The State’s Use of Force by Meredith Lauren Blank
A dissertation submitted in partial ful llment of the requirements for the degree of Doctor of Philosophy (Political Science) in the University of Michigan 2017
Doctoral Committee:
Professor Christian Davenport, Chair Associate Professor John David Ciorciari Professor James D. Morrow Assistant Professor Yuri Zhukov Meredith Lauren Blank
ORCID iD: 0000-0002-9340-2580
© Meredith Lauren Blank 2017 Dedication
Finally, the time has come Everything up ’til yesterday was a prologue.1
To William Patterson, who always believed in me to run farther, climb higher and complete this doctorate.
1Radwimps. Kimi No Na wa, 2016.
ii Acknowledgements
Raising a family is di cult enough. Raising a Ph.D. can be just as involved. This dissertation would not have been possible without the incredible support from my advisors, parents, mentors, friends, colleagues and a number of academic organizations. I am deeply indebted to my committee who have guided me over the course of many conversations, classes and zip codes. I must start by thanking James Morrow, professor of WorldPolitics at the University of Michigan. Through all of this, from my rst course on international relations in undergraduate school to the completion of my dissertation, he taught and mentored me. He showed me how to be bold in intellectual pursuits and that the best work came from asking the hardest questions. I owe tremendous gratitude to Christian Davenport, who has been instrumental in shaping my research and way of thinking. Through his guidance and wisdom, he taught me to push the boundaries of our discipline. Yuri Zhukov and John Ciorciari provided expertise, ceaseless encouragement and showed great patience throughout the multiple drafts of this project, from its earliest incarnation to nal submission. Many mentors also provided guidance on this project during my time at University of Michigan. Allan Stam broadened my perspective of international aairs from merely considering current events to pursuing a wider understanding of history. Scott Tyson deserves special thanks for the countless hours he spent reading chapters, rereading chapters and never giving up on an idea. Rob Sayre provided me with as much advice as anyone; he gave valuable expertise on security aairs, military history and martinis. I’m grateful to Leanne Powner for showing me how to organize a seemingly chaotic writing process. I owe thanks to Mark Dincecco, Christopher Fariss, Mai Hassan, Ragnhild Nordås and Philip Potter for their comments and insight. I also wish to thank Robert Mickey, who has mentored and supported me since I was an undergraduate and throughout my doctoral training; it was Rob Mickey who informed me I had
iii been accepted to Michigan’s graduate program, and for that I will always be thankful. I owe much thanks to my graduate student colleagues at University of Michigan, who I am lucky enough to also consider close friends. Isaac Jenkins, Thomas O’Mealia, Anita Ravishankar, Laura Seago, Jessica Sun, Alexander Von Hagen-Jamar and Alton Worthington provided guidance, encouragement and−as needed−housing. I’ve had the pleasure of calling both the University of Michigan and the George Washington University home. I want to extend my thanks to Charles Glaser and Stephen Biddle at the George Washington University Elliot School of International Aairs for providing me with the expertise, direction and resources to succeed. Elizabeth Saunders, Caitlin Talmadge, Alexander Downes, Yonatan Lupu and participants of the Research-in-Progress workshop at the Institute for Security and Con ict Studies provided indispensable advice on multiple versions of this project. Ryan Baker deserves special thanks for his insights on military aairs; this dissertation is stronger because of the rich series of dialogues that occurred over many happy hours and half-priced beers. My data collection would not have been possible without institutional support from Rackham Graduate School at the University of Michigan, the Elliott School of International Aairs at the George Washington University, and the William A. and Shirley P. Yolles Graduate Fellowship for Research on Humanitarianism. Friends and family, too, contributed to this work, in ways they may not even realize. Ben VanWagoner and Sarah Waitz provided valuable editorial assistance with even the roughest drafts. Bill, Sue and Catherine Patterson along with Elizabeth Bacon, Marlene Tyner-Valencourt and Dorian Waitz gave unwavering support throughout the process. I would not have been able to stay positive throughout this journey without Remi. Being a dog, he showed me how to stay calm under pressure−and enjoy naps. I would not have made it through without the cooking of Ann Sayre, the wisdom of Karen Shen, the humor of Clare Tyson, or weekly climbs with Ethan and Wendy Zorick. I am indebted to Tedfor creating a Gibson original on my dissertation’s behalf−three parts rye, one part applejack, one part Campari. I, of course, have to give a special acknowledgment to my brother Jonathan Blank along with my parents Ira and Karen Blank. They instilled in me early the passion for arguing about politics.
iv Everyone listed here helped me take this from a discussion at the dinner table to a dissertation, and is part of my family. Thanks for raising me. See you at family reunions.
v Contents
Dedication ii
Acknowledgements iii
List of Figures viii
List of Tables ix
Abstract xi
Chapter 1 Introduction to the state’s use of force1 1.1 Question and motivation...... 1 1.2 Conceptualization of security and scope conditions...... 4 1.3 Internal security forces: Tradeos in cost and control...... 7 1.4 Plan of the dissertation ...... 12
Chapter 2 Internal security forces: Conceptual re nements and a new dataset 17 2.1 Introduction...... 17 2.2 Current data limitations...... 19 2.3 Data re nements and coding rules...... 21 2.4 Sources for data re nements...... 25 2.5 Potential measurement error ...... 31 2.6 Comparison to other measures ...... 34 2.7 Conclusion ...... 38 2.8 Dataset appendix ...... 40
Chapter 3 Construction of coercion: Violence, the international system, and internal security 49 3.1 Introduction...... 49 3.2 Models of rivalry and state formation...... 52 3.3 Theory: Violence and the international system ...... 59 3.4 Hypotheses and empirical strategy...... 63 3.5 Empirical ndings...... 67 3.6 Conclusion ...... 79
Chapter 4 Dancing in the dark in civil con ict: The state’s information environment and internal security 88 4.1 Introduction...... 88 4.2 The information environment in civil con ict...... 91
vi 4.3 Internal security as a response to information challenges ...... 95 4.4 Hypotheses on information challenges ...... 104 4.5 Empirical strategy and data ...... 107 4.6 Results...... 112 4.7 Conclusion ...... 123
Chapter 5 Bringing the police back in: Internal state security and repression 133 5.1 Introduction...... 133 5.2 Conventional wisdom on state repression...... 135 5.3 Theory: Centralization and repression in limited democracies ...... 139 5.4 Hypotheses ...... 145 5.5 Research Design...... 148 5.6 Results...... 157 5.7 Conclusion ...... 171 5.8 Appendix ...... 174
Chapter 6 Conclusion: Democratizing domestic security 182 6.1 Summary ...... 182 6.2 Contributions...... 186 6.3 Directions for future research: Democratizing security ...... 188
Works Cited 190
vii List of Figures
2.1 Militias and gendarmerie organizations worldwide ...... 30 2.2 Countries with militaries, militias and gendarmeries...... 31 2.3 Country years with militias and gendarmeries by regime type...... 32 2.4 Countries with internal security forces...... 35 2.5 Countries with regulated militias ...... 37 2.6 Countries with unregulated militias...... 37 3.1 Security organizations in new states, according to non-violent and violent independence (1950-2010)...... 69 3.2 Security organizations in new states, according to geographic region (1950-2010) ...... 70 3.3 Predicted probabilities, security organizations in newly formed countries (1950-2010) . . . . 74 3.4 Kaplan-Meier survival estimates: Gendarmerie creation...... 77 3.5 Kaplan-Meier survival estimates: Militia creation...... 77 3.6 Kaplan-Meier survival estimates: Military creation ...... 78 3.7 Kaplan-Meier survival estimates: Former power with gendarmeries, militias...... 85 3.8 Gendarmerie creation...... 85 3.9 Militia creation ...... 85 3.10 Average marginal eects of violent independence on security onset...... 87 4.1 Spectrum of internal security forces...... 96 4.2 Irregular and regular militias in civil con ict...... 114 4.3 Gendarmeries in civil con ict ...... 115 5.1 Average marginal eects of centrally controlled internal security on the probability of state repression...... 161 5.2 Average marginal eects of centrally controlled internal security over executive constraints . 164 5.3 Average marginal eects of centrally controlled internal security over judicial independence 164 5.4 Average marginal eects of gendarmerie personnel over excluded ethnicities ...... 168 5.5 Average marginal eects of militia personnel over excluded ethnicities ...... 168
viii List of Tables
2.1 Descriptive statistics of internal security forces...... 29 2.2 Countries uniquely identi ed as having militias per dataset...... 36 2.3 Militias and Gendarmeries by Country, 1970-2011...... 40 2.4 Cross tabulation of country years with militias, 1981-2007...... 46 2.5 Cross tabulation of country years with state-associated militias, 1981-2007...... 46 3.1 Security organizations in violently formed countries, 1950 - 2010 ...... 68 3.2 Multivariate probit: internal security organizations in newly formed countries ...... 73 3.3 Univariate probit: security organizations in newly formed countries ...... 74 3.4 Cox proportional hazard, security organization onset (1970 - 2010)...... 76 3.5 Non-violent independence, 1950-2010...... 83 3.6 Non-violent independence, 1950-2010 (continued) ...... 84 3.7 State regular militias in newly formed countries...... 85 3.8 Gendarmerie in newly formed countries...... 85 3.9 Logistic regression, security organization onset ...... 86 4.1 Emergence of irregular local militias in civil con ict, 1970-2010 ...... 116 4.2 Correlation matrix of security personnel in civil con ict, 1970-2010 ...... 117 4.3 Seemingly unrelated regression, security personnel in civil con ict (1) ...... 118 4.4 Seemingly unrelated regression, security personnel in civil con ict (2) ...... 119 4.5 Personnel in internal security forces, 1970 - 2010...... 126 4.6 Descriptive statistics for civil con ict, 1989-2009...... 126 4.7 Correlation matrix: External support and security personnel in civil con ict ...... 126 4.8 Multivariate probit coe cient estimates, security personnel in civil con ict ...... 127 4.9 OLS coe cient estimates (FE); personnel in security forces (1989-2009) ...... 128 4.10 Population measure, coe cient estimates (FE) ...... 129 4.11 Seemingly unrelated regression, personnel in security forces (absent state failure) ...... 130 4.12 Seemingly unrelated regression, personnel in security forces (state failure) ...... 131 4.13 Seemingly unrelated regression, personnel in security forces and the Cold War ...... 132 5.1 Central internal security and political terror (Amnesty)...... 158 5.2 Central internal security and political terror (Amnesty)...... 159 5.3 Selection of Centrally controlled internal security and state repression (latent human rights) 166 5.4 Centrally controlled internal security, the Cold War and state repression (PTS[A]) . . . . . 170 5.5 Descriptive Statistics...... 175 5.6 Ratio of Centralized Security and Latent Human Rights ...... 176 5.7 Security Personnel (ln), Democracy and Latent Human Rights ...... 177 5.8 Security Organizations (Binary) and Latent Human Rights ...... 178 5.9 Security Personnel (ln) and Political Terror Scale (Amnesty) ...... 179 5.10 Security Organizations (Binary) and the Political Terror Scale (Amnesty) ...... 180
ix 5.11 Ratio of Centralized Security and Political Terror Scale (State) ...... 181
x Abstract
How do governments organize internal power? What are the consequences of these actions for human rights? I explore the creation, evolution and repercussions of the state’s organization of internal security, centering on the national government’s competing needs to maximize control over security and minimize costs. I argue that legacies of violence in uence the state’s organization of internal armed forces and the subsequent eects of those organizations. Counterinsurgency campaigns and anti-colonial wars lead to the creation and expansion of local security, allowing governments with limited capacity and limited local area knowledge to engage in irregular warfare and population-centric operations. However, governments using local forces during domestic operations may be unable to demobilize them; these forces can remain long after wars end. The short-term strategic bene ts of local security forces undermine prospects for an enduring peace. In addition, I show that limited democratic constraints on national leaders will unintentionally encourage them to adopt centralized but publicly unaccountable security. This organization of force provides national leaders with the capability and command authority to engage in widespread repression against their own population. I test these arguments using an original global dataset on internal security forces that covers 198 countries from 1970 through 2011. The organization of internal security ultimately emerges from factors historical and immediate, domestic and international, and has lasting consequences on citizens within the state.
xi Chapter 1
Introduction to the state’s use of force
1.1 question and motivation
After police commit violence against civilians, we question whether their actions were excessive. When national guards and national police impede protests, we ask whether protests were unduly suppressed. Following a coup, we question the intent of the military o cers who carried it out. However, we rarely ask what role the national government had in creating, expanding and empowering these forces. Central governments can and will project force through a variety of organizations.1 National governments relied heavily on militias and hired mercenaries to supplement feudal troops prior to the twentieth century (Bradbury 2004). Governments employ a variety of domestic security forces today, with domestic security arrangements ranging from extraordinarily localized to highly centralized under a complex national bureaucracy. For example, in counterinsurgency campaigns against the Kurdish Worker’s Party (PKK), the government of Turkeydeployed poorly trained Kurdish village guards alongside professional, heavily armed gendarmerie (Felbab-Brown 2016). Similarly, the government of Iraq deployed Sunni and Shiite militias with U.S.-trained Special Operation Forces in campaigns against al-Qaida in Iraq and the Islamic State (Biddle, Friedman and Shapiro 2012; Dodge 2013). Yet this runs counter to conventional international relations theories, which would not predict the central government to use local forces to 1By central government, I mean the central or national government of a state, including those that devolve power to regional or local authorities. By projection of force, I refer to the the ability of a central government to eectively deploy and sustain elements of its coercive forces (regular and irregular armed forces). This de nition corresponds with the U.S. Department of Defense concept of power projection, which is “the ability of a nation to apply all or some of its elements of national power−political, economic, informational, or military−to rapidly and eectively deploy and sustain forces in and from multiple dispersed locations to respond to crises, to contribute to deterrence, and to enhance regional stability.” The Dictionary of Military Terms. US Department of Defense. 2009.
1 implement national security policies, as those forces could undermine the state’s control of violence. Even developed democratic countries display a signi cant amount of variation in the arrangement of domestic security, which has implications for upholding human rights and combating extremism. In the United States, local police agencies remain decentralized but have increasing authority over national security threats like terrorism (Marenin and Das 2000). The United States has between 13,500 to 19,000 state and local police agencies, with many agencies engaging in counterterrorism activities (Waxman 2009). In contrast, other developed democracies like France and the Netherlands have a far more centralized approach to domestic security and counterterrorism, with internal security and national police forces tightly administered under national ministries (Lutterbeck 2004). Both local and central security forces have their shortcomings when it comes to upholding human rights as well. Both Afghan local police and Afghan national police committed predatory violence against civilians and engaged in illegal drug tra cking (Goodhand and Hakimi 2014a; Wilder 2007). Abuse against civilians in Turkey has occurred at the hands of the Kurdish village guards as well as the state’s gendarmerie (Asfura-Heim 2014). In France, national police use excessive force against detainees and migrants, with reports of security agents having “beat, kicked, and used pepper spray on migrants and asylum seekers.”2 Across countries and organizations, security forces routinely engage in unauthorized use of violence against civilians they are pledged to protect. Yet it is unknown how these internal forces dier in their motivation for abuse and type of abuse. Given the variation in security organizations and their behavior, central governments confront a strategic and fundamental choice in how to organize internal security. They can focus more on building centrally controlled forces or on supporting local organizations.3 Todate, the preponderance of academic models suggest that states make this decision based on what type of government they have or what region of the world they are in. This dissertation shows that other factors are more in uential in this choice, including legacies of violence as well as domestic and international threats. I explore the causes and consequences of the state 2Country Reports on Human Rights Practices for 2016. “France.” Bureau of Democracy, Human Rights and Labor. https: //www.state.gov/j/drl/rls/hrrpt/humanrightsreport. 3As I explore in this dissertation, states can simultaneously develop both centrally controlled and local organizations, but they are unable to spend limitless resources on both.
2 organization of internal security, centering on the central government’s need to maximize control over security while minimizing costs. I rst examine the variety of ways that governments organize security, detailing the advantages and disadvantages of each type of force. Using new cross-national time-series data on 198 countries from 1970 to 2011, I nd that modern states strategically organize internal security to confront rivals under dierent political and nancial constraints. No single form of security is used exclusively by one type of government or in one geographic region. A central government’s organization of security is part of its strategic response to threats while maximizing control of armed forces and minimizing their nancial or political costs. Along with domestic and international conditions, contemporary and historical security challenges in uence the national government’s organization of domestic security. During anti-colonial wars and counterinsurgency campaigns, national governments (and emerging ones) have limited resources but must compete against internal coercive rivals. Operating under severe constraints, these governments turn to local security to engage in targeted irregular warfare tactics and population-centric information operations.4 In contrast, centrally controlled and regular armed forces provide national leaders (who have the resources) with the authority and capability to engage in widespread domestic operations.5 While more costly to create and maintain, these forces can engage in more lethal warfare operations and direct more extensive repression against the public, a useful strategy for national leaders facing challenges at the polls in limited democracies. Takentogether, my dissertation shows that governments rely on an intricate set of local and centrally controlled actors for internal security. These security arrangements have both intentional and unintentional negative consequences for the civilians they are assumed to protect. 4Population-centric operations refer to tactics in warfare−speci cally in counterinsurgencies−where the population and not the enemy is the focus. This includes tactics where the population is considered the most important source for intelligence and necessary for denial of enemy capabilities. Gentile, Gian P. A Strategy of Tactics: Population-centric COIN and the Army. Army War College. 2009. 5I use the term “regular” to refer to security forces who are o cially armed forces of the state with standardized equipment, training and a set organization. “Irregular” forces do not o cially represent the state and lack standardization or a set organization.
3 1.2 conceptualization of security and scope conditions
I propose that states organize internal power through domestic security forces and that these forces vary according to their degree of (1) localization/centralization and (2) regulation by the government. The scope of this project includes domestic security forces, meaning armed agents of the state that engage in internal security operations. These forces may have jurisdiction over civilian policing and violent crime, but their primary responsibility is maintaining internal security, such as preventing and combating domestic terrorism, insurgencies and civil unrest. Internal security forces are (1) managed and organized within the territorial jurisdiction of the state, (2) given explicit authority by the state to engage in internal security operations, (3) not part of the military, and (4) managed through national, regional or local governments. I additionally distinguish between regular forces that have standardized equipment, training, and a set organization that operates o cially as part of the state (such as state militias) and irregular forces that lack such standardization and often operate in an uno cial capacity (such as certain village defense forces). While these forces exist on a continuum, governments may employ relatively more centralized or local internal security forces. More centralized forces are generally subordinate to a national political authority and possess the formal status of the national government. Centralized forces also have greater command and control capabilities, a hierarchical chains of command (including in the execution of government directives), and consolidated funding and recruitment practices (Campbell and Brenner 2002; Perlmutter and Bennett 1980). National budgets provide these forces with the infrastructure and resources required for specialization and procurement of lethal armaments, including heavy weaponry like grenade launchers (Lioe 2011). In many ways, centralized organizations like national police and gendarmerie represent the classic “public provision” of domestic coercion as they are controlled by the national government and nanced through national public resources.6 While variation exists between centrally controlled forces, central command authority makes them distinct from other internal security organizations. Centrally controlled or centralized internal forces often have the oversight of the national government but also the 6Bates(2010). Even historically, centralized forces were the traditional European model for maintaining internal order and conducting civilian policing (Tilly 1992).
4 authority and capability to engage in a range of coercive activities. In contrast, local forces have limited regional or local jurisdiction with more constrained capabilities. They are managed by local leadership (or have allegiances to local actors), tend to have local funding and recruitment practices, and engage in limited geographic areas of operation (Peic 2014). Local security forces include militias, village guards, and civil defense forces. Given localized budgets and resources, they usually have lighter weapons and less specialization than centralized forces (Asfura-Heim 2014). Local forces may have more experience with the community in which they operate and local oversight mechanisms, but they lack the national oversight and capabilities of centrally controlled forces.
Scope and conditions
This investigation into internal security forces is certainly not all encompassing, nor is it an attempt to understand the full array of armed actors within the modern state.7 I exclude foreign forces, private military contractors and municipal police for theoretical and practical reasons. First, private security contractors share more in common with multinational companies than security forces. Private security contractors operate as legal entities with conventional corporate structures and, in many cases, have no easy foreign or domestic divide (Holmqvist 2005).8 Private security contractors often conduct non-coercive support functions as well, like logistics or supply support (Avant 2004; Holmqvist 2005). For these reasons, private military contractors do not t within this dissertation’s discussion of internal security forces and the localization or centralization of security within the state. Similarly, governments may use foreign armed actors, such as foreign militaries, foreign security contractors or foreign rebel groups (Morrow 1993; Singer 2008; Salehyan 2010). While states can and do generate coercive capabilities through foreign forces, I am interested in the state’s generation of security capabilities within its territorial jurisdiction and−therefore−only focus on the state’s use of domestic providers of violence. While other work investigates the state’s use of foreign forces (i.e. Salehyan 2010), 7As John Keegan asks in A History of Warfare, “Bodyguards, regulars, feudatories, mercenaries, military colonists, conscripts, serf militias, remnants of warriors tribes from the steppe...can we impose any order on this medley?” (1993, p. 222). 8In his study of private military contractors, Singer(2008, p. 90) nds that “a [security] rm that one day is considered international could quickly close shop and then open domestic a liates the next day...”
5 reasons why states actively undermine their own monopoly on power or limit their presence only in certain regions of their country is less clear. Finally, municipal police may engage in security operations in some countries but their primary responsibility remains civilian policing. Municipal police agencies in the United States have expanded their activities to include counterterrorism (Waxman 2009). However it is not clear to what extent this experience applies to other countries or remains an exception to the trend, particularly in Europe and Latin America where national police are increasingly involved in security operations (Frühling, Tulchin and Golding 2003; Lioe 2011). While civilian policing and internal security are related, civilian policing consists of law enforcement tasks like preventing crime, protecting property, and arresting oenders of the law rather than eliminating insurgents. The blurring of policing and security tasks does occur, as in the case of crime prevention and counterterrorism (Bigo 2000). However the question this dissertation examines is not whether these tasks are related but how political threats and other political factors drive the organization and distribution of internal security forces. In addition, empirical data on municipal police varies from municipality to municipality, with cross-national data even more prone to inconsistencies in the de nition and measurement of local police.9 For these reasons, I focus my discussion and empirical tests on internal security forces such as militias, village guards, gendarmeries and national police that have explicit authority over security operations. Furthermore, this dissertation’s examination of diametrically opposed security forces−local and central security organizations−allows for a better understanding of tradeos between dierent types of internal security forces, including the dierent logics behind vastly dierent security arrangements. My investigation into internal security forces provides a theoretically informed way to explore how states organize internal power and causes of variation between states. 9Maguire, E.R. and C.D. Uchida (2000). “Measurement and Explanation in the Comparative Study of American Police Organizations.” In D. Duee (ed.), Criminal Justice 2000: Vol 4. Measurement and Analysis of Crime and Justice. National Criminal Justice Reference Service.
6 1.3 internal security forces: tradeoffs in cost and control
Each type of internal security oers its own advantages and disadvantages for the national government. I describe these dierences in terms of costs and control below, since national governments face the competing need to maximize control over security while minimizing its costs.
Costs
Compared to local forces, centrally controlled or centralized forces provide national governments with more capable, more controlled but more ( nancially) costly security. Generating and expanding centralized state forces involves funding by the national government, and the national government must be able and willing to raise su cient national revenue (Regan 1994). National governments in some cases receive funds and resources from foreign aid but this is not always the case.10 Expanding centralized forces also requires mobilizing more personnel into national security services (Tellis 2001). This alone generates substantial administrative and overhead costs for the national government, including funding for national recruitment and training programs, provision of support material, along with providing additional pay and bene ts for new recruits. For example, the French Gendarmerie and National Police represent a substantial portion of France’s total government spending, “accounting for a signi cant number of state employees (240,000)” and 12% of total government employee spending.11 The national government in some circumstances may forgo these administrative and overhead costs by turning instead to local armed forces and local governments.12 Previous studies on militias emphasize that localized forces are nancially cheaper than gendarmeries or militaries (Byman 2005; Salehyan 2010). Alvarez(2006)’s research on the Rwandan Interahamwe nds that the Rwandan national government turned to local Hutu militias in part because the government needed to increase domestic capabilities quickly but lacked the resources and funds to do so. Similarly, Salehyan(2010, p. 502) describes that 10For example, France provides aid to Togo for security system reform and trains Togo’s gendarmerie. The Securitization of Foreign Aid. Ed. Stephen Brown and Jorn Gravingholt. Palgrave Macmillan. 2016: p. 75. 11 Cour des Comptes. “National Police and Gendarmerie: Remuneration costs and working time issues.” March 2013. 12Not all national government always have a choice between national or local armed forces, including situations where stronger foreign powers or interveners direct the government’s policies (i.e. Yergin, Daniel. The prize: The epic quest for oil, money & power. Simon and Schuster, 2011.) While not the only factor, this dissertation simply asserts that a national government’s preferences and policies remain important factors determining a country’s security landscape.
7 “one of the most important functions of delegation [of security] is as a cost-saving device. Delegation allows principals to internalize costs associated with arm’s-length transactions as well as realize e ciency gains associated with specialization.” In other words, delegation to local providers of security might allow national governments to implement security policies in a less nancially burdensome and less resource intensive manner. Even in case of regular civil defense units, these forces have relatively light weaponry and less training, meaning a less hefty nancial budget and lower salaries. Turkey’sVillage Guards received only two weeks of training, carried at most automatic ri es and radios, and received a monthly stipend of around 250 US dollars (USD). (Biberman 2016). In contrast, Turkey’s gendarmerie personnel receive training for a full year from formal Gendarmerie Training Command schools.13 In extreme cases, national governments may not even have resources available for central security forces or the administrative infrastructure necessary to raise and redistribute revenue (Besley and Persson 2011). Even with the assistance of external aid, ‘broke” governments cannot necessarily aord the recruiting, training or arming necessary to establish and maintain central armed forces (Francis 2005; Singer 2008).Even signi cant foreign military aid may not provide enough resources for a national government to create an eective centralized force. Consider the United States’ eorts in Afghanistan, where the U.S. Government appropriated approximately $64 billion from between 2002 and 2016 to train and equip Afghan security forces that included the Afghan National Police. Despite these eorts, the Afghan National Police display considerable shortcomings in terms of human rights abuse, corruption and lack of eectiveness against Talibaninsurgents. Special Inspector General for Afghanistan Reconstruction. “Quarterly Report to the United States Congress.” April 2017, p. 23. In the context of failed states, generating capabilities through a regular central force is simply not possible. Somalia was without a central government after the collapse of the Siad Barre dictatorship in 1991 and lacked basic institutions required to develop centrally controlled security forces.14 Expansion of central forces imposes direct nancial costs on a state that may be unable or unwilling to bear it. 13Gendarmerie Schools Command. “Training and Education System.” Gendarmerie General Command. 14Lyons, Terrence, and Ahmed Ismail Samatar. Somalia: State collapse, multilateral intervention, and strategies for political reconstruction. Brookings Institution Press. 1995.
8 Control
While centrally controlled forces cost more nancially, they likely oer the national government command authority over security along with improved military command and control (C2) capabilities within those forces. Separate from political control, command authority refers to the legitimate power of national leaders to direct security forces and have security forces take action within the scope of their position.15 Command and control refers to the capability wherein delegation of authority within security forces ensures mission requirements are met (Allard 1996). In centralized forces with functioning hierarchical chains of command, directives of commanding o cers are passed down and carried out by lower level o cers, and those lower level o cers can communicate back up the chain of command.16 Military justice systems within these central security organizations provide further ways to discipline o cers, holding trials and punishing those who fail to obey orders or regulations. In addition to the national government having command authority over these forces, central forces like gendarmerie and national police generally have greater command and control capabilities compared to local alternatives. While C2 capabilities vary between gendarmeries and national police, their typical command and control structure is hierarchical and is similar to the military, allowing for centralized coordination of resources and operations (Lutterbeck 2004). Command and control is an important part of a functioning security apparatus, ensuring that lower level o cers eectively carry out directives from leadership (Perlmutter and Bennett 1980; Regan 1994).17 Given that the government can expand its central armed forces, both a national government’s command authority over its security forces and C2 capabilities are advantageous for combating other conventional forces and when engaging in trans-border con icts. First, greater command and control 15Army, United States. Field Manual 7-21.13: The Soldier’s Guide. 2004. In the United States, command authority formally “originates with the President and may be supplemented by law or regulation. It is the authority that a leader exercises over subordinates by virtue of rank and assignment to a position of leadership” (Ibid.) 16As discussed throughout this dissertation, principal-agent arguments emphasize that lower level o cers or agents may still fail to obey the orders of their superiors. Given the potential for security o cers to slack or not ful ll orders, organizations where command authority is accompanied with eective command and control capabilities (C2) should curb disobedient behavior compared to organizations that lack such capabilities. 17As discussed later, greater military command and control can threaten political leadership when civil-military relations are poor. Under certain circumstances, providing more resources to centralized forces might further jeopardize civilian control (Feaver 2009). However, command and control is an important part of a functioning security force and bene cial where there exists civilian control of the military, as in cases of fully consolidated democracies (Pilster and Bohmelt 2012).
9 enables rapid marshaling and deployment of resources and forces to quickly meet challenges from domestic rivals (i.e. rebels in the Spanish Civil War or American Civil War).18 Hierarchical command structures also allow rapid ow of information and instructions and e cient employment of varied capabilities throughout the country. National governments will have command authority over central armed forces but not necessarily the same authority over local forces, which are usually reporting and accountable to sub-state government authorities. In terms of organizational structure, national police and gendarmerie report to national government authorities who have oversight of security forces’ activities as well as the ability to monitor policy implementation, conduct investigations or reviews into o cer conduct and to oversee the resources, operations and performance of centralized forces (Goldsmith and Lewis 2000). In contrast, local forces are accountable to local governments and local communities. Even in countries with strong civilian control throughout all levels of government, local security forces still tend to report to local rather than national civilian authorities.19 Case studies of village guards and militias similarly nd strong allegiance to local leaders (Ahram 2011; Blocq 2014). For example, in Peru local communities formed and managed self-defense organizations (rondas campesinas) in the Andean highlands. While eventually incorporated into the state’s broader counterinsurgency strategy against the Shining Path, these self-defense committees ultimately reported to communities and not the national government (Fumerton 2001). While use of local forces might curb expenditures and provide operational advantages during counterinsurgencies, local forces decrease the national government’s control since the wielders of violence are not directly under their authority. Even if local security forces do not existentially threaten a state, local forces may be less likely to accept the national government’s authority compared to centralized forces who formally exist under a national administration and must report to national command authorities (Mitchell 2004). When governments generate military capabilities through local forces, they risk decreased oversight, non-compliance or unauthorized use of violence. 18Kalyvas(2008, p. 92) notes that domestic rivals can acquire regular armed forces if “an existing army splits, either because of a failed coup (e.g. the Spanish Civil War) or because a unit of a federal or quasi-federal state, which can claim control over a substantial part of the state’s armed forces, attempts to secede.” 19 Kohn(1997, p. 3) provides a succinct de nition of civilian control: “In theory, civilian control is simple: All decisions of government, including national security, are to be made or approved by o cials outside the professional armed forces, in democracy, by popularly elected o ceholders or their appointees.”
10 In addition, local forces may not be as easily demobilized as centrally controlled troops. While there are instances of centrally controlled troops being di cult to demobilize, demobilization of local forces on average presents more di culty due to the combination of highly autonomous units (a lack of hierarchical command structure) and lack of national government oversight (Asfura-Heim 2014). In Colombia, state-sponsored paramilitary organizations did not demobilize after the central government passed laws requiring it. Successor armed units have remained in a variety of forms; some units continued, unauthorized, as paramilitaries whereas other units transformed into criminal gangs.20 Colombia’s case among others suggest that local security forces may not be as easy to demobilize as they are to mobilize.21 This is not to say expanding centrally controlled forces is always a low-risk strategy for the national government. Central forces can create political costs for a national government wary of a coup or, in the case of established democracies, intense public scrutiny. In countries with poor civilian control of security forces, expansion of forces with more eective command and control can increase the risk that civilian national leaders are deposed by the military and gendarmerie.22 Nationalized forces with more resources do seize power more easily from the existing head of government (Feaver 2009; Pilster and Böhmelt 2011; Powell 2012). In a leading essay on civil-military relations, Feaver(2009) argues that “coercive power often gives [security forces] the capability to enforce its will on the community that created it” (1999, p. 214). Empirical trends support this claim; coup attempts have occurred in every major region of the world between 1950 and 2010 (Powell 2012, p. 4). Providing the national police or gendarmerie with greater coercive capabilities has been proven risky for certain governments, particularly those that lack domestic legitimacy. In these cases, expansion of central security forces can be risky for the government and incentivize them to instead empower local or regional sources for generating capabilities.
In summary, central armed forces oer the national government command authority, greater 20Human Rights Watch. Paramilitaries’ Heirs: The New Face of Violence in Colombia. 2010. 21Militias in Libya similarly did not demobilize following the overthrow and death of Muammar Gadda in 2011, and rival groups continue to engage in political violence with each other over control of Tripoli. See Raghavan, Sudarsan. “A day in Libya’s capital, just as the civil war reignites.” The Washington Post. 26 May 2017. 22Gendarmerie units took part in the 2016 coup attempt in Turkey against President Erdogan. See Foster, Peter. “How Turkey’s military launched coup against President Erdogan and government.” The Telegraph. 16 July 2016. Coup risk generally tends to increase among countries with lower income per capita, fewer democratic mechanisms, less regime legitimacy and a weak civil society (Pilster and Bohmelt 2012; Belkin and Schofer 2003).
11 command and control capabilities and specialization but at a higher nancial cost and even higher political risk for governments with poor civil-military relations. In contrast, local forces are nancially less expensive but are less capable and often are not fully controlled by national authorities. For better or worse, local forces are more likely beholden to local rather than national leaders. Therefore, this dissertation examines the international and domestic conditions under which national and local forces are more common, and the eects of these organizations on human rights. By explaining the conditions under which dierent security forces are adopted and the consequences of that adoption, this dissertation answers broader questions on how governments organize force under severe constraints and when confronting a multitude of challenges to their governance. I nd that challenges of anti-colonial wars and counterinsurgency campaigns are associated greater reliance on local security forces, providing governments with limited resources the ability to engage in irregular warfare and population-centric information gathering. States using militias during these campaigns may be unable to demobilize them, though; I nd that these groups remain long after wars end. Second, democratic but limited constraints on national leaders unintentionally encourage the adoption of centralized security for repressive purposes. I show that centralized internal security within limited democracies (where national leaders are not completely subordinated to other political institutions) are associated with a higher likelihood of widespread repression. Internal security forces both shape and are shaped by the political institutions and domestic populations they operate within.
1.4 plan of the dissertation
In the following chapters, I provide theoretical and empirical evidence that national governments organize security based on the need to confront or minimize dierent armed challenges under nancial and political constraints. First, I contribute to the growing body of research on the state coercive apparatus by presenting a new global dataset on internal security forces and re ning the conceptualization and measurement of domestic force. My dataset distinguishes between forces centrally controlled by the national government (gendarmerie or gendarmerie-like paramilitaries) and those relatively localized (such as militias). This
12 includes whether security forces have more national or more local jurisdiction and whether they have clear reporting to a national administration. Based on paramilitary data from the Military Balance’s annual reports (1970-2011), my resulting dataset provides yearly information on the manpower and names of gendarmeries and militias within each country between 1970 and 2011. Given that militias can exist autonomously within the state−such as the tribal militias of Somalia, Yemen or Oman−I include additional information for whether the government has institutional or legal control over the security force. This dataset allows for a more systematic examination of the tradeos across forces and exploration of why states rely on certain security arrangements, including using forces that are beholden to local−rather than national−authorities. I provide an improved operationalization of security forces in panel studies, present new measures of internal armed forces, and describe how it allows us to better understand conditions under which a more sustainable and democratically-oriented security apparatus emerges. Through measuring dierent types of internal security forces, this dataset helps us understand why there exists variation in the state security apparatus and the consequences this has for human rights. After providing a new framework to understand domestic security, I move to examine the emergence of state security arrangements through the longer lens of history. I explore how the pathway to statehood itself in uences the creation and mobilization of internal security forces. The experience of frequent interstate war and state-building in early modern Europe suggests that con ict contributes to the centralization and consolidation of security. Under the European model, external competition encourages sovereigns to wage centralized campaigns and consolidate state security. The creation and maintenance of centrally controlled security also contributes to the sovereigns ability to eliminate domestic challenges and pacify opposition internally. Yet, the role of violence in recent anti-colonial wars suggests another dynamic; rebels engaging in anti-colonial wars against stronger colonial powers must ght insurgency campaigns, relying on guerrilla tactics that necessitate irregular localized forces. I argue that an underlying logic uni es these two models. New states organize security strategically in response to rivals, but this response is dependent on the type of rival and international system competition occurs within: whether burgeoning states are primarily concerned with external competition
13 and waging centralized campaigns against peers or, alternatively, are engaged against stronger imperial powers in decentralized campaigns. In the latter case, post-revolutionaries must deal with irregular and localized former insurgents, whereas more interstate competition with peers promotes consolidation and regulation of forces. My empirical ndings suggest that wars of independence among post-colonial states are associated with local security arrangements, whereas non-violent independence is associated with more centralized forces. The process and international system through which national governments emerge have lasting consequences on the organization of internal security within the state. Factors more immediate also in uence how the national government strategically organizes force. Governments often rely on a variety of armed actors especially in civil con ict, devolving force to irregular local actors even when they have more regular centrally controlled forces available. To understand why states develop risky security arrangements, I disaggregate the challenges governments confront in civil con ict and explore the advantages of dierent types of security forces. I argue that state’s organization of internal security during domestic con ict re ects the tension between central control of security forces and the tactical advantages of local forces that understand their local environment. Governments confronted with rebellions emerging from isolated subpopulations or certain ethnic groups will rely more on irregular local forces, who are better positioned to provide local area knowledge, access valuable social networks and provide defense within their communities. Yet when needing to deploy troops across their country, increase the lethality or complexity of operations and collect information through more technical means, governments mobilize more centrally controlled organizations. I use my novel panel data on internal security forces in civil con ict to evaluate these claims; my ndings show that national governments engaged in con ict with isolated subpopulations do rely more heavily on irregular local forces, whereas regular centrally controlled forces increase with con ict’s distance from the capital. This chapter describes previously unacknowledged but critical variation in the information environment of the state. The type and magnitude of these challenges, in turn, explains the organization of state security during civil wars. After examining why governments organize internal security in the way they do, I turn to the consequences these domestic arrangements have for human rights abuse. While previous research has
14 explored the relationship between domestic political institutions and abuse, the in uence of coercive arrangements remains relatively unknown. This chapter uses my new measures of internal armed forces to argue that centrally controlled security organizations provide national governments with more e cient ways to domestically deploy forces for repression, especially when constrained from using other coercive institutions like the military. National governments with regular centrally controlled forces have the capacity, and sometimes the incentives, to engage in widespread repression. Findings from an empirical analysis of internal security support this claim; centrally controlled security is positively associated with state repression, especially when the state has only limited constraints over its executive. My results oer strong support that democracy alone does not eliminate human rights abuse but only changes the organizations and methods for engaging in abuse. Having only some democratic constraints over national leaders is not enough to discourage repression but requires transparency and public accountability.
I conclude this project by providing a summary of the dissertation’s contributions, discussion of its limitations and avenues for future research. This project presents a new typology and measurement for internal security forces. Rather than considering one type of internal security organization, my work examines the costs and bene ts of dierent types. This approach provides a broader framework for considering how states organize internal security and a more nuanced perspective on the nature of state power. We began by outlining the questions that we have often posed immediately after violent acts in the past. It is now time to consider the future. What questions should we ask if another Middle East state begins to fail? Should we ask “what type of government do they have?” Should we ask “what is the ethnic background of the population?” I argue these questions oer incomplete answers. We should instead turn to the domestic and international threats governments face and organizations of force they use to address such threats. By examining the institutions of violence, academics and policymakers will be better prepared to respond to civil war, terrorism, state failures, and other situations that cause mass violence. Finally, my dissertation contributes to research on the formation and evolution of state power. Governments will continue using a variety of internal forces to address current and emerging threats,
15 though the organizations they adopt have substantial and lasting consequences. We should consider the relationship between the government and its armed forces as ever adapting and evolving.
16 Chapter 2
Internal security forces: Conceptual re nements and a new dataset
2.1 introduction
Why is there variation in how states organize internal security? Speci cally, why do some governments rely on local and often amateur forces to combat threats and maintain internal stability but others use centralized highly professional forces? Existing scholarship oers incomplete answers at best, with most works excluding the role of domestic armed troops entirely and limiting their analysis to the military (Hartzell and Hoddie 2003; Levine and Smith 2000; Vasquez 2005). Some research has begun to examine the state’s use of internal armed forces (Carey, Mitchell and Lowe 2013; Cohen and Nordås 2015; Greitens 2016; Stanton 2015). However, even this scholarship focuses overwhelmingly on one type of armed force (such as militias or secret police) or treats irregular forces as all equal when, in fact, these armed groups dier by their capabilities, organization, operations, resources and oversight by the national government (Carey, Mitchell and Lowe 2013; Pilster and Böhmelt 2011). There has yet to be a rigorous cross-national study on dierences between internal security organizations or of the causes and consequences of various arrangements. I contribute to this growing body of research on the state’s internal coercive apparatus by presenting a new typology and related new global dataset that distinguishes between organizational characteristics of internal security organizations: those under national government control (gendarmeries or gendarmerie-like paramilitaries) and those relatively localized (militias).1 Armed forces that are localized 1Not all armed forces called “militias” may have local management and operations. For the purposes of understanding tradeos between local and national forces, I only consider security organizations to be militias when they display local management and conduct local operations.
17 / local or centralized / centrally controlled may vary by other attributes, including mechanization and recruitment patterns. For coding purposes, I focus on whether the security force is organized and managed under the authority of the national government or sub-nationally organized and managed at regional or local levels, meaning whether forces report to a national ministry (such as the Ministry of Interior or Ministry of Defense) or report to a local / regional administration.2 Based on paramilitary data from The Military Balance (1970 - 2011) annual reports, my resulting dataset provides yearly information on the manpower and names of gendarmeries and militias within each country between 1970 and 2011. Given that some security forces exist autonomously within countries−such as Libya−I include an indicator for whether the government has some form of institutionalized control over the security force. This information is incorporated from Jane’s Sentinel Security Assessments(2010), which provide in-depth information on the organizational attributes of the same security forces found in The Military Balance. Jane’s Sentinel Security Assessments(2010) include information on the force’s geographic jurisdiction, legal responsibilities and whether there exist formal accountability mechanisms to the government. This project focuses on militias and gendarmeries since they conduct similar duties−a combination of internal security and policing activities−but are distinct in terms of their localization/centralization under the state. This allows for a systematic examination of tradeos between forces and structured exploration for why governments rely on certain organizations of internal security, including forces that are beholden to local (rather than national) authorities. Besides militias and gendarmeries, states may have other domestic armed forces not measured in this dataset, such as special operations forces (i.e. India’s Rashtriya Ri es) or military reservists (i.e. the National Guard of the United States).3 While these other types of forces in uence the state’s military capabilities, an examination of diametrically-opposed internal 2This does not explicitly account for recruiting or funding behavior. 3Rajagopalan, Rajesh. “Innovations in counterinsurgency: the Indian Army’s Rashtriya Ri es.” Contemporary South Asia 13(1) 2004: 25-37. Doubler, Michael Dale. The National Guard and Reserve: A Reference Handbook. 2008. Similarly, The United States Federal Bureau of Investigation (FBI) is not included in The Military Balance (2011) and, subsequently, not included in this dataset. However, the FBI could be classi ed as a gendarmerie based on its counterterrorism operations and relatively centralized organization under the national government, reporting to both the Attorney General within the Department of Justice and the Director of National Intelligence. Even though the FBI serves primarily as a federal law enforcement agency, it also engages in domestic counterterrorism investigations. I plan to incorporate it in future iterations of this dataset. Bjelopera, Jerome P. and Mark A. Randol. “The Federal Bureau of Investigation and TerrorismInvestigations.” Congressional Research Service. 2011.
18 security forces provides an initial examination of why governments choose particular ways to organize coercive capabilities. In this chapter, I provide re nements to the conceptualization of internal security, present new measures of internal security forces and describe how these new measures allow us to understand conditions under which a sustainable democratically-oriented security apparatus emerges. In addition to having implications for policy-oriented work, research that only examines one type of security force or ignores the structure of security organizations (i.e. Carey, Mitchell and Lowe 2013; Petersohn 2014) might fail to account for the eects of other coercive forces, dierent combinations of forces and the eects of organizational variation among armed forces. Providing separate measures of internal security organizations, this dataset helps us explain variation in the state security apparatus, the tradeos between diering types of internal forces, and the consequences these dierences have for future stability and human rights.
