Desertion, Control and Collective Action in Civil Wars Theodore

Home , Antonio Aranda, Battle of Santander, Iron Column, Military sociology, Radical Republican Party

Desertion, Control and Collective Action in Civil Wars

Theodore McLauchlin

Department of Political Science

McGill University, Montreal

June 2012

A thesis submitted to McGill University in partial fulfillment of the requirements of the degree

of Doctor of Philosophy

Abstract

This dissertation develops and tests a new theoretical synthesis for understanding how armed groups keep their combatants fighting rather than deserting or defecting. It examines two basic methods of limiting desertion: keeping coercive control over combatants, and fostering norms of mutual cooperation among them. It argues that the effectiveness of each approach is conditioned by the degree to which combatants value the common aim of the success of the armed group.

Norms of cooperation require a commitment to this common aim to be effective. Control can be effective even when combatants are uncommitted, but loses effectiveness with severe disagreements among combatants.

This approach provides an advance on past work on the requirements for armed groups in civil wars. Some assume, unrealistically, that common aims drive individual behaviour directly.

Others focus exclusively either on individual rewards and punishments or on norms of cooperation. This dissertation, in contrast, sees each as important and as contingent upon the prior consideration of whether combatants share a common aim.

A qualitative analysis of armed groups in the Spanish Civil War examines micro-level evidence about common aims, the provision of control, and the emergence of norms of cooperation. The dissertation then tests its major hypotheses statistically using two original datasets of soldiers from that war, based on the author’s archival research. It conducts further statistical tests against a new dataset of defection from government armies in 28 civil wars during the 1990s. It concludes with a discussion of new directions.

Résumé

Cette thèse élabore et met à l’essai une nouvelle synthèse théorique permettant de comprendre comment les groupes armés arrivent à faire en sorte que leurs membres continuent de se battre au front plutôt que de déserter ou de faire défection. Elle examine deux méthodes traditionnelles permettant de limiter la désertion, soit l’exercice continu d’un contrôle coercitif sur les combattants et l’encouragement de normes de coopération mutuelle entre eux. Elle soutient que l’efficacité individuelle de ces approches est déterminée selon l’importance accordée par les combattants à l’objectif commun de la réussite du groupe armé. Les normes de coopération nécessitent un engagement envers cet objectif commun afin de pouvoir être efficaces. Si le contrôle peut être utile même lorsque les combattants ne sont pas engagés, son efficacité est réduite lorsqu’il y a des désaccords profonds entre ces derniers.

Cette approche présente une avancée sur des travaux antérieurs portant sur la présence

nécessaire de groupes armés dans un contexte de guerre civile. Certains savants croient à tort que

ce sont les objectifs communs qui influencent directement les comportements individuels alors

que d’autres ne pensent qu’aux récompenses et punitions individuelles, ou alors aux normes de

coopération. Quant à elle, cette thèse reconnaît l’importance individuelle de ces deux méthodes

et considère qu’elles sont liées à la considération antérieure cherchant à savoir si les combattants

partagent un objectif commun.

Une analyse qualitative des groupes armés de la guerre civile espagnole traite de données

détaillées en lien avec les objectifs communs, la disposition de contrôle ainsi que l’émergence

des normes de coopération. La thèse met ensuite ses hypothèses principales à l’essai sur le plan ii statistique à travers l’usage de deux bases de données originales de soldats tirés de cette guerre, basés des recherches d’archives de l’auteur. Elle réalise des tests statistiques additionnels à partir d’un nouvel ensemble de données sur la défection d’armées gouvernementales dans 28 guerres civiles au cours des années 1990. Pour conclure, elle ouvre un dialogue portant sur de nouvelles directions.

iii

Table of Contents

Abstract i Résumé ii Table of Contents v Acknowledgements ix List of Tables xiii List of Figures xiv List of Abbreviations xv

Chapter 1. Introduction: Explaining Desertion in Civil Wars 1 1.1. Introduction 1 1.2. Defining Desertion 4 1.3. Theoretical Approach: Common Aims, Control and Cooperation 6 1.3.1. Common Aims 6 1.3.2. Control 12 1.3.3. Cooperation 16 1.3.4. Summary and Micro-Level Implications 26 1.3.5. Macro Implications of This Approach 27 1.4. Alternative Approaches 28 1.4.1. The Cause 29 1.4.2. Complex Motivations and Control 31 1.4.3. The Collective Action Problem and Selective Incentives 36 1.4.4. Social Homogeneity and the Collective Action Problem 41 1.4.5. Alternative Approaches: Summary 48 1.5. What Is to Come 49

Chapter 2. Background to the Spanish Civil War 54 2.1. Introduction 54 2.2. The Roots of War 57 2.3. The Collapse of Central Authority in the Republic 63 2.4. The Process of Reform in the Republic 70 2.5. Nationalist Spain 75 2.6. Conclusion 82

Chapter 3. Common Aims: Recruitment and Factionalism 86 3.1. Introduction 86 3.2. Voluntarism and Opportunism in the Militias 88 3.3. Voluntary Recruitment in Nationalist Units 107 3.4. Conscription and the Expansion of Motivations 112 3.5. Factionalism in Republican and Nationalist Spain 115 3.6. Conclusion 119

Chapter 4. Control and Desertion 121 4.1. Introduction 121 4.2. Control in the Republican Militias 123 4.3. The Evolution of Control in the Republic 132 4.4. Nationalist Spain: Control in the Rearguard and the Front Line 150 4.5. Conclusion 160

Chapter 5. Desertion, Collective Action, and Norms of Cooperation 163 5.1. Introduction 163 5.2. Collective Action and the Republic’s Militias 164 5.3. Collective Action in the Popular Army 180 5.4. Collective Action in Nationalist Spain 187 5.5. Conclusion 190

Chapter 6. Desertion and Control of Hometowns in Santander Province 193 6.1. Introduction 193 6.2. Desertion, Control, and Hometowns 195 6.3. The Setting: Santander Province, Spain 199 6.4. Hypotheses and Control Variables 202 6.5. Potential Biases from Case Selection 207 6.6. Data and Method 211 6.7. Results 223 6.8. Qualitative Evidence from Santander 228 6.9. Conclusion 232

Chapter 7. Collective Action and Desertion in Santander 236 7.1. Introduction 236 7.2. Group Influences and Desertion in Civil Wars 238 7.3. Military Units in Santander 241 7.4. Hypotheses 252 7.5. Method 255 7.6. Results 257 vi

7.7. Qualitative Evidence 263 7.8. Conclusion 269 7.9. Appendix 271 7.9.1. Robustness Checks 271 7.9.2. Discrepancies between Chapter 6 and Chapter 7 276

Chapter 8. Defection in Civil Wars, 1990-1994 280 8.1. Introduction 280 8.2. From Micro to Macro 281 8.3. State Capabilities and External Support 284 8.4. Militias, Agency, and Control 292 8.5. Motivations and Factionalism: Conscription and Coup Attempts 298 8.6. Empirical Analysis 304 8.7. Results 306 8.8. Potential Control Variables: Economics and Institutions 313 8.9. Conclusion: States and Civil Wars 316

Chapter 9. Conclusion: Desertion and the Dynamics of Civil Wars 321 9.1. Introduction 321 9.2. Testable Hypotheses 324 9.2.1. Initiation of Civil Wars 325 9.2.2. Civil War Dynamics: Strategy and Tactics Selection 326 9.2.3. Civil War Outcomes 329 9.3. Improving the Theory 331 9.3.1. Voice, Not Just Exit and Loyalty 332 9.3.2. Types of Desertion 333 9.3.3. Bilateral Interactions 335 9.4. Implications of Contingent Control and Cooperation 339 9.4.1. A New Synthesis 339 9.4.2. The Diversity of Civil Wars 342 9.4.3. State Building 344 9.4.4. The International Context 348 9.5. Conclusion 353

Bibliography 355

vii

viii

Acknowledgements

I must first thank Hudson Meadwell, my supervisor, for all his help in travelling this long road. His constant and perceptive critiques, demand for sharp analysis, and continuous encouragement improved this dissertation immensely and made writing it much easier than it would otherwise have been.

In Spain, Juan Díez Medrano at the Institut Barcelona d’Estudis Internacionals was extremely generous in providing me a home base and in advice and encouragement in the archives. The late Gabriel Cardona graciously shared his unparalleled insight into the military history of the Spanish Civil War. Laia Balcells helped me get oriented to studying the Spanish

Civil War. Joan R. Roses helped me navigate historical Spanish economic and wage data.

Édouard Sill and I swapped thought-provoking ideas about desertion in the Spanish Civil War.

Robert, Katie, and Elsie Kissack gave me considerable moral support in Spain whenever I needed it. They, and David Convery, Pat Cullen, Diego Funes, Rosy Rickett, Niall Smith, Pere

Soler, and Nehir Sönmez, enriched my life and work in Spain immensely.

The staff were consistently helpful at the Centro Documental de la Memoria Histórica,

Salamanca, the branches of the Archivo General Militar in Ávila, Madrid, and Segovia, the

Arxiu Nacional de Catalunya, the Biblioteca Municipal de Santander, the Biblioteca Provincial de Cantabria, and the libraries at the Universitat Pompeu Fabra, the Universitat de Barcelona, and the Universidad de Cantabria. Miguel Solla Gutiérrez graciously provided me a copy of his

Ph.D. thesis work, and conversations with Enrique Menéndez Criado helped with the politics of

Santander during the war.

Juan Andrés Blanco Rodríguez at the Universidad de Salamanca put me in touch with my research assistants in the archives. Assistance in coding the data in Spain was generously provided by Juan Carlos García, Braulio Pareja, Daniel Blanco, Patricia García, and Manuel

Talaván. Back in Montreal, Amanda Beerworth-Gervais, Louis Fouquet, Linda el Halabi, Kevin

Lu, Kayleigh Metviner, Alexis Nigro, Kartiga Thavaraj, and Marla Tyler helped me to code the cases for Chapter 8.

I gratefully acknowledge the support of many different institutions. Funding for my research came from a Canada Graduate Fellowship and a Michael Smith Foreign Study

Supplement from the Social Sciences and Humanities Research Council of Canada, a PhD

Fellowship from the Security and Defence Forum, research travel grants from the McGill-

Université de Montréal Institute for European Studies and the Faculty of Arts, and an Alexander

Mackenzie Fellowship and Graduate Excellence Fellowship from the Department of Political

Science.

I received extremely helpful feedback on this work in presentations at the International

Studies Association, the Households in Conflict Network, and Yale University’s Program on

Order, Conflict and Violence. I must thank Laia Balcells and Stathis Kalyvas for organizing the latter two venues, and Shane Barter for putting together a great panel on paramilitary forces at

ISA. I especially want to thank Ana Arjona, Laia Balcells, Shane Barter, Rob Blair, Thomas

Boccardi, Stathis Kalyvas, Bethany Lacina, Janet Lewis, Jon Monten, and Andreas Wimmer for their perceptive comments and helpful suggestions at these conferences.

Its faculty make the Department of Political Science a place of thriving and rigorous intellectual engagement. In particular, Michael Brecher made me want to be an academic in the first place. Steve Saideman provided helpful advice, encouragement and good humour all along

x the way. The germ of this dissertation came from a paper for Rex Brynen that became a journal article. T.V. Paul’s influence has constantly prompted me to think of the big picture. The dissertation was greatly aided by statistical training from Stuart Soroka. The research also benefited from methods courses with Steven Rytina and Erik Kuhonta, and from summer courses at the Inter-University Consortium for Political and Social Research in Ann Arbor, Michigan, with David Armstrong and Christopher Zorn, and the Quebec Interuniversity Consortium for

Social Statistics, with Richard Wanner.

My brilliant colleagues in the graduate program and at the Social Statistics Laboratory have been a wonderful, sustaining group of friends and comrades-in-arms. Above all, I am supremely lucky to count Ora Szekely among my great friends. Aisha Ahmad, Ece Atikcan,

Marc-Andre Bodet, Adam Chalmers, Erin Crandall, Mark Daku, Andy Dawson, Nicole Denier,

James Devine, Ginger DiGaetano, Dan Douek, Doug Hanes, Julia Honnaker, Donovan Jacobsen,

Sarah Malik, Julie Moreau, Kerem Öge, Chuk Plante, Jeff Sachs, Melanee Thomas, and Jess

Trisko made graduate school terrific. They and Guy-Philippe Bouchard, Jon Bracewell, Kat

Childs, Jen Dickson, Ali Glaser, Justin Mizzi, Mark Ordonselli, Ian Ratzer, and Philip Lemieux have propped up my sanity more times than I care to admit, but they probably knew that anyway.

Hobbit and the late Pepper reminded me of the important things in life, like feeding them.

Every now and again they graciously accepted being petted, a solid mental health break.

Otherwise they were no help at all, and in true feline fashion they don’t care.

It was hearing the voice of my late father, David McLauchlin, over the radio and talking with him about what he had seen as he reported from around the world that awakened my interest in politics at a young age. His example has been a constant companion. My family, every branch of it, is unstinting in their love and help. Lynn and Matt McLauchlin, my mother and

xi brother, are always there for me, whenever I have needed encouragement, love, and moral support. I am lucky enough to have wonderful in-laws in Jane and Bob Bracewell, who kept up a ready supply of good wishes, good cheer, and good food.

I don’t really know what to say to my wonderful wife, Jen Bracewell. I’d thank her for putting up with all this, because I know this dissertation has been hard to live with, as have I. But that wouldn’t do justice to the love that she’s given or the grace she has shown. There were points where I would not have kept going with this if she wasn’t there. Through her patient and perceptive critiques, she helped me think through the ideas behind this dissertation. Ever the archaeologist, she exhorted me to go get data at a time when that was exactly the right piece of advice. Thanking her isn’t enough, though I do, profusely. I only hope I can do the same for her, on her own research and on her own road. I owe her this, and I dedicate it to her.

xii

List of Tables

Table 1.1. Effectiveness of approaches for reducing desertion 27 Table 2.1. Empirical summary, Chapters 3 through 5 83 Table 4.1. Survival rate for attempted defectors, 1 Army Corps 147 Table 6.1. Descriptive statistics: individual-level variables 219 Table 6.2. Descriptive statistics: hometown-level variables 220 Table 6.3. Multilevel logit results 222 Table 7.1. Descriptive statistics, company characteristics averaged over time 248 Table 7.2. Occupation distribution by conscription rate 249 Table 7.3. Descriptive statistics 257 Table 7.4. Multilevel logit results 259 Table 7.5. Predicted probability of desertion – interaction effects 261 Table 7.6. Group analysis 263 Table 7.7. Robustness checks 273 Table 7.8. Robustness checks 2 274 Table 7.9. Individual- and company-level conscription 278 Table 8.1. External support, coup history, and defection 307 Table 8.2. Interaction effects 308 Table 8.3. Defection in regular and militia forces 311 Table 8.4. Connections between defection in regular and militia forces 311 Table 9.1. Defection and war outcomes 331

xiii

List of Figures

Figure 2.1. Spain, 25 July 1936 62 Figure 4.1. Monthly Rate of Defection Attempts, 1 Army Corps 146 Figure 6.1. Location of Santander Province in Spain 200 Figure 6.2. Estimated rate of desertion by municipality (% of soldiers), 202 Santander province Figure 6.3. Relief map of Santander Province 208 Figure 6.4. Predicted probabilities of desertion according to home municipality 224 steepness index and vote share, for a typical soldier, with 95% confidence interval Figure 6.5. Desertion rates by steepness decile (with standard error of estimate) 225 Figure 7.1. Conscription rate by company over time, showing average across companies 250 Figure 7.2. Changing patterns in volunteer units 250 Figure 7.3. Changing average company composition over time 251 Figure 7.4. Changing county fragmentation over time 252 Figure 7.5. Predicted probabilities of desertion 261

xiv

List of Abbreviations

AGMAV – Archivo General Militar, Ávila

CDMH – Centro Documental de la Memoria Histórica, Salamanca

CEDA – Confederación Española de Derechas Autónomas , conservative political party

CES – Cuerpo de Ejército de Santander , Santander army corps

CNT – Confederación Nacional del Trabajo , Anarcho-Syndicalist trade union confederation

FAI – Federación Anarquista Ibérica , Anarchist political organization

FET-JONS – Falange Española Tradicionalista y de las Juntas de Ofensiva Nacional

Sindicalista , joint Falange-Carlist political party

JSU – Juventudes Socialistas Unificadas , joint Socialist-Communist youth wing

MAOC – Milicias Antifascistas Obreras y Campesinas , Communist-organized militias

PCE – Partido Comunista Española, Spanish Communist party

POUM – Partido Obrero de Unificación Marxista , revolutionary Marxist party

PSOE – Partido Socialista Obrero Española , Spanish socialist party

PSUC – Partit Socialista Unificat de Catalunya , Catalan Socialist-Communist party

SIM – Servicio de Investigación Militar, military internal security and counter-espionage service

UGT – Unión General de Trabajadores , Socialist trade union federation

UMRA – Unión Militar Republicana y Antifascista , pro-Republic officers’ organization

UR – Unión Republicana , centrist political party

xvi

Chapter 1 Introduction: Explaining Desertion in Civil Wars

1.1. Introduction

This dissertation sets out to explain how armed groups in civil wars are able, or not, to prevent desertion. Sometimes—very often in some cases—combatants leave. They go home, they switch sides, or they start new armed groups. Others stay and fight, despite the hardships of war. How do some armed groups keep their soldiers fighting over long periods of time? Why do others fall apart to desertion and defection? Do the successful ones have members who are committed to the cause and trust in each other, working together for a collective goal? Does reliability grow out of the barrel of a gun, out of coercing soldiers?

In this dissertation, I take up this puzzle, focusing on its political dimensions—on the organization of force and on political conflict and cooperation within armed groups. I consider two basic ways of keeping combatants from deserting: keeping control over them and facilitating norms of cooperation among them. I argue that the effectiveness of each is conditioned by the degree to which the combatants in the armed group value the armed group’s success. When they do share this common aim, control and norms of cooperation can both be effective. Where there is less consensus, norms of cooperation are ineffective and rewards and punishments—including the exercise of control—are necessary. And when disagreement over aims is severe, mistrust prevails over norms of cooperation and control can even be undermined.

This approach to the puzzle of desertion helps us deal with certain pressing theoretical and empirical problems. Civil wars and the armed groups that fight them appear to be diverse

1 phenomena, with a few major competing images. Some see civil wars as clashes of grand societal projects or group interests. Others argue that the self-interest of individual combatants in civil wars makes such group interests effectively irrelevant, and what matters instead is the capacity of armed groups to reward and punish individuals. Still others see armed groups as potentially cooperative community endeavours, with group interests and individual egoism reconciled through norms of mutual cooperation. A final approach sees civil wars as involving concatenations of multiple and overlapping interests, with armed groups pieced together from fragile alliances and contingent upon coercion. These are images of civil wars, but also, at a basic level, images of the armed groups that fight them: how they begin and how they remain together despite the rigours of war. My account helps us see the degree of common aims among members of armed groups as a set of empirical possibilities, rather than fixed by assumption. I purport to provide a more complete image of the diversity of civil wars, showing that armed groups vary according to the preferences of their combatants and that this variation conditions the effectiveness of different approaches to limiting desertion—norms of cooperation and control.

A good understanding of desertion in civil wars can be helpful in several respects. It is fundamental to the project of thinking through the diversity of civil wars and the organizations that fight them. Desertion tells us about how well armed groups can put fighters in the field for long periods of time, a feature essential to the feasibility of a civil war in the first place. 1 Hence, desertion rates can have an important effect on how wars end. From the Russian Revolution 2 to

Libya close to a century later, 3 rebellions have often been greatly facilitated by the unravelling of government armies. D.E.H. Russell finds, reviewing fifteen rebellions, that the key predictor of

1 Collier, Hoeffler, and Rohner 2009. 2 Skocpol 1979. 3 International Crisis Group 2011. 2

success or failure is the cohesion of the government forces. 4 As for the cohesion of the rebel side, a recent survey of over eighty insurgencies shows that an important sign of their impending defeat is a wave of desertion and side-switching. 5 A case in point is the Tamil Tigers, who

suffered a wave of defection in advance of their final defeat at the hands of the Sri Lankan

government.6

Beyond outcomes, patterns of desertion have important, and complex, ramifications for civil war dynamics. Deserters switching sides can sometimes be associated with shorter wars. A wave of defection might bring a war to a swift conclusion. 7 On the other hand, the more armed groups fighting a civil war, including through splits in existing armed groups, the more difficult it is to make peace among them. The creation of new armed groups from deserters thus makes the task of conflict resolution considerably more difficult. 8 Finally, the threat of desertion strongly affects the lives of soldiers and civilians. As I argue throughout this dissertation, the threat of desertion creates large incentives for control—that is, the threat and use of violence— against soldiers and their families. This is an under-explored form of violence. The risk associated with desertion and threats made against deserters also give individual combatants strong incentives to try to “pass,” and their ability to navigate this realm of political identification can have an important effect on their lives during wartime.

Much of this dissertation explores micro-dynamics. It examines processes of conducting recruitment, instituting control, and fostering cooperation, and the impact of these processes on desertion. I carry out this micro-level analysis in a specific empirical setting, the Spanish Civil

War (1936-1939). However, I also explore the macro-level consequences of my approach. I

4 Russell 1974. 5 Connable and Libicki 2010. 6 Staniland 2012. 7 Fearon 2004. 8 Doyle 2002; Cunningham 2006. 3 argue that the account of desertion offered in this dissertation enables a clearer understanding of the international dimensions of internal armed conflict, addressing central questions of international involvement. What are the ramifications of external support to actors in armed conflicts? When does such support facilitate the construction of stable armed forces that are able to prevent desertion? When is it inefficient, undone by unreliable local agents? Illuminating this question promises a better understanding of the international politics of civil war, particularly the consequences of external support for armed groups. 9 It is a question of sharp importance to third

parties such as NATO, supporting one side or another, as in Afghanistan after 2014; actively

intervening, as in conflicts like Libya; or contemplating such intervention in conflicts like Syria.

It is also relevant to international organizations attempting to determine how best to conduct

security-sector reform in post-conflict environments. I argue that external military support can be

of help in preventing defection, because it facilitates the provision of rewards and punishments

including control. I argue as well, however, that the mistrust that appears to dominate in some

armed groups does pose a serious barrier to even the short-term effectiveness of support. And I

also caution that external support could pose long-run problems for the sustainability of control.

In the remainder of this introductory chapter, I provide a definition of desertion, outline

the theoretical approach that informs the rest of this dissertation, and situate my work within past

approaches to the subject. I conclude with an outline of the chapters to come.

1.2. Defining Desertion

I define desertion as a member of an armed group (a combatant) voluntarily ceasing to fight for that group. The one exception is failure to re-enlist at the end of a term of service. A

9 Regan 2000; Balch-Lindsay and Enterline 2000; Saideman 2001; Balch-Lindsay, Enterline, and Joyce 2008; Salehyan 2010; Salehyan, Gleditsch, and Cunningham 2011. 4

deserter can return to civilian life or join another armed group. I call the former “desertion

proper,” and the latter “defection.” Defection could entail either changing sides to a previously-

existing armed group (side-switching) or creating a new armed group (fragmentation).

The definition I offer follows Gary Goertz’s approach to social science concepts. For

Goertz, “to develop a concept is more than providing a definition: it is deciding what is

important about an entity.” 10 In line with this maxim, I have included under a single heading

three distinct destinations for deserters (civilian life, a pre-existing armed group, and a new

armed group), because I want to focus theoretical attention on the decision to leave and the

efforts of the armed group to prevent soldiers from leaving. There will be differences among

these three phenomena. For example, defection rests on a perceived advantage in joining a

different group, whereas desertion proper may just engage a desire not to fight at all. Soldiers

facing a particularly powerful adversary would be much more inclined towards side-switching

than starting a new group. However, all three entail exit , and so all three raise the question of

what barriers to exit exist. I do consider the decision about where to go at times, however. In

addition, in Chapter 8, conducting a broad overview of civil wars in the 1990s, I focus my

attention specifically on defection both because of its particular importance (it pits a combatant’s

fighting power actively against his former armed group) and for reasons of data availability.

This concept of desertion abstracts from legal definitions contained in codes of military

justice. It does so because such legal definitions vary, even if desertion “has always been

regarded as an offence of the greatest magnitude.” 11 For example, the practice of leaving a unit

(but not the army) to avoid onerous or dangerous duty without intent to remain away

permanently, known as “short” as opposed to “straight” desertion, steadily became defined as a

10 Goertz 2006, 27. 11 Avins 1963, 91. 5 form of desertion over the late 19 th and early 20 th centuries in Anglo-American law. The specific conditions identifying an instance of this action vary from country to country. 12 Indeed, in some

armed groups in civil wars, especially those that are regularly organized, rules against desertion

may be codified or not, just as formal codes of conduct could exist or not.13 Abstracting from legal definitions allows a baseline for comparative purposes.

Desertion is an action that is distinct from other forms of acting against one’s own armed group, such as illicit abuse of civilians, a failure to follow orders, or espionage. These are certainly related concepts, since, like desertion, they involve individual combatants acting at cross-purposes to the armed group. What is distinct about desertion is that it entails ceasing to fight for the group. A combatant may abuse civilians but also fight; disobey some orders but follow others; or spy for one side while fighting for the other. Desertion, in contrast, entails exit.

It therefore entails crossing the barrier between being in an armed group and being out of it, in a way that the other cited acts do not.

1.3. Theoretical Approach: Common Aims, Control, and Cooperation

What, then, accounts for different patterns of desertion? I explore the implications of three interrelated considerations: the degree to which combatants value the armed group’s success; the armed group’s exercise of coercive control over its combatants; and norms of cooperation among combatants themselves. I discuss each in turn.

1.3.1. Common Aims

I begin with the assertion that combatants will vary in the value they place on their armed group’s success on the battlefield. I refer to this as sharing the armed group’s common aim.

12 Avins 1961. 13 Weinstein 2007. 6

Valuing the armed group’s success is the minimal condition for a soldier to have a collective aim

at stake, that is, an aim shared among multiple individuals (unlike army wages, adventure,

personal glory, looting, vengeance, rape, and leadership ambitions, to name several individual

aims that are often sought in wartime).

Some combatants will place considerable value on their armed group’s military success.

This can be instrumental to a variety of further aims. Most directly, some could value battlefield success for the sake of rather direct gains: the defense of a town against a feared invader; the acquisition of a lucrative plot of land; reflecting in personal glory from unit success. Indirectly, battlefield success can be instrumental towards victory in the civil war and the implementation of a political program. If one’s armed group succeeds in battle, it increases the likelihood that that political program will come to fruition. 14 I exclude from this definition conditions in which armed group success is regarded as directly and immediately necessary for a soldier to remain alive. In those circumstances, fighting increases one’s chances of survival, rather than decreasing them.

It is not a given that soldiers have any particular preference for battlefield success. Many combatants may be apolitical or neutral between different political programs. Even on a more local level, they may consider that there is no real difference between one side ruling over their hometown and another. Still others may be caught on the “wrong” side of the war, in territory controlled by the side that in principle they would rather see lose. The number of people whose support for a cause is intense and unyielding, prompting action regardless of circumstance, is

14 Because of this variation in final aims, soldiers may vary in the circumstances in which they actually care about their armed group’s success. Someone who just wants to defend his hometown may not care one way or another about a battle occurring hundreds of kilometers away. For this reason, those with a strong attachment to the armed group’s political program may be particularly useful fighters, because they retain a commitment under a broader array of circumstances. They can be sent anywhere, and still value what their military unit can achieve. I do not distinguish much between these final aims in this dissertation; what matters mainly is whether the combatant seeks the armed group’s success, not the reasons for it. But this is a potentially interesting avenue for future research. 7 generally quite small; most people’s actions will vary greatly depending on their circumstances.15 We can therefore think of soldiers as arrayed on a continuum of intensity of preferences: at one end are combatants who strongly value the success of their side; at the other end are combatants who strongly value the success of the other side; in the centre are combatants who are wholly indifferent. 16

Variations in preference mean that the lukewarm have a higher baseline likelihood of deserting or defecting than hard-core supporters. Still likelier are those who are strongly opposed to the group and only join under duress, for example. 17 At a basic level, strong supporters will more highly value their marginal contribution to the armed group’s success. When a soldier is indifferent about the group’s success, and hence only fights for the balance of individual rewards and penalties he faces, he is likelier to desert if the balance goes awry—when facing a difficult battle, for example, or the cold, or a cut in wages. Those soldiers cannot be sustained through tough times by the prospect of contributing to a common aim. Of course, as Mancur Olson insists, 18 in a large group like an army, any given individual’s contribution to the likelihood of victory will tend to be extremely small. But in addition to the expected benefit from increasing the marginal likelihood of victory, individuals who share the common aim may be more likely to realize certain benefits obtained from the mere act of participating, such as Elisabeth Wood’s concept of “pleasure in agency.” 19

To summarize, other things equal, combatants should be less likely to desert the more they share the armed group’s common aims, but this effect may not necessarily be very large. By

15 DeNardo 1985; Kuran 1991; Lichbach 1995. 16 James DeNardo (1985) makes use of a similar continuum as a basic theoretical move in studying revolutions. 17 This argument underpins theories of side-switching in revolutionary situations in DeNardo 1985; Kuran 1989; Kuran 1991. 18 Olson 1971. 19 Wood 2003. 8

itself, the direct effect of different preferences about the armed group’s success is not terribly

important to my account. What matters more is how combatants’ aims, and whether they share

the common aim of the group’s success or not, are mediated through control and norms of

cooperation, two points that I develop further below.

The question then becomes how armed groups can successfully ensure that they have

committed combatants. I focus on two basic considerations: recruitment and factionalism. Who,

then, is committed, and who is uncommitted? The armed group can have serious difficulty in

answering this question. An individual inclined to leave has very strong reason to keep this

preference hidden until the chance comes. Given the likelihood of preference falsification in

recruiting and managing soldiers, armed groups face a basic principal-agent problem: 20 combatants typically know more about their own intentions than does the armed group’s leadership. 21 This does not always mean that soldiers are fully informed even about themselves: it is extremely difficult to know what combat is like until one has experienced it, so one’s intentions about joining and staying may change. 22 But in terms of basic orientations, the asymmetry of information likely holds.

Actors in civil wars sometimes try to use different kinds of observable indicators of people’s preferences or intentions. “Where political parties associated with one side of a conflict boycotted elections, as in post-WWII Greece and Colombia, electoral registers became depositories of information about each person’s loyalties... During the Russian Civil War, the

Whites sometimes determined who was a Bolshevik by looking for callused hands.” 23 Hutu

20 Spence 1974; Laffont and Martimort 2002. 21 Weinstein 2005. 22 Grossman 1996; Gill 2010. 23 Kalyvas 2008, 1047. 9 militias infamously used Rwandans’ identity cards to determine if they were Tutsi. 24 And, as I detail in later chapters, armed groups in the Spanish Civil War frequently used individuals’ political membership as an indicator of loyalty to the cause, a method of screening out supposedly unreliable individuals.

How useful are such labels in ensuring that only committed combatants are able to join?

The key issue is whether they are actually consistent in identifying friend from foe. The primary

danger they are used to avoid is the false negative—treating an individual as a likely friend when

he is in fact a foe. A visible indicator such as a membership card, an entry on a voter roll, or a

label on a passport, only works as an indicator of loyalty to the extent that its presence

consistently indicates loyalists rather than hiding opponents. To the extent that there were, for

example, Marxist journalists or intellectuals with smooth hands, the Whites risked false

negatives in using callused hands as their indicator of Bolshevism.

One important influence on whether the signal in question improves the accuracy of

judgments is whether it is costly or cheap to send. If a signal is cheap—and I argue that joining a

political party or union frequently was a cheap signal in Spain—then the problem collapses to

the initial problem of preference falsification. The head of the armed group does not believe my

commitment until I do X, but X is very cheap to do, so I can do it whether I am actually

committed or not. It is only as a signal gets costly that it is really informative. 25 Using cheap

signals, therefore, is likely to provide little benefit in identifying likely deserters.

Weinstein argues that the nature of the armed group that an individual joins means that

joining can constitute a costly signal in and of itself. If joining the armed group is more costly

and difficult than its alternatives, then it indicates a relatively strong commitment on the part of

24 Mamdani 2001, 222. 25 Spence 1974. 10 those who join. For example, volunteers do not have to join, but they do; for conscripts, in contrast, joining is often less costly than not joining. As such, it is generally easy to anticipate that conscripts will be less committed to the cause than volunteers. In addition, however, the analysis of cheap versus costly signals helps us understand distinctions among volunteers themselves. If the armed group places severe demands upon its recruits, requiring them to fight at the front, for example, rather than serving as a local militia, and guaranteeing them little in the way of loot or personal wealth, then the uncommitted may seek other options: remaining in private life, for example, or joining an armed group with better provision for short-run gains. 26

To join an undemanding armed group that provides significant monetary reward is a cheap signal of commitment, in contrast, because the risks are low and offset by short-run gains. We should therefore expect demanding, disciplined armed groups with low short-run rewards to attract more committed recruits than lax groups offering substantial short-run rewards.

Moreover, combatants may come from multiple different political factions in alliance with each other. Like allies in international conflict, they may share the aim of defeating a common foe, but conflict about much else. 27 For example, the government armed forces in

Rwanda’s civil war and genocide in the early 1990s included southern Hutus who sought both

the defeat of the rebel Rwandan Patriotic Front and an increased share in government power. 28 In the Kurdish nationalist movement in Iraq, there have been significant internal divisions over land reform and rivalries among local notable families. But such divisions certainly do not have to have anything to do with matters of substance; the two leading Kurdish nationalist groups in

Iraq, for example, have frequently acted just as rival leadership groups each seeking to advance

26 Weinstein 2007. 27 Snyder 1997, 165. 28 Lemarchand 1994, 600; Kakwenzire and Kamukama 1999, 68; Jones 2001, 33. 11 its own power within the movement. 29 In consequence, alongside seeking battlefield success,

members of different factions may seek to promote their own faction’s particular interests at the

expense of others. For example, they may try to gain preferred access to military supply and

payment, to secure the promotion of allies within the military hierarchy, or to denounce rivals.

They may have competing ideas about how to run the war, and have to divide burdens between

each other. Disputes about such matters frequently impede military efficiency. Arms may be

dispensed, promotions conferred, and military justice meted out based on factional membership

rather than military necessity.

More generally, there are broader questions of which faction leads the armed group.30 I assume each faction is ultimately out for its own victory. Faction A may try at some point to seize control of their side and put it towards its own aims, marginalizing Faction B’s agenda. Or

Faction A may be willing to make a separate peace with the adversary. At that stage, then,

Faction B’s military efforts are exploited by Faction B: it pays the costs of fighting but does not reap the rewards. In addition, members of Faction B may simply believe that Faction A is plotting these forms of exploitation. Hence, factionalism can present additional aims that interfere with the common aim of overall military success. The greater the degree of conflict between these aims, the more likely the common aim of defeating the opponent can be overwhelmed.

1.3.2. Control

Thus, armed groups may vary greatly in the degree to which their combatants hold the common aim of the armed group’s military success. Some combatants may only have a weak

29 Romano 2006, 197. 30 This distinction between particular issues under dispute and the more general issue of fears of exploitation owes much to Glenn Snyder’s analysis of alliance management. Snyder 1997, 170–185. 12 preference, others may be neutral, and still others may have a strong preference for the other side’s victory. In addition, some combatants may pursue narrow agendas in temporary alliance with the armed group as a whole, with some interests that align with other factions and other interests in competition. And each combatant, whatever his aims, will pay serious costs to fight.

Faced with these sources of variation in common aims, the armed group can alter combatants’ incentives and to make it in their self-interest to fight, if it has the power to issue rewards for fighting and punishments for desertion. If they are sufficiently large, then rewards and punishments, I argue, can limit desertion whether the combatant in question actually shares common aims or not. In the bulk of this dissertation, I focus on control, by which I mean the threat of physical force (that is, coercion, in Thomas Schelling’s formulation 31 ) to deter desertion. Armed groups vary in the degree to which they can impose central control over soldiers. A credible threat of significant harm for desertion can induce compliance even among individuals without any commitment to the group’s common aims.

But control raises an important concern. How can combatants trust that their leaders will inflict punishments based on actual behaviour? To the extent that punishments are meted out arbitrarily, there is less perceived benefit in remaining fighting rather than deserting: the combatant might be punished anyway. And there are occasionally incentives for such arbitrary justice. Leaders could punish political rivals, for example. They could seize a soldier’s goods by denouncing him as a traitor. They could punish soldiers on suspicion rather than on proof in order not to have to pay the costs of actually monitoring soldiers’ real behaviour. Soldiers can denounce comrades they do not like. Thus while intensive use of internal security forces has to some extent helped regimes in the Middle East keep reluctant soldiers fighting for their regimes, the frequent use of group-based stereotypes—suspicions of Shi’ite or Kurdish officers in Iraq or

31 Schelling 1966. 13

Sunnis in Syria—has prompted members of those groups to fear arbitrary and discriminatory punishment unrelated to their actual actions, and merely to advance a sectarian or ethnic elite. 32

Thus control involves a dilemma similar to the sort analyzed in cooperation theory: a leader can create incentives for individuals to contribute to a team,33 but that leader could also mete out punishments arbitrarily for his own benefit at the expense of common aims. At that stage, there is little point in cooperation for many team members; they pay the costs of working, and still endure the punishments meted out to those who shirk. 34 In political life, the power of the central state poses a similar dilemma for subordinates: the same power that can protect them from harm, punishing those who threaten social order, can likewise be used arbitrarily to enrich and strengthen the powerful. 35 In civil war studies, an analogous problem emerges in the issue of

indiscriminate (as against selective) violence against civilians in war. As Kalyvas outlines,

indiscriminate violence is generally counterproductive, provoking civilians to support one’s

opponents because they have little to gain from supporting the armed group that is

indiscriminately persecuting them. 36

Increasing control has countervailing consequences for the decision to desert. On the one hand, it increases the chance that a combatant will be punished; on the other, it has the potential to provoke desertion out of fears of arbitrary force. In an environment of particularly severe mistrust arising from disagreement about aims—where there are many political opponents fighting for the “wrong” side, or multiple political factions pursuing their own agendas—fear of arbitrary punishment is often rife. Perceived supporters of the adversary have obvious reasons to fear such persecution. In factionalized environments where multiple groups come together for

32 Brooks 1998; Quinlivan 1999; Bellin 2004; McLauchlin 2010. 33 Alchian and Demsetz 1972. 34 Bianco and Bates 1990; Miller 1992, 155; van der Heijden, Potters, and Sefton 2009. 35 North and Weingast 1989; Saideman and Zahar 2008. 36 Kalyvas 2006, 114. See also Goodwin 2001. 14 the common aim of defeating an opponent, their other disagreements can present serious problems of mistrust as well: one faction might fear that efforts at control will really just be the arbitrary exercise of violence against them. Therefore, particularly intense disagreements about aims can partially undermine control by creating incentives to abuse authority and increasing fears among combatants that such abuses will occur.

This is not to suggest that control becomes wholly ineffective in such circumstances.

After all, a combatant who bears a significant risk of persecution even if he does not desert may still believe that he bears a much higher risk of punishment if he does desert. The argument here is just that the effectiveness of control is considerably reduced in this environment of severe disagreement, not that it is necessarily wholly eliminated. The implication is that even if improvements to control can help armed groups prevent desertion in a wide array of soldiers’ preferences, control is not equally effective for all distributions of preferences. Its effectiveness can be sharply reduced in circumstances of severe disagreement about aims.

Other sources of variation in the effectiveness of control can be imagined. Notably, for example, scholars have discussed limited institutions under checks and balances as ways of keeping leaders’ commitments to their followers credible. 37 One could therefore suppose that military justice is more effective to the extent that it is dispensed according to regular procedures with limits on the arbitrary use of punishment. There is much value in such a perspective, but I do not examine it here, leaving it open for future research.

In Chapter 8, looking at general patterns of civil war in the 1990s, I relax the focus on control alone, and also examine material rewards and the payoffs of fighting on a winning side.

The logic is, in essence, the same: these are top-down ways that the armed group can affect the combatant’s cost-benefit analysis of serving versus deserting. They can also be undermined by

37 North and Weingast 1989; Weinstein 2007; Acemoglu and Robinson 2012. 15 disagreement about aims and factionalism, since material rewards can be distributed to the advantage of some and not others, and since the “winning side” might get hijacked by one’s political opponents. But for most of the dissertation, examining the Spanish Civil War, I focus on control as an example of rewards and punishments, as it was the main source of variation across military forces and over time within that war.

This account of control generates some basic hypotheses, then:

H1. The more extensive the armed group’s control over its soldiers, the less they are likely to desert, provided that combatants perceive that punishments are meted out for actual actions and not arbitrarily. Control is exercised in two areas: • the front, with monitoring of combatants by their superiors; • civilian life, with monitoring behind the lines.

H2. Intense disagreement about aims, such as when the armed group includes many supporters of the other side and when there is intense factional competition, increases the expectation that punishments will be handed out arbitrarily.

1.3.3. Cooperation

Even if combatants share the common aim and thus value the success of the armed group, they still might desert. The costs of fighting a war may strike them as too high, and their own contribution to likely success marginal. By the arguments above, I merely mean to imply that combatants more committed to the common aim have a higher tolerance for the costs of fighting a war, not that they—apart from a very small minority—have an infinite tolerance. In this sub- section, however, I develop the major reason for an association between combatants’ common aims and lower desertion rates: groups of combatants who share a commitment to the common aim can develop norms of mutual cooperation to carry on fighting rather than deserting. Such norms of cooperation have larger ramifications beyond the committed combatants as well, reducing the costs of fighting for everyone in the unit and providing the basis for social sanctions

16 against deserters. This line of analysis takes the story past the bilateral relationship between an individual combatant and the armed group’s leadership, and into the realm of multilateral relationships among groups of combatants.

The costs of fighting a war to its participants are generally assumed to be quite high

compared to the costs of remaining in civilian life. There is a constant risk of death, the

expectation of enduring harsh conditions to fight, and the personal trauma of killing. Even those

who share the common aim may find the costs too high to bear in comparison to the contribution

that each individual makes to the success of the armed group. They can therefore free-ride on the

contributions of others, reaping the benefit from the success of the armed group (if it occurs)

without paying the costs of participating. This is the classical collective action problem as

Mancur Olson presents it. 38 The key assumptions of the problem are: (1) contribution is costly to each player; (2) each player benefits the more others contribute; (3) the benefit of others’ contribution is large enough that each player does better if everyone (including himself) contributes than if no one at all contributes. The set of games covered by these assumptions includes the familiar two-person Prisoner’s Dilemma, widely used across political science to capture problems of cooperation. 39 It also includes multi-player cooperation problems like the

difficulty of production in teams, multi-player Prisoner’s Dilemma, and the Tragedy of the

Commons. 40 The “dilemma” in these cases is this: each player has clear individual incentives not to contribute, because of the costs of contribution (assumption 1). But if everyone recognized this, and contributed, they could achieve a better outcome, as seen in Assumptions 2 and 3.

This applies well to a military context in which soldiers value the armed group’s common aims sufficiently. Desertion—both to the rearguard and to another armed group—can easily be

38 Olson 1971. 39 Jervis 1978; Axelrod 1984. 40 Hardin 1968; Alchian and Demsetz 1972; Schelling 1973; Ostrom 1990. 17 seen as non-contribution. Assumption 1, then, is fulfilled by the costs of fighting to each combatant. Assumption 2 may be fulfilled if a combatant benefits by others’ contributions to group success—that is, if he shares the common aim. Assumption 3 requires that the combatants benefit enough from the armed group’s success that each would prefer everyone contributing

(including paying the costs of contributing himself) to no one contributing.41

Thus the question of fighting or deserting can resemble a collective action problem, specifically a multi-player Prisoner’s Dilemma. We might therefore conclude, as does much of the literature on collective action, that attachment to common aims is effectively irrelevant: what really matters, in a group the size of a typical armed group, are direct rewards and punishments such as a soldier’s wages and the prospect of getting caught when one deserts.42 At that stage, there is little reason to move beyond a theory of desertion based solely on those rewards and punishments. I do accept that the latter can, indeed, help to overcome this collective action problem, and I focus specifically on control as an instance of this.

However, a second approach presents itself for groups of combatants who share a common aim. This is to devise a reciprocal norm of cooperation to keep fighting. Such norms— which I understand as an expectation, shared among a group, of mutual cooperation—have received considerable attention in the formal literature on collective action. Much of the work is in iterated two-player Prisoner’s Dilemma. 43 Here, each player cares about the future

consequences of his cooperation behaviour. Each can threaten to not contribute in response to the

other player’s non-contribution. These threats can make mutual cooperation an equilibrium,

41 The condition of holding a common aim does not preclude defection under this scenario; if the other side offers the individual defector a sufficient reward, then the payoffs from different decisions are essentially the same. 42 Tullock 1971; Collier and Hoeffler 2004. 43 Taylor, however, argues that the division of lumpy (as opposed to continuously divisible) goods can be better represented using the chicken game. Taylor and Ward 1982; Taylor 1987. 18 provided players care sufficiently about the future. Mutual non-cooperation 44 remains an equilibrium, but it is now one among many, including cooperation. 45 This solution, then, is

interactive rather than top down. Scholars such as Russell Hardin and Elinor Ostrom 46 have

explored how the development of the principle of reciprocity in groups can result in cooperative

outcomes.

In a warfighting setting, then, norms of cooperation can help to reduce desertion. Such

norms establish that each will fight if others do. This sort of arrangement requires, however, that

a sufficiently large group of combatants values the armed group’s success. Otherwise,

Assumptions 2 and 3 are more likely to be violated. Consider this theoretically first. In the

iterated two-person game, if a player is indifferent between mutual cooperation and mutual non-

cooperation or indeed prefers the latter, then the prospect of gains from future cooperation does

not at all deter him from non-cooperation. Assumptions 2 and 3 are violated directly. Indeed, the

game stops being Prisoner’s Dilemma, and starts being some other game entirely, like Deadlock:

a situation in which players prefer mutual non-cooperation to mutual cooperation. 47 If a player is not strictly indifferent between the two but has only a marginal preference for mutual cooperation, then he has to care quite greatly about the future for that prospect of cooperation to be preferable to non-cooperation. Hence the conditions for cooperation are highly restrictive if a player’s preference for mutual cooperation is only marginal.48 According to Tsebelis, the same

44 I use the term “non-cooperation” rather than “defection”, despite the former’s infelicity and the common use of the latter term in these games, so as to avoid confusion with defection as a specific act of exiting an army to join an adversary. 45 The most prominent treatment is Axelrod’s (1984). The “Folk Theorem,” though, makes clear that any payoff better than mutual non-contribution is an equilibrium provided discount rates are sufficiently high. Friedman 1971; Fudenberg and Maskin 1986. 46 Hardin 1982; Ostrom 1990. 47 For a set of two-person games as arrayed by preferences, see Tsebelis 1990, 62–68. 48 Formally, as Morrow (1994, 265–266) shows, for the “grim trigger” (permanent non-cooperation) strategy to be effective, each player’s discount rate must be higher than (T – R) / (T – P), where T is the temptation from cheating while the other player cooperates, R is the reward from mutual cooperation, and P is the punishment from mutual 19 argument holds when each player can adopt contingent strategies, based on their subjective assessments of the probabilities of what the other player will do. In those circumstances, the likelihood of cooperation increases with the rewards from cooperation and decreases with higher payoffs from non-cooperation. 49

In a multi-player context, the prospects of cooperation depend on the composition of the group in question. Schelling’s multi-player Prisoner’s Dilemma game 50 requires that there be a number k such that if k players cooperate, those players do better than if everyone had not cooperated. This is equivalent to my Assumption 3. But assume that there are two sets of players.

One set of players does indeed do better if they cooperate, provided a sufficient number of other players do so as well, than if everyone else had not cooperated. But for another set, this is not the case: there is no scenario in which they prefer to be part of a cooperating group than for everybody not to cooperate. The larger the proportion of the latter type relative to the whole, the less the likelihood of cooperation emerging. The prospect of everybody doing better and achieving the common aim has no real weight for this latter set, so they cannot be included in a group that prefers to cooperate together over everyone not cooperating. Thus the more players who are indifferent about the common aim, the harder it is for the first set to find a group of size k of cooperative players.

In the military context, groups of combatants who share the common aim may have an encouraging, desertion-reducing effect on each other. Such a group can establish a norm of reciprocity: combatants fight rather than desert if others do too. But in order for these norms to be sustained, each soldier must expect that others will fight. This is particularly important

non-cooperation. Hence, as R and P get closer together, the discount rate required approaches 1, so the range of discount rates in which there is a cooperative equilibrium steadily disappears. 49 Tsebelis 1990, 70–71; Aumann 1974. 50 Schelling 1973. 20 because of the interactive nature of the battlefield. The presence of other soldiers reduces the cost of fighting to each soldier. Combatants can provide covering fire for each other, protect each other’s flanks, and share information. A soldier is, therefore, likely to place a very high premium on how reliable he believes the other soldiers in his unit to be. If he trusts that they recognize a shared aim, and that they see that they will all be better off if they all contribute, then he can anticipate lower costs for fighting rather than deserting. His own contribution will not be betrayed by the non-participation of others. He will then be more likely to fight, and in fighting, will encourage others to fight as well.

This is not a guarantee that groups of committed soldiers will necessarily develop such norms. A fear that others will desert makes each soldier more likely to desert. Just as mutual cooperation is only one equilibrium among many in theory and mutual non-cooperation is an equilibrium as well, it is possible that each soldier could anticipate that others will desert and so desert himself—confirming the others’ suspicions. Common aims make norms of cooperation more likely; they do not make them certain.

In contrast, in groups of mainly indifferent soldiers, the prospect that others around you do not share the common aim is likely to have a discouraging effect. There is little chance to appeal to others’ desire for a common benefit from cooperation. It will not be of interest to uncommitted combatants that they could achieve something together if they all agreed to cooperate. Indeed, if soldiers prefer to leave and this preference becomes known as a common aim, then norms of cooperation can help them to desert together rather than fight together. Even short of this, a soldier’s inability to count on the men around him makes combat more risky. If he keeps on fighting, he expects that he could be put in particularly strong danger. Desertion becomes a much more attractive prospect.

Norms of cooperation may also be undermined in the context of factionalism. Even if combatants share the common aim of defeating the opponent, they may also have goals in competition with each other, as discussed above. The collective action problem is more severe than when such competitive interests do not divide the combatants. In a group of combatants who are divided between different factions in severe competition against each other, it is likely to be particularly difficult to recognize that a common aim exists; that common aim can be overwhelmed by the prospective gain from marginalizing an opponent. Rather than earning a reciprocal effort from an ally, a combatant’s decision to fight may be rewarded with efforts to marginalize his political program, pass him over for promotion, or even denounce him as a traitor. Intense factional competition therefore makes it less likely that norms of cooperation will be put in place, and more likely instead that combatants will mistrust each other and desert rather than suffer the consequences of exploitation.

Thus norms of cooperation can take root in a group of soldiers who share a common aim, as a solution to their collective action problem. But the presence of a large group of combatants committed to the common aim can also have second-order effects on uncommitted combatants, in two ways. First, as noted above, the costs of fighting depend on how many others will fight rather than desert. Consider, then, a combatant who is indifferent about the success of his unit.

His decision about whether to desert, then, has nothing to do with the prospect of achieving a common goal in coordination with others. It instead has to do with a balance between the costs and benefits of fighting and the costs and benefits of deserting. Holding constant the other costs and benefits at play (such as the wage from fighting and the risk of punishment from deserting), the lower the costs of fighting, the more likely that the soldier will prefer to fight over deserting.

If he knows that there are many soldiers in the unit who share the common aim and who have

22 established a norm of cooperation to uphold it, then he anticipates that fighting will be less costly than if he is surrounded by the uncommitted. This establishes that the presence of committed troops can reduce the likelihood that uncommitted troops will desert.

In addition, when a group develops a norm of cooperation, it can enforce that norm through socially, multilaterally enforced sanctions against violators. Most importantly in the military context, combatants can choose to actively monitor each other or not, and to report deserters or feign ignorance about their whereabouts. As Mancur Olson rightly points out, social sanctions are direct punishments: they distinguish between those who violate the accepted social practice and those who do not.51 These social sanctions should therefore affect the decisions of

the uncommitted as well as of the committed. If there are collectively enforced punishments for

deserters, then any combatant, committed or not, pays these costs when he deserts. Thus the presence of individuals committed to the common aim, who can develop norms of cooperation and the social sanctions that help to uphold those norms, can keep the uncommitted in line.

Uncommitted soldiers can thus be encouraged to keep fighting by the presence of committed soldiers. The difficulty, of course, is that uncommitted soldiers can affect the committed as well. The more uncommitted soldiers there are, the more the committed ones will anticipate that if they fight, other men in their unit will let them down. The issue becomes, then, a balance between the committed combatants who want to maintain a norm of cooperation, and the uncommitted combatants who weaken that norm. The larger the proportion of committed personnel in a unit, the lower the likelihood that any combatant in that unit will desert.

A final point about norms of cooperation relates to social homogeneity—similarity among combatants on social characteristics. If combatants share a common aim, norms of cooperation can be further facilitated by social homogeneity. First, social homogeneity can entail

51 Olson 1971, 60–61. 23 knowing specific other individuals or having connections to them. People who know each other well have a stronger basis for reciprocity. Knowledge of others’ interests, of whether they are aware of the game being played, and of the credibility of their threats all help in enforcing agreements. 52 Ties among members of a group facilitate the provision of such knowledge. Ties also form the basis of iteration, because they stretch interaction among group members over time. Hence, for Taylor, the stability and persistence of a community are vital to the enforcement of cooperation within that community. 53 In Axelrod’s terms, communities carry with them the

“shadow of the future.” The necessity of future interaction serves as the basis for tit-for-tat strategies: one contributes out of the belief that others will in the future. 54 The facilitating role of direct social ties led Coleman to popularize the term social capital to refer to such connections. 55

Beyond specific knowledge of other individuals, social homogeneity can facilitate trust.

It is important for individuals to gauge whom they can trust—that is, to identify if someone else has a common interest, and that the other person understands this to be the case. 56 If group

members have difficulty maintaining trust, they are more likely to anticipate that their own

contributions will not be matched by others’. Some scholars have argued that social homogeneity

provides the basis for trust. 57 For example, ethnically diverse communities have been found to

suffer from a lower degree of public goods provision, such as the construction of wells, than

comparable ethnically homogeneous communities. 58 A field experiment in Kampala employing

Prisoner’s Dilemma games finds that this is because individuals from different ethnic groups

often do not trust each other well enough to demonstrate reciprocity: there is too much fear of

52 Hardin 1982, 183. 53 Taylor 1982, 91. 54 Axelrod 1984. 55 Coleman 1990, 300–321. 56 Cook, Levi, and Hardin 2009. 57 Glaeser et al. 2000. 58 Alesina, Baqir, and Easterly 1999; Miguel 2004; Miguel and Gugerty 2005. 24 non-cooperation. In contrast, within ethnic groups, there is a stronger anticipation of reciprocity.

This applies even if the individuals in question do not actually know each other personally. 59

Foddy and Yamagishi find similar results in an experiment conducted with ties that are apparently much weaker than ethnicity: university affiliation and majors within universities.

They argue that greater in-group trust has little to do with a stereotype of superiority, and much more to do instead with the expectation of generalized reciprocity: trust and cooperative behaviour will be rewarded within in-groups, but not as much with out-group members. 60 Within armed groups, then, individuals with different characteristics may be less inclined to trust each other—and, in particular, to trust that they will help each other manage the rigours of war.

Sensing that others will not help them enough to make fighting bearable, soldiers may come to believe that it is no longer worth fighting.

However, I argue that for social homogeneity to reduce desertion rates, a common aim is required in the first place. It is important to note that experimental studies such as Prisoner’s

Dilemma games assume shared goals by design. 61 However, if there is strong disagreement about

goals, social homogeneity should not be of any assistance. 62 If there are common goals, norms of reciprocity can be powerful motivators to individuals to carry on fighting despite the costs, and those norms can then underpin social sanctions at the margins. Social homogeneity, in this setting, can facilitate trust by easier identification of common aims, and can also facilitate the infliction of social sanctions. In contrast, without such common goals, social homogeneity will

59 Habyarimana et al. 2007. 60 Foddy and Yamagishi 2009. 61 Habyarimana et al. 2007; Foddy and Yamagishi 2009. 62 Social heterogeneity could, in some settings, just imply different goals. Alesina et al (1999), for example, treat ethnic diversity as just an indicator of different preferences about public goods in American cities, and then it is this goal heterogeneity that drives their analysis. Hence, findings linking social heterogeneity to lower public goods provision may just be picking up goal heterogeneity instead; the two concepts tend not to be clearly separated in the empirical literature. I argue, in contrast, that such separation is analytically important, and I attempt it empirically in Chapter 7 by examining the distribution of conscripts and volunteers, and of different political organizations, within companies. 25 be no help. There is no common interest to help identify (or if there is, it is a common interest to desert), nor any particular reason to use social ties to inflict social sanctions.

This discussion produces several basic hypotheses relating to collective action:

H4. The more the soldiers in a unit share the common aim of the success of the unit, the more likely that a reciprocal norm of cooperation, to fight rather than desert, will emerge. This relationship occurs because of two mechanisms: • the presence of committed soldiers indicates that collective success is a possibility if all contribute; • competing aims, such as with intense factional rivalry, provoke mistrust of other factions.

H5. Such a norm should influence soldiers not to desert. It should do so by the following mechanisms: • increasing the sense that if they fight, others will as well, and so the armed group’s success is likelier; • increasing their perceived likelihood that others will fight and so reducing their estimates of the costs of fighting; • creating social rewards for fighting and social sanctions, including monitoring by one’s peers, for desertion.

H6. If common aims exist, social homogeneity among a group of soldiers increases the likelihood that norms of cooperation will emerge.

1.3.4. Summary and Micro-Level Implications

Thus the dissertation investigates two basic ways of limiting desertion: through control and through norms of cooperation. It situates those mechanisms in the context of the preferences of combatants. It argues that norms of cooperation depend on combatants sharing the armed group’s aims. Control does not have such strong limitations (although intense disagreement can shift it to persecution and make it less effective), and thus it is the more broadly and generally available solution.

Table 1.1. Effectiveness of approaches for reducing desertion

Typical preference for armed group’s aims Approach for reducing desertion Strong agreement Neutrality Strong disagreement

Control Effective Effective Partially effective

Norms of cooperation Effective Ineffective Ineffective

Table 1.1 summarizes my broad expectations. Comparing across rows, it suggests that control is the more widely applicable approach, since it is useful even when combatants are uncommitted. Comparing across columns, it suggests that when combatants share a preference for the armed group’s success, both control and norms of cooperation can be effective at reducing desertion. This is therefore generally the best situation for armed groups—not only because individual soldiers are, taken in isolation from each other, less likely to desert, but also, and mainly, because norms of cooperation are now available in addition to control.

1.3.5. Macro Implications of This Approach

The approach above is a theoretical argument mainly at the micro level, about specific practices such as recruitment policies, alliances among multiple factions, coercion, persecution, and cooperation within armed groups. However, it produces macro-level implications as well. It suggests that armed groups are more likely to suffer serious desertion problems if they have greater disagreements about aims and if their capacity to control is particularly weak; and it suggests that if disagreements are particularly severe, then even the capacity to control is less effective. This argument is important for understanding the international politics of armed groups. External military support can give the armed group’s leadership greater capabilities to

27 enact rewards for cohesion and punishments for defection—including policies of control. Such assistance helps even when combatants are generally uncommitted, as in conscript armies.

However, when the armed group goes beyond a general lack of commitment and is instead severely divided to begin with, such assistance is rendered less effective. It can, in fact, constitute a waste of money, as the capabilities for reward and punishment are put to private factional interests rather than to the interests of the armed group as a whole, provoking others to leave. Factionalism can potentially be reinforced by the possibility of external support as well, as different groups may favour or oppose support based on whom it might advantage in factional infighting.

It is possible to see the purchase of this argument in contemporary problems in Iraq,

Afghanistan and Libya. External support may be the only thing holding those governments’ various armed services together. According to some, withdrawing such support could result in collapse,63 as it did in Afghanistan after the withdrawal of the crumbling Soviet Union. 64 But

analysts have raised reasonable fears about the waste that such support could represent if the

governments in question are factionalized anyway. 65 My approach signals that this is a genuine dilemma without a particularly easy solution. In Chapter 8, in a broad macro-level comparison across multiple civil wars, I demonstrate that the twin factors of external support and factionalism have had real significance on defection rates in contemporary conflicts.

1.4. Alternative Approaches

How does this approach, then, relate to past work on desertion? What does it offer as an advance on the literature? In this section, I consider several basic alternative accounts.

63 Simon 2008; Peceny and Bosin 2011; Amnesty International 2012. 64 Rubin 1995; Sinno 2008; Giustozzi 2009. 65 Byman 2006. 28

1.4.1. The Cause

One possible approach to desertion is to suppose that individuals will line up where their

group interests tell them. A default position in thinking about civil wars is to consider them as

arising from grievances.66 Much criticized throughout the 1970s in theoretical reactions that

favoured a focus on the organization’s ability to mobilize resources 67 and the political

opportunity structure that organizations face,68 the grievance view enjoyed a resurgence in more

recent scholarship on ethnic conflict. Scholars have wondered what it is about ethnicity and

nationalism that makes people willing to die for their groups; for example, Benedict Anderson

evokes the Tomb of the Unknown Soldier at the outset of his seminal work on the development

of nationalism. 69 Barry Posen considers this issue with specific attention to desertion. He argues

that the development of nationalism in France and Germany in the late 18 th and 19 th century was

in part an active response by governments to ensure that there were soldiers who would carry on

fighting even as warfare evolved to favour large armies and dispersed skirmishers rather than

smaller numbers of men using lock-step drilled infantry tactics.70

More recently, scholars have seen ethnic conflicts in the contemporary world as dividing armies along group lines, because of the powerful pull of ethnic identity.71 Looking at armed groups that make ethnic claims, they then read in ethnic motivations to the individuals fighting in those groups. Behaviour and ethnic identity thus line up neatly, with minimal slippage between the two.72 Even some constructivist scholars, who reject treating identities as fixed, enduring

66 Gurr 1970. 67 Tilly 1978. 68 Skocpol 1979. 69 Anderson 1991, 9–10. See also Mosse 1990. 70 Posen 1993. 71 Horowitz 2000. 72 Kaufmann 1996a; Kaufmann 1996b. 29 traits exerting a powerful and continuous pull on individual behaviour, 73 are interested in understanding the construction of different alignments of group identities and in identifying the circumstances in which they become particularly powerful predictors of action across whole groups .74 Ethnic conflict is regarded as a potentially temporary instance of the power of group

identity, but given ethnic conflict, individual differences are not given much consideration; group identity is treated as a trump card. 75 Desertion and defection are limited, on this view, to circumstances in which soldiers find themselves fighting on the wrong side. 76 Desertion and defection from one’s “own” ethnic group are expected to be rare.

However, considerable evidence has accumulated suggesting that individuals vary substantially in their attachment to any given cause. According to Kellett’s study of the psychology of combat motivation, there is indeed some small proportion of individuals who actively seek to fight, another minority disaffected enough to want to leave, and a majority in the middle. 77 Jeremy Weinstein works with a distinction between “investors,” seeking a long-term

goal from the armed group’s success, and “consumers,” looking only for short-term benefits. 78

Lichbach argues that “in any dissident group, one will find many more constituents than sympathizers, sympathizers than members, members than activists, and activists than militants.” 79 Kalyvas, surveying the literature on this subject, finds considerable anecdotal and occasional systematic evidence to support Lichbach’s claim. 80 Given this variation in

preferences, the less attached a combatant to the cause, the more likely he is to desert. The costs

of fighting will more easily outweigh the perceived benefits for those who are uncommitted than

73 Against, for example, Smith 1986; Connor 1994. 74 e.g. Anderson 1991; for a review see Fearon and Laitin 2000. 75 This critique is developed in Kalyvas 2008. 76 For example, see Horowitz 2000, 443. 77 Kellett 1982, 291, 334. 78 Weinstein 2007. 79 Lichbach 1995, 17. 80 Kalyvas 2006, 102–103. 30 for true believers. 81 Statistical analyses of data from the U.S. Civil War, where “the cause” has

underwritten much scholarship about motivations, 82 confirm this variability across individuals.

Studies have consistently found that individuals who came from counties that voted strongly

Republican (in the North) and Democratic (in the South) tended to desert less often than others, and that conscripts deserted more often than volunteers.83

It is natural to expect soldiers to vary in their likelihood of desertion according to the

intensity of their commitment to the armed group’s common aims. However, the array of

powerful critiques to this perspective, outlined below, suggests that a focus on a soldier’s

commitment to common aims alone does not go far enough. I treat variation in the value placed

on the armed group’s success as a baseline for a combatant’s preference to fight or desert. The

effects of this commitment to the group’s common aims are mediated through the armed group’s

control and through the mix of committed and uncommitted combatants. I further draw on

Jeremy Weinstein’s approach to recruitment and signalling, in which more demanding and less

materially rewarding armed groups tend to recruit more committed followers. However, I add to

this the recognition that especially in civil wars, with multiple different agendas, intense

disagreement and factional competition may often be as important as individual self-interest in

explaining why groups of combatants often fall short of common aims.

1.4.2. Complex Motivations and Control

What happens if combatants do not particularly share the armed group’s goals—if they are indifferent, support the other side, or have other interests that are only in temporary and fragile alignment with those of the armed group as a whole? In that scenario, control is vitally

81 DeNardo 1985. 82 McPherson 1997. 83 Bearman 1991; Giuffre 1997; Costa and Kahn 2003. 31 important. Stathis Kalyvas builds an approach to understanding civil wars that rests on the assumption of multiple and conflicting preferences. As Kalyvas emphasizes, civil war situations are surrounded by a fog of uncertainty about what others intend to do. 84 Individuals who join armed groups may do so for many reasons. Scholars have variously argued that individuals join to follow a grand cause, 85 to respond to social pressures from those to whom they are connected, 86 to pursue fundamentally local agendas like vengeance against personal rivals, 87 to seek loot and other forms of financial gain, 88 or out of fear of the threat posed by a perceived ethnic other, 89 violence in the rearguard, 90 or the indiscriminate repression of the state. 91 Surveys

and interviews with ex-combatants find numerous different self-reported reasons for joining up

armed groups in the first place, ranging from community defense to self-interest to adventure,

and several different reasons for deserting as well. 92 Even in conflicts with supposedly clear lines

such as ethnic conflicts in Rwanda, the former Yugoslavia or Iraq, people have joined armed

groups out of thuggery or local interests with little reference to ethnicity. 93 People fight for the

“wrong” side of apparently ethnic civil wars, notably in Sri Lanka, 94 and internecine conflicts have been a common feature of even very intense ethnic wars such as in Bosnia, 95 Palestine and

Iraq. 96

Following Kalyvas, therefore, we can think of civil wars as complex concatenations of multiple different, private, often local aims, with warring groups knit together by temporary and

84 Kalyvas 2006. 85 McPherson 1997. 86 Gould 1995. 87 Kalyvas 2003; Kalyvas 2006. 88 Keen 1998; Collier 2000. 89 Petersen 2002. 90 Kalyvas and Kocher 2007b. 91 Goodwin 2001. 92 Peters and Richards 1998; Humphreys and Weinstein 2004; Arjona and Kalyvas 2006. 93 Mueller 2000; Kalyvas and Kocher 2007a. 94 Staniland 2012. 95 Christia 2008. 96 McLauchlin and Pearlman 2012. 32 often shifting alliances that rest on power and by the use of coercive force.97 Kalyvas takes the

importance of private and local incentives as an assumption, and then argues that the

fundamental criterion for preventing defection is control. An armed group can keep a civilian or

a subgroup from defecting to the extent that it, and not its adversaries, is able to coerce. Hence,

the pattern of geographic control by armed groups is the primary factor explaining defection and

hence violence rates in civil wars. 98 In other work in a similar vein, Scott Gates explains the decision by combatants to join and to switch sides according both to a reward offered by each side and to the distance of the agent from the headquarters of the armed group, an indicator of the group’s ability to punish defection. 99

There is certainly existing evidence for the importance of control. In an empirical confirmation of Gates’ approach, Johnston finds that armed groups in Liberia and Sierra Leone that were spread out over great distances suffered much larger defection problems than those that operated in confined areas, and that the problem of distance could be temporarily overcome through, for example, the use of helicopters and radios. 100 Taking the very long view, as

European states increased in strength, steadily taking power out of the hands of potential internal rivals and building up unrivalled capacities to extract capital and thus finance armies, they were steadily able to reduce desertion and defection rates. Lords who financed their own private armies to be called into service at particular points in time steadily lost this autonomy and ability to defect at will. 101 The issue of desertion then appears to be a bilateral relationship between the

armed group’s leadership and the combatant, with the former threatening the latter to prevent

desertion.

97 Kalyvas 2003; Kalyvas 2006. 98 Kalyvas 2006. 99 Gates 2002. 100 Johnston 2008. 101 Finer 1975; Tilly 1985; Tilly 1992; Mann 1986; Dandeker 1990; van Creveld 2004; Howard 2009. 33

Kalyvas draws a crucially important distinction between indiscriminate and selective

violence. He argues that indiscriminate violence undermines deterrence: at an extreme, it means

that the civilians who suffer indiscriminate violence have no reason not to collaborate with their

persecutors’ opponents, because they have a considerable chance of dying either way. 102 The

difficulty is analogous to Robert Jervis’ analysis of deterrence and the security dilemma: efforts

to deter can, by rendering the target less secure, provoke that target to strike first instead,

especially when the target cannot distinguish between defensive and offensive capabilities. 103 In

line with this analysis, Goodwin emphasizes the importance of indiscriminate violence in

galvanizing revolution. 104 Saideman and Zahar treat deterrence and assurance as posing a

dilemma: strong states can both deter internal threats and fail to reassure their citizens that they

are secure, provoking rebellion. 105 I build on this distinction here. I argue that when efforts at control are pre-emptive strikes against soldiers based on their characteristics rather than their actual efforts to desert, they are likely to provoke desertion rather than preventing it. Moreover, the fear of such persecution, which emerges with particularly severe political disagreements, can

undermine these efforts at control.

Kalyvas’ approach has much to recommend it. First, by departing from the straightforward “master cleavage” of, say, Sunnis vs. Shi’ites vs. Kurds in Iraq, or Communists vs. Capitalists in Vietnam, he usefully reminds us that there are very many other conflicts going on in civil wars. Actual groups of combatants are more or less variable alignments of groups with multiple interests, who may join together to defeat a common foe but share little in common beyond that potentially temporary interest. In addition, the logic of control is itself compelling. I

102 Kalyvas 2006, 144. 103 Jervis 1976, chap. 3; Jervis 1978. 104 Goodwin 2001. 105 Saideman and Zahar 2008. 34 argue that it provides a straightforward understanding of why individuals may sometimes fight for sides of a civil war to which they appear to have very little attachment in principle. In this dissertation, I affirm the importance of direct rewards and punishments, especially of control, in explaining how individuals who do not really share the armed group’s goals can be induced to keep serving. I also share Kalyvas’ concern to distinguish between indiscriminate and selective violence.

