Studien zur Neuen Politischen Ökonomie Herausgegeben von T. Bräuninger, Mannheim, Deutschland G. Schneider, Konstanz, Deutschland
Eva Bernauer Identities in Civil Conflict How Ethnicity, Religion and Ideology Jointly Affect Rebellion
Eva Bernauer Mannheim, Germany Dissertation zur Erlangung des Doktorgrades der Sozialwissenschaften an der Universität Mannheim unter dem Titel Civil Conflict in Three Dimensions - Ethnic, Religious and Class-Based Support of Rebel Groups in the National and International Domain Tag der mündlichen Prüfung: 02.10.2015 Erstgutachter: Professor Dr. Thomas Bräuninger Zweitgutachter: Professor Sabine Carey, Ph.D. Drittgutachter: Professor Nicolay Marinov, Ph.D. Studien zur Neuen Politischen Ökonomie ISBN 978-3-658-14151-6 ISBN 978-3-658-14152-3 (ebook) DOI 10.1007/978-3-658-14152-3 Library of Congress Control Number: 2016939109 Springer VS Springer Fachmedien Wiesbaden 2016 This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, express or implied, with respect to the material contained herein or for any errors or omissions that may have been made. Printed on acid-free paper This Springer VS imprint is published by Springer Nature The registered company is Springer Fachmedien Wiesbaden GmbH
Acknowledgments Having worked for more than four years on this dissertation while studying at the Graduate School of Economic and Social Sciences at Mannheim University, I would love to thank many people. First and foremost, I want to thank my supervisor Prof. Dr. Thomas Bräuninger, who combines excellent academic and analytic skills with a great personality, and who let me pursue my own research interests. I am extremely grateful for the time and patience that he afforded, especially at the beginning of this dissertation project, when there was most need for it. As my original interest was primarily in civil conflicts and quantitative methods, I very greatly thank Thomas for broadening my horizon by insisting on thorough at best formal theorizing. Last but not least, I thank Thomas for assisting me in gaining financial support throughout the entire research period. I am thankful to Prof. Dr. Gerald Schneider who supported me in applying for the Graduate School of Economic and Social Sciences in Mannheim in the first place. I also thank Prof. Sabine Carey, Ph.D., who shares my interest in civil conflicts, for serving as a secondary reviewer, and Prof. Nikolay Marinov, Ph.D., for serving as a tertiary reviewer. I would love to thank all my colleagues and friends from the CDSS and from the Chair of Political Economy at Mannheim University for their great companionship. It was good to have so many fellows struggling with the same challenges when writing a dissertation, being able to discuss technical issues at workshops or in D7, and at the same time spending a great nonacademic time over lunch or at night. I especially thank Susanne Michalik and Nathalie Giger for a joyful time in Mannheim and very good friendship. Most of all, I thank my husband Andreas for providing me with moral and practical support. In difficult times, Andreas placed his trust in me, while at the same time holding my ambitions in check. Andreas also had helpful tips regarding programming, the R Language and Latex, for which I am grateful. Our three wonderful children were born during the time of writing this dissertation. It was great that Andreas went on parental leave for one year so that I was able to finish this book. Finally, I am deeply indebted to the Graduate School of Economic and Social Sciences (GESS) in Mannheim and the Mannheim Centre for European Social Research (MZES) for their generous financial support. Without their support, this research would not have been possible. 5
Contents List of Tables 9 List of Figures 11 1. Introduction 17 I. National Political Exclusion and Support of Rebel Groups 23 2. Ethnic, Religious and Class-Based Civil Conflicts 25 2.1. The Literature......................... 25 2.2. Patterns of Favoritism and Conflict.............. 30 3. Favoritism and Conflict along Identity Lines 41 3.1. The Argument......................... 41 3.2. Hypotheses........................... 49 3.3. A Multidimensional Perspective................ 52 4. Testing the Argument on Exclusion and Conflict 63 4.1. Definitions, Data and Variables................ 63 4.2. Coverage and Methods..................... 76 4.3. Empirical Results........................ 81 5. A Model of Violent Political Competition 109 5.1. Model Assumptions and The Game.............. 109 5.2. Case Studies.......................... 123 5.3. Empirical Results........................ 132 II. Transnational Support of Rebel Groups 145 6. Ethnic, Religious and Class-Based Rebel Support 147 7
