Rebel Financing in Civil Wars: A Quantitative Analysis with. Three Case Studies of Civil Wars in the Philippines. Matthew A.

Transcription

1 Rebel Financing in Civil Wars: A Quantitative Analysis with Three Case Studies of Civil Wars in the Philippines by Matthew A. Testerman Submitted in Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy Supervised by Professor Gretchen Helmke Department of Political Science Arts, Sciences and Engineering School of Arts and Sciences University of Rochester Rochester, New York 2012

2 For Jennifer and the Boys. ii

3 iii Curriculum Vitae The author was born in Chicopee, Massachusetts in He attended the United States Naval Academy from 1989 to 1993, and graduated with the Bachelor of Science degree, with distinction, in He received the Master of Arts degree from the University of Maryland in He was a military instructor of political science at the United States Naval Academy from 1999 to He came to the University of Rochester in the Fall of 2008 and began graduate studies in Political Science. He received the United States Navy Permanent Military Professor fellowship for doctoral studies from 2008 to He pursued his research in comparative politics and international relations under the direction of Professor Gretchen Helmke and received the Master of Arts degree from the University of Rochester in 2011.

4 iv Acknowledgements This dissertation was completed with the assistance and support of many people, only a few of whom I am able to mention in this brief space. Gretchen Helmke has been a source of sage counsel and wisdom in this process and throughout my entire time at the University of Rochester. As well, I have relied extensively on comments, advice, and encouragement from Hein Goemans and Bethany Lacina. It is not excessive to thank the entire department of political science because this close-knit faculty and staff are an integral part of what makes this program unique and challenging for graduate students. I am also indebted to those who have guided me throughout my militaryacademic career. Ellie Malone, Arthur Rachwald, Priscilla Zotti, and Linda Hull have been encouraging, realistic, and supportive in assisting my endeavors to become a more permanent member of the U.S. Naval Academy faculty. I am also grateful for two leaders who helped me at the beginning of this long professional journey and near its end, Captain Rob Leeds and General Mark Welsh. Both were inspirational and integral to me arriving at this juncture in my career. In confronting the challenges of the Rochester program, I have been in good company and in constant awe of my graduate student cohort. Brenton Kenkel, Hyesung Kim, Miguel Rueda, and Kristin Rulison are outstanding scholars and good people on whom I have relied repeatedly. I owe a tremendous debt of gratitude, of devotion, of love, to my wife, Jennifer - a debt I can never repay. She remains my touchstone, my best friend, and my reality check in every one of life s challenges. Finally, I must thank my boys, Dean, Henry, Joshua, and Sam, who constantly give me perspective and bless me with their love.

5 v Abstract This dissertation investigates why rebels decide to continue fighting costly wars instead of seeking the benefits of peaceful settlement. It explores the possibility that the means rebels use to exploit economic opportunities during civil war makes continued war more likely. The central argument is that civil wars are longer when economic opportunities offset rebel costs of war. Beginning with a basic model of bargaining and costly war, the theoretical framework describes how opportunity for rebels to profit during war reduces the potential range of bargaining settlements, thereby making continued war more likely. As an extension of this finding, permanent and intermittent means of financing rebellion are examined through a theory of warlords and bandits. Longer wars are those in which rebels pursue economic opportunities through regularized means of financing in the manner of a stationary bandit. A quantitative, cross-country analysis of civil wars is used to test hypotheses derived from this theory. An original dataset of civil wars since 1946 catalogs four principal sources of rebel economic opportunity: looting, taxes, natural resource exploitation, and external support. Civil wars are longer when rebel groups impose a system of taxes and when rebel groups receive external support. Next, three case studies of civil wars in the Philippines are used to further develop the theory and investigate the means of rebel financing. In the Huk rebellion ( ), rebels acted as roving bandits by looting to gather resources needed to fight the war. The Moro National Liberation Front took on the role of Robin Hood in the Moro war ( ), relying on external support and acting as the primary conduit of such aid within the southern Philippines. In the war with the Communist Party of the Philippines and the New People s Army ( ), rebels established a system of taxes and installed shadow governments to provide public goods and protection. The case studies thus not only confirm the phenomenal regularity of rebel financing, but also confirm the postulated causal mechanisms of the stationary bandit theory as applied to rebel groups in a civil war.

6 vi Contents 1 Introduction The Central Question The Central Argument Why is this important? Literature Review Economic Opportunity Costs State Capacity Ethnicity Grievance Plan for the Dissertation Conclusion Theory Introduction Bargaining The Costs and Benefits of War The Bargaining Model Bandits Stationary versus Roving Bandit Profits Rebel Financing Hypotheses Endogeneity Conclusion Quantitative Analysis Introduction Data Description Defining Civil War The Set of Civil Wars Under Consideration

7 vii Sources of Rebel Funding Rebel Financing in the Data Model Specification Univariate Analysis of Primary Variables of Interest Potential Additional Covariates Final Model Specification Results Robustness Checks Counter-Examples Conclusion The Huk Rebellion Introduction A History Of Resistance The Huks Early Violence, Civil War, Reform of the AFP The HMB/Huks Civil War Huks and Theory of Civil War Duration Huks as Roving Bandits Alternative Explanations Conclusion The Moro War Introduction A Brief History of Muslims in the Philippines Pre-colonial History Spain and the Moros US and Moros Republic of the Philippines and Moros The Jabidah Massacre as Watershed Event Financing the Philippine-Moro War Start of the Civil War The MNLF MNLF as Robin Hood Civil War End of the War Alternative Explanations

8 viii 5.7 Conclusion Civil War with CPP-NPA Introduction Communism in the Philippines Jose Sison and the CPP Rectification Start of the War Financing the War Alone and Unafraid Stationary Bandits Expanding the Tax Base End of the War Alternative Explanations Military Ineffectiveness Sustained Grievances Conclusion Conclusion Theory Findings Future Extensions State Formation Role of the International System Other Aspects of Economic Opportunity Bibliography 193 Appendix A 210

9 ix List of Tables 1.1 Philippine Civil Wars: duration and source of rebel financing Median Duration of Intrastate Wars COW conflicts omitted from analysis Observations of Rebel Financing Conflicts outside revenue construct sorted by duration Tests of Equality of Survivor Function Univariate Cox Test Results Cox Proportional Hazards Model without Controls Civil Wars by Region Cox Proportional Hazards Model of Civil War Duration Grambsch & Therneu Test of Proportional Hazards Assumption Cox Model for Original COW Data Civil Wars with Durations Less than One Calendar Month Outliers by War Number and War Name Short Wars with Tax and External Support

10 x List of Figures 2.1 Bargaining Range Reduced Bargaining Range Collapsed Bargaining Range Survival Probability Plot for Civil Wars, Histogram of Civil War Duration Estimated Survivor Function for all Civil Wars Survivor Functions for Civil Wars by Source of Funding Comparing Distribution of Civil War Durations in Eastern Europe with those in the rest of the world Survivor Functions for Civil Wars by Region Survivor Functions for Civil Wars during Cold War Plot of Scaled Schoenfeld Residuals Cox-Snell Residual Plot with 45 reference line Influential Observations Map of the Philippines from CIA World Fact Book Huklandia map from Greenberg 1987, Rebel Strength in Thousands derived from Greenberg Map of Southern Philippines from philippinemaps.ph Map showing location of Sabah from bbec.sabah.gov.my Map of Cotabato from McKenna (1998) Map showing Isabela province from wikipilipinas.org

11 1 Chapter 1 Introduction 1.1 The Central Question The rebel decision to keep fighting is the focus of this analysis of how civil wars endure. In internal wars, the government is generally understood to have the advantage in the use of violence against weak rebel challengers. The government armed forces are organized and often better equipped to fight the war (Wagner 1993, 254). The state, as the incumbent player, is a force in being... the insurgents, by contrast, are a force in development (McCormick, Horton and Harrison 2007, 322). This starting military disadvantage for the rebel groups draws attention to their decision to continue the fight in the face of low odds of achieving a military victory. The statistics on civil war outcomes confirms the poor chances confronting rebels. The most likely outcome of a civil war is a state victory. In contrast, only 12% of civil wars since 1945 have ended in a victory for the rebels. This is only a slightly higher percentage than that for civil wars whose outcome is unknown for lack of data (McCormick, Horton and Harrison 2007, 324).

12 Section The initial decision to fight could be based on misperceptions of the rebel group s chances for victory. After several years of fighting, though, it is difficult to sustain the belief that belligerents lack accurate information regarding the enemy s capabilities or resolve (Fearon 2004, 290). As a war continues, a rebel victory becomes even less likely (Brandt et al. 2008, 416). The most probable outcome from a prolonged war is that the fighting will end without a peace deal, or a clear winner (Cunningham, Gleditsch and Salehyan 2009, 588-9). It seems unlikely that rebels are eager for defeat, or that rebels are desirous of incurring the costs of conflict without a result. Why do rebels decide to continue fighting in such a situation instead of seeking the benefits of a peaceful settlement, and avoiding the costs of continued war? In this dissertation, I explore the possibility that the means by which rebels exploit economic opportunities during civil war changes their decision-making calculus, making continued war more likely. 1.2 The Central Argument The central argument of this dissertation is that civil wars are longer when there are economic opportunities that offset the rebels cost of war. Beginning with a basic model of bargaining and costly war, the theoretical framework describes how the opportunity for rebels to profit during war reduces the potential range of bargaining settlements. Ceteris paribus, the reduced range of peaceful alternatives makes the rebel choice of continuing the war a more likely outcome. As an extension of this finding, permanent and intermittent means of financing rebellion are examined through

13 Section the lens of a theory of warlords and stationary bandits. A regularized system of taxation is then shown to be the most lucrative method for rebels to extract resources from areas under their control. Such a permanent means of financing rebellion provides the rebels with benefits that exceed those available from intermittent sources like looting, or plunder. Hence, a civil war will be of greater duration when rebels pursue economic opportunities in war through taxes, taking on this theoretical role of a stationary bandit. Other means of rebel financing are looting, exploitation of natural resources, and receipt of external support. Through an accounting of recorded observations of rebel behavior in this regard, this analysis finds that taxes and external support are positively correlated with civil war duration. Case studies are then used to better develop the causal mechanisms at work and provide additional evidence in support of the central argument that a permanent means of rebel financing contributes to the increased duration of civil war. 1.3 Why is this important? Civil war has been the most common type of warfare since the end of World War II. In the latest catalog of wars of all types by the Correlates of War project, intrastate wars are the largest category and, within this group, civil wars are the largest subcategory (Wayman and Sarkees 2010). Civil wars, on average, are also more than twice as long as interstate wars (Brandt et al. 2008, 416). Despite the overall downward trend in the number of new conflicts and in the battle-related deaths resultant from conflicts in general

14 Section since the end of the Cold War (Human Security Report Project 2005), their longer durations make civil wars more deadly (Brandt et al. 2008, 417), and this intensity motivates the perception that civil wars are also the most brutal type of conflict (DeRouen Jr., Bercovitch and Wei 2009). For these reasons, the study of civil war duration is an important undertaking as it pertains to understanding the factors that perpetuate violent conflict. In many cases, international attention and pressure for intervention to limit the violence or support the rebel cause often only occurs after the on-set of civil war. While there is certainly merit in research that attempts to mitigate the outbreak of war, there is an equal need to understand what sources of leverage exist to intervene in civil wars once they have begun. By presenting both a cross-national study and a series of case studies, this analysis seeks to address these needs and attempts to answer this important question regarding the duration of civil wars. 1.4 Literature Review Research on civil war duration is a small sub-set of the expansive literature exploring civil war in general. Much of the recent work in this area has been devoted to the effect of third party interventions on the duration of conflicts, including neutral peacekeeping, counter-insurgency, occupation, and biased interventions (see, for example, Lyall and Wilson 2009; Brandt, et al. 2008; Fortna 2004; Regan 2002; and Balch-Lindsay and Enterline 2000). Indeed, in the Box-Steffensmeier and Jones (2004) primer on the methodology of survival analysis in political science, third-party intervention in civil wars

15 Section was a main example. The examination of civil war duration, however, is not an independent enterprise. The study of war on-set, termination, and duration are deeply intertwined. The question of what factors influence the length of a war is inextricably linked to the questions of war causation and peace. Blainey (1988) described their inter-relationship explaining that to concentrate one s analysis on war duration is simply to focus attention on the kind of influence which terminate wars; to survey what terminates wars is of course to examine the influence that bring peace; and to examine the causes of peace is essentially to turn the causes of war upside down (Blainey 1988, viii). The preponderance of the weight of effort in civil war studies, however, has been focused in the area of war on-set. Thus, while much of the research that will be discussed here addresses civil war on-set, these works are also relevant to the study of civil war duration. There are four general conceptual approaches to studying civil war: from the perspective of economic opportunity and costs of war; the influence of state capacity; war as derivative of ethnic, religious, or other divisions; and conflict as the violent manifestation of grievances. While much has been made of what is termed the greed versus grievance debate, most recent publications have delved into economic factors, state capacity, and ethnicity, with only a few studies bringing empirical methods to address grievances. Despite wide variation in empirical results, the main debates in civil war research are about the causal mechanisms at work in connecting these covariates to on-set and duration. Before the literature on civil war is reviewed, I briefly discuss two of

16 Section the field s key findings in regards to civil war s correlation with poverty and population. The purpose in doing so is, paradoxically, to highlight the primary weakness in civil war research the considerable distance between postulated causal mechanisms and their observable implications. The measures most often used as correlates of civil war for poverty and population can serve as proxies for many things. This dissertation, by coding documented observations of actual rebel behavior, will at least partially address this shortcoming. The two enduring findings in the empirical literature on civil wars are that poverty and population size are linked to the incidence, on-set, and duration of civil war. This almost unique source of agreement in studies of civil wars shows that risk of on-set is inversely correlated with poverty and positively correlated with population (Hegre and Sambanis 2006, 509). These results appear repeatedly in the literature as proxies for a wide range of phenomena (Kalyvas 2007, 422). Poverty can proxy for grievances of those affected. Hirshleifer (2001) theorized that poverty makes people more politically bellicose while those with recent economic gains are more accommodating (Hirshleifer 2001, 17). Poverty can be indicative of poor government capacity, or represent political and economic corruption in the state (Fearon and Laitin 2003, 80). Poverty can also be thought of as lowering the opportunity costs for individuals to join a rebellion (Kalyvas 2007, 419). Likewise, the correlation with population could be resultant from increased demands on state capacity created by larger states with dispersed populations. It could also stem from the unique demands of large states with concentrated urban populations (Hegre and Sambanis 2006). The wide variation in the

17 Section substantive interpretation of these two findings is illustrative of a recurring problem that confronts different aspects of the civil war research. These problems are reviewed in the subsequent section Economic Opportunity Costs A natural starting point for discussing economic opportunity costs in war is the work of Collier and various co-authors. Their research formed the foundation of the greed-side of the current greed versus grievances debate. As well, Collier contested the idea that civil war was simply another stop on the continuum of political conflict. Collier argued that a rebellion could not be thought of simply as the evolution of political protest. Civil war requires a quantum difference in financial resources (Collier, Hoeffler and Rohner 2008, 6). The critical component is the means to rebel that is, the means to acquire the arms, ammunition, supplies, food, and patronage. Civil war is then about the economics of rebel motivations and rebel opportunities. Grievances, or motivations, are based on desire for economic redress whereas opportunity is rooted in access to financial resources necessary to secure the means of rebellion. From such a perspective, it is then useful to begin with an examination of people s economic desires and their relation to the use of force. Hirshleifer (2001) theorized that predation and conflict during war time are the natural complement to the production and exchange of peace time economies (Hirshleifer 2001, 1). He postulated that the desire for economic gain motivates economic activity by peaceful and violent means. The challenger to the state resorts to violence as a means of leveling a political and

18 Section economic playing field that is dominated by the state. As a rational economic actor, a rebel s decision to fight the state is predicated on perceptions of the likelihood of outcomes and their associated payoffs. The rebels must then anticipate economic gain either as the end result of the fighting, or as a derivative of the fighting itself. The argument that rebels see potential gain from war is countered by Fearon (1995) who laid the groundwork for a widely held assumption that war is costly. The resources and effort exerted to wage war, in addition to the physical cost in the destruction of life and property, incur a cost by the belligerents sufficient to make fighting inefficient. Collier (2000) estimated that civil war reduced a state s GDP by 2.2%, on average, for each year of conflict (Collier 2000, 101). War also exacerbated poverty and inequality (Blattman and Miguel 2009, 56). Wood (2000) cited assets destroyed, costly strikes, increasing security costs, investment suspended, and taxes increased as costs incurred by economic elites during civil war (Wood 2000, 15). In the case of El Salvador, Wood found that foreign investment losses approached a value of approximately 11.4% of GDP in a single year of fighting based on the risk of expropriation, loss of coffee lands, and the nationalization of industries (Wood 2000, 54). Keen (1998) identified four general categories of costs from civil war. Civil war deepens poverty, damages agriculture, damages industry, and interferes with the provision of social services as the state spends more on the military (Keen 1998, 46). If war is costly and the decision to fight is based on rational expectations, then there must be some mechanism inhibiting the accurate assessment of potential outcomes (Wagner 1993, 242). Fearon (1995, 2004) proposed the

19 Section existence of barriers to more efficient and peaceful alternatives to conflict, including issues of credible commitment, indivisible stakes, and incentives to misrepresent (Fearon 1995). Schelling (1960) and Walter (1997) amongst many others also focused on the factors that would make rational the prima facia irrational decision to fight. The assumption that war is always costly was challenged by Collier based on potential economic opportunities presented during conflict to belligerents, both individually and as a group. As Martinez (2000) noted in his study of the Algerian civil war, war, seen as a destructive phenomenon, dramatic in human terms and economically unproductive, can turn out to be a profitable activity (Martinez 2000, 16). To effect economic gain requires a change to the existing peacetime economy that enables the rebels to have access to, control of, and profit from economic resources. The challenge to the assumption that war could only be costly provided considerable theoretical and empirical space for others to conduct research about the potential sources of profit. Collier examined how conflict could be economically feasible by examining changes during war such as reduced time horizons for actors, increased criminality, disrupted markets, and increased predation (Collier 2000, 101-2). Other researchers viewed this economic change less as a disruption, or breakdown, of the system, but rather as the emergence of an alternative economic system revolving around profit, power, and protection (Keen 1998, 22). Keen suggested that war may be a continuation of economics by other means centered on violent private accumulation and involving the legitimization of criminal economic activity, scarcity, increased risk, and distor-

20 Section tional spending in the form of outside aid and increased military spending (Keen 1998, 29-30). Martinez described how war transformed the economy of Algeria in the civil war with the Algerian Islamic Front. Insurgents sabotaged aspects of the economy beneficial to non-muslims who supported the government, including public transportation, pharmaceutical plants, and cement manufacturing facilities. In destroying them, control of these resources was redistributed to rebel-affiliated entrepreneurs who gained personally from these ventures and were thus better able to provide money to the rebels (Martinez 2000, ). Wood (2000), using El Salvador s civil war as a case study, also found that insurgent social movements could provide economic and political opportunities as incentives for support during war (Wood 2000, 12). Thus, an alternate economic system can emerge in war that enables rebel groups to benefit from people and businesses by providing protection and other public goods. These potential benefits from participation in civil war can then be compared to the economic welfare of potential participants to examine the collective action problem. A reduction in an individual s opportunity costs occurs when the expected utility of not fighting is minimized through lowered expectations of peacetime gains. As mentioned previously, various studies have examined economic factors that characterize the state, or regions within the state, prior to the outbreak of civil war. Among these are analyses of economic growth in terms of per capita income, gross domestic product, or changes in either (see, for example, Collier and Hoeffler 2004; Hegre 2003). The logic is straight forward arguing that in poorer states, where individuals have limited earnings potential, a person is more likely to join in a rebellion

21 Section for the opportunity it might provide. Conversely, countries with higher levels of development should be less susceptible to the occurrence of civil war as the opportunity cost to participants is higher (Sambanis 2004, 836). Similarly, measures of social welfare have been demonstrated to relate to rebel participation. Social indicators, such as infant mortality, could proxy for future economic expectations (Goldstone, et. al. 2005). Thus, if measures of health and longevity are low, or declining, then individuals may discount the future and more willingly engage in risky and violent rebellion. Likewise, if educational opportunity, or social mobility are low, then individuals opportunity costs for rebeling against the state would be low (Collier and Hoeffler 2004). Again, however, an important critique of such studies is the difficulty of differentiating the effects of these outcomes from poor governance that creates conditions for them to take root, or from poverty more generally. Collier also inferred the economic agenda of rebel groups by evaluating the characteristics of the state in which the rebellion took place (Collier 2000, 93). In particular, Collier examined cross-national measures for the export of primary commodities thought to be easily exploitable by rebels. Perhaps the most oft cited examples of this phenomena are African civil wars in Liberia, Angola, and Sierra Leone where rebels and the government vied for control of diamond mines (see, for example, Outram 2004; Lebillon 2004; and Keen 1998). Collier and his co-authors findings have been subject to substantial criticism and refinement. Criticism has focused on shortcomings in the measure used as a proxy for rebel economic opportunity and also on the range of potential explanatory mechanisms. The linkage between civil

22 Section war and the basket of goods contained in the primary commodities measure could also be through their vulnerability to external shocks, the unequal distribution of land, or forced migration that derive from a state economy excessively dependent on natural resources (Kalyvas 2007, 421). Extensions of this work have attempted to answer such critics with more narrowly focused research. An entire branch of civil war research has been spawned that examines the correlation between civil war and specific natural resource endowments. Lebillon (2004, 2005, 2008) analyzed the microfoundations of how resources can be more, or less lootable for rebel groups. He found that lootable resources are those requiring minimal infrastructure and institutions to harvest, transport, sell, and tax (Lebillon 2005, 30). Natural resources with the most potential for rebel exploitation are diamonds, other gemstones, forests, and agricultural commodities such as cattle and coffee. These are in contrast to resources that require capital intensive operations such as petroleum, or mineral resources including copper and gold (Lebillon 2005, 32-37). The natural resource findings, however, are often contradictory. For example, Fearon (2005) and Lujala (2010) found that oil, or hydrocarbon, based resources are among the more important variables in predicting conflict, but disagree on the underlying causal mechanisms. Lujala s findings supported the mechanism of rebel economic opportunity and even argued that the potential benefit generated by the mere presence of accessible, on-shore oil production increased the risk of civil war on-set by 50% and significantly prolonged the conflict (Lujala 2010, 16). Fearon also concluded that there was a relationship between oil exports and civil war duration, but

23 Section attributed this to weak states and not rebel economic opportunities (Fearon 2005, 503) State Capacity Instead of the opportunity, or incentive, to rebel, another research agenda for civil war addresses the ability to rebel. More specifically, these studies assess the state s capacity to maintain a monopoly of the use of violence within its borders. Even if motivated to attack the state, civil war cannot result unless the state is sufficiently weak that it cannot prevent rebels from organizing and equipping themselves for rebellion. The most direct measure of state capacity for violence is, of course, the size of the security apparatus. This could include the police force, national constabulary, and the military. The size can be evaluated in terms of the number of personnel, or alternatively by the amount of money allocated to security as a fraction of the state s budget, or GDP. Hegre and Sambanis (2006) find that the size of a state s military the most readily available cross-national measure either as thousands of personnel, or as a percentage of the population, has a robust negative correlation to risk of civil war on-set (Hegre and Sambanis 2006, 529). More general measures of state financial and administrative capabilities, proxied by per capita income and GDP, are also thought to be correlated with lower risk of violent rebellion. In addition to improved military capacity, an effectively governed state can provide information on those contemplating armed rebellion and provide alternative means of confronting them short of war (Fearon 2004). Wealthier states are also thought to have a more robust infrastructure, enabling the state to reach potential threats at the pe-