However, I argue that it is best not to treat multiple agendas as an invariant property of armed groups in civil wars. Some armed groups can be more factionalized than others. Variation in the factionalism of groups in civil wars can help us to understand the degree to which combatants share the same aims. As Kalyvas acknowledges, the welter of micro-level, local agendas that make up a civil war can undermine unity: “local cleavages may even subvert central ones, causing factional conflicts within supposedly unified political camps.” 106 The difficulty is

that Kalyvas does not then theorize the implications of variation in factionalism, instead treating

such divisions as a ubiquitous feature of civil wars that prompts the theoretical importance of

control alone. However, scholars have begun to investigate the impact of political factionalism

on defection and fragmentation in civil wars. The implication is that competitive aims can clearly

interfere with common aims. Staniland argues that when a single faction attempts to take over a

whole opposition movement, it can prompt “fratricidal flipping”: a willingness to fight for the

other side. The Sunni Awakening in Iraq in 2007 is the most prominent example of this

phenomenon, but Staniland also finds considerable evidence for it among the Tamils in Sri

Lanka. 107 Tensions between factions within the Palestinian and Kurdish nationalist movements

106 Kalyvas 2006, 375. 107 Staniland 2012. 35 have also prompted the fragmentation of those movements. 108 My approach treats political agreement and competition as one of two major sources of commonality of aims among combatants, along with recruitment policy. I argue that by analyzing variation in factionalism, we do a better job of identifying when control is necessary. However, I also argue that particularly intense factionalism undermines control, making it extremely difficult to trust that an effort towards control will in fact be used selectively. This is a prospect that Kalyvas leaves unexplored. Finally, studying variations in factionalism can also help us identify when norms of cooperation can obtain as well. Such norms further depart from Kalyvas’s scheme, since they do not treat the alignment of individuals in civil wars as just the consequence of a bilateral relationship between them and the armed group’s leadership, but instead as a product of multilateral relationships with other combatants as well. I discuss these points further below.

1.4.3. The Collective Action Problem and Selective Incentives

As I argue above, a prominent line of argument suggests that attachment to collective aims should make no difference at all to individual behaviour, because egoism gets in its way.

This approach has had a profound influence on scholarship on internal conflict in the years since the publication of Mancur Olson’s The Logic of Collective Action .109 Scholars began to see organization, rather than grievance, as the fundamental criterion for the occurrence of rebellion and revolution. 110 Indeed grievance seems largely irrelevant: “grievances are neither necessary

nor sufficient for rebellion…As the extent and intensity of grievances in a collectivity increase,

108 McLauchlin and Pearlman 2012. 109 Olson 1971. 110 Tullock 1971; Gamson 1975; Oberschall 1973; McCarthy and Zald 1977; Tilly 1978; Popkin 1979; Lichbach 1995. 36 collective dissent may or may not increase.” 111 The collective action problem has persisted as an analytical framework, brought from this earlier concern for social movements and revolution into the more contemporary frame of civil war. In an enormously influential paper, Collier and

Hoeffler argue that grievances are ubiquitous but civil wars are rare; hence the capacity of the group to overcome collective action problems has a great deal more leverage in explaining individuals’ decisions to join rebellion than does the cause in question. 112

For Olson, group interests are successfully defended only by organizations that emerge as

a “byproduct” of other factors, particularly their ability to provide selective incentives. In turn,

“the only organizations that have the ‘selective incentives’ available are those that (1) have the

authority and capacity to be coercive, or (2) have a source of positive inducements that they can

offer the individuals in a latent group.” 113 Thus selective incentives essentially consist of carrots

and sticks.114 The literature on organizing for internal conflict has paid particular attention to the former.115

Crucially, when attention is restricted to selective incentives, personal commitments to collective aims matter little. If contribution to a group interest has only a tiny impact on an individual’s decision to participate in collective action, then there will only be slight differences between those who care and those who do not, and even between those who care and those who strongly believe in the other side. In not contributing to a common aim, an individual is understating the value he places on that common aim, eroding any difference between him and someone who places no value on it at all. 116 Consider rewards to participation, the focus of much

111 Lichbach 1995, 288. 112 Collier and Hoeffler 2004. 113 Olson 1971, 133. 114 Gamson 1975. 115 Olson 1971; Tullock 1971; Popkin 1979; Collier 2000. 116 On the problem of preference mis-revelation in public goods, see Samuelson 1954, 388–389. 37 scholarly attention. For Collier, rebellion is a “quasi-criminal activity”; the decision to join an armed group is a matter of a monetary offer that is just better than whatever the prospective recruit can hope to earn outside of rebellion. 117 Collier and Hoeffler, then, juxtapose “grievance”

with “greed” in their famous article, and while they find some evidence for the importance of

grievances, the focus is ultimately on economic inducements to fight.118 The terminology indicates the profound shift that the focus on selective incentives entails: it is not just that such incentives make action towards a group goal possible, it is that they make the group goal irrelevant to the question of how rebellions are initiated and sustained. There is no serious

difference between someone who cares about the group goal and someone who does not; their

egoistic motivations dominate. There is then no difference between asking for a monetary reward

to offset the personal costs of fighting for a cause one happens to believe in, and fighting just for greed, for monetary gain, with no interest in the group goals one way or another.119

On this view, desertion is more likely to occur when carrots and sticks are lacking: if structures of coercion are insufficient, or if the balance of economic incentives lies against military service. Collier and Hoeffler thus argue that the importance of economic incentives explains why, in the Russian Civil War, desertion correlated with harvest season—with economic opportunities outside of the army. 120 And the coercive tools mentioned above are fully

consistent with this line of argument, providing incentives that are specifically tied to a

combatant remaining in an armed group. One important implication of this line of argument is

that when combatants share the armed group’s goals, control does not typically become

irrelevant. The collective action problem reminds us that even when individuals seek a common

117 Collier 2000. 118 Collier and Hoeffler 2004. 119 Perhaps realizing that they may have overreached, Collier and Hoeffler more recently express their analysis in terms of the feasibility of rebellion. Collier, Hoeffler, and Rohner 2009. 120 Collier and Hoeffler 2004, 569. 38 aim, they can still free-ride on each other’s contributions, and top-down rewards and punishments can induce them to cooperate.

However, it seems too far a departure from reality to assume that individuals’ preferences

about group goals are irrelevant. James DeNardo pithily summarizes the intuition behind this

critique: the theory “implies that socialists will gladly participate in fascist demonstrations, and

vice versa, if the organizers simply provide coffee and doughnuts to the marchers. After all, why

not enjoy the selective incentives when it is obvious that one extra person will not affect the

outcome of the demonstration?” 121 Agreeing with DeNardo, I take it that individuals will vary in the utility they derive from the armed group’s success, and therefore in a baseline preference to desert or to fight. Further, as I argue above, rewards and punishments do not exhaust the set of solutions to the collective action problem; norms of cooperation are an additional solution provided that combatants share a common aim . We therefore cannot simply dispense with group aims, as the selective-incentives literature seems to urge.

I therefore take a contextual approach to the collective action problem, and emphasize norms of cooperation as an important solution alongside rewards and punishments. But does the collective action problem even apply to many wars? Kalyvas and Kocher argue that in irregular war, the costs of non-participation are frequently higher than the costs of participation, because a combatant can frequently find safety from victimization in an armed group. 122 This is an

important challenge. If it is correct, then the collective action problem may just be inapplicable

to irregular wars, because participation would not imply paying particularly high costs.

Kalyvas and Kocher’s claim rests on data from the United States’ Phoenix Program in

Vietnam, which tracked its killing and capturing of suspected Vietcong. It found “unconfirmed”

121 DeNardo 1985, 56. 122 Kalyvas and Kocher 2007b. 39

Vietcong were captured and killed at a much higher rate (52.53%) than “confirmed” Vietcong

(5.88%). To be classified by the United States as “confirmed” Vietcong an individual had to have three denunciations against them, or one denunciation from an “irrefutable” source. 123

Kalyvas and Kocher argue that this is because if an individual was “confirmed” he or she was in fact more likely to actually be Vietcong.

This is a possible interpretation, but the argument supporting it is problematic. First, we do not have any information about the proportion of actual Vietcong or non-Vietcong who were killed. Kalyvas and Kocher do attempt to estimate these proportions, but their attempt to do so appears to turn on assuming the conclusion. To obtain estimates, they assume that the ratio of the rate of innocence among unconfirmed to the rate of innocence among the confirmed is the same as the ratio of the rate of victimization among the unconfirmed to victimization among the confirmed. Given that they know, beforehand, that the unconfirmed were victimized at a much higher rate than the confirmed, this assumption amounts to building in a correlation between innocence and victimization into their estimates. They justify the link between being unconfirmed and being innocent as follows: “this assumption tells us that it is the composition of the two categories, confirmed and unconfirmed, and in particular their respective ratios of

Vietcong and innocents that account for the observed difference in rates of victimization.” 124 In other words, they are assuming that the reason the unconfirmed were killed and captured more often was that they were more likely not to be Vietcong. But this is exactly what they must prove, since they are trying to establish the vulnerability of the innocent as the explanation for the victimization patterns of the confirmed and unconfirmed. It is not surprising that they find a

123 Ibid., 194. 124 Ibid., 197. 40 high rate of victimization among the “innocent” when they construct this category of “innocent” to correspond to a known, high rate of victimization.

In the absence of any real information about actual membership in the Vietcong, there is a simple plausible alternative. Confirmed Vietcong may have been killed less often because they knew they were being denounced, since it took three denunciations (or one denunciation from an unimpeachable source). Someone who knew he was exposed might have been expected to take greater care to ensure that he was not captured—to flee from areas of contention, for example.

Notwithstanding my objections to Kalyvas and Kocher, is still quite possible that civilians will be victimized by war at a higher rate than combatants, and that the approach here does not easily apply beyond the regular-war context of the Spanish Civil War. This is a matter for further investigation. However, combatants need not compare the costs of fighting only to the costs of remaining in civilian life in a particular location. A deserter need not return to a hometown under threat: he could attempt to flee the area entirely, or, despite his convictions, defect to a side offering substantially greater protection. In an armed group, it will often be the case that the combatant will be required by his superiors to undertake serious danger. For a collective action problem to attend on participation and desertion in civil war, it is necessary (in addition to common aims among a group) that a combatant have one option that is less costly

than fighting. I therefore maintain the assumption held by the bulk of the literature that the risks

from participation are higher than the risks from non-participation, and that this is a central

problem for the armed group to overcome if it is to maintain participation and prevent desertion.

1.4.4. Social Homogeneity and the Collective Action Problem

Top-down rewards and punishments are not the only way to solve the collective action problem. As discussed above, theorists have also explored the ways in which norms of 41 cooperation or reciprocity can do so. In the context of internal war, scholars such as Michael

Taylor and Mark Lichbach have considered reciprocity an important resource in organizing for revolutionary collective action. 125 I argue above that reciprocity also has strong appeal in studying desertion, because combatants are more likely to fight if they believe that others will as well.

Interpersonal trust, as a basis for norms of reciprocity, has driven the most thorough recent research on desertion in civil wars. But the focus has so far been on social homogeneity alone as a generator of trust. Scholars have not explored how social homogeneity’s desertion- reducing effects could be contingent on having a common aim. In an important article and book, 126 Costa and Kahn find that punishment cannot be the explanation for the relatively low

desertion rates in the Union Army during the U.S. Civil War. Rather, their explanation is that

soldiers fight because they have commitments to other soldiers with whom they can identify.

Costa and Kahn find that more socially homogeneous units—companies where soldiers share

hometowns and occupations—tended to have lower desertion rates than more heterogeneous

units. In such homogenous units, they argue, there was a stronger basis for norms of cooperation.

Soldiers who deserted suffered social costs, such as shame and opprobrium, from people whose

opinion they valued. Indeed, social homogeneity often implied that soldiers joined up together in

groups from the same community, and they would suffer particularly strong sanctions from their

peers if they deserted. Elsewhere, Costa and Kahn find that deserters endured shame and

ostracism in their communities, with Union Army deserters often having to leave home later.127

In contrast, familiarity breeds the possibility of trust and the possibility of social sanctions for violating group norms of collective action.

125 Taylor 1988; Lichbach 1995. 126 Costa and Kahn 2003; Costa and Kahn 2008. 127 Costa and Kahn 2004. 42

In addition to collective action theory, Costa and Kahn also appeal to altruism among soldiers. They argue that affective ties reduce desertion rates, and that those affective ties are facilitated by homogeneity. In brief, “Union army soldiers...were loyal to men who looked like themselves.”128 This argument alludes to a literature that indicates the importance of social

cohesion among soldiers—that is, interpersonal affective ties among them—for military

performance in general. Classic studies of German and American soldiers in the Second World

War suggested that soldiers fought for their war buddies. When such affective ties were weak—

for example, when their platoon or squad was thrown hastily together in an ad-hoc fashion—

soldiers tended to desert or surrender much more easily. 129 This link became a trope, and in fact a

stereotype. 130 Thus Costa and Kahn argue that social homogeneity helps reduced desertion by facilitating social cohesion among troops.

However, affective ties alone do not appear to be sufficient to explain low desertion rates.

The link between affective ties among soldiers and low desertion is empirically dubious. If socially cohesive units had lower desertion rates, we should expect them to perform more effectively on the battlefield. But subsequent studies of social cohesion in armed forces have often found little discernible independent effect of affective ties on military performance, net of reverse causality in which success breeds interpersonal affect. 131 According to such critiques,

soldiers’ resentment of others on the basis of social diversity has no apparent empirical benefit

for combat effectiveness. 132

Indeed, social cohesion can reduce military effectiveness. It can do so via disobedience and even desertion. Here, scholars of cohesion in combat argue that social cohesion is not nearly

128 Costa and Kahn 2008, 220. 129 Marshall 1947; Shils and Janowitz 1948; Stouffer et al. 1949. 130 Mullen and Copper 1994; MacCoun, Kier, and Belkin 2006. 131 Madej 1978. 132 MacCoun 1993; Kier 1998; MacCoun, Kier, and Belkin 2006. 43 as important as task cohesion , the commitment of the members of a group to the group’s task.

Scholars hint also that the impact of social homogeneity may in fact be contingent on task cohesion, 133 though there has yet to be a systematic study of this contingency. Madej suggests that this phenomenon could account for “mass-surrenders or group disobedience of orders.” 134

For example, “fragging” among U.S. troops in Vietnam—attacking officers with fragmentation grenades—was a group activity, routinely discussed with others beforehand and bolstered by their encouragement. 135 Shils and Janowitz, in fact, found instances of mass desertion among

cohesive Wehrmacht units at the end of the Second World War. 136 Roger Gould’s study of mobilization in the Paris Commune found that community ties among National Guardsmen could reinforce them in a decision to fight for either the government or for the insurrection. 137 Social

cohesion can cut multiple ways.

Consonant with the military sociology literature’s critique of social cohesion, and the possibility that it is contingent on task cohesion, I investigate the possibility that the effect of social homogeneity on desertion rates is contingent upon sharing a common goal. Kier first raises this suggestion, arguing that the contingency of social cohesion can help explain an important finding in the small literature on desertion in civil wars: 138 Bearman’s evidence that

more homogeneous North Carolina units—contra Costa and Kahn—had higher desertion rates

than more heterogeneous units by the end of the U.S. Civil War, even though more homogenous

North Carolina units had lower desertion rates at the beginning of the war.139 Costa and Kahn

133 Kier 1998, 16. 134 Madej 1978, 243. 135 Bond 1976. 136 Shils and Janowitz 1948, 286–87. 137 Gould 1995, 179–181. 138 Kier 1998, 16. 139 Bearman 1991; Costa and Kahn 2008. This is a more compelling basis for Bearman’s finding than Bearman himself provides: he argues that homogeneous units failed adequately to socialize soldiers to loyalty to the Confederacy as a nation. However, this finding does not account well enough for why homogeneous units had lower 44 acknowledge Bearman’s finding but argue that it may have had something to do with different preferences between highland and lowland soldiers. 140 This suggests immediately a prospect that

Costa and Kahn leave unexplored: that the impact of social homogeneity may be contingent on a

shared preference. Indeed, Costa and Kahn study an army with a relatively high likelihood of

shared preferences. Only 2 and 6% of Union soldiers were draftees and substitutes. 141 Matching this, 91% of Costa and Kahn’s sample were volunteers. 142 In contrast, there are strong reasons to believe that the overall Confederate commitment declined substantially over time, accounting for the over-time shift Bearman finds. Towards the end of the Civil War the Confederacy was clearly losing. Conscription began in 1862, and some 21% of Confederate soldiers were conscripts. 143 Costa and Kahn thus restrict the domain of their analysis to an environment in

which a common motivation is quite likely. Their finding should not be taken as generally

representative of the impact of social homogeneity on desertion rates. That relationship, instead,

may be contingent on commitment to the task at hand.

On my line of argument, then, social homogeneity only really works to lower desertion

rates when soldiers value their armed group’s success. When they do not share this common aim,

social homogeneity should be no help in lowering desertion rates, and may in fact increase them.

Costa and Kahn quote a soldier thus: “I myself am as big a coward as eny [sic] could be...but me

the ball [bullet] before the coward when all my friends and comrades are going forward.” 144 But

what if one’s friends and comrades do not go forward? Thus the link between social

homogeneity and low desertion rates is an important finding, and I expect it to hold up, but only

desertion rates at the start of the war. A change over time in motivation, with more conscripts and with enthusiasm steadily losing out to resignation to defeat—is plausible (Weitz 2005) and accounts for the data much more clearly. 140 Costa and Kahn 2003, 540, fn. 21. See also Giuffre 1997. 141 Chambers and Anderson 1999, 181. 142 Costa and Kahn 2003, 529. 143 Chambers and Anderson 1999, 181. 144 Costa and Kahn 2008, 215. 45 in some circumstances. This approach helps us make sense of the apparent discrepancy between the two most thorough systematic studies of individual desertion decisions we have, Bearman’s and Costa and Kahn’s.

Not all solutions to the collective action problem are created equal. Control, as I have argued above, does not depend on individual combatants’ preferences to nearly the same extent that norms of cooperation do. Unlike norms of cooperation, it can also apply to situations well beyond the classic collective action problem: specifically, it can apply even where soldiers place very little value on the common aim. In essence, then, when group aims are shared, the set of solutions helping to prevent desertion expands.

The approach outlined here is quite similar to Jeremy Weinstein’s. 145 Weinstein examines the degree to which rebel groups are able to recruit committed “investors” versus short-run egoist

“consumers” on the basis of an initial distribution of resources. Material resources (such as access to loot or foreign finance) favour the recruitment of consumers, while social resources

(such as community networks) favour investors. The proportion of each type of fighter then affects the armed group’s organizational choices: for example, their ability to rely on norms of cooperation, the responsiveness of their rule over civilians, and their use of violence.

Do I provide anything new, beyond Weinstein? There is an obvious affinity between

Weinstein’s work and my own. Like him, I take the existence of a shared, common aim as a vital source of variation; like him, I use the logic of costly signalling as a method of identifying committed combatants; and like him I suggest that it affects the applicability of norms of cooperation in keeping soldiers fighting. I part ways from Weinstein on several counts. First, I treat the commonality of aims as a function not only of recruitment policies but also of factionalism within the armed group. Second, Weinstein argues that rebel groups that do not

145 Weinstein 2005; Weinstein 2007; Humphreys and Weinstein 2006; Humphreys and Weinstein 2008. 46 possess norms of cooperation or common aims also tend to lack clear rules for combatants’ behaviour. I allow these two approaches to vary independently of each other. My account also does not rest on the idea that economic endowments result in less committed combatants despite the best efforts of armed groups to recruit more committed followers. My approach is more catholic about the origins of different approaches to recruitment.

More importantly, I attempt to generalize beyond the domain of Weinstein’s analysis.

Weinstein’s argument rests on there being a point in time in which a rebel group emerges; at that point in time the group’s distribution of resources is assessed. In turn this analysis is restricted, as

Weinstein explains, to circumstances in which the state is not so strong that it can quash any threat, nor so weak that there are essentially no barriers to organizing a rebel group. 146 In

focusing just on the connection between distribution of motives and organizational mechanisms

for preventing desertion, and in resting on a more general understanding of recruitment and

factionalism that extends beyond material resource endowments, I allow the contingency of

norms of cooperation to be applied beyond Weinstein’s analytical range. This is illustrated by

my Spanish case study: I consider two sides, Republican and Nationalist, that did not begin the

war as new entities but built on previous organizations—branches of the military and political

groupings that existed before the war—as well as pursuing new recruitment policies.

1.4.5. Alternative approaches: summary

In the approach to desertion developed here, control helps to reduce desertion even when many combatants are uncommitted, whereas norms of cooperation are contingent upon soldiers sharing the armed group’s aims. This synthesizes several parts of the literature. It takes as its starting point the variability of individuals’ motivations, a point frequently made against the

146 Weinstein 2007, 14–15. 47 grievance school of understanding civil war. However, it takes seriously that perspective’s claim that common aims matter. It proposes, however, that the impact of common aims is mediated through control and the influence of other combatants. Unlike Kalyvas’ view of civil wars, which sees them as complex welters of private interests to which only coercion offers a clear solution, it regards as meaningful the variation in the degree to which an armed group’s combatants share the group’s goals, and it gives more credence to the collective action problem. Unlike approaches to the collective action problem that focus only on selective incentives, in which individual attachments to public goods are irrelevant in the face of egoistic motives, it does not assume that the collective action problem essentially makes common aims irrelevant. Hence, against both of these views, it argues that while top-down rewards and punishments—particularly control—may be important, they are not the only tool for limiting desertion; norms of cooperation can work as well. But I contextualize norms of cooperation, arguing that they are contingent upon the prior existence of common aims—a point suggested in the military sociology literature, but under- developed even there, and absent in recent approaches to desertion in civil war.

Ultimately, then, my approach bridges the gap between different images of civil war: as competition among multiple local aims, as clashing collective projects, and as contests of groups to be the more effective organizers. I argue that treating the commonality of aims as a variable rather than fixing it by assumption, and interpreting the organizational tasks a group faces as contingent upon the commonality of aims, improves our understanding of civil wars by grappling with their diversity and by placing organizational problems in context.

1.5. What Is to Come

This dissertation pursues a multi-method examination of the hypotheses explored here.

The bulk of the dissertation is an examination, both qualitative and quantitative, of the Spanish

Civil War. I use the qualitative material to furnish evidence about the mechanisms of desertion and the plausibility of my approach. Do cheap signals really do perform worse than costly signals at indicating individuals’ preferences? Do combatants actually respond to credible threats of punishment by deserting less often? Are armed groups with committed followers really able to foster norms of cooperation in a way that armed groups with uncommitted followers are not?

And do severe disagreements and factionalism undermine control by making combatants fear that force will be used arbitrarily?

In examining the plausibility of my approach, the qualitative analysis takes advantage of several sources of variation. On both the Republican and Nationalist sides in the first year of war, different armed groups emerged. These different groups, including civilian militias, attempted different formulas to attract and keep recruits. In particular, they employed costly signals and control, and suffered from factionalism, to differing degrees. I highlight differences and similarities across these armed groups. I also conduct a comparison across the two sides, because the Nationalist side had much stronger central control and much less factionalism than its

Republican adversary. Finally, I compare over time within the Republican side, because that side implemented much stronger central control and changed its recruitment policy as time went on, changing from voluntarism to conscription. The comparative analysis therefore takes advantage of several sources of variation. The Spanish Civil War is, hence, an empirically rich setting for establishing the plausibility of my theoretical approach.

I conduct statistical tests of my major hypotheses in a micro-level setting within the

Spanish Civil War. Specifically, I study novel data on individual soldiers’ desertion decisions, data drawn from Santander province. I examine whether greater control helped to make soldiers’ desertion less likely, in particular examining control as the ability to extend into the rearguard and monitor soldiers’ hometowns. I find that mountainous terrain, an index of the ability of a deserter and his family to hide in the rearguard, was indeed strongly correlated with desertion. I further investigate whether norms of cooperation really are contingent upon the existence of common aims among soldiers. I look at the composition of soldiers’ companies as an influence on their decisions about whether to desert or not. I find that soldiers in companies with a higher proportion of volunteers (vs. conscripts) deserted less often. I find also that this was in part because social homogeneity—commonalities among soldiers in terms of location of origin, age, and occupational category—was more effective in mostly-volunteer units. I also find that soldiers in companies that were polarized between the two great union confederations in the

Spanish Republic also deserted more often.

The combination of quantitative and qualitative micro-level data from a single setting allows for a rich and vivid illustration of causal mechanisms and for statistical tests of their

relevance in predicting individuals’ behaviour. With the qualitative material, I follow the recent

trend in political science toward multi-method research. Responding to the potential for case

study research to shed light on the causal mechanisms underlying larger correlations, 147 scholars

have suggested that this role is ultimately an area where micro-level qualitative research has a

distinct comparative advantage. 148 The qualitative material here is not sufficiently structured to

constitute a test of my hypotheses, however. Using comparative case studies for hypothesis

147 Elster 1989; Hedström and Swedberg 1998; Tilly 2001; McAdam, Tarrow, and Tilly 2001. 148 Laitin 2003; Fearon and Laitin 2008. 50 testing is a well-established practice, despite the skepticism of some,149 but all are in agreement

that for case study research to provide hypothesis testing requires structure and careful attention

to case selection. 150 However, the available data on Republican militias are often too scattered to conduct a structured comparison across clearly identified and specific militia groups; and in any case, any comparisons within Spain would be subject to Galton’s problem, the interdependence of observations. 151 The purpose of the qualitative material is somewhat more modest, then, than

hypothesis testing: it is to illustrate the plausibility of the mechanisms underlying my approach,

indicating their purchase in micro-level behaviour. Hypothesis testing, in contrast, enters with

the quantitative work.

Chapters Two through Seven present my analysis of the Spanish Civil War. In Chapter

Two I give essential background to the case, important for contextualizing the remarks that

follow. In Chapter Three I discuss how different groups in the war varied in terms of policies of

recruitment, including their use of cheap signals and self-selection, and the degree of

factionalism that each suffered from. The chapter therefore looks at variation in the distribution

of commitment to common aims among combatants.

I follow this discussion of motivations with my analysis of their effects in terms of

control and cooperation. Chapter Four turns to control. It indicates the clear variation in control

policies instituted by different Republican militia, the important shift towards greater

centralization of control over time on the Republican side, and the general advantage enjoyed by

the Nationalist side in terms of control. In Chapter Five, I move on from the discussion of

individual combatants considered in isolation to an examination of how soldiers affected each

149 Lieberson 1992. 150 Lijphart 1971; Eckstein 1975; Skocpol and Somers 1980; Ragin 1987; King, Keohane, and Verba 1994; Ragin 2000; George and Bennett 2005; Mahoney 2007; Ragin 2008. 151 Ross and Homer 1976. 51 other. It shows the pernicious influence of political disagreements on desertion rates, and the potential for norms of cooperation to emerge in relatively united groups of volunteers. It indicates that these norms of cooperation were much less likely among conscripts on both sides and when military units were split between rival political factions.

Chapters Six and Seven then conduct more hypothesis-testing. The qualitative analysis of control in Chapter Four sets up a quantitative hypothesis test in Chapter Six, in which I find that terrain, an indicator of the ability of the group to control soldiers’ hometowns, was an important predictor of soldiers’ decisions to desert or not. Chapter Seven then confirms my hypotheses about cooperation in statistical tests. It establishes that unit composition, especially the proportion of conscripts and volunteers, was a significant predictor of individual decisions to desert, in part because in volunteer units social homogeneity could have a greater effect.

In Chapter Eight, I turn to the broad, macro implications of the approach considered here.

The final empirical study that this dissertation pursues is an analysis of its broad, international implications across twenty-eight civil wars in the 1990s. I examine the incidence of defection from government armies in those conflicts, with reference to a combination of three primary predictors that illustrate the capabilities available to provide rewards and punishments, lack of commitment, and serious disagreement: the receipt of external military support by a government, its use of conscription rather than voluntarism, and whether its forces experienced a military coup attempt within the previous five years. External support helped to reduce defection rates, regardless of whether a government used conscription. However, when a recent coup had taken place, external support had little ability to reduce defection rates. This analysis thus establishes that the interaction between the capability to enact rewards and punishments and the distribution of motivations helps explain the reliability of armed group’s forces across a much broader array

52 of cases. It complements the historical research of the Spanish Civil War with a much more contemporary analysis, bridging a potential gap between “old” and “new” civil wars. 152 And it demonstrates how the logic of desertion affects the decisions that policy-makers in the contemporary world must make, as they decide how best to go about facilitating stable security services in conflict settings.

Finally, in Chapter Nine, I present this study’s conclusions and future directions for research. I argue that the evidence presented here compels us to consider how armed groups and civil war environments vary from one another in terms of the degree of commitment of the fighters to a common aim. We have to confront the reality of multiple different kinds of civil wars, relaxing strong assumptions one way or another and instead treating the distribution of preferences as an empirical matter. Beyond this, the perspective offered here indicates some tradeoffs, especially among size, reliability, and military tactics. If an armed group wants to increase in size beyond those who generally prefer to see it succeed, it must rely on control and cannot necessarily trust its soldiers with the freedom to operate as autonomous agents. The general applicability of control to preventing desertion suggests renewed attention to a particular kind of civil war violence: violence against combatants by their own armed groups. And, finally,

I consider the international implications in terms of a tradeoff between the short-run benefits of external support for military cohesion against the long-run concerns for the institutional decisions needed to create military services that are self-sustaining.

152 Kalyvas 2001; Kaldor 2006. 53

Chapter 2 Background to the Spanish Civil War

2.1. Introduction

The Spanish Civil War had at its heart the issue of military loyalty. The war began with an attempted military coup by right-wing officers against the Popular Front government of the

Second Spanish Republic, on 18 July 1936. Beginning in Spain’s Moroccan colonies and proceeding throughout the country, rebel officers attempted to raise their military garrisons, with mixed success. Small extant civilian militias on both the rebel (or “Nationalist”) side and on the loyalist (or “Republican”) joined with soldiers and officers, and new militias were organized by each. Over time—relatively easily on the Nationalist side, much more arduously on the

Republican—each side converted its armed force into a unified conscript army. The character and organization of the opposing forces had an important impact on patterns of desertion on each side and hence on the lives of soldiers and civilians in Spain and ultimately on the outcome of the war. Military organization was certainly not, however, the only factor leading to a Nationalist victory. Notably, Italy and Germany supplied the Nationalists with personnel and military supplies that far outstripped the Soviet contribution.1 But of the other factors influencing the outcome, military organization was highly important, and desertion a central part of that story. In this chapter and the next three, I outline the general patterns of desertion prevalent in the war.

The present chapter sets the stage, drawing out the background to the war and highlighting the variation that I examine. In the following three chapters, I then turn to an analysis of how the

1 Alpert 2004. For a good recent overview of causes of Republican defeat, see Casanova 2010, 335–339. 54 existence of common aims conditioned the operation of control and norms of cooperation among the various armed groups of the Spanish Civil War.

The principal focus of these chapters is the Republican side; the Nationalist side serves as a complement for comparative perspective. I choose this approach because there is a somewhat richer story to tell on the Republican side. First, the Republican side was rather more heterogeneous than the Nationalist side in the forms of military organization that emerged at the beginning of the war. This creates a source of intra-Republican variation that I can exploit for

my analysis. Differences certainly existed on the Nationalist side—particularly between the

regular military and the elite Army of Africa—and I will discuss them as well, but they are rather

more straightforward. Second, the evolution of military organization provoked stronger reactions

and greater turbulence on the Republican side than on the Nationalist side. This is a phenomenon

I attempt to explain, but also a reason to spend rather more time detailing Republican

developments.

In focusing on the requirements of military success in explaining important features of

the war policy of the Republic, I follow two recent trends in the historiography of the wartime

Spanish Second Republic. Contemporary analyses of the Republic are now more thoroughly

drawing out the pressures of fighting the war, turning away from an earlier literature’s focus on

the Republic’s politics as a function of Moscow’s manipulation. 2 Graham argues that “the

overarching influence that shaped the evolution of the Republic between 1936 and 1939 was the war itself”: not all-powerful machinations of Soviet foreign policy but the requirements of

fighting and winning a brutal civil war. 3 Over these four chapters I attempt to explicate an important aspect of these requirements: maintaining a large number of men in the field over time.

2 See especially Bolloten 1991; Radosh, Habeck, and Sevostianov 2001. 3 Graham 2002, xi. 55

The Republican government’s increasingly tight control over its army helped that army to overcome its early ineffectiveness, including the problem of desertion. At the same time, factionalism undermined cooperation and undermined control by making it, in perception and often in reality, a vehicle for factional violence.

The second theme that I follow upon is the attempt by Michael Seidman 4 to emphasize

the role of egoism and opportunism, and the differential ability of the Republican and Nationalist

side to manage individual self-interest. Seidman serves as a useful corrective to the overly heroic

accounts of a previous generation of scholars, much as Ella Lonn’s work 5 did for the American

Civil War. Seidman’s Republic of Egos served as a precursor to Pedro Corral’s recent

Desertores ,6 the first full-length treatment of desertion in the Spanish Civil War. These books contain a wealth of valuable information in unsystematic accounts rooted in vignettes. Paul

Preston’s critique of Seidman is apposite: 7 Seidman jumps around in evidence from one part of

the war to another. While Corral picks out themes and discusses changes over time in a clearer

fashion, his account still lacks unity. Neither scholar develops a systematic account of the origins

of desertion. I offer a somewhat more analytical take, and develop specific hypotheses for testing

against quantitative data in Chapters Six and Seven.

The remainder of this chapter proceeds in several stages. First, I discuss the essential

political background to the war and identify the central political players involved. I then move to

an outline of the major developments in the military organization of the war itself. I first outline

the period in which the Republic’s militia forces were dominant (roughly from the start of the

4 Seidman 2002; Seidman 2011. 5 Lonn 1928. There is an interesting coincidence: both Lonn’s book and Seidman’s Republic of Egos (2002) appeared 63 years after the close of the wars they discuss, wars long mythologized by heroism on both sides. On the mythologizing of old wars, see Kalyvas 2001. 6 Corral 2007. 7 Preston 2007, 340. 56 war to the late autumn of 1936). I then turn to the process of replacing those militias with a regular army from those militias and the operation of that eventual force. Finally, I discuss simultaneous developments on the Nationalist side. After each section, I discuss how that part of the war provides variation that I use in the following chapters to develop my arguments. I conclude with a summary and preview of the chapters to come.

2.2. The Roots of War

The coup attempt of 18 July 1936 had its roots in the conflicts that had dominated the short life of the Spanish Second Republic.8 Founded in April 1931 out of the ashes of Miguel

Primo de Rivera’s dictatorship (1923-1930) and the abdication of King Alfonso following

Republican-dominated local elections, the Second Republic was beset by intense political rivalries, and by forces that never fully accepted its legitimacy. The Republic was led from 1931 to 1933 by moderate left-wing Republicans. The Partido Socialista Obrero Española (PSOE), the Socialist party, actually had the largest bloc of seats in the first Republican general elections in June 1931, with 115 out of 470, but the more centrist Republicans were better positioned to pull together a coalition. This government, supported as well by the moderate wing of the

Socialist and PSOE-affiliated Unión General de Trabajadores (UGT) union confederation, embarked on a reformist program, pushing for land reform, labour rights, secularization in public life and especially in education, civil status and the rights of women, and reductions in the privileges of the clergy and in the size and social power of the officer corps.

Various conservative groups feared the newfound potency of the left under republican institutions that permitted the latter much greater representation than had the monarchy or the dictatorship. Wealthy landowners and industrialists were threatened by the left’s projects for land

8 This summary draws upon Payne 1967; Salas Larrazábal 1973; Thomas 1994; Graham 2002; Preston 2007. 57 and labour reform, and convinced the large number of smallholding farmers across the North that their interests were equally menaced. The Republic also endured the dogged opposition of the

Catholic Church. Indeed much of the highly religious public at large was wary of the Republic’s efforts to enact secular reforms in education and civil status.

The armed forces had been wholly humiliated in 1898 by the United States and the

Cuban rebellion and bogged down in counterinsurgency warfare in Morocco through the 1910s

and 1920s. There was broad resentment at what many regarded as the Republic’s antimilitarism

and willingness to permit disorderly labour mobilization and to negotiate increased autonomy for

Catalonia and the Basque Country. Beyond this the armed forces were divided. So-called

Africanista officers, particularly those who volunteered for service in Africa, sought, via a

decisive victory in Morocco, the restoration of earlier Spanish glories and power—with

themselves at the forefront. The group ultimately included many key leaders of the eventual

rebel side in the Civil War, including General Francisco Franco himself. Other officers, called

peninsulares , regarded serving as a means of social advancement rather than for personal glories.

Content to remain in service in garrisons on the peninsula, they sought to preserve promotion by

strict seniority rather than by war record. A smaller group opposed both colonial adventurism

and the inefficiency of the Spanish military, with an officer corps bloated by sinecures. The

Minister of Defense, Manuel Azaña, embarked on a project of military reform to reduce the size

of the officer corps and the influence of patronage. These efforts were direct threats to

peninsulares and viewed by the Africanista camp as threats to the armed forces’ honour.

What was to be done about the left-wing program? Some elements of the right were

willing to work within the Republic, regarding forms of government as a mere “accident” (hence

their label “accidentalists”) when what was really at issue was the policies that they pursued.

They organized political parties, most notably Acción Popular , which drew large support from smallholding farmers and was the core of the later Confederación Española de Derechas

Autónomas (CEDA) party. The “accidentalists” were certainly willing to delegitimize the

Republic in order to gain popular support, however. Others, “catastrophists,” were wholly irreconcilable to the Republic and believed that a catastrophic action against it was necessary.

Officers plotted coups and sometimes—notably in the 1932 Sanjurjada coup attempt led by Gen.

José Sanjurjo—tried to carry them off. Alfonsist monarchists hoped for the restoration of King

Alfonso’s monarchy while Carlists—supporters of a dynasty that had been ousted in the 19 th century—sought not only the restoration of their pretender but for an intensely Catholic, conservative revolution. To this end the Carlists maintained a militia known as the Requeté or

Red Berets. Finally, the fledgling Falange party, avowedly fascist, combined right-wing

conservatism, a fetish for a decisive military dictator, and an attempt to appeal to labour

grievances and to shift worker support to the right.

At first, these forces bided their time, recognizing that the failures of the monarchy and of

the Primo de Rivera dictatorship gave the Republic considerable popular appeal. In the interim,

the right was united in seeking to block the left’s attempts at reform whenever possible, through

parliamentary maneuvering, removing wealth from the country to prevent its redistribution, and

fomenting popular fears of social change from the pulpit. They also attempted to actively impede

the implementation of reform, even when laws were passed, through the resistance of entrenched

local power brokers such as landowners and detachments of the Civil Guard, the feared rural

paramilitary police.

There were strong political forces on the Left that were intensely suspicious of the

Republic as well, considering it a bourgeois democracy. In particular, the state was opposed by

59 the anarcho-syndicalist union confederation Confederación Nacional del Trabajo (CNT), which

had a strong presence among workers in Barcelona, Zaragoza, and Sevilla and among the

landless rural masses of Andalucía in southern Spain. The anarchist movement, especially as it

was increasingly dominated by the shadowy militant group Federación Anarquista Ibérica

(FAI), explicitly called for the downfall of the Spanish Republic and its replacement with

confederated workers’ communes, and organized continual strikes. The CNT had a bitter rivalry

with the UGT trade unions, competing to organize and dominate the Spanish labour movement.

In contrast, though there was certainly a revolutionary wing of the PSOE and the UGT, for the

time being the moderate Socialists seeking to work for labour interests within the Republic were

in the ascendancy. The small but tightly organized Partido Comunista Española (PCE) followed the general line taken by Stalin and the USSR, in principle revolutionary but explicitly willing to work with other, more moderate left-wing forces in a Popular Front to contain the advance of fascism in Europe. And, following the divisions within European communism, the PCE was most bitterly opposed to the Partido Obrero de Unificación Marxista (POUM), a small, breakaway party promoting the Trotskyist line of global revolution, though not affiliated with

Trotsky himself. The success of the right at blocking various attempts at reform in the Republic’s first two years left the left wing mistrustful of centrist politicians and dissatisfied by what it saw as bourgeois, not true, democracy. The FAI increased its power within the anarchist movement at the expense of reformist moderates. The revolutionary wing of the UGT gained in appeal as well, with the moderates delegitimized by the failures of the Republic.

The 1933 elections left the largest share of seats in the hands of the right-wing CEDA party, accidentalist but willing to play on skepticism toward the Republic. CEDA’s uneasy relationship to the Republic led the President, the moderate Niceto Alcalá-Zamora, to nominate

60 the centrist Radical Republican Party under the opportunist Alejandro Lerroux as the head of a new governing coalition manipulated from outside by CEDA. Even without CEDA and its

“Jefe,” Jose María Gil Robles, directly at the helm, the following two years became known as the

Bienio Negro —the Two Black Years—for the rolling back of the progressive agenda of 1931-33 and for the degree of repression that the left endured. The core tension was ultimately between the labour movement and hardline sections of the armed forces and the Civil Guard.

This conflict boiled over when, in October 1934, CEDA was included in the coalition formally for the first time. A general strike and revolutionary commune were proclaimed in

Asturias, complemented by a general strike that fizzled out elsewhere in Spain and the proclamation of a short-lived autonomous Catalonia. The revolt in Asturias was brutally suppressed, the state calling in battle-hardened units from Morocco to finish the bloody job and to enact direct military control over that province. When the Popular Front re-emerged to win the elections of 15 February 1936, the stage was set for a further violent confrontation. The

Socialists decided to remain aloof from the new government. Competition with the CNT led the

Socialists to want to bolster their appeal to the increasingly revolutionary sentiment of the labour movement rather than be seen as too close to the Republic. In parallel, military officers began planning a rising. Street battles broke out frequently over the following months between cells, still small in number, organized by the UGT, CNT, and PCE on the Left and the Falange on the right.

The conflict crested in mid-July. The murder of a prominent Republican officer in

Madrid, Lt. José Castillo, on 12 July, was followed the next day by the revenge killing of the conservative parliamentarian José Calvo Sotelo.

Figure 2.1. Spain, 25 July 1936

Days later, on 17 July 1936, right -wing officers launched their long -planned coup attempt. It started—a day too soon because of miscommunication —in the units stationed in

Morocco, which quickly got behind the revolt. It was then followed the next day by at tempted risings in garrisons throughout the whole of peninsular Spain, with mixed success. In general, with crucial exceptions, the map of Spain after the rising followed the electoral map of February

1936: the rising succeeded where the right wing was str ong, and failed where it was weak (see figure 2.1). The Republic held the key cities of Madrid and Barcelona with their strong labour

62 movements, as well as a broad swath across the center-south of the country from Extremadura in the West through New Castile and to the Levante in the East, and the northern coastal provinces of Asturias (apart from a determined rebellious garrison in Oviedo), Santander and the Basque

Country. The Nationalists took Galicia in the far northwest, and the conservative cities and countryside of Old Castile and Navarre across the center-north of Spain. In addition, they scored some surprising successes in anarchist strongholds such as Sevilla in Andalucía and Zaragoza in

Aragón. The latter two regions, divided between the Republic and the Nationalists, were the site of much of the fighting of the first days.

The problems that the Republic faced in managing desertion and defection in the early months of the war had their roots in the way the war began: as a coup attempt. The state lost much of its ability to control armed force and was desperate for any help to combat the rising.

Whether seized or freely given, arms went into the hands of the workers’ militias, often in haphazard fashion. This presaged the openness of recruitment in the militia period: new recruits were often very badly screened, leaving ample room for opportunists. That there were multiple components of the Left with deep disagreements and a history of bitter competition, in addition to military personnel who remained under a cloud of suspicion, entailed quite serious factionalism among combatants. Such suspicions made it extremely difficult for individuals to trust each other, often leaving little room for norms of cooperation to emerge.

2.3. The Collapse of Central Authority in the Republic

To the Republican government, the outcome of the coup attempt was far from certain at first. José Martín Blázquez, an officer serving in the Presidential Guard in Madrid, captures this uncertainty on the night of 17-18 July: “For an hour we would be deeply pessimistic. For the

63 next hour we would be buoyed up by an illusory optimism. Everything was uncertain, and all the news was contradictory.” 9 The outcome depended on many unknowns, prominent among them

the decisiveness and hesitation of the coup plotters and the preferences of the officers that they

sought to bring onside. For example, the behaviour of the Civil and Assault Guards—

paramilitary police forces for the countryside and cities respectively—was decisive in many

locations, including the key cities of Madrid, Barcelona and Valencia. 10

As a consequence of this uncertainty, the Left Republican-led, Socialist-supported

government of Prime Minister Santiago Casares Quiroga did not apparently realize the full

extent of the threat. Casares attempted to deal with the rising largely through publicly

downplaying its extent and dismissing Franco and other coup leaders from the military. As bad

news poured in, Casares realized that his efforts were in vain and resigned on the 18 th . Manuel

Azaña, President since April, then called upon Diego Martínez Barrio, the head of the centrist

Unión Republicana (UR), to form a government to attempt to compromise with the coup leaders.

Rebuffed, Martínez Barrio resigned within one day, replaced by the Left Republican activist José

Giral. 11

The government security forces were shattered by the rising, which took tens of thousands of officers and men out of the government’s disposal. Salas Larrazábal estimates that out of 15,343 officers officially listed in the Anuario Militar (Military Yearbook) of 1936, 7,624

served in garrisons that ended up in Republican territory.12 On 18 July, Casares had also released

soldiers from the duty of obeying their officers in order to stem what was still just a coup

attempt. This had the unintended consequence of provoking a flight of the uniformed rank and

9 Martín Blázquez 1939, 101. 10 Cruells 1974, 11–14; Mainar Cabanes 1998, 17–18; Salas Larrazábal 1973, 129. 11 Accounts of these developments are in Thomas 1994, 224–230; Salas Larrazábal 1973, 89–91. 12 Salas Larrazábal 2006, 264. 64 file. They were no longer available to be ordered. Michael Alpert thus estimates that of the

90,114 regular-army soldiers serving, 46,188 remained in Republican territory but a very large proportion of those were not at the government’s disposal. 13 The same went for the various

quasi-military police forces. The Civil Guard began with 34,320 officers and men across Spain

and lost between 14,000 and 20,000 in the rising, between those in Nationalist territory and

subsequent deserters and defectors. Thousands of men were also lost from the Assault Guard,

which was established in 1931 to take over policing functions in Spain’s cities and defend the

Republican regime, and from Carabineros, under the direction of the Finance Ministry and

serving as guards of Spain’s ports. 14

Left-wing organizations were already active in Spain, however, with a view to combating the military and right-wing militias should the need arise. Therefore, with the collapse of central authority, the workers’ militias were well-placed to fill the vacuum. As noted, the months prior to the rising saw a significant increase in violent incidents among various militias of the left and right and the forces of public order. The militias had made preparations in case of a military rising. Concentrated in Madrid, the Communist Milicias Antifascistas Obreras y Campesinas

(MAOC) were a clandestine militia of at least 1,500 before the war, with ties to sympathetic military officers. The Communist leader Enrique Líster exaggerates their readiness in saying that they had started to form five battalions by 17 July, the day before the rising; but that they were on high alert is clear. 15 The Socialists in Madrid also prepared for an armed response early,

seizing arms to supply their group La Motorizada as a shock force. 16 Anarchist militias in

Barcelona under the Nosotros leadership group were similarly prepared, hoping to draw upon

13 Alpert 1989, 21–22. 14 Ibid., 25–27; Bolloten 1991, 48. 15 Líster 2007, 99; Blanco Rodríguez 1993, 26–28; Fraser 1979, 49; Alpert 1989, 18. 16 Zugazagoitia 1977, 58–59. 65

2,000 militants. 17 The different groups of union and party militants formed the committed, motivated core of subsequent militias. For example, the MAOC established a headquarters at

Francos Rodriguez Street in Madrid and sent militia volunteers there to be organized into militia units. This process was the origin of the Communist-led Quinto Regimiento or Fifth Regiment,

an important militia force discussed at some length in the chapters that follow.18 The pre-

existing Anarchist command in Madrid had, already on 19 July, the organizational capacity to

order a small contingent away from the city to get ready to intercept General Emilio Mola’s

forces coming in from Old Castile.19 That kind of organized nucleus allowed CNT militias to be

built up starting on 21 July, with “a trustworthy militant” at the head of each column. 20

Indeed the early days saw the organization of a panoply of new workers’ militias.

Thousands joined, from the key cities of Madrid, Barcelona and Valencia, and from the countryside around. Militias manned the forward defense of Madrid in the Sierra north of the city, and headed out from Barcelona and Valencia to retake Aragón and its capital, Zaragoza, from the Nationalists. They were joined by those regular soldiers, Civil Guards, Assault Guards and Carabineros who had remained loyal to the regime, formed into makeshift columns that included both soldiers and militiamen in varying proportions. On the Aragón front there were some 17,000 men in the first months; 21 in the centre, according to Ramon Salas Larrazábal, some

30,000 militiamen and a similar number of men from the uniformed security services by

September 1936. 22 The government in Madrid attempted some forms of coordination of the militias. Most important was the creation of the Inspección General de Milicias on 8 August,

17 Fraser 1979, 62–63. 18 Blanco Rodríguez 1993, 28–29; Modesto 1969, 25. 19 Montoliú 1998, 61. 20 De Guzman 67. 21 Maldonado 2007, 70. 22 Salas Larrazábal 2006, 548. 66 which now coordinated and attempted to regularize the process of requisition; 23 indicated to the

parties and unions who raised the militias where they were most needed; and paid a wage of ten

pesetas per day as of 15 August, disbursing the funds through militias themselves.24 These steps

later formed the basis of more serious centralization, as the Inspección General started to add

conditions to its provision of payment and materiel. However, the militias retained their essential

autonomy for the time being. There was no serious centralization in practices of recruitment,

with each militia column going about recruitment its own way. Military coordination was

extremely poor; the parties and unions often did not respond to the government’s efforts at

sending the columns where they were needed. 25 Further, Catalonia, the Basque Country,

Santander and Asturias maintained their own, autonomous regional governments, sometimes responding very little to the government’s directives.

The collapse of central, uniformed military and police authority gave local workers’ committees considerable power. On 20 July, two days after the rising began, two incidents in

Barcelona illustrated the power that the militias now had. The Anarchist leader Buenaventura

Durruti insisted that his CNT forces attack the Atarazanas barracks, “the last rebel bastion” in

Barcelona, and his forces succeeded in turning aside a Civil Guard company who had come to take control of the situation and proceed more carefully. On that day, as well, Police Chief

Frederic Escofet informed the head of the Catalan government, Lluís Companys, “that the commanders of the civil and assault guards could no longer rely on their men to restore order in the streets. Even if the attempt were made it would mean fighting the libertarians [anarchists], and the ensuing battle would be as heavy as, if not heavier than, the one which had just been fought in the streets. The chances of ‘restoring order’ were virtually nil.” At that point,

23 Martín Blázquez 1939, 122. 24 Salas Larrazábal 2006, 544–545. 25 Martín Blázquez 1939, 128. 67

Companys met with Anarchist leaders and offered to resign or to stay on, as the Anarchists wished, and offered in addition to form a combined Popular Front-CNT government. 26 As central authority crumbled, a social revolution emerged, especially in Catalonia and Aragón. Local workers’ committees, especially led by the CNT, took power from the established civil authorities. They expropriated property, established worker control over factories and collectivized agriculture. 27 This was much to the dismay of the moderate bourgeois Republican forces. At the same time, the PCE took the position that much more moderation was necessary to keep the middle classes on the side of the Republic and win the international favour of the

French and British. 28

It was in a context of severely weakened central authority and considerable autonomy and power for the workers’ militias that the Republic now confronted war. As the Socialist activist Julián Zugazagoitia put it, “[t]he power of the state lay in the street, pulverized, and a fragment of it lay in the hands and at the disposal of every antifascist citizen, who used it in the manner that best suited his temperament.” 29 A brutal wave of violence against civilians followed, conducted on the initiative and in the interests of anyone who could organize an armed patrol and give it a sheen of revolutionary legitimacy. 30

The militia forces that emerged from this period, confronting Nationalist advances in

Aragón, Andalucía, Asturias, the Basque Country, and the central front on the approaches to

Madrid, varied widely in their approaches to warfighting. Many allowed their soldiers

considerable leeway: officers were often elected, their commands subject to discussion and

debate, and soldiers were allowed to return to the rearguard quite regularly. Others, notably the

26 Fraser 1979, 110. 27 Casanova 1985. 28 Bolloten 1991, chap. 7. 29 Ibid., 48–49. 30 Juliá 1999. 68

Communist-led Fifth Regiment, were more demanding of their members, insisting upon discipline and obedience. I exploit this variation to help explain variation in desertion rates in the militia period in terms of recruitment and control. Chapter 3 argues that the better-disciplined militias took advantage of self-selection, attracting those with a greater commitment to the cause, while opportunists sorted themselves into less demanding forces. In turn, Chapter 4 argues that better-disciplined militias were then better able to enact control when the will of their combatants wavered. And, finally, the more tightly recruited militias were better able than the ones that attracted more opportunists to maintain norms of cooperation, a theme I explore in Chapter 5.

The Republic’s war efforts also endured the legacy of factionalism from the politics of the previous years. Since individual militias were organized by political parties and unions, competition among these parties and unions created problems of cooperation between militias.

Factionalism emerged most strikingly, however, in the treatment of regular military personnel.

Militias varied in how well they cooperated with the regular army. Some, with the Fifth

Regiment again notable among them, had a highly positive and cooperative relationship with army officers, actively seeking their input and commitment and considering their interests essentially aligned. Many other militias, especially anarchist units emerging from Catalonia, deeply mistrusted the army because of its longstanding role in repressing the labour movement and because of the coup attempt itself. Regular soldiers thus faced different levels of mistrust and suspicion. I exploit these differences in order to explore the effects of factionalism in the militia period. As I argue in Chapter 5, where militias saw their interests as aligning more closely with those of regular personnel, norms of cooperation among them could emerge. Where militias and the military were rivals, however, the latter often feared for their safety and defected.

2.4. The Process of Reform in the Republic

The first efforts at military reform in the Republic were begun soon after the start of the war, in August of 1936, with a move that attempted to provide a disciplined, centrally led alternative to the militias but did not directly coerce them. The Giral government announced the creation of “Volunteer Battalions” at the beginning of August. The militias were to convert to these units, which were to be officered by retirees and by inactive non-commissioned officers whose loyalty had been vouched for, and under central command and direction by the

Republican government. Again, however, self-selection emerged here, with the vast majority of militiamen preferring to remain in the popular militias and with new recruits joining the former rather than the new volunteer units; the project was discarded as a failure. 31 On 17 August the government tried again, this time to create a “Volunteer Army” drawing on reservists rather than militia, with slightly more success; four battalions were created. 32

However, the new government of the Socialist union leader Francisco Largo Caballero, taking office on 5 September 1936, conducted more forceful efforts toward creating a centrally directed, disciplined army, called the Popular Army of the Republic. He did so under considerable pressure from the Communists now in cabinet and from the several hundred Soviet and Comintern advisers. The advisors were alarmed at the disorganization of the war effort. As the Comintern delegate André Marty reported on 10 October, “anyone can walk into the War

Ministry unimpeded and unchecked. At the reception for the ministry there is always a crush: officers, militia commanders, union workers helping everyone there, sometimes on highly secret matters.” 33 The Soviet advisors, in response to the reigning disorganization, pushed along with the PCE for a new, regular army to be formed out of the militias. In some areas, the Soviet

31 Bolloten 1991, 250; Azaña 1967, III:488; Salas Larrazábal 2006, 540–541. 32 Salas Larrazábal 2006, 543–545. 33 Radosh, Habeck, and Sevostianov 2001, document 15, 41–42. 70 advisors’ contribution to reorganization was direct: Daniel Kowalsky credits them with a considerable direct influence in planning the defense of Madrid, for example planning where defensive fortifications were to be placed and coordinating the mobilization of militias. 34

The steps toward a regular army accelerated after the fall of Toledo, just southwest of

Madrid, on 27 September. 35 Largo began by appointing career officers with a clear preference for

command and discipline to key posts, particularly in the General Staff.36 The government then

attempted first to subject the militias to greater centralized control, and second to convert them

into new, reorganized military units. At the end of September the government declared that, as of

10 October in the central zone and 20 October elsewhere, the militias would be subject to the

Code of Military Justice; those who refused would submit their names to the Militia Command

to be struck from their lists.37 On 16 October, Largo Caballero, as Minister of War, took formal command of all of the Republic’s forces. 38 Two days later, the government ordered the creation

of the Mixed Brigades, the units that would be the basis of the Republic’s new army, and the

incorporation of the militias into these new units. The name “Mixed Brigades” referred to the

practice of including both infantry and other services (cavalry, artillery, mortars, and logistics

services) at the brigade level, rather than the more traditional division level, with a view to

allowing greater flexibility in the employment of these forces. 39 Six were created immediately,

and within a month there were twenty Mixed Brigades in operation alongside five International

Brigades of forces from outside Spain. 40

34 Kowalsky 2001, para. 591–592. 35 Salas Larrazábal 2006, 634. 36 Alpert 1989, 64–72. 37 Ibid., 72–74. 38 Salas Larrazábal 2006, 649. 39 Alpert 1989, 74–82. 40 Salas Larrazábal 2006, 652–655. 71

To bring about the militarization, the government used both carrots and sticks. Though the Inspección General de Milicias , mentioned above, was plagued by problems early on, it was

able over time to increase the government’s bargaining power with militias as its requisitions

were increasingly regularized. Now, any militia unit hoping to receive materiel or wages from

the Ministry of War now had to accept the militarization order. The Inspección General began by

covering the centre, southern and Valencia zones; for the time being the government still could

not enforce its writ in Catalonia and Aragón. 41

Access to the military resources that the Republic was now getting from Moscow would

therefore run through the state. Much of the materiel distributed by the Inspección General now, indeed, came from the Soviet Union. Moscow did sign the infamous Non-Intervention

Agreement on 23 August 1936 in order to keep favour with the French and British and draw them into an anti-German collective security arrangement. However, German and Italian breaches of this agreement prompted Stalin to reverse course and start to ship Soviet arms to

Spain. The first large-scale shipment arrived on 15 October 1936. By the end of 1936, twenty- three journeys by Soviet ships and ten by others had carried guns, planes, tanks, and Soviet military advisers, tank drivers and pilots. 42 Over the course of the war, the Soviets provided

several hundred aircraft and tanks, over 1,000 artillery pieces, over 12,000 machine guns and

around 300,000 rifles—more rifles than had been left to the Republic after the coup attempt. 43

The militia wages themselves were now administered directly by agents of the Ministry, rather than given in lump sums to the militia leadership, a system that had produced much abuse.

The stick was the threat of police and military action: against certain recalcitrant militias in Catalonia and Aragón, the regime deployed its security forces to force them into militarization

41 Graham 2002, 138–139. 42 Smyth 1996, 90–93; Alpert 2004, 76. 43 Howson 1998, 28, 142. 72 in the spring of 1937. Factionalism and mistrust undermined centralization in the short run: the centralization efforts were often characterized as Communist attacks on the Anarchist CNT and the “Trotskyite” POUM, which they often were. The militarization order thus provoked running street battles in Barcelona at the beginning of May. 44 However, these efforts at control ultimately

resulted in a more centrally directed and controlled force that would fight more reliably when

ordered to do so, with regularized procedures for punishing desertion.

At the same time as the government was centralizing and controlling militia units at the

front, it was conducting a similar process in the rearguard. Shortly after coming into power,

Largo Caballero created a Vigilance Militia to incorporate the independent rearguard militias and

patrols. They would also coordinate rearguard security with the official police. The latter, the

Assault Guards, Carabineros, and Civil Guards (renamed the National Republican Guards)

expanded over late 1936 and early 1937. A militia acting as a police force that did not join the

Vigilance Militia would be regarded as “disaffected elements,” and their arms could be seized;

but they would be given priority to join. Many local anarcho-syndicalist committees protested

the new order, but steadily over the fall of 1936 and spring of 1937, local militias and patrols

were disarmed or centralized. 45 The rate of autonomous, uncontrolled rearguard violence

declined substantially. 46

While control was expanding and strengthening, the Republican government also began to expand recruitment widely, calling up various draft classes for military service. Twenty-one- year-olds had long been subject to military service. Thus draft classes were designated by the years in which the men in question turned 21 and would normally have to serve, so that, for example, those in the 1936 draft class were already enlisted at the start of the war. The Republic

44 Orwell 1989; Bolloten 1991; Graham 2002. 45 Bolloten 1991, 216–219. 46 Balcells 2010; Juliá 1999. 73 called up the 1932 and 1933 draft classes on 30 September, followed by the 1934 and 1935 years on 7 October.47 The draft decisions in the autumn were not well enforced: Largo Caballero was

caught between pressures from the Communist Party to implement the draft, and pressures from

organized labour to desist. 48 The resistance, in fact, required the Republican government to repeat several of these draft orders. 49 However, by May 1937 the Republic had called up all draft

years from 1931 to 1937—that is, draft classes among men aged twenty to twenty-six—a total of

650,000 men. The Republic would continue to add draft classes over the course of the war, 26 in

total, to a total of 1,750,000 men. As Pedro Corral notes, this was an early resort to large-scale

conscription, compared to the American Civil War, where both parties waited at least a year, and

the Russian Civil War, in which Trotsky waited almost a year. 50

The Popular Army that emerged from the process of militarization was thus unified and

conscripted. The variation within the army was somewhat more marginal than the wide variation

during the militia period. Towns varied in the degree of control the armed forces could exercise,

with terrain serving as an important source of variation; and units varied in their propensity to

cooperation, with soldiers prone to desertion when serving among other conscripts. I explore

these sources of variation more fully in Chapters 6 and 7. I exploit the better record-keeping of

the regular era to conduct hypothesis-testing via a statistical analysis of desertion in one

particular location, Santander province.

However, beyond these sources of variation, the militarized period as a whole offers

some relatively clear points of contrast to the militia period. Desertion rates rose at first as

combatants reacted against being militarized; they then fell with the implementation of control,

47 Corral 2007, 95. 48 Bolloten 1991, 346–347; Graham 2002, 139–40. 49 Salas Larrazábal 2006, 645–646. 50 Corral 2007, 95–96. 74 and stabilized. But they increased substantially toward the end of the war. The Republic’s leadership exercised much greater control, both in the front and in the rearguard. However, conscription broadened the base of recruits to include many who did not have much interest in the Republic’s victory, as I argue in Chapter 3. I argue in Chapter 4 that part of the reason for the increasing desertion rate over time was an erosion of control. As the Republic lost ground and combatants gained motives to desert, the Republic’s existing capabilities for control were less able to handle the increased demands placed on them. Officers, those tasked with actually monitoring and inflicting punishments, sometimes became less willing to do so. More generally, mistrust among soldiers abounded. Under the militia period, mistrust certainly was substantial, but was concentrated: it was directed towards the regular army and other political factions. In the conscript army, mistrust was much more general, with soldiers realizing that their fellows did not share their common aims and could not necessarily be counted upon to do their fair share.

2.5. Nationalist Spain

In general the Nationalist army experienced fewer desertions than its opponent. Ramón

Salas Larrazábal estimates a ratio of five Republican deserters for every Nationalist, but the source of this estimate is unclear. 51 However, there is a general consensus that the desertion rate was considerably lower than on the Republican side. Across the board—in motivation, control and collective action—the Nationalist side enjoyed advantages over the Republican side. This is not to say that it achieved perfection; indeed the Nationalists experienced similar challenges, but in much reduced degree. Ultimately, these differences come down to the domination of the armed forces in the organization of the Nationalist side, in contrast to the disintegration of organized military power on the Republican side.

51 Seidman 2002, 208; citing Salas Larrazábal 1973, 1580. 75

The Nationalist forces were coordinated at first by a National Defence Junta formed a week after the rising, operating from Burgos and nominally led by General Miguel Cabanellas, though its most influential member was General Mola. At first, the coordination of the

Nationalist side was rather loose: Mola maintained his own command in the north, General

Gonzalo Queipo de Llano ran Sevilla and Nationalist Andalucía as a sort of personal fiefdom, and Franco and his subordinates directed the steady advance of the Army of Africa through southern and western Spain towards Madrid. While they disagreed on certain military decisions, on the basic principles of the need for control and unity the senior officers leading the rising were united. Over time, Franco gained in power among the officers leading the rising. He was assisted in this by the successes of the Army of Africa and by his important role as the key point of contact with the rebels’ crucial patrons, Nazi Germany and Fascist Italy. On 28 September he took the positions of Generalísimo of the armed forces and Head of State in the Nationalist zone, headquartered in Salamanca. 52 This placed him in ultimate command of all the various

Nationalist forces. In contrast to the fragmentation of the Republican side, it was with relative

unity in the apparatus of state power that the Nationalists approached war.

This was also how they approached repression. Nationalist Spain endured an enormous

degree of political violence behind the lines. The core difference with violence in the Republic,

as many have observed, is that it was largely directed from above, by the agents of the state,

rather than from below, by agents acting outside the state. The Nationalist leadership deployed

violence to purge Spain of those supposed to be of the left or their sympathizers, though a

considerable portion of the violence was conducted, as in the Republican zone, for the private

interests of individuals who stood to benefit. Altogether tens of thousands died. 53

52 Preston 1994b. 53 Juliá 1999; Richards 1998. 76

Against this backdrop of terror behind the lines, a heterogeneous Nationalist force confronted its rival. The components of the Nationalist war effort varied in their self-selection for difficult military service. First were those sections of the regular conscript army stationed in mainland Spain that joined the rising, across northern Spain from Galicia to Aragón and in western Andalucía, as well as pockets surrounded by besieging Republican forces in Toledo and

Oviedo. The Nationalists did not make the mistake that the Republicans had made of dismissing enlisted men. They did, however, lose the services of those officers and enlisted men who resisted the rising. Hence, these regiments were not entirely intact. In addition, beginning some three weeks after the start of the war, the Army of Africa, the military units stationed in

Morocco, began to be transported to the mainland. This was a substantial force, comparable to if not larger than the peninsular force available to the Nationalists. Ramon Salas Larrazábal estimates that the Nationalist army totaled 62,275 from the peninsular land forces, and 45,000 from the Army of Africa. Their force also included thousands of Civil and Assault Guards; again according to Salas Larrazábal, some 27,000. Michael Alpert gives the much lower estimate of

43,926 among peninsular land forces; Payne gives the still lower figure of 30,000 peninsular troops and 32,400 in the Army of Africa. 54

As noted earlier, the officer corps had been divided, under the Republic, into those who

had sought peninsular service, often as an entrée into the middle class, and those who had

actively sought more dangerous postings in Morocco with the prospect of combat and personal

glory. These differences between peninsular officers and those in the Army of Africa were

mirrored, to a certain extent, by the rank and file. The garrisons of Spain were filled with men

performing their required military service, but not going beyond this to volunteer for African

service. Franco began to call up draft classes once the assault on Madrid stalled in November

54 Salas Larrazábal 2006, 262–263; Alpert 1989, 21–22; Payne 1967, 346. 77

1936, and in response to low numbers of volunteers across much of Nationalist Spain—in

Galicia, at an extreme, less than 1% of the male population aged 20 to 40 volunteered. By March

1937, some 350,000 new conscripts were added. 55 In contrast, the key units of the Army of

Africa had been increasingly staffed with volunteers during the 1920s, after conscripts acquitted themselves poorly in 1919-1920. 56 The Spanish Foreign Legion, organized by Major José Millán

Astray and commanded by Franco in 1921-22, recruited among foreigners and, principally,

Spaniards from marginal social groups. Convicted criminals, for example, were quite common.

They could receive redemption for their crimes and opportunities to commit further brutality against Moroccans, but in return they agreed to accept brutal discipline in the sense of blind obedience enforced with the whip. 57 Franco also recruited thousands of Moroccan Regulares , mainly on a voluntary basis, with the promise of payment and land and out of alliances with local leaders in Morocco. The Regulare units, again, were subject to strict control and discipline. 58

According to Payne, the Legion numbered 4,200 at the start of the war, the Regulares 17,000, and there were in addition 11,000 regular army troops who had volunteered for African service.59

The experience and effectiveness of these forces, and their reputation for brutality against civilians, gave the Army of Africa a fearsome reputation in the Republican front line and rearguard.

Finally, the radically traditionalist, Catholic and monarchist Carlists contributed their own militia, the Requeté , and the fascist Falange organized new militia units as well, which grew at a very fast pace through the summer of 1936. As with the workers’ militias in the Republic, the Falange and the Carlists were prepared to support the rising when it came. Individual Falange

55 Payne 1967, 387–388; Seidman 2011, 26. 56 Cardona 1983, 171–172; Payne 1967, 155–157; on prior problems with conscripts, see Balfour 2002, 22. 57 Payne 1967, 156–157. 58 de Madariaga 1992; Balfour 1997, 278. 59 Payne 1967, 346. 78 cells had, in the past, been given instructions to support such a rising, and aided the rising in its success in a few cities, quickly organizing militia. In fact the Falange had been, according to

Payne, itching for a rising. The Falange’s leader, José Antonio Primo de Rivera, the son of the old dictator, demanded that Mola speed up the timeline. 60 For their part, the Carlists had built up a substantial militia force over time, and over the course of difficult negotiations with Mola, ultimately promised to contribute 10,000 trained militiamen for the assault on Madrid. Similar to workers’ militias discussed above, Carlists had laid a kind of surreptitious groundwork: for example, one Carlist organizer in Burgos hid rifles in his father’s bakery, which he then retrieved once the coup attempt began. In the Carlist heartland of Pamplona, 6,000 men were mobilized almost immediately after the rising began. Carlist units assisted the military rebels in Álava,

Burgos, and Zaragoza. Falange and Requeté units played an important role on the front from the first, particularly among Mola’s forces in the North.61

Estimates of the eventual size of each militia vary substantially. For the Carlists,

Blinkhorn cites a maximum figure of about 70,000 Requetés at any one time, while Casas de la

Vega’s figures never rise above 24,000. For the Falange, the party itself estimated some 80,000

militia by the end of 1936, while again Casas de la Vega has a much smaller figure. In October

1936, he estimated that there were 36,809 against 22,107 Carlists, out of a total of 188,581 total

in the Nationalist armed forces at that stage. 62 Militia members were therefore a substantial force, particularly in the northern campaigns out from Old Castile to Madrid to the south and the

Basque country to the north. Relations between the key army officers and the Carlists especially had long been strained, because the Carlists suspected that the officers preferred a military dictatorship to the true restoration of monarchy. As for the Falange, its fascist personality cult

60 Ellwood 1987, 30–31; Payne 1961, 114–115. 61 Payne 1961, 115–120; Blinkhorn 1975, 252–254; Fraser 1979, 55. 62 Blinkhorn 1975, 256; Ellwood 1987, 33, citing; Casas de la Vega 1974; Payne 1961, 146. 79 centering on Primo de Rivera suggested a different basis for political power for the new state than the officer corps itself. However, both the Carlists and Falange were essentially willing to put aside these disagreements to the end of unity, so that their factionalism was limited. The militias also did not enjoy nearly the degree of autonomy that the Republican counterparts did.