2.2 current data limitations
This project argues that cost, control and capability tradeos exist over internal security forces, which shape how national governments organize internal security and project force domestically. I examine how these tradeos dier across two main dimensions of internal security: (1) centralization/localization and (2) regular/irregular armed force. These dimensions incorporate a range of security forces that operate within the state’s territorial jurisdiction but are legally separate from the military and police.4 For this reason, this dataset distinguishes between gendarmeries or gendarmerie-like paramilitaries and local militias. Gendarmeries are centrally organized and directly report to a national administration (such as a Ministry of the Interior), have a hierarchical and military-like chain of command, and are nationally coordinated (Lioe 2011). In contrast, local militias operate within speci c territories and have self-contained units, often with no hierarchical chain of command and no clear reporting back to the central government (Ahram 2011; Carey, Mitchell and Lowe 2013). This is not to say that all forces called 4Many irregular forces are unique in that they are “inward” or domestic focused rather than “outward” or externally focused, and relatedly having dual policing and security duties (Davenport 2007c; Lutterbeck 2004).
19 “militias” are geographically restricted but only that militias as de ned in this project refer to locally organized and locally managed security organizations. Similarly, Ahram(2011, p. 13) nds that militias are dierentiated primarily by geography whereas the military is dierentiated by specialization.5 Despite organizational dierences between internal security forces, most cross national studies that quantify state coercive capabilities only examine the military−regular armed forces intended for external security operations and for combat with foreign ghters (Lutterbeck 2004). These measures use military manpower, military expenditures, conventional arms transfers or heavy weaponry as indicators of coercive capability.6 And rather than measuring coercive capability directly, many other studies assess security indirectly through composite scores like the Correlates of War’s Composite Index of National Capability (CINC) (Singer 1988).7 While CINC’s subcomponents provide useful information on non-combat factors, such as urbanization and energy production, the composite score excludes irregular or internal security forces. In other words, it explicitly removes forces identi ed in internal security operations−such as militias and gendarmerie−because “they were not integral to the regular military forces” nor “intended for combat with foreign forces.”8 By only measuring armed forces intended for foreign engagement, the CINC composite score may misrepresent or underestimate coercive capabilities of states, particularly for states most concerned with internal (rather than external) security threats. In other words, the CINC composite score might also only assess potential power for interstate con ict rather than actual coercive power created by the state given the domestic and foreign challenges it faces.9 Scholarship that excludes internal security forces might be missing one of the most important factors related to coercive capacity, including eectiveness against domestic insurgents and the ability to repress challengers to the regime. For quantitative analysis, exclusion of internal security forces could bias models of great interest to international security scholars, including models of battle eld eectiveness, con ict 5In my framework, the ideal militia type is envisioned as a “stationary bandit” or an armed group operating within a consistent area (per Olson(1993)) rather than an armed group moving from place to place. Paramilitaries generally −and militias speci cally−likely vary to the extent they are “stationary” or “roving.” A next step for this project could be to examine the roving behavior of militias. See Beardsley, Gleditsch and Lo(2015) for a related discussion on roving and stationary rebels in civil con ict. 6Hartzell and Hoddie(2003); Kinsella and Tillema(1995); Levine and Smith(2000); Sechser and Saunders(2010). 7These studies include Buhaug(2010) and Hegre(2008). 8Correlates of War Project, National Material Capabilities Data Documentation Version 4.0. 2010: p. 11 9Some studies even argue that state use of regular and irregular forces has changed over time, with internal security forces more frequently used for external operations in recent decades and regular armed forces used for internal operations (Bigo 2000).
20 recurrence and instability. I provide a re ned conceptualization of the state’s security apparatus and new measures of internal security forces that better evaluate theories on political violence over time and across countries.
2.3 data refinements and coding rules
I develop two major re nements to The Military Balance’s paramilitary listings. First, I distinguish whether security forces are gendarmeries or militias through cross-checking paramilitaries in The Military Balance with Jane’s Country Risk Assessments (1999 - 2015, Jane’s) and country-speci c references. Second, I include a variable that designates whether the forces are regular or irregular, meaning whether internal security forces o cially represent the state, have state standardized training and equipment or, alternatively, operate autonomously without o cial management by state authorities.10 Ideally, gendarmeries are centralized under the national government whereas militias are localized and operate through local governments (Ahram 2011; Lioe 2011). In practice, these dierent types of forces might have closer institutional arrangements, including dual accountability to local and national governments, similar responsibilities or shared geographic jurisdictions. Militias and gendarmeries can share other characteristics as well, such as weaponry, tactics or recruits. Yet more relevant to my study on the national government’s capacity for internal force, these groups display key organizational dierences−administration at the national or local level and level of regulation or standardization by the government. My theory on internal security anticipates that these dierent organizations emerge or expand as a strategic response by national governments who face various security challenges under political or nancial constraints. In the following section, I describe the unique organizational characteristics of gendarmeries and militias, which represent alternative sources of security for the national government, and my coding rules for each.
Gendarmeries. Gendarmerie refers to military-like forces with the national command authority to engage in internal security operations and policing of the civilian population. These forces have 10National governments may provide covert assistance to local irregular forces but that does mean these forces are overtly or o cially maintained by the government.
21 hierarchical chains of command like the military and report to national authorities in interior and defense ministries (Andrade 1985; Lioe 2011; Lutterbeck 2004).11 In addition, gendarmeries may operate alongside or in place of the military for major internal security tasks, including border control, domestic peacekeeping and counterinsurgencies (Dunn 1996; Lutterbeck 2003, 2004). Whereas the military traditionally address external security challenges, centrally controlled regular forces like gendarmeries focus on security challenges within the state (Lutterbeck 2004).12 While variation certainly exists among these forces, my de nition of gendarmeries consist of armed forces within the territorial jurisdiction of a state that:
1. have the national authority to exercise jurisdiction over the civilian population,
2. are under the direct control or command authority of the Ministry of Interior and/or Ministry of Defense,
3. are legally separate from regular armed forces and local civilian police,
4. are legally sanctioned to participate in both domestic policing and internal security.
In addition to forces explicitly called gendarmeries in the Military Balance, common aliases include “carabinieri / carabiners” (as in Italy, Chile or Colombia), “civil guard” (as in Spain or South Korea), “national guard” (i.e. Tunisia) or the “national police” (i.e. Bolivia). However I exclude nationalized forces that have only civilian policing duties or certain narrow non-combat duties (such as exclusively guarding borders or acting as custom agents), since I am interested in forces that engage in both internal policing and security.13 Similarly, national police that exclusively engage in civil law enforcement rather 11A Ministry of the Interior usually refers to the o ce that has federal administration over internal security, national emergencies and public administration. 12Gendarmeries and the military can overlap in their activities. One trend in the civil-military relations literature is to claim that security forces are increasingly blurring their roles and responsibilities compared to earlier eras; the military addresses more internal threats and gendarmeries address more external threats (Bigo 2000). Contrary to this assumption, this dissertation discusses in Chapter 3 that “blurred” security arrangements are more the norm than the exception historically; sovereigns in early modern Europe regularly used militaries internally and included irregular local forces in external wars. The United States in the nineteenth century similarly used local militias in interstate con icts (Cooper 2002). 13Even though these forces engage in less coercive activities, border guards and custom agents represent another way to consider the decentralization or specialization of the state security apparatus. They should be considered in future research on how governments organize force, especially for studies examining more subtle forms of state power.
22 than security operations are not counted as gendarmeries nor paramilitaries that only participate in armed con icts and operate as special operation forces (i.e. Thailand’s Thahan Phran).14 Why choose to measure gendarmeries over other centralized security forces? Like militias, gendarmeries represent a unique blurring of military and civilian operations, with internal responsibility for security and policing. In addition, gendarmeries are organizationally unique within the state security apparatus; they are centrally controlled under the national government, have a command structure similar to the military but remain separate from the military and have the authority to engage in domestic policing. Of substantive importance, gendarmeries may represent a trend toward centralization of internal security (Lutterbeck 2004). In cases of relatively well funded and well staed gendarmeries, it might be this type of security organization rather than others (such as militias or militaries) that critically in uences a state’s dometic coercive capabilities and practices.
Militias As with paramilitaries generally, previous research on militias provides a vague conceptualization of these forces that make subsequent measurement and inference problematic. For example, Stanton(2015, p. 900) uses the term militia to refer to “armed groups operating in support of the government but outside of the government’s regular police or military forces.” Similarly, Tresch (2010, p. 4) de nes militias as “any organization of citizens whose task is to protect society, the political system, the territory, and property” or “forces that are beyond the government’s control.” These conceptualizations include a vast mix of armed groups that vary by their geographic jurisdiction, primary responsibilities, capabilities and administration under the state. Therefore, my main criteria for militias is that these forces are locally organized and locally managed (with varying levels of regulation and standardization).15 Paramilitaries are coded as militias if they are:
1. managed and organized within the territorial jurisdiction of the state, 14According to Jane’s 2010 (Thailand Security and Foreign Forces) p. 2, “The Thahan Phran are commanded by regular army o cers and NCOs [and] are employed as light infantry; they carry out patrol duties on the frontiers in order to allow regular battalions to be kept as formed units in their rear.” 15This also extends Carey, Mitchell and Lowe(2013, p. 250)’s de nition in which pro-government militias are “pro-government or sponsored [groups] by the government (national or sub-national), not part of the regular security forces [and] is armed and has some level of organization.”
23 2. armed and having the ability to engage in violent acts and active combat,
3. not part of the regular armed forces or civilian police, and
4. managed by the local government.
Common terms of these localized internal security forces include:
• Tribal/Local/Regional Militia
• Village Defense Forces
• People’s Militia
• Local Defense Militia
• Home Guard Forces
• Provincial Forces
Even by this de nition, militias can dier substantially in terms of the government’s level of regulation and standardization over them.16 Some national governments work with amateur irregular militias that remain at arm’s length from any legal institution, whereas others have well-organized, state-established and state-regulated (regular) militias (Francis 2005). For example, the national government of Jordan formally created its militia system in 1985 and o cially has its militias report under the command of the Armed Forces General Sta, in addition to recruiting militia members directly from the military (Jane’s 2002, Jordan p. 3). In contrast, Oman’s militias (tribal home guards or firqats) consist of autonomously formed, self-trained and self-armed tribesman, many of whom were previously rebels (Jane’s 2003, Oman p. 3). Given the variation in the regulation of these forces under the state, it might be the case that irregular local forces drive violence and abuse, whereas regular local forces are more likely to protect civilians.17 16Militias also may vary according to their level of mechanization and recruitment practices. While currently no cross-national data exist on equipment among militias, Cohen(2011) and Cohen and Nordås(2015) examine the in uence of forced versus voluntary recruitment practices on patterns of violence (including forced recruitment of child soldiers). 17I explore and test the relationship between internal security forces and state repression in Chapter 5.
24 As such, I include an indicator variable for whether the local force is irregular or regular.18 This is a slight deviation from previous measurements of a militia’s relationship with the state. Carey, Mitchell and Lowe (2013) make a distinction between whether the militia is simply legally sanctioned by the government. I use a higher threshold; I distinguish whether a government o cially manages a militia under a local or regional administration.
2.4 sources for data refinements
I use The Military Balance (1970 - 2011) to provide an initial global sample of local militias and gendarmeries. Published by the International Institute of Strategic Studies, The Military Balance gives an annual assessment of global military and defense capabilities, including country-speci c information on security forces, personnel and observed equipment holdings of armed forces.19 For each country, information on security forces sometimes includes length of conscript service and type of specialized units. Other research has begun to explore the importance of mechanization within regular armed services, in addition to conscription or forced recruitment relative to volunteerism (Cohen 2011; Sechser and Saunders 2010; Vasquez 2005). While valuable information, The Military Balance omits substantial information of these other variables for internal security forces. For this reason, I rely on the presence of militias/gendarmeries within each state in addition to the estimated manpower of these organizations. While I am interested in gendarmeries and militias, The Military Balance includes a variety of paramilitary forces, such as troops that engage predominately in policing or only operate as reservists. For example, The Military Balance lists three paramilitary organizations in Belarus for 2010 that operate under the Ministry of Internal Aairs: Border Guards, the Militia, and Ministry of Interior Troops. The Military Balance (2010) does not provide information on the level of centralization of these forces, 18I do not directly measure standardization and professionalization such as training but assume that regular forces are more standardized and professionalized, whereas irregular forces are less standardized and less professionalized. See Kamrava, Mehran. “Military Professionalization and Civil-Military Relations in the Middle East.” Political Science Quarterly 115(1): 2000, p. 67-92. An important exception to this rule is the Sons of Iraq, some of whom were highly experienced and trained former Sunni insurgents and/or former Iraqi military o cers under Saddam Hussein (Biddle, Friedman and Shapiro 2012, p. 37). 19According to The Military Balance (2011, p. 494), “force-strength and equipment-inventory data are based on the most accurate data available, or on the best estimate that can be made...The data presented re ects judgements based on information available to the IISS at the time the book is compiled.”
25 their primary responsibilities or how tightly regulated these forces are by the government. I use Jane’s Security: Country Risk Assessments to better classify these internal security forces.20 Jane’s consist of detailed reports on individual security forces across 195 countries.21 It documents all countries covered by The Military Balance and almost all paramilitaries listed within each country by The Military Balance. Through incorporating Jane’s and other external sources, I better classify each security force. While only having access to Jane’s since 1999 (due to data restrictions), the assessments include valuable background material on most internal security forces. Jane’s provides descriptive summaries of the organization’s primary objectives and jurisdiction, hierarchy and reporting structure under the government. While not found in political science literature as commonly as other sources, international security research frequently incorporates Jane’s assessments and data collections into their analysis.22 Returning to the Belarus example, the description in Jane’s makes it clear that the Belarusian Militia is actually a police force with the name a holdover from its Soviet experience:
“The Belarusian police force is based on the structure of the old Soviet militsiya (militia). The force is under the authority of the Ministry of Internal Aairs (Ministerstvo Vnutrennikh Del: MVD) in Minsk. Each province has a regional Internal Aairs Directorate coordinating the work of local police forces. The MVD has a wide range of additional responsibilities, including issuing IDs and passports to Belarusian nationals and managing migration.”
- Jane’s Security and Foreign Forces, Belarus (2010), p. 1
Jane’s also clari es the role of the Ministry of Interior Troops,identifying that they have a centralized chain of command under the Ministry of the Interior and broad security responsibilities. As such, I designate the Ministry of Interior Troops as gendarmeries. 20A complete list of external references is provided at the end of this chapter. 21IHS. Jane’s Security: Country Risk Assessments. https://www.ihs.com/products/janes-country-risk.html 22Bellamy, Alex J., and Paul D. Williams, eds. Providing peacekeepers: the politics, challenges, and future of United Nations peacekeeping contributions. Oxford University Press, 2013. Cohen, Eliot A. Foreign Aairs. 78(6) 1999: p. 149.
26 “The Ministry of Interior troops (Vnutrenniye Voiska: VV) is part of the MVD...The VV is a motorized (and partly mechanized) security force responsible for supporting the Special Designation Police Units in riot control, protection of government facilities and counter insurgency operations.”
- Jane’s Security and Foreign Forces, Belarus (2010), p. 3-4
I use the information provided by The Military Balance (1970 - 2011) to develop an initial listing of militias and gendarmeries within each state and personnel estimates for each organization. Acknowledging that many names of these forces are ambiguous or potentially misleading, I then veri ed each organization using Jane’s and external sources.23
Dataset description
The resulting dataset covers covers 118 militias across 94 countries and 88 gendarmeries identi ed in 72 countries from 1970 through 2011. While the unit of analysis in this dataset is the country-year, I preserve separate manpower estimates for each internal security force. Given adequate information on each security force, one could repurpose this information for militia-year or gendarmerie-year studies. Using information from the The Military Balance (1970 - 2011), I construct the following variables:
1. Gendarmerie personnel total: estimated total personnel in all gendarmeries for each country per year. This is the sum of personnel across gendarmeries for countries with more than one gendarmerie.
2. Gendarmerie size: estimated personnel in each gendarmerie within a country per year. Size receives a -999 when there is a known organization but not available personnel information.
3. Militia personnel total: estimated personnel in all militias for each country per year.
4. Militia size: estimated personnel in each militia within a country per year. Size receives a -999 when there is a known organization but not available personnel information. 23See the Appendix 2.8 for sources.
27 5. Gendarmerie count: number of distinct gendarmerie organizations within each country by year. This receives a value of 0 when there are no gendarmeries identi ed in the country.
6. Militia count: number of distinct militias within each country by year.
7. State managed: whether or not militias within the country are formally organized and managed through the state. Countries with non-state managed militias receive a “1”, countries with state managed militias receive a “2”, countries with militias of both types receive a “3,” and cases where it is unclear receive a “-999.”
Descriptive statistics: Variation across regimes and time
Table 2.1 shows the descriptive statistics on the constructed variables for internal security forces. All countries at some point between 1970 and 2011 have a military, though not all countries have militaries at all times.24 Among states with militias, 24 states have unregulated militias, 55 states have regulated militias and two states have militias that changed their status over time (Mali and Tanzania). My dataset identi es three countries that concurrently have unregulated and regulated militias: Indonesia, Iran and the United States. From the 1970’s through the 1990’s, Indonesia had unregulated militias (civil defense groups) loosely sponsored by the military in addition to state regulated militias (Kamra) (Ahram 2011). In Iran, irregular tribal militias and home guards co-existed alongside the Basij Popular Mobilization Militia that were established and managed by the government under Ayatollah Khomeini (Golkar 2015).25 Last, the United States has both state-managed militias (Volunteer Defense Corps or State Defense Forces) in addition to autonomous militia groups that are legally sanctioned but not managed by the state (i.e. Michigan Militia Corps).26 24Seventeen countries lacked militaries during at least one year: Afghanistan, the Bahamas, Barbados, Belize, Bosnia and Herzegovina, Costa Rica, Cyprus, Grenada, Haiti, Iceland, Macedonia, Mauritius, Panama, Saint Vincent and the Grenadines, Somalia, Tajikistanand Uganda. 25Iran’s “tribal militias” likely refer to militias associated with komitehs or local revolutionary committees, where local priests or mullahs managed these forces but these forces were allowed by the state (Ahram 2011, p. 110). 26The National Guard Regulation manual 10-4 (p. 5), National Guard Interaction with State Defense Forces, states that “any State, Territory,or District of Columbia that creates a SDF under USC Article 109 is solely responsible for the establishment, training, equipping, funding, management and employment of that SDF in accordance with its law.”
28 Table2.1: Descriptive statistics of internal security forces
Variable Obs Mean Std. Dev. Min Max Gendarmerie Personnel Total 1,807 52,544 160,414 350 1,850,000 Militia Personnel Total 1,310 676,507 4,673,010 250 94,500,000 Count of Gendarmerie Organizations 7,180 0.27 0.48 0 2 Count of Militia Organizations 7,180 0.21 0.45 0 3 State-Managed Designation 1,353 1.763 0.490 1 3 Militia1 - Personnel 1,417 388,379 1,057,986 -999 7,000,000 Militia2 - Personnel 102 702931 1,566,615 -999 6,000,000 Militia3 - Personnel 15 17,600,000 36,200,000 -999 87,500,000 Gendarmerie1 - Personnel 1,843 49,400 158,694 -999 1,850,000 Gendarmerie2 - Personnel 110 35,168 25,383 340 65,000 Year 7,180 1992 11.95 1970 2011
Both militias and gendarmeries display substantial variation in manpower that is right skewed (means being higher than the medians). This is not surprising given variation in populations across countries, in addition to likely variation in how heavily states can and will recruit their citizens into internal security forces (including dierences in recruitment mechanisms). For example, Shambaugh(2002, p. 173) notes that “the Chinese Communists have always seen the local militia as an essential part of the people’s war strategy of drowning any invader in a sea of humanity. This was particularly the case during the 1960’s and 1970’s, when China faced possible invasion from the Soviet Union and the population was mobilized for war.” China’s militias numbered up to 94.5 million personnel during the 1970’s. In contrast, militias in the Philippines numbered around only 400 personnel during the same time period. This could partially be due to measurement error, such as militia personnel being overestimated for China and underestimated for the Philippines, or due to dierences in recruitment mechanisms. Whereas the government of China mass mobilized its population, the Philippines selectively established rural defense forces during this time.27 The number of gendarmeries and militias change over time, speci cally with the change in the international system at the end of the Cold War. Figure 2.1 presents trends in the total number of militias and gendarmeries from 1970 through 2011. Militias become more common throughout the Cold War with a peak around 1987. These organizations then decline at the end of the Cold War and throughout 27Human Rights Watch. The Philippines: Violations of the Laws of War by Both Sides. 1990: p. 60.
29 Figure 2.1: Militias and gendarmerie organizations worldwide
the 1990’s and 2000’s (with a notable increase in 2001 and 2002). Gendarmeries experience a sharp rise from 1985 to 1986. As I show in the following chapters, the rise of gendarmeries and fall of militias during the late 1980’s is likely due to changes in the types of rivalries confronted, external support provided and growth of limited democracies or democratizing states throughout the 1980’s until the end of Cold War.28 Figure 2.2 provides trends in internal security arrangements by country year. Countries with only militaries (as opposed to those with militias or gendarmeries) are the most common across time. However countries with only militaries experience a drop in the 1980’s as countries with both militaries and militias increase. Countries that concurrently have all three forms of security forces (militias, gendarmeries and militaries) are relatively less frequent. As I explore below and throughout the dissertation, this could be due partially to measurement error (internal security forces not being recorded by The Military Balance) or due to changes in the international system and types of security challenges national governments 28Strand et al. nd increases in democratizing states from the 1980’s through the early 1990’s, followed by a reverse wave in the 1990’s (speci cally in Sub-Saharan Africa and former Soviet Union states). Strand, Havard, Havard Hegre, Scott Gates, and Marianne Dahl. “Why Waves? Global Patterns of Democratization, 1820-2008.” PRIO. 2012. Though the resolution of civil con icts could drive changes in security arrangements (Fearon and Laitin 2003), the absence of civil con ict does not imply the absence of other internal political threats (including mass protests, coups or domestic terrorism).
30 Figure 2.2: Countries with militaries, militias and gendarmeries
confront. Last, Figure 2.3 displays the variation in militias and gendarmeries across regime types (-10 being the most autocratic and +10 being the most democratic).29 Both state-managed and non-state managed militias appear associated with autocracies. In contrast to militias, gendarmeries are common in both autocracies and democracies.30 As Chapter 3 and Chapter 4 show, gendarmeries are in part a function of a state having the national bureaucracy and resources for centralizing force, whereas states lacking in infrastructure or central administrations lack the ability to raise and fund national forces.31
2.5 potential measurement error
Despite its global coverage, The Military Balance and my subsequent indicators of internal armed forces may have random and nonrandom measurement error associated with them. My measures may fail to accurately determine the presence of gendarmeries and militias in a given country year, mis-categorize security forces and/or misestimate the number of personnel within each force. For gendarmeries and 29Information on regime type is from Marshall and Jaggers(2005)’s Polity IV project. 30I examine the relationship between domestic political institutions and internal security forces in the subsequent chapters. 31However as explored in detail within this dissertation, central capacity is not the sole determinant for whether states have gendarmeries or not. Peaceful decolonization, adoption of colonial security arrangements, and domestic democratic institutions are positively associated with the emergence and expansion of gendarmeries.
31 Figure 2.3: Country years with militias and gendarmeries by regime type
militias, personnel estimates are sometimes missing in The Military Balance (designated as -999). While I discuss estimation strategies to minimize bias in the following chapters, I provide a brief overview of the likely reasons for measurement error in this dataset and how I address these potential errors.32 First, The Military Balance only includes paramilitaries in its assessments when it identi es an armed force whose “training, organization, equipment and control suggest they may be used to support or replace regular military forces” (2014, p. 559-560). In the absence of an identi ed paramilitary, The Military Balance excludes the paramilitary section from its annual assessment for that country. In my subsequent dataset, the count and dummy variables receive a “0” in the absence of any gendarmerie or militia organization listed under The Military Balance. My internal security force variables thus receive 32 For example, I use the presence of internal security forces in robustness checks, since security force presence presents far fewer missing observations and potentially more accurate indicators than personnel numbers that have more missingness and are potentially misestimated for certain countries.
32 a “0” under three circumstances: (1) if the country actually does not have a militia/gendarmerie, (2) militias/gendarmeries exist but The Military Balance fails to identify them, or (3) The Military Balance identi es paramilitaries but I mis-categorize them as a gendarmerie or militia. Militias in particular could be missed or mis-categorized for a variety of reasons. Militias can be di cult to identify if they operate independently from the government, including where militias only receive covert support from the government (Campbell and Brenner 2002). Unregulated militias may align and coordinate with the government during some times but not others (Jentzsch, Kalyvas and Schubiger 2015). Due to di culty in detecting state-aligned militias, my indicator variable might underestimate the presence of these organizations (particularly unregulated militias) relative to gendarmeries. Second, certain internal security forces might be missing based on how The Military Balance de nes paramilitaries. Since The Military Balance focuses on armed groups “used to support or replace regular military forces,” its de nition may not always include forces managed by local communities or local governments (even if these forces are legally sanctioned by the national government). The Colombian government in the 1960’s passed legislation that legally allowed civil defense groups to form and subsequently armed those groups to combat leftist guerrillas (Holmes, Piñeres and Curtin 2006). The Military Balance may exclude internal security forces created by local communities, even when they are aligned and allowed by the national governments. Consequently, my corresponding indicators would underestimate local forces relative to national forces.33 Last, the The Military Balance likely underestimates the number of distinct internal security forces within a state, particularly among localized and less regulated internal forces. For this reason, I have both a binary indicator and count variable for the presence of internal security organizations. The Military Balance will often identify the presence of a “militia” for a country that in reality combines many distinct militias. For example, The Military Balance identi es for Afghanistan “several regional militias” under one heading between 1985 and 1990, in addition to “tribal militias” in 1991 and 1992. This listing of “militia” and related militia personnel in Afghanistan combines multiple, distinct tribal militias into a 33Studies focused on localized non-government security forces should incorporate information from alternative cross-national data sources, including Carey, Mitchell and Lowe(2013)’s Pro-Government Militia Dataset and Cunningham, Gleditsch and Salehyan(2009)’s Non-State Actor Dataset that both measure autonomous armed forces.
33 single organization. Therefore, my measure for Afghanistan militias captures the presence of militias and total personnel across militias rather than the true count of separate militia organizations. This is less an issue for state-managed militias and gendarmeries, which are more identi able through publicly accessible and government-provided information.34
2.6 comparison to other measures
How does my internal security force data compare to existing global measurements of militias and gendarmeries? No such measure currently exists that explicitly identi es gendarmeries, but I can compare my militia indicators to the Pro-Government Militia Dataset.35 Acknowledging the variation between autonomous and more regulated militias, these datasets express dierent aspects of internal security; the Pro-Government Militia Database better identi es distinct, non-state (but state-aligned) armed organizations whereas my data better identify state-regulated institutions and the extent to which these institutions are localized or nationalized. Dierences in these datasets suggest that international relations research must consider institutional and even bureaucratic aspects of the coercive apparatus rather than considering dichotomous and often arti cial “state” versus “non-state” delineations. Carey, Mitchell and Lowe(2013)’s dataset provides a useful comparison as it also documents the presence of irregular organizations in the state between 1981 and 2007. The dataset codes in all state-aligned paramilitaries as a pro-government militia, regardless of whether the militia is managed locally or nationally. As shown in Table 2.2, the Pro-Government Militia dataset even identi es groups as militias that are actually nationally-managed and nationally-coordinated (such as gendarmeries or national police). For example, the Pro-Government Militia dataset identi es Argentina’s death squads throughout the Dirty Wars (1974-1983) as “militias,” even though these internal security forces were operating nationally under the Argentine Military Government (Lewis 2002).36 In contrast, The 34Governments provide information on their internal security forces online. For example, see https://www.jandarma. tsk.tr/ing/start.htm for information on Turkey’s gendarmerie, http://www.fiep.org/member-forces/ french-national-gendarmerie/ for France’s gendarmerie, and https://www.scdf.gov.sg/content/scdf_ internet/en.html for Singapore’s civil defense forces. 35Carey, Mitchell and Lowe(2013). 36The Pro-Government Militia’ coding rules even mention that these death squads were created by the military government to protect national institutions. See http://www.sowi.uni-mannheim.de/militias-public/data/pgag/238/.
34 Military Balance identi es security forces that are formerly associated with the government, more commonly leaving out armed forces that exist autonomously within a country. For these reasons, I compare observations of militias and gendarmeries in my dataset with militias in the Pro-Government Militia Database. Then, I compare measures of state managed militias: identi ed as “state managed” in my internal security force dataset and “semi-o cial PGMs” in Carey, Mitchell and Lowe(2013). Finally, I examine measures of autonomous militias: “non-state managed” in my dataset and “informal PGMS” in Carey, Mitchell and Lowe(2013). The analysis suggests the importance of careful conceptualization of internal security forces.
Figure 2.4: Countries with internal security forces
As shown in Figure 2.4, both datasets identify a similar number of countries with militias over time. The internal security force dataset identi es 89 countries with militias and the Pro-Government Militia dataset (PGMD) identi es 88 countries. However, the datasets disagree on approximately 30% of the countries as shown in Table 2.2. The PGMD identi es militias in 33 countries that I do not nd militias in. In addition, the PGMD identi es militias present in 11 countries that I only identify as having gendarmeries. National paramilitary organizations like gendarmeries are sometimes counted as militias in the PGMD. My dataset also identi es militias present in 22 countries that the PGMD does not include. The Military Balance better identi es state-regulated internal security forces or internal security forces with the explicit authority of the state, whereas the PGMD better identi es state-aligned but not state
35 Table2.2: Countries uniquely identi ed as having militias per dataset.
Mil. Balance PGMD Only Only Cuba Haiti Suriname Panama Brazil Bolivia Uruguay Paraguay GDR Chile Poland Argentina Slovakia U.K. Bulgaria Belgium Azerbaijan France Mali Spain Benin Ger. Fed. Rep. Chad Macedonia Somalia Greece Lebanon Estonia Yem. Arab Rep. Belarus Yemen P.R. Armenia Kuwait Georgia Oman Denmark Taiwan Sierra Leone North Korea Togo Vietnam Nigeria Singapore Kenya Malawi South Africa Lesotho Madagascar Turkey Israel Tajikistan Nepal East Timor Papua New Guinea Solomon Islands
Bolded countries have gendarmeries identi ed by the The Military Balance.
36 Figure 2.5: Countries with regulated militias
Figure 2.6: Countries with unregulated militias
managed armed forces. These two datasets overlap on 67% of the observations, including on the United States (1983-1988), Mexico (2000-2007), Guatemala (1983-1993), Netherlands (1981-1990), Switzerland (1985-1990), China (1981-1987), Libya (1983-1993), Sudan (1993-2007) and Indonesia (1981-2004). However, the Pro-Government Militia dataset identi es more country years overall with militias than I identify. The Pro-Government militia database incorporates more observations of internal security forces but at the risk of including non-state actors that lack alignment with the government and additionally fails to distinguish
37 between localized versus centralized forces. My internal security force data have fewer observations of militias but a more concise de nition and consistent measurement process.37
2.7 conclusion
In this chapter, I explored previous measurements of internal military capabilities, shortcomings of these measures when used for statistical inference, and better ways to consistently measure and assess internal security forces. I expanded on previous data collection eorts by incorporating information from the The Military Balance (1970 - 2015) and Jane’s to develop a spectrum of forces according to their degree of regulation and centralization under the national government. These conceptual re nements to internal force and the corresponding dataset provide three major contributions for international relations. First, the data will help academics determine which security arrangements promote responsible use of force in the long term, including which security arrangements promote peace and stability in post-con ict contexts and developing countries. Second, I provide re nements to the conceptualization of the state’s projection of power, identifying and distinguishing between national versus local sources of internal force. Last, I generate new cross-national annual measures of the state’s internal security capabilities, which researchers can use to evaluate and compare cases over time and across countries. Overall, this research aids both academics and policymakers in understanding why governments create certain security arrangements and the long-term consequences of these arrangements for the state. Beyond this dissertation, this dataset can be used to investigate the choices and tradeos central governments face when creating and expanding armed forces. This project also contributes to studies on the formation and evolution of state power, allowing researchers to examine how organizational aspects of internal security in uence the government’s power projection and battle eld eectiveness in conventional or non-conventional campaigns. Finally for studies on institutions, this project provides researchers with the empirical evidence to understand how domestic or international political institutions shape and are shaped by coercive organizations. While widely assumed that democratic institutions are 37The Pro-Government Militia Database relies on news sources, which introduces reporting bias between countries and over time. Eck, Kristine. “In data we trust? A comparison of UCDP GED and ACLED con ict events datasets.” Cooperation and Conflict 47(1): 2012. p. 124-141.
38 always associated with more accountable and less abusive countries, it may be the level of democracy or speci c aspects of democracies that in uence the organization of internal force. As shown throughout this dissertation, understanding the various types of internal security forces and organizational dierences between them provides a way to understand and to test the complex relationship between national governments and their sources of power.
39 2.8 dataset appendix
Table 2.3 provides the names of each paramilitary force from the Military Balance coded as either a militia or a gendarmerie. Countries not listed had no evidence of either type of force.
Table2.3: Militias and Gendarmeries by Country, 1970-2011
State Militias State Gendarmeries
Managed
Afghanistan Tribal, Regional Militias No
Khalki Youth, Sarandoli Defense No
Village Defense No
Algeria Legitimate Defense Groups No Gendarmerie Nationale
Angola Org. of Popular Defense, Yes
Popular Vigilance Brigades
Argentina Gendarmeria Nacional Argentina
Austria Federal Gendarmerie
Azerbaijan Karabakh People’s Defense Yes
Bangladesh Ansars Yes Bangladesh Ri es
Barbados National TaskForce
Belarus Ministry of Interior Forces
Belgium Belgian Gendarmerie
Benin People’s Militia Yes Gendarmerie Nationale
Bhutan Royal Bhutan Police
Bolivia National Police
Bosnia & Herz. Serbia Militia Yes
Brazil State Militia Yes National Public Security Force
Brunei Gurkha Reserve Unit Yes
Bulgaria Civil Defense Force Yes Internal Forces
40 Burkina Faso People’s Militia Yes Gendarmerie Nationale
Burundi Local Defense Militia No Gendarmerie
Cambodia Home Guard, Unknown Royal Gendarmerie of Cambodia
Local Forces, Provincial Forces Yes
Cameroon Gendarmerie Nationale
Canada Royal Canadian Mounted Police
Central Af. Rep. National Young Pioneers Yes Gendarmerie
Chad Village Militias No Gendarmerie Nationale
Chile Carabiniers of Chile
China Civilian Militias, Yes People’s Armed Police
Armed Militias, basic Militias
Colombia Rural Militia No Corps of Carabiners,
National Police Force of Colombia
Comoros Gendarmerie
Congo People’s Militia No Gendarmerie
Costa Rica Guardia Civil (Civil Guard)
Croatia Croatian Defense Army Yes
Cuba TerritorialTroop Militia, Yes
Youth Labour Army
Czechoslovakia People’s Militia Yes
DRC Civil Guard Yes Gendarmerie
Djibouti Gendarmerie
Dominican Republic Gendarmerie
Egypt Central Security Forces
El Salvador TerritorialCivil Defence Force Yes
Equatorial Guinea National Police
Ethiopia People’s Militia No
France National Gendarmerie
Gabon Gendarmerie Nationale
41 Ger. Dem. Rep. Workers Militias, Yes
Civil Defense
Ghana People’s Militia No
Greece Greek Gendarmerie
Guatemala Territorialmilitias No
(civil defense patrols)
Guinea People’s Militia Yes Gendarmerie
Guinea-Bissau Guarda Nacional
Guyana Guyana People’s Militia (GPM) Yes
Haiti Civil Militia Yes Police Nationale d’Haiti
Honduras Civil Guard Yes Fuerza de Seguridad Publica
Hungary Worker’s Militia Yes Public Safety Service
Civil Defense Force
India Home Guard, Yes Corps of Military Police,
Defense Security Yes Central Reserve Police Force
Indonesia Militia (Ambigious) Both POLRI
Kamra (People’s Security) No
Iran Bassej Volunteers Yes Gendarmerie
Border Tribal Militia No
Hezbollahi (Home Guards) Unknown
Iraq National Police
Israel Nahal Militia Yes Israel Border Police
Italy Corps of Carabineers,
Guardia Di Finanza (Finance Guard)
Ivory Coast Militia No Gendarmerie
Jordan Civil Militia Yes Gendarmerie
Laos Militia Yes
Lebanon Private Militias No Internal Security Forces
(El Darak)
42 Liberia Militia No
Libya People’s Cavalry Force Yes
Muslim Youth Yes
Lithuania Ri emen’s Union Yes
Luxembourg Grand Ducal Gendarmerie
Madagascar Gendarmerie Nationale
Malaysia Local Defense Corps, Yes Police Field Force
People’s Volunteer Corps Yes
Mali Militia (-1995) No Gendarmerie Nationale
Militia (1995-2005) Yes
Mauritania Gendarmerie Nationale
Mauritius Special Mobile Forces
Mexico Rural Defense Militia no Genderarmeria Nacional
Moldova Carabinier Troops
Mongolia Militia Yes
Civil Defense Force Yes
Morocco Royal Gendarmerie
Mozambique Provincial People’s Militias No
Myanmar Armed Village Defense, No
People’s Militia No
Netherlands Home Guard, Civil Defense Yes Royal Marechaussee
Nicaragua Militia Yes Gendarmerie
Niger National Gendarmerie,
Internal Security Forces
North Korea Workers & Peasant Red Guard, Yes People’s Border Guards
Young Red Guard Yes
Norway Civil Defense Home Guard Yes
Oman Tribal Home Guard No Gendarmerie
Pakistan Civil Armed Forces No Pakistan Rangers
Peru Rondas Campesinas Yes Republican Guard,
National Police Force
Philippines Self defense units Yes Constabulary,
CAFGU Yes National Police
Poland Citizen’s Militias Yes
Portugal Guarda Nacional Republicana
Romania Militia Yes Gendarmerie
Patriotic Guard Yes
Russia MVD Security Troops
Rwanda Local Defense Forces Yes Gendarmerie
San Marino Corps of Gendarmerie
Saudi Arabia General Civil Defense Yes
Senegal Gendarmerie Nationale
Seychelles People’s Militias Yes National Guard
Singapore Home Guard, Yes
People’s Defense Yes
Slovakia Civil Defense Forces Yes
Somalia People’s Militia No
South Korea Civilian Defense Corps Yes
Student Homeland Defense Yes
Spain Civil Guard
Sri Lanka Home Guard No
Sudan Popular Defense Force No Gendarmerie
Suriname National Militia Yes
Sweden Civil Defense Yes
Switzerland Civil Defense Yes
Syria Baath Workers Militia, Yes Gendarmerie
People’s Militia Yes
44 Taiwan Militia Yes
Tanzania Citizen’s Militia (1975-1983) No
Citizen’s Militia (1983-1997) Yes
Thailand Volunteer Defense Corps Yes
Togo National Gendarmerie
Tunisia National Guard
Turkey Gendarmerie
Uganda People’s Militia, Yes Uganda National Police
Local Defense Units Yes
Ukraine Civil Defense Force Yes National Guard
USA Civilian Volunteer Force Yes
State Militias No
Venezuela National Guard
Vietnam People’s Regional Force (Militia) Yes
Local Forces (Militia) Yes
Yemen Tribal Levies No
Yem. Arab Rep. Tribal Levies No
Yemen People’s Rep. Popular Militia Yes
Yugoslavia Civil Defense Yes
Worker’s Militia Yes
Zambia Home Guard, Yes
TerritorialForces Yes
Zimbabwe National Militia No Republic Police Force
45 Table2.4: Cross tabulation of country years with militias, 1981-2007.
Presence of PGMS No Yes Militias All No 2,387 896 Yes 520 522
Table2.5: Cross tabulation of country years with state-associated militias, 1981-2007.
Semi-O cial PGMS No Yes State Managed No 2,917 708 Yes 404 296
46 Sources for coding
Abrahams, Ray. “Sungusungu: village vigilante groups in Tanzania.” African Aairs, 1987: 179-196. Ahram, Ariel. Proxy warriors: the rise and fall of state-sponsored militias. Stanford University Press, 2011. Andrade, John M. World police and paramilitary forces. Macmillan Publishers Limited, 1985. Collelo, Thomas. Angola, a country study. Vol. 550. No. 59. Government Printing O ce, 1991. Curtis, Glenn Eldon, ed. Yugoslavia, a country study. Vol. 550. No. 99. Headquarters Dept. of Army, 1992 Fleisher, Michael L. “Sungusungu: State-sponsored village vigilante groups among the Kuria of Tanzania.” Africa 70.2, 2000: 209-228. Haberfeld, M. R., and Agostino Von Hassell. “A New Understanding of Terrorism.” 2009. Haggarty, Richard A. “El Salvador: A Country Study. Washington: Federal Research Division.” Library of Congress. 1988: 229. Heald, Suzette. “Domesticating leviathan: Sungusungu groups in Tanzania.” 2002. Heinz, Wolfgang S., and Hugo Frhling. Determinants of Gross Human Rights Violations by State and State Sponsored Actors in Brazil, Uruguay, Chile and Argentina: 1960-1990. Vol. 59. Martinus Nijho Publishers, 1999. Historical Dictionary of the Central African Republic. No. 93. Scarecrow Press, 2005. p. 106 Hills, Alice. “Warlords, militia and con ict in contemporary Africa: A re?examination of terms.” Small Wars & Insurgencies 8.1, 1997: 35-51. Jane’s, I. H. S. “Jane country risk assessment.” 1999-2010. Keita, Kalifa. “Con ict and con ict resolution in the Sahel: The Tuareg insurgency in Mali.” Small Wars & Insurgencies 9.3, 1998: 102-128. Kisangani, Emizet Francois, and Scott F. Bobb. Historical dictionary of the Democratic Republic of the Congo. Vol. 112. Scarecrow Press, 2009: 350-351. Konings, Piet. “University students’ revolt, ethnic militia, and violence during political liberalization in Cameroon.” African studies review 45.02, 2002: 179-204.
47 Lungu, Hanania, and Naison Ngoma. “The Zambian military-trials, tribulations and hope.” Evolutions and revolutions: A contemporary history of militaries in Southern Africa, Institute for Security Studies Pretoria, 2005. O’Toole, Thomas, and Janice E. Baker. Historical dictionary of Guinea. Scarecrow Press, 2005: 272. Sloan, Stephen, and Sean K. Anderson. Historical dictionary of terrorism. Scarecrow Press, 2009: 672. Tresch, Tibor Szvircsev. “The transformation of Switzerland’s militia armed forces and the role of the citizen in uniform.”Armed Forces & Society,010. Williams, Kieran. The Prague Spring and its aftermath: Czechoslovak politics, 1968-1970. Cambridge University Press, 1997.
48 Chapter 3
Construction of coercion: Violence, the international system, and internal security
3.1 introduction
Be it thy course to busy giddy minds With foreign quarrels; that action, hence borne out, May waste the memory of the former days.
- William Shakespeare, 2 Henry IV, 4.3.372-5.1
Violence is considered a key element of state formation and state capacity. Across studies on development, the rise of the modern sovereign state is intimately tied to violent processes.2 This includes the monopolization of the legitimate use of violence by controlling internal coercive challenges and eliminating external rivals (i.e. Bates 2010; Krause 1996; North, Wallis and Weingast 2009).3 Tilly(1992) even identi es modern states as essentially “coercion wielding organizations.” Yet states have historically displayed remarkable variation in the organization of force, relying on local forces despite severe limitations in controlling these units and ensuring their loyalty. In medieval England (pre 1500’s), sovereigns relied heavily on the private armies of noble elites, mercenaries and urban 1Stephen Greenblatt et. al. eds. Norton Shakespeare (New York: W.W. Norton, 2016). Henry of Bolingbroke deposed Richard II to become King Henry IV of England (1399 -1413) and the rst king of the House of Lancaster, further pushing England toward civil war over succession of the crown. Tucker,Spencer C. The Roots and Consequences of Civil Wars and Revolutions: Conflicts that Changed World History. ABC-CLIO, 2017: p. 63. 2North, Wallis and Weingast(2009, p. 13) nd that “no society solves the problem of violence by eliminating violence; at best, it can be contained and managed.” 3Weber(1922, p. 4) de nes the modern state as “a compulsory association which organizes domination. It has been successful in seeking to monopolize the legitimate use of physical force as a means of domination within a territory.”
