noted that some public goods, like security, are very difficult for the public to measure in real time, this remains a useful simplifying assumpt ion for the current study. 27 trade - loser to the ideal point of one societal group poses the very real possibility of moving sufficiently far away from the other the other) to violently lash out against the incumbent. On the other hand, in a world more marked by constraints, the hard times a state is facing leads to a lack of resources for the incumbent such that they are unable to offer bundle of policies that satisfy the demands of the mili tary, rebel groups and the broader population, thus endangering her hold on power from all segments of society. Following Bueno de Mesquita et al. (2003), I assume that each of the three societal groups -- military, public and rebel group -- have perfect information and know their own ideal point for military spending. However, all actors including leaders are uncertain about where the ideal points of the other societal groups lie. In general, the military would like as much funding as it can get. 22 How ever, it recognizes that at some point the next spending increment will not go as far as the last one, and the funding will be subject to diminishing returns. This suggests that the military's preference structure will be strictly increasing, such as in a positive logarithmic function it will increase quickly and then level off, but will not decrease. The only time the military is prone to take action against the state is when military spending fails to reach the military's minimally acceptable funding le vel. In general, less military spending is better for the rebel group: at the point of no military spending, the rebel group stands a very high chance of defeating the military. 23 As military spending increases from zero, the rebel group's utility initially decreases monotonically -- each 22 The military also has non - economic preferences, such as being valued and respected by the members of the society it serves. To this end, the military prefers to be used in ensuring domestic defense against external threats. Be ing used against domestic targets may take an eventual toll on members of the military as individuals as well as on the or ganization as a whole. In these cases, the military is no longer protecting society writ large, but has rather singled out a particular group or groups against which to take action. 23 ite lack of funding would alter the relative balance of power in a situation such as this. 28 additional spending increment on the military decreases the chances of a rebel victory. In situations when the public is extremely sympathetic to the goals of the rebel group, increases in military spending above some optim al point may actually promote the rebel group's utility marginally. This can occur because overspending on the military, at the expense of other publicly beneficial policies can increase inflation, reduce welfare spending, promote inequality, and decrease the underspending below some optimal point and overspending, above some maxima, on the military. Underspending results in an underprepared state that is less a ble to repel a rebel attack. Overspending, on the other hand, decreases public benefits such as economic growth (Cappelen et al. 1984), increases the public's dissatisfaction with the state and opens up additional recruitment opportunities for rebels. The public must decide what they expect to be an appropriate balance between military spending and the provision of domestic public goods like education and infrastructure. Too little military spending and the country is at risk of being a target of an inters tate attack by a predatory state or of an intrastate attack by the rebel group (assuming for the time being that the rebel group does not elicit public sympathies). Too much military spending, though, and public goods spending is eroded. This erosion may l ead members of the public to support or even join the rebel group as they seek a reallocation of state funds. With the public's ideal amount of military spending come both security -- but not at the cost of economic growth -- and other public goods that ac tively improve the well - being of the populace, such as increases in healthcare quality and availability. 29 Interactions The preferences of these actors shape the expected interactions between the groups in each of the three models. By understanding what interactions are expected across the three models, we can identify observable hypotheses that allow us to compare the usefulness of these theories. Trade - Offs In a political world marked by trade - offs, the fundamental relationships between actors within the selectorate theory is different than those identified in Bueno de Mesquita et al. (2003), where the only meaningful differentiation between the disenfranchised, the selectorate and the winning coalition is the chance any individual has of being in the future. This carries with it the further assumption that all members of the winning coalition are However, if, as I argued above, the military as an organization is assured membership in every winning coalition, then the military is not l and Ordeshook (1985) in noting that although public goods might be public in consequence, the production of su ch goods is still often private in nature. It makes sense that different types of groups would benefit differently from the production of different public goods. For instance, civilians in a winning coalition with no ties to anything remotely resembling th e defense industry would not benefit from the allocation of lucrative defense contracts in the same way civilians with such ties or military officers would. Likewise, those defense civilians or military officers would 30 not benefit equally from education ove rsight authority as other groups more tied to the provision of education. Given this, it is instructive to think of the winning coalition as being comprised of two distinct groups: the military and those members of the winning coalition who are not in the military. As will be discussed in more detail below, given this distinction within the winning coalition, it may become in the interest of the civilian members of the winning coalition to align themselves with the other civilians, not in the winning coali tion. In the pursuit of their own interests, these civilian members of the winning coalition may find that their interests are not wholly compatible with those of the military, and that the military, as armed kingmaker, is more likely to have its interests satisfied than is the rest of the winning coalition. In this sense, then, the trade - offs leaders face are at a macro - level, between blocs within their winning coalition that may have different preferences, rather than micro - level trade - offs, where particu lar individuals may be privileged or neglected, but the winning coalition can be thought of as a unitary actor with a single set of preferences. As a result of this relational configuration between actors, members of the winning coalition are more likely to be at odds with one another than if the members of the winning coalition had uniform interests, as in Bueno de Mesquita, et al. (2003). With these potentially - opposed preferences, the incumbent faces a delicate balancing act: in trying to maintain the l oyalty of one component of the winning coalition, the incumbent may sufficiently incentivize the defection of the other component that she faces overthrow from that segment. Politics in states facing considerable constraint is discussed in more detail in the next section; the trade - offs model operates in a political and economic environment with room for at 31 least some flexibility 24 . In these environments, where states have even slightly more capacity than the weakest states, incumbents will have the choice and ability to distribute a bit more widely than incumbents in states with the lowest levels of capacity. Enacting policies that benefit one societal group does not prohibit concessions to other groups in states with slightly more capacity. Politics in st ates with a modicum of capacity and flexibility is more than just a series of crises of varying severity strung together. Incumbents do not just have long - term policy preferences, they also have the resources and capacity to begin to enact those policies. Meaningful reform might be slow, in order to avoid sociopolitical or economic shockwaves that markedly allocation decisions may still be contentious, but the choic e of one policy will not completely remove the possibility of some steps toward another policy in the same way as in a state with very limited capacity. In terms of the situations leaders in these states are faced with and the policy choices available to likely and what sectors of society she perceives as the most dangerous and threatening. For instance, given that Svolik (2009) finds that popular revolutions only very rarely succeed in removing a leader from power, if confronted with a dual risk of military and popular overthrow, an incumbent should be more concerned with ensuring the military has little incentive to stage a coup than trying to halt a popular uprising t hat is unlikely to result in a loss of power (at least in the short - survived a coup attempt are particularly unlikely to be unseated by a coup in the future, leaders 24 Although politics (and especially redistributive politics) by definition involve trade - offs and the pursuit of some policy avenues at the cost of the pursuit of o thers, when constraints are considerable, the sheer limitations on the range of options available to incumbents have the potential to outweigh the impacts of budgetary trade - offs. 32 who h ave survived a coup attempt but fear another, as well as a civil war onset, should focus their efforts on preventing the civil war. Of course, there are factors, such as having been in office for only a short period of time, that may make leaders more vuln erable to misinterpreting the level of risk of each event, and thus of implementing counterproductive policies. From this perspective, the policies leaders enact are in response to the political climate as they perceive it, and are more broadly a reaction to the conditions they faced upon their arrival in office; the incumbent tries to continue building up factors that seemed favorable at the outset while trying to minimize those that had the greater potential to be detrimental to their tenure. Regardless of the level of ultimate control the incumbent has on the overall economic well - being and performance of her state, she can certainly make spending decisions based on the actual economic health of her state and adjust spending priorities according to her threat perceptions. threats facing that leader 25 , affecting how likely a leader is to face a coup attempt and/or a civil war onset. 26 Sound national defense i - military portions of society are much less likely to want an over - equipped military, especially when that funding comes at the expense of things like healthcare and education spending. Greater social spendin g can help address economic inequalities or other situations that, if left unaddressed, might promote resentment among the non - military portion of the winning coalition as well as among broader 25 face, leaders do face particular ex ante likelihoods of certain events. Their political choices will affect these likelihoods, potentially in ways the incumbent was not expecting, depending on the accuracy with which the incumbent had perceived the initial situation facing them. 26 For each of the following hypotheses, the impact discussed is an overall impact; even if a leader does ultimately face a particular event, she still might have decreased the likelihood of experiencing that even through her policy decisions. Policies can be broadly considered to be public or private; likewise, spending decisions can be broken down into two broad categories military and non - risk environment . 33 society. Greater healthcare or infrastructure spending, for in stance, will generate private goods for portions of the winning coalition excluded from defense - related private - goods spending, as well as make broader society healthier and better able to engage in economic activity and generally traverse the geographical territory of the state, among other things. With potential economic grievances being mitigated through increased social spending, there is less reason for society groups to rebel, thereby lessening the chances of a civil conflict onset. In both cases, th ough, the prioritizing of one type of spending and goods provision for a particular segment of society results in the other type of spending and goods provision for other segments of society folio. The neglected group or groups, in turn, have reason to seek a policy revision, and different avenues by which to seek those revisions. When the military is aggrieved, it should be a relatively straightforward process to rally enough troops to topple the neglectful leader, especially considering that coup attempts are typically carried out by very small groups of officers. When the civilian portion of the winning coalition is being neglected, though, their lack of weapons and institutional organizatio n means that they will not by themselves be able to unseat an incumbent who, after all, would, in this case, still have the support of the military. Rather, those malcontent civilian members of the winning coalition know that the masses are also disaffecte d by the emphasis on perks for the military, and will thus have an easier time recruiting members of the public to participate in action against the incumbent. These civilian members of the winning coalition can foment anti - regime activity, up to and inclu ding the onset of a civil war. This dynamic of placating one group (e.g., the military) by de - emphasizing support of the other (e.g., the civilian component of the winning coalition) leaders to the following hypotheses: 34 H 1 : Budgetary allocations favoring t he military will decrease the likelihood of a coup attempt, but increase the likelihood of a civil conflict onset. H 2 : Budgetary allocations favoring social spending will increase the likelihood of a coup attempt, but decrease the likelihood of a civil co nflict onset. While budgetary allocations are the most visible component of the political environment in a state, they are not the only component; leaders also have non - economic policy options at their disposal. Looking within their own regimes, incumben ts have alternatives they can pursue to interest to shrink the size of the winning coalition while increasing the size of the selectorate (Bueno de Mesquita et al. 2003, 377 - 378). Doing so markedly increases the loyalty of the winning while simultaneously enabling her to skim off more state resources for her o wn discretionary use, because even this reduction in private goods produces an amount larger than the challenger can only probabilistic. Anyone still in the wi nning coalition after it is shrunk obviously also benefits from this decision by the leader and in a world of trade - offs, the military is more likely to see a greater increase in welfare after this type of institutional reform since it is assured a spot in the winning coalition, while it is members of the civilian portion of the winning coalition that would be purged by the incumbent. Despite these incentives for leaders, however, the strength and durability of existing political institutions in a state m ay preclude a leader from being able to completely uproot the existing political system and replace it with a much narrower one Given the size of the population excluded from the winning coalition relative to the size of the winning coalition, even in stat es that 35 already have fairly large winning coalitions, the incumbent faces a much broader threat, and potentially multiple threats, from those outside the winning coalition than from those within the winning coalition, relying on those military and civilian members of the winning coalition to have enough buy - in to the system to help defend it against these societal threats. In such cases, an expansion of the winning coalition toward more inclusive institutions can placate society by making the political pr ocess more transparent and participatory. With broader swathes of the population now involved in the political process, the disenfranchised and the members of the selectorate previously not included in the winning coalition have less incentive to begin any kind of rebellion. Likewise, the civilian component of the winning coalition, through the shift toward a more public - goods - oriented policy portfolio, also increases its wellbeing relative to the military component of the winning coalition, which will see a decrease in the provision of some of its private goods. This shift from a small - coalition to a larger - coalition system necessarily limits and weakens the influence of the military, which can be problematic in states that have not yet cemented civilian - do minated civil - military relations and a norm of non - interference by the military. Leaders are unlikely to completely disempower the mainstays of their old winning coalition in one fell swoop, though, as that would encourage a preemptive strike against them, so at least for a while, until the transition to a civilian - centric system is complete, the military will still be in a somewhat privileged position, benefitting from some institutions that are still authoritarian - leaning even as they lose prestige and a degree of power under the institutions that are more democratic. Thus, H 3 : Mixed political institutions will decrease the likelihood of a civil conflict onset, but increase the likelihood of a coup attempt. 36 Regardless of political institutionalization, however, requiring the use of the military may have different effects among the different components of the winning coalition, depending on the type of missions to which the military is assigned. Although a leader cannot fully control whether her state has a rival, or whether her state is the target of some kind of interstate dispute, she does have control over how her state responds to that rival or dispute. Orienting the military to handle ale, as they are being tasked with the type of responsibilities for which they are best equipped and organized and often given considerable autonomy to wage those campaigns most effectively and efficiently. Because it is difficult to know exactly how m uch is required in the way of resources to secure victory but avoid superfluous spending, leaders will err on the side of superfluous spending in order to shore up a victory, which will have the added benefit of helping secure her hold on power by providin g that little extra infusion of private goods to the military, further increasing its loyalty. This increase in goods provision and authority of the military will make civilian segments of the population, including the segment in the winning coalition, w ary, however. The additional budgetary resources going to the military and the accompanying likelihood of overspending on the military (that is, spending more than is necessary to simply secure victory), as well as whatever additional legislation or polici es may be required to help the military conduct its campaigns most effectively necessarily means a reduction in the provision of budgetary allocations to policies more directly benefitting those outside the military, and laws facilitating the waging of war may bring with them domestic limitations, all serve to disaffect the general population within a state. The desire for more societal goods and policies that place limits on society while increasing the relative standing of the military will provide the ci vilian - component of the winning coalition with points of protest that can be used to recruit those excluded from the winning coalition altogether. 37 Moreover, the externally - oriented attention of the incumbent and the military will provide such civilians wit h an opportunity of relative surprise in an attack by an anti - government movement - policy - oriented portfolio that again narrows the privilege - gap between the military and non - mili tary portions of the winning coalition and increases the well - being of the segments of the population not included in the winning coalition. In these cases, H 4 : Engagement in an international dispute will decrease the likelihood of a coup attempt, but inc rease the likelihood of a civil conflict onset. Although the military is likely to benefit from participation in external crises, and in turn reward the incumbent with their continued loyalty, being used in domestic crises, especially when such use is fr military may benefit from repression in terms of increased allocation of resources and materiel, the individual members of the military must pay the psychological costs of vio lently suppressing their fellow citizens time and again. It is one thing to benefit at the expense of others in terms of policies that favor the military over society at large; it is quite another to benefit at the expense of others in terms of increased b udgets in exchange for physical harm against the population the military is ostensibly tasked with protecting (as in the case of external crises). These costs are particularly high when the ethnic and class composition of the military mirror that of broade r society, so that members of the military see their own reflections in the people they repress. In this case, although the military and non - military members of the winning coalition, and those excluded from the winning coalition, operate in different sphe res of the political and economic environment much of the time, the fundamental differences between the different sectors of society are reduced, 38 and as the military is able to more readily identify with the non - military components of society, the loyalty the military feels for the incumbent in the face of orders of repression is reduced. In the short - term, before these psychological costs are fully borne by individual members of the military, repression absolutely helps the leader develop a strengthened hold on power, which provides greater assurance to members of the winning coalition about the continuation of the provision of private goods into the indefinite future. It is in the long - run that the costs to individual members of the military begin to out weigh the organizational benefits, arising first in the form of lowered morale. Such consequences of long - term repression have been evident at various points in history when militaries have refused to engage in continued violence against the civilian popul ation, effectively abandoning the leader and removing the legitimacy of that incumbent to rule, as occurred in Romania in 1989 and Egypt in 2011. Likewise, while short - term repression might instigate a civil war by an already - agitated public, in the long - run the nonmilitary portion of the winning coalition and the broader population are going to see any potential for reform slipping away as their efforts are continually quashed by the military, acting on behalf of the government, and as their repressors ar e continually rewarded for their heavy - handedness. The opposition, like the military, becomes demoralized through the persistent use of repression, though with a different outcome than when the military becomes demoralized: H 5 : Sustained repression will in crease the risk of a coup attempt, but decrease the risk of a civil conflict onset. In the next section I will explore more fully how the selectorate theory plays out in a political and economic world marked by severe constraints, and derive observable i mplications from the dynamics of the theory in such a setting. 39 Constraints In a political world marked by constraints, the basic relationships between societal actors are as described by the original formulation of the selectorate theory, but within a m uch tenser environment. The fundamental differences between groups depend on their relationship to the winning coalition and what options the different groups disenfranchised, selectorate and winning coalition have for increasing their chances of being included in future winning coalitions. While the military, as in the world marked by trade - offs, is always a member of the winning coalition, its preferences are not in tension with those of other components of the winning coalition. The biggest modificat ion of the selectorate theory in a world of constraints is the role that those constraints greatly diminishing it which in turn has a bearing 2003, 99) opening it up to a wider variety of risks than it might ordinarily face. As briefly mentioned above and in contrast to politics in a world marked by trade - offs, the biggest differences that exist in society are between the different nested s ectors of society. The winning coalition can be thought of as a unitary actor with the single preference of maximizing the quantity of private goods it receives from the incumbent. Members of the selectorate and disenfranchised alike both want to belong to the winning coalition, but only members of the selectorate stand any chance of actually doing so in the future. Thus, the clashes of interest are between these groups as the winning coalition seeks to protect its supply of private goods and those excluded from the winning coalition seek to increase the emphasis the incumbent places on public - goods provision. A political world marked by constraints is heavily influenced by those state characteristics primarily, though not exclusively, economic and envir onmental that are exogenous to an 40 incumbent who has just taken power and that remain difficult for an incumbent to bring firmly under her sole control 27 . These factors and conditions, such as limited economic and political capacity, loom large on the poli tical horizon, and limit the degree of agency an incumbent has to enact any kind of significant policies that would generate both public and private goods. Whether access to the centers of political power, and thus no control or influence over the state of the country when she enters office. However, she has come to power on the tide of the promises she made as a challenger about resource allocation and policy provi sions, and the other societal groups will judge her leadership and apportion their loyalty on the basis of her ability to fulfill her promises. The performance of the economy and the degree of willingness of disaffected social groups to work with the gov a handle on negative conditions in the country and sustain, if not improve, any already - positive conditions. However, in an extremely vulnerable state, it is possible that the le ader will never have an opportunity to gain enough control over the political system and environment to be able mitigate immediate development crises enough to have any kind of longer - term impact or consequences 28 . When an incumbent takes over a state with limited capacity, she may be fairly powerless to affect, let alone improve, her chances of a longer tenure. In a world marked by constraints, the situations that arise must be dealt with in succession. There is no strategic balancing in the long - term bec ause immediate crises that threaten the leadership and can lead to coup attempts or civil conflict are occurring regularly. Given this, the 27 although over time there is the potential to change some of these underlying conditions, in the short - term it is nearly impossible to do so, and these enviro nmental features of the state can play large roles in the dynamics between the incumbent and the winning coalition and those excluded from the winning coalition, and thus the ultimate success or failure of a leader. 28 Although not all policies are equally costly in terms of either financial or political capital, they do require at least a degree of forward thinking that remains difficult to accomplish in the most vulnerable states. 41 winning coalition and those excluded from the winning coalition, both those in the selectorate and even the disenfr anchised, share commonalities and even have similar motivations in particular situations. Although the groups do share motivations, it is still important to note that economic suffering at the country level can hurt the winning coalition and the rest in di fferent ways, especially in autocracies, where those in the winning coalition are entrenched in a system of relative privilege and spoils, and those excluded from the winning coalition, and particularly the disenfranchised, are stuck in a system of neglect . Nonetheless, economic shocks, especially large 29 , and discontent over goods provision can lessen the difficulty any member of the population, whether in the winning co alition or not, faces in trying to mount a challenge to the incumbent should that member become sufficiently dissatisfied with the status quo. Vulnerability to both coup attempts and civil conflict onsets may be structural as opposed to reactions to stra tegic choices on the part of the leadership. An example of this extremely limited agency by incumbents is the length of time it takes failing states to recover to even the weakest level of non - failure 30 . Although Somalia is the most notorious failed state a nd an extreme outlier in terms of the length of time it has been a failed state, the climb out of the trap of state failure is a slow and painful slog for all polities (Chauvet and Collier 2008; Rotberg 2003). After years of 29 However, extreme resource constraints can lead to drastic policy choices . For instance, in the poorest countries, spending choices may come down to paying civil servant salaries or the military's top brass, as was the case in Nigeria in early 2012 (Campbell 2012). As circumstances become direr, even small fluctuations in econo mic performance -- positive or negative -- can have major repercussions, as funding certain programs becomes either feasible or all but impossible. Hyperinflation in Zimbabwe in the late 2000s left the government unable to raise the salaries of healthcare workers; the police force was also substantially underfunded, with the specter of rioting raised by the capital's police chief (Wines 2007); in late January 2013, Zimbabwe had just $217 left in the bank after paying civil servant wages (Smith 2013). Health care or education cuts may be necessary to rein in overburdened budgets, while military spending remains stagnant or also decreases. Although without mutinous consequences, dire economic circumstances left the Russian government unable to pay the military in 1998, and in fact in broader stretches throughout the 1990s (Holmes 1998). 30 30 At this level of recovery, relapses to failure are easy and a broader recovery is by no means assured. 42 suffering under (possibly sever al) regimes that have all proven incapable of raising the state up, and the decreasing patience of the population over time, it may be difficult for new leaders to keep loyalty of the public and military for a long enough period of time to begin to see the fruits of sound policies materialize in the form of a stronger, more stable and economically functional state: H 6 : States with low capacity are more likely to experience both coup attempts and civil conflict onsets. Given that the factors that limit and inhibit opportunities for the development of a strong state also increase the likelihood of both coup attempts and civil wars, weak leaders seem doomed from the outset, victims of the political and economic environment in which they come to power in the f irst place; with the odds stacked so strongly against incumbents in these regimes, deeply disruptive and violent instability carries with it an air of inevitability. Despite the ongoing an unresolved debate about whether ethnic diversity is by itself dete rminative of policy in any way, it is safe to say that in states with a large number of ethnic groups, there are, quite simply put, more groups that can potentially be excluded from any kind of power 31 , repressed or otherwise dissatisfied with the status qu o. The impact of this ethnic diversity can cut across societal groups, as entire groups may form part of the winning coalition or be disenfranchised, providing additional stratifications of society. 32 This ethnic dimension to politics thus affects all group s in society; the manner in which it influences those excluded from the winning coalition is fairly straightforward, though the winning coalition is not immune from the effects of ethnic diversity. 31 mmon in European states like Switzerland (Lijphart 1977, 1999), such societies typically have just a few larger ethnic groups. Additionally, such European states also have high levels of state capacity and the ability to absorb these different groups into broader society without destabilizing the economy or political system. 32 This assumption about the importance of ethnicity as a method of dividing society is an alternative to the view that the constituent segments of the winning coalition are different fr om one another in fundamental ways that impact the extent to which they really have mutual interests. 43 Because ethnicity is used by leaders as a heuristic 33 for loyalty or disloyalty the loyalty/disloyalty of particular ethnic groups resembles a self - fulfilling prophecy. Members of the particular ethnic groups included in the winning coalition are the current recipients of private goods, and can reasonably exp receiving private goods, into the indefinite future (Bueno de Mesquita et al. 2003, 60). This winni ng coalition, since that challenger would likewise use ethnicity as a heuristic and assume that ethnic groups loyal to the incumbent would be less loyal to him, and thus draw on other ethnic groups to help build his base of supporters. Without some kind of additional, negative, intervention by the incumbent, then, these members of the winning coalition have little incentive to defect to the challenger. Just as those already in the coalition can expect to remain included, so too can those not included in the winning coalition expect to be excluded (Bueno de Mesquita et al. 2003, 85 - 86). Cut off from whatever private goods are being provided by the incumbent and receiving only a diminished supply of public goods, these individuals have a stronger incentive to defect. Groups, especially ones that are already organized to an extent, will have an easier time persuading those excluded from the winning coalition that the chances of victory and with it, broad policy revisions are high enough that the risks involv ed in a rebellion are worth bearing. The credible possibility of improvement is important, as not even the disenfranchised will want to sign on to a hopeless cause that has no chance of being able to follow through on its promises. Nor is the winning co alition immune from the impact of ethnopolitics. While members of the military who belong to a particular ethnic group will certainly be members of the winning 33 The idea of ethnicity being used as a heuristic is not new. For more on ways in which ethnicity is used as a heuristic by scholars, see, for instance, Khan and Eriksen (1992). For more on ways members of societies use ethnicity as heuristic for evaluating public policies, see, for instance, Lieberman and McClendon (2013) and see Schnapper (2004) for the use of ethnicity as a heuristic for elites. 44 coalition, the incumbent has limited options in terms of her ability to truly exclude members of the armed forces from other ethnic groups. While they as individuals might not be particularly important to her political survival, to the extent that they comprise part of the organization of the military, they are still provided with some kind of access to private goods. While it is not as was the case in Burma and Nigeria, among other places the leader does not have cart blanche to alter the military strictly as she sees fit. Effe ctively barring members of some groups from the officer corps or simply not recruiting members of some ethnic groups at all has a direct bearing on the efficacy of the military in this way, those order to preserve the integrity of their organization. This suggests that H 7 : States with high levels of ethnic diversity are more lik ely to experience both coup attempts and civil conflict onsets. As argued above, ethnic diversity may contribute to the presence of rebel groups within a state. Not all rebel groups are ethnic in nature, though, and may form along ideological lines as we ll. Regardless of the identity foundation of a rebel group, though, the mere presence of a group that is already engaged in rebellious activities or is sufficiently organized to be able to do so without facing additional high barriers (such as those facing nascent groups seeking to initially formally organize) poses particular threats to states and leaders. With some degree of organization already in place, these organizations face much lower costs in trying to recruit members from those excluded by winning coalition by promising them at least some access to private goods and additional public goods, and lower costs in terms of mounting a full - blown revolt. 45 Again, in its role as the defender of the state against harm, the military is attuned to broader soc iety, from which attempts at harm may originate, and will take note if a leader proves incapable of dealing with groups threatening such harm. A rebellion or persistent credible threat of a rebellion signals the incompetence of the leader, giving the m ilitary pause. Members of the armed forces are unlikely to fully support an incompetent leader who may end up needlessly risking their lives in an otherwise - avoidable confrontation with a malcontent domestic opposition group. Rather than allow the situatio n to deteriorate to this point, the military will step in to restore social order and ensure its own interests and members are protected. Thus, H 8 : The presence of a credible rebel group in a state makes both coup attempts and civil conflict onsets more l ikely. While the reasons a specific state experiences a coup attempt at a specific time as opposed to experiencing a civil war onset may well be idiosyncratic, once that last veneer of state stability is ripped away by the first event, reaching a large en ough threshold of actors to carry out the other act (or an additional instance of the first) in the future will be less difficult, given that the first act already occurred, and occurred at a time when the state was stronger (even if only slightly) (Kuran 1989). 34 Put differently, although a loss of legitimacy may initially emerge and be isolated among members of one segment of society, a loss of legitimacy among any segment of the population indicates the vulnerability of the incumbent. This vulnerability w ill make any subsequent and groups unlikely to rebel as the sole voices of dissent to join an already - existent rebellion (Kuran 1989). This vulnerability also makes any subsequent opposition more threatening, since 34 Rega rdless of ultimate success, both coup attempts and insurgencies, if not full - blown civil wars, are extremely stave off any such future attemp ts. 46 the more protests a state faces, the more resources it must expend to defend against these protests. The more resources expended, the fewer resources the state has at its future disposal, and the mor e likely any given attempt to dislodge a leader is to succeed. As these weaknesses compound and previously excluded and powerless segments of society may also b e able to extract and extort concessions from the leader in exchange for loyalty (or at least not active opposition) that they previously would have been unable to, or perhaps to remove her from power altogether. Given the potential for low - level instabili ty in the form of protests or strikes to escalate, H 9 : States experiencing lower - level instability are more likely to also experience both coup attempts and civil conflict onsets. Leaders of states with limited capacity do not just face internal threats , although those may seem the most pressing. A number of states also face external threats of various kinds, whether in the form of a rivalry or a generally unfriendly neighborhood, and leaders must be cognizant of external conditions as well, which might require the few resources a state has in order to fend that threat off. Leaders face an especially difficult task in the face of external threats. On the one hand, and lives are at risk beca use of this threat and ensure that the military is ready should the threat escalate into a more imminent, active danger. On the other hand, even as the leader is trying to protect society, that external threat, especially if it takes the form of a rival o r rebel groups in neighboring states, might be using members of the winning coalition and excluded members of society alike to try to undermine the leader and gain a foothold of power and influence within the state. In the case of fledgling rebel groups wi thin a state, external powers could gain traction by providing the resources to engage in 47 conflict as well as perhaps training and an external rebel base. Via - a - vis the military, it may be sufficient for that external power to meddle at a higher level, con vincing the military that they (the military) are being undermined by the leader and that incumbent and the military have opposing interests. This latter method was undertaken by the United State in Chile, with a successful coup in 1973 overturning Allende - oriented economic structure. In addition to the states and groups behind external threats trying to woo domestic groups, if discontent is high enough within the country, these domestic group s might seek out external actors with whom to collude in order to gain some leverage over the incumbent. The greater the danger posed by domestic groups colluding with external threats, the higher the level of resources that will need to be expended in ord er to stave off the threat as well as win back the minimal loyalty of the rebellious factions in order to prevent future threats. Given the pressures that constraints place on the state as it tries to engage in constructive policy, and the interplay betwe en these constraints, leaders will have a more difficult time ensuring the readiness of the military. While in the face of an external threat, the military would be a logical place to channel increased resources, the leader must strike a delicate balance. On the one hand, more resources could help buy the loyalty of the military, or possibly make it much easier for the military to carry out a successful coup should they decide to break away from the leader. On the other hand, the already - tough times preclud e the ability to shuffle all that many resources; as members of the winning coalition, the military is already receiving a higher level of goods than many others in society, and there is very little slack in the remaining resources of the state to be able to divert. 48 External threats can increase the risk of civil conflict onset by providing domestic opposition groups with a convenient time to strike against the state, if not also provide those groups with an increased capability to directly engage the stat e in battle. External threats can encourage military disloyalty and defection away from the leader toward one of their own. Even if the military is inclined to remain loyal, however, the inability of the incumbent to provide additional resources to the mil itary reduces its sense of efficacy and increases frustration and feelings it is being tasked with the impossible: protecting the state and staving off external threats without sufficient resources to be able to do so effectively or successfully. Therefore , H 10 : States faced with external threats are more likely to experience both coup attempts and civil conflict onsets. The final layer of the constraints argument centers on the fact that typically, though not always, correlated with level of development and capacity is regime type. Although democracies certainly encounter tough economic times the Eurozone crisis being a case in point and other similar challenges, the fact that incumbents focus on the provision of public policy goods that are nonriva lrous and nonexcludable means that everyone in society feels the hurt in similar ways and to a similar extent. The structure of democratic systems, and their focus on public policy rather than private goods, provides them with more flexibility in terms o f policy options when confronted with economic, geographic and political hardships. And to the extent that the winning coalition or members of the public may feel particularly aggrieved, the system is structured such that the replacement of the leader is a relatively straightforward, institutionalized process that can occur at regular intervals. No drastic or dire steps are needed beyond the routine, institutionalized process. In small - coalition systems (i.e., autocracies), tough times mean even fewer publ ic goods for the general population, giving them an increased incentive to rebel, and fewer private goods 49 for the winning coalition, giving them an incentive to defect to a challenger, if to recruit among themselves, namely from the military, to replace th e leader directly. Autocracies tend to lack these strongly institutionalized political processes, so that there are rarely standardized, constitutionally - provided - for routines for the replacement of an incumbent. In these systems, leaders tend to be depose d on an ad hoc basis by challengers who rely on institutions only as much as they minimally require to secure their hold on power and weaken the institutions they do not require but might pose a threat, who are themselves eventually turned out of office on an ad hoc basis 35 . Based on the broad institutional strengths and successes of democracies and broad institutional weaknesses and failures of nondemocracies, I expect that H 11 : Nondemocracies are at greater risk for both coup attempts and civil conflict onset. Independence of Events The independence of events model suggests a much simpler set of interactions and empirical expectations as compared to the trade - offs and constraints perspectives. In this model, distinct sets of structural factors and poli cies lead to coup attempts and civil conflicts. Therefore, the risks underpinning one phenomena are exogenous to the risks underpinning the others. The coup and civil war literatures, explored in more detail in Chapter 1, identify factors expected to impac t the occurrence of each event. If coup attempts and civil conflict onsets really are independent of one another, the variables that are unique to explanations of coup attempts should provide no leverage in better understanding civil conflict onsets, and v ice versa. 