24 Section riphery of the state and in densely populated urban environments (Fearon and Laitin 2003). Again, though, GDP and other income measures can proxy for a range of mechanisms that are difficult to unbundle (Hegre and Sambanis 2006, 524). Exclusively linking such variables to state capacity is thus problematic (Fearon and Laitin 2003, 83). An alternative means of evaluating the ability of groups to rebel is by measuring conditions that favor insurgency. Such conditions would lend advantage to the insurgents as a relatively weak force attacking a comparatively strong state. Terrain features, vegetation, climate, distance from the center of the state, and limited infrastructure reduce the effectiveness of conventional military forces (Buhaug and Gleditsch 2008, 557). Such physical and geographic features can also mask a rebel group s location, size, strength, and base of support (Fearon and Laitin 2003, 79-80). Buhaug, Gates, and Lujala (2009) evaluated a conflict s distance from the capital city and whether it was fought near state borders. They found both measures increased civil war duration. Fearon and Laitin (2003) introduced rough, or mountainous, terrain as being significantly correlated with risk of civil war. More mountainous regions provide areas in which rebels can hide from the government, train, and re-equip. Scott (2009) described such areas as ungovernable, or at least requiring greater government capacity to control their inhabitants. As an example, Chechen rebels used mountains to great effect in both of their civil wars with Russia to shield themselves from the overwhelming conventional force advantage of the Russian army (Smith 1998). Other studies have employed alternative specifications for rough terrain,

25 Section including average elevation of a country, average elevation of regions in conflict, and indicator variables for terrain type at the conflict site (see, for example, Lyall and Wilson 2009; Buhaug, Gates, and Lujala 2009). This, in part, answered Sambanis (2004) critique that in many instances civil wars are not fought in the mountainous regions for which a state is coded (Kalyvas 2007, 432). Beyond mountains, others have looked to correlate civil war on-set with the presence of jungles, forests, and other geographic, or climatic features that might aid insurgents. Kalyvas (2007), however, emphasized the multiple potential explanations for these empirical correlations. He commented that it is not clear whether these features are indicative of poor governance, inhibited state capacity, or other mechanisms Ethnicity The study of ethnic civil war peaked in the 1990 s after the conflicts in Yugoslavia and Rwanda. These civil wars appeared to break down easily along distinct ethnic lines Serbs versus Croats verus Bosniaks, Hutus versus Tutsis. Each group exhibited extreme brutality that could only be rationalized through a prism of ancient and primordial hatreds (Brubaker and Laitin 1998). Civil war researchers investigated a wide range of indicators for the presence of ethnic cleavages, or how groups might coalesce along cultural lines. Toft (2003) reduced ethnic conflict to being primarily driven by the desire to define favorable settlement patterns, that is, where groups live and whether they are concentrated in a homeland and a majority, or minority (Toft 2003, 2). She found that two-thirds of civil wars involved conflict derived from issues of ethnicity. Such wars were also observed to be

26 Section more violent and more intense than other civil wars. As well, ethnic wars were more susceptible to spillover effects involving ethnic kin in neighboring states (Toft 2003, 2-3). Fearon and Laitin (2003) disagreed. Their empirical analysis using various measures for ethnic, linguistic, and religious fractionalization found no statistically significant effects to the risk of civil war onset (Fearon and Laitin 2003, 84). Under the umbrella concept of ethnic fragmentation, Hegre and Sambanis (2006) evaluated a dozen alternative specifications for cultural cleavages and found none that were statistically significant in their correlation with civil war (Hegre and Sambanis 2006, 516,529). Counterintuitively, they did find that extreme measures of ethnic polarization are correlated with armed conflict at lower levels of intensity (i.e., conflicts with battle-related death thresholds of 25 instead of 1,000), but not with civil wars (Hegre and Sambanis 2006, 516). Taking stock of ethnicity as an alternative explanation for civil wars, Mueller (2000) concluded that ethnicity is important only as an ordering device, but not as the motivational force for rebellion (Mueller 2000, 62). The empirical findings undermine arguments that ethnic differences are primordial and serve to entice conflict by their mere presence. The salience of ethnic cleavage is more aptly construed as constructed. When politically useful to leaders, ethnic cleavages are made salient by political entrepreneurs for example, to organize a base of support in rebellion against the state (Posner 2004). Kalyvas concluded that ethnic cleavages are made meaningful because of conflict, but are not the cause of conflict (Kalyvas 2007, 419).

27 Section Grievance A final approach to the study of civil war is as violent conflict emerging from grievances. This area of conflict studies, described as presently unfashionable (Kalyvas 2006, 422), was pioneered by Gurr (1971) based on his relative deprivation theory of how rising, unmet expectations cause rebellion. Toft (2003) described the mechanisms in Gurr s theory as the formation of groups in reaction to declines in economic and political conditions followed by violent competition between groups over resources. Although a compelling story, Toft found inadequate means for evaluating individual and group perceptions of grievances sufficient to test the validity of the theory in explaining civil wars (Toft 2003, 5). Boix (2003) attempted to address this by analyzing economic inequality as a proxy for inter-group grievance. Other researchers have turned to evaluating political institutions as a measure of the state s ability to satiate grievances. Democracies are believed to provide non-violent means for remedying grievances and thus should be negatively correlated with civil war on-set. Other extensions of this hypothesis suggest that greater civil rights and less discrimination (again, both positively associated with democracy) should act to dampen grievances and reduce the risk of civil war (Fearon and Laitin 2003, 79). Additional variations of factors with potential impact on grievances include proportional representation and political decentralization, or federalism (Reynal-Querol 2002). The only persistent and robust finding in these analyses is that weak regimes partial democracies, anocracies, or illiberal democracies are more susceptible to civil war than either authoritarian regimes, or lib-

28 Section eral democracies. Even so, Fearon and Laitin (2003) found that the rate of onset of civil war in democracies is indistinguishable from the rate of onset in non-democracies (Fearon and Laitin 2003, 84). Similarly, Goldstone et. al. (2005) developed a global model of vulnerability to the onset of instability and found no variation in the risk of civil war amongst five, non-authoritarian regime types. Civil war sits at the cross-roads of comparative politics, international relations, and, increasingly, economics. Thus, the literature on civil war contains elements of each field and their theoretical and methodological perspectives. Each of these is, in turn, applied to the multiple phases of civil war including onset, duration, termination, post-conflict peace, and conflict recurrence. This review of civil war literature demonstrates the breadth of the field and the central challenge researchers face in substantively interpreting the empirical results. This dissertation adds to the economic opportunity literature by examining civil war duration and its correlation with the changed economic dynamics that are created by war. 1.5 Plan for the Dissertation This analysis is organized in six chapters. In the next chapter, a theory of civil war duration is developed based on rebel economic opportunity and means of financing. The opportunity to profit from war is shown to decrease the range of potential peaceful settlements and make violent conflict a more likely outcome of bargaining. Rebel economic opportunity is then shown to be maximized through behaviors of a stationary bandit who endeavors to

29 Section establish a regular and permanent means of extracting resources while also providing protection and other public goods to the people in the area under the bandit s control. To have such regular access to tax individuals and businesses requires that rebels organize the greatest capacity for violence in the area, displace the state monopoly of violence, and develop the means to sustain this violent capacity. The theory will thus show how longer wars are those in which rebels pursue economic opportunities through regularized means of financing in the manner of a stationary bandit. In chapter three, a quantitative, cross-country analysis of civil wars is undertaken to test the hypotheses derived from this theory. An original dataset is derived from a review of civil war histories and catalogs observations of four principal sources of economic opportunity in civil wars: looting, taxes, natural resource exploitation, and external support (money, arms and supplies, training, safe haven, direct military support). This large-n analysis is presented in five sections. Section 3.2 on data description provides the nuts and bolts of the definitional choices made in constructing the dataset and the operationalization of the four sources of rebel economic opportunity. Section 3.3 examines the Cox proportional hazards model and the methods used for accurately specifying the model. The results of the model are presented in section 3.4 alongside diagnostic methods for assessing fit, form, influential observations, and the proportional hazards assumption. The results are then subjected to robustness tests and tests of alternative specifications in section 3.5 and counter-examples are examined in section 3.6. To preview these findings, civil wars are longer when rebel groups impose a systematic tax in areas under their control and when rebel groups

30 Section receive external support. Looting also appears to be linked to increased duration, but the finding is tentative and sensitive to alternative specifications. Rebel exploitation of natural resources does not have a statistically significant correlation with duration. These results are robust to alternative specifications with other covariates of civil war duration, including regional effects, conflict neighborhoods, country specific effects, the effect of the Cold War, population size, and per capita income. In chapters four through six, I present three case studies of civil wars. The method of selecting these three cases follows from two principal guidelines, variation in the dependent variable (duration) and finding most similar cases. This is derivative of Flyvbjerg s (2006) recommended strategy for determining the effect of a particular variable using multiple cases. To do so, the cases should be identified as cases of maximum variation along the dimension of interest, but otherwise similar (Flyvbjerg 2006, 230). While it is relatively straight-forward to draw from the set of all civil wars a subset that includes wars of different lengths, it is a more difficult endeavor to find such a subset that is similar in most other regards. The case studies chosen for this analysis are all drawn from the Philippines from 1950 to The most significant deviation from Flyvbjerg s recommendation is that the three wars also differ in the type of rebel financing observed during the war (see table 1.1). An alternative case selection strategy would be to find cases of varying duration and consistent financing, then attempt to account for the dissimilarities of the contexts in which they were fought. By selecting wars fought within the Philippines, the variation along this second dimension of rebel financing is embraced as a means of investigating the different

31 Section Years Duration Natural External Fought (months) Looting Resources Taxes Support Huk Rebellion Moro War CPP-NPA War Table 1.1: Philippine Civil Wars: duration and source of rebel financing types of economic opportunities pursued by rebels, within a more controlled context of wars fought within the same country. Table 1.1 shows the duration of each of the three civil wars and the observed sources of rebel economic opportunity during the fighting. The CPP-NPA war of 1981 to 1992 was the longest war and the principal means of rebel financing was a tax system on individuals, families, landowners, and businesses. The Huk Rebellion was the shortest civil war and one in which the rebels were forced to rely on looting to gather money, food, arms, and other supplies to fight. The Moro War had a duration of nearly seven years and the rebels were sustained almost exclusively by external support. State level characteristics for each of these conflicts, such as political history, infrastructure, and state capacity, are held relatively constant. The wars are fought within the broad context of the Cold War, which has important implications for continued US assistance to government forces. The military technology of the rebels is largely static across all three wars. The wars all have roots in disputes over land reform. The wars are fought in many of the same regions and in predominantly rural settings with heavily forested/jungled and mountainous terrain. All three rebel groups are faced with similar problems in terms of the unique geography of the Philippines

32 Section as an archipelagic state, including the tremendous difficulty of movement between islands and the significant distance over water to other states. There are, however, a number of differences that need to be taken into account with each case. The government of the Philippines swings from a partial democracy from 1948 to 1971, to an autocracy during the era of martial law 1972 to 1981, and becomes a full democracy in 1986 when President Marcos is removed from power. The rebel groups also differ from one another, with two based on peasant and workers movements, and the third based on religious and cultural identity. The Huks and CPP-NPA wars were fought for control of the central government, while the Moros fought a war of secession. The leadership of the Moros and the CPP-NPA are both products of the unique surge in student activism of the 1960 s, while the Huks emerged from the anti-occupation guerrillas of World War II. As would be expected, there are also a number of significant changes in economic indicators throughout this period. Overall, however, examining these three civil wars within the Philippines still allows many factors to be held constant and for these limited differences to be identified and considered without excessive distraction from the primary intent of the research. Chapter four presents the case study of the Huk Rebellion. The Huks, in 1948, were a guerrilla army rebeling against the abuse of power by the central government and the persistent inequality of the tenant-farmer system that dominated much of Central Luzon. Though successful in gaining the popular support of the peasants, the Huks did not have a national political strategy for their growing rebellion. Conversely, the Philippines Communist Party (PKP) were a revolutionary political organization lacking a fighting

33 Section force due in part to the government s anti-communist policies of the previous 30 years. The ill-fitting merger of the PKP and the Huks was an effort to rectify these mutual shortcomings, but divergent goals, strategies, and membership made the merger unsuccessful. The PKP central committee adopted a plan to conventionally confront the government of the Philippines believing that the perceived revolutionary situation would enable a rapid success. As such, the rebels did not develop a financial means of supporting the rebellion. The Huks returned to looting as a short term fix to their problem of economic opportunity, but this was insufficient to sustain the fight and make an enduring war. The short length of the war fits well with the economic opportunity theory of civil war duration and the causal process of permanent financing. The Moro civil war is examined in chapter five. Building on a longstanding political and religious cleavage, the Moro nationalist movement organized for violent conflict with the state in the late 1960s with support from ethnic and religious sympathizers in the Sabah province of Malaysia. The conflict reached civil war intensity in March In early battles, the government of the Philippines was accused of waging a genocide against the Muslims in the southern Philippines. The subsequent international attention gained the Moro National Liberation Front (MNLF) the financial and military support of Libya and other Arab states. With such substantial external support, the MNLF did not organize to develop domestic means of financing the rebellion. Instead of acting as a stationary bandit, the MNLF took on the role of Robin Hood as the nearly-exclusive conduit of external support to the Moros. When Libya and other states and organizations with-

34 Section drew their support for an independent Moro state the MNLF had no means to continue financing the war and the fighting rapidly ended in January Finally, the Philippine civil war with the Communist Party of the Philippines (CPP) is analyzed in chapter six. The CPP emerged from the student radicalism of the 1960 s and purposefully built on lessons learned from the failed communist-led Huk rebellion of the 1950 s. The main critique of the Philippine Communist Party (PKP) and the Huks was that the leadership was disconnected from the mass movement. The CPP and its military arm, the New People s Army (NPA), thus pursued goals of increasing their base of popular support and also increasing the size and capabilities of the NPA. As part of this effort, the CPP-NPA developed a system of taxation in which the party was able to extract needed resources from the people in exchange for the services of a shadow government. This system of finance was extended to include taxes on businesses and landowners in order to sustain the resources necessary for increased combat operations. The CPP-NPA acted as a stationary bandit, with a long time horizon guiding their choice of the means to finance the war that aligned with a similarly long time horizon for the protracted struggle to overthrow the central government. Both are significant factors in sustaining this 11 year civil war. The dissertation concludes with chapter seven. The results and findings of the preceding chapters are reviewed and drawn together to show how they answer the central question, why do rebels decide to continue fighting instead of seeking the benefits of a peaceful settlement, and avoiding the costs of continued war? Issues and questions arising from this analysis are

35 Section also discussed in the context of potential future extensions of this research into civil wars. 1.6 Conclusion The most problematic aspect of civil war research is bridging the gap between causal mechanisms and their observable implications. In many of the cited works, the measured variable can proxy for multiple potential mechanisms. Returning to the most often cited correlation of poverty and civil war, the various measures of GDP and income could proxy for low state capacity, poor governance, wide spread grievances, or low opportunity costs. In the empirical work and case studies that follow, I endeavor to more effectively link economic opportunity to civil war duration. I do so by coding historical accounts of civil wars in which rebels are actually observed to engage in looting, taxation, or the exploitation of natural resources, or the rebels are recipients of foreign aid. In so identifying how rebels are actually financing their armed rebellion against the state, I remove some ambiguity from prior studies that assume the character of rebel financing based on country-level economic indicators. The results from the cross-national analysis then provide a more concrete basis for the further investigation of the theoretical underpinnings and micro-foundations of civil war duration.

36 26 Chapter 2 Theory 2.1 Introduction A theory is needed that focuses specifically on the rebel decision to continue fighting instead of seeking the benefits of a peaceful settlement, or capitulating and avoiding the costs of continued war. To craft a theory of civil war duration based on economic opportunity in fighting, I turn to two theories of conflict. The bargaining theory of civil war duration, first proposed by Fearon (1995, 2004), establishes a basic understanding of how the potential costs of war influence decisions to fight or settle. From this I arrive at the conclusion that economic opportunity for rebels in war makes war less costly, thereby decreasing the range of possible peaceful settlements, and making continued war more likely. I then examine Olson s (1993) theory of the stationary bandit to explore two potential types of financing available to rebels. Here I find that rebel leaders can realize greater profits from longerterm and systematic exploitation of the area under their control then they can from short-term predation. Thus rebel groups (and their wars) endure

37 Section longer with this type of permanent revenue source. 2.2 Bargaining The basic (canonical) bargaining model for analyzing conflict between or within states begins with the assumption that war, or violent conflict, is costly (Fearon 1995, Wagner 2000, Powell 2002, etc.). Both sides could achieve greater gains by dividing the contested spoils between them without having to first deduct this cost. The resulting intuition is that conflict is an inefficient or irrational outcome when there always exists a range of potential settlements that have greater utility to both players. Hence the focus of much of the literature is to analyze the barriers to efficient outcomes such as incentives to misrepresent, credible commitments, and indivisible goods. I depart from this initial assumption that war is costly, as have others such as Collier, Hoeffler, et al., and consider the consequences when a group can gain from the process of fighting sufficient to make choosing to continue the fight a more likely choice.

38 Section The Costs and Benefits of War There are three basic categories for the cost of war. 1 These are the human toll, supplying fighting forces, physical damage, and adverse economic consequences from conflict. Beginning with common database coding procedures, conceptualizing the cost of war usually addresses first the human cost in terms of battle-related deaths (e.g. Correlates of War, Uppsala Conflict Data Program Armed Conflict Dataset). This is often labeled the intensity of violence. Expanding on this, one can find reference to genocide, human rights violations, terrorist attacks, and other security fears as measures of the negative impact of fighting on belligerents and non-belligerents (Human Security Report Project 2005). The cost of war is also the demand for resources. From foraging requirements of early large armies of the Napoleonic era to logistics chains of 1 In this note, I briefly address the moral argument against violent conflict and war. The moral aversion to war is based on the unobjectionable premise that violent conflict inherently involves casualty, or death. The harm done to a person along the scale of physical and mental wounds is sufficient that any conceptualization of costs of war should always be negative. Sebastien Junger puts it thus in his 2010 book, War, War is so obviously evil and wrong that the idea there could be anything good to it almost feels like profanity (Junger 2010, 233). Yet there have certainly been conflicts fought in which the death, or injury, of persons is regarded as a necessary sacrifice, a regrettable yet acceptable cost. There are wars where such human tolls are cavalierly dismissed, or ignored as inconsequential in relation to the gains secured by the violence. While the normative argument against fighting is acknowledged, it is also noted that this argument has not risen to the level of a moral imperative to avoid conflict. Leaders in and out of government, and their followers as well, portray and use violence as an often necessary component of efforts to effect political, economic, and social ends. This moral aversion to war is a sliding and relative scale in which low numbers of casualties can have tremendous effect (e.g., the withdrawal of US forces from Somalia after the death of 18 US Army Rangers, the emergence of a Muslim identity in Mindanao after the death of dozens of military trainees in 1968) and enormous casualty figures might have no purchase on policy making (e.g., the declaration that a genocide was underway in Darfur, or Rwanda). Given this wide range of reactions, a rational calculation of the utility of violence must move beyond an abhorrence of violence and attempt, as many others have done previously, to evaluate costs and benefits within the unique context of warfighting.

39 Section modern warfare, sustaining the fielded armed force includes supplying material and personnel, provisioning fighters, providing fuel for transportation be it mules, or trucks and compensating fighters (Van Creveld 1977; Blainey 1998). In the fourth year of fighting the Moros in the Southern Philippines, the annual cost to the government had increased by a factor of four, since the first year of fighting, to nearly $325 million and the casualty rate amongst the Armed Forces of the Philippines in 1975 was over one hundred each month (Yegar 2002, 299). Thus, minimally, war consumes stockpiles of material, personnel, food, fuel, and cash. Bueno de Mesquita and Lalman (1992), amongst others, discuss the economic impacts broadly as loss of life and property (297). Longer term costs are deepened poverty, damage to agriculture, damage to industry, and the opportunity costs of foregone social services due to spending on the military. Some cite these as components of the cyclical conflict trap of civil wars (Keen 1998, 46). Wood (2000) argues that conflict dampens returns for the elite via assets destroyed, costly strikes, increasing security costs, suspended investment suspended, and increased taxes (15). During Liberia s many civil wars from 1989 until 2003, companies lost hundreds of millions of dollars in equipment and inventory looted from logging operations (Truth and Reconciliation Commission 2009, 22). Violent conflict, in many ways, is inherently costly. Offsetting these costs are potential benefits derived from the war-making process. Rebel groups can profit from fighting by employing the same tools of violence used to combat the government as means to forcibly extract resources. The Mozambique National Resistance Organization (RENAMO),

40 Section for example, is noted for the extreme brutality with which it looted resources from the local population (Roesch 1992, 468). During the Huk rebellion in the Philippines, the rebels specialized in stealing from municipal treasuries, post offices, and payroll offices. The Huks also robbed trains and hijacked trucks (Kerkvliet 1977, 215). Groups can also adapt to the changed economic processes brought by the fighting to derive more sustainable revenues from fighting. After the 1962 coup in Burma and the ensuing economic collapse, rebels took advantage of the opportunities to exploit the rapid expansion of the black market. As the black markets boomed, it was insurgent groups that controlled most frontier posts and were in strategic position to benefit (Smith 2007, 19). Keen (2000) described this as the emergence of an alternative system of profit, power, and even protection and lists several types of economic benefits including fruits of pillage, protection money, control of trade, and profit from foreign aid (Keen 2000, 22,29-31). Similarly, Weinstein (2007) finds that there are economic endowments which rebel leaders may access to raise capital to support their rebellion during war. These include natural resource extraction, taxation, criminal activity, and external patronage (Weinstein 2007, 7). The state can also profit from fighting. From the government perspective, Bates (2008) describes potential benefits as payoffs to state elites from the use of state controlled violence to predate on resources. When elites are unable to control the means of violence (i.e., the military and constabulary), government forces can take advantage of opportunities to benefit personally through looting and extortion (Soderlund 1970, 50). The state can also profit from foreign aid flows, as can the rebels. In many cases this aid is

41 Section connected to the on-going conflict. If aid is expected to be reduced when the violence ends, there can arise a perverse incentive for the recipients to continue fighting. The war itself changes the dynamics of systems for economic distribution and political gain in such a way that benefits may be realized from the war. There are thus both potential costs and potential benefits that shape belligerents payoffs for a decision to continue fighting. Given the wide range of disparate types of payoffs, it would be difficult, if not impossible, to attempt to scale the costs and benefits in order to arrive at some true, or net payoff to war. Still, these types of economic opportunity in war need to be further examined to understand their effect on bargaining and the decision to continue fighting. To do so, the costs and benefits of fighting are abstracted and reduced to a horizontal range from 0 to 1. This sets the stage for the canonical bargaining model. It is reviewed in the follow section to illustrate the potential effect when the cost of war is offset by a benefit to fighting The Bargaining Model Let G be the government of a state and R be a rebel group engaged in a war. 2 Suppose the potential settlements for issues being contested are represented by a horizontal line that ranges from 0 to 1. The government, G, prefers settlements that have a result closer to 1. The rebels, R, prefer an outcome closer to 0. The range (0, 1) can be thought of as the range of 2 This discussion draws heavily from Fearon s (1995) discussion of when two rational states would prefer war to negotiation (Fearon 1995, 386-7).

42 Section G s value for war Bargaining Range R s value for war G s value for an outcome x R s value for an outcome x 0 p + k G x p 1 - ( p-k R ) 1 R s preferred outcome G s preferred outcome Figure 2.1: Bargaining Range potential power sharing arrangements (see figure 2.1). If the disagreement continues as violent conflict, then each side has a utility, k i, from the fighting. If k i < 0 then this represents that war imposes a cost to the belligerent, i, and if k i > 0 then this represents that war provides a benefit to the belligerent, i. If the two sides are unable to agree on a peaceful settlement to resolve the issue, the war will continue. The outcome of the war is probabilistic and the government will prevail with some probability p. A victory in war is considered to be decisive such that if the government wins then it secures its most preferred outcome of 1, and if the rebels win they secure their most preferred outcome of 0. With a probability of winning, p, and a cost from fighting, k G, the government s expected utility from continued war is p + k G. The government will thus not settle for a deal that gives them anything less than this value.