6.1. The Literature......................... 148 6.2. Episodes of Transnational Support.............. 152 7. Transnational Support Along Identity Lines 157 7.1. The Argument......................... 157 7.2. Hypotheses........................... 164 8. Testing the Argument on Links and Conflict 165 8.1. Data and Variables....................... 165 8.2. Empirical Results........................ 166 9. Conclusion 189 9.1. The Arab Spring How Does It Relate?.......... 189 9.2. Exclusion and Support in Three Dimensions......... 204 A. All Models with Control Variables Displayed 209 8
List of Tables 3.1. How the Identity Setup Changes with the Underlying Dimensionality........................... 53 3.2. How the Quantities of Interest Change with the Underlying Dimensionality......................... 57 4.1. Key Independent Variables on Exclusion........... 70 4.2. Correlation Table for the Excluded Population....... 72 4.3. Correlation Table for the Largest Excluded (Sub-)Group.. 72 4.4. Logit Estimations: Effect of Excluded Population Without Interaction on Conflict Onset................. 83 4.5. Logit Estimations: Effect of Largest Excluded Subgroup Without Interaction on Conflict Onset............ 84 4.6. Logit Estimations: Effect of Excluded Population With Interaction on Conflict Onset.................. 85 4.7. Logit Estimations: Effect of Largest Excluded Subgroup With Interaction on Conflict Onset.............. 86 4.8. Robustness Checks: Autocracy................ 99 4.9. Robustness Checks: Onset2, Onset3............. 101 4.10. Robustness Checks: Onset2, Onset3, Autocracy....... 103 4.11. Influential Obs., Onset1.................... 105 4.12. Influential Obs., Onset2.................... 106 5.1. Morocco s Identity Setup (in %): Left and Right...... 125 5.2. Morocco s Identity Setup (in %): Adding a Centrist Category125 5.3. Morocco s Opposition to a White Government........ 127 5.4. Morocco s Opposition to a White, Leftist, Rightist Government127 5.5. Morocco s Opposition to a White, Centrist, Rightist Government............................... 127 5.6. Guatemala s Identity Setup (in %).............. 129 5.7. Thailand s Identity Setup (in %)............... 131 5.8. Logit Estimations: Effect of Formally Derived Variables Without Interaction on Conflict Onset............ 133 9
5.9. Logit Estimations: Effect of Formally Derived Variables With Interaction on Conflict Onset................. 134 5.10. Robustness Checks: Formal Part............... 139 8.1. Transnational Links Without Interaction........... 168 8.2. Logit Estimations: Effect of Transnational Links With Interaction on Conflict Onset.................... 170 A.1. Logit Estimations: Effect of Excluded Population Without Interaction on Conflict Onset................. 210 A.2. Logit Estimations: Effect of Largest Excluded Subgroup Without Interaction on Conflict Onset............ 211 A.3. Logit Estimations: Effect of Excluded Population With Interaction on Conflict Onset.................. 212 A.4. Logit Estimations: Effect of Largest Excluded Subgroup With Interaction on Conflict Onset.................. 213 A.4. [continued]........................... 214 A.5. Robustness Checks: Autocracy................ 215 A.6. Robustness Checks: Onset2, Onset3............. 216 A.7. Robustness Checks: Onset2, Onset3, Autocracy....... 217 A.8. Influential Obs., Onset1.................... 218 A.9. Influential Obs., Onset2.................... 219 A.10.Logit Estimations: Effect of Formally Derived Variables Without Interaction on Conflict Onset............ 220 A.11.Logit Estimations: Effect of Formally Derived Variables With Interaction on Conflict Onset................. 221 A.12.Transnational Links Without Interaction........... 222 A.13.Logit Estimations: Effect of Transnational Links With Interaction on Conflict Onset........................ 223 A.13.[continued]........................... 224 10
List of Figures 3.1. Schematic representation of the argument........... 42 4.1. Box plots on the size of the largest excluded (sub-)group... 71 4.2. Fitted heat map line for Model 1 in Table 4.5......... 87 4.3. Fitted heat map line for Model 1 in Table 4.7......... 88 4.4. First difference analyses of the effect of exclusion in noncompetitive regimes....................... 89 4.5. Predicted probability of armed conflict onset and 90% confidence interval for varying values of the variable ERI max, separately for non-competitive and competitive regimes.... 91 4.6. Predicted probability of armed conflict onset and 90% confidence interval for varying values of the variable EI max, separately for non-competitive and competitive regimes.... 92 4.7. Predicted probability of armed conflict onset and 90% confidence interval for varying values of the variable W all, separately for non-competitive and competitive regimes...... 93 4.8. First difference analyses of the effect of exclusion in competitive regimes........................... 