35 While Hosni Mubarak came to power in Egypt constitutionally, he systematically engaged in the weakening of Egyptian political institutions so that they could not be utilized against him as he excluded greater portions of the population and built up his own hold on power. 50 Conclusion Influential research has assumed that civil conflicts and coup attempts as independent events. Only more recent research has begun to explore the link between coup attempts and civil conflict. While Roessler (2011) argues that civ il wars may be long - term alternatives to coups, building on this work, I argue that the reverse relationship may also be true. Leaders exist in a world of political, economic and geographic pressures that constrain the choices they make. When faced with s evere resource constraints, leaders are unlikely to have coalition and those excluded from the winning coalition suffer, and thus both groups have incentives to see k a change in leadership. As a state becomes somewhat more developed, and there are sufficient resources to deal with the most pressing existential issues facing a state, leaders may find themselves presented with a risk trade - off between the potential thr eats they face from within their own winning coalition and from those selectorate members excluded from the winning coalition and from the disenfranchised. In making one phenomenon less likely by devoting resources toward the mitigation of that risk, they may, as a result, be making the other phenomenon more likely due to that resource diversion. Leaders divert resources from the military and society at their own peril; if they fail to strike a suitable balance between the two or shift too many resources aw ay from one, they may fall victim to both a coup attempt and civil conflict. Tables 2.1 - 2.3 summarize each of the three frameworks 36 . The next chapter will provide a statistical test of the hypotheses outlined above and assess wh ether and what kind of specific trade - offs in their quest to keep their countries peaceful. Chapters 5 and 6 examine the cases of Nigeria from 1960 - 1967 and Burma from 1948 - 1962 in 36 All tables appear in an appendix at the end of the chapter. 51 more detail, respectively, to explore the microprocesses underpinning event s in those countries. These qualitative analyses provide an opportunity to carefully examine data that is not readily quantified or available for all countries in comparable forms, and thus serve as a complement to the quantitative analyses undertaken in C hapter 3. 52 APPENDIX 53 Table 2.1. Summary of Expectations of the Trade - Offs Framework General Expectation Specific Hypotheses Factors that make one phenomenon more likely simultaneously serve to make the other less likely. 1. Budgetary allocations favoring the military will decrease the likelihood of a coup attempt, but increase the likelihood of a civil conflict onset. 2. Budgetary allocations favoring social spending will increase the likelihood of a coup attempt, but decrease the lik elihood of a civil conflict onset. 3. Mixed political institutions will decrease the likelihood of a civil conflict onset, but increase the likelihood of a coup attempt. 4. Engagement in an international dispute will decrease the likelihood of a coup attemp t, but increase the likelihood of a civil conflict onset. 5. Sustained repression will increase the risk of a coup attempt, but decrease the risk of a civil conflict onset. 54 Table 2.2. Summary of Expectations of the Constraints Framework General Expectati on Specific Hypotheses Factors that make one phenomenon more likely simultaneously serve to make the other phenomenon more likely 6. States with low capacity are more likely to experience both coup attempts and civil conflict onsets. 7. States with high lev els of ethnic diversity are more likely to experience both coup attempts and civil conflict onsets. 8. The presence of a credible rebel group in a state makes both coup attempts and civil conflict onsets more likely. 9. States experiencing lower - level instab ility are more likely to also experience both coup attempts and civil conflict onsets. 10. States faced with external threats are more likely to experience both coup attempts and civil conflict onsets. 11. Nondemocracies are at greater risk for both coup attempt s and civil conflict onsets. 55 Table 2.3. Summary of Expectations of the Independence of Events Framework General Expectation Specific Hypotheses The unique variables used to explain one phenomenon will provide no empirical leverage in trying to explain the occurrence of the other phenomenon n/a 56 CHAPTER 3 QUANTITATIVE ASSESSMENTS OF THE RELATIONSHIP BETWEEN COUP ATTEMPTS AND CIVIL CONFLICT ONSETS Introduction The theoretical framework developed in Chapter 2 identified unique observa ble expectations from three distinct models of coup attempt and civil conflict onset risk that can help us determine which models is operable in the political world. This chapter contains a quantitative analysis of these expectations. Such an analysis is a dvantageous for several reasons. Statistical analyses allow for the assessment of large quantities of data with parsimonious models, and are able to concretely describe the impact of independent variables on the probabilistic occurrence of the dependent va riables (Braumoeller and Sartori 2004; King et al. 1994). This provides a more tangible conception of the relationships that exist among the concepts of interest. Such methodologies also allow researchers to be more systematic and uniform in their analyses than do qualitative research designs, which can be conducted according to a common framework, but which will necessarily vary in their contexts and the types and forms of information available. In quantitative analyses, if data are missing along any dimen sion in any case, that entire case is removed from the analysis, ensuring that all cases are assessed based on identical criteria. The quantitative analyses conducted in this chapter are implemented in several steps. In the next section, I will describe the primary statistical method utilized throughout the chapter, the bivariate probit model, including a comparison to the univariate probit model as well 57 as why the multiple equations are more appropriate in the present context. Following a discussion of t he methodology I employ, I identify and describe the variables used in my analysis. I first present the logic underlying the inclusion of the specific variable, drawing connections between the conceptualization and the testable expectations identified in C hapter 2, and then discuss the operationalization of that concept and the data source for the variable. The remainder of the chapter is dedicated to the empirical analyses themselves and a discussion of the results. The development and assessment of the empirical models to test the expectations of the theoretical framework proceeds in several parts. As an initial, prima facie test of whether the independence of events model is sufficient or if the constraints and trade - offs models are worth further explor ation, I analyze existing models of coup attempts (Powell 2012) and civil conflict onsets (Buhaug 2006), both as single - equation analyses and as part of a bivariate probit analysis, which explicitly tests the relationship between the two outcome equations. In the next step of the analysis, I explore additional variables that are suggested by the discussion in the previous chapter on trade - offs and constraints. As noted there, comparing and contrasting the coup and civil war literatures yields unique testa ble hypotheses. I test these to shed light on the benefits of bridging coup and civil war studies. Third, in addition to updating the list of measurements, I also update the data in order to extend the empirical analysis and move beyond simple replicati ons of previous models. The larger sample, which primarily includes more recent data than those included in the previous studies, will provide an illustration of the ongoing applicability of the theoretical framework as a way to understand some forms of po litical violence. The analyses in this third step, then, are the test of which model in the theoretical framework is most plausible given the observable trends across space and time in the occurrence of coup attempts and civil conflict onsets. 58 Fourth, I discuss the fit of the various models I test throughout the chapter to determine the explanatory gains made with my new modeling approach. AIC and BIC measures of model fit consistently demonstrate a better fit to the data of the models that explicitly con sider coup attempts and civil conflict onsets as part of the same underlying process. After conducting all the analyses, I engage in a general discussion of the results and the insights those results provide. I conclude with a brief recap of goals of the chapter and the broad outcomes of the quantitative analyses as well as an introduction to the logic underpinning the qualitative analyses undertaken in Chapter 5 and Chapter 6. Primary Methodology: Bivariate Probit The basic structure of the bivariate pr obit model is similar to that of the univariate, or binomial, probit model (Long 1997) 37 . I will thus briefly describe the mechanics of the univariate probit model before expanding out to a discussion of the characteristics unique to the bivariate probit, r elative to the univariate probit. The univariate probit model has a binary dependent variable, expressed as a function of a continuous and unobserved latent variable. The structure of this latent variable is as follows: (1) where y * is a vector of the underlying continuous variable for each observation, x is a matrix that holds the measurements of the independent variables, ß is the vector that hol ds the coefficients to 2 . Because the observed variable takes 37 to the number of independent variables within the equation. 59 on only two values, however, 0 and 1 , there must be some way to transform the latent variable into a dichotomous variable. This is done in such a way that the observed value of y i is 1 if the latent variables passes some threshold (indicated here as 0), and is observed as a 0 otherwise. That is: (2) In order to be able to identify the model, the threshold must be set at a specific value. By convention, that threshold is set at 0 with no loss of generality. In the bivariate probit model, there are two such dependent variables, each with the same structure of latent variable and the link between the latent variable and the observed value. It can be represented as: (3) Further, in a bivariate probit, the error terms (and the relationship between them) are represented as: (4) Because I am interested in the occurrence of both coup attempts and civil conflict onsets, utilization of the bivariate p robit is appropriate. Indeed, examining the two events in separate probit equations may produce biased coefficients, because they could be omitting relevant information on the correlation across outcomes. While a multinomial logit model may seem like a via ble alternative to the bivariate probit because of the existence of multiple categories of the dependent variable, the multinomial logit is nevertheless inappropriate in the present case. First, unlike the dependent variable categories in a multinomial log it, the two dependent variables in this study are not mutually exclusive: states may neither event, one phenomenon or the other, or both. 60 Second, the correlation across event types is of particular interest here, rather than simple classification into cate gories and placement in one category or another. The unique feature of the bivariate probit that makes it particularly useful in gaining insight to the theoretical model developed in Chapter 2 is the fact that the error terms of the two equations are cor that both processes share a common, but exogenous, component that is not otherwise observed in the model. In terms of testing the hypotheses derived in the pr evious chapter, if the error terms are indeed correlated, it is evidence that the unobserved processes of coup attempts and civil conflict onsets are related, and that the independence - of - unobserved compon ents of the two equations are positively correlated, and coup attempts and civil conflict onsets are being affected by that unobserved component in the same way, this would lend support to the constraints model. This could occur because unobserved, but imp ortant, structural constraints on leadership choices make both coup attempts and civil conflict onsets more likely. affected in opposite ways by the unobserved component, this would provide support for the trade - offs model. In contrast to the constraints model, this would be consistent with latent leadership strategies that improved one risk, but tended to worsen the other, systematically. Powell and Buhaug Rep lications samples of each modified so that they are ide ntical, thus making it possible to compare the results of the univariate probit employed for each with the results obtained in the bivariate probit analysis 61 that combines the two models together. In this section I discuss my rationale for using these two m odels as the foundation for my analysis, as well as discuss the research designs of each before conducting the univariate replications and the new bivariate analysis. While there is considerable research on both coup attempts and civil conflict onsets to choose from for replication analyses, the two pieces I have chosen, Powell (2012) for coup attempts and Buhaug (2006) for civil conflict onsets, offer distinct advantages. In the case of Powell, especially, the primary benefit is that this is a very recen t study examining coup attempts because of the sheer quantity and temporal relevance of the cases included in his sample, as well as the way he chooses to operatio nalize the biggest explanatory factors for coup attempts, namely background country conditions, including capacity, and variables directly impacting the military, such as military size and expenditures. Given that the military is the actor that stages coup attempts, understanding components of the military and their impact on the propensity of the military to attempt a coup is vital. what has become the standard s et of civil war covariates. 38 Given that studies of civil conflict tend parsimony the inclusion of variables that are generally accepted to be important in the explanation of civil conflict onset 39 without numerous extraneous variables less relevant to the current project. 38 These covariates were identified and employed in Fearon and Laitin (2003), and served to establish the norm of what variables should be included in future examinations of civil wars. 39 Ward, Greenhill and Bakke (2010) identify GDP and population as the only two variables consistently and robustly ass the original models I develop later in this chapter. 62 One challenge in examining these two pieces of scholarship is the need to maintain the same sample for the purposes of any kind of model com was states with a population over 500,000 from 1961 - 2000. Buhaug (2006) had a similar population constraint, but examined the years 1946 - 1999. In order to produce the same sample across both studies, the time frame under study in my replications is 1961 - 1999. Variables Employed in Powell (2012) and Buhaug (2006) Replications In order to conduct the replications of Powell (2012) and Buhaug (2006) as faithfully as possible, I utilize the variables used by each schol ar in his original analysis. Although there are conceptual similarities between the two models, such as regime type and GDP per capita, the operationalizations of those variables differ slightly and the replication data provided by each author is used for the analyses involving that model. So, for instance, although both studies use models. This maintains consistency within each model and faithfulness to the operationalization of the variables developed by each researcher. In the original analyses I conduct later in the chapter, I update the data used for some variables so that I can examine more recent data than either Powell (2012) or Buhaug (2006) were able to. Variables Employed by Powell (2012) Coup Attempt Powell examines both successful and unsuccessful coup attempts, and in operationalizing both variables he relies on the de finitions provided by Powell and Thyne (2011). An attempted coup is 63 an effort by elites within the governing apparatus, namely members of the military, to replace the incumbent via unconstitutional methods. Change in Military Expenditures This variable i s measured as the percent change in the level of military expenditures provided by the Correlates of War Project (Singer, Bremer and Stuckey 1972) from year t - 1 to year t . Soldier Quality In order to tap the concept of the quality of the military, Powell the totality of military expenditures divided by the number of military personnel, providing the amount of military expenditures per soldier. Military Personnel Once again relying on the Correlates of War Project, Powell provides the log of the number of military personnel in a state. Effective Number of Military Organizations In order to provide a gauge of the degree of coup - proofing that has occurred within a state, Powell employs the measure of the effective number of military organizations. This mea sure is discussed in greater detail below. 64 Change in GDP Per Capita (t - 1) Much like his measure of change in military expenditures, Powell provides the percent change in GDP from year t - 1 to year t , then lags this value one year. Powell uses real 1996 dollars, obtained from Gleditsch (2002). GDP Per Capita (t - 1) - year lag of GDP per capita for each country - year. Instability , including the following types of events: government purges, riots, strikes, protests and assassinations. Democracy and Autocracy These measures of regime type consolidate data from the Polity IV scale (Marshall, Gurr and Jaggers 2014) into dummy variab - 5 or lower in a particular year. Powell excludes anocracies as the comparison category. Military Regime This variable is a simple indicator of whether a military regime is in power in a given country - year, with information again taken from Banks (2001). 65 Peace Years and Peace Years Splines r of years since the previous coup attempt and cubic splines of that counter. Variables Employed by Buhaug (2006) Civil Conflict Onset Buhaug uses the UCDP/PRIO armed conflict data, and includes conflicts with a threshold of 25 battle deaths and greater. Democracy (t - 1) and Mixed Regime (t - 1) Like Powell, Buhaug uses the Polity IV dataset (Gurr, Jaggers and Moore 1989) to create indicator variables of regime type. Buhaug uses autocracies as the reference category, and codes democracies as states with a score of 6 or greater on the Polity scale and mixed regimes as states with Polity scores between - 5 and 5. He lags this variable, designed to tap state capacity, by one year. GDP Per Capita (t - 1) capabilities, and is also lagged one year. Oil Exporter States that receive at least one - third of their export revenue from fuel exports in a given country - year are indicated with a 1; country - years that do not meet this condition are coded with a 0, fol lowing Fearon and Laitin (2003). 66 County Land Area from the World Bank (2002). Ethnic Fractionalization The measure of ethnic fractionalization is taken from F earon and Laitin (2003) and ranges from 0 to 1, denoting the probability that two randomly selected people will be members of different ethnic groups. Conflict Onset Decay Function revious civil war or year of independence to capture country - Additional Observable Indicators for Testing the Constraints and Trade - Offs Models Dependent Variables Civil Conflict Onset I have used the civi l conflict data contained in the UCDP/PRIO Armed Conflict dataset (Themnér and Wallensteen 2014; UCDP/PRIO 2014) to code all conflict - related variables. Because the PRIO/Uppsala dataset codes two different levels of conflict, it allows a degree of nuance n ot available in other conventional datasets like the Correlates of War Intrastate War dataset, which only codes conflicts that have reached the 1000 battle - deaths threshold. In general, I have coded an onset of an episode of violence (and there may be seve ral distinct episodes of violence within 67 the same, larger conflict) with the highest level of violence achieved during that episode, even if the highest level of violence did not occur in the first year of that episode. If an episode of violence did have o ne or more years of 1000 battle deaths per year, the dynamics of that episode are different than if there were no years of violence that produced 1000 or more battle deaths. 40 I expect that episodes of violence that reach the level of a civil war (with the higher casualty threshold) more accurately measure the rebel threat discussed within the constraints and trade - offs models. If more than one distinct conflict was occurring within a state in the same year, I employ the highest intensity of conflict. I code conflicts using both the 25 battle - deaths threshold and the 1000 battle - deaths threshold. Coup Attempts successful and unsuccessful coup attempts. They define a (252); because my dissertation is interested in what encourages militaries to revolt against the incumbent, the attempts need not be successful to provide an adequate test of the theoretical framework identified in Chapter 2. This is also consistent with the coding of civil conflict, as a civil conflict does not have to successfully topple the government to be counted. 40 The distinction between 25+ and 1000+ battle - deaths thresholds in the literature is com mon, with scholars like Collier and Hoeffler (2004) and Fearon and Laitin (2003) arguing in favor of only 1000+ and other scholars, such as Buhaug (2006), Gleditsch et al. (2002) and de Soysa (2002) taking the position that insight can be gained from studi es of 25+ battle deaths. 68 Independent Variables Each of the independent variables can be tied to both the constraints model and the trade - offs model. I briefly describe the connection to each before discussing the measurement of each variable. Ongoing Civil Conflict Continuin g civil conflicts might make coup attempts and new conflict onsets more likely by demonstrating the ongoing inability of leaders to deal with internal challenges. Conversely, although rebel organizations may be inspired to begin new conflicts, the military is quite obviously needed and its ability to defend the leadership puts it in a good position to seek concessions without having to dislodge the leader in order to obtain them. In order to account for correlation across years within the same state, I ind icate whether there was an ongoing conflict in a given country - year. If a conflict episode was contained within conflict spans from one year into another, onse conflict that began in the previous year carried over into a new year. Because a state can experien ce more than one conflict at a time, there are cases where both the onset and ongoing indicator variables are coded 1 for a given country - year. In this case, onset is coded 1 because a new conflict episode began, and ongoing is coded 1 because a conflict e pisode that had begun previously has carried over into that year. 69 Previous Civil Conflict Involvement Previous involvement in a civil conflict signals past difficulties on the part of incumbents in dealing with domestic tensions and might make future c onflict more likely. It is also possible that this past reliance on the armed forces have increased the prestige of the military and made it a more powerful actor with greater access to resources, making a coup attempt less likely. This variable provides a counter for the number of previous civil conflict onsets a state has experienced in any given year. Because past civil conflict occurrence has a positive impact on future civil conflict occurrence, it is important to control for this participation. Cr edible Rebel Group The presence or absence of a group that has the potential to credibly threaten the authority and reach of the state is hypothesized to make both a civil conflict onset and a coup attempt more likely; civil conflict is more likely becaus e of the risk of domestic rebellion, while a coup attempt is more likely because appeasement of the rebel group by the incumbent may involve the diversion of resources from the military. It could also be the case, however, that reliance of the leadership o n the military in countering such a threat makes coup attempts less likely even though the risk of civil conflict onset remains high. In order to most accurately assess whether a state is faced with the threat of a potential rebel group, I use data from the Global Terrorism Database produced by the National Consortium for the Study of Terrorism and Responses to Terrorism (2013) to identify attempted terror attacks (like with the measurement of coup attempts, the success or failure is distinct from what le ads to the attempt in the first place) carried out by an organization that claims responsibility for an attack. 70 To that end, attacks carried out by individuals, unknown perpetrators or suspected perpetrators are not included. While the last exclusion may s eem somewhat controversial, a group seeking concessions from the state is likely to ensure the state knows its identity (and thus to whom to allocate any concessions). That is, any group that may openly challenge the state is going to be just that open. 41 International Threats and Disputes A tough international environment can make coups and civil conflicts more likely because the rebel leaders. They al so have the potential to increase the chances of a domestic opposition at is tasked with protecting the state against such threats and disputes. volvement with their external environments are measured in three ways. Two are based on whether a state was involved in an international dispute in a given year. Two different thresholds are used; one dichotomously measured as participation in a militarize d interstate dispute that reached the level of a display of force, the other dichotomously measured as participation in a MID that reached the level of a use of force or higher. Data are taken from the Correlates of War Militarized Interstate State Dispute dataset, version 4.1. 41 Groups who engage in terror attacks have incentives to make sure their identities are known both to the state and to the broader public. By making themselves known to the state, terror groups are able to ensure that any concessi ons that come about as a result of the terror attack are appropriately distributed to the responsible parties. Likewise, being known to the broader public can help with recruiting, especially when several groups are vying for support in the same general ge ographical space (Pedahzur and Perliger 2006). For instance, in the Philippines in 2013, the Moro National Liberation Front stepped up attacks on the state at the same time the government was negotiating with the Moro Islamic Liberation Front, an organizat ion with similar goals for increased autonomy for the island of Mindanao but rival interests (Jennings 2013). With any deal leaning in favor of the group engaging in negotiations, the MNLF had the incentive to utilize tactics that brought it back to the fo refront of the public eye, reminding the government that it still has the capacity to create difficulties for the state and populace alike. 71 The second measure is whether a state has a contiguous rival. As discussed by Salehyan (2007, 2008) and Salehyan and Gleditsch (2011), contiguous states can serve as extraterritorial bases for rebels, and rival states would have even more incentive to provide such shelter or perhaps outright meddling in the affairs of the other, such as through government financing of rebel movements. Contiguity data are taken from Stinnett et al. (2002), while rivalry data are obtained from Thompson and Dreyer (2011). Political Exclusion By decreasing the size of the winning coalition and possibly the selectorate, political exclusion makes coup attempts and civil conflict onsets more likely by increasing the chances that key individuals in society w ill be excluded on the basis of particular group memberships (e.g., ethnic affiliation). However, by favoring some groups and individuals over others, the leadership of a state curries favor with the included groups, helping ensure their loyalty, while fos tering resentment among and an increased chance of retaliation by those excluded from power. These data come from the Ethnic Power Relations data set (Wimmer, Cederman and Min 2009), and provide two measures of the population excluded from government. The first is the total number of excluded ethnic groups, while the second is a measure of the proportion of the population that belongs to those excluded ethnic groups. While the notoriously poor - performing ELF (see Collier 2001; Elbadawi and Sambanis 2002; S ambanis 2001) provides a simple measure of diversity, with no commentary on what diversity means , the Ethnic Power Relations is a direct indication of the ethnic dynamics in a state. Those ethnic groups excluded from power have no guarantee of a meaningful voice in the state, and with no routine voice, groups would have only more extreme methods available to them to make their demands known. The greater the portion of 72 the population excluded from power, the more stark are the differences between those inclu ded in the selectorate and those excluded from it and as discussed in Chapter 2, the greater are the incentives for the incumbent to completely ignore the disempowered and to instead focus on providing spoils to those already entrenched in power. Governm ent Spending Low levels of government spending are likely to leave all groups dissatisfied. However, higher levels of spending on public goods are more likely to satisfy the public while leaving the military dissatisfied, while elevated levels of military spending are likely to have the opposite effect. The first measure of government spending comes from the US Arms Control and Disarmament Agency/State Department World Military Expenditures and Arms Trade publication and reflects military expenditures. Mi litary expenditures are defined in the WMEAT publications defense agencies for military programs; expenditures for the military components of such mixed ac tivities as atomic energy, space, and research and development; military assistance to foreign countries; military stockpiling; retirement pensions of career personnel; and expenditures as y 1975, 10 - 11). 42 The WMEAT measures had the advantage over Correlates of War military expenditures data of being measured in constant US dollars, which facilitates comparison of expenditures across time and space. In the data under examination here, milita ry expenditures are reflected as a percentage of GDP so that it is more readily compared to the social welfare spending variable described below. 42 e included (US Arms Control and Disarmament Agency 1975). 73 Better - funded militaries have greater capacity to maintain their fighting fitness through the acquisition of m ore sophisticated weaponry, more recruits, and higher - quality officers. The second measure of government spending is total government expenditures on social welfare programs as a percentage of GDP, compiled by Taydas and Peksen (2012). This variable inclu des government expenditures on health, education and social security, obtained from Burgoon (2006), Kugler, Feng and Zak (2002) and the World Bank (2009) and has had its missing values filled in using multiple imputation (Taydas and Peksen 2012, 278). Mi litary Readiness A well - equipped and autonomous military may discourage both coup attempts, out of lack of necessity, as well as civil conflict onsets, by decreasing the chances any such rebel movement would succeed. Conversely, such a military would have a higher probability of success were it to attempt a coup, and be able to secure for itself more resources, though the chances of a rebel victory would remain low, again discouraging an insurrection at all. Military readiness is measured in three ways, al l designed to tap the extent to which the military is adequately maintained as a fighting force with the capability to carry out its defense mandate. Military spending obviously helps proxy the materiel readiness of the military, while these variables acco unt for the effective allocation of personnel. The first measure is from the Correlates of War Project, and is the number of military personnel in a state. These data are combined from the US Arms Control and Disarmament Agency and The Military Balance , p ublished by the International Institute for Strategic Studies, and are provided up to 2007. The second measure uses this data, as well as population data, to create a ratio of the number of military personnel to the population as a whole. States with a hig her 74 proportion might have more intense security requirements, but those militaries may also be at least somewhat bloated, and be larger than is strictly required for defense purposes, hinting instead at bribery by the incumbent to keep the military as an o rganization loyal. The third indicator is taken from Pilster and Böhmelt (2011, 2012), and is a measure of the effective number of military organizations. This takes into consideration the extent to which the armed forces in a state have been undermined b y the creation of paramilitary forces and the pulling of personnel away from the primary military into these other organizations. Higher numbers of - proofing efforts in the form of creating an artifici 340). Regime Type The regime type of a state has a direct bearing on the extent to which the public has a voice in the political process. In nondemocracies, the publi c has not meaningful voice in the policy process. In mixed regime types, there is some room for societal participation, which may decrease the risk of a civil conflict onset, while simultaneously increasing the risk of a coup by decreasing the prominence o f the military in the winning coalition of such regimes. I measure regime type based on the standard Polity Scale (Marshall, Gurr and Jaggers themselves contain indicators of political violence. Because I am interested in explaining forms of political violence, it would be inappropriate to retain these components of the original Polity measure in my measurement of regime type. While the original Polity scale is 2 1 points and ranges 75 from - 10 to +10, the measure of regime type with the problematic components removed ranges from - 6 to +7. 43 Capacity Capacity is expected to have a similar effect on both coup attempts and civil conflict onsets, such that lower levels of state capacity will increase the risk of both types of events, while increased capacity will decrease the risk of both types of events, as incumbents have more resources at their disposal with which to implement policy. Three variables are used to capt ure the idea of state capacity, which has a direct bearing on a state has overall, the less it will be able to use fiscal policy to try to ensure the loyalty of any groups within broader society. GDP data are taken from Gledtisch (2012), and lagged one year, change over time, rather than instantaneously. That is, state weakness is more of a culmination of hard times than the result of a one - time shock; the time lag helps account for this gap between income and performance. The second and third variables are both known correlates of civil conflict onset, and are included for control purposes. The population of a state is measured in thousands, and data are drawn from Gleditsch (2012). Additionally, higher proportions of mountainous terrain in a state 43 Another measure of regime type, the Scalar Index of Polities (SIP), developed by Gates et al. (2006) combines non - oter turnout. As Vreeland (2008) notes, however, although these measures have the advantage of being free of contamination from measures of political violence, neither is found to be significantly correlated with civil war (though Gleditsch et al. [2008] f ind more et al. (2006). 76 also aff ects conflict onset (Fearon and Laitin 2003) and duration (De Rouen and Sobek 2004), as it is more difficult for any state to project its authority over such topography. As the US experience with al - Qaeda in Afghanistan just after September 11 or Pakistan much more recently, mountainous terrain provides excellent cover for groups seeking to avoid the reach of the state (and in former case, two states, as neither Afghanistan nor the United States was able to dislodge the al - Qaeda operatives from the Tora Bor a region) (Corera 2011; Schorzman and Nazish 2014). Instability Acts of rebellion such as protests and strikes have the ability to increase the risk of both coup attempts and civil conflict onsets by making the regime more vulnerable in general. Instabi lity data are taken from Banks and Wilson (2014), and measure anti - government instability indicators number of strikes, number of riots and number of anti - government demonstrations because they represent anti - government actions that are not necessarily direct predecessors to other variables already captured in my dependent variables, such as guerrilla warfare and revolutions. More specifically, Banks and W any strike of 1,000 or more industrial or service workers that involves more than one employer and that is aimed 100 citizens involvi 100 people for the primary purpose of displaying or voicing their opposition to government policies or authority, excluding demonstrations of a distinctly anti - 77 Repression Unlike instability, which may increase the risk of both coup attempts and civil conflict onsets, repression is expected to increase the possibility of a civil conflict onset by further alienating the targeted societal groups, but decrease the risk of a coup attempt by raising the prestige of the military, and making more resources and funding available to the military as a result. Terror Scale data is comprised of two scales from Amnesty International and the United States Department of State -- measured from 1 - 5, with higher values indicating more wonton disregard Sample of Cases in Original Quantitative Analyses Due to the constraints of data availability for some of the key independent variables, the sample under consideration here spans from 1975 to 2005 and includes states with populations larger than 500,000. Modified and Extended Replication Results In the first step of the analysis, I ran modified replications of Powell (2012) and Buhaug (2006). The replications are modified in both cases because of the constrained sample size, and in the case of Buhaug because I employ a probit analysis rather than a logi t. These modified replications produce some results that differ from those presented in the original articles. I discuss those differences below. 78 Table 3.1 presents the results of the modified replication of Powell (2012) 44 . Country - years military expenditure per soldier, the log of the numbe r of military personnel and the number of years since the most recent coup attempt were statistically significantly and negatively associated with coup attempts. Instability and the presence of a military regime in a state were both significantly and posit ively associated with coup attempts. In my modified replication, the log of military personnel, instability and the presence of a military regime retained their signs and ases under examination here, and the concomitant reduction in the number of observed coup attempts could account for the altered results; nonetheless, the results of the modified replication still suggest that manipulation of the military has an impact on the propensity of the armed forces to attempt a coup. Table 3.2 displays the results of Buhaug (2006) modified replication. As with the Powell replication, constraining the sample to allow for comparisons alters the overall composi tion of the - years have a conflict onset. In the constrained sample used here, about 4 percent of country - years do. In this case too, the results from the modified repli cation differ somewhat from the results from the original analysis. In his article, Buhaug finds that mixed anocracy, being an oil exporter, country size, ethnic fractionalization and a recent conflict onset all increase the likelihood of a conflict onset, while higher levels of GDP per capita make conflict onsets less likely. In the current analysis, country size and recent conflict both make conflict onset more likely, while higher GDPs sults, the results here are 44 All tables and figures appear in an appendix at the end of the chapter. 79 consistent with other civil conflict research; indeed, Sambanis (2001) finds that GDP per capita is one of only two variables that has a consistent impact on conflict onsets across studies. 45 Table 3.3 pre sents the results of a bivariate probit analysis that combines the separate Powell (2012) and Buhaug (2006) equations into a single model. This bivariate probit is treated as a seemingly unrelated regression, with each equation retaining its unique variabl es, rather than being combined in such a way that each equation included all the variables under consideration. That is, even though they are now included as a bivariate analysis, the Powell and Buhaug equations are distinct in their identification. For t he most part, the results within the bivariate probit framework are the same as in the single - equation frameworks. In the Powell equation, the log of the number of military personnel and a higher GDP per capita decreases the likelihood of a coup attempt, w hile instability and military regimes increase the likelihood. In the Buhaug equation, higher GDP per capita makes civil conflict onsets less likely while larger countries and states that have recently experienced a civil conflict are more likely to experi ence a conflict onset. unobserved factors that affect both equ ations, and that they should be modeled together, rather here, the error terms of the two equations are positively correlated, which means that those unobserved factors are affecting each equation in the same way. This lends support to the constraints model identified and discussed in Chapter 2 the unobserved factors that make one phenomena more likely also 45 The other variable identified by Sambanis as a consistently robust predictor was population size, a variable which Buhaug did not include. 