43 Section Similarly, the rebels will not settle for less than their expected utility from continued fighting, 1 (p k R ). These values are represented on the top line of figure 2.1 for a war that is costly to both sides (i.e., k i < 0). The familiar result is that the range of values (p + k G, 1 (p k R )) represents the range of values for a potential peaceful settlement that both sides should rationally prefer to continued war. As an example, consider the situation where a settlement of value x is proposed. The government prefers x to their expected utility from further fighting if x > p + k G. Likewise, the rebels prefer x to their expected utility from further fighting if x < 1 (p k R ). Therefore, for any settlement x, where x (p + k G, 1 (p k R )), continued costly war is an inefficient and irrational outcome. Suppose now that the rebels can obtain a net profit from fighting such that k R > 0. This increases the rebel s expected utility from choosing to continue to fight as is shown by the enlarged value R has for war on the top line of figure 2.2. The effect of this increase is the reduction of the bargaining range. Put another way, there is a reduction in the range of potential settlements that both sides prefer to their expected utility from fighting. Intuitively, ceteris paribus, the war is more likely to continue when the set of potential agreements is reduced (Fearon 1995, 404). This is true even if the benefits of war do not wholly offset or outweigh the costs of war. The reduction in cost realized by these benefits is sufficient to reduce the bargaining range. This reduced range of values for potential settlements makes a peaceful outcome less likely and continued fighting more likely. In the most extreme example, consider a situation where the benefit

44 Section G s value for war Bargaining Range R s value for war G s value for an outcome x R s value for an outcome x 0 p + k G x p ( p-kr ) R s preferred outcome G s preferred outcome Figure 2.2: Reduced Bargaining Range from fighting is so great that the bargaining range effectively collapses. If k R > 0 and k R > k G, then the rebel s expected utility from fighting is greater than any settlement that would be agreeable to the government. Conversely, the government s expected utility from fighting is greater than any settlement the rebels would agree to. The rational choice for both sides is to continue the war as any settlement amenable to the other side would be of less value than their reservation value for fighting. As depicted in figure 2.3, the bargaining range no longer exists. The second civil war in Sierra Leone, , is an illustrative and extreme case in which the bargaining range collapsed. The conclusion of the first civil war in 1996 ended a substantial proportion of foreign aid that had been flowing to the state and also imposed a tight monitoring regime on diamond production that had been split, during the war, between the rebels and the military. The military returned to a state of war, and supported

45 Section G s value for war R s value for war G s value for an outcome x R s value for an outcome x 0 p + k G x p ( p-k R ) R s preferred outcome G s preferred outcome Figure 2.3: Collapsed Bargaining Range the rebels in doing so, in order to restore their profits. The government resumed fighting by providing arms to the rebels and even staged mock battles in order to restart international aid for the crisis and to regain control over the diamond industry. The rebels cooperated so that they too could profits from diamond sales that had been interrupted by the peace agreement (Keen 2000, 35-6). This exposition of the effects of changes in rebel economic opportunity on the decision to continue fighting yields the following hypothesis: H 1 : Civil wars are more likely to continue when economic opportunities to profit from fighting at least partially offset the costs of fighting for rebels. Having demonstrated the intuition behind the hypothesis that the existence of economic opportunity in war increases the probability that war will continue, I next turn to an examination of the different methods for making

46 Section war profitable. Olson s (1993) theory of the stationary and roving bandits is useful in this endeavor. 2.3 Bandits Olson frames his theoretical discussion with an account of the Chinese warlord Feng Yu-hsiang. Feng is attributed with some of the common traits of a warlord. Feng would conquer a territory, establish himself as a permanent warlord, levy heavy taxes on the population, and then use his army to defeat the threat of passing bandits, such as White Wolf, who sought to loot and plunder the people under Feng s control (Olson 1993, 568). Olson examines why it is that the people appeared to prefer a stationary bandit, like Feng, over roving bandits, like White Wolf. The stationary bandit repeatedly extracts resources and property from the people, whereas the roving bandit steals in a single episode and then leaves. While Olson s primary intent is to examine the preferences of the conquered, in the present discussion of civil war duration it is of greater interest to explore the motivations of the bandits. If one conceptualizes rebel groups as bandits, the central question then becomes, why would a rebel group prefer to act as a stationary bandit rather than as a roving bandit? First, the concept of a bandit and how a bandit can be equated with a rebel group warrants further explanation. Bandits are actors who operate outside of government coercive controls. If the state is, as Weber defines it, a human community that (successfully) claims the monopoly of the legitimate use of physical force within a given territory, then to operate outside

47 Section of government control implies that the bandit has the ability to contest the state s monopoly of violence (Weber 1991, 78). Rebel groups can certainly be thought of as having the capacity for violence. It is this characteristic that serves as the main distinction between rebellion and protest. In a civil war, a rebel group is contesting the government s monopoly of violence within a given area, such as the territory of the entire state, or some region thereof. This contest is won by the side that can organize the greatest capacity for violence (Olson 1993, 568) Stationary versus Roving Bandit Profits In keeping with the rational actor assumption of the bargaining model, the next step is to assign rational preferences to the bandits and their subject people. People acting in their own self-interest want to keep as much of their property as they can. Meanwhile, the bandit desires to maximize their theft of the people s property. In Olson s conception, the roving bandit plunders once, while the stationary bandit is engaged in a system of repeated theft. Returning to the motivating example for Olson s theory, if Feng rationally chooses to be a stationary bandit, then it must be the case that this system of continuous stealing or taxation, is more profitable than the alternative of looting. The roving bandit, with very short time horizon, extracts a high tax, but this source of revenue is intermittent. The roving bandit heavily discounts any future earnings and is thus incentivized to steal as much as they can now. While they may take nearly all that a person possesses, this leaves the people with nothing to use to amass additional wealth or property. Thus,

48 Section the roving bandit s utility from a subsequent round of looting these same people is nearly non-existent. Furthermore, the roving bandit, having taken all that they can, departs the area. The roving bandit obviously cannot credibly commit to protect people from theft by other bandits or the state. Without providing this public good, and with already depleted resources, the people have even less incentive to engage in productive activity when they foresee that their property could again be stolen by someone else. The stationary bandit, in contrast, has a long time horizon. The stationary bandit takes actions necessary to cultivate a more permanent revenue stream from the people. Only part of the people s property is stolen or extracted as a tax. This leaves the people sufficient income, property, and possessions to continue production such that a tax can again be levied in a future period. Olson theorizes that their incentive to produce is derived from the knowledge that they can keep whatever proportion of their output is left after they have paid their taxes (Olson 1993, 568). In addition to only taking a portion of their production, the stationary bandit must also protect the people from looting by other bandits or theft by the state. The stationary bandit thus builds a more permanent means of gathering resources from a foundation of people and production that can be taxed repeatedly. This repeated extraction over time results in a far larger take for the bandits than could be had as a roving bandit (Olson 1993, 568). Thus, a rational and self-interested rebel group will act as a stationary bandit to maximize their profits. To be a stationary bandit, a rebel group needs to maintain their capacity for violence. Any bandit must first establish control of an area, by contesting

49 Section the monopoly of violence. This requires some initial violent capacity with which the rebels can confront the state. As a stationary bandit, however, the rebels must also sustain this capacity for violence to defend against the incursions of other groups or the return of the state. The profit motives of the rebels, as stationary bandits, are thus not solely to pursue personal enrichment, but also to ensure that the group has the resources necessary to sustain their capacity for violence. 2.4 Rebel Financing With rebel motivations thus theoretically established, we next consider how one might actually observe the financing of rebel groups. For this, the writings of Weber (1978) on The Financing of Political Bodies is instructive. Weber s work is in relation to the political bodies of a state, but rebels as an organized, armed group with political objectives also fit well into this discussion. Radtke (2006) similarly connects rebel financing with Weber s ideal types in her discussion of the role of the diaspora in civil wars. Weber proposes two ideal types of financing for political bodies that closely align with the practices of roving and stationary bandits described by Olson (Weber 1978, 194-9). The two foundational categories of financing are intermittent and permanent. Political bodies obtain intermittent financing through irregular, nonsystematic, or occasional contributions (Weber 1978, 194). These donations can be either compulsory or voluntary, as the main focus is on whether or not the financial contribution will be recurring. In a voluntary sense, such

50 Section a contribution could take the form of begging. In a compulsory sense, the donation would be forcibly extracted as in looting. Such was the case in Mozambique during the civil war between Mozambique National Resistance Organization (RENAMO) and the Marxist government of the Front for the Liberation of Mozambique (FRELIMO). While RENAMO received substantial military aid from the U.S. and other states, for food the rebels depended almost exclusively on stealing from peasants and the looting of rural agricultural businesses. It is interesting to note that the need to steal food was due in part to the government s ability to force the rebels to shift its bases frequently in essence, forcing them to act as roving bandits (Roesch 1992, 471). Permanent financing, as one might expect, takes the form of taxes, fees, or imposts collected regularly by the group (Weber 1978, 196). These taxes can be linked to economic production, but are not necessarily so. Thus, they can be levied on individuals, receipts, possessions, production, services, and transactions. These taxes can be collected as monetary payment or by the contribution of goods-in-kind (Weber 1978, 197). Revolutionary taxes are often associated with Maoist communist insurgencies as in the Vietnamese civil war. Family farms were taxed based on their rice production, with those producing over 10 to 146 bushels taxed at a rate of 5% while larger producers were taxed at a rate of up to 25% (Lanning and Cragg 2008, 129). In the same war, a type of tax-in-kind was levied on civilians by forcing them to provide three months of labor each year, carrying the wounded, transporting ammunition and rice, or working on roads (Lanning and Cragg 2008, 125). Weber s purpose in describing these means of financing was to assess

51 Section their relevance to the future character of economic markets. Here these ideal types are adapted to better understand the types of resource extraction rebel groups might engage in when they act as roving or stationary bandits. 2.5 Hypotheses From the bargaining model we derived the result that economic opportunity reduces the probability that rebels will choose to peacefully settle a civil war with the state. This was expressed in the first hypothesis, H 1 : Civil wars are more likely to continue when economic opportunities to profit from fighting at least partially offset the costs of fighting for rebels. Olson s theory was examined next and it was determined that the most lucrative economic opportunities accrue to rebels that act as stationary bandits. As stationary bandits, the rebel group imposes a permanent system of financing that, in an amalgamation of Olson and Weber, involves the regular collection of taxes and fees from individuals, businesses, and commercial transactions within the area the rebels control. From this theoretical analysis, the following additional hypotheses are thus derived. H 2 : Civil wars in which rebels act as stationary bandits will be longer than other civil wars. H 2 a: Civil wars are longer when rebels employ a permanent means of financing, such as a system of regular taxes and fees.

52 Section Endogeneity One might question the potential for endogeneity in this theoretical explanation. Could it be that longer wars give rebels more opportunity to develop a permanent system of finance, instead of rebel permanent financing contributing to longer wars? Two causal stories fit this possibility. The first is that it takes time to design and implement a system of taxation. Thus, the longer a war endures, the more time the rebels have to craft taxes and put them into place. From only a slightly different perspective, another causal story could be derived from the organizational sophistication required for an effective tax system. Efficient organization is a necessary component of successful military action. In fighting an armed insurgency, a rebel group acquires and improves their organizational skill. This improved skill set then, as a residual, enables the creation of a tax system within the rebel organization. Thus, the longer a war endures, the more that rebels fight, and the more likely it is that a rebel group can implement a system of taxation. To refute this direction of causality, recall that a pre-requisite for extracting resources from people in an area is first to have control of that area. Controlling an area requires that a rebel group first has the violent capacity to establish control, and then the means to sustain this capacity in order to maintain control. If the argument is that a longer war that is, prolonged control of an area leads to more opportunity to engage in taxation, then a means of sustaining the rebel s violent capacity must be found independent of taxation. In other words, to justify the assertion of endogeneity, it must be shown that there exists a means of financing that is

53 Section the antecedent of a bandit maintaining control of an area. Three scenarios thus need to be examined for how rebels might acquire the resources necessary to sustain the means of control. These are: 1. the rebel group already possesses the required financial resources; 2. the rebels plunder the area under their control to gain the necessary resources; or, 3. the rebels tax the area under their control to gain the necessary resources. Case number three is tautological: rebels develop a tax system to maintain control over an area so that the rebels can develop a tax system. Case number one is interesting, but largely irrelevant to the discussion of endogeneity. If the rebel group is somehow endowed with the means to maintain control over an area without having to further extract resources from that area, then the rebels have no need, and perhaps no incentive, to gather additional resources through plunder or taxes. This situation could arise when the rebels receive a large amount of foreign aid to finance their rebellion. The final scenario, number two, plausibly requires that the rebels plunder the area under their control to amass the resources needed to maintain control. However, if the rebels steal all, or nearly all, of the people s property, then the people are left without the means of further production. In addition, having lost all of their property, the people do not rationally perceive a benefit to further production. In this scenario, the rebels have depleted their tax base and cannot extract further resources even if a tax system were

54 Section put into place. Thus, while the causal story of endogeneity is interesting, it is not consistent with the theoretical framework advanced here to support its causal direction. 2.7 Conclusion The central hypothesis derived from this theoretical analysis is that civil war, on average, will be longer when rebels profit from fighting. This chapter engages in an effort to unbundle this using two different approaches to understanding war. The first is that profit from war alters the basic assumption that war is costly, thereby reducing the potential range of bargaining settlements, and making continued war more likely. The second is that rebels prefer a sustainable means of raising resources to fight. This type of permanent financing requires a longer time horizon by the rebels, but is ultimately more lucrative, enticing rebels to choose to continue fighting in order to sustain this revenue stream. In the next chapter of quantitative analysis, I define potential types of observable economic activities available to rebels within the unique dynamics of conflict. This set of observations provides evidence of a phenomenal regularity confirming the correlative claim of my theory of civil war duration. In subsequent chapters, case studies of civil wars are used to investigate the events that link cause and effect to ascertain their alignment with the theory s causal claims.

55 45 Chapter 3 Quantitative Analysis 3.1 Introduction Economic opportunity is both the ability for rebels to profit from war and also the ability to gather the resources necessary to sustain combat (i.e., the means to obtain arms, ammunition, food, supplies, and pay fighters). The four principal sources of rebel financing in civil wars provide for more focused investigation of economic opportunity. These are looting; taxes on people, businesses, and transactions; exploiting natural resources; and receiving external support (money, arms and supplies, training, safe haven, direct military support). In this chapter, a quantitative, cross-country analysis of civil wars is undertaken to test these hypotheses. This large-n analysis is presented in five sections. The first section is data description to provide the nuts and bolts of the definitional choices made in constructing the dataset and the operationalization of the four sources of rebel economic opportunity. The second section examines the Cox proportional hazards model and the methods used for accurately spec-

56 Section ifying the model. The results of the model are presented in section three alongside diagnostic methods for assessing fit, form, influential observations, and the proportional hazards assumption. The results are then subjected to robustness tests and tests of alternative specifications in section four. An examination of counter-examples is presented in section five. To preview my findings, civil wars are longer when rebel groups impose a systematic tax in areas under their control and when rebel groups receive external support. Looting also appears to be linked to increased duration, but the finding is tentative and sensitive to alternative specifications. Civil wars in which rebels exploit natural resources do not have a statistically significant correlation with duration. These results are robust to alternative specifications with other covariates of civil war duration, including regional effects, conflict neighborhoods, country specific effects, the effect of the Cold War, population size, and per capita income. 3.2 Data Description Defining Civil War The not-so-obvious first hurdle in a cross-national analysis of civil wars is to develop and apply a definition of civil war. The range of answers to this question, what is a civil war, is evidenced by basic disparities in academic research datasets. A simple count of the number of civil wars between 1960 and 1993 yields a range of answers from 58 to 116, with a correlation between the dozen datasets from 0.42 to 0.96 (Sambanis 2004, ). This ambiguity is also readily apparent in popular discourse where the news

57 Section of violent internal conflict frequently turns to the question, is this a civil war? An internet news search of civil war and any of the countries involved in the 2010 Arab Spring uprisings will yield a myriad of phrases adopted by decision-makers, journalists, and participants such as, being on the brink of civil war, at risk of an all-out civil war, civil war erupting, preventing civil war, and burgeoning civil war. While there does not appear to be consensus about a definition of civil war, the distinction is not a trivial one for the term civil war connotes an intensity of conflict that has the potential to yield serious consequences for the destruction of life and property. In the social sciences, the definition of civil war should then be one which conveys the qualitative and quantitative differentiation of such conflicts from lower intensity political violence in a state. For this analysis, I adopt the Correlates of War (COW) rules for defining and coding civil wars. In general, a civil war is a conflict fought between a recognized state and a non-state actor that involves effective resistance of an organized armed force and that is fought primarily within the territorial borders of the state (Wayman and Sarkees 2010, 337; UCDP/PRIO, 2). The state is defined in the COW data as a classification of entities based on territory, population, sovereignty, independence and diplomatic recognition (Wayman and Sarkees 2010, 18). A non-state actor is an entity that fails to meet one or more of these criteria. Both sides are politically and militarily organized with political objectives. As articulated by Sambanis (2004), the latter criteria is part of a broader concept of effective resistance that is either qualitatively derived from the context of the war, or quantitatively arrived at via thresholds on the level and dispersion of violence. I address thresholds of violence

58 Section separately, but here it warrants mention that one-sided violence by the government against a group that is not organized to resist does not necessarily make the conflict a civil war. In 1947, the Nationalist governor of Taiwan, Chen Yi, appealed for assistance from the Beijing Nationalist government of Chiang Kai-shek to put down widespread protests by ethnic Formosans agitating for government reform. The protests evolved in reaction to government actions, but never organized politically to articulate a coherent set of political goals, nor did the protestors present evidence of organized or armed opposition. The resultant violent crackdown by the Chinese Nationalist Army ended with the deaths of an estimated 10,000 Formosans (Smith 2008; Kerr 1965). This conflict is one-sided violence and not a civil war. 1 Another common characteristic of civil wars is that they are fought primarily within the territorial borders of the state (Wayman and Sarkees 2010, 337; UCDP/PRIO, 2). Colonial wars are not always considered civil wars as they can fail to meet the standard of being near the center of government, or metropole, of the state (Wayman and Sarkees 2010, 45). Similarly, imperial wars are, by definition, fought in non-state territory and thus outside of state boundaries (Sambanis 2004, 828). Extra-state, or extra-systemic, 1 In a civil war, the government is also principal combatant, ruling out wars fought between non-state actors. Non-state wars formed the basis of four new categories of intrastate wars as introduced in 2010 as part of version 4.0 of the COW data. These are regional wars, inter-communal wars, non-state wars in non-state territory, and non-state wars across state borders. They are distinct from civil wars in that the neither of the major belligerents is the state government (Wayman and Sarkees 2010, 340). The competition for primacy between rival Marxist independence organizations in Eritrea erupted in violent infighting between the Eritrean Liberation Front and the Eritrean Popular Liberation front. The violence caused an estimated 3,000 battle-related deaths between 1972 and 1974 (Wayman and Sarkees 2010, 429). This inter-communal conflict, absent the participation of the state as a belligerent, is not coded as a civil war.

59 Section wars are often not included as civil war for practical reasons involving the complexities of obtaining data specific to the colony or territory, for the most common covariates of civil wars, such as per capita GDP and polity scores (Sambanis 2004, 827). Fearon (2004) and Sambanis (2004) make arguments for the inclusion of some extra state wars. But while these wars share many common traits with civil wars, the Correlates of War and other studies of civil war exclude colonial and imperial wars from the intrastate civil war category as do I (Fearon 2004, 216; Sambanis 2004, 825; Kalyvas 2007, ). While there is debate about the use of an absolute threshold for violence that characterizes a civil war in contrast to conceptualization involving a continuum of violence, the most commonly used datasets employ an absolute threshold in their definition of civil war (Kalyvas 2007, 418; Sambanis 2004, 822; Gates and Strand 2004, 5). The threshold used to differentiate civil wars from other internal armed conflict is 1,000 battle-related deaths. This threshold value, however, is characterized by multiple units of measure. As originally envisioned by Singer and Small in 1972, civil wars are those conflicts with 1,000 or more battle-related deaths per year. In addition to reflecting a violent intensity commensurate with war, this sustained combat served to help distinguish civil war from one-sided violence and large-scale massacres (Wayman and Sarkees 2010, 49). Fearon (2004) sets his threshold for civil war as 1,000 or more cumulative deaths over the conflict with a lower yearly average of 100 deaths (Fearon 2004, 278). The Uppsala Conflict Data Program (UCDP) distinguishes between minor armed conflict and civil war when violence exceeds 1,000 deaths in the calendar year of conflict

60 Section (UCDP 2009, 7). Sambanis (2004) identifies civil wars as starting in the first year with between 500 and 1,000 deaths, or when the three year cumulative death toll exceeds 1,000. By these disparate standards, Fearon codes as a civil war the conflict between the Irish Republican Army and the British government from , while UCDP/PRIO classifies this as a minor armed conflict, Sambanis observes a civil war from , and COW does not include it in their dataset. Beyond signifying a level of intensity in assessing whether or not a conflict is a civil war, these coding rules are important for determining if two conflicts are coded as a single war or separate wars. Both COW and UCDP use a year (although COW uses conflict-years and UCDP calendar-years) without 1,000 deaths as indicating the end of a war and the start of another (Wayman and Sarkees 2010, 55; UCDP 2009, 9). Sambanis rule is that civil war ends with a three year period resulting in fewer than 500 conflict related deaths, or six months to two years of peace following a peace treaty, decisive victory, or ceasefire (Sambanis 2004, 830-1). Fearon differs in that he codes the end of a war as a settlement, demobilization, or victory followed by a minimum of two years of peace (Fearon 2004, 278). Again, this combination of different thresholds for deaths and peace can have significant impact on the data. For instance, COW identifies four separate civil wars in Chad from Fearon codes these as a single war. Sambanis lists three civil wars from UCDP codes three episodes as being civil wars with the remaining years as minor armed conflict. These selection rules reflect multiple different theoretical and empirical approaches to the quantification of civil war. Equally important, these se-

61 Section lection rules obviously affect any subsequent analysis. Given the unsettled debate over which of these sets of rules most accurately reflects true civil war, any choice of datasets will be subject to critique. For this analysis, the Correlates of War dataset of intrastate civil wars since 1946 is used and the COW coding rules are adopted. COW is well used by other conflict researchers and is regularly updated. Conflict measurements are based on conflict-time and not the calendar year. As well, the COW data set covers wars since While not germane to the present analysis of civil wars since 1946, future extensions of this work will include earlier periods of conflict. Such research will yield more comparable results if the adopted coding rules are consistent over time. Restricting the set of civil wars to those fought since 1946 reflects the changes in the international system wrought by World War II. Different researchers refer to World War II and its effects as an exogenous shock, or qualitatively describe it as the beginning of the modern era or of a postcolonial era. The supposition of a break in the data at World War II can be traced to a number of significant changes to the conduct of civil wars including the development of guerilla and counter-guerilla warfare as a result of partisan resistance movements in areas occupied by Japanese and German forces (Beckett 2001, 55). The fusion of nationalism with communism by Mao Tse-Tung also contributed to a new era of rebel strategy and motivation that were consciously emulated by many of those who fought insurgencies after World War II (Beckett 2001, 70). These enhancements to the viability of rebellion suggest that civil wars after World War II should be substantively different from civil wars prior to World War II.

62 Section Structural Break in Survivor Functions at Duration in Months post46 = 0 post46 = 1 Figure 3.1: Survival Probability Plot for Civil Wars, The structural break in the data at World War II is evidenced by the difference in observed durations. Examination of the survival plots for civil wars prior to 1946 and after 1946 lends additional credibility to this break in duration data (see figure 3.1). There is a clear distinction between the survival curves of the two groups, with the dashed line representing conflicts after 1946 and indicating a lower probability of the conflict ending (i.e., a higher probability of the conflict surviving) throughout the range of observed durations. A calculation of median conflict duration from COW data is presented in table 3.1. Conflicts beginning prior to the end of World War II have median durations that, at a 95% confidence level, are 42% shorter than those beginning after the war. The preponderance of analysis of civil war has thus focused almost exclusively on the post-world War II era with an

63 Section Number of Median Duration 95% Confidence Intrastate Wars Months Interval Source: Correlates of War V4.0, intra-state wars Table 3.1: Median Duration of Intrastate Wars assumption (either explicitly stated or implied) that there is a fundamental shift at the end of World War II that created the circumstances by which civil wars have become more prevalent and longer. Kalyvas (2009) describes the use of post-1945 data as the main method for analysis of civil war. My analysis confirms this structural break in the data and focuses on wars since The Set of Civil Wars Under Consideration There are 242 intrastate wars since 1946 in version 4.0 of the COW dataset. In the course of researching the variables for this analysis, each civil war observation was evaluated for consistency with the stated COW coding rules and for accuracy. While the companion text, Resort to War, provides a single entry and very brief narrative description for each conflict, the COW dataset is built on a dyadic approach often resulting in multiple observations for a single war. As an example, COW provides three observations on the Angolan civil war, , to represent the direct military intervention of Cuba and South Africa. I reduce these to a single observation between the principal belligerents, i.e., the state (Angola) and the rebels (UNITA).