From an early stage, Requetés were attached to units of the regular army and accepted operational instructions from military officers. Falange units were starved for effective officers and had to look to the regular army for leadership. Many of the best Falange units were co-opted and placed under army command, leaving less effective and committed militiamen to people the remaining militia. Toward the end of December 1936 Franco ultimately ordered that the Falange and Carlist militias be fully subordinated to the army, and in April he announced the fusion of the two parties, under his own command, and the dissolution of all other political forces. Though this finally provoked a backlash among Carlist and Falangist leaders seeing their autonomy slipping away, the coercive power of the security forces ultimately had the last word: dissident

Falangist and Carlist leaders were quickly rounded up and unity was preserved. 63 This was the

last gasp of any serious political competition in Franco’s Spain: there was now a single political

party, and Franco stood unrivalled among the military leaders as well.

Principally, then, Nationalist Spain offers an instructive contrast to the Republican side,

with fewer problems of motivation, control, or collective action. In terms of motivation, I argue

in Chapter 3 that the vast majority of Nationalist forces were subject to strict discipline from an

early stage, so that they experienced much smaller problems of self-selection of opportunists into

ill-disciplined units than did the Republican side. However, there was still some analytically

useful variation within the Nationalist camp as well. There were some exceptions to strict

discipline among civilian militias organized by the Falange, which did provoke some pernicious

63 Blinkhorn 1975, 267–268; Payne 1961, 142–145; Ellwood 1987, 39–40. 80 self-selection effects. As with the Republic, self-selection operated in a positive direction as well.

Well-disciplined volunteer units such as the Army of Africa and the Carlist Requeté militia, like the Communist units on the Republican side, enjoyed a comparative advantage in the commitment of their members over conscript and Falange units. Like the Republicans from militarization onward, much of the Nationalist force was conscripted, and thus of dubious motivations and reliability. Unlike the Republican side, however, the Nationalists rarely faced severe losses that prompted a loss of control and large-scale flight among these conscripts.

Instead, as I argue in Chapter 4, control was maintained rigorously and consistently across

Nationalist forces, and so even those without much commitment to the cause faced significant barriers to deserting successfully.

Finally, the Nationalists suffered less from factional competition. This meant that there was less opposition to central control by the state, as I argue in Chapter 4. The campaign of terror did produce major problems of mistrust of purported leftists, in some respects the mirror image of the Republican side. Those suspected of leftist tendencies had very good reason to desert. The

Nationalist command occasionally attempted to limit the damage that this approach caused to its available manpower, telling those with a left-wing past that they could be redeemed through committed and loyal action on the front lines. However, this policy proved to be in tension with the violence of repression, and the promise of forgiveness was often broken—provoking desertion or defection. In any event, however, control was severe enough that those with a leftist past often had no real ability to desert or defect.

2.6. Conclusion

The following chapters, then, employ several sources of variation in order to examine the plausibility of the mechanisms that my approach proposes. I summarize the sources of variation and the areas of analysis in Table 2.1. Ultimately, the empirical questions for the Spanish case are as follows. First, why did some Republican militias experience much larger desertion problems than others? Second, why did the Republican army in the militarized period experience desertion rates that were initially lower than under the militia system, but increased dramatically over the course of the war? And, finally, why did the Nationalists have, overall, lower desertion rates than the Republicans?

I answer the three questions as follows. First, those militias that had higher desertion rates were those that employed weaker policies of control towards their combatants. This not only weakened the ability of those militias to prevent soldiers from leaving but also allowed the self- selection of unreliable recruits. Those units, in turn, suffered from weaker norms of cooperation.

Second, militarization resolved much of the problem of lack of control, but generated a higher potential for desertion over the long run by weakening the overall commitment of soldiers and undermining cooperation. This potential was realized once military defeats undermined control and provoked collective desertion. Finally, the Nationalists had less difficult problems of desertion because control was much more readily maintained and political competition did not undermine control or norms of cooperation to nearly the same degree as in the Republic.

The next three chapters develop the answers to these questions by explaining the three logics of common aims, control, and cooperation. They thus provide empirical detail for the central components of the dissertation’s theoretical argument.

Table 2.1. Empirical summary, Chapters 3 through 5

Areas of analysis Sources of variation Common Aims Control Cooperation (chapter 3) (chapter 4) (chapter 5) Variation among Voluntarism and use of A common problem of Voluntarism and self- Republican militias political membership little rearguard control. selection allow strong across all units allows Control within units norms of cooperation to for false negatives. enables some militias to emerge in some militias. Some militias employ prevent desertion. Presence of opportunists tighter discipline and and suspicion of army thus enjoy self-selection personnel in others for more committed creates problems with recruits. Factionalism is norms of cooperation a considerable problem and prompts desertion. with regular forces in some units but not others. Competition among parties and unions creates factionalism across militias as well. Changes in Republic Adoption of Efforts to create central Norms of cooperation over time conscription leads to control are undermined are undermined by (militarization) less committed troops. in the short run by conscription; in extreme Factionalism persists suspicions of cases, norms of over time. Communist plots, but cooperation among limit desertion over the prospective deserters long run. Control over favour mass desertion as rearguard and within defeats increase. units strengthens at first, then weakens with defeats by the end of the war. Nationalists vs. Generally disciplined Control is much Nationalists do not have Republicans militias mean generally stronger on Nationalist nearly as large problems less difficulty with self- side than Republican of political competition selection of opportunists side at both rearguard as Republic, but still than in the Republic, and front line. Weaker experience problems of with some exceptions. factionalism and mistrust of leftists and More reliable troops stronger control mean among conscripts. self-select into Army of efforts to centralize Africa, and out of control over militias are regular army. As in not nearly as Republic, conscription problematic as in the weakens regular army Republic. motivations.

In Chapter 3, I argue that some forces on both the Nationalist and Republican sides did, from time to time, discipline their recruits and hence screened for a relatively high degree of commitment on the part of their soldiers. Attempts by both sides to screen by political labels may have helped to a certain degree, but generated significant problems of false negatives. It was much more effective to screen by self-selection—maintaining voluntary enlistment and ensuring that enlistment implied a very high commitment on the part of volunteers. In contrast, opportunists took advantage of units with laxer standards of discipline. Conscription, of course, implied much more widespread problems of motivation. Finally, Chapter 3 discusses the problem of factionalism on each side.

Chapter 4 then takes up the variation in the degree of control over soldiers. Since the variation in this regard is principally on the Republican side, I focus on Republican developments while including the Nationalist side as a complement. I show that many

Republican militias failed to maintain control and thus to prevent desertion. I then demonstrate how the steady establishment of control on the Republican side first provoked some desertion because of mistrust, then improved the prospects of preventing desertion before unravelling at the end of the civil war. I revisit how control operated in the rearguard in Chapter 6, in a study of the impact of terrain on desertion rates.

In Chapter 5, I examine the operation of norms of cooperation. I demonstrate that soldiers

frequently responded to the preferences of the men around them: combatants could be strong

influences on each other. Among groups of committed combatants, as in tightly recruited

volunteer forces, norms of cooperation could be quite strong. I then demonstrate the problems of

factionalism. Combatants who were actively mistrusted by their peers, as in the case of

uniformed military and security personnel on the Republican side, knew they could not count on

84 their colleagues and so frequently deserted. The use of conscription on both sides, in bringing in so many uncommitted soldiers, threatened norms of cooperation. Indeed, some soldiers who did not want to serve cooperated to desert together. I take this subject up again in Chapter 7, with a more detailed analysis of cooperation under the Republic. By the end of these chapters, I hope to have demonstrated the richness of my theoretical approach, its ability to elucidate features of a case in some depth.

Chapter 3 Common Aims: Recruitment and Factionalism

3.1. Introduction

In this chapter, I examine how the different armed groups of the Spanish Civil War came to have common aims to greater or lesser degrees. I focus on two major influences: their practices of recruitment and their degree of factionalism. In examining recruitment policies, the chapter explicates some of the consequences of the armed group’s shortage of a priori information about soldiers’ commitment. Given such a lack, a particularly telling and effective way of gauging soldiers’ preferences is to rely on them to self-select. This logic is borne out in comparing conscription to voluntarism, and, among volunteers, examining the differences between those who are committed and those who are not. The first issue is obvious: conscripts are much likelier to be uncommitted than volunteers, because the latter took on the decision to fight even without as much of a threat of physical violence. On both the Republican and

Nationalist sides, volunteer units coexisted with conscript units, and conscripts steadily replaced volunteers. With forced recruitment came a massive dilution of the commitment of the individuals fighting: as long as men had the choice about whether to fight or not, the indifferent at least had the option of self-selecting out. This distinction is basic enough, however, that it does not really require extensive treatment here; I deal with it briefly at the end of the chapter. Its brief treatment, however, should not be taken as a sign that it was unimportant; quite the contrary.

Because this point is stronger, it is more obvious; the differences among volunteers require more extensive discussion.

Despite the basic distinction between forced and voluntary enlistment, not all volunteers were equally committed to the cause. It is true that even enthusiasts who joined in the first rush of war often lacked a clear idea of what it would entail, and deserted out of fear when they met the first rigours of war. A commitment to the common cause does not guarantee non-desertion; this is the main lesson of Mancur Olson’s collective action problem. But beyond the enthusiasts, other volunteers were opportunists, who joined to seek personal gain. Efforts to keep the uncommitted out of volunteer units by checking their credentials beforehand often failed: such visible signals as a membership card can often be faked, and so constituted a cheap rather than a costly signal. Ultimately, such cheap signals do not resolve the fundamental imbalance in information: that the recruit knows much more about his own preferences than the armed group leader. More telling than membership cards, because it responded to the imbalance of information, was self-selection. Opportunists would join those units where they could reap the benefits of service, such as payment, looting, and violence, without paying its costs, such as having to fight. That is, they joined units that were lax in discipline. Militias developed such reputations by word of mouth and public statements. When a volunteer unit imposed severe discipline, however, its combatants joined knowing they would have to fight. Opportunism would consequently be much more difficult. Volunteers who were more interested in military success thus tended to join more disciplined units. And if an ill-disciplined unit wanted to achieve military aims, it would have to change its policies, requiring its soldiers to fight; but given the motivations of the recruits who joined such units, to do so was to risk further desertion.

Both sides also suffered from factionalism, the Republican side much more severely than the Nationalist side. There was severe mutual mistrust between regular security forces and civilians. And there were deep disagreements, competition and mistrust among various factions

87 of the Left, and particularly the conflict between the bourgeois Republic, its Communist supporters, and the more fire-breathing Anarchists and revolutionary Socialists. Factionalism set up severe problems down the line: the various groups in the Republic were frequently working at cross-purposes to each other. On the Right, there were certainly substantive disagreements and competition, but these were less severe than those on the Republican side, and had provoked little past violence. Mistrust, therefore, was a much less severe problem for the Nationalists.

This chapter proceeds by comparing across volunteer units on the Republican side, building on some fruitful comparisons with the Nationalist side, and finally drawing out the distinctions between conscripts and volunteers.

3.2. Voluntarism and Opportunism in the Republican Militias

Recruitment was voluntary in the militia units of both the Republican and Nationalist sides. Volunteers would generally be more committed to the cause than conscripts, but that commitment was not universal. Many joined for self-interested reasons and sought to shirk difficult duties. When they experienced the costs of fighting, then—when they encountered combat or their commanders decided to impose discipline so that they would have to fight—such individuals were inclined to desert. On the Republican side, militiamen joined for the high wage of ten pesetas per day. On both sides, others joined to loot or to commit revenge, all in the name of the cause. Still others joined in order to keep up appearances in a context of persecution and violence. Opportunists would tend to join those units where less would be demanded of them, because discipline was more lax. This process of self-selection created a basic difference among the Republican militias. And since many more Republican militias adopted loose discipline than did militias on the Nationalist side, the Republican side faced concomitantly greater challenges.

In the Republic, many joined the militias during the popular reaction against the coup attempt. The coup was not defeated by ordinary people in the streets; it was defeated by the concerted acts of loyalist soldiers and the small numbers of long-standing union and party activists that supported them. The public did show its enthusiasm for the cause once the rising had been defeated, however. In the display of enthusiasm, with workers coming out in huge numbers into the streets for marches and rallies, many were swept up enough to join the new militias and head to the front. Josep Cercos, an Anarchist metalworker, discusses the enthusiasm in Barcelona to head out west to Zaragoza, capital of Aragón, to stop the Nationalist advance:

“We were all workers. There was a tremendous fever to reach Saragossa, to take it.” 1 A POUM

militant, Wildebano Solano, confirms: “The people’s revolutionary instinct was amazing…They

knew they had to inflict one defeat after another, move ahead every minute. There wasn’t a

moment to lose. The cry went up – ‘To Saragossa.’”2 In the anti-fascist fever, some individuals displayed immense courage. Militia put up fierce resistance in the mountains to the northwest of

Madrid and in some instances in Andalucía. 3 Rosario Sánchez Mora, one of many women who served in the militias, describes the faith that her unit had in eventual victory: “We were young and we thought that the young, alone, would end the war, that we were going to devour the enemy.” She became a public hero in Madrid for having her hand blown off while laying dynamite; there was a poem written about her entitled “Rosario, dinamitera.” 4

It is important to emphasize that enthusiasm was no guarantee that a soldier would carry

on fighting. They learned quickly that there were serious costs of war. And those costs inspired

some to desert. This process of learning the costs of war is fairly general. John Keegan leads his

1 Fraser 1979, 119. 2 Ibid., 120. 3 Seidman 2002, 54. 4 Montoliú 1999, 64–65. 89 classic study of battle like this: “I have read about battles, of course, have talked about battles, have been lectured about battles and, in the last four or five years, have watched battles in progress, or apparently in progress, on the television screen…But I have never been in a battle.

And I grow increasingly convinced that I have very little idea of what a battle can be like.” 5 The

Republic’s militiamen underwent a process of discovering the costs of war. A French military

attaché put the basic problem in a report to Paris about the situation in Madrid: the militiamen

“are dressed, after a fashion…they are armed; they leave, without understanding, for the front,

where they discover, too late, that war is serious.” 6 The columns that headed out to Aragón from

Barcelona, fuelled by this popular fervor, soon felt the costs of war. Juan García Oliver, an

Anarchist leader from Barcelona, raised a column called Los Aguiluchos in late August 1936 to

break the deadlock that had emerged in Aragón. García Oliver recounts considerable discontent

among his men the first night, setting camp in the woods with cold rations only because the local

townsfolk in Grañén, Huesca province, refused to billet them. In his view, his men “were taken

at a stroke away from the comforts of home. Against blunt reality, the revolutionary fantasy had

ended, and they felt themselves caught in a trap.” 7

García Oliver’s own soldiers got used to the new conditions, but others did not. Battle exposed the fragility of initial enthusiasm in maintaining a willingness to serve. Air attacks were particularly intimidating. 8 Narciso Julián, a Communist railwayman who had joined the

Communist-organized Del Barrio column, recalls the extreme panic that set in after an air raid in

Huesca province on the Aragón front. “‘Everyone, including the anarchist machine-gunner, leapt

for cover; I never saw him again. The engine driver was the only one to show any control…’ So

5 Keegan 1976, 15; see also Grossman 1996, 2. 6 Thomas 1994, 434. 7 García Oliver 1978, 261–262. 8 Seidman 2002, 52; Cirre Jiménez 1937, 46. 90 many men left the column after the raid that they had to remain in the township of Grañén to reorganize.” 9 Like aircraft, Moroccan units, presented in Republican propaganda as the Moorish

hordes of old, provoked particularly panicked responses. 10 Many militia columns fell apart to desbandadas (disbandments): instances of panicked flight and, often, large-scale desertion.

These occurred especially early on when units had their first taste of battle, for example when two Communist columns fled from the Siétamo area in Huesca around 26 July—that is, eight days after the start of the war. 11 The Anarchist commander Victor de Frutos vividly captures the excuses given for a disbandment: “Sometimes the pretext was a lack of munitions; others, feeling sick and looking for a corner to relieve one’s bowels, and, other times, the men yelled openly,

‘They’re cutting us off!’ This spirit would become a contagion, until the retreat became universal.” 12

Some commanders recognized the hardships that their militiamen endured. Manuel

Uribarry, a Civil Guard captain, organized a column in mid-August in Valencia and headed out to the Sierra de Guadalupe, southwest of Madrid. The column was composed of 1800 militiamen plus several companies and squadrons from various military and police units. Despite considerable propaganda surrounding this unit and a supposed emphasis on discipline, it suffered a serious desbandada in Guadalupe in August, with apparently massive desertions. 13 In Valencia

again in October, Uribarry delivered a speech in which he expressed understanding for why

many of his men had deserted: “One might say that men fled for a lack of fighting spirit, but I

must confess to you that many times, when I saw them flee and I saw their unshod feet covered

in sores, saying to me, ‘I cannot continue,’ I said to myself, ‘Perhaps I would have fled sooner

9 Fraser 1979, 134. 10 For example, see de Frutos 1967, 37–40. 11 Arcarazo García 2004, 111. 12 de Frutos 1967, 42. 13 Cardona 1996, 49; Mainar Cabanes 1998, 33–35. 91 than you.’” After all, Uribarry continued, his feet were fine, and he had not had to deal with hunger or cold. 14

The costs of fighting a war meant that enthusiasm was no guarantee that a soldier would not desert. It was costly to participate, frequently less costly not to. Effectively the Republic’s enthusiastic militiamen faced the classic collective action problem. The common aim of battlefield success, the source of their enthusiasm, would be sustained if everyone fought, but each individual soldier would pay serious personal costs to fight. Since control and norms of cooperation are solutions to the collective action problem, it was relevant whether the soldiers in question were adequately controlled, and whether they developed norms of cooperation. I discuss these issues, respectively, in Chapters 4 and 5.

But those militiamen at least had the advantage of shared aims. Many, in contrast, joined for much less public-minded motivations. It is true that opportunists—individuals who seek to take advantage of war situations to loot and to commit violence for their own private ends— could often find their opportunities in the rearguard rather than at the front, with revolutionary committees appropriating the assets of wealthy landowners and plenty of individuals taking advantage of the collapse of central authority for the purposes of theft. 15 Others murdered,

joining irregular patrols in the rearguard and hunting down old enemies whether rightists or

not. 16

The rearguard opportunities available generated a selection effect that, interestingly, selected out unreliable elements from the front-line militias to a certain extent. If there were better options for personal advance in the rearguard, then those who went to the front were often

14 “El importante acto de la C.N.T. en el Principal,” Fragua Social (Valencia), 10 October 1936, pp. 2-4. 15 Bolloten 1991, ch 5; “¡Ni pillaje, ni saqueo, camaradas!” Solidaridad Obrera (Barcelona), 24 July 1936, p. 1; “Un caso bochornoso de pillaje,” Solidaridad Obrera (Barcelona), 29 July 1936, p. 1. 16 Graham 2002, 161. 92 those who eschewed such opportunities entirely. The anarchist leader Frederica Montseny acknowledges that “[t]he most audacious and the most idealistic went to the front....The type of man that was left [for the rearguard patrols] was neither intrepid nor capable.” 17 Eduardo Pons

Prades recalls that the Barcelona woodworkers’ union “lost at least half of its best militants, between those killed in the streets of Barcelona and those – the greater number – who went to the

Aragon front.” 18 Opportunists would thus often avoid service in the militias altogether.

However, there were clearly many opportunistic reasons to join the front-line militias as well. There were material rewards to militia service. The first were formal: the Republican government announced on 18 August 1936, after one month of war, that militiamen would earn ten pesetas per day—between two and five times the typical wage for a rural day-labourer in much of the Republic, and indeed over five times higher than soldiers’ wages before the war. 19

George Orwell, who generally insisted upon the “straightforwardness and generosity…[the] real largeness of spirit” of the POUM militiamen with whom he served in Aragón, also noted that

“boys of fifteen were being brought up for enlistment by their parents, quite openly for the sake of the ten pesetas a day which was the militiaman’s wage; also for the sake of the bread which the militia received in plenty and could smuggle home to their parents.” 20 In Madrid, in the context of severe wartime hunger, the Communist-affiliated aid organization Socorro Rojo

Internacional (International Red Aid), apparently only delivered food aid to the families of militia members. 21

17 Bolloten 1991, 773. Note 73. Citing Pons 1977, 141-145. 18 Fraser 1979, 222. 19 Corral 2007, 86–87. 20 Orwell 1989, 10. 21 Gutiérrez Rueda and Gutiérrez Rueda 2003, 60. 93

In addition, some militia units offered other kinds of opportunities for enrichment.

Militiamen in Madrid enjoyed some material privileges. 22 They traded on the social value of appearing to be heroic, for example gaining the ability to eat in cafes without paying. 23 Wives

and girlfriends—and prostitutes—followed militiamen to the front. 24 The war experience was eased considerably by the fact that militiamen could often return home at night. 25 At the militias’ siege of the Alcázar of Toledo in July-September 1936, “many were ‘tourists’ of war, who drove out with their wives or girl friends from Madrid for an afternoon’s sniping.” 26

Looting and murder were hardly the exclusive preserve of popular justice, local committees and patrols. Militias in various locations requisitioned goods and even conducted looting and pillaging. Militiamen who served where they were from had clear opportunities to conduct personal vendettas. For example, militiamen in Madrid killed an officer friend of José

Martín Blázquez’s. The officer was “far from being a reactionary”; instead, according to Martín

Blázquez, “the motive for the crime was personal vengeance.” 27 When militias served far from

home, some, particularly the Catalan militias in Aragón, sometimes looted in the name of

requisitions for the war effort.28 The Columna de Hierro or Iron Column gained particular infamy for looting and violence, both in Valencia province (where it was from) and in Teruel province in Aragón (where it was often stationed). Begun by anarchist firebrands in Valencia, it counted hundreds of prisoners from San Miguel de los Reyes Penitentiary among its ranks.

Although some of these were already anarchist activists and others had joined the movement while in prison, most “were hardened criminals who had undergone no change of heart and had

22 Seidman 2002, 35; Martín Blázquez 1939, 209–210. 23 Martín Blázquez 1939, 125–126. 24 Thomas 1994, 325. 25 Fraser 1979, 117. 26 Thomas 1994, 325. 27 Martín Blázquez 1939, 135. 28 Casanova 1985, 110–111; Seidman 2002, 36; Modesto 1969, 33–34. 94 entered the column for what they could get out of it, adopting the Anarchist label as a camouflage.” The Iron Column would steal precious metals from jewelers’ shops to pay for war material, and cut a swathe of “purifying” violence through Valencia province. 29 Looting is, of course, a common consequence of war of all kinds, 30 and the regular Popular Army conducted its own abuses against the civilian population much later on in the war, for example with uncompensated “requisitions” of food in Aragón. 31 But those abuses take on a different cast in a

conscript army than in a volunteer army: in the latter they probably did not affect recruitment, as

individuals had to join anyway. The analytical weight of looting, for my purposes, comes in the

context of the volunteer army, where it clarifies what was in it for someone to join a militia,

given that they could freely choose to do so or not.

Violence against suspected right-wingers in the Republic generated further self-interested reasons to join the militias: in order to pass. Some rightists caught on the Republican side joined militias deliberately in order to cross the lines. This apparently happened with some regularity in the first days, as rightists attempted to flee the Republican zone. In Barcelona, Solidaridad

Obrera reported eight fascists in a militia who defected to the Nationalists at the beginning of

August. 32 José Cirre Jiménez, a rightist artillery officer who found himself on the Republican side at the outset, remembers meeting a man in his unit who had joined the militia because in his home province of Murcia, he was in danger of persecution for known right-wing sympathies. He went to the front first to escape this persecution and second to try to defect to the Nationalists. 33

Thus there were many opportunistic motivations for joining militias on the Republican

side. In turn, the militias themselves often either could not or did not have the ability to keep

29 Bolloten 1991, 333–334. 30 Van Creveld 2004. 31 Graham 2002, 352–353. 32 "¡Alerta! Hay fascistas enrolados en las milicias," Solidaridad Obrera (Barcelona), 1 August 1936, p. 6. 33 Cirre Jiménez 1937, 92–93. 95 such opportunists out. The fragmentation of power at the beginning of the war led to a chaotic process of arming the public, and thus most anyone could form a patrol or a militia. Later, militias began to regularize the process of recruitment. Even though many militias had policies of screening, recruiting only those with union and party membership cards, competition among various parties and unions meant that different political groupings had incentives to maximize their membership, so such cards were not difficult to find.

From the first, the major labour unions called on the government for arms. The two short- serving prime ministers Casares Quiroga and Martínez Barrio, who served on the first two days of the war, refused this demand. President Azaña wanted to avoid it as well. In the conventional wisdom, this was out of their fear of revolution and disorder from the left. 34 Regulo Martínez, a schoolmaster affiliated with the Left Republicans, saw Azaña during the early days, and later reported the President’s fears that “assassinations and pillaging” would result from such a step. 35

In Barcelona, the Generalitat —the semi-autonomous government of Catalonia—also refused, for the same reason. For President Lluís Companys and Chief of Police Frederic Escofet, “[t]he

CNT...posed as serious a threat to the Republican regime as did – from the opposing camp – the military revolt. And what, moreover, if the CNT’s presence in the streets caused the guardia civil, whose attitude was uncertain, to join the rebellion?”36

Despite the government’s reluctance in the first two days, the workers still gained their

arms. Many were handed out by sympathetic officers, notably the Socialist head of the Madrid

Artillery Park, Lt. Col. Rodrigo Gil. Although the arms were supposedly distributed according to

five improvised but accredited battalions, each led by a uniformed officer, the battalions existed

34 Bolloten 1991, 39–40; Graham 2002, 24; Montoliú 1998, 54; Salas Larrazábal 1973, 126; Thomas 1994, 227; Zugazagoitia 1977, 57. 35 Fraser 1979, 74. 36 Ibid., 62–63. Citing Escofet 1973, 184, 189. 96 only on paper, and the arms made their way to the workers’ militias. 37 Others were seized by

force, often with little clear sense of who would wind up with them. CNT leader Cipriano Mera

recalls that at an arms deposit in Madrid on 19 July, he saw people taking the rifles they wanted

without any sense of control; the following day, when he accompanied militias in taking over the

army base at Carabanchel, he had to give arms to the mass that had gathered around, “because to

deny them that would be to risk being lynched.” 38 In Barcelona, the CNT availed themselves of

thousands of rifles from the Parque d’Artilleria Sant Andreu. 39 There, according to Andreu

Capdevila, a CNT activist:

They started taking whatever arms they could lay their hands on. More and more began to

arrive from all over the city, in cars, lorries, any form of transport. Everyone was mad to

get arms.... ‘We don’t know who these people are,’ I said to my companions. ‘They may

be fascists for all we know.’... There was total disorder. We formed a commission, and

thereafter all arms were handed out only to revolutionary organizations. 40

The disintegration of central power thus created opportunities for those who wanted to seize some measure of power for themselves—in this case, a rifle. Capdevila’s concerns, and Mera’s, were a first sign that the centrifugal forces taking power out of the hands of the state and putting it in the street went beyond the wishes of the militia leaders. It was immediately unclear that those who had arms and thus could fight had any desire to do so. After the workers’ militias had begun seizing arms on their own, the Giral government made a virtue of necessity and began authorizing these arms transfers. Ultimately, then, the availability of guns at the start of the war

37 Bolloten 1991, 39; Martín Blázquez 1939, 112; García Venero 1973, 382–383. 38 Mera 1976, 17–18. 39 Cruells 1974, 29–30; Fraser 1979, 71. 40 Fraser 1979, 71–72. 97 helped to give opportunists an early start, able to use weapons for their own purposes—which, as we have seen, included revenge and looting.

Later, as the process of recruitment became more regular, individuals who wanted to join militias would usually have to present some sort of guarantee: either a union or party membership card, or a voucher from a trusted union or party member. This was in the context of enormous growth in the unions especially. According to a claim made to the anarcho-syndicalist

International Workers’ Association in June 1937 by one of its representatives, CNT membership increased from 150-175,000 before July 1936 to over one million; UGT membership increased from 30,000 to over 350,000 according to a Catalonian UGT official. 41 In far more conservative

Santander province, where the unions had quite limited pre-war presence, the UGT-affiliated

Federación Obrera Montañesa increased its membership fourfold, from 16,502 immediately

prior to the war to approximately 70,000 during the war; the CNT increased its membership from

2,545 in April 1936 to 19,845 by February 1937. 42

It is clear that many of these new recruits had little real commitment to the left. Rather,

the need to appear anti-fascist generated an enthusiasm for the cause that was more apparent than

real. Felix Carrasquer, a FAI militant in Barcelona, remembered the reaction of the public with a

note of bitterness: “Where there were 2,000 of us libertarians who rallied to put down a fascist

coup...by 8 a.m. the next day there were 100,000 in the streets.” 43 Cipriano Mera found a rather dubious response in a tour of villages around Cuenca (in New Castile east of Madrid) some ten days after the start of the war. In many of those villages, there was no real indication of pro-

Republican sentiment, but once the militia arrived, CNT and UGT supporters suddenly appeared.

41 Seidman 1990, 94. 42 Solla Gutiérrez 2006, 102, 115–116. 43 Fraser 1979, 107. 98

Mera was convinced that many locals, despite what they might profess, lacked left-wing convictions. 44

The difficulty with a screening method based on some label, such as a membership card, rather than a set of actions or self-selection, is that there are frequently ways of faking the label involved, including simple theft. Using membership cards was therefore prone to false negatives.

The CNT organ Solidaridad Obrera in Barcelona contained numerous reports of lost or stolen

CNT membership cards, publishing lists of names in an effort to ensure that those cards could

not be used. 45 One could obtain credentials through connections, as a kind of privilege. Marcial

González Bonet, a Falangist student organizer before the war, claims that a member of the joint

Socialist-Communist youth wing, the Juventudes Socialistas Unificadas (JSU) could, once established, obtain a card that said he had joined before the rising. 46 In addition, parties and unions had incentives to maximize membership. Some local unions, such as the CNT in Aragón, in fact, forced workers to join in order to keep their jobs. 47 Maximizing membership

strengthened union and party demands for access to funding, arms, cabinet posts, and other

resources of the central government. As such, each had an incentive to allow lapses in their

screening and enable those of dubious commitment to the Republic to join. The Anarchists

apparently had particular difficulty in screening because of their loose, confederal cell structure;

it meant that individual union locals could act on their own account in distributing

memberships.48 Precisely because they were so useful, a market for membership cards opened up. Solidaridad Obrera reported that fascists in Madrid were attempting to buy CNT

44 Mera 1976, 26. 45 For example, Solidaridad Obrera (Barcelona), 29 July 1936, 2 August 1936. 46 Montoliú 1999, 199–200. 47 Fraser 1979, 352–353. 48 Graham 2002, 87; Cervera 2006, 112; Montoliú 1999, 139, 149. 99 membership cards for 100 pesetas. 49 In response to such problems, certain CNT unions only started handing out membership cards to those who were also vouched for by an existing member. 50

Even if it was particularly pronounced among the CNT, the problem of falsifying

membership cards was general, cutting across all political factions. In fact, each side accused the

others of allowing opportunism into the ranks of the left: “Socialists blamed republicans and

communists; republicans blamed the communists, and everyone blamed the CNT, who, in turn,

blamed the communists and republicans.” 51 In Madrid, hidden right-wingers tended to join the

JSU, which had expanded quickly just as the CNT had. 52 González Bonet, the Falangist

organizer, claims that the JSU never checked into his political background. 53 Gerardo Martínez

Lacalle, another Falangist organizer, joined the JSU also: “I don’t know who told me it would

help me to join them. But I found that in that group we were all fascists.” 54 Sócrates Gómez, a

Socialist youth militant in Madrid, claimed that the Communists attracted the uncommitted with

promises of better jobs: “Accept a communist party membership card – and promotion. I don’t

say this lightly, I know what I’m talking about. The communist party grew strong on this

procedure. A membership card, a post. It made a big impact on people with no particular

political loyalty.” 55 Eventually, once the concerns about the reliability of these new entrants

became clear, union militants expressed regret. In an internal document in 1938, a Barcelona

CNT power-company manager noted: “one of the principal errors of the unions was to force the

49 “Los fascistas quieren comprar carnets,” Solidaridad Obrera (Barcelona) 27 July 1936, 1. 50 “Horas de depuración,” Solidaridad Obrera (Barcelona), 8 August 1936, 2. 51 Graham 2002, 293. 52 Ibid., 176. 53 Montoliú 1999, 199–200. 54 Ibid., 209. 55 Fraser 1979, 333. 100

workers to join one of them. We are not really sure about many of the huge numbers of new

workers, although it’s not worthwhile to discuss this outside of the union.” 56

The saga of Republican membership cards illustrates the perils of screening through

labels. It was simply too easy to obtain one, via theft or corruption. There were thus too many

false negatives: individuals who appeared committed because of this label but were not. Self-

selection proved more effective: the uncommitted tended not to join militias where discipline

would demand much of them. Militias differed widely in the degree of discipline imposed. In

some militia units, militiamen frequently argued about orders. In many, the men elected their

officers. These delegates then suffered from a lack of authority. They were under severe pressure

to reduce their militiamen’s duties in order to keep their posts, since “all delegates could be

removed as soon as they failed to reflect the wishes of the men who had elected them.” 57 Thus

militiamen often refused orders that they were given. 58 Instead, in some units, they held a vote to

decide whether to attack or retreat. 59 The lack of discipline eventually led Montseny, the

Anarchist leader, to complain in Solidaridad Obrera of how long it took to make any sort of

decision. 60 Militiamen who preferred light duty without much discipline, in order to take advantage of militia service for their own aims, sorted themselves into less-disciplined units, just as right-wingers joined those unions who were least demanding of political credentials. 61 For

example, they would often attempt to leave front-line units for militia units remaining close to

home. 62 Such an approach would allow them to gain the benefits of service—ten pesetas per day,

56 Seidman 1990, 94. 57 Bolloten 1991, 261–262. 58 García Oliver 1978, 266; Brusco 2003, 71–73. 59 Martínez Reverte 2004, 9–10. 60 Martínez Bande 1989, 71–72. 61 Cirre Jiménez 1937, 37–38. 62 Seidman 2002, 56. 101 perks of service, the opportunity to appear to be a hero, or political cover—without having to pay the costs of fighting, since they could frequently avoid doing so.

Anarchist propaganda did anarchist militias few favours in self-selection. Instead, light

discipline was practically advertised in Anarchist publications. At the beginning of the war,

Anarchist newspapers printed the requirements for militia service, indicating clearly that they

were rather lax—for example, explicitly noting that militiamen had the “liberty to enter and

leave as free men.” 63 The CNT in Valencia published a resolution in its mouthpiece, Fragua

Social , making it clear to new recruits that their units were different from others: “When a

comrade enters the CNT barracks, he must understand that the word barracks does not signify

subjection to odious military regulations consisting of salutes, parades, and other trivialities of

the kind, completely theatrical and negating every revolutionary ideal.” 64

Self-selection worked in the opposite fashion as well. Some individuals were convinced

of the need for military discipline, and actively joined those units that instituted it. The Quinto

Regimiento or Fifth Regiment demanded immediate and unquestioning compliance with orders from the very beginning of the war. Led by the Communists from a headquarters on Francos

Rodriguez Street in Madrid, the Fifth Regiment served principally as a focus for training and organizing militiamen of many political stripes on the central front.65 Many military officers joined the Communists and the Fifth Regiment. Those officers, “though far removed from

Communist ideology, were attracted to the party because of its moderate propaganda, superior

63 Comité de la Confederación Nacional de Trabajo de Cataluña, “La actitud de la organización obrera ante la llamada a filas de los reemplazos,” Solidaridad Obrera (Barcelona), 5 August 1936, pp. 1-2. This claim was reprinted verbatim, suggesting José Peirats’ authorship, in Peirats 1971, 1:186–187. 64 Bolloten 1991, 261, quoting Fragua Social , 18 November 1936. 65 Colodny 1958, 158. 102

discipline, and organization and because it alone seemed capable of building an army that could

carry the war through to victory.” 66

The Fifth Regiment’s reputation spread through word-of-mouth and through signals in its

own publications. Bolloten notes that José Martín Blázquez was one such officer, attracted by the

discipline of the Communist Party and noting how impressed his officer colleagues were:

It often happened that, when I came across a man who was just leaving for the

front, I asked him:

‘But why did you join the Communist party? You were never a Communist, were

you? You were always a Republican.’

‘I joined the Communists because they are disciplined and do their job better than

anybody else,’ was the answer. 67

The appeal extended to ordinary civilians. A teacher in Madrid, Leopoldo de Luís, stopped

giving classes to the children of militiamen and joined the Fifth Regiment with a group of

friends, “because they were the best organized battalions, those best able to face the fascist

rebellion.” 68 Domingo Malagón relates a similar story, joining the Fifth Regiment with a group of colleagues from the La Paloma art school:

I cannot say I was a communist at the time, although before the uprising I had done

fundraising for Socorro Rojo Internacional , but we thought that the best thing to do was

to go to the Fifth Regiment, since we saw that that group hadn’t disappeared in one or

two weeks. In fact, a few days after the rising, my teacher told me: “This will be a Civil

66 Bolloten 1991, 269. 67 Martín Blázquez 1939, 205, cited by Bolloten 1991, 269-270. 68 Martínez Reverte 2004, 65. 103

War and we will need to face them with an Army of our own.” I believe that, at the time,

he was one of the few who thought that way. 69

The Fifth Regiment actively attempted to exploit self-selection. Milicia Popular , its publication, discouraged individuals from joining up if they were unwilling to subject themselves to discipline. In laying out the requirements for joining the Fifth Regiment-affiliated Acero companies, Milicia Popular made an effort to make those requirements appear strenuous, including that militiamen “commit to submit to a rigid discipline.” It validated individuals’ contributions outside the militias in order to give those who sought to validate their revolutionary credentials incentives to select themselves out: “Not everyone can meet these obligations. One can be a good militiaman, a good revolutionary, a good anti-fascist, but still not have the preparation needed to meet these obligations.” 70 Again, on 20 September, Milicia Popular

dissuaded those who disagreed with discipline from joining: “If there is anyone who does not

agree [that indiscipline must be punished], it is better that they stay home.” 71

Over time, Milicia Popular was joined by Socialist and even Anarchist publications in its

stance of appealing to those who would accept discipline. At the end of September, El Socialista ,

in Madrid, juxtaposed a small number of committed forces and a large number of uncommitted

troops, asking for “five thousand hardened men…Five thousand decisive men will win the war;

twenty-five thousand men will lose it.” 72 Fragua Social , in Valencia, later published an article that, much like Milicia Popular over two months earlier, explicitly asked the uncommitted not to join and validated that decision:

69 Montoliú 1999, 39. 70 Milicia Popular (Madrid), 6 August 1936. 71 Milicia Popular (Madrid), 20 September 1936. 72 Martínez Bande 1976, 33, quoting El Socialista , 29 September 1936. 104

If someone wants to help bring about the triumph of the Revolution, he must first

examine what he would accomplish as a combatant, and if he does not find in himself the

degree of self-sacrifice necessary to be a good combatant, he must abstain from seeking

entry into the columns, and instead intensify his zeal and labour in the rearguard, where

he can also serve the Revolution, with more loyalty than those who march to the Front

knowing that they will not resist the first shot they hear fired.73

Ultimately, then, the discipline policies of different militia units had downstream effects

on recruitment. They enabled a self-selection process, allowing different militiamen with

different degrees of commitment and opportunism to sort themselves into units accordingly. The

different militias went as far as to explicitly indicate to new recruits the degree of discipline that

they could expect once serving.

When a combatant did not share the Republic’s aims, it could lead directly to desertion.

Some rightists joined up or, if they were already in the regular army, remained serving in order to get to the front to defect. José Cirre Jiménez, the right-wing artillery officer, stayed in his unit

and bided his time. His own unit was stuck behind the lines for much of the early fighting, but

when finally it joined the battle around Madrid, Cirre took advantage of the opportunity to

defect. 74 Cirre’s story was repeated elsewhere, as officers and soldiers of the regular army whose

sympathies lay with the Nationalists waited until their units were sent to the front to rebel and

defect. 75

But beyond this scenario, opportunism often produced desertion when conditions changed. Since many joined the militias out of opportunism, self-selecting into ill-disciplined

73 E. Gimeno Ortells, “Los Cobardes en la Vanguardia y los Derrotistas en la Retaguardia.” Fragua Social (Valencia), 29 October 1936, p. 13. 74 Cirre Jiménez 1937, 44, 80, 186–7. 75 Maldonado 2007, 64; Mainar Cabanes 1998, 19–20. 105 units in order to reap the benefits of service without paying the costs of fighting, waves of desertion often occurred when unit commanders tried to demand more of their troops. Day fighters deserted when they could not go home at night, for example when their units moved away from their towns of origin.76 Certain Anarchist units that had once advocated opposition to the state and the armed forces tried to impose discipline, and the result was often a wave of desertion. 77 In essence they self-selected out now. Ultimately, what this meant was that in any ill-

disciplined unit, where opportunists tended to join, there was a brutal tradeoff. Achieving serious

military aims required issuing orders to soldiers. To the extent that units were populated by

opportunists, such orders would be far more likely to interfere with the personal opportunities

that they sought from service. As long as unit leaders were content with remaining close to their

point of origin, letting militiamen sleep at home at night, or allowing them to loot, they could be

maintained with little desertion. Once more self-sacrifice was required of them, opportunists

deserted.

Thus the Republican militias varied substantially by their recruitment policies. They generally used union or political party membership cards as a method of screening, but these were cheap signals, easily falsified, and so little help. Better disciplined units, on the other hand, tended to prompt unreliable individuals to self-select out and to attract those willing to commit to fighting. They therefore enjoyed more reliable recruits. I argue in Chapter Four that these better- disciplined units tended also to have more substantial control, and so to be able to keep their combatants fighting even when the costs of doing so rose. Other militias, with less control, could not stop their fighters from deserting. Moreover, the militia with more reliable combatants could also enjoy more effective norms of cooperation, an argument I develop in Chapter Five.

76 Broué and Témime 1970, 174. 77 Fraser 1979, 133. 106

3.3. Voluntary Recruitment in Nationalist units

Just as on the Republican side, the Nationalist side was characterized by a rush of initial

enthusiasm. Such enthusiasm was particularly concentrated in certain communities. In Navarre,

the centre of Carlism, the rising was greeted with “scenes of religious enthusiasm, combined

with warlike zeal”; men poured into the provincial capital, Pamplona, from outlying villages

singing Carlist songs, hailed by the crowds. 78 Some 15 to 30% of men of military age volunteered for Nationalist militias in Navarre and the neighbouring regions of La Rioja and

Aragón. 79 One village, Artajona, “sent 775 out of 800 eligible males into battle.” 80

Egoistic motivations to join volunteer Nationalist units also emerged. However, since

Nationalist volunteer forces were generally subject to much stricter discipline, volunteers generally knew that to volunteer entailed fighting. It was possible for soldiers to fulfill their own egoistic aims, particularly through the rape, murder, and looting of leftists. But strict discipline meant that they had to be willing to fight. In this way, the Nationalists effectively enabled egoists to join without risking desertion.

Being in the army, by itself, provided some rewards. Where Republican militiamen were

valorized as heroes of the working class, Nationalists were valorized as heroes of the nation and

the faith. They could earn some social rewards: for example, women were encouraged to spurn

suitors who were not in uniform. 81 Wages were, it is true, considerably lower on the Nationalist

side than on the Republican. A typical soldier’s daily pay remained at the prewar 1.90-peseta

rate, with an additional 1.10-peseta campaign bonus—thus less than a third of the Republican

78 Thomas 1994, 239. 79 Seidman 2011, 26. 80 Blinkhorn 1975, 259. 81 Seidman 2011, 232–233. 107 militiaman’s wage. Army of Africa troops were paid somewhat more—5 pesetas per day. 82

Economic motivations mattered much more, however, among Moroccan troops. A drought in summer 1936 in Morocco led many to join the Regulares for a bonus of “clothing, four kilos of sugar, a can of oil, bread proportional to the number of their children, and two months’ pay in advance,” as well as generally more abundant food in the ranks. 83

Beyond the direct rewards offered by the Nationalist authorities for service were

opportunities to obtain the fruits of violence. Soldiers, Civil Guards, and Falange and Carlist

militias perpetrated a vicious campaign of repression.84 One could, as in the Republican zone, pursue one’s own personal vendettas, denouncing one’s opponents as “reds” on scanty evidence.

Joining a militia, then, gave one the chance to enact one’s own violent agenda, with official approval. Further, it gave leftists and the undecided crucial political cover.

Nationalist troops, especially Moroccans, were allowed to loot leftists’ property in just- captured Republican towns. They were also permitted to rape women identified as left-wing.

Indeed Nationalist leaders encouraged rape, perhaps most notoriously in the lurid radio broadcasts of General Gonzalo Queipo de Llano in Sevilla. Rape by the Legion carried on in an awful tradition of the abuse of Moroccan women in Spain’s colonial campaigns. 85 As in other civil wars, 86 rape was employed as a weapon, to inflict terror upon an opponent. Moroccan

troops generated a particular fear of sexual violence and pillage, evoking tropes of marauding

Moorish hordes. Republican morale often broke when there were rumours that the Army of

Africa had appeared. 87 However, consistent with their use as weapons of terror, rape and pillage

82 Corral 2007, 90. 83 Seidman 2011, 40–41. 84 Juliá 1999. 85 Balfour 2002, 181. 86 Wood 2006. 87 Richards 1998, 52; Seidman 2011, 40–41. 108

were subject to certain controls, directed against the left, not against the right. In general, looting

was not as common on the Nationalist side as on the Republican, and certainly just-conquered

towns endured far more looting than the Nationalist rearguard. Property owners were a core

Nationalist constituency. Hence, Moroccan troops could be sentenced to death for looting the

property of “respectable” citizens, and soldiers who raped women from right-wing families were

typically subject to severe punishment. 88

As on the Republican side, it was often not difficult to join a militia. While the

Republican side had nominal attachment to political screening in the form of union and party

membership cards, Nationalist militias often dispensed with such credentials altogether. On the

contrary, in the rapidly expanding Falange, leftists were encouraged to join in order to redeem

their left-wing past. Falangist documents circulated, stating that, in the words of Stanley Payne,

“voluntary enlistment for active duty was a clearer sign of loyalty than was ideological purity.”

According to the editor of the Falangist newspaper in Oviedo, fully twenty percent of the

Falange in Asturias were ex-leftists, equal to the proportion of Falangist true believers. 89

Dionisio Ridruejo, a Falangist organizer in Valladolid, was attracted to the party in part because he had been sympathetic to socialism but was too attached to religion to accept the left’s secularism. At first, many of the thousands who joined the Falange were right wing, but over time, Ridruejo and like-minded colleagues pushed for the entry of a greater number of leftists:

For the hostile there was a powerful initial argument for joining: the repression. In the

Falange a man had a chance: he was either accepted or shot.

88 Seidman 2011, 41, 47–48, 167, 228. 89 Payne 1961, 146. 109

Cases of left-wingers joining became so notorious that the right called the Falange

the FAIlange. Firstly because our flag was the same red and black; secondly because of

our pseudo-revolutionary demagogy, and lastly because we accepted everyone. 90

Left-wing recruitment into Carlist militias was not as pronounced as in the Falange, but the expansion of the Carlist militias still brought in those who were not fully committed to the cause.

Martin Blinkhorn considers it likely that only a minority of new Requeté recruits had been active

Carlists before the war. Carlists set up new branches where they had previously been inactive.

Many of the new recruits were thus “unavoidably deficient in their grasp of the essentials of

Traditionalism,” to use another word for Carlism.91 As for the Moroccan Regulares , political

screening was effectively irrelevant, since there was little connection in Morocco to the politics

of left and right. In any case, Moroccans were typically recruited on a decentralized basis by

their local notables rather than by an army agency.92

However, the greater discipline of the Nationalist side implied that those who joined

volunteer units would have to fight more often than some of their Republican counterparts. The

Legion was deeply ingrained with a sense of immediate obedience of orders, with failure to

comply punished with the whip. Beatings and whippings were also a standard practice within the

Regulares ; they were administered by Moroccans rather than by Spanish officers so as to avoid

one form of inter-ethnic resentment. 93 Requeté units were placed under military command, as,

frequently, were Falange militias. Indeed, the more effective Falange militias were directly

incorporated into the army. 94 However, it does appear as though Falange units, especially, were subject to looser controls in the first few months of war. As I note in the next chapter, Falangists

90 Fraser 1979, 315–316. 91 Blinkhorn 1975, 257–259. 92 Seidman 2011, 40–41. 93 Balfour 2002, 278; Seidman 2011, 43. 94 Payne 1961, 145; Blinkhorn 1975, 254–255; Ellwood 1987, 32. 110

were often able to return home at night, just as their Republican counterparts were. In the early

days the Falange militias were not subject to the code of military discipline, though Franco

subjected them to it in December 1936. Falange units, with somewhat lower discipline standards,

thus endured particularly strong self-selection problems. Falange militias attracted those who

wanted to avoid the military discipline that they would face if they joined the regular army.

Falange units gained a reputation for performing worse in battle, and were often employed in the

rearguard instead of the front line. In turn, this practice encouraged further self-selection of

individuals who wanted to avoid difficult duty. 95 Interestingly, just as on the Republican side, the more lenient practices of the Falange militias resulted in desertion once new norms were put into place. At the end of December 1936, three youths who returned home were denounced as deserters from the 5 th Bandera of the Falange. It was ruled, however, that they had not had a chance to learn and adapt to the new rules in place. 96

The forms of opportunism in voluntary recruitment on the Nationalist side apparently did

select for some who would later desert. Leftists concerned about the repression in the rearguard

would often join militias, especially the Falange. As I discuss in Chapter Four, their hopes for

cover were sometimes dashed: persecution of suspected leftists frequently extended to the front

lines, prompting them to desert. Moroccan soldiers, many of whom had joined for economic

reasons, deserted in a pattern consistent with this source of opportunism: in order to pursue

business opportunities in the rearguard. They sold food, clothing, and other items, much of which

they had looted. This indicates that the desertion of Moroccan troops was consistent with self-

95 Payne 1961, 145; Seidman 2011, 232. 96 Corral 2007, 101–102. 111 selection for economic incentives. Nationalist authorities tolerated this practice apparently because it helped reduce material privation in the rearguard. 97

However, greater discipline on the Nationalist side implied greater self-selection.

Someone who wanted to loot and rape could do so by joining a volunteer unit, but the best

opportunities for such depraved egoistic motivations existed in the shock troops of the Army of

Africa. To seek such personal, egoistic rents, a man had to be willing to accept that his unit

would fight in the most difficult battles under pain of wholly brutal discipline. Shirking from

combat would be extremely difficult. Hence there were rather more limited opportunities for true

opportunists, for those who wanted to extract the benefits of joining up but not pay its costs—

that is, individuals who, when finally called upon to fight, would refuse to do so. The advantage

enjoyed by the volunteer Nationalist units was not that they did not recruit egoists, but that their

egoism did not, to the same extent, prevent them from accepting that they had to fight. The

Nationalists proved more successful than the Republic at channeling egoism to military action.

3.4. Conscription and the Expansion of Motivations

Even if volunteer units were prone to opportunism, conscription had a much stronger effect in bringing in combatants who had little preference to fight. To enforce conscription, both sides used the same Code of Military Justice ( Código de Justicia Militar ), dating from 1890; it

was formally applied to the Republican militias in October 1936. Under the Code, someone who

failed to enlist once called up would be declared a desertor simple , and, if caught, penalized with

four additional years of military service on top of the required two. The Republic expanded these

penalties successively. On 18 June 1937 it ruled that draft evasion was equivalent to desertion to

the enemy, to be penalized with between six and twenty years in a work camp—but the guilty

97 Seidman 2011, 43. 112

individual would still have to complete his military service in a disciplinary combat battalion.

The following year the penalties increased again: on 8 April 1938, the government ordered all

men who had not joined up when called to present themselves within three days, or face the

death penalty as a traitor. Local authorities were required to denounce such individuals. 98

Both sides faced problems of individuals seeking to shirk their duties by obtaining exemptions, and both sides cracked down. Franco dealt with the problem of individual requests for exemptions swiftly, issuing an order on 1 November 1936 that any such request received from a draftee’s family or friends be destroyed unread. In addition, as noted, many on the

Nationalist side joined the Falange militias rather than be drafted into regular units. As part of the process of integrating these militias into the army, on 24 April 1937 Franco ordered that every militia member serving far from the front be incorporated into the army on pain of being considered a deserter. On the other side, unions could frequently obtain exemptions for their members, declaring their work vital to the war effort, for example. In late October 1937 this led

Indalecio Prieto, the Minister of Defense, to cancel all exemptions not ordered by his own department. 99

Many, of course, could not get such exemptions, and many evaded the order to join up.

Groups of draft dodgers, on both the Republican and Nationalist sides, would sometimes hide in the hills. Civilians could be prosecuted for offering them aid and shelter. Indeed civilians were generally caught in the middle; some groups of Catalan draft dodgers, for example, assassinated mayors for denouncing them. Both sides pursued serious efforts to track down draft dodgers, however. In August 1937, the Republican government set up a new organization, the Servicio de

Investigación Militar (SIM), tasked with counter-espionage behind the lines, but also with

98 Corral 2007, 98, 106–110. 99 Ibid., 121, 124–126. 113 finding draft dodgers and deserters who had hidden. Still, however, the Republic sometimes sent in army and security force units to eliminate pockets of draft dodgers who held out. This mirrored conditions on the Nationalist side, where bands of draft dodgers hid in the hills especially in Andalucía and in Asturias. The Nationalist authorities would send units of Falange and police after these bands; for example, in December 1937 in Asturias, a deadline was set for any draft dodgers to present themselves, after which public order forces would, according to the ultimatum, “open fire on everyone found hidden in the hills.” Suffering the privations of a life on the run, and fearful of reprisals should they be caught, many hidden draft dodgers turned themselves in, especially after the death penalty was instituted in April 1938. 100

With such coercive tactics pursued to intimidate individuals into serving and to roust out those who fled from service, conscription had a deleterious impact on the commitment of the troops fighting for both sides. César Lozas, a Republican sympathizer in Valladolid, on the

Nationalist side, had a chance at an exemption because he was of dual French-Spanish nationality. However, he joined anyway in the winter of 1936-37, out of fear of reprisals against his Republican father. “The new regime might not consider [his dual nationality] sufficient.

There was no way out but to serve. And that, he believed, was what the majority of peasant recruits in his infantry company – he was the only student – felt. They were not politically motivated to fight.” Another student serving in the Nationalist army on the Andalucía front found, similarly, that the peasantry “joined up with resignation – the sort of attitude that the great mass of people always shows in these situations.” 101 Republican commanders found serious fault with the new recruits. Santiago Álvarez, the political commissar of the 11 th Division, worried on the eve of that division’s assault on Teruel that the hundreds of new conscripts who had joined

100 Graham 2002, 344–345, 375; Corral 2007, 116–120. 101 Fraser 1979, 284. 114

were “mostly politically indifferent” and hence unlikely to fight well in battle. Another report, in

summer 1938, found that the new conscripts were “accustomed to a tranquil and placid prior life,

totally out of step with the current moment.” 102 Among such uncommitted combatants, control

became particularly important, for it was very difficult for norms of cooperation to emerge.

3.5. Factionalism in Republican and Nationalist Spain

In addition to problems of opportunism, the Republican war effort under the militias felt the legacy of factionalism quite strongly. The many different factions on the Republican side had a long history of conflict; up until 18 July, in fact, the Republic and the CNT had considered themselves enemies. Josep Cercos, the CNT-affiliated metalworker from Barcelona, expressed quite clearly that politically committed men could fight against Franco without any strong loyalty to the regime: “We didn’t give a damn about the republic, we were concerned only about the revolution. I wouldn’t have gone to the front if not to make the revolution.”103 This approach confronted the much more moderate preferences of the liberal Republican and Communist parties, preferring to focus on organizing the war effort and eschew the redistribution of property from the middle classes.

The multiple different interests and jockeying for power among these factions generated problems of coordination and cooperation among them. The problems of cooperation were apparently particularly strong among the militias on the Aragón front. In the village of Angüés, a bit behind the lines near Huesca, a CNT-affiliated peasant observed the anarchist Roja y Negra column on one side of the village, and the POUM column on the other: “When the former went into action, the latter sat back with their hands in their pockets, laughing. When the POUM was

102 Corral 2007, 157–158. 103 Fraser 1979, 120. 115 in combat the anarchists, I have to admit, did the same. That's no way to fight a war, let alone win it. They should have got together to fight the common enemy.” 104 Major Aberri, a

Republican officer sent to Aragón to assist in the organization of the front, reports a striking

example of the absurdity of this situation: “I was once in a position where there were several

10.5 guns, but there were no munitions. These were in the possession of a nearby column, which

refused to part with them although it had no artillery itself.” 105

Even more intense mutual mistrust emerged from combining civilian militias with regular

uniformed security personnel. The coup attempt, on top of years of conflict between the Left and

the army, now put military officers under deep suspicion. The risk was clear: many officers may

have remained on the government’s side not out of any preference but merely because, in their

garrison, the rising had failed. The ambiguity of the coup attempt, in large part, emerged from

the ambiguity about the loyalties of officers. Several incidents during and immediately after the

coup attempt confirm the difficulty of identifying loyal officers from rebels. Martín Blázquez

relates the sense of ambiguity within the armed forces in Madrid. Rightist officers would trick

some of their leftist subordinates by shouting “Long Live the Republic!” before taking advantage

of an opportunity to defect. 106 This occurred elsewhere as well. In the second day of the rising, a boy in Oviedo, the capital of Asturias where the upheaval of October 1934 had been centered,

“saw lorry-loads of civil guards approaching. They were giving the clenched fist salute and shouting ‘ ¡Viva la República! ’ … The guardia civil, only eighteen months after the October revolution, giving the clenched fist salute here, in Oviedo! ‘What a change has come over them,’ he heard a passer-by say.”107 However, this was apparently trickery. The garrison commander in

104 Ibid., 135. 105 Bolloten 1991, 257. 106 Martín Blázquez 1939, 111. 107 Fraser 1979, 69. 116

Oviedo, Antonio Aranda Mata, put up an appearance of loyalty long enough to ensure that the

militia columns departed the city, and then took it over with the help of most the same Civil

Guards. 108

Defections by Nationalist officers continued in the early days. Officers would take advantage of the movement of militias to the front to defect. 109 In a prominent incident, on 29

July, Civil Guard members of a column from Valencia rebelled, took over the column, and

defected to the Nationalists at Teruel.110 In the Eixea-Uribes Column in Aragón, of whose 3500 men some 950 were from the regular military, several were right-wing sympathizers, including a

Captain Frigola who was able to defect to the Nationalists. 111 The problem of right-wing sympathizers among uniformed armed services persisted for months. María Martínez Fernández, the wife of a Civil Guard, says that her husband’s unit had been planning on defecting entirely,

“with their lieutenant at their head,” if called to the front, but early on they were left in the rearguard. Eventually ordered up in September, he was able to defect during the Battle of

Brunete, southwest of Madrid, in July 1937. 112 It was clear to many, therefore, that members of

the army and paramilitary police had quite different interests from most militiamen. Many did

want to switch sides and fight for the Nationalist cause.

On the Republican side, then, factionalism was a powerful force with two main forms:

competition among political parties and unions, and especially the suspicion of the different

interests of regular army officers. In Chapter 4, as I turn to control, I examine how factionalism

could undermine control in the short run. In Chapter 5, discussing norms of cooperation, I

discuss the extreme difficulty of establishing such norms under conditions of factionalism.

108 Salas Larrazábal 1973, 111–113. 109 Alpert 1989, 61. 110 Maldonado 2007, 64; Mainar Cabanes 1998, 19–20. 111 Maldonado 2007, 66–67. 112 Montoliú 1999, 183. 117

In Chapter 5, I argue that in the early days of the militias, competition among political parties and unions could cause problems of desertion when militias from multiple factions attempted to work together, but since militias were generally organized separately by different factions, militiamen generally did not have to serve alongside individuals from a different political faction so the impact of this form of factionalism was attenuated. The suspicions of army officers had quite important effects on cooperation in the militia period as well, since army officers were attached to many militias. They were frequently under suspicion from their peers, making it extremely difficult for norms of cooperation to emerge between them and the militiamen in their units. Instead, mistrust prevailed, a mistrust that induced many officers who otherwise would have served to defect.

Political tensions had a much stronger impact on desertion in Republican Spain as centralization took hold, a theme I discuss in Chapter 4. The efforts to impose central control on militiamen were read as efforts by the Communist Party to dominate the other factions, so that control was effectively undermined, backfiring on the Republic’s central planners. And now that members of different parties and unions served side by side much more often, soldiers could be easily denounced for political gain. This both undermined control by rendering violence arbitrary and undermined norms of cooperation by fostering suspicions among soldiers.

One clear difference between the Nationalist and the Republican sides lay in the course of political competition. As on the Republican side, there were several political tendencies in

Franco’s zone, and a central authority attempting to centralize. None of the factions on the

Nationalist side had fought openly the way that Anarchists had opposed the Republic. No political disagreements pitted the necessities of the war effort against one grouping’s interests to nearly the degree that this occurred in the Republic’s debate between war and revolution. While

118

there was opposition to the principle of centralization on the part of the Carlists, their reluctance

to submit to central authority did not prevent them from seeing the necessity of discipline to

make war: their own militias, after all, required quite strict adherence to command. The

Nationalist side also did not have a clash over anything so fundamental as private property and

submission to the state, the core interests that animated the Anarchist-Communist conflict in the

Republic. Indeed, one of the appeals of Franco over his military rivals, Gens. Emilio Mola and

Gonzalo Queipo de Llano, was his lack of a defined political leaning; Mola and Queipo were

both perceived as too secularist and anti-monarchist to bridge the Right. Franco’s political

program evolved over time, skillfully combining an appeal to hierarchy and military strength (to

keep the various military chiefs together), restoration of official Catholicism (to appeal to the

Carlists), a corporate state (to appeal to the Falange) and a cordial but arm’s length relationship

between the military and an exiled monarchy (to bridge the gap between anti-monarchist officers

and those favouring restoration). 113 In general, therefore, the Nationalist side suffered from a much less intense degree of factional competition than its Republican adversary. I argue in the next chapter that this made control easier to accept: Carlist and Falangist combatants could generally be satisfied that increasing control would not leave them seriously vulnerable to domination by their adversaries.

3.6. Conclusion

This chapter has investigated how recruitment and factionalism affected the degree of common aims among combatants in units in the Spanish Civil War. Self-selection meant that generally, volunteers were more committed than conscripts. Not always, however: some opportunistic volunteers joined with the hope that they could enjoy the benefits of war but not

113 Preston 1993; Payne 1987. 119 pay its costs. In addition to recruitment policies, both sides of the Spanish Civil War brought together more or less disparate factions, with much more severe problems of factionalism on the

Republican than the Nationalist side. Factionalism, too, undermined the degree to which combatants shared common aims.

Common aims, in turn, affected the operation of both control and norms of cooperation in reducing desertion rates. When armed groups had a high degree of consensus, there was no guarantee that soldiers would remain fighting: war is a highly costly enterprise to the individual, and the collective action problem is powerful under these circumstances. However, consensus at least meant that norms of cooperation were possible. I turn to the operation of such norms, and how they depended upon common aims, in Chapter 5.

But beforehand, in the next chapter, I turn to the more general of the two solutions: control. I examine how coercive control is vital when there are disparities in common aims.

However, I also recognize that if mistrust is particularly severe, as in cases of factionalism, it can seriously undermine the operation of control. It can make control appear to just be a threat. The next two chapters, then, illustrate the power of common aims. Mediated by control and the development of norms of cooperation, common aims can have an important influence on desertion.

120

Chapter 4 Control and Desertion

4.1. Introduction

The clashing armies of the Spanish Civil War recruited men through both voluntary enlistment and obligatory service, and often included many different factions. Once serving, these combatants were subject to varying degrees of control, via coercion, to prevent them from deserting and to punish them for leaving. Of the two major approaches to preventing desertion, norms of cooperation required common aims, while control could be more useful when combatants were uncommitted. In the previous chapter, I analyzed various strategies to try to limit recruitment only to those who were committed, noting in particular that the most effective strategy was to rely on the asymmetry in information about preferences, letting men self-select out of particularly demanding voluntary units. Other approaches, such as the use of identity cards, were less effective because they were subject to much abuse to maintain asymmetric information. In any event the whole question of screening became to a large extent moot with conscription. But beyond recruitment policies, armed groups suffered from a lack of consensus about aims to the extent that they had multiple different factions. Enthusiastic fighters still had to pay quite severe personal costs to fight. Thus the armies of both sides managed men who might be inclined to desert for a variety of reasons. That is, they had the motive to leave. It was the role of control to limit the opportunity to do so. At the same time, however, severe disagreement about aims could undermine control itself.

121

In broad outlines, the war pitted a Republic with very weak control at first against a much more thoroughly controlled Nationalist camp. Militia forces fighting on the Republic’s side frequently did little to prevent men from leaving, and once in the rearguard there was little action taken against deserters. The Republic’s efforts to impose greater control were undermined by mistrust, since those efforts depended on monopolizing force in a way that was seen by many as factional and arbitrary. Efforts at control therefore provoked desertion initially. However, control also helped to overcome this factional conflict and in compel acquiescence. The Republic was able to control much more effectively over time, policing the rearguard and the front lines much more effectively starting in the winter of 1936-37. By the end of the war, however, its control mechanisms were not equal to the task that it faced.

From the outset, the Nationalist side was much better equipped to control desertion. Its security forces had not been shattered as the Republic’s had by the coup, nor was armed force really held in multiple, wholly autonomous hands. This control did mean mistrust of leftists to a great degree, provoking leftists to defect even if they had been willing to try to live under the

Nationalist regime. However, repressive as control frequently was under the Nationalist regime, it meant that prospective defectors often lacked any chance to switch sides. A much more attenuated degree of political competition and mistrust meant that control was not regarded as a severe threat to large swaths of the coalition fighting against the Republic, in contrast to the tensions in the Republic itself; those who did contemplate defecting in response to factional conflict were intimidated against doing so by the central coercive capability of the Nationalist leadership.

Since the Republic varied much more than the Nationalist side, I discuss the Republican side in much greater detail here. This chapter proceeds by examining variation among different

122

Republican militias amid the general condition of Republican disorder in the rearguard, before

turning to the changes wrought by militarization and the undermining of control in the Republic

at the end of the war. It then offers a brief contrast to the Nationalist side.

4.2. Control in the Republican Militias

As noted in the previous chapter, individuals joined the militias for a variety of reasons, many of them self-interested. In addition, there was often very little control to prevent militiamen from deserting. In the previous chapter, I noted that self-selection drove much of the difference in commitment across militia units: because men would submit themselves to stricter discipline in the latter two units, those units selected for militiamen who were willing to pay the costs of combat when serving. Militias also varied substantially in the desertion problems that they faced.

Many suffered from desbandadas or disbandments in the face of battle, others a steady trickle of militiamen deciding to quit the fight. A few, such as the Communist-organized Fifth Regiment, maintained greater cohesion in the face of the Nationalist advance. In this section, I outline how different approaches to discipline corresponded also with different approaches to putting punishments in place to prevent desertion. Since control in the rearguard was lacking, control in militia units themselves was particularly important in explaining how the Fifth Regiment, and other units such as the Anarchist Durruti Column, avoided desertion.

Lack of control began, like charity, at home. There were few efforts to track down those who had left the front and punish them for doing so. The Republican officer José Martín

Blázquez, working in the War Ministry, found that there were noncombatants with rifles all over

Madrid. 1 According to him, “The streets of Madrid swarmed with militiamen promenading with

1 Martín Blázquez 1939, 175–176. 123 rifles, but when we needed them only few presented themselves.” 2 Indeed, early on the CNT

declared its opposition to any effort to round up arms in the hands of militiamen at the

rearguard 3—an early indication of the way in which factionalism could undermine control. That many of these men with rifles were deserters becomes clear in the conflicts in some militia units over soldiers returning home with their weapons. Victor de Frutos, commander of the CNT

Primero de Mayo battalion, very reluctantly allowed his companies to issue permissions to return home at night. However, many men attempted to bring their arms home, and stopping them doing so was, for de Frutos, “the costliest fight of the first part of the war.”4 Such problems were encountered in other units as well. Narciso Julián served in the Del Barrio column organized by the PSUC ( Partit Socialista Unificat de Catalunya , the Catalan Socialist-Communist party). He relates that after an air raid, when many men wanted to leave and the column had to reorganize, he was assigned to ensure that the men did not leave with their rifles:

The heart went out of us all when we saw the numbers coming to hand in their names to

leave….Without exception, those who wanted to go refused to surrender their arms, even

when told that hundreds upon hundreds of peasants were waiting to use them. Finally, I

pointed at the two machine-guns. One of the men threatened to shoot me all the same.

“Go ahead, not one of you will be left alive.” Trueba, the political commissar, harangued

them and at last they handed in their weapons. Then they were put on two trains and sent

back to Barcelona. 5

These two commanders thus had serious difficulty in convincing their militiamen to surrender

their arms, even though their returning to the home front—permitting them to stop serving—was

2 Ibid., 128. 3 “Barricadas,” Solidaridad Obrera (Barcelona), 24 July 1936. 4 de Frutos 1967, 28–29. 5 Fraser 1979, 134. 124

a relatively organized process. In a mass flight, it would have been all the more difficult to

prevent soldiers deserting from keeping their weapons. Thus the frequency with which rifles

were found in the rearguard, and the authorities’ inability or unwillingness to corral those rifles,

indicated a lack of control in the rearguard. Individuals who left militia units, even with rifles,

faced few sanctions in the rearguard from doing so.

Complementing the general condition of a lack of control in the rearguard, some militia

units placed few restrictions on their members at the front: they could challenge or dispense with

orders. There was, correspondingly, little control put in place to prevent desertion. A typical

Anarchist view was that militiamen needed to be free to come and go, even if, while serving,

they had to accept some discipline if only to ensure that the unit operated together effectively.

The CNT committee in Catalonia argued that the regular conscripts of the Spanish military

wanted to return to their barracks to fight for the Republic, “but only when they enter into the

barracks as militiamen and have the liberty to enter and leave as free men, who voluntarily

accept the discipline necessary for joint action, but not as automata stripped of human

personality.”6

Freedom to come and go also often involved allowing soldiers to return home at night, a widespread practice. In Madrid, the militias were often day fighters:

In the morning there’d be shouts. ‘Pablo! Pedro! Manolo!’ and the men came out of their

houses with their rifles in their hands. Under the other arm they had the lunches their

wives had prepared. They set off for the sierra as though they were going on a Sunday

6 Comité de la Confederación Nacional de Trabajo de Cataluña, “La actitud de la organización obrera ante la llamada a filas de los reemplazos,” Solidaridad Obrera (Barcelona), 5 August 1936, pp. 1-2. This claim was reprinted verbatim, suggesting José Peirats’ authorship, in Peirats 1971, 1:186–187. 125

outing, to shoot rabbits….How amazing it seemed when in the evening they all came to

spend the night at home. The next morning the scene was repeated… 7

According to Broué and Témime, to undermine these rights was to risk mass desertion: “a

column that strayed from its home base lost most of its militiamen: they liked to sleep home at

night.” 8

In part, the rejection of military discipline was out of philosophical opposition.