80 make the other phenomena more likely. For example, an unexpected ec onomic crisis would be modeled as making both coup attempts and civil conflict onsets more likely. Figures 3.1 and 3.2 illustrate the coefficients and 95% confidence intervals produced by the separate models and the Powell /Buhaug combination bivariate probit. Figure 3.1 shows how the results of the probit estimating coup attempts compares to the bivariate probit, and Figure 3.2 displays the results of the civil conflict model compared to the bivariate probit model. Although the results of each model compared to the bivariate probit seem similar to one another, it is apparent that there are differences in the estimates produced by each model, indicating that considering the two equations together in the same model does produc e tangibly different results than the two probits run separately. This also helps build the case that pursuing an understanding of coup attempts and civil conflict onsets through the use of a bivariate probit model is a useful enterprise, and does shed a d ifferent light on the occurrence of these events. The results of the Powell and Buhaug replications with a more - constrained sample that will facilitate comparison with the original model I develop are presented in Tables 3 .4 and 3.5. The sample is decreased by approximately 800 cases, but the results remain largely consistent with those found in Tables 3.1 and 3.2. The major difference is that in the Powell model, GDP per capita is no longer statistically significant in the more constrained sample while the peace - year counter gains statistical significance. The Buhaug results are consistent across the two sample sizes. Likewise, Table 3.6 presents the results of the bivariate probit with the more rest ricted sample size to facilitate comparison with the original bivariate probit model later in the chapter. In the Powell equation, the number of years since the most recent coup attains statistical significance (which it had lost in the bivariate combinati on with a larger sample). Instability and military regimes retain their significance and signs, while the number of military personnel loses its 81 statistical significance. In the Buhaug equation, GDP per capita remains negatively associated with civil confl ict onsets, larger countries are still at greater risk for conflict onset and states that have recently experienced an onset are more likely to experience another. The largely similar results across the more - and less - restricted sample models contributes t o the validity of examining the relationship between coup attempts and civil conflict onsets as part of a bivariate probit as opposed to separately - estimated probit models. Figures 3. 3 and 3. 4 further corroborate the validity of this exercise. Much like in Figures 3.1 and 3.2, the coefficients and confidence intervals from the bivariate probit clearly differ from those from the single - model probits. That is, doing nothing more than looking at the two models as part of the same underlying process, rather than as two distinct processes, changes the statistical results. Extensions to the Powell and Buhaug Models In models developed and tested in the following section, I operate from the premise that if coup attempts and civil co nflict onsets are indeed not independent of one another, as the previous analyses have demonstrated, it may be fruitful to develop a model that explicitly considers the two phenomena together, rather than still treating them as independent in seemingly unr elated bivariate probit analyses, as was the case in the models presented in Table 3.3 and Table 3.6. In those models, each of the two equations had some variables unique to it. To continue testing the utility of the bivariate probit enterprise, I develop new empirical models. All variables are included in both equations of the bivariate model in order to test the hypotheses identified in Chapter 2 and provide an indication of whether the constraints model (i - 82 Throughout the analysis, I consider not only the bivariate probit equations, but also test each equation in the bivariate models as their own separate probit models . This allows me to check that given the new covariates, the bivariate probit is still the more appropriate model choice. Because of the previous strong pattern in support of the bivariate probit methodology and the continuation of that pattern in the mode ls that follow, I will briefly summarize the results of the separate probit analyses, but focus the majority of my attention on the bivariate probit models. The sample sizes for the models presented in Tables 3.7 - 3.9 match the sample size for the previous several equations to provide a basis for comparison, examined in more detail in the following section. Table 3.7 displays the results of the coup attempt probit with additional covariates. As past research has suggested (see, for e xample, Belkin and Schofer 2003, 2005; Jackman 1978; McGowan 2003), the greater the number of previous coup attempts a state has experienced, the more vulnerable it is to coups in the present. Additional, ongoing civil conflict also makes a coup attempt mo enough control over society to prevent rebel movements signals incompetence to the military. The higher the number of effective military organizations, the lower the likelihood of a coup, perhaps because the military is too divided to muster enough coherent support within itself to pull off a coup attempt (Belkin and Shofer 2003, 2005). More consolidated regimes face a lower risk of coup attempts, while states engaged in repression face a higher risk. In looking at the causes of civil conflict onsets, the past vulnerability of states to coups does not have a significant impact, but the presence of an ongoing conflict and larger numbers of prior c ivil conflict onsets both make states more likely to experience a current onset. Likewise, the presence of a credible rebel group makes the state more susceptible to a conflict onset, as does an international dispute involving the use of force. Unlike in t he coup attempt analysis, military 83 manipulation and regime strength do not have much impact. States that employ repression are also more vulnerable to civil conflict onsets, although the continued use of repression over time mitigates that vulnerability, a nother departure from the coup attempt analysis. Moving to the bivariate probit model with the additional covariates, if the independence - of - events framework is the most helpful, we should see no specific relationship between the co the constraints or trade - positively signed for the constraints model and negativel y signed for the trade - offs model. Likewise, the relationship between the coefficients across the two equations should also differ based on whether the constraints or trade - offs model is more useful. If the constraints model has stronger support, the coeff icients should be similarly signed factors that make one phenomenon more likely should make the other more likely as well. If, however, the trade - offs model is more strongly supported, coefficients should be oppositely signed factors that make one phen omenon more likely should make the other less likely. That is, we should see evidence of one framework of the model. The results of the inclusive original model w ith sample size restricted to overlap with the Powell and Buhaug modified/extended replications are presented in Table 3.9. This restricted sample size is also designed to make comparison between the models possible. Table 3.3 demonstrated that combining t he Powell and Buhaug probit models into a single bivariate probit model was a useful enterprise. Based on that, it is helpful to think about what this original model, that introduces a number of new variables, provides us in terms of leverage as compared t o the simple combination bivariate probit. 84 In looking at the substantive results of the analysis, only ongoing civil conflicts affects both the likelihood of a coup attempt and the likelihood of a civil conflict onset, in each case making the phenomenon u nder consideration more likely. The number of previous coup attempts is strongly related to the likelihood of a coup attempt: more consolidated regimes, be they autocratic or democratic, are less likely to experience a coup attempt. Additionally, states wi th a greater effective number of military organizations that is, states that engage in more counterbalancing, are also less likely to experience a coup attempt, in large part because the actual armed forces in such states have been stripped of some of th eir capability. In the civil conflict equation, the number of previous conflict onsets is very strongly linked to the likelihood of outbreak of a new conflict episode. The presence of credible rebel groups and participation in an international dispute that involves the use of force also makes conflict onset more likely. Short - term use of repression makes civil conflict onset more likely, while longer - term use of repression by a regime makes such an onset less likely, suggesting that it may be possible to k eep even disgruntled populations quiescent through sustained violation of civil and political liberties. The failure of some variables in particular to reach statistical significance was surprising. None of the standard control variables GDP per capita, population and mountainous terrain was significant in either equation, suggesting that when the variety of other factors examined here are taken into consideration, they are not such robust predictors after all. I use a Wald test to examine the joint im pact of those variables on the explanatory power of the model. A non - significant result of the Wald test would indicate that the inclusion of the variables has no significant bearing on the model. Indeed, when I perform the Wald test on the three variables , the null of no contribution cannot be rejected, meaning the variables do not jointly significantly 85 contribute to the model fit across both equations. With six degrees of freedom, the is only 5.47, far below the threshold for statistical significan ce. More surprising from a theoretical standpoint was the poor performance of the spending and military variables. The spending variables were at the core of the trade - offs model, with increased spending on one group expected to increase the animosity of the other, resulting in an attempt at the removal of the incumbent by that disaffected group. The military - oriented variables, namely the size of the military and the extent to which incumbents engage in counterbalancing, were expected to affect the loyalt y and readiness of the military, which would have a bearing on both coup attempts and civil conflict onsets. A Wald test on the spending variables reveals that these variables too are jointly non - significant: the , with four degrees of freedom, is 6. 11. The military manipulation variables fare little better: with six degrees of freedom, the is 6.78, again indicating that these variables do not jointly make a significant contribution to the fit of the model. positive, and it is statistically significant, indicating that using a bivariate probit model was necessary in order to avoid inefficient and biased coefficient estimates. This holds with the findings from both Powell and Buhaug bivariate probit models. Although the individual variables shed only limited light on the phenomena at hand, this is a broader problem in conflict literature more broadly, as both studies on interstate conflict (Gartzke 1999) and ci vil conflict demonstrate (Ward, Greenhill and Bakke 2010) indicate; because there have been so few studies on coups in recent years, model fit and predictive power have not been as widely examined, but the extent to which rom previ ous approaches in both model and variables indicates that there exists even less consensus in the coup literature than in the conflict literature as to what 86 is remaining positive and significant across the different bivariate probit models is perhaps more telling. While there is more work to be done in explaining these phenomena, an issue I will in more detail later in this chapter, it is becoming quite clear that it is detrimental to maintain that they are wholly distinct events and that understanding one event will provide no assistance in understanding the other. Figures 3.8 and 3.9 are consistent with previous figures demonstrating that coefficient and confidence interval estimates for variables do indeed differ between the separate and bivariate probit models. This again provides evidence that looking at the two phenomena as independent, wholly distinct events produces flawed estima tes as compared to looking at them as part of a similar underlying process. The last component of the statistical analysis presented in this chapter is to expand the sample under consideration for the model with additional covariates. The data for the mod el with additional covariates ranges from 1976 - 2005, beginning and ending later than either the Powell (2012) or Buhaug (2006) datasets. When the sample restriction is relaxed so that the models will no longer be compared with those from the Powell and Buh aug modified replications, there is a gain of more than 1000 cases. In the cases of both the separate probit analyses and the bivariate probit analysis, a number of interesting results emerge in the results obtained from this sample size that depart from t he findings obtained from the smaller sample size. Looking at the separate coup attempt probit model, the results of which are presented in Table 3.10, most of the variables that were significant in the smaller sample number of pr evious coup attempts (+), ongoing civil conflict (+), regime consolidation ( - ) and repression (+) are again significant and in the same direction in the larger sample. Differences in the findings between the two sample sizes emerge, however. The number o f effective military organizations loses its 87 significance, no longer dampening the risks of a coup attempt. GDP per capita and years of repression gain significance with a negative impact, and instability becomes statistically significant with a positive i mpact. With the larger sample size comes a more nuanced view of the world. In the results from the separate civil conflict onset model, displayed in Table 3.11, all of the statistically significant variables in the restricted - samp le model are again statistically significant. With one exception, discussed below, the variables also retained their direction of significance: number of previous conflict onsets (+), credible rebel group (+), international dispute involving the use of for ce (+), repression (+) and years of repression ( - ). Ongoing conflict had a positive impact on current onsets in the smaller sample size, but has a negative impact on conflict onsets in the larger sample. With more data points available come more conflic ts and the potential for more ongoing conflicts, so the exact relationship between the two can be more thoroughly assessed in the larger sample. Many countries had more than one conflict ongoing at any given time, and as conflicts accumulate, new rebel gro ups may be less all the other groups vying for contention and concessions. Such a process is highlighted in the struggles of some groups in the Philippine s to gain traction against the state in the face of stronger groups already in contention with the government (Jennings 2013) with more data points available, has a negative impact on conflict onsets. Two additional variables gained statistical significan ce in the larger sample size. Social welfare spending has a negative impact on civil conflict onset, lending credence to the argument and the state and lessen t he willingness to rebel. Additionally, instability became significant and 88 has a positive impact on the onset of civil conflict. Given that instability involves the active display of dissatisfaction with the state on the part of the populace, it makes sense that conflict of a more violent nature might follow from this lower - level show of resistance. Table 3.12 contains the results of the bivariate probit with additional covariates. In the coup attempt equation, a history of previous attempts make an attempt in any given country - year more likely. Ongoing civil conflicts also make coup attempts more likely, lending support to the argument that the military views civil conflicts as an indication of the incompetence of the incumbent sh ort conflicts may flare up for reasons beyond the direct control of the incumbent, but the inability to bring a conflict to a close is a symptom. The - 6 to +7 measure of regime type fails to reach statistical significance, meaning there is no linear relati onship between regime type and coup attempts. However, that same measurement squared so that more consolidated regimes, be they autocratic or democratic, have higher values is negative and significant, suggesting that regime type in itself matters less than regime strength. GDP per capita has a significant negative impact on the likelihood of coup attempts as well, lending support to the view that state capacity matters. Instability and repression both make coup attempts more likely, while longer - term r epression makes coup attempts less likely. This is counter to the expectation identified in Hypothesis 5, which was that sustained repression would increase the risk of a coup attempt. It is possible that the initial burden of repressing the public might b e traumatizing to members of the military, making a coup attempt more likely, but as such activities become part of the standard operating procedures of the military, and thus more routine, they becomes less of a psychological burden on the armed forces. In the civil conflict onset equation, a history of coup attempts and ongoing civil conflict actually make a new civil conflict onset less likely, counter to expectations. In both cases, it could 89 be that potential rebels see the state already breaking down internally (in the case of coup attempts) or facing strong external threats (from other rebel organizations) and feel less need to try to overcome the collective action problem. If the goal is a new regime, it is less costly in terms of both time, lives an d resources to let the military or another organization bring about that change than to take on the effort oneself. However, a history of civil conflicts makes a new onset in any given country - year more likely, in line with previous findings. The presence of a credible rebel group is also strongly associated with civil conflict onset, as is participation in an international militarized dispute involving the use of force. Social welfare spending significantly lowers the likelihood of a civil conflict onset, suggesting that the public does respond to spending oriented to their needs. Instability and repression both increase the likelihood of a civil conflict onset, while long - term repression makes a rebellion less likely. This is in line with the expectation in Hypothesis 5, but since that hypotheses predicted a trade - off situation, the hypothesis as a whole must be rejected. Instead, the evidence supports a constraints explanation. As with the previous b supporting the constraints framework. Consistent with the other graphical evidence presented thus far, Figures 3. 7 and 3. 8 again demonstrate the addition al leverage gained in coefficient and confidence interval estimation by modeling coup attempts and civil conflict onsets as part of a bivariate probit rather than separate, independent probit models. Model Fit One way to gauge the utility of the modified and extended replications as well as the original model developed here is to compare the fit of the models to one another. If the models where coup 90 attempts and civil conflict onsets are modeled separately from one another fit the data better than the mod els where they are modeled as part of the same underlying process, there would not have been no meaningful information gained from the latter. If, however, the models where the two phenomena are modeled together fit the data better than when they are model ed separately as is indeed the case then modeling civil conflict onsets and coup attempts together sheds new light and reveals new information about both phenomena. Although the coefficients on the independent variables do contain information about rel ationships and how processes unfold, the measure of model fit speak to the utility and improvements in understanding gained by the overall pursuit carried out here. The assessments in this section are why co nsistent sample compositions were particularly i mportant the extent to which the models fit the data can only be meaningfully compared if the same set of data is under examination in each model. In this section I will consider the fit of each model and compare across models based on two criteria: the log - likelihood and the Akaike and Bayesian Information Criteria (AIC/BIC). Tables 3.13 - 3.15 provide a summary of the log - likelihood, AIC and BIC for each of the models discussed above, sorted by sample size to highlight com parison groups. Probit models are often estimated using maximum likelihood estimation techniques, which seek to identify the value Thus, models with higher l og - likelihoods (a greater maximized likelihood) fit better than models with lower log - likelihoods. The concept of the AIC is to measure the information lost between the ion that the number of parameters is taken into consideration; models with more parameters may 91 explain more variance, for instance, but may not do so as efficiently as a model with fewer parameters that perhaps also explains less variance. The BIC is akin to the AIC, but imposes a harsher penalty for each parameter than the AIC. This is designed to penalize models that might overfit the noise in the data. Figure 3.9 shows the in - - Buhaug seemingly unrelated bivariate regression model in separation plots. Separation plots have quick and easy - to - understand summary of a model Sacks 2011, 995) because contained within each separation plot is an array of information, including: 1. The r elative number of 0s and 1s in the actual data. This provides an indication of the sparsity of the events in the data used to generate the model. 2. The range and degree of variation among the predicted probabilities generated by the model; 3. The degree to which high predicted probabilities correspond to actual instances of the event, and low predicted probabilities correspond to nonevents in other words, a visual summary of the fit of the model; 4. The total number of events predicted by the model. (G reenhill, Ward and Sacks 2011, 995) The data in separation plots are arranged in order, from left to right, of increasing predicted probabilities. Actual occurrences of a coup attempt are indicated by a vertical red line, and the predicted probability of cases are displayed with a line. The better the fit of the model, the more completely separated are occurrences and non - occurrences, and occurrences will be concentrated among cases with higher predicted probabilities of that event, while non - occurrences w ill be concentrated among cases with low predicted probabilities of the event. - Buhaug seemingly unrelated bivariate probit separation plot in Figure 3.9b, we can obtain info rmation 92 about the worst - fitting cases in each plot (that is, coup attempts that occurred despites having very low predicted probabilities given the model) as well as how those cases fit in the other model. The five worst - fitting cases from the Powell model were the United Arab Emirates in 1987 (predicted probability of 0.0047), Paraguay in 1989 (0.0057), Qatar in 1995 (0.0057), Tunisia in 1987 (0.0066) and Turkey in 1971 (0.0068). In the Powell - Buhaug Seemingly Unrelated Bivariate Probit Model, those same c ases had the following predicted probabilities: United Arab Emirates in 1987 (0.0061), Paraguay in 1989 (0.0051), Qatar in 1995 (0.0062), Tunisia in 1987 (0.0058) and Turkey in 1971 (0.0076). The five worst - fitting cases from the Powell - Buhaug model were P araguay in 1989, Tunisia in 1987, the United Arab Emirates in 1987, Qatar in 1995 and Venezuela in 1992. While Qatar in 1995 and the United Arab Emirates in 1987 fit better in the bivariate probit, Paraguay in 1989, Tunisia in 1987 and Venezuela in 1992 fi t better in the separate Powell probit model (Venezuela in 1992 in the separate Powell model had a predicted probability of 0.0101). Similar, albeit more modest, results can be seen in the separation plots comparing the model fit - Buhaug Seemingly Unrelated Bivariate Probit model. The five least accurate predictions in the separate Buhaug model were the United Kingdom in 1998 (predicted probability of conflict onset of 0.0047), Trinidad and Tobago in 1990 (0.0068), Lebanon in 1975 (0.0072), the United Kingdom in 1971 (0.0084) and Saudi Arabia in 1979 (0.0101). The Powell - Buhaug Seemingly Unrelated Bivariate Probit model had the same worst - fitting cases in the same order : the United Kingdom in 1998 (0.0048), Trinidad and Tobago in 1990 (0.0066), Lebanon in 1975 (0.0079), the United Kingdom in 1971 (0.0083) and Saudi Arabia in 1979 (0.0102). Of these five cases, only Trinidad and Tobago in 1990 and the United Kingdom in 19 71 fit better in the separate Buhaug probit model. 93 Figure 3.11 illustrates the Powell single - equation probit model, Powell - Buhaug seemingly unrelated bivariate probit model, and the model with additional covariates presented a s both separate single - equation probit models and a bivariate probit model in separation plot form for the more restrictive sample that allows for model - fit comparison across all models. Looking at Figure 3.11, which focuses on coup attempts, the bivariate models appear to fit the data better than the single - equation probit models, for both the replication model and the model with additional covariates. Like in Figure 3.9 above, there is heavy overlap in the worst - fitting cases for the separate Powell model and the Powell - Buhaug Seemingly Unrelated Bivariate Probit model, and in the cases from the single - equation model with additional covariates and the bivariate probit model with additional covariates, although the two sets of equations are more different t han similar in terms of the cases they fit worst. Given the very different sets of variables in these separate models, this makes sense. However, it can be noted that the very lowest predicted probability in the models with additional covariates is higher than the lowest predicted probability from the replication models. Figures 3.12 and 3.13 display the point predictions for each of the four models for the worst fitting cases from the separate Powell and Powell - Buhaug bi variate probit models (Figure 3.11a and Figure 3.11b) and from the separate and bivariate probit models with additional covariates (Figure 3.11c and 3.11d). The five worst - fitting cases in the Powell separate - equation model were Paraguay in 1989 (0.0038), the United Arab Emirates in 1987 (0.0058), Tunisia in 1987 (0.0060), Argentina in 1988 (0.0079) and Syria in 1982 (00087). In the Powell - Buhaug Seemingly Unrelated Bivariate Probit model those cases fit as follows: Paraguay in 1989 (0.0031), the United Ara b Emirates in 1987 (0.0084), Tunisia in 1987 (0.0052), Argentina in 1988 (0.0091) and Syria in 1982 (0.0111). The fits for the model with additional covariates are as follows, with the predicted 94 probabilities for the separate equation listed first and from the bivariate probit model second in each case: Paraguay in 1989 (0.0343, 0.0300), the United Arab Emirates in 1987 (0.0125, 0.0117), Tunisia in 1987 (0.0386, 0.0398), Argentina in 1988 (0.0544, 0.0623) and Syria in 1982 (0.2625, 0.2931). Without fail, th e model with additional covariates, whether in separate - equation or bivariate form, fit those cases better than the replication models. For the five worst - fitting cases from the Powell - Buhaug Seemingly Unrelated Bivariate Probit model, I will list the pre dicted probabilities from that model, the separate Powell equation, the separate - equation model with additional covariates and the bivariate model with additional covariates, respectively, to facilitate comparison across models. The five worst - fitting case s form the Powell - Buhaug Seemingly Unrelated Bivariate Probit model were: Paraguay in 1989 (0.0031, 0.0038, 0.0343, 0.0300); Tunisia in 1987 (0.0052, 0.0060, 0.0386, 0.0398); the United Arab Emirates in 1987 (0.0084, 0.0058, 0.0125, 0.0117), Argentina in 1 988 (0.0091, 0.0079, 0.0544, 0.0623) and Venezuela in 1992 (0.0099, 0.0150, 0.0297, 0.0311). With the exception of the United Arab Emirates in 1987, there was no overlap in the sets of worst - fitting cases between the replication model and the model with a dditional covariates. For the worst - fitting cases in the separate - equation model with additional covariates, I will present the predicted probabilities for that model, the single - equation Powell model, the Powell - Buhaug Seemingly Unrelated Bivariate Probit and the bivariate model with additional covariates, respectively: Nigeria in 1983 (0.0054, 0.0855, 0.0742, 0.0063), Madagascar in 1992 (0.0082, 0.0371, 0.0320, 0.0101), Spain in 1981 (0.0091, 0.0260, 0.0195, 0.0169), the Philippines in 1987 (0.0125, 0.187 9, 0.1572, 0.0222) and the United Arab Emirates in 1987 (0.0125, 0.0058, 0.0084, 0.0117). 95 Nigeria in 1983, Madagascar in 1992 and the United Arab Emirates also proved particularly difficult for the bivariate model with additional covariates to accurat ely predict. For the worst - fitting cases in the bivariate model with additional covariates, I will present the predicted probabilities for that model, the single - equation Powell model, the Powell - Buhaug Seemingly Unrelated Bivariate Probit and the separate - equation model with additional covariates, respectively: Nigeria in 1983 (0.0063, 0.0855, 0.0742, 0.0054), Madagascar in 1992 (0.0101, 0.0371, 0.0320, 0.0082), the United Arab Emirates in 1987 (0.0117, 0.0058, 0.0084, 0.0125), Lesotho in 1994 (0.0154, 0.1 178, 0.1142, 0.0163) and Panama in 1989 (0.0157, 0.1703, 0.1841, 0.0180). Although the bivariate model with additional covariates fit Nigeria 1983 and Madagascar 1992 better than the separate - equation model with additional covariates, the separate equation fit the United Arab Emirates in 1987, Lesotho in 1994 and Panama in 1989 more accurately. Except for the case of the UAE, the replication model both single - equation and bivariate probit fit these data points better than the model with additional covar iates. Figure 3.14 illustrates the Buhaug single - equation probit model, Powell - Buhaug seemingly unrelated bivariate probit model, and the model with additional covariates presented as both separate single - equation probit models and a bivariate probit model in separation plot form for the more restrictive sample that allows for model - fit comparison across all models. Figure 3.14 follows the same pattern as Figure 3.11, though for conflict onsets rather than coup attempts. There is on ly slight variation in which cases fit most poorly across the four models, with most of the same cases appearing in each. Just as in Figure 3.11, comparisons can be made between the separate - equation Buhaug replication and the Powell - Buhaug Seemingly Unrel ated Bivariate Probit model, and between the model with additional covariates in both its separate conflict onset probit equation and as a bivariate probit. The lowest predicted probability from either of the versions of the model 96 with additional covariate s is higher than the lowest predicted probability from either of the versions of the replication model, by more than 1 percentage point. Figures 3.15 and 3.16 follow the pattern of Figures 3.12 and 3.13, illustrating th e worst - fitting cases for the separate Buhaug and Powell - Buhaug bivariate probit models and the separate and bivariate models with additional covariates. The five worst - fitting cases from the separate Buhaug probit are: the United Kingdom in 1998 (0.0041), Trinidad and Tobago in 1990 (0.0050), Yugoslavia in 1991 (0.0083), Saudi Arabia in 1979 (0.0101) and Haiti in 1989 (0.0117). The predicted probabilities from the Powell - Buhaug Seemingly Unrelated Bivariate Probit model, separate probit model with addition al covariates and bivariate probit model with additional covariates, respectively, are: the United Kingdom in 1998 (0.0044, 0.0626, 0.0657), Trinidad and Tobago in 1990 (0.0051, 0.0129, 0.0108), Yugoslavia in 1991 (0.0093, 0.0334, 0.0366), Saudi Arabia in 1979 (0.0098, 0.0063, 0.0060) and Haiti in 1989 (0.0131, 0.0086, 0.0082). The first four worst - fitting cases from the Powell - Buhaug Seemingly Unrelated Bivariate Probit model are the same as in the separate Buhaug probit model. The fifth worst - fitting ca se in the bivariate probit, however, is Spain in 1980 (0.0118). The predicted probability for this case in the separate Buhaug probit, separate equation model with additional covariates and bivariate probit model with additional covariates is 0.0122, 0.143 2 and 0.1387, respectively. For the two versions of the model with additional covariates, Saudi Arabia in 1979 and Trinidad and Tobago in 1990 continue to be poorly predicted. Haiti in 1989, Spain in 1985 and Ghana in 1981 are among the worst - fitting in both versions of the model with additional covariates, although they were not consistently worst - fitting among the versions of the replication model. For the separate equation model with additional covariates, the five worst - fitting were: Saudi Arabia in 1 979, Haiti in 1989, Spain in 1985 (0.0112), Trinidad and Tobago in 1990 and Ghana in 1981 97 (0.0129). Although Spain (0.0109) and Trinidad and Tobago switched places in terms of fit in the bivariate model with additional covariates, Ghana in 1981 (0.0115) r emained the fifth - most - poorly fit case. The predicted probabilities from the separate Buhaug probit and the Powell - Buhaug bivariate probit were: Spain in 1985 (0.0050, 0.0051) and Ghana in 1981 (0.0346, 0.0323). Figures 3.17 and 3.18 provide a point of comparison between the separate - equation and bivariate versions of the model with additional covariates. Figure 3.17 presents separation plots for coup attempts, and Figure 3.18 presents separation plots for civil conflict onse ts. Each set of worst - fitting cases is same across the two versions of the model: the same five worst - fitting coup attempt cases in the separate - equation version are the five worst - fitting cases in the same order in the bivariate version, and likewise with the worst - fitting conflict onset cases. Three of the worst - fitting coup attempt cases fit better in the bivariate version of the model, while only one of the worst - fitting conflict onset cases does so. The five worst - fitting coup attempt cases from the separate - equation and bivariate models, with predicted probabilities, were: Spain in 1981 (0.0033, 0.0041), Pakistan in 1999 (0.0053, 0.0050), Nigeria in 1983 (0.0054, 0.0065), the United Arab Emirates in 1987 (0.0103, 0.0093) and the Philippines in 1989 ( 0.0114, 0.0145). The five worst - fitting civil conflict onset cases from the separate - equation and bivariate models, with predicted probabilities, were: Saudi Arabia in 1979 (0.0027, 0.0027), Turkey in 2005 (0.0031, 0.0029), Venezuela in 1982 (0.0085, 0.007 5), Spain in 1985 (0.0087, 0.0089) and Chad in 1976 (0.0099, 0.0094). Discussion While the findings reported in the previous sections hinted back to the three theoretical models and which one seems to be most useful in gauging the relationship between co up attempts and 98 civil conflict onsets, this section will focus on a more explicit discussion of the findings as a whole and the theoretical frameworks. To that end, this section will proceed in two primary parts: first, identification and discussion of the overall trend in the statistical models and the theoretical framework that receives the most support, and second, the development of a deeper understanding of the relationship between the two phenomena within that broad theoretical framework. Although the variables and cases vary across models, there has been a consistent finding of a positive and statistically significant rho. The rho and associated confidence interval from each bivariate probit model the results of which have been presented in Tables 3.3, 3.6, 3.9 and 3.12 is depicted in Figure 3.19. This suggests that the constraints framework is the most useful in understanding the relationship between coup attempts and civil conflict onsets, with the unobserved factors ac ross the two equations having a similar effect on each equation the unobserved factors that make one event more (less) likely also make the other event more (less) likely. It is also possible to examine the effects particular variables have on the like lihood of each phenomenon. To do so, I have chosen three independent variables from the model with additional covariates with the unrestricted sample (presented in Table 3.12) that have statistically significant impacts on each dependent variable instabi lity, the use of repression and the number of years of repression use to also demonstrate the consistent effects across equations. I look at the marginal effects of each of these equations to show more precisely the impact that these specific variables h ave, holding dichotomous variables at their mode and all other variables at their median. The consistency of results for these variables also provides support for the constraints model, with the same observed factors that make one event more (less) likely also making the other event more (less) likely. 99 Figure 3.20 shows the marginal effect of instability on the probability of coup attempts and civil conflict onsets. Larger numbers of events like strikes and protests markedly incre ase the probability of both coup events and civil conflict onsets. Indeed, the probability of each event exceeds 50% for greater numbers of events. Again, this is the expected pattern given the constraints framework: instability weakens regimes and makes t hem more vulnerable to attack, both from the military and rebel groups. That is, a regime that has lost legitimacy and protests, strikes and riots are a symptom of that loss of legitimacy faces risks to its survival from various quarters of society, bo th within and outside the regime itself. Figures 3.21 and 3.22 show the short - and long - term marginal effects of repression on the probability of coup attempts and civil conflict onsets. In the short term, the use of re pression makes both events more likely, albeit less dramatically than instability. In the longer - term, however, sustained use of repression makes both events less likely. Both results are consistent with a political world characterized by constraints. On a day - to - day basis, acts of repression do not overwhelm those targeted, and may heighten their resolve, increasing their inclination to take aim at the leader. w ho are forced to take action against their fellow citizens, increasing frustration with the incumbent. Over time, however, the use of repression becomes part of the basic mission of the military, an expected and routine mode of operation. Likewise, the su stained use of repression continually reinforces to society that anti - regime will not be tolerated, and that any such activity will be dealt with harshly. This discourages would - be dissidents by increasing the costs of collective action. 100 Taken together, th e model - level - level results of consistent effects point to the constraints framework as an effective lens through which to view the processes of coup attempts and civil conflict onsets. Conclusion This chapter contained a statistical examination of the three theoretical frameworks developed in Chapter 2 and sought to identify which of the three models was the most helpful in understanding the relationship between coup attempts and civil conflict onsets. To build this investigation, I f irst conducted modified replications of two prominent studies one examining coup attempts (Powell 2012), the other civil conflict onsets (Buhaug 2006) to test whether a more detailed examination had merit. I first ran these two very different models te sting different expectations about different events separately to create a baseline sets of results. I then combined the two single - equation probit models into a single bivariate probit model to check whether the error terms were correlated. They were, ind icating that similar unobserved factors were having an impact on both equations. Moreover, this correlation was positively, suggesting that those unobserved factors were affecting both equations in the same way. I then developed an original bivariate prob it model with a wider range of covariates to test the various relationships hypothesized in Chapter 2. Using an unrestricted sample, instability, repression and years of repression had strong and consistent effects on coup attempts and civil conflict onset s instability and repression make both events more likely, while more sustained use of repression makes both less likely. Although factors like the presence of a credible rebel group or participation in a militarized international dispute involving the u se of force positively affected the likelihood of a civil conflict onset, they did not have a significant effect on the likelihood of a 101 coup attempt. Likewise, stronger regimes were less likely to experience coup attempts and states with higher GDPs per ca pita were less likely to experience a coup attempt, but neither of these When looking at the overall fit of the models to the data, the analyses in which coup attempts and civil conflict onsets were explicitly modeled together consistently had lower AIC and BIC values that is, fit better than those models where they were assumed to be independent events. This was true of both the modified/extended replications as well as of t he original analyses. These results point to a better understanding of coup attempts and civil conflict onsets when they are assumed to be part of the same underlying process than when they are assumed to be independent of one another. Taking the empirica l enterprise undertaken in this chapter in its entirety, all results point to the limited use of the independence of events model model after model some comprised of very different samples and very different covariates demonstrates that there is a sy stematic relationship between coup attempts and civil conflict onsets. Further, these models demonstrate that both the observed and unobserved factors affect the two phenomena in similar ways, so that factors that make one event more likely also make the o ther event more likely, and vice versa. This points to the constraints framework as a consistent reflection of the political world. Given the consistent rejection of the independence of events model and the coarse nature of country - year statistical data, the next chapter introduces the main issues under consideration in the case studies, and the subsequent two chapters present the case study investigations themselves of Burma from 1948 - 1962 and Nigeria from 1966 - 1967 of the constraints and trade - offs frameworks to get a more nuanced view of the political decisions and processes in states facing a 102 higher risk of these events. This will help illuminate whether the constraints framework is sufficient or if the trade - offs model gains any traction at a more micro level of investigation. 103 APPENDIX 104 Table 3.1. Powell (2012) Modified Replication Variable Name Coefficient Std. Error Change military exp. - 0.058 (0.128) Mil. exp. per soldier - 0.019 (0.066) Military personnel - 0.092* (0.0 38) Effective # mil. orgs. - 0.070 (0.095) Change GDP t - 1 0.192 (0.770) GDP per cap. t - 1 - 0.205** (0.078) Instability 0.083*** (0.013) Democracy - 0.183 (0.176) Autocracy - 0.118 (0.120) Military regime 0.949*** (0.127) Peace years - 0.069 (0.044) Pea ce yrs. spline 1 - 0.001 (0.001) Peace yrs. spline 2 0.001 (0.001) Peace yrs. spline 3 ^ - 0.007 (0.260) Constant 0.342 (0.611) N 2837 Log - Likelihood - 349.472 AIC 728.945 BIC 818.202 * p <0.05 , ** p <0.01, *** p <0.001 ^ coefficient/standard error*10 0 105 Table 3.2. Buhaug (2006) Modified Replication Variable Name Coefficient Std. Error Democracy t - 1 0.003 (0.124) Mixed regime t - 1 0.027 (0.121) GDP per cap. t - 1 - 0.062** (0.020) Oil exporter 0.208 (0.128) Country land area 0.096** (0.035) Ethni c frac. 0.307 (0.186) Conflict onset decay func. 1.229*** (0.138) Constant - 3.174*** (0.437) N 2837 Log - Likelihood - 392.579 AIC 801.159 BIC 848.763 * p <0.05 , ** p <0.01, *** p <0.001 106 Table 3.3. Seemingly Unrelated Bivariate Probit Combination of Powell (2012) and Buhaug (2006) Modified Replications Powell Equation Buhaug Equation Variable Name Coefficient Std. Error Coefficient Std. Error Change military exp. - 0.062 (0.127) Mil. exp. per soldier - 0.018 (0.065) Military personnel - 0. 