64 Section The COW dataset of intrastate wars includes 75 additional observations on wars already recorded that I choose to omit. My coding is sensitive to conflicts in which there are multiple rebel groups fighting the state. A civil war in Burma is coded by COW as a single observation, , involving the Karens and Communists fighting the state. A closer investigation reveals two separate insurgencies are active during this time frame in two different parts of the country. The first is the Communist rebellion that erupted after their exclusion from government and the assassination of communist leader Aung San on 19 July The second is the Karen rebellion that began in January 1949 after their ethnic minority group was not granted a semi-autonomous state within the newly independent Burma (Charney 2009; Fredholm 1993; Lintner 1994). Further, Fredholm (1993) finds that it was not until late 1952 that a ceasefire between the Communist Party of Burma and the Karen National Union was agreed upon and not until 1959 that the two groups entered into a military alliance (Fredholm 1993, 211). Two civil wars are added to the dataset based on these criteria. The resulting dataset to be used for the remainder of this analysis has 146 observations of civil war from Of 242 COW intrastate war observations, 75 are repeated entries, and an additional 21 are omitted after having reviewed case histories. There were four general reasons for excluding these conflicts from this analysis. Ten of these conflict did not involve the central government as a belligerent. Three conflicts lacked effective resistance by the non-state group and were more akin to one-sided violence than civil war. Four conflicts had battle-related death totals below the threshold

65 Section for a civil war. Four were merged with earlier civil war observations when there was sufficient evidence to believe these represented a single civil war. Phase 2 of the Chinese Cultural Revolution is given the reason code other as this complex conflict involved the central government putting down a renegade branch of the Red Army engaged in a regionally confined pogrom, but did not otherwise match the definition of a civil war. Table 3.2 lists the intrastate wars that do not appear in my data and provides the general reason code for their omission. The COW war number and start/end years are included for reference. Modeling results will be tested using the original COW data as a partial substantiation of these choices. The 146 civil wars occur in 68 different states. 34 countries experience a single civil war. The other 34 countries experience multiples of the remaining 112 civil wars. The most conflict prone states are Indonesia (8 civil wars), Iraq (7), the Philippines (6), Chad (5), Burma (5), and Ethiopia (5). The potentially confounding effect on modeling of repeated conflicts in a single state is accounted for in the section on robustness checks. After reviewing each of the 146 civil wars, all observations had a start and end date precise to month and year. A complete list of wars is included in the appendix. Ten observations are missing the day for the start and end of the conflict. An additional 14 are missing just the start day and 12 are only missing the end day. For measurement purposes, missing days were coded as the first of the month. Because of this lack of precision, duration is measured in months. Only one of the wars in the dataset is on-going at the cutoff date, that is the Eighth Colombian civil war beginning in Five civil wars have a duration of zero months based on the spreadsheet function

66 Section COW War War Start End Reason Number Name Year Year Code 726 Taiwan Revolt a 768 First Uganda b 772 Chinese Cultural Revolution Phase a, c 776 Chinese Cultural Revolution Phase e 778 Naxalite Rebellion b 788 Eritrean Split c 808 Second Ogaden Phase d 823 Nigeria - Muslim War b 846 Inkatha - ANC c 858 Romania a, b 866 SPLA Division War c 868 Jukun-Tiv War c 890 Iraqi Kurd Internecine War c 910 Moluccas Sectarian War c 911 First Nigeria Christian-Muslim War c 920 Fourth Rwanda d 923 Ethiopian Anyuaa-Nuer c 930 Third Aceh d 931 Second Nepal Maoist War d 933 Second Nigeria Christian-Muslim War c Reason Codes: (a) One-sided violence; (b) Does not meet thresholds of civil war violence; (c) Conflict not involving the central government; (d) Merged into a single civil war with earlier observation; (e) Other. Table 3.2: COW conflicts omitted from analysis

67 Section that extracts the number of months from the start and end dates. Again, this anomaly is considered in later robustness checks of the model. Density Civil Wars since Duration in Months Figure 3.2: Histogram of Civil War Duration The median duration of civil wars is 24 months with an estimated 95% confidence interval of months. This is a Kaplan-Meier estimated median duration of the time at which 50% of the wars will have ended. The distribution of durations has a positive skew (mean=37 months, skewness=1.93) and significant initial peak (kurtosis=6.99) with the first 25% of the data having less than 8 months duration. In contrast, the longest 25% of wars range in duration from 53 to 222 months. The longest duration war is the second Tamil war in Sri Lanka, July February Figure 3.2 provides a histogram overlayed with a kernel density plot of the distribution. The Kaplan-Meier plot of the estimated survival probability of a war beyond

68 Section All Civil Wars Duration in Months 95% CI Survivor function Figure 3.3: Estimated Survivor Function for all Civil Wars a given time is also provided in figure 3.3. Interpreting the data using these two plots jointly, we observe, for example, that only approximately 7% of civil wars will be of greater than 100 months duration, and there is only a roughly 10% probability that a war that has endured for 100 months will continue into the next month Sources of Rebel Funding I operationalize Weinstein s (2007) four potential economic endowments (looting, taxation, natural resources, and external support) to enable the coding of civil wars according to the source of rebel funding. As these are not mutually exclusive, it is certainly possible that a rebel group may derive support from any, all, or none of these sources. Looting is evidenced by

69 Section rebels supporting their organization through the illegal and forceful seizure of supplies, money, and other private goods. This could include reported criminal activity by rebels or their agents, reports of destruction of property or acts of violence to coerce support, elite rebel statements endorsing such actions, or allowing criminal activity as payment to supporters. In the Philippines, during the Huk rebellion, the Huks targeted banks, government offices, wealthy landowners, and even medical facilities to steal money, weapons, food, and medical supplies (Taruc 1967, 86). In contrast, taxation is when a rebel group engages in the gathering of monetary revenue and other supplies through a systematic tax on people, property, and businesses in areas under rebel control. This could demonstrate a longer range vision for procuring resources necessary to sustain the rebel fight than would be obtainable via plunder. In the second/third Aceh civil war of the Free Aceh Movement (GAM) was elaborately organized including official documents for tax collectors, receipts, and lists of contributions. Small and large businesses were taxed such that GAM profited from every sort of economic activity in the province (Apsinall 2009, 179). As just one example, a fertilizer factory is reported to have paid approximately one million dollars to GAM in 2000 (Apsinall 2009, 178). Hence, observables for taxation might include reports of fiscal tax levied and collected on individuals, taxes on the sale or transport of goods, documentation of tax revenues, requirements to regularly provide food or other supplies, to rebels, and expected contributions from family or ethnic group members overseas (diasporas). The Tutsi Rwanda Patriotic Front (RPF) imposed levies on the Banyarwanda diaspora in Zaire, Uganda, Tanzania,

70 Section and Burundi that raised more than two million dollars for uniforms, food, supplies, and armament (Watson 1992, 52). The fund raising was coordinated by Aloysia Inyumba, the RPF Financial Commissioner, and included regular contributions from those in exile (Prunier 1995, 117). An alternative means that rebel groups can employ to provide revenue for fighting is the harvesting or extraction of natural resources in areas under their control for purposes of generating profits through their legal or illegal sale. Collier, Hoeffler, et al. (2004), through their emphasis of economic greed as a motivation for rebellion, have brought considerable analytic attention to the exploitation of natural resources as a means of funding rebel movements. This has been expounded upon by LeBillon, Ross, and others. In their analysis, wars are fought when rebels and/or the state are presented with opportunities for economic gain from control of resources (such as gems, minerals, timber, and other commodities). Blood diamonds are the most oft-cited example of such exploitation as in the first Liberian civil war, , where Charles Taylor s National Patriotic Front of Liberia (NPFL) derived approximately $100 million from control of diamond mines (Truth and Reconciliation Commission 2009, 27 8). Observables include reference to control of trade in the region, exploitation of labor involved in extraction of resources, claims to land that holds resources, or confiscation of such lands. In the case where agricultural goods are exploited for the direct supply of a rebel group (such as the provision of rice), I differentiate between natural resource extraction and taxation in that natural resources primarily concerns the exploitation of resources through export and sale. Finally, external support is material, monetary, or training support pro-

71 Section vided from a foreign entity. The Uppsala armed conflict data project recently released a dataset on external support to belligerents in internal armed conflict from which documents several types of support. These are the provision of troops, access to territory, access to intelligence, weapons, material support, and funding or economic support (Stina Hogbladh and Themner 2011). In 1958, the PRRI (Revolutionary Government of the Republic of Indonesia) rebeled against the Indonesian government, in part supported by regular weapons drops including heavy weapons (bazookas, antiaircraft guns, crew served machine guns) from the US via airlift flown out of Taiwan. In addition, rebels were provided military training in Taiwan (Conboy and Morrison 1999, 61, 69 70). Other researchers have more broadly defined categories of external assistance as foreign military aid (supplies, training, weapons), economic aid to rebel groups (monetary contributions, development assistance), and safe haven (Lyall and Wilson, III 2009). To evaluate the dichotomous variables for rebel funding, I examined brief narratives of the conflicts and used these as a research springboard to scholarly articles, case histories, government agency reports, and other references to extract expert observations on how the rebel groups were funded. Wayman and Sarkees (2010) Resort to War, Clodfelter (1992,2002) Warfare and Armed Conflicts: A Statistical Reference to Casualty and Other Figures, and the Uppsala Conflict Data Project website served as the primary encyclopedic reference works. In most cases, specific mention of the sources of funding and support are able to be identified. Additional fidelity, however, regarding amounts of revenue generated from different sources, or when in the conflict the support was provided, is not as easily discerned. The ques-

72 Section tion of the temporal specificity of the data is relevant to the decision to code the data as observations on the entire war as opposed to coding calendar years as is done in other datasets, such as UCDP. Even if values could be assigned by year, the duration data is accurately coded by month and the mismatch would prove problematic if attempting to introduce time-varying covariates to model the intra-conflict variation Rebel Financing in the Data Thirty-nine civil wars are observed to have rebel groups reliant on looting for rebel funding. In 45 cases, the rebels are noted to adopt some sort of system of taxation. Natural resources are exploited by rebels in 32 civil wars. By far, the largest source of rebel group funding is external support with 81 instances of such support observed in the dataset. Rebels use multiple means of gathering resources and this is reflected in table 3.3. The most frequent combination of sources is taxation and external support. The least frequent are wars in which rebel groups are found to have used all four means. This combination occurs only once. In the Second Turkish Kurds war, July 1991 to February 1999, the Kurdistan Workers Party (PKK) supplemented revenue obtained from drug smuggling and taxes on the German based diaspora with extortion of families, looting of Turkish businesses, and outside assistance from Iran (Criss 1995, 28; Yildiz 2005, 5). There are 22 civil wars about which no observations are found on any of these sources of revenue (see table 3.4). A closer look shows a few different types of conflicts where this is the case. The first is the fanatical religious such as the Holy Spirit Movement in Uganda. The movement was effectively

73 Section Observations in 146 Type of Financing civil wars since 1946 Looting 39 Taxation 45 Natural Resources 32 External Support 81 Table 3.3: Observations of Rebel Financing a cult that employed suicidal tactics based on mystical beliefs and using primitive weapons literally sticks and stones (Rake 2004, 1168). It is a charismatic cult that rejects material trappings and there is no evidence of attempts to garner monetary revenue, or solicit support beyond gaining new recruits (Jackson 2002, 38 40). The second type of quad-zero case are civil wars that are very brief, such as a successful coup or a rebellion that is quickly put down by the state. Four of these conflicts have a duration of less than one month, 14 have a duration less than six months. These conflicts, generally speaking, are fought with weapons on hand or weapons seized in initial victories against government forces. In some cases, these wars are initiated by a faction within the military that is already armed for combat, as in the case of the Argentine Military war in The First Sri Lanka-JVP war is just over a month s duration, 6 April 16 May There are no indications of looting for resources or exploitation by the rebels of the gemstone resources present in the region (Hierman 2006, 5). It could be hypothesized that in such situations the rebels swift military defeat did not afford them time to be confronted with supply problems, or to develop solutions to these unrealized problems.

74 Section COW War War Start End Duration Number Name Year Year (in months) 737 Bolivia < Chilean Coup < South Yemen < Shiite and Kurdish < Yemeni Imamate Costa Rica First Sri-Lanka JVP First Burundi Georgia War Argentine Military Dominican Republic Paraguay War Algerian Revolutionaries War Overthrow of the Shah Second Guatemala Holy Spirit Movement First Iraqi Kurd Second West Papua Hukbalahap Rebellion First West Papua Anti-Khomeini Coalition Second Uganda Table 3.4: Conflicts outside revenue construct sorted by duration There are two cases in which the rebellion is funded by substantial financial gifts from individuals, thus not qualifying as foreign aid from another state, nor as a regular tax. In the civil war in Costa Rica, March to April 1948, the rebel army was trained and equipped by Jose Figuera at his property in the south of Costa Rica. The approximately rebels attacked the government forces and forced the government to capitulate in approximately five weeks (Hoivik and Aas 1981). In February to March of the

75 Section same year, the Yemeni Immamate war was waged by a reform movement apparently dependent on the 1946 donations of a group of wealthy Yemeni families (al Abdin 1979, 38). The rebel groups in both cases gained their resources prior to the outbreak of the conflict and there are no evident efforts to gain additional materiel or military support during the brief conflicts. The National Resistance Army (NRA) of Uganda provides a rare example where a revenue poor rebellion did not prey on the local population to generate revenue and supplies to support the rebellion and in which this tattered group was able to sustain itself over a lengthy, six years of civil war in Uganda from December 1980 to March 1986 (Schubert 2006; Clodfelter 1992; UCDP 2011). NRA soldiers were poorly equipped, underfed, and operated with old ancient weapons. Only in the final year of the war was the NRA able to arm all of its guerrilla fighters (Schubert 2006, 103). The NRA did not have secure rear-areas to which they could retreat, forcing the guerrillas to adopt a code of conduct to encourage the cooperation of the local population that essentially placed a prohibition on looting (Schubert 2006, ). The NRA was also loathe to adopt a system of taxing peasants for fear that the system would be abused (Weinstein 2007, 178). The NRA was thus almost wholly dependent on donations of food, clothing, and medicine (Weinstein 2007, 69). The success of the NRA is also notable as the first in which a post-colonial African rebel movement succeeds without substantial support from the outside (Schubert 2006, 96). Drawing on findings discussed in the literature review, additional covariates are included in the dataset for conditions that favor rebellion and insurgency. These include the size of the state s population and per capita

76 Section gross domestic product as the variables most robustly correlated with civil war on-set (Hegre and Sambanis 2006, 509). Population is a potential proxy for state capacity to control violence while the per capita income measure can be a proxy for rebel opportunity costs. States involved in civil war are also coded by conflict neighborhood to capture the spillover effect of conflict in adjacent states. Regional controls are added for potential effects correlated with the geographic clustering of conflicts. Finally, a cold war indicator variable is adopted to control for possible changes wrought by the end of the US-USSR superpower competition that included substantial involvement in civil wars from 1946 to Model Specification A Cox proportional hazards model is used to estimate the effects of these covariates on civil war duration. 2 The variables of interest are selected for their linkage to the theoretical motivation for the analysis. Theory, however, does not provide predictions regarding a particular parametric form for the duration dependency of civil war on these variables. As a modeling technique that does not make assumptions about the shape of the hazard function over time that is, the likelihood of the war ending at a particular time the Cox proportional hazards model is appropriately employed in this analysis (Cleves et al. 2008, 129). Further emphasizing the rectitude of this choice 2 In developing and testing the Cox Proportional Hazards model for this analysis, I relied primarily on three technical manuals. These are Box-Steffensmeier and Jones (2004) Event History Modeling: A Guide for Social Scientists, Cleves, et. al. (2008) An Introduction to Survival Analysis Using Stata, and the on-line statistical computing seminar provided by UCLA Academic Technology Services, Survival Analysis with Stata, accessed most recently in March 2012.

77 Section is the methodological handbook of Box-Steffensmeier and Jones (2004) that finds there are few instances in social science event history applications in which a Cox proportional hazard model would not be the preferred model (Box-Steffensmeier and Jones 2004, 66 7). The hazard rate for the ith observation is, h i (t) = h 0 (t)exp(β x) and gives the instantaneous risk of failure at time, t. Again, failure in this context is interpreted as the end of the civil war. The baseline hazard function is h 0 (t), x is the matrix of covariates, and β are the regression parameters. The Cox model makes the proportional hazards assumption that a covariate s effect is invariant to when in the process the value of the covariate changes (Box-Steffensmeier and Jones 2004, ). If the assumption is tested and holds true, the form of the baseline hazard rate is unparameterized (Box-Steffensmeier and Jones 2004, 49). The proportional hazards assumption will be tested for all variables found to be of statistical and substantive significance. A second assumption of the Cox model is that the baseline hazard is the same for all observations. Thus, while having an unparameterized baseline hazard means that the hazard can increase, or decrease, in any combination, each observations hazard is a multiplicative replica of another s (Cleves et al. 2008, 129). Robustness checks are performed in this analysis to evaluate the plausibility of this assumption. Duration times are calculated in months (Cleves et al. 2008, 33). This

78 Section lack of precision in time measurement (e.g., months instead of days) results in the creation of ties in the data where multiple civil wars fail, or end, after the same number of months. The Cox model depends on ordered duration times. Thus, ties present a problem whereby the observations are either lost, or a means must be employed to regain the efficiency of the model and break the ties. The underlying assumption is that these ties arise from imprecise measurement along a time continuum and not that the wars all ended at precisely the same time. From this, the conditional probability of these tied failures is calculated using the exact partial likelihood method (Cleves et al. 2008, 150) Univariate Analysis of Primary Variables of Interest The first method to evaluate discrete variables of interest for their inclusion in, or exclusion from, the model was to visually examine the Kaplan-Meier plots of survival functions. Kaplan-Meier estimates of the survivor functions for civil wars in which each type of rebel funding is present are provided in figure 3.4. The dashed line represents the survivor function estimate when the respective source of funding is present for a civil war. For all four variables, the plots tentatively confirm the theoretical expectation that their presence is correlated with longer duration wars. The plot for taxation visually presents the most distinct separation between that type of rebel group and all other cases, indicative of this variable having a potentially statistically significant effect on civil war duration. The survival functions for external support, looting, and natural resource exploitation also show separation, but there is some ambiguity in the month range. In addi-

79 Section Looting Duration in Months loot = 0 loot = Taxation Duration in Months tax = 0 tax = Natural Resources Duration in Months nat = 0 nat = External Support Duration in Months aid = 0 aid = 1 Figure 3.4: Survivor Functions for Civil Wars by Source of Funding tion, all four survivor functions appear roughly parallel to the plot for other civil wars giving an initial positive indication that the proportional hazards assumption may hold. To further test these variables for inclusion in the model, I employ a nonparametric log rank test of the equality of the survivor functions. Observed failure times for civil wars in the specified group are compared against an estimate of expected failure times if the group shared the same survivor function as the rest of the observations. A simple chi-squared test is used to test the null hypothesis that the survivor functions are the same. Only taxation definitively rejects the null hypothesis. External support approaches significance in the log rank test, while natural resources and looting fail to

80 Section Variable Log Rank χ 2 Wilcoxon χ 2 Name (Pr > χ 2 ) (Pr > χ 2 ) Looting (0.3993) (0.2381) Taxation (0.0325) (0.0182) Natural Resources (0.2283) (0.2142) External Support (0.1025) (0.2379) h 0 : survivor functions are the same Table 3.5: Tests of Equality of Survivor Function reject the null hypothesis of equality of survivor functions (see table 3.5). While the log-rank test places equal emphasis along the entire range of observed durations, a Wilcoxon variant of the same test places greater emphasis on observations occurring earlier in time. This reflects the skew of the data and that more civil wars are at the risk of ending at earlier times. Again, the results are summarized in table 3.5. Of note, the looting group appears to gain significance with the weighting of earlier failures while external support loses significance with the Wilcoxon test. Natural resources is consistent in both tests. As a final preliminary estimate of the statistical relevance of these four variables, a univariate Cox model is run for each. The results are listed in table 3.6. All four variables have negative coefficients indicating an effect that increases duration of civil wars. Only taxation is statistically significant at the p < 0.05 level. External support, natural resources, and looting do

81 Section Variable Coefficient Name (Std. Err.) z Pr> z Looting (0.191) Taxation (0.185) Natural Resources (0.205) External Support (0.176) Table 3.6: Univariate Cox Test Results not have a statistically significant effect in the univariate Cox model. The results of these tests are only conclusive for the inclusion of taxation. However, based on the theoretical point of embarkation for this model, I choose to include all four sources in an initial model. The results of this model, without control variables, are presented in table 3.7. With all four variables included, looting and external support are now very nearly statistically significant at the p < 0.05 level. Natural resources is not significant in its correlation with length of civil wars. All of the coefficients are again negative and the associated hazard ratios are also listed to provide an initial estimate of the size of the effect. Given the few number of variables that are potential candidates for inclusion in the model, I next test all possible combinations of these variables for interaction effects. When interaction terms are constructed with the natural resources variable, the coefficient for natural resources changes from negative to positive while still not attaining statistical significance at even

82 Section Std. Hazard Std. Coef. Error Ratio Error z Pr> z Looting Taxation Natural Resources External Support N 140 log-likelihood χ 2 (Pr> χ 2 ) (0.0117) Table 3.7: Cox Proportional Hazards Model without Controls the p < 0.10 level. This sensitivity to specification and lack of significant correlation leads me to conclude that natural resources should be omitted from the final model. Combinations of looting, taxation, and external support do not have statistically significant interactions. Given these results and a lack of theoretical motivation to suppose an interaction effect, these interaction terms are not included in the model Potential Additional Covariates I next evaluate other variables as potential controls for effects induced by factors exogenous to my theoretical construct. These are proxies for grievance and state capacity (poverty and population), regional effects, and temporal effects of the end of the Cold War. Using tests as described above in evaluating the resource variables, I investigate the potential effects of civil wars fought in five regions of the world. Descriptive data is provided in table 3.8 in descending order of civil war prevalence. Only the ten civil wars fought in Eastern Europe and the

83 Section former Soviet Union exhibit an estimated median and confidence interval that differ appreciably from the entire set of civil wars. This result echoes similar findings by Fearon (2004). A closer examination of the data reveals that the range of observed civil war durations in Eastern Europe is much more constrained than in other regions. The longest civil war in Eastern Europe is just 49 months, while all other regions duration data are bounded on the right by conflicts of months (see figure 3.5). The Kaplan- Meier plots in figure 3.6 illustrates the distinctiveness of this pattern in the data. Number K-M estimated 95% Conf. Region of Wars median Interval Sub-Saharan Africa Asia N. Africa/Middle East Latin America E. Europe/Former Soviet Union All Civil Wars (19 30) Table 3.8: Civil Wars by Region Despite its apparent significance, there are two reasons to withhold Eastern Europe as a control variable. The first, and most important by far, is the lack of any theoretical justification for its inclusion. Critiques of the Fearon typology 3 focus on the confusion of analytic criteria for his different types civil wars (Kalyvas 2007, 426). This categorization only captures wars 3 Fearon (2004) identifies five distinct types of civil wars based on patterns in the duration data. These are anti-colonial wars, wars fought in Eastern Europe, wars in which rebels trafficked in contraband, coups/revolutions, and wars fought by minority ethnic groups over agrarian issues on the periphery of a majority dominated state.

