Beyond the Civil Democratic Peace: Subnational Political ... · Beyond the Civil Democratic Peace:...

321
Beyond the Civil Democratic Peace: Subnational Political Institutions and Internal Armed Conflict Tore Wig [email protected] Department of Political Science Faculty of Social Sciences University of Oslo Advisors: Professor H˚ avard Hegre Professor Carl Henrik Knutsen April 2015

Transcript of Beyond the Civil Democratic Peace: Subnational Political ... · Beyond the Civil Democratic Peace:...

  • Beyond the Civil Democratic Peace:

    Subnational Political Institutions

    and Internal Armed Conflict

    Tore [email protected]

    Department of Political Science

    Faculty of Social Sciences

    University of Oslo

    Advisors:

    Professor H̊avard Hegre

    Professor Carl Henrik Knutsen

    April 2015

  • Contents

    List of Figures 9

    List of Tables 11

    Acknowledgements 3

    1 Institutions and Civil Peace in Weak States 7

    1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

    1.1.1 Crucial definitions . . . . . . . . . . . . . . . . . . . . . . . . . . 10

    1.2 The institutional causes of conflict: The case for a subnational approach 11

    1.2.1 Subnational institutions in weak states: Beyond methodological

    statism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

    1.2.2 Customary institutions: Ashanti in Ghana . . . . . . . . . . . . . 19

    1.2.3 Formal local government institutions: The Boko Haram insurgency

    in Nigeria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

    1.3 Conceptual framework: Institutional attributes, ethnic groups and conflict 23

    1.3.1 Institutions as rules . . . . . . . . . . . . . . . . . . . . . . . . . . 23

    1.3.2 Institutional attributes . . . . . . . . . . . . . . . . . . . . . . . . 24

    1.3.3 Ethnic groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

    1.3.4 Internal armed conflict . . . . . . . . . . . . . . . . . . . . . . . . 27

    1.4 Theoretical framework: The conflict ladder and the role of institutions . 27

    1.4.1 The individual level: Rationality and motivations . . . . . . . . . 28

    1.4.2 The organizational level: Groups as unitary actors . . . . . . . . . 29

    1.4.3 The bargaining level: Obstacles to agreements short of war . . . . 30

    1.4.4 How institutions solve bargaining problems . . . . . . . . . . . . . 32

    1.5 Measuring institutions and conflict . . . . . . . . . . . . . . . . . . . . . 34

    1.5.1 Data on subnational institutions . . . . . . . . . . . . . . . . . . . 34

    1.5.2 Varieties of internal conflict data . . . . . . . . . . . . . . . . . . 36

  • 1.6 Reaching conclusions: Philosophy and methodology . . . . . . . . . . . . 38