088* (0.037) Effective # mil. orgs. - 0.084 (0.094) Change GDP t - 1 0.200 (0.755) GDP per cap. t - 1 - 0.177* (0.077) Instability 0.070*** (0.013) Democracy - 0.182 (0.176) Autocracy - 0.105 (0.119) Military regime 0.890*** (0.125) Peace y ears - 0.079 (0.043) Peace yrs. spline 1 - 0.001 (0.001) Peace yrs. spline 2 0.001 (0.001) Peace yrs. spline 3 ^ - 0.016 (0.026) Democracy t - 1 - 0.013 (0.123) Mixed regime t - 1 0.034 (0.119) GDP per cap. t - 1 - 0.060** (0.019) Oil exporter 0 .190 (0.127) Country land area 0.091** (0.035) Ethnic frac. 0.278 (0.185) Conflict onset decay func. 1.244*** (0.135) Constant 0.198 (0.606) - 3.105*** (0.438) N 2837 0.421 (0.087) Log - Likelihood - 727.245 AIC 1502.491 BIC 1645.303 2 =29.613 * p <0.05 , ** p <0.01, *** p <0.001 ^ coefficient/standard error*100 107 Figure 3.1. Plot of Coeffi cients and Confidence Intervals for Coup Attempts in Separate Powell Model and Simple Powell - Buhaug Combination Model 108 Figure 3.2. Plot of Coefficients and Confidence Intervals for Conflict Onsets in Separate Buhaug Model and Simple Powell - Buhaug Com bination Model 10 9 Table 3.4. Powell Modified Replication with More Restricted Sample Variable Name Coefficient Std. Error Change military exp. - 0.019 (0.142) Mil. exp. per soldier 0.028 (0.092) Military personnel - 0.083 (0.049) Effective # mil. o rgs. - 0.055 (0.117) Change GDP t - 1 0.193 (0.938) GDP per cap. t - 1 - 0.189 (0.102) Instability 0.069*** (0.017) Democracy - 0.065 (0.212) Autocracy - 0.080 (0.153) Military regime 0.900*** (0.160) Peace years - 0.121* (0.054) Peace yrs. spline 1 - 0.002 (0.001) Peace yrs. spline 2 0.001 (0.001) Peace yrs. spline 3 ^ 0.005 (0.040) Constant - 0.104 (0.772) N 2028 Log - Likelihood - 220.102 AIC 470.204 BIC 554.426 * p <0.05 , ** p <0.01, *** p <0.001 ^ coefficient/standard error*100 110 Table 3.5. Buhau g Modified Replication with More Restricted Sample Variable Name Coefficient Std. Error Democracy t - 1 - 0.018 (0.141) Mixed regime t - 1 - 0.135 (0.148) GDP per cap. t - 1 - 0.066** (0.022) Oil exporter 0.203 (0.149) Country land area 0.115** (0.044) Ethni c frac. 0.317 (0.222) Conflict onset decay func. 1.316*** (0.159) Constant - 3.389*** (0.542) N 2028 Log - Likelihood - 277.267 AIC 570.534 BIC 615.453 * p <0.05 , ** p <0.01, *** p <0.001 111 Table 3.6. Seemingly Unrelated Bivariate Probit Combination of Powell (2012) and Buhaug (2006) Modified Replications with More Restrictive Sample Powell Equation Buhaug Equation Variable Name Coefficient Std. Error Coefficient Std. Error Change military exp. - 0.023 (0.140) Mil. exp. per soldier 0.034 (0.09 2) Military personnel - 0.077 (0.048) Effective # mil. orgs. - 0.083 (0.117) Change GDP t - 1 0.259 (0.916) GDP per cap. t - 1 - 0.150 (0.100) Instability 0.059*** 0.017 Democracy - 0.096 (0.209) Autocracy - 0.072 (0.152) Military regime 0.8 11*** (0.158) Peace years - 0.131 (0.053) Peace yrs. spline 1 - 0.002 (0.001) Peace yrs. spline 2 0.001 (0.001) Peace yrs. spline 3^ - 0.006 (0.038) Democracy t - 1 - 0.048 (0.140) Mixed regime t - 1 - 0.126 (0.146) GDP per cap. t - 1 - 0.062** (0.022) Oil exporter 0.163 (0.149) Country land area 0.102* (0.044) Ethnic frac. 0.294 (0.221) Conflict onset decay func. 1.347*** (0.157) Constant - 0.326 (0.766) - 3.213*** (0.545) N 2028 0.443 (0.087) Log - Likelihood - 486.654 AIC 1021.309 BIC 1156.064 2 =21.429 * p <0.05 , ** p <0.01, *** p <0.001 ^ coefficient/standard error*100 112 Figure 3.3. Plot of Coeffi cients and Confidence Intervals for Coup Attempts in Separate Powell Model and Powell - Buhaug Combination Model with More Restricted Sample 113 Figure 3.4. Plot of Coefficients and Confidence Intervals for Co nflict Onsets in Separate Buhaug Model and Pow ell - Buhaug Combination Model with More Restricted Sample 114 Table 3.7. Coup Attempt Analysis with Additional Covariates with More Restricted Sample. Variable Name Coefficient Std. Error # prev. coup attempts 0.089*** (0.015) Ongoing conflict 0.991* (0.428) # prev. conf. onsets - 0.013 (0.039) Credible rebel group - 0.006 (0.173) Intl. dispute - 0.065 (0.173) Excluded population 0.029 (0.254) Social welfare spend t - 1 ^ 0.052 (1.971) Military spend t - 1 - 0.039 (0.025) Military personnel ^ 0.011 (0.041 ) # effective military orgs. - 0.255* (0.118) Xpolity - 0.019 (0.028) Xpolity sq. - 0.016* (0.007) GDP per cap. t - 1 - 0.140 (0.096) Instability 0.006 (0.026) Repression 0.325* (0.016) Years of repression - 0.008 (0.016) Mtn. terrain - 0.053 (0.047) Po pulation - 0.058 (0.047) Mil. - pop. ratio - 0.771 (16.389) Constant 0.577 (1.320) N 2028 Log - Likelihood - 227.593 AIC 495.186 BIC 607.482 * p <0.05 , ** p <0.01, *** p <0.001 ^ coefficient/standard error*100 115 Table 3.8. Civil Conflict Onset Analysis with Additional Covariates with More Restricted Sample. Variable Name Coefficient Std. Error # prev. coup attempts - 0.025 (0.017) Ongoing conflict 0.656* (0.316) # prev. conf. onsets 0.207*** (0.031) Credible rebel group 0.501** (0.160) Intl. disput e 0.452** (0.151) Excluded population 0.509 (0.284) Social welfare spend t - 1 - 0.038 (0.022) Military spend t - 1 - 0.011 (0.022) Military personnel ^ - 0.030 (0.022) # effective military orgs. - 0.011 (0.107) Xpolity - 0.002 (0.027) Xpolity sq. 0.002 (0. 007) GDP per cap. t - 1 - 0.008 (0.091) Instability 0.013 (0.020) Repression 0.597** (0.199) Years of repression - 0.031* (0.015) Mtn. terrain - 0.040 (0.055) Population - 0.055 (0.081) Mil. - pop. ratio - 8.431 (15.460) Constant - 1.344 (1.373) N 2028 Lo g - Likelihood - 203.229 AIC 446.457 BIC 558.754 * p <0.05 , ** p <0.01, *** p <0.001 ^ coefficient/standard error*100 116 Table 3.9. Coup Attempt and Civil Conflict Onset Bivariate Probit Analysis with Additional Covariates with More Restricted Sample Coup Attempt Equation Conflict Onset Equation Variable Name Coefficient Std. Error Coefficient Std. Error # prev. coup attempts 0.088*** ( - 0.015) - 0.026 (0.017) Ongoing conflict 1.109** (0.405) 0.619* (0.315) # prev. conf. onsets - 0.021 (0.040) 0.204 *** (0.031) Credible rebel group 0.048 (0.168) 0.523*** (0.159) Intl. dispute - 0.082 (0.171) 0.469** (0.149) Excluded population 0.068 (0.252) 0.496 (0.282) Social welfare spend t - 1 - 0.049 ^ (1.962) ^ - 0.036 (0.021) Military spend t - 1 - 0.044 (0.026) - 0. 012 (0.023) Military personnel ^ - 0.014 (0.045) - 0.025 (0.021) # effective military orgs. - 0.243* (0.115) - 0.036 (0.107) Xpolity - 0.023 (0.025) - 0.005 (0.027) Xpolity sq. - 0.014* (0.007) 0.002 (0.007) GDP per cap. t - 1 - 0.152 (0.095) - 0.004 (0.091) I nstability 0.006 (0.025) 0.014 (0.020) Repression 0.301 (0.160) 0.616** (0.201) Years of repression - 0.008 (0.016) - 0.032* (0.015) Mtn. terrain - 0.055 (0.046) - 0.038 (0.054) Population - 0.036 (0.077) - 0.046 (0.080) Mil. - pop. ratio 1.238 (16.438) - 7.29 4 (15.360) Constant 0.294 (1.317) - 1.510 (1.370) N 2028 0.517 (0.097) Log - Likelihood - 419.860 AIC 921.720 BIC 1151.927 2 =21.923 * p <0.05 , ** p <0.01, *** p <0.001 ^ coefficient/standard error*100 117 Figure 3.5. Plot of Coefficie nts and Confidence Intervals for Coup Attempts in Separate Probit Model with Additional Covariates and B ivariate Probit Model with Additional Covariates with More Restricted Sample 46 46 The original military - population ratio variable has been multiplied by 100 so the scale of coefficients and confidence interval is more comparable to those of the other variables for the purposes of this graph. 118 Figure 3.6. Plot of Coefficients and Confidence Intervals for Conflict Onsets in Separate Probit Model with Additional Covariates and B ivariate Probit Model with Additional Covariates with More Restricted Sample 47 47 The original military - population ratio variable has been multiplied by 100 so the scale of coefficients and confidence interval is more comparable to those of the other variables for the purposes of this graph. 119 Table 3.10. Coup Attempt Analysis with Additional Covariates with Unrestricted Sample Variable Name Coeffici ent Std. Error # prev. coup attempts 0.082*** (0.011) Ongoing conflict 0.477*** (0.147) # prev. conf. onsets - 0.010 (0.023) Credible rebel group 0.162 (0.123) Intl. dispute - 0.005 (0.121) Excluded population - 0.108 (0.199) Social welfare spend t - 1 - 0.014 (0.016) Military spend t - 1 - 0.012 (0.015) Military personnel ^ - 0.007 (0.029) # effective military orgs. - 0.055 (0.082) Xpolity - 0.028 (0.021) Xpolity sq. - 0.012* (0.005) GDP per cap. t - 1 - 0.144* (0.065) Instability 0.034** (0.013) Repressio n 0.370** (0.131) Years of repression - 0.037*** (0.010) Mtn. terrain - 0.057 (0.036) Population - 0.106 (0.058) Mil. - pop. ratio - 10.188 (11.459) Constant 1.140 (0.998) N 3402 Log - Likelihood - 402.818 AIC 845.637 BIC 968.279 * p <0.05 , ** p <0.01, * ** p <0.001 ^ coefficient/standard error*100 120 Table 3.11. Civil Conflict Onset Analysis with Additional Covariates with Unrestricted Sample Variable Name Coefficient Std. Error # prev. coup attempts - 0.038** (0.013) Ongoing conflict - 0.908*** (0.15 0) # prev. conf. onsets 0.143*** (0.014) Credible rebel group 0.592*** (0.107) Intl. dispute 0.428*** (0.102) Excluded population 0.266 (0.205) Social welfare spend t - 1 - 0.049** (0.016) Military spend t - 1 ^ 0.015 (1.174) Military personnel ^ - 0.026 (0 .015) # effective military orgs. - 0.054 (0.073) Xpolity 0.017 (0.019) Xpolity sq. ^ 0.042 (0.465) GDP per cap. t - 1 - 0.084 (0.055) Instability 0.023* (0.010) Repression 0.726*** (0.153) Years of repression - 0.023** (0.008) Mtn. terrain - 0.046 (0.03 8) Population - 0.044 (0.054) Mil. - pop. ratio 1.484 (7.484) Constant - 0.653 (0.946) N 3402 Log - Likelihood - 456.261 AIC 952.523 BIC 1075.165 * p <0.05 , ** p <0.01, *** p <0.001 ^ coefficient/standard error*100 121 Table 3.12. Coup Attempt and Civi l Conflict Onset Bivariate Probit Analysis with Additional Covariates with Unrestricted Sample Coup Attempt Equation Conflict Onset Equation Variable Name Coefficient Std. Error Coefficient Std. Error # prev. coup attempts 0.081*** (0.011) - 0.042** (0 .014) Ongoing conflict 0.478*** (0.145) - 0.913*** (0.150) # prev. conf. onsets - 0.025 (0.025) 0.142*** (0.014) Credible rebel group 0.204 (0.120) 0.604*** (0.106) Intl. dispute - 0.008 (0.118) 0.443*** (0.101) Excluded population - 0.065 (0.196) 0.258 ( 0.203) Social welfare spend t - 1 - 0.017 (0.016) - 0.048** (0.016) Military spend t - 1 - 0.012 (0.015) - 0.011 ^ (1.183) ^ Military personnel ^ - 0.009 (0.294) - 0.029 (0.014) # effective military orgs. - 0.050 (0.081) - 0.063 (0.072) Xpolity - 0.024 (0.021) 0.015 (0.019) Xpolity sq. - 0.012* (0.005) 0.000 (0.005) GDP per cap. t - 1 - 0.151* (0.065) - 0.081 0.055 Instability 0.036** (0.013) 0.023* (0.010) Repression 0.362** (0.130) 0.740*** (0.154) Years of repression - 0.036*** (0.009) - 0.023** (0.008) Mtn. terra in - 0.060 (0.036) - 0.045 (0.038) Population - 0.090 (0.058) - 0.034 (0.054) Mil. - pop. ratio - 9.863 (11.310) 2.212 (7.431) Constant 0.964 (0.985) - 0.843 (0.944) N 3402 0.544 (0.072) Log - Likelihood - 838.137 AIC 1758.275 BIC 2009.927 2 =41.885 * p <0.05 , ** p <0.01, *** p <0.001 ^ coefficient/standard error*100 122 Figure 3.7. Plot of Coeffi cients and Confidence Intervals for Coup Attempts in Separate Probit Model with Additional Covariates and Bivariate Probit Model with Additional Covariates with Unrestricted Sample 48 48 The original military - population ratio variable has been multiplied by 100 so the scale of coefficients and confidence interval is more comparable to those of the other variables for the pu rposes of this graph. 123 Figure 3.8. Plot of Coefficients and Confidence Intervals for Conflic t Onsets in Separate Probit Model with Additional Covariates and Bivariate Probit Model with Additional Covariates with Unrestricted Sample 49 49 The original military - population ratio variable has been multiplied by 100 so the scale of coefficients and confidence interval is more comparable to those of the other variables for the purposes of this graph. 124 Table 3.13. Bivariate Probit Log - Likelihood and Information Criteria for Combination Powell and Buhaug Modified Replications Model Name Sample Size - 2*Log - Likelihood AIC BIC Combination if No Correlation between Equations 2837 1484.10 1530.10 1666.97 Seemingly Unrelated Combination 2837 1454.5 1502.49 1645.30 125 Table 3.14. Bivariate Probit Log - Likelihood and Information Criteria for Combination Powell and Buhaug Modified Replications and Model with Additional Covariates with More Restricted Sample Model Name Sample Size - 2*Log - Likelihood AIC BIC Powell/Buhaug Modified Replications if No Correlation between Equations 2028 994.76 1040.73 1169.88 Seemingly Unrelated Powell/Buhaug Modified Replications 2028 973.3 1021.31 1156.06 Model with Additional Covariates if No Correlation between Equations 2028 861.64 941.64 1166.24 Model with Additional Covariates 2028 839.72 921.72 1151.93 126 Table 3.15. Bivariate Probit Log - Likelihood and Information Criteria Information for Model with Additional Covariates with Unrestricted Sample Model Name Sample Size - 2*Log - Likelihood AIC BIC Model with Additional Covariat es if No Correlation between Equations 3402 1718.16 1798.16 2040.44 Model with Additional Covariates 3402 1676.28 1758.28 2009.69 127 Figure 3.9. Separation Plots of Powell Coup Attempts Figure 3.9a. Separate Powell Probit Model Figure 3.9b. Powell - Buhaug Seemingly Unrelated Bivariate Probit Model 128 Figure 3.10. Separation Plots of Buhaug Conflict Onsets Figure 3.10a. Separate Buhaug Probit Model Figure 3.10b. Powell - Buhaug Seemingly Unrelated Bivariate Probit Model 129 Figure 3.11 . Separation Plots of Cou p Attempts with More Restricted Sample Figure 3.11a. Separate Powell Probit Model Figure 3.11b. Powell - Buhaug Seemingly Unrelated Bivariate Probit Model Figure 3.11c. Separate Probit Model with Additional Covariates Figure 3.11d: Bivariate Probit Model with Additional Covariates 130 Figure 3.12. Fit of Each Model on Five Worst - Fitting Coup Attempt Cases from Powell and Powell - Buhaug Models 131 Figure 3.13. Fit of Each Model on Five Worst - Fitting Coup Attempt Cas es from Separate and Bivariate Mo dels with Additional Covariates 132 Figure 3.14. Separation Plots of Conflict Onsets with More Restricted Sample Figure 3.14a. Separate Buhaug Probit Model Figure 3.14b. Powell - Buhaug Seemingly Unrelated Bivariate P robit Model Figure 3.14c. Separate Probit Model with Additional Covariates Figure 3.14d. Bivariate Probit Model with Additional Covariates 133 Figure 3.15. Fit of Each Model on Five Worst - Fitting Civil Conflict Onset Cases from Buhaug and Powell - Buhaug Models. 134 Figure 3.16. Fit of Each Model on Five Worst - Fitting Civil Conflict Onset Cases from Separate and Bivariate Models with Additional Covariates. 135 Figure 3.17. Separation Plots of Coup Attempts with Unrestricted Sample Fig ure 3.17a. Separate Probit Model with Additional Covariates Figure 3.17b: Bivariate Probit Model with Additional Covariates 136 Figure 3.18. Separation Plots of Civil Conflict Onsets with Unrestricted Sample Figure 3.18a. Separate Probit Model wit h Additional Covariates Figure 3.18b: Bivariate Probit Model with Additional Covariates 137 138 Figure 3.20. Marginal Effect of Instability on the Probability of C oup Attempts and Civil Conflict Onsets 139 Figure 3.21. Marginal Effect of Repression on the Probability of Coup Attempts and Civil Conflict Onsets 140 Figure 3.22. Marginal Effect of Years of Repression on the Probability of Coup Attempts and Civil Conflict Onsets. 141 CHAPTER 4 INTRODUCTION TO THE CASE STUDIES Introduction The previous chapter contained the statistical analyses of the three theoretical frameworks developed in Chapter 2. The following two chapters will contain in - depth case study analyses examining the micro - foundations of those frameworks in Nigeria and Burma. Although the statistical tests allow for findings that are typically generalizable across a large number of cases, one of the strengths of case studies is that they ar e able to examine processes in much greater detail, exploring nuances that may not be repeated in any other specific instance. To clarify, though, just because particular nuances may be unique to a particular case does not mean that there are not patterns that occur across cases. Nonetheless, the case studies provide an opportunity to zoom in on the micro - level processes and probe the interconnections between events and frameworks. In this chapter, I outline the logic of the case studies presented in the next two chapters. I provide a general discussion of the broad expectations of each of the three theoretical frameworks. I then identify the major components of each case study and the general pattern of presentation that they will follow. I conclude with a discussion of the key finding of the case studies that will be highlighted more thoroughly in the individual analysis chapters and in the conclusion, but which is also useful to consider at the beginning. 142 Case Selection I used three primary criteria f or selecting the cases under consideration in the next two chapters. First, I wanted cases where there was a co - occurrence of events. While this means the cases were not selected randomly, they do provide an excellent opportunity to explore the dynamic pro cesses that led to the occurrence of each event as well as being able to pinpoint the ways the frameworks, especially the constraints and trade - offs frameworks, interact with one another. Second, I selected cases that were not included in the samples for t he original statistical analysis. Given that the cases were not randomly selected, I wanted to ensure that I was not selecting outlier cases that were driving the statistical results without being in some way representative of the rest of the sample. Final ly, I selected cases where enough information was available to produce a qualitative assessment of what happened and the reasons why. Without sufficient documentary and historical evidence available, the in - depth case analyses required for this micro - level examination would have been impossible. Background/Brief Historical Lead - Up to the Period under Consideration Because both Burma and Nigeria were colonies shortly before the period of primary concern, I will provide a brief overview of the colonial era, including the sociopolitical challenges the colonial power was faced with and the policies it put in place. The general condition of the economy will be noted, and special focus will be paid to the state of political institutions as well as the nature of the social landscape (demographic features, etc.) at the time of independence for each country. For instance, various regions of both Nigeria and Burma were ruled separately from one another, and the forced combination later in the colonial era did not alw ays make for smooth political and societal functioning. Thus, even though the events of primary concern here all 143 occurred while Nigeria and Burma were independent states, the policies implemented and general sociopolitical environment during the colonial e ra profoundly shaped the early years of independence of each country. And it was how each state reacted to and handled these early conditions that shaped the nature and timing of political turmoil in later years. Broad Expectations of the Three Framework s Although the specific nuances of how each of the three theoretical frameworks fits the data of the case studies varies across cases, it is nonetheless possible to characterize the basic expectations of each framework. Independence of Events When think ing about the independence of events framework, it may be more instructive, especially in the first years of each case study, to think of the independence as referring to risks rather than necessarily to actual event occurrences. For the independence of ev ents framework to have explanatory traction in the case studies, regardless of the timing and/or sequence of coup attempts and civil conflict onsets, we should find no tangible link between the risks of each phenomenon. Any risk or actual occurrence of one phenomenon should be unrelated to any risk or occurrence of the other. Admittedly, this framework faces a high initial evidentiary burden, as both the previous chapters and Roessler (2011) provide evidence and strong arguments that there are indeed link s that exist across such risks and events within a given country. The specific case studies chosen for further exploration here further heighten the burden, since the occurrence of both types of events 144 in each country prima facie supports the possibility o f a connection between coup attempts and civil conflict onsets. Constraints The constraints framework broadly hypothesizes that the risks and occurrences of coup attempts and civil conflict onsets are not independent of one another, and that the links be tween them are low state capability and a challenging geopolitical environment and climate. In order for this framework to have explanatory power in Nigeria and Burma, we should see economic and sociopolitical hardships in each state. While GDP per capita is the traditional cross - national measurement of state capability, much more nuanced and idiosyncratic indicators become available at the level of the individual case. And these indicators need not be identical across cases; because the exact political, e conomic and social configurations differ between the two cases, it is both plausible and likely that the functioning and level of state capacity will be revealed in somewhat different ways. For instance, while ethnic diversity and weak political institutio ns created more complicated environments within which incumbents operated in each case, there were nuanced variations in those indicators. In Nigeria, ethnic diversity is most aptly described as three primary ethnic groups, one of which is a numerical majo rity across the country, with several small ethnic minorities. In Burma, on the other hand, there is one overwhelmingly dominant ethnic group and several notable minority groups. And while both states began at independence as parliamentary democracies, Nig came from intra - organizational turmoil in the primary umbrella political organization, the Anti - s make both coup attempts and civil 145 conflict onsets more likely, and in their presence, we should not be surprised to see one or both phenomena actually occur. Trade - Offs The trade - offs framework, like the constraints framework, expects a systematic conn ection between coup attempts and civil conflict onsets, in this case in the form of problematic or heighten the risk of the other, making themselves vulnera ble to both. In Nigeria, for example, as an attempt to deal with the after - effects of a coup in January 1966 and counter - coup that July, the incumbent who gained power in July undertook policies designed to stabilize the political system. While he did so s ufficiently not to face another coup attempt, these stabilization policies had the consequence of further upsetting already - tenuous societal relations between two of the major ethnic groups and regions. This worsening of inter - ethnic relationships culminat ed in the Biafran Civil War of 1967 - 1970. In Burma, on the other hand, the incessant threat of insurgencies and uprisings by various rebel and ethnic groups created political tensions and divisions that lead to a complete breakdown of the parliamentary sys tem that had been in place since independence. As a coping strategy, the civilian regime voluntarily turned over power to a military - controlled - 1960. Power was returned to a civilian regime without incident, only for a resu rgence in rebel activity to again push the political system into petty infighting and subsequently broader turmoil, yielding a coup in 1962. 146 Narrative of Period under Consideration The historical narrative of the progression and deterioration of events forms one half of the core of the case study chapters. The historical narrative is divided into three primary sections: the lead - up to the first event, the interlude following the first event and the short - term lead - up to the second event. In Nigeria, the period of time from the occurrence of the first event (the January 1966 coup) to the occurrence of second event (the Biafran Civil War) was approximately 19 months. The process was much more protracted in Burma, taking more than a decade to completely unra vel from civil conflict onset in 1948 to a coup in 1962, including the peaceful regime changes in the interim. Despite the difference in the timeline, each series of events can be thought of as following this three - part sequence. Assessment of the Framew orks Following each section of historical narrative, including description of the colonial era, I engage in an analysis of the fit of each framework to the events during that time span, constituting the second part of the core of the chapter. I consider th e extent to which the data lend support to one framework over another, and what components of each time span could be interpreted as compatible with each framework. There are four such analysis sections corresponding to the historical narrative, as well as an overall assessment of the frameworks given the case in its totality. Nested Frameworks Probably the biggest insight from the case studies is that rather than being mutually exclusive frameworks, the constraints and trade - offs frameworks co - exist, the latter nested within the former. 147 attempt or civil conflict onset, it is improbable incumbents would have found themselves in such a precarious position without such an unfavorable environment. Indeed, the incumbents in power during the periods under consideration in the case studies found themselves in situations where one miscalculation destabilized the state; the overarching, pre - existing environment of econom ic and sociopolitical hardship ensured such fragility of the political system. Once a state faces serious constraints, then, we should expect that they are more prone to further political destabilization as a result of trade - offs. In these situations, the efforts of leaders to mitigate one risk may do more harm than good, not only failing to prevent that first risk from coming to fruition, but also putting themselves at heightened risk of the other phenomenon. This was the case in both Nigeria and Burma, a nd it is not unreasonable to believe that similar dynamics unfold in other cases where states are already economically, socially and politically vulnerable. Conclusion The case study analyses in the next two chapters will explore in greater detail the micro - processes tying together the risk and occurrence of coup attempts and civil conflict onsets. Although the specific details of each case vary, both case studies are comprised of four sections of historical narrative, four accompanying assessments of t he three theoretical frameworks and one overall assessment of the frameworks based on the case in its totality. Additionally, the underlying conclusions of the two cases are the same: the trade - offs framework is nested within the constraints framework, and both shed light on the difficulties facing incumbents confronted with the risks of coup attempts and/or civil conflict onsets, since the risk of one can ultimately produce the risk and even the occurrence of the other. 148 CHAPTER 5 NIGERIA, 1966 - 1967 In troduction This chapter examines in - depth the series of events in Nigeria that led to an initial coup in January 1966, a counter - coup in July of that year and the slide to a civil war which began in July 1967. Four historical periods are outlined: the col onial period and early years of independence, the deterioration leading to the January 1966 coup, the period from the January coup to shortly before the civil war and finally the run - up to the civil war. Each historical period is analyzed with respect to t he three frameworks discussed in Chapter 2. After these narrower analyses, the chapter concludes with an overall assessment of the performance of each framework. Colonial Period and First Years of Independence Historical Background Although hailed at in overt strife. It is comprised of three larger ethnic groups the Hausa - Fulani in the North, the Yoruba in the West and the Ibo in the East accounting for about 2/3 of the population and hundreds of smaller ones, constituting the final 1/3 of the population (Diamond 1988). With each group and each region having experienced, at a minimum, a sligh tly different history, Nigeria epitomizes diversity on a variety of dimensions. Most important of these dimensions are those concerning the structure of society and government, an issue that Nigerians have struggled with 149 since the colonial era (Kirk - Greene 1997; Nigeria 1965 ; Olugbade 1992; Tamuno 1970; Ukpabi 1989). Britain first established a serious foothold in what it now Nigeria in 1861 with the annexation of Lagos, but it was not until 1914 that Nigeria came to encompass its current territory, with t he amalgamation of the Protectorates of Northern Nigeria and Southern Nigeria. The description as an amalgamation rather than a unification is apt (Ukpabi 1989), for Northern and Southern Nigeria had been administered quite differently, and the legacy of t hose differences arguably even into the present. 50 The biggest cleavage in colonial Nigeria existed between the Northern and Southern territories (Amuwo 199 2; Anglin 1965; Nigeria 1965 ; Siollun 2009; Tamuno 1970; Ukpabi 1989), and perhaps the biggest distinction between them was the extent to which the ethnic groups in the territories adopted many of the sociopolitical customs (language, religion, educational and economic structures, etc.) of the colonial power. The Northern part of the country, inhabited primarily by the Hausa - Fulani, actively resisted such assimilation, and thus trailed behind the Southern part of the country populated primarily by the Yor uba and the Ibo on most development indicators leading up to and after Nigeria gained independence in 1960. Initial recommendations of quotas favoring the North with a plurality of seats in the legislature 51 were rejected by the North, which insisted on at least 50% of the seats, lest it secede from greater Nigeria (Amuwo 1992; Anglin 1965; Oyeweso 1992; Siollun 2009; Tamuno 1989). 50 50 50 Boko Haram, for i nstance, operates much more actively in the northern, primarily Muslim, portion of the country, targeting those Muslims deemed to be insufficiently conservative. The southern portion of the country, on the other hand, is predominantly Christian. 51 Under t his plan, the East and West (the Mid - West did not yet exist) would each receive a smaller, equal number of seats. 150 In the end, the North did not secede, and had a degree of representation greater than the other two regions combined. Souther n politicians rejected the 1952/1953 census figures that this division of representation was based on, as they were convinced the British had inflated population figures in the North (Anglin 1965). Attempts to conduct a new census in 1962 and again in 1963 were fraught with similar controversies (Anglin 1965), though the results of the latter were eventually accepted. the large number of Southern civil servants di spatched to the North because so few Northerners met the education and training requirements for the Nigerian Civil Service (Jinadu 1985; Nigeria 1965 ). Indeed, the North had been trying to combat the development gap for quite some time, shifting the prev ailing anti - education and training was now not a threat to emirate class structure but a condition essential for northerners would be (Diamond 1983, 473). Such decisions proved hard on the political institutions at the time, and the decision to privilege qualified northerne rs over similarly qualified southerners marked a break in the government structure at the time, indicating a need for a more flexible structure. With all this contention between regions, care had to be taken to establish a political system in which all R egions could coexist, even if they would never fully cooperate. In 1954, all parties ethnic groups could stay together within a geographical entity called Nigeri - 297). 151 Nonetheless, as already suggested above, these debates o ver what form the post - independence government should take were frequently contentious, with the regions frequently raising the possibility of opting out of the Nigerian system through secession. The North had threatened secession over the question of rep resentation in 1950. The West, during the 1954 Resumed Conference on the Nigerian Constitution advocated for the right of secession to be included in the constitution. Even within the federal, noncentralized structure, itical organization of ethnic minorities gathered momentum in the twilight of colonial rule, spurred by mounting apprehension of political repression, socioeconomic discrimination, and even cultural extinction by the majority groups when they took unfetter 474). Had these groups gained their sought - after concessions, the Nigerian political landscape would have been significantly different from the form it took (Diamond 1983). Of course, r adically altering the structure of the Nigerian government was the whole point, since even under a federal system, minority groups within each region were largely at the mercy of the political whims of the majority groups. Advocacy for these separate state s never completely subsided, and the desire to shore up minority support for the national government played a role in the declaration of a twelve - state federation in 1967, a development that will be discussed in greater detail below. It was North - South c onflict, however, made starker by the wide gulf in the level of development, and centered on the degree of representation each Region (North, West, East and later Mid - West) would receive in the national - level House of Representatives that garnered the most attention in the years leading up to independence. But while this was the moment prominent division in the colonial era, intra - South tensions and divisions existed as well: within the Action Group, the most prominent political party in the Western Region (Madiebo 1980); between the 152 Yoruba and Ibo (Diamond 1983; Nigeria 1965 ); and between the Ibo and surrounding minority ethnic groups (Ukpabi 1989), tensions that will be discussed more in the next section. Assessment of the Frameworks Independence of Even ts Given the volume of past literature, discussed in more detail in Chapter 2, that highlights the importance of low state capacity as being a consistent factor in the occurrence of both events, there is reason to believe that Nigeria did face the risk of both types of events in this historical period. In the first years surrounding independence, the divisions and cleavages between regions fostered broad social discontent. Chief Awolowo of the West presciently noted prior to not a nation. It is a mere geographical expression. The word Nigerian is merely a distinctive appellation to distinguish those who live within the boundaries of Nigeria t the lack of social cohesion within Nigerian society, and suggests the real possibility of upheaval along the dividing lines of society the Regions. With subnational units and territories garnering more loyalty than national identity, and with those sub national territories closely tied to ethnicity, differences and inequalities would come to be drawn along regional and ethnic lines. Although these regional and ethnic affiliations crept into the military, they were not the most serious issue facing the cohesion of the armed forces. Rather, the early stages of reliance on the military for sociopolitical stability created and repeatedly affirmed the proactive and interventionist role required of the military by the political leadership (Adekanye 1989; Inya ng 1989; Luckham 1971; Madiebo 1980; Otubanjo 1989). The initial barrier separating the military from society had been broken and a culture of active interference in, rather than protection of, the 153 home front had been introduced. The threshold for future m ilitary involvement had been lowered, and continued to be lowered with each repeated use of the military in such situations, until no real division between the military and the civilian elites tasked with maintaining law and order existed, which would make efficiency. The increasing politicization of the military by elites, then, through multiple deployments to contain civil unrest, alienated both the military, which was being us ed for purposes outside its normal purview, and the public, which was being confronted with multiple occasions when the military was repressor rather than protector. Based on its role in keeping the civilian leaders in power and the country in a state of ( fragile) peace, it is clear that the potential for civil war and the durability of the regime were inextricably linked together by the military. Given this, there is thus ematically unrelated risks, and, in fact, strong evidence against it. Constraints At independence, and the years following it, Nigerian elites were faced with very challenging social, economic and political landscapes. The biggest of these challenges wer e the cleavages between regional and ethnic groups and the ripple effect of this fractionalization. The ability of the North to dominate the national political scene, discussed in further detail below, caused discord in the West and East, which were signif icantly more developed. The educational and economic dominance of the South, then, was incongruous with the political reality, which created tensions in both portions of the country: the South believed it should have political power to match its prestige i n other areas of national life, and the North believed it should receive South - subsidized 154 assistance in its effort to catch up to the South in these areas ( Nigeria 1965 ). These social and economic inequalities between regions had major implications for the level of political inequality in Nigeria, both between North and South and within East and West between the majority and minority ethnic groups. Perhaps the biggest consequence and difficulty for political elites created by regional and ethnic divisions for example, Diamond 1988). These were particularly affected by the composition of the winning coalition. Because of the apportionment of representation and the strength of the North in national East or West would be helpful in presenting a united front, neither en bloc was absolutely necessary for any incumbent, and both toge ther would still be insufficient without the support of the North. This simple political reality led the East and West to believe that the national - level government was more or less a puppet of the North, and thus did not strongly support it, instead fav oring strong and fairly autonomous regional governments. And the North, although it did have this political dominance, was perpetually wary of the South, worried it would somehow wrest away control of the government, in addition to its continued advantage in the realms of education/training and the economy. Ethnic minorities, too, had qualms with the current governing structure, as they felt they had been abandoned and left to struggle to have their voices heard by the majority ethnic group in each region. With all these rivalries, national politics, and especially any policies impacting budgetary allocations, were quite contentious. However, disputes were less about categories of spending and more about the proposed recipients. Consensus might have exist ed about the need to spend on both the military and on more publically - oriented policies, but not about what portions of the public 155 should be privileged by certain spending. In this case, spending that benefitted one region or group more than the others wa s typically viewed as essentially providing private goods, rather than public goods. Just as with the political structures, great care had to be taken with any budgetary allocations. While spending more on the North would keep the North content and loyal, it could also encourage rebel groups to form and take up arms against in the state in other regions. Ethnic diversity very much permeated all realms of Nigerian society and politics, and though it was not an especially prosperous country, it was non - budget ary concerns that ultimately trumped concern over economic allocations to the military and various areas of public policy. Distribution of money was certainly a concern, but not as big of one as the distribution of power across regions and groups. It was t his distribution of power that made civil conflict more likely, and also facilitated the use of the military for domestic purposes, making a coup attempt more likely as well. In this first period, the constraints model performs well: Nigeria faced a wide variety of challenges, especially in terms of demography and political institutionalization. These challenges all served to narrow the range of issues leaders could realistically focus on (namely allocation of political power and development efforts), as well as limit their options in how to deal with those North were at risk of being rejected immediately). Trade - Offs An analysis of the underlying risks of coups and civil conflict allows us to consider how the decisions leaders made served to mitigate or bolster one risk over another. As the colonial era wound to a close and in the first years of independence, it was clear that national - level elites were con cerned with social cohesion and the peaceful interaction between ethnic groups and regions. 156 The numerous compromises on issues such as representation and structure of the national government as a federal system are a testament to this concern. While it is difficult to know if coup attempt risk or civil conflict risk was being impacted more, there is support for the argument that the actions taken by the national - level government only served to make the social and political situation worse, as the leadershi p of the First Civilian Regime adopted a policy of benign neglect or even non - - decisions turn out to be decisions for the consequences for the system. One unsavoury consequence in this case was that narrow partisan and sectional interests exploited the absence of a positive national policy to sow seeds of discord and disintegration. (Onyeoziri 1989, 37). This discord and disintegration certainly affect ed popular politics and the social landscape. During this period of time, clear evidence for the trade - offs model has yet to emerge, as the extent to which the military was affected by the decisions of the civilian leadership did not begin to manifest itse lf until a few years later, a process discussed more in the next section. At the beginning various parts of the country was not just a possibility, but a realit y that had already emerged on more than one occasion. Deterioration Leading to the January 1966 Coup Historical Background The North had broader disunity issues, namely the outbreak of protests over time by the Tiv, one of the minority ethnic groups in t he Middle Belt region of the North. Even before threatening secession in 1965, the Tiv had had a history of unrest: protests had erupted in Tivland in 1929 and 1960, and when they recurred in 1964, they prompted a more general crisis. The Nigeria Police Fo rce alone was insufficient to bring the area under control, and portions of the Army eventually 157 another symptom of the increasing politicization of the Armed Forces by the government (Adekanye 1989; Luckham 1971). In the South, the pr imary divisions were between within the Western Region, between the Yoruba and Ibo and between the Ibo and the ethnic minorities residing in the East. The former occurred primarily in the pre - independence era, but profoundly shaped the nationalist movement economic and political dominance of the Yoruba by the rapid rise of the Ibos along these dimensions], the Yoruba elite organized politically and culturally in defense of their privileged - colonial wars betw een [Ibo] groups and their neighbors and the feeling of insecurity which could only be assuaged through their being given their own separate 9, 111). This insecurity contributed to ongoing tension between the Ibo and ethnic minorities and also bred discontent among the ethnic minorities, who were not granted any kind of separate territories until very shortly prior to the onset of the civil war in 1967. In the West, the splintering of the Action Group and friction between its two most powerful members culminated in 1962 when the elected leader of the Region, Chief Akintola, was ejected from the party, accused of conspiring with Northern politic ians against the good of the West 158 dissolution of the legislature of the Western Region, the disbandment of its executive, and the (Otubanjo 1989, 50). The expulsion and its fallout led to the disempowerment of the Action Grou p and its leader, Chief Awolowo, creating upheaval by members of the party when it was marginalized in the 1964 federal election 52 and the 1965 Western Regional elections. The Action Group had already been pushed out of the coalition at the national - level f ollowing the 1959 federal elections by the parties from the North and East, respectively. During its period of relegation, the Action Group made sure to oppose all substantive legislative output by the national government and the NPC - NCNC coalition (Otubanjo 1989). The conduct of the Regional elections in 1965 were hotly contested, with the Action Group hoping to regain the power lost in 1962 and voters generally hop ing to replace Chief Akintola, who had grown increasingly unpopular in the interim (Dare 1989). For most of the post - the final straw that - election may have been a significant milestone in the demise of the First Republic, it would be another few months before it officially came t o an end. The use of the Nigeria Police Force and Army to regain control over situations of unrest became a perennial feature of the Nigerian political landscape, occurring several times between 52 The legitimacy of the 1964 federal election was more generally contested, with the leader of the East advocating 159 the gaining of independence and the January coup: in the Wes t twice, in 1962 and again in 1965; in the wake of two disputed rounds of census results (1962 - 1963 and 1963 - 1964); in the Mid - West region shortly after its creation as ethnic groups jockeyed for control; in the midst of a general strike in 1964 and confli ct over the awarding of a key position at the University of Lagos in 1965; in the national - level general elections of 1964 - 1965; and in riots in Tivland, which had occurred on and off since 1960, but were especially intense after 1964 (Adekanye 1989, 191 - 1 92). This persistent reliance on the military is argued to have muddled the societal role of the military in the minds of both society and the members of the armed forces (Luckham 1971), making military interventionism increasingly the norm, rather than th e exception. In early January 1966, the government began picking up rumors of a planned uprising in the West for the latter part of the month and possible plans by the leaders of the Northern and Western Regions to lead some kind of government takeover. The compounding of the prior litany of military interventions and this new round of speculation of unrest resulted in a coup staged by a number of Majors of Eastern origin on January 15 - whether anything less t han the complete seizure of power would have enabled [the highest - ranking military officer in the country, Major General] Ironsi to regain control of the army. The revolt left ham 1971, 25). Enlisted soldiers in the North were disinclined to obey their commanding officers, taking hours if not days for more senior officers to eventually regain control over those troops. The problem of insubordination was not limited to enlisted - ranking officers who, though they knuckled under to threats or calls to discipline from Major General Ironsi after he had shown his hand, were said to have been sympathetic to the political uckham 1971, 26). Once Ironsi had announced, at midnight on January 160 16, 1966, that he had assumed control of the country, the participating and sympathetic officers were reined in and deferred to General Ironsi. While the timing of the coup coincided with renewed potential for unrest, the plot itself Southern i ntellectuals, tended to attribute all of the political shortcomings of the regime to the Not just the 1964 and 1965 election crises, then, but also the way in which the national - level government responded (declaring a state of emergency in the West in 1962 after a relatively - narrow incident of violence in the legislature but not in the face of widespread rioting and protests after the 1965 Regional election). Assessment of the Frameworks Independence of Events The Independence of Events framework continues to perform poorly in this second time. If anything, with the passage of more time, the framework falters even more. As time goes by, it becomes increasingly clear that the coup was a reaction to the deteriorating political situation, which was itself partially a function of continued competition between the regions. Leading up to the January 1966 coup, it was obvious that Nigeria continued to face social and political unrest in a variety of locations and for a variety of reasons. The North - South cleavage continued to exist and was joined by a splintering in the West in which opposing political factions seemed incapable even of mere coexistence without inciden t. As a consequence, there seemed to be an increased risk of more general civil conflict onset stemming from this tension and unrest. 