84 Section fought in Yugoslavia and in former republics of the Soviet Union during the 1990 s. There is no apparent connection between this designation of a subset of civil wars and the economic opportunity costs theory under discussion. Additional discussion of counter-examples, which include three of these ten wars, will be presented in the final section of this chapter and perhaps sheds some additional insight into this group of short civil wars. Density Civil Wars in Eastern Europe Duration in Months Density Civil Wars since Duration in Months Figure 3.5: Comparing Distribution of Civil War Durations in Eastern Europe with those in the rest of the world The second issue is the small size of the group. With only ten wars, drawing substantive causal inference is problematic. After also evaluating the regions using the log-rank and Wilocoxon tests, then including it in the model alongside the resource variables where it had no noticeable effect, the regional controls are excluded from the model. As continuous variables, state population and per capita gross domestic product are not amenable to testing by Kaplan-Meier plots or log-rank tests. I thus conduct a univariate analysis with a Cox proportional hazards model as a check of their potential import. Using the natural log of per capita GDP in the start year of the conflict, the coefficient is positive, indicating

85 Section All Civil Wars Duration in Months 95% CI Survivor function Wars in Eastern Europe Duration in Months eeurope = 0 eeurope = Wars in Latin America Duration in Months lamerica = 0 lamerica = Wars in Asia Duration in Months asia = 0 asia = Wars in Sub Saharan Africa Duration in Months ssafrica = 0 ssafrica = Wars in North Africa & Middle East Duration in Months nafrme = 0 nafrme = 1 Figure 3.6: Survivor Functions for Civil Wars by Region

86 Section that as per capita GDP increases the length of civil war decreases. The result, however, fails to attain statistical significance (β = 0.093, p = 0.383). Because of the general agreement of scholars on the importance of a measure of income, per capita GDP is again tested as a control variable after the final model specification, but again it fails to yield a statistically significant effect on duration nor does it appear to have any measurable effect on the coefficients or significance of other variables. The univariate Cox model of state population, by contrast, has a negative effect (β = 0.097, p = 0.134). As described in the literature review, the correlation between civil war duration and a country s population size could, in part, be linked to this 1,000 deaths threshold. Intuitively, a state with a larger population could more easily wage a conflict with this level of deaths than could a state of smaller size. I thus include population as a control variable in the final model using the log transformation of population in calculations. Adding population to the Cox model brings the coefficient for external support to statistical significance at the p < 0.05 level, but does not impact taxation or looting. Interaction terms composed of a resource variable and population size are tested and were not statistically significant. A dummy variable for civil wars fought during the Cold War is also evaluated. The range of dates is and the variable is observed at the on-set of the conflict. The Kaplan-Meier plots for the two time periods show some degree of separation, with conflicts during the Cold War being more likely to endure at a given time (see figure 3.7). The log-rank test for the equality of survivor functions in both time periods, however, fails to reject the null hypothesis that the hazard is the same in both time periods.

87 Section A simple correlation check of the external support and Cold War variables shows that there is not a significant correlation between the two that is, external support is not more prevalent during the Cold War than after the Cold War. 42 of 84 civil wars during the Cold War have external support for rebels and 39 of 62 civil wars after the Cold War are observed to have external support. What is not revealed by the external support dummy variable, however, is the amount of aid provided in each period. Such a comparison of how much aid is even more difficult given that the external support variable is constructed from observations of a range of different types of support, including monetary aid, provision of arms and supplies, direct military intervention, and safe haven for training. The Cold War variable is then left out of the specified model, but because of lingering concerns resultant from insufficient data, additional checks will be performed on the model after estimation to confirm this choice Final Model Specification The main hypothesis under consideration is that civil wars will be longer when rebels profit from fighting. To test this hypothesis, a Cox proportional hazard model will be used to estimate the effects of variables from a dataset of civil wars since A final model of civil war duration is constructed with the risk of a civil war ending at a given time being dependent on rebel revenue derived from looting, a system of taxation, and external support, while controlling for the effect of the size of the state s population. The preceding analysis demonstrates how other variables under consideration fail to have either the substantive or statistical significance to warrant their

88 Section Civil Wars during the Cold War Duration in Months coldwar = 0 coldwar = 1 Figure 3.7: Survivor Functions for Civil Wars during Cold War inclusion in the model. In the following sections, I report the results of this model, provide evidence of its robustness to alternative specifications, confirm the assumptions of the model, then discuss substantive take-aways from the results. 3.4 Results Table 3.9 lists the results of the Cox proportional hazards model of civil war duration. The coefficients for taxation and external support are negative and statistically significant at the p < 0.05 level. This confirms the theoretical expectation that when observations on these variables are positive there is a statistically significant decrease in the likelihood of war ending in the

89 Section ensuing month. Therefore, such wars are longer, on average, than other civil wars. Interpreting these coefficients using the hazard ratio, a civil war in which the rebels are observed to put in place a taxation system have only approximately two-thirds the likelihood of ending in a given month as do other wars. The presence of external support to rebels has a very similar, although slightly smaller, falling hazard that also implies longer survival times. Std. Hazard Std. Coef. Error Ratio Error z Pr> z Looting Taxation External Support log population N 140 log-likelihood χ 2 (Pr> χ 2 ) (0.0141) Table 3.9: Cox Proportional Hazards Model of Civil War Duration The looting variable does not return a statistically significant effect at the p < 0.05 level. The coefficient, however, is negative revealing an effect in the same direction as taxation and external support prolonging conflict. Thus, while looting does not have a statistically significant effect on war duration, the result, in general, confirms the economic opportunity cost conceptualization of war in that looting can be expected to provide a marginally positive incentive to continue the conflict. All of the variables are next tested to verify that they do not violate the proportional hazards assumption of the Cox model. The modeling assump-

90 Section Looting Taxation scaled Schoenfeld loot Time bandwidth =.8 scaled Schoenfeld tax Time bandwidth =.8 External Support Test of PH Assumption scaled Schoenfeld aid Time bandwidth =.8 scaled Schoenfeld lnpop Time bandwidth =.8 Figure 3.8: Plot of Scaled Schoenfeld Residuals tion is that the effect is invariant to when in the process the value of the covariate changes (Box-Steffensmeier and Jones 2004, ). An analysis of Schoenfeld residuals is central to testing this assumption. The Schoenfeld residuals represent the difference between the observed and expected values of the variables at each failure time (Box-Steffensmeier and Jones 2004, 121). The Grambsch and Therneu test of the proportional hazards assumption tests the null hypothesis that the hazard-ratio function is constant over time using Schoenfeld residuals (Cleves et al. 2008, 201). In addition, plots of the scaled Schoenfeld residuals are examined visually for an approximate zero slope indicating the effect of the variable is not time dependent (Box- Steffensmeier and Jones 2004, 121). The Grambsch and Therneu test results

91 Section are presented as table For each variable and globally, the test fails to reject the null hypothesis. The plots of the Schoenfeld residuals are in figure 3.8. They confirm visually that all variables have a near constant log hazard ratio function over time. ρ χ 2 df Pr> χ 2 Looting Taxation External Support log population Global Test H 0 : hazard-ratio function is constant over time Table 3.10: Grambsch & Therneu Test of Proportional Hazards Assumption An analysis of Cox-Snell residuals is next used to evaluate the goodness of fit of the model. Cox-Snell residuals are calculated from the estimated cummulative hazard and, if the model is correctly estimated, the residuals should have a standard exponential distribution with a hazard ratio of 1 for all t (Box-Steffensmeier and Jones 2004, 124). The resultant plot should approximate a straight 45 line. Variability is expected in the right-hand tail of the plot as this is where estimation uncertainty is the greatest (Box- Steffensmeier and Jones 2004, 121; Cleves, et. al. 2008, 216). From figure 3.9, the model fits the data reasonably well.

92 Section H(t) from Cox Snell Residuals Cox Snell residual Figure 3.9: Cox-Snell Residual Plot with 45 reference line 3.5 Robustness Checks Having invested a substantial amount of effort into cleaning up the COW dataset, the first check of the model is to evaluate its sensitivity to these changes in the dataset. While the coding of the independent variables remain consistent between my data and version 4.0 of COW, here I maintain all of the original COW observations and their associated parameters. The results are very similar and are presented in table The only noticeable difference is that in using the original COW data, the magnitude of the coefficients for taxation is increased by 9.8% and for external support by 5%. Thus, in comparison, my model understates the correlation between these variables and civil war duration. Otherwise, the effect is in the same direction (prolonging conflict) and statistically significant. From a similar concern about sensitivity to coding decisions, cases with

93 Section Coef. Std. Err. Haz. Ratio Std. Err. z Pr> z Looting Taxation External Support log population N 220 log-likelihood χ 2 (Pr> χ 2 ) (0.0083) Table 3.11: Cox Model for Original COW Data War Number War Name Month Year 737 Bolivia April Chilean Coup September Saur Revolution (Afghanistan) April South Yemen January Shiite Kurd March 1991 Table 3.12: Civil Wars with Durations Less than One Calendar Month zero duration are addressed. There are five such cases in which the start and end dates are contained within the same month resulting in their being assessed as having less than one month s duration. These cases are listed in table With zero duration, these observations technically fail (or end) prior to entry in the database and are excluded from the model. To return them to the model and evaluate their effect, the duration is set to one month, and the Cox proportional hazards model is run. There are no significant differences in the results due to this minor change in coding. The model is also run with and without a Cold War indicator variable. Employing a dummy variable for civil wars fought during the Cold War did

94 Section not significantly change the outcome of the model in terms of the coefficients for other variables or their significance. Further, an interaction between Cold War and the external support variables was tested. This interaction term was statistically significant, but the coefficient was very nearly one, indicating almost no substantive effect (β = 1.02). In the present analysis, then, I conclude the model is robust to this alternative specification. This, however, does not provide sufficient evidence to claim the Cold War was without effect on civil war duration. As discussed previously, the coding for external support does not contain an important aspect of Cold War effects, namely the magnitude of the aid provided by the superpowers. Returning to potential regional effects, an analysis was performed on multiple models that included regional control variables. Only the indicator variable for Eastern Europe and the former Soviet Union was statistically significant. Even so, it did not substantively or statistically alter the results of the original model. The coefficient was positive, correlating with shorter duration civil wars. Asia, Sub-saharan Africa, North Africa/Middle East, and Latin America control variables did not produce relevant results. The seemingly persistent anomaly of wars in Eastern Europe is deserving of further study, but its connection to an economic opportunity cost model of civil war remains unclear and hence it remains outside of this analysis. The concept of the conflict neighborhoods and the potential spillover effects of civil war have a more robust theoretical justification. Such contagion is possibly linked to the dispersed economic impact of factors that might lead to civil war or to physical spillover of refugees flowing into adjacent states (Salehyan and Gleditsch 2006). As well, neighborhood effects

95 Section can be observed when ethnic groups straddle political borders and conflict by the group in one state is echoed by violent conflict in an adjacent state (Buhaug and Gleditsch 2008). Fifteen conflict neighborhoods were defined and used to test for this type of effect. Since conducting hypothesis tests using Kaplan-Meier plots and log-rank tests is cumbersome for this number of variables, I instead construct a discrete variable for the 15 conflict neighborhoods and employ a stratified model. As I am more interested in the effects of the revenue source than I am in the conflict neighborhood effects per se, this method allows me to test the robustness of the revenue source findings. A Cox model stratified by conflict neighborhood allows the baseline hazards for each conflict to take different shapes instead of being just a multiplicative version of the overall baseline hazard (Cleves et al. 2008, 194). In running this alternative model on the data, I find that the statistical significance of the covariates are not adversely affected by conflict neighborhood. Another potential confounder of the model are the effects of multiple wars being fought in a single state. In such instances, there is potential for state characteristics entering the data repeatedly to bias the results. Because of the large number of states (34 with multiple civil wars), including individual state effect variables is unfeasible. Instead, a shared frailty model is employed that is the rough equivalent of a random effects Cox model for the estimation of within group correlation. In this case, the groups of observations are identified by the country code of the state engaged in the conflict. Within group correlation is measured by the parameter θ with θ = 0 implying no correlation. For this data, θ =4.41e 21, or very nearly zero, and statistically insignificant. As such, the shared frailty model fails

96 Section Looting Taxation dfbeta loot _t dfbeta tax _t dfbeta aid External Support _t 835 dfbeta lnpop Population _t Figure 3.10: Influential Observations to reject the null hypotheses that there are no differences induced by states experiencing multiple civil wars. To check for the effect of influential observations, an approximation using efficient score residuals is employed. The evaluation is based on approximating the difference between coefficients for the model estimated first with full data then with the ith observation removed (Cleves et al. 2008, 217). Intuitively, if the difference between the β s is close to zero, then the removed observation has minimal influence. When the difference is large, the ith observation is considered influential. Outlier plots for looting, taxation, external support, and population are presented in figure The observations are plotted by the COW war number. Table 3.13 follows with

97 Section War Number War Name Duration (in months) 731 Seventh Colombia ( La Violencia ) Angolan Control War Mozambique Second Uganda First Sri Lanka Tamil Second Afghan Mujahideen Kashmir Insurgents Tajikistan War Table 3.13: Outliers by War Number and War Name the relevant war numbers defined. The influential observations for each variable are omitted and the model is run repeatedly on the reduced datasets. As well, the model is run with all of the influential observations dropped. In none of these combinations does the revised data change the result of the model. 3.6 Counter-Examples While the model and the data provide strong support for the hypothesis, there are counter-examples in the data in which the regularity of these statistical correlations fail. To understand the context in which these failures occur, a closer examination is conducted of unusually short civil wars where rebel taxation and external support are observed. There are six such cases in the data and these are listed in table While it is not expected that the causal mechanisms developed in the previous chapter can account for these failures, counter-examples can provide a point of departure from which alternate hypothesis might be derived for further inquiry.

98 Section War Number War Name Duration (months) 782 Pakistan-Bengal Croatian Independence Nagorna-Karabakh (Azerbaijan) Abkahzia Revolt (Georgia) Sixth Iraqi Kurds Fifth DRC 7 Table 3.14: Short Wars with Tax and External Support Unusually short civil wars are defined as those with duration of less than 19 months. 19 months is the lower bound of the 95% confidence interval for median duration of civil wars in the dataset (median=24). In five of these six cases, the principal rebel group either begins the war as a de facto state, or rapidly assumes control of regional institutions and replaces the state in the contested area. In East Bengal, the Awami League party that formed the main body of the independence movement, won 160 of 162 seats in the 1970 election. They took control of the regional parliamentary government and its ability to tax and profit from natural resources primarily sold to India (Johnson 2005, 149). In advance of their war of secession, the Republic of Croatia, as a state within Yugoslavia under the 1974 constitution, had the capacity to make independent political decisions, raise revenues, and regulate its own military and police force (Meier 1999, 6). Nagorno- Karabakh was considered by 1989 to be operating a parallel government to that of Azerbaijan and this parallel government was recognized by Moscow and had the power to legislate and tax (Kruger 2010, 20). Abkhazia, Georgia was also an autonomous republic (ASSR) with the institutional trappings of

99 Section a state and the ability to legislate and make policy. Thus in 1992, the Russians supplied them with small arms, air support, and artillery to fight what was portrayed as a Georgian assault on their already established autonomy (Cornell 2002, 198,263). Here then the rebel group operated as a state in nearly every regard, fighting not to gain autonomy, but to retain it. Why were these wars short? Why were they not longer given the rebels ability to generate revenue and garner external aid? The answer may lie in the international system s perception of their place. Either the system is willing to allow them a place as a fully independent and separate state (as in the case of Croatia and Bangladesh), or the system fears their further devolution of the existing state structure (as in the case of Nagorno-Karabakh and Abkhazia). Thus it could be thought that the causal mechanism at work here is an exogenously imposed constraint, or relief from constraint, on the ability of the group to move beyond their autonomous position to one of sovereignty. Conceptualized in this manner, one then must consider the role of the most prominent regional actors on three of these cases. The role of Russia is overtly obvious in its importance in the latter two cases. The role of India is central, physically and politically, to the conflict in Bangladesh. In only a slightly more abstract way, the role of the western international community and its enthusiasm for post-cold War independence played a decisive role in Croatia. As rebel groups become only an epsilon s difference from a state, the interests of the international community are piqued and its influence becomes more pronounced. Ultimately, the conflict is then ended more rapidly in either pushing the rebellion to successful independence, or containing the

100 Section rebellion within the existing state structures. While the quantitative outcomes are at odds with the results of the model presented here, the qualitative assessment allows room for other causal processes to be at work in making these civil wars shorter than expected. 3.7 Conclusion The Cox proportional hazards model of civil war duration as a function of sources of rebel revenue produces results consistent with the assumptions of the model. These results are robust to alternative specifications, the use of various control variables, and are also robust to minor differences in coding criteria and influential observations. The results suggest that rebel access to sustainable means of economic revenue, specifically a system of taxation or support from a state external to the war, are correlated with an increased risk that the conflict will survive and continue into the next period. Civil wars in which rebels employed a system of taxes and civil wars where rebels received foreign economic, military, or material support are, on average, less likely to end in the following month than other civil wars. This provides support for the economic opportunity costs literature and for my theory of civil war duration through a direct observation of rebel sources of funding in civil wars since Subsequent chapters provide case studies of three civil wars in the Philippines. These in-depth examinations of conflict move beyond the strong statistical link between sources of funding and civil war duration provided by the quantitative analysis. They will provide an evaluation of the causal

101 Section mechanisms and causal processes postulated by the theory. The combined quantitative and qualitative approaches give a more robust accounting of civil war as economic opportunity.

102 92 Chapter 4 The Huk Rebellion 4.1 Introduction Consistent with the long history of resistance to Spanish and American colonization, the Huks emerged as an armed guerrilla group to contest the Japanese occupation of the Philippines during World War II. After the war, the Huks became the vehicle for a violent phase of continued conflict over agrarian reform. The growth of the Huks in the three years after Philippine independence led to their merger with the communist party of the Philippines and the outbreak of civil war. The divergent preferences of the urban communists and rural Huks were manifested in a strategy that failed to establish control of barrios and develop a permanent means to finance the war. Throughout this civil war, the Huks took on the role of roving bandits, extracting the resources for war predominantly by looting. As a result, the rebellion rapidly collapsed.

103 Section A History Of Resistance The path to independence for the Philippines was drawn out over three centuries of colonization and occupation by Spain, the United States, and Japan. From the very beginning, the people of the Philippines engaged in a wide range of resistance to these occupiers. Ferdinand Magellan was the first European to lay claim to the islands in March 1521 on a voyage for the Spanish crown. After a few weeks, he also became one of the first victims of violent resistance by Filipinos. In a brief battle between 60 of Magellan s men and 1,500 warriors of Mactan chief Lapulapu, the Spanish were defeated, and Magellan was killed (Abinales and Amoroso 2005, 48-49). Over the next fifty years, the Spanish continued their expeditions to the islands until, in 1571, Miguel Lopez de Legazpi conquered and settled Manila. Manila became the foothold for Spanish military rule and the influx of Catholic missionaries. Named in honor of crown prince Philip, later King Philip II, the Philippines became a point of exchange for silver from Spanish colonies in the Americas and Asian luxury goods, such as Chinese silk (Abinales and Amoroso 2005, 61). Once a year under tight regulation by the Spanish government, a Spanish galleon from Acapulco, Mexico would arrive to disgorge its load of silver as payment for a wide range of goods imported from China and destined for Spain (Barrows 2009, 195; Karnow 1990, 56). This economic construct for the colony had two important effects. The first was that it created the basis for an urban-rural split. The concentration of galleon trade in Manila was accompanied by the concentration of colonial bureaucracy and manpower in the capital city. This left most of

104 Section the rest of the rural Philippines to the unimpeded rule of the friars of the Catholic church with only the occasional interaction with Spanish government authorities (Abinales 2005, 67; Barrows 2009, 237). Second, it allowed the Spanish to maintain their colony with a limited military presence, again based primarily in the city. This facilitated a British seizure of the colony in the late 18th century and later the US takeover at the beginning of the 20th century. The British seized Manila in 1762 as part of the Seven Years War with almost no opposition by the Spanish. A force of 2,000 Filipinos loyalist provided the only resistance to the British assault (Abinales and Amoroso 2005, 71). When Spain resumed control of the islands, the galleon trade was in decline. The final ship laden with silver arrived in Over the next 100 years as a Spanish colony, the Philippines economy transitioned to one primarily dependent on export agriculture. By the time the galleon trade ended, Governor-General Jose de Basco y Vargas had imposed a plan for the cultivation of cash-crops such as tobacco, sugar, and hemp (Barrows 2009, 233). Nearly all of the Philippines export revenue was generated from these crops by the 1840s (Abinales and Amoroso 2005, 76-82). A hierarchical agrarian society evolved with peasants subsisting on plots managed by local elites, or principalia. The land, however, was ultimately owned by friars who held large tracts referred to as friar estates (Steinberg 2000, 22). This shift in the distribution of economic power to concentrated landholders generated tensions that were seed for later rebel movements. By the late 1800s, the economic development of the Philippines also created a class of wealthy, urban Filipinos. They began sending their scions to be educated in Europe. The most prominent of these was Jose Rizal who

105 Section returned to the Philippines to start a political reform movement. He was arrested and executed by the Spanish in 1896 after a parallel, but peasantled group under Andres Bonifacio attempted a short-lived armed rebellion. Emilio Aguinaldo, another well-educated son of a wealthy family, replaced Rizal as leader of the propaganda movement. Aguinaldo was exiled from the Philippines in an agreement with the Spanish government that was rumored to have paid him 800,000 pesos for leaving (Karnow 1990, 77). Although these revolts were of substantial importance to the trajectory of Philippine history, at the time they did not pose a significant threat to Spanish colonial power. The result did, however, reinforce a key cleavage amongst Filipino revolutionaries between the rural poor and the urban, educated elites. Spanish rule came to a quite sudden end in Just a month and a half after the Spanish-American war was triggered by the explosion of the USS Maine in Cuba, the ships of the US Asiatic fleet sailed into Manila Bay. Eleven of twelve Spanish warships were quickly dispatched by Commodore Dewey without damage or casualty to US forces. From his exile in Hong Kong, Aguinaldo had met with US officials in to discuss US intentions with regard to the possible independence of the Philippines. 1 Whilst the Filipino leader hoped for a swift transition to independence, the US moved forward with more imperial ambitions. After their defeat in Manila bay, the Spanish signed an agreement ceding control of the colony to the US. A dispute over the status of the Philippines ensued between the US and pro-independence groups under the leadership of Aguinaldo. The follow- 1 There is considerable debate about these events (see Silbey 2008; Goodno 1991; and Mahajani 1971).

106 Section ing year, this conflict between the US and the Philippines turned violent after months of tense standoff in the areas surrounding Manila. Aguinaldo fought a brief conventional war against US troops, but his limited forces were quickly overwhelmed and he fled to the mountains. From there he waged an insurgent campaign that would last for less than two years. Aguinaldo s rebellion was crippled by a lack of support from both the urban elites and rural village chiefs. Their objections to US plans for a colonial government were quickly dropped when they realized they were being offered the opportunity to share political power and assume the estates previously owned by the friars (Abinales and Amoroso 2005, 117). The urban elites and the rural chiefs thus had little incentive to support a rebellion that jeopardized these payoffs. Thus Aguinaldo survived as a roving bandit in unpopulated areas without access to means to supply his insurgent army. These economic opportunities for a select few at least temporarily displaced ambitions for independence. More importantly, the policies and practices of this new class of wealthy landowners would create an enduring cleavage in the Filipino political landscape. In the 30 years after the US-Philippines war, an agrarian reform movement gained momentum in the rice producing region of Central Luzon that surrounded the capital of Manila (see figure 4.1). This movement peaked in the mid to late 1930 s with the National Peasants Movement (KPMP). The KPMP drew its power from organizing tenant farmers to contest unfair and predatory practices of landowners, complicit police forces, and to a lesser degree the central government. It is important to note here that the KPMP arose separately from the primarily urban communist party again reflect-

107 Section Figure 4.1: Map of the Philippines from CIA World Fact Book

108 Section ing an urban-rural split in the opposition movements of the Philippines. Communist parties are often associated with peasant/tenant uprisings, but in the case of the Philippines the Communist Party of the Philippines (PKP) had its origins in the trade unions of Manila (Abinales and Amoroso 2005, 149). While there were PKP members who devoted considerable time to the peasant movement, the Lava brothers who formed the nucleus of PKP leadership were from a well educated landowning family who did not place great emphasis on rural matters (Kessler 1989, 29; Kerkvliet 1977, 97). The Quezon government outlawed the activities of the PKP almost at its inception in As a result, their membership dwindled and their organizational capacity to expand and incorporate the peasant movement at this time was limited (Kerkvliet 1977, 50). 4.3 The Huks The Japanese occupation of the Philippines during World War II transformed the existing structures of peasant-based reform movements into an armed insurgent opposition. The day after the attacks on Pearl Harbor, Japanese aircraft bombed American bases in the Philippines. Several weeks later, Japanese ground forced landed at Lingayen Gulf at the north end of the island of Luzon. On 11 March, 1942, MacArthur fled the Philippines leaving behind over 60,000 American and Filipino troops cornered on the Bataan Peninsula (Karnow 1990). In April 1942, these forces surrendered to the Japanese, ending conventional resistance to their invasion of the Philippines and marking the beginning of the Japanese occupation.