Antimilitarism had been fundamental to prewar Spanish anarchism. 9 Indeed, in a congress of the

CNT on 1 May 1936 at Zaragoza, “the proposal to create a libertarian militia to crush a military uprising was rejected almost scornfully, in the name of a traditional anti-militarism. Instead, much time was devoted to outlining what life would be like under libertarian communism.” 10

The question of anti-militarism is difficult to dissociate from anarchist and anarcho-syndicalist philosophy: the opposition between the liberty of the individual and the state’s efforts to make the people fight and die for it, and inflict brutal punishment if they refused. The Anarchist historian José Peirats claimed, for example, that “military discipline was more a humiliation to human dignity than an effective tactic.” 11

Anarchists regarded military discipline as a way in which the process of war threatened to

undermine the very goals of that war. Many Anarchists desired revolution as a primary aim;

making war against fascism was a subsidiary goal if goal it was. Josep Cercos, the CNT

metalworker from Barcelona, expressed this view clearly:

We had come out to fight in the streets of Barcelona because we had no option; but to go

and fight the military in Aragon – no, I wouldn’t have gone simply for that. It was the

7 Fraser 1979, 117. 8 Broué and Témime 1970, 174. 9 Alpert 1989, 8. 10 Fraser 1979, 101. 11 Peirats 1974, 157. See Peirats 1974, 155-175, for a general Anarchist account of Spanish military history. 126

fever of the revolution which carried us forward. We had preached anti-militarism for so

long, we were so fundamentally anti-militaristic, that we wouldn’t have gone simply to

wage war. That was something we couldn’t envisage. 12

Thus warmaking was subordinate to and at the service of a fundamental reshaping of society, to liberate the working class. Much of the Anarchist activity in the civil war was directed to creating agricultural and factory communes and in redistributing property locally. From an anarchist perspective, then, there was an extremely difficult tension: what if fighting the war required severe infringements on their liberties? They had long opposed the Republic, regarding it as a mere bourgeois democracy, not a truly liberated regime. Now the Republic seemed to be demanding their submission in order to defend against an even more repressive system. Indeed, the language of Republican leaders gave anarchists little reason for greater optimism than this.

Manuel Azaña, President of the Republic and a “bourgeois” Republican, argued in his memoirs:

The strongest threat was no doubt the military rising, but that rising’s principal strength

came, for the moment, from the fact that the lawless masses left the Government unarmed

to face the enemies of the Republic. Reducing those masses to discipline, making them

enter the State’s military organization, with commanders dependent upon the

Government, in order to make war according to the plans of a General Staff, was the

capital problem of the Republic. 13

In the face of the demands of war, many Anarchists maintained their philosophical antimilitarism. The argument that the civil war required a disciplined army was, according to

Peirats, “impressive,” but ultimately a deception, “aimed towards one end that could not be

12 Fraser 1979, 120. 13 Azaña 1967, III:487. 127 waived: the disarming of the people” and thus their vulnerability to domination by the State. 14

Others, such as the CNT journalist Eduardo de Guzmán, were more willing to accept military discipline:

This was not a revolutionary riot. This was a war, with all the pains and demands of a

war. To win, neither enthusiasm, nor faith, nor heroism was enough. To win, it was

necessary to organize. To win, it was necessary to act without vacillation or dismay, with

serenity and invincible energy...We are antimilitarists; we continue to be so. But today,

against the dramatic realities of a war which we can do nothing but accept, we must adopt

warlike methods. If it is necessary, we must bury our ideas to defend them heroically. 15

However, it was still unclear just what that would mean: “We are not asking for a barracks discipline. But we are demanding a minimum of priceless responsibility, according to the needs of the war.” 16

I explore, below, the effects of the mistrust that Azaña and Peirats’ words indicated, undermining the process of centralization. In the meantime, the tension within anarchism between anti-militarism and the needs of the war affected desertion by prompting different approaches to control within anarchist militias themselves. Many Anarchist units allowed soldiers to leave, at least formally, but some made it much more difficult than others and hence suffered fewer departures. The Torres-Benedito column, organized by the CNT in Valencia and serving in Aragon, prided itself on its lack of hierarchy. According to a profile in the Anarchist newspaper Fragua Social , “There are not chiefs properly speaking, but comrades with different tasks. For example, Mirasol and Marcelino Pérez, heads of a division of several centuries, have, instead of a hierarchy imposed on a mass of men, the confidence and esteem of all.” This helped

14 Peirats 1974, 159. 15 de Guzmán 2004, 83. 16 Ibid., 84. 128

make Torres-Benedito , in the words of the propaganda, “a model of discipline and a guarantee of

effectiveness.” 17 Despite the glowing image, however, the column suffered serious indiscipline

problems, and men continually returned from the front to the rearguard at night. 18

In contrast, the Anarchist leader Buenaventura Durruti attempted to impose order within

his unit. He insisted that militia delegates be obeyed, and took measures to ensure that each

militiaman was pulling his weight, for example investigating whether a soldier claiming illness

was really sick or just malingering, with the latter given extra work. Finally, though following

standard Anarchist practice of permitting militiamen to leave the unit, he made this difficult and

humiliating: for example, a soldier who wanted to return home (and home was potentially as far

as Barcelona, a few hundred kilometres away) would have to go on foot. Durruti also forcefully

detained men attempting to leave the front with arms. Apparently, few soldiers deserted from the

Durruti column under such circumstances. 19 For the Durruti column, then, taking a firm stance against exiting the unit was part and parcel of a broader departure from typical Anarchist practices of free movement.

As time went on, other anarchist leaders increasingly turned to military discipline. The

Anarchist leader Cipriano Mera’s memoirs indicate, explicitly, a change in outlook. As a political leader of the CNT del Rosal column, he relied upon persuasion to convince his militiamen of the need to keep fighting. In a speech to the men on 1 August 1936, he referred to

“the necessity of imposing on ourselves a self-discipline stronger than military discipline,”20

However, as time went on, Mera became increasingly convinced of the need for discipline that went beyond an appeal to personal convictions. After a shelling that killed some of his friends,

17 “De nuestro enviado especial,” Fragua Social , 31 October 1936, p. 11. 18 Maldonado 2007, 65. 19 Francisco Oliva, “La Gesta del Proletariado,” Solidaridad Obrera (Barcelona), 2 August 1936; Bolloten 1991, 263; Corral 2007, 99–100; Oña Fernandez 2004, 63. 20 Mera 1976, 28. 129

Mera reflected on the problems of the militias in general, and resolved on the need for command and for planning. He believed his militiamen were not following self-discipline, but were instead

“making use of an improvised liberty.” On 19 August, several men presented themselves to Mera and requested that they be able to return to Madrid to change clothes and spend some time in more comfortable surroundings, and Mera felt himself obliged to threaten them with force to make them return to sentinel duty. Another CNT leader, Teodoro Mora, faced much the same problem. On 20 August, Mera gave a speech to his assembled militiamen to inform them that, once at the front, they would not be permitted to leave. 21

On 3 October 1936 the CNT Defense Committee in Madrid published Reglamento de las

Milicias Confederales , a set of rules for their militiamen. They were enjoined to “obey the rules

of the Battalion Committee, Century Delegate, or Group Delegate”; a militiamen could not “act

on his own account in war matters, and shall accept, without discussion, the posts and positions

he is assigned, both in the front and the rearguard.” It defined certain acts as “grave breaches” of

the rules, to be punished by the Battalion Committee: these included “desertion, abandoning of

one’s post, pillage, and speaking demoralizing words.” Ultimately the tension between anarchist

liberty and the needs of the war was resolved as follows: “Every militiaman must know that he

has voluntarily joined in the militias, but, having joined, as a soldier of the revolution, must

accept and comply...Militiaman! These rules of action and conduct are not barracks discipline.

This is a force of all, jointly, united and disciplined. Without such cohesion of energy, no

triumph is possible.” 22

However, many others within the Anarchist movement resisted this new approach.

Writers in Solidaridad Obrera asserted a clear preference for a lack of discipline, with one

21 Ibid., 28–35. 22 de Guzmán 2004, 117–118; Bolloten 1991, 264. 130

author juxtaposing “absurd and antiquated discipline” with “true camaraderie.” 23 Others, especially Anarchist hardliners in the Columna de Hierro or Iron Column, insisted that military life remained the exploitation of the many by the few: “Barracks and prisons are the same thing…Who can claim that fighters, once they are militarized, are stronger, more willing to fill battlefields with their blood?” 24

In contrast to these Anarchist units, the Communist-organized Fifth Regiment insisted

upon discipline, including strong controls to prevent desertion, from very early on. As I argued in

the previous chapter, the Fifth Regiment’s discipline policies enabled it to enjoy the fruits of self-

selection, apparently able to avoid most of the opportunism that plagued other units. However,

fighting war was still, of course, an extremely costly and difficult enterprise, and militiamen in

the Fifth Regiment felt those costs the same as anyone else. 25 They could thus be tempted to

desert, even if they had joined up out of a sense of the cause rather than out of opportunism. The

collective action problem still exists among people with an attachment to a shared goal. As such,

policies of control remained important. Via its central publication, Milicia Popular , the leadership of the Fifth Regiment indicated a preference for control that contrasted strongly from

Anarchist views. In an early issue it published the Promise of the Popular Militiaman:

I, child of the people, citizen of the Spanish Republic, freely take the status of Militiaman

of the People’s Army....I commit to keep, and to ensure that others keep [ guardar y hacer

guardar ], the most rigid discipline, precisely obeying all the orders of my chiefs and

hierarchical superiors....If I voluntarily fail in these solemn commitments, may the

23 G. Séguero, “Impresiones de un viaje al frente de Bujaraloz,” Solidaridad Obrera (Barcelona) 28 August 1936. See also Alfonso Martínez Rizo, “De Barcelona a Zaragoza: una asemblea de milicianos,” Solidaridad Obrera (Barcelona) 8 August 1936. 24 Anonymous 1997, 14, 19, 21. 25 Fraser 1979, 258. 131

disrespect of my comrades fall upon me, and may I be punished by the implacable hand

of the law. 26

An internal Fifth Regiment document, entitled “Camaraderie and Discipline,” noted that there were militiamen who demonstrated insufficient discipline: they “advance and retreat as they like, mock their superiors, sow confusion and indiscipline. With such militiamen, it is necessary to speak seriously, and if they persist, they must be expelled from the militias.” While insisting on obedience, however, it also required commanders to exercise restraint in how they treated their subordinates. It noted that some commanders thought that they had “many rights and few duties”; a chief should instead “never forget that his militiamen are his companions, volunteers like him; and if he is chief, it is because of them.” Hence, the responsibilities of commander and subordinate were reciprocal. “A chief who treats his militiamen like automata is a bad chief; a militiaman who does not obey his chief is a bad militiaman. Neither has the right to be in the militias.” 27 The Fifth Regiment was thus able to maintain cohesion and fighting effectiveness on

its fronts in a manner that other militias were not able to.

4.3. The Evolution of Control in the Republic

Over the winter of 1936 and the spring of 1937, the Spanish Republic arrogated to itself the means of control in a thoroughgoing fashion. In this section, I argue that active mistrust from factionalism provoked desertion as control expanded, since some militiamen—especially

Anarchists and POUM members—saw control as persecution. However, in the medium term the imposition of control did help to reduce desertion. The ability of the new Popular Army of the

26 Milicia Popular 4 August 1936. 27 “Camaradería y Disciplina,” Fifth Regiment Historical Document no. 1, p. 14, reprinted in Vidali 1975, 94–96. 132

Republic to maintain control was a vitally important factor in its efforts to prevent desertion. By

the end of the war, however, its control began to unravel.

Militarization entailed a clarification and regularization of control. As noted, the

Republic maintained little control over the rearguard after the coup attempt stripped it of much of

its security forces and left the remainder in organizational disarray, allowing men with rifles to

pervade villages and towns. There was no central, consistent capability for tracking down

deserters. This changed over time, as the central government steadily asserted its control,

disarming the local militias and individual workers.

The central government declared that responsibility for rearguard security, including the

tracking down and punishing of deserters, was now the province of the official security forces

rather than ad-hoc militia committees. The militias themselves were subject (at least formally) to

the 1890 Code of Military Justice as of October 1936. According to this code, a deserter to the

rearguard could be punished with four years in prison, with an additional two years given for

deserting with arms. Defection constituted treason, subject to the death penalty. 28 Military courts were established to enforce these penalties, with, for example, the Auditoría de Guerra de Gijón set up in mid-December 1936 to apply the code in Asturias. Over half of its cases were, eventually, for desertion. 29 Over time, penalties for desertion increased. As of 18 June 1937, desertion was punished with a sentence of at least 12 years in a work camp, and the death penalty could now be given even for desertion to the rearguard. 30

The process of implementing these new control measures, however, was undermined by mistrust. Where, as in Madrid, the Communists were popular and the war effort generally cooperative, rearguard control was relatively easy to enact. In other areas, such as Catalonia, the

28 Corral 2007, 285. 29 Almendral Parra, Flores Velasco, and Valle Sánchez 1990, 191; Murillo Pérez 1990, 206. 30 Corral 2007, 293. 133

CNT and the POUM were much stronger, and many of their members were deeply suspicious of these efforts at control as a Communist plot. Therefore, the implementation of control, both in the rearguard and in the militias, provoked desertion in the latter but not nearly to the same extent in the former.

Madrid was notable in its implementation of rearguard control. The private use of violence was reduced first: by late August 1936, according to Franz Borkenau, “workers with rifles, but in their ordinary civilian clothes, [were] quite exceptional” in Madrid. 31 The Madrid

Defence Council, set up in early November 1936 to coordinate the defense of the city against the

Nationalist siege while the Republican government transferred to Valencia, imposed further control. José Cazorla, the Public Order councilor after December 1936, consolidated the security services and reduced the frequency of violence by autonomous patrols to an incidental level by early 1937.32 However, Cazorla still noted in a council meeting of 15 April 1937, “today it is

easier to be a thief than at any other time, because we do not have an organized State.” 33 Control

over the rearguard assisted the defense of Madrid against the siege by closing off the spaces in

which deserters could live in the rearguard. Eventually, when a man walked into a café swinging

a rifle around, he was greeted by demands that he get to the front. 34

Just as rearguard control was put into place relatively early in Madrid, central control of

the militias was pursued relatively readily in that city and especially among the locally quite

influential Communists. I noted earlier the much greater willingness of the Fifth Regiment to

institute discipline among its forces, a tendency that led rather naturally to militarization.

31 Borkenau 1963, 123. 32 Graham 2002, 193–194. This is not to say that rearguard violence was over. Far from it: early November 1936, when the Madrid Defence Council was up and running, saw the massacre by elements of the security services of defenseless prisoners, including many rightists, from the Model Prison. 33 “Acta de la Sesión Celebrada el Día Quince de Abril de Mil Novecientos Treinta y Siete,” reprinted in Aróstegui and Martínez 1984, 446. 34 Martín Blázquez 1939, 175–176. 134

Militarization had long been advocated by Communist Party leaders, many leaders of the first

Mixed Brigades were Communists, the new army had many Soviet advisors who were well-

known to favour the Communist Party, and the Brigades themselves used a system of political

commissars inherited from the Fifth Regiment.

In other regions, notably Catalonia, it was much more difficult to impose centralized

order. Local CNT committees were much stronger in Catalonia, and resisted efforts to control the

rearguard. Thus, for example, on 27 October 1936 the Republican government ordered that long

guns such as rifles and machine guns be handed over to municipal authorities, which were

themselves ordered reconstituted on 9 October to replace local committees. In Catalonia,

however, this order provoked skirmishes between the paramilitary police and resistant local

patrols, notably along the French border. 35 The clashes over control of the rearguard and

militarization of the militias came to a head with running battles in Barcelona on 2-5 May 1937.

After the defeat of rebellious CNT and POUM forces in Catalonia, control was finally enacted.

The CNT and the POUM felt that moves toward more regular control of the rearguard

were part and parcel of the liberal Republic’s bourgeois counter-revolution, in alliance with

Stalin’s pressure to rein in revolution in order to appeal to Britain and France to form an anti-

Fascist alliance. In this context, replacing local committees as the wielders of force would be a

tool for rolling back the communal farms and factories and local redistribution of wealth that the

CNT had enacted in Catalonia and Aragón. 36 And indeed the Communists had appealed for

caution and restraint, pitching themselves as the best guarantors of middle-class property-

holders. 37 The militarization of the militias, as well, was regarded as an effort to pursue this counterrevolutionary agenda. The Anarchist Diego Abad de Santillán spoke for many when he

35 Graham 2002, chap. 5, passim. 36 Ibid., 254–261. 37 Bolloten 1991, 83–84. 135 said that militarization “was not due to considerations of military order, but to the political calculus of counterrevolution.”38 Therefore, members of the key political forces that believed

themselves most targeted by the PCE and by the Soviets—the CNT and the POUM—often

believed that by fighting in a militarized army, they were fighting against their own cause.

Over the course of the spring of 1937, as the Republic’s efforts at centralization and

control continued, they provoked immense tension and often desertion. The ability to control the

behaviour of soldiers was regarded by many as the ability to wield violence for particular

repressive aims rather than for the real control of the war effort. Josep Costa, a CNT textile

leader, recalled with bitterness: “little by little we were being reduced to mere spectators of our

own slaughter.” 39 The Anarchist Iron Column experienced a wave of over 400 desertions of men who preferred not to militarize, in March of 1937. Many of these may have been among the prisoners that the unit had recruited, who for obvious reasons self-selected against military discipline, but it is also the case that the Iron Column included many Anarchist hardliners who saw service under Republican command as furthering a counterrevolution. Thus it is telling that many of the deserters from the Iron Column later took up arms in Barcelona two months later. 40

The Durruti Column was similar even though it was better-disciplined than the Iron Column.

Though it in general maintained cohesion when it was militarized, it still had men who left because they refused to fight for what they saw as a Communist-dominated government. These

“Friends of Durruti” advocated armed resistance in Barcelona. 41

Many more considered desertion than actually left. The attitude of the CNT leadership was quite important in this regard. The CNT leadership ordered its militants to remain fighting

38 Abad de Santillán 1975, 195. 39 Fraser 1979, 380. 40 Bolloten 1991, 338; Casanova 2005, 125. 41 Graham 2002, 274–275. 136

rather than deserting, out of a combination of concern for the common cause of defeating Franco

and out of fear of the state. Costa remembers: “the CNT wasn’t prepared to order troops to leave

the front, for that would have let the enemy through.” 42 Helen Graham’s analysis concurs, but also cites intimidation by state power:

the Republican government was already poised to intervene: had it been faced with an

all-out CNT challenge, it would surely have drafted in far greater numbers of troops and

police to take on ‘revolutionary Barcelona.’ Otherwise it could not have guaranteed the

Aragon front or retained liberal state control over Catalonia's war industries...The

Republic itself might well not have survived such a massive escalation of armed

internecine conflict, but, either way, the CNT would certainly have gone down in the

blood bath. 43

The leadership’s orders were important in the decisions of some not to desert. Ricardo

Sanz, commander of the Durruti column after Durruti’s death, had 500 men heading to the

Aragón front, where he was ready to link up with other CNT militants and return to Barcelona to fight. But he was ordered to stay at the front by Juan García Oliver, a key national leader in the

CNT and in fact the Republic’s minister of justice as of late 1936. Sanz’s troops obeyed. Sanz recalled later, “My personal feelings didn’t matter; I was a disciplined man, a military commander…” 44

Beyond the leadership, others who contemplated desertion were also ultimately intimidated by the state and its coercive capabilities. In the Iron Column, those who did not desert evoked the new power of the state to explain why they acquiesced to militarization. They regarded their ultimate conscription as now inevitable, and chose to have the unit be militarized

42 Fraser 1979, 380. 43 Graham 2002, 272–273. 44 Fraser 1979, 380. 137 so that it could remain together, as a coherent entity. 45 Perhaps suspicious of the influence of

other political forces, the CNT leadership, which accepted militarization, still insisted that its

units remain, in the majority, Anarchist, a move that the government accepted lest it provoke

further rebellions against militarization. 46 Despite this reluctance, Abad argues that the

leadership’s overall acceptance of militarization was folly: “They approved their own suicide!” 47

Ultimately, however, the increasing control power of the state caused the CNT senior leadership to acquiesce in the supplanting of its local committees and to restrain its troops from deserting.

While increasing central control provoked some to desert out of mistrust, over the long run, it succeeded in compelling many others to continue fighting.

However, where mistrust was more severe, the problem of control provoking desertion

rather than quelling it was much greater. This was the case with the POUM. The POUM

experienced even greater persecution than the CNT, persecution that indicated to its members

rather clearly that they had no political future in a Republic under heavy Communist influence.

Though Marxist, the POUM broke with Moscow and with the PCE over the revolution in

Catalonia and Aragón. Nor did it help its relationship with Moscow that its leader Andreu Nin

had been Leon Trotsky’s secretary, though the two were now estranged. 48 The POUM’s

leadership had been more willing to rebel than the CNT’s in Barcelona in May 1937, but the

CNT’s reluctance and the recognition of the likelihood of defeat convinced its leaders to stand

down. The POUM leadership ordered its men to stay at the front during the May Days, and the

POUM column itself had accepted the militarization order, becoming part of the 29 th Division. 49

However, the impression of rebellion lingered, and on 16 June, the POUM’s leaders were

45 Bolloten 1991, 337–342. 46 Ibid., 330–331. 47 Abad de Santillán 1975, 208–209. 48 Graham 2002, 235. 49 Ibid., 274; Casanova 1985, 113–114. 138

arrested. Subsequently Nin was kidnapped and murdered, accused of being a fascist spy; the

circumstances surrounding his death are still shrouded in mystery. 50 POUM militants could not believe the accusation, still less given the appearance of a Communist hidden hand. George

Orwell, who had fought with a POUM unit, vividly evokes the sense of disillusionment with the whole Republican cause that the anti-POUM putsch provoked. 51 Ultimately, the POUM column’s officers were imprisoned, most of its men deserted and the unit was dissolved. It was the most extreme example of the desertion of committed anti-fascists that emerged from the heterogeneity of interests and the twisting of control towards factional interests.

After this wave of desertions that responded to the implementation of control, the

Republic’s new measures allowed it a much greater capacity to coerce soldiers into remaining to fight. Under the newly constituted security apparatus, capturing and punishing deserters in the rearguard was the task of the public order forces such as the greatly expanded National

Republican Guard (the reconstituted Civil Guard) and Assault Guard, as well as the shadowy new Military Investigation Service ( Servicio de Investigación Militar , or SIM). SIM, established on 9 August 1937 “to combat espionage, prevent acts of sabotage and carry out duties of investigation and vigilance within the armed forces,” gradually took it upon itself to extend its writ to the rearguard. 52 And this included finding and punishing deserters and draft dodgers, who would serve their required sentences in SIM-run work camps. 53

Government attempts at controlling the rearguard to prevent desertion wrapped soldiers’

family members in the climate of coercion. Family members would of course be questioned as

the first step in a typical investigation of a deserter. But they were also made to bear the

50 Graham 2002, 284–292. 51 Orwell 1989. 52 Bolloten 1991, 600. 53 Graham 2002, 375–377. 139 punishment for desertion, and especially for switching sides. The government issued a new order to this effect at the beginning of June 1938. Noting, with sinister overtones, an “excessive generosity of Republican sentiment,” it ordered military units and recruitment centres to gather data on soldiers’ close relatives: mother, father, brothers, sisters, and wife. The data were to include their names, ages, and place of residence. When a soldier defected, that information would be sent on to the SIM. One male member of the family would then be made to take the defector’s place in the line. Other males would serve in fortification, communication and other auxiliary services. The place of women in the Republic’s conception of combat motivation was made clear: female members of a defector’s family would be detained until they could attest, with sufficient proof and testimony from parties or unions, that they had done all they could to convince the deserter not to leave.54

All in all, the Republic now made a much more concerted attempt to find and punish

deserters on the home front than had ever been implemented. But it was clearly of variable

effectiveness. Deserters were often difficult to find. Official provincial bulletins frequently

printed orders to particular deserters to present themselves, and according to Pedro Corral, in

1938 these constituted some 90% of announcements in some provinces. Interrogated by

Francoist agents about whether they feared reprisals against family members, many defectors

“averred calmly that there were many who deserted without anything happening to their

families.” Rough terrain made it particularly easy to hide and resist government capture.

Thousands of deserters hid in the hills of La Mancha, requiring public order forces to conduct

operations against them in the spring and summer of 1938. 55 Soldiers from hill country possessed the local knowledge to be able to evade government capture for long periods of time, and were

54 Corral 2007, 334–336. 55 Ibid., 119, 314, 338. 140

thus better able than others to desert. Indeed, in Chapter 6 I present systematic evidence that

soldiers from mountainous regions were more likely to desert than lowlanders, and that they and

their families took specific advantage of the terrain to evade punishment. By pointing out

variations in desertion that follow on variations in control, this reinforces the general theoretical

argument of control’s importance.

Control, however, was also exercised at the front. Men attempting to cross the lines

would typically be warned to stop, and if they did not, they would be shot. Executing men in the

act of desertion also applied to desertion to the rearguard: as of 21 August 1937, according to

documents seized by the Nationalist army, the Republican Army of the East had set up machine

guns behind the lines. In some units—such as the central front’s 27 th and 138 th Mixed Brigades in the period October 1938 through February 1939—more men were killed trying to desert than by enemy fire. 56 Vigilance in preventing desertion was the province, initially, of the political commissar. In the militia period, the Fifth Regiment had imported the Red Army’s political commissar system into its units, to maintain the motivation of the men to fight fascism and to monitor the political sympathies of the officers and men. They were implemented throughout the armed forces upon militarization from the company level up, until, in October 1937, the ministry of war restricted political commissars to the brigade level and above. Desertion in a commissar’s unit would be held against that commissar, as a testimony to the poverty of his political work. At different points, officers and commissars were ordered not to sleep at night but to patrol the trenches in order to prevent desertion. 57 Later, the SIM took over many of these functions.

Under the risk of being shot trying to defect, typically a defector would try to separate himself from his fellows one way or another. Common pretexts in the First Army Corps were

56 Ibid., 54, 295. 57 Graham 2002, 331–332; Seidman 2002, 157–159; Alpert 1989, 184–189. 141 going to gather firewood or to answer a call of nature. Pedro Talón Bollver managed to cross the lines on 6 August 1938 when he was separated by a crag from his partner on guard duty. 58

Defectors needed to be particularly wary of their non-commissioned officers such as cabos

(corporals). Defection reports frequently indicate that soldiers left when an NCO was arranging for sentinel posts to be relieved. Much, in fact, depended on the reliability of those NCOs: soldiers whose NCOs defected gained a huge opportunity to defect as well, and there are occasional reports of a whole guard post—for instance, one cabo and two soldiers—crossing the lines together. 59

Thus soldiers’ chances at desertion depended critically on how tightly they were monitored. If they were under continual surveillance, getting away would be extremely difficult.

It was, however, difficult to maintain constant vigilance at all times over all soldiers, and so units tried to prioritize. Men who gave particular grounds for suspicion would thus come in for tight surveillance, for example assigned guard duty always with more reliable soldiers. If a defector managed to escape without being shot, questions would arise: had he been under enhanced vigilance? If so, how was it broken? If not, why not? For example, the 26 th Mixed Brigade had

standardized forms for investigating desertions, asking about the deserter’s previous behaviour in

the unit and what was known about his political history and sentiments. It also asked whether the

man in question had been placed under tight surveillance beforehand. Such a report of two men

from the 101 st Battalion, 26 th Mixed Brigade, indicated that both had been under close

surveillance for their demoralizing comments; one had apparently threatened a corporal with

death. But they took advantage of the absence of their section’s sergeant to escape. 60

58 “Desertores,” 1 Cuerpo, 15 October 1938, folio 30. Centro Documental de la Memoria Histórica [CDMH], Salamanca, Serie Militar [SM], caja 421, folios 29-31. 59 “Desertores,” 1 Cuerpo, 10 February 1939, folio 77. CDMH, SM, caja 421, folios 74-81. 60 “Filiación e informe”, 26 Brigada Mixta, 24 September 1938. CDMH, SM, caja 607, folios 3-4. 142

Soldiers under particularly tight vigilance were apparently more likely to be shot trying to

escape than others. Reports of shot defectors from the 1 st Army Corps on the Central Front often

indicate that the men were under particularly tight surveillance because of their past behaviour or

political history. Such suspicions allowed the men tasked with surveillance to be wary of cover

stories and keep alert. For example, José Marco Puig, who attempted to defect on 7 October

1938 and was shot, had been friends with men who had defected the previous day. The unit

command, concerned that Marco might attempt to defect as well, assigned him to guard duty

with one Enrique Sánchez, who was held in much greater confidence. When Marco attempted to

defect, leaving his post with the claim of having to answer a call of nature before running to the

Nationalist lines, Sánchez was prepared; he witnessed Marco’s flight, demanded that he stop, and

shot him when he did not. 61 Other soldiers, who were under less suspicion, were consequently

not watched, and they seemed to be able to defect more easily. For example, Vicente Suárez

Benarre, a soldier from 1 st Company, 393 rd Battalion, 99 th Mixed Brigade, defected on 24

September 1938 to the surprise of his unit: he had served in the assault on Teruel, one of the

most brutal battles of the Civil War to that point, and had lost his toes to frostbite, “which,” the

report stated, “he boasted as a contribution he was willing to have the war against fascism

impose upon him.” 62

Generally, predicting who would desert was an important part of an approach to control.

That there were false negatives is illustrated vividly by the fact that a man who had lost his toes

to fascism was willing to desert. Since it is impossible to have complete information about

soldiers’ intentions, it would have surely been more effective, in preventing desertion, to watch

all soldiers at all times. But that would always strain resources and inhibit military operations.

61 “Desertores,” 1 Cuerpo, 15 October 1938, folio 30. CDMH, SM, caja 421, folios 29-31. 62 “Desertores,” 1 Cuerpo, 14 October 1938, folio 34. CDMH, SM, caja 421, folios 32-35. 143

Some predictions were necessary. The Republic’s standard policy was, first, to base predictions on demonstrated behaviour rather than on cheap signals, and, second, to use the information to decide whom to watch, rather than to decide whom to actively and pre-emptively punish.

When predictions fell short of these two conditions, however, they could prompt desertion. Occasionally, units carried on using membership cards to decide whom to monitor: some units had data about the date on which their soldiers joined their party or union, assigning vigilance accordingly. But it was still possible to purchase a membership card with a date prior to the start of the war, for a price. As such it constituted a cheap signal, and deserters took advantage of it. 63

In addition, sometimes suspicions of a soldier did not provoke a decision to watch that soldier more closely, but instead active hostility and persecution. Sometimes soldiers were detained; in rare cases, they were shot. The story of the Falangist Vicente Pozuelo Escudero is fascinating from this point of view, indicating both that a right winger could serve for a long period of time provided that he felt himself relatively safe, and that the sudden realization that he was not would provoke him to leave. Pozuelo had been a stellar medical student in Madrid, where Juan Negrín, at the time a professor at his department, knew him to be a Falangist. As a soldier conducting his required military service at the start of the war, Pozuelo continued to serve the Republic as a medic, but was under some suspicion because his education suggested that he was upper-class. However, he continued to serve until April 1938. Then, he received word that he would be killed. A friend of Negrín’s—who was now the Prime Minister—told him that the

Communists had found his name on a list of Falangists at the medical school. Negrín had ordered this man to tip off Pozuelo because Spain would need him in the future: “I have orders to tell you that men are coming to kill you tonight. According to Negrín, you’ll know how to deal with

63 Corral 2007, 215–216. 144

them.” He defected that evening. 64 Thus it was surely possible for many defectors—even committed Falangists—to carry on serving the Republic. In Pozuelo’s case, he had served for close to two years. But the immediate prospect of death quite naturally provoked him to leave.

However, the policy of the Republic was to try to avoid such active persecution. Soldiers would often still serve while they were being watched, with punishment often only inflicted once the soldier attempted to desert. In the meantime, the army enjoined its commissars to be discreet about monitoring. 65 The decision to punish a soldier, then, was generally contingent upon an

actual attempt to desert.

Control is not costless: as noted, constant vigilance can reduce military efficiency. Hence

the capacity of an armed group to control its soldiers depends in part on its ability to handle the

number of individuals attempting to desert. As the Republic’s losses mounted and individuals

gained more reason to risk deserting, the Republic proved that it could not meet this increasing

strain. Desertion was relatively low in 1937, increased in 1938, and ultimately the wave of

defection crested after the fall of Catalonia at the end of January 1939. Much can be learned

about this pattern from the First Army Corps’ reports of defection over this time period. I

compiled 49 summaries of defection, generally appearing every five to seven days and covering

the beginning of June 1938 through mid-March 1939, a total of 358 reports of 832 defectors.66 I combined these with summaries of defection compiled by the corps and summarizing the periods

May-November 1937 and December 1937-March 1938, to give a comprehensive overview of defection from ten months into the war until its end. Figure 4.1 charts the total number of

64 Montoliú 1999, 215–227. 65 For example, a commissar in Santander was ordered to proceed in his vigilance over a particular soldier with “maximum discretion so as to avoid a lamentable mistake.” Jefe, Sección Segunda, Estado Mayor, to Jefe del Batallón 115, 20 January 1937. CDMH, Serie Político-Social [PS] Santander, caja L436, carpeta 13, expediente 19. 66 “Información: Deserciones,” CDMH, SM, 3 September 1938 through 18 March 1939, caja 421, folios 1-208, and “Información: Deserciones,” 22 March 1939, CDMH, SM, caja 781, folios 1-9. 145

Figure 4.1. Monthly Rate of Defection Attempts, 1 Army Corps

250

200

150

100 lived 50 died

*Monthly rate estimated from period totals

attempted defectors by month, indicating both the number who were shot trying to escape and the number who successfully escaped. There were two peaks: in August-September 1938 and in

February 1939. The crucial events occurring on the Republican side during the first wave were the stalling of the Republican assault on Nationalist positions on the Ebro and the beginning of a

Nationalist counterattack; the latter, much larger, wave occurred after the fall of Catalonia in

January 1939.

More and more soldiers gained reasons to switch sides over this period. The Republic’s successive defeats put victory out of its grasp and brought Nationalist triumph ever closer; men who fought to the end would not be safe in the new order. The Republic was steadily exhausting its supply of food and other essentials. And men were increasingly finding that their homes and families were now on the other side of the Nationalist line, and felt a responsibility to protect them. There is no doubt that a secular increase in motivation to defect took place over this period. Several reports on the desertion of various soldiers suggest that they were demoralized by

146

Table 4.1. Survival rate for attempted defectors, 1 Army Corps

Period Defectors per Survival rate (%) month 1 March – 30 November 1937 37* 70.5* 1 December 1937 – 31 March 1938 30 74.2 June 1938 52 88.5 July 1938 37 83.4 August 1938 83 89.2 September 1938 86 90.7 October 1938 50 84.0 November 1938 45 91.1 December 1938 57 94.7 January 1939 99 97.0 February 1939 215 90.2 March 1939 108 96.3 *Estimate; see text

bad news from the Catalan front, including one, Francisco López Cervera, who had a particularly extensive knowledge of Catalan geography and was therefore able to follow—and become preoccupied by—war reports as they came in. 67

These figures also indicate that the Republic’s capabilities for control were increasingly unable to meet the demands placed on those capabilities. If the increase in the defection rate was solely due to an increase in the motivation to defect, with the army’s ability to control adapting to this shift, then we should expect that the risks to defection stayed constant: it was simply that more men wanted to leave, and so more men took the risk. This is not what we find. For example, the Republican army got worse at shooting defectors over time, as seen in Table 4.1.

Before November 1938, 85-90% of defectors survived; afterward, 90-95% did. In fact, on average the death rate declined by one percentage point per month. The general downward trend is confirmed with prior data as well: from December 1937 to the end of March 1938, the survival rate for attempted defectors from the 1 st , 2 nd and 69 th Divisions, First Army Corps, was only

67 “Desertores,” 1 Cuerpo, 10 February 1939, folio 78. CDMH, SM, caja 421, folios 74-81. 147

74.2%. 68 (Earlier figures are more difficult to compare; the summary reports of desertion from the First Army Corps from March through November 1937 lump together deserters to the rearguard and defectors. 69 If we assume that 41% of these were defectors, as was the case from

December 1937 through March 1938, and is a generous estimate considering that more soldiers

would have deserted rather than defected earlier in the war when the Republic stood a greater

chance, then the survival rate for defectors from March through November would have been

around 70%.) The evidence, then, suggests that Republican control at the front lines was steadily

declining.

The faltering of control had two basic manifestations in this period. Not everyone could

be watched at all times. As a result, when the unit did not know that a soldier was likely to

desert, he had a greater opportunity to escape. Desertion increased among men who, by the lights

of the unit leadership, were not suspected of desertion. For example, Emilio Salvador Peirò, a

soldier from the 3 rd Company, 103 rd Battalion, 26 th Mixed Brigade, had always been an

enthusiastic soldier. He continually spoke ill of the enemy, and was under little suspicion.

However, after the Nationalists conquered his hometown of San Agustín, Teruel, in southern

Aragón, he became noticeably worried about his family. He defected shortly after, on 11

November 1938. 70 Reports abound toward the end of the war, in February and March 1939, of soldiers who had always been reliable now deciding to switch sides. What was going on was not just that more soldiers wanted to leave: it was that the army was ill-equipped to identify those soldiers. Demonstrated enthusiasm was no longer enough. There were too many false negatives.

68 69 División, “Relación numérica,” 8 April 1938; 2 División, “Relación numérica,” 13 April 1938; 1 División, “Relación de bajas habidas,” CDMH, SM, caja 5379. 69 1 Cuerpo de Ejército, “Estado numérico,” n.d. CDMH, SM, caja 5379. 70 “Desertores,” 1 Cuerpo, 15 November 1938, folio 164. CDMH, SM, caja 421, folios 163-165. 148

Further, control was weakened when officers began leaving in larger numbers, permitting

the men serving under them to leave as well. Earlier, men would typically leave singly or in

small groups. In the First Army Corps, the defection of officers allowed mass defection. In one

incident on 30 January 1939, a full section—sergeant, three corporals, and eleven men—

defected. One month later, in the largest single defection incident in that corps, a full platoon

switched sides: lieutenant, six sergeants, 18 corporals, and 35 privates. The incident report took

on a note of resignation: “In none of the men on the foregoing list had there ever been observed a

motivation that would have allowed one to predict that they would desert to the enemy. It is

supposed that the causes motivating this incident are due to the current circumstances which

influenced the men in a demoralizing fashion.” 71 There was little more to be done: when officers were leaving, masses of men could as well.

Control on the Republican side was thus difficult to enact, undermined by suspicions and factionalism. However, centralized control, in combination with a common interest in defeating

Franco, did in most cases overcome this factional resistance. Once it was set up, control succeeded in coercing soldiers to remain rather than desert. In order to get away, soldiers needed to devise various methods of evading monitoring and capture. Control permitted a large expansion in the size of the army through conscription. However, as time wore on, control became less effective, lacking the capability to handle an increased number of deserters. As soldiers gained new reasons to desert and defect, it became much more difficult to anticipate who would try to leave. And the control system unraveled as officers themselves started to defect.

Thus the unraveling of control had its own part to play in the ultimate crumbling of the

Republican army.

71 “Desertores,” 1 Cuerpo, 4 February 1939. CDMH, SM, caja 421, folios 125-129. “Desertores,” 1 Cuerpo, 22 March 1939. CDMH, SM, caja 781, folios 1-9. 149

4.4. Nationalist Spain: Control in the Rearguard and the Front Line

Control of desertion in the Nationalist rearguard was part and parcel of the general control that the authorities exercised over society. Unlike the “uncontrolled” violence characteristic of the Republican zone, violence in the Nationalist rearguard generally occurred as the consequence of central direction and as a manifestation of central state power over society. 72

There is some disagreement about this point: Stanley Payne, for example, argues that this is a distinction made by “partisans of the left.” In fact, he argues, “in the early months the Nationalist repression was not at all centrally organized” for regional and local military authorities were in control of the exercise of violence in the first several months. 73 There was an understanding that

an element of disorder was inevitable: in the instructions preparing the rising, it was argued that

“certain disorders under the supervision of armed civilians must be permitted in order that a

number of specified persons can be eliminated and revolutionary centres and organisms

destroyed.” 74 Thus executions were conducted at night and after procedures of denunciation,

drawing a certain resemblance to the violence of the “uncontrollables” of the Republic. 75

These elements of decentralization and private violence should not be overstated, however. Nor do they seriously undermine the general point that Nationalist authorities had much more extensive control over society than their Republican counterparts, a control that would enable them to punish desertion more effectively. On the Republican side, as we saw, the phenomenon of violence, decentralized to local authorities and in private hands, created the conditions in which deserters could escape persecution, and in fact retain their rifles as rearguard patrols. In contrast, on the Nationalist side, private violence never reached that extent. Far from

72 Juliá 1999; Richards 1998. 73 Payne 1987, 211–212. 74 “Top Priority Orders by the Junta de Gobierno,” in del Castillo and Álvarez 1958, 164–165; quoted by Raguer 2007, 130. 75 Sánchez 1987, 111–112. 150

the chaotic distribution of arms in the first days and the resistance to disarmament by the CNT in

Barcelona, the Nationalist authorities vigorously disarmed civilians from an early date. Ten days

after the start of the rising, for example, General Gonzalo Queipo de Llano, administering Sevilla

with a significant degree of autonomy from the other Nationalist leaders, ordered “intensive

searches” for arms, which had to be handed over to the Civil Guard on pain of execution. 76

Queipo’s actions were indicative of a broader phenomenon: decentralization to local military authorities, such as Queipo, did not in any sense mean a less thorough or credible top-down control of society by the state. Every garrison that rose immediately declared martial law in its respective region and replaced the existing civil authorities, and on 28 July 1936 martial law was declared throughout the whole of Spain. 77 In addition to the order to turn in guns, martial law also gave a legal justification, however thin in principle, to persecute those who objected to

Nationalist rule as having committed the “crime of rebellion,” expansively defined. 78 In any event, by March 1937 Franco was signing off personally on all executions in the Nationalist zone. 79

Thus the Nationalist authorities had an extensive ability to round up and kill the opponents of the new regime. In conducting this violence, the leaders of the rebellion spoke of a

“cleansing” of enemies of Spain that needed to take place: the purpose was not merely to win a war but “the permanent suppression of the enemy.” 80 The victims were designated by local committees involving locally influential individuals including priests. They drew up blacklists of local leftists, for example permitting landowners to enact a bloody revenge on landless peasant activists in Andalucía. While this permitted the settling of private scores that had little to do with

76 Raguer 2007, 132. 77 Payne 1987, 211. 78 Raguer 2007, 143. 79 Payne 1987, 215. 80 Richards 1998, 34–35. 151 maintaining state power, even those acts of opportunism were typically conducted by petitioning the local military authority rather than on one’s own account. 81 Ultimately, the Nationalist state could prevent and punish far more private acts of resistance than could its Republican adversary.

However, it was not just that leftists defected because they were motivated to join the

Republic by their ideology. Instead, the officially sanctioned mistrust and persecution of leftists provoked some to desert, even despite official attempts to give leftists a chance. Thus, to a certain extent, persecution undermined control. A notable case is the Legion unit Bandera

Sanjurjo , created in August 1936 in Zaragoza. It made a specific appeal to leftists in the area and encouraged them to “do meritorious actions that can redeem their pasts.” Men joined the unit in order to protect their families from reprisals. But the unit was wracked with desertion in late

1936 and early 1937. Nationalist officers deemed it likely that many had not given up their left- wing pasts and had intended instead to defect at the first opportunity. However, this was not true of all of them. General Miguel Ponte, the head of the 5 th Army Corps under which the unit served, noted that the recruits’ families had been persecuted in the rearguard, the bonus that they were due for the Legionaries’ service unpaid. Ponte clarified the effect this would have had: the men had been “assured that by [joining the Legion], if their conduct was honourable and loyal, they would wipe out the stains and errors of their former lives, which contrasts vividly with reality, as this [the repression of their families] shows that it is no longer just individuals but the

Authorities of their home villages who work against their desire for liberation and their material interests.” In effect, the new Francoist state had undermined its own assurance strategy: serve with loyalty, and you will be safe. Franco ordered local authorities to cease this ill-treatment of family members immediately. And the Bandera Sanjurjo ’s experience was repeated elsewhere,

with former leftists of the 12 th Division induced to desert when they heard from family members

81 Sánchez 1987, 111–112; Richards 1998, 36–38. 152

of the persecution they were enduring. 82 Men thus chose to desert when mistrust and suspicion

meant that they could not really trust that they would not suffer if they remained. However,

despite the provocative nature of persecution, ultimately control on the Francoist side was so

thoroughgoing that, even when provoked, individuals frequently had little recourse.

For the Francoist state’s extensive control included the prevention and punishment of desertion. The control wielded by the Nationalist state allowed it to find and to execute deserters in the rearguard. Deserters would be tried and punished under courts martial organized within their division or corps, administering the Code of Military Justice. But the punishment of deserters involved the rearguard quite extensively as well. The local Civil Guard and Falange officials from the soldier’s hometown would conduct an investigation after a soldier was denounced for desertion, including an extensive inquiry into the individual’s political and social history. Tight central control, for example over safe-conducts and permission to travel, made it easier for the authorities to prevent a deserter from getting very far. 83 Even if the deserter himself could not be caught—for example, if he switched sides and thus escaped Nationalist punishment—his family could pay the price. In turn, family members were subject to punishment for desertion as well. For example, Nationalist authorities in La Coruña ordered the arrest of all the adult male family members of two men who had fled to the Republican camp with their arms. 84 The repression of those family members was a bureaucratic endeavor engaging the military unit, with its records of soldiers’ hometowns and family members, and the local military authorities who were tasked with detaining soldiers’ families. 85 This apparatus of punishing family members was akin to its counterpart on the Republican side, but, given the

82 Corral 2007, 238–239, 271–272. 83 Ibid., 286–288. 84 Seidman 2011, 238. 85 Corral 2007, 319–320. 153 greater power of the Nationalist state over society, more effectively maintained. In the previous chapter I noted that César Lozas, a Republican sympathizer in Valladolid, accepted the conscription order rather than dodging the draft out of fears of reprisals against his father. Once in the army, Lozas found that the other conscripts in his unit were intimidated from deserting by the policy of control in place: “They lacked the ideological awareness that would have motivated a serious attempt to desert. Moreover, any such attempt could cost you your life – and even if you succeeded, reprisals could be taken on your family. There was considerable terror in the rearguard…” 86

There were, however, ways in which deserters could slip through the cracks of this system. Where it was more difficult to maintain top-down control over society, deserters could hide more easily. Not only draft dodgers but deserters as well hid out in bands in the hills of

Huelva, Sevilla and Asturias, facing death from the Civil Guard and other rearguard authorities if they were caught. 87 The role of terrain is interesting here: as on the Republican side, it appears

anecdotally as though desertion was more common among soldiers from mountainous areas than

among soldiers from flat land, because the former could take advantage of the difficulty of

moving around and their knowledge of the rough terrain to evade the Nationalist authorities. 88 I provide more systematic evidence on this point in Chapter 6.

Home could prove an important resource for a deserter who was hiding out. Among his family or friends, he could receive support and shelter. While many denounced deserters out of fear of further reprisal, others refused to do so. 89 Indeed, the repression of families of those who

86 Fraser 1979, 284. 87 Corral 2007, 116–117; Nieto 2007, 31. 88 Seidman 2011, 239. 89 Ibid., 213. 154

had hid out in the hills sometimes induced those family members to escape to flee as well. 90

Recognizing the appeal of home, the Nationalist authorities sought to make it as difficult as possible for soldiers to return home. One tactic was physical distance. On 23 September 1936, for example, Nationalist authorities issued an order that militiamen must serve outside their home regions in order to receive their pay. 91 Within the regular army, various orders over the

course of the war transferred soldiers away from their home regions. With Catalan soldiers

crossing the lines on the Aragón front to head to Catalonia, and with Galician soldiers on the

northern front (Asturias, Santander, Basque Country) deserting back home, in April 1937

Franco’s headquarters authorized a swap of Catalans for Galicians between the two fronts. 92

Very often, soldiers could not even take advantage of home because they could not make

it out of their units in the first place. Those attempting to defect could be shot in the attempt, and

in December 1936, Trench Councils permitted summary executions at the front rather than

having to wait for headquarters, though documentation on these proceedings is scarce. 93 As in

the Republican army, the Nationalists employed particularly tight vigilance against those who

were dubious, relying upon more proven veterans to administer front-line monitoring. For

example, the army assigned one corporal and two veteran soldiers to watch over sixty conscripts

at the Battle of the Ebro in 1938, paying particular attention to the most recent draftees. 94

Desertion and defection apparently varied by the degree to which these policies of front-

line control could be put into place. In some circumstances defection was more difficult. For

example, crossing no-man’s-land was a particularly dangerous proposition when an elite unit

90 Nieto 2007, 31–32. 91 Seidman 2011, 27. 92 Corral 2007, 145. 93 Ibid., 286–287. 94 Seidman 2011, 74. 155 with good snipers, such as the Legion, was serving at the front. 95 In contrast, defection was facilitated when such monitoring could not be established. Officers in northern Spain, for example, were frustrated that “the extensive and sparsely covered front prevented competent surveillance,” enabling defections to the Republicans.96 As in the Republican army, soldiers took

advantage of occasional lapses in monitoring to leave. For example, Santiago Crespo Pelayo, of

the 1 st Navarre Division, left his unit supposedly to enroll in the Legion, but instead went home. 97

In addition, desertion was more possible among the volunteer Falange and Carlist militias

early on in the war, since they were frequently not subject to the same rules about desertion as

the regular military. Faustino Sánchez, a Falangist youth from Asturias who fled to Galicia when

the rising failed back home, reported that, just like on the Republican side, Falange militiamen

could go home at night: “For us volunteers there wasn’t much discipline. After we took my home

town, I used to leave the column and go home for the night when I felt like it.”98 Carlist militiamen were also apparently able to return home from time to time—but this was apparently not as much “when they felt like it,” but restricted to certain moments, particularly after victories or when the harvest needed to be taken in. 99 It is unclear whether such behaviour changed after

Franco’s order of September 1936 requiring militiamen to serve in a different province than where they were from in order to receive their pay. In any case, as of December 1936 the militias were militarized and subject to the regular Code of Military Justice that the other units of the

Francoist army were required to follow. Despite some variations in the extent of Nationalist control, which permitted instances of desertion, on the whole the Nationalist side suffered

95 Corral 2007, 52. 96 Seidman 2011, 59. 97 Corral 2007, 31. 98 Fraser 1979, 145. 99 Seidman 2011, 237. 156

considerably less from the disorder that afflicted its adversary and thus offered soldiers far fewer

opportunities to desert.

The centralized coercive capacity of the new Nationalist state was also able to overcome

differences in political agendas. Factionalism on the Nationalist side was not as deep as on the

Republican, lacking deep substantive differences or a history of violent conflict. However, there

were still, among some, programmatic differences. To the extent that these differences existed,

however, central control was able to prevent them from producing large-scale defection.

Over the course of the autumn of 1936 and the spring of 1937, Franco institutionalized

unity in the new regime, in two steps: unified military command and a unified political entity.

First, the coequal collective decision-making body of the generals, the Junta de Burgos , was replaced on 29 September 1936 by a unified command with Franco himself as head of the military and head of the new state. Franco had two key advantages in this process: after the deaths of key rivals in the first days of the war, he was the most senior general who would be acceptable to all. Vitally, in addition, he had the support of the Army of Africa. At the key meeting of generals outside Salamanca at which Franco’s command was decided, Col. Juan

Yagüe “pointed out forcefully that his legionaries and Moroccan troops wanted Franco as supreme commander.” 100 The strong preference for unity among the military hierarchy was

probably enough to ensure Franco’s ascendance, but the need for Yagüe to intervene to highlight

his support among the elite forces of the Nationalists indicates the long-term advantage Franco

had over the other generals.

The Army of Africa also lingered in the background in the more difficult process of

political unification. On 18 April 1937, from his headquarters in Salamanca, Franco announced a

decision to unify the Carlist Traditionalist Party and the Falange in a single entity with the

100 Fraser 1979, 203. 157 unwieldy name Falange Española Tradicionalista y de las Juntas de Ofensiva Nacional

Sindicalista (FET-JONS), and to ban all other political parties. There had been mixed feelings among both Carlists and Falangists about the prospects of unification. Before the war, the

Falange’s founder, José Antonio Primo de Rivera, had had discussions with Manuel Fal Conde, the leader of the Carlist militias, on cooperation in a rising. During the war, Carlist and Falange leaders had broached the idea of political unification and published an exploration of points of agreement and disagreement between the two tendencies. Certainly members of both parties saw the need to unify for the sake of victory in war. Both would now come closer to real political power than they ever had. 101 But there were elements of each that were antagonistic toward cooperation with the other, disagreeing essentially between Carlist decentralized rural traditionalism and Catholicism and the Falange’s preference for centralizing, modernizing, corporatist revolution. 102 In addition, Primo de Rivera had feared domination by the military,

believing that it would co-opt the Falange with few guarantees for the implementation of its

program. Fal Conde, in the preparation for the rising, had been concerned that Carlist support

would help ensure military dominance without a clear guarantee that the Carlist monarchy would

be restored. Both recognized that in a confrontation with the armed forces, they stood little

chance. 103

When Franco announced political unification, there was opposition within both camps, and both Falange and Carlist militia leaders contemplated a revolt against Franco. Although there were scattered mutinies on both sides, and the senior leadership that opposed unification was arrested, others contemplated revolt but decided against. Their reasons are telling. Dionisio

101 Payne 1961, 109–110; Ellwood 1987, 41, 45. 102 Blinkhorn 1975, 167; Payne 1961, 152. 103 Blinkhorn 1975, 242–245; Payne 1961, 110. 158

Ridruejo, a Falange student, contemplated seizing Salamanca and taking Franco prisoner. But he

was dissuaded:

Naturally, the Foreign Legion would have moved in the next day and captured us. But,

much more important, such a move on our part would have brought the war to a halt. I

didn’t dare take the step – who would have? In other circumstances, if there hadn’t been a

war, the Falange would have killed Franco. None of us could accept our forced

capitulation, our unification in which we had not a word to say.

The Marquess de Michelina, a Carlist commander, had a similar reaction:

‘It was evident from the moment Franco took over political and military power that the

state was becoming falangist. The latter we saw as an extension of German and Italian

fascism. Nothing could be further from our ideals than totalitarianism.’ But what could

they do? Had they risen to oppose unification they would have been shot – worse, the war

might have been lost. Troubled, worried, they continued to collaborate. 104

If efforts at centralization prompted much less desertion on the Nationalist side, then, it

was ultimately testimony to two basic logics. There was apparently a greater potential for

cooperation on the Nationalist side, a more widespread recognition that winning the war required

submission to central authority. And this was facilitated by the fact that the nationalist program

was able to offer both the Falange and the Carlist programs substantive gains. Unlike among

many Anarchists, there does not appear to have been a widespread sense on the part of either that

they would have to give up their whole program to win the war. The balance of common versus

competing interests favoured the former to a greater degree than it did on the Republican side.

However, lingering in the background was the sense that there was very little alternative: that the

coercive apparatus of the military would bring unification into being by force.

104 Fraser 1979, 318–319. 159

4.5. Conclusion

In both the Republican and Nationalist camps, then, the occurrence of desertion had much to do with the coercive measures that each side had in place to control desertion: limiting opportunities for soldiers to leave and deterring them from doing so by finding them and punishing them, and their families, for the act. The Republic began with little rearguard control and considerable variation among militias in the control they enacted on the front lines, while factionalism meant that resolving these problems by instituting greater control provoked desertion in the short run. However, the Republic was able to intimidate recalcitrant political groups into accepting greater control. Control improved as the militias were militarized and the rearguard policed, but this control unraveled over the later part of the war as Republican monitoring capabilities were less and less able to meet the demands placed on them. It was difficult to determine who should be monitored and as the agents of monitoring, such as officers, deserted as well. The Nationalist side implemented much more rigorous and thoroughgoing top- down control from a much earlier stage. Even though its persecution of leftists prompted some to leave and undermined control policies, the extent of Nationalist control also meant that soldiers had very little opportunity to desert.

The chapter has not attempted to give a detailed analysis of the origins of different degrees of control. However, the Spanish case gives some hints about the importance of organizational hierarchy and coercive capabilities. The Nationalist advantage in control had much to do with its ability to keep control over arms and its clearer military hierarchy at the outset of the war. Its army, and especially its core Army of Africa units, provided an important source of threats to keep its militias in line. In contrast, the Republic’s coercive capacity and organizational structure was shattered by the military rising and the fragmentation of state

160

authority. It was able to build control over time in part through institutional adaptation to its new

circumstances, such as the innovation of organizations like the Inspección General de Milicias . It improved its coercive capabilities through the training of new, tightly disciplined military units the threat of which proved persuasive others, and through the expansion of security personnel in the rearguard. However, one important source of variation may be idiosyncratic. Philosophical anti-militarism in the Anarchist movement had a role in the Spanish case that it may not have elsewhere.

Rather than examining the origins of control, this chapter has attempted to give a basic overview its impact. It has noted control’s correspondence to the general patterns of desertion in the Spanish Civil War and has provided qualitative evidence of its role in explaining particular instances of desertion. In Chapter 6, I conduct a more systematic analysis of control and desertion. I focus on control in the rearguard, exposing how variations in terrain affected the ability of the Republic to track down and prosecute deserters, and thus shaped its desertion rates.

Next, however, I turn to cooperation, the final dimension in my analysis of desertion in the

Spanish Civil War. I examine how the dynamics of trust and mistrust affected cohesion within the Republican and Nationalist camps. Much of the time, the armies fighting in Spain could not rely on mutual trust—and this made control particularly important.

The importance of control reminds us that an attachment to the common cause is alone not a sufficient explanation for joining and remaining in armed groups. Schemes of rewards and punishments, including coercive, top-down control, permit the armed group to avail itself of the services even of people who do not share its aims. And it also allows the group to outweigh the personal costs of fighting endured by combatants, which tempt even those who are committed to desert. A focus on control also broadens the military dimension of civil war past manpower,

161 resources, and strategy. It takes seriously that civil war engages at least two parties in a clash of segmented areas of political authority—competing projects for the state. 105 Within that context,

the prevention of desertion as the implementation of coercive control over front line and

rearguard takes on a special significance, as part of a state building project. Many other things

equal, 106 the party to the civil war best able to impose such control is best placed to win.

At the same time, the evidence that factionalism placed limits on the effectiveness of control, making efforts at control appear to be efforts at arbitrary violence, reminds us that the existence of common aims is not meaningless. It is not wholly trumped by top-down, individual incentives such as the threats implied by control. Control can handle many uncommitted troops, but it can falter in the face of deep factionalism and persecution. This chapter, therefore, has illustrated the need for a new theoretical synthesis between a focus on combatants’ commitment to the cause and a focus on their self-interest.

105 Kalyvas 2006, 17. 106 See Cunningham, Gleditsch, and Salehyan 2009 for the most comprehensive analysis of sources of victory and defeat in civil wars. 162

Chapter 5 Desertion, Collective Action, and Norms of Cooperation

5.1. Introduction

Thus far, we have seen how different forces in the Spanish Civil War varied by the degree to which their combatants shared common aims, and how control helped to induce combatants to remain fighting even when they had no particular interest in their unit’s success. In this chapter, I now move beyond control and turn to a second approach: the cultivation of norms of cooperation among soldiers. I argue that combatants can have a heavy influence on each other, and can induce each other to fight or to desert. The existence of common aims helps us understand when that mutual influence is likely to favour fighting and when it is likely to favour deserting. When common aims exist, norms of cooperation among soldiers can induce them to fight together. Such norms create a sense that if one soldier fights, others will as well, and they can achieve a collective aim together. The perceived costs of warfighting decrease, because a combatant is more likely to anticipate that he can rely on others in combat to fight as well. And groups evolve social rewards for fighting and social penalties for deserting. These auxiliary rewards and punishments help to make desertion less attractive. Provided common aims exist, friendship and social homogeneity can facilitate the emergence of norms of cooperation.

However, when combatants have no particular attachment to the group’s common aims, norms of cooperation can fail to emerge; indeed, norms of lax behaviour, including cooperating to desert, can emerge instead. And under severe disagreement about aims, as with factionalism, conflicting aims interfere with common aims and combatants fear betrayal by members of other

163 factions. They are thus more likely to desert, rather than pay the costs of fighting and not have that decision reciprocated by others.

This chapter first discusses reciprocity in the militias of the Republic. It notes that the

Republic’s enthusiastic militiamen could encourage or discourage each other. But among groups of opportunists, the latter was much more common, and a climate of severe mistrust could induce combatants to desert or defect. As the Republic shifted towards conscription and hence more uncommitted soldiers, negative mutual influences increased concomitantly as well, and norms of cooperation were much more difficult to achieve. I conclude by discussing the Nationalist side more briefly, noting that it experienced many of the same dynamics as the Republican side.

5.2. Collective Action and the Republic’s Militias

As I noted in Chapter 3, many militiamen joined up in an initial rush of enthusiasm and with a strong passion to defeat the military rising. This kind of enthusiasm was clearly preferable to opportunism; other things being equal, opportunistic soldiers would be more likely to desert than ones who valued their units’ success on the battlefield. However, enthusiastic new recruits still suffered the pains of war fighting, and their enthusiasm was difficult to sustain in combat.

As I argued in Chapter 3, this meant that militiamen who shared a common aim faced the collective action problem: they preferred that their units succeed, but faced severe personal costs in contributing to that success, costs that became worse as those around them deserted rather than fighting. In this section, I discuss how norms of cooperation could help them address this collective action problem. I explain how these norms depended upon common interests. I conclude the section by discussing how norms of cooperation were facilitated as well by social homogeneity.

164

Militiamen’s enthusiasm could break down in the face of combat and its rigours.

Confronting these problems, men frequently reacted to others around them. An example of

courage in the face of such terrors could inspire others, indicating that it was possible to resist,

and providing an example of self-sacrifice that could evoke a willingness on another’s part to

sacrifice as well. Self-sacrifice could become a norm, inducing others to acts of courage as well.

A vivid instance was individual attacks on tanks in the defense of Madrid. Lacking anti-tank

weapons, the militia could only defend against them with a grenade assault. Victor de Frutos, of

the Anarchist Primero de Mayo battalion defending the Carabanchel suburb south of Madrid,

relates that his men were profoundly afraid of tanks until they saw the Soviet film Sailors of

Kronstadt , which depicted militia resistance to tanks. The day after seeing the film, militiamen

were willing to try a grenade assault on a tank, which they accomplished heroically. 1 Militiamen in other units likewise feared tanks terribly: “as the Romans must have feared Hannibal’s elephants,” according to Julián Vázquez, a Communist union leader. To attack them required immense courage. Narciso Julián, a Communist leader from Barcelona serving at Madrid, relates how a militiaman who had blown up a tank described it at a rally:

‘Well, look, it’s – it’s very easy. You lie on the ground with a bomb in your hand and you

let the tank get to within three metres and you throw the bomb at the tracks. If it

explodes, the tank blows up on the spot. And if it doesn’t – if it doesn’t explode –’ He

stopped, not knowing how to continue. ‘If it doesn’t,’ he added finally, ‘the tank crushes

you...’

When asked to explain to an American journalist how he had blown up a tank, another militiaman replied, “echando cojones al asunto ”—applying courage (literally testicles) to the matter, according to the Left Republican leader Régulo Martínez who set up their interview.

1 de Frutos 1967, 62–71. 165

Martínez relates, “A week later, I was shown a copy of an American paper in which I read that

Madrid militiamen had invented a new anti-tank device called ‘echando cojones al asunto. ’” It

did not occur to the journalist that there was a way of defeating tanks that was not through a

special weapon but through audacity alone. In this setting, the example of militiamen could make

all the difference. Vázquez saw a group of naval orderlies in the Usera suburb charging down a

group of tanks. “One of these sailors, a man he later learnt was Antonio Coll, threw himself on to

the ground in the path of three enemy tanks, let them nearly reach him and then threw his bombs.

Two were blown up, the third turned tail and fled. [Vázquez:] ‘Coll was killed; but tanks were no

longer seen as invincible juggernauts.’” 2

The attitude of civilians was also important in reinforcing willingness to fight. Civilians served as the agents of social rewards and penalties. There could develop among civilians a norm of mutual, personal sacrifice that led them to want to join up: if others were sacrificing, they would as well. For example, an elderly couple attempted to join a UGT militia, according to a union member: “The president of the socialist party's East Circle came to ask what they were doing. ‘We've answered the call, we’ve come to do what everyone else is doing – ’ He asked them to go home; they refused. They had to be taken off almost by force.” 3 Civilians also shamed soldiers who were retreating into returning to the lines: Lorenzo Iñigo, a member of a

CNT metalworkers’ union, heard women yelling at a few militiamen retreating from the front west of Madrid: “Cowards, chickens, where are you going?... If you aren't brave enough to be in the front line, then give us your rifles because we haven’t got any. We’ll go down to take your place.” 4

2 Fraser 1979, 267–268. 3 Ibid., 261. 4 Ibid., 264. 166

On the other hand, soldiers could furnish negative examples to each other. War was an

interactive endeavor, and men relied upon each other. For Timoteo Ruiz, a young militiaman

from Toledo province in the Fifth Regiment, “the worst was always feeling alone, exposed. ‘You

could never be sure whether the column to your right or left was still there, or whether you were

in danger of being encircled.’” 5 Militiamen could decide that desertion was better than trusting their lives to those around them when doing so was risky. Based on this logic, the willingness of some to flee in the face of battle increased the risks for everyone else, and in doing so risked provoking a mass exodus. When others around them were panicking as well, militiamen would lose heart and try to leave too. A militia commander writing in September 1936 about desbandadas (panicked mass flight) near Sig üenza, Guadalajara province, noted the importance of confidence in one’s comrades. When such confidence was lacking, for example in new units where militiamen had not had the chance to learn whether they could rely upon their colleagues, a desbandada was much more likely. They “began with the spontaneous retreat of a few groups of scared men,” provoking the rest to flee as well. To remedy the situation, this commander recommended not only the imposition of harsh penalties for sowing discontent and indiscipline, but also instructing militiamen “that those who fled to avoid their own death or injury created situations that magnified the possibilities of casualties. Soldiers should be drilled to instill confidence, strengthen obedience, and induce group trust.” 6

Thus men’s behaviour swung dramatically from enthusiasm to terror and back again.

General Vicente Rojo, the commander of the defense of Madrid, portrayed the militia as “highly variable in morale, which went from a passionate ideological extreme to the edge of easy

5 Ibid., 258. 6 Seidman 2002, 50. 167 collapse in critical situations.” 7 A story recalled by Eduardo de Guzmán, a CNT journalist in

Madrid, provides an example of how the mutual influence of soldiers could provoke such

variation in morale. “Mikháil Koltsov, the Pravda correspondent, said to me one day that he

didn’t understand the Spanish. ‘As soon as they see Moroccan cavalry they start to run. And then

one says to the other: “You're a coward,” and the other replies, “I’ve got more balls than you,”

and he stays there and allows himself to be killed. How do you explain that?’ Koltsov was right:

how do you explain it?” 8 Ultimately, one answer to Koltsov’s question was a norm: men would hold fast when others were holding fast. But they could flee when others were fleeing too. The machismo or male ethos of the situation emerges from the interaction of the two men. Fleeing initially seems acceptable if the other is doing so too. But then one soldier invokes a norm about what is obligatory behaviour for a man, and they stand their ground.

But in order to achieve reciprocal cooperation and hold the unit together, a common aim was necessary as well. While haranguing his troops outside of Madrid, Cipriano Mera, the

Anarchist commander, attempted to promulgate a cooperative norm on the basis of his men’s commitment to the cause:

Our discipline needs to correspond to the conviction we have in our ideas, and, according

to our ideas, we cannot fight one moment and do what we like the next. Such actions

reflect badly on both ideas and men. We who are gathered here have come on our own

account and we have made a commitment among ourselves. He who would withdraw

without first defeating the fascists across that mountain would be a traitor to this

commitment. 9

7 Rojo 1967, 36. 8 Fraser 1979, 264. 9 Mera 1976, 28. 168

Mera’s attempt thus explicitly linked the aims to the group’s value of mutual commitment and honouring one’s responsibilities, while alluding to the unit’s immediate objective and the project that the men in question could realize in common together. In Mera’s appeal, the commitment to a common cause is mean to underpin a norm and to induce men to carry on fighting.

When common aims were lacking, in contrast, it was much more difficult for those who wanted to fight to trust others, and those who had little interest in fighting consequently had little interest in promulgating norms of cooperating to fight. Instead, in some units with many opportunists, norms of shirking were more common. According to Broué and Témime, in some units “you went back home between two turns at guard duty, and you were regarded as a madman if you refused to sleep when you were on guard at night.” 10 Col. Mariano Salafranca, in

a report about the difficulties encountered by the Oropesa column fighting in late August 1936

on the main highway controlling the southwestern approach to Madrid, wrote that many of that

column’s difficulties had to do with the variety of different motivations within the militias:

Their retreats had as a cause the organization and heterogeneity in constitution of the

militias, which brought together masses in which, in a disorganized fashion, were mixed

noble, valiant spirits, passionate about the cause they defended, and men with completely

opposite characteristics…and…an amorphous mass, disposed, according to

circumstances, to follow one or the other, without their own feeling, looking, egoistically,

in difficult and dangerous circumstances,…the easiest route for the conservation of their

own life. 11

Thus the mixture of those who were deeply committed to the cause and those who had no such commitment conditioned the men around them to fight, or not to.

10 Broué and Témime 1970, 174, emphasis added. 11 Alpert 1989, 57. 169

Since there were multiple political factions within the Republican camp, it was often hard to see that they were working to a common aim. As I argued in Chapter 3, competition among various political parties and unions meant that the Republic’s militiamen fought with multiple different aims in mind. And their competition could prompt desertion, as soldiers became disheartened at the lack of cooperation among their peers.

An illustrative example is the fiasco of the assault on Huesca, a key provincial capital in

Aragón, in November 1936. According to a Major Aberri, a Republican officer sent to the

Aragón front to assist in the organization of the Republican war effort there, Huesca was poorly defended and ripe to fall. But the assault was totally undermined by a lack of coordination among the militias. At a meeting where the commander of the Aragón front presented the militia leaders with a clear plan to take the city, “those present listened to his plan, which was discussed in detail, but unfortunately they finally decided to consult their respective trade-union organizations before accepting anything. In the end the discussion took a very regrettable turn because the commander’s request that some of the columns should hand over to other units the additional material they needed was rejected out of hand.” 12 Ultimately, this lack of coordination doomed the assault on Huesca when an anarchist unit attacked too early. José María Aroca, an

Anarchist student and militia volunteer, admitted that he decided to leave the militias because the failed Huesca assault ended the last vestiges of his confidence in the unit’s organization and leadership: “I decided to abandon the militias when the occasion to do so ‘with dignity’ presented itself. That is how far I had been discouraged by the mess that reigned in the militias, a mess that could play with criminal callousness with the lives of three hundred men.”13 Thus

factionalism, interfering with commonality of aims, produced desertion. One party would fear

12 Bolloten 1991, 257. 13 Aroca Sardagna 1972, 114–115. 170

that the success of another would translate into weakness down the road. When they saw that the

men around them were willing to pursue factional interests at the expense of the common cause,

men could decide that the risk of fighting was too high, and leave.