161 The leadership was also coming to rely heavily on the military, repeatedly drawing them into strictly domestic situation s to impose a return to law and order. Given the problems very politicized militaries can create in states, from increased intervention to a split in civilian elites (Finer [1962] 2009; Huntington 1965; Janowitz 1964), the coup risk also seems to have been increasing during this period of time, as the military became the pillar the regime relied on for its survival. Constraints The constraints Nigeria faced at independence only grew more problematic as they were left unchecked and in some cases, as di scussed in more detail below, exacerbated in the years following independence. As previously described and discussed, the regional and ethnic cleavages in Nigeria, which served to reinforce one another, posed a considerable challenge for all governmental entities. Tensions between groups threatened not only the delicate balance between regions, but also raised the potential of unrest within regions as minority groups tried to break free of control by the majority ethnic groups (as the unrest in Tivland il lustrates). Moreover, the majority ethnic groups could not always remain united among themselves (highlighted by the splintering of the Action Group). The conflicts also manifested themselves in nonviolent ways; the refusal of the Southern regions to accep t various rounds of the census, for instance, indicated a refusal to work within the existing political system. Leading up to the coup, Nigerian politics operated in many ways in lines with the predictions and expectations of Bueno de Mesquita et al.'s ( 2003) selectorate theory. Northerners knew they were safely in the winning coalition so long as the North retained its privileged political position vis - à - vis representation. However, a shift in the power dynamics would almost certainly 162 decimate the influe nce of at least the Hausa - Fulani, one of the big three ethnic groups, given Southern resentment of the North. Although it was not necessary by the numbers since the North ght additional support outside the North, creating competition between the primary Eastern and Western political parties for inclusion in the last few national governing spots. The split in the Action Group also fostered loyalty in the faction currently in cluded in the government, though, since a change in the political fortunes of the excluded faction would likely block the former faction from remaining in office. At least some cooperation with the East and Northern elements would be necessary to maintain a majority hold on power. Nonetheless, aligning with these select Northern elements (possibly the oft - ignored minorities) would not be the same as aligning with Western collaborators who conspired with the North to disempower fellow Westerners. All membe rs of the winning coalition, then, had an incentive to remain loyal, especially given the quasi - private public goods allocations on the basis of region. A national newspaper ight that more federal Northerners are being reasonable in insisting that for some time to come they should receive some preference in federal public service Nigeria 1965 , 2). Noble though this view may have been, it was shared by neither the elites nor the public at large in the southern regions. Despite a fairly stable national winning coalition and the resolution of individual crises, bigger issues always simmered near the surface: Northern domination, Yoruba disunity and Eastern aggressiveness were all highlighted; as were bitter personal animosities, the ugly scramble for jobs, the 163 resort to force and fraud, the glaring inequalities of opportunities and the growing economic disparities which remain to plague this troubled land. (Anglin 1965, 1 87). Accompanying this regional and ethnic cleavage were increasingly weak political institutions, in terms of both stability and legitimacy. After the 1962 split in the Action Group and the 1965 Regional elections, the government in the West had all but completely broken down in what was Nigeria 1965 , 44). Although all institutions face the potential for breakdown, the possibility that fragile institutions will suffer such a fate is m uch more likely (Diamond 1999; Huntington [1968] 2006). Given the various sociopolitical tensions in the country, even as fragile and in need of reform as those institutions were, no guarantee existed that any alternative set of institutions would be more amenable to all the relevant factions. This left Nigeria with a weak government and no concrete possibility of a more stable one; no group in the winning coalition was willing to give up the certainty of inclusion for the slight change that they would cont inue to be included following any substantive reforms to the system. Trade - Offs Leading up to the January 1966 coup, it was clear that leaders were aware of the regional and ethnic cleavages that divided Nigeria, evidenced by the repeated use of the mil itary to quell episodes of social unrest. Conflict within broader society, which perhaps would be more difficult to contain, even with the military, was certainly a concern. The use of the military and perhaps more importantly, the inconsistent use of th e military only served to foster more problems. For instance, the flare - up between the two factions of the Action Group led to the declaration of a state of emergency and the dispatching o federal military troops to the West in 1962, while the widespread riots after the 1965 Regional election festered for significantly longer before any 164 national - level action was taken. Although large number of troops were eventually sent in and came to essentially occupy the West,that action was not nearly as quickly fort hcoming as in 1962, and the rationale behind the delay was not made clear (Madiebo 1980). While each use of the military may have succeeded in restoring law and order to whichever affected area, the continued use of the military primarily for that purpose , coupled with the fact that it was not always called in, only served to further politicize the military and create and confirm a domestically - interventionist mission. So while there may have been relative societal peace as a result of this military action , it also became increasingly clear to the members of the armed forces ere equally sufficient number of officers decided that the existing political order was no longer acceptable, there was little political infrastructure that coul d have withstood their desertion; when they abandoned their loyalty to that order, there were no non - military forces strong enough to prop it up. As the regions continued to be the locus of many political decisions in Nigeria, the role of the Federal g overnment became more contentious. Attempts by the national and regional governments to deal with some of the challenges facing Nigeria at independence in many cases simply further exacerbated the underlying issues. Often the elites at the various levels o f government tended to focus on strategies of exclusion and personal/group political structural frame, if the political actors had been more accommodating, it would still have been 165 (Dare 1989, 121). And as the southern regions, particularly the East, gained stronger autonomy, anything the national government did counter t o its wishes only served to further confirm its view that the center was dominated by the North (which was viewed as concerned only with its own interests, rather than a broader national interest). As a result, actions taken by the national government were inherently divisive, continually pitting region against region and calling into question the stability of the federation. With the same policy producing both strong support and strong opposition, national political actions and decisions were doomed to be contentious. With not even public goods being viewed as benefitting Nigerians across the whole country, leaders were not doing much to dampen the risk of civil conflict, but the fact that the dysfunctional status quo largely remained in place, the overall risk may not have been increasing as much as the risk of coup brought on by the frequent use of the military in domestic settings by the politicians and their socialization as an interventionist force. Indeed, in light of the state of Nigerian society an d political structures, the January coup 1989, 51 - 52) . In addition to the non - security policies of the incumbent elites, the increasing reliance on the military to regain social control did have a marked influence on the interventionist proclivities of the military. Previously considered a coherent instituti on within the Nigerian state (Luckham 1971), the reliance on the military in times of heightened unrest continued to increase the chances that at some point the army would decide that actually being the government rather than just its lynchpin would be mor e efficient. Strong evidence for the trade - offs framework begins to emerge surrounding the January coup. The price of years of heavy reliance by the government on the military for survival in the 166 face of societal unrest revealed itself in January 1966. E very time the government needed the military to restore law and order in the face of disturbances in the regions, the military became a little more politicized and a little less removed from the actual governance of the country. While full - blown civil conf lict may have been avoided as a result, it became apparent that the military was the true governing apparatus. The coup simply removed the last façade of civilian control of the government. Without this use of the military by the civilians, though, the mil itary would have likely remained far less politically active. In trying to keep the peace internally by dispatching the military, then, the civilian government sowed the seeds of its own downfall one deployment at a time. Fallout from the January 1966 Co up and the Slide to Civil War Historical Background been preventing the ordinary awarding of promotions, leading to a backlog, and Eastern and Mid - Western o fficers (both predominantly Ibo) were worried that their own career advancements would be interfered with by Northern politicians, contributing to hostility on the part of these soldiers. With the advent of the coup in the face of these political developme nts, the army, which had long been relied on to keep the population in order and the Federation thus united, was now divided against itself, Northern troops resisting the command of officers of Eastern origin. During the staging of the coup, the Federal Pr ime Minister, Tafawa Balewa, had been killed, as had the premiers of the Northern Region and the Western Region, the Sardauna of Sokoto and Chief Akintola. Conversely, no prominent Easterners had been attacked, leading to speculation that the coup was the result of a broader Ibo conspiracy. 167 Those in the North, especially, were inclined to think that the coup was a specifically Ibo - driven conspiracy against the rest of the Federation. The views that the coup was aimed at destroying the political domination of the North (Luckham 1971) and that it was an Ibo conspiracy are not inherently compatible with one another. The South more broadly construed opposed Northern domination, and the desire to weaken its hold on the political system of the entire country was not just an Ibo sentiment. The fact that most of the officers involved were Ibo also contributed to this view, but most officers in the Nigerian military were Ibo (in 1961, 75% of officers were Ibo [Peters 1997]), a function, in part, of the development ga distinguish between the roles played on 15 January and afterwards by Major General Ironsi and - particip ants on the one hand, and the Majors on the other, it is understandable that the Northern Officers and NCOs tended to assimilate the two in their definition of the - al with nationalist conspiracy, a calculated attempt to eliminate the top Northern military and political class. For apart from the preponderance of Igbo officers i n the planning and execution in the coup, Although Ironsi did ultimately install himself as the Head of the Federal Military Governemnt of Nigeria on January 16, 19 66, and seize control over the rest of the country by asked tw o days after the coup by a journalist whether he would remain in charge of the government 168 about aggrandizement of his own position, but rather an attempt to stop the further spread of did not show favouritism towards his own tribe [the Ibo] in appointments and promotions; and indeed surrounded himself with Northern aide s. Northern ministers present at the first meetings of 15 January vouchsafed that Ironsi was in tears and appeared genuinely upset by the whole affair. Thus the most convincing reason for his seizure of power on the 16 th must still be his organisational lo of the situation in which he found himself, with only two exceptions, all major army p ositions were held by Ibos for the first month after the coup. Even once Ironsi had appointed some Northerners to important roles, top positions were still predominantly held by Ibo officers. Nonetheless, once the rebellious troops had been brought back un der control, Ironsi appealed to re of a politically moderate supreme commander, tenor and diminished the importance of tribalism and prevalence of corruption (Sklar 1967, 531) proved much more contentious and divisive. 169 The January Conspirators Another sourc e of confusion and tension between the military government and members of the armed forces and public at large from the North and West was the treatment of the Majors who had led and participated in the January coup. This confusion and tension was also bou nd up in views of an Ibo conspiracy as the driving force behind coup Those from the North, and lower - ranking soldiers, felt they should be harshly punished for their insurrection and the murder of leading national politicians. In the East, those Majors wer e considered by some to be heroes. At a minimum, general public opinion in the East favored the release of the Majors without having been administered much in the way of disciplinary action. The Ironsi regime balked on this matter, failing to do anything d This lack of political courage was evidenced by the fact that an investigation into the conspiracy behind the coup an d the officers responsible for it took more than four months to complete, and for reasons that are not clear (Luckham 1971), nothing was ever done with the findings of the investigation. 53 Additionally, one of the primary conspirators, Major Nzeogwu, public ly announced that upon his surrender to Major General Ironsi, he received assurances that none of the conspirators would be formally punished for their participation. The disparate views of the conspirators in the Nigerian public, as both heroes and traito rs, made any decision concerning the consequences all the more complicated for the military government. The stories and rumors that circulated about the deaths of the Prime Minister and Northern and Western premiers, and the lack of any public statement by the military government, did not help either. 53 The officer in charge of the investigation, Lt. Colonel Gowon, would become the military leader following the July counter - coup 170 The decision on the part of the military government that the conspirators would be dealt development of suspicio ns in the North that the government was at one with the conspirators; that Ibo blood was thicker than water between Ironsi and the Majors and that the North had been unduly ver, it is also unclear to what extent the Majors were actually dealt with through normal military channels while they were not punished, neither were they released at any point in the short - term, still in detention in May. Ironsi and the Unification D ecree Probably the most consequential policy implemented by the Ironsi regime was Decree No. 34, which abolished federalism and replaced it with a unitary system. A commission to investigate the feasibility of a unitary system was launched in mid - February, the Decree itself was announced in late May, and officially implemented on June 2. By removing the Regions, along which political considered to be one of the majo r causes of instability and insecurity in the Balewa years: decree a bolishing a number of political organizations, most notably political parties, and announcements in June and July of further reforms to shore up the centralization of power. terns of communications and control within both the military and the administration during the weeks 171 Even though there was nominally an enormous change in the gover nment of Nigeria, much of the functioning remained the same, and the change from a federal system to a unitary system - Brick 1970, 15). For example, the word structure was left to function much as before. Regional authorities were not abolished but replaced by Military Governors who inherited their legislature and executive powers in a modified ral authority retained a quasi - federal character and the regional administrators - Brick 1970, 15 - 16). This provided some continuity for the people of Nigeria, who w ere confronted with a number of other major changes in the political landscape, from the almost immediate loss of Northern supremacy to life under a military dictatorship. Despite such continuity, however, vings all over the country and, consequently, May 1966 Anti - Ibo Riots Following the announcement of the Unification Decree, riots broke out both among students and in Northern trust within the military, with Ibo officers suspecting the deliberate involvement of Northern officers, leading the former to desire a tighter grip over the national government. The violence against Ibos in the North also facilitated the rise of Lieutena nt Colonel Ojukwu, military governor of the Eastern 172 genocide a - 10). - for greater Ibo rights under Gowon and eventually guided the East to se cession and civil war, discussed more below (Tamuno 1989, 9 - 10). July 1966 Counter - Coup After the large changes put in place by the unification decree and the subsequent protests in the North, Ironsi recognized the importance of selling the country on t he benefits and strengths of the unitary system. To that end, he had embarked on a nation - wide tour to extoll the virtues of the Decree and follow - on policies. Ironsi had further announced in mid - July 1966 several new policies that would result in greater centralization of the government. In addition to this announcement, which had only served to rattle the North even more, Northern officers were also hearing rumors of shake - ups sometime in August, including the replacement of Ironsi and the removal of many adverse reactions to it in Northern Nigeria, as well as suspicions concerning his role and that of other Igbo officers in the January 1966 , resulted in his own assassination soon after he and [Lt Colonel] Fajuyi were kidnapped by soldiers of northern origin, in Ibadan [in the Western soldiers staged a successful counter - coup. Although the North had successfully regained control over the central government, in the days and weeks immediately following the counter - coup, popular opinion in the North and especially among soldiers and officers favored the secession of the North from the rest of Nigeria. 173 The North also favored Lieutenant Colonel Gowon, the highest - ranking officer of Northern origin, as the new head of state, which would have involved skipping over several higher - ranking (primarily Ibo) officers. It was clear, however, that without such a concession, the Northern troops still in rebellion would be unwilling to halt their mutiny and insist instead on full secession from Nigeria (Elaigwu 1986; Luckham 1971). However, even after Gowon had quelled the mutiny in the North and declared himself the Supreme Commander and Head of the Military Government, Lieutenant Ojukwu of the East never acknowledged the legitimacy of Gowon to hold that position, heightening tensions between North and East. In the days followi ng the restoration of order after the July counter - coup, the Northern troops who had previously advocated for secession began to advocate instead for the return to a strong Federation. Indeed, the swathe cut through Revenge was certainly one of the motives. Separation from the South, in the early stages, was another, but this changed as tempers cooled and senior Northe rn officers reasserted their control over the mutinous rank and file. It then became evident that if the North was not going to secede, it was determined to fill the power vacuum in Lagos and the empty slot at the head of the army. (De St. Jorre 1972, 75 - 7 6) In further attempts to consolidate his power, Gowon soon announced the abrogation of the Unitary Decree and promised a phased return to civilian rule. He also released a number of political prisoners, although the decision to do so had been taken under Ironsi, just not implemented. Public and political debate about the future of Nigeria began, and 174 proliferated on all sides, and people began to question the continued existence of the Aside from the swing in the North from secession to a stronger Federation, the other sig and his government were transformed from probably the most criticised of all the new hifts and the recognition for a more deliberate debate on the future of Nigeria, Gowon alternatives for the future government of Nigeria: federalism is a strong national governmen new arrangement which may be peculiar to Nigeria and which has not yet found its way con flicting preferences for the structure of the national government. With the exception of the Mid - West, which favored a federal system with a strong national government, the ity ethnic group in each region without much regard for the minorities who also resided within them. September 1966 Anti - Ibo Riots Riots broke out around the North in September 1966. These were similar to the May riots, though Ibos were much more speci fically targeted in September. These riots coincided with the occurrence of the ad hoc Constitutional assembly, which was itself designed to encourage the Eastern Region to fully rejoin the social and political life of the country, halting its drift away f rom the center. 175 contingency plan one of several alternatives forces within the Eastern government used them in such a w ay that agreement in the deepening Assessment of the Frameworks Independence of Events The January coup had been a reaction to the deterioration of the sociopolitical situation after independence, and the unification decree had been an attempt to remove what Ironsi saw as the major stumbling blocks to social and political cooperation. Likewise, the July counter - coup was a reaction to the unification decree. Coup begot counter - coup, then, a claim which is not all that contentious, given past research on the vulnerability of regimes to future coup attempts after a successful coup has been carried out (see, for example, Belkin and Schofer 2003, 2005; Jackman 1978). Howeve r, the military was not the only societal group bothered by post - independence events and the turn the country took after January 1966. Other components of the winning coalition, especially in the North, were also concerned. After the July counter - coup, the East, power, was more deeply troubled by events than the North. In this case, too, though, the impact was broader than just the military or civilian portions of t he winning coalition. The public and armed forces alike were being affected by the sociopolitical policies being implemented by elites. Members of the public, namely members of the various minority groups that had long been marginalized, and Easterners, wh o found themselves increasingly disregarded after the July countercoup, had reason to try to change their circumstances. Likewise, members of the military, 176 inte rest in pushing for reforms that would ensure increased existential security for the East and the other regions. Since both groups can be agents of change and had incentives to try to effect changes e the argument that the risk of civil conflict by minorities and Easterners would be truly independent of the risk of backlash by non - Northern members of the military. Constraints The development of a unified social culture has long proven difficult for Nigeria, both before and well after independence (Ukpabi 1989, 108). In light of the societal divisions that reinforced regional and ethnic acrimony and distrust between the winning coalition and the portion of the selectorate excluded from the winning coa lition, Nigeria faced particular hurdles in establishing sufficiently strong political and economic institutions and structures to keep Nigeria politically united even in the face of societal discord. That was, after all, the motivation and rationale for t he installation of the federal system leading up to independence, though even it was not resilient enough in the early days of the country. The coups only served to further complicate the already convoluted political landscape. It has even been claimed th political unity and put the country on the path to the thirty - The breakdown of political equilibrium keeping incumbents in power and society reasonably stable preceding the January coup and the inability of the Ironsi regime to patch it back together put in place the conditions that made the July counter - coup possible. Indeed, with a total loss of legitimacy in the existing political institutio ns both those in place before the January coup and 177 meaningful action taken within that governmental structure became virtually impossible. The winning coalition did not have much support of any part of society excluded from it, but these excluded actors favored trying to press for some change as strongly as possible, since a change to the status quo might carry with it large favorable changes to their political fortunes, while a continuation of the stat us quo would only foster further discord (see, for example, Omotola 2010). There is also speculation that the anti - Ibo riots in the North following the July counter - coup were supported by foreign powers (Balogun 1973), making the task of maintaining inte rnal stability all the more difficult. 54 And while certain government actions may have contributed to the outbreak of more riots and protests, once they began, they simply provided yet another hurdle the Ironsi and later the Gowon regimes would have to over come in order to regain legitimacy at political problems remained intractable. These included the unequal size of the regions making up the Federation, the politic al aspirations of the minorities and the healing of the wounds brought the basic problems facing Nigeria remained the same. Compounding these problems further , the July counter - coup, even more so than the January coup, decimated the Nigerian military. In addition to the tensions arising between troops disorganization b y the sudden withdrawal of nearly all Ibo and a very high percentage of other officers of Eastern origin...While the majority of rank and file were of Middle - Belt and Northern 54 Nigeria and Ghana had ongoing diplomatic tensions during this time perio d, and Nigeria sought the assistance of 178 eated by the - 80). soldiers alike. Those who felt overlooked in the regional battle, namely the minority groups, tried to draw attention to their plight in other ways, often unrest, while the military was not strong enough to withstand the infiltration of these social and political divisions into the armed forces. As a consequence, soldiers grew more restive as civilians and senior officers tried to manipulate the army to suit their own needs with seemingly little regard, from the collective perspective, for the good of the country as a whole. 55 Trade - Offs Ironsi faced a number of challenges when he assumed power in January 19 66. Following the coup came a growth in political conflict in the army because it became the focus of all political allocations, the military leaders having become the government. This contributed to the spread of indiscipline and to a loss of solidarity b etween military Western officers in the January coup and the foreshortening of the hierarchy of command that resulted; both tending to generalise revolt from the right knot o f conspiracy to a wider circle of mass rejection of authority at all lower levels of command. The result was the July counter - coup, an outburst of revolt against authority from below by junior officers and NCOs in an attempt to effect a transfer of politic al power but also as an outburst of hostility against a particular ethnic group, the Ibos. (Luckham 1971, 83). had been set aside, Ironsi was unable to really reg ain full control over the country after he assumed 55 It is clear that different groups had different views of what would be good for the country as a w hole, but one also gets the sense that many, especially the ethnic minorities, felt completely alienated from the political system and process, with little presumption that politicians had anything but parochial concerns at heart. 179 power. When coupled with the differences of opinion that existed between the various groups (majority vs. minority) and regions (namely East vs. North) on a wide variety of dimensions, any decision he took was likely to rankle some societal and military groups even if there were many others who supported it. The unification decree proved particularly divisive in this regard, and the by Northern officers to stage a coup and remove Ironsi from power. Indeed, rather than mitigating regionalism and tribalism, as the unification decree had been intended to do, its implementation has been called and friction which were ultimately to consume the Ironsi regime itself and threaten the survival of In the interregnum between January and July 1966, just as i n the lead - up to the January coup, the underlying risk factors favoring civil conflict onset were not addressed and reduced, but neither were they severely stoked. Much more obvious and problematic were tensions within the military, directed toward both th e civilian regime (which was perceived as incompetent) and senior (namely Ibo) officers. So while the civil conflict risk remained, it also remained fairly steady, while the risk of a coup attempt continued to rise. Anti - Ibo protests, especially in the N orth, following the unification decree and again in the months following the counter - coup had the effect of planting the seeds of suspicion among the Ibo that the North could not be trusted at all. This viewpoint was primarily the result of the deliberate targeting of Ibo in the region. Although the May riots were problematic, the September riots were particularly troubling. That Gowon did not take a fast and firm stand against them much like Ironsi had done in the case of the January conspirators only further confirmed to the Ibo their precarious situation in Nigeria and the kinds of potential threats they faced in the future. This 180 increased suspicion that the Gowon did not care about their wellbeing gave the Ibo pause, and reason to contemplate other p olitical arrangements for the future so they could be assured of their safety as well as some political autonomy (see, for example, McKenna 1969). There was widespread mutinying and revolting by soldiers in the wake of the counter - coup, and it was not im mediately clear how these troops could be brought back into a state of semi - discipline. Strong leadership by many junior officers kept the mutinies from spreading and began to bring calm, but the efforts to keep Northern soldiers at bay alienated required concessions, namely his assumption of power, that alienated the broader Ibo public, though especially the de facto withdrawal from 1971, 79). But his unorthodox assumption of power was only the first such move on his part that pushed the Ibo f urther and further from a Nigerian identity and more entrenched in ethnic loyalty After the counter - n still faced risk from Eastern soldiers, but the military action involved in the counter - coup had a much more profound effect on the public in the East, which almost at once began to distance itself from the center in favor of a stronger Eastern identity with increased autonomy and decreased association of any kind with the North. The trade - offs model expects that decisions by the incumbent to mitigate one risk will have the consequence of increasing the other risk. By taking power, Gowon managed to effec tively eliminate the risk of continued (and perhaps increased) rebellion by Northern troops, and the risk of that rebellion spreading to the Northern public at large. However, this same action did not 181 diminish the risk of civil conflict altogether. Rather, it primarily shifted it from North to East; his becoming leader angered Eastern military contingents, and his lack of response to anti - Ibo rioting raised fear and anger among the broader population of the East. It was Lt. Colonel Ojukwu who basically ensu red there would be no follow - on coup by the East (but with the overall consequence that Gowon was able to consolidate his hold on power without strong fears of reprisal by the military). However, Ojukwu simultaneously fomented secessionist support by essen tially breaking the Nigerian military along regional lines and stoking Ibo fears and concerns. In this period, although the relative risks Gowon faced did shift from coup to civil conflict, the actions of the military governor of the East was perhaps equal ly, if not more, responsible for this shift, as he garnered the loyalty of those under his command and the public in the East. Thus, his troops listened when he chose insulation and preparations for secession as an alternative to a reprisal coup as a respo - Ibo sentiment and actions afterwards. Final Descent to Civil War Historical Background In the midst of the Ad Hoc Constitutional Conference and the strengthening preferences of each regi found himself faced with the increasingly difficult task of trying to keep Nigeria united as a single country, but also ensuring readiness in case the East did secede. Th e skipping of the normal chain of command required for Gowon to take control of the government had outraged the Military Governor of the East, Lt. Colonel Ojukwu, who in turn supported a breakup of the country more strongly than the idea of accepting Gowon as leader of the country. Despite this, Ojukwu and Gowon did try to negotiate a resolution to the crisis, in the form of the Aburi Accords. The Aburi 182 Accords, concluded in January 1967, were aimed at de - escalating relations between the East and the center , and included the renunciation of the use of force in the conduct of relations between the two. Despite the formal agreement of the Aburi Accords, and the promulgation of Decree No. 8 on the part of the Federal government, debate over the implementation of Aburi and its contents continued. For nearly two months, the Accords existed on paper but not in practice. In late Accords if Gowon did not do so by March 31, 1 967 (De St. Jorre 1972, 102). This speech was 1972). On the heels of this public broadcast by Ojukwu, Gowon privately addressed a group of visiting African dignit aries that he would use force is necessary to keep the federation together, despite assurances in Aburi that force would not be used under any circumstances. As uncertainty about what, if any, form a united Nigeria would take in the future, minorities an d diaspora groups from the North began advocating for the creation of new states in September - West held the balance politically if not militarily. Neither Region had any vital interest in giving outright su pport to Lt - demand for a loose confederation of existing Regions or to Lt - - Brick 1970, 44). After months of vacillation and uncertainty concerning the is sue, on May 27, 1967, he announced both a state of emergency effective throughout the entire country and also the creation of 12 new states: six created out of the old Northern Region, three out of the East, part of the West carved out to become part of La gos State, West and Mid - had varying amounts of popular support throughout the history of late - colonial and independent 183 Nigeria. With the notable exception of the Ibo, the announcement was strongly supported by a number of groups in Nigeria, in part because in creating the new states, al - was an anti - climax. The Federal Government had expected it. Gowon had made up his mind that violence could no longer be avoided. (Elaigwu 1986, 104) The East made two major an nouncements of its own around this time. The same day, May 27, the Eastern Assembly passed a resolution demanding that Ojukwu declare an independent emergency. On May 30, 1967, the old Eastern Region formally seceding from the Federation of Nigeria as the Republic of Biafra. War, however, did not begin until July 6, 1967 with the Federal n the part of the Federal government to arrest Lt. Colonel Ojukwu, but quickly expanded from that narrow scope - West and march into the Western Region on August 9, 1967. On August 11, Gowon declared total war against the Republic of Biafra. Assessment of the Frameworks Independence of Events Although the independence of events framework has performed poorly in each of the historical time periods, it performs especially poorly here, as there continues to be strong eviden ce for the constraints model and clear evidence emerges supporting the trade - offs model, as will be discussed in further detail below. 184 It is obvious that Gowon did not have much maneuvering room upon his assumption of maneuvering space shrank further as the Ibo, led by Lieutenant Colonel Ojukwu, tried to regain as much autonomy as they could from the central government. These attempts and their general thwartin g by Gowon also began increasing the possibility of some kind of broader social conflict, almost certainly along regional lines. The circumstances under which Gowon found himself taking control of Nigeria, and the reaction of the East to the decisions and actions he took to try to preserve the unity (but not unitary system) of the country, then, lend themselves to the ready dismissal of the independence - of - events framework for understanding the final political deterioration in Nigeria and the outbreak of ci vil war. As time wore on, events not just risks did begin to piece themselves together in a more nuanced way, further negating the usefulness of the independence of events framework in understanding events in Nigeria in the lead - up to the Biafran war. Constraints As in previous historical periods, Nigeria faced the same types of constraints it had in the past, though the intensity of them continued to grow: as the Ibo grew more bellicose, many minorities in the East began to ally themselves more with the federal government, generating more discord in the East and raising the potential for a second conflict based around the region. This conflict would take the form of the East against the center, and the ethnic minorities against the Ibo. In the afterma th of the July counter - coup and the flip - flop between North and East in level of support for centralized versus federal government, political institutions broke down even further, especially in light of the refusal of at least some Eastern officers, most n otably the Military Governor of the East, to accept the leadership of Gowon as legitimate. 185 In the lead - up to the secession of the East, Nigeria found itself confronted with a new constraint. The evidenced willingness of foreign countries to sell weapons and other military technology and of foreign companies to negotiate contracts with the soon - to - be Biafrans indicates The fact that the Biafrans could count on o ther countries for at least some material support was a big problem for the national government. Continued rivalry along ethnic lines and on the basis of the old regions, coupled with tates, limited the chance that Gowon would be able to use policy to maneuver out of his tight position. The external support for the Biafran movement further compounded these difficulties, as the national government would by no means have an assured prepon derance of military strength in the approaching fight, raising the prospects of a Biafran victory in the impending conflict. Trade - offs The more Gowon tried to keep his options open in dealing with the East, the further he seemed to alienate Lt. Colonel Ojukwu and the Ibo, and the more he seemed to push them out of the cillation in the reaction to the massacres of Ibo and inconsistent upholding of the Aburi Accord by the (De St. Jorre 1972, 115). In light of this, 186 actions taken by the national government, first concerned with yet another coup and only later perhaps too late with the disintegration of the country. Shortly after Gowon named himself Supreme Commander of the Military Government, Ojukwu said that as a result of the events of the counter - belligerent and hostile approach to the South, including the favoring of force to bring it in line, he that we were heading for something terrible In such a situation, there seemed to be little hope that any future decisions by Gowon would do anything other than ratchet up the already - high tensions between the East and the central government, pushing the East closer to war. The carving up of the regions into new states, for instance, very deliberately disempowered the Ibo, and led to their almost immediate secession: the Ibo went from being in the majority in a however, the calculation of the costs and benefits of the creation of the states did favor the national government: the creation of states specifically for ethnic minorities in the East helped to win their loyalty, and although the creation of states was the final move the East was looking for to leave, pting to exit the federation alongside the East. The invasion of the Mid - cause, as this violation of agreements between the leaders of the Southern Regions also helped ensure t hat neither the West nor Mid - West would secede in solidarity with the East, even though the West had been leaning strongly in the opposite direction as recently as April 1967. So while Gowon had some success in limiting the scope of the conflict, by the ti me he felt comfortable with 187 his own hold on power, he had ignored the festering societal discord for too long and could not successfully prevent a conflict altogether. 