109 Section The occupation provided the impetus for the peasants, the communists, and other groups to join forces in a common armed struggle as the Hukbalahap (Kerkvliet 1977, 97). The People s Anti-Japanese Army, the Hukbo ng Bayan laban so Hapon, or Hukbalahap, or simply the Huks (pronounced hook-s ), was formed as a guerrilla army to counter the Japanese military occupation and oppose the collaborationist forces of the Philippines government. They hoped to sustain their resistance to the Imperial Army until the Americans returned with sufficient force to defeat Japan (Kerkvliet 1977, 99). The Japanese occupation provided a common enemy for this united front and consolidated the membership of the KPMP, and other groups, as Huks (Kerkvliet 1977, 97). Around thirty percent of the peasants in San Ricardo {a village in Central Luzon} were KPMP members. But during the Japanese time, ninety percent or more of the people supported the Hukbalahap (Kerkvliet 1977, 67). The main source of weapons and ammunition for the Huks were those left behind by the retreating US and Philippine armed forces. In ambushes and raids of small Japanese units, the Huks were able to collect additional weapons. There are few reports of supplies from the US during the occupation. There was at least one delivery of weapons via submarine during the war, but such events were intermittent contributions and not part of a regularized support system. Even with these limited supplies, the Huks were able to establish areas of rebel control in Central Luzon (Abinales and Amoroso 2005, 162). Their

110 Section ability to do so was facilitated by three main factors: landlord absenteeism, the limited value of money, and the independence of each Huk unit. Perhaps most importantly, many land owners fled to Manila. From there, along with the collaborationist government and Philippine Constabulary, they made weak attempts to enforce the conditions of tenant agreements. Their absence allowed the Huks, as a peasant based movement, to govern in a manner more beneficial to the tenant farmers. The isolation of these liberated zones, also reduced the value of money, and revived a barter system. Thus, the local Huk government provided protection and other public goods in exchange for food, clothing, and other support that the peasant population could supply (Kerkvliet 1977, 67-77). The decentralized Huk organization was also a key factor in Huk success during the Japanese occupation. This allowed town and village level units to operate and support themselves independently. Huk national leadership provided guidance and some coordination, but did not attempt to extract local resources to supply a national organization. The Huk army was organized as local squadrons of 100 to 200 fighters. By the end of the war, there were over 70 squadrons in Central Luzon totalling more than 10,000 guerrillas. Each squadron was encouraged to engage in three combat actions per month, but records show that an average squadron had less than a dozen encounters each year. This was consistent with the overall objective of sustaining the resistance until the Allies returned to defeat the Japanese. This low level of activity also reduced the need for resources and hence the demand on the Huk system of extracting resources (Kerkvliet 1977, 83-96). The Huks were thus able to act as stationary bandits and, with their limited

111 Section political objectives and the limited demands for supplies, they could sustain their rebellion through this system of taxes-in-kind. MacArthur s return to the Philippines in 1944 marked a return to a Philippine government dominated by urban and landed elites as well as a return to the uneven distribution of land and abuses of the tenancy system. Any reforms of the political and economic systems of the Philippines, or punishment of political leaders and landowners who collaborated with the Japanese, were sacrificed in favor of stability. From 1944 until August 1945, the US focus was on the use of the Philippines as a staging area for the invasion of Japan (Walens 1994, 8). The Huks refused to acknowledge the US backed government and maintained their claim of governorship of three provinces in Central Luzon (Lansdale 1972, 8). The new government took military action to disarm the Huks, delegitimize its leaders, and harass its supporters (Lanzona 2009, 79). The Huks were also classified as communist (Lansdale 1972, 7). Their motives of agrarian reform were interpreted as the work of propagandists to conceal their communist intentions (Valeriano 1961, 44). 2 The devastation of the Philippines by the occupation and then the Allied liberation created a chaotic governance scenario. As a result, President Truman ordered that those involved in continued agrarian unrest, principally 2 Despite this merger with elements of the communist party, the Hukbalahap remained separate from the Philippines Communist Party. Undoubtedly, some of them {the PKP} helped the Hukbalahap to become what it did. But it would be incorrect to say that these persons controlled the resistance in Central Luzon, as so much of its origins can be traced to local organizations and far more numerous were leaders who were not in the PKP (Kerkvliet 1977, 104). In contrast to the enlarged membership of most component groups of the Huks, the communist PKP shrank during the occupation. The communist PKP was specifically targeted by Japanese forces who proved successful in disrupting the party leadership, membership, and organization (Kerkvliet 1977, 100).

112 Section the Huks, be dealt with ruthlessly (Walens 1994, 16). This exclusion of the Huks, and the variety of peasant and workers organizations that had joined with them to resist the Japanese occupation, is a key factor in the unrest that would characterize the post-war years. It is also a basic element in the re-emergence of the Huks as an armed rebel group to fight a civil war in Early Violence, As the Philippines moved toward independence in 1946, many of the former Huks demobilized and joined the National Peasants Union (PKM). The PKM was a political party organized to press for agrarian reforms (Kerkvliet 1977, 121). In order to effectively compete in upcoming national elections for Congress, in an echo of the origins of the Hukbalahap, the PKM formed a loose association with the Lawyers Guild, the Hukbalahap s Veterans League, Congress of Labor Organizations, Fishermans Union, the Communist Party of the Philippines, and others (Kerkvliet 1977, ). The Democratic Alliance, as the association was known, won six seats in the April 1946 elections. When Congress convened in May, however, President Roxas and the Speaker of the House refused to allow these six congressmenelect to be seated (Lanzona 2000, 79). Charges of voter intimidation and election fraud were alleged as justification for the refusal (Kerkvliet 1977, 150). Among those refused their seat in Congress were Luis Taruc, the military leader of the Huks during World War II, and Jesus Lava, who would become general secretary of the communist party in After this, the

113 Section conflict between the state and the workers and peasants parties turned increasingly violent under the motto of bullets not ballots. Peasants who agitated for change were subject to intimidation, violence, and kidnapping. The main perpetrators of the violence were thugs working for landowners and the national police who either allowed the violence or directly participated. Residents of Central Luzon reported incidents of theft and rape in operations by the government that targeted armed rebels (Taruc 1953, 243). This early violence also included sporadic attacks against rural barrios, often with the help of the army (Kerkvliet 1977, 158). The Huks were accused of instigating the violence, primarily in Central Luzon. The Huks attacked police posts, kidnapped state sympathizers, burned their homes, robbed government offices, and ambushed police patrols (Lansdale 1972, 9). However, much of the Huk strategy remained on the defensive: evasion, hiding, and movement (Taruc 1953, 247). The Huks were armed mostly with weapons from the time of the Japanese occupation. The Huks depended substantially on popular support in the agricultural regions of Central Luzon, much as they had during World War II. So pervasive was the support for the Huk movement at this time that the Manila Times referred to these areas as Huklandia (see figure 4.2). Locals provided food, information, and other goods to assist the rebels. The poverty that pervaded the region, and the needs of an armed rebel group with more ambitious political objectives, forced the Huks to loot and steal money from the government and wealthy citizens who opposed their cause. As these funds were insufficient to purchase needed ammunition and supplies, the Huks also relied on raids of military and police supplies during

114 Section Figure 4.2: Huklandia map from Greenberg 1987, 46 this time. While the Huks had a base of popular support in the barrios of Central Luzon, they took on the role of roving bandits to finance this early stage of conflict. This early violent contact between the Huks and the state was sporadic and interspersed with lengthy, often months long, breaks of relative peace. The violence was sufficiently intermittent that, in Kerkvliet s (1977) interviews with villagers, they recalled with great specificity those few months in which attacks did occur. Most of the Huk actions were characterized as hit-and-run raids designed primarily as a show of force. The actions of the government of the Philippines were likewise inconsistent. In March 1948,

115 Section President Roxas declared the Hukbalahap and the PKM illegal organizations and called for the Huk annihilation (Kerkvliet 1977, 159). By May 1948, the government reversed its position and ordered a stop to operations against the Huks. Attempts to complete a settlement to satisfy the demands of the Huks then collapsed in late August 1948 and fighting resumed (Taruc 1953, 262). 4.5 Civil War, It was not a decision by either side that escalated the conflict to a civil war. Rather, it was actions by both the rebels and the state that elevated the stakes and intensified the fighting. Discontent over failed negotiations of 1948 and the election outcomes of 1949 generated rapid growth for the Huk guerrilla army and expanded support for the communist party. The increased size brought the Huks into a closer partnership with the communist party. These changes in the rebel group took place in the shadow of major shifts in the international system that had substantial impacts on how the government of the Philippines engaged the rebels. The success of the Chinese communist revolution in 1949 and the invasion of South Korea by North Korea in 1950 had significant repercussions. Instability in the Philippines was recast as part of a global communist push that threatened US interests in southeast Asia and Japan. The larger and more aggressive Huks then had to wage war against a US-backed military with new equipment, training, organization, and leadership.

116 Section Reform of the AFP The behavior of the government of the Philippines changed markedly as a result of actions taken by its main supporter, the US. The level of assistance allocated by the US for the Philippines is illustrative of this substantial shift in US policy. In the fall of 1947, President Truman denied a request for $9.4 million in aid for rebuilding the Philippines, citing a lack of domestic support for continued funding to a country that was on the periphery of US interests. As late as mid-1949, the CIA assessed that the Huk rebellion did not pose a significant threat to either US interests or the Philippine government (Walens 1994, 46). This changed after the Nationalist government of China fell to the communist forces of Mao in For the fiscal year 1950 budget, Congress approved $12 million in material assistance to the Philippine army. The invasion of South Korea in June 1950 further spurred the US to increase its support for the government of the Philippines. With the war in Korea and the sudden perceived vulnerability of Japan, the US actively sought to secure other territories in the western Pacific as bases to ward off further communist inspired adventurism. In this light, the Philippines became central to US interests and hence the US made extensive efforts to neutralize the instability caused by the Huks. The US greatly increased its economic and military aid to the Philippines. By 1952, an additional requests for nearly $50 million in aid for the AFP would easily win approval from the President and Congress (Greenberg 1987, 109). In this way, the

117 Section US embedded itself in the Philippines counter-insurgency. 3 Equally important as the deluge of materiel was the overhaul of the structure of the fighting forces. The AFP took the lead role in the conflict and was reorganized into 26 Battalion Combat Teams (BCTs), each responsible for a specific geographic sector. 23 of the 26 BCTs were concentrated in Huklandia to more aggressively engage the Huks. A military intelligence command was created to better develop information on enemy disposition and coordinate counter-insurgency operations. Filipino soldiers and officers were also provided training in the US at basic and advanced courses in nearly every military specialty, including communications, supply, maintenance, infantry, leadership, and strategy (Greenberg 1987, 104). The AFP were transformed from a post-world War II organization intended to police the reconstruction of the Philippines to a fighting force focused on counter-insurgency The HMB/Huks The number of armed insurgents fighting for the Huks peaked in 1950 (see figure 4.3). Both official figures from military intelligence estimates and those reported by former rebel leaders place the strength of the Huks at approximately 15,000 in 1950 (see Kerkvliet 1977, Taruc 1967, Pomeroy 1963, Greenberg 1987). The PKP estimated the size of the movement s popular base of support at nearly 1 million in Huk leaders, such as Luis Taruc, saw the need for greater organizational structure to harness the 3 Almost literally in bed with the Filipinos as the new Secretary of Defense Raymond Magsaysay and US Army advisor Lt. Colonel Edward Lansdale shared a bedroom on the JUSMAG compound in the early months of the war.

118 Section Huk/HMB strength Strength in Thousands Year Figure 4.3: Rebel Strength in Thousands derived from Greenberg 1987 rebel movement. Thus, at a critical time when armed conflict had begun and the intensity and scale of armed rebellion was increasing the Huk army lacked a national political party to provide the structure and organization necessary to move from a rebellion of self-defence to a sustained armed struggle for control of the state. Concomitantly, the communist party saw great potential in the idea of incorporating a pre-existing guerrilla army into its revolutionary movement. Thus the two groups merged in what appeared to be an ideal partnership for civil war. The Huks changed their name to the People s Liberation Army, or Hukbong Magpapalaya ng Bayan (HMB) when they became the military arm of the PKP, but they were still widely referred to as the Huks. This reflected the essentially unchanged, and primarily rural, composition of the armed

119 Section insurgents (Lansdale 1972, 12). With the addition of a well-seasoned military arm, the PKP politburo convened in early 1950 to identify a strategy for their revolution. They decided that the time was ripe to shift from a military strategy of guerrilla warfare and for the HMB to directly confront the military in a war for control of the central government (Kerkvliet 1977, ). The PKP was optimistic that Manila would be captured as early as the end of 1950, but no later than the end of 1952 (Taruc 1967, 87). Other than a failed assassination attempt against President Quirino after the November 1949 elections, this was the first time that the Huks, now the HMB, had targeted the national government of the Philippines. These moves did not resolve the substantial friction between the PKP and the HMB leadership. Luis Taruc, the overall military commander of the Huks, felt the changes were premature and reflected the flawed conclusions of a politburo that lived in Manila whilst the war was waged outside the cities (Taruc 1967). Jose Lava, general secretary of the PKP, believed a revolutionary situation existed that supported these changes. Despite these differences, Taruc consented to the shift in goals away from earlier limited demands for agrarian reform and toward a civil war that targeted the central government in Manila (Greenberg 1987, 67). At the top tier of leadership, the PKP took the dominant role with the majority of decisions generated by the general secretary, a politburo, and a secretariat. The secretariat was comprised of departments for education, communication, the military, and finance. Below this, the Huks were reorganized into regional commands (Recos) to align with the regional committees of the political education department. This allowed the PKP to

120 Section coordinate military and political activities across the regions and down to the village level. In pushing acceptance of the existence of a revolutionary situation through the politburo, Jose Lava was also able to gain approval for shifting military and political control of the PKP/HMB to himself as general secretary (Taruc 1967, 74). The accompanying re-labeling and re-organization of the Huks as the HMB created regional commands that flowed finances to the politburo in Manila who claimed to have unique knowledge of how to finance the war (Kerkvliet 1977, 230). In October ,000 pesos was stolen by the PKP/HMB from a payroll railroad car and half of this was is forwarded to the national finance department in Manila (Kerkvliet 1977, 215). The new organization yielded some early and impressive results. The HMB was able to generate simultaneous attacks in most of its ten regional commands. These attacks commenced in August 1950 and were ostensibly set to commemorate the failed 26 August 1896 Philippine Revolution against the Spanish (Kerkvliet 1977, 214). Estimates of armed insurgents involved in the multiple attacks range from 500 to 2,000. Several hundred Huks raided a military hospital in Tarlac province where they killed at least 30, raped nurses, and executed the camp commander. As well, the rebels looted medical supplies, food, clothing, weapons, and 140,000 rounds of ammunition (Lansdale, 26; Taruc 1967, 86). In Laguna province, Huks of Reco 4 captured the city of Santa Cruz for a night and stole 86,000 pesos from the municipal treasury (Greenberg, 66; Kerkvliet 1973, 214; Taruc 1967, 86). Thus even after the merger of the PKP and the HMB/Huks, the rebels were still acting as roving bandits to finance their war. Other regional

121 Section commands also took part in these anniversary attacks, raiding garrisons of the Philippine Constabulary, briefly occupying town centers, burning homes, and ambushing PC patrols. For most observers of the conflict, these changes represented a new phase in the conflict and the start of a civil war Civil War The start of the civil war in 1950 is really the start of the end of the Huk rebellion. With the peak in membership and popular support in 1950, accompanied by the largest coordinated attacks, the war for the rebels in 1950 was, for the most part, going well. The area in which the rebels failed was in creating economic opportunity in the war. The earlier promises of the Huks had received broad support. This was the promise of displacing the existing landowners and relieving tenant farmers of their debt burden. The new national strategy of the PKP-HMB bypassed this phase of establishing control of barrios and delivering payoffs to supporters. The PKP-HMB needed to identify an alternative means of financing, but were ultimately unsuccessful in doing so. The tax that was imposed on families was a tax-in-kind. Families would put aside a cup of rice each day for the Huks. Additionally, the Huks would collect a portion of the rice from each family s yield at the harvest. These contributions, however, were intermittent and were often only provided when there was enough for the family to spare. They yielded an insufficient amount of rice to support the army of insurgents (Kerkvliet 1977, 167). The PKP-HMB did not have the capacity to further exploit these natural resources on which their supporters were dependent for their

122 Section minimal livelihood. The raid tactics of the HMB meant that barrios did not have a sustained and overt armed insurgent presence to keep landowners, police, and the military at bay. Thus, the PKP-HMB could not control the domestic market for production and sale of crops to exploit them for profits. Likewise, they did not demonstrate an ability to control the export market. Reported instances of levying a rebel fee or tax on the transportation and production of goods in Central Luzon were limited and episodic. External support for the PKP-HMB was also lacking. In this early stage as a communist state, the People s Republic of China was not in a position to provide foreign aid to the PKP-HMB. Whilst the Huks looked to the Maoist revolution in China for inspiration and emulation, there is scant evidence that the PRC contributed significantly to the Huk rebellion. There is also little to suggest that the Kremlin provided military or economic aid to the Soviet inspried PKP. It is quite plausible that Soviet Union opted to stay out of the conflict based on an assessment that the overwhelming interest of the US in keeping the Philippines as a quasi-colony provided little room for successful intervention. Even within the Philippines, but outside of Luzon, the PKP-HMB found it difficult to garner support, popular or monetary, for their revolution. Thus, the PKP-HMB continued to rely on looting as their principle source of financing, much as the Huks had done for the previous four years. The PKP-HMB organized Economic Struggle (ES) units in the political structure of each region. They specialized in robbing landlords, merchants, and politicians, thought to be hostile to the Huks. Their main tactic was to dress in the uniform of government forces to infiltrate towns (Pomeroy 1963, 163-

123 Section ). The result was predictable in that the rebels could not purchase needed weapons, ammunition, or supplies. They continued to fight with arms acquired during World War II. New recruits were not issued weapons, having to wait until they could take one from a dead government soldier (Kerkvliet 1977, 211). Even when the Huks were able to acquire machine guns from the AFP or PC, they lacked funds to purchase black market ammunition to use them (Pomeroy 1963, 159). The Huks were restricted to the use of small arms and did not regularly use grenades, mortars, or any other heavy weaponry. Insufficient ammunition, guns, money, and supplies impeded the employment of the full strength of the HMB. Luis Taruc estimated that no more than 4,000 of 15,000 armed insurgents could be mustered for an assault. Resources also constrained the purchase of much needed communications equipment to coordinate quickly amongst the regional commands (Taruc 1967, 88). The simultaneous rebel attacks of August 1950 were built upon a common date, but the Huks did not have the ability to organize a rapid counter-action against later AFP operations. In addition to these financial shortcomings, the PKP-HMB suffered some considerable operational setbacks. In October 1950, the government conducted a well planned raid in Manila that effectively decapitated the organization. The entire PKP politburo was arrested, including General Secretary Jose Lava. In addition, tactical plans and detailed notes on the structure and composition of the PKP and HMB were found. Yet even with this loss of leadership and the organizational setbacks, the new General Secretary (and brother of the previous one) Jesus Lava pushed to continue the

124 Section military and political strategy as originally planned (Pomeroy 1963, 105). This perceived recklessness by the secretariat served to renew the cleavage between the military and the political sides of the movement (Taruc 1967, 93) In response to the October politburo raids, the Huks launched a reprisal attack against a hostile village in November They massacred nearly the entire population including women and children (Taruc 1967, 89; Greenberg 1987, 128). In response, the government launched the first of a series of large, offensive operations. The operation targeted Huk camps near Mount Arayat (Greenberg 1987, 129). The AFP also at this time began to employ observation aircraft provided by the US to find and harass PKP-HMB forces near their forest production bases. The disruption of leadership and limited supplies negated the organizational gains of the PKP-HMB alliance. The Huks were again forced to fight as small bands, spending much of their time simply acquiring subsistence supplies and guarding against AFP attacks (Greenberg 1987, 140). The government s counter-insurgency strategy also included a substantial socio-economic component. The Economic Development Corps (ED- COR) program helped identify and develop new land for the re-settlement of farmers. 4 Surrendering or captured, Huk soldiers and their families were allowed to homestead on modest plots with government support. Military engineers helped clear land and build roads (Lansdale 1972, 54). Although 4 The public land used for this project was predominantly in the southern province of Mindanao. The homesteaders of EDCOR were viewed by the Muslim population of Mindanao as an attempt by the central government to Christianize the region. This directly contributed to the outbreak of violence leading to the First Moro civil war.

125 Section small in scope (only three EDCOR projects from were undertaken with approximately 50 former Huk families per project), the psychological impact was widespread. Thus, as the PKP-HMB lost their ability to operate in populated areas to defend against the abuses of the state, the government exacerbated the Huk s situation by providing at least a perception of better conditions in areas of state control. The size of the insurgent force decreased rapidly. By early 1954, the AFP estimated fewer than 1,500 armed rebels remained. The largest military operation to date was launched in April 1954 to root out remaining pockets of resistance. With more than 5,000 soldiers, the AFP pushed further west of Mount Aryat. The operation culminated in the surrender of Luis Taruc, the former supreme commander of Huk forces. His May 1954 capture marked the end of effective resistance by the PKP-HMB. The military continued to pursue the rebels to the Pacific coast for an additional two months (Greenberg 1987, 140). In approximately three and a half years of civil war, an estimated 9,700 Huks and 1,600 government forces were killed (Wayman and Sarkees 2010, 112). 4.6 Huks and Theory of Civil War Duration The economic opportunity theory of civil war duration provides a plausible explanation for the events that unfolded in the Huk rebellion. Without making the transition to being stationary bandits, the Huks failed to generate the resources necessary to continue financing the fight. The rebels did not control areas sufficiently to provide public goods to their supporters and con-

126 Section sequently the were unable to establish a more permanent means of finance. The Huks could not develop the capacity to sustain the civil war with the state. Consequently, the duration of the Huk rebellion is much shorter than the subsequent wars fought by the Moros, and the New People s Army. I next present a discussion of the Huks as roving bandits followed by a brief examination of alternative explanations for the relatively short duration of the war Huks as Roving Bandits The HMB relied almost exclusively on raid techniques. They did not establish a long term, or in many cases even an overnight, presence in barrios and villages. Even at the height of the strength of the Huk army, they withdrew to bases in the forest before the sun rose (Pomeroy 1963, 66). They did not hold liberated areas nor did they confiscate land as they had done during the Japanese occupation (Kerkvliet 1977, 230). Retreating to the forest exposed weakness in the rebellion, leaving the people naked to suppression... helpless before abuse by government forces (Pomeroy 1963, ). It is important to differentiate between popular support and control of an area. The PKP estimated that as many as 1 million people supported the rebels. More conservative government estimates put the number between 250, ,000. Either estimate, however, indicates broad support for the Huks. Popular support translated to intermittent provision of resources, including the rice tax, information, and the rare cash contribution. Without control of an area, though, the Huks could not create an environment necessary for a more regular extraction of resources. Even in pro-huk villages,

127 Section tenant farmers were still struggling to satisfy other bandits - namely the land owning principalia with the backing of the state. Businesses looted by the Huks still had to pay taxes and bribes to government officials. By not providing this public good of protection, the Huks as roving bandits could not enter into an agreement with the people and businesses to extract more regular contributions as a tax. I propose two plausible explanations for why the PKP-HMB did not understand the need to become stationary bandits. The first is a failure in leadership. During World War II, the PKP were adamantly against the Huk strategy of overtly opposing the Japanese by establishing local Huk governments in the barrios of Central Luzon. Vincente Lava, general secretary during World War II, had attempted to persuade the Huks to retreat for defense, hide in the mountains, and avoid battle (Kerkvliet 1977, ). When the PKP was ascendant in the late 1940 s, and the Huks became their military arm, General Secretary Jose Lava again was dismissive of the Huk strategy to establish secure areas and local bases of support. The PKP directed instead that they move to an assault on the national government. Given the nepotism of the urban, educated, and elite PKP leadership, from Vicente to Jose to Jesus Lava, it is plausible to suppose a distorted view of the Philippines political landscape existed in the communist party. From their perspective, rural peasant villages had some limited utility in providing foot soldiers, but they were second to the urban proletariat as the vanguard of the revolution. Hence there was minimal need to focus effort on providing rural peasants a benefit from fighting in the form of a secure PKP-led local government.