49 militias (Bradbury 2004).4 These heterogeneous forces not only switched sides frequently but ultimately contributed to the weakening of the English crown and strengthening of local magnates (Tucker 2017, p. 65).5 The problems reliance on local forces can create are apparent in wars and state development.6 Local forces weaken central institutions, hinder coordination in security operations and threaten the state itself (Cooper 2002; Greitens 2016; Marten 2007). In Iraq during the early 2000’s, Prime Minister Nouri al-Maliki relied on the support of Muqtada al-Sadr’s Jaysh al-Mahdi militia even though these forces destabilized regions, increased criminal activity and ultimately “played the roles of both ‘arsonist and re ghter”’ (Boduszyński 2016, p. 116). Local forces simply lie about their loyalty to the national government as well. In the early 2000’s, Afghan “anti-Taliban” militias had been known to switch sides and undermined the authority of Hamid Karzai’s government in their regions (Giustozzi 2007). Whereas some scholars have noted the rise in centrally controlled armed forces in recent decades (Lutterbeck 2004), many countries seem to adopt this risky strategy of employing local irregular organizations (Avant 2005; Carey, Mitchell and Lowe 2013).7 Given the risk posed by local security, why might national governments create and even expand these arrangements? And why might some states centralize security to a greater degree, while others do not? Existing research provides only partial answers to these questions, focusing on the role of violence and emergence of security organizations in speci c historical and international contexts. In early modern Europe (1500-1800), nation states consolidated security to vanquish external rivals and gain territory (Avant 2000; Finer 1975; Tilly 1992). European states elded standing armies to acquire or defend territory from rival states elding their own centralized forces (Kestnbaum 2017, p. 308). Yet after the 4English armies employed on average three militia archers for every infantryman (Bradbury 2004, p. 272-274). 5Elton, Georey Rudolph. England under the Tudors. Routledge, 2012. 6The term “local” or “localized” is consistent with previous chapters: state or state-aligned armed forces that operate within a relatively speci c or narrow geographic region and have an explicit tie to the local government and/or local power holders. The term “modern” in this context applies to the post-World War II era. 7Even some democracies employ local and regional armed organizations. Though democracies are assumed to have consolidated and constrained coercive institutions (North, Wallis and Weingast 2009; Pilster and Bohmelt 2012), the most developed and democratic countries employ private military contractors for internal security tasks (Holmqvist 2005). The United States has a long, varied history with highly disaggregated militias, municipal police and private military contractors (Shy 1963; Singer 2008). Israel has established and sponsored autonomous village guards to protect controversial settlement communities (Gordon 2008), and the government of Peru has allowed for self-defense committees or rondas campesinas in the Andean highlands to assist with policing and counterinsurgency operations (Fumerton 2001).
50 decolonization of the Middle East and Asia in the twentieth century, newly independent states often relied on local and irregular forces (Ahram 2011; Krause 1996). In wars of liberation, rebels usually ght stronger colonial powers, using guerrilla tactics and highly mobile small units with cellular organizations (Mao 1937). In these recent cases of post-colonial state formation, new countries turn to local forces as part of a strategy to violently remove colonizers (Hironaka 2009). In this chapter, I argue that an underlying logic uni es these otherwise separate patterns of state formation and internal security arrangements. In both early European and post-colonial contexts, new states organize internal security as a strategic response to competition and rivalries. However, state responses are not uniform over contexts and time periods. Responses of new states depend on the rivals they confront and the international system within which this competition occurs.8 In cases where states compete with peers over territory−as in international systems without norms of territorial integrity−external rivalries stimulate the centralization of security. In modern contexts where state formation is relatively rare and occurs within an established international system, new states form violently through wars of liberation against stronger internal powers, not against external peers. To win, revolutionaries must ght colonial or imperial powers using insurgent tactics and guerrilla warfare. Internal con ict against stronger ruling powers favors localization of security.9 Thus the international system and type of opponent shape the way new states organize security. This chapter begins with a discussion of how violent con ict during state formation in uences the post-con ict arrangement of domestic security and why these arrangements dier across historical and international contexts. In section three, I propose a new theory uniting existing theories on the state’s internal security arrangements. I argue that growing states organize security as a strategic response to the types of coercive rivals they confront.10 The rivalries and competition states confront is also shaped by broader international arrangements. Therefore, the organization of security is dependent on two interrelated factors: rivalries and the international system. Section four presents hypotheses based on 8As discussed in the theory section, the presence or absence of adversaries is associated with the presence or absence of external allies. 9In the twentieth century, state formation also occurs through peaceful transitions from a former colonial ruler. During peaceful transition, recently decolonized states may retain centralized−and still highly repressive−coercive arrangements of their former powers. I explore this logic throughout the chapter. 10By coercive rivals, I mean both internal and external rivals that have the ability to use coercive force and make coercive threats.
51 this theory and provides the empirical strategy for testing them. Section ve presents cross-national empirical trends and the results of regressions, which oer evidence in support of my theory on the relationship between coercive rivalries and subsequent security arrangements. This chapter concludes with a discussion on how the past experiences of states explain today’s organization of security.
3.2 models of rivalry and state formation
Historical studies of security organizations suggest that violent con ict during state formation in uences subsequent security arrangements, though types of rivalries diverge dramatically across contexts. The experience of Western Europe suggests that external competition during state formation and consolidation gives rise to security forces controlled by the central government (Avant 2000; Finer 1975; Tilly 1992). This argument follows from the understanding that, in early modern Europe, polities that engaged successfully in territorial conquest had centrally controlled armed forces (Tilly 1992, p. 63). Alternatively, twentieth century decolonization movements in the Middle East and Asia suggest that con ict during post-colonial state formation contributes to local, less capable forces (Ahram 2011; Krause 1996). Even with the the assistance of foreign support, states emerging from wars of decolonization struggle converting former insurgents−including dierent ethnic groups of insurgents−into a coherent central force.11 Both historical experiences suggest that violent con ict during state formation favors the adoption of certain security arrangements over others. Yet these periods−early modern Europe (1500-1800) and the post-World War II era (1950-present)−seem to provide con icting paradigms of the role of rivalries and war: whether rivalries contribute to the centralization and nationalization of security or, alternatively, whether they lead to the localization of security and ultimately inhibit the state’s coercive capacity. Below, I brie y describe these diering paradigms and then oer a theory to unify them. 11Foreign support can even decrease the national government’s incentive or ability to invest in and maintain national forces, especially if foreign interests back one political party or leader over another (Hironaka 2009).
52 The European model
How might con ict during state formation in uence the government’s post-con ict choices of the coercive apparatus? Arguments based on the European model of state formation suggest that external rivalries are associated with the adoption of nationalized consolidated security. In early modern Europe (1500-1800), the demands of inter-state rivalry and territorial conquest led sovereigns to centralize forces and minimize the use of irregular local options (such as mercenaries and militias) that were harder to coordinate and control over long distances (Andreski 1968; Avant 2000; Finer 1975; Tilly 1992). Because nation states only emerged after the submission or defeat of local lords, centralizing security provided sovereigns with the capacity to deter domoestic rivals, maintain internal control and acquire new territory (Bates 2010; Finer 1975).12 Examining the rise of standing armies in France, Britain and Prussia, Tilly (1992, p. 57) emphasizes external rivalry and preparation for external war as crucial to the consolidation of security: “War wove the European network of nation states, and preparation for war created the internal structures of the states within it.” The need to compete with external rivals for territory drove sovereigns to adopt centralized organizations. A large standing force oered sovereigns numerous advantages over mercenaries and militias when needing to deploy and coordinate troops against external rivals. Mercenaries could be notoriously unreliable, consisting of troops both foreign and domestic, and militias often were amateur and part-time. Local forces lacked standardized chains of command and permanent o cers, which gave rise to command and control problems. Dupuy and Dupuy(1993, p. 501) argue that marshalling these forces before battle was “a long, complicated process with much shouting and confusion.” A lack of professional o cers compounded coordination problems. With no clear leadership oversight or command and control capability, these forces had many of the same weaknesses witnessed among modern militia and mercenary variants. While militias and mercenaries provided sovereigns with a exible and cost-eective means for raising 12Contrary to the early modern period, the feudal medieval period (1100-1500) had more localized security as local noble elites had more in uence relative to their sovereigns. For example, noble elites in feudal England “recognized the king as their superior, but his supremacy was a purely theoretical one, and the only bond which continued−not very eectively−to bind them to the crown was the fact that they were its vassals” (Ganshof 1964, p. 60).
53 military forces, they hindered coordination, reliability, professionalization and−ultimately−battle eld eectiveness. Dupuy and Dupuy(1993, p. 501) describe the challenges of directing these forces in combat:
“Rulers relying so heavily on mercenaries had many headaches. Though usually well-disciplined and perfectly willing to ght (particularly the Swiss and Germans), the mercenaries had no patriotic feelings and were frequently untrustworthy. It was not uncommon for them to ask for bonuses on the eve of battle−threatening to abandon the army, or to join the enemy, if the demand was not met. [And] the undisciplined, poorly organized feudal militia would have been a detriment rather than a help in any oensive operations, aside from the expense of paying them and of supporting them logistically.”
In contrast to mercenaries and militias, standing armies could be out tted, trained and controlled under a national administrative structure with a standardized chain of command (Bradbury 2004). For example, Charles VII of France initially used mercenary companies like the ecorcheurs or “ ayers” during the Hundred Years’ War (1337 - 1453), but these forces engaged in such widespread atrocities and extortion that Charles VII abandoned them in favor of more reliable (though more expensive) standing armies (Rogers, Caferro and Reid 2010). Similarly, the Prussian Duke Frederick William established a standing force to defend against external rivals and maintain Prussian territories internally (Finer 1975, p. 121). Despite the nancial and administrative costs of standing forces, polities with “large, durable military forces from their own population” outperformed other forces in territorial con icts (Tilly 1992, p. 63). While not all polities consolidated security under a centralized administration, polities that did (such as France and England) engaged more eectively in battle and in acquiring new territory. Why did not all states adopt the model of France, Prussia and Britain? Examining the in uence of external rivalries and scal capacity, Gennaioli and Voth(2015) nd that exposure to warfare weakened already fragmented states like Spain and Poland. External rivalries may have encouraged security centralization but not all states were able to centralize. Van Zanden, Buringh and Bosker(2012) similarly nd “a little divergence” in early modern Europe, where state institutions and economic activity increased among northern states throughout the sixteenth and seventeenth century but declined among central
54 and southern states. Polities that failed to centralize institutions experienced a deterioration in economic development and battle eld eectiveness. In his study of state-building in medieval and early modern Europe, Ertman (1997, p. 31) nds that the early Polish parliament (the Sejm) “repeatedly block[ed] attempts to construct a modern state apparatus that might strengthen royal authority and more eectively defend the country against Russians, Swedes, Turksand Prussians in favor of retaining more power at the local level, where it could easily be appropriated by the magnates.”13 Even in spite of external competition, some states lacked the ability to centralize due to the strength of local governments and their perceived lack of need to arm against such external threats. In contrast, states like Prussia and France consolidated armed forces after generating su cient capital and successfully bargaining with local elites (Finer 1975; Tilly 1992). Thus, the national government’s ability to extract resources from local elites and compatible incentives of both local and national elites were required for national security consolidation. International factors also shaped the organization of security in early modern Europe. The centralization of security occurred under an anarchic system lacking modern norms of territorial integrity (Hironaka 2009). States did not confront international constraints against inter-state warfare, such as international recognition of xed borders and territory.14 The consolidation of armed forces also occurred during the Military Revolution (1550-1650). In previous centuries, the earlier medieval soldier “needed little training and less skill: if he inclined his pike in the correct alignment and leaned heavily on the man in front of him, he had done almost all that could be required of him” (Roberts 1956, p. 14). In contrast to the medieval soldier, early modern infantrymen required more training, equipment and ultimately funding. As states like France, England and Prussia engaged in inter-state warfare, the European Military Revolution introduced changes in technology that led to larger, more trained and more expensive armed forces, dynamics that−in turn−in uenced the state’s arrangement of security (Parker 1996; Roberts 1956). Siege warfare methods developed to overcome improved and more elaborate forti cations, which meant larger and more professional armies requiring even greater funds, complex equipment and logistical support (Parker 1996, p. 146). Krause(1996, p. 325) notes that “changes in the 13Ertman, Thomas. Birth of the Leviathan: Building states and regimes in medieval and early modern Europe. Cambridge University Press, 1997. 14Jackson, Robert H., and Carl G. Rosberg. “Why Africa’s Weak States Persist: The Empirical and the Juridical in Statehood.” World Politics 35(01) 1982: p. 1-24.
55 scale of warfare and concomitant revolutions in tactics and organization required vast state investments that were beyond the reach of many local rulers, [which] catalyzed the emergence of the modern state.” Centralization of security occurred under unique domestic and international conditions that encouraged larger and better nanced forces, which only sovereigns with centralized bureaucracies and a revenue base could provide. More broadly, this historical experience suggests that anarchic international systems and external rivalries emerging from such systems in uence the organization of state force and encourage formation of a centrally controlled security apparatus. The early modern period (1500-1800) was strikingly war-prone; state sovereignty was not the norm, states engaged in frequent external ghting, and state formation was a common phenomenon. By and large, these experiences do not apply to modern and speci cally post-World War II contexts. After World War II, state formation occurred rarely and only through the breakup of an existing nation state or removal of a colonial power rather than through territorial conquest or consolidation over political sub-units (Gibler 2017; Tilly, Davis and Pereira 2003). Changes in territory and state boundaries remain relatively rare after World War II compared to the early modern period (Hironaka 2009). Finally unlike the rising states of France and Prussia in the early modern period, there is no reason to expect that new states have the nancial capacity or political incentives to consolidate security like some of their earlier European predecessors. New states may more closely resemble early modern Poland rather than early modern France, with national governments having to cede authority to local leaders and rely heavily on local forces. However the emergence of states in Europe provides insight into the bene ts and risks of dierent internal security arrangements. Though paid for by a state sovereign, mercenaries and local militias lacked political accountability, standardized training and the ability to coordinate on the battle eld. On the other hand, guns-for-hire provided sovereigns with a exible, cost-e cient way to generate force, and urban militias were remarkably well-suited to engage in localized defense operations. Though the international system is extraordinarily dierent between early modern and modern contexts, the costs and bene ts of internal security organizations are remarkably consistent over time.
56 The Decolonization model
State formation in non-European cases after War World II suggest a dierent dynamic. Con ict during post-colonial state formation more likely leads to state reliance on local rather than centrally controlled security. Postcolonial states also confront unique development challenges, created by long-term and extensive resource extraction by colonial rulers (Ahram 2011; Krause 1996; Herbst 2014). Compared to earlier periods, territorial expansion and international war is relatively rare after World War II (Hironaka 2009; Herbst 2014). While European states emerged through territorial consolidation and centralization, new states of the late twentieth century formed through decolonization or secessionist movements (Gibler 2017; Thies 2005).15 Where weak states in earlier periods may have been vanquished by stronger external rivals, new modern states are often supported by third parties and foreign military and nancial aid.16 In examining the birth of states after World War II, con ict does not seem to contribute to security centralization. In fact, many contemporary examples show the opposite: Angola and Mozambique both had localized security organizations following wars of independence from Portugal.17 Similarly, local militias retained a signi cant role in Croatia along with Bosnia and Herzegovina following independence from Yugoslavia.18 These wars of independence suggest that rivalries with stronger imperial powers contribute to localized internal security (Ahram 2011). What contributed to these dynamics? After World War II, the limited number of states that did emerge successfully from wars of liberation had to form security forces from former insurgents and with the aid of external powers. During the Cold War, the Soviet Union and the United States provided support to liberation movements aligned with Communist or anti-Communist ideologies respectively. Hager Jr. and Lake(2000, p. 19) argue that this pattern represented “competitive decolonization” or an attempt by established powers to increase global in uence, where “the Soviet Union [would] support national liberation movements on several continents and the Reagan Doctrine back[ed] anticommunist 15Also see Correlates of War Project. Colonial Contiguity Data, 1816-2016. Version 3.1. 16Examining the in uence of the international community, Hironaka(2009, p. 132) even nds that “dual-sided interventions have been a major factor sustaining and prolonging civil wars of the post-World War II period.” 17Minter, William. Apartheid’s Contras: An inquiry into the roots of war in Angola and Mozambique. William Minter. 1994. 18Alvarez(2006).
57 ‘freedom ghters’ wherever they might arise.” In the presence of substantial military aid and support of an external power, new states may use foreign forces and foreign resources rather than investing in their own national organizations and national bureaucracies. More importantly, new central governments may have to work with multiple and even competing former rebel organizations. For example, Indonesia formed a largely militia-based security system after it “had emerged from revolutionary chaos” of the national revolution that removed Dutch colonizers in 1949 (Ahram 2011, p. 26). Newly formed central governments can lack the ability to unilaterally consolidate power if local leaders they relied on during anti-colonial wars resist. New states can decide not to consolidate security for strategic and bargaining purposes as well, including prevention of military coups or abuse by centralized powers. The United States after the Revolutionary War relied heavily on local civil militias throughout the nineteenth century until the Spanish American War of 1898 (Cooper 2002). Demobilization of the United States Continental Army and return to civil militias was perceived as less risky for the newly formed national government. By relying on local security, states prevent abuse by a centralized armed force (Tyson 2017). Fear of a powerful centralized force, availability of external support and the need to negotiate with multiple armed groups can all limit the national government’s ability and willingness to invest in strong national administrative structures, including military organizations. Unlike in the European model, internal rivalries and dynamics of the post-colonial world drive localization of security within new states. Yetlocalization of security does not imply more democratically-oriented or less abusive organizations. National leaders may strategically localize security to simply protect themselves, rather than their citizens, against internal coercive rivals and popular challenges to rule. These dynamics are evident in cases like the Democratic Republic of the Congo and Angola, where national leaders actively sought to undermine the military and created alternative, competing sources of coercive power that coincided with less reliance and less investment in formal national institutions (Reyntjens 2009). Whereas pressures of inter-polity rivalry incentivized early modern sovereigns to invest in national administrative structures, governments emerging from twentieth-century rebellions must protect against internal rivalries and former insurgents while dealing with (and potentially undermined by) diverse global powers. Emergence
58 from colonial rule−a form of “arti cial” state creation−likely limits the development and strength of national representative institutions (Hironaka 2009, p. 2). External support from state patrons further incentivizes national leaders to reduce investment in national institutions. Thus leaders in the central government can rely on alternative and even foreign sources of security. In contrast to rebellions in the postcolonial world, decolonization through non-violent settlement allows for the adoption of existing security arrangements from former colonial powers. National governments emerging from peaceful transitions do not have the challenge of incorporating former rebel groups into a single organization and are not beholden to local violence wielders (Ahram 2011, p. 4). New governments emerging from peaceful settlements may also adopt the former colonial power’s governing institutions and security arrangements. For example, nation states peacefully transitioning from French colonial rule, such as Niger, continue to use national French-style gendarmeries and receive French security support.19 However, claims that violence during decolonization impacts the state’s delegation of force has yet to be tested cross nationally. It is unknown whether the dynamics of the early modern period apply to modern experiences. Finally, recent decolonization wars are very dierent from interstate violence in early modern Europe. Below I propose a unifying paradigm and generate hypotheses to test its observed implications.
3.3 theory: violence and the international system
Acknowledging the dierences noted above, my theory provides a uni ed explanation of a national government’s reaction to violence during state formation and development. State security arrangements are a strategic response by governments to diering forms of violent con ict during formation. The eect of these challenges is not constant, though, but dependent on two interrelated factors: the relative strength of rivals and the international environment within which states emerge and interact. When the international system is relatively anarchic, as in early modern Europe, states engage 19Gopfert, Mirco. “Surveillance in Niger: Gendarmes and the Problem of ‘Seeing Things’.” African Studies Review 59(02). 2016: 39-57.
59 in frequent warfare with peers (Parker 1996). In these situations, emerging nation states tend to consolidate security so they can successfully compete with peer challengers. Even prior to complete internal consolidation, states shifted from feudal systems to more centralized systems during the fteenth century.20 In England, the War of the Roses−the English civil wars over the throne from 1455 to 1485−further promoted the change from feudalism to centralized security as nobility and their private armies consolidated with either York or Lancaster forces (Tucker 2017, p. 65).21 The in uence of local nobility relative to the sovereign also decreased, as nobles were no longer allowed to have private armies and many were secretly tried and punished following the end of the war.22 The few states that formed after World War II did not emerge into an anarchic international system, but into an established system of states characterized by bipolarity and relatively few interstate con icts. These new states arose through the defeat or withdrawal of colonial powers rather than internal consolidation. The violent path to post-colonial independent statehood often meant competing against a stronger existing state, which led to the adoption of guerrilla or insurgency tactics by rebels. In these revolutionary cases, having a centrally controlled security apparatus was not feasible.23 Yet as in early modern Europe, some governments do have the means and incentives to centralize armed forces under a national administration. The relatively stable system of states after World War II and support from former colonial powers can allow newly independent states to consolidate security under certain circumstances. Governments in decolonized states emerging through peaceful processes may rely on centralized institutions of former colonial powers and protection of the regime by those 20Feudalism was a decentralized system of reciprocal legal and military obligations (from a noble holding land in exchange for labor of a vassal), where governance and customs evolved around a “specialized military class occupying the higher levels in the social scale; an extreme subdivision of the rights of real property...and a dispersal of political authority amongst a hierarchy of persons who exercise in their own interest powers normally attributed to the state and which are often, in fact, derived from its break-up.” Ganshof, Francois Louis. Feudalism. University of TorontoPress. 1964: p. xv. 21The consolidation or semi-consolidation of forces occurred in part because of the increasing practice of executing nobles during this period. As with militias in modern civil wars, nobles in premodern England “on occasion switched sides, and some key battles were decided by treachery. Nobles were the primary targets for the opposing side. Whereas during the Hundred Years’ War troops might have been ransomed for large sums, if taken in the War of the Roses, speedy execution as traitors was a more likely fate” (Tucker 2017, p. 65). 22Elton, Georey Rudolph. England under the Tudors. Routledge, 2012. 23Mao(1937) viewed guerrilla warfare as an initial strategy in rebellion that would eventually give way to conventional operations: “These guerrilla operations must not be considered as an independent form of warfare. They are but one step in the total war, one aspect of the revolutionary struggle.” This is not to say that post-colonial states could do this shift from guerrilla to conventional tactics even if they should.
60 powers (Ahram 2011). During British colonial rule in India, Britain relied on a minority “martial class” of Punjabi and Pashtun in the colonial army (Wilkinson 2015). After independence from Britain, the Indian government inherited this professionalized albeit ethnically imbalanced British-style military that had strong leadership oversight, trained o cers, established recruitment channels and armaments (Wilkinson 2015, p. 10). Similar dynamics occurred among former Soviet states. Ukraine inherited its security arrangements, tactics and personnel directly from the Armed Forces of the Soviet Union.24 Finally, many decolonized states still receive aid from their former colonial rulers and peacefully decolonized states can adopt their security organizations and practices, including recruitment and training practices.25 In addition to having the option of consolidating security forces, peacefully decolonized states do not have the challenges of demobilizing former insurgents or unifying groups of guerrilla ghters. This is not to say that peacefully decolonized states lack problematic ethnic and religious divides, but only that these peacefully formed states can more easily form a coherent and cohesive national force. This includes states that previously had armed forces strategically composed of a single ethnic group (Horowitz 1985). In fact, many peacefully decolonized states (and their former colonial powers) purposefully developed ethnically imbalanced or minority-controlled national forces to protect the regime (Wilkinson 2015). Upon independence, Sierra Leone already had Mende o cers making up a quarter of all security forces, and the second prime minister of Sierra Leone then doubled representation of the Mende in the military at the expense of fewer non-Mende recruits.26 Peacefully decolonized states may inherit institutions and practices from former powers that allow regimes to adopt centralized security models. Newly formed national governments may centralize security in response to external rivalries or border con icts as well. Centrally controlled armed forces proves an eective strategy when engaging in conventional con ict against a peer and protecting territory from external rivals. The centralization of forces in early modern Europe granted new nation states the means to control and manage their security (Parker 1996). Similarly in modern times, centrally controlled armed forces tend to have 24Mychajlyszyn, Natalie. “Civil-Military Relations in Post-Soviet Ukraine: Implications for Domestic and Regional Stability.” Armed Forces & Society. 28(3) 2002: 455-479. 25Alesina, Alberto, and David Dollar. “Who gives foreign aid to whom and why?” Journal of Economic Growth 5(1) 2000: 33-63. 26Cox, Thomas S. Civil-military Relations in Sierra Leone: A Case Study of African Soldiers in Politics. Cambridge, MA: Harvard University Press. 1976.
61 advantages in interstate combat, including standardized chains of command, specialization, and heavy armaments−attributes less feasible for local and underfunded forces. However, as national-level control of security forces increases, so does the cost of coordinating and maintaining those forces. European states could only generate centrally controlled forces after creating national administrative structures for taxation, which required successful bargaining with local elites for revenue (Tilly 1992). The ability to raise su cient revenue and work with local elites constrains a government’s ability to consolidate security within new nation states. For example, France only managed to consolidate security after removing coercive power from local governments, a process that was accelerated and only completed during the French Revolution (1789-1799).27 Even as France and other European states successfully formed professional armies, they still augmented these forces with auxiliary troops (Bradbury 2004). In the post-World War II era, newly formed states faced even greater challenges forming conventional forces, lacking the funds, infrastructure and even incentives at times for such tasks. For example, following independence from Britain, the new government of Nigeria inherited a military that severely lacked adequate resources and materials, in addition to being structured and trained more for internal policing tasks than external security.28 New states may prefer conventional armed forces rather than mercenary companies and militias, but raising such a force is not always feasible. When states emerge after a violent struggle for independence from a stronger power, new central governments confront additional challenges of disarming and co-opting former rebel organizations. After declaring independence from Yugoslavia in 1992, the Republic of Bosnia and Herzegovina faced the di cult task of forming a single coherent security force from distinct ethnic groups hostile to one another. As Rohan(2014, p. 180) describes, “The new country was a complex creation dominated by mistrust among entities and the ethnic groups it contained. Under [the General Framework and Agreement for Peace], the state possessed no armed forces, but the entities retained large, conscript-based armies...with 27In the historical examination of France’s security arrangements, Bayley(1990, p. 67) argues that France “was formed by conquest as kings in the Ile de France extended sovereignty by forces of arms. Powerful feudal lords and regional parlements resisted unitary institutions for years...Louis XIII and Cardinal Richeliue, his chief minister, concluded that regional aristocracies could not be entrusted with government authority.” 28Barua, Pradeep. The Military Eectiveness of Post-colonial States. Brill, 2013. Barua (p. 20) argues that “the military that the Nigerians were left with was essentially a glori ed police force.”
62 well-separated Bosniak (Bosnian Muslim) and Croat components.” New states must consolidate not only organizationally distinct groups of ghters but ethnically distinct groups as well. Finally, central governments may have limited ability to wrest power from local warlords. After the Philippines’ independence from the United States in 1946, militias became an important and consistent feature of its security landscape.29 Given internal challenges, including powerful local elites and geographic factors, and internal threats associated with consolidated security (such as coups), local security represents an imperfect but expected response to the central government’s limited control over the internal use of force. A revolution may have been strong enough to put together a state but perhaps not strong enough to centralize its security.
3.4 hypotheses and empirical strategy
I will evaluate hypotheses from the theory outlined in the previous section, determining to what extent violence experienced during modern state formation in uences a state’s choice between more local or more centrally controlled security organizations. I will also examine to what extent the adoption of a colonial power’s security forces may occur in new states.
• Hypothesis 1: Violent independence is associated with a greater likelihood of local security.
• Hypothesis 2: Security organizations are associated with a former colonial power’s security organization.
To test these hypotheses, this chapter rst examines the in uence of violent and non-violent independence on internal security arrangements across 110 countries formed after World War II, between 1950 and 2010. I examine to what extent successful anti-colonial wars are associated with more local or central security organizations. Second, I include a restricted timeframe of countries formed between 1970 through 2010−the same time period for which I have annual data−to determine how long after state creation certain security organizations are reportedly formed and expand. The results of these empirical 29Clarke, Gerard. Civil society in the Philippines: Theoretical, methodological and policy debates. Vol. 11. Routledge 2012, p. 139-140.
63 tests suggest that violent independence during the twentieth century is associated with subsequent reliance on local security and, absent con ict, countries tend to adopt the security arrangements of their colonial rulers.
Dependent Variables: Internal security organizations
The dependent variable is the presence of four dierent state or state-aligned security organizations: gendarmeries, regular local militias, irregular local militias or a national military. Gendarmeries and the military provide measures of centrally controlled security, whereas militias provide measures of local security. Since governments may employ multiple organizations simultaneously, I include binary dependent variables for all four security organizations. For the time-series analysis, I also include annual personnel measures to evaluate how diering legacies of violence in uence the relative size of these security organizations. The four types of security forces examined here correspond to my typology discussed in earlier chapters:
1. Military − regular national external forces,
2. Gendarmerie − regular national internal forces,
3. Unregular Militia − regular local internal forces, and
4. State-Managed Militia − irregular local internal forces.
As noted previously, The Military Balance (2011) and Pilster and Böhmelt(2011) provide the initial population for paramilitaries and personnel estimates, from which I distinguished between dierent types using Jane’s Country Risk Assessments (1999-2015) (Jane’s) and country-speci c references.
Main Explanatory Variables
To specify whether violence was associated with the process of state formation, I rely on data from the Issue Correlates of War (ICOW) Colonial History Dataset.30 This dataset identi es “the process 30Paul R. Hensel (2014). “ICOW Colonial History Data Set, version 1.0.” Available at http://www.paulhensel.org/ icowcol.html.
64 by which independence occurred, and the colonial power(s) that in uenced the entity or entities that would eventually become the independent state” (Hensel 2014, pg. 1). The data distinguish between four separate independence processes: formation, decolonization, secession and partition, and designate whether these processes occurred through violent rebellions or peaceful transitions.31 As de ned by the Colonial History Data Set, general formation occurs where the nation state was formed through other political entities that have “no direct analog in the Correlates of War interstate system.” For example, England is listed as becoming independent in 1066 AD and without forming from existing political entity in the Correlates of War database. Second, decolonization broadly refers to instances where the nation state was previously governed by a foreign power, including as a colony, protectorate or part of an empire. Decolonization is the most common type of state birth in both the ICOW History Dataset and the restricted sample. 60% of the total countries in the COW state system are identi ed as emerging through decolonization, with 80% of new countries between 1950 and 2010 emerging from colonial status. Third, secession refers to situations where “the entity was part of another state before achieving independence, with the original state surviving in reduced form... ([i.e.] Eritrea leaving Ethiopia...” Finally, the Colonial History Data Set designates to what extent states may become independent through partition, as in the case of the two states of the former Czechoslovakia.
Controls
I include controls in both the cross-section and cross-section time-series analysis to account for other factors that likely in uence the relationship between state formation processes and security institutions. For the cross-section analysis, this includes the former colonial power’s security organizations. I generate a measure for the colonial power’s security arrangements based on whether the former colonial power had a militia or gendarmerie using my security force data. In addition, post-independence con ict and country-speci c characteristics could drive the creation of dierent types of paramilitaries. I include measures for post-independence civil con ict, post-independence interstate con ict, GDP, population, executive constraints, and total area as controls. 31Independence/decolonization is considered violent if it “occurs through, or as the result of, large-scale armed violence” (Hensel 2014, pg. 4). This does not include political violence that occurs immediately following independence.
65 Civil con ict could further prevent the central government from consolidating security or discourage the government from relying on an army raised from the broader population. I incorporate measures for post-independence civil con ict from the UCDP Monadic Con ict Onset and Incidence Dataset (UCDP), which identi es the number of civil con icts a central government participated in for a given year.32 As shown in the European model, interstate con ict could drive governments to consolidate security to better engage with external rivals. Measures of post-independence interstate con ict include the number of interstate con icts the state was annually engaged in, de ned as the use of force or war as provided from the Militarized Interstate Dispute Dataset.33 Along with the security environment, country-speci c characteristics may in uence the state’s arrangement of security. Countries with larger populations or larger areas may tend to have local security arrangements, while wealthier countries may have the resources to establish nationalized and consolidated security. Therefore, I use the estimated measure of total population and gross domestic product per capita, logged from Kristian Gleditsch’s expanded GDP data and based on Penn World Table values.34 Finally, new states with greater executive constraints may rely more on central armed forces (Carey, Colaresi and Mitchell 2015). I include a variable on executive constraints to account for the in uence of democratic institutions. The measure for executive constraints is from the Polity IV regime characteristics data (Marshall and Jaggers 2005). Below, I present the results of empirical trends for state formation and security arrangements on a cross-section of countries from 1950 through 2010. Then, I use a restricted timeframe of countries formed from 1970 through 2010−the period for which I have annual data−to examine how long after state birth security institutions are formed. 32Melander, Erik, Therese Pettersson & Lotta Themner. “Organized violence, 1989-2015.” Journal of Peace Research 53(5). 2016. 33Palmer, Glenn, Vito D’Orazio, Michael Kenwick, and Matthew Lane. “The MID4 Data Set: Procedures, Coding Rules, and Description.” Conflict Management and Peace Science. 2015. 34Gleditsch, Kristian S. “Expanded Trade and GDP Data,” Journal of Conflict Resolution 46. 2002: p. 712-24.
66 3.5 empirical findings
Overview of data
I examine the emergence of security organizations across 110 states that achieved independence between 1950 and 2010. The majority of these countries emerged in Sub-Saharan Africa (48%), followed by Asia (20%), Central and South America (12%), Eurasia (10%), and−last−the Middle East and North Africa (10%). 16 states generated both militias and gendarmeries, with 27 states only generating militias and 16 states only generating gendarmeries. Among these newly independent countries, 21 states (19%) achieved independence through violent means (with the majority achieving independence through decolonization wars). Only six states emerged after violent secessionist movements: Bangladesh, Bosnia and Herzegovina, Croatia, Eritrea, Kosovo and Slovenia. Table 3.1 provides a complete description of the countries that achieved independence through violent means, including their subsequent security organizations, their year of independence and their former colonial or imperial power. Most countries after World War II achieved independence through non-violent means (82%), likely due to the collapse or disintegration of colonial or imperial rules. Table 3.5 in the Appendix provides a description of these countries that achieved independence through peaceful processes. Among them, the overwhelming majority (72 countries) emerged through non-violent decolonization and only 16% (14 countries) achieved independence through non-violent secession. In general, states achieving independence through secession are post-Soviet cases: Azerbaijan, Belarus, Georgia, Kazakhstan, Kyrgyzstan, Tajikistan, Turkmenistan, Ukraine and Uzbekistan. Only two countries achieved non-violent independence through partition (the Czech Republic and Slovakia from Czechoslovakia), and Yemen is considered as gaining independence through “formation” in 1990. In terms of colonial powers, the majority of countries separated from either the United Kingdom (41%), France (20%) or Russia (10%).
67 Table3.1: Security organizations in violently formed countries, 1950 - 2010 ∗ Country Gend. Irregular Militias Regular Militias Former Power Type Algeria Yes Yes No France decolonization Angola No No Yes Portugal decolonization Bangladesh Yes No Yes Pakistan secession Bosnia & Herz. No No Yes Yugoslavia secession Cambodia No No Yes France decolonization Cameroon Yes No No France decolonization Croatia No No Yes Yugoslavia secession Cyprus No No No United Kingdom decolonization Dem. Rep. Congo Yes No Yes Belgium decolonization East Timor No No No Indonesia decolonization Eritrea No No No Ethiopia secession Kenya No No No United Kingdom decolonization Kosovo No No No Yugoslavia secession Laos No No Yes France decolonization Mozambique No Yes No Portugal decolonization Namibia No No No South Africa decolonization Republic of Vietnam No No No France decolonization Rwanda Yes No Yes Belgium decolonization Slovenia No No No Yugoslavia secession Dem.Rep.Vietnam No No Yes France decolonization Yemen People’s Rep. No No No United Kingdom decolonization ∗ Note: Pr. Militias refer to private or unregular militias.
68 Figure 3.1: Security organizations in new states, according to non-violent and violent independence (1950-2010)
Trends in independence and internal security
Figure 3.1 shows the presence of security organizations within newly independent states according to violent and non-violent independence. Figure 3.2 shows the presence of security organizations within newly independent states according to geographic region. Regular local militias emerge more commonly through violent independent processes compared to centralized militaries and gendarmeries. With the exception of Sub-Saharan Africa, local militias are also more common than gendarmeries in new states.35 Slightly less than half of the countries that achieved independence through violence had regular local militias. In contrast, only a quarter of the countries that emerged from violent independence had gendarmeries. This trend tentatively supports the claim that violent revolutions and rivalries against stronger powers are associated with the emergence of local internal forces. In contrast to the trend in local militias, gendarmeries appear closely associated with colonial legacies and non-violent independence. Among the 31 countries in the sample that generated gendarmeries, 26 developed gendarmeries after non-violent independence movements. The majority of cases occurred 35Gendarmeries are present in 55% of newly independent countries within Sub-Saharan Africa, whereas local militias are present in only 45% of newly independent countries in Sub-Saharan Africa. This is likely due to the greater number of colonial powers within Sub-Saharan Africa that had gendarmeries, including France, Belgium and Portugal.
69 Figure 3.2: Security organizations in new states, according to geographic region (1950-2010)
in situations where the former colonial power had a gendarmerie.36 Even in the few cases where gendarmeries formed after violent independence, the former colonial powers−Belgium, France, and Pakistan−all had gendarmeries.37 Overall, this suggests that the presence of gendarmeries among former colonial powers is in uential. Particularly in the absence of con ict, countries adopt centralized security arrangements of their former powers. Unsurprisingly, the data suggest that newly independent states most commonly create militaries. Ninety among the 110 newly independent states generated militaries. While states develop militaries in both non-violent and violent scenarios, it is worth noting that every incidence of violent independence is associated with a new state having a military. The 20 cases of states without militaries are exclusively from non-violent decolonizations and tend to be small island states.38 The exception to this rule is Swaziland. Though not being an island, it only has a domestic defensive force−the Umbutfo Swaziland Defense Force−that is nominally responsible for external security but really provides protection for the royal family and engages in domestic activities.39 This is not to argue that these militaries are eective, entirely loyal to the central government or refrain from human rights abuses. This trend simply suggests that 36Former colonial powers with gendarmeries include Belgium, France, Portugal, Russia and Spain. 37These ve cases are Algeria, Bangladesh, Cameroon, Democratic Republic of Congo and Rwanda. 38Countries that lack a military consist of Comoros, Dominica, Fed. States of Micronesia, Kiribati, Maldive Islands, Marshall Islands, Mauritius, Nauru, Palau, Samoa, Sao Tome and Principe, Solomon Islands, Swaziland, St. Kitts and Nevis, St. Lucia, St. Vincent and Grenadines, Tonga,Tuvalu, Vanuatu, and Zanzibar. 39United States Department of State Human Rights Report. Swaziland. 2015.
70 militaries remain a core organization within newly formed states. In terms of evaluating the primary hypothesis−that violent independence after World War II encourages localization−these patterns so far suggest a tentative relationship between con ict and subsequent internal security in newly formed states. As expected by the literature on decolonization, wars of independence against colonial powers are associated with local security. This applies to both secessionist and decolonization con icts. While militaries remain common across new states, the proliferation of local security is more frequent after wars of independence. In addition, local militias are more common than gendarmeries among new states, and this trend applies across dierent regions of the world. These ndings run contrary to expectations about the rising trend in gendarmeries (Lutterbeck 2004) and claims that local security arrangement are limited to certain geographic areas or only failing states (Francis 2005). While gendarmeries may be more common among established countries engaged in low-intensity civil con ict, local security is more frequent in newly independent states. Centralized internal security does appear when the former colonial power had a similar organization. This suggests that security arrangements are not created in isolation but also emerge from the varying environments and challenges states face. In addition, dierent internal security forces are not created in isolation from one another; multiple organizations can be created at once, and underlying processes may drive the simultaneous emergence of these organizations. Acknowledging the limited number of countries provided in the data, I brie y turn to more rigorous tests of these dynamics, including regression and survival analysis.
Econometric Analysis: 1950 through 2010
First, I test the relationship between violent independent struggles and subsequent state organization of security forces across the 110 countries gaining independence between 1950 and 2010 using a multivariate probit, where the unit of analysis is the country. I begin with an analysis of these 110 countries from 1950 through 2010, since this covers the major break up of colonial empires after World War II. Second, as an additional series of tests, I use a restricted timeline of countries formed between 1970 and 2010−the time period covered by both annual state formation data and annual security data−to examine when security
71 organizations are created. For the rst speci cation, I rely on the multivariate probit that allows for the estimation of multiple dichotomous variables where errors are likely correlated across equations.40 The multivariate probit is similar to a seemingly unrelated regression but permits correlated binary outcome variables to be jointly estimated. While primarily investigating the role of violence on internal security arrangements, I anticipate that the presence of one type of internal security force is associated with the presence (or absence) of another type of internal force. If the rst hypothesis holds, violence-creates-localization, then I anticipate violent independence to be positively associated with local militias and negatively associated with gendarmeries. Table 3.2 summarizes the results of this rst regression model with country-level data. In the model, the correlation coe cients ρ2,1 and ρ3,2 for the cross-equation error terms are statistically signi cant, suggesting that the multivariate probit regression is appropriate over separately estimated equations. Results with separate univariate probit equations are also provided in Table 3.3, which displays substantively similar results. Figure 3.1 displays the predicted probabilities from the rst regression model. Acknowledging the small sample size, the results suggest that violent independence is positively and signi cantly associated with regular local militias. In contrast, the negative and weakly signi cant coe cient estimate for gendarmeries suggest that violent independence has a negative eect on centralizing internal security. The predictive margins provide even sharper evidence of the role violence has on subsequent internal security arrangements. If states emerge through non-violent independence, then the probability of forming a regular local militia is 20%, whereas the predicted probability of forming a regular local militia after violent independence is over 40%. This implies that violent decolonization or violent secession doubles the chances of states having local militias after independence. In contrast, states have only an 18% probability of forming gendarmeries after violent independence but a 28% probability of forming them after non-violent independence. This provides support for the rst hypothesis that violent independence−speci cally violent contestation against a stronger colonial power−is associated 40The model was unable to converge when including all four equations for security organizations. Since I am interested in the emergence of internal security forces (and since there is little variation in the presence of militaries), militaries were excluded from this regression analysis.
72 Table3.2: Multivariate probit: internal security organizations in newly formed countries
(1) (2) (3) VARIABLES Regular Militia Irregular Militia Gendarmerie
Violent Independence 0.658** -0.120 -0.625* (0.326) (0.427) (0.344) Decolonization -0.295 0.517 (0.317) (0.397) Colonial power w/ 0.755 0.394 militia (0.711) (0.574) Colonial power w/ 1.474*** Gendarmerie (0.274) Constant -0.662** -1.637*** -1.336*** (0.298) (0.382) (0.219)
ρ2,1 -0.0997 (0.223) ρ3,1 0.349** (0.176) ρ3,2 0.686*** (0.230) Observations 108 108 108 Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 Likelihood ratio test: χ2(3) = 6.42003; Prob > χ2 = 0.0929
with local internal security.
This examination of the full cross-section of countries strongly suggests that dynamics of state formation in uence the subsequent organization of internal security. Yet states may also form (or eliminate) organizations based on the subsequent security environment (including the relative strength of internal or external rivals) and other country-speci c factors. As such, I brie y turn to countries created between 1970 and 2010 to identify how long after initial state creation these internal security organizations emerge and how post-independence dynamics in uence their presence and relative size.
73 Figure 3.3: Predicted probabilities, security organizations in newly formed countries (1950-2010)
Table3.3: Univariate probit: security organizations in newly formed countries
(1) (2) (3) VARIABLES Regular Militia Irregular Militia Gendarmerie
Violent Independence 0.668** -0.108 -0.475 (0.323) (0.423) (0.356) Decolonization -0.194 0.747** (0.317) (0.368) Colonial power w/ institution 0.716 0.595 1.586*** (0.697) (0.678) (0.297) Constant -0.745** -1.833*** (0.292) (0.336) Observations 109 109 108 Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
74 Econometric Analysis: Security onset 1970 through 2010
While the post-World War II era was a major period of decolonization, further examination of modern state formation provides insight into the legacies of violence while accounting for subsequent security challenges. More immediate factors may in uence the security apparatus, such as internal or external violent challenges and years spent in internal or external con ict.41 To examine the in uence of state formation and more contemporary violence, I turn to an analysis of countries formed between 1970 and 2010. I use two dierent speci cations of the security apparatus: survival analysis and a logistic regression model of security organization onset. Results of the Cox proportional hazard model are provided in Table 3.4 and results of the logistic regressions for security organization onset are provided in Table 3.9. Estimated results from these two model speci cations provide substantively similar support that violent independence, along with the security organizations of colonial powers, in uence a state’s subsequent security arrangements. I examine factors associated with the onset of three dierent types of security organizations: militaries, regular local militias and gendarmeries.42 I estimate a Cox proportional hazard model since it does not rely on assumptions about the shape and nature of the baseline hazard rate. Robust standard errors account for heteroscedasticity. I include variables for violent independence and whether the former colonial power had a militia or gendarmerie. Since it may not be the immediate conditions at independence that in uence the emergence of dierent security organizations, I estimate a second set of models that include controls for post-independence con ict, in addition to the immediate post-independence population and economy. Con ict variables consist of the number of years (prior to formation of the security organizations) that the country spent in an interstate or intrastate con ict.