The Republic’s militias were often beset by deep divisions and severe mistrust. This was

a manifestation of the severe violence and mistrust behind the lines. Rearguard violence emerged

from close political competition between left and right before the war. 14 In this context, being a

suspected rightist was often a sufficient condition for murder or other punishment. For example,

Gerardo Martínez Lacalle, an open Falange activist before the war, reports returning to Madrid

from El Escorial shortly after the start of the war, to be told by his building’s doorman: “‘What

are you doing here? They came to look for the señorito [an often derogatory term meaning “little lord”] Gerardo.’ He told us that the CNT had come with a list of those who belonged to the

Falange and were taking them to the fields of the Casas de Campo to be killed.” 15 But evidence

of right-wing sympathies could be more arbitrary, such as the possession of monarchist

symbols. 16 Luís María Lorente Rodrigáñez, the son of a Republican banker, was taken from his

home by a Communist patrol, for the crime of having been seen playing field hockey, saying it

was “a sport for señoritos. I told them we played against teams of railway and tram workers, so it

wasn’t a señorito sport, and they let me go.” Later he was denounced as a rightist by his

building’s doorman and assaulted. 17

Above all the other forms of suspicion that undermined trust in the new armed forces, and hence cooperation, was the tension surrounding the regular army. This illustrates vividly the problem of suspicion in a climate of factional divisions. Many officers and soldiers, of course,

14 Balcells 2010. 15 Montoliú 1999, 208–209. 16 Ibid., 40. 17 Ibid., 98. 171 preferred to serve the Nationalist cause, often strongly so. This led to their defection directly, when they had the opportunity to do so, but it also had further knock-on effects on desertion. By increasing the suspicion between uniformed personnel and the militias, such disloyalty sowed the seeds for further defection problems.

Since officers were the chief instigators of civil war and since some carried on defecting into the war, those who remained suffered guilt by association. During the siege of the Montaña

Barracks in Madrid at the beginning of the war, Republican officers inside the barracks who wanted to surrender to the militias raised the white flag, but Nationalist officers carried on firing at the assembled workers anyway. The militia members thought that the white flag had been pure trickery, and killed many pro-Republican officers in taking revenge. 18 The sense of vengeance

and mistrust carried on at the Montaña. On 20 July, Manuel Carabaño, a fifteen-year-old

anarchist, “saw a group of men in shirt-sleeves who were trying to hide the fact that they were

officers by crying ‘ ¡Viva la República! ’ A group of militiamen surrounded them shouting

‘Fascists!’” They were then executed. 19 On 8 October 1936, retired officers in Madrid were

called upon to assemble, and faced a dilemma: if they did not go and were captured, they were

likely to be shot. But if they went, they might be presumed disloyal anyway since they had not

already started fighting for the Republic, and could be imprisoned in any case. According to

Martínez Reverte, “the majority who presented themselves ended up in the Model Prison. And

the majority of them were executed.” 20

Left-wing forces sometimes looked to signals from soldiers to determine where their

loyalties lay. These signals put officers on a knife edge. At Caspe, Catalonia, a young libertarian

saw a group of Guardia Civil wearing red neckerchiefs. “One of the guards came up to me. ‘Ah,

18 Martín Blázquez 1939, 116. 19 Fraser 1979, 78. 20 Martínez Reverte 2004, 56. 172

my friend, when you see a civil guard without one of these round his neck, shoot him. Only those

wearing them are on the republic’s side.’” 21 It could be extremely difficult for an officer under suspicion to know what signals to send, since protestations of loyalty were often cheap talk.

Colonel José Villalba, the chief of the garrison at Barbastro in Huesca province, had been in contact with both the coup plotters and Republicans before siding with the latter because they had a stronger presence within his garrison. He came under the suspicion of local militias from the first, as a known Africanista .22 Later, Juan García Oliver, the CNT delegate for war in

Catalonia and generally one of the Anarchists most sympathetic for the need for military

discipline, considered Villalba the most qualified candidate for the job of overall commander of

the militias operating in Aragón. But he did not fully trust Villalba’s politics. García Oliver

relates his strategy for dealing with this dilemma:

I made a decision. I would call Col. Villalba and ask him bluntly his opinion of the war.

If he responded that he had always been a leftist and republican, I would order him

imprisoned on the Uruguay [a prison ship in Barcelona harbour] to be tried for treason.

But if with complete frankness he answered that he did not understand politics, but was

just a professional officer, then…we would present him as Chief of Operations of the

Aragón Front. 23

Villalba apparently passed this test, but he could be forgiven for thinking it somewhat

Kafkaesque.

Mistrust led some on the Republican side to dismiss army and paramilitary officers, or

worse. Efforts to ensure loyalty thus also removed many actually loyal officers from the

Republic’s disposal. The Civil Guards came under particular attack. In Barcelona, the loyalty of

21 Fraser 1979, 120. 22 Arcarazo García 2004, 97–101. 23 García Oliver 1978, 269. 173 the Civil Guards was crucial in defeating the rising, but the government still dismissed 40% of its officers. 24 In Santander, the Militia section of the Popular Front war committee denounced the assassinations of the Civil Guards on 13 August: “The fact that, in different places in Spain, the

Guardia Civil have joined the fascists, cannot justify treating them all as disloyal, because in other places they have joined the people and with the people are fighting against the traitors.” 25

The siege at the Montaña barracks northwest of Madrid had consequences that reverberated through the regular military in the city. Francisco Abad, a Communist soldier who, before the coup, had helped to organize clandestine groups of pro-Republican soldiers and non- commissioned officers in an infantry unit in Madrid, later worked at finding loyal officers to lead militia columns. His job was made much more difficult by the Montaña siege:

However, there weren’t many trusted officers available for the 6,000 men we sent out. In

my own regiment, the majority were arrested after the fall of the Montaña, which was a

mistake. A great number of the junior officers could have served the republic. Suspicion

and distrust combined to make us lose a potentially useful force. Career officers weren’t

treated justly nor used properly in those first months.” 26

The climate of suspicion was systematized early on. Within the War Ministry, an effort began,

under Captain Eleuterio Díaz Tendero, to classify the Republic’s remaining officers as

Republican, indifferent, or fascist. Martín Blázquez compares Díaz Tendero to Robespierre for

the purifying zeal with which he approached his work. 27 Still, however, it was unclear that this effort was accurate. Antonio Cordón was involved in this effort, and claims that the office often lacked the information it needed: “We lost time uselessly, most of the time not knowing whom to

24 Alpert 1989, 25. 25 Saíz Viadero 1979, 69. 26 Fraser 1979, 117. 27 Martín Blázquez 1939, 121–122. 174

contact for the information we wanted.” He estimates that the standard claim that 20% of the

officers were loyal was actually too high. 28

The suspicion of officers led to considerable tension within the columns that formed in

the first months of the war. Jesús Pérez Salas, a Republican officer and the military chief of the

Macià-Companys column in Aragón, claims: “When, out of absolute necessity, [the working-

class organizations] had to make use of us…they employed only the minimum of loyal officers

strictly indispensable to their needs; these were kept under constant vigilance and were, in

addition, menaced because of their alleged fascist sympathies.” 29 In Gijón, Asturias, Dr. Carlos

Martínez served as an unofficial liaison between a column led by Captain José Gállego and the

Popular Front’s War Committee, and had to speak to Gállego’s men to increase their trust in him.

“While the committee trusted Gállego, amongst the militiamen there was a certain lack of

confidence in all the army officers. Nothing discouraged an officer more than to know that he

was not trusted by the men he was trying to lead. ‘He needed support, wanted me, as a

republican, to bear witness to his conduct, his loyalty and courage which were all quite

remarkable.’” 30 According to Helen Graham, mistrust of officers served as a frame to interpret

the misfortunes of a militia unit: “At any time, problems caused by shortage and dislocation

could be turned into accusations of treason and crypto-fascism.” 31

Faced with such circumstances, mistrust became something of a self-fulfilling prophecy.

Anti-militarist militiamen would sometimes desert if ordered to serve under officers. For example, Saturnino Carod, whose militia column on the Aragón front included eighty Civil

Guards, decided to reorganize his column after a bad defeat. He attempted to divide the column

28 Cordón 1971, 250. 29 Bolloten 1991, 250; quoting Pérez Salas 1947, 259. 30 Fraser 1979, 246. 31 Graham 2002, 144. 175 up into smaller units with a command structure. “The result was a near disaster; the militiamen abandoned the column and he was left with almost the guardia civil alone.” Many of those who

left later returned, but there were persistent tensions: “‘You can rejoin the column, but first you

will have to do a fortnight’s training. And your instructors will be the guardia civil .’ Imagine telling a CNT militant that he had to accept orders from a guardia ! But I wasn’t going to back down.”32

For their own part, officers were more likely to defect because of the mistrust that they faced. Major Jaime Solera, “a self-styled liberal democrat without political affiliation,” was serving as a staff officer at the war ministry in Madrid when the war broke out, and carried on serving the legally constituted government, as he saw his duty to be. He further believed that “the majority of officers in Madrid…shared his view.” However, given the public’s suspicion of all army officers, Solera claims, “many officers were not only in terrible danger of their lives, they were killed. Living in fear, they tried to escape.”33 Jesús Pérez Salas says that in the militia units from Catalonia, officers were under constant suspicion, and could be killed “for the slightest hint of a lack of enthusiasm,” ultimately giving those officers a strong reason to try to defect. 34 In the confusion of the early part of the war, when effective control was minimal, such defections were not terribly difficult.

As with military discipline, the Communists in Madrid represented an important exception to these patterns. They had laid the groundwork for cooperation with officers prior to the civil war,35 with a close association with the pro-Republican officers’ organization the Unión

Militar Republicana y Antifascista (UMRA), with more than 200 members in Madrid and a few

32 Fraser 1979, 133. 33 Ibid., 118. 34 Pérez Salas 1947, 106. 35 Montoliú 1999, 76. 176

hundred others scattered through other columns. 36 In organizing the Fifth Regiment, then, the

Communists made it a standard practice to use loyal officers as much as possible, in sharp

contrast to other political groupings, which seemed to employ those officers as little as

possible. 37 The receptive attitude of the Communist Party for officers made it an attractive

political entity, whether or not they actually joined the party ultimately. 38 Thus while the

factional division between the uniformed forces and the militias was a source of mistrust across

much of the Republic’s war effort, Communist cooperation indicated that it was possible to

overcome this mistrust, and build norms of cooperation, through concerted efforts to identify

common aims.

Norms of cooperation were assisted when men knew each other. Prewar ties, for example based on locality or political membership, were a way in which someone else’s self-sacrifice would provide an example, serving both as a basis for recruitment and a way of solidifying a unit. The small pre-existing militias of the different parties and unions, which had engaged in running street battles in the run-up to the war in the spring of 1936, served as the nuclei of the new militia, and local unions joined the militia en bloc . Militia often took their names from the

groups that formed them, such as the Artas Blancas bakers’ militia, or Voluntarios de Andalucía among expatriates from that region in Madrid. 39 All of José María Aroca’s friends and colleagues in the Anarchist youth league Juventudes Libertarios in Barcelona joined the Durruti column early on, and this became, in turn, an obsession for him as well. 40 Timoteo Ruíz, a young peasant from Toledo province, had been attracted to Socialism and to the Soviet Union in the years before the Civil War because of the highly unequal distribution of land in his village. He

36 Modesto 1969, 13–14. 37 Líster 2007, 102. 38 Graham 2002, 145. 39 Blanco Rodríguez 1993, 241. 40 Aroca Sardagna 1972, 11, 13. 177 joined the Fifth Regiment in Madrid with friends: “When I came from my village in Toledo province with a few other local youths who like me wanted to fight, we met a lad in Madrid from the village who had joined the communist youth. He told us that the best unit was the 5 th

Regiment which the communists were forming.” 41

In general, serving with others that one knew well made war more bearable, and militiamen frequently sought out that chance. According to Antonio Cordón, the Republican officer, this preference helped to account for the failure of the attempts in August 1936 to create new, government-directed volunteer units. “As for the rank and file, the militiamen preferred to stay with their militia units, with their comrades and brothers in arms, over joining the new battalions.” 42 Victor de Frutos, the Anarchist commander, argues that the fact that his Primero de

Mayo battalion was largely recruited from the same neighbourhood in Carabanchel Bajo, southwest of Madrid, gave it a strong sense of mutual reliance in the first days. A man from elsewhere still generally joined only if he had friends in the unit, and “in this way became a man we could rely upon.” Later, a man who had fled Nationalist repression in Sevilla joined Primero de Mayo but was a liability, as an unknown quantity. Not knowing much about his past, the unit

could only rely on the impression he made, and as a drunk that impression was poor: “we did not

know to what extent he was anti-Franco, but what we did learn in a short time was his love for

Bacchus.” Once his secret store of wine was discovered, the man denounced the Republic and

was put under close watch from that point forward—but still tried to defect and was shot in the

attempt. 43 Apart from this man, however, the sense of defending one’s own home gave Primero

de Mayo particular cohesion when the battle got to Carabanchel Bajo itself. The unit requested

41 Fraser 1979, 256. 42 Cordón 1971, 249. 43 de Frutos 1967, 73–77. 178

permission to be stationed there, and now “[t]he militiamen had a stronger sense than at any time

in the past four months that they were defending something that was their own.”44

A dissimilar social background, in contrast, could fuel mistrust. For example, class divisions persisted and interfered with trust. Aroca, the young Anarchist student and militiaman from Barcelona, recalls being treated with skepticism as a student. He was almost prevented from joining because the man known as “el Murcia,” who examined his Juventudes Libertarias credentials, suspected him of being a rightist since he looked like a cleric and not like a worker.

El Murcia was “a typical example of a case that was repeated innumerable times in the

Militias:… of the illiterate who, with a happenstance position of authority, takes pleasure in humiliating all those who know how to read and write.” El Murcia jibed him further about being a dilettante student rather than someone who had to work from age seven. Once in the ranks,

Aroca discovered that class tensions undermined the unit, provoking mistrust and accusations.

“On one occasion, I myself was labeled a fascist by a comrade who did not brook my bourgeois custom of cleaning my teeth.” 45

Thus the militia period offers instances of impressive cooperation, bolstering the willingness of combatants to fight and to maintain resistance in the face of depressing odds.

However, there were plenty of exceptions: units with opportunists and units divided by factionalism had severe difficulty in engendering norms of cooperation, and instead were beset by mistrust. Ultimately, however, the fact that, at bottom, the men of the militias were volunteers meant that many did share a desire to fight, and offered the chance of recognizing that that was the case. When the Republic shifted to militarization and conscription, it gained many soldiers

44 Ibid., 53–54. 45 Aroca Sardagna 1972, 20–21, 27. 179 who lacked such a desire, and in any case made it more difficult to recognize a common aim if it did exist.

5.3. Collective Action in the Popular Army

With the creation of the mainly-conscripted Popular Army, the efficacy of norms of cooperation was considerably reduced. The army leadership and other soldiers had a strong belief that conscripts did not place much value in their units’ success—a justifiable sense.

Indeed, desertion tended to be concentrated in groups of conscripts. Lt. Col. Juan Perea, the head of the 5 th Corps of the Republican Army, reported “repeated defections” on the part of soldiers in his 156 th Battalion, 138 th Mixed Brigade, in June 1937 at the Guadalajara front. That unit, he said, suffered from low morale, because they were “forces that came from the last call-ups almost in their entirety.” 46 Army leaders also recognized that such combatants could influence

others to desert, convincing them that the cause was lost or that the unit would not hold together

under fire—that is, demoralizing them. Deserters themselves were a source of demoralization,

serving as a signal to others of disaffection within the unit. An analysis of desertion written up by

the 19 th Army Corps in December 1938 serving in eastern Spain noted desertion to the rearguard

as a demoralizing force both in the rearguard and in the army itself. 47 When men feared that

others would desert, they were themselves much less inclined to serve, believing that to do so

would only increase their risk and make their survival less likely. For example, as Asturias was

falling in October 1937, the leadership in that region concluded that defections and losses meant

that the fight was lost, and planned their escape; men who remained were, in turn, demoralized

46 Corral 2007, 158–159. 47 Ibid., 194. 180

by the rumours that their leaders were about to flee. 48 With conscription bringing in men who

had no interest in fighting, there was a shrinking basis for reciprocal collective action in support

of the fight.

Beyond the influence of desertion on other soldiers, demoralizing talk was policed

because of its potentially pernicious effect on others, including otherwise loyal men. When men

deserted together, their commanders would attempt to determine whether one, through his words,

could have influenced the other to leave. For example, in the report of three soldiers defecting

together from the 99 th Mixed Brigade on 2 February 1939, it was noted that one of them had been a defector from the Nationalist side in the first place. Because he faced the death penalty at the hands of the Nationalists if he were to return, the report supposed it likely that he was an agent provocateur deliberately sent to demoralize Republican troops. Therefore, the defection of the other two men—who had been “of great confidence and of impeccable conduct” up to that point—must have been due to his demoralizing influence, the report concluded. 49 Responding to

the potential of demoralizing talk to encourage desertion, the Republican command issued orders

to prevent such talk. One day after the fall of Teruel in late February 1938, the Minister of

Defense, Indalecio Prieto, recognized that men had abandoned their positions without resistance,

and orders severe punishments handed out for “any words that are defeatist or can demoralize

commanders and troops.” 50

There was thus a clear concern that demoralized sentiment could spread even to relatively committed combatants, because they could not rely on their comrades. In early 1937, the collapsing Córdoba front was plagued with low morale. The Republican leadership faced a

48 Seidman 2002, 153. 49 “Desertores,” 1 Cuerpo, 10 February 1939, Centro Documental de la Memoria Histórica [CDMH], Salamanca, Serie Militar [SM], caja 421, folios 74-81. 50 Corral 2007, 198. 181 dilemma about whether to reinforce the militias fighting there: the General Staff “feared that assistance to Republican forces in the south would create a vicious circle: to send resources to demoralized troops might mean throwing good money and men after bad, but to refuse aid would lead to further demoralization and flight. The general staff realized that battle desertions were contagious. To restore discipline, it absolutely insisted upon the disarming of defective units and punishment of the disobedient. Their weapons would be turned over to more reliable forces in the area.” 51 This approach was repeated later and turned into a general policy by Prieto in June

1937: “When the Leaders of the Army consider that a Unit, for its low morale, insufficient

instruction, or for repeated acts that reveal inaptitude for combat, does not offer the required

guarantees to be employed in combat, they can proceed to disarm [that unit], dispersing its

components to the remaining Units of the Army.”52

Dissolving units was a response to extremes of demoralization, however. In less extreme

cases, in contrast, the army believed that the presence of committed troops could reinforce the

fighting will of others. Lt. Col. Perea, head of the Republic’s 5 th Corps, whose difficulties with

conscripts were noted above, asked to swap some officers, NCOs and privates with a “unit with

high morale and spirit, who by their conduct awake the enthusiasm and emulation of the

reluctant,” requesting a transfer specifically with the 90 th Mixed Brigade. 53 In late 1937, the 148 th

Mixed Brigade on the Teruel Front experienced a similar problem, and its officers came to a similar conclusion, that it could be addressed by mixing veteran anti-fascists in with new recruits. Indeed this was a generally popular method to “promote a combative spirit.” However, as Michael Seidman notes, “Although this method could be effective in boosting morale, it ran the inverse risk of contaminating good soldiers who might begin to imitate the unenergetic and

51 Seidman 2002, 75. 52 Corral 2007, 155. 53 Ibid., 158. 182

selfish ways of the uncommitted.” 54 Both the positive influence of the committed and the

negative influence of the uncommitted are consistent with a norms-of-cooperation approach. The

greater the presence of the committed, the more each soldier would be influenced to remain; the

greater the presence of the uncommitted, the more each soldier would be induced to desert. The

dissolution of extremely demoralized units and the influx of more committed soldiers in less

extreme cases suggests that the Republican command generally believed that there was some sort

of tipping point or critical mass between these two influences: if there were enough committed

soldiers, the remainder could be induced to cooperate, while if there were too few committed

soldiers, those few who were committed could be induced to desert. That this was a dilemma at

all can be understood by the argument that the mutual influence of soldiers could be a positive or

a negative one for cohesion.

Beyond these general changes with conscription, competition among various political

factions undermined a sense of common purpose to the detriment of cooperative action. Once the

Popular Army was established, political competition continued to cause tension within the ranks.

Soviet support for the Republic, according to many, was leveraged to the benefit of the

Communist Party. The other parties believed the Communists were able to secure high military

commands for their officers, and to bypass restrictions on newsprint so that they could flood the

front with their propaganda. 55 The political commissars, responsible for political education and monitoring morale within units, were closely associated with the Communists in much of Spain, and were regarded as a catspaw of Communist influence within military units. Prieto, the

Socialist Minister of Defense, sceptical of the Communists, eliminated commissars at levels below the brigade, tried to replace Communist commissars with Socialists, and gradually shifted

54 Seidman 2002, 159. 55 Fraser 1979, 462. 183 the monitoring of political sentiment to the Servicio de Investigación Militar (SIM. discussed in the previous chapter). But the competition for control over the political monitoring of the military just shifted to the SIM. Much of this competition remained at a rarefied level, and in many ways may not have affected life at the front. But the political commissars and the SIM both maintained networks of vigilance among the rank and file. Within these systems of monitoring, men were encouraged to denounce their fellows. 56 And these accusations, from time to time, could gain a political inflection: when a CNT-affiliated soldier was prosecuted for having deserted, the CNT, in turn, would level accusations of persecution at the Communists. 57 Thus

factional competition and mistrust further undermined cooperation.

Beyond making it more difficult for many other men to fight and to survive, the presence

of uncommitted soldiers changed the nature of friendship and social homogeneity among them.

Whereas, in the militia period, a network of soldiers from the same background could bolster

collective action by allowing them to recognize deeply held shared interests, such ties would

have this effect less and less often as more soldiers were uncommitted. In fact, social

homogeneity could allow individuals to recognize others’ lack of commitment. For example,

Juan Brines was well-known as a Falangist in his hometown of Simat de la Valdigna in Valencia

province, and fled to the city of Valencia soon after the start of the war. Drafted in April 1937,

however, he joined the new Republican army. “At the Valdepeñas barracks…I had the

misfortune to meet the son of the mayor of my village. He told the commander that I was a

fascist and that I needed to be killed then and there.” He only survived because the commander

required him to repeat the denunciation in front of Brines himself, which, out of fear of reprisal

after the war, he was unwilling to do. Thus the hometown connection only enabled mistrust: it let

56 Seidman 2002, 159. 57 Graham 2002, 374. 184

this mayor’s son know that Brines clearly opposed the movement’s goals, and could hardly have

helped the two of them to fight side by side had they had to do so.58

Indeed, such connections might be worse than no help: they might facilitate desertion.

Because secrecy was at a premium, and men would be in great danger if their intentions to desert were known, desertions were most frequently done alone. 59 However, desertion could generally

be easier in cooperation with another if this were feasible. Because soldiers were rarely alone,

often patrolling or keeping watch with others, they would have to take advantage of rare lapses if

they were to get away on their own. Coordinating with another could thus make it easier to

survive: a hopeful deserter in a two-man post or patrol could much more easily cross the lines if

he convinced his partner to cross too, for example. In turn, because it was so crucial to a

prospective deserter that he not be found out, such soldiers were typically inclined to rely only

on men they knew well. Mutual recognition was, in general, easier among men who had more

longstanding connections. 60 There are thus frequent instances of men from the same hometown

deserting together. For example, two men from the 7th Mixed Brigade, 1 st Army Corps, defected together on 2 February 1939. They were both married peasants from Boniches, a village in

Cuenca province, 34 and 35 years old. Their personal connections to each other may have made it particularly easy to trust each other. A few days before, in the 99 th Mixed Brigade, two men

from Chulilla in Valencia province defected within one day of each other, and the report of the

second man’s defection speculated that they could have made a pact to defect or that the first

man’s defection had convinced the second to switch sides too. In another instance from the same

brigade, four men defected on 31 August 1938. Two were from Belmonte, Cuenca province, two

others from Moncalvillo, also in Cuenca province. One was made a corporal in the field because

58 Corral 2007, 236. 59 Ibid., 37. 60 Ibid., 30–31, 37–40. 185 of the lack of NCOs, and the four of them were on guard duty together when they crossed sides.

The report of their defection mentioned that they had been seen having a discussion together two days prior, but as all of them had observed good conduct beforehand, this discussion had not raised any suspicions. 61 Thus the connections among soldiers in the first days of the militias, which had enabled militiamen to recognize the commitment of others and to forge a bond of trust among them, now failed to serve such a purpose. In fact, from time to time they enabled desertion and defection.

Indeed, friendships with deserters were a common source of interest in monitoring within units. Desertion reports from 1 Corps asked whether the deserter had any suspect friends within the unit. When a man deserted, his friends within a unit came under immediate suspicion, and some were quickly placed under surveillance. In more than one instance, reports from 1 Corps indicate that this vigilance of friends of defectors enabled the unit to shoot these soldiers when they later attempted to defect themselves. 62

Thus the shift to a conscripted army, in which most men did not have a preference to fight, undermined the first requirement for reciprocal maintenance of cohesion: the requirement of a common aim. Trust could not easily be maintained, and social bases of recognizing a shared commitment to fight, such as coming from the same hometown or friendships among soldiers, could now serve as a basis of recognizing a shared commitment to desert.

61 “Desertores,” 1 Cuerpo, 10 February 1939, CDMH, SM, caja 421, folios 74-81; “Desertores,” 1 Cuerpo, 16 September 1938, CDMH, SM, caja 421, folios 181-183. 62 “Desertores,” 1 Cuerpo, 15 October 1938, CDMH, SM, caja 421, folios 29-31; “Desertores,” 1 Cuerpo, 10 February 1939, CDMH, SM, caja 421, folios 74-81. 186

5.4. Collective Action in Nationalist Spain

The insurgent armed forces, no less than those fighting for the Republic, confronted problems of collective action. They had to manage the possibility that their combatants had different goals. If soldiers shared the common aim, norms of cooperation could emerge and help to maintain cohesion and the fighting will of their forces. However, among men who did not value their units’ success, norms of cooperation would either not emerge or would actually work in favour of desertion or defection. In addition, men who were under a cloud of mistrust would come to believe that they had very little to gain from those around them by acting cooperatively.

There was, in essence, little basis for reciprocity, and hence greater incentives to defect. The mechanisms of cooperation in operation were thus similar on both sides of divided Spain.

As on the Republican side, volunteer forces, where individuals self-selected for service when it was not strictly necessary, witnessed instances of strong cooperation. Volunteers recognized in each other a willingness to fight, so that norms of cooperation could function.

Francisco Gutiérrez del Castillo, a young Falangist fighting in the Guadarrama mountains north of Madrid, remembered a clear instance of inspirational self-sacrifice as a spur to action when his unit took a mountain pass in an early battle: “A Jesuit priest advanced in front of us, carrying the cross, and urging us on. Like madmen we followed him. What bravery he displayed – without touching a rifle! I'm convinced that it was his example that gave us our victory.” 63 The Spanish

Legion had a longstanding institutional history of reciprocity stretching back to its founding in

1920. Legionaries were supposed to be able to call for help under any circumstance with the cry

“¡A mí la legión!” Its founder José Millán Astray worked assiduously to convince his men of his

personal commitment to them, and in this regard his own wounds helped give him an almost

63 Fraser 1979, 116. 187 mythic image. Its subsequent leaders, such as Franco himself and, during the Civil War,

Lieutenant Colonel Juan Yagüe, attempted to continue in this tradition. 64

Within this volunteer context, social connections from hometowns helped to reinforce a sense of collective action. This played the clearest role in the Carlist Requeté units from Navarre, where Carlism was a longstanding political tradition. As mentioned in Chapter 3, the Requeté took great pride in its capacity to recruit whole villages in that province. This capacity, and the resulting mutual commitment and willingness to fight, played a large part in Carlist propaganda:

A leaflet released in 1937 contained a conversation between a Falangist and a Red Beret;

asked who is to be informed should he die in battle, the Red Beret replies: “Tell José

María Hernandorena, of the tercio [unit] of Montejurra, aged 65. He’s my father.” And if

he should prove to be…unavailable? “Then tell José María Hernandorena, of the tercio of

Montejurra, aged 15. He’s my son.” 65

Thus the Carlists played on the sense of a whole community collectively defending its supposed political traditions.

As voluntarism waned and conscription took hold, however, cooperation was as difficult to maintain on the Nationalist side as it was on the Republican. In January 1938, a note from the head of the 22 nd Division to General Gonzalo Queipo de Llano noted, “Each recently organized

unit that arrives at the front for the first time is sifted, producing desertions that are difficult to

avoid.” Fidel Dávila, head of the Army of the North, confronted the same reality in August of

1938 with new conscripts from the Balearic Islands, including ten deserters in one week. They

lacked training and “do not demonstrate sympathy for the Glorious National Movement, in

64 Balfour 2002, 277–281; Jensen 2002; Preston 1993, 26–28; Preston 1998, 69–72. 65 Blinkhorn 1975, 259. 188

contrast with the other troops of this division.” 66 As it suffered losses and replaced its old volunteers, the Legion suffered a decline in the commitment of its troops. This was particularly true on the Aragón front, quiet during late 1936 and much of 1937. Frank Thomas, a British volunteer in the Legion, deserted out of his dismay at the willingness of Spanish Legionaries to live and let live. Such examples from time to time induced Nationalist commanders to be wary of mixing volunteers with conscripts, with the fear that the latter would bring down the fighting will of the former. Yagüe complained that requiring draft dodgers to serve in the Legion, a policy adopted in June 1937 as a deterrent to draft evasion, was “prejudicial” to the Legion. The policy was rescinded in January 1938, with the draft dodgers in question dispersed throughout the military. However, the Nationalist commanders were caught in a dilemma: units that were full of conscripts were more likely to have poor morale and to fall apart. Thus other commanders, such as at the battles of Jarama in February 1937 and the Battle of the Ebro in 1938, preferred to mix volunteers and conscripts so that the former could bring up the morale of the latter. 67 The

inconsistency of these two approaches bore a striking resemblance to the inconsistency in

Republican practice. Mixing and separating conscripts and volunteers were opposite reactions to

a common phenomenon: men tended to respond to those around them.

As on the Republican side, desertion was from time to time facilitated when soldiers

trusted each other. Men were, of course, assigned to monitor each other, so a man’s chances at

desertion often depended on the intentions of another. A corporal in the 105 th Division, deployed southeast of Zaragoza in Aragon, allowed eleven different soldiers serving under him to defect to the Republicans. It seemed that every night he was on sentinel duty, one of his men would cross the lines. Eventually his superiors uncovered this pattern and the corporal himself was driven to

66 Corral 2007, 158–159. 67 Seidman 2011, 52, 55, 74; Corral 2007, 109, 159. 189 defect in November 1937. Friends could therefore help friends desert. For example, two corporals in the 1 st Army Corps were very close, and often talked about deserting together, but

were never assigned sentinel duty together. On the night of 7 August 1937, however, the sergeant

who was supposed to go on duty with one of them suffered an extremely painful insect bite in the

foot, and the two friends were finally assigned together; they did not hesitate to leave. 68 Because of the need for trust in such situations, men engaged in a delicate process of steadily discovering that a compatriot was a leftist, as did five Legion members, leftists from the north, in the winter of early 1938 near Teruel. 69 It was much easier when they served with others whom they knew to

be like-minded: for example, two brothers and another from the same hometown deserted on the

same night, 7 June 1937, in the Zuera sector in Zaragoza. 70 Trust among soldiers was therefore

not necessarily an aid to limiting desertion. Frequently, trust served as the basis for desertion

itself. Social homogeneity could therefore have either positive or negative connotations for a

unit’s desertion rate. It could allow the recognition that other men shared the unit’s common aim;

a recognition that they actually did not; or even the recognition that another was like-minded in

trying to leave the unit.

5.5. Conclusion

The qualitative evidence in this chapter confirms the importance of norms of cooperation among soldiers, and offers support for the argument that a shared aim is a vital precondition for such norms to take hold. However, this evidence is principally anecdotal. I provide a more rigorous statistical test in Chapter 7.

68 Corral 2007, 30–32. 69 Fraser 1979, 465–466. 70 Corral 2007, 214–215; Seidman 2011, 239. 190

This chapter does not explore sufficient conditions for the emergence of norms of cooperation, only a necessary condition (a common aim) and a facilitating factor (social homogeneity). It should be clear that I do not consider that norms of cooperation will always emerge given a common aim, or even given a common aim and social homogeneity. There is no guarantee, for example, that some combatant will give an example of courage to inspire those around him. Sufficient conditions for such norms in general require more investigation. But the existence of common aims and social homogeneity do, I argue, offer analytical leverage.

The evidence in this chapter indicates a new way of integrating the grievance approach to understanding civil war with the collective action approach. A greater commitment to a common aim probably does, other things equal, make a combatant more likely to continue to serve rather than desert. But its principal effects go beyond that one individual. Since each combatant affects those around him, the degree to which combatants share the group’s common aims conditions which approaches are available to keeping soldiers fighting. Among soldiers who share common aims, the collective action problem as classically conceived holds, with its full array of solutions: both norms of cooperation and control. Without common aims, however, the collective action problem is not the issue, and norms of cooperation are not really an available approach.

Collective action and group interests are thus far from opposing viewpoints, the way the “greed and grievance” account of civil war presents them. The collective action problem, in fact, gets its very analytical force from the existence of a group aim, and the difficulty in really achieving it.

The similarity between the Nationalist and Republican sides is striking. On both sides, there were examples of heroic self-sacrifice inspiring courage in their fellow militiamen; there were cases of norms of lax behaviour setting in among men with little interest in fighting; common communities helped to maintain a sense of fighting for home among volunteers on each

191 side; and social commonalities also facilitated desertion among men who did not share their side’s cause. The similarity in the mechanisms suggests two possibilities for interpreting the overall, macro effect of norms of cooperation on the outcome of the Spanish Civil War. On the one hand, Nationalist Spain could take advantage of somewhat greater commonality in aims.

Franco’s most committed units, such as the Legion, were also very often their most militarily talented, whereas the Republic faced a tradeoff: its most committed combatants (militiamen) generally had very limited experience, whereas its career officers and other military men were under deep suspicion. Further, the relatively lower degree of political factionalism may have made it easier for, for example, Carlists, Alfonsist Monarchists, and Falangists to serve alongside each other in a way that was perhaps more difficult for the mutually suspicious factions of the

Republic. In general, however, the most striking difference between the two sides as regards desertion was the significant Nationalist advantage in control, outlined in the previous chapter. It is therefore unclear that norms of cooperation were a decisive difference between the two sides.

But whatever their impact on who won and lost, it is clear that norms of cooperation, when they operated, could keep some soldiers fighting rather than deserting.

192

Chapter 6 Desertion and Control of Hometowns in Santander Province

6.1. Introduction

The three previous chapters have given qualitative case-study evidence for the theoretical synthesis that I outlined in Chapter 1. Armed groups in the Spanish Civil War varied by the degree to which their combatants shared their common aims. I argue in Chapter 3 that this variation emerged from their recruitment policies and their factionalism. This variation then set up the effectiveness of different approaches to limiting desertion. In the last chapter, we saw that in uncommitted units, it was very difficult for norms of cooperation to take hold. However, as I argued in Chapter 4, control could still have an important effect even when combatants were uncommitted, provided disagreements were not too severe.

The account of desertion in the Spanish Civil War that I have outlined gives an important

role to control. The militias in Republican Spain varied in their degree of control, and over time

the Republic improved its capacities for control in both the rearguard and the front line. This was

a difficult process, proceeding in fits and starts. But I argue that control was vital to the ability of

the Republic to prevent desertion, particularly as the proportion of uncommitted soldiers

expanded through conscription. The Nationalist side carried an advantage over the Republic in

preventing desertion largely because of its more effective control from an earlier stage. Thus far,

we have seen qualitative evidence that suggests that deserters needed to actively evade capture,

and that whole political groups were intimidated against rising up in rebellion by the coercive

193 power of the Republic and of the Nationalist side. But thus far the evidence remains qualitative and suggestive.

In this chapter, I conduct a statistical test of the ability of control to prevent desertion. In particular, I focus on control in the rearguard. Other accounts of desertion in civil wars have also discussed control,1 but this chapter is innovative in that it goes beyond control within armed

groups and discusses the control that an armed group maintains over soldiers’ hometowns. I find

statistical evidence that an important correlate with desertion rates was control of a soldier’s

hometown: the easier that a town is to control, the more difficult it is for a soldier from that town

to desert. Efforts to prevent desertion thus go well beyond the faction itself, and extend into

civilian life.

I examine control by focusing on terrain, especially altitude and steepness. Difficult terrain is difficult to police. However, difficult terrain is also associated with political and economic marginality. Rough terrain might be correlated with high desertion rates because of a stronger preference not to serve, rather than a lack of control. I do find evidence that soldiers from difficult-terrain locations had different political preferences and economic interests than soldiers from easier ground. However, I also demonstrate that the high desertion rates of difficult-terrain communities were not merely an artifact of marginality. Net of these effects, and in light of considerable qualitative evidence, it appears that terrain has its impact through the most obvious of mechanisms: it is easy to hide in the hills. This chapter develops its argument via a statistical analysis of data drawn from Santander province (now the autonomous community of Cantabria) in the Spanish Civil War. It combines characteristics of soldiers and characteristics of their hometowns in a multilevel model. The chapter supplements the statistical analysis with qualitative evidence confirming that terrain was an important predictor of the

1 Gates 2002; Johnston 2008. 194

ability of the Republic to control a location, and that soldiers considered the possibility of control in terms of terrain when deciding whether or not to desert.

6.2. Desertion, Control, and Hometowns

As discussed in Chapter 4, control can be wielded within armed groups, preventing soldiers from deserting by monitoring them and threatening them with punishment if they leave.

But control can also vary over soldiers’ hometowns. First, deserters often head back to home districts as a matter of first resort, relying on families or friends for help and support, including shelter and food. Second, deserters’ families often bear punishment for their desertion. If a deserter is not himself caught, an armed group may decide to resort to other coercive tools at its disposal, creating a deterrent. The punishment of families also means that hometowns are important for defection as well as for desertion proper; if a soldier switches sides, his family can still be punished. If credible, that threat is likely to deter defectors.

Armed groups in civil wars vary in their ability to control the rearguard. What factors help ensure local control? There may be a role for positive cooperation. Armed groups in civil wars may vary in the degree to which they have cooperative relationships with civilians, as

Jeremy Weinstein argues. Such cooperation can provide armed groups with various different kinds of support, including supplies, information, and shelter. 2 And it can also help with the

finding and capture of deserters. In the U.S. Civil War, for example, the civilian population on

the Union side bought in to the cause to a large extent, such that deserters faced social sanctions

rather than social rewards when they returned home.3 In Spain, a group of women in the town of

2 Weinstein 2007. 3 Costa and Kahn 2004. 195

Guijuelo, Salamanca province, wrote a letter to General Franco himself, alerting him to the presence of several young men who were draft dodgers and deserters. 4

However, beyond cooperation, finding deserters in particular rearguard locations may be

a matter of control. Just as states can vary in their ability to compel citizens’ compliance, 5 armed groups in civil wars can vary in this ability as well. Sheer muscle—the use of large numbers of capable forces in investigative duties—may be one reason for this variation. Forces with a large number of troops covering a smaller area should be better able to control towns, and hence prevent desertion, than forces with a small number of troops covering a larger area. For example, in the U.S. Civil War, Confederate efforts against desertion improved dramatically after General

Gideon Pillow formed a special military branch to try to find deserters, rather than relying on the ill-equipped and undermanned Conscript Bureau. 6 And, as argued in Chapter 4, the re- monopolization of force in the hands of the uniformed security services was an important contributor to the re-establishment of order in the Republic’s rearguard.

One very important structural feature of control is geography, specifically difficult terrain. It has historically been quite difficult to exercise control over high-altitude locations. It is, of course, easier to hide in the folds of the earth than on flat land. 7 In addition, the classic loss-

of-strength gradient, based on the difficulty of transporting men and the exponential difficulty of

transporting supplies over distance, 8 is compounded by the friction of terrain. 9 Thus heading to the hills is close to a ubiquitous feature of desertion. Confederate deserters hid in bands in the hills to resist capture. When the Confederate War Office was analyzing the problem of desertion,

4 Corral 2007, 107–108. 5 Mann 1986; Migdal 1988; Tilly 1992; Herbst 2000. 6 Lonn 1928, 53. 7 Hence the use of terrain to evade firepower in modern warfare. Biddle 2004. 8 Boulding 1962. 9 Scott 2009, 43–48. 196

it noted that “[t]he condition of things in the mountain districts of North Carolina, South

Carolina, Georgia, and Alabama menaces the existence of the Confederacy as fatally as either of

the armies of the United States.” 10 Similarly, civilians have frequently headed to the hills to avoid the depredations of warfare. In massive numbers, civilians in Southeast Asia long took to the hills to flee conscription and the pillaging of marauding armies, and more recently in response to Myanmar’s counterinsurgency campaign against the Karen.11 The increasing ability

to cross rough terrain, especially through road-building, is therefore a vitally important part of

state-building conceived of as the projection of control over peoples at a distance. 12

Those best able to take advantage of the opportunities to hide should be those who are

actually from hill communities. Possessing an unparalleled knowledge of the terrain, soldiers from the mountains will be able to exploit the opportunities for physical concealment that mountains provide. Moreover, hill peoples can often evade monitoring through social practices that help maintain social “illegibility,” on top of physical concealment. Nomadic pastoralism, for example, allows hill peoples to evade the state and still eat, since they are not bound to one place for food.13 The friction of travel in the hills also tends to make language, identity classifications

and structures of social control extremely heterogeneous and diverse. In consequence, lowland

state officials are often unable to understand the social relationships that dominate in the hills,

such as who will be loyal to whom; hence state efforts to categorize and classify peoples. 14 Such

“illegibility” can allow deserters from hill communities to evade the agents of the armed group,

who will often not know where to begin looking or whom to ask. Finally, unlike lowland

soldiers, hill-country soldiers do not have to decide between receiving help from their families

10 Lonn 1928, 25. 11 Scott 2009, 94–95, 146–149. 12 Herbst 2000. 13 Scott 2009, 184. 14 Anderson 1991; Hobsbawm and Ranger 1992; Scott 2009. 197 and taking to the hills; they can do both. And they can anticipate that they and their families will be able to hide relatively easily if need be, protected by their knowledge of the physical and human terrain.

This basically intuitive argument has important implications for theoretical debates in the literature on civil wars. The chapter provides a reminder that when soldiers join armed groups in civil wars, they do not leave everything behind. Instead, they maintain important connections with their hometowns. They remain embedded within communities that they value. The embeddedness of soldiers in civilian life is already a well-established point in the context of irregular warfare. In that context, the very nature of warfare, with guerrillas hiding among a civilian population, blurs the distinction between military and civilian spheres.15 This chapter, by examining desertion in the context of a conventionally fought conflict (the Spanish Civil War), suggests that the intermeshing of the civilian and the military realms is actually a more general feature of civil wars, whether irregular or conventional.

Control is not the only important factor in preventing desertion. In Chapters 5 and 7, I make clear the importance of norms of cooperation, for example. However, in this dissertation, I depart from approaches to desertion that focus only on motivations or only on cooperation. I agree with critics of these views that control is significant, particularly when, as in the mainly conscripted Popular Army, common preferences are wanting. The analysis of the impact of terrain on desertion allows for a relatively clear test of the impact of control, beyond cooperation.

Past work has found a link between terrain and desertion, but explains that link by differences in motivations. Kathleen Giuffre attributes the high desertion rates of upland North

Carolina soldiers to their disconnection from the slave economy, and thus to their lack of

15 Petersen 2001; Kalyvas 2006. 198

motivation to fight to defend that system. 16 This is certainly a plausible explanation. However,

one kind of motivation is easily countered by another: from a different point of view, poor

highlanders, disconnected from the economic centre, may actually depend more on the wages

offered by rebellion than comparatively better-off lowlanders.17 Rather than focus on particular motives, my approach focuses on the opportunities presented by a lack of local control. The theoretical work above suggests that the high desertion rates of uplanders may have instead to do with their ability to hide, rather than their unique goals.

Control is not the be-all and end-all of preventing desertion. In the next chapter, I go on to indicate the importance of norms of cooperation, and their ambiguous, conditional effects, giving further evidence from the same setting as this chapter. However, as the present chapter and Chapter 4 argue, control is essential, especially when combatants often do not share the armed group’s aims and these norms of cooperation do not avail.

6.3. The Setting: Santander Province, Spain

This chapter focuses on Santander province (now Cantabria), between Asturias to the west and the Basque Country to the east on Spain’s northern coast (see Figure 6.1). Santander followed the general pattern of the Republic in some respects, but not others. 18 Cut off by the

Nationalist zone from the bulk of the Republic, with its capital first in Madrid and later in

Valencia, Santander developed an autonomous government and armed force, as did the Basque

Country and Asturias next door. However, unlike the other two areas, Santander’s war effort

followed the Republic’s lead rather closely. Beginning with a loose militia system based on

16 Giuffre 1997. 17 Collier and Hoeffler 2004. 18 The following account draws on Salas Larrazábal 1973; Saíz Viadero 1979; Martínez Bande 1972; Martínez Bande 1980; Solar 1996; Solla Gutiérrez 2005; Solla Gutiérrez 2006. 199

Figure 6.1. Location of Santander Province in Spain

political parties and the remnants of the uniformed security forces that remained loyal to the

Republic, it gradually developed a regular, conscripted army, the Cuerpo de Ejército de

Santander (CES). In this, the autonomous government followed the letter of the directives laid

down by the capital in the creation of the Popular Army of the Republic. Moreover, its formal

political forces closely mirrored those of the Republic in general in the period 1936-1937, with a

Socialist-dominated government; an uneasy partnership with a weaker but still highly active

presence of the Anarcho-Syndicalist CNT; and a small but rising Communist Party whose

increasing political strength raised suspicions among the other factions. As elsewhere, political

power fragmented among local committees at first, and it was a long and difficult process to

reassert it at the centre. In some ways, therefore, Santander province was the Republic in

miniature.

But in one vital respect it was wholly different from much of the Republic: the prewar

strength of conservative political forces. Of all Spain’s provinces that were part of the Republic,

Santander had returned the one of the largest proportions of votes for right-wing parties, and in

fact a right-wing majority, in the elections of February 1936. 19 And it had a weaker union

presence than most other Republican provinces—certainly much less than the key union centres

19 Linz and de Miguel 1977, 43. 200 of Madrid, Barcelona and Asturias. Despite its conservatism and the weak presence of organized left-wing forces, it was kept for the Republic because of a series of fortuitous circumstances.

During the coup attempt, its location, between the anti-Franco Asturias and Basque Country and with a mountainous southern border with Nationalist territory, meant that the local Nationalist forces had little prospect of external reinforcement. The local military commander vacillated while Republican officers demonstrated greater agility in seizing key communications sites during the coup. And the fact that the Republic’s forces held on to a warship in the capital harbour led to an early threat of bombardment of the garrison if it went over to the Nationalists.

After the coup attempt, a stable front was quickly set up on Santander’s southern frontier, which remained much quieter than the Asturias and Basque Country fronts. However, this front was less a continuous line than a rather porous series of positions. Santander’s militias patrolled this border and fought in the two neighbouring regions. The various militias were gradually brought under the single command and organizational structure of the CES, and the draft began in earnest. In the fall of 1936, the Republic attempted offensives across the whole northern front to relieve the siege of Madrid, but they failed; Santander’s contribution, towards Vitoria, was a particular fiasco, prompting the first major desertion problems in the CES. The northern front stabilized until April of 1937, when the Nationalist forces, with their Italian and German allies, carried on with the conquest of the Basque Country and Santander troops conducted another failed attack towards the key Burgos-Vitoria highway; again, its offensive was plagued with desertion. Bilbao, capital of the Basque Country, fell at the end of June; at this point it was clear that it was only a matter of time before Santander faced a Nationalist invasion. This began in mid-August and was over within two weeks: the CES’ resistance collapsed in the face of the

201

Figure 6.2. Estimated rate of desertion by municipality (% of soldiers), Santander province Results are given by quintiles. Source: author’s data.

assault. Some CES forces remained in Asturias to fight on until October; others took to guerrilla warfare, which lasted sporadically until the late 1950s; but most surrendered.

6.4. Hypotheses and Control Variables

Over the course of this fighting, desertion rates varied widely by soldiers’ hometowns.

Figure 6.2 maps the rate of desertion—deserters divided by the weighted estimate of soldiers

from the same municipality. What explains this variation? I suggest that an important reason why

municipalities varied in desertion rates is that the faction varied in the degree to which it could

impose its control on those municipalities.

A challenge in measuring the degree of control over different municipalities is that, in

some respects, this control is likely to be endogenous to desertion. This is particularly true of the

numbers of security forces able to search for deserters in a particular location. Obviously, the

armed group is likely to deploy its security forces where they are needed: to locations that are

202

particularly problematic in terms of desertion. Reports of large groups of deserters, such as an

April 1937 report of twelve deserters from the municipalities of Solórzano and Entrambasaguas,

would lead to the deployment of security forces for a search (in this case, by a unit of twenty).20

This means that comparisons over space in the deployment of forces may not be particularly effective at indicating the underlying likelihood of capture, since forces will tend to shift to where they are needed. Comparisons over time may work more effectively, tracking the evolution of the overall capacity of the state, but I could find no strong evidence of over-time variation on this point in Santander province.

I turn instead to a strongly exogenous factor in control over a hometown: its terrain. I measure this with two indicators: altitude and steepness. Centres of power are often located on low ground, and this was certainly the case in Santander province, with its capital in Santander city on the coast and with its left-wing Socialist-dominated government based in its lowland industrial core. Altitude implies difficult access, dramatically increasing transportation costs and thus the costs of projecting power over distance. Moreover, rough terrain can create places to hide, and makes transportation even more difficult. As I outline above, my chief interpretation of terrain is its ability to allow deserters and their families to evade capture. I therefore hypothesize as follows:

Hypothesis 1. The higher the altitude of a soldier’s home municipality, the higher his likelihood of desertion.

Hypothesis 2. The steeper the terrain of a soldier’s home municipality, the higher his likelihood of desertion.

Of course, communities on the geographic periphery may often have very different social, economic, and political conditions than the centre. Mountainous terrain may therefore be

20 Radio Comunista de Secretario Político and Agrupación Socialista de Riaño to Frente Popular Santander, 30 April 1937. Centro Documental de la Memoria Histórica [CDMH], Salamanca, Serie Político-Social [PS] Santander L, caja 412, carpeta 22, expediente 22. 203 associated with desertion rates not because of opportunities to desert, but because people’s economic and political preferences differ substantially to those of the centre. Marginality may lead them to believe that the civil war is just not their fight; it is driven instead by lowlander agendas. This is Giuffre’s argument in attempting to account for higher desertion rates in North

Carolina hill country. 21 If true, this would tend to support a strictly motives-based approach and

not a control approach. Indeed, in Santander province, the development of the late 19 th and early

20 th century was very much concentrated in the lowlands: in the ports of Santander and Astillero

and the surrounding mining basin. Although railroads permitted closer integration of rural

communities into the broader economy, with the shipping of livestock and cattle, 22 many

mountainous areas did not have easy access to highways or railroads. 23 Since mining and

industry, the two major early targets of left-wing unionization, were largely absent from

mountainous districts, the Left had little political appeal. 24 The Socialists’ efforts to unionize

agriculture were belated and of limited success. 25 Literacy rates indicate the degree of

development in different areas, since literacy enables a wider array of economic opportunities.

These varied quite substantially by municipality, from under 50% to over 80%. According to the

1930 census, Spain’s overall literacy rate in 1930 was 56%; it was many years away from

universal literacy. 26 Santander’s literacy rate was 72%, the highest in the country. 27 Political support for the Right, as indicated by its vote share in the February 1936 elections (five months before the start of the war), is a straightforward indicator of the degree of local political opposition to the Republican cause.

21 Giuffre 1997. 22 de Puente Fernandez 1992. 23 Obregón Goyarrola 2007, 30–31. 24 Gutiérrez Lázaro and Santoveña Setién 2000; Toca 2005. 25 Obregón Goyarrola 2007, 36–37. 26 Dirección General del Instituto Geográfico, Catastral y de Estadística 1930. 27 Solar 1996, 74. 204

Now, economic and political marginality may still be consistent with my control approach, since marginality may have served to reduce the ability of the Republic to monitor the countryside. Popular Front organizations like unions were central to monitoring in the rearguard during the Civil War: a soldier’s union local would often report to the military headquarters on his political leanings, and the local committees of the Popular Front ran the local governments that were tasked with finding deserters. If the Left had very limited pre-war support in a place, this may have reduced the Republic’s ability to monitor in the long run by reducing access to the local information they would need in order to investigate well. However, economic and political marginality cannot stand on their own as evidence of my control story, because they are both also consistent with a theoretical perspective focused on motives rather than on control.

I therefore examine literacy rates and political support for the Right as control variables for the major hypothesis about geography. If high altitude and steep terrain have a substantial correlation with desertion even after controlling for these two variables, it indicates that terrain has a relationship with desertion goes beyond its association with people’s motives. I suggest, then, that the best explanation is the most intuitive one: that rough terrain makes it easier to hide.

That is an interpretation consistent with a theory based on control. If altitude and steepness have no correlation with desertion after controlling for literacy and right-wing support, this could be due either to the control properties of economic and political conditions, or to people’s different motives. There would not be enough evidence to warrant accepting a control story over a motives story.

Finally, an indicator of lack of control may be the local rate of violence. This is fundamentally ambiguous, however: lack of control can allow for more violence to occur, or violence could be the result of an attempt to impose control on a place. In the context of the

205

Spanish Civil War, it is often argued that violence against civilians was the consequence of lack of control over undisciplined militias ( incontrolados ) in the Republic, and a consequence of

deliberate state policy in the Nationalist camp. 28 While I generally agree with this interpretation—violence in the Republic was wielded by multiple local hands—it is also certainly the case that, over time, the central state wielded considerable violence in the rearguard as well, often to find deserters themselves. Indeed, violence may emerge as armed groups try to move away from this lack of control. Cohen, Brown and Organski capture this ambiguity by terming state building “the violent creation of order”; they find, at a macro level, that state building produces violence in the first instance but long-run order once the state is well- established. 29 Balcells, in a statistical study of violence in Catalonia during the Spanish Civil

War, finds that rates of local violence were correlated with pre-war political competition, so that closer election results predicted higher rates of violence. 30 This may indicate the pursuit of local rivalry and competition outside of the Republican leadership’s interests, and hence a lack of

Republican central control; or, from another point of view, it may indicate efforts to eliminate pockets of resistance and hence to assert central control. Because local violence may be an ambiguous indicator of state control of a location, I include a measure of local violence as a statistical control, rather than to test a theory-derived hypothesis.

Soldiers from rough-terrain communities may also desert less often because of differences in individual characteristics. They may be less likely than their lowland comrades to want to serve in the first place. They may more frequently be agricultural workers, with their families depending on them for survival. Rural soldiers may be more likely to be married and to feel the pull of home more strongly than others. I conduct a multilevel analysis, therefore, to

28 Beevor 1982, 106–107. 29 Cohen, Brown, and Organski 1981. 30 Balcells 2010. 206

control for these individual characteristics systematically. I examine soldiers’ circumstances of

joining up—whether as a volunteer or as a conscript, and the date of joining—as a useful

indicator of a general preference for serving. Whether or not a soldier has an official party or

union affiliation may also indicate his general degree of political commitment to the left. I also

include age, marital status, and occupational category, to capture several demographic

characteristics occasionally thought to correlate with a propensity to desert. 31

6.5. Potential Biases from Case Selection

In general, as noted above, Santander province followed the basic patterns of the

Republic, with the major exception of its political conservatism. How does the right-wing character of Santander province bias the results of my analysis? It makes it a most-likely case for a strictly motives-based approach to desertion. Though reliable overall figures on desertion rates in Spain do not exist, Santander province is generally acknowledged to have among the highest.

The province’s conservatism is the most frequent explanation, offered both by the Left and the

Right: it is a claim made by the last Republican commanding officer in the north, General

Mariano Gámir Ulibarri, and by the pro-Franco historians Salas Larrazábal and Martínez

Bande. 32 In the Spanish context, it is strong ground for a strictly motives-based story.

The dynamic of coastal core and mountainous periphery is particularly pronounced in

Cantabria (see Figure 6.3). The geography of the province ranges from coastal plains to high

cordillera. Moreover, as indicated, the heart of the Republican government in Santander province

was the capital and towns in the lowland industrial centre. This suggests the classical pattern of a

lowland state ruling over a mountainous periphery. It is possible that a faction rooted in the

31 Bearman 1991; Costa and Kahn 2003. 32 Gámir Ulibarri 1939, 37; Salas Larrazábal 1973, 358–359; Martínez Bande 1980, 153, 167. 207

Figure 6.3. Relief map of Santander Province Source: Government of Cantabria.

mountains themselves may not have as much difficulty controlling highland dwellers. It also suggests that geography may be more highly correlated to desertion here than in provinces with less variable altitude, so any positive relationship cannot necessarily be extrapolated to other contexts. However, the pattern of lowland authority and highland periphery is common enough because of the wealth and power of lowland urban cores that this analysis should be applicable fairly widely. Within Republican territory alone, the pattern also holds for Catalonia and Aragón, for instance. As well, even where authority emerges from the mountains themselves, it tends to be very limited in geographic scope because of the difficulty of transportation.33 Therefore, even

when armed groups emerge from the mountainous periphery itself, they may have considerable

difficulty in controlling territory and preventing desertion.

The prospect of draft dodging in Santander province complicates an analysis of the role

of control in desertion. People from rough terrain should be better able to evade conscription

than people from flat land, for essentially the same reasons as they are better able to desert. They

33 Scott 2009. 208

would therefore be less likely than lowlanders to make it into my dataset in the first place. In a

conscripted armed force like the CES, then, it is worth asking what impact this selection effect

has on my analysis. Indeed, there is substantial anecdotal evidence of a high rate of draft evasion

in hill districts. 34 In addition, much of mountainous terrain in Santander province is located along

the province’s borders with Nationalist-held Palencia and Burgos provinces. At the beginning of

the war, civilians were able to take to the hills to cross the lines before they even got in to the

army, removing them from the ambit of the army and thus from my dataset. 35 What effect does this selection process have on my data and analysis?

First, one might be tempted to ask why a soldier, especially a probable future deserter, would join at all if it were easy to evade the draft. One answer is that “easy” does not imply

“utterly frictionless,” for low state presence does not imply no state presence. Both successful draft evasion and successful desertion are probabilistic events. An intuitive scenario is one in which a draft-age individual from a town where it is relatively easy to hide intends to evade the draft but fails to do so through chance events: he is simply in the wrong place at the wrong time.

Incorporated into the army, he still sees that there is a good chance for successful desertion— better than for other soldiers—and may decide to desert. The desertion rate is just the conditional probability of desertion, given that a soldier is in the armed group in the first place; and this conditional probability can still be affected by control even if control also affects the likelihood of being in the army in the first place.

But more importantly, the selection process suggests that there are fewer soldiers from easy-to-hide municipalities than there would be if selection was unaffected by control in the first instance. This selection effect actually introduces a conservative bias into my analysis. Think of

34 Obregón Goyarrola 2007; Obregón Goyarrola 2009. 35 García Guinea 2005; Obregón Goyarrola 2007. 209 draft evasion and desertion as two different ways of avoiding military service, at two different points in time. A lack of control allows men more easily to avoid military service, by whatever means. In the counterfactual scenario in which draft evasion has no relationship to control, then all of control’s relationship with the avoidance of military service occurs by means of desertion.

None of it occurs by means of draft evasion. This implies that in my analysis, where both paths

are open, some of the potential effect of a lack of control on desertion goes unmeasured because

of the initial selection effect. Compared to what it could be if draft evasion were unaffected by

control, the effect of control on desertion is thus underestimated by my analysis.

Is the type of soldier who is drafted systematically different under different conditions of control? Quite possibly. Those who are drafted may have different motives, such as a weaker aversion to fighting, than those who evade the draft. In a town where it is easy to evade the draft, those who fight must want to, or at least not be totally averse to it, compared to towns where it is hard to evade the draft. In the latter, proportionally more people will join just because they got caught. This continues to strengthen the conservative bias of my analysis, because it means that those who were sampled from places where it was easy to evade the draft should actually be the most committed (or least uncommitted) people available locally. If even they desert more often than soldiers from a town where people fight just because they got caught by the draft board, then this speaks even more strongly in favour of a control approach.

Now, there may be some countervailing, highly specific motives that arise from the movement of civilians in mountainous terrain prior to conscription. The scenario outlined above, where families crossed the lines at the beginning of the war, may suggest that some high-altitude soldiers were not going home when they deserted but crossing the lines to reunite with families who had already evaded the draft or left the province—indicating a strong motive. This is a

210

highly specific scenario, however. In any case, it would not necessarily challenge the broad

theory at stake, because it is also a scenario in which a soldier’s family is out of the faction’s

control and cannot be investigated or punished for their relative’s desertion, and so it is

consistent with my control-based approach.

6.6. Data and Method

The focus of this chapter, on soldiers’ hometowns, would perhaps be best treated with a hometown-by-hometown sampling method. However, this proved impractical. No complete set of hometown-by-hometown lists of soldiers exists, and so there was no real basis for sampling this way. Instead, therefore, I decided to sample on individuals.

Data on deserters and other soldiers were obtained from the archives located in the

Centro Documental de la Memoria Histórica , formerly the Archivo General de la Guerra Civil , in Salamanca. I began by identifying deserters on the basis of army records. The army used the following definition of a deserter: a soldier who failed to appear for three consecutive roll calls without leave. Where a subsequent report indicated that the initial report was in error—for example, a leave unreported because of a failure in communication—the individual in question was excluded.

I identified deserters from three kinds of document. First, the army compiled lists of deserters’ names, by battalion and month, beginning with a retrospective list in February 1937 that went back to November 1936, and continuing in each subsequent month; the last set of lists was created in July 1937. 36 Second, beginning in January 1937, the army produced individual

reports of desertion, including the deserter’s name, background details, and circumstances of

desertion. The preservation of these reports is unfortunately highly inconsistent: only those with

36 CDMH, PS Santander A, caja 180, carpeta 7; CDMH, PS Santander A, caja 190, carpeta 4. 211 last names beginning A through L, and scattered others, are preserved. 37 Finally, the army produced a card for each of its soldiers indicating his background information. These cards also included notes on his subsequent military career, such as battalion reassignments, leave, hospital visits, and desertion. I compiled every soldier with a desertion report in all of the over 32,000 army cards that have been preserved. Unlike the deserter files, these cards appear to have no serious inconsistency in preservation, aside from under-sampling some military units.38 There

were other deserters that are not included here. Scattered and disorganized military reports refer

to reported deserters who are not elsewhere referenced. As a particularly unfortunate example,

one report from the municipality of Miera indicated a list of 7 men reported by their battalion to

have deserted; only two of them had files and could be included in my dataset.39 There are also

numerous ambiguous cases. For example, there is a case of a soldier from Battalion 130 who

disappeared and, after a long paper trail, could not be found. 40 Time did not allow for a systematic accounting of each of these, unfortunately. A further, more labour- and time-intensive effort is needed to compile a more definitive list of deserters in Santander province.

To preserve as much information as possible about deserters, I included every deserter as defined above, totalling 1,313. This almost certainly understates the true number of deserters. In addition, I took a simple random sample of 1,305 of the remaining personnel files, following a rare-events research design.41

There are at least two validity concerns with this approach to desertion. The first arises

from the time period covered by the data. Reports of deserters only really begin in earnest in

37 CDMH, PS Santander C, caja 26, carpetas 1-4; CDMH, PS Santander M, caja 7, carpeta 12; CDMH, PS Santander A, caja 188, carpeta 4. 38 The large bulk of these files are contained in CDMH, PS Santander A, cajas 1-72. 39 CDMH, PS Santander L, caja 444, carpeta 13, expediente 19. 40 CDMH, PS Santander L, caja 412, carpeta 20, expediente 21. 41 King and Zeng 2001. 212

January 1937. This means that anyone who joined and deserted prior to that date is not captured by my approach. Therefore, any volunteers who remain from the militia period have already failed to desert on multiple prior occasions that are not captured in my analysis. Volunteers in my data set should thus be particularly unlikely to desert. This should exaggerate the differences between conscripts and volunteers. In addition, the Santander Army Corps crumbled in the face of Nationalist offensives over the course of about one week in late August 1937. Because reporting would have become unreliable at the same time, the wave of desertion that apparently occurred is not captured either. My data then capture desertion during a relatively stable period,

January to mid-August 1937. As I argue in Chapter Four, it appears as though the Republic’s ability to control soldiers steadily eroded as the Republic’s soldiers gained additional incentives to desert. In such situations, the ability to control the home front may have eroded as well, as the agents of control fulfilled their duties less and less often. This creates the plausible suggestion that the results of the present analysis do not apply to such situations.

A second concern with this measure of desertion arises from the incentives of those doing the reporting. A commander may have had an interest in telling the General Staff what it wanted to hear. For example, commanders would have wanted to report that desertion rates were low, which may account for the surprisingly low figure of 1,313 deserters, a rate of about 4%, compared to anecdotal evidence that suggests a much higher rate. An underreporting bias certainly existed at a high level: General Toribio Martínez Cabreras, serving in the North at the time, reported to the Republican government on 15 July 1937 that in Asturias and Santander,

“despite certain problems of political order,” the armed forces, now provided with regular leadership and organized into regular units, “improve each day at responding to norms of

213 discipline.” 42 This statement was not at all true of desertion in Santander: in the month in which it was written, according to my data, desertions had steadily increased in frequency and were at a peak. In and of itself, the tendency to underreport desertion does not immediately suggest a systematic bias in the data, just that the baseline rate of desertion was lower than true, and that some members of my sample of non-deserters may actually have deserted.43 However, we can

get at possible under-reporting in another way. According to Pedro Corral, soldiers who deserted,

especially on the battlefield itself, were frequently listed as having disappeared. 44 I therefore also

re-ran the analyses counting the disappeared as deserters. Doing so did not substantially alter the

results.

A graver concern, since it deals with validity and not just reliability, is that commanders might also have wanted to report that the profile of deserters was as the General Staff expected it to be. For example, since the General Staff saw hidden right-wingers as a major problem of desertion, it would find right-wingers who deserted unsurprising. But this would also imply that a report of the desertion of a good party comrade would reflect badly on a commander who could not even keep those individuals in line. The commander’s career could then be jeopardized. He would therefore be inclined to quash such a report: either to not deliver it at all, or to report some other reason why the soldier was not present to be reviewed: a disappearance, for example. This is a plausible logic, and it suggests that the picture of deserters may be artificially biased in favour of the army’s conventional wisdom. That conventional wisdom did include the stereotype that soldiers from mountainous communities—in particular, the Pasiego nomadic-pastoralist

42 General Toribio Martínez Cabreras, “La Lucha en el Norte,” Valencia, 15 July 1937. Archivo General Militar de Ávila [AGMAV], Documentación de la República [DR], Armario 63, Legajo 853, Carpeta 7, Documento 4, pp. 27- 28. 43 This underreporting of desertion is found in other cases as well, such as the Confederacy. Weitz 2005. 44 Corral 2007, 195. 214

group—deserted frequently. However, the conventional wisdom was much more dominated by

the idea that soldiers deserted because they were not attached to the cause.

There may be a countervailing bias from reporting as well, however. The CES pursued

various policies in an attempt to constrain those soldiers who were thought likely to desert. If

such a soldier deserted, it might raise uncomfortable questions about why that soldier was not

under sufficient vigilance in the first place. Consider the case of one soldier from Battalion 115

who deserted on 27 July 1937. An apparent right-winger with no official political affiliation, he

had been placed under “discreet” watch by the captain of his company; the report says that he

was able to defect to the Nationalist front because there were only two sentinels on duty that

night since it was an unimportant front. 45 This suggests an effort to determine how a likely deserter had been allowed to escape, casting aspersions on his immediate superior’s abilities.

Therefore, there is reason to expect a counter-bias from commanders’ reports: the inclination to under -report unsurprising desertions, because unsurprising desertions might be taken to mean

that the commander was negligent. The prospect of validity problems from reporting is,

therefore, inconclusive.

Soldiers’ hometowns were given on their files. I classified these hometowns according to

the 102 municipalities in Santander province. I include only soldiers from within Santander

province (92% of the total), because data on steepness were only readily available for that

province. I examined two measures of terrain: altitude and steepness. Altitude is defined as the

altitude of the municipal seat in meters above sea level. For steep terrain, I constructed an index.

There are data available for all but two of Santander’s municipalities (the geographic survey is

not yet complete) on the proportion of the municipality’s terrain that is between 30 and 50% in

45 Oficial Informador, Batallón 115, to Jefe Sección Información Santander, 30 July 1937. CDMH, PS Santander L, caja 412, carpeta 16, expediente 1. 215 grade, and the proportion greater than 50%. I combined these two measures in a single index of steepness, defined as follows:

(0.3 * proportion of territory 30% to 50% in grade) + (0.5 * proportion >50% in grade)

In this expression, the coefficients equal the minimum grade of the terrain in each category of steepness. The data are given by the Instituto Cántabro de Estadística .46 The municipality-by- municipality results of the February 1936 elections were published in the Boletín Oficial de la

Provincia de Santander .47 Literacy data are drawn from the 1930 census. 48

Data on local violence were compiled by Jesús Gutiérrez Flores. 49 They indicate the

number of deaths in local violence against individuals residing in the municipality. This means

that a death was attributed to a community even if it did not occur there; according to Gutiérrez

Flores, much of the violence in Santander province during the Civil War took place with

kidnappings followed by executions elsewhere. The seizure itself of an individual, however, is

quite relevant: it indicates the effort to eliminate an individual from a community, wherever the

person’s execution ultimately took place. To construct my measure of local violence, I divided

this number of deaths by the total population according to the 1930 census. I limited the violence

measure to the period prior to January 1, 1937, because desertions were only measured in earnest

after that point; this allows me to avoid potential problems of endogeneity arising from the use of

violence in efforts to capture deserters and punish their families. I removed any executions at the

front, since this would tend to be quite endogenous to desertion.

46 Instituto Cántabro de Estadística 2009. 47 Boletín Oficial de la Provincia de Santander , 26 February 1936. 48 Dirección General del Instituto Geográfico, Catastral y de Estadística 1930. 49 Gutiérrez Flores 2006. 216

Data on conscription versus voluntarism were gathered from the card’s indicator Soldado o miliciano—soldier or militiaman. This indicates whether the soldier was a member of the ad-

hoc volunteer militias that had emerged at the beginning of the war, or was regularly

incorporated via the draft. Political affiliations were also listed on the card. A left-wing

affiliation was defined as follows: indicated membership in one of the two large unions, the

Socialist UGT or the anarcho-syndicalist CNT, or affiliated unions; in a Left Republican party

(the Izquierda Republicana and the Unión Republicana ), the Socialist party, the Communist party, the anarchist FAI, or the joint Socialist-Communist youth wing, the JSU. Non-membership was often indicated explicitly (e.g. “ ninguna ” or none), but soldiers were also coded as unaffiliated if there was a line drawn through the cell for political affiliation. The reason is that the field was often left blank; if affiliation was truly unknown, then, the clerk could leave it as it was. (For obvious reasons, no-one was listed as having a right-wing affiliation.)