128 Section A second explanation also speaks to blindness in the communist party. The regional political committees spent much of their time attempting to educate Huks and rural supporters of the validity of the communist cause. They invested considerable effort into the creation of propaganda and its distribution. The core belief was that there was a tipping point after which the conversion to true believers would be self-sustaining and contagious, engulfing the Philippines, and ushering in the communist state. In this sense, military control of an area was unnecessary. Such an event was even believed to be able to occur independent of the armed struggle. Soon enough the people would exclude the government. Reflecting on the disconnect between the communist leadership and the people, William Pomeroy, a native of Western New York and leader of the education committee in the PKP politburo, wrote We had thought that the people moved at our pace, to the rapid click of the mimeograph machine. 5 We had thought that the morale and the discipline in this camp was the morale and discipline everywhere. We had thought that by the leaders setting a high tempo we could set a high tempo of the revolution. We have been living in a fool s paradise (Pomeroy 1963, 157). The Huks remained roving bandits and became increasingly bitter about the lack of evolving support from the people. Huk attacks no longer targeted landowners, the wealthy, and government officials. Increasingly, whole 5 The mimeograph was the main means of reproducing communist propaganda at forest outposts.

129 Section villages were suspected of being unsympathetic to the rebellion and thus became the targets of Huk raiding. Canned goods, clothing, shoes, anything useful in the mountains were stolen from villagers (Pomeroy 1963, 66). Rocamora (1978) argues that because the military aspects of the rebellion were not successfully tied in with its political aspects, and because the rebellion failed to generate basic changes in relations of production in the countryside, there was no basis for peasants to continue fighting and hence even popular support for the rebellion declined (Rocamora 1978, 404). 4.7 Alternative Explanations According to Collier, Hoeffler, and Soderbom (2004), the Huk rebellion should have been less likely to end early. They reference a theory of war as business. If the gross costs of rebellion to the rebel group are low (low opportunity costs for recruits) and if rebel revenues during conflict are high (high commodity prices), rebels may be better off than prior to the conflict and therefore under little pressure to reach a settlement (Collier, Hoeffler and Soderbom 2004, 267). The opportunity costs for recruits was low in the Philippines with GDP per capita at just $147 in The market trends were also quite favorable for a lengthy war during this time. The world export price of rice rose approximately 15% during the four years of conflict from US$137 to US$158 per ton. Likewise, the amount of rice harvested in the Philippines increased by 15%

130 Section from 2.8 million tons in 1951 to 3.2 million tons in 1954 (International Rice Research Institute 1987, 196). 6 Despite this, the support for rebellion in the rural population and the number of armed insurgents decreased rapidly during the war (see figure 4.3). Whilst derivative of the same fundamental approach to civil war duration, that wars are longer when rebels have an opportunity to profit, the Collier, Hoeffler, and Soderbom findings do not accurately capture the processes at work in the Huk rebellion. The grievance approach also has minimal purchase in this case study. Grievances can be addressed in two ways. The first is by a direct payment to the aggrieved (the peasants). The most prominent of this type of program was the Economic Development Corps program of homesteading former Huks and their families. Its impact was limited to fewer than 200 families, representing a tiny proportion of the Huks and their supporters for whom land reform remained the central issue for the rebellion. The second means of addressing grievances is through policy changes that impact living and working conditions. Wages during this time, however, remained static (Kessler 1989, 26). The endless cycle of dependency of tenant farmers on the landowners also did not significantly change by war s end. Kerkvliet (1977) documents nearly identical claims and charges levied against landowners by tenant families in 1954 as there were during the war (Kerkvliet 1977, 268). If anything, the persistence of grievances should have lengthened the war. I thus conclude that a grievances approach is not convincing as a causal explanation for the duration of the civil war statistics on rice production were unavailable.

131 Section A final possible alternative explanation would be to attribute the brevity of the civil war to the counter-insurgency efforts of the Philippine government and the assistance of the US. Certainly, the theoretical conclusions of this case study are not exclusive and allow for counter-insurgency to be seen as contributing to the end result. 7 To say that counter-insurgency success is the primary explanation, however, ignores a significant aspect of the historical context. The US exported the lessons learned from defeating the Huks to the growing insurgency in Vietnam. Many of the same strategies and tactics were implemented by many of the same practitioners, most notably Lt.Colonel Lansdale, with obviously different results in the Vietnam war. Thus, while an effective counter-insurgency effort by the state is perhaps necessary to shorten a conflict, it is at least equally important to take into consideration the strategic actions of the rebels in financing their rebellion. 4.8 Conclusion Although initially successful as a military force, the Huks were unable, or unwilling, to control populated areas. Absent such a monopoly of the use of violence in particular areas, the Huks could not establish a permanent system of financing. Without this profit potential, the Huk leaders and their followers could not sustain support for the rebellion and the Huks were unable to acquire the resources necessary to sustain the fight. The ill-fitting merger with the Philippines communist party (PKP) was an effort to rectify 7 Collier, Hoeffler, and Soderbom would disagree, arguing that external military intervention on the side of the state is ineffective in reducing the length of a civil war (Collier, Hoeffler and Soderbom 2004, 267).

132 Section this shortcoming, but divergent goals, strategies, and membership made the merger unsuccessful. The Huks returned to looting as roving bandits in an effort to create a short term fix to their problem of economic opportunity. Ultimately this was unsuccessful and led to the shorter duration of the war. The short length of this war fits well with the economic opportunity theory of civil war duration.

133 123 Chapter 5 The Moro War 5.1 Introduction The influence of Islam in the Philippines pre-dates Spanish colonization. It was the Spanish, however, that made salient the Muslim-Christian cleavage during their three hundred year rule of the Philippines, setting the stage for an independence movement derived from this distinction. The Moro nationalist movement organized for violent conflict with the state in the late 1960 s with essential assistance from sympathetic Muslims in Malaysia and allies in the Arab states. The lead organization for Muslim Filipinos, the Moro National Liberation Front (MNLF), depended nearly exclusively on this considerable external support to finance their war with the government of the Philippines. Instead of exploiting domestic sources of financing, as either a roving or stationary bandit, the MNLF acted as Robin Hood with its almost exclusive control over the inflow of external aid. When this aid diminished, the MNLF was forced to settle for an agreement of autonomy within the Philippine state and the war rapidly concluded in January 1977

134 Section after six-and-a-half years of fighting. 5.2 A Brief History of Muslims in the Philippines Drawing from conflict between the Spanish and Muslims in North Africa and Europe wherein Muslims were referred to as Moors the Spanish labeled the Muslims in the Philippines, Moros. By the time of the Spanish arrival the Muslims had established a distinct cultural identity and built governing institutions at the village and sultanate level. The Moros effectively resisted Spanish efforts to subjugate them. In the course of this conflict, the Spanish made the Christian-Muslim cleavage a politically salient and persistent feature of the Philippine political landscape Pre-colonial History Geographically, the southern Philippines is described as the southern-most island group of the Philippines that is dominated by the island of Mindanao. It also includes the island provinces of the Sulu archipelago, Tawi-Tawi, and Basilan (see figure 5.1). As much as 400 years prior to Magellan s discovery of the Philippines, Muslims had used the islands in the south, most notably Jolo, as way points in trade routes between India and China (Yegar 2002, 185). Jolo is the largest island in Sulu and served as a key link between Muslims and the broader Muslim community, particularly in Malaysia. This connection to other Muslim communities in Asia and to Arab Muslim communities is an essential component of the Philippine-Moros war in the 20th century.

135 Section 5.2 Figure 5.1: Map of Southern Philippines from philippinemaps.ph 125

136 Section Obviously not indigenous to the islands, Islam developed atop the linguistic, ethnic, and religious patterns of the region. Islam expanded predominantly through intermarriage and assimilation with local tribes and customs. Even though Islam became the overwhelming majority religion earliest estimates are that 90% of the population belonged to the Muslim faith the region remained far from homogenous. There are thus some 13 ethnic language groups and a broad range of orthodoxy and paganism amongst Muslims in Mindanao and the Sulu archipelago (Yegar 2002, ). The largest ethnic groups are the Tausug, Maguindanao, and Maranao. 1 Village chiefs, or datu, were the religious, political, and economic leaders of the local communities and this structure was incorporated into the politico-religious hierarchy of Muslim sultanates that were established throughout the region Spain and the Moros From the perspective of the Muslims, the arrival of Spanish explorers and colonizers halted the northward spread of Islam in the Philippines. From the perspective of the Catholic friars, the Moros frustrated their efforts to completely Christianize the indigenous populations. The Spanish crown, however, remained committed to extending its rule throughout what they had claimed as the Philippines. Muslim piracy presented a persistent, although relatively low cost, irritant. As well, the Spanish colonial government was keenly interested in exploiting the economic potential of Mindanao s natural resources, most notably its vast supply of timber (Yegar 2002, 200; George 1 The Tausug are from Sulu and Zamboanga, the Maguindanao from Cotabato, and the Maranao are from the Lanao Lake area.

137 Section , 221). Over three centuries, the Spanish and the Moros engaged in a range of violent conflict in which the Spanish were largely unsuccessful in conquering the Moros. The main military force employed by the Spanish in these battles, however, were native Philippine soldiers, or indios. This is the term applied to the indigenous Christian population, converted by the Spanish friars. 2 Thus, not only did the Spanish import the Christian-Muslim conflict from Europe, but by using indios to fight the Moros they deeply embedded this cleavage in the culture of the Philippines. The animus that evolved from this conflict also served to further develop a common Muslim identity amongst the disparate tribes. This Muslim identity survived repeated attempts to incorporate the people of the Southern Philippines into a broader Filipino nationalism. Of note, Muslim leaders refused Emilio Aguinaldo s request for support in fighting the Americans at the beginning of the twentieth century. In their eyes, Spanish, American, and Filipino Christians were equally bad. The Moros referred to all three as gobrirno a sarwang a tao the government of an alien people (Canoy 1987, 147) US and Moros The United States replaced the Spanish as colonial rulers of the Philippines after a very brief battle in Manila Bay in The US renewed efforts to 2 The Spanish and the friars of the Catholic church used indios as one of many terms to distinguish between classes of citizens. Peninsulares were Iberian born Spaniards and criollos were Spaniards born in the colonies. Persons of mixed-heritage, with at least one Spanish parent, were mestizos (Karnow 1990, 62).

138 Section draw the southern provinces more completely into the Philippines. Although initial pronouncements indicated a willingness to allow the Moros greater autonomy than under Spanish rule, US policy in Mindananao soon fell in line with policy toward the Philippines in general. The Philippine colony was to be guided toward democratic rule and eventual independence. In practice this meant that traditional power structures, such as the datu, were to be replaced with a westernized vision of indigenous democratic control. For the US, the efforts to civilize the Moros were again conducted forcefully. Intense guerrilla fighting took place from 1901 to The fighting was done almost exclusively by US soldiers and, by 1913, the region was considered to be effectively subjugated to US colonial rule. Then, beginning in 1914 under the guidance of President Wilson, rule over the region was transferred to Philippine politicians and bureaucrats. This was in-line with the changed makeup of the national Philippine government to one in which Filipinos held majority control, but that was still under the executive power of a US governor-general (Yegar 2002, 220). Filipino control, however, did not equate to Muslim-Filipino empowerment. Rather, the officials who assumed positions of authority in Mindanao and Sulu were overwhelmingly Christian Filipinos from the northern island, and seat of national power, Luzon (George 1980, 78). The effect of US and guided Filipino governance of the south was largely the same. Their policies served to undermine the traditional datu power structures in the South. The datu lost the power to socially manage their communities and adjudicate disputes. More importantly, the datu lost their sole power to tax their subjects. Further, with the privatization of land

139 Section ownership, the datu were also deprived of their ability to manage economic affairs (Yegar 2002, 225) Republic of the Philippines and Moros After the Philippines became an independent state in July 1946, the new government pursued policies of national integration to consolidate and grow the state. The Philippine government treated Mindanao as the last frontier and the most easily exploitable area for expansion. The newly independent Philippine government continued a policy begun by the US to more effectively exploit the natural resource potential of Mindanao through the re-settlement and homesteading of northern Filipinos in the South. These agricultural colonists, as the US had termed them, included some 100 American plantation owners in the 1920 s that were essentially imported to better develop timber reserves, agricultural crops, and mineral wealth (Yegar 2002, 222; George 1980, 221). The resettlement greatly accelerated in the 1950 s. As discussed in the case study of the Huk rebellion, the government of the Philippines established a program to redistribute land to tenant farmers of central Luzon. Through this effort, some 1,500 immigrants from the north took up residence in Mindanao from 1950 to In the ensuing four years, 23,400 families would be re-settled with government sponsorship as part of the Economic Development Corps program (Yegar 2002, 247). This, however, represented a fraction of the total demographic shift taking place in the south. In 1903, a US census reported that 76% of the population in Mindanao was Muslim. Approximately 700,000 Filipinos then moved to Mindanao from 1903 to Nearly double that number, 1.25

140 Section million, would follow by 1960 (Yegar 2002, 246). By the 1980 s, Muslims were a minority of only 20% to 30% of the population (Yegar, 247; Majul, 30). This marginalization of the Moro population, coupled with the breakdown of the datu structure, would present barriers to Moro nationalist groups organizing and financing rebellions. Parallel to the government s re-settlement policy was a national scholarship program for Muslims. As part of the Commission on National Integration (CNI), higher education scholarships were provided to Muslim students to enable them to attend university in Manila. From 1958 until 1967, 1,210 students were funded by the Philippine government (Yegar 2002, 245). The intent of the program was to educate the next generation of Muslims in the hopes of bringing them into conformance with Christian Filipino attitudes and practices. Paradoxically, many of these CNI students, along with middle-class Muslim students who attended university in Arab countries, would become the nucleus of nationalist and Islamist independence movements in the 1960 s and 1970 s (Majul 1988, 33). In a sense, it is the training provided by the state that enabled Muslims to organize a rebellion based on cleavages made salient by state efforts to otherwise displace Muslims. As the future leaders of Moro separatists were receiving intellectual training at university, there was also on-the-job training in violent conflict occurring in Mindanao. Low level violent confrontations between Moro and Christian settlers increased in frequency and intensity. Philippine landowners for years had employed katiwala, or overseers, to manage the tenants. Increasingly, however, katiwala had become enforcers to insure adherence to the landowner s policies and collection of debts (Kerkvliet 1977, 10). During

141 Section the 1960 s in Mindanao, these enforcers transitioned from one or two thugs to private militias hired by wealthy landowners to protect their property in disputes with Moros. In response, Muslims organized to defend themselves, challenge homesteaders, and enrich themselves through looting and plunder (George 1980, 155). The Philippine Constabulary, widely viewed as siding with land owners, estimated there were some 20,000 active gang members in Mindanao (George 1980, 141). Describing the chaos of this environment of gang-on-gang violence, Canoy (1987) wrote, the American wild west was nothing compared to Mindanao (167). 5.3 The Jabidah Massacre as Watershed Event The first Moro group to demand independence from the Philippines was the Muslim Independence Movement (MIM), later renamed the Mindanao Independence Movement. MIM was formed in May 1968 by Datu Udtug Matalam as a means to publicize his and other datu s dissatisfaction with the current state of affairs and to press for a greater role of Islam in the region s administration (Yegar 2002, 253). 3 The Jabidah Massacre was the trigger event for the group s formation and the public declaration of their intent to create an independent state. Two aspects of this event are essential to understanding the war that would eventually follow. The first is that the perpetrator of the killings, and hence the object of Muslim animosity, was the Philippine state. The second is that the killings occurred within a Philippine government plot to attack a neighboring Muslim state. Together, 3 Datu in this usage is a title to indicate Matalam s stature within the community even though he was, in actuality, no longer the political head of Cotabato.

142 Section this recasts the conflict in the southern Philippines and opened the door to assistance from outside states and organizations. In 1967, the Armed Forces of the Philippines (AFP) began a secret program to recruit Muslim Filipinos to be trained as commandos for the eventual invasion of Sabah. Sabah is the northern and easternmost province of Malaysia, on the island of Borneo, and at the southern end of the Philippine Sulu archipelago (see figure 5.2). The region was believed by President Marcos to be part of his conceptualization of a greater Philippine nation (Vitug and Gloria 2000, 4). Oplan Merdeka, as the secret operation was named, involved several hundred trainees. When the training moved out of Mindanao to the island of Corregidor in Manila Bay, the Muslim trainees began to agitate for back-pay, contact with the outside, better food, and improved living conditions. In March 1968, possibly expecting a mutiny, the AFP executed the trainees. Reports vary regarding the number of deaths, from 11 according to a widely discredited government investigation to 68 according to a Muslim separatist group s accounting (Majul, 40; Vitug and Gloria, 14; George, 123). At least one of the trainees escaped and the story of the massacre rapidly spread. It is not hyperbolic to describe the Jabidah Massacre as a watershed event in the chain of events leading to civil war. Building on the Christian- Muslim cleavage that had erupted into low level violent conflict in the previous decade, it provided the spark that united Muslims in opposition to the Philippine central government. As the target of Oplan Merdeka, Malaysians were also moved to action by this event. Jabidah provided the motivation for Muslim Malaysians and the Malaysian government to actively support

143 Section Figure 5.2: Map showing location of Sabah from bbec.sabah.gov.my the Moro rebels. Most critically, this support included military training for Filipino rebels. Those selected for training were predominantly of the new generation of politically aware Moros who had been student activists at universities in Manila and abroad. Ultimately, the Jabidah Massacre put in place the external support necessary to bridge the gaps in the Moro community amongst popular supporters, violent gangs, and new leadership. In so doing, the Moros transitioned from contesting the property rights of private citizens in Mindanao to violently rebeling against the legitimacy of rule by the Philippine state. 5.4 Financing the Philippine-Moro War Start of the Civil War Amongst the initial batch of Muslims sent to Malaysia for training was Nur Misuari, a young political science professor at the University of the Philip-

144 Section pines in Manila. As a student and instructor, Misuari had organized multiple groups and participated in a wide range of protests in favor of various issues of importance to Filipino Muslims (George 1980, 198). By 1968, Misuari came to believe that these issues could only be fully resolved by the separation of Muslims from the Philippines and the creation of an independent Moro nation (Yegar 2002, 267). In the aftermath of the Jabidah massacre, Misuari was drawn into the larger Moro movement when his activities gained the attention and financial support of other Muslim leaders such as Datu Matalam, the founder of the MIM (McKenna 1998, 147). In the late 1960 s, these Muslim leaders had united as the Bangsa Moro Liberation Organization (BMLO) to funnel outside resources for the training and equipping of an armed liberation force (Che-Man 1990, 78). The BMLO was not a popular movement, but a coordinating umbrella organization comprised primarily of the existing political leadership of different Muslim organizations (Majul 1988, 87). The BMLO leaders saw the potential of Misuari s leadership, activism, and intellectual skill and selected him to participate in the first round of guerrilla training. Toward the end of 1969, Misuari joined 90 Muslims headed to Pangkor, Malaysia for training by the Malaysian Army in guerrilla warfare (Yegar 2002, 253). The training was financed by the Malaysian government and facilitated by the chief minister of Sabah, Tun Mustafah (McKenna 1998, 148). When these recruits returned to the Philippines they formed the core of a training cadre for uniformed Muslim armed militias essentially changing the character of what until then had been gangs. The most prominent of these groups were the Black Shirts, widely viewed as the military arm of the MIM and

145 Section Datu Matalam, and the Baracudas, affiliated with Muslim congressman Ali Dimaporo (George 1980, 162). These and other Muslim militias and their Christian counterparts engaged in a tit-for-tat terror campaign against members of the opposing religious group. Their attacks evolved into more organized and more lethal violence in On the Muslim side this was directly attributable to the greater resources provided by Malaysia. For example, in August 1971 the Philippine Constabulary (PC) was dispatched to the town of Buldun to quell fighting between Muslims and Christians. After initially being repulsed by the Black Shirts, the PC called for reinforcement by the Armed Forces of the Philippines (AFP) to engage the well armed and well organized enemy numbering between 750 and 1,500 (McKenna 1998, 153; George 1980, 167). Notably, a battalion of AFP soldiers and a supporting artillery unit had been deployed to the region in June 1971 precisely to support the PC in such operations against the ever more potent Moro rebels (Yegar 2002, 255). Christian militias similarly demonstrated increased intensity in their attacks in Of equal importance, this violence was linked to both local and national government officials. The Ilagas, literally the rats, was perhaps the most notorious of the Christian militias (Majul 1988, 48). The Ilagas was led by Feliciano Luces, a settler who operated under the nom de guerre of Kumander Toothpick and had cultivated a near-mythic status as being extremely brutal and fanatical in attacking Muslims (George 1980, 147). The group was employed by local political leaders to terrorize the Muslim population. When these attacks took place, the absence or complicity of

146 Section the PC, a national police force, appeared obvious to Muslims (Noble 1976, 410; Majul 1985, 51-2). The most infamous of these attacks took place in the town of Manili in June 1971 where up to 70 Muslim men, women, and children were herded into a mosque and killed (Che-Man 1990, 75). Muslim leaders issued a joint statement in July 1971 declaring that there is a strong reason to believe, of connivance with, or support by, certain sectors of the armed forces and local Christian politicans (Majul 1988, 114). The perception of complicity by the Philippine central government gained further credibility when Kumander Toothpick was flown to Manila to confer with President Marcos after another series of brutal attacks in 1971 (George 1980, 151). While the scene in the Southern Philippines at this time was one of utter anarchy, most researchers agree that the period from March through December 1971 contained a heightened level of violence in which an estimated 1,500 were killed in the fighting and another 100,000 Filipinos are forced from their homes (George 1980, ; Yegar 2002, 255-7; McKenna 1998, ; Majul 1985, 57; Che-Man 1990, 75-6). The deaths were distributed nearly equally amongst Christians and Muslims and included PC and AFP casualties. The civil war in Mindao is thus assessed to have begun in March of The intensity of the fighting as reflected in battle-related deaths, the organized opposition of the Moros with declared political intent to secede, and the active involvement of the central government in the fighting are determining factors in this assessment.