94 4.9. First difference analyses of the effect of exclusion, when comparing non-competitive and competitive regimes..... 95 4.10. Receiver Operating Characteristic (ROC) curve, separately for a model that includes ERI max and its interaction with regime type, and for a zero model without these variables.. 96 4.11. In-sample predictive power for a model that includes ERI max and its interaction with regime type.............. 97 4.12. Robustness check: First difference analyses of the effect of exclusion in autocratic regimes................. 100 4.13. Robustness check: First difference analyses of the effect of exclusion in non-competitive regimes, when onset2 or onset3 is used as a dependent variable................. 102 4.14. Robustness check: First difference analyses of the effect of exclusion in autocratic regimes, when onset2 or onset3 is used as a dependent variable.................. 104 11
4.15. Robustness check: First difference analyses of the effect of exclusion in non-competitive regimes, when influential observations are excluded....................... 107 5.1. Fitted heat map line for Model 1 in Table 5.8......... 135 5.2. Fitted heat map line for Model 1 in Table 5.9......... 136 5.3. First difference analyses of the effect of exclusion in noncompetitive regimes for the formally derived exclusion variables.137 5.4. Predicted probability of armed conflict onset and 90% confidence interval for varying values of the variable ERI, separately for non-competitive and competitive regimes.... 138 5.5. Receiver Operating Characteristic (ROC) curve, separately for a model that includes ERI and its interaction with regime type, and for a zero model without these variables.. 140 5.6. In-sample predictive power for a model that includes ERI and its interaction with regime type.............. 141 5.7. Robustness check: First difference analyses of the effect of exclusion for the formally derived exclusion variable ERI for different scenarios...................... 142 8.1. Predicted probability of armed conflict onset and 90% confidence interval for varying values of ethnic/religious links, when exclusion is held constant and non-democratic regimes are considered.......................... 171 8.2. Predicted probability of armed conflict onset and 90% confidence interval for varying values of ethnic/religious links, when exclusion is held constant and democratic regimes are considered............................ 172 8.3. Predicted probability of armed conflict onset and 90% confidence interval for varying values of ideological links, when exclusion is held constant and non-democratic regimes are considered............................ 173 8.4. Predicted probability of armed conflict onset and 90% confidence interval for varying values of ideological links, when exclusion is held constant and democratic regimes are considered.............................. 174 8.5. First difference analyses of the effect of ethnic links in nondemocratic regimes for varying values of exclusion (ERI all ). 175 12
8.6. First difference analyses of the effect of religious links in nondemocratic regimes for varying values of exclusion (ERI all ). 176 8.7. First difference analyses of the effect of ethnic/religious links in non-democratic regimes for varying values of exclusion (ERI all )............................. 177 8.8. First difference analyses of the effect of ideological links in non-democratic regimes for varying values of exclusion (ERI all )............................. 178 8.9. First difference analyses of the effect of ethnic/religious links in democratic regimes for varying values of exclusion (ERI all ).179 8.10. First difference analyses of the effect of ideological links in democratic regimes for varying values of exclusion (ERI all ). 180 8.11. First difference analyses of the effect of exclusion in nondemocratic regimes for varying values of ethnic/religious links.181 8.12. First difference analyses of the effect of exclusion in nondemocratic regimes for varying values of ideological links.. 182 8.13. First difference analyses of the effect of exclusion in democratic regimes for varying values of ethnic/religious links.. 183 8.14. First difference analyses of the effect of exclusion in democratic regimes for varying values of ideological links..... 184 8.15. Receiver Operating Characteristic (ROC) curve, separately for a model that includes ethnic/religious links and the respective interaction terms, and for a zero model without these variables............................. 185 8.16. In-sample predictive power for a model that includes ethnic/religious links and their interaction with regime type and exclusion............................. 186 8.17. In-sample predictive power for a model that includes ideological links and their interaction with regime type and exclusion.187 9.1. Map of the MENA region.................... 190 13