56 Overall Assessment of the Frameworks Based on the narratives above, it is clear th e January 1966 coup was related to the July 1966 counter - coup, and both were related to the July 1967 secession of Biafra and the ensuing civil war. Indeed, those events have consistently been viewed as inextricably linked to one another, with scholars arg t. Jorre 1972, 43); b) (Peters 1997, 15). On the basis of th ese strands of evidence, the independence of events framework seems to offer very little by way of a greater understanding of events in Nigeria between January 1966 and July 1967, and fairly easily be rejected. The overall assessment of the constraints and trade - offs frameworks, however, is more complex. It is clear that Nigeria faced a number of constraints at independence, and that those constraints were not dealt with in a constructive way and thus persisted throughout the period under consideration in this case study. However, given that these constraints were in place, it is equally clear that they affected the range of possible actions available to elites. And those actions 56 Given that the final o utcome was total war between Nigeria and the Republic of Biafra, limiting the scope of the impacts on the intensity and type of risk t hey face. 188 and decisions built upon one another, affecting the range of possible actions available to elites in the future. Without these initial constraints of regional and ethnic rivalry and weak political institutions in place, therefore, it is likely elites would have found themselves with a broader array of potential courses of action at their disposal. If there had been meaningful agreement on the form of government and the degree of representation of each region, and less tension between regions and the different ethnic groups in those regions, the resort to fraud and electoral manipula tion by the civilian incumbents would have been less necessary as a condition for remaining in power. Rather, there was so little support among the population and politicians for the status quo system that conducing politics backed by legitimacy was essent ially impossible. Without the development gap between North and South and the ensuing resentment between regions, the population dominance of the North, and its subsequent control over the federal government, might have been less threatening to the South, and particularly the East. Given the failures of the civilian regime to effectively address any of these issues, both Ironsi and Gowon found themselves in more complicated positions, since these old problems were compounded by high passions and tensions f ollowing each military coup. The two coups were followed by demands for a re - ordering of the political system, but the military leaders had even less widespread legitimacy than the politicians had had. This made their decisions more controversial and divis ive, further reinforcing the preexisting mistrust and suspicion between ecree outraged best option for survival was complete exit from the existing system. 189 In order to more fully understand the interconnections between events in the late years of the colonial era, through the early years of independence to the coup, counter - coup and outbreak of civil war, both the constraints and trade - offs frameworks must be considered together. The constraints framework does an excellent job helping provide an understanding of the social, political and economic contexts states find themselves in. The trade - offs framework, meanwhile, helps account for the types of decisions that incumbents and elites make, and the ripple effects that those decisions h ave. While each framework by itself is incomplete, viewing the trade - offs framework as nested within the constraints framework brings into clearer focus the layers of complexity and hardship present in Nigeria since the colonial era and extending even beyo nd the period considered in this chapter. Conclusion In this chapter, I have carried out a qualitative analysis of Nigerian politics from the period before independence to 1967 in order to identify and assess the micro - processes leading up to a coup in J anuary 1966, a counter - coup six months later and the outbreak of civil war in July 1967. As the colonial period ended, Nigeria was endowed with a parliamentary democracy, but also a vast array of social, economic and political challenges. Nigerian politics unfolded along largely regional lines, and those geographic regions more or less coincided with the major ethnic divisions within the country. In this climate of regional rivalries, public goods like development assistance that tended to benefit one regio n more than the others were viewed with suspicion and as semi - private goods by the other regions, making almost all policies extremely contentious. Despite the deeper divisions these rivalries sowed, regional elites were generally content to exploit them f or their own political benefit, pitting the regions and primary political parties against one another to ensure their 190 incumbency. At the national level, the military became a standard tool of internal enforcement as it was dispatched to various areas of un rest throughout the country. It was against this backdrop that the first coup occurred, in January 1966. As Major General Ironsi sought to stabilize the country through policies of centralization, namely the unification decree, the North panicked that it s domination of the center under the democratic rules would be immediately ripped away. Despite evidence that the political system still functioned similarly in a number of ways under Ironsi as under the civilian regime before him even after some centraliz ation measures were implemented, the North fought back against this feared possible future by staging a counter - coup in July 1966. Ironsi was replaced by Lieutenant Colonel Gowon, the darling of the North, as leader, and. Gowon quickly undid many of the po licies aimed at centralization. While lauded by the North, these same policies antagonized the East, which viewed itself as the victim to ease tensions between th e East and the national government, made the situation worse through lack of clarity on the part of the Accord to ambivalence over implementation on the part of the two sides. Further actions by Gowon, especially the declaration of new states in April 1967 to replace the old regional system, pushed the East to secession and the country as a whole to civil war. With a different set of political, economic and demographic factors in place at independence, and a different colonial legacy, Nigerian elites woul d not have faced quite so many major problems at the onset, and would have had a wider set of alternative policies to choose from in trying to get a handle on the political process in the country. At a minimum, without the regional and ethnic rivalries, a far broader range of redistributive and development policies could have been and policy decisions raised one risk at particular times while leaving the other ri sk more or less 191 stable, in support of the trade - offs model, it also shows that elites only took these courses of action when it seemed there were very few others. Incumbents did often make matters worse, but things were hardly ideal to begin with, lending support to the constraints framework. To develop a fuller understanding of how and why politics played out in the way they did in the 1960s, neither the constraints nor the trade - offs framework by itself is sufficient. Both are necessary for insight into not just the decisions that were made, but why they were made how the regional/ethnic troubles and disparities in the levels of development in different areas limited the range of possible policy avenues so that the chosen courses of action were the mos t appealing and rational at the time. For instance, it is in a dire environment that policies made by the military decree and the minority support for the ne w states) were the best decisions that could have been reached. Each framework provides only one slice of the explanation, but together they can be layered on top of one another for a clearer and more sophisticated view of the past. 192 CHAPTER 6 BURMA, 1 948 - 1962 Introduction This chapter examines the social and political events in Burma that resulted in an outbreak of civil conflict by two different groups shortly after the achievement of independence in 1948, a gradual breakdown of the political system resulting in a rather dramatic temporary turnover of power to a military regime for 19 months beginning in 1958 and the military voluntarily turning power back over to civilians in 1960, only to seize it permanently in 1962. In order to provide the necess ary context for an understanding of these events, I will consider the impact of the colonial era, as well as the dominance of political life in Burma by a few key personalities. As with the case study of Nigeria, I will also provide an assessment of the th ree theoretical frameworks from Chapter 2 in light of events in Burma. The timeline will be broken into four primary sections: the colonial era, the period immediately surrounding the outbreak of hostilities by the Communists and Karen, the political moras s that slowly unfolded throughout the 1950s, and the state of affairs that had a more had turned power back over to the civilians. I also include an overa ll assessment of the performance of each of the theoretical frameworks to determine which are most helpful in understanding how events unfolded in Burma. 193 Colonial Era Historical Background The British began ruling Burma in 1885, administering the Ir rawaddy Delta and southern plains separately from the northern hills and frontier areas. The south was populated overwhelmingly by portion of the country. This g eneral pattern of governance was resilient, surviving transitions from Burmese kings to the British in 1886 to the Japanese in 1942 and back to British control in 1945. The direct rule of the lowlands and plains and indirect rule of the frontiers and hills that would be the hallmark of British rule in Burma was largely inherited. When the British authority replaced that of the kings in 1886, this differentiation was already in place. Efforts at Burmanization in the hills under the kings had been none too ef service in the Burman armies, they were allowed to retain their identity, language, customs, dress predecessors, the British established much more effective rule over all the peoples under their - frontier divisions already tacitly guiding national policy. This division between the Delta/plains and the hills/frontiers was the dominant one in Burmese society, and one that would feature prominently in discussions prior to independence about the scope and form of government in Burma, and the extent to which the distinct historical and political heri tage of the minority groups would be acknowledged and respected (Lintner 1984; Silverstein 1959; Silverstein 1990; Selth 1986). The nationalist movement in Burma did not really begin until the late 1920s and into the 1930s. It was dominated by two big pe rsonalities: Aung San and Thakin (later U) Nu, both of whom were active in the student movements in the early 1930s and would go on to be the faces of 194 movement was b orn well before World War II, most progress was not made until after the British returned, following their hasty flight from Burma when the Japanese invaded in 1942. There are two primary issues of particular note in this pre - war period. First, until 1930, the nationalist movement never made an effort to contextualize the traditional differences in governance between Burma proper and the hill/frontier areas, essentially conceding that separate rule made sense (Selth 1986). This was probably due in part to t he fact that the nationalist movement was primarily Burman national dress, and Burmanization of the schools and the unification of all the peoples in the country groups cling more tightly to British rule, which they viewed as protection against Burman encroachments. Second, the Burmese military prior to World War II was almost en tirely comprised of ethnic minorities, in particular the Karen. Indeed, Burmans had been effectively barred from participation in the regular armed forces by the British after 1925, when all Burmans already in the military were dismissed and recruitment ef forts came to focus on the Karen, Kachin and Chin. - 489). Thus, when the war started in 1939, the Burmese armed for ces were overwhelmingly comprised of non - Burmans. The composition did not change dramatically even after the 1941 declaration of war by the Japanese, and many of the Burmans who had joined the military deserted after the Japanese invasion in 1942 to join t he Burma Independence Army, the military wing of the almost - exclusively Burman nationalist movement. 195 an invasion and occupation as an opportunity to further its own goa ls. The Japanese were viewed as a potential ally in the Burman quest to overthrow British colonial power, and multiple organizations reached out to initiate contact with them. At this same time, the Japanese were close of 1939 it was suggested to the Burmese groups - 5). Contact between nationalists and the Japanese was made prior to the invasion, and young nationalists were covertly tr ansported to and from Japanese - controlled territory to receive military training for the fight against the British, forming the Burma Independence Army. On the eve of the invasion, Thakin Nu and Aung San tried to extort independence from Britain, naming it as the price for Burmese loyalty and assistance in the war effort. When the British did not agree on this point, the nationalists moved forward with their plans to collaborate with the Japanese, and managed to secure the formation of a subordinate govern ment in 1942 and a declaration of independence in 1943, although Japan had not been honest with the nationalists about how quickly that independence would be forthcoming, given that the nationalists had agreed to help the Japanese with the invasion on the condition that Burma be declared independent once Japan reached the city of Moulmein in 1942. Although the Japanese had deceived the nationalists, the nationalists were quick to return the favor. Nationalists, most prominently Thakin Nu, who served in t he Japanese puppet government in part as a means of protection for the Burmans began working and conspiring against the Japanese as early as 1942 (the year of the invasion). At this point, the nationalists, primarily through contacts with the Communists an d the Karen, reached out to the British to plot against Japan. The Japanese forces had eventually grown suspicious of the Burmese, and by 1945 the 196 majority of Burman troops were located near the capital, surrounded by Japanese troops. It was, however, from this position that General Aung San asked for and was granted permission to march out of Rangoon with the Burma Independence Army (now renamed the Burma Defense Army) to attack incoming British paratroopers. Before he departed, Aung San made sure to leave some of the nationalists behind, as ostensibly loyal partners, to retreat with the Japanese. However, they were really there to help rescue any nationalists who were captured by the Japanese after Aung Burmese cooperati on with the Japanese, however limited or opportunistic, was scorned by many British. Following the Japanese invasion, the Allied Powers viewed malcontent minorities, who were becoming increasingly convinced that their best hope for future autonomy and prot ection was under British rule, as crucial to their war effort, an internal weapon against the Japanese in control in Rangoon and their Burman collaborators. The Burma Independence Army had engaged in brutal, targeted violence against some minority groups, particularly Indiana, Chinese and Karens, who were none too eager to welcome the Burmans with open arms. And while active moves against retreating British had been limited, they were more than enough to get post - war conciliation efforts and independence ne gotiations off to a rocky start. Once the British returned to Burma in 1945, they continued to actively encourage distinctions between the Delta/plains and the hills/frontier areas. When approached by the Burman nationalists about a united Burma, Britai n insisted that any possible joining of the two areas would er 197 sharper distinctions between its ethnic groups than had exis ted over one hundred years ago at the returned, the slaughter of some minority communities by the Burma Independence Army on the heels of the Japanese invasi [seeing] a greater hardening of divisions between the majority Burmans and minority hill peoples, as well as the development of new rifts that were to lead directly to the many attack s on the In addition to the collaboration between the nationalists and the Japanese, the other major political development for the nationalists during the war was the formation of the Anti - Fascist edom League (AFPFL), a conglomeration primarily of Army, Socialists and Communists. It was recognized by the British colonial administration as the foremost political organization operating in Burma at the time, as it was the most popular, inclusive and la rgest of any ethnic persuasion, and as such was the primary vehicle of representative government moving forward through negotiations for independence. 57 While there was an elected Constituent Assembly and other bodies, these were all dominated by the AFPFL. The first two years after the end of the war were spent by both the British and the nationalists jockeying for bargaining power and leverage. The British did not want Burma to outpace India on the road to independence, but it was also untenable to hold Burma back simply for the sake of letting India go first. The vast majority of the negotiations, and particularly the most contentious, were headed by Aung San. After a shake - up in the British colonial administration in Burma, negotiations progressed much more quickly and smoothly, with the Burmese receiving 57 Although the AFPFL came to be recognized as such, it was not immediately recognized, with the apparent view that the British did not want to provide any legitimacy to the primarily Burman organization to use as leverage in any post - war negotia tions. 198 most of the concessions they sought. The pace quickened considerably more in January 1947, when the British Prime Minister and Aung San agreed to the rapid timeline for the creation of an independent Bu rma. The Delta/plains and hills/frontier would be governed together, with the consent of the various minority groups needed in order for the Burmans to exert political authority over them. The Panglong Agreement, negotiated in February 1947 between Aung Sa n and the Chin, Shan and Kachin, exchanged autonomy and representation for the minority groups in Burma for cooperation in presenting a united front against continued British rule. While not all groups felt represented by the agreement (the Karen, in parti cular, had been hoping for British support in pressing for a separate, independent state), the united front it presented was invaluable to Aung 1947, Thakin Nu be came the president of the AFPFL and continued to court the minority groups in order to win their approval and support for the constitution that would govern independent Burma. However, the Karen especially were divided on the question of cooperation with t he Burman leadership, with some continuing to support the AFPFL and others withdrawing and preparing for conflict. As 1947 wore on, the milestones needed for independence to be achieved in January 1948 were achieved, and Thakin Nu announced Burma would fol low a leftist ideology Assessment of the Frameworks Independence of Events Alt hough Burma had yet to experience a coup attempt or civil conflict onset during the colonial era, trouble was certainly brewing, even while the British were still on the ground. At a macro 199 ighting that took place there during World War II, and limited capacity has been shown time and again to raise the risks of both types of phenomena. The bombing and ground fighting decimated agricultural production, which was the backbone of the economy. E ven for wealthier countries, like Britain, the cost of rebuilding after the war was enormous and burdensome; this was even more the case for smaller, less affluent societies. And Britain and Burma were not on the friendliest of terms after the war. Burma r esented having its nominal independence, granted by Japan, revoked, and the British resented that Burmese soldiers had taken part in violence against British troops and government officials during their exodus from the country surrounding the Japanese inva sion. By the time independence arrived, the economy was beginning to show small and tentative signs of improvement and some infrastructure was being repaired, though things were still much worse than they had been in the pre - war period. And the country cer tainly faced a unique sociopolitical situation, with the military as one of the three core groups in the dominant indigenous political organization (the AFPFL) and two distinct sets of potential insurgents, the ethnic Karen and the Communists, who were the mselves dived into two factions: the Red Flag group, which even in the colonial era took a very oppositional stances to the AFPFL, and the White Flag group, which had had the roller coaster relationship with the AFPFL (expelled by Aung San, reconciled with U Nu and expelled again). The portion of the established were questionably loyal to the civilian regime, as will be discussed in more detail in the following sectio n. Although no threats had been made by the PVO against the civilian regime, reliance on one portion of the military as part of the AFPFL tripod and a paramilitary organization of unknown loyalty did not offer any sense of reassurance that the civilian gov ernment would be able to carry out its work unimpeded. 200 Constraints By the end of the colonial era, Burma had secured paper unification of the country as a single quasi - federal entity 58 and stable leadership (U Nu formerly Thakin had taken over as leade r of independence, as well). Despite this seeming success, it also faced a number of hardships, including continued encouragement by the British of the minority grou ps to push for a separate existence. Indeed, The state of affairs in Burma when it achieved its independence in 1948 could hardly have been worse. The country had suffered some of the severest air - strikes in Asia during the war; the countryside was ravaged and the infrastructure almost destroyed. The inner circle of competent leaders had been murdered even before Prime Minister, U Nu, was a talented, intellectual politician but c riticized for not being the strong statesman Burma needed during its first difficult years of independence. Army units rose in mutiny, the Karen minority took up arms and demanded a separate state and the Communists went underground to organize guerrilla f orces. (Lintner 1984, 408 - 409) The conflicts involving the PVO, Karen and Communists will be discussed more in the next historical background section, so preceding challenges (including acrimony leading up to the outbreaks of conflict) will be the focus o f this subsection. The troubled relationship between the Burmans and the many ethnic groups was perhaps the most encompassing challenge facing Burma as it neared independence, permeating many other facets of society. Not only had the hill/frontier areas been administered separately for essentially the entirety of colonial Burmese history, those hill and frontier groups were wary of any potential Burman acquisition of power in those areas. For instance, as early as the 1880s the Karen were pushing for a te rritorial entity distinct and separate from the rest of Burma (Silverstein 1990). The 58 Burma proper was ruled as a unitary state while federal principles were applied to the minority areas. 201 treatment of the minority groups, especially the Karen, by the Burma Independence Army as they swept through the countryside after the Japanese invasion only further exac erbated concerns about what might happen under Burman control. In the early days of the Japanese occupation, entire villages were placed under arrest and subjected to violent repression for their cooperation with the British by members of the Burma Indepen dence Army, which was comprised almost exclusively of Burmans. The degrees of support for a united Burma were varied among the different groups, even those that participated in the Panglong Agreement. The Shan, for instance, pressed for and received a con stitutional right to secession after 10 years. Most groups did not receive such an assurance, and thus had less recourse for protection from pushes for Burmanization. With these intergroup differences in rights and privileges and intragroup differences of opinion in the degree of support that should be given to the government and the AFPFL, the uncertainty about where the minority groups really stood both in outlook and in policy vis - à - unresolved in the pre - independence pe (Silverstein 1959, 100). Also problematic were the divisions between the two factions of the Communists and between the Communists (broadly considered) the other major political units in the country. Especially significant among the latter were the divisions between the Communists and the other two legs of the AFPFL tripod, the Socialists and the military. Early on in the nationalist effort, Marxism was understood broadly and generally accept ed. However, as time went on the core of the AFPFL leadership came to see itself as Socialist, and to see Communism as being in conflict 202 with its goals. 59 Viewing Communism as incompatible with the mission and goals of the AFPFL, Aung San had carefully remo ved White Flag Communists from positions of power the League before expelling them completely. By this time, the Red Flag Communists were already in the process of becoming an underground organization and preparing for an assault on the state. After Aung S Communist Party back into the political fold, a move supported by the broader population. This attempt was short - lived, though, and the White Flags were expelled again sh ortly thereafter; at this point, they too began to seriously consider an organized resistance against the government, which refused to allow their legal participation in the political process. Challenges on the political front were closely tied to challen ges on the military front, and centered on the role of the latter in the former. These challenges can be understood in part from the viewpoint that a key feature of pre - independence Burmese political life was the kha piet saung tat , or pocket army (Badgley 1958). Life under the Burmese Kings was clouded with unrest during intervals when central authority was weak. Provincial groups would form their own military force, for both protective and aggressive reasons, usually under an inspired leader. The British fought for years against renegade forces of this type and were never able to Japanese after 1942 the young Thakin Burmese history, organized most Burman and minority leaders with military propensities into the Burma Defense Army. It was largely these men that General Aung San utilized in his own post - Organization) which was so effective an instrument of latent force in the independence negotiations with the British in 1947. (Badgley 1958, 338). 59 under the egis of the Burma C ommunist Party, remaining on slightly better terms with the government for a slightly longer period of time. 203 were reached for independence and Aung San was assassinated, creating som e problems for the civilian government, to whom they were not as strongly loyal. Indeed, it refused to give up its paramilitary identity, and viewed itself as distinct from the other coalition members in the League. This in particular could pose serious pr oblems in the future, since the PVO and other former members of the various incarnations of the Burma Independence Army were a necessary component of the AFPFL and could not be removed as easily as the Communists. If anything, the remaining members of the initial AFPFL coalition were all the more reliant on these military forces now, following the removal of the Communists as a secondary balancing group. The dubious loyalty much more to Aung San than to the League in its entirety of the PVO was problema tic as well, as they were an integral component of the winning coalition with only one well - known preference: to remain armed and a powerful political force. Even among minority groups pocket armies were prevalent, with each traditional ruler in the 30 pri ncipalities of the Shan States having his own private military force, capable of causing trouble if the state took unwanted action (Lintner 1984). There were two additional challenges facing Burma on the eve of independence. First, although U Nu was highl y popular among Burmans and minorities alike, he seemed questionably committed to being the leader of the country. Before Aung San was assassinated he had tried to step away from politics, but was called back following the assassination. Once the remaining details surrounding independence were shored up, he tried to exit again, only to once more be encouraged by various groups to serve as the leader of the Constituent Assembly, which would become the Parliament upon independence. Each time he tried to remov e himself from the political arena he was persuaded to return, but this, coupled with the fact that the AFPFL was the only 204 significant mass political organization meant that the functioning of the government rested on the ular appeal as an individual as much as in his role as a politician. With functionally just one party, Burma was sparsely and weakly institutionalized as it approached independence. Trade - Offs In the lead - up and conduct of World War II, the nationalists were faced with a choice ally themselves with the British, from whom they sought independence, or ally with the Japanese in exchange for the ability to self - govern, long refused by Britain. Siding with Britain would have brought eventual independence, an d perhaps a more coherent plan for incorporating the frontier and plains areas that had for so long been governed separately. Alliance with the Japanese meant an earlier declaration of independence and entrenchment of the ethnic Burmans and their military forces as the key actors with whom the returning British would have to interact. The explicit empowerment of the Burmans by the Japanese elevated their negotiating position vis - à - vis the British, but also guaranteed tensions with the groups who had subsequ ently been overlooked or Burman components of the military as an organization and the ethnic minorities as a result of World War II. Three main sources of risk faced the civilian regime at the onset of independence: the two Among the ethnic minorities, some actively opposed being united within a single union; others were, at least for the time being, quiescent; and still others were fairly content, in part because they had won constitutional provisions permitting secession after trying the quasi - federal experiment 205 for 10 years. As for the PVO, it had been extremely loyal to Au ng San, and was, as indicated regime, was unknown. Given that Aung San was the one who had initially sought the expulsion of the Communists from the AFPFL, U Nu w ould have to tread lightly in seeking any rapprochement or conciliation with the Communists lest the PVO feel the wishes of their martyred leader were being betrayed. On the whole, while it is unquestionable that Burma faced an imminent threat of civil c onflict onset as it approached independence, the exact inclinations of the armed forces, still largely comprised of non - Burmans, and the paramilitary PVO, which was almost entirely Burman, were simply unknown. This is perhaps the most difficult position to be in, since there is no indication of whether particular actions are helping or hurting the already - uneasy relationship between government and military. While it thus cannot be said that Burma also faced the risk of a coup attempt on the eve of independe nce, it can be said that the actions of the civilian regime would be extremely important in coming years in determining whether the military and paramilitary forces and civilian regime would come together or be pulled further apart from one another. Imme diate Civil War Historical Background With the sociopolitical situation deteriorating in Burma even before independence was achieved in January 1948, Thakin Nu had proposed in November 1947 a political coalition that would ganization, the Burma Communist Party and the Socialist Party. This offer was rejected by the Communists, however, who opted instead to go underground and 206 prepare for an uprising. The PVO would eventually break into factions the White and Yellow Bands based on the question of the inclusion of the White Flag Communists in the political system. While it was still a united organization, Nu had proposed its disarmament and the creation of a new political party with them as equal partners. Disarmament was un appealing, though, and no agreement on the issue was reached, eventually sparking the split in the PVO and the alignment of the White Band faction with the Communists and the Yellow Band faction with the AFPFL. were particularly difficult, with three distinct armed resistance movements breaking out. The Red Flag Communists and portions of the White Flag Communists rebelled first, three months after independence, followed by the White Band faction of the PVO and m embers of the armed forces who had been part of the Patriotic Burma Forces (formerly the BIA) in July and August 1948 and by the Karen (and military troops loyal to the Karen) in April 1949. The Communists and the PVO had access to various arms caches thro ughout the country that had been stockpiled during the war, some under the authority of Aung San, and never hauled in by the government. The leading Karen political organization, the Karen National Union, had rejected the subnational state offered to them within the Burmese Union prior to independence, preferring more territory than included in the proposal. A group of Karen hardliners then went on to form the Karen National Defense Organization, which advocated for full independence, and won strong support , even among some Karen moderates. These insurrections sparked further divisions, contributing to the mutiny of several components of the Burmese military forces. The White Band PVO, the KNDO and other Karen organizations and Burma Communist Party joined forces as the Joint Political Committee, making them that much more powerful and that much more difficult for the government to handle. Moreover, by the end of 1949 the various 207 rebel organizations controlled almost 75 percent of Burmese towns and posed a s erious danger to the capital. By the end of 1950, however, the Burmese military had seized on slowed rebel momentum to successfully recapture considerable territory it had previously lost to the rebels. In this offensive against the insurrectionary movemen dependent upon the military for its survival. Hence, the ruling politicians incurred a heavy debt to some conflict of goals amongst the members of the Joint Political Committee, as the ethnic minorities in Burma tended to want a federal system with more autonomy for their territories, while the Burma Communist Party wanted a strong unitary system. As a result, the BCP fo ught against both the minorities and the Burmese military as part of its struggle. In addition to the domestic rebellions, fleeing Chinese Nationalists first entered the border regions of Burma in 1949, and clashed with the military throughout the 1950s. The Kuomintang incursions were primarily into the Shan States, and members of the Burmese military were shipped in to deal with this foreign threat. The necessity of having troops on the ground in the Shan States to combat the KMT was another burden for a government and military already stretched thin by multiple rebellions, but also caused greater friction among the locals, who had been shielded from interaction with the Burmans during the colonial period. Indeed, to those in the Shan countryside who had h nationalist movement based on the unique characteristics of a Shan identity (Lintner 1984, 412). government became more assured of its bare survival, the insurrections were not completely eradicated, but rather lingered on, with the potentia l to flare up again at any time (Wallinsky 1965 - 2 08 1966, 271). Following the defections of large portions of the military, the armed forces were restructured somewhat to help ensure the loyalty of the overall organization: some senior officers of ethnic minor ity option were relieved of their command, and an ethnic Burman, General Ne Win, was promoted to the position of commander - in - chief of the armed forces. And it should be noted that those first three domestic insurrections were not the only three to have ev er occurred. In addition to the Karen, portions of the Mon, Kachin and Arakanese rebelled as well in the first few years of independence, as did a Muslim movement based in the Arakan region but distinct from the ethnic Arakanese rebellion. And more broadly , beginning with the Karen uprising in 1949, and frequently with several Assessment of the Frameworks Independence of Events Wi th the onset of civil war so quickly after independence, the risk that brewed as the colonial era came to a close was realized. This civil war had broader reaching impacts, as well, bringing into sharp relief the relationship between the military and the g overnment. Large segments of the military broke away from the government side and joined the rebel movements, some on the basis of ethnicity, others on the basis of ideology. The defections of the military were directly contingent on those of the rebels, s ince with no rebellion there would have been no immediate cause for defection and no pre - existing organizations for the defectors to join. And while it was not clear at this point if disloyalty in the military extended beyond those who willingly left to jo in up with the rebels, it was all too clear that anti - government action by members of the military was a very 209 distinct possibility, indicating the threat posed by members of the armed forces to the survival of the government. Constraints The rebellion o f some segments of the armed forces comprised of the ethnic minorities only served to further underscore the precarious and fragile relationship between the majority Burmans and the minority groups. Care would have to be taken to ensure that other groups w ere not pushed to revolt out of a sense of solidarity with those groups which had already rebelled. If sociopolitical dealings from the colonial era contributed to the onset and continuation of ot make things easier: Several factors combined to make the pacification of the countryside a protracted village and jungle areas. The monsoons made it difficult if not impos sible to campaign from June through October. The government was severely handicapped by lack of trained manpower and materiel, especially in the early years of fighting. majori ty of them actually went over to the rebels. (Trager 1966, 116) The configuration of the countryside and the weakness of the government also made it all the easier for China to provide support to the Red Flag Communists and for Kuomintang troops who had been defeated in China to make incursions into the hill areas and wreak havoc there. Heavy reliance on the military to handle the domestic and foreign insurrections made the h now had fewer members of the PVO as the White Band decided to join the movement against the military itself as well as its supposedly civilian masters in the AFPF 210 vanquished by any one of the various rebel movements. To help ensure a military victory, a majority of national revenue in the first few years after independence was spent on combatting the insurrection, and although the military lacked sufficient personnel and equipment at the height of the conflict, after the worst of the fighting was over, the civilian regime made sure that military 134). It should be noted that the ability of the military to secure Bur partially relieve some constraints. The government, with the protection of the military, had proven itself strong enough to survive the multi - faceted rebellion and perhaps move on to other issue areas now that the biggest had been handed. However, the issue of the rebellions was not completely resolved, and thus remained capable of causing more problems in the future. Nonetheless, once it was clear the government would not fall to the rebels, some attention could be turned to p ursuing balanced between agriculture and industry and a foreign policy centered on neutrality (willing to accept help from both superpowers while being official ly tied to neither). Trade - Offs While the survival of the state in its fight against the rebels was still in doubt, the government of Burma faced tough decisions in terms of how to try to bring the conflict to an end through political, rather than just m ilitary, means, but also to what extent these political options should be pursued, and the extent to which the military should be supported at all costs. One such decision about possible political solutions was whether to integrate the non - rebelling portio n of the White Flag 211 Communists or keep them exiled from the political sphere. The latter course might halt the rebellion of the other White Flags by providing them with an institutionally empowered voice, but the divisions just within the AFPFL on the issu e of Communism (the early alliance, but later expulsion) showed just how incendiary the issue was, and no guarantees could be made about the potential reaction by the military or the public to bringing the White Flags back in after some of them had been ac tively fighting against the state. With these questions lingering, in an apparent effort to give U Nu maneuvering room in negotiations with the Joint Political Committee to end the insurrections, the Socialist and Yellow Band PVO members of the government resigned en masse from the Cabinet. In reality, this was in itself a political crisis and led Nu to form a non - party Cabinet with the General Ne Win, commander - in - chief of the Burmese military, as a member. The military as an organization had been importa nt to the survival of the government before, but now a specific member of the military became part of the government itself, foreshadowing a similar but more encompassing development later in the 1950s. The importance and prominence of the military grew wi th the seriousness of the ongoing civil war, and without this military leadership and a small group of loyal field officers and junior officers and men, the Union could not have survived the initial attacks of the Communist and Karen insurrectionists. Yet the army leaders were not only a military elite. Like the Socialist AFPFL political leadership, they participated in the struggle for independence and were known in the countryside as Burmese patriots. They could and did attract support even after the mili tary units of the Karens and Communists defected. Typical of these men was their leader, General Ne Win, who had been a Thakin before he became a Bo, or military leader, and who since 1932 had participated in nearly every phase of the development of the ar dent nationalist movement. During the desperate and intricate peace maneuvers between the AFPFL government and the insurrectionists in the spring of 1949, he agreed to and Defens e. (Trager 1966, 133 - 134) 212 Although portions of the military remained loyal throughout the worst of the civil war and the 1950s, the seriousness of the defection of so many personnel cannot be understated. 60 Without the onset of conflict in the first place , troops would not have defected to join the conflict on the side of the rebels. But a bigger issue was the loyalty of the military to the regime. Clearly, a sizeable portion of the military was in no way loyal, and left as soon as a justifiable opportunit y arose. But it is unknown whether the remaining troops were really more loyal than those who defected at that time, or just had not felt compelled to desert over that particular issue. Those who defected initially did so in opposition to government respon se to the rebels and in support of the rebel ideology and cause. Could and would another issue arise that would call into question the loyalty of the remaining troops? With no way to know, the civilian leadership could not formulate a comprehensive policy toward the rebels and feel secure in the continued fight of the military on whether keeping such considerations in mind would even have an impact on troop loyalty in the long run. And while defection was certainly one potential response for disloyal troops, it is not the only option: rather than join forces with groups opposing the state, such members of the military would certainly have the ability and access to w eaponry to directly challenge the state themselves. Thus, the onset of civil war in the immediate wake of independence confirmed the risk of civil conflict, but also very much highlighted the very weak ties between portions of the military and the state, a nd the risk that members of the military would take action against the state. 