Occupational codings were quite difficult. I broke down the sample into different economic sectors: day-labourers and unskilled workers; yeomen farmers ( labradores ); miners and industrial workers; service-sector workers (other than businessmen and professionals); students; and a combined business-professional category. I broke down occupations in this way in order to capture broad income categories. In line with Collier and Hoeffler’s argument that soldiers should desert when better economic opportunities exist outside the army,50 I attempted

to estimate the effect of wages directly by drawing on contemporary government data on

industrial wages 51 and Bringas Gutiérrez’s data on agricultural wages, gathered from the Spanish

Statistical Yearbooks, 52 but this was a questionable approach. The base of information for

industrial wage rates was often quite small, measuring only one or two enterprises in a given

50 Collier and Hoeffler 2004. 51 Ministerio de Trabajo y Previsión 1931. 52 Bringas Gutiérrez 2000. 217 province; many different occupations were not covered by Spanish government records, especially professionals and businesspeople; much of workers’ compensation was informal

(especially for agricultural workers); 53 agricultural workers were not included under the same surveys as industrial workers; and there was no practicable way of distinguishing among different classes of agricultural worker, such as between yeomen and hired farmhands.

Considering these sources of random and systematic error, it seemed more plausible and fairer to an important control variable to employ broad income categories in the analysis. In particular, there were important differences between rural day-labourers and yeomen farmers in political preference. The latter were often excluded from Spanish left-wing coalitions, even where, as in

Cantabria, they formed a large proportion of the rural poor. 54 Therefore, it seems wiser to employ a measure that distinguishes between these groups. In any event, when included as a regressor instead of occupational categories, income had no discernable relationship to the likelihood of desertion. 55

Date of enlistment and age were straightforwardly coded from the soldiers’ files. I included a square term for each, because there is a plausible non-linearity. In the Spanish Civil

War, the very young—especially those under 21—and the old were thought to desert more often than those in their mid-20s. The young were thought to be fearful and vulnerable, the old to be too committed to their families. There was a greater cultural presumption that those in their mid-

20s would serve. 56 As for date of enlistment, I included a square term because two effects may be present: an early date of joining may indicate greater commitment, but a late date potentially

53 Ministerio de Trabajo y Previsión 1931; Malaquer de Motes and Llonch 2005, 1205. 54 Malefakis 1970, 112–121. 55 In addition, since my data cover only the period January-August 1937, there was no way of testing the effect of the harvest on the likelihood of desertion. Bearman 1991; Collier and Hoeffler 2004. This is unfortunate considering the prevalence of agricultural workers in my sample. 56 Corral 2007, 162–163. 218

Table 6.1. Descriptive statistics: individual-level variables

Analyzed sample Whole sample

Categorical variables Frequency Percent Frequency Percent Non-deserters 838 66.83 1305 49.85 Deserters 416 33.17 1313 50.15

Conscripts 856 68.26 1200 68.49 Volunteers 398 31.74 552 31.51

Unaffiliated to left-wing organization 345 27.51 443 26.62 Affiliated to left-wing organization 909 72.49 1221 73.38

Single 862 68.74 1450 70.15 Married 392 31.26 617 29.85

Occupation: Unskilled/day-labourer 287 22.89 420 20.04 Yeoman farmer 421 33.57 799 38.14 Mining/industrial 332 26.48 506 24.15 Services 174 13.88 279 13.32 Student 13 1.04 33 1.58 Commercial/professional 27 2.15 58 2.77 Home county: Área de Santander 354 28.23 488 22.74 Asón-Ramales 60 4.78 104 4.85 Cabuérniga-Tudanca 55 4.39 85 3.96 Castro Urdiales 41 3.27 66 3.08 Costa Occidental 96 7.66 128 5.96 Laredo 48 3.83 82 3.82 Liébana 48 3.83 95 4.43 Pas-Castañeda 100 7.97 159 7.41 Reinosa-Campoo 138 11.00 220 10.25 Santoña-Miera 142 11.32 286 13.33 Torrelavega-Besaya 172 13.72 259 12.07 Other provinces 174 8.12

Continuous Variables Mean Minimum Mean Minimum (SD) Maximum (SD) Maximum Age 25.89 16 25.75 16 (5.42) 62 (5.41) 62

Date of enlistment (days since 17 July 1936) 195 1 188 1 (81.07) 384 (84.46) 391

219

Table 6.2. Descriptive statistics: hometown-level variables

Mean Minimum (SD) Maximum Altitude (metres above sea level) 237 5 (280.82) 957

Steepness index = (proportion of municipality’s territory 30- .0688 0 50% in slope * .3) + (proportion >50% * .5) (.084) .326

Violence (deaths per 10,000 population as of 1930) .170 0 (.218) 1.3

Literacy rate (proportion) .710 .467 (.0573) .814

Right-wing vote share (proportion) .670 .302 (.149) .950

subjects the soldier less to exhaustion and the burdens of war. It is possible, therefore, that date of enlistment has an inverted-U relationship with desertion rates. Marital status was also taken directly from soldiers’ files. I included widowers under the category of married, because this variable is intended to capture soldiers’ attachments at home, and widowers would have attachments to children and in-laws.

Summary statistics for each of these variables are given in Tables 6.1 and 6.2. Table 6.1 gives summary statistics of individual-level data, both for the sample of soldiers that is included in the statistical models and for the full sample before geographic limits and missing data delete cases. It also reports the eleven counties of Santander province, which aggregated its 102 municipalities, to show the basic geographic distribution of soldiers. Table 6.2 gives descriptives for the hometown-level variables, measured at the level of the 102 municipalities. One municipality had only six soldiers sampled, and they dropped out due to missing data, leaving

101.

220

Of the 2618 soldiers in the sample, over half are deleted from the eventual logit analysis; this reduces the dataset to 1254. Two further municipalities, with 43 soldiers, leave the analyses that include the steepness index, because it is not available for all municipalities. This large loss of cases is as a consequence mainly of missing data, but also because of the geographical limitation I placed on the dataset. Because of this large missing data problem, there is some concern about systematically missing data. As it is possible to see from Table 6.1, missing data has a large effect on the outcome variable, desertion rate. Deserters are much more likely to have missing data than non-deserters. Many deserters’ military cards, especially before January 1937, are blank apart from their name, battalion number, and a report of their desertion; this indicates that the cards may have been created after the fact, as placeholders in a file-keeping system rather than as a substantive data-gathering operation. More substantively, this raises the concern that deserters about whom little is known are systematically different from deserters about whom much is known. It is obviously very difficult to determine this after the fact, but in any event the data must be interpreted with this possibility in mind: this chapter can tell us about soldiers the army had records on, but not others. Thus, for example, it can say very little about soldiers who deserted prior to January 1937. Otherwise, yeomen farmers are under-represented; unskilled workers, day labourers, industrial workers and those from the capital region are over- represented. This indicates an urban bias in the eventual sample: more is apparently known about urban soldiers than about rural soldiers. This is consistent with my overall approach to the core and periphery, but it represents something of a selection bias.

221

Table 6.3. Multilevel logit results

Dependent variable: 0 = Model 1 Model 2 Model 3 Model 4 did not desert; 1 = Estimate Estimate Estimate Estimate deserted (s.e.) (s.e.) (s.e.) (s.e.) Volunteer -1.529*** -1.727*** -1.813*** -1.826*** (.258) (.271) (.255) (.256) Left-wing affiliation -.870*** -.803*** -.775*** -.775*** (.165) (.167) (.157) (.157) Married -.121 -.113 (.190) (.193) Age .330* .241 (.165) (.161) Age squared -.00579 -.00416 (.00310) (.00302) Date of enlistment .0247** .0249*** .0283*** .0283*** (.00621) (.00632) (.00617) (.00619) Date of enlistment squared -.000054** -.0000564*** -.0000665*** -.0000666*** (.0000156) (.0000158) (.0000152) (.0000153) Occupation: Yeoman .180 .183 (.210) (.216) Occupation: Mining or .488* .562* industrial (.224) (.230) Occupation: Services .0436 .143 (.271) (.279) Occupation: Student 1.602* 1.677** (.698) (.709) Occupation: Commercial .0944 .249 or professional (.524) (.522) Hometown altitude .000592+ (.000315) Hometown steepness a 3.698** 3.345** 2.705* (1.081) (1.026) (1.342) Hometown literacy rate -.668 1.238 (1.745) (1.681) Hometown violence .0672 -.0498 (.395) (.378) Hometown right-wing .0255*** .0207** .0199** .0201** vote share a (.00619) (.00636) (.00603) (.00596) Right-wing vote * .0508 steepness a (.0703) Constant -9.767*** -9.982*** -5.763*** -2.542*** (2.451) (2.384) (.693) (.602) Variance in constants .309* .186 .224 .201 (.172) (.300) (.229) (.263)

Log likelihood -624.326 -596.569 -603.882 -.603.623 BIC statistic 1377.065 1320.924 1264.558 1271.14

N of individuals 1254 1211 1211 1211 N of hometowns 101 99 99 99 + p < .1; * p < .05; ** p < .01; *** p < .001. a In Model 4, right-wing vote share and steepness are centered at the grand mean.

222

6.7. Results

I first ran two multilevel logit models, corresponding to altitude and steepness. Results are in Table 6.3. In Model 1, the relationship between altitude and desertion is positively signed, as expected, but this relationship falls just short of statistical significance at the conventional .05 level ( p ≈ .06) . In Model 2, however, rough terrain has a much stronger, and statistically significant, association with desertion. Despite the fact that 43 cases and two municipalities drop out of the data set because of unavailable data on steepness, Model 2 has a lower Bayesian

Information Criterion (BIC) statistic, indicating a better fit to the data. The result for steepness is also substantively important. Based on Model 2’s estimates, I predicted probabilities of desertion for a typical soldier—mean age and date of enlistment, from a hometown with mean literacy, violence, and modal values on the other variables. This hypothetical soldier is thus 26 years old, joined 195 days after the start of fighting, and is an unmarried, conscripted yeoman farmer who is affiliated with a Popular Front organization, from a hometown with .17 deaths per 10,000 pre- war population and a literacy rate of 71%. Predicted probabilities of desertion at different levels of steepness are then given in Figure 6.4, with predicted probabilities at different levels of right- wing vote share given for comparison purposes. As it indicates, a soldier from the steepest municipality had, on average, 3.15 times the likelihood of desertion as a soldier from the flattest—an association comparable in magnitude to, and perhaps slightly stronger than, that of right-wing vote share. For the sake of further comparison, a typical soldier, if a conscript, had a likelihood of desertion 4.2 times higher than that of a volunteer. Steepness therefore had a strong association with desertion, and one that is comparable to other important indicators of the likelihood of desertion.

223

Figure 6.4. Predicted probabilities of desertion according to home municipality steepness index and vote share, for a typical soldier, with 95% confidence interval Based on estimates in Model 2.

(a) By steepness (b) By vote share .16 .16 .14 .14 .12 .12 .1 .1 .08 .08 .06 .06 .04 .04 .02 .02 0 0 Predicted probabilitydesertion of Predicted probability desertion of 0 .1 .2 .3 30 40 50 60 70 80 90 100 Municipality steepness index Municipality right-wing vote share

Thus it appears to be the roughness of the terrain, rather than altitude, that drives the association between geography and desertion rates. The result for altitude is substantially weaker. Steepness may be a better indicator of the opportunity to hide than altitude, since it captures the ruggedness of terrain rather than its elevation. Mountains should have a comparative advantage over plateaus in the opportunity to hide, as should rough coastal terrain over coastal plains. Steepness distinguishes rough terrain from flat terrain, while altitude does not. (However, see the Appendix to Chapter 7 for discrepant data from a separate sample, in which altitude appears to have considerably more importance, steepness and right-wing vote share less.)

We can better characterize the relationship between steepness and desertion through a visual representation. Reporting the raw percentage of deserters in my dataset across each steepness decile (for example) would not be useful in giving raw rates of desertion, because my sampling method, by design, inflates the proportion of deserters. Instead, in order to generate as close to true rates of desertion as possible, given my sampling method, I ran a logit model of desertion against deciles of the steepness index alone, with no other covariates. I then adjusted its estimates of the desertion rate according to King and Zeng’s “prior correction” method for

224

Figure 6.5. Desertion rates by steepness decile (with standard error of estimate)

.16 .14 .12 .1 .08 .06 DesertionRate .04 .02 0

1 2 3 4 5 6 7 8 9 10 Steepness Deciles

adjusting the constant term to account for the sampling technique I employed. 57 This gives an estimate of percentage of deserters at each decile, adjusted to reflect the fact that the true proportion of deserters is much lower in the population than in my sample. The results are given in Figure 6.5. They indicate that the desertion rate increases considerably from the lowest to the highest decile, but that variation in desertion rate across the flattest half of the municipalities is better characterized as trendless fluctuation. Terrain appears to have its effect at the upper reaches of steepness, rather than at the lower end.

Local right-wing vote share had a robust, positively signed association with desertion.

This indicates that, in general, the stronger the local right wing, the likelier desertion was among soldiers from a given municipality. However, it is not immediately obvious how to interpret this result. It might be taken as an ecological indicator, suggesting that individual soldiers from right- wing locations were more likely to hold right-wing attitudes themselves. The association with desertion rates would then be explained by the stronger motivation of the soldiers in question not to fight for the Left. However, because of the logical difficulties of ecological analysis, other possibilities exist. Right-wing communities may have provided help to deserters of any political

57 King and Zeng 2001. 225 stripe, and may have been places where the left had a weak local organizational presence. Both factors suggest that such communities were places where soldiers would have a better chance at deserting without being caught. Adjudicating this relationship appears to be an important area for future research.

Local violence, included for investigative purposes rather than for hypothesis testing, had

no apparent relationship with the likelihood of desertion. This is unsurprising since, as noted

above, there is a plausible case that violence can indicate either lack of control or efforts to

reassert that control. Similarly, literacy rates did not have any clear association with desertion

rates. This suggests that even if local development patterns gave individuals different economic

interests in fighting, this did not translate into a decision to desert or not to desert.

As for the control variables, voluntarism versus conscription was substantively important

across all models, as was membership in a popular-front organization. The negative sign on the

square term indicates an inverse-U-shaped relationship. The likelihood of desertion first

increased among those who joined later, and then declined among those who joined later still.

Interestingly, the demographic variables of marital status and age had no clear relationship with

desertion. There is also very little clear association between occupational categories and

desertion, with the exception of frequently-deserting students. This finding fits an interpretation

in which those with better economic opportunities outside the armed group desert more often:

students (especially military-age students, who would generally be in college or university) came

from elite families. But the lack of other consistent relationships with occupational categories—

especially yeomen farmers, who were clearly better-off than day-labourers—suggests that this is

not a robust conclusion. The lack of an apparent association between occupations and desertion

could be an artifact of the limitations of my occupational categorization scheme. It is unclear that

226

the occupational codings add much information for the investment of degrees of freedom: the

BIC statistic for a model without the occupations (not shown) is lower, 1297.051 as compared to

1320.924 for Model 2. In fact, a stripped-down model, removing the apparently non-influential

demographic indicators of marital status, age, and occupation, and the apparently unimportant

hometown conditions of violence and literacy rates, performs rather better given the investment

of degrees of freedom (Model 3): the BIC statistic clearly declines, from 1320.924 in the best-

fitting saturated model to 1264.558.

Does rough terrain facilitate desertion in all municipalities, or only in some political

settings? This question essentially addresses the uniformity of the relationship between terrain

and desertion, allowing us to characterize the result more fully. The more general the result, the

more universal the importance of terrain; it would suggest that terrain is not just something that

affects some municipalities’ desertion rates and not others. In Model 4 I explore an interaction

effect, using the stripped-down Model 3 as a baseline for clarity’s sake. This model follows

Enders and Tofighi’s advice 58 to handle interaction effects among second-level variables by centering them at the grand mean. The interaction term is not statistically significant. The results indicate that the relationship between terrain and desertion is relatively general; it offers opportunities for anyone to desert, regardless of local political conditions.

Motivation differences need not be limited to the left-right master cleavage. Histories and ethnographies of mountain peoples, notably Scott’s recent important contribution, frequently indicate a strong preference for autonomy from lowland political authority. 59 One might then argue that such a systematic pattern is not controlled for by my data, and could account for the observed relationship between steepness and desertion. However, a stronger preference for not

58 Enders and Tofighi 2007. 59 Scott 2009. 227 being subject to an outsider’s authority should also translate into a resistance against joining up in the first place. Soldiers with that preference should thus be more likely than other soldiers to be conscripts. The voluntarism variable should therefore control for much of the effect of a difference in values. In addition, one would expect a cultural preference for autonomy from the centre to be associated with a broader disengagement from the centre’s politics. I therefore estimated a separate model, not reported here, controlling for turnout to vote in the February

1936 elections as a measure of political alienation; there is no apparent relationship to desertion rates.

6.8. Qualitative Evidence from Santander

The statistical analysis above is strongly suggestive of an association between difficult terrain and high desertion rates. Does this association occur for the suggested reason—because deserters from towns with rough terrain, and their families, find it relatively easy to hide? There is suggestive anecdotal evidence in this regard. In this section I draw heavily on the work of

Fernando Obregón Goyarrola, whose interviews in Cantabria have led to multiple volumes of local history of the war.

Deserters did worry about capture, attempting to develop a good knowledge of where army posts and patrols were located in order to avoid being caught. 60 This suggests the

importance of whatever could help a soldier hide. The hills, of course, could help them hide, and

the advantages of the hills loomed large in the calculations of deserters. One civilian witness

from a valley distinguishes between “ arriba ” (above) and “ abajo ” (below, where he lived) in describing the importance of the hills: “Those from [the village of] Esles up above... did not want to go the militia ...Around here, below, there wasn’t anyone who was hiding out.” The mountains

60 Obregón Goyarrola 2005, 90; Obregón Goyarrola 2007, 131–132; García Guinea 2005. 228

of the Peña Herrera massif, near Esles, served as a refuge for dozens, and, according to Obregón,

“it was said that the militiamen never came up looking for them.” 61 Those who were actually

from hilly areas had particular advantages. A good knowledge of the terrain could assist with hiding. The Pasiegos, a group of nomadic pastoralists in the Pas area, 62 typically made their way

in itinerant fashion among a network of cabins in the hills. When Pasiego soldiers deserted—

which was often, according to a June 1937 letter from the commander of the 2 nd Division 63 — they tended to hide in those very cabins. 64 In contrast, those who did the searching were

frequently lowlanders, according to one witness: “If they caught you they shot you, but they [the

Republican military authorities] did not know [the hills].” 65 Noted the commander of the 2 nd

Division, the Pas area “even in normal times is difficult to control.” 66

In addition, local contacts in the hills were of vital importance. Where deserters did not

know the terrain, they could attempt to arrange a local guide, for a fee. 67 Help from family and

friends could be quite important for deserters, and they could make more credible commitments

to help and not turn the soldier in. For example, in the Pas region, one witness remembers

keeping his brother in a hay loft for several days early on in the war. Another remembers that her

fiancé was kept hidden in a sort of sepulchre in his mother’s house for nine months, his mother

advising him whenever a car—indicating a patrol—was coming. 68 Neighbours also helped: “if

61 Obregón Goyarrola 2005, 90, 157. 62 Freeman 1979. 63 Jefe, 2a División, to Sr. Jefe del C. de E. de Santander, Ontaneda, 23 June 1937. AGMAV, DR, armario 63, legajo 855, carpeta 1, documento 1, page 3. 64 Obregón Goyarrola 2009, 125; Obregón Goyarrola 2005, 90. 65 Obregón Goyarrola 2005, 90; bracketed notes are Obregón’s. 66 Jefe, 2a División, to Sr. Jefe del C. de E. de Santander, Ontaneda, 23 June 1937. AGMAV, DR, armario 63, legajo 855, carpeta 1, documento 1, page 3. 67 Obregón Goyarrola 2007, 176. 68 Obregón Goyarrola 2009, 125, 192. 229 anyone came from Santander [city] to look for us, our neighbours warned us. We lived this way almost a year, until [the Nationalists] took Santander.” 69

Families were also very important because, as elsewhere, family members were punished

for desertion. It was standard bureaucratic procedure in Santander province to initiate a local

investigation and attempt to apprehend family members. The standard order to a local municipal

government was to detain the deserter in question, but if he could not be detained within 48

hours, the government was directed to seize his family and goods. 70 This applied both to deserters and to defectors. In a typical example, when a soldier in Battalion 118 deserted, the head of the Information Section of the General Staff wrote the Popular Front committee of his town of residence, Torrelavega, to request that an investigation take place to determine if his family or friends was sheltering him or if one of them might have encouraged him to defect to the enemy. 71 If a soldier could anticipate that his family was likely to suffer from his defection,

therefore, he might be less likely to defect. Deserters’ families were apparently able to escape

punishment more easily if they lived in mountain districts. Whole families in Reinosa, a mid-size

town in southern Santander province nestled in a valley, took to the nearby hills for safety at the

outset of the war. 72 Civilians also took to the hills to flee in the Pas district, where they were aided by the thick fogs that were characteristic of hill country. 73

There is evidence, in addition, that mountain districts had distinctive attitudes to politics as a consequence of their isolation. As noted, the left had very little presence in mountain districts, where industrialization and the labour movement had not penetrated very far. Right-

69 Obregón Goyarrola 2005, 90; bracketed notes are Obregón’s. 70 For example, Jefe de Investigación, II Cuerpo, Ejército del Norte to Alcalde-Presidente del Consejo Municipal, Los Tojos, Santander, 9 August 1937. CDMH, PS Santander L, caja 412, carpeta 18, page 25. 71 Jefe, 2ª Sección, Estado Mayor, Cuerpo de Ejército de Santander, to Frente Popular de las Izquierdas, Torrelavega, Santander, 22 June 1937. CDMH, PS Santander L, caja 406, carpeta 5, page 57. 72 García Guinea 2005, 17. 73 Obregón Goyarrola 2009, 132–134. 230

wing villages occasionally provided refuge for hidden deserters, and deserters sometimes

conceived of physical space in political, left-right terms. One deserter from Lloreda remembers:

“We were hidden, there were many who were hidden, Lloreda was a very right-wing village.” 74

In other places, the right wing itself had little presence either; some districts found themselves quite isolated from politics. In Liébana, among the most mountainous regions of the province, where there was very little highway access at the time, a common expression captures removal from the war and its agendas because of isolation: “the war happened on the highway.” 75 While

Liébana was certainly right-wing politically, this alignment was often, it seems, more a default

option than the consequence of strong mobilization by the right: before the war, Catholic

agricultural syndicates, a key locus of right-wing organizing, had as little presence as left-wing

labour unions. 76 The hills could just be isolated , rather than strongly motivated for the other side.

Indeed, right-wing sentiment, though prevailing over left-wing attitudes in the mountains,

did not determine mountains’ political role. An interesting piece of evidence comes from

Liébana as well. This was both a district of very high desertion from the Republican side and a

key locus of eighteen years of anti-Franco guerrilla fighting after the war. 77 The guerrilla activity was not just guerrilleros from other districts taking advantage of the hill country: in fact,

Lebaniego soldiers themselves staffed the guerrilla force in this district, drawing once again on families and friends and their local knowledge to evade capture. While their activity generated local resentment for the repression that it produced, and while places that identified as right-wing often denied the guerrilleros aid, the fact that the guerrilleros got assistance from villages at all says something telling about hills. They help people hide, no matter what “side” they are hiding

74 Obregón Goyarrola 2005, 90. 75 Obregón Goyarrola 2007, 23. 76 Ibid., 38. 77 Ibid., 334–335, 340–349. 231 from. It is their association with opportunities and the evasion of capture, rather than their association with the left-right master cleavage of the war, that produces hills’ association with desertion.

6.9. Conclusion

This chapter has advanced the dissertation’s argument through a statistical test of the importance of control in the prevention of desertion. Despite the fact that Santander province is an easy case for a motives-based approach, it has provided good evidence that desertion rates were not a consequence only of motives. In particular, the chapter has argued that while rough- terrain locations may have different distributions of preferences than easy-terrain locations, those different preferences are not the reason for their higher desertion rates. Rather, the higher desertion rates of rough-terrain locations hold net of local political preferences and the commitment of soldiers, as indexed by local vote share and soldiers’ status as volunteers versus conscripts.

The chapter has established, further, that control is not only maintained within the army, but over the rearguard as well. Chapter 4 noted the role of control over the home front elsewhere in the Spanish Republic in the transition to a better-controlled army. It found hints that control was particularly difficult to maintain in rough terrain as well. The evidence reinforces the embeddedness of soldiers in hometowns. When they serve in factions in civil wars, soldiers are not simply isolated. In a sense, they carry their hometowns with them. In the next chapter, I extend this analysis further by discussing how soldiers respond when they serve with other soldiers from the same hometown or from different hometowns.

232

This finding can potentially be replicated in other instances. There is one other

systematic, large-N study of desertion that examines the impact of terrain. As noted above,

Giuffre attributes the high desertion rates of upland North Carolina soldiers to their

disconnection from the slave economy, and thus to their lack of motivation to fight to defend that

system. 78 This is certainly a plausible account, but the more frequently the finding is replicated

elsewhere, the more we would be forced to abstract from particular motivation stories and shift

instead to opportunity structures.

Beyond replication, the finding suggests other testable hypotheses. We can test whether

the finding holds in broad, cross-national terms. Much effort has recently been deployed at

gathering data on geographic features of conflict zones on a broad basis. 79 Thus we can test the hypothesis that if local forces are employed in a rough-terrain conflict zone, they should be more likely to desert than local forces in easier ground. In addition, scholars are beginning to investigate the sources of the fragmentation of armed groups in civil wars; in particular, Fjelde and Nilsson suggest that fragmentation comes as a result of structural opportunities for a sub- faction to break away. 80 Terrain may be one source of such opportunities, though Fjelde and

Nilsson do not investigate it. We can expect, however, higher rates of fragmentation in civil wars

fought in more difficult terrain.

The chapter’s central finding suggests some important theoretical openings. First, it

proposes a new view of control over the rearguard. Whereas Kalyvas argues that there is strong

control in zones that are uncontested by opponents,81 this chapter suggests that this control may

vary even in such zones, with important consequences for the war effort. In turn, terrain may

78 Giuffre 1997. 79 Buhaug, Gates, and Lujala 2009; Raleigh et al. 2010. 80 Fjelde and Nilsson 2011. 81 Kalyvas 2006. 233 have important downstream effects on the process of negotiation between armed group and civilian community. If it is particularly expensive for an armed group to impose its control over a place, the armed group may be willing to make greater concessions in negotiating an alliance with the civilian community working there: for example, greater autonomy, lower taxation rates, or more favourable terms of service for that community’s young men. Mountainous terrain thus may create a sharp tradeoff for the armed group’s leadership: between the cost of imposing control and the cost of creating incoherent, chaotic organizational structures. 82 In turn, the

fragmentation of armed groups may create difficulties in resolving conflicts in peripheral zones

over the long term, because of the proliferation of veto players. 83

The chapter suggests a further sharp tradeoff around local knowledge. Local knowledge is a precious military asset, especially in counterinsurgency, 84 but soldiers who have good local knowledge can use it to desert as well as to fight. I have made the point that deserters with good knowledge of mountainous terrain have an advantage over members of the armed group tasked with finding them, but the point may be more general: if a soldier knows where the enemy is hiding, he knows where he may hide. If he knows which civilians shelter the adversary, he knows where he may find shelter. Do armed groups consciously limit their soldiers’ autonomy and choose not to recruit soldiers with local knowledge in order to guarantee their reliability?

There an analogous hint in international conflict that states limit their armed forces’ capabilities if they fear the coup potential of such capabilities. 85 Future research can assess the degree to

which that decision really exists, and the circumstances in which different armed groups are

likely to play the tradeoff differently.

82 Cunningham, Gleditsch, and Salehyan 2009. 83 Cunningham 2006. 84 Kilcullen 2010. 85 Biddle and Zirkle 1996. 234

In a broader sense, the chapter’s findings about terrain help us understand the role of

peripheries in insurgency. It is a trope of the literature to focus on the revolt of the periphery

against the centre, with the insurgency employing guerrilla tactics, swimming in the sea of the

people and enjoying home-field advantage. 86 This chapter suggests that, sometimes, peripheries play other roles: peripheral to the war itself . The mountain villages of Santander province found themselves caught between two major political entities based in large cities and fighting it out for control of the state centre. Though most peripheral villages certainly had a preference for the right over the left, this preference does not appear strongly associated with a hometown’s propensity for desertion net of its geography. This suggests that a peripheral position can help people avoid the war entirely, not just enable them to resist the government side. As new civil war projects compare the local geographic conditions of wars, therefore, they should be cognizant of this ability of peripheries to remove themselves from the war. If soldiers do not want to fight, they can head for the hills.

86 Arreguín-Toft 2006; Kilcullen 2010. 235

Chapter 7 Collective Action and Desertion in Santander

7.1. Introduction

The clashing armies of the Spanish Civil War varied considerably in the degree of unity of aims that their members showed. Volunteer units often, though not always, displayed considerable enthusiasm for the cause, while others included many opportunists, and conscription on each side brought in many who were decidedly uncommitted or even opposed to their side’s victory. In Chapter 3, I discussed how recruitment and factionalism set up these different preferences. In Chapter 4, I analyzed how control could operate as a countervailing factor when soldiers did not want to serve, and in Chapter 6, I demonstrated the link between control and desertion statistically by investigating the effect of rough terrain on desertion rates in

Santander Province. Chapter 5 discussed, in qualitative terms, the impact of common aims on the operation of norms of cooperation. Where soldiers shared the armed group’s aims, norms of cooperation could take hold, and were further sustained by social homogeneity. Where soldiers were frequently indifferent or opposed, or supported multiple different political agendas, norms of cooperation could not take root to keep soldiers fighting; indeed, soldiers could often cooperate to desert together rather than to fight together. Social homogeneity, in consequence, was no help in limiting desertion, and in fact frequently facilitated soldiers leaving together.

With norms of cooperation unavailable, armed groups with uncommitted soldiers had to fall back on control.

236

This chapter subjects my arguments about norms of cooperation to a statistical test. I

examine how the composition of military units affected desertion rates in Santander. This chapter

complements the previous chapter’s focus on top-down control in Santander by a closer

examination of what might be called, in contrast, the multilateral influence of soldiers on each

other. The core finding extends the analysis of voluntarism in the previous chapter. It was not

just that volunteers who joined up in the first days of the war were, considered as individuals,

less likely to desert than conscripts in the new army. They also influenced each other. Soldiers

who served among more volunteers were less likely to desert.

To support these claims, this chapter reports on the results of a further data-gathering exercise from Santander. Whereas the previous chapter’s data were gathered at the individual level, by a full census of deserters (to the extent that it was possible to gather data about them) and a random sample of non-deserters, I began this chapter’s dataset by randomly sampling companies within the CES. This new sampling strategy allowed me to gather data about the composition of each of these companies and test the importance of various features of company composition. The new empirical strategy thus yields a new perspective on desertion patterns in the CES.

The chapter proceeds as follows. I begin by recapitulating the theoretical issues at stake in considering the influence of unit composition on desertion. I then consider the composition of military units in Santander, and in particular I conduct a qualitative and quantitative exposition of the changing patterns of recruitment and unit composition in the province over time. I then present hypotheses and the results of my statistical analysis. I complement the quantitative analysis with qualitative material. An appendix then considers further statistical tests.

237

7.2. Group Influences and Desertion in Civil Wars

As I outlined in more detail in Chapter 1, the major argument of the dissertation is that the distribution of motivations within an armed group affects the effectiveness of different mechanisms for limiting desertion. Specifically, while control is generally effective across a wide array of conditions, even when many soldiers are uncommitted, norms of cooperation depend on whether combatants generally want to see the armed group succeed. If they do, then they can value norms of cooperation as a way of achieving that goal. When they know that their fellows are committed to the cause, combatants can more easily trust that their own contributions will be matched by those around them. Social homogeneity can help this process because those who resemble each other tend to be more likely to trust each other. Even uncommitted combatants will find the costs of fighting lower if they are only a few in a group of committed combatants, since fighting is less risky the more others are dependable. However, in a unit with many uncommitted combatants, norms of cooperation cannot take hold so easily. Combatants who do not share the common aim place little value on the existence of a norm of cooperation, and even those who do share the group’s aims will fear being abandoned if they are surrounded by men that they cannot trust. Norms of cooperation can even come to facilitate desertion, if put to the goal of leaving and evading capture rather than staying. And social homogeneity should be consequently of little help.

This view takes a new approach to understanding group interests and collective action in civil war settings. It challenges four major perspectives in the literature, and puts the collective action problem on a surer empirical footing. First, it goes well beyond a motivations-only approach to desertion, in which individuals are just more likely to desert if they are, as individuals, unmotivated to fight. As I argued in Chapter 1, this approach simply leaves too

238 much out: problems of organization and of the interaction of individual combatants are much more important than naïve grievance theorists can claim. If individuals’ decisions do not just depend on their own preferences but also on those of others around them and hence on their ability to overcome the collective action problem, then the link between motivations and action is not as simple as motivation or grievance theorists claim.

However, at the same time, I avoid the conclusion that the collective action problem renders preferences about common aims irrelevant. This is effectively Mancur Olson’s conclusion and that of the “greed” approach to civil war studies. If individuals suffer significant personal costs to participate and make only a marginal contribution to success, and if they make their decisions in isolation from each other, then their decisions really just depend on selective incentives. In Olson’s view, selective incentives effectively overwhelm preferences about group goals; the latter are so marginal that they cannot affect decision-making. Hence, there is no real difference between those who value the common aim in question and those who do not; both can be bought, and both can be coerced. Since Olson and his intellectual descendants focus exclusively on selective incentives, they should expect norms of cooperation not to matter at all.

However, other scholars have since proposed norms of cooperation as a way of overcoming the collective action problem. But unlike control or other individual rewards and punishments, norms of cooperation really do seem to depend on common aims. I thus draw a basic distinction between two different solutions to the collective action problem: whereas control applies to more situations than just a common aim, norms of cooperation do not. This theoretical move allows me to reconcile group aims and the collective action problem. It reminds us that the collective action problem was originally formulated to deal with situations with common aims, and that arguments based on selective incentives do not necessarily need to

239 assume those aims. And it means that armed groups whose fighters share the group’s goal have more resources available to combat desertion.

The prospect for norms of cooperation, based on a common aim, is also in tension with a third perspective. Like Olson, Kalyvas gives pride of place to individualized incentives, focusing on control while leaving rewards very much out of his story. This focus is justified by a different perspective on group aims, however. For Kalyvas, the interests at stake in civil wars are enormously variable and local, and difficult therefore to discern. As such, it is really threats of harm that keep civilians in line; by extension, the same should go for combatants. If this is correct, again, motivations should effectively not matter, or at any rate the apparent indicators of motivations should generally be too deceptive to really admit of clear conclusions. Together, this perspective and Olson’s strict selective-incentives approach to the collective action problem would support the null hypothesis of no relationship between apparent commitment to common aims and decisions to desert, and hence no effect for norms of cooperation.

Finally, however, I also offer an amendment and clarification of the most important and systematic recent research into desertion in civil wars. Costa and Kahn find that social homogeneity was crucial in reducing desertion rates in the Union Army during the U.S. Civil

War. They offer two different logics for this explanation. In one, a logic of altruism, soldiers felt greater affective ties to those who were like them, and fought for them. In another logic, social homogeneity underpins norms of cooperation. The first contention has received considerable criticism in the literature on military sociology, however: social cohesion (that is, interpersonal affect) is regarded as having little independent effect, and there are hints that it may have effects that are contingent upon commitment to a common task. This idea is much more in line with the second logic that Costa and Kahn offer, a logic in which social homogeneity underpins norms of

240

cooperation. But under that logic, as I have argued, it is vital to recognize that norms of

cooperation depend on having a common aim in the first place, which Costa and Kahn fail to do.

Therefore, I dispute the altruism logic. Rather than expecting, as an altruism approach does, that

soldiers should desert less often from socially homogenous units regardless of common aims, I expect that soldiers should desert less often in units that share a common aim, and I expect the effect of social homogeneity to be contingent upon the existence of that common aim.

Thus the theoretical stakes are relatively high. How we interpret soldiers’ decisions to desert says a good deal about the basic nature of armed groups and their aims in civil war settings. It is my effort to shift away from strong assumptions about whether common aims exist or not, and to treat that instead as a variable. It is also my aim to reconcile the logic of collective

action with the felt importance of individual motivations, to demonstrate that the collective

action problem does not imply the irrelevance of group aims, and finally to better specify the

scope of the collective action problem. More practically, the argument here sheds light on how

different armed groups work. Some are more amenable to cooperation than others. Some, lacking

recourse to norms of cooperation, must rely on coercion to a greater degree if they are to

succeed. I turn now to examining how the motivations and social homogeneity evolved in my

empirical setting, Santander Province.

7.3. Military Units in Santander

As elsewhere in the Spanish Republic, the pattern of military units shifted over time in

Santander Province. Initially, the Republic’s forces were improvised groups of volunteers supplemented by the remnants of the army and public security personnel. At the outset of the military rising on 18 July 1936, union and party militias and groups of workers began to

241 congregate at key points in the province’s major towns in order to indicate that the Popular Front had the upper hand. 1 From that point on, the Popular Front forces proceeded to organize new units. The militia volunteers had nuclei of recruitment centering on particular geographic locations and particular political and syndical organizations, but this geographic and political concentration was not wholly strict. For example, a militia column formed on 21 July to set out from Santander to forestall a rumoured invasion from Burgos province. The unions had already begun to form militias in the capital, but the Popular Front brought together both Socialist UGT and Anarchist CNT militants into this column. 2 The capital was the base for other militia units as well, including a 100-strong force of the Left Republican Juventudes de Izquierda Republicana that was created on 27 July. 3 In August of 1936, the political parties and unions carried on

forming their own militia battalions. For example, the Regimiento Lenin , consisting of two

battalions, was organized by the Communist Party, the Libertad battalion by the Anarchists, and

the Malumbres and Lina Odena by the Socialist youth league Juventudes Socialistas Unificadas

(JSU). 4 The Anarchist leader Francisco Fervenza formed an Anarchist militia unit mostly from

outlying neighbourhoods of Santander city. The men were generally CNT union members or FAI

party members, but were highly heterogeneous in terms of occupation particularly. 5 There was also local recruitment elsewhere in the province, including a group of sixty volunteers from

Liébana who formed about half of a company within a larger militia battalion. 6 However, there

was less popular enthusiasm and voluntarism here than in neighbouring Asturias or the Basque

Country, and by 5 September 1936 there were only 1,470 militia registered in the province. 7

1 Saíz Viadero 1979, 33–34. 2 Solla Gutiérrez 2005, 133–135. 3 Saíz Viadero 1979, 41. 4 Solla Gutiérrez 2006, 385. 5 Gutiérrez Flores and Gudín de la Lama 2005, 212–213. 6 Obregón Goyarrola 2007, 112, citing Álvarez 1988, 21. 7 Salas Larrazábal 1973, 380. 242

As elsewhere, in Santander the Popular Front leadership attempted to retain central

control of the new political and military forces that had emerged. On 23 July the Executive

Committee of the Popular Front issued orders that no militia could operate without its

authorization, nor conduct home searches or confiscations without a mandate. Those who

violated these orders would be detained and disarmed. 8 The local leadership included prominent senior military commanders like Lt. Col. José Maria García Vayas, who encouraged the implementation of military discipline. 9 However, Santander also experienced some of the

Republic’s standard problems with military organization in the militia period: a lack of junior

officers and NCOs, 10 and chronic problems of coordination among the various independent militias despite the Executive Committee’s statements demanding that militias operate only under its direction.11 In addition, the relatively small force raised in the province suggested a need for conscription.

The Santander government therefore shifted towards a centralized, regularly organized, partially conscripted army corps. It formed a Militias Secretariat on 12 August and a

Commissariat of War at the beginning of September; the latter was given sole authority in the organization and conduct of the Republic’s military effort in the province. It divided up the militias into five columns with responsibility for different sectors and a sixth mobile column, and began to keep more thorough records of numbers of men and arms. 12 The Army of the North was officially created on 27 September 1936 as an umbrella organization for the various units in

Republican territory in the north of Spain; it consisted of the I Corps (Basque Country), II Corps

8 Solla Gutiérrez 2005, 136. 9 Saíz Viadero 1979, 56. 10 Martínez Bande 1980, 170–172. 11 Solla Gutiérrez 2006, 385. 12 Ibid., 385–386. 243

(Santander), and III Corps (Asturias). 13 On 20 November this Army instituted a numbering

system for battalions, with 1-100 for the Basque Country, 101-200 for Santander, and 201-300

for Asturias, in an effort to regularize those battalions and reduce their independent identity. 14

Thus, for example, Fervenza’s Anarchist unit became Battalion 122, the Regimiento Lenin became Battalions 124 and 125, the Anarchist Libertad battalion became Battalion 126 (though its militants carried on committing small acts of autonomy: on its soldiers’ files, the Battalion field was frequently filled out “126 ‘Libertad’”). A measure of the centralized bureaucracy developing over time is the regularization of battalion rosters. In November and December 1936, battalions started sending lists of their personnel to the Commissariat, and pay records, apparently on the basis of these lists, began to be created. The lists themselves were created on an ad-hoc basis, using apparently whatever paper was available, but as of February and March

1937 the battalion’s lists were highly uniform. 15

Beyond such administrative procedures, the army attempted to centralize military

operations. A document entitled “Instructions for Combat,” published in 1936 and apparently

issued to unit commanders, argued that the worst thing was a lack of coordination among

different forces. It justified central direction by arguing that only the army’s leadership was in

possession of information about the whole strategic situation. 16 Commanders differed in their implementation of military discipline. Some, such as Eloy Fernández Navamuel, commander of the Corps’ 3 rd Division as of November 1936, insisted on military discipline with its standard

13 Salas Larrazábal 1973, 397. 14 Martínez Bande 1980, 172–173. 15 For example, lists in Centro Documental de la Memoria Histórica [CDMH], Salamanca, Serie Político-Social [PS] Santander L, caja 271, carpeta 17; CDMH, PS Santander L, caja 239. 16 “Instrucciones para el combate.” Editorial Montañesa, Santander, 1936. CDMH, PS Santander L, caja 273, carpeta 4, p. 2. 244

trappings, including such standard military procedures as the salute, whereas others, like

Fervenza, eschewed the privileges of rank and, for example, ate with their men. 17

Conscription accompanied this shift to a regular military. The Boletín Oficial de la

Provincia de Santander began printing orders the draft classes of 1932 through 1935—that is, men who were aged 22 through 25—to join up, following orders from the central government in

Madrid. Further draft classes were added throughout the spring. 18 A soldier who resisted the draft

could be forced into a disciplinary battalion if caught, his family detained and his goods seized. 19

Through forced recruitment, and through conscription’s knock-on effect of encouraging individuals to join up as volunteers before they would be required to join up as conscripts, the

CES expanded rapidly, to over 20,000 by December 1936. 20 The implementation and enforcement of conscription were entrusted to local government. Local Popular Front committees produced lists of the members of different draft classes who lived locally. 21

However, despite the local basis of conscription, it is ultimately somewhat unclear how different soldiers were assigned to different battalions within the regular army. It does appear that soldiers from different places were brought together in more diverse units. As an analysis of my data suggests, homogeneity by geographic origin declined with the shift to a conscripted army, but there was very little change in homogeneity by occupation or age.

My data gathering began by randomly selecting 34 of the 43 infantry battalions in

Santander province, and then randomly selecting one of the five companies from each of these

17 Gutiérrez Flores and Gudín de la Lama 2005, 181, 216. 18 Boletín Oficial de la Provincia de Santander , 12 October 1936, 29 October 1936, 17 February 1937, 16 May 1937; Solla Gutiérrez 2006, 387. 19 CDMH, PS Santander L, caja 444, carpeta 8, expediente 7; CDMH, PS Santander L, caja 444, carpeta 10, expediente 2. 20 Martínez Bande 1980, 183. 21 For example, Consejo Municipal de Rionansa, “Relación de soldados de los reemplazos de 1921, 22 y 39, residentes en este término que deben efectuar su presentacion en Santander” 19 June 1937. CDMH, PS Santander L, caja 443, carpeta 7, p. 41. 245 battalions. The CES began systematically keeping lists of soldiers from each company in January of 1937. I recorded all soldiers present on the lists at two-month intervals, from January through

August 1937. I then matched the names on the lists to the soldiers’ individual files and to the monthly lists and individual reports of desertion gathered for the dataset in Chapter 4. This ultimately yielded a sample of 4764 soldiers in 34 companies, organized into 111 company- periods.

The soldiers’ files, as in the previous chapter, contained information on the soldiers’ volunteer status; age; number of children; marital status; occupation; place of residence; parents’ names and place of residence; rank; and some details of his military career. I constructed each company’s conscription rate as the number of conscripts divided by the total number of soldiers whose conscription status was indicated. I treated members of the military who had already been serving on 18 July 1936 as conscripts, because the conscription variable is meant to capture whether or not an individual indicated the willingness to join a common cause in the first days of the war.

I measured heterogeneity, first, by various kinds by fragmentation. Political heterogeneity

was measured as fragmentation by union membership. Heterogeneity of origin was measured by

fragmentation in municipality of origin, with soldiers from outside Santander province grouped

as follows. Soldiers from Burgos and Palencia had separate categories; the areas of these

provinces under Republican control were under Santander’s government and had considerably

higher numbers than any other non-Santander province. Asturias and León were grouped

together, since areas of León were under the control of the Asturias regional government.

Soldiers from the Basque Country (Vizcaya, Guipúzcoa and Álava) were likewise grouped

together. All others were placed in a residual group. I also measured fragmentation by county—

246

one level of aggregation higher. Heterogeneity of occupation was given by fragmentation by

occupational category: unskilled workers, yeomen, resource workers (mining and fishing),

industrial, services, a combined category for an economic elite (students, merchants and

professionals), and finally those of unknown occupation. In a few cases, battalion lists in a given

month did not break down the officer corps by company. In those cases, I substituted the whole

battalion’s officer corps, in addition to the rank and file, to measure conscription rates and

heterogeneity.

Following Costa and Kahn, I defined fragmentation as one minus the sum of squares of

the proportion of each category (union, county or occupation) in a given company-period, so:

= 1 − where is the measure of fragmentation in question; is a company-period; is a particular union, county or occupation, assuming there are such categories in the company; is the number of soldiers within that union, county or occupation in the company in the period; and is the total number of soldiers with non-missing values for the categorization in question. The expression that is subtracted from 1 essentially gives the probability that any two soldiers randomly selected (with replacement) from the company would come from the same category. In addition to these measures of fragmentation, I measured heterogeneity by age in terms of the coefficient of variation of age within the company in question.

There was important, but limited, variation across companies in these measures of heterogeneity. Table 7.1 gives descriptive statistics for unit composition. Averaged over the four time periods, conscription rates had by far the highest variation, with a standard deviation of .36.

The measures of fragmentation varied rather less, with a standard deviation around .1.

247

Table 7.1. Descriptive statistics, company characteristics averaged over time

Mean Std. Dev. Minimum Maximum Conscription rate .544 .359 0 1

Union .308 .119 .017 .510 fragmentation County .742 .103 .426 .861 fragmentation Municipality .867 .094 .484 .954 fragmentation Occupation .737 .054 .564 .820 fragmentation Age variation .178 .048 .092 .253

N=34

Volunteer units were built on political organizations in particular locations. This is reflected in correlations among company conscription rates and fragmentation by union affiliation, home county and municipality, all averaged across the four time periods ( = .52, .34 and .52 respectively). That is, the higher the proportion of conscripts, the more fragmented the company across unions and geographic origins. That there was a higher correlation coefficient for municipalities than for counties is suggestive: volunteer companies drew on certain very specific locations. Interestingly, the correlation was negative for heterogeneity by occupation category ( = -.32) and for variance in age (-.82). The latter is immediately interpretable as the consequence of the age-sensitivity of conscription. The job-heterogeneity of volunteer units requires a little more unpacking. Table 7.2 indicates the proportion of each job category by

conscription rates, measured over all four time periods. The proportions of industrial, service,

elite and unknown-occupation workers are remarkably constant across the four categories of

conscription rates. The real differences are in the three most common categories: unskilled,

yeomen and resource workers (mining and fishing). The higher the conscription rate, the higher

the proportion of yeomen. This is consistent with the historical record: yeomen farmers were

248

Table 7.2. Occupation distribution by conscription rate

Proportion conscripts % occupation: Less than .25 .25 to .5 .5 to .75 More than .75 Unskilled 32.48 31.97 32.36 18.81 Yeoman 12.22 24.13 28.75 39.74 Resources 16.26 10.69 6.65 5.31 Industrial 20.96 17.52 17.51 19.76 Services 11.70 9.99 9.63 10.15 Elite 2.87 2.58 2.84 3.49 Unknown 3.51 3.12 2.26 2.74

among the least enthusiastic individuals for the Left. There are also far fewer resource workers in conscript units. In contrast, volunteer units are somewhat more spread out among unskilled, yeomen and resource workers.

As more draft classes were added to the army and more regular units were formed, the

CES continued its shift from a predominantly volunteer to a predominantly conscript army.

Based on my sample, I can track the evolution of the Santander Army Corps over time. Taking each company as a single observation, the average company conscription rate increased from

34% in January 1937 to 61% by July (see figure 7.1a). The conscription rate of the whole sample, irrespective of company, increased at essentially the same rate, from 34% to 60%. Partly this was because new units created in late March and early April 1937 were heavily conscripted

(above 80% conscripts). However, on average each company shifted its composition as well.

Figure 7.1b replicates the findings but adjusts them so that zero is each company’s overall mean conscription rate. It indicates an increase within companies: the overall increase in conscription in the army was not just due to the addition of conscript units. The increase was statistically significant. Regressing the conscription rate against the month of measurement and including company fixed effects basically estimates a regression with a separate intercept, or starting point, for each company. This analysis shows that, irrespective of company, the conscription rate

249

Figure 7.1. Conscription rate by company over time, showing average across companies

(a) Raw rates (b) Differences from company means 1 +.2 .8 +.1 .6 0 .4 -.1 Conscriptionrate Conscriptionrate .2 (0 = (0company mean) -.2 0 January March May July January March May July Observed Mean across companies Observed Mean across companies

Figure 7.2. Changing patterns in volunteer units

(a) Conscription rates among volunteer units (b) All units, mean trends indicated 1 1 .8 .8 .6 .6 .4 .2 Conscriptionrate .4 0 Conscriptionrate January March May July .2 Observed

0 Mean across companies January March May July Mean across volunteer companies

increased an average of 2.5 percentage points per month, or twenty percentage points over the eight months covered by my data.

Finally, another way of seeing the increase in conscription rates is by tracking particular units over time. Figure 7.2a indicates several units that entered the data in January 1937 with a conscription rate less than 20%. It indicates that, over time, many of these units gained a large

250

Figure 7.3. Changing average company composition over time .8 .6 .4 .2 0 Mean acrosscompanies January March May July Conscription rate Union frag. County fragmentation Job fragmentation Age variance

proportion of conscripts. As Figure 7.2b shows, the trend in the average conscription rate of those volunteer units differed very little from the trend in the overall average conscription rate.

Changes in unit composition over time did not much extend to heterogeneity. Figure 7.3 indicates changes to the average rate of conscription alongside average fragmentation by union membership, county of origin, and occupational category, and the standard deviation in age.

Figure 7.3 shows that as conscription increased, heterogeneity remained remarkably stable: only geographic heterogeneity increased (by .02 per two-month period, on average—a statistically significant increase). In fact, however, as Figure 7.4 shows, the increase in geographic heterogeneity was concentrated in the five companies which I have identified; they entered the sample as particularly homogeneous and lost their particular homogeneity over time. In addition, regressing measures of heterogeneity against time and using company fixed effects reveals that there was a statistically significant increase in age variance over time within companies—that is, when the different starting points of different companies are accounted for.

251

Figure 7.4. Changing county fragmentation over time 1

102 102 .8 102 111 124 111 124 114 123

.6 114 111 124 123 123 123 111

.4 102 124 114

.2 114 Countyfragmentation 0 January March May July Observed Mean across companies

The figures indicate, then, one major change in unit composition, an increase in the average conscription rate. There was also an increase in fragmentation by county of origin; this increase was concentrated in a few units that became more heterogeneous over time. Overall, therefore, the analysis of conscription rates that follows tracks differences both over time and across units, while the analysis of homogeneity and heterogeneity mainly tracks differences across companies.

7.4. Hypotheses

If the collective action problem operates in the way that I believe it does, contingent upon the existence of a common aim, then we should see soldiers as less likely to desert to the extent that the men in their military unit share a commitment a common aim. I measure this commitment, first, according to the conscription rate of the soldier’s company. As I argue in

Chapter 3, conscripts are considerably less likely to place value on the armed group’s success than are volunteers, and hence less likely to develop norms of cooperation. Surrounded by

252

uncommitted individuals, a serving soldier should be less likely to trust his fellow soldiers, less

likely to believe that they will support each other in combat, and thus more likely to desert.

Hypothesis 1 (commitment to cause). The higher the proportion of conscripts in a soldier’s company, the more likely that soldier should be to desert .

My dataset offers a unique opportunity to assess a further way in which common aims

may not exist. With the political affiliation (party or union) of each soldier given on his file, it is

possible to assess the association between desertion and factionalism within a company. This

allows for an evaluation of the two principal sources of a lack of common aims in an armed

group, as set out in Chapter Three: recruitment policies and factionalism. If indeed there was a

belief that members of other political organizations did not share the same objectives, it may

easily have been more difficult to establish trust among soldiers of different political stripes.

Hypothesis 2 (political heterogeneity). The higher the heterogeneity of political membership in a soldier’s company, the more likely that soldier should be to desert.

In contrast, the selective incentives approach to the collective action problem, and the emphasis on control alone that arises if individual motivations are extremely chaotic and close to impossible to discern, both suggest the null hypothesis. The existence of a common aim within an armed group should have no relationship to the decision of a soldier to desert.

Hypothesis 0 (null). The likelihood that an individual soldier will desert should not vary by the proportion of conscripts or the heterogeneity of political membership within a soldier’s company.

As noted, Costa and Kahn draw on two very different kinds of argument to explain their finding linking homogeneity and low desertion rates. From an altruistic perspective—that affective ties among soldiers, driven by similarity, cause soldiers to want to fight for each

253 other—serving in a homogeneous unit should reduce the likelihood of a soldier deserting regardless of who his fellow soldiers are.

Hypothesis 3 (altruism). The higher the degree of heterogeneity in a soldier’s company, the more likely that soldier should be to desert .

However, from a rational collective action point of view, homogeneity should be associated with lower desertion rates (that is, heterogeneity should be associated with higher desertion rates) when soldiers share a commitment to the common aim, but not otherwise. When they do, homogeneity can be a prop to reinforce the basis of collective action. When they do not, however, homogeneity should be irrelevant (in a weak formulation) or positively associated with desertion (in a strong formulation).

Hypothesis 4a (rational collective action – weak). The higher the rate of conscription in a company, the lower the relationship between heterogeneity and desertion. Among units with a high proportion of conscripts, heterogeneity should be irrelevant to desertion.

Hypothesis 4b (rational collective action – strong). The higher the rate of conscription in a company, the lower the relationship between heterogeneity and desertion. Among units with a high proportion of conscripts, heterogeneity should be associated with lower desertion rates.

For hypotheses 3, 4a and 4b, I measure heterogeneity according to fragmentation in occupational category and home county, as described above, and according to the standard deviation of age.

We can get a further handle on the relationship between homogeneity and desertion by isolating the influence of the presence of soldiers who resemble a given soldier. Heterogeneity is an ecological measure. For any given level of county heterogeneity, for example, it is unclear whether there is any difference between soldiers who come from well-represented counties and soldiers who come from particularly scarce counties. Costa and Kahn’s two theoretical justifications, altruism and rational collective action, may have different predictions with regard to the specific influence of other soldiers. If soldiers fight specifically for others who are like them, then their likelihood of desertion should decline the more other soldiers there are in the

254

unit who share certain characteristics. If, on the other hand, soldiers are less likely to desert when

norms of trust build up in the military unit in general, then it may be possible for the trust that

exists among certain soldiers in a company to extend beyond that group. If soldiers’ decisions

about whether or not to desert are influenced by their perceptions of the commitment of others,

and if those others have a particularly high level of mutual commitment because of homogeneity

and common goals, then their commitment may extend beyond the in-group and become a

general property across the whole company. This approach, then, is more open to the possibility

of general company-level effects. (Costa and Kahn test the influence of the ratio of other soldiers

from the same birthplace on desertion rates and find a negative, statistically significant

relationship, indicating that in their sample soldiers were influenced specifically by others like them. 22 ) We can break this down as follows:

Hypothesis 5 (specific influence): Soldiers should be less likely to desert the more other soldiers there are in the company who specifically resemble them.

If Hypothesis 5 holds, then we should give more credence to the altruism of resemblance. If not, and there are homogeneity effects in general, then the evidence is stronger for homogeneity’s

role within a rational collective action framework.

7.5. Method

Because my data are time-varying across four time periods, with a binary dependent variable, they have a binary cross-section time-series (BTSCS) format. 23 I could not reject (at the

.1 level) the hypothesis that one’s likelihood to desert depended on how long one had been

serving. I therefore included a series of dummy variables for a soldier’s tenure in the army, as

recommended by Beck et al. Since there were only four time periods, it was not necessary to

22 Costa and Kahn 2003, 537. 23 Beck, Katz, and Tucker 1998. 255 employ any of the more complicated possible methods for accounting for duration dependence, such as cubic smoothing splines, that Beck et al suggest. I also included a random-effects component, allowing the intercepts to vary by company-period, in order to take account of the company-period-level variation that is central to my analysis.

I mean-centered the various company-level variables, because none but the conscription rate included a zero point, and because Hypotheses 4a and 4b require the assessment of interaction effects. It is generally wise to mean-center when estimating interaction effects in non- linear analysis such as logit, because the non-linearity means that interaction effects will be different when measured at different points of each variable. Measuring them at their means ensures that the results are relevant to the maximum proportion of the data. Doing so also provides clear and easily interpretable estimates of the main effects and interaction terms. In addition to the company-level variables that are of primary interest, I controlled for a series of individual-level covariates (soldier’s own conscription status, and whether that soldier had any political affiliation at all; rank in the month in question; age; marital status; occupation). Finally,

I controlled for the impact of hometown altitude, steepness and right-wing vote share. Each of these variables was assigned zero if the soldier in question came from outside Santander or his hometown was not identified; I added a dummy variable, coded 0 if a soldier was identified as coming from Santander province and 1 if not. Because of the logic of interaction terms, then, the coefficients on the altitude, steepness and right-wing vote share variables give the association between those variables and desertion rates conditional on the dummy variable equaling zero, that is, conditional on a soldier coming from Santander province. Descriptive statistics are given in Table 7.3, at the individual, company and hometown levels.

256

Table 7.3. Descriptive statistics (a) Individual-level variables, time-invariant

N Mean Std. dev. Min Max Individual-level variables, time-invariant Deserted 4764 .052 .223 0 1 Conscript 4438 .546 .498 0 1 Left-wing affiliation 4225 .815 .388 0 1 Age 4629 25.57 5.01 16 60 Date of incorporation 4764 165.298 85.58 0 348 Married 4627 .314 .464 0 1 Occupation: Unskilled 4764 .253 .435 0 1 Occupation: Yeoman 4764 .273 .445 0 1 Occupation: Resources 4764 .094 .292 0 1 Occupation: Industrial 4764 .191 .393 0 1 Occupation: Services 4764 .119 .324 0 1 Occupation: Elite 4764 .033 .180 0 1 Occupation: Unknown 4764 .037 .189 0 1

Company-period-level variables Conscription rate 111 .507 .364 0 1 Union fragmentation 111 .298 .130 .016 .515 County fragmentation 111 .739 .130 .191 .882 Occupation fragmentation 111 .736 .059 .438 .825 Age variation 111 .179 .052 .037 .264

Hometown-level variables Right-wing vote share 102 67.005 14.81 30.168 94.992 Steepness index 100 .095 .090 0 .326 Altitude 102 238.89 280.05 5 957 Population 102 3577.127 8484.655 295 85117 Population Density 102 61.203 152.69 4.436 1107.883

7.6. Results

The main hypotheses are tested in the models in Table 7.4. Model 1 leaves out the interaction effects examined in Hypotheses 4a and 4b for the moment, to focus instead on

Hypotheses 0 through 3: the direct associations between desertion rate and social and political heterogeneity. It indicates possible support for Hypothesis 1: a lack of commitment among other members of a company, measured by a high conscription rate, is associated with a higher

257 likelihood of desertion. Interestingly, the conscription status of the individual soldier has no statistically significant effect, net of the conscription rate within the unit at large. This is a discrepancy from the findings in Chapter 6, and I discuss this—and other discrepancies—at length in the Appendix to this chapter. However, the non-significant effect of the individual-level conscription indicator should not be taken to suggest that a soldier’s personal motivation is irrelevant. Whether a soldier had a left-wing affiliation or not was a robustly important predictor of a propensity to desert. In any event, assignment to majority conscript companies was, of course, nonrandom: one was obviously likelier to be in a majority-conscript unit if one was a conscript, so unit composition is, to a certain extent, an indicator of an individual’s commitment.

There is also some support for Hypothesis 2: soldiers in companies that were highly fragmented by union membership—that is, polarized between the two union federations—were more likely to desert than soldiers in more concentrated units. We can therefore also reject the null hypothesis, Hypothesis 0, that there is no relationship between the distribution of motivations in a company and the decisions of soldiers in that company to desert or not. As well, there is mixed support for Hypothesis 3: with the exception of heterogeneity by occupation, the social heterogeneity of a military unit does not clearly increase the likelihood that soldiers will desert. The coefficients are, as expected, positively signed, but only the coefficient for fragmentation by occupation approaches statistical significance at the .05 level. In separate analyses, not shown here, I also considered each of these measures of fragmentation independently of each other, and none had a statistically significant relationship with a soldier’s likelihood of desertion.

258

Table 7.4. Multilevel logit results

Model 1 Model 2 Model 3 b (s.e.) b (s.e.) b (s.e.) Conscript .268 (.357) .304 (.363) .437 (.344) Left-wing affiliation -1.289 (.199) *** -1.289 (.199) *** -1.343 (.187) *** Officer -.865 (.496) + -.851 (.496) + -.914 (.481) + Non-commissioned officer .129 (.244) .144 (.244) .076 (.233) Occupation: Yeoman .231 (.264) .203 (.264) Occupation: Resources -.333 (.461) -.350 (.461) Occupation: Industrial .381 (.286) .357 (.286) Occupation: Services .136 (.367) .116 (.368) Occupation: Student / commercial / -.022 (.545) -.048 (.546) professional Occupation: other / unknown .190 (1.069) .140 (1.067) Age .045 (.023) * .045 (.023) + .039 (.021) + Date of incorporation .001 (.002) .001 (.002) Married -.144 (.204) -.139 (.204) Company conscription rate 2.068 (1.072) + 3.436 (1.212) ** 3.583 (1.199) ** Company union fragmentation 4.595 (2.094) * 5.352 (2.112) * 5.445 (2.102) ** Company county fragmentation 3.625 (2.240) 5.527 (2.715) * 5.500 (2.707) * Company job fragmentation 7.290 (4.077) + 13.806 (6.781) * 14.224 (6.786) * Company age variation 3.406 (6.005) 15.349 (9.307) + 15.657 (9.261) + County fragmentation x conscription -8.340 (6.582) -8.178 (6.563) Job fragmentation x conscription -25.689 (15.868) -27.534 (15.826) + Age variation x conscription -31.340 (20.001) -30.022 (19.851) Home outside Santander .168 (.582) .164 (.583) .127 (.394) Hometown altitude .001 (.000) *** .001 (.000) *** .001 (.000) *** Hometown right-wing vote share .000 (.007) -.000 (.007) Hometown steepness .244 (1.229) .272 (1.226) Time since entry: 2 months .573 (.254) * .606 (.254) * .587 (.251) * Time since entry: 4 months .664 (.357) + .687 (.355) + .594 (.334) + Time since entry: 6 months .140 (.598) .160 (.594) .034 (.556) Constant -6.757 (.853) *** -7.384 (.918) *** -6.900 (.736) *** Company-level variation .600 (.307) + .406 (.320) .399 (.318) N 8965 8965 8981 Log-likelihood -622.294 -617.757 -620.183 BIC 1490.316 1508.547. 1422.423 *** p < .001; ** p < .01; * p < .05; + p < .01. Dependent variable in all models: 0 if soldier did not desert in time period; 1 if he did.

Model 2 adds interactions between conscription rate and the measures of social homogeneity in order to test Hypotheses 4a and 4b. While the interaction terms themselves are not statistically significant, they are, as expected by both hypotheses, negatively signed: as the conscription rate increases, the relationship between heterogeneity and desertion rates declines.

Interestingly, adding them into the model seems to have an important effect in clarifying the

259 relationship between social heterogeneity and desertion rates: the main effects become statistically significant here. Therefore, it appears to be important to include interaction terms in order to better capture the link between unit composition and desertion rates. Various control variables—occupation, date of incorporation, marital status, steepness, and right-wing vote share—appeared to have little relationship to desertion decisions, so I excluded them for Model 3 in an attempt to reduce noise. The results are essentially the same, but more efficiently estimated.

In any case, those excluded variables tell us very little: in a likelihood-ratio test, we cannot reject the null hypothesis that Model 2 does not fit the data any better than Model 3 ( p >> .05).

As is common with interaction effects in non-linear analyses, much can be clarified by simulating predicted probabilities. First, we can compare the impact of conscription rate and social homogeneity. In addition, predicted probabilities are vital for determining whether heterogeneity simply has no relationship to desertion in conscript units, or whether it has a negative relationship. This will distinguish between Hypotheses 4a and 4b, and thus establish if it is simply that the relationship between heterogeneity and desertion weakens at high levels of conscription, or if it reverses direction entirely. I generated predicted probabilities on the basis of

Model 3, for a typical soldier: the other values are at their means (in continuous variables) and modes (in categorical variables). The hypothetical soldier in question is thus a 26-year-old conscript, a member of a left-wing organization, in the ranks rather than a non-commissioned officer or an officer, and from a Santander province district 115m above sea level. His company is, except when I manipulate these values, at the mean conscription rate (46.6%) and mean heterogeneity across each measure of heterogeneity. Figure 7.5 shows the marginal difference in predicted probability of desertion in a given two-month period associated with a shift across the

260

Figure 7.5. Predicted probabilities of desertion .008 .006 .004 .002 0 Predicted of prob.desertion Conscription County frag. Age var. Union frag. Job frag.

Interquartile ranges

Table 7.5. Predicted probability of desertion – interaction effects

Cell values are the ratio of predicted probability at the 75 th percentile to predicted probability at the 25 th percentile of the indicator of social heterogeneity, with 95% confidence interval.

Social heterogeneity Volunteers only Mean (50.7% conscripts) Conscripts only indicator County fragmentation 4.914 2.606 1.258 [0.859 28.043] [1.038 6.532] [0.440 3.560] Job fragmentation 6.119 2.662 1.029 [1.001 37.289] [0.988 7.158] [0.334 3.133] Age fragmentation 13.569 4.116 1.055 [2.993 61.372] [1.701 9.946] [0.247 4.427]

inter-quartile range of my various company characteristics. As is readily apparent, the conscription rate has the largest substantive effect of any of the unit characteristics.

Table 7.5 indicates the proportional change in predicted probability of desertion

associated with a change from the 25 th to 75 th percentile of heterogeneity by county of origin, occupation and age, in volunteer-only units, units with a mean proportion of conscripts and conscript-only units. A value of 1 thus indicates no difference between the 25 th and 75 th

261 percentile. These results are for hypothetical companies with mean conscription rate, all volunteers, and all conscripts. The table thus shows the change in probability associated with social heterogeneity, and how that change itself is different at different levels of conscription.

The results show, as expected by both Hypotheses 4a and 4b, a substantial decline in the relationship between heterogeneity and desertion rates the higher the proportion of conscripts in a given company. However, the results also are clearly more favourable to Hypothesis 4a than to

4b: the relationship between heterogeneity and desertion is much weaker at higher conscription rates. Among all-volunteer units, there are clear and statistically significant differences in desertion rates between socially homogenous and socially heterogeneous companies. For example, a soldier was over six times likelier to desert from a company that was highly heterogeneous by occupation than he was to desert from a highly homogeneous company, provided that company only had volunteers. In contrast, there are no such clear differences in fully conscripted units: the ratios of predicted probabilities are much closer to 1.

In general, then, there is solid support for the hypothesis that the impact of heterogeneity is contingent on the prior existence of a common goal. This is consistent with a rational collective action approach to heterogeneity. In Table 7.6, I unpack the impact of heterogeneity further, testing Hypothesis 5—that a soldier should be less likely to desert the more other soldiers there are in his company who resemble him. In Model 4 I test this hypothesis using the raw number of other soldiers of the same county, job, and age group, 24 and in Model 5 I use

proportions of the soldier’s company. In neither model does Hypothesis 5 receive any clear

support. Nor does testing each group at a time (not shown here) indicate any clear result. It

24 Age groups were defined as soldiers within two years in age of the soldier in question. However, because of the relative scarcity of soldiers above age 40, I grouped all those soldiers together. 262

Table 7.6. Group analyses

Covariate Model 4 Model 5 b (s.e.) b (s.e.) Conscript .327 (.349) .332 (.349) Left-wing affiliation -1.323 (.190) *** -1.332 (.190) *** Officer -.848 (.484) + -.865 (.485) + Non-commissioned officer .072 (.235) .068 (.235) Home outside Santander .046 (.022) * .046 (.022) * Hometown altitude .001 (.000) *** .001 (.000) *** Company conscription rate 1.577 (.740) * 1.597 (.746) * Company union fragmentation 3.955 (1.950) * 3.920 (1.969) * Number of soldiers in company from home county -.007 (.007) Number of soldiers in company with same job -.005 (.006) Number of soldiers in company in same age group .006 (.005) Proportion of soldiers in company from home county -.622 (.688) Proportion of soldiers in company with same job .297 (.675) Proportion of soldiers in company in same age group .614 (.483) Time since entry: 2 months .592 (.255) * .605 (.254) * Time since entry: 4 months .613 (.337) + .618 (.337) + Time since entry: 6 months .047 (.561) .066 (.560) Constant -6.311 (.673) *** -6.326 (.673) *** Company-level variation .632 (.306) * .657 (.305) * N 8947 8947 Log-likelihood -625.090 -625.470 BIC 1404.865 1405.624 *** p < .001; ** p < .01; * p < .05; 1 p < .01. Dependent variable in all models: 0 if soldier did not desert in time period; 1 if he did.

appears, then, that homogeneity and heterogeneity have their impact at the level of the whole company rather than through the specific influence of serving with others whom one resembles.

7.7. Qualitative Evidence

The statistical analysis thus upholds the conclusion that soldiers deserted more often when their units had a higher proportion of conscripts, that differences in political affiliation were also associated with higher desertion rates, and that the association between social heterogeneity and desertion was contingent on the proportion of conscripts in the unit.

Qualitative evidence confirms that the CES underwent an important change with the addition of conscripts.