41While this form of analysis may introduce multicollinearity, potential post-treatment bias is less of a concern since I am interested in the direct eect between violent independence and internal security rather than the total eect of violent independence on internal security (an eect that is likely attenuated by the introduction of variables that occur after state independence). See Gelman, Andrew and Jennifer Hill. Data Analysis Using Regression and Multilevel/Hierarchical Models Cambridge University Press. 2007: p. 188-192. 42While inclusion of irregular militias would have been preferred, there were too few instances of irregular militias in the restricted timeline.
75 Table3.4: Cox proportional hazard, security organization onset (1970 - 2010)
(1) (2) (3) (4) (5) (6) VARIABLES Gendarmerie Gendarmerie Militia Militia Military Military
Violent Indep. 0.346 -0.724 1.361* 0.558 0.950*** -0.330 (1.135) (2.534) (0.769) (0.898) (0.356) (0.573) Former Power 2.231** 1.821* w/ gendarmerie (1.012) (0.998) Former Power w/ 1.302* 1.742 militia (0.696) (1.255)
Interstate con ict -0.375*** -0.135 -0.171 (0.116) (0.132) (0.210) Civil con ict -0.458 0.500* -0.00362 (0.448) (0.273) (0.289) GDP_ln -1.049* 1.102*** 0.487*** (0.548) (0.403) (0.104) Population_ln 2.075** -1.037* 0.222* (0.952) (0.639) (0.121) Observations 55 51 54 50 50 46 Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
76 Figure 3.4: Kaplan-Meier survival estimates: Gendarmerie creation
Figure 3.5: Kaplan-Meier survival estimates: Militia creation
Based on the estimates shown in Table 3.4 and the Kaplan-Meier survival estimates (Figure 3.4, Figure 3.5 and Figure 3.6), the positively signed and statistically signi cant coe cient estimates for both regular local militia and military imply that the risk of these security organizations forming increases among states gaining independence through wars of liberation. The coe cient estimate suggests that states formed through violent means are almost four times more likely to generate a militia compared to states that form through non-violent means.43 Similarly, states formed through violent means are over twice as likely to form militaries compared to states that achieve independence through non-violent means. The coe cient estimate for militia onset also remains positive in the presence of controls, though is not statistically signi cant. This is likely due to the relatively few number of cases in the 43For the coe cient estimate 1.36, the corresponding hazard ratio is 3.9005.
77 Figure 3.6: Kaplan-Meier survival estimates: Military creation
data. For gendarmeries, the coe cient estimate is neither consistently signed nor statistically signi cant. While countries established gendarmeries more commonly after peaceful transitions, there is not enough evidence to suggest whether violence during independence is related to the absence of these centralized security organization. Turning to adoption of security forces, there is strong evidence to suggest that gendarmeries are inherited from former colonial powers and weaker evidence to suggest that militia-style organizations are similarly inherited. Even in the more restrictive model the hazard ratio suggests that newly established states are over six times more likely to establish a gendarmerie when the former power had a gendarmerie. It is not necessarily just violence that contributes to a state’s organization of force but the organization of force within former colonial or imperial powers. Finally, the results also suggest that violence following independence in uences the creation (or absence) of internal security forces. First, the presence of interstate con ict appears to be negatively associated with the emergence of gendarmeries. The estimated coe cient for fatal militarized interstate disputes (subsequent years engaged in interstate disputes) has a corresponding hazard ratio of 0.687, which suggests a 31% decrease in the likelihood of a gendarmerie forming. Among the 36 countries that experienced interstate disputes after state formation, only two generated a gendarmerie (Moldova and Ukraine). As gendarmeries are relatively centralized, this nding goes against the traditional wisdom that interstate con ict acts as centralizing mechanism for the state. It could be that interstate disputes
78 discourage redundant security forces or parallel security to the military. Alternatively, the role of war in the pre-modern era may have had a unique eect on security arrangements compared to the post-World War II era, where large-scale conventional interstate war is not as common as “internationalized” intrastate con ict−civil con icts where at least one or more states provide troops to at least one of the warring sides.44 In the modern international system with relatively established state boundaries, central governments more often confront domestic rivals (who happen to have external support) than state peers. This nding supports my theory that the type of rivalries new states confront and the international system these rivalries occur within in uence the organization of security. As a nal test, I estimate a cross-section time-series analysis of the countries that formed between 1970 and 2010. Results of the logistic regressions are provided in Table 3.9 and average marginal eects are provided in Figure 3.10. I use a cross-section time-series since it might be that post-independence dynamics contribute to the emergence of certain security organizations. Consistent with the trend analysis and estimates of the survival analysis above, wars of liberation are positively associated with the establishment of local militias and negatively associated with the establishment of gendarmeries. The eects are consistent with previous models, though the standard errors increase when a set of controls are included that account for population, wealth, area of the country, and level of democracy. As the point estimates are in the same direction and consistent with previous models, the increase in standard errors is likely a result of inherent limitations of the data (namely the limited observation of countries and security organizations) rather than strong evidence against the theory of violence furthering localization.
3.6 conclusion
How do rivalries during state formation in uence the subsequent arrangement of domestic security? To answer this question, this chapter explored two dierent historical models of state formation and domestic security, and then proposed an alternative theory to unify them. The European experience suggests that engagement with interstate rivals under an anarchic system leads national governments to 44Pettersson, Therese, and Peter Wallensteen. “Armed con icts, 1946-2014.” Journal of Peace Research 52(4). 2015: 536-550.
79 consolidate an otherwise heterogeneous combination of security forces. These forces are consolidated from lower level polities under a national administration, allowing the central government to develop a professional national force. While this arrangement of forces requires greater funds and the elimination or acquiescence of lower-level elites, early nation states bene ted from having control over security throughout their territory. Centrally controlled security outperformed the local armed forces of rivals and allowed early nation states to eliminate coercive threats. Yet the experience of state formation after modern wars of liberation suggest that violent internal con ict and foreign support contribute to a localized security apparatus. Unlike many WesternEuropean countries, recently decolonized states lack the incentives and resources to establish centrally controlled internal security. Operating within a more stable international order, these decolonized states have greater access to external aid and face fewer interstate rivals compared to early modern counterparts. Finally, I argued that post-revolutionary countries in the modern era face internal challenges to security consolidation, having to transition former insurgents into a coherent national force. These new states can also lack the incentives to consolidate and centralize security, as they have fewer opportunities for territorial expansion, and they may rely on external aid rather than their own resources. Thus, both the structure of the international system and violent con ict (internal and external) in uence the constraints and preferences of national governments as they organize domestic security. To determine the lasting in uence of violence, I explored the relationship between state formation and the subsequent organization of the security apparatus among newly formed countries after World War II. I tested whether modern violent independence processes resulted in more local security organizations like militias. The empirical ndings suggest that modern wars of independence are associated with local security, whereas colonial legacies−like gendarmeries−are associated with centrally controlled security. Both violent con ict against colonial forces and colonial legacies shape the subsequent security arrangements of the state. This is not to say that local security is only an attribute of post-revolutionary states in the twentieth century. Rather, this trend re ects a continuation of states localizing security in response to stronger rivals and bargaining with local political units. Even developed countries like the United States have a great
80 degree of devolution and localization of security forces, as shown through its highly localized municipal police, private prisons and private security guards. These arrangements represent the di culties in consolidating internal force in the presence of relatively powerful sub-national political units and strategic bene ts of localized security. Local security develops within states composed of relatively strong local political units, weary of central armed forces and those engaging in irregular warfare. Whether medieval England, early modern Poland, or the contemporary United States and Iraq, local security forces emerge from similar challenges and incentives. The origin of the state is necessary to understand its current coercive apparatus and practices.
81 appendix
82 Table3.5: Non-violent independence, 1950-2010
Country Gend Pr. Militias State Militias Ind From Type Antigua and Barbuda No No No United Kingdom decolonization Armenia No No No Russia secession Azerbaijan No No Yes Russia secession Bahamas No No No United Kingdom decolonization Bahrain No No No United Kingdom decolonization Barbados No No No United Kingdom decolonization Belarus Yes No No Russia secession Belize No No No United Kingdom decolonization Benin (Dahomey) Yes No Yes France decolonization Botswana No No No United Kingdom decolonization Brunei No No Yes United Kingdom decolonization Burkina Faso (Upper Volta) Yes No Yes France decolonization Burundi Yes Yes No Belgium decolonization Cape Verde No No No Portugal decolonization Central African Republic Yes No Yes France decolonization Chad Yes Yes No France decolonization Comoros No No No France decolonization Congo (Brazzaville) Yes Yes No France decolonization Cote d’Ivoire (Ivory Coast) Yes Yes No France decolonization Czech Republic No No No Czechoslovakia partition Djibouti Yes No No France decolonization Dominica No No No United Kingdom decolonization Equatorial Guinea Yes No No Spain decolonization Fed. States of Micronesia No No No United States of America decolonization Fiji No No No United Kingdom decolonization Gabon Yes No No France decolonization Gambia Yes No No United Kingdom decolonization Georgia No No No Russia secession Ghana No Yes No United Kingdom decolonization Grenada No No No United Kingdom decolonization Guinea Yes No Yes France decolonization Guinea-Bissau Yes No No Portugal decolonization Guyana No No Yes United Kingdom decolonization Jamaica No No No United Kingdom decolonization Kazakhstan No No No Russia secession Kiribati No No No United Kingdom decolonization Kuwait No No No United Kingdom decolonization Kyrgyzstan No No No Russia secession Lesotho No No No United Kingdom decolonization Libya No No Yes United Kingdom decolonization
83 Table3.6: Non-violent independence, 1950-2010 (continued)
Country Gend Pr. Militias State Militias Ind From Type Macedonia No No No Yugoslavia secession Madagascar (Malagasy Rep.) Yes No No France decolonization Malawi No No No United Kingdom decolonization Malaysia Yes No Yes United Kingdom decolonization Maldive Islands No No No United Kingdom decolonization Mali Yes Yes Yes France decolonization Malta No No No United Kingdom decolonization Marshall Islands No No No United States of America decolonization Mauritania Yes No No France decolonization Mauritius Yes No No United Kingdom decolonization Moldova Yes No No Russia secession Montenegro No No No Yugoslavia secession Nauru No No No Australia decolonization Niger Yes No No France decolonization Nigeria No No No United Kingdom decolonization Palau No No No United States of America decolonization Papua New Guinea No No No Australia decolonization Qatar No No No United Kingdom decolonization Samoa (Western Samoa) No No No New Zealand decolonization Sao Tomeand Principe No No No Portugal decolonization Senegal Yes No No Mali decolonization Seychelles No No Yes United Kingdom decolonization Sierra Leone No No No United Kingdom decolonization Singapore No No Yes Malaysia secession Slovakia No No Yes Czechoslovakia partition Solomon Islands No No No United Kingdom decolonization Somalia No Yes No Italy decolonization St. Kitts and Nevis No No No United Kingdom decolonization St. Lucia No No No United Kingdom decolonization St. Vincent and Grenadines No No No United Kingdom decolonization Sudan Yes Yes No United Kingdom decolonization Suriname No No Yes Netherlands decolonization Swaziland No No No United Kingdom decolonization Tajikistan No No No Russia secession Tanzania(Tanganyika) No Yes Yes United Kingdom decolonization Togo Yes No No France decolonization Tonga No No No United Kingdom decolonization Trinidad and Tobago No No No United Kingdom decolonization Turkmenistan No No No Russia secession Tuvalu No No United Kingdom decolonization Uganda No No Yes United Kingdom decolonization Ukraine Yes No Yes Russia secession United Arab Emirates No No No United Kingdom decolonization Uzbekistan No No No Russia secession Vanuatu No No No France decolonization Yemen No Yes No Yemen Arab Republic formation Zambia No No No United Kingdom decolonization Zanzibar No No No United Kingdom decolonization Zimbabwe (Rhodesia) Yes Yes No United Kingdom decolonization
84 Table3.7: State regular militias in newly formed countries
Independence Type New State Non-violent Violent No state militia 71 12 (65%) (21%) State Militia 17 9 (16%) (8%)
Table3.8: Gendarmerie in newly formed countries
Independence Type New State Non-violent Violent No gendarmerie 61 16 (56%) (15%) Gendarmerie 26 5 (24%) (5%)
Figure 3.9: Militia creation
85 Table3.9: Logistic regression, security organization onset
(1) (2) (3) (4) (5) (6) Variables Gend. Gend. Militia Militia Military Military
Violent independence -0.830 -2.981* 2.357** 2.687*** 1.871*** -0.977 (1.719) (1.605) (0.927) (0.837) (0.687) (0.824) Peace Years -0.0924 -0.0280 0.00140 -0.00527 -0.191*** -0.0405 (0.0568) (0.0820) (0.0306) (0.0398) (0.0526) (0.0318) Violence∗Peace Years 0.111 0.0528 -0.639* -0.682* 0.188*** 0.0293 (0.105) (0.131) (0.377) (0.375) (0.0664) (0.0758) Civil con icts 0.108 0.0901 -0.382 -0.520 1.931** 0.407 (0.989) (0.812) (0.758) (1.052) (0.770) (0.774) Interstate disputes -0.579 -0.610 -0.183 -0.0239 -0.0971 0.669 (0.813) (0.745) (0.262) (0.339) (0.445) (0.483) Gendarmerie - 1.118 -0.0964 Former Power (0.691) (0.681) Militia - 1.203 1.395 Former Power (1.342) (1.824) Decolonization -0.310 0.779 (1.261) (1.685) GDP_ln -0.903* 0.674** 0.656*** (0.519) (0.295) (0.156) Area_ln -0.582 -0.375* 0.0540 (0.390) (0.196) (0.133) Pop_ln 2.457** 0.0707 0.586** (1.009) (0.550) (0.299) Executive Constraints 0.0300 -0.0450 -0.106 (0.169) (0.132) (0.124) Constant -4.858*** -1.417 -5.009*** -3.320 -1.743*** -11.58*** (0.830) (4.575) (1.333) (3.934) (0.349) (2.391) Observations 1,228 1,183 1,245 1,192 578 473 Robust standard errors in parentheses, clustered on country *** p<0.01, ** p<0.05, * p<0.1
86 Figure 3.10: Average marginal eects of violent independence on security onset
87 Chapter 4
Dancing in the dark in civil con ict: The state’s information environment and internal
security
4.1 introduction
This gun’s for hire Even if we’re just dancing in the dark.
- Bruce Springsteen (1984)
Central governments engaged in civil con ict have already lost the monopoly on the use of force, but some governments further choose to employ local and irregular actors when they have centrally controlled and regular forces available. The Turkish government relied in part on pro-Turkey Kurdish militias in their counterinsurgency campaign against Kurdish separatists, though they also deployed the military, gendarmerie and police (Biberman 2016; Kalyvas 2008). After the fall of President Suharto in Indonesia, the Indonesian military itself loosely sponsored and allowed an array of irregular local militias (Ahram 2011). During the Colombian civil con ict, the national government used regular armed forces, the National Police, rural militia and right-wing paramilitary forces to combat guerrillas (Dugas 2012). Traditionaltheories on state projection of force fail to explain this behavior: why states develop seemingly less capable security arrangements even when they have more centrally controlled regular forces available. Conventional wisdom suggests that governments turn to local irregular forces when they lack the capacity to create or expand regular armed troops (Hills 1997). Yet that cannot explain situations where central governments turn to local militias even when deploying the military and centrally controlled
88 national police. The potentially abusive and destabilizing behavior of local militias makes state delegation to them even more puzzling (Alvarez 2006; Koren 2015). Recent examples of militias in Iraq, Afghanistan and Libya all suggest that governments have di culty controlling and demobilizing localized forces once created.1 When mobilizing local irregular forces, central governments risk deploying troops that are not easily coordinated, monitored or controlled in the long term (Mitchell 2004). To understand why national governments at times allow and even support these security arrangements, this chapter explores the information challenges governments confront in civil con ict and the information advantages of dierent internal security forces. Throughout this chapter, I use the term “internal security force” to refer to any armed force that is (1) acting on behalf of the government, (2) within the territorial jurisdiction of the state, and (3) distinct from the military. Combining insights from studies on both civil-military relations and counterinsurgencies,2 I argue that variation in a con ict’s information requirements−establishing local area knowledge, including accessing social networks of rebel support, and collecting information through technical means and monitoring one’s own security forces−helps explain why central governments at times prefer local irregular armed forces. Organizations like local militias can be better positioned to gather information on the local environment and communities, leveraging local knowledge, language skills and access to social networks that regular central forces lack (Lyall 2009). Local knowledge can be exceptionally important during counterinsurgency campaigns. As noted by the United States Army Field Manual 3-24 (8-14): “During counterinsurgency operations, a great deal of intelligence is based on information gathered from people.” On the other hand, national governments often have greater technical capabilities for information gathering (such as signals intelligence) with regular central forces, which allow leaders to more eectively target national security operations, leverage specialized forces, receive technical information and monitor their own force’s activity (Feaver 2009; Lioe 2011). The national government’s competing needs for information−on the population, over enemy operations and on its armed forces−and type of information (including human or technical means of gathering it) shape its subsequent security arrangements. I test this theory with a panel of con ict-exposed countries between 1989 through 2009, using my 1Cigar(2014); Goodhand and Hakimi(2014 a); Kirkpatrick(2012). 2i.e. Biberman(2016); Feaver(2009); Lyall(2010); Staniland(2012).
89 new data on internal security forces. I nd that irregular local forces grow when central governments confront challenges gathering information on the local environment and local population. Contrary to previous studies that view locally raised and locally controlled forces as a response to challenges in the country’s periphery (Hills 1997), this chapter nds that con icts waged farther away−on a state’s periphery−are actually associated with the expansion of centrally controlled forces like gendarmeries that have more specialized and mobile capabilities. These results suggest that variation in the information environment and challenges in the geographic and human terrain in uence the subsequent allocation of security resources. This occurs above and beyond structural factors traditionally highlighted in the literature. Section two discusses the conventional wisdom on the state’s organization of security in civil con ict. Section three presents a new theory for understanding the government’s distribution of forces as a strategic response to informational challenges. It explores how internal armed forces address dierent information requirements. My theory argues that national governments engaged in civil con ict and faced with severe resource constraints strategically organize and mobilize internal security to mitigate information challenges posed by unique local political, social, economic and physical environments. Section four presents hypotheses based on this theory of security allocation and explains the empirical strategy for testing them. Finally, section ve presents cross-national empirical trends and the results of regressions, which strongly support the in uence of information challenges and related issues of population-centric information operations.3 This chapter concludes with a discussion on these short-term strategic challenges of civil con ict compel governments to rely on “out-of-control” forces. Just as important, challenges requiring greater coordination, mobility and technical intelligence incentivize the national government to invest in more centrally controlled security forces. 3Per earlier chapters, population-centric operations refer to tactics in warfare−speci cally in counterinsurgencies−where the population and not the enemy is the focus. Gentile, Gian P. A Strategy of Tactics: Population-centric COIN and the Army. Army War College. 2009.
90 4.2 the information environment in civil conflict
Governments commonly turn to local irregular forces to implement national security policies within civil con icts, even in the presence of regular armed forces. India employs Muslim special operations units to ght Pro-Pakistani organizations in Kashmir alongside the Indian Army and the Central Reserve Police Force (HRW 1996; Staniland 2012). The International Security Assistance Force (ISAF) in Afghanistan has raised and trained the Afghan Local Police in conjunction with the Afghan National Police and Afghan Army (Goodhand and Hakimi 2014a). The current government of Iraq depends on militias to combat insurgents.4 Recent works have highlighted the presence of both regular and irregular security organizations in civil con ict.5 Yetit remains unclear why central governments engaged in con ict allocate force dierently, including why some governments rely heavily on local and often amateur forces. Despite policy practitioners examining the role of irregular local forces,6 political science research to date focuses overwhelmingly on the role of the military rather than alternative security forces (Davenport 2007c; Hartzell and Hoddie 2003; Levine and Smith 2000; Vasquez 2005). Works that do examine the state’s use of internal armed forces do not distinguish between important organizational dierences among these groups, which has led to considerable disagreement over the nature and consequence of particular paramilitary forces like local militias (Carey, Mitchell and Lowe 2013; Koren 2015; Stanton 2015). Finally, many studies simply consider the presence of internal security organizations rather than the relative resources or manpower allocated to them (Carey, Colaresi and Mitchell 2015; Clayton and Thomson 2015). These approaches conceal key variation in strategies for state projection of force and the ability of the national government to apply coercive instruments of power during civil con ict. In addition, conventional wisdom on state projection of force anticipates that weaker states or states engaged in civil con ict simply rely less on regular armed forces. Failed states are by de nition governments that lack the institutions and resources required to recruit, train and arm a national force (Francis 2005; Singer 2008). Consequently, political o cials in these states turn to alternative sources of coercion. For example, Somalia lacked a central government after the collapse of the Siad 4George, Susannah. “Breaking Badr.” Foreign Policy. 2014. 5Carey, Mitchell and Lowe(2013); Peic(2014); Pilster and Böhmelt(2011). 6i.e. Asfura-Heim(2014); Goodhand and Hakimi(2014 a); Jones(2012).
91 Barre dictatorship in 1991 and did not have even the basic bureaucracy required to develop a military.7 Foreign militaries, Local military contractors and pro-government militias took over the roles and responsibilities previously assigned to regular armed forces.8 In situations of complete state collapse, alternative organizations can act as substitutes for regular armed forces. Yet instances of complete state collapse or near-total failure of state authority remain relatively rare (Marshall, Gurr and Har 2009). Even states characterized by weaker institutions vary considerably in their coercive capacity and capability (Rotberg 2010). Why governments use dierent combinations of internal security forces or rely on dierent distributions of those forces remains unknown. Moving beyond structural explanations, research on civil-military relations and counterinsurgencies suggest that national governments confront critical information challenges in the course of con ict that inhibit battle eld eectiveness.9 Using a principal-agent paradigm, studies on civil-military relations emphasize the problems that government principals face when delegating to security forces.10 The government principal and its security agents may have divergent preferences and security agents can withhold information (Byman and Kreps 2010; Feaver 2009). Since the central government cannot perfectly observe its agents, security agents can shirk and act in their own interests rather than in the interests of the government. Security agents could themselves directly threaten the government (Belkin and Schofer 2003; Pilster and Bohmelt 2012). As observed by Feaver(2009, p. 214), “the very institution created to protect the polity is given su cient power to become a threat to the polity.” Central governments require scrutiny and knowledge of the activities of their security forces, who may be shirking in their duties or working to directly undermine the state. Also using the principal-agent paradigm, some research argues that governments use their imperfect 7Lyons, Terrence, and Ahmed Ismail Samatar. 1995. Somalia: State collapse, multilateral intervention, and strategies for political reconstruction. Brookings Institution Press. 8Singer(2008). Human Rights Watch. 2012. https://www.hrw.org/news/2012/03/28/ somalia-pro-government-militias-executing-civilians. 9By battle eld eectiveness, I mean the security organization’s ability to carry out speci c combat tactics during battles. This is similar to the conventional concept of battle eld eectiveness for militaries, whereby units can apply both basic and complex combat techniques in conventional wars to defend or acquire territory (Talmadge2015, p. 4). 10 Eisenhardt(1989, p. 58) de nes the principal-agent framework as “the ubiquitous agency relationship, in which one party (the principal) delegates work to another (the agent), who performs that work...and the problems where (a) the desires or goals of the principal and agent con ict and (b) it is di cult or expensive for the principal to verify what the agent is actually doing.”
92 knowledge over security agents to their advantage (Acemoglu, Robinson and Santos 2013; Alvarez 2006; Cunningham 2004). Government principals strategically delegate to security agents not directly accountable to them to “conveniently conceal knowledge”over repressive activities (Mitchell 2004, p. 48). Yet many of these studies only examine the use and behavior of death squads, security forces covertly used by the government for the explicit purpose of committing extrajudicial killings and other illegal activities against civilians (Campbell and Brenner 2002, p. 2). This research does not consider alternative types of internal armed forces that engage in non-abusive behavior, including civil defense activities like protecting local communities and denying rebels access to the population and needed resources (Clayton and Thomson 2015). For example, pro-government militias in the Aceh region of Indonesia secured their own villages (Barter 2013). More recent research has found that states with abusive militias likely have regular armed forces engaging in similarly brutal activities (Cohen and Nordås 2015; Stanton 2015). If both regular armed forces and militias abuse civilians, then the state may rely on militias for reasons other than deniability. Research on counterinsurgencies emphasizes a second information problem governments confront in civil con ict: the ability to gather intelligence on rebels and their bases of support. In this framework, the central government and its central armed forces lack the local area knowledge necessary to disrupt and counter enemy operations, including knowledge about the identity of rebels, rebel sanctuaries and relevant social networks (Kalyvas 2006, 2008; Lyall 2010; Salehyan 2010). In both political science and policy research, this type of intelligence is viewed as critical in the conduct of successful operations against insurgents. As stated in the United States Army Field Manual 3-24 (2006):
“Without good intelligence, counterinsurgents are like blind boxers wasting energy ailing at unseen opponents and perhaps causing unintended harm. With good intelligence, counterinsurgents are like surgeons cutting out cancerous tissue while keeping other vital organs intact. Eective operations are shaped by timely, speci c, and reliable intelligence, gathered and analyzed at the lowest possible level and disseminated throughout the force.”
Intelligence failures and information challenges can arise whenever the national government and
93 its armed forces are separated socioculturally and/or geographically from the population rebels rely on for support and safe havens (Lyall 2009; Tollefsen and Buhaug 2015; Wucherpfennig et al. 2011). Governments may have di culty working with populations that are geographically reachable but remain socially isolated, as in the case of some ethnic minorities. In these cases, national governments may receive information on the movement of individuals or resources through technical means (such as through signals intelligence) but lack the local area knowledge and cultural understanding to determine whether these individuals and their actions are hostile.11 Even if the central government can identify noncombatants, noncombatants may be unwilling to inform national state forces on rebel activities or reluctant to denounce rebel combatants (Kalyvas 2006, p. 91). For these reasons, Civil Self-Defense Patrols in Guatemala and Home Guards in Oman both conducted continuous surveillance on behalf of the centrally controlled armed forces and regularly informed on rebel activity (Asfura-Heim 2014; Jones 2012). As in Oman, ethnic communities in autonomous territories can also hide rebels from the government’s formal reach and provide safe havens (Fearon and Laitin 2003). While central governments may have the ability to deploy and use national forces, these forces do not necessarily provide the state with the intelligence it requires to deny and disrupt rebel operations or precisely employ their resources and capabilities. Taken together, insights from civil-military relations and counterinsurgency studies imply that national governments confront substantial information problems meeting internal security challenges.12 National governments face two kinds of information and intelligence problems. First, they confront challenges gathering information on local conditions and local challengers. Second, they can have di culty monitoring and coordinating their own security forces. National governments can address these dierent challenges with dierent internal forces. Yet the extant literature leaves key dynamics unexplored. Informational challenges may occur simultaneously or separately. Information may be gathered through technical means or through 11McFate (2005, pg. 44) shows that misunderstandings of cultural cues at the operational level can lead to unintentional ring against noncombatants; United States Marines in Iraq misinterpreted hand gestures and symbols as hostile when they were actually signs of surrendering and halting. McFate, Montgomery. The military utility of understanding adversary culture. O ce of Naval Research. Arlington VA,2005. 12Rebels confront information challenges as well (i.e. Weinstein(2005)).
94 non-technical means. Governments may end up relying on the same internal forces to address a variety of dierent intelligence problems. The aforementioned research also relies on subnational and qualitative studies, failing to explore whether these dynamics are systematic across countries experiencing con ict.
4.3 internal security as a response to information challenges
The previous section suggests that national governments confront dierent types of challenges in civil con ict. In conducting operations, governments may face challenges in (1) generating local area knowledge related to rebels and rebel activities and in (2) monitoring the actions of their own security forces and developing technical means of information collection.13 Previous research has explored the importance of these problems separately, but research to date has not examined how these issues simultaneously or in combination in uence the national government’s domestic distribution of force. It does not explore whether these challenges impact the number of personnel within certain security organizations or personnel across internal forces. Central governments may have di culty coordinating their own armed forces and in gathering information on rebel activity. Armed forces better suited to address one problem may be ill-suited to address another problem. Finally, governments face resource constraints when needing to mobilize and deploy internal security personnel, further complicating their calculus.14 A state’s distribution of internal security forces, therefore, can be understood as the strategic attempt to address and balance informational requirements under limited resources.15 13As opposed to gathering human intelligence (information derived from human sources), governments may increase their capability for gathering information through intercepting electronic information and communication (i.e. signals intelligence). See Gortney, W. E. “Joint Publication 1-02: Department of Defense Dictionary of Military and Associated Terms.” Washington, DC: Joint Chiefs of Sta. 2012. 14Governments face even greater resource constraints in the context of civil con ict. 15National government can opt into security arrangements that seem neither rational nor optimal in the long term. However, even these choices−such as delegating to local militias−may be rational responses to immediate dilemmas. This project does not attempt to perfectly predict internal security arrangements, but only show that incorporation of the information environment better explains patterns in internal security arrangements.
95 Figure 4.1: Spectrum of internal security forces
Organizational attributes of internal security−the balance between regulation under the central government and local control−help explain why governments rely on dierent armed forces to address information requirements.16 As discussed in previous chapters, regular refers to the extent armed forces o cially act on behalf of the state, have state standardized training and equipment and a set organization. Regular armed forces exercise legitimate state use of force and are subordinate to a government command authority (Feaver 1999; Perlmutter and Bennett 1980). Irregular armed forces lack standardization, institutionalized political oversight and have limited accountability back to the government. The national government may legally allow irregular armed organizations to exist but not recognize them as an o cial arm of the state (Carey, Mitchell and Lowe 2013). For example, the Colombian government formally authorized the creation of defense forces and vigilante organizations in 1994 to help combat insurgents but these forces did not operate under the command authority of the central government.17 Regular armed forces include gendarmeries, national police and state militias. Irregular forces include local Local (but government aligned) militias, security guards that report directly to a leader or part of a government but not the central government as whole, and other forces that the government lacks command authority 16In terms of other ways of classifying armed forces, Finer(1975) describes security forces as ad hoc (forces employed for a nite time period or campaign) or permanent like standing armies. Avant(2005) dierentiates between corporatized or non-corporatized security actors, and Biddle(2004) examines dierences in forces across modern and non-modern militaries. 17Aviles, William. Global capitalism, democracy, and civil-military relations in Colombia. SUNY Press 2012, p. 118.
96 over. The second attribute of interest, localization, describes the degree to which armed forces operate within a speci c or narrow geographic region. Local forces conduct activities in certain geographic areas or have an explicit regional or local jurisdiction. They tend to recruit from immediate areas of operation and are managed by local authorities or local leaders (Peic 2014). Local forces tend to have lighter weapons with greater reliance on small arms and less technical capabilities or specialization (Asfura-Heim 2014). In contrast, centrally controlled forces more likely have the infrastructure, budgets and specialization required for more technical information gathering, mobility and the use of more lethal complex weapons systems. For example, Turkey’s Gendarmerie has its own intelligence wing, JITEM, which is used to gather intelligence internally against the PKK along with other extremist groups.18 Turkey’sGendarmerie has its own logistical support units, aircraft for transportation and observation, mortars, and armored personnel carriers.19 In addition, central armed forces have national jurisdiction and command authority from the national government, meaning the authorization to act from the national government or national leader and the ability to operate throughout the country. Examples of centrally controlled forces include gendarmeries and many special operation forces, who have the authority and capability to deploy to dierent parts of a country regardless of where they are permanently stationed. Local forces, in contrast, lack national jurisdiction and may have more limited capabilities. Local governments rather than central governments may also have command authority over these forces. Local forces include groups like local militias and police. Assessing internal forces in terms of regular/irregular and local/central (Figure 4.1) provides important insights into why governments choose diering ratios of internal security forces in civil con ict. Below, I discuss the costs and bene ts of internal forces across these dimensions and with respect to the information environment. I argue that di culty in monitoring and coordinating security forces, along with gathering information through more technical means, encourages national governments to expand regular and centrally controlled armed forces. These forces grant the central government greater oversight over operations while allowing for more exible deployments and technical capabilities to 18IHS Jane’s. Jane’s Sentinel Security Assessment; Turkey; Security and Foreign Forces, 2015. 19Ibid.
97 gather information over greater distances. On the other hand, information disadvantages over local area knowledge and local networks encourage the government to expand local irregular sources of security, which provide the central government with intelligence capabilities it cannot match, especially greater capability to gather information from the civilian population and better knowledge of the local environment.20 Finally, the government may allocate security resources between dierent types of internal security forces, balancing the need for greater capabilities and control with the need for local area knowledge, including knowledge over and access to local social networks. The national government strategically allocates internal security resources to mitigate distinct and potentially competing challenges.
Variation in information advantages across internal armed forces
In civil con ict, central governments confront problems related to coordinating their own forces and in dierent ways to gain knowledge about rebels. Given resource constraints that they confront, especially when engaged in civil con ict, central governments must balance resources between central and local forces to best address their internal security environment. The nature of con ict will determine the mix of forces and capabilities deployed. Below, I discuss the advantages and disadvantages associated with dierent types of internal security. I provide the rationale for why varying government distribution of security resources depend on the information environment. Governments expand centrally controlled and regular internal security when the need to coordinate or monitor security personnel is greater, along with increasing need for technical capabilities, whereas governments expand local irregular forces when the need for local area knowledge is more important.21 Local irregular forces and central regular forces mitigate distinct information challenges. Fielding an unsuitable mix of security forces degrades the central government’s ability to address other informational requirements. For example, enlisting local irregular units (like militias) to 20This is not to say that central armed forces never provide local area knowledge but only that local forces may have certain advantages in certain situations, such as within ethnic enclaves or autonomous ethnic territory. 21Regular and centrally controlled security forces may have their own intelligence units, though even these units may have limited ability to gather intelligence on the local area or limited ability to work with the local population. Turkey’s Gendarmerie has an intelligence command but still have local Kurdish Village Guards inform on PKK activity. Gurcan, Metin. “Arming civilians as a counter-terror strategy: The case of the village guard system in Turkey.” Dynamics of Asymmetric Conflict. 8(1) 2015: 1-22.
98 monitor and inform on rebel activities could leave the central government unable to prevent those militias from pursuing objectives counter to the central government’s interests.
Centrally controlled regular security. Governments strategically expand centrally controlled regular forces when they seek improved oversight over the coercive apparatus and more technical capabilities. When confronting domestic challenges, central governments may need to rapidly and reliably deploy armed forces to areas of combat far from main operating bases (Hammond 2009). To a lesser extent, governments concerned about coups can expand these forces when requiring a counterbalance to the military (Belkin and Schofer 2005). The central government can more easily monitor, move and coordinate forces that are well regulated and nationalized under a central authority or administration (Dunn 1996; Lutterbeck 2003). These internal security forces can consist of special operation units, gendarmerie and national police.22 How might standardization, specialization and central control of security improve the government’s access to information over its own security forces and the security environment? These armed forces tend to have robust command and control (C2) capabilities, institutionalized political oversight mechanisms and military-like logistical support (Avant 2005; Perlmutter and Bennett 1980). First, a hierarchical chain of command and robust C2 provide higher ranked o cers with the authority and ability to direct lower level units and receive reports on those units (Andreski 1968; Tellis 2001). Theoretically this command structure improves the ow of information in both directions; higher ranked o cers communicate orders to lower-ranked personnel, and lower-ranked personnel report back to o cers on the execution of those orders. Second, political oversight mechanisms provide the central government with oversight over security operations to induce security o cers and their units to comply with policies (Perlmutter and Bennett 1980; Regan 1994).23 Security forces report to administrative authorities, like a Ministry of Interior, and the central government has the ability to audit or investigate security operations (Feaver 22Gendarmerie refer to forces with the authority to engage in policing of the civilian population and internal security operations (Lutterbeck 2004). The National Guard of the United States represents an interesting borderline case, since it is under the dual control of the federal and state government. As a reserve force, I do not consider it in the same category of these internal security forces. See U.S. Code 12107 - Army National Guard of United States. 23Many gendarmeries even have duel reporting requirements to both a ministry of interior and ministry of defense.
99 1999). For example, regional governments in Belgium report to the central government on the security operations or potential violations of the federal police.24 Lastly, centralized and regular forces often have more technical capabilities and better logistical or mobility support to deploy personnel and provide technical information back to leaders on their unit’s status and operations of enemy combatants (Hammond 2009). Studies on military logistics similarly emphasize the importance of providing realtime or near realtime data on deployments and equipment, including the capability to collect, process, and report technical information related to enemy operations.25 Forces like gendarmeries have transportation support like armored cars and helicopters that allows for faster deployment across wider territorial ranges (Lutterbeck 2004). For example, Morocco’s Gendarmerie Royale has both air and paratrooper squadrons, which conduct counterinsurgency operations in Western Sahara.26 Similarly, the Philippines’ gendarmerie have light transport aircraft.27 Governments likely emphasize regular and centrally controlled internal security forces when needing to deploy and coordinate security, particularly over great distances, and to increase technical capabilities for gathering intelligence on the enemy. Centralized internal security forces provide the national government with C2 and logistical capabilities that local irregular forces lack. These armed forces are not necessarily available or preferred in all contexts. Regular and centrally controlled armed forces often do not have the local knowledge, language skills or access to social networks required to gather intelligence on rebels, rendering technical intelligence and information less useful if it lacks proper context.28 Regular central armed forces may have access to technical information such as signals intelligence but still lack su cient knowledge to separate enemy combatants or enemy operations from civilians and civilian events. For example, the United States conducted a drone strike on 24Belgian Federal Police have a counter-terrorism unit and special forces unit deployed in cases of terrorism and for conducting domestic reconnaissance. Perdikaris, John. Physical Security and Environmental Protection. CRC Press, 2014. 25Tripp, Robert S., Mahyar A. Amouzegar, and Ronald G. McGarvey. Sense and respond logistics: Integrating prediction, responsiveness, and control capabilities. Vol. 488. Rand Corporation, 2006. 26Cordesman, Anthony H., and Aram Nerguizian. The North African military balance: force developments in the Maghreb. Vol. 31. CSIS 2009, p. 26. 27The Military Balance (2010, p. 377-440). 28United States military doctrine emphasizes the shortcomings of raw information that lacks proper context. Marine Corps Doctrinal Publication (MCDP) 2, Intelligence, notes that “even if we are able to collect vast amounts of information, information alone does not equate to knowledge or understanding, which are ultimately the product of human cognition and judgment...Technologyby itself does not produce eective intelligence” (1997, pg. 20).
100 a Yemen wedding procession by accident in 2013, assuming it was an enemy operation.29 More generally, regular armed forces explicitly associated with the incumbent government may have di culty gathering information from civilians distrustful of the state or fearful of rebel retaliation (Lyall 2010). Forces like gendarmeries sometimes lack the jurisdiction and civilian support required to work eectively in autonomous territories. Finally, national governments may not have the resources available to establish, expand or adequately fund central armed forces. In these cases, the national government might turn to irregular and less expensive local internal security actors.
Irregular local security. Governments confronted with greater information disadvantages relative to rebels on local area knowledge expand irregular local security forces. Local organizations can better gather and assess information on the local environment, including on the identities and support networks of rebel groups, and provide social or cultural context for rebel greviances and actions.30 With members recruited from and serving in their native area, irregular local militias can leverage social networks on behalf of the state (Fearon and Laitin 2003; Kalyvas and Balcells 2010). They may also move freely in areas where regular forces have limited access (Kalyvas 2008; Peic 2014). The case of Oman and the Dhofar Rebellion provides a sharp example of the information advantages of local forces. With the assistance of British forces, the government of Oman raised militias from local tribes in Dhofar (Ladwig 2008). These militias, also known as tribal home guards or firqats, assisted the military and gendarmerie with reconnaissance and defense in Oman’s southernmost province during the Dhofar Rebellion (1965-1976), an area where Omani armed forces historically lacked experience and knowledge (Asfura-Heim 2014). As noted by Ladwig(2008, p. 73), “ firqats functioned not only as scouts and guides for SAF [the Sultan’s Armed Forces] but also as ‘home guards,’ consolidating and defending tribal areas on the Jebel after they had been swept clean of insurgents.” Firqats provided information to the state’s national forces and encouraged defection because of their established social ties with the local community.31 More recent examples also illustrate this. The International Security Assistance Force 29Friedersdorf, Conor. “The Wedding That a U.S. Drone Strike Turned Into a Funeral.” The Atlantic. 9 January, 2014. 30McFate, Montgomery. The military utility of understanding adversary culture. O ce of Naval Research. Arlington VA, 2005. 31Clegg, Will. “Irregular Forces in Counterinsurgency Warfare.” Security Challenges 5(3) 2009: p. 1-25.
101 (ISAF) created the Afghan Local Police out of local groups under the assumption that local militias−in contrast to the Afghan National Security Forces−“knew the local terrain and they can generate eective intelligence−all critical factors in counterinsurgent operations” (Goodhand and Hakimi 2014a, p. 15). Central governments use irregular local forces to collect intelligence and work with civilian populations in a way that regular central forces cannot. Yet irregular local forces generate substantial risks for the national government. The atomized and localized nature of these security organizations make them di cult for the central government to monitor, control and coordinate (Ahram 2011). The localization of security that allows for valuable information gathering also makes these forces ill-suited for engagements across distances as in conventional con ict. The American colonial experience with militias in the French and Indian War illustrates these dynamics. Colonial militias suered from a form of “myopic localism”−local recruitment strategies, loyalty to townships and intercolonial rivalries−that hindered their ability to act as a well-coordinated force (Cave 2004; Cooper 2002). Attributes that made colonial militias eective defenders of local security made them less eective at attacking conventional, dispersed rivals and more likely to engage in disputes with each other. In Afghanistan during 2010, the Afghan Local Police were similarly drawn from rival communities and at times “turned on each other instead of ghting the Taliban” (Felbab-Brown 2016, p. 270). Irregular local forces may have con icting loyalties and interests, in addition to poor mobility and poor technical capabilities, making them di cult to coordinate and less reliable to deploy. The government can choose to use these local irregular options to gain local area knowledge or to more cost-eectively project force in the short term. Yet the national government risks deploying forces who are unwilling to work with each other and who may exacerbate the security situation.
Other security arrangements. I brie y discuss two other internal security arrangements that national governments may expand during civil con ict: regular local security and irregular centralized security. These types of forces act as strategic “compromises” in addressing information problems associated with gathering information from the security environment and controlling one’s own forces.
102 Regular local security. First, governments may create or expand regular local security when needing to project force in a relatively inexpensive manner and maintain some oversight over armed forces. In this manner, both government budget constraints and information challenges in uence allocation of force. For example, the Philippines President Corazon Aquino established the Civilian Armed Force Geographic Units (CAFGUs) in the late 1980’s as a way to cheaply establish a presence and combat insurgences throughout rural villages.32 The central government raised these local militias in communities vulnerable to insurgent attacks and in rural areas that otherwise lacked central armed forces or police (Asfura-Heim 2014). Central governments may nd mobilizing locals, even untrained civilians, useful for defense or denial operations (Biberman 2016; Francis 2005). However, cost savings come at the cost of decreased control by the central government and decreased conventional capabilities. Local administrations often have some, if not all, authority over local security forces. The state-managed militias of the Soviet Union were accountable to both local districts and the central Ministry of Internal Aairs.33 Local security can also lack the specialization, heavy armaments and training of national forces (Lutterbeck 2004; The Military Balance 2010). As Ahram(2011, p. 12) notes, local armed forces tend to be “dierentiated by geography rather than specialization.” Lacking the logistical support of gendarmeries or the military, these groups are less able to conduct complicated deployments or to be redeployed from main operating bases. Regular local security provides the government with a way to project limited capabilities in certain districts but at a cost of decreased national accountability, specializaton and mobility.