147 Section The MNLF The Manili massacre and on-going violence in Mindanao was widely publicized and gained the attention of Libyan leader Muammar Qaddafi. A personal phone call to President Marcos yielded denials of any type of repression going on in the Southern Philippines. In August 1971, the Libyan delegation to the United Nations told the Human Rights subcommittee that a possible genocide was being perpetrated by the Philippine state against Muslims (Yegar 2002, 255). Qaddafi s sympathy for the plight of the Moros as an extraterritorial constituency of Muslim brothers moved his regime to actively support the Moro rebels (George 1980, 245). Reports differ, but it appears that as early as July 1971, Libya began delivering economic, military, and materiel support to the BMLO (Majul 1988, 55). Despite being recruited into the BMLO, during their training in Malaysia seven of the initial cadre of Moro guerrillas including Misuari created a separate and, at the time, clandestine Muslim organization - the Moro National Liberation Front (MNLF) (Che-Man 1990, 77-8). The purpose in starting the MNLF was to distance themselves from what they viewed as the traditional leadership of the Moros who were complicit in the current state of affairs. There are varying accounts of the timing of the formation of the MNLF, but all agree that the MNLF was initially a covert effort until approximately mid-1971 when a meeting of guerrilla leaders in Zamboanga City convened to reject what they viewed as the collaborationist leadership of the MIM and BMLO in favor of the MNLF (Buendia 2005, 116 fn 29). The MNLF called for Moro independence from the Philippines through

148 Section armed struggle and established a headquarters in Sabah to direct the effort. Their godfather was Tun Mustapha who had originally provided arms, cash, and training to the BMLO. Mustapha was a native of Sulu and had fought in the Philippines against the Japanese during World War II. Afterward, he settled in Sabah and became a wealthy businessman with investments in timber, mining, and agriculture (Canoy 1987, 183). Angered by the policies of the Philippines and with perhaps his own agenda for greater power in Sabah, Mustapha became the primary conduit for weapons to the MNLF (Noble 1976, 413). As the guerrilla war continued throughout 1972, the Marcos administration attempted to blunt the criticisms of its handling of the Moro problem in the Philippines. Repeated invitations were made to Muslim countries to visit Mindanao and Sulu to survey conditions in the provinces. Ambassadors from Egypt, Saudi Arabia, Libya, Malaysia, Indonesia and other states participated in these visits. While they did not find evidence of a genocide, they remained convinced that the government of the Philippines was complicit in a religious war against Muslims (George 1980, 246). In September 1972, President Marcos declared martial law throughout the Philippines, citing the secessionist movement in the South as the principal reason. Despite Marcos public statements to the contrary, most see the declaration of martial law as an aggravating factor in the violence in the South and not a consequence of it (Vitug and Gloria 2000, 30). One month later, several hundred Muslim fighters attacked and occupied the university, the constabulary headquarters, the radio station, and the airport in Marawi, Mindanao (Majul 1988, 60). They successfully held off police and

149 Section AFP forces for approximately three days. Despite their success, the attacks were not coordinated by any of the existing Moro rebel groups. Instead, a lesser known set of religious leaders had managed to harness the considerable popular anger toward the regime in a tide of nearly spontaneous protest (Yegar 2002, 259). This vacuum in the leadership of the war provided an opportunity for Misuari and the MNLF to challenge the leadership of the BMLO. Misuari correctly assessed that the source of power in the civil war was the ability to control the flow of aid and armament into the South. Thus the leadership of the MNLF traveled to Libya to present the MNLF s vision of a pathway to a separate Moro homeland bangsa moro and to solicit the diversion of Libyan aid to the MNLF (Che-Man 1990, 78). The radicalism of the MNLF in pursuing not only a separate state, but also in displacing the traditional Muslim leadership, appealed to Qaddafi and earned his support. While obviously receiving harsh criticism from the BMLO leadership, who were at the same time in Jeddah asking for the support of the Organization of the Islamic Conference (OIC), this strategic move to secure financing was successful and largely cemented the position of the MNLF as the lead rebel group in the civil war MNLF as Robin Hood The MNLF quickly became the exclusive conduit of foreign aid not only from Libya, but also from Syria, Pakistan, Egypt, and the OIC. Other partisan groups soon began to join forces with the MNLF and, by the end of 1973, nearly every armed Muslim group was loosely united with the MNLF

150 Section (Noble 1991, 1098). The prime reason was his {Misuari s} status as the most effective link with foreign supply sources, without which the fighting in Mindanao could not have continued (George 1980, 231). Thus instead of acting as a bandit exploiting domestic sources of financing, the MNLF acted as a Robin Hood distributing largesse to secure support. The MNLF structured the rebel belligerents through a revitalization of a Moro identity and the goal of reclaiming a homeland independent of the Philippines (Buendia 2005, 115). The organization of the movement on paper appeared to reflect a unified political organization supported by a similarly unified military arm. In reality, the political organization of the MNLF essentially ended below the Central Committee (Che-Man 1990, 82). For instance, the National Congress convened in 1974 and not again until The Central Committee was focused almost exclusively on building, sustaining, or gaining the support of Muslim states and international organizations. This perception was confirmed at the end of 1972 when the MNLF moved its headquarters from Sabah to Tripoli a clear indication that the MNLF leadership valued the cultivation of external support more than the development of domestic sources of support (Yegar 2002, 274). Across multiple written histories of the Moro struggle by academics, journalists, members of the AFP, and members of the MNLF, there are almost no references to efforts to exploit internal sources of finances from either the people or businesses of the Southern Philippines. The first exception is Che Man s (1990) suggestion that the Moros used zakat, or Muslim alms giving, donated by locals either in cash or in kind (Che-Man 1990, 84). This is, however, counter to other research that shows how this diversion of zakat is

151 Section more characteristic of later Islamist separatist movements, such as the Moro Islamic Liberation Front (MILF) that split from the MNLF in the 1984 (Vitug and Gloria 2000, 129; McKenna 1998, 229). It is not until 1981 that international Islamic organizations recognize the legitimacy of using zakat to support the Palestine Liberation Organization (PLO) for humanitarian or charitable purposes (McKenna 1998, 229 fn 27). It is thus unlikely that a decade prior to this ruling that supporting Muslim rebels in Southeast Asia was an accepted practice. Still, Yegar (2002) also suggests that the MNLF used zakat as a means of financing their armed struggle during later stages of the war. Again, though, the contributions were predominantly made in kind, as donations of rice instead of money (Yegar 2002, 273). While a system of taxation cannot wholly be ruled out, based on this information, it would be exceedingly generous to refer to zakat as even a secondary source of financing for the MNLF. Another reference to a possible system of taxation is made in one of the more widely cited books on the war, George s (1980) Revolt in Mindanao. In it, he describes how Muslim refugees who sought safe haven in Malaysia were, until 1974, paid a daily stipend. The stipend was paid by the provincial Malaysian government. With as many as 40,000 Filipino refugees, this represented a potentially substantial cash subsidy and it is likely that some of these funds went onward to MNLF rebels in the Philippines. George, however, describes as doubtful the likelihood that these refugees made a significant impact on the insurrection (George 1980, 236). Even if it were to be supposed that the MNLF placed a levy on the refugees stipend, it would not be inconsistent to consider this money as foreign aid as if it had

152 Section flowed directly from the Malaysian government to the rebels instead of via refugees. This conclusion also aligns with the remainder of the chapter that George devotes to Economics and Revolution in which he makes no other mention of rebels raising revenue domestically. The MNLF instead relied almost exclusively on external support to wage the civil war. This support took a variety of forms from the military aid already mentioned to food, medical aid, and even money to build mosques and Islamic centers (Yegar 2002, 233). All of this aid flowed through the MNLF for dispersal to provincial revolutionary committees. These committees, however, did not so much have a political purpose, but instead functioned to assist in recruiting, training, and equipping Moro fighters for their respective regional commands of the Bangsa Moro Army (BMA), the military arm of the MNLF. Below the level of the headquarters cadre, the MNLF was dependent on local leaders and their ties to armed groups in their areas of concern (Che-Man 1990, 83). Although dependent on the MNLF for financial and material aid, these local leaders were often unaware of the larger political agenda of the movement (George 1980, 229). Nevertheless, the substantial flow of money and materiel proved sufficient to at least unite these disparate groups in armed opposition to the central government Civil War In a display of their military capability, the MNLF launched a largely conventional military campaign in February 1973 to take over the province of Cotabato and to capture Cotabato City. Cotabato refers to the south-

153 Section Figure 5.3: Map of Cotabato from McKenna (1998) west region of Mindanao even though this area is divided into four separate provinces (see figure 5.3). The region has significant symbolism for the Moros as being comprised of two of the principal sultanates from pre-colonial times (McKenna 1998, 25). The lowlands of the Cotabato plains are a source of rice, corn, and coconut production while Cotabato City is the commercial center of the region (McKenna 1998, 33). The AFP general who opposed the rebels in this campaign, General Fortunato Abat, described the rebel campaign plan as the first step in conventionally carving out the Bangsa Moro Republic (Abat 1999, xxi). Rebel units of up to 1,000 armed guerrillas attacked government positions and displayed striking power and vastness of resources unperceived before (George 1980, 213). In less than two weeks, approximately three

154 Section fourths of the province was under MNLF control and influence (Abat 1999, 10). The title of General Abat s account of the conflict is telling: The Day We Almost Lost Mindanao. A similar battle took place the following year in which some 2,000 rebels seized control of the island of Jolo and its five towns (Canoy 1987, 197). In these and other battles the rebels demonstrated their capacity to control the countryside, seize towns, and infiltrate and bombard cities (Noble 1976, 418). If this were the case, then why did the rebels not use this capability to become stationary bandits and develop a permanent means of financing the civil war by exploiting domestic resources? The first answer could be that the the rebels were outgunned by the military capacity of the Philippine state. Thus while the rebels demonstrated an ability to seize control of areas, they were unable to hold onto those areas for long enough to realize economic exploitation. In 1973, the MNLF fielded approximately 13,400 armed rebels compared to the AFP deployment of 73,500 troops to the Southern Philippines. By their overwhelming strength, the AFP were able to regain control of Cotabato and declare the central plains operation complete within three months (Abat 1999, 89). In Jolo, the Philippines military displayed the depth of its military arsenal. The AFP employed naval gunfire support, aerial bombardment by jet aircraft, helicopter gunships, and the dispatch of several battalions of soldiers to defeat the rebels in less than two weeks (George 1980, 217). By 1975, 75% of the total AFP strength of 330,000 soldiers would be deployed to the Southern Philippines (Yegar 2002, 276). Thus despite the impressive conventional capability of the MNLF, the rebels were physically unable to retain control of an area to exploit it for financing the war.

155 Section Yet even at the outset of the war, when the strength of the Philippine forces was a fraction of that of the Muslim militias, the rebels did not appear to make a concerted effort to control territory. Prior to June 1971, the main national government presence for law enforcement or suppression of rebellious activities was a detachment of the Philippine Constabulary (PC) of only 400 lightly armed constables. By this time, several thousand armed rebels were active. Thus, prior to the outbreak of the war - and even during the initial calendar year of the war when only 600 AFP soldiers are sent as reinforcements - armed Moro groups were well positioned to establish control of villages and barrios. Even given the opposition of civilian Christian militias, the BMLO could have potentially coordinated the establishment of a safe area in Mindanao to begin developing a means of domestically financing the war. It was not the case that the state already had in place an overwhelming capacity in Mindanao such that it could deny Moro groups from extracting financial resources. Thus, at least anecdotally, it is noted that state capacity is not an omitted variable at the outset of this war. The door was open for rebel exploitation of domestic sources of financing, but the rebels did not signal an intent to proceed further in this regard. A second and more convincing answer could be that the amount of support being provided by international donors exceeded expectations of potential benefit from any conceivable system of taxation on businesses, commercial transactions, or families. Noble (1976) suggests that one could infer from the conventional tactics and strategy of the rebels in confronting the Marcos regime in a war of attrition that the MNLF firmly believed in its ability to sustain the war so long as it could maintain its external connec-

156 Section tions (Noble 1976, 420). Estimates of contributions to the MNLF suggest that there were multiple sizable payments throughout the war in addition to military equipment, ammunition, and training. Libya alone is reported to have provided 4.4 million Philippine pesos in 1971 (Majul 1988, 55), $3.5 million in 1972 (Yegar 2002, 273), and 2 million in 1974 (George 1980, 232). 4 The OIC provided at least another 1 million in humanitarian aid, again all directed through the MNLF (Vitug and Gloria 2000, 60). With a close personal relationship with Qaddafi; the public backing of the OIC; and hundreds of recruits being trained in Malaysia, Syria, Pakistan, and at PLO camps in Libya, the rebels perceived a permanence to the financial arrangements put in place by Misuari and thus chose not to pursue domestic means of financing. 5.5 End of the War Unsurprisingly, the civil war ended when the external supporters of the MNLF dropped their support for Moro independence and pressured the rebels to settle for an agreement granting them limited autonomy within the Philippine state. The reasons for the shift in support for the Moros had more to do with state-to-state politics than with conditions on the ground in Mindanao. President Marcos recognized the delicacy of a situation in which his government was fighting an Arab-backed insurgency in the Southern Philippines whilst the Philippine economy was overwhelmingly dependent on 4 The diversity of currency denominations reflects nothing more than the preferences of the various authors in presenting the data. The intent here is to demonstrate that large sums of money were flowing between Libya and the MNLF.

157 Section imports of Arab oil. Marcos also worried about the cost of keeping 250,000 soldiers in the field and about the further vulnerability of the Philippine economy to changes in the more than $1 billion received annually from the estimated 100,000 Filipino workers in Arab states. Marcos engaged in a wide variety of activities to curry favor with Arab leaders including a pledge of Philippine support of a United Nations resolution calling for the withdrawal of Israeli forces from areas occupied by Israel in wars with its Arab neighbors (Abat 1999, 171):...the Philippine government experimented with a variety of tactics. It made and shuffled appointments of Muslims; brought rebel leaders together for peace talks; created new offices and agencies; announced additional programs and appropriations; and spread rumors about factionalism, corruption, and foreign influence among the MNLF leadership. By mid-1975, before a meeting of the Islamic Foreign Ministers Conference in Jidda, Marcos was proposing to restructure the southern Philippines into regions with virtual autonomy (Noble 1981, 1099). One of the most important results from these efforts was a statement from the OIC that there would be no punitive actions taken in regard to the flow of oil to the Philippines. This reduced the diplomatic pressure on Marcos from the Arab states (McKenna 1998, 168). Ultimately, the Conference of Muslim Foreign Ministers voted to support the MNLF in negotiations with the government of the Philippines in working out conditions of autonomy but not independence (Yegar 2002, 286-7).

158 Section Additionally, relations between Malaysia and the Philippines improved. The central government of Malaysia dropped its support for Tun Mustapha, the rebel financier and chief minister of Sabah. Mustapha was defeated in the 1975 elections and replaced with a minister who ended the use of Sabah as a waypoint for the flow of arms to Mindanao (Noble 1976, 413). The Tripoli agreement that would end the war began with negotiations between Imelda Marcos and Qaddafi in Philippine representatives, the OIC, and Misuari agreed to a plan that called for a cease fire to be in force as of January After this, the southern provinces would hold a referendum on a range of terms for autonomy. The final status of what would become the Autonomous Republic of Muslim Mindanao would not be resolved until well after Marcos was displaced as leader of the Philippines. Despite the many setbacks in determining the final status of Muslims in the Philippines, the enthusiasm of Arab leaders for the MNLF had largely dissipated (George 1980, 256). In a mid-1977 meeting with Islamic foreign ministers in Tripoli, Muammar Qaddafi made it obvious that his interests had shifted away from support for the MNLF (Noble 1981, 1102). The cease fire held and the first Philippine-Moro civil war effectively ended. For the next ten years, the problem in Mindanao receded into an almost dead issue (Abat 1999, 195). Subsequent to the Tripoli agreement, there was sporadic fighting with levels of violence well below civil war levels (Canoy 1987, 235).

159 Section Alternative Explanations An alternative explanation for the termination of the Moro civil war is that the Tripoli agreement provided resolution to grievances through the creation of a framework for autonomy in the southern Philippines. The principal grievances of the MNLF, as expressed by Misuari in a proposal to the Conference of Muslim Foreign Ministers in 1975, were that the extension of the Philippine Republic to include the Bangsa Moro homeland in Mindanao, Basilan, Sulu, and Palawan was illegal and could only be remedied by selfrule. An agreement would minimally require the withdrawal of government forces, return of land expropriated by Christian settlers and the government, and the incorporation of Muslim law and customs into governing institutions (Yegar 2002, 298). The main points of the original agreement appear to substantially address these grievances. The 1976 Tripoli agreement established an autonomous area that included 13 of 21 districts sought by the MNLF. The area would remain within the sovereign state of the Philippines. The central government would be responsible for foreign policy and national defense, and retain control over mines and natural resources within a revenue sharing agreement. The Muslims could establish a court system based on shari a law, establish schools and universities, create their own administrative system, administer their own economic and financial system, elect a special legislature and executive, and participate in the establishment of a special security force for the region (Yegar 2002, 304-5). While this agreement did establish a means to rectify the political grievances

160 Section of the Moros, it did much less in terms of addressing the economic grievances of people in the southern provinces. While the agreement held out the promise of government assistance in resolving issues of land distribution, tenant-farmer agreements, and poverty, there were few changes in economic indicators at the time or in subsequent years. Even twenty-five years later, Muslim leaders and scholars continue to point to the socioeconomic inequality that persists in the region. In particular, they highlight the continued benefits accruing to non-muslim owners of large agricultural businesses and the profiteering of outside corporations that exploit natural resources in the region (Tan 2000). Also, many former rebels of the MNLF returned to towns and cities as their allowances from the MNLF ended. The influx of unemployed and armed young men had the predictable consequences of increased crime and violence that ultimately reduced economic and physical security in the region (McKenna 1998, 180). A second counter to the grievance explanation is that, subsequent to the Tripoli Agreement, President Marcos greatly manipulated the terms of its implementation. The governing institutions that resulted were cosmetic creations with no real legislative authority and no independent operating budget and thus served only as an additional regional bureaucracy that had little impact on the lives of Muslims (McKenna 1998, 168). In his first effort to undermine the agreement, Marcos announced a popular referendum in which the districts could vote to become part of the autonomous area. The demographic shifts in the region that made Muslims an overall minority had also made Christians a majority in eight of the thirteen districts of the proposed autonomous area. Outrage over this proposal led to further

161 Section negotiations over implementation until a provisional government that was majority Christian was appointed by Marcos (Yegar 2002, 307). It was not until 1987, when President Aquino pushed for its inclusion in the new Philippine Constitution, that the Autonomous Region of Muslim Mindanao was implemented in the spirit of the Tripoli Agreement. Even after this, the election of a regional legislative council, first promised in 1976, did not occur until A grievance-based explanation for the end of the war is not a plausible alternative. This approach postulates that the anticipation of improvements in political and economic conditions resultant from the Tripoli Agreement reduced violence in Mindanao below civil war levels. The actual conditions in the region, however, do not improve economically or politically. This again highlights the main shortcoming of a grievance-based approach. It is exceedingly difficult to accurately measure individual or group expectations. Even if such measures were reliable and accessible, the researcher would be confronted with having to weigh the importance of individual perceptions relative to their actual conditions. 5.7 Conclusion Building on a long-standing political and religious cleavage, the Moro nationalist movement organized for violent conflict with the state in the late 1960s with support from ethnic and religious sympathizers in Malaysia. The conflict reached civil war intensity in March 1971 and continued until the cease fire of the Tripoli Agreement went into effect in January The

162 Section MNLF was almost entirely dependent on the financial, materiel, and training support of sympathetic states. The MNLF acted as Robin Hood, instead of a stationary bandit, and organized to sustain external support instead of developing a domestic means of financing the war. The dependence of the Moro war effort on this source of aid rendered the political objectives of the MNLF subject to the desires of their benefactors. When these states dropped their support for an independent Moro state, the MNLF was forced to settle for an agreement on Moro autonomy within the Philippine state. The MNLF lacked an alternative, domestic source of extracting resources and was thus unable to continue the war.

163 153 Chapter 6 Civil War with CPP-NPA 6.1 Introduction The defeat of the Huk army in 1954 and the Anti-Subversion laws that banned the communist party forced the Philippines Communist Party (PKP) into hiding. It was not until a new wave of communist adherents became active that new life was breathed into the communist movement in the Philippines. The new communist leaders of the Communist Party of the Philippines (CPP) rejected the Soviet-style outlook of the older generation and embraced a Maoist strategy for revolution that emphasized the building of a mass movement for a protracted struggle. From the birth of the CPP in late 1968 until the start of the civil war in approximately January 1981, the CPP and its military arm, the New People s Army (NPA), worked to increase their base of popular support, increase the size and capabilities of their army, and build a means of domestically financing the civil war. The geographical dispersion of the CPP-NPA and their ability to financially sustain combat operations through an elaborate system of taxation enabled the

164 Section communists to continue the war for more than eleven years. 6.2 Communism in the Philippines There are two currents in the history of conflict in the Philippines that combine to set the scene for this second episode of communist-inspired civil war in the Philippines. The first is the memory of, and lessons learned from, the failed civil war of the Philippine Communist Party (PKP) and the Huks that ended less than two decades prior. The second is the student based activism of the 1960 s that provided a new generation of leaders to persistent Philippine political movements. This new leadership would turn the violent conflict of the Moro independence movement into the Philippine-Moro civil war. The story of the second civil war with the communists begins with the new cadre of activist leaders becoming politically engaged in the 1960 s and attempting to correct what they believed to be the shortcomings of the Huk rebellion Jose Sison and the CPP Jose Sison is the leader at the center of the revival of the communist movement in the Philippines. His early biography is remarkable for its parallels to that of Moro National Liberation Front (MNLF) leader, Nur Misuari, who began his political activities at university and then was recruited by the existing old guard leadership of the movement. From 1956 to 1961, Sison was a student at Ateneo de Manila University and the University of the Philippines (UP). After graduation as an English major he continued at

165 Section UP as a teaching assistant. Sison formed the Students Cultural Association of the University of the Philippines (SCAUP). SCAUP channeled student radicalism to protest the government s investigation into anti-filipino activities in the late 1950 s a primarily anti-communist inquiry similar to the United States House Un-American Activities Committee (Kessler 1989, 37). Sison s activism likely played a role in his being let go by the English department and in 1961 he traveled briefly to Indonesia where he worked with their communist party (PKI). When he returned to the Philippines in 1962, the PKP sought him out to take on a party leadership position (Abinales 2001, 17). Sison was recruited to become a member of the PKP executive committee. In keeping with the 1957 single file policy enacted to avoid arrest under the Philippines Anti-Subversion Law, the PKP were conservative and extremely cautious in their activities for fear of further government suppression (Nemenzo 1984, 74). The PKP executive committee held weekly, or sometimes monthly meetings chaired by Jesus Lava, the third Lava brother to be General Secretary of the PKP. In these meetings the main enterprise was to attempt to reactivate the first generation of communist members still in hiding instead of recruiting new ones (Nemenzo 1984, 74; Abinales 2001, 19). It was not until 1964 that Sison was able to begin the role for which he had been brought into the PKP the reinvigoration of the youth wing of the party. While working to create an activist Patriotic Youth (KM) organization, Sison was scathing in his assessment of the weekend warriors of the executive committee they were such mediocre men (Abinales 2001, 19). It was during this time that Sison further developed a criticism that

166 Section would become the central theme of his later split from the PKP. Sison believed that the major failing of the communist movement in the Philippines was that the PKP leadership had no connection whatsoever with the mass movement (Abinales 2001, 19). The tensions between the tightly controlled conservatism of the PKP and the Maoist radicalism of the younger generation led to Sison s split from the PKP. Amongst the KM, Sison had cultivated a group of followers that supported his rectification campaign for the communist party. In December 1968, Sison gathered a dozen trusted comrades to launch the Communist Party of the Philippines (CPP) (Reid 2000, 14). Their immediate task was to build the party and Sison believed an integral part of this strategy was an army to protect the base and wage war against the state (Jones 1989, 27). Sison was soon introduced (possibly by Benigno Aquino) to the leader of a small group of armed insurgents, Bernanbe Dante Buscayano. Dante had been trained by the remnants of the PKP-Huk movement in the 1960 s (Davis 1989, 73). Despite his rapid rise through the ranks, he broke away in 1968 over ideological and ethical differences with what had become largely a criminal gang of former Huks operating around US bases in the Philippines (Nemenzo 1984, 80). Dante accepted the mantle of commander of the New People s Army (NPA) that at the time was nothing more than his band of 50 former Huk fighters and just 35 weapons (Jones 1989, 32) Rectification The CPP Central Committee, including the leadership of the NPA, met again in May 1969 (Jones 1989, 34). With a new party and a nascent

167 Section military arm, the central committee adopted a general report, prepared by Sison, to document the lessons learned from the failures of the PKP. In so doing, the CPP both presented a roadmap for the building of the party base and also codified the grievances of the younger activists with the PKP. This process of rectification is important to an understanding of how, over the next 13 years, the CPP grew its base and army into a position from which it could finance and wage a decade-long civil war against the government of the Philippines. The report outlined a Maoist agenda for contesting political control of the Philippines by establishing a mass base of support, engaging in guerrilla warfare, and launching a protracted struggle (Abinales 2001, 25). Initially, and in keeping with the guiding Maoist teachings, the CPP hoped to create a Yenan-style safe haven and base. Yenan, or Yan an, was revered as the impregnable Shensi province base Mao built following the Long March and in which Mao remolded the communist movement (Jones 1989, 38). The CPP and NPA thus moved out of Manila to establish such a base in the foothills of the Sierra Madres mountains in the province of Isabela (see map, figure 6.1). While this was a change from the Manila-based strategy of the PKP, the entirety of the CPP-NPA remained on the island of Luzon. It would have been difficult to more greatly disperse the fledgling party and army at this early stage. The concentration in Luzon was in part responsible for a series of defeats and setbacks at the hands of the Philippine Army from 1971 to These early defeats destroyed the Isabela base and led to a reevaluation of how a Maoist struggle could be enacted in the environs of the Philippines. In 1974, the CPP central committee outlined the unique traits

168 Section Figure 6.1: Map showing Isabela province from wikipilipinas.org