263

Conscripts had difficulty mustering much enthusiasm to fight. “When we were ordered to advance, some fainted, others shot themselves in the hand,” 25 recalled one soldier in Battalion

105, whose fellows were almost all conscripts (and whose 3 rd Company had a conscription rate

of 95% in my dataset). The army command clearly thought that the presence of volunteer

veterans would help prevent desertion. On 8 May 1937, the Chief of Operations sent a message

to the division commanders noting recent cases of desertion and abandonment of positions. It

referred specifically to an incident two days prior, when “a small post of conscripts abandoned

their position....Having studied the causes, it is believed that the crime arises from insuperable

fear, since, just before it occurred, an enemy battalion arrived.” They were ordered, therefore, to

transfer 25 veterans “of demonstrated valour and loyalty” to conscript units, and to transfer 25

conscripts out. 26 Ten days later, a follow-up message noted that there were still cases of

desertion, though their numbers were “not alarming.” The division commanders were therefore

instructed to direct their officers, NCOs and veterans to remain in close contact with troops at all

times to keep up morale. 27 (This appears to be one of the ways in which the conscription rate of

volunteer units increased between March and July, as noted above.)

In addition, the process of conscription brought into the CES a variety of individuals of

suspect loyalties. This had been a problem early on in the war in volunteer units as well, with

rightists passing themselves off as loyal Popular Front supporters. Around 27 July 1936 a report

appeared in the newspaper El Cantábrico that fascists in the village of Vallejo had seized thirty union membership cards; these were of an old type, and the report noted how differences in size

25 Obregón Goyarrola 2007, 114–115. 26 Jefe de Operaciones to Srs. Jefes de la 1a, 2a, 3a Divisiones and 11a Brigada. Santander, 8 May 1937. Archivo General Militar, Ávila [AGMAV], Documentación de la República [DR], Caja 686, Carpeta 1, Documento 1, page 1. 27 Unknown [Jefe de Operaciones?] to Srs. Jefes de la 1a, 2a, 3a Divisiones and 11a Brigada. Santander, 18 May 1937. AGMAV, DR, Caja 686, Carpeta 1, Documento 1, page 2. 264

and shape could be used to detect the counterfeit documentation. 28 The committed Falangist

Rivero Solozábal recounts that he was able to join the Popular Front in Escalante because he

appeared locally useful as a highway engineer. He was able to use his position to gain safe-

conducts and move around the province, organizing a clandestine network for the transportation

of right-wing civilians to Nationalist territory. 29 In fact, there was a clear recognition of the fragility of political identification. A song common among volunteers in Liébana showed awareness that in a climate of political violence, right-wingers would attempt to pass themselves off as left wing:

If the friars and priests knew The beating we’re going to give them They’d go up to the choir, singing, “Liberty, Liberty, Liberty.” 30

Such problems of political identification meant that volunteer units often had trouble keeping out right-wing “volunteers” who used their position to defect to the Nationalists. Solano

Palacio, an Anarchist activist, writes in an epitaph for the Republican North that there were

“hidden traitors, touting some charge of responsibility in order to defect to the enemy when they considered it convenient,” who “enjoyed official protection and tolerance from us.” 31 In Liébana, some right-wingers joined the militia specifically to get closer to the front in order to defect. 32

However, such stories are few in number.

The growth in the armed forces in autumn 1936, and particularly the use of conscription, intensified this fear for the political loyalties of soldiers. Local Popular Front committees were tasked with reporting on the political pasts of members of each draft class. Soldiers could be

28 Saíz Viadero 1979, 40–41. 29 Rivero Solozábal 1941, 19–20, 30. 30 Obregón Goyarrola 2007, 113. 31 Solano Palacio 1938, 164. 32 Obregón Goyarrola 2007, 114. 265 denounced for having done propaganda for right-wing political parties in past elections, membership in Falangist syndicates before the war, having worn Carlist political paraphernalia, or having been ousted from a left-wing union in the past. 33 Letters from a local union or Popular

Front committee could be damning indications of disaffected political opinions, on evidence such as having leafleted for right-wing parties in the February 1936 election.34

Dubious political commitments posed very difficult dilemmas for the armed forces, undermining the ability of a cooperative, mutually supporting set of norms to emerge in military units. Political commissars, appointed at the battalion and company levels and appearing on battalion rosters as such starting in March 1937, were tasked with supervising the political attitudes of the men in their units. Their instructions captured the dilemmas of maintaining cooperation and control. Keeping military discipline was highlighted as “the most important and most difficult” task of the commissar. To maintain such discipline, the document went on, commissars needed to keep a positive, cooperative relationship among the troops. The document contrasted explicitly the difference between “our discipline and Fascist discipline,” highlighting the cooperative, persuasive nature of the former. However, some situations required authoritative action; commissars were tasked with “knowing the moment to impose oneself over all.”

Commissars also needed to keep a close watch on soldiers’ political loyalties, isolating agents provocateurs within the unit and “seeking out secret collaborators in the heart of the unit who would maintain vigilance over any suspicious elements.” 35 Maintaining cooperative relationships

33 Comisario General de Guerra to Jefe del Batallón Disciplinario, 6 April 1937. CDMH, PS Santander L, caja 562, carpeta 6, p. 19; Agente de Movilización y Control, “Rafael Gutiérrez Fernández,” n.d. CDMH, PS Santander L, caja 436, carpeta 17, exp. 3-5; file on Ricardo Iglesias Cheda, CDMH, PS Santander L, caja 412, carpeta 21, expediente 3. 34 For example, representatives of Popular Front Organizations of Riotuerto to Jefe del Estado Mayor, Ejército del Norte, 27 April 1937. CDMH, PS Santander L, caja 444, carpeta 4, p. 1; PSOE, Agrupación de San Salvador, to Comité Ejecutivo del Federación Socialista, 27 May 1937. CDMH, PS Santander L, caja 444, carpeta 13, exp. 10. 35 “Organización del comisariado y actuación del comisario.” No author, date, or location. CDMH, PS Santander L, 544/5/1-4. 266

among soldiers was difficult under such circumstances. Soldiers frequently found that they had

to be careful about their words lest fellow soldiers denounce them, and they sometimes decided

to leave when the climate of rumour became too problematic. Luís Ayestarán Ayestarán, a

soldier in Battalion 134, was denounced by other soldiers, who said he celebrated Francoist

victories and insulted his compatriots. In his trial for desertion, he attested that he was driven to

leave the unit by these rumours about him. 36 Another soldier, Francisco García Álvarez, told

other soldiers that the war was lost if Bilbao were to fall; a fellow soldier alerted the company

commander, who, on the basis of this and other incidents, ordered García Álvarez detained,

whereupon he tried to defect to the enemy lines. 37

Such problems of cooperation, emerging from dubious political commitments, could reach a crisis point. Francisco Fervenza, commanding the 12 th Brigade, faced serious tensions with militiamen accusing conscripts of being secret right-wingers, including sinister hints of getting rid of those conscripts. Fervenza gathered all 2,000 men of the brigade in a field at

Ruerrero in Valderredible, giving a speech in which he announced that he would shoot any man who killed another in the unit. Those who attempted to defect would have to be tried and, if found guilty, executed, but this justice dealt with acts and not thoughts: “no one is guilty just because of his way of thinking. No one!” Subsequently, according to Gutiérrez Flores, several right-wing soldiers in his unit were actually decorated for valour. 38 But such an approach was

relatively uncommon.

The impact of homogeneity and heterogeneity among soldiers also differed among

conscripts, especially as the war went on. One soldier from Battalion 125, who went with others

from Liébana to Santander to volunteer at the start of the war, remembers: “In Battalion 125 we

36 Case of Luís Ayestarán Ayestarán, CDMH, PS Santander L, caja 406, carpeta 6, pp. 42-46. 37 Case of Francisco García Álvarez, CDMH, PS Santander L, caja 406, carpeta 8, expediente 6. 38 Gutiérrez Flores and Gudín de la Lama 2005, 215. 267 were mostly Lebaniégos, fighting to defend our own.” 39 (The First Company of this battalion, which appears in my dataset, recorded 17 Lebaniégo soldiers: fifteen volunteers, two conscripts.)

However, connections between soldiers could also increase the propensity for desertion. There were several plots among soldiers to desert together, and these plots rested on trust. Reports of the detention of prospective deserters sometimes noted that the deserters had attempted to bring others in on a plot to leave together, but were then betrayed to a commanding officer. 40

Connections among soldiers, therefore, could sometimes facilitate the success of these plots.

Friendship and family ties could indicate problematic motives. For example, cousins of deserters would be detained on suspicion of deserting themselves. 41 The head of the Engineers’ Battalion

wrote, explaining why he wanted to sideline the aide to the commander of the Engineers’ unit at

Reinosa (who had defected): “As is logical, a man who has won the confidence of a deserter and

spy cannot win our confidence.” 42 Groups of soldiers from the same hometown became sources of suspicion. For example, the head of Battalion 131 wrote to the head of the General Staff on 24

June 1937, reporting that on 19 April many soldiers joined his battalion, desiring to serve together since they were from the same hometown. He believed that this may have been a plot to desert all together. 43

Thus, with the influx of conscripts into the CES came fundamental changes in the

dynamics of military units. The initial enthusiasm of the small number of volunteers from the

early days was now joined by the forced service of the uncommitted. They intensified a climate

39 Obregón Goyarrola 2007, 122. 40 Letter from Comisario Delegado, 1st Division to Comisario General de Guerra, Santander, 16 June 1937, CDMH, PS Santander L, caja 406, carpeta 7, expediente 3; trial record, Tomas Cosio Gonzalez, April 1, 1937, CDMH, PS Santander L, caja 406, carpeta 8, expediente 5. 41 Oficial informador 3 Division to Jefe 2 Secn EM, 23 July 1937, Reinosa, CDMH, PS Santander L, caja 406, carpeta 5, p. 101; 42 Comandante Jefe and Capitán Ayudante, no date, recipient unknown. CDMH, PS Santander L, caja 436, carpeta 14, expediente 1-2. 43 Letter from Mayor Jefe, Batallon 131, to Jefe, Estado Mayor, 24 June 1937. CDMH, PS Santander L, caja 406, carpeta 7, expediente 11. 268

of suspicion around their convictions, suspicion which undermined the ability of soldiers to trust

each other. The occasional infectious enthusiasm of a group of volunteers from the same place or

union all joining together to fight against Fascism now had a counterpart in networks of deserters

helping each other out instead.

7.8. Conclusion

The findings presented in this paper have, in general, confirmed the importance of a shared commitment to the faction’s goals. Such a shared commitment is considerably more likely among volunteers than among conscripts, and the higher the proportion of conscripts in a soldier’s company, the higher his likelihood of desertion. In addition, factionalism was also found to spur higher desertion rates, as soldiers in politically polarized units were more likely to desert. Social heterogeneity, in contrast, has an association with desertion that is contingent on the proportion of conscripts in the unit. Social homogeneity is only associated with lower desertion rates where there is a large proportion of volunteers, not where there is a large proportion of conscripts. In the latter scenario, it appears that social homogeneity is no help in keeping soldiers fighting together. Social ties may even, from time to time, facilitate deserting together.

Thus norms of cooperation can have an important effect reducing desertion rates. But

they depend upon the existence of a common aim, which it is untenable to assume across units;

instead, it is an important variable condition, to be assessed empirically. The relevance of

common aims and of norms of cooperation therefore challenges two powerful perspectives on

armed groups: a focus on selective incentives alone as solutions to the collective action problem,

and a focus on control as the only real way of maintaining cohesion in an environment in which

269 motivations are chaotic and difficult to discern. Finally, the contingent importance of social homogeneity confirms a finding long argued in military sociology, that social cohesion by itself does not keep a group fighting; task cohesion, the existence of a common aim, is primary.

The restrictions on the effectiveness of norms of cooperation suggest that groups that share common aims enjoy more options than groups that do not. The latter must depend on control to keep a low rate of desertion. While control can indeed be effective at solving the collective action problem, the analysis in this chapter suggests that armed groups with a high degree of consensus on a common aim do not need to rely on control at all times. They can also

employ norms of cooperation. If control has its own costs, then armed groups that can develop

norms of cooperation can better afford to eschew control. This argument suggests an elective

affinity between styles of recruitment and styles of military operations. Some military tasks, such

as guerrilla warfare and commando raids, allow for less oversight. They therefore require greater

degrees of trust. 44 If, as my argument suggests, large, disparate and especially conscripted armed groups require greater degrees of monitoring to maintain cohesion, such forces may be more inclined to conventional war than to guerrilla war. The argument developed here also suggests one important way in which social ties intersect with warfare: a combatant’s attachment to aims of the armed group mediates how he or she draws on local ties. Such ties neither persist unchanged into soldiering life, nor are superseded by the military experience.

Finally, my argument suggests some thoughts about the broader context of civil war.

Sharing a common aim can just mean political narrowness. Members of different unions did not apparently work well together in the same companies. By the same token, for example, ethnically narrow armed groups might be able to hold together well. The more armed groups represent the full array of political opinion and social diversity in their ranks, the more they may

44 Weinstein 2007. 270

experience desertion. Bashar al-Assad’s narrow base may aid the cohesion of his forces faced

with a widespread but fractious opposition, and the popularity of ousting Muammar Gaddafi in

Libya may make it harder to build cohesive security institutions after his exit. It is a depressing

thought: political narrowness can be a military advantage.

7.9. Appendix

7.9.1. Robustness Checks

The core findings are generally robust to alternative specifications. The first such alternatives are given in Table 7.7. I added fixed effects for the time period in Model 6 to control for the impact of events at different periods of time during the war. The results remained nearly unchanged, except that the county and age heterogeneity variables lost statistical significance. I measured alternate codings for heterogeneity in political goals. First, I grouped soldiers together by affiliation, so that, for example, if a soldier was a member of the UGT or the PSOE, he was coded as a Socialist. I thus grouped the soldiers by five basic affiliations: Socialist, Communist,

Anarchist, Republican and none. As is apparent in Model 7, this variable is likewise statistically significant. However, using fragmentation by political party instead of union changes the result substantially. Unfortunately, since some companies had no party members—that is, all of their soldiers were members of unions rather than parties—there was no clear way of measuring fragmentation in such cases. Thus almost 500 soldier-periods drop from the analysis. The party fragmentation variable itself is negatively signed and statistically insignificant. Ultimately it appears as though it was fragmentation by unions that drove any relationship between heterogeneity in political organizations and desertion rates. In Model 9, I employed an alternative coding of conscript versus volunteer. Whereas in previous models I had coded

271 soldiers who were serving in the Republican military before the start of the Civil War as conscripts because they typically had been conscripted to join in the first place, here I code them as volunteers. Doing so renders the age and county fragmentation variables statistically insignificant.

Further robustness analysis is given in Table 7.8. I re-ran model 3 using a different coding of the dependent variable (Model 10), including only those who were listed as deserting, not those who were listed as disappeared. The company’s conscription rate retains a statistically significant, positive relationship with desertion. However, union fragmentation does not, and an alternative model (not shown here) employing fragmentation by broad political affiliations rather than unions shows no statistically significant effect either. The homogeneity variables and interaction effects with conscription also diminish in relevance, losing statistical significance across the board. This difference is of interest. It is clear that there was some inconsistency in record keeping on the Republican side. Returning to the data from Chapter 6 based on a full census of deserters as given by monthly lists and individual desertion reports, of the 1305 deserters listed, there were 56 individuals who had a report of a disappearance on their war record; of those, 37 had war records that only had a report of disappearance, saying nothing of desertion. This indicates some inconsistency in reporting across different sections of the army, with confusion over who had deserted and who had disappeared for other reasons. What impact was this inconsistency likely to have on results?

In this context, there is some evidence to suggest that large-scale incidents of desertion were recorded by Republican record-keepers as disappearances rather than desertion.

Statistically, the rate of disappearance varied quite substantially more than that of desertion

272

Table 7.7. Robustness checks

Model 6 Model 7 Model 8 Model 9 Time fixed effects Alternate affiliation Alternate affiliation Alternate conscription coding coding coding b (s.e.) b (s.e.) b (s.e.) b (s.e.) Conscript .444 (.344) .441 (.343) .430 (.341) .294 (.318) Left-wing affiliation -1.348 (.187) *** -1.347 (.187) *** -1.338 (.187) *** -1.381 (.186) *** Officer -.928 (.481) + -.918 (.480) + -.920 (.481) + -.909 (.489) + Non-commissioned officer .076 (.234) .070 (.233) .073 (.233) .067 (.236) Age .037 (.021) + .040 (.021) + .040 (.021) + .040 (.021) + Company conscription rate 3.636 (1.264) ** 4.122 (1.198) *** 4.655 (1.312) *** 2.753 (1.080) * Company union fragmentation 5.180 (2.095) * 5.283 (2.097) * Company affiliation fragmentation 4.606 (2.092) * Company party fragmentation -.284 (1.477) Company county fragmentation 4.960 (2.655) + 4.647 (2.544) + 3.825 (2.417) 3.188 (2.260) Company job fragmentation 12.995 (6.646) + 17.547 (6.792) *** 16.128 (6.726) * 16.651 (6.810) * Company age variation 16.590 (9.630) + 12.763 (9.136) 18.389 (10.008) + 3.024 (6.887) County fragmentation x conscription -7.679 (6.490) -5.994 (6.339) -6.427 (6.212) -3.431 (7.450) Job fragmentation x conscription -23.292 (15.646) -31.696 (15.624) * -33.693 (15.856) * -31.457 (15.659) * Age variation x conscription -30.112 (19.510) -27.113 (19.923) -37.418 (21.012) + -6.608 (18.093) Home outside Santander .144 (.393) .135 (.394) .101 (.394) .133 (.392) Hometown altitude .001 (.000) *** .001 (.000) *** .001 (.000) *** .001 (.000) *** Time since entry: 2 months .681 (.260) ** .609 (.252) .612 (.252) * .579 (.252) * Time since entry: 4 months .597 (.341) + .614 (.335) + .624 (.335) + .595 (.335) + Time since entry: 6 months .232 (.581) .052 (.557) .054 (.557) .040 (.557) March-April 1937 .029 (.891) May-June 1937 .955 (.865) July-August 1937 -.115 (.984) Constant -7.179 (1.060) -6.951 (.743) *** -6.928 (.745) *** -6.484 (.700) *** Company-level variation .258 (.329) .428 (.318) .454 (.314) .463 (.315) N 8981 8981 8508 8981 Log-likelihood -616.565 -621.147 -620.763 -623.081 BIC 1442.496 1424.352 1422.500 1428.219 *** p < .001; ** p < .01; * p < .05; 1 p < .01. Dependent variable in all models: 0 if soldier did not desert in time period; 1 if he did.

273

Table 7.8. Robustness checks 2.

Model 10 Model 11 Model 12 Alternate dependent Rare-events sampling, Rare events sampling, variable coding simple logit company random effects b (s.e.) b (s.e.) b (s.e.) Conscript .772 (.481) .166 (.399) .341 (.473) Left-wing affiliation -1.330 (.254) *** -1.715 (.246) *** -1.806 (.302) *** Officer -.994 (.741) -.700 (.559) -1.059 (.643) + Non-commissioned officer -.157 (.363) .353 (.313) .151 (.369) Age .047 (.027) + .064 (.024) ** .047 (.030) Company conscription rate 2.820 (1.190) * 4.216 (.917) *** 4.001 (1.384) ** Company union fragmentation 3.129 (2.009) 2.525 (1.351) + 3.402 (2.281) Company county fragmentation 2.660 (2.428) 5.847 (2.126) ** 5.455 (3.123) + Company job fragmentation 9.337 (6.420) 12.855 (4.736) ** 13.371 (7.051) + Company age variation 14.442 (9.147) 16.226 (6.133) ** 16.515 (10.277) County fragmentation x conscription -2.196 (6.656) -5.845 (5.076) -8.999 (7.533) Job fragmentation x conscription -27.055 (14.938) + -27.402 (11.048) * -27.009 (16.604) Age variation x conscription -7.541 (20.116) -30.921 (13.048) * -30.350 (21.965) Home outside Santander -.268 (.591) -.216 (.464) -.207 (.539) Hometown altitude .001 (.000) *** .001 (.000) * .001 (.001) * Time since entry: 2 months .750 (.300) * .644 (.260) * .679 (.350) + Time since entry: 4 months .403 (.494) .070 (.366) .205 (.463) Time since entry: 6 months .050 (.804) -.629 (.579) -.255 (.723) Constant -7.258 (.925) *** -3.546 (.764) *** -3.747 (.959) *** Company-level variation -.002 (412) .500 (.385) N 8981 688 688 Log-likelihood -616.565 -264.227 -244.832 BIC 1442.496 652.596 620.341 *** p < .001; ** p < .01; * p < .05; 1 p < .01. Dependent variable in all models: 0 if soldier did not desert in time period; 1 if he did.

narrowly defined: while the average desertion rate was 1.2% with a standard deviation of 2.3%, the average disappearance rate was 1.0% with a standard deviation of 3.9%. As much as 30% of one company—the 2 nd company of Battalion 139—disappeared in a single time period. In

addition, there appears to be some association between disappearance and battle. The mean death

rate in companies where any soldier disappeared was 2.1% (s.e.: 0.6%), and the mean death rate

in companies where no soldiers disappeared was 0.55% (s.e.: 0.1%), a statistically significant

difference. There was no such difference between companies where at least one soldier deserted

and companies in which no-one deserted. It may be that this is entirely explained by taking

prisoners of war after a battle. However, from qualitative accounts, there was a relationship

274

between battles and desertion proper, such as in the November and April offensives towards

Vitoria: soldiers seemed to desert more frequently just before, during, and just after battles. 45 In addition, much turns on one’s interpretation of particular incidents. The case of two soldiers from

Battalion 105, in June 1937, generated a large paper trail, including requests for more information to determine whether the soldiers deserted or just disappeared.46 In Battalion 116

there was a list of 44 soldiers “whose whereabouts are unknown” but who “according to reports”

disappeared on the Basque Country front.47 On 3 July 1937, a list of 37 soldiers from Battalion

124 was issued, reported as having disappeared. However, the investigation section of the

General Staff issued a note on 10 August 1937 indicating scepticism that this number of soldiers could have disappeared at the same time, and asking for further information from the battalion commander about the circumstances of the incident and about the soldier’s political pasts, in order to determine whether they were deserters or not. The same list of soldiers was later reproduced, with “disappeared” changed to “deserters”. However, the individual requests for information and vigilance of friends and family that were issued to the mayors of the soldiers’ hometowns on 10 August listed some soldiers as deserters and others as disappeared. Whether the army reported a soldier as a deserter or as disappeared seemed, then, to be an unclear, complicated, and difficult process, and one potentially affected by the army’s own beliefs about the political origins of desertion.48 Ultimately, therefore, there is a strong enough suspicion that

“disappearance” was a cover for desertion—frequently mass desertion—that it is sensible to treat disappearance as desertion.

45 Martínez Bande 1980, 221; Solar 1996, 85–86. 46 CDMH, PS Santander L, caja 443, carpeta 8, expediente 3. 47 CMDH, PS Santander E, caja 84, carpeta 3, folio 75. 48 CDMH, PS Santander L, caja 443, carpeta 9, expediente 4. 275

As a final robustness check, I reran the analysis as a rare-events logit (Models 11 and 12, table 7.8). As noted in Chapter 6, because of the nonlinearity of the logit function, there is a bias that emerges when the proportion of “successful” events (1s) is very small. Mirroring the analysis in Chapter 6, I took all 172 deserters (including soldiers who disappeared) on whom I had full information and a random sample of 516 non-deserters (three for every deserter).

Rerunning Model 3 from Table 7.1 showed no substantial difference whatever in the results.

7.9.2. Discrepancies between Chapter 6 and Chapter 7

The results thus far presented show some important, and puzzling, differences from the

results in Chapter 6. First, referring back to Models 1 and 2 in Table 7.4, the hometown-level

variables that were so important in Chapter 6—local terrain steepness and the local left-right vote

split—have coefficients of dramatically reduced size, and are not statistically significant; and in

this chapter’s data, altitude has a much more robust relationship to desertion. Second, whether a

given soldier was a conscript or a volunteer was apparently not relevant.

The discrepancies have three possible sources. First, the new analysis may account for

variables that are unexplored in Chapter 6, suggesting that Chapter 6’s analysis is incomplete

because it is spurious on variables identified here. Second, the difference in sampling methods

may generate sample selection biases, such that the relationships that are found are limited to

different portions of the data. Finally, the discrepancy may be a result of random error: different

samples, different results; in which case, the results are simply not robust.

In the case of hometown-level variables, we can eliminate the first explanation. I re-

analyzed the data from this chapter removing the company characteristics, so that the model is

left with the individual characteristics and local political variables as in Chapter 6. I used several different strategies for estimating such a model. I first used soldiers as my unit of analysis,

276

grouped within municipalities, as in Chapter 6, thus not taking account of their tenure in a given

company or the cluster sampling by company. I used an individual-level model with standard

errors clustered by company to take account of the sampling method. I used the soldier-period as

my unit of analysis, using the method employed in Chapter 7. The results are unequivocal in the

case of municipality indicators: in each case, altitude had a statistically significant, positive

relationship to desertion, but neither steepness nor local right-wing vote share did. Thus it is not

that paying attention to military unit composition eliminates the influence of hometown

characteristics. Steepness and right-wing vote share have little association with desertion here

even when unit composition is left out of the picture.

There may be other explanations for the discrepant findings. Was there a bias in the sampling? This chapter restricts its sample to infantry companies, excluding engineers, medical personnel and others, and I do not include soldiers in the plaza mayor or general staff of each battalion. This is a bias but it seems unlikely that it would have a large effect on my results: personnel from these excluded units were 23% of my sample from Chapter 6. In any case, rerunning Chapter 6’s models on a sample restricted only to include soldiers in infantry companies and not those in the plaza mayor of the battalion, the results are effectively no different.

However, is entirely possible that sampling a small number of companies allows for discrepancies in sampling to emerge. Consider the county of Liébana, among the steepest areas of Santander province on average. I estimate the average desertion rate of Lebaniego soldiers as

8.82% according to my data in Chapter 6 (weighted to account for the probability of sampling).

In contrast, in the data for this chapter, I have two deserters out of 138 Lebaniego soldiers, or

1.45% (though five disappeared bring the percentage to 5.07%). Sampling error seems like the most plausible source of difference between the two chapters’ results. The consequence is that

277

Table 7.9. Individual- and company-level conscription Model 13 Model 14 b (s.e.) b (s.e.) Conscript .958 (.318) ** .415 (.340) Company conscription rate 2.458 (.698) *** Left-wing affiliation -1.348 (.189) *** -1.369 (.188) *** Officer -.844 (.478) -.934 (.480) + Non-commissioned officer .125 (.233) .055 (.233) Age .042 (.021) * .042 (.021) * Home outside Santander .002 (.392) .149 (.391) Hometown altitude .001 (.000) *** .001 (.000) *** Time since entry: 2 months .607 (.252) * .638 (.253) * Time since entry: 4 months .662 (.337) * .642 (.336) + Time since entry: 6 months .104 (.562) .085 (.559) Constant -6.452 (.660) *** -6.238 (.662) *** Company-level variation .881 (.291) ** .752 (.299) * N 8981 8981 Log-likelihood -636.317 -629.624 BIC 1381.869 1377.586 *** p < .001; ** p < .01; * p < .05; 1 p < .01. Dependent variable in all models: 0 if soldier did not desert in time period; 1 if he did.

we can have less confidence in the robustness of the results for right-wing vote share and steepness, and more confidence in the robustness of altitude.

In contrast to hometown effects, the association between a soldier’s conscription status and his likelihood of desertion may be spurious on unit composition. Comparing a model which only includes individual and hometown variables to one with those variables and adding the company’s rate of conscription (see Table 7.9), the coefficient on the individual-level conscript vs. volunteer variable is no longer significant, while the overall conscription rate is, indeed, positive and significant. Thus there is a significant chance that the association between being a conscript and deserting is spurious on the overall conscription rate, a finding that casts doubt on one of the most consistent results from Chapter 6. Substantively, if correct, this proposition would add considerable support to arguments derived from the context of serving and the

278 preferences of others. Even a conscript has a relatively low chance of deserting if he serves in a unit that is mostly volunteers; even a volunteer has a relatively high chance of deserting if he serves in a unit that is mostly conscripts.

279

Chapter 8 Defection in Civil Wars, 1990-1994

8.1. Introduction

Thus far the dissertation has had much to say about the Spanish Civil War at a micro level. In the previous chapters, I outlined the micro-dynamics of differentiating among reliable and unreliable recruits, of maintaining control over potentially untrustworthy soldiers, and of facilitating mutual trust and reciprocity among committed soldiers. I laid out the plausibility of recruitment and factionalism as factors in the distribution of motivations in armed groups, and demonstrated how control and norms of cooperation worked to keep soldiers from deserting, contingent upon the initial distribution of motivations. I then subjected my hypotheses about the general importance of control and the contingent importance of norms of cooperation to micro- level statistical tests.

In this chapter I expand the scope of the dissertation by turning to macro patterns in the reliability of soldiers. I examine the relationship between a state’s material capabilities and the mass defection of its soldiers in civil wars. State capabilities indicate the rewards and punishments that states can implement from the top down—the threats that soldiers face if they choose to defect, and the rewards they can receive if they do not. More capable states can more credibly threaten soldiers with punishment for defection. They can maintain material rewards, such as wages, more consistently. And since these governments appear stronger in the first instance, defectors are more likely to be caught on the losing side. The earlier chapters of the dissertation focus on control as a crucial instance of top-down rewards and punishments, but in

280

this chapter I broaden the discussion, with state capabilities—specifically those provided by

external actors—as a broader indicator of the capacity to dispense rewards and punishments.

Principally, the chapter finds that in a set of 28 civil wars in the 1990s, governments that

received military assistance from abroad were generally more successful than others at

preventing desertion. This effect held even if soldiers were more uncommitted, as in conscript

armies; but it did not hold if factionalism was particularly severe, as in countries that had

recently experienced a coup attempt. This confirms the dissertation’s main hypotheses about

rewards and punishments, which I examined in the previous chapters by focusing on control. I

presented a theoretical argument in Chapter 1 for why rewards and punishments should work

even when combatants are uncommitted, provided that disagreements are not too severe. The

empirical findings in this chapter confirm the general importance of this argument.

The mechanisms outlined in the previous chapters are highly general in nature. This

means that, as the previous paragraph hints, the move to the macro level to generate more

substantive predictions about broad patterns requires some theoretical groundwork. In the first

section, then, I engage in further theoretical development. Across civil wars, what general

features indicate state capabilities? Second, I move on to outline the empirical study underlying

this chapter. I then discuss my results in the third section, before concluding with suggestions

about where the findings in the chapter leave us.

8.2. From Micro to Macro

The purpose of this chapter is ultimately to see whether the dissertation’s central arguments provide insights for the understanding of contemporary conflict across an array of cases. Specifically, it examines defection from the government side. Government defection is

281 particularly important because government armies generally—though, importantly, not always— begin civil wars with a preponderance of power. 1 Early scholars of military behavior in the face of uprisings, such as Katherine Chorley and D.E.H. Russell, argued that this preponderance meant that rebels could not easily succeed without defection. 2 More recent scholarship has

focused instead on how rebels can overcome the gap in capabilities by adopting good strategies

that focus on harassing government forces while buying time to strengthen, and getting third-

party counterinsurgents to exit the country through eroding their political will to keep up an

intervention. 3 Work on the government side has examined how governments counter these

strategies and how successful they are at it, with the major focus placed on the strategy of

counterinsurgency and its implementation. 4 Scholarship on defection in civil wars has focused

much more on defection from the rebel side, especially on the issue of why members of ethnic

out-groups in rebellion fight for government forces such as in Sri Lanka.5

This chapter focuses on defection—instances where soldiers split off from the government to a rebel group, whether one that is already in existence (side-switching) or to create a new rebel group (fragmentation). It does not discuss desertion to the rearguard, unlike the rest of the dissertation in which all three destinations (rearguard, the other side, and a new group) are lumped together. The reason is that defection is a particularly important, and particularly visible, form of unreliability. It not only poses more significant dangers than desertion, because it entails new enemies and not just the loss of fighters, but also leaves more visible empirical traces. In a low-information environment, which is what scholars of civil war

1 Cunningham, Gleditsch, and Salehyan 2009. 2 Chorley 1943; Russell 1974. 3 Mack 1975; Merom 2003; Arreguín-Toft 2006. 4 Arreguín-Toft 2006; Downes 2007; Lyall and Wilson 2009; Lyall 2010. For practical concerns in counterinsurgency, see Kilcullen 2010; US Department of the Army 2007. 5 Kalyvas 2008; Connable and Libicki 2010; Staniland 2012. 282

frequently have to deal with, defection allows for considerably easier study. Importantly, though,

this means that much unreliability will not be captured here.

Because “recruiting committed followers” and “top-down rewards and punishments” are

relatively general, micro-level social mechanisms, 6 further theoretical arguments are necessary to

understand how these bits of theory generate robust empirical patterns. For example, what

“committed followers” are, and what means exist for identifying them, is likely to vary

considerably by case. Where Spanish Anarchists might designate someone as a possible Fascist

for brushing his teeth (see chapter 5), other civil wars will engage other identifying markers. The

Bolsheviks in the Russian Civil War looked for callused hands to identify supporters. 7 Identity groups can be based on language, accent, religion, or physical resemblance, and can be of highly variable import. 8 Moreover, some of the factors important to the Spanish Civil War are unlikely to travel very far; some groups will be much better able to put aside class and political differences than the Spanish Republic, for example, while, in other contexts, lootable natural resources and international smuggling networks will play a greater role in creating incentives to join armed groups. 9 Finally, some variables of vital interest at the micro level, for example the intersection between the proportion of conscripts in a company and its social homogeneity (see

Chapter 7), will be too data-intensive to really be useful in generating cross-case generalizations.

Accordingly, this section fleshes out what recruiting unreliable followers, and providing rewards and punishments, are likely to mean across different cases. I begin with top-down rewards and punishments.

6 Elster 1989; Hedström and Swedberg 1998; Tilly 2001; McAdam, Tarrow, and Tilly 2001. 7 Kalyvas 2008, 1048. 8 Brubaker 1996; Horowitz 2000; Fearon and Laitin 2000; Brubaker 2004. 9 Keen 1998; Collier 2000; Collier and Hoeffler 2004; Weinstein 2007. 283

8.3. State Capabilities and External Support

A state’s material capabilities, such as its revenue base, military budget, and stock of military hardware, have a large impact on its ability to punish defection and reward loyalty to keeping soldiers fighting. There is, indeed, a long theoretical tradition linking the state’s ability to dispense rewards and punishments with its ability to get its agents to do its bidding. Tilly, for example, identifies two routes to state building, coercion-intensive and capital-intensive: the latter permits the state to purchase what it needs, the former to compel it.10 Similarly, Wintrobe considers material rewards and punishments two alternative methods for dictators to induce compliance among citizens.11

The link between material capabilities and the provision of rewards is straightforward: the larger the budget, the more agents can be paid for their services. Hence, a strong strand of literature on state capacity focuses on raising revenue. 12 In the context of defection, what is of

particular importance is whether the budget can be maintained reliably. For instance, if a state’s

budget collapses and it cannot pay its troops any more, then the balance for a combatant can

easily tip against serving and towards defecting, especially to a warlord with the ability to pay.

This is one reason for the frequently-noticed correlation relationship between economic crisis

and authoritarian breakdown, for instance in Indonesia in 1998 following the Asian financial

crisis of the previous year.13

In addition, the state’s capabilities strongly affect its ability to control. The state’s control by coercion over its agents is squarely in the Hobbes-Weber tradition of statism. As Weber puts

10 Tilly 1992. 11 Wintrobe 1990. 12 Levi 1988; North 1982; Olson 1982; Olson 1993. 13 O’Donnell and Schmitter 1986; Bratton and van de Walle 1997; Levitsky and Way 2010; on Indonesia see Lee 2005. 284

it, “the use of physical force...is neither the sole, nor even the most usual, method of

administration of political organizations....But, at the same time, the threat of force, and in the

case of need its actual use, is the method which is specific to political organizations and is

always the last resort when others have failed.” 14 For Tilly, the performance of the state’s core tasks “depend on the state’s tendency to monopolize the concentrated means of coercion.” 15

Force helps not only to end rebellion directly but also to keep agents from defecting in the course of doing so. Tilly identifies coercion as a crucial way of ensuring that agents actually perform the tasks that they are entrusted with:

Other authorities...are much more likely to confirm the decisions of a challenged

authority that controls substantial force; not only fear of retaliation, but also desire to

maintain a stable environment recommend that general rule. The rule underscores the

importance of the authority’s monopoly of force. A tendency to monopolize the means of

violence makes a government’s claim to provide protection, in either the comforting or

the ominous sense of the word, more credible and more difficult to resist. 16

Control, however, is costly to impose. Monitoring soldiers means diverting other

soldiers’ activities from direct combat to policing their fellows. It can also mean taking

suboptimal tactical or operational approaches, such as failing to delegate to autonomous and

flexible units. In Chapter 4, I discussed how commanders in Republican Spain allocated

monitoring by trying to determine who needed to be monitored—an indication that control was

not costless. If control is costly to impose, then an armed group should be more able to impose

control to the extent that it has greater material capabilities at its disposal.

14 Weber 1978, vol. I, p. 54. 15 Tilly 1985, 181. 16 Ibid., 171–172. 285

Material military capabilities also measure the general capacity of the state to intimidate

potential opponents with the prospect of losing. The more powerful the state, the more it will be

able to inflict punishments upon those who defy it, including defectors no less than the initial

rebel group. Entry into rebellion against the state, via side-switching or fragmentation, is

consequently likely to be considerably less rewarding to the extent that the state is powerful.

Again, there is substantial evidence of these dynamics in the Spanish Civil War: the greater

willingness of Republican soldiers, including officers, to defect as the Nationalists racked up

victory after victory; and the fact that both Anarchists (on the Republican side) and Falangists

and Carlists (on the Nationalist side) were deterred from active rebellion in the rearguard by the

power of the Republican and Nationalist armies generally. Hence the capability of the armed

groups affected both side-switching and fragmentation.

How should capabilities be measured? One might first be inclined to examine the

government’s existing military capabilities, such as the number of men under arms or the amount

of military equipment at its disposal. One could also examine the military balance between

rebels and states. Kalyvas, for instance, argues that civilians’ defection decisions follow the

military balance between rebels and government in the territory where the civilians are living. 17

Cunningham, Gleditsch and Salehyan confirm a straightforward claim: the balance of state and rebel power, construed as a property of the state-rebel group dyad, is of vital importance to the prediction of civil war duration and outcome. 18 As rebels’ power grows, and the prospect of rebel victory increases, the agents of the government may become more tempted to defect. However, the difficulty with just examining these indicators is their endogeneity. As defection increases, the military’s stock of men and equipment obviously declines. In addition, since many rebel

17 Kalyvas 2006. 18 Cunningham, Gleditsch, and Salehyan 2009. 286

groups have begun with military defection, 19 and since defectors can be an important source of rebel military strength, it will generally be difficult to assess whether it is the military balance that is causing the defection or the other way around. One could, potentially, isolate two moments in time—the military balance at the moment of the outbreak of war, and subsequent defection down the line. However, this temporal strategy may not fully eliminate endogeneity concerns. Civilians, anticipating future defection from the government at t2, could be more likely to join a rebel group at t1. In that scenario, the tendency of a state to suffer defection may be the cause of a military balance unfavourable to the incumbents even if it follows that balance in time.

Rather than looking at the balance between the two sides, then, I examine the characteristics of the government side alone. In particular, I focus on external military support.

Because the dependent variable is the behaviour of the military, it is the armed forces’ material capabilities that are most at issue in judging whether the state can dispense rewards and punishments to its soldiers. In contrast, general fiscal indicators like the overall budget or GDP per capita, while surely relevant, are one step removed. The unreliability of military budget data, especially in countries prone to conflict and most particularly in the context of state collapse, also tells against official budget statistics as an indicator.

External military support is an important source of government military capabilities, affecting both the dispensing of rewards and punishments. In the contemporary world of internal conflicts, located principally in the developing world, an important source of material capabilities is external military support. Experiencing an armed conflict is associated with a substantial increase the value of arms transferred to a state. Although interpretations differ as to

19 Kalyvas and Balcells 2010. 287 whether this is an indicator of demand for arms or an indicator of the ability of arms to facilitate conflict, 20 this appears to be splitting hairs: presumably, arms would not be demanded if they

were not useful in armed conflict. Arms embargoes may sometimes be put into place to help

dampen a conflict. However, such embargoes are routinely violated, including, very often, by the

five permanent UN Security Council powers. The globalization of production also makes arms

transfer controls extremely difficult to implement.21

Thus few barriers really exist to preventing the transfer of arms to states at war, and states

experiencing civil wars are much more likely to arm. As Mohammed Ayoob argues, “the internal

dimension of security, which is inextricably intertwined with the process of state making, is the

core variable that determines the Third World state’s security problematic.” 22 Given the

predominance of internal conflict over international conflict since 1945, 23 and the overriding

importance of retaining power internally, 24 states vulnerable to such conflicts tend to treat internal war as much more important than international war in their foreign-policy behaviour.

For example, during the Cold War, developing-world states frequently made alignment decisions with a view to obtaining the most effective support for the defeat of internal enemies,25 while

leading powers used such transfers to obtain amenable foreign policies. 26 If a particular group in

an armed conflict has access to significant external military support, this support can often help it

to defeat its rivals. 27 In contrast, when states do not have access to external support, their military capacity correspondingly suffers.

20 Craft and Smaldone 2002; Comola 2012. 21 Cooper 2006; Wallensteen 2008; Moore 2010. 22 Ayoob 1995, 21. 23 Holsti 1996; Themnér and Wallensteen 2011. 24 Bueno de Mesquita et al. 2003. 25 David 1991. 26 Krause 1991; Sislin 1994; Kinsella 1998; de Rouen and Heo 2004. 27 Balch-Lindsay, Enterline, and Joyce 2008; Peic and Reiter 2011. 288

Thus a major shift occurred with the end of the Cold War: it implied a substantial

reduction in the capacity of states to fight their internal conflicts. The Cold War had seen

enormous superpower military assistance to states around the world, 28 and after 1990, Soviet clients especially suffered a dramatic decline in military aid. Thus, for example, Jeffrey Herbst argues that many African armed forces declined precipitously in capacity once the pipeline of superpower aid was shut off, with many paper military units no longer able to function effectively in the face of rebellions. 29

Since it increases state military capacity dramatically, external military assistance may limit defection. It can give regime elites the material resources—military equipment and supply, funding, and foreign troops—necessary to keep agents in line. Toward the end of the Cold War, states facing dramatic cutbacks in external support were no longer able to pay state agents or to credibly threaten them, and so state agents were able increasingly to engage in illicit anti-regime activity. 30 Stedman argues that the end of the Cold War “undermined the external sources of support for Africa’s patrimonial regimes and left some with no legs to stand on.” 31 It is relatively

straightforward to extend this to the military context, and anticipate that the state’s agents—

soldiers and officers—should be more likely to defect from their principals. One large-N finding

that bolsters this view is that arms transfers have tended to reduce the likelihood of coups

d’état. 32 Kalyvas and Balcells, in the context of civil wars, uncover a further suggestive finding:

the end of the Cold War was accompanied by an increase in the frequency of conventional as

opposed to guerrilla civil wars. They argue that this was because rebels could now hope to obtain

28 Westad 2005. 29 Herbst 2004. 30 Reno 1998. 31 Stedman 1996, 236, quoted in Kalyvas and Balcells 2010, 423. 32 Wang 1998. Diplomatic support can have a similar effect as well, apparently. Thyne 2010. 289 conventional capacity from disintegrating armies. The link to external support is bolstered by the fact that conventional civil wars were particularly common in new post-communist states such as

Azerbaijan, Georgia, and the former Yugoslavia. 33

The link between external support, capacity, and lower defection rates also has suggestive evidence in the Spanish Civil War. German and Italian assistance was crucial for the strength of the Nationalist army. Notably, German and Italian aircraft were instrumental early on in transporting the Army of Africa across the Straits of Gibraltar. And, again, the power of the armed forces, and especially of the Army of Africa, was an important factor in keeping the

Falange and Carlist militias from rebelling. The increasing strength of the Popular Army of the

Republic in late 1936 and spring 1937 had much to do with arms transfers from the Soviet

Union, such as more than doubling its supply of rifles. 34 Since the Republic used access to arms

and supplies as a crucial tool to bring militias in line, and since its rearguard control activities

involved significant growth in its security services, Soviet arms supplies may be regarded as an

important background condition for the establishment of Republican control, although, as I argue

below, factionalism meant that that control occurred in a climate of severe mistrust and may not

have assisted much.

Is external military support necessarily exogenous to defection? The possibility that states

will attempt to avoid delegating to defection-prone armies suggests that it may not be. Patrick

Regan argues that external states often choose to intervene on the basis of pragmatic concerns,

notably whether their intervention will actually have an impact on the outcome of war. 35 Daniel

Byman further notes that the unreliability of local armies has been a serious difficulty for

33 Kalyvas and Balcells 2010, 422, 426. 34 Moradiellos 1999; Howson 1998, 28, 142. 35 Regan 2000, 43–44. 290

American interventions in the developing world. 36 In effect, supporting an army prone to defection could be equivalent to pouring money into a pit. States attempting to avoid this outcome should avoid aiding defection-prone armies. Therefore, high external military support may be endogenous to a reliable army.

In this chapter, I cannot eliminate the possibility of an endogeneity problem. However, there are reasons to believe that external support can occur and persist in spite of problems of reliability. For example, states support different sides of ethnic conflicts in part because of domestic constituencies, and hence the domestic political benefits they reap from doing so may have little enough to do with the efficiency of the action construed in pure foreign policy terms. 37

In addition, even if external interveners acknowledge that much is wasted from supporting fractious government forces, they may consider the alternatives—such as suspending support— worse. The very fact that, as Byman argues, this problem has persisted in American foreign policy suggests the difficulties of ending such a relationship or of identifying credible and reliable partners in the first place, or a willingness to carry on supporting governments regardless of their unreliability. Loss aversion on the part of leading states has induced them to maintain military positions, including interventions in internal conflicts abroad, long past when it was sensible to do so. 38 In contemporary Afghanistan, for example, there may be few good options:

there is a broadly unreliable Afghan National Army, but one that is likely to crumble further if

and when NATO withdraws. 39 In any case, this question can, to some extent, be assessed

empirically: does external support decline when other indicators of wastefulness are present?

36 Byman 2006. 37 Saideman 2001. 38 Jervis 1992; Levy 1992; McDermott 1998; Snyder 1991; Copeland 2000. 39 Peceny and Bosin 2011; Christia 2012. 291

There is thus a good case for using external military support as an exogenous indicator of a state’s military capabilities. If a government receives external military support, I expect it to have a generally lower incidence of defection.

Hypothesis 1. Governments receiving external military support should experience mass defection less often than governments without such support.

However, the point is well taken that local agents may be divided enough that increasing coercive resources do not actually help much, because they are undermined by severe disagreements arising from factionalism. Such factionalism might therefore erode the effect of external military support at the margins. I discuss this possibility further below.

8.4. Militias, Agency, and Control

Control over combatants might be contingent on institutional arrangements in the

military. If government forces enjoy autonomy and delegation, they may be able to defect more

easily. Important in this regard is the use of civilian militias, which are frequently—though, I

argue below, not always—decentralized forces. The shift from ad-hoc forces raised in a

decentralized, delegated fashion to centralized standing armies was an important part of the story

of state-building in Western Europe. 40 However, states fighting civil wars in the 20 th and 21 st centuries have frequently made use of civilian militias. Out of 130 insurgencies from 1945 to

2011 surveyed by Seth Jones, the government side employed civilian militias in 103. 41 Thus

militias are a considerable part of states’ efforts at fighting civil wars. 42

On the basis of some analysts’ work, we might expect civilian militias to be particularly

prone to defection. This would just be part and parcel of standing outside the state’s military

40 Finer 1975; Tilly 1985; Mann 1986; Dandeker 1990; Tilly 1992. 41 Jones 2012. 42 Joes 2004, 113. 292

hierarchy. For Ariel Ahram, militias manifest M-form (multidivisonal) organization, in which an

agent runs a division that dominates essentially all local functions, rather than U-form hierarchies

in which agents control divisions differentiated by functional form. In an M-form organization,

each unit can operate essentially independently of the others. 43 This type of decentralization

“makes it harder for the state to monitor and control independent militia leaders.” 44 This

autonomy has led to significant concern that militias pose a long-term threat to the integrity of

states. For numerous scholars, the disruption to the Weberian monopoly of violence that

paramilitaries represent is their central theoretical significance. 45 Kimberly Marten argues that

civil militias can descend into warlordism, as in Somalia under Siad Barre and Afghanistan

under Najibullah. The consequence is that the power to kill is in the hands of individuals who

only have incentives to invest in their own patch of turf (if that), rather than the whole state. 46

Hughes and Tripodi note the potential for militias to become involved in illicit economies in which violence carries a comparative advantage, such as in Colombia and Northern Ireland. 47

And scholars have frequently noted militias’ propensity to commit violence for their own purposes once given license to commit violence at all.48 John Mueller writes that “as disciplined, conventional warfare has become comparatively rare, we are increasingly left with the ravages, often savage ones, of irregulars.” In such a context, war is more a criminal activity than anything, with violence frequently committed for private aims. 49 Martin van Creveld echoes this view of

contemporary conflict: “Armies will be replaced by police-like security forces on the one hand

43 Chandler 1966; Williamson 1975. 44 Ahram 2011, 13. 45 Kaldor 2006, 41; Kalyvas and Arjona 2005, 35; Francis 2005, 20; Marten 2006, 43–44; Ahram 2011, 2–3. 46 Marten 2006. See also Olson 1993. 47 Hughes and Tripodi 2009. 48 Shannon 2000; Francis 2005, 20; Kalyvas 2006, 108. 49 Mueller 2004, 86. 293 and bands of ruffians on the other, not that the difference is always clear even today.” 50 There does indeed appear to be an increase in “symmetric non-conventional” civil war, irregulars vs. irregulars. 51 The ultimate resolution of this situation, according to some, requires the reassertion

of centralized and organized violence in state hands. 52

If militias represent a limit to the state’s monopoly of violence, one way this could manifest itself is in defection. Ahram argues that militias’ autonomy means that “they can refuse to comply with the state’s demands, either by neglecting to target those the state identifies as dangerous or by attacking those groups the state identifies as friendly. They can amass their own power base by setting up a ‘state-within-a-state,’ by seceding, or even by marching on the capital to depose the regime.” 53 Hughes and Tripodi argue similarly that “there is no guarantee that surrogate forces…will not resume hostilities against the government and SSF [state security forces] at a time of their own choosing.” Thus, they give the example of the French-sponsored

Force K, which defected to the Algerian nationalist movement in 1956. 54 The Najibullah regime

in Afghanistan became increasingly reliant upon militias during and after the withdrawal of its

Soviet backers. Those militias ultimately turned on Najibullah, their defection prompting his

ouster in 1992. 55 Apart from switching sides to previously existing rebel movements, the arming of civilians can create significant threats of fragmentation, the creation of new rebels.

However, militias might not always be defection-prone. Stathis Kalyvas reports that in preparing his magnum opus, he “was able to find few reports of militiamen collaborating with

50 van Creveld 1991, 225. 51 Kalyvas and Balcells 2010. 52 Kaldor 2006, 11. 53 Ahram 2011, 15. 54 Hughes and Tripodi 2009, 15, 23. 55 Rubin 1995; Sinno 2008; Giustozzi 2009. 294

rebels,” which he finds “surprising.” 56 Given the breadth of his work, this is a statement to be

reckoned with. As Seth Jones argues, militias vary in their propensity to defect. In any case, so

do regular armies. 57 A government’s most determined opponents are quite frequently its own

regular soldiers. 58 As an extreme instance, while members of the regular Sierra Leone armed

forces were operating as “sobels,” soldiers by day and rebels by night, defecting back and forth

at will, the irregular Kamajor militia fought much more reliably for the government for much of

the war. 59 Byman points out that we should be particularly likely to see unreliable state armies in countries experiencing a rebellion, because it is precisely where states are weak that rebellions occur. 60 Unreliable state armies should be particularly unable to snuff out rebellions before they

are capable of mounting a civil war. And, of course, some civil wars begin with the defection of

a section of the armed forces, in cases from Spain in 1936 to Libya and Syria in 2011. 61 We may

well ask if there is really any difference between regular security personnel and state-sponsored

militia in terms of reliability. 62 There has, as yet, been no systematic investigation of this

question; the evidence that we have to rely on is based on only a few cases. Are militias, indeed,

more defection-prone than regular forces?

It is particularly important to understand the behavior of civilian militias because they

have recently received considerable attention as a policy alternative: the delegation of military

force to civilian militias has been an important part of U.S. strategy in Iraq and NATO strategy

in Afghanistan. These policies, however, have provoked serious debate. The “Anbar

56 Kalyvas 2006, 109, fn. 50. 57 Jones 2012, 9–11. 58 Luttwak 1969; Nordlinger 1977; Biddle and Zirkle 1996; Quinlivan 1999; Belkin and Schofer 2005; McLauchlin 2010. 59 Reno 1998; Johnston 2008. 60 Byman 2006. 61 International Crisis Group 2011b; Holliday 2012. 62 In parallel fashion, the involvement of civilian militias in illicit economies should not obscure the fact that regular soldiers are often embedded in such economies as well. Reno 1998; Howe 2001, 12. 295

Awakening,” the co-optation of Sunni fighters who wished to break with al-Qaeda in Iraq, is regarded as a core reason for the decline in violence in Iraq after 2007, 63 but others raise

concerns that the co-opted militias will just defect from the government now that U.S.

occupation is over. 64 Similar skepticism has emerged about NATO’s dealings with warlords to

reconstruct an Afghan state, a central part of its strategy there. 65 Finally, the multiplicity of

militias in post-Gaddafi Libya leads some analysts to wonder whether NATO’s intervention

merely had the result of fostering long-run chaos there. 66 When state leaders arm civilian

militias, are they just “using fuel to put out a fire”? 67

Despite the view that militias are automatically more autonomous and hence more defection-prone than regular armies, it is worth pointing out that they may vary in their autonomy. This is, of course, the way in which civilian militias are supposed to differ from regular government forces: the latter are, in theory, much more tightly linked to a hierarchy. But the difference does not lie just between civilian militia and regular personnel. Instead, militia themselves vary in the degree to which they are controlled by states. The stereotype of a militia in terms of control, in fact, used to be quite different. Writing over forty years ago, Maurice

Duverger defined militia as “a kind of private army whose members are enrolled on military lines, are subject to the same discipline and the same training as soldiers, like them, wearing uniforms and badges... They must always be ready to hold themselves at the disposal of their leaders.” 68 This is a far cry from the stereotypical ragtag militia that so animates scholars such as

Mueller or Kaldor. Nor were ill-disciplined militias as new a development as the latter suggest.

63 Smith and MacFarland 2008; Ricks 2009. 64 Simon 2008; Long 2008. 65 Sinno 2008, 273; Peceny and Bosin 2011. 66 International Crisis Group 2011a; Amnesty International 2012. 67 Francis 2005, 19. 68 Duverger 1967, 36–37. Quoted in Francis 2005, 1. 296

Variation in fact had existed when Duverger was writing. As Chapter 4 made clear, many militia

units in the Spanish Civil War, for example, were often subject to very little discipline or

oversight by state agents, while others were much more tightly controlled. 69

States might therefore sometimes be able to assert strong, centralized authority over militias, punishing them if they get out of hand. But their ability to do so may turn on whether the state has other strong and reliable regular forces. Seth Jones notes that under the Musahiban dynasty, Afghan militias were considerably more reliable servants of the state than they were under the Soviet-supported Najibullah regime. The reason, according to Jones, lies in variations in autonomy. Alongside its militias, the former regime also took careful steps to create a centralized army, which was able to deal with militia revolts as they occurred and to regulate inter-tribal disputes. In contrast, by 1991, militia numbered 170,000 as against 160,000 regular government personnel, the latter rapidly shrinking in number. In certain provinces, the difference was vast: for example, 28,000 militia versus 5,000 regular troops in Herat. 70 Similarly, the growth in centralized power in the Spanish Republic’s army helped to ensure that the fragmentation of May 1937 was not more extensive: Anarchist militia leaders, knowing that they could not afford a confrontation with the state, chose not to rebel and the revolt was limited to small numbers of militants. A similar process occurred among the Republic’s Nationalist adversaries and their own civilian militias. It is plausible to suppose that militias are better able to defect when the state is less powerful. Thus the formal separation between regular military and civilian militia may be a poor indicator of relative autonomy. Rather, such autonomy may be

69 Ahram (2011, 2) likewise argues that militias’ limits to the monopoly of state violence have plenty of historical antecedents, and for Mueller (2004, chap. 2), “criminal” (i.e. undisciplined) war is historically more common; its current dominance is owed to the decline in war among disciplined armies. See Kalyvas (2001) for a general critique of the “new wars” approach. 70 Jones 2012, 21–26. 297 endogenous to the relative coercive power of the regular army. Because of this variation in autonomy, I do not expect civilian militias to be any more defection-prone than regular armies.

8.5. Motivations and Factionalism: Conscription and Coup Attempts

This dissertation’s theoretical approach places rewards and punishments and norms of cooperation in the context of the value that combatants place on the armed group’s success, holding that the former should be generally effective in a much wider array of situations than the latter. Top-down rewards and punishments should often be effective, that is, even when combatants in general do not share the group’s aims. However, when disagreements about aims are particularly severe, for example because of factionalism, top-down rewards and punishments may have attenuated effects. Soldiers are likely to view increases to central state capabilities with suspicion, fearing that only cronies will be rewarded and only rivals will be punished. How can we measure commitments to common aims among a government’s soldiers? I suggest two measures, with different implications for the distribution of motivations in a government armed force: conscription and recent coup history. The former, I argue, is an indicator of moderately lower commitment to the government’s success; the latter, an indicator of much more severe disagreement. Hence, I expect the role of external support to hold up in conscript armies, and to weaken substantially in armies with a recent coup attempt.

The first question is whether the military employs conscription or all-volunteer service. I consider this to be an indicator of moderately lower commitment among government troops.

Generally, as outlined in Chapter 3, I expect conscripted armies to be less committed to the armed group’s aims than volunteer forces, other things being equal. However, there are also reasons to believe that the difference between conscripted and volunteer forces might be less stark in general than it was in the Spanish Civil War. First, the Republic and the Nationalists

298

called up many more draft classes than had been standard practice in peacetime. Most

importantly, however, their volunteer forces either joined up after the war had begun, or, in the case of the Legion, joined up for military service in Africa in a force closely identified with the political right, and one that had already been used to put down a left-wing rebellion in Asturias in

October 1934. Although, as I indicated in Chapter 3, many volunteers were relatively uncommitted to the cause as a consequence of opportunism, they had at least volunteered knowing that they would have to fight a civil war, and had volunteered for a specific side rather than for service in general. In the general case of an all-volunteer army, volunteers who joined up before a civil war began may have their loyalties profoundly tested by the emergence of civil war itself. There may not be very strong differences between volunteer and conscript armies in general. However, there is still an aspect of coercion involved in recruiting conscripts that does not exist with volunteers. Civil wars generally do not emerge from nowhere, but build from earlier political tensions. Hence the decisions of individuals to join an all-volunteer force may be contingent, in part, on an awareness that they might have to fight for the government in a civil war. Thus there is a case to be made that conscription indicates a moderately lower degree of commitment to the success of the government than voluntarism does, but I acknowledge a lack of clarity in this indicator.

Because my theoretical approach considers the influence of combatants’ motivations as mediated by top-down rewards and punishments, I do not have strong expectations for the impact of conscription by itself. In contrast, in the analysis of unit composition in Chapter 7, because the analysis was conducted within a single army, the army’s capability to dispense rewards and punishments was effectively held constant, and so the impact of conscription within military units in undermining norms of cooperation could be more clearly identified. An

299 approach based on combatant motivations alone, regardless of capabilities and rewards and punishments, might instead expect a clear association between conscription and defection.

Below, instead, I suggest hypotheses about the interaction between conscription and external military support that more clearly capture my dissertation’s focus on the commitments of soldiers as mediated by rewards and punishments.

Hypothesis 2. Governments with conscripted armies should have higher defection rates than governments with all-volunteer forces.

Conscription by itself, though suggesting a lack of common aims, does not necessarily entail the widespread, severe disagreement and mistrust that can really undermine top-down rewards and punishments. Factionalism, however, can certainly have this undermining effect.

Factionalism in the Spanish Republic, for example, set serious limitations on the effectiveness of

Soviet aid. While Soviet assistance helped to increase the Republic’s capabilities and control, the fact that the Soviets strongly preferred Communist Party dominance hardly helped in convincing

Republican soldiers that Soviet intentions were benign. Indeed Soviet involvement likely made suspicion of the Communists by other factions worse , confirming their fears that the Communist

Party had ulterior motives: the elimination of perceived enemies of Communism generally, and not just the defeat of Franco. Factionalism thus helps us to specify the conditions in which external assistance can contribute to rewards and punishments and when it is undermined—or possibly even turned into a negative influence—by factionalism.

It is not easy to gauge the various factional conflicts that may exist within a government armed force, but one important indicator may be a history of coup attempts. 71 Recent coup

attempts, whether successful or not, can indicate quite intense factionalism and disagreement

71 This paragraph and the next draw heavily on McLauchlin 2010, 335–336. 300

within an armed force. There are two logics underlying this conclusion. First, coup attempts

themselves can emerge from divided armed forces. A section of the army may support the

incumbent government even while the coup attempt is taking place. Some officers may well hold

that the regime being challenged is in fact legitimate, even when their fellows are rising against

it. 72 Coup-proofing strategies, including steps that are popular in the Middle East such as placing

loyalists in key positions and maintaining multiple rival internal security services, each

monitoring the other, 73 mean that a regime that fears a coup will generally build in conditions that make such coups, when they come, extremely contentious. Thus it is rare that a coup actually has the full underlying support of the officer corps. There may certainly be bloodless palace coups, but coups involving shootouts among rival groups of officers are certainly common as well; Spain itself is an excellent example. An indicator of the divisions that coup attempts often indicate is the fact that purges often follow coups, such as the Egyptian Free

Officers’ Coup of 1952. 74 Levitsky and Way therefore treat coup attempts as indicating an underlying lack of cohesion in the armed forces. 75

Second, coups themselves can engender further mistrust. Coups tend to arise in conditions of secrecy. As Rapoport puts it, “a great majority of the officers will be apprehensive because the coup necessarily originates in a secret plot by a tiny minority, and those not invited to participate must be asking themselves why.” 76 Purges may certainly eliminate rivals from the armed services, but a regime will typically fear that it did not go far enough. The legacy of suspicion from each coup is one reason why coups engender further coups, beyond structural

72 Huntington 1957. 73 Brooks 1998; Quinlivan 1999. 74 Hurewitz 1969, 152. 75 Levitsky and Way 2010. 76 Rapoport 1982, 257. 301 factors such as economic or civil society conditions. 77 Indeed, the pattern of successive coups in

countries such as Syria, Nigeria, or Guinea-Bissau led commentators to give such regimes a

category and a label: praetorian regimes. 78 In extreme cases of praetorianism, according to

Rapoport, “virtually all governments experience military conspiracies.” 79 Each coup engenders

mistrust of further political action by army personnel.

Thus two logics—a selection effect and a causal effect—suggest that a recent coup

attempt is likely to indicate a high degree of mistrust within the armed services. The case of

Spain is again illustrative. It had had a long history of coup attempts or pronunciamientos during

the late 19 th and early 20 th century. 80 The Republic, in fact, emerged in 1931 after the collapse of the Primo de Rivera dictatorship, which had itself been installed in a coup. The unsuccessful

Sanjurjada coup attempt of August 1932 illustrates both logics. It arose from right-wing officers’

fears about the reformist policies of the new Republican government. 81 And after its failure,

General Sanjurjo and other right-wing officers were sent into exile. But the coup attempt itself,

and subsequent left-wing attempts to limit the influence of right-wing officers, also served to

crystallize Republican fears of right-wing sentiment within the armed forces. Rival organizations

within the military, supporting the Republic and plotting against it, emerged over the next few

years. 82 Thus the Sanjurjada both indicated factionalism and reinforced it.

We should expect that governments that have experienced recent coup attempts should

suffer from quite serious factionalism. Since factionalism sets limits to the operation both of top-

down rewards and punishments and of norms of cooperation, I hypothesize that a recent coup

77 Belkin and Schofer 2003. 78 Huntington 1968. 79 Rapoport 1982, 256. 80 Payne 1967; Boyd 1979. 81 Preston 1994a, 51–62. 82 Payne 1967; Cardona 1983; Alpert 1989. 302 should make defection more likely. Specifically, I expect that a recent coup should resemble a sufficient condition for a high defection rate, with few countries with a recent coup attempt and a low defection rate.

Hypothesis 3a. Governments with a recent coup attempt should have higher defection rates than governments with no recent history of coup attempts.

Hypothesis 3b. A recent coup attempt is a sufficient condition for a high defection rate.

In my theoretical approach, whether combatants share common aims does have an important impact, but mainly indirectly, by conditioning the effectiveness of top-down rewards and punishments, notably control, and norms of cooperation. I argue that rewards and punishments can help even when many combatants are unmotivated—provided those combatants are not severely divided or factionalized. Because conscription indicates a reduced degree of commitment but not necessarily the severe disagreements implied with factionalism, rewards and punishments should still reduce defection rates even when conscription is employed. I therefore expect that military capabilities should be associated with a lower defection rate even if the country employed conscription. In a stronger formulation, I expect that the effect of military capabilities should actually be stronger where conscription is employed, because there may be a somewhat lower likelihood that troops in conscript armies can develop norms of cooperation.

Hypothesis 4a. External military support should be associated with lower defection rates even if conscription is employed.

Hypothesis 4b. The relationship between external military support and lower defection rates should be particularly strong when conscription is employed.

However, external military support should be less effective where there is severe factionalism, as indicated by coup attempts. It should be most effective where there has not also been a recent coup attempt.

303

Hypothesis 5 . The relationship between external military support and lower defection rates should be stronger where there has not been a recent coup attempt than where there has.

8.6. Empirical Analysis

This section outlines very preliminary findings from a broad overview of defection patterns among government and government-supported forces in civil wars beginning in the

1990s. This analysis of 28 civil wars identified by Sambanis 83 as beginning in the period 1990-

1994 is the first step in a larger research project that intends to cover civil wars since 1945. The

early 1990s provides an important frame for studying the reliability of government armies,

especially in the context of some of the variables identified above: conflicts in the 1990s have

variously been said to have emerged from the withdrawal of superpower support for states 84 and

to disproportionately involve unreliable civilian militias acting as gangs and thugs. 85 The

hypotheses implying that the character of civil wars changed substantially with the end of the

Cold War suggest an over-time comparison. However, for present purposes, considering the

dearth of data about defection from government armies, it can be useful to assess whether such

arguments apply to the post-Cold War period to begin with. This is an “easy” period for both the

arguments emphasizing external support and militias, since both of these lines of argument were

developed to discuss post-Cold War changes in patterns of conflict. I do not include three

contentious cases in Sambanis’ dataset (Kenya, 1991-1993; Haiti, 1991-1995; and Chad, 1994-

1997). They are not coded as armed conflicts by the Uppsala Armed Conflict Dataset, 86 the basis

for Cunningham, Gleditsch and Salehyan’s dataset on government-rebel dyads, which in turn is

83 Sambanis 2004. 84 Posen 1993b; Kalyvas and Balcells 2010. 85 Kaldor 2006; Mueller 2004. 86 Gleditsch et al. 2002; Themnér and Wallensteen 2011. 304

my source of data about external support for government armies. The Uppsala dataset does not

list Kenya in the 1990s at all and restricts Haiti to the coup of July 1991, while the time period

given by Sambanis for conflict in Chad is instead a period without armed conflict according to

Uppsala.

On the basis of scholarly accounts, IGO, NGO and news media reports, I coded whether

government forces experienced an episode of mass defection in each civil war, including both

regular and militia forces. Defection, here, includes both side-switching to an existing rebel force

and fragmentation, breaking off to start a new rebel force. For inclusion as mass defection, there

had to be clear evidence that more than 10% of the force, whether militia or regular, defected in

at least one calendar year of war. If there was no clear evidence that this occurred, then the case

was not coded as experiencing mass defection. The coding is thus for the whole war, but it refers

to whether there was a single year in which more than 10% of the armed force defected. I

approached coding in this way because of the difficulty of obtaining precise information about

defection for many cases. In such circumstances, it made sense to have a blunt indicator to

reduce sensitivity to error. In most cases it is considerably easier to assess whether an armed

group cleared a 10% hurdle than just what its defection rate was. A more intensive over-time

analysis within a civil war, for example a different coding for each year, would be preferable, but

lack of information again makes it difficult to produce reliable codings for each year of war. This

is a problem that can be overcome with greater resources, but for a pilot project a single coding

for a civil war will suffice.

Data on external military support were drawn from Cunningham, Gleditsch and

Salehyan’s data on civil war dyads. 87 The variable includes both military assistance and troop

87 Cunningham, Gleditsch, and Salehyan 2009. 305 deployment in support of a government. It is coded 1 if either of those two forms of aid were rendered, and 0 if either no aid was received or the only support was non-military. Data on conscription policies came from Nathan Toronto’s coding of military recruitment policies worldwide, and is just a binary indicator of whether military recruitment involved conscription or was all-volunteer. 88 Data on coup attempts were drawn from the Center for Systemic Peace

Coups d’Etat, 1946-2010 dataset. 89 I measured whether the country experiencing civil war had experienced a coup attempt in the five years prior to the start of the war. I excluded coups that occurred after the beginning of the war. I did not distinguish between successful and failed coup attempts, since mistrust should be a difficult problem either way.

8.7. Results

The approach outlined above suggests that defection should be associated with a lack of external military support and with a coup attempt in the previous five years. In addition, a motivations-based approach would hypothesize that conscription should be associated with elevated defection rates, while I propose that any such effect should be mediated by capabilities.

Mass defection occurred in half of the twenty-eight cases of civil war. Table 8.1 indicates the bivariate relationships with defection. In the first place, government forces with external military support do appear to have a somewhat lower incidence of defection than those without, with 90% confidence. Almost 70% of governments without external support experienced mass defection, while only one-third of those with support did.

There is little support for Hypothesis 2, relating conscription to defection: conscript armies experienced mass defection only slightly more often than all-volunteer forces, and the

88 Toronto 2005. 89 Marshall and Marshall 2011. 306

Table 8.1. External support, coup history, and defection

(a) External support and defection

Mass defection External military support No Yes No 4 (30.8%) 9 (69.2%)

Yes 10 (66.7%) 5 (33.3%)

χ2 = 3.59 ( p < .1)

(b) Conscription and defection

Mass defection Recruitment policy No Yes Conscription 7 (46.7%) 8 (53.3%)

All-volunteer force 7 (53.9%) 6 (46.2%)

χ2 = 0.705. Column percentages do not sum to 100% due to rounding.

Mass defection Coup in last five years No Yes Yes 2 (28.6%) 5 (71.4%)

No 12 (57.1%) 9 (42.9%)

χ2 = 1.90

difference was not statistically significant. There is somewhat more support for Hypothesis 3a, but the difference between armies that had experienced coup attempts and armies that had not is not statistically significant. As for Hypothesis 3b, that a recent coup attempt should be a sufficient condition for mass defection, again I find lukewarm support: two out of seven cases with a recent coup still retained cohesion during the civil war.

Table 8.2 examines interaction effects. In Panel (a), I test Hypotheses 4a and 4b, examining whether external military support is associated with lower defection rates even when

307

Table 8.2. Interaction effects

Cell values are frequencies with which armed groups experience mass defection.