Irregular central security. Finally, governments may turn to irregular centralized security forces, such as praetorian guards who protect national leaders or serve a particular faction of the government. The fear of a coup might lead governments to create or expand these forces (Quinlivan 1999). In contrast to gendarmeries or special operation forces, praetorian guards are responsible to a speci c person or speci c faction of the governmention rather than to the entire central government (Belkin and Schofer 2005). For 32Institute of Bangsamoro Studies [IBS], and the Centre for Humanitarian Dialogue [HD], Armed Violence in Mindanao: Militias and Local Armies, Geneva: HD, 2011. 33Williams, James L.; Serrins, Adele S. “Russian Militia: An Organization in Transition.” Police Studies: The International Review of Police Development 16.4 (1993): 124-128
103 example, Saddam Hussein established the Special Republican Guard for his personal protection rather than rely on potentially disloyal members of the Iraqi military (Talmadge 2015, p. 157). Leaders within central governments expand these armed forces when distrusting regular armed forces. Outside the context of personalist regimes, irregular central forces often lack institutionalized oversight or accountability, which can make them di cult for the central government to coordinate, monitor and control for combat operations. For example, the United Self-Defense Forces of Colombia (AUC) acted in collaboration with the Colombian national government but was not part of the Colombian government. The Colombian government had no institutionalized way to monitor operations of the AUC or means to eectively demobilize those units.34 Acting outside the government’s formal control and sometimes pursuing their own objectives, these well-armed roving security forces provided the central government with relatively poor oversight, potentially threatening the regime, creating political di culties and undermining operations against rebels.35
4.4 hypotheses on information challenges
The theory outlined above claims that the government will allocate internal security resources in response to speci c informational challenges in civil con ict. Drawing on the logic of information asymmetries, I argued that diering mixes of centrally and locally controlled forces provide the state with dierent information advantages. This theory generates a number of testable hypotheses on variation in the information environment within civil con ict and personnel allocated across internal security forces. First, my theory anticipates that irregular local forces expand when the national government needs to collect information on the local environment but centrally controlled forces lack jurisdiction or reliable informants within rebel support bases. Regular centrally controlled armed forces are at a distinct disadvantage when key populations are relatively isolated or hostile to the national government (Lyall 2010). State forces confront even greater di culty gathering information on rebels based in autonomous 34“The guerrilla groups in Colombia.” United Nations Regional Information Centre for Western Europe. United Nations. 2015. 35Paramilitaries generally−and militias speci cally−likely vary to the extent they are “stationary” or “roving.” A useful next step for projects on internal security would be studying the movement of these forces. For a discussion on roving versus stationary rebels in civil con ict, see Beardsley, Gleditsch and Lo(2015).
104 territories. In such environments, the central government has little to no institutional presence, and local populations can easily harbor combatants or withhold knowledge of rebel activities. Based on this rationale, I anticipate that irregular forces under local control are elded when the central government cannot easily work with or gather information from relevant local populations. This includes minorities in areas of the country where the national government has fewer personnel and resources, including insu cient infrastructure, bureaucracy or central forces.36 I test the relationship between ethnic minorities in autonomous regions and the type of security forces elded, as minorities in autonomous regions present a signi cant information challenge to governments lacking institutional links with certain domestic populations.
• Hypothesis 1: During civil con ict, irregular local security forces are positively associated with populations in autonomous regions.
My theory also anticipates that regular centrally controlled security expands when central governments need to monitor and coordinate armed forces, particularly when these forces are deployed to distant theaters of operation.37 Such deployments increase the di culty of eectively exercising command authority, including monitoring the performance of deployed forces and ensuring that deployed forces carry out instructions. Command and control capabilities and centralized political oversight increase the likelihood that security forces do as they are told and that their operations support the policies of the central government. In addition, regular central forces tend to have better logistical capabilities and technical means to gather certain types of intelligence compared to irregular local forces, meaning that there is less of a decline in logistical support (including food, fuel and ammunition) as operations move farther from main operating bases and less of a decline in the ability to intercept an 36This further distinguishes between two types of inaccessibility on the part of the national government: physical and social inaccessibility. While interrelated, physical inaccessibility refers to the physical or geographic challenges governments confront whereas social inaccessibility encompasses the ethnic or human terrain. As Tollefsen and Buhaug(2015, p. 10) describe, “Areas inhabited by people with deviating traditions and preferences regarding language, minority rights, wealth redistribution, local autonomy, etc., will be harder to penetrate by the central government. For this reason, rulers with less extensive territorial ambitions may decide to leave backward zones alone: not investing in infrastructure or bureaucratic and socioeconomic institutions, and refraining from providing costly public goods that serve no greater political purpose.” 37This corresponds with the traditional loss-of-strength gradient model, where a national government’s peak strength is at its home base and declines as military power is projected farther across distance. Boulding, Kenneth E. Conflict and defense: A general theory. New York: Harper. 1962.
105 enemy’s communications through technical means (Hammond 2009). I expect that governments expand regular centrally controlled forces when needing to deploy forces to combat areas farther away or across more di cult terrain.38 This leads to my second hypothesis:
• Hypothesis 2: During civil con ict, regular centrally controlled security forces are positively associated with conflict’s distance from the capital and mountainous terrain.
Finally, I argue that internal security forces provide the government with command and control capabilities and other advantages above and beyond that of the military but not in place of the military. Central governments are more likely to use internal security forces to supplement (rather than as a substitute for) regular armed forces. Countries with the bureaucracies and budgets to maintain a standing military have the resources and means to mobilize other forces, including security forces with explicit domestic, rather than external, jurisdiction.39 This is in contrast to traditional theories of state weakness, which anticipate the expansion of internal security forces in the absence of regular armed forces and in situations of state collapse. Based on my theory that internal security forces supplement but do not substitute regular armed forces, I should observe a positive relationship between expanding regulated internal security and the military. My nal hypothesis tests this relationship:
• Hypothesis 3: Regular internal security is positively associated with the military.
Alternative explanations
In contrast to my argument on information advantages of internal security, some theories argue that the state’s limited capacity entirely explains the government’s deployment of internal security in civil con ict (Francis 2005). These arguments suggest that central governments rely on irregular local forces such as 38As discussed in the conclusion, only central governments above a certain capacity or resource threshold likely have the ability to adopt this strategy of regulated national force. The argument does not necessarily apply to countries experiencing total state collapse. For example, Somalia’s state security forces are unable to maintain control over large parts of the country occupied by clan militias and al-Shabaab, and Somali forces lack the resources to eectively deploy, monitor and secure these areas. However, instances like contemporary Somalia (states experiencing collapse) are relatively rare. Bureau of Democracy, Human Rights and Labor. “Somalia Human Rights Report.” Country Reports on Human Rights Practices. U.S. Department of State, 2016 39The distinction between internal and external jurisdiction is discussed in greater detail in Chapter 5.
106 militias when they lack the capability to project force by other means (Butler, Gluch and Mitchell 2007; Carey, Mitchell and Lowe 2013). If arguments of state capacity better account for the distribution of security, we would anticipate that governments in less developed states rely more heavily on irregular local security, especially when needing to project force in the periphery (Herbst 2014; Tollefsen and Buhaug 2015). Less developed states may have substantial geographic barriers to identifying and rooting out rebels; they may have fewer roads and less infrastructure, making logistical support and deployment of forces harder (Francis 2005). Given these dynamics, we would expect regular centrally controlled forces to decrease when con icts are waged in the periphery of less developed states. Finally, some theories argue that the risk of coup leads central governments to create and expand internal security forces as a counterbalance to the military (Belkin and Schofer 2005) and that internal security forces expand when central governments are at a greater risk for a coup. According to theories of “coup proo ng,” internal security forces oset the in uence and power of the military (Belkin and Schofer 2005; Pilster and Bohmelt 2012). In contrast, my theory anticipates that only centrally controlled security forces expand when the government fears a coup, since national forces provide the central government with greater control over the coercive apparatus and a better counterbalance than local militias, which would be unable to prevent a military-backed coup.
4.5 empirical strategy and data
Totest these hypotheses, I use the country-year as the unit of analysis and examine the relative distribution of internal security between countries and over time. First, I examine the relative distribution of personnel across internal security forces from 1970 through 2010. Second, I use an abbreviated timeframe, 1989 to 2009, to include covariates associated with the state’s information environment that do not otherwise span the full time period examined. I employ a seemingly unrelated regression (SUR) model to account for correlation of the error terms across models of internal security personnel. Underlying dynamics that in uence personnel in one security force, such as government budgets and available manpower, are likely
107 to simultaneously in uence personnel in the other security forces.40 Covariates are lagged one year since I expect the distribution of personnel among internal security forces to be slow moving, particularly among formalized and regulated forces like gendarmerie. Government polices and processes associated with troop mobilization of forces, such as changes in budgets and legislation, take time to implement. Robust standard errors are correlated across equations to account for non-independence of errors between models of security force personnel. In the next section, I outline the variables and data used to test my hypotheses.
Dependent variables: Personnel in internal security forces
The dependent variable is the estimated personnel in each type of internal security force per country-year. I use my new data on internal security forces to dierentiate between types. I include personnel measures on internal security forces for over 125 countries from 1970 through 2010, after initially examining 198 countries and territories. I provide descriptive statistics for the full panel of internal security forces in the Appendix, in Table 5.6. I measure personnel across three types of internal security forces that correspond with my typology:
1. Gendarmerie − regular centrally controlled forces,
2. Local Militia − irregular local forces, and
3. State-Managed Militia − regular local forces.41
As discussed in Chapter 2, The Military Balance (2011) and Pilster and Böhmelt(2011) provided the initial population for internal security forces and personnel estimates, from which I distinguish dierent types using Jane’s Country Risk Assessments (1999-2015) (Jane’s) and country-speci c references. Gendarmeries include forces like France’s Gendarmerie Nationale, Italy’s Arma dei Carabineri, and the 40Alternative xed eects speci cations can be found in the appendix. Country xed eects account for time-invariant characteristics of the state that in uence the distribution of security and year xed eects account for widespread shocks. I employ a multivariate probit estimation to jointly estimate the presence, rather than personnel, of internal security organizations. As the presence or absence of a security organization is binary, the multivariate probit model allows for joint estimation of several correlated binary outcomes (i.e. the presence or absence of a militia and the presence or absence of a gendarmerie). 41Due to data constraints, I explore three of the four categories initially discussed. While currently excluding irregular central forces, I plan to incorporate measures on this fourth type of security force in future work.
108 Philippine’s National Police that conduct both policing and domestic counterinsurgency operations. Irregular local militia consist of forces outside the government apparatus acting on behalf of the government, like Oman’s tribal home guards and Peru’s Rondas Campesinas. Regular militias include regionally based and government managed forces like Iran’s Basij Popular Mobilization militia and Jordan’s civil militia that have standardized units and a set organization under the government. Additional coding rules and a complete list of these armed forces are provided in Chapter 2.
Main explanatory variables: Information requirements
To account for variation in ethnic populations within autonomous territories, I use measures from Vogt et al.(2015)’s integrated Ethnic Power Relations dataset. It provides information on ethnic group settlement patterns within a country’s autonomous territories. Using the Ethnic Power Relations (EPR) dataset (Cederman, Wimmer and Min 2010), Vogt et al.(2015) provides information on the population within autonomous territories and whether ethnic populations from those territories are involved in con ict or associated with rebel groups.42 Populations are considered living in regionally autonomous areas when they have a “meaningful and active regional executive organ that operates below the state level (e.g. department or provincial) [and] the regional entity has de facto political power” (Vogt et al. 2015, p. 1331-1332). Populations in autonomous territories are considered involved in con ict when at least one rebel organization is linked to that group. I incorporate into the models both a measure for the total population within a country’s autonomous territories and whether populations were previously associated with rebel activity.43 I use measures of geographic distance from the PRIO Con ict Site dataset for information relevant to con ict sites and distance (Hallberg 2012). This dataset provides global data on geo-referenced armed con icts from 1989 to 2010. It makes available the latitudes and longitude of con ict zones and the approximate area of con ict zones in kilometers. From this, I develop a measure for the maximum distance between a con ict site and the country’s capital in kilometers. I include a measure for the 42Nils-Christian Bormann, Luc Girardin, Philipp Hunziker, Manuel Vogt. 2015. GROWup Research Front-End Documentation RFE Release 2.0. http://growup.ethz.ch/[email protected]. 43I separately examine location and geographic area of territorial autonomy.
109 maximum area covered by con ict. More ethnically diverse societies may decrease the national government’s ability to access local information and social networks and simultaneously encourage irregular local forces to emerge along ethnic cleavages. In addition, the ethnic diversity or composition of a country could hinder the national government’s ability to have local area knowledge using regular armed forces (Tollefsen and Buhaug 2015). Therefore, I incorporate a measure for ethnic diversity using the ethnolinguistic fractionalization (ELF) index (Fearon and Laitin 2003) based on data from the Atlas Narodov Mira (1964). Other factors of geography could impact the ability of a government to collect information on rebels and monitor its own forces, including rough or mountainous terrain. As noted by Tollefsen and Buhaug (2015, p.13), “the availability of safe havens, by means of rough terrain or neighboring territory, increases con ict risk by making prospective rebels harder to detect and defeat through conventional military means.” Rebel organizations operating in rougher terrain should be more di cult for the government to identify and root out through regular forces, in addition to making it harder to coordinate and monitor one’s own forces. To address these other geographic challenges, I rely on the commonly used indicator for mountainous terrain by Gerrard (2000) and extended by Fearon and Laitin 2003, which incorporates information on plateaus, altitude and landscape ruggedness.44 Finally, for identifying the population of countries experiencing civil con ict, I rely on the indicator variable for con ict from the UCDP Monadic Con ict Onset and Incidence Dataset (1946-2014). It provides an aggregated country-year version of the UCDP/PRIO Armed Con ict Dataset (ACD), designating whether a country experienced intrastate con ict in a given year (Gleditsch et al. 2002b; Pettersson and Wallensteen 2015). Intrastate con ict is de ned as any incidence between a government and another party within its territory that results in at least 25 battle-related deaths (Gleditsch et al. 2002b). From the same dataset, I include a variable for whether the previous year experienced con ict associated with the autonomous territory and the magnitude of violence. The magnitude variable distinguishes between whether there were 25 to 1,000 battle-related deaths or over 1,000 battle-related deaths. 44Gerrard, John. “What is a Mountain?” Development Economics Research Group. World Bank. 2000.
110 Controls
I include variables to evaluate alternative hypotheses and control for factors relevant to the state’s information environment and its distribution of internal security. I use total population and gross domestic product per capita, logged from Kristian Gleditsch’s expanded GDP data and based on Penn World Table values.45 I also include a state failure indicator following Cohen and Nordås(2015), which allows me to test whether state collapse is associated with certain types of internal security forces. This measure from the Political Instability Task Force (PITF) consists of a 4 part ordinal score, with 0 indicating no state collapse, 1 indicating failure of a state authority in a limited party of the country, 2 indicating that a country experienced failure in a substantial part of the country, and 3 indicating a complete collapse or failure of the state authority (Marshall, Gurr and Har 2009). Some works argue that democracies are more likely to have regular armed forces, whereas non-democracies have more irregular armed forces (Carey, Colaresi and Mitchell 2015). I include a variable on executive constraints to account for the in uence and strength of democratic institutions. The measure for executive constraints is from the Polity IV regime characteristics data (Marshall and Jaggers 2005) and is one of the four primary components of the democracy score. The executive constraint variable accounts for the amount of authority accountability groups have over the executive and, therefore, could be related to both decisions to expand security forces and certain types of security forces in response to domestic threats. In addition, the executive constraint variable does not include aspects of political violence in its measure (Hill and Jones 2014), making it ideal to include in studies of con ict and security organizations. The measure consists of a 7-point categorical scale that indicates the extent of institutional constraints on the decision-making powers of the chief executives (Marshall and Jaggers 2005). The lowest score [1] represents unlimited authority of the executive, and the highest score [7] indicates instances when accountability groups have equal or greater authority than that of the executive. I include a measure of military coup risk from Powell(2012). This risk score is compiled from several factors, including economic growth, military spending and personnel, regime type, and years since 45Gleditsch, Kristian S. 2002. “Expanded Trade and GDP Data,” Journal of Conflict Resolution 46: 712-24.
111 last coup.46 Finally, regional dummy variables are used in some speci cations to account for unobserved unit heterogeneity and address time-invariant characteristics of each region that might be associated with the dynamics of civil con ict and internal security (such as other unaccounted factors of regional geography, development and border con ict).47
4.6 results
Does the information environment in civil con ict in uence how governments organize and distribute internal security? While previous work has explored whether militias likely increase the incidents of political violence (Carey, Mitchell and Lowe 2013), relatively little scholarship has explored how the dynamics of civil con ict in uence the expansion of these and other internal security organizations. I analyze personnel in internal security forces across a panel of countries experiencing civil con icts to evaluate these dynamics. I present empirical trends and the results of regressions to evaluate my hypotheses. First, I anticipate that ethnic minorities within autonomous territories are positively associated with personnel in unregulated local forces. Second, I expect that con icts that occur farther from the capital are associated with greater personnel in regulated national forces. Using a cross-country time series, I examine personnel across three types of internal security forces: (1) gendarmeries for regular centrally controlled forces, (2) irregular local militias for irregular local forces, and (3) regular or state-managed militias for regular local forces. Measures for security personnel, total country populations and distance are log transformed due to being highly skewed. Robust standard errors are clustered on the country. I include country xed eects to account for time-invariant characteristics of the state that might change. Year xed eects account for widespread shocks. 46In the appendix, I include an alternative measure of coup risk−the previously attempted number of coups−from Powell and Thyne(2011). 47World regions consist of Western states, Eastern Europe, Latin America, Sub-Sahara Africa, Asia and North Africa & the Middle East. North Africa & the Middle East is the reference category in all models with region indicators.
112 Empirical trends in internal security forces
While all three internal security forces appear in civil con ict, gendarmeries are the most common form of internal security. Among the 27 states experiencing civil con ict in 2000, gendarmeries are present in 10 (37%) con ict-exposed countries (as shown in Figure 4.3). State-managed militia occur in four countries experiencing con ict in 2000, and irregular local militias occur in ve countries experiencing con ict. These geographic distributions of internal security personnel and con ict are provided in Figure 4.2 and Figure 4.3. The relative frequency of gendarmeries compared to local forces is consistent with the broader pattern between 1970 and 2010. Among countries experiencing con ict during this timeframe, gendarmeries are present in 42% of the country-years, whereas irregular militias are present in 20% of country-years and state-managed militias are present in only 15% of country-years. Contrary to previous studies that focus on militias, the data reveal that gendarmeries are common in con ict and more common than other types of internal security forces. This is consistent with my theory on national security forces; compared to local militias, gendarmeries provide the national government with greater command and control, specialized capabilities and logistical support. While local security is certainly present in civil con ict, these ndings suggest that national governments more frequently rely on regular centrally controlled forces to address internal coercive challenges. Interestingly, the frequency of gendarmeries and irregular local militias are relatively similar in countries experiencing civil war (con icts with over 1,000 battle-related deaths). Both irregular local militias and gendarmeries are present in approximately 30% of countries involved in civil war. In contrast, state-managed militias occur in only 10% of countries involved in civil war. The relative increasing occurrence of irregular militias in civil war compared with lower intensity con ict−along with a decreasing occurrence of state-managed militias and gendarmeries−suggests that the magnitude of political violence in uences the distribution of internal forces. This nding lends support to conventional theories on state capacity; when countries experience civil war, the central government may be less able to mobilize more national regulated forces. I also examine when internal security forces are established in the course of con ict. The empirical
113 Figure 4.2: Irregular and regular militias in civil con ict
evidence suggests that central governments do not mobilize irregular local militias as a rst response to civil con ict. Among countries that experience multiple years of civil con ict, irregular local militias begin later in con ict and often in the presence of more regulated forces. Local militias begin on average eight years after con ict onset, whereas most gendarmeries appear within the rst four years of con ict. Local militias also emerge in countries with pre-established national forces. Among the seventeen cases of local militias that emerge during civil con ict, the central government had an established gendarmerie in nine of those instances. For example, local defense militias (Tutsi urban militias) rst appeared in Burundi around 2004, eleven years into the Burundian Civil War and formed from former rebel groups who acted as informants for the more o cial gendarmerie. At that time, Burundi had an established gendarmerie for over twenty years (and the gendarmerie itself likely a legacy of former Belgian rule). Historical accounts
114 Figure 4.3: Gendarmeries in civil con ict
suggest that these local defense militias did complement gendarmerie operations, providing them with information and conducting defensive tasks.48 These defensive tasks typically include area defense, where armed forces deny the enemy access to speci c terrain and attempt to wear down the enemy over time.49 Similarly, Oman created the firqats in 1974−twelve years into the Dhofar Rebellion−and long after the establishment of its gendarmerie. Contrary to previous studies, irregular local militias are not an exclusive cause of con ict but can emerge long after con ict begins. The central government prioritizes the mobilization of regular centrally controlled forces in civil con ict, likely turning to irregular local forces only when other strategies have failed (including when 48As noted by Jane’s, “The tutsi militia usually provided information to the Gendarmerie, but its main task was the administration of protected villages.” Jane’s Sentinel Security Assessments. “Burundi.” 2007: p. 2. 49Army Doctrine Publication (ADP) 3-90. Oense and Defense. Department of the Army. 2012.
115 other information-gathering strategies were unsuccessful and resources for internal security have become more constrained). Fielding irregular local militias means ceding governmental authority to locals, a political cession not given easily.
Table4.1: Emergence of irregular local militias in civil con ict, 1970-2010
Start Previous Country Militia Name Year Con ict Years Colombia Rural Militia 2005 36 Guatemala TerritorialMilitia 1983 14 Burundi Local defense Militia 2004 11 Mozambique Provincial People’s Militias, Local Militias 1986 10 Sudan Popular defense Force 1991 9 Peru Rondas Campesinas 1989 8 Ethiopia People’s Militia 1977 8 Algeria Communal Guards / Legitimate defense Groups 1997 7 Afghanistan Village Defense Force 1984 7 Iran Hezbollahi (Home Guard) 1983 5 Oman Tribal Home Guard (Firqat) 1974 5 Afghanistan Pashtun Tribal Militia 1982 5 Myanmar Armed Village defense and Militia 1973 4 Liberia Militia 2002 3 Iran Border Tribal Militia 1981 3 Ghana People’s Militia 1983 1 Lebanon Local Militias 1982 1 Pakistan Civil Armed Forces 1974 1 Note: The earliest observed date is chosen in cases where security forces repeatedly drop in and out of the sample.
116 Table4.2: Correlation matrix of security personnel in civil con ict, 1970-2010
Military Size Gend. State-Managed Militia Local Militia Military Size 1.0000
Gendarmerie 0.1777*** 1.0000 (0.000)
State-Managed Militia 0.0936** 0.1903*** 1.0000 (0.0263) ( 0.0258)
Local Militia 0.0887** -0.1120*** -0.1856*** 1.0000 ( 0.0434) (0.0082) (0.000)
*** p<0.01, ** p<0.05, * p<0.1
Finally, the allocation of personnel for one force in uences the distribution of personnel across other forces. While the mobilization of personnel are not necessarily drawn from the same population, correlations of security personnel suggest that the relative expansion of one security organization is associated with changes in personnel among other organizations as shown in Table 4.2.50 However, increasing internal security forces does not mean decreases in the military personnel. Personnel across all three internal forces are positively correlated with personnel in the military. This provides evidence that internal security (including irregular local militias) can expand even as regular armed forces expand. Thus the central government has simultaneous strategies for addressing security challenges. Yet the expansion of regular internal security is signi cantly associated with the contraction of irregular internal security. This is likely due to the national government’s preference for security forces it has direct control over and aversion to ceding command authority. National governments prefer to expand forces that they have command authority over. When the national government has limited access and information over certain areas, such as in autonomous ethnic territories, in must then turn to irregular local (and likely co-ethnic) forces. I now turn to the results of regressions to test this logic.
50For example, citizens that compose the militia may be more
117 Table4.3: Seemingly unrelated regression, security personnel in civil con ict (1)
Dependent Variable: Local Militia Gend. State Militia Military
Autonomous population 0.386*** 0.0207 0.0374 0.013*** (0.133) (0.167) (0.135) (0.003) Mountainous terrain -0.218 0.961*** -0.458** 0.015*** (0.184) (0.231) (0.187) (0.004) Con ict distance -0.442*** 0.0993 0.300* 0.00257 (0.158) (0.198) (0.159) (0.004) Previous con ict 0.918*** -1.052*** 0.856*** 0.030*** (0.200) (0.250) (0.202) (0.005) Max intensity 1.488*** -1.535*** -0.500 0.005 (0.449) (0.561) (0.452) (0.011) Coup Risk 11.33*** -20.54*** 3.278 -0.924*** (3.976) (4.938) (3.988) (0.094) State Failure -1.088*** (0.394) post 1990 1.359 (1.001) Russian support 1.291 (1.463) Foreign support -0.045*** (0.010) Constant 0.866 3.282 -1.080 2.280*** (2.109) (2.735) (2.134) (0.051) Observations 378 378 378 378 Note 1: Breusch-Pagan test of independence: χ2 = 61.838, Pr = 0.000 Note 2: Covariates lagged one year and standard errors in parentheses, correlated across equations.
Regression Results
As shown in Table 4.3 and Table 4.4, results of the models suggest that variation in the state’s information environment in uences the allocation of internal security.51 Even after accounting for confounding factors, including GDP and executive constraints, there is a signi cant positive association between ethnic populations in regionally autonomous territory (hence “autonomous populations”) and personnel in 51Descriptive statistics are provided in Table 4.6 in the appendix, along with estimations for binary indicators of the dependent variables (Table 4.8) and xed eects estimation (Table 4.9). Samples with and without state failure are provided in Table 4.12 and Table 4.11 respectively. While other sensitivity analyses could be conducted, I estimate models that test the most prominent arguments in the literature and require the most scrutiny.
118 Table4.4: Seemingly unrelated regression, security personnel in civil con ict (2)
Dependent Variable: Local Militia Gend. State Militia Military
Autonomous population 0.583*** 0.0965 -0.258* 0.0212*** (0.160) (0.169) (0.154) (0.00318) Mountainous terrain 0.0390 0.396* -0.851*** 0.0147*** (0.206) (0.218) (0.195) (0.00429) Ethnic fractionalization -1.336 -7.018*** 1.537 -0.0111 (1.004) (1.057) (0.954) (0.0208) Con ict distance -0.0169 0.275 -0.153 0.00726** (0.168) (0.178) (0.160) (0.00347) Previous con ict 1.240*** -1.203*** -0.245 0.0218*** (0.243) (0.256) (0.232) (0.00501) Max intensity 1.376*** -1.638*** -0.230 0.0113 (0.446) (0.470) (0.424) (0.00932) Coup risk -2.108 6.049 15.80*** -0.293*** (4.887) (5.163) (4.649) (0.100) Population -1.339*** 1.488*** 1.330*** (0.241) (0.253) (0.230) GDP per capita -8.24e-05 3.81e-05 -3.31e-05 6.34e-06*** (6.48e-05) (6.88e-05) (6.16e-05) (1.36e-06) State failure -1.281*** (0.394) Executive constraints -0.326** 0.588*** 0.0960 -0.00145 (0.129) (0.135) (0.121) (0.00259) Post 1990 0.747 (0.821) Russia support 1.787 (1.320) Foreign support -0.0407*** (0.00905) Western Europe, North America 3.163 -11.68*** -2.487 -0.0256 (2.457) (2.610) (2.335) (0.0506) Eastern Europe -1.763 -1.084 -0.0900 -0.0149 (1.104) (1.154) (1.051) (0.0230) Latin America -0.486 -0.946 2.286*** -0.0706*** (0.827) (0.872) (0.786) (0.0171) Sub-Saharan Africa -1.639* -0.327 0.0950 -0.103*** (0.841) (0.889) (0.800) (0.0173) Asia 0.354 -3.141*** 2.863*** -0.0235 (0.820) (0.864) (0.784) (0.0171) Constant 19.06*** -21.83*** -16.26*** 2.112*** (3.901) (4.098) (3.691) (0.0510)
Observations 366 366 366 366 R-squared 0.219 0.432 0.289 0.592 Note: Covariates lagged one year and standard errors in parentheses, correlated across equations. *** p<0.01, ** p<0.05, * p<0.1
119 local militias the following year. Con ict-exposed countries with more autonomous populations are associated with more personnel in local militias; a 10% relative increase in autonomous populations is associated with a 4% relative increase in Local militia personnel. For a country like Myanmar with 35,000 militia personnel in 2008, for example, the coe cient estimates suggests that militia personnel would increase by around 1,330 personnel in 2009 if the autonomous population increased from 14,800,000 people to 16,280,000 people. Even absent failure of the government in certain parts of the country, autonomous populations are positively and signi cantly associated with expansion of unregulated local security (Table 4.11).52 Taken together, this supports Hypothesis 1 that autonomous populations are positively associated with irregular local security in the course of civil con ict. These ndings are broadly consistent with my theory on the strategic expansion of local forces; irregular local security can gather information or intelligence from local populations that the central government and its regular centrally controlled security cannot. In terms of the central government deploying forces across rugged terrain and distances, the coe cient estimate for mountainous terrain is positive and statistically signi cant for regular centrally controlled forces but negative for militias. These ndings are statistically signi cant in models that exclude state failure (Table 4.11). Similarly, the coe cient estimate for distance of con ict from the capital is negatively associated with personnel in irregular local militias and positively associated with regular forces. This supports Hypothesis 2 that regular centrally controlled forces increase when the government needs to deploy forces across wider or more di cult terrain. While mobilizing centrally controlled forces like gendarmeries is more expensive or requires greater resources, these forces are easier to deploy, coordinate and support across regions. With centralized and institutionalized oversight, including greater C2 and logistical capabilities, regular central forces are a rational strategy for governments confronting rebellion or civil con ict in the geographic periphery or across di cult terrain. However, these dynamics do not extend to cases of state failure, where the national government has failed to prevent behavioral challengers from achieving full control in certain parts of the country 52In cases with a functioning national government across the country, autonomous populations are still positively associated with gendarmeries. This could be due to national forces and local forces being concurrent strategies in civil con icts that national governments use to address rebels emerging from autonomous ethnic populations.
120 (Table 4.12).53 In cases of state collapse in certain parts of a country, mountainous terrain and the distance of con ict from the capital is signi cantly associated with contraction of gendarmeries and state-managed militias but expansion of the military.54 This implies that only countries above a certain threshold of administrative capacity or state control have the option of deploying regular armed forces. Generally militias are likely less able and less willing to deploy to distant areas compared with centrally controlled forces (like gendarmeries) who have increased mobility through helicopters and transport aircraft (Lutterbeck 2004). Local militias are also unlikely to be eectively away from their home areas. Given limited training and capabilities, their use outside their home areas is likely to prove counterproductive.55 Rugged terrain appears to constrain the mobilization and deployment of state-regulated militias but increase the number in gendarmeries. Contrary to previous work, my results suggest that con ict in the periphery of a country and over di cult terrain is associated with the mobilization of regular central security.56 Mountainous terrain and other geographic challenges can incentivize the central government to rely on national sources of security, which have the mobility and logistical capabilities necessary to support such operations. Similarly, the coe cient estimate for state failure is in the opposite sign that theories on state capacity and local militias would predict; its negative sign suggests that states experiencing failure are associated with a relative contraction in local militia personnel the following year. Acknowledging that The Military Balance (2011) only measures militias explicitly aligned with the central government, it could be that The Military Balance (2011) is less likely to identify local militias as pro-government when central governments are dissolving or have dissolved. Alternatively, it may be that state collapse is associated with a change in the loyalty and alignment of local armed forces; either militias aligned with a functioning government in one year are not necessarily aligned with that government once it collapses or a change in loyalty by those 53As de ned by the Political Instability Task Force, state failure occurs when there is at least “failure of state authority in a limited part of the country, e.g., secession or rebel control of, or anarchic conditions in, one or several regions that do not include the core area of the country or its capital” (p. 12). Political Instability Task Force - Problem Set Codebook version 2014. 54Due to the small number of cases, it is unclear how robust these ndings are for state collapse. 55Cooper(2002). 56It may be that local security forces are also less easily identi ed and measured within peripheral con icts, in which case con icts in the periphery are not necessarily associated with the contraction of local security but only limit our ability to measure it.
121 militias precipitates state collapse. Other internal threats drive changes in the relative allocation of internal security. Countries with greater risk of coups are associated with the expansion of local forces and contraction of central forces. Previous research has emphasized the establishment of alternative security organizations to “counterbalance” the military (Belkin and Schofer 2005). Building on those results, these ndings suggest that governments not only develop alternative sources of security but speci cally expand local security forces (rather than central forces). This nding is consistent with the experiences of post-revolutionary countries like the United States where civilians and local governments were distrustful of a strong military that could intimidate and repress citizens.57 Regarding domestic political factors, I nd evidence consistent with previous research on democratic institutions and security forces. The coe cient estimates on executive constraints suggest that constraints over national leadership are positively associated with regular security forces and negatively associated with irregular security forces. Having fewer constraints on the executive could be associated with less institutionalized governance generally, including fewer restraints and regulations over the adoption of certain forces. In addition, constraints on the executive can lead the national government to adopt alternative national organizations (such as the gendarmerie or national police rather than simply relying on the military).58 Finally, there is evidence to suggest that the external environment shapes the state’s distribution of security.59 External support (such as foreign military aid) is negatively associated with military personnel but positively associated with personnel in internal security forces.60 This suggests that national governments may substitute militaries with foreign military support but at the same time expand their internal security forces. This could be due to needing more internal security (particularly during civil 57Browne, Susan M. “War-Making and US State Formation.” Irregular Armed Forces and their Role in Politics and State Formation. 2003: p. 232. 58Chapter 5 discusses this logic in greater detail, showing that national governments with limited constraints over the executive tend to have more repressive national security forces. National governments may have enough administrative capacity to generate and regulate centralized forces but have weak enough democratic constraints that they can use these forces for repressive purposes. 59The role of the Cold War is less clear from these tests since the vast majority of observations in the main speci cations occur after the Cold War (1988 to 2010). Alternative speci cations for testing the Cold War are found in the Appendix, Table 4.13, and suggest that both gendarmeries and irregular local militias expand after the Cold Warwhereas regulated militias contract. 60The correlation matrix between external support and security personnel is found in Table 4.7 of the appendix.
122 con ict) or strategically investing foreign aid in internal forces for political purposes.61 In addition, the end of the Cold War is associated with the expansion of local militias and gendarmeries (Table 4.13). The end of the Cold War is associated with a change in the distribution of personnel across internal forces during civil con ict; more personnel in irregular local militias and gendarmeries. After the Cold War, national governments likely turn more to these two complementary forces as internal con icts became more protracted, military expenditures among NATO and NATO allies decrease, and/or border and transnational challenges more frequently emerge (Lutterbeck 2004).62
4.7 conclusion
Why is there variation in how states organize their internal security during civil con ict? Why do some central governments rely on poorly trained local forces to combat rebels rather than highly professionalized standing forces? This chapter argues that a country’s information environment in uences the distribution of internal security, leading to the national government’s strategic adoption of forces like local militias that would be irrational to promote under other circumstances. The challenges of population-centric operations and information gathering−working with and gathering intelligence from local populations−encourage the mobilization of irregular local forces when the central government lacks the capability to do so. However, local militias are neither the most frequent nor least risky strategy for the central government. Fielding local militias means that the central government must cede authority (along with control and lethal capability) to local powers. Gendarmeries are more common, more easily controlled and more easily coordinated by central authorities in states experiencing civil con ict. This is due to their relative advantages in command and control (and related advantages in mobility and specialization) compared to irregular local forces. The central government’s need to coordinate and deploy security forces over di cult terrain or in the 61For example, countries such as Indonesia and the Philippines invest heavily in domestic security forces, including the Coast Guard, as these forces are viewed as less threatening than the military by regional rivals. Morris, Lyle J. “Blunt defenders of sovereignty: the rise of coast guards in East and Southeast Asia.” Naval War College Review 70(2): 2017, p. 75-113. 62In terms of changing military expenditures since the Cold War, Sandler and George (2016, p. 174) nd “rising shares of East Asia & Paci c and the Middle East & North Africa and recent falling shares of North America and Europe & Central Asia.” Sandler, Todd, and Justin George. “Military expenditure trends for 1960-2014 and what they reveal.” Global Policy. 7(2). 2016: p. 174-184.
123 geographic periphery, along with elding forces with more technical means of gathering information, lead to the expansion of highly centralized forces. In contrast, local militias have relative advantages in local knowledge with access to ethnic kin and other populations regular armed forces cannot access. My ndings show that central governments are far more selective and strategic in their mobilization of forces during civil con ict than initially presumed. The domestic challenges a government confronts−particularly information challenges−strongly in uence the organization of internal force and personnel allocated to them. As the population of a country is seen as critical to intelligence-driven operations,63 the relationship between civilians and the state’s security forces remain an important part of understanding potential successes or failures within counterinsurgencies. In addition, this chapter shows that the expansion of local militias occurs well after con ict’s onset and alongside regular forces. This nding challenges the conventional wisdom on pro-government militias, which views militias exclusively as a cause of con ict (Carey, Mitchell and Lowe 2013). The causal relationship between militias and con ict may, in part, go the other direction; central governments strategically delegate to local militias to address challenges that regular armed forces are unable or ill-suited to address during civil con ict. While this chapter does not explicitly test causal mechanisms, the ndings are consistent with my theory regarding variation in the state’s information environment and the advantages of dierent types of organizations. Governments create or expand local militias when they lack local knowledge and information within communities of rebel support. Governments expand gendarmeries when they need to deploy, coordinate and monitor troop activity across di cult terrain or far from main operating bases. Other geographic measures related to civil con icts, such as porous borders and roads, can provide additional insight into the conditions where local militias and gendarmeries are mobilized. Results from this chapter suggest that a fruitful next stage of research would be to examine how variation in logistical challenges−namely the ability to supply deployed forces−in uences the deployment of security forces. Finally this chapter shows the bene ts of using continuous personnel measures rather than binary speci cations of security forces. The re ned conceptualization and measurement of internal security 63Sims, Christopher. The Human Terrain System: Operationally Relevant Social Science Research in Iraq and Afghanistan. U.S. Army War College Carlisle. 2015.
124 allows for a comprehensive understanding of the state’s internal coercive capabilities. States engaged in civil con ict have already lost the monopoly on the use of force, but governments strategically delegate the use of force to irregular local forces to gain information capabilities they otherwise lack. However, these forces are not the only or even the most likely option. Central governments turn far more frequently to national police and gendarmeries to combat rebel forces. This chapter discusses the delegation of force to local forces by the central government. The theory presented and ndings reached suggest that a rationale exists for using these forces, but that regular centrally controlled forces are still more common and useful for central governments dealing with insurgents. In domestic con ict, rather than irrationally or haphazardly devolving security, the central government strives to maintain its monopoly of force.
125 appendix
Table4.5: Personnel in internal security forces, 1970 - 2010
Variable Obs Mean Std Dev. Min Max Gendarmarie 6,955 13,171 84,295 0 1,850,000 Local Militia 6,943 2,293 27,607 0 900,000 State-Regulated Militia 6,990 117,342 2,036,472 0 94,500,000 Military 6,995 126,224 348,028 0 4,750,000 Note: Raw personnel estimates provided.
Table4.6: Descriptive statistics for civil con ict, 1989-2009
Variable Obs Mean Std. Dev. Min Max Military_ln 561 11.215 2.157 0 14.635 Gendarmerie_ln 600 4.106 5.033 0 14.221 State Managed Militia_ln 609 1.597 3.848 0 14.221 Local Militia_ln 591 1.772 3.897 0 13.305 Autonomous Pop._ln 607 5.225 7.193 0 18.272 Con ict Area_ln 556 11.646 1.808 6.111 14.655 Capital Distance (km)_ln 556 12.458 1.601 0 14.584 Prior Con ict w/ Auto. Pop. 607 0.448 0.978 0 4 Max Intensity 609 1.271 0.445 1 2 Coup Risk 530 0.054 0.056 0 0.317 GDP per capita 609 4496.915 6720.287 203.27 43499.89 State Failure 609 0.251 0.770 0 3 Executive Constraints 563 3.471 2.388 0 7 Note 1: Logged personnel estimates provided. Note 2: States that lack a military at some point during this period consist of Afghanistan, Bosnia and Herzegovina, Somalia, Tajikistan,and Uganda.
Table4.7: Correlation matrix: External support and security personnel in civil con ict
External Support Military Gendarmerie State Militia External Support 1 Military -0.187 1 Gendarmerie 0.065 0.164 1 State Militia 0.022 0.156 0.047 1 Local Militia 0.036 0.0571 -0.188 -0.201 Note: Logged personnel estimates provided, 1975-2009.
126 Table4.8: Multivariate probit coe cient estimates, security personnel in civil con ict
DV: Local Militia Gendarmerie State Militia
Regional autonomy 0.323*** -0.0615 -0.115** (0.0732) (0.0458) (0.0535) Mountainous terrain 0.112* 0.150** -0.218*** (0.0623) (0.0608) (0.0701) Con ict distance -0.0647 -0.0248 0.0106 (0.0614) (0.0615) (0.0500) Previous con ict 0.332*** -0.300*** 0.148** (0.0914) (0.0647) (0.0668) Max intensity 0.315* -0.432*** -0.301 (0.178) (0.142) (0.194) Coup risk -0.515 -3.484** 3.257** (1.638) (1.392) (1.431) Pop. -0.427*** 0.0848 0.433*** (0.104) (0.0756) (0.0727) State failure -0.767** (0.303) post 1990 0.361 (0.269) Russia support 0.545 (0.412) Constant 2.738** -0.332 -6.570*** (1.382) (1.113) (1.189) Observations 394 394 394 Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1 Breusch-Pagan test of independence: χ2 = 8.595, Prob > χ2 = 0.0352
127 Table4.9: OLS coe cient estimates (FE); personnel in security forces (1989-2009)
Equation [1] [2] [3] D.V: Local Militia State-Managed Gendarmerie
Conflict factors Autonomous territory 0.133*** -0.145 -0.0692 (0.0389) (0.0986) (0.164) Con ict area 0.127 -0.103 0.0161 (0.0899) (0.217) (0.200) Con ict distance -0.162 0.0265 0.638*** (0.103) (0.0512) (0.106) Prior con ict -0.689** 0.014 -0.986 (0.304) (0.785) (1.263) Max intensity -0.502** -0.081 -0.151 (0.238) (0.584) (0.211) Coup risk 2.564 6.239 -6.019 (3.744) (3.815) (5.977) Security forces Military_ln 0.140 0.243 0.0123 (0.206) (0.178) (0.251) Gendarmerie_ln 0.0421 0.0224 (0.0421) (0.0886) State Managed Militia_ln -0.0104 -0.112*** (0.019) (0.0319) Local Militia_ln 0.0211 0.0593 (0.0242) (0.0497) State factors GDP per capita -0.00018* 0.00004 -0.0002 (0.000104) (0.000) (0.0002) State Failure -0.252 -0.375 -0.232 (0.347) (0.378) (0.205) Executive Constraints 0.0301 -0.342* -0.181 (0.217) (0.200) (0.179)
Constant 1.392 1.194 -0.0230 (2.094) (2.463) (1.652)
Country FE Yes Yes Yes Year FE Yes Yes Yes Observations 380 381 381 R-squared 0.151 0.121 0.205 Number of countries 52 52 52 Robust standard errors in parentheses, clustered on country. Covariates lagged one year. *** p<0.01, ** p<0.05, * p<0.1
128 Table4.10: Population measure, coe cient estimates (FE)
Equation [1] [2] [3] DV: Local Militia State Militia Gend.