60 While the Independence of Events framework would hold that the systematic turn of portions of the military away from the state and toward the re bels is unrelated to the causes of the rebellion, this is a hard case to make. The defectors mishandling of the rebel risk provided impetus and opportunity for the military to show its disgust at the situation. 213 The Uneasy 1950s Historical Background Continued Malaise Throughout the 1950s, insurrectionary activity continued, especially by the Karen, Communists and Kuomintang troops, a just after independence, following a long and aggressive campaign by the armed forces on a variety of fronts. In an effort to help in the fight against the insurgents, the governme nt created its own pocket army, the pyusawhtis , or local defense units at the township level. They were designed to augment, rather than replace, the regular armed forces at the local level. While the intended purpose of these units were to assist in the o verall effort against the Communists, they came to be associated with excessive violence in their communities and were viewed in generally as poorly defense forc es were of national consequence only because of the popular sentiment they created settlements with the rebel groups, the government continued to drag its feet on t he Karen question, with a constitutional amendment for a separate Karen state within the Union only introduced to Parliament in 1951 and administrative authority for several Karen territories remaining in the hands of the central government until 1954. A s the 1950s progressed, various groups continuously vied for power. Political parties had initially held the position of the most privileged and prime actors in the political process; however, as the decade wore on, the presence of other groups trying to p articipate in this process challenged both the primacy and legitimacy of the political parties. Except for the insurrectionaries, who 214 in influencing the power of this assault on its dominant position in society and the political arena, the AFPFL sought to exercise its control patron - client network cultivated by the polit ical leadership inevitably led to abuses of power, merit being replaced by those with ties to the nationalist movement or individual high - ranking members of th e AFPFL (Aung - Thwin and Myint - U 1992, 70). Despite general elections held from mid - 1951 to early 1952 in which the AFPFL won an overwhelming majority to both houses of Parliament, the ongoing troubles in the 1950s were further compounded by the outbreak of a Shan rebellion, partially egged on by the KMT and simultaneous economic and political crises in the Burman heartland, described more below. In addition, independence was closely followed by a recession, and U Nu had not been successful in his efforts to establish a welfare state in Burma. 1956 Elections The National Unity Front, a pro - Communist party, became the first major party challenger to the AFPFL in 1956, taking far more seats than the latter had anticipated in the election that year, even ami d allegations by the NUF that the AFPFL engaged in violent tactics leading up to the April election. In response, U Nu stepped down as Prime Minister in June to focus on reorganizing the AFPFL, which he believed had lost due to poor organization of the par ty structure itself and rampant corruption on the part of its members. In July, U Nu announced he would resign from 215 increased tensions among the party leaders who se powers lay in the state executive and those who - 28), presaging more catastrophic divisions in the coming years. When U Nu resigned as Prime Minister, he appointed junior colleagues to take turns filling the post. While he was still in charge of the AFPFL and on his crusade to reorganize it and and instead be made into a figurehead, with nominal, but not actual, power. Upon hearing of this proposal second - hand (U Nu himself was not present at the meeting where the suggestion was made and was only told later by a colleague in attendance), U Nu vowed to return to the premiership as soon as possible. Upon his return in Feb ruary 1957, he stepped up his crusade against corruption, which increased tensions between different factions in both the AFPFL and the government, who felt Nu was playing favorites in terms of which factions were being more severely targeted for mass arre sts. AFPFL Split Nu ultimately joined one of the budding rival factions in the AFPFL, but prior to that, he authorized mass arrests of members of both factions. The Third All - Burma National Congress of the AFPFL had been an opportunity and attempt to re concile the two sides, but to no avail, and in early May the League was formally split into two rival factions: The Clean faction led by U Nu and Thakin Tin, and the Stable faction led by Ba Swe and Kyaw Nyein. After siding with Thakin Tin, Nu continued to use his authority to order various forms of repression against his political rivals, particularly a number of mass arrests, ostensibly related to corruption, throughout the spring of 1958. After the split in the political - party component of the AFPFL, all the accompanying 216 organizations at all levels of government split apart as well. Groups previously deemed undesirable or less important minorities and communist sympathizers were suddenly courted by both factions as the key to a majority vote and the r etention or reclaiming of power. Perhaps more than for any other group, the split in the AFPFL was a victory for the Communists, who viewed the split as an opportunity to potentially regain legitimate political power as an ally to one of the factions. For - Tin faction came government concessions to the Communist insurgents, an exchange which deeply concerned the Army. However, in the immediate aftermath of the split, the Army remained neutral in terms of support fo r the two factions. Both factions were warned not to use violence in their attempts to gain the political upper hand. The Army remained very much opposed to the Communists and any attempt to incorporate them into the political process or surrendered Commun ists into the armed forces, but Ne Win warned against the military as an organization becoming involved, but he also warned the officers and soldiers against becoming involved as private individuals. As part of the stipulations surrounding the split, an emergency session of Parliament would be convened on June 5, 1958, to determine which faction would control the government until new elections were held. In the lead - up to a vote of no confidence in the Parliament on June 9, 1958, both factions did whateve r they could to secure support. In an effort to secure the loyalty of the minority groups, Nu recanted previously stated opposition to the creation of more ethnic states by openly supporting calls for an Arakanese state. The protracted showdown on virtuall y all societal fronts between the two factions brought government functioning to its knees, as no political or governmental actor or entity was willing to anger either faction, lest they ultimately back the winning faction after the power struggle was decided. 217 On the day of the no - the Secretariat in which compound the Parliament is situated. Around it were drawn heavy armed guards comprising the Army, Navy, Air Force and Police. Magistrates were also on stand - by to give orders to shoot in case trouble broke out. Fire - engines, ambulances, armoured cars, and the ote by 8 votes, mostly because he had secured the allegiance of the NUF and many of the minority groups. After surviving the vote, U Nu shored up his control over the government and continued to heighten tensions: he did not call the Cabinet again (which still had some members of the opposing faction) between the vote of no confidence and the takeover of the caretaker government in late October 1958. Even so, both the Clean/Nu - Tin and Stable/Swe - Nyein factions continued to jockey for position for the upcom ing elections scheduled for November 1958. Part of the Clean/Nu - more loathed even than the various rebel groups who had been waging war on the state for a decade. While this may have initially been prompted by a few specific military officers who publicly supported members of the Stable/Swe - Nyein faction, the Nu - Tin faction went much further , seeking to undermine the credibility of the organization and foster acrimony toward the military in the broader population. Nu - Tin had already undertaken such campaigns against the police, Union Military Police and the civil service without much backlash . Given the general popularity of the military because of its success against the insurgencies in the early 1950s and the successful ridding of other state organs of followers of the Stable/Swe - Nu - Tin judged at the time that it was easier to weaken the strong Army than to strengthen their 218 While this did have the impact of reducing somewhat the popularity of the military, it was much more successful at alienating and annoying the military, which (rightly) viewed itself as the reason survival in the first years of independence. usually passed by Parliament in Septem faction than the more evenly - split Parliament was. The Stable/Swe - Nyein faction strongly opposed this plan, as it would strip them of the remnants of their political power. Accompanying the possib ility of this maneuver were rumors that after doing so U Nu would make big changes in the high ranks of the Army, including sending some leading officers out into the heart of the country, away from the center, as well as possible assassinations of perceiv ed political and military opponents, further increasing apprehension across all sectors of society. Caretaker Government As tension throughout the country in the aftermath of the split came to a head, it became apparent that elections conducted on schedu le would produce considerable violence by those in both factions, and that the government as it currently stood would be unable to accomplish anything with just an 8 - vote cushion of support over the rival faction. Moreover, the Army was alarmed by rumors c irculating in late September 1958 about the leaders of the Clean/Nu - Tin loyal to the Clean faction took up positions throughout Burma, including the capital. In response, an increasingly nervous military heightened its security positions in the capital and other major 219 mistrust and suspicion on the part of the military and the Nu - friction between the two, leading to preparations by both sides. The situation then resembled the international politics where world powers justify their armaments as precautionary measures for 59, 70). Thus was the situation when two close associates of Ne Win helped broker the details of the invitation by U Nu to Ne Win to take control of the regime as head of a caretaker government. As will be explored in further detail below, while there were complex The invitation to form a caretaker government was f ormally extended to Ne Win on September 26, 1958, and he was unanimously voted into the office of Prime Minister on October 1949 - 1950, during the height of the ins urrections, as qualification for taking over as Prime Minister, especially with the nonpartisan civilian government he would appoint to provide him premiership was that the elections originally scheduled for November 1958 could be legally delayed for six months. After Ne Win was ensconced as Premier, the Parliament was able to return to a more normal state of affairs, meeting several times, in part to discuss the bud get. However, perhaps even more so under Ne Win than U Nu, the real power lay in the hands of the Prime Minister and those he deemed to be his closest advisers, rather than in the Parliament. Upon taking over as Prime Minister, Ne Win engaged in a major r eorganization of the Cabinet, including a dramatic reduction in the number of ministers. Ne Win outlined the caretaker 220 for free and fair elections within six months, if law and order were restored and the political parties cooperated; lowering the cost of living; and seeking remedies for the economic chaos in the support i n the Burmese public in the early days of his caretaker administration, relieved that the most major crisis facing Burma had been peacefully resolved and that the continued issue of the insurrections was being addressed. Indeed, the military focused most o f its attention on the governing style. While significant attention wa s paid to the insurrections, the caretaker government accomplished more than a reduction in the number of rebels and the readiness of the country for governmental eff however, the administration frequently used vigorous methods which were a marked depar ture from the easy - and von der Mehden 1960, 145 - 146). Importantly, the Ne Win re gime took great pains to adhere to the electoral rules and guidelines of the constitution. He came to power constitutionally, and when it became apparent that the country would not be ready for elections by the end of his constitutionally - set six months in office, Ne Win resigned. The Constitution was then amended so that an individual who was not a member of parliament could hold the post of Prime Minister for more than six months, and Ne 221 Win was reelected as Prime Minister to continue the work his caretak er administration had begun. Enough progress had been made for Ne Win to announce in September 1959 that elections would be held in February 1960, and the elected government would take power in April of that year. 1960 Election Campaign The 1960 electora l contest was strictly between civilians the military - led caretaker government did not attempt to remain in power by electoral means beyond the time limit it had specified. Although there were a number of political parties and independent candidates runn ing in the February 1960 election, the main contest was really between just two: the Stable/Swe - Nyein and Clean/Nu - Tin factions of the AFPFL. Despite the hard campaigning by the Clean and Stable AFPFL factions in the lead - up to the elections against the opposing faction in both cases and against the military in the case of the Clean AFPFL Although it did not compete outright, th e military was deeply involved in the 1960 elections. During the rule of the caretaker government, members of the Stable AFPFL had the tacit and sometimes even open backing of the military, which the faction believed would work in its favor in the February 1960 elections. In contrast to the Stable AFPFL, which courted the military, 61 the Clean AFPFL spent the duration of the caretaker regime opposing, if not openly antagonizing the armed forces. Indeed, U Nu had been preparing a civil disobedience campaign s urrounding the 61 There is debate on the extent to which the military outright supported the Stable faction. Butwell (1969) holds the view that the military did not strongly support the Stable faction as much as it preferred it to the Clean faction as the least awful of two thoroughly unappealing options; however, regardless of the level of support of the Stable faction, the military was quite actively opposed to the Clean faction, arresting its follow ers and trying to defame U Nu (Butwell 1969). 222 complained [including the defamation of his char acter] were not remedied. The major speech of Aung San and those assassinated with him were buried]. One hundred thousand persons turned out, and U Nu was repeated - handed tactics of the military in fighting the insurrections and accomplishing its other goals was days were considered a lesser evil, in spite of its frequent corrupt and inefficient character, than the Army - bably protest votes against army The Clean faction overwhelmingly swept the 1960 election, winning more than a 2/3 strengthened by the elimination of those disruptive elements in the pre - split AFPFL, which had trongest - 71), an issue discussed more in the next section. While the Clean AFPFL had certainly been united for the highly educated, retired bureaucrats and some retired military officials into his party, creating three differe 223 official return to the premiership. The biggest exception to the Clean AFPFL sweep of the country was in the Karen state, which went almost unanimously for the Stable fac tion. Following the victory by the Clean AFPFL, the Ne Win regime handed power back without incident to the Clean faction, headed by U Nu. While this was not the preferred electoral outcome of the military, which had spent its tenure governing effectivel y (if sometimes harshly) and building the capacity of the state, it nonetheless accepted the will of the populace. that the military would be weakened by protracted neglect of i ts main mission, Ne civilian government by most of the top Army leaders, and probably fear that U Nu would launch a civil disobedience campaign (which would not serve any of t he - rulers also may have realized that they were becoming more like the civilian politicians they had displaced, slowly compromising their standards and differing among themselves the result in part of the environment o f political problems and traditions that they inherited. The fact that the elections were held, however, was a tribute both to the selflessness of the Assessment of the Frameworks Inde pendence of Events As the 1950s wore on, so did the insurrections that broke out just after independence. While they did not pose an existential threat to the Burmese state after about 1951, the military remained an especially important political actor. Th e military publically voiced concern when it felt the government was making too many concessions to the rebels, and warned U Nu on multiple occasions to not give up too much to the Communists as he fought members of his own political party to remain in pow er. The relative standing of the rebels vis - à - actions toward the rebels, and these actions were subject to scrutiny by the military. To the extent 224 that conciliation with the rebels would bring them into the legitimate political arena, the government faced retaliation from the military, which refused to have anything to do with the rebels, especially the Communists. Repeated warnings not to grant amnesty or allow rebels into the regular armed forces suggests t hat the military was quite serious in its scorn for the rebels. The condemnation of the rebels is also evidenced by the actions of the caretaker government, which spent considerable time and energy grinding down the insurgencies and weakening them sufficie ntly that continuing to exclude them from the political process posed no major consequences. Throughout the 1950s, it became increasingly clear that the military would not tolerate any incorporation of former rebels into politically - active society, and t hat it viewed efforts to do so as a direct threat to the military. Thus, as the insurrections wore on and the government tried a variety of tactics, including conciliation, to bring them to a close, the government faced threats posed by ongoing civil confl ict, but also the risk of military action to prevent or stop particular interactions with the insurgents. Constraints Although the actual existence of Burma was more or less shorn up by the end of 1951, other issues continued to face the country. The eco nomy remained sluggish in the early to middle years of the ec onomy remained in non - reached pre - war levels. Remote areas disrupted by the war did not completely recover throughout - Thwin and Myint - U 1992, 70). This generally poor economic 225 performance continued into the later years of the decade, when 1957 brought both droughts and floods, harming the overall agricultural output that year as well as the 1958 price crop, and lower prices for other commodity exports like minerals dampened economic performance even more. More damaging than this economic uncertainty was the governmental uncertainty produced by the AFPFL split, which led to lower economic output and increased government inefficiency as businesses and bureaucrats alik e opted to slow activity rather than anger either faction. The general outlook was bleak in Burma leading up to the handover to the caretaker government. years of in holding the allegiance of the various hill peoples, especially the militarily significant Kachins and so be detrimental, especially if the individuals to whom the masses are loyal come into conflict with one another, as they did in the mid - to late - Another problem facing Burma after its basic existence was assured was the capability of its leaders to competently serve as such: terribly damaging, ha d a modicum of such experience and ability been present among his colleagues, and had he been inclined to delegate authority reasonably commensurate with the responsibilities assigned. Neither of these was the case. Of any experience worth mentioning in the world of affairs, and even his was not great. Below ministerial level, the civil services unfortunately were also weakly staffed, even when measured only by the requirements of traditional government activities. But the socialist and development drives had burdened the government apparatus with a host of new, non - traditional responsibilities constituting an almost impossible burden for even an experienced, first - class civil service to which an appropriate degree of au thority had been conveyed. U Nu simply could not recognize the need for administrative 226 reorganization to improve the quality of the public services and their managements, or to remove the bureaucratic shackles inherited from the colonial administration so that the services could at least perform up to the level of their limited capabilities. (Wallinsky 1965 - 1966, 274 - 275) Between the limited political experience of the politicians, continued insurrections, the surprise performance of the NUF in the 1956 e lections and the perceived threat to his continued (Trager 1966, 122). While this may have been the goal, the preceding circumstances and other factors conspired against Nu, making achievement of that ideal very difficult. The pyusawhtis , or government - formed pocket armies, also proved to be an ongoing challenge for the civilian regime, with U Nu himself stating at one point that without reform, they - committees existed at the district level headed by senior members of the military to provide pyusawhtis were difficult to control and in some districts they became unmanageable vigilante groups, committing crimes in the name of justice. In two or three areas they became out In some ways, the assurance of the continued existence of Burma opened the door for new problems, since the government now had some maneuvering room. It no longer had to focus all of its attention on mere surviv al, but once it came time to govern, the lack of political experience by those in power became a major hindrance in moving the country forward normally. In this circumstance, the fact that the AFPFL was less a fully - developed political party and more a con duit for political patronage became more obvious and was in a position to create greater difficulty. Once the government was in a position to actually govern, the lack of necessary internal 227 infrastructure to do so caused the government to falter, and preve nted it from carrying out much needed development tasks. For instance, appointment to top party positions (as opposed to election by party members) was relatively common and at the local level, power allocation was determined in large part by the strength have helped the AFPFL remain united for a time, it hindered its ability to adapt to the new needs Burma would encounter as it developed and grew from barely - maintained existence to a fu lly - functioning country. Additionally, there was no meaningful counterweight to the AFPFL in the government. After 1956 the NUF was an opposition party in the parliament, but the parliament itself was viewed by U Nu as mostly serving an affirming function he would make the primary governmental decisions and the parliament would sign off on them without question. The functional process by which policy was made centered on the Cabinet and the Executive Council of the AFPFL, so very few important matters wer e ever posed to parliament, meaning that government policy was almost wholly at the discretion of U Nu. The survival of the regime and the country into the heart of the 1950s was an impressive feat, given the economic and demographic challenges it faced, to say nothing of the ongoing civil conflicts by multiple societal groups. Compounding these initial challenges, though, were difficulties stemming from the lack of political infrastructure in the state that affected not just the government as a whole, bu t also the AFPFL, which for the first several years of the decade was the only organization that existed with sufficient strength to populate and run the government. Trade - Offs g the destiny of the 228 destroyed on the whims of the individuals in power. Because U Nu was the dominant political player during this period, much of what hap pened throughout the 1950s was a direct consequence of actions by U Nu or reactions by other elites to those actions. Up until the AFPFL split in 1958, (Butwell 1969, 156). This view was held by colleagues and competitors alike, and Nu was able to proceed on a variety of policy programs as a result. As this faith in Nu faltered when he chose control of the government by any means necessary after doing so, so too did Nu lose the ability to carry out any actions whatsoever without appearing to his critics play favorites. In this light, the risks Nu faced from both the rebel groups and the milita ry, especially beginning in the mid - to late - 1950s, were entirely of his own manufacture: there was no way for Nu to satisfy both the Communists and the Army, but his frequent attempts to court the favor of one group at the expense of the other, combined w ith his repeated flipping of support from one group to the other, prevented him from enjoying either themselves with the opposing Stable/Swe - Nyein faction o f the AFPFL. results of that year, to reorganize and purge the AFPFL organizations in the districts the apparent motive or rather to demonstrate to his colleague s that they could not get along without him? The truth is, I fear, that they did not miss him very much, during this absence. In many ways, the - 1966, 279). Prior to the vot e of no confidence on June 9, 1958, U Nu met with senior army officials to discuss the security and political situation in the country. The fighting between the two AFPFL 229 factions had permeated every level of society, and with the government distracted fro m the ongoing insurgencies and in some cases hoping for the cooperation and support of some of the insurgent forces, namely the Communists, security in Burma was tenuous at best. Faced with this, proposed formation of a 6 - man inner circle with three Army Bosses and three from his faction to take a joint hand at the helm of the States. It was an astounding offer an unofficial Coalition Government with the Army! But the Army leaders declined the of off the further spread of unrest and hostile political competition, U Nu, months before the caretaker government was actually formed, courted the military as a means to retain at least a marginal hold and concessions promised in exchange for support minority and leftist po level of activity by pacifist groups and Communist - front organizations clamouring for u nconditional cessation of the civil war increased the anxieties of the tatmadaw [Burmese armed the political split, the Army had faith extension of excessive conces 230 respect to the insurrection movements. Because of the courting of and heavy reliance on the lo yalty of the rebels by U Nu between his victory in the vote of no confidence and the handover to Ne 62 rebels who were on their last legs and suing for peace and leniency j ust before the two factions broke up The possibility of complete disloyalty by the military as a reaction to the anti - military rumors and propaganda originating fro m his followers was not lost on Nu, who had a speech broadcast addressing this very issue as he left on an official trip through Upper Burma. In the speech, he stated that the allegations against the Army by the Nu - Tin Convention had been made owing to th e ignorance of the Nu - Tin leaders of the role of the Army. He clutches of the insurgents when it was only a two - finger breadth from the edge of the abyss and it fought and dro ve out the Chiang Kai - shek intruders; that the Army was a benefactor of the people and his AFPFL must defend the honour and prestige of the Army which had excelled itself in courage, perseverance and sacrifice, U Nu would be enough recompense for the damage done to the Army. But the Nu - making more District Army Officers become increasingly suspicious of Nu - Tin ith crypto - Communists. The relations between the Army, and the Nu - Tin politicians, especially in the districts, turned from bad to worse. (Win 1959, 76 - 77) Win prevent ed broader society from completely disintegrating (Win 1959). While this may have been the case, U Nu was not an enthusiastic supporter of the caretaker regime, and in fact felt the 62 This is a reference to the propensity of the various ideological rebel groups in Burma to incorporate a color into Volunteer Organization, etc. 231 caretaker government took on too wide an array of actions, rather than lim iting itself to preparing for the elections. Nu tried to push for elections even before the initial six - month term was over, promising desirable Cabinet positions to members of the military if his party won the subsequent elections. However, Ne Win disagre ed that the country was ready, and did not call for any such , and during the caretaker regime, to barter with the military for his own hold on power, including making the military an integral component of civilian adminis trations. In the course of dealing with the insurrectionaries, U Nu increased concern among members of the military that he would make too many concessions in trying to bring the conflicts to a close. The conflicts wore on, though, and even as the milita ry was growing increasingly wary of each faction to secure societal support increased the importance of the rebel groups, and as such increased the consterna tion of the military when efforts at reconciliation with the rebels were broached. In order to prevent the complete political breakdown of society, U Nu agreed to turn the government over to General Ne Win, the Commander in Chief of the Burmese armed force s, in part to prevent a coup that could have easily been carried out in the fall of 1958. The rebel movements were severely weakened during that caretaker regime, but U Nu continued to perturb the military even while out of power. Nonetheless, U Nu won a h uge electoral victory in 1960, and the military, though preferring a different outcome to the election, handed power back to U Nu, albeit it with some skepticism and reluctance. In various efforts to deal with the ongoing civil conflicts, then, U Nu manage d to earn the distrust and in some cases disloyalty of the armed forces. 232 While the distrust and disloyalty were very real, they were also the direct result of the actions U Nu took in the political realm, particularly with regard to the various rebel movem ents and autonomy - seeking minority groups, in addition to his electoral strategy of demonizing the military, which only served to sow deeper seeds of discord, as he made the differences in governing style and preferences very personal. Final Decline to t he 1962 Coup Historical Background When U Nu returned to power as Prime Minister in April 1960, he had been granted a huge electoral mandate by the people to carry out his promised policy proposals, including the creation of states for the Mon and Arakanes e ethnic groups and the institution of Buddhism as the official state religion. In part because of the promises of statehood, once he returned to office, the Mon and Arakanese began pressing hard for their promised autonomy, and other ethnic groups also co ntinued to seek similar autonomy, self - determination and even outright independence. To the chagrin of the military and non - Buddhist minority groups, U Nu also sought to follow through on his other major electoral promise, to make Buddhism the state relig ion of Burma. He passed two constitutional amendments in 1961 to achieve this, one officially acknowledging it as the state religion and the other assuring protections for religious minorities. This move in particular angered the Kachin, who outright rebel led as a result of the proposal. Additionally, the number of Shan rebels, who aligned themselves with the ongoing insurrection of the Karen, more than tripled between 1959 and 1961. By 1961, rebels controlled 10 percent of the country, a far cry from the c ase in the first few years of independence, but still a substantial amount of territory lost by the government. 233 Grumblings about the policy pursuits of U Nu were particularly prominent in February 1961, with senior officers openly questioning the civilia n leadership and musing that the Army should step into the political scene once more. U Nu approached General Ne Win about this subversion, and Ne Win remained loyal to civilian rule. A number of the high - ranking officers involved in the rumblings were nam ed to political positions, re - stationed outside the country, while still others resigned, ostensibly to enter the private sector or the political arena. escalation of ethnic dissent, economic uncertainty, and increasing demands for internal peace. The populace became disillusioned with the ruling party and developed apathy to the political process isintegration of the Union election campaign, the party had been splitting into three distinct factions, two of which aligned with one another against the th AFPFL had in 1958. In December 1960, U Nu announced that he would step down as leader of the party at the next national party congress. Rapidly from that point forward, the party continued to factionalize and in - fight, culminating with the functional and practical (though not official) split in the Union Party when individuals from the Thakin faction (organized around the idea of strong, personalized leadership of the party) won the battle f or the leadership positions in the party in February 1962. ethnic minorities to discuss possible reforms to the federal structure of the Union in order to accommodate each - determination to the greatest extent possible, the army, once again led by General Ne Win, staged a coup, placing U Nu under arrest, 234 abrogating the constitution and with it the guaranteed right of some ethnic groups to secede from the state religion, that pushed the military to int ervene (Aung - Thwin and Myint - U 1992; Chang 1969; McCarthy 2008; Min 2008; Schock 1999; Silverstein 1990; Trager 1963). However, although not discussed as vocally as the prior two policies, Probably no less important, the leading soldiers were fed up with w hat they - style democratic political institutions had been twice tried and found wanting. And Burmese socialism had been too long delayed in its imple and the still economically important Indian and Chinese communities. (Butwell 1972, 901) Although Ne Win made no efforts to assure the public of the temporary nature of this mili hostility to the coup. If anything, there was a feeling of relief: at last, the slide downward Assessment of the Frameworks Independ ence of Events The independence - of - events framework continues to perform poorly in this final historical period, the military, and the degree to which U Nu paid heed to the preferences of the military relied heavily on the potential gains he viewed as a result of a final settlement with the various ethnic groups. Regarding the former point, there were certain policies that were simply unacceptable as far as the military was concerned, such as open acceptance of Communists as part of the 235 government. U Nu understood that there would be swift and dire consequences if such policies were attempted. Vis - à - vis the latter point, U Nu perhaps paid less attention to th e warnings of the military about his plans at revising the federal structure and trying to end the conflict with the minority groups because he perceived an enormous electoral and popular boost to himself and his colleagues if he ended at long las the conf lict even the military government, with all its efficiency, had been unable to bring to a close. And while the victory of U Nu, with the help of General Ne Win, in the showdown with the military in early 1961 helped the regime survive to continue pressing 231). While the resignations and reshufflings neutralized the immediate t hreat of specific mutinous officers, they left the underlying issue of chilly civil - military relations unaddressed. Constraints Although his sweeping electoral victory appeared to be a feather in the cap of U Nu, it belied the troubles and struggles stil l facing the state. For instance, despite his victory, several of the minority ethnic groups were still, and increasingly, restive, in part as a response to the increased laxity of the U Nu regime following that of the strong - armed methods of the military caretaker government. Once again, the threat of rebellion by multiple groups, coupled with the ongoing rebellions of others, constrained the range of options open to U Nu. He could not carry on as though there was peace throughout Burma, because this was s imply not the case, but it was also unclear what tactics, other than those backed by brute military force, could succeed in quieting the insurrections. 236 The economy had improved under the caretaker government, and a rebound in agriculture and world food p rices further helped produce a stronger economic output than had occurred over the past decade. In the political, arena, though, all was not well. A number of overlapping factors combined to cause some uneasiness, and in the end and in combination, they br the promulgation of Buddhism as the state religion; the rise of insurrectionary activity, particularly among the Karen National Defense Organization (K NDO) and other minorities and this problem was complicated by the Kuomintang (KMT) particularly among the Shans, demands which were interpreted as an indirect attack on the i his great popular appeal to restore order where it was lacking or where it was under attack. (Trager 1963, 311) While some of those factors, like the pursuit of Buddhism as the s direct control and are discussed further below, others were beyond the realm of his control, but nonetheless constrained him and increased tension among other groups in the winning coalition, namely the military. With the strong emphasis on individual personalities in Burmese politics, the personality characteristics of the individual leaders can also be taken into consideration as assets or liabilities in their efforts to guide the state through the various challenges fac ing it. In the case of U Nu, his devout adherence to Buddhism was problematic, as it raised tensions with the military and further alienated the already - disgruntled non - Buddhist ethnic minorities. This especially became the case after his return to power f somewhat disorganized, never fully using his popular appeal, always preaching to leader groups d no 237 longer reconcile the requirements of political leadership with his desire to achieve Buddhist Trade - Offs After returning to office in April 1960, U Nu was cau ght between a rock and a hard place. The ethnic minorities to whom he had promised statehood in 1958 and again in 1960 were demanding follow - through. Portions of the military officer corps were openly hostile to U Nu and speculated on the necessity of anot her Army intervention. Moves to placate the minorities, some of whom were gearing up for a rebellion, caused further consternation in the military, and any bowing to military preferences by U Nu, while not really attempted, surely would have rankled the mi norities, since the Army was in general opposed to new states or any other concessions to the ethnic minorities that would divide the Union. Stemming from its (albeit fairly recent and sometimes strained) role as one of the few multinational organizations in the country, the military also opposed the sanction of Buddhism as the official state religion, against the popular viewpoint among the ethnic Burmans, who constituted 75 percent of the population, and most of whom were Buddhist. Indeed, the military c onsistently favored unity in Burma opposing federal reforms that would transfer considerable authority to subnational units as well as the culturally divisive state religion proposal that underscored differences between the majority and minority groups. Throughout his second regime, fatal to his hold on power: Once returned to office, U Nu seemed to pay scant regard to maintaining his position: he antagoni zed the army by his insistence that Buddhism be made the State religion and failed to keep in check the wild schemes of federalism on the part 238 a fateful mistake: and on 2 M arch 1962 the army acted. (Cook 1970, 259 - 260) fraught with appr 993, 35 - U Nu could have proceeded constructively from this point forward (early 1961), he would have maintained control of the situation, for the country had rallied in the face of an external aggressor, sic ], and had soured on the claims of the Shan and Karen separatists and rebels who made common cause with the hated Kuomintang Chinese. But, once again, party friction and factionalism broke out Finally, previous decisions of U Nu continued to have consequences on the functioning of the state. The centrality of the military to the survival of the country in the early days of independence and as its government from 1958 - 1960 meant that even once U Nu was restored to be partly made by the politicia had not carried out his policies with this consideration in mind, he would have perhaps bee n more successful in retaining control of the government, though constrained in what he could hope to accomplish, if he had. And although the circumstances at the time may have justified the bringing in of the military as a caretaker government in 1958, do 239 (Lovell and Kim 1967, 117). The precedent set in the name of saving the country, then, had proven to be a dangerous o ne. Overall Assessment of the Frameworks There can be little doubt that the colonial era, and especially World War II, produced a tough environment for Burma. Divisions had existed between the ethnic Burman majority and the numerous minority groups even before the British arrived, but the British and Japanese policies of pitting the different sets of groups against one another exacerbated those divisions. The destruction war only created greater hardship for the fledgling government. The issue of minority dissatisfaction weighed heavily on the leadership in the negotiations leading up to independence as the Burmans needed the support of the minorities as leverage agains t the British. The issue continued to preoccupy leaders in the first years of independence, when several distinct groups rebelled and nearly succeeded in bringing about the destruction of the country. This dire situation would be diminished but never elimi nated, and continued as one of the primary issue areas throughout the entire parliamentary period. In light of the ethnic troubles, the primary leader throughout this period, U Nu, faced the continuous political tightrope between the rebels and the milit ary. Leaning too far in the direction of one would provoke the other. U Nu consistently erred on the side of the former, with the result not only of continued insurgent threats (as he did not completely court the rebels nor acquiesce to their demands), but also a disgruntled military that came to lose confidence in his ability to govern in a manner consistent with the national interest. 