(a) External support and recruitment

Recruitment policy External military support All-volunteer force Conscription No 3/5 (60%) 6/8 (75%)

Yes 3/8 (37.5%) 2/7 (28.6%)

(b) External support and coup history: five-year threshold

Coup in last five years External military support No Yes No 7/10 (70%) 2/3 (66.7%)

Yes 2/11 (18.2%) 3/4 (75%)

Coup in last twenty years External military support No Yes No 4/5 (80%) 5/8 (62.5%)

Yes 2/9 (22.2%) 3/6 (50%)

conscription is applied. As hypothesized, the association between support and defection does not weaken in conscript armies. If anything, the association appears to be slightly stronger , supporting Hypothesis 4b. Indeed, the difference in defection rates with external support is statistically significant in conscript armies ( p <.05) but not in all-volunteer armies. Interestingly, this suggests perhaps that the presence of unmotivated combatants makes strong material capabilities more necessary, by reducing the potential for norms of cooperation.

308

The interaction between external support and recent coup history, in panel (b), is particularly strong, bearing out Hypothesis 5. In the seven countries in that experienced a recent coup attempt, external military support appeared to have no association with defection rates.

However, in countries without a recent coup attempt, defection rates were substantially lower where there was external support, a statistically significant difference (p <.01). In addition, defection rates were uniformly high when a country had experienced either a coup or did not enjoy external support, or both. The discrepant situation is when a country both had external support and had not experienced a recent coup. Only two of those eleven countries—Georgia

(Abkhazia) and Tajikistan—experienced mass defection, and while both enjoyed some external support, they were also dealing with the fallout of a sharp decline in support from Moscow after the collapse of the Soviet Union. Because there were only seven cases of countries with a coup attempt in the previous five years—perhaps too low a number of observations to sustain my conclusions—I also repeated the analysis using twenty years as a threshold (panel (c)). The results show the same basic pattern, but somewhat weaker, as might be expected if coups’ effects diminish over time. External military support was seemingly more effective when a coup attempt had taken place within twenty years than just in the past five. Again, however, the association between external support and lower defection rates was not statistically significant with a past coup attempt. In contrast, where there had not been a coup attempt in the past twenty years, there was a strong association between external support and lower defection rates ( p < .05).

As noted above, it is possible that the relationship between external support and low defection rates could be explained with a reverse logic. Notably, foreign sponsors may withhold support for a country’s army if they think it likely that that army will experience defection, because of the wastefulness of doing so. Above, I offered the argument that foreign sponsors

309 seem to intervene for different reasons, and to maintain their intervention despite the unreliability of the army in question. The empirical data appear to bear this out. If foreign sponsors were, in fact, sensitive to the problems of waste in the military, we might expect external support to correlate negatively with a history of coups, because of the fractiousness of such armies and because, as established above, external military support has a very weak relationship with defection in the context of a military coup. However, this is not what we find. The rate of external support was 52% for countries that had not experienced a coup in the past five years, and 57% for countries that had. It therefore does not appear that foreign sponsors were particularly sensitive to the risk of supporting a divided army. It might still be that foreign sponsors are, in fact, sensitive to lower prospects for aid effectiveness when their clients are defection-prone. The lack of sensitivity to a history of coups tells against this interpretation.

However, without over-time data or an instrument for external support, it cannot be ruled out.

Is there a difference between militia and regular forces? Table 8.3 shows the frequency with which regular and militia forces experienced mass defection. Overall, the findings indicate that militias were no more unreliable than regular forces. If anything, they were somewhat more reliable. Thus far, it appears as though disparaging judgments of civilian militias may apply equally to regular forces. One problem with this analysis was the lack of information in some cases. If there was no evidence of defection, the case was coded as 0. But there is much more information about regular forces than about militias. This is likely to induce a pro-militia bias to the data. Excluding the cases where there was just no evidence, however, does not change the finding substantially; the combined incidence of defection is 42% for the militia cases for which there is clear evidence one way or another, and 45% for regular forces.

310

Table 8.3. Defection in Regular and Militia Forces

Mass defection? Regular Militia No 16 (59.3%) 17 (65.4%)

Yes 11 (40.7%) 9 (34.6%)

Total 27 26

Table 8.4. Connections between defection in regular and militia forces

Defection from militias Defection from regular No Yes forces No 12 (85.7%) 2 (14.3%)

Yes 5 (45.5%) 6 (54.6%)

χ2 = 4.59 (p<.05). Column percentages do not sum to 100% due to rounding.

Militias thus do not appear to be more defection-prone than regular forces. I argue that one reason may be that regular army defection and militia defection are connected. Table 8.4 indicates that there is a statistically significant, positive relationship between regular army defection and militia defection. While they experienced mass defection less than 14% of the time when regular forces had low defection rates, this rate jumped to over 50% when regular forces also experienced mass defection. The image of unreliability that militias sometimes have masks the defection that regular forces experience: when militia forces were defecting or fragmenting, it was much more often than not the case that regular forces were doing so too. From this point of view, we might instead reconceptualize central material power and control as regards militias.

We might think of the reliability of the regular army as a key intervening variable driving militia behaviour, with the former in turn driven in part by external military support.

311

Beyond the numbers, there is ample case-study evidence indicating that the withdrawal of external support at the end of the Cold War hampered regimes’ ability to keep their agents in line, and led to their reliance on militias that they could not keep in check. This was particularly true in post-Communist regimes, where the removal of Soviet military support led to highly dubious armies.90 In Afghanistan, the last remnants of the Afghan National Army defected from

the new government of Burhanuddin Rabbani to warlords like Abdul Rashid Dostum, who could

pay rather more reliably. Lacking much of a regular force to speak of, Rabbani carried on

Najibullah’s reliance on alliances with militias and their leaders. Many of these militias broke

their alliance with the government and allied instead with Gulbuddin Hekmatyar, and eventually

with the rising Taliban starting in 1994. 91 Without a Soviet military presence, the only regular

force Georgia had in facing rebellion in South Ossetia in 1991 was its quickly improvised

National Guard, supplemented by several other militia units. The weakness of Tbilisi’s

bargaining position essentially forced it to rely on forces with their own access to funding and

thus a much stronger ability to break off and defect as they needed to. When a faction of the

National Guard led by Tengiz Kitovani split off to rebel against the government of Zviad

Gamsakhurdia in September 1991 through January 1992, it was joined by members of various

different militia forces from the South Ossetia front.92

The new state of Azerbaijan had similar difficulties to its Georgian neighbour as it faced

the attempted secession of Nagorno-Karabakh to Armenia. Upon independence, Baku had to deal

with the desertion of thousands of troops from the Soviet 4 th Army, which was supposed to be

90 Fairbanks 1995. 91 Giustozzi 2009; Sinno 2008. 92 Human Rights Watch 1992; O’Ballance 1997, 106–109; Ozhiganov 1997, 356; Zürcher, Baev, and Koehler 2005, 266–267; Zürcher 2007, 139; Cheterian 2008, 178. 312

under its jurisdiction. 93 It effectively had to start a new regular army from scratch. As a

supplement to a very small (but growing) regular army, it relied on militia leaders such as Surat

Husseinov and Iskender Hamidov to fight in Nagorno-Karabakh. These forces had begun

organizing as the Soviet Union was falling apart, anticipating a power vacuum, and they entered

an alliance with the state. The militia constituted almost half of Azerbaijan’s forces in 1991. 94

The militia units had considerable autonomy in these circumstances. Their actions were often very badly coordinated, with various factions fighting the enemy on their own, and sometimes each other. 95 This autonomy facilitated two clear instances of fragmentation, with militias and

units of the army entering into open rebellion against the government. In May 1992, many

paramilitary forces, including the Gray Wolves, stormed the parliament building in Baku to

overthrow the government. One year later, in June 1993, Husseinov’s own forces led a “march

on Baku,” a putsch against the government, that provoked many other militia and regular units

fighting in Nagorno-Karabakh to rebel against the government as well. 96

8.8. Potential Control Variables: Economics and Institutions

The relationships identified so far appear to be relatively clear, but they could potentially be spurious on other indicators. Given the relatively small N, it is not really possible to implement statistical control directly. However, we can have a look at the relationship between control variables and defection rates to see if these indicators might mitigate the relationships outlined above.

93 Mary Dejevsky, “Yeltsin Plays for Time Amid Collapse of Military.” The Times (London), 29 January 1992; Goltz 1993. 94 Cornell 1999, 25; Zürcher 2007, 171, 179. 95 Cornell 1999, 33; De Waal 2003, 165; Cheterian 2008, 133. 96 Cornell 1999, 33–35; De Waal 2003, 182, 214–215; Zürcher 2007, 171. 313

A country’s economic conditions may influence its defection rate in important ways.

States’ fiscal base can be important to state capabilities: 97 states in fiscal crisis sometimes cannot pay their soldiers. Hence, for example, one reason for the frequently-noticed correlation relationship between economic crisis and authoritarian breakdown. 98 As I argue above, a state’s

fiscal standing is one step removed from military capabilities, but is worth considering as an

alternative source of capabilities from external support. In addition, however, economic

conditions can affect the ability of rebel groups to recruit combatants; according to Collier and

Hoeffler, the lower the GDP per capita, the fewer economic opportunities are available in the

regular economy, and the more individuals may be tempted to join rebel groups for illicit

economic activities. 99 The same logic may apply to government soldiers and militiamen: they may switch sides if a rebel group can offer them a better deal.

Economic data, drawn from the World Bank’s World Development Indicators, 100 were available for 24 of my 28 cases. The missing data are important: in addition to Iraq, they include cases like Afghanistan and Somalia, in extremes of state failure, and the new state of Bosnia-

Herzegovina. Apart from Iraq there may therefore be a correlation between missing data and low state capacity. I found no relationship between GDP per capita at purchasing power parity and defection, either in the first year of war or in the year before. Interestingly, however, I did find a clearer relationship between GDP growth and defection: the average growth rate for countries without mass defection was -5.1%, while for countries with defection it was -14%. This was a statistically significant difference ( p < .1). The difference between overall economic conditions, as indexed by per capita GDP, and changes to those conditions, as indicated by growth, is

97 Levi 1988; North 1982; Olson 1982; Olson 1993. 98 O’Donnell and Schmitter 1986; Bratton and van de Walle 1997; Lee 2005. 99 Collier and Hoeffler 2004. 100 World Bank 2011. 314

interesting: it suggests that as economic conditions worsen, soldiers’ calculations may change to

favour defection. GDP growth therefore appears to be an important control variable for future

analysis with a larger number of cases. It suggests a further reason for caution in interpreting this

chapter’s primary findings.

Additionally, the quality of the state’s political institutions may matter to military

defection. In weakly institutionalized regimes, we might expect the rules of the political game to

be unsettled and unclear, and hence to create greater incentives and possibilities for armed forces

to get involved in politics.101 The current standard approach to weakly institutionalized regimes is to consider such regimes in terms of their institutional inconsistency: combining elements of authoritarianism and democracy, hybrid regimes offer opportunities for political mobilization without the legitimate, regularized rules of political competition to channel it. 102 (However, more

recent research suggests that any relationship between hybrid regimes and defection may be

spurious: faced with an unreliable military, autocratic leaders might create competitive

legislatures in order to compensate.) 103 I tested a relationship between anocracy and defection using both the Scalar Index of Polities measure of anocracy 104 and the absolute value of the state’s POLITY score (since scores around 0 indicate more inconsistent regimes, -10 is full dictatorship and +10 is full democracy). There was no apparent statistical connection between defection and anocracy in the year prior to war or in the first year of war, for either measure. I reserve judgment about whether regime type mediates the effects of the other variables tested here until there is a large enough N to support multivariate analysis, but in bivariate terms there is no apparent relationship between regime type and defection rates.

101 Huntington 1957; Huntington 1968. 102 Hegre 2001; Gates et al. 2006; Vreeland 2008; Levitsky and Way 2010. 103 Gandhi and Przeworski 2007; Gandhi 2008; Geddes n.d. 104 Gates et al. 2006. 315

8.9. Conclusion: States and Civil Wars

What this chapter tells us is that external military support is associated with lower defection rates in government armies, but only for non-factionalized armies. The interpretation I provide is that this is because of the importance of capabilities in providing top-down rewards and punishments to induce soldiers not to defect. I must urge caution in this interpretation, because there are alternatives that cannot be eliminated on the data here. If correct, however, it indicates that external support has possibilities but also serious limitations in facilitating the cohesion of government armed forces.

External support was associated with a lower rate of defection even—in fact, especially—in conscripted armies. Since conscription may indicate a greater lack of commitment to the government side, this provides suggestive evidence in favour of the broad effectiveness of top-down rewards and punishments. However, despite the general importance of top-down rewards and punishments, factionalism seems to be a limiting factor to the effectiveness of external support: the clearest decline in the incidence of defection occurs when a country had both external military support and no coup in the previous five years, while external support had

a substantially lower impact in countries with a recent coup, whether within the previous five or

the previous twenty years. In turn, when regular security forces are reliable, it appears that militia

forces are generally more reliable also. The resulting image is of material capabilities transferred

to a central government from outside, and radiating outward to its security services. Rather than

the formal institutional distinction between uniformed personnel and militias, the material

aspects of power appear to be much more relevant in conditioning defection patterns.

This chapter has begun to place defection in an international context, but there is

considerably more work to do to develop these connections. As noted, it is possible that external

316

patrons are more reluctant to support governments with defection-prone forces. I have given

some arguments and evidence to suggest that this may not be the case, but I do not claim to have

completely settled such fears. With over-time data, one could examine whether an incident of

mass defection prompts a withdrawal of external support, for example. An instrumental variable

for external support would also be of considerable interest. This chapter also does not consider

the basic nature of the external military assistance being granted, for example not distinguishing

between providing military vehicles, small arms, and training or advice. Finally, this chapter

elides distinctions among different external sponsors. If a government has multiple patrons, it is

worth considering whether those patrons’ competing agendas interfere with the unity of a

government force as they bring multiple different agendas to bear. This is therefore far from a

complete analysis of the international dimensions of the cohesion of armed groups.

In addition to these caveats, it is vital to emphasize that my data only speak to the period

of civil war itself, and hence to the short run. There may be serious differences between the

ability to keep an army together to stay in power in the short run, on the one hand, and long-run

state-building, on the other.105 It may be that even if external military support helps to prevent defection, this does not actually help to build strong states in the long run. Even when, as in Iraq in 2003, a foreign power occupies a country and monopolizes the use of force, it is often unable to create highly centralized or effective security force, or unwilling to put the effort and expense that would be required to do so. This is because, as Alexander Cooley argues, military occupations are often run on the cheap and in a highly decentralized fashion. Such arrangements

105 Migdal 1988; Bueno de Mesquita et al. 2003; Bates 2008. 317 often fail to seriously reform institutional structures in the country in question. 106 As such, the

prospects for external states to build up stable structures that can persist into the future are quite

limited. To do so requires an enormous commitment that will often prove too costly for third

parties.

Moreover, to the degree that the external state provides sources of coercion to prevent

defection, it can serve as a substitute and a source of moral hazard, allowing temporizing local

state leaders to fail to invest in long-run structures. External support may well fail to produce

stable domestic sources of reward and punishment. Indeed, according to Robert Jackson and

Jeffrey Herbst, the international community’s willingness to prop up weak states has allowed the

leaders of those states to avoid long-term efforts at state building. 107 Charles Tilly argues that the availability of external military aid has led to states in which the military apparatus dominates at the expense of popular legitimacy through institutions—a source of long-run weakness. 108

It is useful to highlight how different this is from past state-building: indeed, Tilly argues that the key difference between contemporary developing-world state-building and its antecedents is that later state-builders are able to acquire military power from outside. 109 External

support is not one of the canonical parts of the story of European state-building, in which states

emerged by a steady and painful accretion of power at the centre, a process of defeating internal

rivals and centralizing the security apparatus. 110 However, as many other scholars have noted, the

contemporary international environment in which state-building is embedded differs

considerably from that of European state-building in past centuries. The imperative is not, now,

106 Cooley 2005. Despite many differences among them, scholars of European colonialism in Africa generally agree that indirect rule via intermediaries was an important political form of empire. Robinson and Gallagher 1961; Young 1994, 107; Mamdani 1996; Herbst 2000. 107 Jackson and Rosberg 1982; Jackson 1990; Herbst 2000. 108 Tilly 1985, 186. 109 Ibid., 185–186. 110 Finer 1975; Tilly 1985; Tilly 1992; Dandeker 1990. 318

to build a strong state or disappear at the hands of a stronger competitor; states rarely die.

International warfare has become considerably less frequent over time;111 boundary change has

drastically declined as the terms of settling international conflict; 112 and states rarely disappear to conquest. 113 The U.S.-led international order has generally enforced a norm against explicit

conquest, upholding juridical sovereignty whatever pragmatic manipulations have gone on.114

External powers continue to support internal claimants to power, even as conquest itself has declined. 115 External state actors have thus emerged as crucial actors in internal conflicts. 116 On

balance, then, the relative importance of foreign powers as supporters of one side or another now

outpaces their role as conquerors of vulnerable weak states.

As state policymakers and the international community contemplate security sector

reform, approaching the coercive institutions of the state with a view to ensuring that they

provide order, stability and predictability, the findings of this chapter are worth considering. It

may be that external military assistance can have an important effect in preventing defection.

The withdrawal of that assistance can be a catastrophe for a government. On the other hand,

whatever effectiveness emerges from external support is seriously undermined in the context of

factionalism. The importance of these issues is made manifest by current problems facing the

United States and its NATO allies in Afghanistan. At present, U.S. and Afghan officials estimate

that the funds available to pay and equip the Afghan military will decline from $7 billion

annually to $4.1 billion,117 and that number is still contingent on non-U.S. donors tripling their

111 Holsti 1996; Themnér and Wallensteen 2011. 112 Rosecrance 1986; Zacher 2001. 113 Fazal 2007, 23–28. 114 Jackson and Rosberg 1982; Jackson 1990; Krasner 1999; Zacher 2001; Fazal 2007. 115 Salehyan 2010; Salehyan, Gleditsch, and Cunningham 2011; Saideman 2001; Regan 2000. 116 Balch-Lindsay and Enterline 2000; Balch-Lindsay, Enterline, and Joyce 2008. 117 Thom Shanker and Alissa J. Rubin, “Afghan Force Will Be Cut After Taking Leading Role,” New York Times , 10 April 2012. 319 current contributions. 118 Of that figure, the Afghan government would contribute only $500

million. That would still constitute about half of its current tax revenue (as of 2009). 119 This is a fairly clear indicator of Afghanistan’s continued dependence on external donors. It is believed that the cuts to the military budget will require a reduction from a peak of 352,000 troops in the

Afghan National Security Forces (ANSF) to 230,000 troops by 2017. However, the Afghan regime and its armed forces are highly factionalized, with multiple competing local leaders and a considerable body of opinion that is opposed to ongoing NATO influence, reducing the effectiveness of NATO assistance in furnishing cohesion in the ANSF.

The options are therefore poor, a dilemma captured in this chapter’s findings. The Karzai regime is hardly an attractive client; its fractiousness helps make outside support less effective.

But with a reduction in external support, the prospects of defection to the Taliban are only likely to increase as the ability of the government to provide rewards and punishments declines dramatically. In turn, as the central force weakens, President Hamid Karzai’s regime, already highly reliant upon warlords with their own sources of revenue and a history of controlling military forces, is likely to become more so. Those militia forces are unlikely to remain loyal over time. 120 Given current financial constraints, it is unlikely that the United States or its NATO partners can avoid these problems over the long run. But if external support is important for keeping defection rates low, then such problems of dependence are likely to be a chronic concern in third-party counterinsurgency campaigns. 121

118 Karen DeYoung, “U.S. Seeks More Money for Afghan Force.” Washington Post , 27 March 2012. 119 World Bank 2011. 120 Mukhopadhyay 2009; Peceny and Bosin 2011. 121 Byman 2006. 320

Chapter 9 Conclusion: Desertion and the Dynamics of Civil Wars

9.1. Introduction

Alejandro Goicoechea was a Captain of Engineers in the Basque army corps that fought for the Republican side in 1936 and 1937. He helped to design the “Iron Ring” of Bilbao, a system of defensive fortifications on which the Republic placed considerable hope as the

Nationalist troops closed in on the Basque Country. In February 1937, he defected to the

Nationalists and brought the plans with him. The plans indicated areas of weakness in the Iron

Ring, a piece of information that proved of considerable use in the Nationalist capture of the

Basque capital in June of 1937. 1

To slow the Nationalist offensive in the North, the Republic attempted a diversionary assault on the town of Brunete west of Madrid in July. 2 However, it was not long before

Santander (in August) and then Asturias (in October) fell, with a cascade of desertion especially in the former. With the collapse of the Republican pocket in the North, Franco was able to concentrate his forces to the south and bring the full weight of his army to bear on the remainder of the Republic. Franco sought, not a quick victory, but a complete, comprehensive one.

Nationalist forces advanced over 1937 through 1939, grinding out a victory, two steps forward, one step back. In bitter winter weather, Republican forces took the town of Teruel in southern

Aragón in January 1938, but were turned back by Nationalist armies that steadily pushed towards

1 Thomas 1994, 613, 691-2; Preston 1993, 240. 2 Preston (2007, 266-300) gives a clear overview of the steps toward Republican defeat. 321 the Mediterranean Sea and, by April, cut off Catalonia from the central Republican zone with

Madrid and Valencia. A bloody, hard-fought Republican campaign along the Ebro River, with the aim of reconnecting its two key territories, started in July and made gains against the

Nationalist forces. The Ebro offensive was turned back by late November. Franco’s armies took

Catalonia in January and February 1939, and flood of Catalan refugees fled across the French border. With each Nationalist advance came more Republican desertion and defection. The

Communist-supported Negrín government intended to fight to the bitter end. Tapping into disgust with the Communists and seeking in vain to sue for a quick peace, Colonel Segismundo

Casado launched a coup in Madrid in early March. Insisting on unconditional surrender, Franco launched a final offensive on Madrid, and declared the end of the war on 1 April 1939.

Desertion and defection had an important influence on the course of the Spanish Civil

War, from the crippling side-switching of much of the officer corps onward. The weaknesses of the Republican militias, which had much to do with their tendency towards desbandadas or

disbandments, permitted Nationalist forces to advance farther faster than they might have

otherwise done. Desertion also undermined the Republic’s attempts to turn back the tide in

offensives, whether in Santander, 3 at Brunete, or along the Ebro. 4

The Republican defeat may well have occurred whatever its rate of desertion or defection. Before the siege of Bilbao, for example, German support for Franco had already begun to have a decisive influence, including the inauguration of the era of saturation bombing at

Guernica on 26 April 1937. Much desertion followed Nationalist military success on the battlefield rather than causing it. But the Nationalist advantages in control and political unity gave it an important edge in preventing desertion.

3 Martínez Bande 1980, 221; Solar 1996, 85–86. 4 Seidman 2002, 111-115, 190-193. 322

In the meantime, as we have seen, grappling with the problem of desertion and defection

generated important political dynamics on each side. The Republican side expended considerable

energy and generated considerable mistrust in centralizing its military force and subjecting it to

military discipline. The suspicions of military officers and their possible treachery led to a wave

of violent reprisals against them. The Republican side lost much military talent this way, while

the Nationalists had the immense advantage that they did not particularly suspect the motives of

their most effective forces, the Army of Africa. On each side, the pressure to prove one’s bona

fides became intense, as individuals looked for markers of loyalty. Each side’s rearguard

monitoring was put to use in tracking down deserters and in persecuting their families. Thus,

desertion in the Spanish Civil War affected its outcome, the way it was fought, and the political

competition and confrontation that occurred on each side. This dissertation has set out, therefore,

to explore the origins of desertion and the ability of armed groups in civil wars to prevent it from

occurring.

This dissertation argues for the general importance of control in preventing desertion, and for the contingency of norms of cooperation on the preferences of an armed group’s combatants.

It has demonstrated the plausibility of these arguments in qualitative analyses of the Spanish

Civil War, and tested them in both micro and macro contexts. It has contributed a theoretical synthesis of past work on the cohesion of armed groups, unique data from Spain, and the first cross-national data set on defection from government forces in civil wars.

To conclude, I draw out some further testable hypotheses that arise from my approach, outline steps for further theory-building, and discuss the implications of this dissertation in terms of how we view civil wars. I explore testable hypotheses at three levels of civil wars. The prospects for desertion should affect actors’ calculations regarding the decision to initiate war,

323 because desertion affects the strength and weakness of armed groups. The likelihood of desertion should also affect the effectiveness of certain tactics that depend on granting soldiers a greater degree of autonomy. And finally, desertion should make an armed group less likely to win. I also consider future directions for theory-building. Three important next steps for theoretical development on desertion and defection, using my work as a starting point, are to break down the types of desertion by destination more thoroughly, to complement the arguments made here with theorizing about the interaction between armed groups, and to consider voice in addition to exit and loyalty as options for dissatisfied combatants. In the meantime, however, my argument makes several contributions to the study of civil war. It helps us to synthesize important theoretical perspectives regarding motive and opportunity as influential factors in civil wars. It helps us to clarify the basic character and requirements of armed groups in civil wars. It can help us think about the state-building consequences of civil wars more carefully and clearly, with a surprising implication: that unified armed groups, because they frequently have a narrow base of support, may actually impede state-building in the long run. And it affects the options available to third-party policy-makers, ranging from state supporters to international organizations seeking to build cohesive and stable security services in the future.

9.2. Testable Hypotheses

The approach offered in this dissertation provides us with new testable hypotheses. In this section, I explore such hypotheses at several different levels: in terms of the initiation, tactics, and outcomes of civil wars.

324

9.2.1. Initiation of Civil Wars

A good understanding of desertion can potentially help us better understand the initiation of civil wars, the subject that has by far received the greatest degree of sustained attention in the recent civil war literature. Desertion affects the opportunity structure available to rebels and governments. Generally, two non-mutually-exclusive perspectives have dominated the study of opportunities for civil war. One focuses on the requirements for organizing a rebellion, especially financing. 5 The other focuses on the ability of the state to respond to such rebellions,

through the capacity to police its terrain, the strength of its institutions, the perverse

consequences of its access to natural resource wealth, and the instability of its neighbourhood. 6

On either of these views, the propensity of government armed forces to defect or desert

can have a large bearing on civil war initiation. An initial defection can form the nucleus of a

rebel group, making rebellion feasible. In addition, desertion diminishes the capacity of the state

to respond to threats. Prospective rebels can calculate that if desertion is probable, their chances

of success are higher. General Emilio Mola, the key planner of the coup attempt in Spain,

vacillated all through the spring of 1936 as he attempted to calculate who would defect and who

would not; realizing that a coup had only a small chance of complete success, the rebel leaders

concluded that they would probably have to fight a civil war, and planned accordingly. 7

Of course, if the government also sees the likelihood of desertion as high, it has incentives to concede more in bargaining to avoid a war. This sort of logic, in which bargaining could allow actors to rationally avoid a costly war, suggests framing desertion in terms of

Fearon’s rationalist explanations for war: private information and commitment problems. 8 The

5 Collier and Hoeffler 2004; Collier, Hoeffler, and Rohner 2009. 6 Hegre 2001; Fearon and Laitin 2003; Goldstone et al. 2010. 7 Payne 1967, 333–334. 8 Fearon 1995. 325 possibility of desertion can create an information failure. Specific soldiers obviously have large incentives to keep their preferences hidden, as I have argued throughout this dissertation. If they become the nucleus of a rebellion, then the connection between hidden information, defection, and war initiation is obvious. But soldiers may also have clandestine contacts with prospective rebels, promising to defect against the government in case of a rebellion, as General Mola did with the Carlists and Falange in the lead-up to July 1936. 9 Rebels may therefore be privy to information about soldiers’ preferences to desert or defect that the government side does not have access to. In addition, desertion or defection, as I have argued in this dissertation, can result from a commitment problem. Soldiers who fear a purge from a different faction, for example, might rebel rather than submit to it. Thus the risk of desertion and defection can work via information failures or commitment problems to bring about a war. I therefore expect that factors making defection more likely, such as a lack of external military support or a recent coup, should also make war initiation more likely, other things being equal.

9.2.2. Civil War Dynamics: Strategy and Tactics Selection

Civil wars also vary by the strategies and tactics that armed groups employ. To frame the debate, in Kalyvas’ scheme, civil wars have included conventional conflicts like the Spanish and

American Civil Wars; irregular wars as in the Algerian and Angolan wars of independence, fought by weak rebels employing guerrilla tactics against strong states; and symmetric non- conventional civil wars, with irregular tactics on both sides, as in Liberia. The technology of warfare can have significant repercussions. Since violence in civil war, in Kalyvas’ scheme, stems from the inability of actors to identify friend from foe, violence against civilians should be

9 Payne 1967, 335–339. 326

greatest in symmetric non-conventional and irregular wars, and least in conventional wars. 10 In addition, one of the primary ways of thinking about victory and defeat in civil war is to consider such outcomes as resting to a large extent on the mix of strategies employed. For Arreguín-Toft, for example, the weak can win wars, provided they employ different strategies than the government; to take the most common example, using guerrilla tactics against a government army that employs conventional tactics. 11 Lyall and Wilson confirm that rebels’ use of guerrilla tactics is more successful the more the government army is beholden to conventional tactics: incumbents are substantially less likely to win the greater the number of vehicles per soldier in their armed forces. 12

Guerrilla tactics involve hit-and-run attacks and hiding, whether in remote territory or within the civilian population. 13 Such tactics, as Jeremy Weinstein argues, present serious agency problems: if an armed group sends a unit to perform a particular military mission, for example attacking a police outpost in a village, the group can have little guarantee that the mission will be performed. Guerrilla war demands that fighters be dispersed, for example into small cells.

Therefore, argues Weinstein, the ability of an armed group to implement such tactics may depend crucially on how much their combatants can be trusted. 14 One interesting question in the history of the Spanish Civil War is why the Republic, on the back heel from at least the winter and spring of 1937 onward, did not make more extensive use of guerrilla warfare. The reasons remain unclear, but the guerrilla units that did emerge did so among the particularly reliable.

There were bands of draft dodgers and deserters behind the lines, locked in to fighting Franco because they had little alternative, and unable to make it back to the Republican lines. As for the

10 Kalyvas 2005. 11 Arreguín-Toft 2006. 12 Lyall and Wilson 2009. 13 Kilcullen 2010; US Department of the Army 2007. 14 Weinstein 2007. 327

Republic’s own active efforts at organizing guerrilla war, they were limited to carefully-selected, highly-trained troops. 15 It is not clear that this is the case, but it may be that the Republic was only willing to entrust these delegation-intensive tasks to its most trustworthy combatants, and otherwise preferred to make use of conventional war, given that it could.

A similar problem may exist in conventional warfare as well. The dramatic increase in firepower at the turn of the 20 th century, according to Stephen Biddle, created strong incentives

for the “modern system” of infantry tactics: using terrain as cover, dispersing forces, and relying

on independent movement by small groups of men. This was a tremendous shift from massed

men in firing lines, for example. And in comparison to masses of infantry, especially those

performing the same task in unison, the dispersal of infantry under the modern system offers far

fewer opportunities to monitor men. 16 Accordingly, just as with guerrilla war, conventional

armies may be limited in their ability to pursue “modern-system” tactics by the degree to which

they trust their men. However, in general, conventional war offers many more opportunities to

monitor and control soldiers than guerrilla wars do. Dispersal of forces is a property of the

battlefield in conventional wars; outside of battles, conventional soldiers may be relatively easy

to control on bases, and the clear separation between each side should reduce the costs of

monitoring. In contrast, in irregular wars, fighters may be dispersed out of contact with their

commanders for long periods of time, without clear lines of control that set points at which to

detect defection. 17

In Spain, the reliability of the Nationalists’ most tactically capable troops, the Army of

Africa, who had trained in flexible mobile warfare in Morocco, meant that Franco did not have to trade off between flexibility and reliability. Franco employed the Army of Africa in swift, mobile

15 Whaley 1969; Pons Prades 1977; Nieto 2007. 16 Biddle 2004. 17 Kalyvas 2005. 328

attacks throughout the first summer of war in 1936. However, Franco shifted his operational

approach soon after his armies made it to Toledo on 27 September 1936, towards a slower and

more methodical approach to offensives to ensure complete victory over the Right’s political

opponents. Still, however, the flexibility of the Army of Africa had given the Nationalists an

important early advantage. 18

Combatant for combatant, therefore, armed groups ought to be better able to deploy

flexible tactics and hence more effective in battle when their soldiers share the group’s common

aims, whether they use conventional or irregular war. Their effectiveness, that is, should not only

stem from disinclination to desert, but also from a greater ability to pursue flexible infantry

tactics. In addition, however, the selection of a technology of warfare might be affected by

propensity to desert as well. If guerrilla warfare requires greater flexibility and has fewer

opportunities for monitoring, then its success may depend more on the reliability of troops than the success of conventional war does. As such, forces with a large number of combatants of dubious commitment may be more inclined to opt for conventional war, and may face greater challenges in implementing irregular war than smaller and more unified forces.

9.2.3. Civil War Outcomes

A principal motivation for studying desertion is the presumption that it has an important effect on the outcomes of civil wars, and I argue above that it played this role in Spain. Keeping combatants fighting is a core requirement for any armed group. Therefore, generally we should expect that armed groups that experience desertion should be less likely to win and more likely to lose, other things equal. Moreover, my specific approach to desertion also provides novel hypotheses. Given the role of external support in explaining defection, it is possible that

18 Cardona 1985; Preston 1994b. 329 defection could account for some of the effect of government-supporting interventions in increasing the likelihood of a government victory. 19 Intervention, however, is only one form of external military support; military assistance to governments short of troop contributions has yet to be studied in the literature in its effects on civil war outcomes. In addition, there has not been comparable research on the role of recent pre-war coups in duration and outcome. To the extent that coups are studied, the question posed in the literature is whether the war itself begins with a coup attempt or not, rather than recent pre-war history of coups. Moreover, while the effects of coups on duration have been examined, with the literature establishing that wars that begin with coup attempts tend to be shorter than others,20 their effects on who wins have not received

attention. There is thus no current research on the effect of recent pre-war coup history on civil

war outcomes. I anticipate that a recent coup should make government victory less likely.

In my cross-national data, there are some hints of a relationship between defection and

outcomes (see table 9.1), using Doyle and Sambanis’ coding of civil war outcomes.21 Of the nine civil wars that involved mass defection from regular government forces and had ended when

Doyle and Sambanis were coding, not a single one ended in a government victory. In contrast, the rate of government victory among civil wars that had ended was 30.8% when regular forces had not experienced mass defection. This difference in likelihood was statistically significant ( p

< .1). However, an outright rebel victory was not any more likely with mass government defection. Nor was there any sort of relationship between militia defection and outcomes. The table indicates that any relationship between defection and outcomes was driven by the likelihood of government victory. If there is a relationship between defection and war outcomes, then it appears possible that it is one of sufficiency: mass defection appears sufficient to prevent

19 Balch-Lindsay, Enterline, and Joyce 2008. 20 Fearon 2004; Cunningham 2006; Cunningham, Gleditsch, and Salehyan 2009. 21 Doyle and Sambanis 2006. 330

Table 9.1. Defection and war outcomes

Mass regular defection Outcome No Yes (War ongoing) 3 2 Rebel victory 2 2 Truce 3 4 Negotiated settlement 4 3 Government victory 4 0

a government victory, but provides little information about other outcomes. The possibility of rebel victory and the feasibility of a negotiated settlement appear to be driven by other factors. In any case, a larger number of cases would be necessary to make strong conclusions about the macro patterns of defection and outcome.

These also are simply bivariate results, not taking into account any of the covariates of civil war outcomes that have been uncovered, such as the military balance between the two sides, 22 or the characteristics of external intervention.23 Further analysis of defections and outcomes will have to use a higher N to allow for effective statistical control. In addition, however, it will be vital when testing the impact of defection on war outcomes to conduct over- time analysis. Soldiers may be more likely to defect if they believe that the war is lost, and certainly there was a wave of defection for this very reason in the final stages of the Spanish

Civil War. By focusing on defection early on in a given civil war, it may be possible to account for this scenario and instead isolate the independent contribution of defection on war outcomes.

9.3. Improving the Theory

Beyond the testable hypotheses that arise from the theoretical perspective developed in this dissertation, the perspective itself could be made more sophisticated in several respects.

22 Cunningham, Gleditsch, and Salehyan 2009. 23 Balch-Lindsay, Enterline, and Joyce 2008. 331

Future theorizing could treat the opportunities for dissatisfied soldiers to voice their concerns rather than just desert or remain silent. It could also attempt to distinguish among types of desertion more carefully, and to consider bilateral interactions among armed groups.

9.3.1. Voice, Not Just Exit and Loyalty

Albert Hirschman’s classic treatment of how individuals who are dissatisfied with an institution—whether consumers responding to a decline in the quality of a product offered by a firm, or members of an organization responding to declining performance—raises the prospect of three options, not just two. Individuals could leave (exit), stay and remain silent (loyalty), or stay and lobby the organization for change (voice). 24 The approach I have taken in this dissertation

only really deals with the first two options. However, the possibility of voice could make exit

(and loyalty) a less popular option. That is, if an armed gives its combatants the chance to

articulate what it is that they find dissatisfying, it could respond and keep those individuals

serving rather than risk them deserting. Jeremy Weinstein argues that institutions within the

armed group that are set up to permit such responsiveness, such as group participation in

decision-making, have an additional benefit in indicating to combatants that they are trusted by

the armed group leadership. However, he also acknowledges that such decentralization requires

common aims to function well: “To make it work, commanders and combatants must share an

understanding of how they will respond to the various situations they encounter.” 25 If combatants

have serious disunity in their aims and preferences, then decision-making fora and other forms of

decentralization could just paralyze an armed group. In the Spanish Civil War, constant

discussion and debate about orders made it quite difficult to accomplish military tasks, as even

certain Anarchist leaders were willing to acknowledge. These constant discussions could actually

24 Hirschman 1970. 25 Weinstein 2007, 137–138. 332

create greater dissatisfaction than they resolved: indeed, for example, José María Aroca deserted

his anarchist unit explicitly for this reason. 26 Moreover, election of officers, as in Anarchist units,

could become a veto for soldiers to avoid serious military duty. As such, elections became part

and parcel of some units’ reputation for lax discipline, encouraging self-selection of

uncommitted individuals into those units. It is therefore unclear that, in the absence of common

aims, institutionalized options for voice could really achieve a reduction in desertion rates.

However, given units with substantial commonality in aims, they may well be a way of

channelling marginal dissatisfactions into reform in the unit rather than desertion.

9.3.2. Types of Desertion

The theoretical approach outlined in this dissertation essentially focuses on dynamics internal to a single armed group. It assesses the armed group’s characteristics and establishes hypotheses that relate those characteristics to the decisions of individual soldiers to leave and the effectiveness of different measures in dissuading them from doing so. This focuses attention on the decision to exit rather than the destination. I lump together three different actions under the heading of leaving—desertion proper (to the rearguard) and defection, consisting of side- switching (to an existing rival group) and fragmentation (splitting off to start a new rival group).

I regard this as a useful move in some respects, focusing attention on the difficulty of knowing what a combatant intends and on efforts to prevent soldiers who want to leave from doing so.

However, this theoretical move leaves considerable room for nuance. If a combatant is only considering one or two of these destinations, then what happens with regard to the other three are of little or no interest to him. For example, a combatant who does not particularly want to fight at all would have utilities as follows: desert > remain > defect. In contrast, the utilities

26 Aroca Sardagna 1972. 333 for a soldier who finds himself on the side he disagrees with might instead be: defect > desert > remain. For the first soldier, efforts to prevent him from defecting, such as control at the front line, are irrelevant; policing the rearguard, much more so. While policing the rearguard is still relevant for the second soldier since his family can be held hostage, relative to the former soldier policing the front lines should be much more effective.

In addition, fragmentation is particularly distinctive from the other two forms of desertion. It is a group activity, never undertaken by individual combatants, since a single individual cannot constitute his own armed group. And it requires certain resources that desertion proper and side-switching do not, especially startup funding. The brief rebellion of CNT and

POUM activists in Barcelona in May 1937, and the consideration of such a rebellion among certain Falange and Carlist activists one month prior, were the key instances of a threat of fragmentation in the Spanish Civil War. In each case, they were provoked by fears about marginalization and repression, but in each case the state’s use of force prevented the rebellions from getting very far. The two instances are still useful illustrations of the tensions of preventing and provoking desertion that I have attempted to uncover in this dissertation, but the incidents also tend to show that fragmentation is a particularly difficult form of desertion to pull off. In this

vein, Weinstein finds that lootable natural resources are associated with a greater number of

armed groups in civil wars. 27 Fjelde and Nilsson similarly argue that greater opportunities to enter the civil war, as given by such features as government weakness and conflict over the composition of the government as opposed to over territory (since, in conflicts over the latter, existing armed groups are likely to have a strong comparative advantage in mobilization) are important predictors of the number of armed groups.28

27 Weinstein 2007, 330. 28 Fjelde and Nilsson 2011. 334

One area in which my dissertation does focus on some types of desertion rather than

others is in Chapter 8, where I explore broad patterns of defection , both side-switching and

fragmentation, and ignore desertion. This was a methodological choice that was, to a certain

extent, forced by the difficulty of obtaining reliable data about desertion from civilian militias. It

is possible that my conclusions in that chapter would not fully translate to desertion proper. For

example, a coup in the previous five years might not necessarily prompt a return to the rearguard.

In particular, coups may create tensions within the officer corps. Officer corps factions, since

they are made up of relatively prominent individuals with a comparative advantage in the use of

force, may be more inclined to defend themselves actively against their adversaries, whether

through fragmentation or through aligning themselves with rebel groups, than to go home.

9.3.3. Bilateral Interactions

Future work on desertion, therefore, could attempt to add nuance to the perspective that I outline here by focusing more specifically on the requirements for each type of desertion. One particularly important area of focus consists of the interactions among adversaries and how they might shape defection especially. A straightforward thesis is that defection patterns might follow military success. Similarly, Kalyvas argues that territorial control by one side or the other tends to shape local patterns of civilian collaboration, and hence violence patterns. When one side is unchallenged in local dominance, local communities tend to acquiesce to its rule. In contrast, if the other side contests control over a locality, then collaboration with the enemy is possible and so violence is employed to prevent it. Kalyvas presents significant micro-level data about violence in the Algerian and Greek Civil Wars and macro-comparative qualitative evidence from a wide array of civil wars to back up this claim. 29 The claim might be extended to defection

29 Kalyvas 1999; Kalyvas 2006. 335 patterns among soldiers. I argue in Chapter 8 that one connection between external support and low defection rates might be that government troops are more likely to see the government as having advantages when it enjoys external support. And elsewhere, I argue that just as civilians might be more inclined to join an uprising as it gets more powerful and more likely to win, 30 soldiers may switch sides in response to the military balance as well. 31 Certainly, in the Spanish

Civil War, Republican soldiers were more likely to try to defect or desert towards the end of the

war.

However, I also argue in my previous work that there are potential limits to who can defect.32 If a soldier fears indiscriminate violence from the other side, defection may be a losing option. This logic is parallel to my argument in this dissertation about the disutility of indiscriminate violence by an armed group in attempting to keep control over its own troops.

Broad labels such as ethnic identity are sometimes used as a low-cost way of identifying friend from foe, but this limits defection: those that an armed group identifies as foes are unlikely to defect to join it. This argument is also in agreement with an argument Kalyvas makes elsewhere: since indiscriminate violence is counterproductive, armed groups will tend towards selective violence provided that they have the resources to be able to employ it. Indiscriminate violence, for example targeting all members of an opposing ethnic group, tends to be employed by groups with low capacity to identify friend from foe. 33

Thus, for example, Alawite soldiers in Syria have feared, both in the Muslim

Brotherhood uprising of the late 1970s-early 1980s, and again in the Arab Spring revolts since

2011, that they would be targeted for persecution because of their close identification with the

30 Kuran 1989; Kuran 1991; Lohmann 1994. 31 McLauchlin 2010. 32 Kuran 1989; Kuran 1991; Lohmann 1994. 33 Kalyvas 2008. 336

regime, and their defection rate has consequently been quite low. Just as commitment problems

can keep wars going longer, 34 these problems are exacerbated if opponents cannot make credible commitments not to repress defectors. Soldiers might fear that they will be mistrusted as possible

spies or as individuals prone to switch sides again. They might also fear that opposing leaders

will have incentives to present themselves as good nationalists by targeting out-group members.

And they might fear that their support is ultimately disposable: useful for the other side to win

the conflict, but not likely to induce its leaders to make substantive concessions to one’s group.

Hence armed groups have pursued various means of trying to convince the other side of their

credible commitments. For example, in Bahrain in 2011, the mainly Shi’ite opposition gave a

prominent public role to Sunni community leaders in order to bridge the gap between the two

confessional communities. The state targeted those Sunni leaders for detention, with the hope of

painting the opposition as a Shi’ite-only movement.35

There is some evidence of these dynamics in parts of the Spanish Civil War as well. Both

sides interrogated defectors and examined their background to determine whether or not they

were spies. 36 One Communist Party member who defected from the Nationalists recalled finding

these interrogations justified, but still very difficult to go through. 37 In addition to a fear of

espionage, the Nationalists attempted to determine the political alignment of soldiers who

defected; records of interrogations include questions about the circumstances in which the

soldiers joined and why they had defected. Practically everyone replied that they were coerced to

join; there is only a small chance that this is true. 38 Nationalist propaganda promised good

34 On commitment problems generally see Fearon 1994; Fearon 1995; Fearon 1998; Powell 2006. On commitment problems as a source of longer duration, see Walter 1997; Goemans 2000; Fearon 2004. 35 Ned Parker, “In Bahrain, Sunni Activist’s Plight Seen as Cautionary Tale,” Los Angeles Times , 25 February 2012. 36 Corral 2007, 146–147, 236–238. 37 Fraser 1979, 466-467. 38 Centro Documental de la Memoria Histórica [CDMH], Serie Militar [SM], caja 700, folio 19. 337 treatment only to defectors “without bloodstained hands,” and instead vowed “inexorable justice” towards “assassins” trying to switch sides. 39 If this induced some Republicans not to

defect out of fear of the consequences of doing so, it might also have convinced some

Republicans not to fight in the first place, or not to fight at all enthusiastically, in order that they

might be able to defect safely.

One particularly interesting case is afforded by the Moroccan regulares . As noted, their

desertion rate was quite small: only about 1,700 out of 100,000 who fought; side-switchers were

still rarer.40 As I argue in Chapter 4, much of this had to do with the strict controls under which

Moroccan soldiers were placed. However, it is also possible that official and popular Republican

antipathy towards Moroccans discouraged them from switching sides. Possibly to motivate

soldiers and civilians out of fear of the consequences of a Nationalist victory, Republican

propaganda deployed racist stereotypes of Moroccans as the rampaging, raping and pillaging

Moor threatening European civilization, with the Communist orator Dolores Ibárruri denouncing

“morisma salvaje ”—savage Moorishness. 41 One cartoon in the Anarchist newspaper Fragua

Social , entitled “ Civilización Cristiana ,” portrays a stereotypical Moor—swarthy, with a fez with a white crescent and star, wine bottle sticking out of his belt—about to run through a white mother and child with a bloody bayonet. 42 Such propaganda may have convinced Moroccan troops not to try to switch sides; a Spanish defector from a Moroccan company told a commissar in the Popular Army that Moroccan troops would not switch sides because they feared shooting or torture. 43 However, layers of irony and complexity did surround these Republican rhetorical

tropes. The stereotypes they used were the same that many of the Africanista officers who led the

39 Corral 2007, 148–149. 40 Ibid., 379, 437. 41 Ibid., 442. 42 Fragua Social , 4 October 1936, p. 6. 43 Corral 2007, 446. 338

Nationalist side had deployed to justify ongoing Spanish colonial rule in Morocco. The

Nationalists during the war, in contrast, found it convenient to rally Muslims in the name of

religion as against secularism, glossing over past characterizations of Muslims as heathens. At

the same time, the Nationalists also portrayed their own cause as a kind of Reconquista , evoking

the fight to oust Moorish rule in Spain and expel Muslims. And through it all, Nationalists also

put the same racist notions of the raping and pillaging Moor to use for themselves, in the belief

that brutal violence by the Army of Africa would stamp out leftist rebellion as it had in Asturias

in October 1934. 44 Overall, then, Moroccan troops were hardly receiving unified messages.

The discussion suggests several important questions. What practices by side A, especially labelling groups of people as loyalists for side B, deter a potential defector from side B? What incentives exist for side A to implement such practices? That is, is it ever rational for an armed group to limit its potential to receive incoming defectors, and if so, why?

9.4. Implications of Contingent Control and Cooperation

9.4.1. A New Synthesis

A classic debate in the study of internal armed conflict pits motive against opportunity, broadly speaking, in trying to understand how rebellions come to be. For some time, this debate has appeared over, resolved in favour of opportunity—of the practical requirements of organizing armed groups and the political opportunity structures available to set them up. The consensus is a sort of two-step: there are many grievances but few rebellions; the practical requirements of organizing rebellion must then be crucial for explaining why rebellions take place, so much so that group interests become effectively irrelevant in the face of individual self-

44 Balfour 2002, 273–274, 280–281; de Madariaga 1992; Preston 1994a, 177. 339 interest. 45 A major exception is ethnic conflict; attention to this subject in the early 1990s especially saw ethnic group interests as powerful motivators for behaviour. 46 However, with the

exception of recent work on ethnic dominance, 47 scholarship on ethnic conflict as well has accepted a form of the two-step consensus: ethnic grievances could be pervasive, but they emerge generally only as a byproduct of other conditions, such as electoral incentives or the structure of civil society. 48 This approach is still distinct from the consensus on other kinds of conflict in that it focuses on how group interests become salient, thus accepting an important

explanatory role for group interests and not just individual interests. However, there is also an

increasing belief that individuals’ apparently ethnically-motivated behaviour belies their use of

identity constructs to suit their own, essentially private, ends. 49 Hence here there is the outline of a fully analogous byproduct theory of ethnic conflict, emerging for private and not really public ends. Thus we seem to want to account for both the language of the cause employed by armed groups in many settings, and for the “real” political and organizational work that seems necessary to maintain political action.

The issue of desertion is central to understanding the role of motive and opportunity, because keeping fighters in the field is precisely a core requirement of fighting a civil war. In turn, my dissertation, posing the effectiveness of control and cooperation as contingent upon the distribution of preferences in an armed group, provides a new avenue for synthesizing these two views. It suggests that the array of effective organizational options available to armed groups is contingent upon the distribution of motivations. Thus motive and opportunity are not really

45 Tilly 1978; Skocpol 1979; Fearon and Laitin 2003; Collier and Hoeffler 2004; Collier, Hoeffler, and Rohner 2009. 46 Posen 1993b; Fearon 1994; Lake and Rothchild 1996; Kaufmann 1996b; Kaufmann 1996a. 47 Cederman and Girardin 2007; Buhaug, Cederman, and Rød 2008; Wimmer, Cederman, and Min 2009; Cederman, Wimmer, and Min 2010; Cederman, Weidmann, and Gleditsch 2011. 48 Horowitz 2000; Fearon and Laitin 1996; Brass 1997; Varshney 2001; Wilkinson 2004; Posner 2004. 49 Mueller 2000; Chandra 2004; Kalyvas 2006; Kalyvas 2008. 340

separate, exogenous conditions. Rather, the distribution of motives will strongly affect

organizational challenges.

This provides a theoretical advance. It acknowledges the powerful critique levelled by

organization theorists that motives cannot be sufficient to explain the existence of armed conflict. It also attempts to be true to the power that individuals’ political commitments often have, avoiding Olson’s troubling conclusion that they do not matter at all. It also avoids relegating motivations only to the role of distinguishing the few true believers, “everyday

Kantians” 50 who may be necessary to get a movement going, from the mass of uncommitted individuals who really matter in a movement’s strength. Instead it argues that motives have a broader, organizational import. In doing so, it attempts to resolve a tension within collective action theory: there are multiple solutions to the collective action problem, but some of them (I have focused on norms of cooperation) require that there be a common aim while others (I have focused on control) do not; therefore the distribution of motivations must , by the lights of organization theory itself, matter.

This synthesis, then, also challenges the idea that there is a single set of requirements for armed group success, whether the one (a strong cause) or the other (selective incentives). Rather, it asserts that there are multiple paths to cohesion and multiple paths to desertion. A sufficiently motivated armed group can rely on norms of cooperation, while an armed group populated by less committed combatants can still fall back on control. My analysis complicates this somewhat, as I argue that severe disagreement can undermine the effectiveness of control to a degree, but the overall argument stands.

Nor, I think, has this perspective unduly multiplied solutions. Almost twenty years ago,

Mark Lichbach called for the recognition that there were multiple solutions to the collective

50 Elster 1985. 341 action problem (he identified some two dozen, grouped in four categories). Lichbach argued that analysts should focus on the conditions in which each operated. 51 I have attempted to do so here,

while still retaining parsimony by keeping focus on two of the most theoretically important and

influential solutions, control and norms of cooperation.

9.4.2. The Diversity of Civil Wars

If there are multiple ways of keeping cohesion, we can better understand and analyze the apparent diversity of groups within civil wars, and perhaps of civil wars themselves. The actors in civil wars evoke multiple different images: committed cells of passionate revolutionaries; ethnic chauvinists targeting their groups’ rivals; communities banding together in fear of a predatory state; and all manner of charlatans, profiteers, and gangsters exploiting conflict to their own ends. There have been several different responses to this diversity, as I argued in Chapter 1.

On the one hand, some theorists have sought to make sense of civil wars in terms of master, overriding narratives: for example, Left versus Right, Contras versus Sandinistas, Serbs versus

Muslims, Sunnis versus Shi’ites versus Kurds. We have seen a thorough critique of this reductionism in recent years. There are many more grievances than there are civil wars, and individual motivations are generally too complex to be easily reduced to a single political program. Thus theorists have instead looked beyond stated common aims and toward other theoretical devices that can help us make sense of civil wars. One group of scholars focuses on opportunity and ultimately on the motive for material gain that seems to render common aims irrelevant. Kalyvas, in contrast, favours building complexity in as an assumption, and then interpreting action in civil wars from the point of view of control and coercion—a force that can cut through this bewildering array of motivations and provide clarity. I have argued that neither

51 Lichbach 1994; Lichbach 1995. 342

the assumption of simple narratives nor the assumption of the effective irrelevance of common

aims is really tenable. It is an oversimplification to presume that individuals can be reduced to

their political leaders’ programs, but attachment to common aims really does seem to vary and to

alter the basic challenges that armed groups face in organizing for war.

One might also see the diversity of civil wars as a new phenomenon. Scholars working in

the late 1990s forwarded a distinction between “old” wars, fought by disciplined armies for

coherent ideological causes, and “new”, post-Cold War civil wars, fought by criminal gangs; to

the extent that there are goals to such conflicts—there frequently are not—they are for ethnic

chauvinism rather than broad ideology. 52 The perspective articulated here is one in which armed

groups may have had common causes at one point but now do not; they have been replaced by

criminal predation.

Kalyvas rejects these easy stereotypes of old and new wars, drawing our attention

instead, in particular, to the elements of self-interest in old wars and the persuasive group appeals

that often exist in new wars. While civil wars still differ from each other, the differences are thus

much less pronounced than they are in the New vs. Old Wars formulation. In particular, there are

elements of complexity in each: local but still political and group interests rather than pointless

wars of greed alone; popular support sitting side by side with alienated civilians; and both

indiscriminate and selective violence. 53 And while the end of the Cold War did mark a shift in the technology of civil war, weakening both rebels and governments while permitting military splits, 54 these macro shifts should not, on Kalyvas’ view, be reduced to mere categories and stereotypes; nor do they imply a contrast between old wars that had a point and new wars that do not. Kalyvas levels a powerful critique of the too-easy labelling of new and old civil wars. He

52 Keen 1998; Kaldor 2006; Mueller 2004. 53 Kalyvas 2001. 54 Kalyvas and Balcells 2010. 343 persuasively argues that a perceived absence of aims in recent civil wars has more to do with the absence of the Cold War analytical category of Capitalist vs. Communist—a category that itself belied significant local conflicts—than with any real changes to the aims of civil wars. We should not, therefore, make sense of variation across armed groups just in terms of eras of war.

9.4.3. State Building

The contingency of control and norms of cooperation also suggests a new view of the state-building and state-destroying aspects of civil wars. Civil wars in the post-Cold War era have been occasionally regarded as part and parcel of “state failure” or “state collapse,” a consequence of a vacuum of authority. 55 Rotberg puts this starkly: a collapsed state “exhibits a vacuum of authority. It is a mere geographical expression, a black hole into which a failed polity has fallen. There is dark energy, but the forces of entropy have overwhelmed the radiance that hitherto provided some semblance of order and other vital political goods to the inhabitants.” 56

However, others note that the picture is more complicated. State building has long been a violent process, as states eliminated rivals through force.57 Violence internal to a particular territory could indicate either the absence of the state or a process of state creation. 58 Within civil wars themselves, authority is not absent , but reconfigured and challenged. Certainly the central state

typically does not have nearly as thorough control over its whole territory as in peacetime, but

this does not imply the negation of statehood. Rather, different state projects are posed against

each other, 59 in what Charles Tilly calls a revolutionary situation. 60 The entities that rebel groups

55 Posen 1993b; Rotberg 2004b. 56 Rotberg 2004a, 9. 57 Tilly 1975; Tilly 1985; Tilly 1992. 58 Cohen, Brown, and Organski 1981. 59 Kalyvas 2006, 17–19. 60 Tilly 1978, 191. 344

create frequently resemble the states that they oppose, 61 enacting authority and governance over

the people living in areas under rebel control. Civil wars thus have the potential to be state-

building enterprises as well as state-destroying events, and can be analyzed in terms of the

political authority structures that they create.

In this context, the contingency of control and norms of cooperation on the distribution of

motivations reminds one of a central tension in the state about who its beneficiaries are, and how

separate it is from other societal actors. In the Hobbesian tradition, scholars have treated the state as an autonomous actor able to stand above contesting social actors and regulate their conflicts, or treated it in terms of its ability to provide public order to all. 62 Alternatively, the state could

favour certain social groups, as in the Marxist frame; recent scholars have noted the important

distributional consequences of the specific institutional arrangements that make up the state. 63

Scholars have also attempted to bridge the gap between these views, conceiving of state strength in terms of the degree to which the state provides public goods or is instead captured by particular, narrow interests. 64

Considering armed groups in civil wars as states in the making, then, common aims

among combatants take on new importance. An armed group may face a tradeoff between

breadth of support on the one hand and reliability on the other. Common aims can be facilitated

by just having a narrow base of support. For example, it is not difficult to attribute common aims

to the Republican Guard or Fourth Armored Division in Syria, almost exclusively Alawite units

led by the President’s brother. Those common aims may lead to collective action within such

groups. However, in the context of a state, a narrow base implies a regime that is more likely to

61 McColl 1969. 62 Skocpol 1985; Olson 1993. 63 Knight 1992; Boix 2003; Acemoglu and Robinson 2006. 64 Huntington 1968; Olson 1982; Bueno de Mesquita et al. 2003; Bates 2008; Acemoglu and Robinson 2012. 345 be exclusivist, to provoke rebellion, and to employ violence to put such a rebellion down. 65 If they derive from a narrow base, common aims, which provide the foundation for norms of cooperation and thus a solution beyond control in a civil war context, can instead imply exclusivity, instability and violence in a state context. States in conflict-prone environments, whether incumbents concerned about rebellions or victorious rebels, may face vicious tradeoffs on these issues. On the incumbent side, according to Philip Roessler, states in Sub-Saharan

Africa have actively narrowed the ethnic basis of their regime in order to prevent military defection in the form of coups, thus proving willing to trade off long-run stability for short-run loyalty. 66 The tradeoff is also evident on the rebel side. While the victory of one side or the other

does seem related to a lower likelihood of civil war resumption than a negotiated settlement or a

truce, because it eliminates the capacity of the other side to resume fighting, 67 if it creates a new

government with a narrow base, it could still prompt future instability. For example, the National

Resistance Army in Uganda, Weinstein’s paradigmatic case of a unified, well-ordered, well-

disciplined armed group with strong norms of cooperation, derived these characteristics in the

first instance from their strong community appeal to the Banyankole and Baganda of the south

and south-west of the country. After its victory over the Okello regime in 1986, however, it

proved difficult to reconstruct this consensual, cooperative approach to governance outside its

base, across the whole country, though it made significant early efforts in this regard. The narrow

base of the new regime told: NRA soldiers have regularly persecuted Acholi civilians in the

north of the country, who have found themselves caught between the brutal Lord’s Resistance

Army and an unresponsive government. 68

65 Gurr 1993; Cederman, Wimmer, and Min 2010; Heger and Salehyan 2007. 66 Roessler 2011. 67 Licklider 1995; Toft 2011. 68 Mamdani 1996, 215–217. 346

In contrast, armed groups without a strong sense of common purpose may actually,

paradoxically, do better as states. That lack of a common purpose may just emerge from a lack of

particularism or chauvinism, and the search for a broader coalition. Even if such an armed group

does not inspire a clear preference among its combatants in the short run, at the very least it may

avoid the problems that come from having a narrow base. In addition, consider the results if such

an armed group wins and becomes a new state. Provided control is exercised by the group

through selective rather than indiscriminate violence, such an armed group may provide its

members with the private good of personal security and hence with the Hobbesian public good of

political order. As such it may actually provide a widespread common aim endogenously.

Having won the civil war, the emergence of a rebellion would constitute a threat to that common

aim, and its soldiers, whatever they think about the distributional consequences of the regime,

may gain an interest in keeping order alive. The victorious armed group essentially comes to

resemble the sort of state outlined in recent theories of statehood: a specialist in violence with an

incentive to provide public goods such as public order, which can obtain the consent of the

governed if it commits not to use its power in an arbitrary fashion. 69 The possibility of a paradox

of state-building, with cooperative armed groups coming to be narrow, violence-prone regimes

and stable states emerging from armed groups that do not engage much enthusiasm, is highly

intriguing. It remains mainly speculative for the moment, but suggests a fascinating line of future

research.

However, the consequences of control in armed groups for state building may have much

to do with the sources of such control, especially as those resources affect the sustainability of

control. As I argue below, the issue of sustainability is crucially important in understanding

international assistance. Since the contemporary environment of state-building now engages

69 North and Weingast 1989; Olson 1993; Bates 2008. 347 much more international assistance than the Western European story on which so much of our thinking rests, 70 the state-building potential of top-down rewards and punishments is now more complicated.

9.4.4. The International Context

The contingency of control and norms of cooperation frames the context for international action in civil wars. International support is extremely valuable for armed groups, for obvious reasons. But does it keep cohesion? Many argue not: external support may, instead, come with moral hazard problems, giving groups that receive support no real reason to enact institutional reforms to become more stable and coherent entities, such as improving the state’s fiscal base through taxation rather than rent-seeking, breaking alliances with dubious warlords, or gathering popular support through broad-based, legitimate government. 71 Instead, external interveners may

need to be careful and selective in the nature of the local armed groups they support; on this line,

U.S. delegation of security to local Sunni leaders in Iraq, so crucial in the short-term success of

the “surge”, created an institutionally incoherent security apparatus based on a temporary local

alliance. For similar reasons, argue these critics, NATO should be quite concerned about

President Karzai’s use of warlords. 72 My argument does build on these sceptics’ insight that it is more wasteful to spend money on some groups than others, noting that if a government experienced a coup in the previous five years, external support seemed to have very little correspondence to defection rates. Beyond coups, sources of factionalism such as the concatenation of local interests or the constant jockeying of different warlords in Afghanistan, may make it clear that some government armies are poor candidates for support.

70 Tilly 1985; Jackson 1990. 71 Byman 2006. 72 Simon 2008; Peceny and Bosin 2011. 348

However, I also find that external support can actually be associated with reduced

defection rates in government armies, even among conscripted armies. This confirms anecdotal

evidence that the withdrawal of support, for example in Afghanistan or in the post-Soviet world,

can lead to massive waves of defection. The finding is essentially short-term; there may still

indeed be problems of moral hazard, with governments failing to find their own long-run sources

of either popular support within the army (for common aims and hence norms of cooperation) or

military capabilities (for rewards and punishments). In a sense, this finding poses the dilemma

that sceptics raise even more sharply: it means that present reliability may mask future

unreliability. If external support mitigates problems of defection in the short run, then that

external support may hide longer-term defection problems. The fear with the “surge” in Iraq then

takes on new dimensions: not just that it built multiple and fractious interests into the security

services, but that the differences among those interests would be difficult to detect, precisely

because of the increased presence of American troops at the time. If American support to the

Iraqi army carries on declining, then, that fractiousness may become newly exposed.

To draw this point and the last together, external support is frequently argued to provide a

special and particularly problematic context for state-building. States enjoying external support

do not have to become more representative or gain a sustainable fiscal base, a weakness that the

end of the Cold War and the withdrawal of such support helped to expose. 73 The long-run

consequences of rewards and punishments for state building may thus have much to do with the

sources of the resources that underpin the dispensing of these rewards and punishments: are the

sources sustainable? And are they set up in such a way as to provide public support and hence

broad-based common aims for a state and its armed servants? External actors intent on

supporting a particular government’s armed forces ought to be aware of the potential tensions

73 Tilly 1985; Jackson 1990; Reno 1998; Herbst 2000. 349 between short-run cohesion and long-run durability. External support ought to be coupled with institutional reforms that help to ensure that violence is wielded in a selective rather than an arbitrary fashion, such as codes and reliable procedures of military justice. It may also not be successful over the long run without steps to improve national governance so as to channel political conflict and hence reduce the reasons for factionalism within a state’s security services.

And its success may also depend on making the fiscal basis supporting the coercive apparatus of the state more sustainable into the future.

External interveners can also play a vital role in civil wars in underwriting negotiated settlements, not just in affecting the character of military institutions on one side or another.

Peace agreements are quite challenging to implement, resulting in a higher likelihood of civil war resumption than when there is victory of one side or the other. 74 Credible commitments among the various parties to conflict are quite difficult to achieve: vulnerable actors will want to take up arms again if they fear that the other side will renege, 75 and a truce or negotiated

settlement can leave them with the capability of doing so. 76 Within this context, defection can

play a large role. Spoilers—splinter groups that stand to lose from a settlement—can split off and

keep a war going. 77 The military integration of rebels into the armed forces can founder, with the armed forces splitting apart again on the path to resumed war. 78 Military defection is, therefore,

an important way in which peace agreements can fall apart.

Consider military defection during a peace process in light of this dissertation, then.

Peace processes are likely to be contentious, with different sub-groups of an armed group standing to gain or lose, and ex-enemies mistrusting each other. Severe disagreement is therefore

74 Licklider 1995; Toft 2010; Toft 2011. 75 Walter 1997. 76 Licklider 1995. 77 Stedman 1997. 78 Glassmyer and Sambanis 2008. 350

quite frequent, though it does vary.79 I have argued in the dissertation that norms of cooperation

are unlikely to emerge on their own in such a setting. I have further argued that the alternative to

such norms, the provision of control, is generally less effective in such a setting than under

greater unity. The exercise of control can be, or appear to be, arbitrary—in pursuit of factional

interests rather than really preventing desertion or defection.

This is quite analogous to the problems facing peacebuilding under a truce or treaty. The

central difficulty of such situations is that unlike decisive victory, there is neither central control

nor real unity among the relevant military forces, a situation that I argue is prone to defection.

And such situations entail credible commitment problems, just as the establishment of control in

environments of severe disagreement is problematic because of the fear of arbitrary or

exploitative violence for narrow, particular aims. In short, post-conflict situations under truces

and treaties are rife with the potential for defection. Thus Greenhill and Major find compelling

evidence that spoilers in peace processes emerge when they have the opportunity (because of

ineffective coercion) or when commitments to them are not credible and so they fear the

consequences of holding to an agreement. 80 And Glassmyer and Sambanis find a weak positive correlation between rebel-military integration and the resumption of civil war, a correlation that seems to be driven slightly by whether they are implemented formally or informally but mainly by whether they are implemented under circumstances conducive to trust or mistrust (such as the damage the war has inflicted in the past). 81 This finding suggests confirmation that whether

former adversaries can serve together in the same force is really contingent on whether the

circumstances foster intense mistrust among them.

79 Doyle 2002; Doyle and Sambanis 2006. 80 Greenhill and Major 2007. 81 Glassmyer and Sambanis 2008. 351

Since the literature on credible commitments in civil wars highlights the importance of external actors in guaranteeing peace, making it costlier to resume fighting, 82 it is worthwhile considering the role that external actors might have specifically in preventing military defection in peacebuilding. This takes the international role beyond support alone. External actors may need to take on the responsibility of responding to factionalism within armed groups, at the same time as they provide assistance to improve the central capability to provide rewards and punishments, including enacting control over soldiers. Their challenge, as I see it, is to assist with the selective use of reward and punishment even as they assist with the capacity to control soldiers. Several items might therefore be usefully on the peacebuilding agenda. Third parties could serve as observers within the armed forces to try to ensure that clear procedures to pursue military justice are actually followed, so that arbitrary denunciation for treason is less possible, and conflicts among soldiers have a regular mechanism for resolution. They could provide oversight to help ensure that processes of promotion are followed by the book. They could try to facilitate after-the-fact military policing, to attempt to find deserters after they leave so that the armed forces do not have to rely on suspicion-prone procedures of anticipating who is likely to desert.

However, it is vital for peacebuilders to respect the limits of what they can do. Intense mistrust is extremely difficult to overcome, especially for a third party. In the long term, serving as a guarantor between two parties may just serve to kick the can down the road, staving off acting on mistrust only because a third party is temporarily involved. The prospect that external support for an armed group can limit defection may then have a somewhat different appearance: as a source of temptation towards policy steps that may not have much effectiveness.

82 Walter 1997; Fortna 2004; Doyle and Sambanis 2006. 352

9.5. Conclusion

Though the incidence of internal armed conflict has been on the wane since its peak in the mid-1990s, we still live in a world in which such conflicts pose the principal challenges in securing human beings. Desertion is an important part of the story of these conflicts. I have attempted in this dissertation to study desertion in order to make sense of the nature of armed groups and hence of civil wars. The wars the world faces often seem as though they are unwieldy concatenations of multiple, shifting agendas, the intentions of the men and women who fight them unclear. Our attempts to make sense of them in terms of master narratives and grand causes can appear naively reductionist in the face of such complexity. I argue that we must recognize that there is no single reality of civil wars, neither the reduction to simple stories nor a complexity so great that only control can really address it. Instead, we can make better sense of armed groups by studying the common aims of their combatants, and in so doing better understand the organizational challenges that armed groups face and the tools at their disposal to address them.

In moving toward peace and encouraging political order, it is unclear whether former foes can really trust each other; if instead there really is a need for a central authority to hold all others in awe; or if the latter is just a ticket to tyranny. I do not claim a clear answer to this question. But I do suggest that control, if it is selective and judicious in its application, can help to keep combatants together even when they disagree. International actors have a difficult task in trying to encourage selectivity over time, even if they can help to provide control. The political aims of warring parties may never really come to be truly common. But they surely share a common interest in political order. I argue that resolving armed conflicts entails the challenge of

353 pulling off a difficult trick. It is to centralize force while avoiding its arbitrary use, to punish those who would break political order while reassuring others of their security.

354

Bibliography