Autonomous territory 0.149*** -0.150 -0.0621 (0.0402) (0.0954) (0.162) Con ict area 0.0904 -0.0787 0.0138 (0.0806) (0.214) (0.202) Con ict distance -0.149 -0.0363 0.630*** (0.0902) (0.0504) (0.101) War history -0.929** 0.152 -1.301 (0.364) (0.843) (1.151) Intensity -0.439* -0.246 -0.440 (0.229) (0.644) (0.316) Coup risk 1.283 3.358 -9.433 (3.670) (4.588) (6.452) GDP -0.0002* -0.0003 -0.0005* (0.0001) (0.0003) (0.0003) State failure -0.294 -0.420 -0.286 (0.328) (0.411) (0.204) Executive constraints 0.0189 -0.314 -0.201 (0.204) (0.191) (0.160) Pop_ln -4.760 0.678 -3.713 (2.924) (3.012) (9.096) Constant 84.39* -5.654 65.15 (50.08) (50.90) (154.4)
Year, Country FE Yes Yes Yes Observations 381 391 390 R-squared 0.157 0.097 0.251 Number of Countries 52 52 52 Robust standard errors in parentheses, clustered on country *** p<0.01, ** p<0.05, * p<0.1
129 Table4.11: Seemingly unrelated regression, personnel in security forces (absent state failure)
Equation [1] [2] [3] [4] DV: Local Militia Gend. State Militia Military
Autonomous territory 0.272** 0.338** 0.0762 0.240*** (0.131) (0.154) (0.125) (0.0424) Distance from capital -0.175 0.385** 0.165 0.107** (0.132) (0.156) (0.127) (0.0430) Mountainous terrain 0.0269 0.654*** -0.725*** 0.177*** (0.169) (0.202) (0.164) (0.0555) Ethnic fractionalization -2.429*** (0.849) War history 0.938*** -1.048*** 0.103 0.296*** (0.211) (0.254) (0.206) (0.0698) Max intensity 1.359*** -1.608*** -0.0442 0.0476 (0.378) (0.454) (0.368) (0.125) Coup risk -2.098 2.497 6.238 -4.560*** (4.226) (5.061) (4.105) (1.392) GDP -5.73e-05 0.000102 -8.35e-05 9.07e-05*** (5.64e-05) (6.73e-05) (5.46e-05) (1.85e-05) Executive constraints -0.324*** 0.604*** 0.125 -0.0582* (0.106) (0.127) (0.103) (0.0349) Western Europe & North America 0.485 -7.192*** 0.376 -0.447 (1.599) (1.920) (1.557) (0.528) Eastern Europe -2.640*** -2.497** 1.143 -1.130*** (0.860) (1.028) (0.834) (0.283) Latin America -0.240 -1.590* 1.810** -0.665*** (0.737) (0.887) (0.719) (0.244) Sub-Saharan Africa 0.141 -2.169** -0.221 -1.005*** (0.771) (0.879) (0.712) (0.242) Asia 0.377 -2.886*** 3.327*** 0.00158 (0.736) (0.857) (0.695) (0.236) Constant 1.805 -3.990* -0.977 7.207*** (1.984) (2.378) (1.928) (0.654)
Observations 440 440 440 440 R-squared 0.175 0.263 0.205 0.463 Note 1: Coe cient estimates lagged 1 year; dependent variable is personnel in security forces (logged) Note 2: “Absent state failure” means that there is no state failure in even a limited part of the country. Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
130 Table4.12: Seemingly unrelated regression, personnel in security forces (state failure)
Equation [1] [2] [3] [4] DV: Local Militia Gend. State Militia Military
Mountainous terrain -0.393 -0.660* -0.316* 0.381*** (0.357) (0.346) (0.189) (0.104) Autonomous territory 0.247 0.238 0.0515 0.116 (0.267) (0.323) (0.176) (0.0969) Ethnic fractionalization 2.273 (1.778) Distance from capital -0.664 -1.470*** 0.00800 0.793*** (0.491) (0.541) (0.295) (0.162) War history 2.646 2.814 -3.221*** (2.842) (1.860) (1.023) Max intensity -0.865 -1.537 0.132 -0.223 (1.080) (1.307) (0.714) (0.393) Coup risk 3.796 15.88*** 5.370* 0.646 (4.995) (5.956) (3.252) (1.789) GDP 0.000194 -3.18e-05 0.000608*** 0.000107 (0.000348) (0.000419) (0.000229) (0.000126) Executive constraints 0.0282 -0.351 0.955*** 0.160 (0.430) (0.516) (0.282) (0.155) Latin America 2.206 14.75*** -7.111** -4.413*** (4.350) (5.267) (2.875) (1.582) Sub-Saharan Africa 2.085 0.202 -0.390 -3.454*** (2.785) (3.301) (1.802) (0.991) Asia 0.815 -3.043 7.410*** -2.657** (3.840) (4.370) (2.386) (1.313) Eastern Europe 0.122 (3.406) Constant 3.978 19.98*** -1.553 1.607 (3.821) (4.619) (2.522) (1.387)
Observations 31 31 31 31 R-squared 0.293 0.713 0.844 0.734 Note 1: Coe cient estimates lagged 1 year; dependent variable is is personnel in security forces (logged). Note 2: “State failure” means that there is state failure in at least a limited part of the country. Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
131 Table4.13: Seemingly unrelated regression, personnel in security forces and the Cold War
Equation [1] [2] [3] [4] DV: Local Militia Gend. State Militia Military
Cold War -0.831*** -1.364*** 0.267 -0.128 (0.277) (0.331) (0.272) (0.0894) Mountainous terrain 0.524*** 0.0292 -0.554*** 0.175*** (0.118) (0.142) (0.116) (0.0377) Max intensity 0.901*** -0.878*** 0.480*** (0.200) (0.246) (0.0668) GDP -5.47e-05* 2.84e-06 -7.89e-06 2.88e-05*** (3.21e-05) (3.84e-05) (3.16e-05) (1.04e-05) Executive constraints -0.126* 0.277*** -0.00631 0.0577*** (0.0652) (0.0780) (0.0642) (0.0217) Pop_ln -0.549*** 0.825*** 0.229* (0.127) (0.153) (0.126) Coup risk -4.199*** (0.889) Western Europe and North America 1.370* -6.269*** -1.347* 0.410* (0.794) (0.951) (0.781) (0.249) Eastern Europe -1.891*** -3.967*** 1.906*** -1.031*** (0.717) (0.858) (0.706) (0.233) Latin America -0.345 -1.118* 1.560*** -1.295*** (0.491) (0.588) (0.484) (0.159) Sub-Saharan Africa -0.182 -3.831*** 0.0399 -1.489*** (0.444) (0.532) (0.437) (0.147) Asia 1.487*** -3.919*** 3.130*** 0.0973 (0.470) (0.563) (0.463) (0.145) Constant 9.520*** -6.475** -1.806 10.94*** (2.103) (2.530) (2.085) (0.200)
Observations 838 838 838 838 R-squared 0.092 0.205 0.156 0.440 Note: Coe cient estimates lagged 1 year; dependent variable is personnel in security forces (logged). Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
132 Chapter 5
Bringing the police back in: Internal state security and repression
5.1 introduction
“Power tends to corrupt, and absolute power corrupts absolutely.”
- John Dalberg Acton (1887)1
Do states with more centrally controlled security arrangements engage in more repressive practices or, alternatively, are local forces responsible for widespread state abuse? On one hand, human rights violations of local forces like the Afghan Local Police and Turkish village guards are well documented (Biberman 2016). Afghan Local Police have beaten, murdered, raped and extorted civilians in the same villages they were meant to protect (Felbab-Brown 2016). Turkey’sVillage Guards, established to ght the Kurdish Worker’s Party (PKK), tortured, abducted and burned down property of villagers who refused to join their units.2 Without fear of punishment or oversight, local security forces can abuse and extort civilians for their own gain. Yet centrally controlled and professionalized forces like gendarmeries and national police have committed mass atrocities (Dallin and Breslauer 1970). During the Great Terrorin the Soviet Union (1937 to 1938), the head of the NKVD, Nikolai Ezhov, under direct orders from Stalin had his regional chiefs brutally and eectively conduct “large-scale purge operations,” arresting 1.5 million people and killing 1Acton’s full letter solemnly warns of the relationship between corruption and authority: “Power tends to corrupt and absolute power corrupts absolutely. Great men are almost always bad men, even when they exercise in uence and not authority: still more when you superadd the tendency or the certainty of corruption by authority. There is no worse heresy than that the o ce sancti es the holder of it.” John Dalberg Acton (Lord Acton). Letter to Archbishop Mandell Creighton. April 5, 1887. 2Amnesty International. “Turkey: Abduction / Fear of Torture.” September 1993.
133 approximately half of those arrested (Jansen and Petrov 2002, p. 80). Soviet repression was systematic during Stalin’s consolidation of power (1929-32) until his death in 1953, including the development of a “brutal administrative bureaucracy” (Dallin and Breslauer 1970, p. 69). Under the Afghanistan Interior Ministry, the Afghan National Police and Afghan Directorate of Security forces engaged in corruption, assaulted protestors, even beat and detained journalists.3 Turkey’s Gendarmerie−Jandarma−have historically committed human rights abuses against Kurds, using a disproportionate amount of force in its operations, and recently killed Syrians seeking asylum 4 Unlike local forces, centralized armed forces may have direct authorization by the national government to repress and more lethal capabilities to do so. Conventional wisdom on state repression fails to explain how repression varies across centrally controlled forces like the Afghan National Police and the Jandarma relative to local forces like the Afghan Local Police and Turkish Village Guards. A lack of oversight might incentivize predatory behavior or, alternatively, security forces might abuse citizens because of directives given from national leaders. Complicating this relationship between security and repression, domestic forces abuse citizens in democratic as well as non-democratic contexts (Campbell and Brenner 2002). Beginning in the late 1970’s, Spain’s Guardia Civil became one of the state’s main instruments for repression, having broad authority to conduct political imprisonments and beatings of Spanish civilians.5 As recently as 2011, the Guardia Civil has been accused of torturing a former member of the Euskadi Ta Askatasuna (ETA) despite Spanish laws prohibiting such practices.6 Abuses across contexts and regimes make it unclear whether centralization ultimately helps or hinders human rights. In this chapter, I show that democratizing states with more centralized security have a greater propensity toward widespread abuse than other types of states and even those with local security, as centralization provides national leaders in limited democracies with the authority and coercive ability 3Human Rights Watch. “Afghanistan: Security Forces Assault Reporters” 2016. Hughes, Michelle. “The Afghan National Police in 2015 and Beyond.” United States Institute of Peace Special Report. 2014. 4Human Rights Watch. “Turkey Border Guards Kill and Injure Asylum Seekers.” 2016. Country Reports on Human Rights Practices for 2011. Turkey. 2011. 5Encarnacion, Omar Guillermo. “Democracy and dirty wars in Spain.” Human Rights Quarterly 29(4): 2007, p. 950-972. 6Country Reports on Human Rights Practices for 2016. “Spain.” Bureau of Democracy, Human Rights and Labor. United States Department of State.
134 to repress popular challenges to their rule.7 In democratizing countries, internal centralized forces like national police and gendarmeries are directly accountable to the national government but not necessarily to the public. Centralized forces oer these national leaders a unique tool for controlled repression, as leaders have direct authority over them but these forces often have limited public transparency. Examining security arrangements and state repression across 149 countries between 1976 through 2010, I nd that states with only moderate executive constraints and limited judicial independence−as in democratizing or partially democratic states−and highly centralized security have a high likelihood of widespread repression compared to fully autocratic or democratic regimes.8 This result holds even when accounting for the most common predictors of state repression. In contrast, local security forces are not associated with widespread repression, though they might engage in predatory behavior or more frequent incidents of minor abuse not accounted for in conventional measures of repression. The chapter ends with a discussion on how countries must have public accountability mechanisms for centrally controlled forces and that robust punishment of abusive practices within those forces is necessary to discourage abuse. Central control of security is not su cient to discourage repression and may even diminish public transparency, encouraging national leaders to use security forces to repress−rather than protect−their population.
5.2 conventional wisdom on state repression
The (contingent) domestic democratic peace
I draw on Goldstein(1978)’s and Davenport(2007 a)’s conceptualization of repression for this chapter. Goldstein (p. xviii 1978) de nes state repression as “government action which grossly discriminates against persons or organizations viewed as presenting a fundamental challenge to existing power relationships or key government policies, because of their political beliefs.” Expanding on this de nition, Davenport(2007 a, p. 2) conceptualizes repression as “the actual or threatened use of physical sanctions 7By popular contestation, I mean cases where the public challenges the regime or national leader through voting for another candidate, engaging in protests or other mass demonstrations. 8I use executive constraints and judicial independence as measures of democracy as these mechanisms should limit the ability of national leaders to order and carry out widespread extralegal repression. These measures have the additional bene t of not including aspects of political violence/repression in their scores (Hill and Jones 2014).
135 against an individual or organization, within the territorial jurisdiction of the state, for the purpose of imposing a cost on the target as well as deterring speci c activities and/or beliefs perceived to be challenging to government personnel, practices or institutions.” Despite the key role of the “agents of atrocity” in the conduct of repressive behavior, studies on repression often examine non-coercive institutions of the state in cross-national analysis,9 ignoring security forces that are the actual sources of physical repression.10 Research examines the relationship between domestic security and repression from a comparative or subnational perspective (Campbell and Brenner 2002; Eck 2015; Greitens 2016), and the empirical evidence remains divided on whether centralized security encourages repression (Linz 2000) or discourages it (Mitchell, Carey and Butler 2014). I brie y review the conventional wisdom on drivers of repression. Then I discuss the disagreement in the literature over whether (de)centralized security results in greater abuse. Most studies on state repression, particularly cross-national studies, argue that various domestic political elements of the state constrain or encourage abusive behavior. One theory proposes that democratic institutions constrain national leaders from directing domestic abuse−promoting a “domestic democratic peace”−as leaders ordering abuse risk loss of popular approval or fail to win re-election (Davenport 2007b; Bueno de Mesquita et al. 2005). Poe and Tate(1994, p. 855) similarly argue that “eective democracy provides citizens the tools to oust potentially abusive leaders from o ce before they are able to become a serious threat.” Many studies nd a strong positive relationship between democratic institutions and protection of personal integrity rights (Davenport 2007b; Hill and Jones 2014; Poe and Tate 1994). While Poe and Tate(1994) fail to determine which democratic institutions restrain abusive leaders, Hill and Jones(2014) nd that stronger executive and legal constraints decrease human rights abuse.11 Other works argue that more expansive surage decreases the most lethal forms 9Mitchell(2004). 10Cohen and Nordås(2015) and Mitchell, Carey and Butler(2014) are notable exceptions, though both focus on one type of organization (pro-government militias) rather than conducting a comprehensive study of security arrangements. Butler, Gluch and Mitchell(2007) examine the relationship between security forces and abuse through the highly problematic proxies of nancial corruption and size of a country’s military, which are unlikely to capture organization of security or variation in regulation over these organizations. 11Hill and Jones(2014, p. 677) also nd that competitiveness of participation is strongly predictive of repression, but this is due to Marshall and Jaggers(2005)’s competitiveness measure incorporating repression itself, namely imprisonment of political competitors.
136 of repression, as constituents can vote out abusive leaders (Davenport 2007b). According to this line of reasoning, democratic political institutions tend to curb repression. Yet democratic institutions do not always dampen repression nor prevent all forms of abuse. Domestic or demographic challenges to the central government−such as popular protest, civil violence, and youth bulges−can overwhelm the pacifying eect of democracy, including the bene cial in uence of mass participation in elections and constraints on national leaders (Davenport 2007b; Conrad and Moore 2010; Nordas and Davenport 2013). Studies nd that democratic mechanisms generally decrease abuse of physical integrity rights while decreasing civil liberty (Davenport 2007b, p. 132). Democratic mechanisms may not even prevent all types of physical integrity abuses. They may simply lead governments to change the tactics and agents for carrying them out (Campbell and Brenner 2002; Cunningham 2004; Rejali 2007).12 Democracy is not a panacea for human rights. The limitations of democratic mechanisms raises key questions: what is the domestic democratic peace conditioned on and how do internal security arrangements dampen or strengthen that eect?
Centrally controlled security as predatory or protective?
In contrast to work focusing on democratic institutions, a separate research agenda has begun exploring whether domestic security arrangements encourage or discourage abuse. Research examining security arrangements and repression remain divided. One theory argues that centralization encourages abuse, allowing for a highly pro cient, lethal and well-coordinated coercive apparatus (Arendt 1951; Linz 2000; Loveman 1998). In this context, centralized security in totalitarian states such as Nazi Germany and the Soviet Union is associated with repression on an “unprecedented scale...the scale in numbers of lives lost, man-years in concentration camps, and the people arrested and subject to limitations of freedom of movement” (Linz 2000, p. 102). This theory asserts that centralized security provides unconstrained leaders with absolute control and direction over the use of force, allowing them to direct widespread terror against domestic populations (Arendt 1951; Linz 2000). Research on military dictatorships and repression nd similar patterns; security centralized 12Rejali(2007) provides a historical analysis of torture techniques, arguing that democracies tend to engage in “clean” torture, meaning torture that leaves no physical marks on detainees.
137 under the state encourages military involvement in domestic aairs and militarizes domestic policing and security (Davenport 2007c; Loveman 1998).13 However, this work focuses overwhelmingly on variation within authoritarian regimes. Centralized security is assumed to increase abuse within military and totalitarian states, but the behavior of centralized security in other contexts (including democracies) remains understudied. An alternative theory focusing on principal-agent dynamics suggests an opposite role for centrally controlled security: centrally controlled domestic security increases accountability and prevents “out-of-control” agents, whereas local security allows for predatory behavior and agency slack (Campbell and Brenner 2002; Krahmann 2010; Mitchell 2004). Local security like militias lack oversight by the national government, allowing them to engage in self-serving activities and pursue their own agenda without risk of punishment (Krahmann 2010; Mitchell 2004). For example, Village Guards engaged in drug tra cking in certain regions of Turkey, which set back central government interests (Biberman 2016). Local forces also engage in abuse to compensate for their lack of size, resources or professionalization, including training (Campbell and Brenner 2002). Studies of police violence against protesters similarly nd that police organizations lacking in resources and adequate training programs increase the likelihood of o cer abuse against protestors (Earl and Soule 2006). However, the police violence literature underscores the importance of training and resources in maintaining public order and does not nd that abuse inherently stems from localization.14 A last argument suggests that local forces fail to contain or discourage domestic oppositions to the government, further escalating violence and atrocities (Reyntjens 2009). It remains uncertain whether localization of security leads to widespread state repression or whether it simply fails to prevent predatory behavior or contain opposition movements. While providing dierent underlying mechanisms, arguments based heavily on principal-agent dynamics suggest that localization is associated with more abuse due to the lack of national oversight and lack of accountability to the national 13Examining variation in repression across authoritarian regimes, Davenport(2007 c, p. 500) nds that “political systems controlled by armed forces are more inclined to use repressive techniques which are more directly within the realm of their area of expertise−physical violence.” 14Some empirical research has found that community-oriented policing in particular appears to curb human rights abuse, community policing being where the police are highly localized, recruited from the same neighborhoods that they regularly patrol and are accountable to the local community (Blocq 2014; Brogden, Nijhar et al. 2013).
138 government. This logic runs contrary to cases of mass human rights abuses in totalitarian regimes, where centralized forces have the capabilities and the orders from on high to commit mass atrocities. In the next section, I consider widespread state repression as distinct from the abuses of extractive, unprofessional or prejudiced security personnel. I examine the in uence of domestic political institutions on the central government’s use (and misuse) of security forces.
5.3 theory: centralization and repression in limited democracies
Tounderstand and expand on the separate logics noted above, I consider mass state repression as distinct from more limited−albeit still atrocious−abuses by security forces who mistreat and abuse citizens not at the behest of a national leader but for their own predatory purposes. Given the disagreement over the role of security organizations and repression, I provide a new explanation of why centrally controlled domestic security forces commit widespread abuse, while local security forces tend more toward predatory abuse, and how varying domestic conditions in uence the behavior of security forces. I argue that centrally controlled domestic security forces are associated with widespread state (and state-directed) repression and that this behavior is most frequent in democratizing states rather than fully consolidated democracies or autocracies.15 In preventing or promoting widespread repression, centrally controlled security acts as a double-edged knife. More centralized security gives the national government control and direction over internal security forces, but also gives the national government the authority and the resources to repress its own population. I argue that when national leaders create and invest in central security institutions, they are acting in their strategic self-interest; creating coercive institutions over which they have more authority relative to local forces, allowing them to direct potentially unpopular or extralegal domestic operations and providing them with the authority to militarize internal security forces. Norms and rules 15I do not directly test the claim that localized forces are more likely to engage in non-state directed abuse, because of the lack of granularity in measures of state repression and lack of data on the behavior of various internal security forces. However, my claim corresponds with recent studies on militia violence, which similarly nd that localized forces engage in predatory violence against civilians due to their organizational characteristics (i.e. lack of unit cohesion) rather than due to orders given by the national government (Cohen and Nordås 2015). Case studies on militias similarly nd evidence that these forces often engage in predation against civilians that is not strategic from the perspective of the state and may further weaken or destabilize the national government (Francis 2005; Reid et al. 2011).
139 of non-intervention by the military, speci cally in democracies, can unintentionally lead to the adoption of internal centralized forces that have legal jurisdiction within the state and report directly to national political authorities. Central security allows national authorities to control, fund and deploy coercive organizations. Since the public lacks complete information and oversight over domestic armed forces−even in democracies (Krahmann 2010)−national political authorities can deploy these organizations to carry out statewide repressive policies. I discuss below the role of national control over security forces in driving state repression particularly in democratizing countries. Centralized organizations provide national governments with the authority to direct and militarize security for repressive policies.16 Militarization may happen at the local level as well, but militarization of centralized forces provides national leaders with a direct tool to repress rather than having to rely on local governments or local forces who may have other interests or loyalties. In addition, weak public accountability mechanisms in a country do not prevent the misuse of security forces by the national government but rather encourage the adoption of institutions outside of the public’s purview and local government’s control.17
National control. Central forces provide the national government with greater control than do local armed forces reporting to local administrations and having local loyalties.18 Beyond notable cases discussed in the totalitarian literature, modern democratic countries like Spain, France, the Netherlands, Argentina, Mali and India have internal security centralized under the national government (The Military Balance 2011).19 Most central governments−especially democratic ones−do not have the 16In certain cases, these centralized internal security forces could threaten the government itself, as in the case of gendarmeries assisting the military with a coup or−short of a coup−in uencing the national government (such as in Iraq, where the Badr Organization controls the Ministry of Interior rather than the Ministry of Interior controlling the Badr Organization). See Mansour, Renad and Felah A. Jabar. “The Popular Mobilization Forces and Iraq’s Future.” Carnegie Middle East Center. 2017. 17By public accountability, I mean the extent to which the domestic public has information over the national leaders’ policies and behavior, and can then reward or punish those leaders conditional on their actions. Stokes, Susan C. 2005. “Perverse Accountability: A Formal Model of Machine Politics with Evidence from Argentina.” American Political Science Review. 2005. 99: p 316. 18This is not to say that issues of agency slack do not exist among national forces but are likely less extreme than when national governments rely on local security. 19The United States and its National Guard is more an exception to this trend in centralized security, since the National Guard has dual accountability to the state and federal government and acts as a reserve force with the majority of guardsmen only part-time. See U.S. Code Title 32. https://www.law.cornell.edu/uscode/text/32.
140 same level of complete and unchallenged authority as the Soviet Union under Stalin or Nazi Germany. However, national governments even in democracies have a relatively easier time ordering operations when they have direct authority over armed forces rather than having to coordinate with local or regional authorities−authorities who may oppose and even violently oppose their orders or rule. France, Japan and Italy all established gendarmeries in response to resistance by regional authorities who wanted to maintain autonomy (Bayley 1990). Cambodia established a gendarmerie in 1994 “with a view to creating a gendarmerie or military police force directly answerable to the Prime Minister.”20 Since its creation, Cambodia’s gendarmerie has violently suppressed opposition party protests, and “regularly applied excessive force in these operations, injuring, sometimes seriously, students, workers, and others.”21 National authorities have more limited ability to order local armed forces to engage in abusive activity, especially if those local armed forces report administratively to local elites. In cases of community policing and civil defense guards, local armed forces may be less willing to commit abuse against their own community (Clayton and Thomson 2015). Community-oriented policing emphasizes the importance of local accountability and local transparency, public oversight mechanisms that central government o cers lack (Brogden, Nijhar et al. 2013). Even in civil con icts, communities establish local civil defense guards for the purpose of protecting civilians from abuse, such as the Aceh militias in Indonesia (Barter 2013). In addition to o cially reporting to local administrations, local armed forces can be loyal to local or regional leaders rather than to the national government. Communities in southwestern Sudan formed militias with strong allegiances to local tribal leaders, not national authorities (Blocq 2014). During and after the Iranian Revolution, defense units formed around local mosques and remained in the service of the mosque’s komitehs or local clerics even after Ayatollah Khomeini took power (Ahram 2011, p. 109). National authorities have greater authority over national armed forces compared to local forces who tend to be accountable to local leaders or their local community.
Resources and weapons. As national leaders have authority over central forces, national leaders can also draw on national funds and defense budgets to provide these forces with more resources for internal 20PoKempner, Dinah, ed. Cambodia at War. Human Rights Watch, 1995: p. 70. 21Human Rights Watch. 2015. “Cambodia: Commander Admits Partisan Use of Force.” https://www.hrw.org/news/ 2015/01/24/cambodia-commander-admits-partisan-use-force.
141 coercive operations. Central armed forces draw on a national base for revenue generation and receive nationally allocated funds, as opposed to local forces with local, more limited funding streams (Waxman 2009).22 For these reasons, local forces generally have lower budgets and resources than national forces. The lack of funds for local security can incentivize these forces to commit abuses against civilians as well, extracting resources for themselves from the local population. For example, Afghan Local Police in certain districts were “perceived locally as a criminal network involved in the forced extraction of zakat [obligatory payments under Islamic law] and kidnappings” (Goodhand and Hakimi 2014b, p. 40). However, this type of abuse is committed for extractive purposes rather than at the direction of the national government. While local forces draw resources from local funding and local populations, the national government can allocate defense budgets and resources to gendarmeries and national police. This, in turn, allows central forces to invest in heavier, more complex and more lethal equipment compared to local municipal police or civil defense guards (Lutterbeck 2004). While some municipal forces may diverge from this trend, for the most part central rather than municipal forces have allocated resources, organization and responsibility to eld and use more lethal armaments and large-scale weapons systems (Lioe 2011).23 For example, the French Gendarmerie use the same standard assault ri es as the military, in addition to armored vehicles and heavy weaponry, whereas local municipal police only have access to handguns and may not be armed at all.24 The national police force of South Africa−the South African Police Service (SAPS)−have access to grenade launchers, machine guns and other military weapons unnecessary for urban policing that allow a disproportionate level of violence by the state.25 The Mexican Department of Defense authorizes its federal police (Policía Federal) to obtain “war weapons” or weapons reserved for the military.26 The federal police’s lethal capabilities are associated with widespread physical abuses, 22This is not to say that municipal police are never well-funded or lack personnel, but only that national armed forces tend to have larger expenditures because they are nationally funded and draw on national defense spending. 23Even municipal forces buying surplus military weapons is more an unintended consequence of the military disposing unneeded equipment to generate funds. For a further discussion, see Grasso, Valerie Bailey. “Defense Surplus Equipment Disposal, Including the Law Enforcement 1033 Program.” Congressional Research Service. 2014. 24Police Weapons: France. Law Library of Congress. 2015. 25Police Weapons: South Africa. Law Library of Congress. 2015. 26Police Weapons: Mexico. Law Library of Congress. 2015.
142 including the torture and killing of Mexican citizens.27 Central security forces even draw on military personnel and military tactics, which increases the potential for lethality and brutality during domestic operations. For example, military o cers led Spain’s gendarmerie (the Guardia Civil) and trained those units in military-style tactics throughout the 1960’s and 1970’s during counterinsurgency campaigns against the Euskadi Ta Askatasuna (ETA), during which the gendarmerie resorted to mass imprisonments, beatings and other abuses against Spanish citizens.28 Similarly, the Argentine National Gendarmerie has the “lethality and mobility of armies.”29 Centrally controlled internal forces receive the coercive capabilities of the military, often without the constraints or training required for responsible domestic use of force. Municipal police may also receive military rearms, but the arming of a police unit within a limited jurisdiction or city suggests a more limited area for abuse. Centrally controlled internal forces functionally militarize internal security operations, expanding the scope, lethality and potential brutality of internal security activities under national leaders.
Regulations against internal military use. Finally, rules and norms that restrict use of the military internally can unintentionally motivate the national government to create parallel organizations with internal jurisdiction. The United States Posse Comitatus Act outlaws “willful use of any part of the Army or Air Force to execute the law unless expressly authorized by the Constitution or an Act of Congress.”30 This act does not prevent against the creation and use of other regular central forces for domestic purposes. The Unites States has actually granted domestic federal agencies broad authority over national security operations and criminal investigations (Cunningham 2004; Doyle 2000). The Federal Bureau of Investigation legally established a domestic Counterintelligence Program (COINTELPRO) during the 1950’s and 1960’s, solely to “expose, disrupt, misdirect, or otherwise neutralize the activities of protest 27Department of State Country Reports on Human Rights Practices for 2016. The Human Rights Practices on Mexico notes that “the National Human Rights Commission accused federal police of executing 22 persons after a gun ght in May 2015 near Tanhuato,Michoacan.” 28de Jong, Michel. “Spanish Security Forces, Anti-terrorism and the Internal and External Security of Spain, 1959-1992.” NL ARMS Netherlands Annual Review of Military Studies (2016): 325-349. 29Pion-Berlin, David, and Harold Trinkunas. “Latin America’s growing security gap.” Journal of Democracy 22.1 (2011): 39-53. 30The Posse Comitatus Act, 18 U.S. Code, Section 1385. This law was created in 1878 to prevent the military from policing elections in former Confederate states. See Matthews, Matt. The Posse Comitatus Act and the United States Army: A Historical Perspective. Army Command and General Sta College. Fort Leavenworth, Kansas. 2006.
143 groups that, in the FBI’s view, engaged in actions that threatened the security of the United States” (Cunningham 2004, p. 37). Since the FBI is not part of the regular armed forces, its domestic deployment of forces does not violate the Posse Comitatus Act. The United States can internally deploy other forces as well. The United States president can deploy the National Guard for natural disasters, civil unrest and “homeland defense activities” that include when “there is a rebellion or danger of a rebellion against the authority of the Government of the United States [and when] the President is unable with the regular forces to execute the laws of the United States.”31 Similar to gendarmeries, the National Guard was created with dual security obligations: to augment military forces and to maintain domestic order.32 Democracies have regulations or norms against the deployment of the military on domestic soil, but have overcome these regulations through developing central armed forces separate from the military and with jurisdiction over internal security operations (Lutterbeck 2004).33 Centrally controlled security organizations are more militarized, have broad internal jurisdiction and direct reporting to national authorities, without being accountable or transparent to the public. This dynamic between internal armed forces and repression is particularly prevalent in democratizing states, which have rules and norms curbing involvement of the military in domestic aairs but only limited constraints over the executive and allow for public contestation of their rule. Lack of public or local regulation allow these national leaders to politicize central armed forces and use these organizations to target domestic challenges, including targeting popular protests or voters during elections. Local armed forces may still commit abuses against civilians but less systematically. Centrally controlled internal forces provide national leaders with the authority and coercive capability to repress. Furthermore in democratizing states, national leaders have the means and the incentives to repress. 31Under Title 10 of the U.S. Code, the president has the authority to domestically deploy the National Guard if the U.S. needs to suppress a rebellion or in the case of foreign invasions. https://www.law.cornell.edu/uscode/text/10/12406. State authority of the National Guard occurs until Title 32. https://www.law.cornell.edu/uscode/text/32/904. 32During the early twentieth century in the United States, proponents of the militia system argued that “a trained and disciplined National Guard could augment the army during wartime, and during peacetime would allow state governments to maintain order during strikes without the need to call for federal troops” (Stentiford 2002, p. 9 - 10). 33Head and Mann(2016) argue that democracies are also decreasing regulation over the deployment of the military on internal soil. While true in the discussed case of Japan, Germany, and Canada (Head and Mann 2016), this argument does not address the broader trend in creating centrally controlled internal forces with wide-reaching authority.
144 5.4 hypotheses
The above discussion suggests several testable hypotheses regarding the relationship between states with centrally controlled internal security, limited national accountability and repression. In general, I anticipate that countries with more centrally controlled security personnel and limited democratic mechanisms are more likely to have widespread repression. Central security forces (relative to local security forces) act on the orders of the national governments and have greater capabilities to repress. And national leaders faced with limited accountability mechanisms have the added incentive to repress public contestation. Centrally controlled internal forces should be associated with greater variance in repressive behavior as well: committing widespread repression when under orders by the national government to repress but more constrained when under national orders to protect civilians and behave in a restrained manner. In contrast, local domestic security should engage in less widespread but more self-serving predatory abuse.34 While this project does not directly investigate predatory behavior of armed forces, my theory has other important testable observations. I should observe countries with more centrally controlled security as having more variance in repression, and more likely to engage in widespread repression compared to countries with relatively greater local security. Countries with strong and developed public accountability mechanisms should constrain the repressive role of central security though, as accountability to the public will constrain national leaders from giving orders to repress.35 This theory of more centralization/more coercion has other testable observations, such as centrally controlled security being associated with abuse across a country and more variance in abuse (conditional on whether the national leader does or does not give the order to repress), whereas local security should be associated with more localized and self-serving acts of abusive behavior and less variance, since their behavior is not dependent on the orders of a national government. For example, Turkey’s Village Guards committed atrocities, but these were limited to the southwest regions of Diyarbakir, Siirt and Hakkari (Biberman 2016, p. 11).36 Understanding the relationship 34Due to the nature of local predatory abuse, I would expect that these instances of abuse are reported less frequently and receive less coverage compared to repression directed by the national government. 35DeMeritt(2012) provides a formal theory for this logic that security forces act on the national leader’s order to repress. 36Subnational studies are well positioned to explore variation in the abusive behavior of state security within countries as well. However, I am interested in this chapter in patterns of repression given a state’s organization of security, which
145 between security organization and state repression is one link in understanding the broader dynamics of state-challenger dynamics, but a highly important and relatively under-theorized link.37 I focus on the eects of security centralization on state repression, including the degree to which constraints over national leadership, public accountability and judicial independence condition these eects. I rst test whether centrally controlled security provides the national government with greater coercive ability in line with Arendt(1951) and Linz(2000), whose works on totalitarian states suggest that centralization generates a highly coordinated, highly pro cient coercive apparatus. I argue, though, that it is not totalitarian regimes that are most likely to use centrally controlled internal forces to repress but national governments facing limited and imperfect public constraints. In weak democracies or democratizing states, national leaders have the means and incentives to repress the population to decrease their likelihood of being punished at the polls or facing popular protests.38
Hypothesis 1: Countries with relatively more centrally controlled domestic security compared to local security are likely to have higher levels of state repression.
Not all governments with centrally controlled internal security may direct their internal security forces to repress or need to use these forces to eliminate internal challengers. Leaders in authoritarian regimes may be able to deploy the military internally to repress, not needing to rely on parallel internal security forces, and leaders in the most developed, democratized states are unable and unwilling to repress due to eective public constraints and accountability.39 Therefore, I expect a direct eect between centralization and repression, but I also expect that democratic mechanisms condition this eect. I rst test whether the eect of centralization is conditional on democratic mechanisms and, second, whether those mechanisms lead to a selection of certain security arrangements. These hypotheses test the claim that in countries with public but imperfect accountability mechanisms and constraints helps us understand which types of arrangements promote less repressive states when accounting for a range of country characteristics. 37For a broader discussion of state-challenger dynamics, see Conrad and Moore(2010); Davenport(2007 b); Ritter and Conrad (2016). 38Acemoglu, Robinson and Santos(2013); Hafner-Burton(2008). 39Edmunds, Timothy. “What Are Armed Forces For? The Changing Nature of Military Roles in Europe.” International Aairs 82(6): 2006, p. 1059075.
146 over national leaders−including limited in uence of the legislature and limited independence of the judiciary−encourage the adoption of parallel internal security organizations. In these limited democracies, national leaders may repress its citizenry through limiting mass mobilization and other forms of popular contestation, such as restricting protests and other forms of mass mobilization.40 Rather than relying on the military, national leaders use national police and gendarmeries to eliminate domestic challengers, limit protests and violently suppress opposition voters during elections. In contrast, national leaders who are unconstrained and unaccountable to the public (as in authoritarian regimes) can more easily turn to the military or their own special forces (such as praetorian guards) to repress domestic dissent.41 I anticipate that weakly democratic countries are more likely to use centrally controlled internal security for widespread repression, whereas centrally controlled internal security in more autocratic countries acts as a counterbalance to the military or−in full democracies−as a way to consistently apply law and order across the country.42 Therefore, gendarmeries dier in their repressive behavior across dierent types of political institutions, where gendarmeries are more repressive in weakly democratic countries but less repressive in other types of regimes. In addition, repression may be more likely in partially democratic countries with centrally controlled internal security forces, because centrally controlled internal forces are more likely to repress and partially democratic countries are more likely to create such forces. I test these selection dynamics as well.
Hypothesis 2: Relatively greater centrally controlled internal security compared to local security is associated with higher levels of state repression when national leaders are only partially accountable to the public and other branches of government.
Alternative dynamics such as demographics and ethnic dynamics may in uence the state’s security arrangements and subsequent repressive practices. Countries, both democratic and autocratic, may 40I am primarily interested in whether weak democracies use centralized security to repress the public, but this theory has applications to theories of autocracies and coups as well. If autocracies are more likely to coup-proof, then I would expect that autocracies are more likely to have gendarmeries for the purpose of developing a check on the military (Pilster and Bohmelt 2012). Gendarmeries may even be correlated with military purges or previous coup attempts. 41Alternatively, autocracies may have centralized internal forces but use them to counterbalance the military (Pilster and Bohmelt 2012). In this case, the presence of internal centralized security in authoritarian regimes would be associated with a higher likelihood of coup-proo ng, but not necessarily with violence committed against large numbers of the public). 42This argument is in line with the More Murder in the Middle hypothesis, where repression is lowest among highly consolidated authoritarian regimes and full democracies (Fein 1995).
147 tolerate and even encourage state forces to repress minorities (Davenport, Soule and Armstrong 2011). Security forces composed of a particular ethnic group (“ethnically stacked” forces) may allow a country’s national and local leaders to use coercive forces for repressive and discriminatory activity without fear of public reprisal (Belkin and Schofer 2003). Even after apartheid in South Africa, white Afrikaner males remained overrepresented in national police leadership positions, whereas the majority of the force’s rank-and- le were black South Africans.43 South African internal security forces still routinely engage in violence against poor black communities, despite eorts at reform and to diversify the forces.44 Depending on the ethnic dynamics between security forces and the population they operate within, security forces may be more willing to commit abuse against civilians, and national leaders may be more willing to give such orders. While not testing recruitment practices directly, discrimination in recruitment could encourage abuse against speci c ethnic groups otherwise unrepresented or underrepresented in security. This suggests that centrally controlled internal security could be associated with widespread repression when a country has a population excluded from the national government and central security forces. Central security forces can enforce discrimination and abuse minority groups at the behest of national leaders. Therefore, I additionally test whether ethnic discrimination or ethnic fractionalization condition the eects of centralization.
5.5 research design
In testing the relationship between security centralization and state repression, I use a cross-sectional time-series research design of 149 countries from 1976 through 2010. The unit of analysis is the country-year, consistent with other studies on both state repression and state security (Carey, Mitchell and Lowe 2013; Nordas and Davenport 2013; Pilster and Böhmelt 2011). As discussed throughout this dissertation, alternative subnational approaches would better examine variation between security forces and repressive practices within a country, thus holding country-level and international-level attributes 43Rauch, Janine. “Police Reform and South Africa’s Transition.” Crime and Policing in Transitional Societies. 2000: 119-126. 44Onishi, Norimitsu. “Police in South Africa Struggle to Gain TrustAfter Apartheid.” The New York Times. 13 August 2016. https://www.nytimes.com/2016/08/14/world/africa/south-africa-apartheid-police-killings. html?_r=0.
148 constant. However, my cross-national approach allows for an exploration of the interrelated factors of domestic institutions (coercive and non-coercive), characteristics of the international system, and global trends in repressive practices. Though my data are not ne-grained enough to provide information on which security force engaged in abuse, information on internal forces and their personnel allows me to test how variation in security arrangements in uences trends in state repression.45 My approach allows for a broader examination of the relationship between global shifts in state security arrangements and abusive practices of national governments. The timeframe of this analysis is from 1976 through 2010, since the scope of data on state repression begins in 1976 and data on security forces end in 2010. I measure security centralization as personnel in centrally controlled internal forces over total internal security forces (which include both local and central forces). I use the ratio of security forces since I am interested in the extent to which security forces are more or less centralized, but countries can simultaneously employ both centrally controlled and local security personnel (Bayley 1990). I provide alternative measures of internal security in the Appendix, including separate measures for central/local security personnel and binary indicators for central/local organizations (see Table 5.8 and Table 5.11 in the Appendix), along with separating local organizations into regular and irregular armed forces (see Equation 2 in Table 5.6 and Table 5.11). I use the Political TerrorScale (PTS) data of Amnesty International as my primary measure of state repression, employing Fariss(2014)’s Latent Human Rights Score and PTS State Department measures as secondary measures. 46
Dependent Variable: State Repression
Since I am interested in physical abuse conducted by the security apparatus on behalf of the government, I measure state repression as physical integrity rights violations by the state: abusive actions by state forces that include “arrests and political imprisonment, beatings and torture, extrajudicial executions, 45Studies that have directly examined the behavior of security forces due so at the expense of a more limited scope, focusing only on one type of security force or one country rather than comparing the security apparatus and abuse across countries (i.e. Cohen and Nordås(2015); Eck(2015); Stanton(2015)). 46Fariss(2014)’s continuous measure is more appropriate in models with the Heckman selection correction that require either binary or Continuous dependent variables in the second equation. Heckman, James J. “Sample Selection Bias as a Speci cation Error.” Econometrica 47(1): 1979, p. 153-161.
149 mass killings, and disappearances” (Fariss 2014, p. 297).47 I measure these violations rst using annual data from the Political Terror Scale (PTS) and, second, through Fariss(2014) for each country. 48 PTS is a 5-point ordinal scale of state violations of physical integrity rights, providing two separate measures of repression based on two primary sources: an Amnesty International scale based on Amnesty International annual country reports and a State Department scale based on State Department annual country reports (Gibney et al. 2012). In both ordinal scales, lower scores represent lower levels of repression.49 For example, Level 1 represents countries “under a secure rule of law, [where] people are not imprisoned for their views, and torture is rare or exceptional. Political murders are extremely rare” (Gibney et al. 2012). Canada and Japan receive a PTS score of 1 for almost all years under both Amnesty and State Department measures. In contrast to countries with a relatively secure rule of law and limited physical integrity rights violations, countries with PTS scores of 4 or 5 have “murders, disappearances, and torture [as] a common part of life” (Gibney et al. 2012). In countries with terror scores of 5, physical integrity rights violations have expanded to the entire population. Such cases cover Rwanda from 1994 through 1998 and Colombia throughout the 1990s. Both State Department and Amnesty International reports may introduce dierent biases though. State Department Human Rights reports exclude the United States from its sample and may rank countries aligned with the United States less harshly on human rights abuse, whereas Amnesty International may report less harshly on socialist regimes (Poe, Carey and Vazquez 2001). Therefore, I estimate two separate models with these two measures. I use the PTS-Amnesty measure as my primary dependent variable and, in a second set of models, use the PTS-State Department measure for robustness tests. My theory anticipates more repression with centralized security in limited democracies or democratizing regimes. If socialist states also tend to be limited democracies, Amnesty International Reports might introduce a downward bias in my ndings, making it more di cult to nd a relationship 47Alternatives like UCDP’s One-sided Violence Dataset measure intentional attacks on civilians by governments. However, the One-sided Violence Dataset only covers the post-Cold Warera (1989-2015) and only countries engaged in intrastate armed con icts (59 countries), excluding the vast majority of countries that have internal security forces deployed in peacetime. Eck, Kristine, and Lisa Hultman. “One-sided violence against civilians in war: Insights from new fatality data.” Journal of Peace Research 44(2). 2007: 233-246. 48Gibney, Mark, Linda Cornett, Reed Wood, Peter Haschke, and Daniel Arnon. 2016. The Political Terror Scale 1976-2015. Retrieved May 2017 from the Political TerrorScale website: http://www.politicalterrorscale.org. 49Please see the Appendix for the interpretation of each level of the Political TerrorScale.
150 between centralization and state repression. In contrast, the use of State Department reports may in ate the relationship between certain regime types and repression if reporting relatively more harshly against leftist or socialist regimes. Given these concerns, I rely on a third measure of state repression as a robustness check. Fariss(2014)’s Latent Human Rights Protection Scores provides a scale for human rights abuse that incorporates existing country-level data and the changing, stricter standard of human rights over time.50 This measure of state repression is estimated using an item-response theory model, using standards-based variables (including the Political TerrorScale) and events-based variables of state repression (Fariss 2014, p. 305). This provides a continuous latent variable of human right protection, where higher and more positive scores imply greater protection of human rights (less repression).
Main Explanatory Variable: Security centralization and localization
I measure centrally controlled security as the ratio of personnel in centrally controlled security forces over total internal security forces.51 I use personnel within gendarmeries for each country-year as a measure of centrally controlled security and provided from my original internal security force dataset produced for this dissertation.52 As in previous chapters, I de ne gendarmeries as not just forces explicitly named gendarmeries but those internal armed forces that (1) have the authority to exercise jurisdiction over the civilian population, (2) are administratively under the Ministry of Interior and/or Ministry of Defense, (3) legally separate from the regular armed forces and local civilian police, and (4) legally sanctioned to participate in both domestic policing and internal security across the country. I use militia personnel as a measure of local security. I de ne a country as having militias if the country has domestic security organizations that are (1) locally managed and locally organized within the territorial jurisdiction of the state, (2) armed and having the ability to engage in violent acts and 50 As suggested by Fariss(2014, p. 300) in his analysis of standard-based measurements of repression like the Political Terror Scale, “overall, the standard of accountability becomes more stringent as the U.S. State Department and Amnesty International look harder for abuse, look in more places for abuse, and classify more acts as abuse.” 51(Central Security) ÷ (Central + Local Security) 52Please refer to “Chapter 2: The Internal Security Forces Dataset” for a fuller description and discussion of the data, including data collection, descriptive statistics and examples.