Summary This book is interested in the prediction of civil conflicts based upon the political exclusion of identity groups and their transnational links to external governments. The innovation lies in a simultaneous consideration of three identities ethnicity, religion and class-based ideology according to which groups may differ, be excluded and have links to external governments. Previous research has almost exclusively focused on ethnic exclusion or ethnic transnational links when predicting civil conflicts. However, empirical case studies provide ample support that civil conflicts are not only started by ethnic groups, but also by religious and class-based ideological groups, while external support patterns also run along all three lines of identity. That said, I spell out in detail what it means if all three identity lines are taken into account. Most importantly, such a focus implies a shift to a society s unique three-dimensional identity setup, based upon which the excluded population and their transnational links can be determined. The goal is to assess whether a three-dimensional perspective on exclusion and transnational links is superior to a purely one- or two-dimensional perspective when predicting civil conflicts. The book comprises two parts: the first part focuses exclusively on national identity constellations and exclusion patterns, while the second part additionally takes transnational links to external governments into account. In the first part, based upon survey data, I assemble an original dataset on the three-dimensional identity setup of each society for 57 countries and with given data on the government s ethnic, religious and class-based ideological identity I subsequently determine two key independent variables: the size of the excluded population and the size of the largest excluded three-dimensional subgroup. I analyze the effect of these variables on the onset of civil conflict in quantitative analyses, as well as assessing the performance of a three-dimensional perspective vis-à-vis a one- or two-dimensional one. Recognizing the a-theoretical nature of both key independent variables, I subsequently develop a game-theoretic model that determines the most likely rebel group and its size based upon expected utility calculations by all individuals in the three-dimensional identity setup. The resulting variables are again tested in quantitative analyses. Ultimately, the first part shows that the shift to three dimensions yields valuable insights. In the second part, I assemble additional data on the number of ethnic, religious and class-based ideological links that the excluded population has 14
to external governments. Following the argumentation of the first part, I develop precise theoretical expectations concerning the effect of each of these variables on the risk of civil conflict, including in interaction with other variables such as a country s regime type and the size of the excluded population. The quantitative analysis examines the merit of considering religious and class-based ideological links in addition to ethnic links and again assesses the superiority of a three-dimensional perspective vis-à-vis a one- or two-dimensional one. One key finding is that in non-democratic regimes the sum of ethnic and religious links significantly increases the risk of civil conflict onset for all levels of exclusion, while ideological links surprisingly have no conflict-enhancing effect. By contrast, in democratic regimes, ethnic and religious links only have a conflict-enhancing effect for low levels of exclusion, as expected, whereas ideological links are again insignificant. The analysis reveals the value of combining ethnic and religious linkages, as well as conditioning the effect of linkages on regime type, thereby confirming key theoretical expectations that have been formulated yet not proven in previous studies. Put briefly, this book predicts civil conflicts based upon the unique ethnic, religious and ideological demographic identity setup of a country, the government s identity, the resulting exclusion patterns and the transnational links that the excluded population has to external governments. With the current conflicts in the Middle East and North Africa, this book deals with a highly relevant and up-to-date topic. It additionally makes a crucial contribution to the research literature, which has increasingly focused on exclusion patterns and transnational links in recent years when predicting civil conflicts, albeit almost exclusively with a mere ethnic focus and thus neglecting religion and ideology. 1 1 Note that this book uses the expressions class-based ideology, ideology, and class interchangeably. 15