240 Since rebel policy was the point of contention between citizens and soldiers, without the rebel groups the civilians wou parliamentary period from 1948 to 1962, then, one must bear in mind the constraints framework, but also the trade - offs framework. The economic and ethnic constraints were the locus of Burmese polit ics, and it was in this environment that everything else came to pass. That is, constraints produced the civil war just after independence, but without the series of decisions and miscalculations in dealing with it, a coup would have been an unlikely respo nse to the normal functioning of the political realm. Conclusion The historical narrative developed throughout this chapter points to the 1962 coup as the final ne gotiation of the Panglong Agreement explicitly acknowledged the presence of serious divisions defection of enormous segments of the military to join the insurrectionary movements, highlighted just how deep these divisions were. Against this backdrop of rebel activity that nearly destroyed government faltered repeatedly. Faced with the troubled post - war economy and tense social and political environments, country had ever known U Nu. Nu was a devout Buddhist, and his high moral ideal di d not always mesh with the realities of running a developing country, bringing him into conflict with rival personalities in his own party as well as with the military establishment. The conflicts 241 lly influenced every decision he made as Prime Minister. strong performance in the 1956 elections, the 1958 Anti - and the 1960 el ection campaign, other political actors and entities came up against the strength for instance, were insufficient to pacify these groups, but were far more ge nerous than other political actors, particularly the military, would offer, and thus earned the scorn of hardliners on both sides of the issue. Tensions within his own political party, the AFPFL, eventually forced a split in the organization, and necessita ted an intervention, albeit a constitutional one, but the Commander - in - Chief of the armed forces. Having already raised the ire of the military once with his perceived intransigence on the rebel issue and alienation of politicians preferred by the milita ry, U Nu rode a popular tide back into office on promises of new states for select minority groups and the designation of Buddhism as the state religion, both of which had levels of support within the public. The military was vocally opposed to both polici es, but U Nu determinedly plunged ahead, despite caution and advice to drop the policies. Although the economy was improving, more ethnic groups threatened to rebel based on the issue of statehood, making that long - standing constraint even more pressing. W ith a path more or less set for at least serious consideration of more states and Buddhism passed as the official religion of Burma, the military opted to remove U Nu from power in 1962 than to see what other problematic policies he would take it upon hims elf to pass. Japanese policy. Ethnic diversity and tension loomed large in the political consciousness, and 242 despite repeated attempts and a number of tactics, the civilian leadership was never able to fully get a handle on the problem, which only continued to fester and flare up periodically. The breakdown of the ruling party based on conflicts between the leading politicians spread throughout Burmese society and p rompted the civilian leadership to briefly hand power over to a constitutionally - approved military leader. The military was much better able to handle the insurgents, but curried no favor with the broader public, who viewed their army rulers as too harsh o n a variety of fronts. U Nu was able to leverage this discontent into a huge electoral victory in 1960, and however reluctantly, the military voluntarily handed power back to the civilians. This cooperative spirit did not last long, though, and as the re bel situation heated back up, so too did annoyance by they seized power, this time with the force and outside the bounds of the constitution. Without having to perpetually address the issue of the rebel groups, U Nu could have economic and infrastructure damage caused by World War II. Instead, he was preoccupied by a problem with no solution backed by consensus; in fact, there existed a wide gulf between civilians and military personnel in terms of the perceived best approach. Whether conscious or not, U Nu had a particular knack for annoying the military, and his ob liviousness to this fact meant that his attempts to deal with the insurrections had the effect of ratcheting up civil - military tensions with no attempts at abatement. The scorn of the military toward U Nu was entirely of his own manufacture, but it was the initial challenge of ethnic diversity and animosity that made the policies on which the two sides disagreed the most, the most salient. 243 direction of the country in the this situation that eventually resulted in the 1962 coup. Both the constraints and trade - offs o f interethnic cooperation to a multifaceted civil war to a constitutional assumption of power by the military, return to civilian rule and at last a coup that would produce a military regime that would rule uninterrupted in 1988, and that following changes in leadership, remains in power to a initial position, and the decisions that were made in that environment, through the lens of the trade - offs framework, accoun t for its slow but assured downfall into multiple forms of political violence. 244 CHAPTER 7 LOOKING BACK AND MOVING FORWARD Introduction At its core, this dissertation has explored the nature of the relationship between two forms of political violence coup attempts and civil conflict onsets that, even if ultimately unsuccessful in their achievement, are deeply destabilizing to the affected societies and institutions. Past literature has tended to examine only one such form of violence at a time. Whil e Roessler (2011) bucks this trend, it is just one example of the avenues still to be explored, not the totality and final word on those avenues. This study, then, has begun to pursue another open line of research, seeking to probe not just what relationsh ip exists between the two phenomena, but also why it exists. The previous chapters set up a progressive examination of the relationship between coup attempts and civil conflict onsets, beginning with the development of new theoretical frameworks, a novel statistical analysis of the observable implications of those frameworks and two case studies to further probe the micro - processes underlying the progression of political violence from one event to the occurrence of the other. Since the primary contributio n of this study has been to reframe the theoretical understanding of the potential connections between these events, I will begin with a summary of this framework and then move to reiterating the key statistical and qualitative findings from my analysis. I will also touch on some of the limitations of the present study and possibilities for future research on this area of interconnection between forms of political violence. I will end with a discussion of some of the implications of this study and the pract ical insights that can be gained from it. 245 Primary Contribution: Theoretical Approach and Nuance The single most significant contribution of this dissertation is my development of a more nuanced and sophisticated theoretical lens for understanding both cou p attempts and civil conflict onsets as part of a single, underlying process as opposed to distinct and unrelated events. Previous research, namely the original incarnation of the selectorate theory (Bueno de Mesquita et al. 2003) and its subsequent i terat ions (see for example Bueno de Mesquita 2007; Bueno de Mesquita and Smith 2009, 2010 ), provides a foundation for the development of this framework. The selectorate theory provides a common language and understanding of actors, preferences and interactions that facilitates theorizing across diverging processes. Although leaders being constrained and unable to affect their political fates seems quite different from leaders having the ability to manipulate the relative risks they face, they can both be underst ood through the language of the selectorate theory, coalition wanting as many private goods as possible driving their interactions. With modifications, the select orate theory can be tailored to more aptly identify actors relevant to the coup attempt/civil conflict onset processes as well as their preferences and predicted patterns of interaction. For instance, while the selectorate theory focuses on incumbents, sel ectorates and winning coalitions as the primary actors, the theory I develop parses those groups further, identifying specific subsets of societal actors, such as the military and potential rebels, that are of direct consequence and relevance to issues and questions of political violence. In terms of the theoretical process, the major insight from this dissertation is to conceive of the potential relationship between coup attempts and civil conflict onsets in three primary ways: as no relationship at all (the independence of events framework), which is the current tacit understanding of most of the literature on either phenomenon; as part of the same underlying 246 process(the constraints framework), where the same types of factors, conditions and circumstance s that cause one form of risk contribute to the other; and as part of contradictory processes (the trade - offs framework), where the factors, conditions and circumstances that make one event more likely actually serve to make the other less likely. Honing in on some of the nuance, while the original selectorate theory (Bueno de Mesquita et al. 2003) constructs the selectorate and winning coalition as unitary actors with uniform preferences, my theoretical argues that there are distinct sets of actors in eac h group, and that the preferences of these disparate groups is not always identical. This leads to the possibility of conflicts of interest within the ruling elite that may manifest themselves in terms of a coup, with the military portion of the winning co alition turning its back on the rest of the government. Likewise, members of the disenfranchised public cannot simply be written off en masse. Here, too, there are distinct societal groups. Members of the potential rebel group will respond much differently to their disenfranchisement than more obsequious members of society, and must thus be considered as distinct from the quiescent portions of the population. Explicitly taking these types of actors into consideration makes it easier to see how leaders would face constraints that would make both groups more likely to revolt, as well as possible trade - offs, where attempts to placate the military increase the animosity of potential rebel groups and vice versa. The conceptualization of the winning coalition and selectorate in this way, leading to the three frameworks of the potential form of the relationship between civil conflict onsets and coup attempts is unique to this dissertation and also opens a host of future research possibilities, discussed later in thi s chapter. 247 Summary of Statistical Findings Chapter 3 focused on conducting statistical analyses of the performance of each of the three theoretical frameworks. For all analyses I employed bivariate probit models, which consider each phenomenon as a depe ndent variable and estimates the regression mo del for each in a recursive error terms of the two equations. This allows for the systematic assessment of the relationship between the two equations that would not be possible using a different statistical technique. In considering the expectations of each of the three frameworks, evidence for the independence of events framework would have been no systematic relationship between the coup attempt and civil conf lict onset equations of the bivariate probit models. A statistically significant, - offs framework. Across all of the model specifications, there was overwhelming and consistent support for the constraints framework. In every model, the rho was statistically significant and positive. This suggests that there is indeed a systematic relationship between coup attempts and civil conflict onsets, and that the types of factors and conditions that facilitate one event likewise facilitate the risk and occurrence of the other. Although the findings vis - à - vis the operable framework were strong and consistent, there was sign ificantly less support for my predictions about the impact of specific variables on each of the two equations. And when looking at the models constructed from existing studies of coup attempts and civil conflict onsets, many of the results that held in the single - equation probit models dissipated when they were included as part of a bivariate model. To a large extent, this shortcoming is consistent with much of the political violence literature more broadly, an issue I discuss in more 248 detail below. To the e xtent that specific variables achieved statistical significance, they were by and large variables more associated with constraints GDP, population and previous instances of political violence than with trade - offs. In the original models I developed, th ere was one key finding with respect to a variable initially expected to indicate with trade - offs, though which performed in a manner more consistent with constraints, with both risks responding to the variable sion in the short - term increased the risk of both events, though longer - term use of repression decreased the risk of both events. Another core component of the statistical analyses was the identification and discussion of model fit. This helped gauge th e modeling utility of considering coup attempts and civil conflict onsets as part of the same underlying process, as opposed to two, independent processes. Calculations of the AIC and BIC repeatedly indicated that explicitly modeling the two phenomena toge ther fit the data better than modeling them separately. This was true for both the replications and the original analysis, providing even stronger support for the theoretical claims advanced here. That is, in addition to the parameter consistently indica ting a systematic relationship between coup attempts and civil conflict onsets, the model fits further revealed that estimating the likelihood of coup attempts with variables typically associated with civil conflict onset models and vice versa increased th e ability of the model to account for both phenomena. That the addition of several variables to each equation did not result in the models overfitting the data is a testament to the potential gains still to be made in understanding the two forms of politic al violence, and points to considering them as part of the same underlying process a useful starting point. 249 Comparison of Case Studies: Similarities between Nigeria and Burma The case studies of Nigeria from 1960 - 1967 and Burma from 1948 - 1962 were desi gned to look more closely at the causal mechanisms linking coup attempts to civil conflict onsets in a more fine - grained way than that possible with the statistical approach. The case studies were able to consider seriously idiosyncratic factors of importa nce in one state but not systematically important in other states that would thus be difficult to measure in a way conducive to statistical examination. The case studies were also able to consider a much longer time horizon, though the exact length varied across the two cases. In Nigeria and Burma alike, both events did not occur within the same calendar year, but there is nonetheless compelling reason to believe a single process ties the two events together. Several insights arise from a comparison of th e findings of the two case studies. First and foremost is that when looking in close detail at the nuances of a specific country over a specific period of time, the trade - offs framework cannot be dismissed, despite the seeming lack of support for it in the statistical analyses. It is certainly true that the constraints framework provides important context for understanding events of the case study. The constraints framework highlights the basic conditions and circumstances in a country and places bounds on what type of decisions by leaders and subsequent outcomes are within the realm of possibility. But it is the trade - offs framework that helps us truly understand why and how situations and climates deteriorated and underscores the air of inevitability of th e occurrence of both coup attempts and civil conflict onsets in Nigeria and Burma. At independence, neither Nigeria nor Burma were on particularly stable ground. Both had weak economies, and in the case of Burma, a domestic infrastructure that had been d ecimated in three years of hard - fought air and ground wars. Both had deep ethnic/regional divisions which, 250 despite federal governing structures, completely consumed politics at the center through the very real and perpetual (and in the case of Burma, reali zed) possibility of armed revolt by one or more of these groups/regions. And neither had particularly strongly institutionalized governments. Nigeria and Burma both had arrived at federalism as a compromise system with the greatest possibility of success, but the system mostly just further increased expectations and subsequent tensions. The Westminster - style parliamentary system, so effective and stable in England, was not as successful in either Nigeria or Burma, in large part because of these other factor s that served to create a deeply divided polity. Also of note is that at some point during the period under consideration in each case, both countries were governed by a leader with a personality that seemed to only breed and foster greater conflict. In Eastern origin in July 1966 created immediate tensions with the military governor of the East. These tensions only continued to grow larger and more entrenched as the two men in teracted with rankling close political allies until even mere coexistence was impossible. But U Nu was also extremely popular with the public and with most of the minority groups. This created a challenge, because the democratically - approved leader (prior to independence, in 1951, 1956 and again in 1960) seemed incapable of maintaining a unified government capable of implementing democracy. When looking throu gh the lens of the trade - offs framework, though, the specific impacts of these constraints can be traced. In Nigeria, regional tensions made democratic government at the center all but impossible, and increased the power of and competition between the regi ons (and their majority ethnic groups) as a result. A military coup in January 1966 and the leader who assumed office afterwards tried to correct some of the perceived problems with federalism by 251 instituting a more centralized governing structure. This onl y heightened tensions, though, and led to a counter - coup in July 1966 by Northern troops who recognized federalism as the key to the following the counter - cou p ignited the conflict between personalities that would gather speed and lead to a total breakdown of Nigerian society, the secession of the East as an attempt at self - preservation and the national - level response of all - out civil war. In Burma, an early outbreak of civil war essentially guaranteed that the question of the ethnic minorities and politically - bothersome Communists would dominate the national agenda. Conflicts began to arise in the governing party (which was also the only meaningful mass polit ical organization for roughly eight years before the establishment of an opposition party capable of winning significant parliamentary seats) over what to do about the various insurgent groups, as well as other issues facing the state. U Nu himself became a participant in this conflict and subsequent power struggle, contributing greatly to the formal split of the ruling party and competition between the two factions for the right to govern. At this point an arrangement was brokered with the military to prev ent further deterioration in the country that might prompt a coup, and U Nu used his time out of power to rally against the military and paint them and their pop ular appeal (and popular discontent with the rougher tactics of the military while it was in power) brought him back to power in 1960, it created an immediately awkward governing situation, with the military once again subordinate not just to the civilians , but specifically to a election campaign included creating factions within his party to increase its appeal with various segments of the population. Once aga in his personality and vision came into conflict with those of 252 some of his colleagues, and as his party fractured apart and his policy pursuits threatened to formally divide Burma among the minorities, the military could no longer stand by idle while U Nu ignored their advice and admonitions. Without the initial civil conflict, though, the path of different focus. One thing that was apparent in both case studi es was that regardless of the exact constellation and severity of the initial set of constraints, the decisions taken by leaders reinforced them and also narrowed the range of possible future policies. This had the effect of creating new risks while never fully or effectively managing the risks that existed when they entered office. Limitations of the Present Study and Possible Future Research Although this dissertation has broken new theoretical groups and used a less - common statistical method to appropri ately assess that theory, like all research, the current study does face some limitations, the most obvious of which lie in the statistical analysis. Using the bivariate probit provided new insights about the relationships between phenomena not captured in previous research, but it was not able to incorporate all potential variables or time horizons. The biggest challenge the study faces is a consequence of carrying out quantitative cross - national research: the vast majority of data is macro - level and often measured, at most, yearly. This allows for only rough approximate operationalizations of the concepts of interest. This can be offset somewhat by the case studies, which can easily incorporate micro - level information and take a holistic approach to long t ime horizons, but there does not exist a remedy that can currently be applied with the same effect to the large - N analysis in its entirety. 253 Despite these shortcomings, there are many options for future research to explore, both quantitatively and qualita tively. On the statistical front, single - country micro - level analyses similar to the conflict studies by Kalyvas (2006) and those utilizing the same general approach (see, for example, Kalyvas and Kocher 2009) could explore systematic relationships within a single country using more localized and, potentially, more frequently measured data. Other statistical methods may also provide additional insights. For instance, multinomial logit analyses along the lines of Bodea and Elbadawi (2007) can provide clari ty on when one form of political violence may be chosen over another. And survival analyses may shed light on the specific timing of events based on their risk configurations. The scope of study could also be expanded although coup attempts and civil con flict onsets are two very prominent and destructive forms of political violence, they are by no means the only forms. And if these two forms are related, there is reason to think that other connections between different forms of violence may also exist. Another limitation is also not unique to the current study. As emphasized by Ward, Greenhill and Bakke (2010) and various coup studies (see, for example, Belkin and Schofer 2003; Jackman 1978; McGowan 2003), we do not have many strong and consistent findin gs in terms of specific variables. In the civil war, literature, for instance, we know GDP and population frequently matter. In the coup literature, only GDP matters with such consistency. Obviously individual studies have more impressive statistical resul ts, but as a broader research enterprise, what we know is still far and away trumped by what we do not know. So while there were few consistent variable findings in this study instability, use of repression and years of repression being the exceptions the new theory and use of method nonetheless ensure a meaningful contribution and simply confirms what we know more strongly than anything else: much more work remains to be done. 254 Implications While the new theory is the foremost contribution of this diss ertation and the qualitative and quantitative assessments of that theory a supplementary contribution, the third is of direct relation to the objects of study. In coming to better understand the relationship between coup attempts and civil conflict onsets, it becomes possible to begin to identify potentially vulnerable states. Once states that may be at risk for one or the other have been identified, policy interventions to prevent the actual occurrence of events, or at least the decline from the occurrence of one event to the occurrence of the other, also become possible. The wealth of coup and civil war literature reviewed in Chapter 1 can help shed light on this first issue. This study has shown that states at risk for one event should be considered at ri sk of the other, given that the same types of unobserved factors contribute to the occurrence of both, and the decisions made and implemented by leaders can heighten additional forms of risk. The statistical findings and certainly the case studies offer assistance with the issue of policy prescription. A top priority should be the development and fostering of strong institutions that are able to weather the storm of tough economic and demographic conditions. At the same time these institutions are being n urtured, efforts should be taken to facilitate societal harmony, which will make institutions stronger through consensus that they are the best institutions, but also diminish a number of conditions that put a strain on any institutions but which are apt t o destroy weak ones. In both Nigeria and Burma, policies that put the various ethnic/regional groups on equal footing with one another would have considerably decreased competition and acrimony between minority and majority groups, but could have also incr - in to the extant institutions. 255 The Yoruba and Ibo in Nigeria probably would have supported the central government more if they were not consumed by the fear of being trampled over by the Hausa - Fulani. Likewise, the minority groups i n each region would likely have been more content if they felt protected from potential excesses by the major groups. Likewise, groups in Burma would not have needed to seek secession as the primary alternative to complete and forceful Burmanization if the government had been structured such that those groups were given states within the country over which they had considerable autonomy, but would have still formed an integral component of a multinational Burmese Union. With these ethnic questions handled c onstructively, the government would then have bolstered institutions and the policy space to try to address the other constraints, especially economic hardships, confronting their state. Additionally, with stronger institutions and a general consensus over derail the functioning of the political process and (b) push portions of the population and/or the military to the point of a revolt. All of this is possible, but knowing which states are most vulnerable and their specific constellation of constraints could help NGOs and aid agencies to make the most effective use of their resourc es. In the end, then, and in combination with past research, the findings of this dissertation can be used to address constraint hotspots before the relevant leaderships have the opportunity to set their states on a deterministic path to multiple forms of political violence as they muddle through one crisis after another that is partially beyond their control and partially of their own design. 256 BIBLIOGRAPHY 257 BIBLIOGRAPHY Nig eria Since Independence: The First Twenty - Five Years. Volume V: Politics and Constitutions. Peter P. Ekeh, Patrick Dele Cole and Gabriel O. Olusanya, eds. Ibadan: Heinemann Educational Books. 186 - 205. Ajagun, Samuel Olushola, and Howardson A.O. Odion. 200 Social Sciences 5(4): 352 - 354. - Diverse Societies." Unpublished manuscript. Perspectives on the Nigerian Civil War . Siyan Oyeweso, ed. Lagos: OAP Publications. 1 - 18. - International Journal 20(2): 173 - 188. Aranson, Democratic Polity. In The Democratic State , ed. Roger Benjamin and Stephen Elkins. Lawrence: University Press of Kansas. 87 - 177. Comparative Political Studies 42(10): 1339 - 1362. Aung - Thwin, Maureen, and Thant Myint - Third World Quarterly 13(1): 67 - 75. ped Countries 1981 - Journal of Peace Research 34(2): 177 - 195. Pacific Affairs 31(4): 336 - 351. - Livermor e National Laboratories. PONARS Policy Memo 4. Balogun, Ola. 1973. The Tragic Years: Nigeria in Crisis 1966 - 1970 . Benin City: Ethiope Publishing Corporation. Banks, Arthur. 2001. Cross - National Time - Series Data Archive. Binghamton, NY: Computer Systems Unlimited. 258 Banks, Arthur S., and Kenneth A. Wilson. 2014. Cross - National Time - Series Data Archive . Jerusalem: Databanks International. Accessed April 7, 2015, via Michigan State University Libraries. Available from http://www.databanksinternational.com . Journal of Conflict Resolution 47(5): 594 - 620. ancing, and International Security Studies 14(1): 140 - 177. American Economic Review Papers and Proceedings 99(2): 292 - 297. Bienen, Henry, and Nicolas van de Walle. 1989 American Political Science Review 83(1): 19 - 34. Journal of Politics 54(3): 685 - 717. Far Eastern Survey 29(5): 70 - 74. Journal of Economic Literature 48(1): 3 - 57. Blimes, Randall. 2009. Great Expectations: Domestic Expectations, Conflict Outcomes, and the Duration of Leadership Tenure. Dissertation. University of Colorado. Braumoeller, Bear, an Models, Numbers and Cases: Methods for Studying International Relations, ed. Detlef F. Sprinz and Yael Wolinsky - Nahmias. Ann Arbor: University of Michigan Press. 129 - 151. World Today 48(8 - 9): 144 - 147. Leaders: A Comparative Study of Regime Types and Politica American Political Science Review 89(4): 841 - 855. American Political Science Review 86(3):638 - 646. 259 Bueno de M Comparative Political Studies 42(2): 167 - 197. Nature of Governme American Journal of Political Science 54(4): 936 - 950. Bueno de Mesquita, Bruce, Alastair Smith, Randolph M. Siverson, and James D. Morrow. 2003. The Logic of Political Survival . Cambridge: MIT Press. International Organization 61(1): 9 - 36. Conflict Management and Peace Science 27(2): 107 - 128. B International Studies Quarterly 52(2): 215 - 233. Bureau of Arms Control, Verification and Compliance. 1975. World Military Expenditures and Arms Transfers 1963 - 1973. Washington: U.S. Arms Control and Disarmament Agency. Bureau of Arms Control, Verification and Compliance. 1984. World Military Expenditures and Arms Transfers 1972 - 1982. Washington: U.S. Arms Control and Disarmament Agency. Burea u of Arms Control, Verification and Compliance. 1985. World Military Expenditures and Arms Transfers 1985. Washington: U.S. Arms Control and Disarmament Agency. Bureau of Arms Control, Verification and Compliance. 1994. World Military Expenditures and Ar ms Transfers 1991 - 1992. Washington: U.S. Arms Control and Disarmament Agency. Bureau of Arms Control, Verification and Compliance. 2003. World Military Expenditures and Arms Transfers 1999. Washington: U.S. Department of State. Bureau of Arms Control, Verification and Compliance. 2005. World Military Expenditures and Arms Transfers 2005. Washington: U.S. Department of State. - Economic Journal of Con flict Resolution 50(2): 176 - 203. Sociological Methods and Research 33(2): 261 - 304. Far Eastern Survey 29(2): 21 - 27. 260 Butwell, Richard. 1969. U Nu of Burma . Stanford: Stanford University Press. Asian Survey 12(10): 901 - 912. Butwell, Richard, and Fre Pacific Affairs 33(2): 144 - 157. No. 781. Africa in Transition . Council on Foreign Relations. https://blogs.cfr.org/campbell/2012/02/09/nigeria - civil - servants - unpaid - journalists - threatened - boko - haram - and - mend - bombing/ . Cappelen, Ådne, Nils Petter Gleditsch, and Olav Bjerkholt. 1 Journal of Peace Research 21(4): 361 - 373. Cederman, Lars - World Politics 62(1): 87 - 119. Ch - Asian Survey 9(11): 818 - 830. Quarterly Journal of Political Science 4(3): 211 - 228. C Conflict Management and Peace Science 25(4): 332 - 348. Intersta War Still Ex Post American Journal of Political Science 48(3): 604 - 619. 55 5 - 570. American Journal of Political Science 56(3): 671 - 689. ival, Rivalry, and Regime International Studies Quarterly 48(4): 713 - 727. 261 American Journal of Political Science 48(3): C Economic Policy 16(32): 127 - 166. Collier, Paul. 2007. The Bottom Billion: Why the Poorest Countries are Failing and What Can Be Done about It. New York: Oxford University Press. Collier, Pau l, V.L. Elliott, Håvard Hegre, Anke Hoeffler, Marta Reynal - Querol, and Nicholas Sambanis. 2003. Breaking the Conflict Trap: Civil War and Development Policy . Washington: The World Bank. Oxford Economic Papers 50(4): 563 - 573 Oxford Economic Papers 56(4): 563 - 595. Feasibil Oxford Economic Papers 61(1): 1 - 27. World Today 26(6): 259 - 266. 21, 2011. Accessed March 22, 201 5, from http://m.bbc.com/news/world - us - canada - 14190032 . Correlates of War Project. 2013. Militarized Interstate Disputes Version 4.1. Accessed March 22, 2015 from http://www.correlatesofwar.org/data - sets/MIDs . Correlates of War Project. 2010. National Material Capabilities Version 4.0. Accessed March 22, 2015, from http://www.correlatesofwar.org/data - sets/national - material - capabilities . The Journal of Peace Research 41(3): 303 - 320. International Organization 67(3): 609 - 627. Nigeria Since Independence: The First Twenty - Five Years. Volume V: Politics and Constitutions. Peter P. Ekeh, Patric k Dele Cole and Gabriel O. Olusanya, eds. Ibadan: Heinemann Educational Books. 106 - 122. American Political Science Review 104(3): 430 - 445. 262 De Rouen, Jr., Karl R., and Dav Journal of Peace Research 41(3): 303 320. Journal of Peace Research 39(4): 395 416. De S t. Jorre, John. 1972. The Nigerian Civil War . London: Hodder and Stoughton. Desch, Michael C. 1999. Civilian Control of the Military: The Changing Security Environment. Baltimore: Johns Hopkins University Press. and the Democratic State: Nigeria, 1950 - Comparative Studies in Society and History 25(3): 457 - 489. Diamond, Larry J. 1988. Class, Ethnicity, and Democracy in Nigeria: The Failure of the First Republic . Syracuse: Syracuse University Press. Diamo nd, Larry J. 1999. Developing Democracy: Toward Consolidation . Baltimore: Johns Hopkins University Press. Elaigwu, J. Isawa. 1986. Gowon: The Biography of a Soldier - Statesman . Ibadan: West Books Publisher Limited. Elbadawi, Ibrahim, and Nicholas Samban Journal of African Economies 9(3): 244 - 269. Ethnic Power Relations 3.01: Access to State Power of Politically Relevant Ethnic Groups, 1946 - 2010. Access ed March 22, 2015, from http://www.epr.ucla.edu . Pacific Affairs 30(4): 299 - 311. Falola, Tonyin, and Matthew M. Heaton. 2008. A History of Nigeria . Ca mbridge: Cambridge University Press. International Organization 49(3): 379 - 414. Institutions and Economic Performance , ed. Elhanen Helpman. Cambridge: Harvard University Press. 292 - American Political Science Review 101(1): 187 - 193. 263 Fearon, James D., and David D. La American Political Science Review 97(1): 75 - 90. Finer, Samuel E. [1962] 2009. The Man on Horseback: The Role of the Military in Politics. New Brunswick, NJ: Transaction Publishers. Fordham, Benjamin O International Studies Quarterly 49(1): 141 - 157. International Organization 53( 3): 567 - 587. and Political Instability: Polity Duration, 1800 - American Journal of Political Science 50(4): 893 - 908. Gelpi, Christopher, and Joseph M. Tenure, and the Targeting of Militarized Challenges, 1918 - Journal of Conflict Resolution 45(6): 794 - 817. Gibney, Mark, Linda Cornett, Reed Wood, and Peter Haschke. 2014. Political Terror Sca le 1976 - 2013. Accessed April 7, 2015, from http://www.politicalterrorscale.org . Gleditsch, Kristian Skrede. Journal of Conflict Resolution 46 (5): 712 - 24. Gleditsch Journal of Peace Research 44(3): 293 - 309. Accessed April 7, 2015, from http://privatewww.essex.ac.uk/~ksg/exptradedata.html . Gleditsch, Kristian Skrede, Han Dorussen, Nils Metternich, and Andrea Ruggeri. 2010. Journal of Conflict Resolution 52(4): 479 - 506. Gleditsch, Nils Petter, Peter Wallensteen, Mikael Eriksson, Margareta Sollenberg, and Håvard - Journal of Peace Research 39(5): 615 - 637. 264 Handbook of War Studies III , ed. Manus Midlarsky. Ann Arbor: University of Michigan Press. 155 - 192. Journal of Conflict Resolution 44(5): 555 - 579. Goemans, HE. 20 Journal of Conflict Resolution 52(6): 771 - 794. Goldstein, Joshua S. 2011. Winning the War on War: The Decline of Armed Conflict Worldwide . New York: Dutton. Gotowicki, Stephen H. 1997. Washington, DC: Institute for National Strategic Studies, National Defense University. Gurr, Ted Robert, Keith Jaggers, and Will H. Moore. 1989. Polity II Codebook . Boulder, CO: University of Colora do. Turkish Studies 6(2): 215 - 231. - Independence Burmese Journal of Southeast Asi an Studies 39(1): 149 - 177. Journal of Peace Research 47(5): 615 - 626. 30 Yea rs Predicted. November 22. Atlanta Journal and Constitution . 4A. Huntington, Samuel P. [1957] 1985. The Solider and the State: The Theory and Politics of Civil - Military Relations . Cambridge: Harvard University Press. Huntington, Samuel P. [1968] 2006. Political Order and Changing Society . New Haven: Yale University Press. Nigeria Since Independence: The First Twenty - Five Years. Volume IV: Government and Public Policy. Tekena N. Tamuno and J.A. Atanda, eds. Ibadan: Heinemann Educational Books. 65 - 83. American Political Science Review 72(4): 1262 - 1275. 265 Janowitz, Morris. 1964. The Professi onal Soldier: A Social and Political Portrait . New York: Free Press. 1957 - American Sociological Review 55(6): 861 - 875. Jenkins, J. Craig, and Augustine International Studies Quarterly 36(3): 271 - 291. Christian Science Monitor . September 12, 2013. Accessed Ma rch 22, 2015, from http://www.csmonitor.com/World/Asia - Pacific/2013/0912/Is - the - rebel - attack - in - the - Philippines - a - publicity - p loy - video . Publius 15(2): 71 - 100. Man 27(4): 873 - 879. King, Gary, Robert O. Keohane, and Sidney Verba. 1994. Designing Social Inquiry: Scientific Inference in Qualitative Research . Princeton: Princeton University Press. Kirk - - Transition without End: Nigerian Politics and Civil Society under Bagangida , ed. Larry Diamond, Anthony Kirk - Greene, and Oyeleye Oyediran. Boulder: Lynne Rienner Publishers. 31 - 53. Coups in Postcolonial Africa, 1957 - American Journal of Sociology 99(1): 126 - 163. International Studies Quarterly 44(4): 667 - 693. Kuran, Timur. 1989. Public Choice 61(1): 41 - 74. - Policy Preference Link in Sub - Comparative Political Studies 46( 5): 574 - 602. Lijphart, Arend. 1977. Democracy in Plural Societies: A Comparative Exploration . New Haven: Yale University Press. 266 Lijphart, Arend. 1999. Patterns of Democracy: Government Forms and Performance in Thirty - Six Countries . New Haven: Yale Univ ersity Press. Contemporary Southeast Asia 5(4): 403 - 450. World Politics 42(2): 1 51 - 183. Long, J. Scott. 1997. Regression Models for Categorical and Limited Dependent Variables . Thousand Oaks, CA: Sage Publications. Pacific Affairs 40(1 - 2): 113 - 1 23. Luckham, Robin. 1971. The Nigerian Military: A Sociological Analysis of Authority and Revolt, 1960 - 1967. Cambridge: Cambridge University Press. Madiebo, Alexander A. 1980. The Nigerian Revolution and the Biafran War . Enugu: Fourth Dimension Publish ers. Marshall, Monty G., Ted Robert Gurr, and Keith Jaggers. 2014. Polity IV Project: Political Regime Characteristics and Transitions 1800 - 2013. Polity IV Dataset version 2013 (p4v2013). Center for Systemic Peace. Accessed April 7, 2015, from http://www.systemicpeace.org/polityproject.html . Handbook of War Studies III: The Intrastate Dimension , ed. Manus I. Mi dlarsky. Ann Arbor: University of Michigan Press. 63 - 99. Australian Journal of International Affairs 62(3): 298 - 314. McGo 1956 - 2001: Frequency, Trends, and Journal of Modern African Studies 41(3): 339 - 370. Survival: Global Politics and Policy 11(9): 287 - 293. Asian Survey 48(6): 1018 - 1037. Montalvo, José G., and Marta Reynal - American Economic Review 95(3): 796 - 816. 267 Mwakikag Nigeria: Current Issues and Historical Background , ed. Marin P. Mathews. Hauppauge, NY: Nova Science Publishers. 17 - 50. - Turkish Studies 1(1): 107 - 127. Nationa l Consortium for the Study of Terrorism and Responses to Terrorism (START). 2013. Global Terrorism Database. Accessed April 7, 2015, from http://www.start.umd.edu/gtd . Nigeria 1965: Crisis and Criticism : Selec tions from Nigerian Opinion. 1966. Ibadan: Ibadan University Press. Comparative Politics 24(3): 293 - 316. Minorities and the State in Africa , ed. Michael U. Mbanaso, and Chima J. Kurieh. Amherst, NY: Cambria Press. Nigeria Since Independence: The First Twenty - Five Years. Volume IV: Gov ernment and Public Policy. Tekena N. Tamuno and J.A. Atanda, eds. Ibadan: Heinemann Educational Books. 30 - 43. Nigeria Since Independence: The First Twenty - Five Years. Volume IV: Government and Public Policy. Tekena N. Tamuno and J.A. Atanda, eds. Ibadan: Heinemann Educational Books. 44 - 64. - Perspectives on the Nigerian Civil War . Siyan Oyeweso, ed. Lagos: OAP P ublications. 95 - 116. Dataset, 2002 - Conflict Management and Peace Science . Panter - Nigerian Politics and Military Rule: Prelude to the Civil War . S.K. Panter - Brick, ed. London: Athlone Press. 14 - 57. Peters, Jimi. 1997. The Nigerian Military and the State. London: Tauris Academic Studies. - Proofing and Military Effectiveness in Interstate Wars, 1967 - 99. Conflict Management and Peace Science 28(4): 331 - 350. 268 - Proof ing? On the Relationship between Regime Type and Civil - Foreign Policy Analysis 8(4): 355 - 372. Pinker, Steven. 2011. The Better Angels of Our Time: Why Violence Has Declined . New York: Viking. Posen, Barry R. 1984. The Sources of Mi litary Doctrine: France, Britain, and Germany between the World Wars. Ithaca: Cornell University Press. Post, Kenneth, and Michael Vickers. 1973. Structure and Conflict in Nigeria 1960 - 1966. Madison: University of Wisconsin Press. Powell, Jonathan. 201 Journal of Conflict Resolution 56(6): 1017 - 1040. Journal of Peace Research 48(2): 249 - 259. International Interactions 33(2): 167 - 193. Lecture Notes on Generali zed Linear Models. Princeton University. Accessed March 20, 2015, from http://data.princeton.edu/wws509/notes . World Politics 63(2): 300 - 346. State Failure and State Weakness in a Time of Terror, ed. Robert I. Rotberg. Washington: Brookings Institution Press. 1 - 2 5. Political Geography 27(2): 761 - 782. ional Rebels: Neighboring States as Sanctuary for Rebel World Politics 59(2): 217 - 242. Journal of Politics 70(1): 54 - 66. and Non - Ethnic Civil Wars Have the Same Causes? A Journal of Conflict Resolution 45(3): 259 - 282. 269 Public Choice 98(3 - 4): 337 - 367. Schnapper, Domi Rethinking Ethnicity: Majority Groups and Dominant Minorities, ed. Eric P. Kaufmann. London: Routledge. 102 - 115. ocial Movement Mobilization Social Problems 46(3): 355 - 375. New York Times . June 30, 2014. Accessed March 22, 2015 from http://www.nytimes.com/2014/07/01/world/asia/Waziristan - Region - of - Pakistan - Has - Long - History - of - Rebellion.html . Sel - Modern Asian Studies 20(3): 483 - 507. Far Eastern Survey 28(7): 97 - 105. Journal of Southeast Asian Studies 21(1): 114 - 134. States, 1816 - International Interactions 14(2): 115 - 132. Singer, J. David, Stuart Bremer, and John Stu and War, 1820 - Peace, War and Numbers , ed. Bruce Russett. Beverly Hills, CA: Sage. 19 - 48. Siollun, Max. 2009. - 1976). New York: Algora Publishing. Government and Opposition 2(4): 524 - 539. Creating Peace in Sri Lanka: Civil War and Reconciliation, ed. Robert I. Rotberg . Washington: Brookings Institution Press. 17 - 40. The Guardian . http://www. theguardian.com/world/2013/jan/30/zimbabwean - government - bank - balance - down . 270 Stinnett, Douglas M., Jaroslav Tir, Philip Schafer, Paul F. Diehl, and Charles Gochman. 2002. Conflict Manageme nt and Peace Science 19(2): 58 - 66. St rem lau, John J. 1977. The International Politics of the Nigerian Civil War, 1967 - 1970. Princeton: Princeton University Press. - Economics and Politics 12(3): 205 - 223. American Journal of Political Science 53(2): 477 - 494. Journal of Modern Afri can Studies 8(4): 563 - 584. - Nigeria Since Independence: The First Twenty - Five Years. Volume VI: The Civil War Years. Tekena N. Tamuno and Samson C. Ukpabi, eds. Ibadan: Heinemann Educational Books. 1 - 84. Journal of Peace Research 49(2): 273 - 287. cy: Leading the Myanmar State Contemporary Southeast Asia 15(1): 24 - 63. - Journal of Peace Research 51(4): 541 - 554. Thompson, William R., and David Dreyer. 2011 . Handbook of International Rivalries . Washington, DC: CQ Press. Journal of Conflict Resolution 50(6): 937 - 961. - Modern Asian Studies 20(3): 461 - 481. Foreign Affairs 37(2): 317 - 327. Review of Politics 25(3): 309 - 328. 271 Trager, Frank N. 1966. Burma: From Kingdom to Republic: A Historical and Political Analysis . New York: Frederick A. Praeger, Publishers. Ukpabi, Sam C. 1986. Strands in Nigerian Military Hi story . Kaduna: Nigerian Defence Academy. Nigeria Since Independence: The First Twenty - Five Years. Volume VI: The Civil War Years. Tekena N. Tamuno and Samson C. Ukpabi, eds. Ibadan: Heine mann Educational Books. 106 - 124. - Journal of Peach Research 37(2): 251 - 265. Journal of Conflict Resolution 52(3): 401 - 425. Wallinsky, Louis J. 1965 - Pacific Affairs 38(3 - 4): 269 - 281. Journal of Peace Research 41( 3): 371 - 388. p - Journal of Peace Research 47(4): 363 - 375. Government, 1960 - In Nigeria Since Independence: The First Twenty - Five Years. Volume VI: The Civil War Years. Tekena N. Tamuno and Samson C. Ukpabi, eds. Ibadan: Heinemann Educational Books. 149 - 161. Wimmer, Andreas, Lars - Erik Cederman, and Brian M American Sociological Review 74(2): 316 - 337. Win, Sein. 1959. Split Story: An Account of Recent Political Upheaval in Burma with Emphasis on AFPFL. Rangoo n: The Guardian, Ltd. New York Times . 1A. American Journal of Political Sc ience 51(4): 772 - 788. World Bank. 2002. World Development Indicators . Washington: World Bank. World Bank. 2009. World Development Indicators . Washington: World Bank.