151 active low- to mid-intensity combat, and (3) legally separate from regular armed forces and local civilian police. I use a continuous measure of local security−total personnel in militias−to capture the size of local security. For robustness checks, I distinguish between two types of militias: irregular militias and regular or state-managed militias. Irregular local militias have no evidence of being formally managed and standardized under a state institution, whereas regular or state-managed militias are formally managed and standardized under a state institution (either local or regional). In robustness checks using separate measures for central and local security, I log transform all security personnel measures as they are highly skewed. Finally, I include in the robustness checks a binary indicator for whether the country has any militias or gendarmeries operating within it for that year. I include covariates to condition the eect of security arrangements on state repression. In particular, I anticipate that public accountability of the national government in uences the relationship between security arrangements and state forces. Accountability mechanisms have been shown to in uence the repressive policies of the national government (Davenport and Armstrong 2004) and therefore should in uence the organization and operations of internal security forces. I include a measure for constraints on the executive (a 7-point score) to capture the in uence of public accountability over national leaders (Marshall and Jaggers 2005), since I am interested in how public accountability modi es internal security forces but need to avoid measures that incorporate repression or political violence (Hill and Jones 2014).53 The constraints on the executive score directly measures the extent to which the executive must defer to accountability groups (Marshall and Jaggers 2005, p. 24), speci cally whether “accountability groups have eective authority equal to or greater than the executive in most areas of activity.” I use domestic legal systems as an alternative measure for accountability, as strong independent judiciaries provide a check against the policies of national leaders and the structure or use of security forces within a country (Mitchell, Ring and Spellman 2013).54 53Many alternative measures for accountability such as “competitiveness of participation” and “executive recruitment” (along with the aggregate polity scores) incorporate state repression (i.e. whether there is imprisonment of political opponents.) See Hill and Jones(2014, p. 677). This is also true for the World Bank’s Voice and Accountability Index. http://info. worldbank.org/governance/wgi/pdf/va.pdf. 54The independent judiciary score is an ordinal measure from Cingranelli and Richards(2010) that indicates the extent to which the judiciary is independent from other political authorities, with a 0 indicating “no independence,” 1 indicating “partial independence” and 2 indicating “general independence.”
152 Rather than accountability mechanisms modifying the expansion and operations of internal security forces within certain regimes, it might be that accountability mechanisms in uence the creation of these forces.55 Democracies may be more likely to create gendarmeries as a way to consolidate national authority when operating under judicial and legislative constraints, speci cally laws restricting the military or municipal police from engaging in internal security operations (Bayley 1990; Lutterbeck 2004). The classical French model of gendarmeries also spread across continental Europe during the nineteenth century and is typical of Francophone states worldwide (Gobinet 2008). Therefore, widespread repression might be more likely in democracies with centrally controlled internal security than non-democracies with similar forces, because democracies are simply more likely to create these forces. I investigate this claim through a selection model, incorporating variables that likely in uence the creation of gendarmeries: how the state gained independence, whether it was a former French colony, and whether coercive or popular challenges occurred prior to the establishment of the centralized force.56
Controls
In line with existing research on security forces and state repression, I control for ethnic and demographic characteristics of the country, wealth, coercive challenges and previous repressive policies by the national government. These policies include control and success from prior eorts at subjugation, including preventing certain groups from serving in the government and having a minority or ethnically stacked security organizations. Ethnic characteristics of a country could in uence the eects of security arrangements on repression (Barter 2013; Blocq 2014). Security forces may be divided along ethnic lines or exclude a minority group from recruitment. In countries with regionally autonomous ethnicities, local armed forces could form along ethnic lines and exacerbate underlying ethnic tensions. While not abusing civilians at the behest of a national government, these local armed forces could of their own accord engage in sectarian violence against civilians. In contrast, national armed forces may have national and more representative recruiting 55As discussed in the previous chapter, this includes creating internal security forces from other armed organizations, including former rebels, former military units or former autonomous civil guards. 56Information on state independence and French colonies is from Paul R. Hensel’s “ICOW Colonial History Data Set” (2014), available at http://www.paulhensel.org/icowcol.html.
153 practices, whereas local armed forces do not.57 Under these dynamics, national armed forces would actually be associated with less abuse. I include two measures for the ethnic makeup of a country: one from Fearon and Laitin’s ethnic-linguistic fractionalization (ELF) index and another measure for ethnic discrimination from the Ethnic Power Relations Dataset (Cederman, Wimmer and Min 2010; Vogt et al. 2015) consisting of the proportion of ethnic populations excluded from central state power. The ELF variable measures a country’s ethnic diversity. However, ethnic diversity may not necessarily condition the eect of security arrangements on repression. The inclusion or exclusion of ethnic groups (rather than ethnic diversity generally) into government may matter more for how security forces interact with civilians, including ability for certain groups to participate in certain security forces. I use Cederman, Wimmer and Min (2010)’s ethnic exclusion variable to account for such dynamics.58 For information on demographics, I use total population measures from Gleditsch(2002). 59 I include information on youth bulges from the United Nation’s Population Division World Population Prospects, following Nordas and Davenport (2013) and Urdal(2006). 60 This measure of youth bulge consists of “large cohorts in the ages 15 to 24 relative to the total adult population” (Urdal 2006, p. 608). Governments facing a youth bulge may both engage in more repressive activity and attempt to recruit more youths into security institutions, simultaneously reducing the supply of potential dissidents and increasing coercive capacity. Theories of popular dissent suggest that greater domestic challenges to the state encourage repression; termed a Law of Coercive Responsiveness by Davenport(2007 a, p. 7).61 Domestic challenges may in uence the type and expansion of certain internal security forces. For example, widespread protests 57This is not always the case though. For example, Irish recruits in the British Armed Forces were underrepresented despite no policy explicitly barring them, though this occurred in part because of threats by Irish dissidents. Riegal, Ralph. “Number of Irish recruits joining British Army stable again following dissident republican death threats.” Belfast Telegraph. January 1 2015. 58Global data on the demographics of internal security forces and recruitment practices would provide a more precise measure of these factors but so far no such data exist (though it is a fruitful next avenue for research). Paul Johnson and Ches Thurber provide ethnic information on militaries of the Middle East in “The Security-Force Ethnicity (SFE) Project: Introducing a new dataset” (Conflict Management and Peace Science, 2016) but their data only cover the Middle East and militaries rather than oering global coverage and information on paramilitaries/internal security forces. 59Total population measures from Gleditsch, Kristian S. 2002. “Expanded Trade and GDP Data,” Journal of Conflict Resolution 46: 712-24. 60The timeframe of analysis is from 1976 through 2000 for equations that include youth bulges, since the scope of data on youth bulges ends in 2000. 61Also see Conrad and Moore(2010); Davenport(2007 b).
154 against national leaders might incentivize the national government to expand and militarize its national police, allowing the government to more eectively repress protestors through mass arrests and beatings (Soule and Davenport 2009). To account for these coercive responses, I include information on popular dissent from Banks’ Cross-National Time-Series Archive (Banks and Wilson 2016). I include annual country data on the count of riots, general strikes and anti-government demonstrations from Banks and Wilson(2016). I rely on an indicator variable for intrastate violent con ict from the Uppsala Con ict Data Program (UCDP) (Gleditsch et al. 2002a). Finally, I include an indicator variable for interstate violent con ict, measured as fatal militarized interstate disputes.62 As these factors might lead to the initial creation of central security forces, I include these variables in the rst-stage equation of the selection model. Finally, a country’s economic development and previous repressive episodes may in uence its security arrangements and current repressive practices. For measures of economic development, I use GDP per capita data from the Penn World Table (Feenstra and Timmer 2015) and re ned by Gleditsch(2002). 63 I include models with and without a one-year lag of the dependent variable, excluding the lag from xed eect models to avoid bias. Prior policies of repression by the state are likely associated with future state policies of repression (Davenport 2007b). Yet lagged dependent variables can underestimate the eects of other variables while overestimating the eect of the lagged dependent variable in cross-sectional time-series data with relatively few time periods.64 Therefore, I exclude lagged dependent variables in models that include country and year xed eects to minimize the risk of the Nickell bias.
Model Estimation
Following Nordas and Davenport(2013) and other studies using the PTS as the dependent variable, I use an ordered probit model since PTS measures are ordered on a 5-point scale. I use time dummies 62See Palmer, Glenn, Vito D’Orazio, Michael Kenwick, and Matthew Lane. 2015. “The MID4 Data Set: Procedures, Coding Rules, and Description.” Conflict Management and Peace Science. Forthcoming. 63Penn World Tableraw data are available for download from www.ggdc.net/pwt. 64 Achen, Christopher H. “Why lagged dependent variables can suppress the explanatory power of other independent variables.” Ann Arbor. 2001. Beck, Nathaniel L., Jonathan N. Katz, and Umberto G. Mignozzetti. “Of Nickell bias and its cures: Comment on Gaibulloev, Sandler, and Sul.” Political Analysis 22(2). 2014: 274-278.
155 to control for year-speci c eects in these models, accounting for changes within a given year like shifts in the detection of repression by nongovernmental organizations and shifts in the global economy (Sikkink 1993).65 I include an indicator variable for the Cold War since characteristics of the international system−such as bipolarity and superpower involvement in a country’s domestic politics−could in uence the relationship between security arrangements and repression. Krahmann(2010) claims that the Cold War drove the development of “centralized warfare states,” while other research suggests the opposite; that interference bt superpowers in domestic countries and decolonization movements during the Cold War drove localization of security (Fearon 2003; Hironaka 2009). In either case, the Cold War and end of the Cold War introduced broad changes in the international system and organization of security.66 I lag all independent variables by one year (t − 1) to limit potential simultaneity bias, since increased repression in one year may be associated with expansion of existing forces. In models without xed eects, I incorporate a lagged dependent variable of repression to account for bureaucratic inertia (Davenport 2007b) and for the possibility that previous repressive episodes are associated with the creation or expansion of certain security organizations. Regional dummy variables are used in all speci cations to account for regional dynamics that could in uence both repression and the organization of internal security, including regional con icts, regional instability and the diusion of security organizations within a region. For example, gendarmeries are considered a growing internal security organization within Europe (Lutterbeck 2004). Standard errors are clustered by country to account for heterogeneity across countries.67 Only certain regimes may initially generate certain security organizations. As mentioned previously, democratizing and democratic countries may be more likely to create centrally controlled internal forces (Lutterbeck 2004). First, countries gaining independence from former colonial powers with gendarmeries−such as France−may be more likely to create centrally controlled internal organizations. 65I address the in uence of time through two other speci cations in robustness checks. First, I use Fariss(2014)’s dependent variable that incorporates the changing human rights standards and detections over time (see Table 5.6, Table 5.7 and Table 5.8 in the Appendix). Second, I incorporate country xed eects into this model of latent human rights (Equation 5, Table 5.6 in the Appendix). 66While acknowledging that the indicator variable is a relatively coarse measure, understanding the relationship between the Cold War, internal security arrangements and coercive practices is beyond the scope of this chapter. 67For example, countries with greater centralized security forces may have more variance in repression compared to countries with less centralized security.
156 Second, national governments needing to consolidate control in the presence of popular dissent and institutionalized public accountability mechanisms (such as elections) could generate gendarmeries (Ahram 2011; Bayley 1990). I use a Heckman Selection model to address these potential selection eects. In the rst stage equation, I estimate the likelihood of a centrally controlled internal security organization being formed and in the second stage estimate how countries with relatively larger central security forces in uence the likelihood of widespread repression.68 Even when accounting for the creation of centralized internal security, countries with relatively larger gendarmeries have a higher likelihood of widespread repression.69
5.6 results
Evaluating Hypothesis 1: Security arrangements and repression
As shown in Table 5.1 and Table 5.2, I nd tentative support for my rst hypothesis that countries with relatively more central security forces compared to local security forces are more repressive.70 Countries with more personnel in central security forces relative to local forces are associated with a higher likelihood of widespread repressive practices. The coe cient estimate on central security personnel is positive and statistically signi cant across model speci cations. The consistency of the positively-signed coe cient estimate and signi cance is notable, since internal security force data are relatively noisy with sometimes few year-to-year changes among countries.71 This result remains consistent whether measuring central internal security as the ratio of security force personnel or simply the presence of central security organizations.72 Finally, these ndings remain robust to measuring state repression using State 68Kennedy, Peter. A guide to econometrics. MIT press. 2003: p. 270 69An alternative method would be an instrumental variable model, with “former French colony” as the instrumental variable since former French colonies should be associated with having gendarmeries but not with state repression. Unfortunately, this approach requires being able to include regressors for “former French colony” that are limited and beyond the contemporary scope of the data project. 70Table 5.5 of the Appendix provides descriptive statistics of these variables. 71This result holds in the xed eects model with latent human rights as the dependent variable (where the negative sign on the security force coe cient indicates that more centrally controlled security is associated with more abuse / fewer protections of human rights). See Table 5.6 in the Appendix. 72For binary indicators of internal security forces, see Table 5.10 and Table 5.8 in the Appendix. These equations also produce results consistent with continuous measures of gendarmeries and militias.
157 Table5.1: Central internal security and political terror (Amnesty)
Eq. 1 Eq. 2 Eq. 3 Eq 4
Central Security Ratio 0.143** 0.245** -0.132 -0.192 (0.0650) (0.111) (0.109) (0.178) Centralized∗Executive Constraints 0.0707*** 0.111** (0.0261) (0.0438) Executive Constraints -0.0808*** -0.166*** -0.101*** -0.197*** (0.0143) (0.0222) (0.0166) (0.0267)
Excluded Population 0.272*** 0.471*** 0.285*** 0.490*** (0.101) (0.180) (0.0998) (0.178) GDP_ln -0.194*** -0.381*** -0.198*** -0.385*** (0.0375) (0.0656) (0.0370) (0.0643) Population_ln 0.300*** 0.564*** 0.305*** 0.568*** (0.0382) (0.0576) (0.0375) (0.0562)
Civil con ict 0.605*** 1.274*** 0.588*** 1.243*** (0.0759) (0.114) (0.0733) (0.109) Interstate con ict -0.0689 -0.117** -0.0672 -0.115** (0.0433) (0.0566) (0.0442) (0.0574) Protests 0.00563 0.0235** 0.00678 0.0254** (0.00753) (0.0113) (0.00747) (0.0110) Cold War -0.0506 -0.0611 (0.0470) (0.0481) PTS(A)(t−1) 1.083*** 1.078*** (0.0533) (0.0531)
Observations 3,728 3,926 3,728 3,926
Lagged DV Yes No Yes No Time Dummies No Yes No Yes Region Dummies Yes Yes Yes Yes Robust standard errors in parentheses, clustered on country. Covariates lagged one year. *** p<0.01, ** p<0.05, * p<0.1
158 Table5.2: Central internal security and political terror (Amnesty)
Eq. 1 Eq. 2 Eq. 3 Eq 4
Central Security Ratio 0.119* 0.210* 0.0826 0.142 (0.0652) (0.113) (0.0858) (0.149) Centralized∗Judicial Constraints 0.0372 0.0687 (0.0718) (0.120) Independent Judiciary -0.153*** -0.319*** -0.162*** -0.336*** (0.0385) (0.0664) (0.0459) (0.0791)
GDP_ln -0.200*** -0.393*** -0.199*** -0.392*** (0.0378) (0.0662) (0.0378) (0.0659) Population_ln 0.304*** 0.576*** 0.303*** 0.574*** (0.0404) (0.0628) (0.0406) (0.0628)
Civil Con ict 0.582*** 1.298*** 0.580*** 1.295*** (0.0727) (0.115) (0.0726) (0.115) Interstate Con ict -0.0396 -0.0811 -0.0396 -0.0812 (0.0475) (0.0623) (0.0474) (0.0621) Protests 0.00425 0.0223 0.00428 0.0224 (0.00959) (0.0154) (0.00957) (0.0153) PTS(A)(t−1) 1.106*** 1.106*** (0.0527) (0.0526)
Observations 3,451 3,644 3,451 3,644
Lagged DV Yes No Yes No Time Dummies No Yes No Yes Region Dummies Yes Yes Yes Yes Robust standard errors in parentheses, clustered on country. Covariates lagged one year. *** p<0.01, ** p<0.05, * p<0.1
159 Department PTS scores.73 These results suggest that countries with more centrally controlled security personnel relative to local security personnel are more likely to experience widespread repressive practices. This nding runs contrary to current expectations that centralized internal security prevents abuse and that repressive behavior occurs mainly through local forces like militias (i.e. Campbell and Brenner 2002; Mitchell, Carey and Butler 2014). In line with my theory on coercive capacity under constraints, states with more central security personnel (larger gendarmeries) appear more likely to engage in widespread abuse. Toprotect and secure their control, national governments can rely on a centrally controlled internal security organization as a means of domestic coercion. The national government is more able and perfectly willing to commit abuses against the domestic population and other domestic armed forces. Based on Equation 1 in Table 5.1, I estimate the predicted probability of a country reaching level 4 on the Political Terror Scale of repression during peacetime (engaged in neither civil con ict nor interstate con ict and all other variables held at their mean). In line with Nordas and Davenport(2013), I examine the predicted probability of a state at the level 4 of repression, as this is the point in the scale where physical integrity rights violations occur against large segments of the population and repressive activity is more lethal, including widespread extrajudicial killings, torture and disappearances. Countries with fewer than 10% of their internal forces composed of gendarmeries have around a 16% likelihood of reaching widespread state repression, whereas countries with over 80% of their internal forces composed of gendarmeries have a 19% likelihood of reaching widespread repression. Graph 5.1 shows the increase in the predicted probability of repression (PTS[A]=4) across security force centralization based on Equation 1 in Table 5.1. The other variables remain consistent with previous research on state repression.74 The coe cient estimates on population, civil con ict, ethnic discrimination and previous repression are positively and signi cantly associated with state repression. The negative and statistically signi cant coe cient estimates on executive constraints, judicial constraints and GDP suggest that more democratic and wealthier states are less repressive on average. As I discuss next, though, the dampening eect of democracy attenuates 73See Table 5.11 for PTS(S) as the dependent variable. 74Hill and Jones(2014).
160 Figure 5.1: Average marginal eects of centrally controlled internal security on the probability of state repression
under centrally controlled internal security. The role of the Cold War remains inconclusive based on the results found in Equation 1 of Table 5.1.
161 Evaluating Hypothesis 2: Internal security and accountability mechanisms
My second hypothesis proposes that states with relatively more centrally controlled internal security compared to local security are associated with higher levels of repression, especially when national governments are only partially accountable. I anticipated that central security is associated with the highest likelihood of widespread state repression when the country has partial or moderate accountability mechanisms over the national government. I argued that national leaders operating under limited public accountability are incentivized to invest and use national internal security, which gives them direct capability to repress opposition parties, voters and popular protests. My results are consistent with this theory; under moderate or partial levels of executive constraints, more central security (relative to local security) is associated with an increased likelihood of widespread repression.75 When the courts are only partially independent rather than not independent or fully independent, increases in central internal security are positively associated with widespread repression (Figure 5.3). Equation 3 and Equation 4 in Table 5.1 and Table 5.2 show the results for this interaction between central security and accountability. When the interaction is not included in the model, central security is a positive and signi cant predictor of state repression (at the 5% and 10% level). Including the interaction of accountability, the constitutive variable for central security is no longer statistically signi cant and even negative in the executive constraints model, whereas the interaction term between accountability and central security is positive and statistically signi cant at the 1% and 5% level. Graphing this interaction (Figure 5.2), holding con ict at 0, shows that the marginal eects of gendarmerie are positively associated with an increased likelihood of widespread state repression (PTS(A)=4) as constraints on the executive increase. This eect increases monotonically except at the highest level of executive constraints, where national leaders are fully subordinate to accountability groups.76 In other words, national leaders must be fully constrained−entirely accountable to others−before centrally controlled internal forces have a 75This result is consistent with latent human rights or the PTS-State Department score as the dependent variable (Table 5.6, Table 5.11 respectively). 76The highest level of executive constraints, level 7, means that “accountability groups have eective authority equal to or greater than the executive in most areas of activity,” whereas level 4 through level 6 means “substantial limitations on executive authority [but] the executive has more eective authority than any accountability group.” Marshall, Monty G., TedRobert Gurr and Keith Jaggers. “Polity IV project: Political regime characteristics and transitions, 1800-2002.” Dataset User’s Manuel. (2015). pg. 25.
162 slightly decreased eect on repression. This eect is similar but less pronounced with the judicial independence variable. When the judiciary is not independent, the marginal eect of a 1% increase in central security is associated with a 3% increase in the likelihood of widespread repression. When the judiciary is fully independent, the marginal eect of a 1% increase in centralized security is associated with only a 2% increase in the likelihood of widespread repression. Yet when the judicial system is partially independent, a 1% increase in central security is associated with a 4% increase in the likelihood of widespread repression. This eect is weakly signi cant at the 10% level. These results suggests that weak or limited accountability mechanisms over national leaders encourage the adoption of centrally controlled security for repressive purposes. In cases of slight or moderate limitations over national leaders, legislatures and courts may block policies that leaders want but not all orders or decrees. National leaders can still turn to internal centralized forces to implement policies or eliminate challengers−actions that constituents, courts and legislatures may otherwise disapprove of. Democracies in particular have stricter laws prohibiting the use of the military within the territorial jurisdiction of the state (Lutterbeck 2004). Unable to deploy the military internally, weakly democratic states can turn to alternative centralized forces like gendarmeries that have the coercive capacity of the military but the jurisdiction and authority to operate domestically. This nding provides strong support for arguments that democratic mechanisms do not eliminate human rights abuses but change the organizations and tactics for engaging in abuse (Conrad and Moore 2010; Cunningham 2004; Davenport 2007b). Alternatively, democracies may be more likely than autocracies to create centrally controlled internal forces, which−in turn−means that democracies are more likely to use those forces for repressive purposes. In contrast, autocracies may refrain from creating gendarmeries or national police if they already have substantial control over the internal use of the military. Therefore, I account for states that eld gendarmeries through a Heckman(1979) selectivity adjustment (selection hazard) and use latent human rights as a continuous measure of state repression. The choice to eld centrally controlled security organizations may aect whether centrally controlled security personnel in uence state repression. I use
163 Figure 5.2: Average marginal eects of centrally controlled internal security over executive constraints
Figure 5.3: Average marginal eects of centrally controlled internal security over judicial independence
164 an indicator for whether the country is a former French colony to identify the presence of gendarmeries in the rst stage model.77 As shown in the previous chapter, countries where the former colonizer had a gendarmerie often have gendarmeries themselves. This is particularly true among former French colonies, where 70% of former French colonies have gendarmeries.78 Sixty-two countries emerged from other former colonizers with gendarmeries but only 47% of those subsequently had a gendarmerie. The second stage model uses latent human rights as the dependent variable, where increased values of human rights imply decreased state repression. The rst stage probit for having a gendarmerie includes factors that encourage national governments to create centrally controlled internal security organizations.79 I estimate the inverse Mills ratio (the Heckman correction for selectivity) from this equation and include it in the second equation for human rights. In the rst stage model, the correlation coe cient ρ indicates that the errors between the rst and second stage equation are negative and statistically signi cant as anticipated; dynamics that determine creation of gendarmeries and dynamics that increase protection of human rights are negative and statistically signi cant. In other words, unobserved factors that increase the likelihood of states generating a gendarmerie are associated with states having decreased human rights. There is a statistically signi cant eect of selection. Having more centrally controlled forces relative to local forces is still positively associated with state repression (negatively associated with human rights). These results are relatively similar to those found in the main models. This nding implies that the eect of having more centrally controlled internal security is associated with more repression, even after accounting for the process by which national governments choose to adopt these forces.
77For identifying former French colonies, I rely on data from the Issue Correlates of War (ICOW) Colonial History Dataset. Paul R. Hensel (2014). “ICOW Colonial History Data Set, version 1.0.” http://www.paulhensel.org/icowcol. html. 78Twenty-four countries in the sample are former French colonies, and 17 of these countries have gendarmeries. 79Explanatory variables in the selection equation include domestic challenges (protests and number of ongoing con icts), international dynamics (interstate con ict and the Cold War) and domestic factors (size of population and violent independence).
165 Table5.3: Selection of Centrally controlled internal security and state repression (latent human rights)
Repression Selection Equation Equation
Ratio of Centralized Forces -0.128* (0.0678) Excluded Population -0.658*** (0.0880) GDP_ln 0.0505** (0.0222) Interstate Con ict 0.248*** 0.0177 (0.0439) (0.0520) Civil Con ict -0.927*** -0.167*** (0.0477) (0.0451) Executive Constraints 0.161*** 0.0281** (0.00978) (0.0112) Protests -0.0280*** 0.0213*** (0.00702) (0.00772) Population_ln -0.326*** 0.268*** (0.0278) (0.0180) French colony 1.296*** (0.0611) Independence Violent 0.0741 (0.0492) Colonizer Power -3.40e-07*** (7.09e-08) External Support 0.453*** (0.0820) Europe 0.110 (0.0768) Cold War -0.0806* (0.0486) Constant 2.252*** -3.255*** (0.197) (0.182)
λ -0.309*** (0.0565) ρ -0.417*** (0.064) Inverse Mills Ratio 1.264 (0.466)
Observations 3,956 Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
166 Cold War and ethnicity: Evaluation of other dynamics that condition centralized security
I examine other factors that may condition the in uence of internal security arrangements on state repression: the Cold Warand ethnic discrimination. I anticipate that centrally controlled internal security is associated with widespread repression when a country has a population that is ethnically discriminated against or speci cally excluded from the national government. Ethnic discrimination against a minority is national-level policy and, therefore, likely to be enforced through national security. Table 5.9 includes equations with ethnic discrimination as an interactive eect. My ndings suggest that ethnic discrimination and security arrangements do not have a conclusive interactive eect. Ethnic discrimination as state policy does not appear to condition the eect of central security on repression. As shown in Figure 5.4, average marginal eects of greater centrally controlled security remains positive even at lower levels of ethnically excluded populations. And as shown in Figure 5.5, the marginal eect of militias remain negative across dierent levels of ethnic exclusion. While ethnic exclusion is itself positively associated with widespread repression, ethnic exclusion from the national government does not seem to condition the eect of security arrangements. Representation across armed forces may matter more or, alternatively, national governments that discriminate against an ethnic population may do so through a variety of forces (both local and central forces). In either case, the relationship between ethnic discrimination and security arrangements requires further investigation. Finally, I examine the role of the Cold War in conditioning the eects of centrally controlled internal security on state repression. I estimate models of repression using the Amnesty International PTS as the dependent variable, examining both the ratio of central security over total internal security in addition to speci cations that include separate personnel measures for gendarmeries or militias. Equation 4 in Table 5.4 evaluates the in uence of centrally controlled internal security during the Cold War, while Equation 5 examines the in uence of centralized security after the Cold War. These results suggest that centrally controlled internal security has a statistically signi cant and positive association with repression during and after the Cold War. The coe cient estimate on the ratio of central security forces is consistently positive across model speci cations and is statistically signi cant. The average ratio of central security compared to local security is also signi cantly higher after the Cold
167 Figure 5.4: Average marginal eects of gendarmerie personnel over excluded ethnicities
Figure 5.5: Average marginal eects of militia personnel over excluded ethnicities
168 War, with 24% of security forces centralized after the Cold Warand only 20% of security forces centralized during the Cold War.80 These dynamics are likely attributable to the decreased number of militias and militia personnel after the Cold War, which−as discussed in the previous chapter−is probably due in part withdrawal of external support to these local forces. Contrary to Kalyvas and Balcells(2010), I do not nd a statistically signi cant eect for the interaction between the Cold War and organization of security. While the international environment contributes to internal security arrangements, there is insu cient evidence that changes in the international environment condition the eect of central security on repression. Central security is positively associated with repression across international systems, even though gendarmeries are more common and larger after the Cold War.
Summary of support for hypotheses
My ndings are suggestive of three important dynamics between internal security arrangements and state repression. First, having more personnel in centrally controlled internal security forces relative to local security has a positive correlation with state repression. Countries with more personnel in gendarmeries tend to have a higher likelihood of engaging in widespread repression. This result provides evidence of centralization as a double-edged sword. Increased control over internal forces grants the national government better oversight and coordination, but it also allows the national government to more eectively repress its own population. Second, my ndings show a positive and signi cant interaction between centrally controlled security and limited democratic accountability. As limited accountability mechanisms over national leaders increase−as in democratizing states−having more personnel in centralized forces is associated with an increased likelihood of widespread repression. Only when national leaders are fully subordinate to other accountability groups does the repressive eect of central armed forces decrease. Partial democratic constraints do not eliminate state repression but change the organizations that national leaders use to repress. Third, I showed that the repressive eect of centrally controlled security holds even when accounting for the selection of these organizations. Even after incorporating dynamics of selection into these models, centrally controlled internal security is still 80A two-sample t-test shows that the ratio of centralized security after the Cold War is higher than during the Cold War at the 0.05 con dence level.
169 Table5.4: Centrally controlled internal security, the Cold War and state repression (PTS[A])
(1) (2) (3) (4) (5) Cold War Post Cold War Sample Sample
Ratio of Central Forces 0.182** 0.311** 0.176 0.224* (0.0724) (0.125) (0.171) (0.118) Gendarmerie_ln 0.0136** (0.00648) Cold War -0.0200 -0.135 -0.00326 (0.0561) (0.0938) (0.0605) Ratio of Central Forces ∗ Cold War -0.110 -0.221 (0.113) (0.172) Gendarmerie_ln∗Cold War -0.00837 (0.00957) Militia Personnel_ln -0.00754 (0.00623) Militia Personnel_ln∗Cold War -0.00384 (0.00741) Controls Executive Constraints -0.0806*** -0.162*** -0.0775*** -0.164*** -0.157*** (0.0144) (0.0221) (0.0134) (0.0340) (0.0252) Excluded Population 0.268*** 0.440** (0.101) (0.181) GDP_ln -0.194*** -0.374*** -0.169*** -0.282*** -0.360*** (0.0375) (0.0645) (0.0299) (0.0872) (0.0565) TotalPopulation_ln 0.300*** 0.561*** 0.268*** 0.382*** 0.593*** (0.0383) (0.0570) (0.0408) (0.0955) (0.0833) Civil Con ict 0.605*** 1.254*** 0.615*** 1.207*** 1.420*** (0.0762) (0.115) (0.0827) (0.194) (0.151) Interstate Con ict -0.0683 -0.115** -0.0488 -0.148 -0.0816 (0.0437) (0.0562) (0.0441) (0.0909) (0.0832) Protests 0.00570 0.0205* 0.00677 0.0223** 0.0430*** (0.00750) (0.0113) (0.00746) (0.0114) (0.0158) Ethnic Fractionalization -0.264** -0.529* -0.516** (0.116) (0.278) (0.246)
Observations 3,728 3,926 3,582 1,352 2,408 Region dummies Yes Yes No No No Lagged PTS(A) Yes No Yes No No Robust standard errors in parentheses, clustered by country *** p<0.01, ** p<0.05, * p<0.1
170 positively and signi cantly associated repression.
5.7 conclusion
In this chapter, I explored whether countries with more centrally controlled internal security arrangements are more likely to engage in widespread state repression. I also asked under what political circumstances these security arrangements are associated with repression. My discussion emphasized the divergent logics in previous theories of repression. One theory draws on totalitarian contexts to argue that security consolidation creates more capable, nationally directed forces, while a competing theory draws on the dynamics of weak states to argue that local security allows for “out-of-control” agents and predatory behavior. In understanding these separate logics, I argued that widespread state repression must be considered as a practice distinct from other self-serving and predatory abuses committed by security forces−meaning as a national policy that occurs at the behest of a national leader rather than exclusively for predatory or extractive purposes by security agents. Centrally controlled internal forces have the authority to engage in both security and policing operations. They may be exploitive but−dangerously−they have the jurisdiction, command authority and resources to conduct a range of coercive activities in the name of national security and under orders from national leaders. In the context of widespread state repression, centrally controlled internal security provides national leaders with the means to repress. Gendarmeries and national police can draw on national resources, funding and more lethal equipment that municipal forces often lack that allows them to conduct more repressive operations. Finally, democratic rules of non-intervention by the military have the unintended consequence of encouraging the creation and use of parallel security forces with domestic jurisdiction. This demonstrates the importance of understanding the relationship between security organizations, political institutions and state repression, including the unintended eects of imperfect democratic constraints. Centrally controlled internal security is particularly detrimental for human rights when leaders are not fully accountable to the public and legislative or judicial authorities. In assessing this theory, I used country-level data on security forces and human rights abuse to rst test whether more centralized security relative to local security encourages state repression. Second, I
171 examined the extent to which the relationship between centralization and repression is conditional on accountability mechanisms and a country’s selection of security organizations. In line with Davenport and Armstrong(2004), I found that states with more personnel in centrally controlled internal security forces have a higher likelihood of experiencing widespread repression and this relationship is particularly signi cant in cases of limited democracies. In other words, central security is associated with repression when state leaders face imperfect constraints on their rule. My ndings hold even after accounting for dynamics that drive certain countries to create centralized security organizations. These ndings have a number of implications for future research on state repression. Centrally controlled internal security is more likely to have a detrimental eect on human rights in democratizing or limited democratic countries, where national leaders have the ability and incentives to use internal security forces to repress domestic oppositions. In these states, local security might better prevent against government-directed national abuse and even prevent against predatory actions when local security is accountable to the local community. Despite concerns of agency slack and “out-of-control” agents, countries with democratic institutions at the local level−meaning security forces that are accountable to the local population and local government−might better curb repressive behavior. That does not mean that local security is always the answer. As shown by cases of police abuse and the continued predatory behavior of militias in a variety of war-torn countries, local security can still lack local accountability and fail to punish abusive security personnel. However, local security forces may dier in the magnitude or type of abuses committed and is a fruitful avenue for future research. While still harming civilians, local security groups may do so in a less systematic or widespread manner. Both central and local security forces can abuse civilians but the scope of abuse likely increases with the jurisdiction and capability of the force. Finally democracy alone does not eliminate human rights abuse but can change the organizations and methods for engaging in violence. Protecting civilians against repression requires strong public accountability over both centralized security forces and the national political leaders who use them. This means not just having laws that prohibit abuse but having eective public oversight mechanisms to investigate and punish security force abuses. At the national level, such mechanisms include independent
172 and impartial judiciaries to enforce these laws and hold security o cers accountable, a legislature ultimately answerable to constituents to ensure adequate regulations, and− nally−making national leaders subordinate to the legislature (and consequently the public) for allocation of resources and changes in security policies. Public scrutiny over internal security forces must occur alongside public scrutiny of national leaders to ensure human rights are upheld.
173 5.8 appendix
Documentation and description of variables
Interpretation of the Political Terror Scale levels:81
• Level 1: Countries under a secure rule of law, people are not imprisoned for their views, and torture is rare or exceptional. Political murders are extremely rare.
• Level 2: There is a limited amount of imprisonment for nonviolent political activity. However, few persons are aected, torture and beatings are exceptional. Political murder is rare.
• Level 3: There is extensive political imprisonment, or a recent history of such imprisonment. Execution or other political murders and brutality may be common. Unlimited detention, with or without a trial, for political views is accepted.
• Level 4: Civil and political rights violations have expanded to large numbers of the population. Murders, disappearances, and torture are a common part of life. In spite of its generality, on this level terror aects those who interest themselves in politics or ideas.
• Level 5: Terrorhas expanded to the whole population. The leaders of these societies place no limits on the means or thoroughness with which they pursue personal or ideological goals.
81Gibney, Mark, Linda Cornett, Reed Wood, Peter Haschke, and Daniel Arnon. 2016. The Political Terror Scale 1976-2015. Date Retrieved, from the Political TerrorScale website: http://www.politicalterrorscale.org/Data/Documentation.html.
174 Table5.5: Descriptive Statistics
State Repression PTS(A) 4,409 2.736 1.102 1.000 5.000 PTS(S) 5,346 2.378 1.160 1.000 5.00 Latent Human Rights 6,942 0.313 1.402 -3.112 4.686
Security Arrangements Central Security Ratio 5,993 0.225 0.406 0 1.00 Gendarmerie_ln 5,805 2.437 4.206 0 14.431 Militia_ln 7,733 1.463 3.725 0 18.364 State Militia_ln 6,035 1.274 3.581 0 18.364 Local Militia_ln 5,793 0.482 2.118 0 13.710
Controls Ethnic Fractionalization 4,682 0.405 0.284 0.001 0.925 Excluded Population 4,915 0.165 0.223 0.000 0.980 Youth Bulge 4,052 18.449 2.501 10.200 25.400 TotalPopulation_ln 5,927 8.457 2.119 2.079 14.087 Protests Count 7,733 0.768 2.531 0.000 49.000 Civil Con ict 5,374 0.177 0.382 0.000 1.000 Interstate Con ict 7,733 0.786 0.410 0.000 1.000 Executive Constraints 4,962 4.054 2.438 0.000 7.000 Judicial Constraints 5,487 1.130 0.799 0.000 2.00 GDP_ln 5,927 9.978 2.279 3.219 16.392
Robustness Tests
In equations with Fariss(2014)’s Latent Human Rights Protection Score as the dependent variable, I use ordinary least squares since the measure is continuous. In line with my previous model speci cations, I incorporate region and Cold War dummy variables. I also examine equations that use Fariss(2014)’s Latent Score with country xed eects.
175 Table5.6: Ratio of Centralized Security and Latent Human Rights
(1) (2) (3) (4) (5)
Ratio of Centralized Security -0.158* 0.185 -0.152* 0.0233 -0.0931 (0.0804) (0.126) (0.0822) (0.105) (0.0660) Ratio∗Executive Constraints -0.0838** (0.0329) Ratio∗Judicial Independnece -0.167* (0.0944) Executive Constraints 0.172*** 0.193*** 0.124*** (0.0186) (0.0213) (0.0167) Judicial Independence 0.354*** 0.396*** (0.0510) (0.0633) Excluded Population -0.501*** -0.511*** -0.531*** -0.532*** -0.209 (0.145) (0.144) (0.144) (0.143) (0.170) GDP_ln 0.382*** 0.380*** 0.393*** 0.388*** 0.362*** (0.0516) (0.0498) (0.0547) (0.0539) (0.0786) Population_ln -0.577*** -0.572*** -0.562*** -0.555*** -0.116 (0.0527) (0.0513) (0.0576) (0.0571) (0.155) Civil Con ict -0.926*** -0.907*** -0.923*** -0.918*** -0.639*** (0.0825) (0.0782) (0.0849) (0.0848) (0.0763) Interstate Con ict 0.178*** 0.176*** 0.166*** 0.165*** 0.0777*** (0.0479) (0.0467) (0.0522) (0.0516) (0.0295) Protests -0.0263*** -0.0277*** -0.0253*** -0.0256*** -0.00799* (0.00799) (0.00777) (0.00936) (0.00929) (0.00459) Cold War -0.140*** -0.131** -0.345*** -0.337*** -0.0839 (0.0521) (0.0534) (0.0550) (0.0556) (0.0628) Constant 1.080*** 0.985** 1.389*** 1.338*** -2.977** (0.398) (0.406) (0.369) (0.371) (1.382)
Observations 4,649 4,649 4,006 4,006 4,649 R-squared 0.715 0.719 0.701 0.702 0.352 Country Fixed Eects No No No No Yes Robust standard errors in parentheses, clustered on country. Covariates lagged one year. All models include regional dummy variables. *** p<0.01, ** p<0.05, * p<0.1
176 Table5.7: Security Personnel (ln), Democracy and Latent Human Rights
Eq. 1 Eq. 2 Eq. 3 Eq. 4 + − (Excluded Pop ) (Youth Bulge )
Internal Security Gend._ln 0.0189 0.0169 0.0119 0.0228* (0.0141) (0.0132) (0.0129) (0.0135) Gend._ln∗Exec. Constraints -0.00929** -0.00854*** -0.00767** -0.0102*** (0.00364) (0.00314) (0.00311) (0.00327) Militia_ln -0.0129 -0.0128 -0.0166 -0.0118 (0.0118) (0.0136) (0.0134) (0.0128) Militia_ln∗Exec. Constraints 0.00207 0.00109 0.00171 0.000254 (0.00318) (0.00351) (0.00345) (0.00382) Exec. Constraints 0.215*** 0.198*** 0.184*** 0.215*** (0.0236) (0.0226) (0.0217) (0.0237) Demographics ELF 0.186 -0.207 -0.282 (0.187) (0.176) (0.176) Excluded Pop. -0.517*** (0.144) TotalPop_ln -0.559*** -0.603*** -0.559*** -0.514*** (0.0604) (0.0617) (0.0503) (0.0591) Youth Bulge -0.0890*** (0.0253) Dissent & Conflict Civil Con ict -0.991*** -0.904*** -0.889*** -0.906*** (0.102) (0.0864) (0.0776) (0.0896) Interstate Con ict 0.164*** 0.171*** 0.182*** 0.167*** (0.0544) (0.0482) (0.0464) (0.0537) Protests -0.0380*** -0.0277*** -0.0273*** -0.0276*** (0.00946) (0.00976) (0.00761) (0.00902) GDP_ln 0.364*** 0.411*** 0.378*** 0.345*** (0.0451) (0.0551) (0.0494) (0.0552) Observations 4,393 4,386 4,649 2,916 R-squared 0.674 0.728 0.724 0.734 Time Dummies No Yes Yes Yes Region Dummies No Yes Yes Yes Robust standard errors in parentheses, clustered on country. Covariates lagged one year. + Eq. 3 uses excluded pop. measures. − Eq 4 includes youth bulge measures. *** p<0.01, ** p<0.05, * p<0.1
177 Table5.8: Security Organizations (Binary) and Latent Human Rights
Eq. 1 Eq. 2 Eq. 3 Eq. 4 ∧ + − (Militas ) (Excluded Pop ) (Youth Bulge )
Internal Security Gend. -0.149 -0.144* -0.163** -0.144 (0.0919) (0.0855) (0.0797) (0.0910) Militia -0.0524 -0.0681 -0.0763 (0.0754) (0.0762) (0.0871) State Militia -0.0357 (0.0867) Private Militia -0.110 (0.136) Demographics ELF 0.246 -0.148 -0.239 (0.187) (0.184) (0.183) TotalPop_ln -0.573*** -0.623*** -0.570*** -0.528*** (0.0616) (0.0635) (0.0520) (0.0612) Excluded Pop. -0.502*** (0.146) Youth Bulge -0.0940*** (0.0258) Dissent & Conflict Civil Con ict -1.015*** -0.919*** -0.908*** -0.931*** (0.112) (0.0956) (0.0830) (0.0953) Interstate Con ict 0.170*** 0.180*** 0.187*** 0.180*** (0.0555) (0.0491) (0.0478) (0.0550) Protests -0.0367*** -0.0264*** -0.0260*** -0.0258*** (0.00960) (0.0100) (0.00791) (0.00925)
Exec. Constraints 0.193*** 0.177*** 0.168*** 0.186*** (0.0186) (0.0200) (0.0186) (0.0219) GDP_ln 0.367*** 0.417*** 0.379*** 0.346*** (0.0454) (0.0575) (0.0515) (0.0563)
Observations 4,405 4,398 4,661 2,928 R-squared 0.667 0.722 0.719 0.725 Time Dummies No Yes Yes Yes Region Dummies No Yes Yes Yes Robust standard errors in parentheses, clustered on country. Covariates lagged one year. ∧ + Eq. 2 separates private and state militias. Eq. 3 uses excluded pop. measures. − Eq 4 includes youth bulge measures. *** p<0.01, ** p<0.05, * p<0.1
178 Table5.9: Security Personnel (ln) and Political TerrorScale (Amnesty)
Eq. 1 Eq. 2 Eq. 3 Eq 4 Cold War Security∗Regime Security∗Ethnic Internal Security Gendarmerie_ln 0.0134* 0.0123** -0.0170* 0.0134* (0.00748) (0.00628) (0.00986) (0.00748) Gendarmerie_ln∗Exec Constraints 0.00744*** (0.00237) Gendarmerie_ln∗Ethnic Exclusion -0.000391 -0.000391 (0.0227) (0.0227) Militia_ln -0.00150 -0.00634 -0.0161* -0.00150 (0.00652) (0.00508) (0.00875) (0.00652) Militia_ln∗Exec Constraints 0.00331 (0.00226) Militia_ln∗Ethnic Exclusion -0.0196 -0.0196 (0.0202) (0.0202) Demographics Excluded Population 0.333*** 0.271*** 0.303*** 0.333*** (0.122) (0.0989) (0.1000) (0.122) TotalPop_ln 0.302*** 0.301*** 0.311*** 0.302*** (0.0378) (0.0377) (0.0378) (0.0378) Dissent & Conflict Protest 0.00714 0.00551 0.00926 0.00714 (0.00718) (0.00756) (0.00724) (0.00718) Civil Con ict 0.620*** 0.612*** 0.601*** 0.620*** (0.0782) (0.0775) (0.0741) (0.0782) Interstate Con ict -0.0732* -0.0709 -0.0681 -0.0732* (0.0431) (0.0434) (0.0443) (0.0431)
Executive Constraints -0.0837*** -0.0824*** -0.114*** -0.0837*** (0.0143) (0.0142) (0.0177) (0.0143) GDP_ln -0.199*** -0.195*** -0.205*** -0.199*** (0.0384) (0.0376) (0.0383) (0.0384) PTS(A)(t−1) 1.091*** 1.082*** 1.085*** 1.091*** (0.0529) (0.0533) (0.0529) (0.0529) Cold War -0.0514 (0.0472) Observations 3,728 3,728 3,728 3,728
Time Dummies Yes No Yes Yes Region Dummies Yes Yes Yes Yes Robust standard errors in parentheses, clustered on country. Covariates lagged one year. *** p<0.01, ** p<0.05, * p<0.1
179 Table5.10: Security Organizations (Binary) and the Political TerrorScale (Amnesty)
Eq. 1 Eq. 2 Eq. 3 Eq. 4 ∧ + − (Militas ) (Excluded Pop ) (Youth Bulge )
Internal Security Gend. 0.0814 0.0836 0.103* 0.0837 (0.0574) (0.0627) (0.0600) (0.0725) Militia -0.0996* -0.0734 -0.0526 (0.0558) (0.0599) (0.0657) State Militia -0.164** (0.0652) Private Militia 0.0705 (0.106) Demographics ELF -0.269** -0.0888 -0.0279 (0.115) (0.127) (0.139) Excluded Pop. 0.286*** (0.0991) Youth Bulge 0.0814*** (0.0207) TotalPop_ln 0.271*** 0.328*** 0.307*** 0.265*** (0.0404) (0.0466) (0.0382) (0.0495) Dissent & Conflict Civil Con ict 0.605*** 0.614*** 0.612*** 0.655*** (0.0837) (0.0824) (0.0780) (0.0918) Interstate Con ict -0.0490 -0.0603 -0.0703 -0.00836 (0.0438) (0.0426) (0.0431) (0.0512) Protests 0.00587 0.00830 0.00727 0.0125 (0.00741) (0.00828) (0.00736) (0.00775)
Exec. Constraints -0.0761*** -0.0849*** -0.0850*** -0.0922*** (0.0127) (0.0147) (0.0143) (0.0171) GDP_ln -0.169*** -0.217*** -0.198*** -0.179*** (0.0295) (0.0427) (0.0382) (0.0441) PTS(A)(t−1) 1.140*** 1.099*** 1.095*** 1.067*** (0.0519) (0.0546) (0.0530) (0.0594)
Observations 3,593 3,590 3,739 2,290 Time Dummies No Yes Yes Yes Region Dummies No Yes Yes Yes Robust standard errors in parentheses, clustered on country. Covariates lagged one year. ∧ + Eq. 2 separates private and state militias. Eq. 3 uses excluded pop. measures. − Eq 4 includes youth bulge measures. *** p<0.01, ** p<0.05, * p<0.1
180 Table5.11: Ratio of Centralized Security and Political TerrorScale (State)
(1) (2) (3) (4)
Centralized Security Ratio 0.0801 0.0643 -0.133 -0.240 (0.0557) (0.102) (0.101) (0.174) Executive Constraints -0.0980*** -0.230*** -0.113*** -0.251*** (0.0143) (0.0231) (0.0148) (0.0260) Centralized∗Executive Constraints 0.0530** 0.0758* (0.0241) (0.0429) Excluded Population 0.393*** 0.724*** 0.401*** 0.734*** (0.100) (0.189) (0.101) (0.190) GDP_ln -0.225*** -0.456*** -0.227*** -0.458*** (0.0322) (0.0607) (0.0321) (0.0606) Population_ln 0.352*** 0.711*** 0.353*** 0.711*** (0.0345) (0.0595) (0.0337) (0.0586) Civil Con ict 0.567*** 1.406*** 0.552*** 1.385*** (0.0783) (0.120) (0.0763) (0.117) Interstate Con ict -0.0807** -0.224*** -0.0796** -0.223*** (0.0400) (0.0604) (0.0401) (0.0602) Protests 0.0121 0.0278** 0.0128 0.0292** (0.00809) (0.0120) (0.00801) (0.0118) PTS(S)(t−1) 1.299*** 1.297*** (0.0469) (0.0464) Cold War -0.365*** -0.374*** (0.0523) (0.0525)
Lagged DV Yes No Yes No Time Dummies No Yes No Yes Region Dummies Yes Yes Yes Yes Robust standard errors in parentheses, clustered on country. Covariates lagged one year. *** p<0.01, ** p<0.05, * p<0.1
181 Chapter 6
Conclusion: Democratizing domestic security
6.1 summary
How do national governments organize and project coercive power internally? This dissertation explored this and related questions through a rigorous global study on internal security forces and their organizational attributes, examining the full spectrum of internal forces and eects of these forces on the safety of citizens within the state. Through a global analysis, I showed the origins and consequences of varying internal state security arrangements over time and across countries.National governments strategically create and expand internal security organizations as they seek to maximize coercive capabilities and control over those capabilities, while simultaneously minimizing political or nancial costs. I explored the types of internal security forces that national governments adopt along with the relative advantages and disadvantages of these forces. Reconceptualizing forces by their organizational attributes−their degree of centralization and regulation−allows for a more precise and holistic examination of how states project force within their territory. Re ning the concept of internal security and improving measurement of such security is essential to understanding dynamics between national governments, internal security forces and the citizens both institutions are assumed to protect. A complex relationship exists between political institutions and strategies for internal security. Many democratic countries, speci cally Western European democracies, have a long history of security consolidation and centralization, which is passed on to former colonies or client states during peaceful
182 transitions of rule.1 Factors attributable to the rise of modern democracies−such as strong national bureaucracies and taxation systems to better compete against rivals−contribute to the formation of regular and centrally controlled security forces. Similarly, national governments in many democratizing states expand centrally controlled internal security to maximize control over the coercive apparatus and minimize the in uence of local or regional actors while maintaining internal order and protecting borders. Gendarmeries and national police increased throughout the late twentieth century, particularly throughout Sub-Saharan Africa and Eurasia after the Cold War. The diusion of gendarmeries from former colonial powers, such as France and Belgium, contributed to this trend in centralizing security. And contrary to conventional wisdom, centrally controlled internal forces are more common globally than local forces in low-intensity civil con icts. While local militias are seen as ubiquitous to civil con ict, national governments are more likely to adopt and expand gendarmeries or national police as an initial strategy for curbing internal dissent or eliminating coercive challenges, relying on local forces only after other strategies have failed and civil con icts have dragged on for a number of years. Short of complete collapse of the national government, countries with at least minimal state capacity rst turn to centralized forces to address internal con ict. Both the international order and domestic dynamics in uence the national government’s ability and incentives to consolidate security. However centralization and national regulation of security does not necessarily decrease state repression. During a transition to democracy, the opposite is likely to occur; national leaders in democratizing states have the incentive to preserve illiberal elements of rule, including a security force beholden to the executive that the executive can then use to repress or threaten voters. In some cases, the adoption of centrally controlled internal security may be done to decrease the risk of a military coup, but−more generally−national elites use central security to ward o domestic threats arising outside of the government itself. Through a global analysis of security forces and repression over four decades, I showed that centrally controlled internal forces increase the likelihood of widespread repression when the populace has only a limited ability to constraint its national leaders. In other words, centralized security is associated with 1While certain democracies like the United States have highly localized municipal forces engaged in security tasks (Waxman 2009), the United States is more of an exception to the rule among developed democratic states.
183 a greater propensity for violence among states that are not fully democratic. This is likely due to the incentives of national leaders in democratizing states, who have the incentive to make the government appear democratic by accepting some constraints (including certain checks and balances by the legislature or judicial system). However, they also have the incentive to develop a force accountable and loyal to the regime rather than to the domestic population. In contrast to gendarmeries, local internal security appears more frequently in autocracies than democracies and have experienced a global decline after the Cold War. While centrally controlled forces emerge from peaceful transitions of rule, local security forces are associated with wars of liberation. Similarly, irregular local forces emerge and expand more often during high intensity intrastate con ict (civil wars) or insurgencies involving autonomous ethnic populations. Both the nature of con ict and the human terrain within con ict in uence the localization and deregulation of internal state security. Irregular local forces like militias are not a central government’s rst strategy nor the most preferred strategy. I showed that local militias often start after con icts have broken out and frequently appearing a number of years into civil con icts. This implies that national governments do not necessarily want militias but adopt them or condone them when other security forces or security strategies have failed. I also showed that these forces have advantages in population-centric operations relative to gendarmeries. Particularly when local militias form around ethnic enclaves, they have an advantage in local area knowledge and dissuading rebel recruitment within local communities over central forces, which may have little knowledge of the area they are deployed to or lack any social capital with the community they operate within. Yet local security forces present central governments with substantial challenges in terms of control. Localization and deregulation of security reduces the central government’s ability to monitor, coordinate and sanction forces. Forces like militias may have fewer startup costs compared to gendarmeries, receive more external aid or foreign support and have comparative advantages in local area knowledge. However these forces can empower local elites at the expense of national institutions and exacerbate ethnic tensions. Central governments also face greater challenges demobilizing local forces compared to centrally controlled forces, as national elites often lack formal administration over them.
184 In spite of the disadvantages of local forces, this dissertation showed that the central government’s adoption or allowance of these forces occurs for strategic reasons that are consistent across historical periods, geographic regions and international systems. In both feudal Europe and the contemporary world, local security provides leaders with a way to combat stronger centralized rivals and−after war−as a check that prevents established central powers from overstepping their bounds. In medieval Europe, the breakup of the Carolingian empire gave way to feudalism, where irregular forces formed around local or regional lords rather than central sovereigns (Ganshof 1964). These local security organizations made for unreliable sources of force in the long term, with multiple overlapping contracts to dierent princes or dukes, and were ill equipped to help sovereigns engage in territorial expansion (Beeler 1971). Similarly, violent disintegration of colonies and empires after World War II resulted in the localization and deregulation of force. After wars of liberation, newly created national governments had to bargain with local elites and their forces over whom national authorities had limited oversight and enforcement over. Local forces in the medieval and modern era gave national elites a form of exible and cost-eective security, though at the cost of decreased reliability and control. Yet local militias in both eras still provided eective defensive capabilities (Clayton and Thomson 2015; Dupuy and Dupuy 1993). Whether private militaries in feudal France, colonial militias in America or civilian volunteer organizations of the Philippines, local forces are more loyal to their local elites than national leaders, providing not only civil protection but also a check against centralized powers. Thus, it should be of no surprise that local security forces are not necessarily associated with a propensity towards state repression like centrally controlled forces. National leaders cannot depend on local forces to carry out nation-wide repressive policies. This is not to say that local forces always curb human rights abuses. On the contrary, it may be that these organizations engage more frequently in abuse against citizens−including unlawful killings or arbitrary arrests−but not systematically or at the behest of national leaders. Further research should explore the type and magnitude of human rights abuse committed by dierent internal forces. Finally this dissertation showed how factors associated with the international system in uence the domestic organization of force. In early modern Europe, an anarchic and bellicose environment
185 characterized the international system. States confronted external rivals over territory and waged war to acquire or defend such territories. The post World War II era, in contrast, strengthened the recognition of international borders, while the superpower rivalry of the Cold War increased ows of foreign aid, thus lessening the incentives for some governments to centralize force even in the presence of security challenges. Both domestic and international factors in uence the strategic calculus of national leaders in their attempt to maximize internal security.
6.2 contributions
This dissertation explored the reasons why national governments create and expand a variety of internal armed forces to combat domestic threats to rule−including immediate and historical causes−and how the resulting security apparatus has unintended consequences for human rights. For example, the centralization of security within democratizing countries grants national leaders an eective tool for repressing voters and curbing popular contestation of their rule. Localization of security, on the other hand, may limit a government’s ability to demobilize and disarm internal forces. This project provided a comprehensive global examination of domestic security. Rather than considering the impact of one policy a state uses to generate force, my work examines the tradeos between potential policies. This approach provides a broader framework for considering how central governments establish the use of force and a more nuanced perspective on the nature of state power. The data and subsequent analysis gives academics and policy practitioners a framework for examining which dynamics lead to unaccountable internal security versus publicly accountable, sustainable and eective security. My work shows which security arrangements generally promote responsible use of force and decreased human rights abuse in the long term. As shown in this dissertation, the establishment of limited constraints on executive authority is not su cient for generating sustainable, democratically-oriented security. In many cases, the development and expansion of centrally controlled security re ects the demands and needs of national elites rather than the safety and security of ordinary citizens. Only in cases where political leaders are fully accountable−to the public, legislatures and independent judiciaries−do centrally controlled security organizations have a decreased tendency to
186 repress. Without eective public scrutiny, national elites use centrally controlled forces for their own bene t. Central security enables them to direct disproportionate lethal violence against a public they were established to protect. This research also shows why, on the other hand, local security does not necessarily re ect the practices and ideals of community policing. Compared to centrally controlled forces, local forces can be more representative of the population they work among and well-equipped to engage in defensive or intelligence operations. However, localization is also not su cient for generating accountable, eective and e cient security forces. Local forces often lack adequate regulation by central or local governments and engage in unsanctioned violence and predatory abuse without risk of punishment. The behavior of the Afghan Local Police in Afghanistan, Kurdish Village Guards in Turkey and even municipal police in the United States all suggest that localization is not su cient for ensuring that internal forces uphold the rule of law and have full respect for human rights. Without public oversight, civilian control and su cient professionalization (including training), local forces can contribute to abuse against civilians, exacerbate con ict and destabilize communities. The role of public oversight, including oversight by independent bodies and civil society, is more important for the maintenance of internal security than previously realized. A closer examination of these accountability mechanisms is required to understand under what circumstances local forces actually protect their community and promote stability. Though mentioned in security sector reform discussions, the importance of these accountability mechanisms is currently overlooked in international relations and security studies and should be a focus of future research.2 Finally, my dissertation contributes to research on the formation and evolution of state power. Central governments with the capacity to create security do not always have the ability to coordinate and control these coercive organizations. Internal security forces must be eective enough to address internal challenges but controlled enough to be regulated and demobilized after con icts end. While this dissertation did not explore the recruitment, training and de-mobilization of security forces, my 2Bryden, Alan, and Vincenza Scherrer, eds. Disarmament, Demobilization and Reintegration and Security Sector Reform: Insights from UN Experience in Afghanistan, Burundi, the Central African Republic and the Democratic Republic of the Congo. LIT Verlag Munster, 2012.
187 ndings suggest that each type of organization comes with its own strengths and weaknesses with respect to these processes. Investigating variation in mobilization, professionalization and de-mobilization across forces will further aid our understanding of the tradeos between strategies for obtaining−and then sustaining−internal security.
6.3 directions for future research: democratizing security
This project not only explained the advantages and risks of centrally controlled security forces but also why local security is not a panacea for a country’s domestic security. The national government may have an easier time controlling and resourcing national police forces, but these forces can be used for illiberal purposes without explicit public accountability. More research is needed to understand the consequences of national police within developing and democratizing states, including situations where eective civilian oversight actually develops. While this dissertation explored the relationship between executive constraints, judicial independence and security forces, it did not consider the role of other accountability and transparency mechanisms at national or sub-national levels. In line with work on community policing (Brogden, Nijhar et al. 2013), consent by and partnership with communities may better ensure that internal forces address the unique challenges of their areas as civilian governments simultaneously maintain control of those forces. Stronger civilian oversight mechanisms may promote more regulated and restrained internal security forces. Understanding the appropriate balance of community governance with national oversight and national provisioning of resources would be a fruitful avenue for further research. Another useful next stage will be to compare recruitment and demobilization practices across internal forces. As governments can have di culty demobilizing both local and central security forces, it may be not only civilian control of these forces but also how demobilizations or disarmaments are done that is vitally important. Given the number of local forces that emerge after civil wars, mechanisms associated with reintegration of ex-combatants into civil society may contribute to whether security forces peacefully demobilize and disarm.3 3Reintegration refers to the process by which former combatants acquire a civilian status, including civilian employment and
188 Finally, the results from this dissertation suggest that dierent internal security arrangements are associated with dierent trends in abuse. Research into human rights abuse should disaggregate the abuses by their speci c perpetrators to have a better understanding of the type of abuse and reasons for abuse associated with each organization. By understanding the abuses internal security forces are likely to commit, reform policies can be better tailored to address, punish and limit such behavior. As the state and its security forces are ever evolving and adapting, so too must the research evolve and adapt. It is not simply whether governments increase defense expenditures or resources but how they allocate expenditures and resources that matters. Knowng what questions to ask, academic and policy work can start to prepare for tomorrow’s security challenges.
civilian income within their local communities. Bryden, Alan, and Vincenza Scherrer, eds. Disarmament, Demobilization and Reintegration and Security Sector Reform: Insights from UN Experience in Afghanistan, Burundi, the Central African Republic and the Democratic Republic of the Congo. LIT Verlag Mnster, 2012.
189 Works Cited
Acemoglu, Daron, James A. Robinson and Rafael J. Santos. 2013. “The Monopoly of Violence: Evidence from Colombia.” Journal of the European Economic Association 11:5–44.
Ahram, Ariel I. 2011. Proxy Warriors: The Rise and Fall of State Sponsored Militias. Stanford Security Studies.
Allard, Kenneth. 1996. Command, control, and the common defense. Technicalreport DTIC Document.
Alvarez, Alex. 2006. “Militias and Genocide.” War Crimes, Genocide and Crimes Against Humanity 2(1):1–33.
Andrade, John M. 1985. World police and paramilitary forces. Macmillan Publishers Limited.
Andreski, Stanislav. 1968. Military Organization and Society. University of California Press.
Arendt, Hannah. 1951. The Origins of Totalitarianism. Schocken Books.
Asfura-Heim, Patricio. 2014. “The Future of Civil Defense Forces and Counterterrorism in an Era of Persistent Con ict.” CNA .
Avant, Deborah. 2000. “From Mercenary to Citizen Armies: Explaining Change in the Practice of War.” International Organization 54(1):41–72.
Avant, Deborah. 2004. “Think again: mercenaries.” Foreign Policy p. 21.
Avant, Deborah. 2005. The market for force: The Consequences of Privatizing Security. Cambridge University Press.
Banks, Arthur S. and Kenneth Wilson. 2016. “Cross-National Time-Series Data Archive. DataBanks International.” http://cntsdata.com/.
Barter, Shane Joshua. 2013. “State Proxy or Security Dilemma? Understanding Anti-Rebel Militias in Civil War.” Asian Security 9(2):75–92.
Bates, Robert. 2010. Prosperity and Violence. W W Norton and Company.
Bayley, David H. 1990. Patterns of policing: A comparative international analysis. Rutgers University Press.
Beardsley, Kyle, Kristian Skrede Gleditsch and Nigel Lo. 2015. “Roving Bandits? The Geographical Evolution of African Armed Con icts1.” International Studies Quarterly .
Beeler, John. 1971. Warfare in Feudal Europe, 730-1200. Cornell University Press.
Belkin, Aaron and Evan Schofer. 2003. “Toward a Structural Understanding of Coup Risk.” The Journal of Conflict Resolution 47(5):594–620.
190 Belkin, Aaron and Evan Schofer. 2005. “Coup risk, counterbalancing, and international con ict.” Security Studies 14(1):140–177.
Besley, Timothy and Torsten Persson. 2011. Pillars of prosperity: The political economics of development clusters. Princeton University Press.
Biberman, Yelena. 2016. “Self-Defense Militias, Death Squads, and State Outsourcing of Violence in India and Turkey.” Journal of Strategic Studies pp. 1–31.
Biddle, Stephen. 2004. Military power: explaining victory and defeat in modern battle. Princeton University Press.
Biddle, Stephen, Jerey A. Friedman and Jacob Shapiro. 2012. “Testing the Surge: Why Did Violence Decline in Iraq in 2007?” International Security 37(1):7–40.
Bigo, Didier. 2000. “When two become one: Internal and External Securitisations in Europe.” International Relations Theory and the Politics of European Integration: Power, Security, and Community p. 171.
Blocq, Daniel S. 2014. “The grassroots nature of counterinsurgent tribal militia formation: the case of the Fertit in Southern Sudan, 1985–1989.” Journal of Eastern African Studies 8(4):710–724.
Boduszyński, Mieczysław P. 2016. “Iraq’s Year of Rage.” Journal of Democracy 27(4):110–124.
Bradbury, Jim. 2004. The Routledge companion to medieval warfare. Routledge.
Brogden, Mike, Preeti Nijhar et al. 2013. Community policing. Routledge.
Bueno de Mesquita, Bruce, Alastair Smith, Randolph Siverson and James Morrow. 2005. The Logic of Political Survival. MIT Press.
Buhaug, Halvard. 2010. “Dude, where’s my con ict? LSG, relative strength, and the location of civil war.” Conflict Management and Peace Science .
Butler, Christopher, Tali Gluch and Neil J. Mitchell. 2007. “Security Forces and Sexual Violence: A Cross-National Analysis of a Principal-Agent Argument.” Journal of Peace Research 44(6):669–687.
Byman, Daniel. 2005. Deadly Connections: States that Sponsor Terrorism. Cambridge University Press.
Byman, Daniel and Sarah E Kreps. 2010. “Agents of Destruction? Applying Principal-Agent Analysis to State-Sponsored Terrorism.” International Studies Perspectives 11(1):1–18.
Campbell, Bruce B. and Arthur Brenner, eds. 2002. Death Squads in Global Perspective: Murder with Deniability. Palgrave MacMillan.
Carey, Sabine C, Michael P Colaresi and Neil J Mitchell. 2015. “Risk Mitigation, Regime Security, and Militias: Beyond Coup-proo ng.” International Studies Quarterly .
Carey, Sabine C., Neil J. Mitchell and Will Lowe. 2013. “States, the Security Sector, and the Monopoly of Violence: A New Database on Pro-Government Militias.” Journal of Peace Research 50(2):249–258.
191 Cave, Alfred A. 2004. The French and Indian War. Greenwood Publishing Group. Cederman, Lars-Erik, Andreas Wimmer and Brian Min. 2010. “Why do ethnic groups rebel? New data and analysis.” World Politics 62(01):87–119. Cigar, Norman. 2014. TribalMilitias: An Eective Toolto Counter Al-Qaida and Its A liates? Technical report DTIC Document. Cingranelli, David L. and David L. Richards. 2010. “The Cingranelli and Richards (CIRI) Human Rights Data Project.” Human Rights Quarterly 32:395–418. Clayton, Govinda and Andrew Thomson. 2015. “Civilianizing Civil Con ict: Civilian Defense Militias and the Logic of Violence in Intra-State Con ict.” International Studies Quarterly . Cohen, Dara Kay. 2011. Causes of Sexual Violence During Civil War: Cross-National Evidence (1980-2009). Minnesota International Relations Colloquium. Cohen, Dara Kay and Ragnhild Nordås. 2015. “Do States Delegate Shameful Violence to Militias? Patterns of Sexual Violence in Recent Armed Con icts.” Journal of Conflict Resolution 59(5):877–898. Conrad, Courtenay Ryals and William H Moore. 2010. “What Stops the Torture?” American Journal of Political Science 4(2). Cooper, Jerry. 2002. The Rise of the National Guard: The Evolution of the American Militia, 1865-1920. U of Nebraska Press. Cunningham, David. 2004. There’s Something Happening Here. University of California Press. Cunningham, David E, Kristian Skrede Gleditsch and Idean Salehyan. 2009. “It takes two: A dyadic analysis of civil war duration and outcome.” Journal of Conflict Resolution . Dallin, Alexander and George W. Breslauer. 1970. Political Terror in Communist Systems. Stanford University Press. Davenport, Christian. 2007a. “State Repression and Political Order.” Annual Review Political Science 10:1–23. Davenport, Christian. 2007b. State Repression and the Domestic Democratic Peace. Cambridge University Press. Davenport, Christian. 2007c. “State Repression and the Tyrannical Peace.” Journal of Peace Research 44:485–504. Davenport, Christian and David A. Armstrong. 2004. “Democracy and the violation of human rights: A statistical analysis from 1976 to 1996.” American Journal of Political Science 48(3):538–554. Davenport, Christian, Sarah A Soule and David A Armstrong. 2011. “Protesting while black? The dierential policing of American activism, 1960 to 1990.” American Sociological Review 76(1):152–178. DeMeritt, Jacqueline. 2012. “International Organizations and Government Killing: Does Naming and Shaming Save Lives?” International Interactions 38:1–25.
192 Dodge, Toby. 2013. “State and society in Iraq ten years after regime change: the rise of a new authoritarianism.” International Aairs 89(2):241–257.
Doyle, Charles. 2000. The Posse Comitatus Act and Related Matters: the use of the military to execute civilian law. DTIC Document.
Dugas, John C. 2012. “Paramilitaries and the economic origins of armed con ict in Colombia.” Latin American Research Review 47(1):205–213.
Dunn, Timothy J. 1996. Militarization of the US-Mexico border, 1978-1992. University of Texasat Austin Center for.
Dupuy, R. Ernest and Trevor N Dupuy. 1993. The Harper Encyclopedia of Military History: From 350 B.C. to the Present. New York: Harper Collins.
Earl, Jennifer and Sarah A. Soule. 2006. “Seeing Blue: A Police-Centered Explanation of Protest Policing.” Mobilization: An International Journal 11(2):145–164.
Eck, Kristine. 2015. “Repression by Proxy: How Military Purges and Insurgency Impact the Delegation of Coercion.” Journal of Conflict Resolution .
Eisenhardt, Kathleen M. 1989. “Agency theory: An assessment and review.” Academy of management review 14(1):57–74.
Fariss, Christopher J. 2014. “Respect for human rights has improved over time: Modeling the changing standard of accountability.” American Political Science Review 108(02):297–318.
Fearon, James D and David D. Laitin. 2003. “Ethnicity, Insurgency, and Civil War.” The American Political Science Review 97(1):75–90.
Fearon, James D; Laitin, DD. 2003. “Ethnicity, Insurgency, and Civil War.” American Political Science Review 97(1):75–90.
Feaver, Peter D. 1999. “Civil-military relations.” Annual Review of Political Science 2(1):211–241.
Feaver, Peter D. 2009. Armed servants: Agency, oversight, and civil-military relations. Harvard University Press.
Feenstra, Robert C., Robert Inklaar and Marcel P. Timmer. 2015. “The Next Generation of the Penn World Table.” American Economic Review 105(10):3150–3182.
Fein, Helen. 1995. “More Murder in the Middle: Life-Integrity Violations and Democracy in the World, 1987.” Human Rights Quarterly 17(1):170–191.
Felbab-Brown, Vanda. 2016. “Hurray for militias? Not so fast: Lessons from the Afghan Local Police experience.” Small Wars & Insurgencies 27(2):258–281.
Finer, Samuel E. 1975. State and Nation-Building in Europe: The Role of the Military. In The Formation of National States in Western Europe, ed. Charles Tilly. Princeton University Press chapter 2, pp. 84–163.
193 Francis, David J. 2005. Civil Militia: Africa’s Intractable Security Menace? Ashgate Publishing, Ltd.
Frühling, H Hugo, Joseph S Tulchinand Heather Golding. 2003. Crime and violence in Latin America: Citizen security, democracy, and the state. Woodrow Wilson Center Press.
Fumerton, Mario. 2001. “Rondas Campesinas in the Peruvian Civil War: Peasant Self-defence Organisations in Ayacucho.” Bulletin of Latin American Research 20(4):470–497.
Ganshof, François Louis. 1964. Feudalism. Number 34 University of Toronto Press.
Gennaioli, Nicola and Hans-Joachim Voth. 2015. “State capacity and military con ict.” The Review of Economic Studies 82(4):1409–1448.
Gibler, Douglas M. 2017. “State Development, Parity, and International Con ict.” American Political Science Review 111(1):21–38.
Gibney, Mark, Linda Cornett, Reed Wood and Peter Haschke. 2012. “Political Terror Scale 1976-2012.”.
Giustozzi, Dr Antonio. 2007. “Auxiliary Force or National Army? Afghanistan’s ‘ANA’and the Counter-Insurgency Eort, 2002–2006.” Small Wars and Insurgencies 18(1):45–67.
Gleditsch, Kristian Skrede. 2002. “Expanded trade and GDP data.” Journal of Conflict Resolution 46(5):712–724.
Gleditsch, Nils Petter, Peter Wallensteen, Mikael Eriksson, Margareta Sollenberg and Håvard Strand. 2002a. “Armed con ict 1946-2001: A new dataset.” Journal of peace research 39(5):615–637.
Gleditsch, Nils Petter, Peter Wallensteen, Mikael Eriksson, Margareta Sollenberg and Hvard Strand. 2002b. “Armed Con ict 1946-2001: A New Dataset.” Journal of Peace Research 39(5).
Gobinet, Pierre. 2008. “The gendarmerie alternative: is there a case for the existence of police organisations with military status in the twenty- rst century European security apparatus?” International Journal of Police Science & Management 10(4):448–463.
Goldsmith, Andrew John and Colleen Lewis. 2000. Civilian oversight of policing: Governance, democracy, and human rights. Hart Publishing.
Goldstein, Robert Justin. 1978. Political repression in modern America from 1870 to the present. GK Hall & Company.
Golkar, Saeid. 2015. Captive Society: The Basij Militia and Social Control in Iran. Columbia University Press.
Goodhand, Jonathan and Aziz Hakimi. 2014a. “Counterinsurgency, Local Militias, and Statebuilding in Afghanistan.” United States Institute of Peace (90).
Goodhand, Jonathan and Aziz Hakimi. 2014b. “Counterinsurgency, Local Militias, and Statebuilding in Afghanistan.” USIP, Washington, DC .
Gordon, Neve. 2008. Israel’s Occupation. University of California Press.
194 Greitens, Sheena Chestnut. 2016. Dictators and their secret police: Coercive institutions and state violence. Cambridge University Press.
Hafner-Burton, Emilie M. 2008. “Sticks and Stones: Naming and Shaming the Human Rights Enforcement Problem.” International Organization 62(4):689–716.
Hager Jr., Robert P and David A Lake. 2000. “Balancing empires: Competitive decolonization in international politics.” Security Studies 9(3):108–148.
Hallberg, Johan Dittrich. 2012. “Prio con ict site 1989-2008: A geo-referenced dataset on armed con ict.” Conflict Management and Peace Science 29(2):219–232.
Hammond, Michael F. 2009. “Sense and respond: logistics at the unit level.” Army Logistician 41(1):6.
Hartzell, Caroline and Matthew Hoddie. 2003. “Institutionalizing Peace: Power Sharing and Post-Civil War Con ict Management.” American Journal of Political Science 47(2):318–332.
Head, Michael and Scott Mann. 2016. Domestic deployment of the armed forces: Military powers, law and human rights. Routledge.
Heckman, James J. 1979. “Sample Selection Bias as a Speci cation Error.” Econometrica 47(1):153–161.
Hegre, Håvard. 2008. “Gravitating toward War Preponderance May Pacify, but Power Kills.” Journal of Conflict Resolution 52(4):566–589.
Herbst, Jerey. 2014. States and power in Africa. JSTOR.
Hill, Daniel and Zachary Jones. 2014. “An Empirical Evaluation of Explanations for State Repression.” American Political Science Review .
Hills, Alice. 1997. “Warlords, militia and con ict in contemporary Africa: A re-examination of terms.” Small Wars & Insurgencies 8(1):35–51.
Hironaka, Ann. 2009. Neverending wars: The international community, weak states, and the perpetuation of civil war. Harvard University Press.
Holmes, Jennifer S, Sheila Amin Gutiérrez Piñeres and Kevin M Curtin. 2006. “Drugs, violence, and development in Colombia: A department-level analysis.” Latin American Politics and Society 48(3):157–184.
Holmqvist, Caroline. 2005. Private security companies. Stockholm International Peace Research Institute.
Horowitz, Donald L. 1985. Ethnic groups in conflict. Univ of California Press.
HRW. 1996. India’s Secret Army in Kashmir: New Patterns of Abuse Emerge in the Conflict. Vol. 8 http://www.unhcr.org/refworld/docid/3ae6a8558.html: Human Rights Watch.
Jane’s Sentinel Security Assessments, IHS. 2010. “Jane’s Sentinel Security Assessments.”.
195 Jansen, Marc and Nikita Vasilevich Petrov. 2002. Stalin’s loyal executioner: people’s commissar Nikolai Ezhov, 1895-1940. Hoover Institution Press.
Jentzsch, Corinna, Stathis N Kalyvas and Livia Isabella Schubiger. 2015. “Militias in Civil Wars.” Journal of Conflict Resolution p. 0022002715576753.
Jones, Seth G. 2012. “The Strategic Logic of Militia.”.
Kalyvas, Stathis. 2006. The Logic of Violence in Civil War. Cambridge University Press.
Kalyvas, Stathis N. 2008. “Ethnic Defection in Civil War.” Comparative Political Studies 41(8):1043–1068.
Kalyvas, Stathis N and Laia Balcells. 2010. “International system and technologies of rebellion: How the end of the Cold War shaped internal con ict.” American Political Science Review 104(03):415–429.
Kestnbaum, Meyer. 2017. The State and the Revolution in War. Cambridge University Press.
Kinsella, David and Herbert K Tillema. 1995. “Arms and Aggression in the Middle East Overt Military Interventions, 1948-1991.” Journal of Conflict Resolution 39(2):306–329.
Kirkpatrick, David. 2012. “Government Issues Order to Disband Libya Forces.” New York Times .
Kohn, Richard H. 1997. “How democracies control the military.” Journal of Democracy 8(4):140–153.
Koren, Ore. 2015. “Means to an end: Pro-government militias as a predictive indicator of strategic mass killing.” Conflict Management and Peace Science .
Krahmann, Elke. 2010. States, Citizens and the Privitization of Security. Cambridge University Press.
Krause, Keith. 1996. “Insecurity and state formation in the global military order: the Middle Eastern case.” European Journal of International Relations 2(3):319–354.
Ladwig, Walter C. 2008. “Supporting allies in counterinsurgency: Britain and the Dhofar Rebellion.” Small Wars & Insurgencies 19(1):62–88.
Levine, Paul and Ron Smith. 2000. “Arms export controls and proliferation.” Journal of Conflict Resolution 44(6):885–895.
Lewis, Paul H. 2002. Guerrillas and generals: the "Dirty War" in Argentina. Greenwood Publishing Group.
Linz, Juan José. 2000. Totalitarian and authoritarian regimes. Lynne Rienner Publishers.
Lioe, Kim Eduard. 2011. Armed Forces in Law Enforcement Operations? The German and European Experience. Springer.
Loveman, Mara. 1998. “High-Risk Collective Action: Defending Human Rights in Chile, Uruguay, and Argentina 1.” American Journal of Sociology 104(2):477–525.
Lutterbeck, Derek. 2003. The Fortress Walls: Policing the EU’s External Borders, 1990-2001 PhD thesis Institut universitaire de hautes études internationales.
196 Lutterbeck, Derek. 2004. “Between Police and Military The New Security Agenda and the Rise of Gendarmeries.” Cooperation and conflict 39(1):45–68.
Lyall, Jason. 2009. “Rage Against the Machines: Explaining Outcomes in Counterinsurgency Wars.” International Organization 63(1):67–106.
Lyall, Jason. 2010. “Are coethnics more eective counterinsurgents? Evidence from the second Chechen war.” American Political Science Review 104(01):1–20.
Mao, Zedong. 1937. On Guerrilla Warfare. Selected Works for Mao Tse-tung: Vol. IX.
Marenin, Otwin and Dilip K Das. 2000. Challenges of policing democracies: A world perspective. Psychology Press.
Marshall, Monty G, TedRobert Gurr and Barbara Har. 2009. “Political Instability TaskForce: internal wars and failures of governance, 1955–2008—PITF problem set codebook.”.
Marshall, Monty and Keith Jaggers. 2005. “Polity IV Project: Political Regime Characteristics and Transitions, 18002004. Data Users’ Manual.” Available at: http://www.cidcm.umd.edu/polity/data/p4manualv2004.pdf.
Marten, Kimberly. 2007. “Warlordism in comparative perspective.”.
Mitchell, Neil J. 2004. Agents of Atrocity: Leaders, Followers, and the Violation of Human Rights in Civil War. Macmillan.
Mitchell, Neil J, Sabine C Carey and Christopher K Butler. 2014. “The Impact of Pro-Government Militias on Human Rights Violations.” International Interactions 40(5):812–836.
Mitchell, Sara McLaughlin, Jonathan J Ring and Mary K Spellman. 2013. “Domestic legal traditions and states’ human rights practices.” Journal of Peace Research 50(2):189–202.
Morrow, James D. 1993. “Arms versus allies: Trade-os in the search for security.” International Organization 47(2).
Nordas, Ragnhild and Christian Davenport. 2013. “Fight the Youth: YouthBulges and State Repression.” American Journal of Political Science 57(4):926–940.
North, Douglass C, John Joseph Wallis and Barry R Weingast. 2009. Violence and Social Orders: A Conceptual Framework for Interpreting Recorded Human History. Cambridge University Press.
Olson, Mancur. 1993. “Dictatorship, Democracy, and Development.” American Political Science Review 87(03):567–576.
Parker, Georey. 1996. The military revolution: Military innovation and the rise of the West, 1500-1800. Cambridge University Press.
Peic, Goran. 2014. “Civilian defense forces, state capacity, and government victory in counterinsurgency wars.” Studies in Conflict & Terrorism 37(2):162–184.
197 Perlmutter, Amos and ValeriePlave Bennett. 1980. The Political Influence of the Military. YaleUniversity Press.
Petersohn, Ulrich. 2014. “The Impact of Mercenaries and Private Military and Security Companies on Civil War Severity between 1946 and 2002.” International Interactions: Empirical and Theoretical Research in International Relations 40(2):191–215.
Pettersson, Therése and Peter Wallensteen. 2015. “Armed con icts, 1946-2014.” Journal of Peace Research 52(4):536–550.
Pilster, Ulrich and Tobias Böhmelt. 2011. “Coup-Proo ng and Military Eectiveness in Interstate Wars, 1967–99.” Conflict Management and Peace Science 28(4):331–350.
Pilster, Ulrich and Tobias Bohmelt. 2012. “Do Democracies Engage Less in Coup: Proo ng? On the Relationship between Regime Type and Civil–Military Relations.” Foreign Policy Analysis 8(4):355–372.
Poe, Steven C, Sabine C Carey and Tanya C Vazquez. 2001. “How are these pictures dierent? A quantitative comparison of the US State Department and Amnesty International human rights reports, 1976-1995.” Human Rights Quarterly 23(3):650–677.
Poe, Steven and Neal Tate. 1994. “Repression of Human Rights to Personal Integrity in the 1980’s: A Global Analysis.” The American Political Science Review 88(4):853–872.
Powell, Jonathan. 2012. “Determinants of the Attempting and Outcome of Coups de’etat.” Journal of Conflict Resolution 56(6):1017–1040.
Powell, Jonathan M and Clayton L Thyne. 2011. “Global instances of coups from 1950 to 2010 A new dataset.” Journal of Peace Research 48(2):249–259.
Quinlivan, James T. 1999. “Coup-proo ng: Its practice and consequences in the Middle East.” International Security 24(2):131–165.
Regan, Patrick M. 1994. Organizing societies for war: The process and consequences of societal militarization. Praeger Westport, Conn.
Reid, Rachel, Sahr Muhammedally, Brad Adams, James Ross and Joseph Saunders. 2011. Just Don’t Call it a Militia": Impunity, Militias, and the" Afghan Local Police. Human Rights Watch.
Rejali, Darius. 2007. Torture and Democracy. Princeton University Press.
Reyntjens, Filip. 2009. The great African war: Congo and regional geopolitics, 1996-2006. Cambridge University Press.
Ritter, Emily Hencken and Courtenay R Conrad. 2016. “Preventing and responding to dissent: The observational challenges of explaining strategic repression.” American Political Science Review 110(01):85–99.
Roberts, Michael. 1956. The military revolution, 1560-1660. M. Boyd.
198 Rogers, Cliord J, William Caferro and Shelley Reid. 2010. “The Oxford encyclopedia of medieval warfare and military technology.”.
Rohan, Maxwell. 2014. New Armies from Old: Merging competing military forces after civil wars. Georgetown University Press chapter Bosnia-Herzegovina: From three armies to one, pp. 180–181.
Rotberg, Robert I. 2010. When states fail: causes and consequences. Princeton University Press.
Salehyan, Idean. 2010. “The Delegation of War to Rebel Organizations.” Journal of Conflict Resolution 54(3):493–515.
Sechser, Todd S and Elizabeth N Saunders. 2010. “The Army You Have: The Determinants of Military Mechanization, 1979–2001.” International Studies Quarterly 54(2):481–511.
Shambaugh, David L. 2002. Modernizing China’s military: progress, problems, and prospects. Univ of California Press.
Shy, John W. 1963. “A New Look at Colonial Militia.” The William and Mary Quarterly 20(2):175–185.
Sikkink, Kathryn. 1993. “Human Rights, Principled Issue-Networks, and Sovereignty in Latin America.” International Organization 47(3):411–441.
Singer, J David. 1988. “Reconstructing the correlates of war dataset on material capabilities of states, 1816–1985.” International Interactions 14(2):115–132.
Singer, Peter Warren. 2008. Corporate warriors: The rise of the privatized military industry. Cornell University Press.
Soule, Sarah and Christian Davenport. 2009. “Velvet glove, iron st, or even hand? Protest policing in the United States, 1960-1990.” Mobilization: An International Quarterly 14(1):1–22.
Staniland, Paul. 2012. “Between a Rock and a Hard Place Insurgent Fratricide, Ethnic Defection, and the Rise of Pro-State Paramilitaries.” Journal of Conflict Resolution 56(1):16–40.
Stanton, Jessica. 2015. “Regulating Militias: Governments, Paramilitaries, and Civilian Targetingin Civil War.” Journal of Conflict Resolution 59(5).
Stentiford, Barry M. 2002. The American Home Guard: the state militia in the twentieth century. Number 78 Texas A&M University Press.
Talmadge,Caitlin. 2015. The Dictator’s Army: Battlefield Eectiveness in Authoritarian Regimes. Cornell University Press.
Tellis, Ashley J. 2001. Measuring Military Capability. RAND Corporation.
The Military Balance. 2010. Vol. 110 chapter East Asia and Australasia, pp. 377–440.
The Military Balance. 2011. The Military Balance 115(1):493–500.
Thies, Cameron G. 2005. “War, rivalry, and state building in Latin America.” American Journal of Political Science 49(3):451–465.
199 Tilly, Charles. 1992. Coercion, capital, and the European states, AD 990-1992. Wiley-Blackwell.
Tilly, Charles, Diane E Davis and Anthony W Pereira. 2003. “Armed Force, Regimes, and Contention in Europe since 1650.” Irregular Armed Forces and Their Role in Politics and State Formation pp. 37–81.
Tollefsen, Andreas Forø and Halvard Buhaug. 2015. “Insurgency and inaccessibility.” International Studies Review 17(1):6–25.
Tresch, Tibor Szvircsev. 2010. “The transformation of Switzerland’s militia armed forces and the role of the citizen in uniform.” Armed Forces & Society .
Tucker, Spencer C. 2017. The Roots and Consequences of Civil Wars and Revolutions: Conflicts that Changed World History. ABC-CLIO.
Tyson, Scott A. 2017. “Centralized Coercive Force and the Political Foundations of the Social Contrac.” Working Paper.
Urdal, Henrik. 2006. “A clash of generations? Youth bulges and political violence.” International studies quarterly 50(3):607–629.
Van Zanden, Jan Luiten, Eltjo Buringh and Maarten Bosker. 2012. “The rise and decline of European parliaments, 1188-1789.” The Economic History Review 65(3):835–861.
Vasquez, Joseph Paul. 2005. “Shouldering the Soldiering Democracy, Conscription, and Military Casualties.” Journal of Conflict Resolution 49(6):849–873.
Vogt, Manuel, Nils-Christian Bormann, Seraina Rüegger, Lars-Erik Cederman, Philipp Hunziker and Luc Girardin. 2015. “Integrating Data on Ethnicity, Geography, and Con ict The Ethnic Power Relations Data Set Family.” Journal of Conflict Resolution pp. 1327–1342.
Waxman, Matthew C. 2009. “Police and National Security: American Local Law Enforcement and Counterterrorism After 9/11.” Journal of National Security Law & Policy 3:377.
Weber, Max. 1922. Economy and Society: An Outline of Interpretive Sociology. Vol. 1 Routledge.
Weinstein, Jeremy. 2005. “Resources and the Information Problem in Rebel Recruitment.” Journal of Conflict Resolution 49(4):598–624.
Wilder, Andrew R. 2007. Cops or Robbers?: The Struggle to Reform the Afghan National Police. Afghanistan Research and Evaluation Unit Kabul.
Wilkinson, Steven I. 2015. Army and nation: The military and Indian democracy since independence. Harvard University Press.
Wucherpfennig, Julian, Nils B Weidmann, Luc Girardin, Lars-Erik Cederman and Andreas Wimmer. 2011. “Politically relevant ethnic groups across space and time: Introducing the GeoEPR dataset.” Conflict Management and Peace Science 28(5):423–437.
200