    1.6.1 Causal explanations . . . . . . . . . . . . . . . . . . . . . . . . . . 38

    1.6.2 Methodological framework . . . . . . . . . . . . . . . . . . . . . . 40

    1.7 Structure of the thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

    1.7.1 Part I: National institutions . . . . . . . . . . . . . . . . . . . . . 45

    1.7.2 Part II: Customary institutions . . . . . . . . . . . . . . . . . . . 46

    1.7.3 Part III: Formal local government institutions . . . . . . . . . . . 47

    1.8 Implications for research and policy . . . . . . . . . . . . . . . . . . . . . 48

    I National Institutions 51

    2 Do Liberal Institutions Pacify? A Predictive Validation Analysis of

    Institutional Arguments for Civil Peace 53

    2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

    2.2 Liberal institutional explanations for civil conflict: Democracy, quality of

    government and the rule of law . . . . . . . . . . . . . . . . . . . . . . . 56

    2.3 Predictive evaluation: Why and how . . . . . . . . . . . . . . . . . . . . 59

    2.3.1 Metrics for evaluating predictive accuracy . . . . . . . . . . . . . 60

    2.3.2 How to evaluate out-of-sample performance . . . . . . . . . . . . 63

    2.4 Research design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

    2.4.1 Measuring institutions . . . . . . . . . . . . . . . . . . . . . . . . 65

    2.4.2 Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

    2.5 Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

    2.5.1 In-sample validation . . . . . . . . . . . . . . . . . . . . . . . . . 69

    2.5.2 Out-of-sample validation . . . . . . . . . . . . . . . . . . . . . . . 72

    2.5.3 Where does the liberal model miss? . . . . . . . . . . . . . . . . . 75

    2.5.4 Alternative explanations . . . . . . . . . . . . . . . . . . . . . . . 80

    2.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83

    II The Role of Subnational Customary Institutions 85

    3 Peace from the Past: Pre-Colonial Political Institutions and Civil Wars

    in Africa 87

    3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88

    3.2 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89

    3.2.1 Classifying pre-colonial institutions: The centralized-decentralized

    dimension . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90

    3.2.2 The modern impact of pre-colonial institutions . . . . . . . . . . . 91

    3.3 Pre-colonial institutions and civil war . . . . . . . . . . . . . . . . . . . . 92

    3.4 Research design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94

    3.4.1 Unit of analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96

    4

  • 3.4.2 Dependent variable: Ethnic armed conflict . . . . . . . . . . . . . 96

    3.4.3 Independent variable: Pre-colonial centralization . . . . . . . . . . 96

    3.4.4 Controls . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97

    3.5 Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98

    3.5.1 Statistical models . . . . . . . . . . . . . . . . . . . . . . . . . . . 99

    3.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106

    4 Which Groups Fight? Customary Institutions and Communal Conflicts

    in Africa 109

    4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110

    4.2 Current explanations of communal conflict . . . . . . . . . . . . . . . . . 112

    4.3 Customary institutions and ethnic groups . . . . . . . . . . . . . . . . . . 113

    4.4 Institutional explanations for war applied to traditional institutions . . . 115

    4.5 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120

    4.5.1 Unit of analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120

    4.5.2 Communal conflict . . . . . . . . . . . . . . . . . . . . . . . . . . 121

    4.5.3 Customary institutions . . . . . . . . . . . . . . . . . . . . . . . . 121

    4.5.4 Controls . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124

    4.6 Estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126

    4.6.1 Controlling for alternative pathways . . . . . . . . . . . . . . . . . 129

    4.6.2 Predictive power . . . . . . . . . . . . . . . . . . . . . . . . . . . 133

    4.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134

    III The Role of Formal Local Government Institutions 137

    5 Subnational Institutional Quality and Local Conflict Violence in Africa139

    5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139

    5.2 Institutional quality and conflict: State of the art . . . . . . . . . . . . . 141

    5.3 Why local institutional quality pacifies . . . . . . . . . . . . . . . . . . . 143

    5.4 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146

    5.4.1 Local institutional quality . . . . . . . . . . . . . . . . . . . . . . 147

    5.4.2 Conflict-related violence events . . . . . . . . . . . . . . . . . . . 149

    5.4.3 Controls . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151

    5.5 Baseline results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153

    5.5.1 Endogeneity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156

    5.5.2 Alternative model specification . . . . . . . . . . . . . . . . . . . 157

    5.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160

    6 The Institutional Legacies of Local Conflict Violence: Perception-Based

    Evidence from the Afrobarometer Surveys 163

    6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164

    6.2 The institutional effects of local conflict: State of the art . . . . . . . . . 165

    5

  • 6.2.1 Institutional persistence . . . . . . . . . . . . . . . . . . . . . . . 166

    6.2.2 Conflict-induced decay . . . . . . . . . . . . . . . . . . . . . . . . 167

    6.2.3 Conflict as a catalyst for institutional development . . . . . . . . 169

    6.3 Research design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170

    6.3.1 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170

    6.3.2 Perceptions of institutional quality . . . . . . . . . . . . . . . . . 171

    6.3.3 Conflict exposure . . . . . . . . . . . . . . . . . . . . . . . . . . . 173

    6.3.4 Empirical strategy . . . . . . . . . . . . . . . . . . . . . . . . . . 174

    6.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177

    6.4.1 Corruption . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178

    6.4.2 Trust in local institutions . . . . . . . . . . . . . . . . . . . . . . 181

    6.4.3 Threshold and cumulative effects . . . . . . . . . . . . . . . . . . 184

    6.4.4 Robustness and conditional effects . . . . . . . . . . . . . . . . . . 188

    6.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189

    IV Appendices and Bibliography 191

    7 Appendix to Chapter 2 193

    7.1 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193

    7.2 Descriptive statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193

    7.3 Additional controls . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193

    7.4 Analyses when democracy is endogenized . . . . . . . . . . . . . . . . . . 195

    7.5 Comparing predictions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197

    8 Appendix to Chapter 3 203

    8.1 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203

    8.2 The dataset . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203

    8.2.1 Matching the Ethnographic Atlas to EPR (the EA2EPR dataset) 203

    8.2.2 Validity and reliability of the Ethnographic Atlas data . . . . . . 206

    8.3 List of variables: Sources and operationalizations . . . . . . . . . . . . . 206

    8.3.1 Afrobarometer variables (used in validity check in Appendix) . . . 209

    8.4 Summary statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210

    8.5 Predictive power . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211

    8.6 Controlling for additional group traits . . . . . . . . . . . . . . . . . . . . 214

    8.7 Multiple imputation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216

    8.8 Removing influential observations . . . . . . . . . . . . . . . . . . . . . . 217

    8.9 Adressing endogeneity . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218

    8.10 Alternative unit of analysis . . . . . . . . . . . . . . . . . . . . . . . . . . 221

    8.11 Controlling for relative size and MEG power-status . . . . . . . . . . . . 223

    8.12 Random intercept instead of country dummies . . . . . . . . . . . . . . . 225

    6

  • 9 Appendix to Chapter 4 227

    9.1 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227

    9.2 Matching the UCDP communal conflict data to EPR . . . . . . . . . . . 227

    9.3 Analysis of FCI branches . . . . . . . . . . . . . . . . . . . . . . . . . . . 230

    9.4 Count models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230

    9.5 Descriptives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232

    9.6 List of regions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233

    9.7 List of matched EPR groups with communal conflict . . . . . . . . . . . 233

    10 Appendix to Chapter 5 237

    10.1 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237

    10.2 Creating the dataset . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237

    10.2.1 Descriptions of main variables . . . . . . . . . . . . . . . . . . . . 239

    10.2.2 Controls . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 240

    10.2.3 Factor analysis investigating local institutional quality dimension 241

    10.3 Descriptive statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243

    10.4 Matching diagnostics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245

    10.5 Sensitivity tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247

    10.5.1 Alternative functional form . . . . . . . . . . . . . . . . . . . . . 247

    10.5.2 Removing extreme cases . . . . . . . . . . . . . . . . . . . . . . . 249

    10.5.3 Removing low-respondent cases . . . . . . . . . . . . . . . . . . . 251

    10.5.4 Additional control variables . . . . . . . . . . . . . . . . . . . . . 253

    10.5.5 Parsimonious models . . . . . . . . . . . . . . . . . . . . . . . . . 255

    10.5.6 Additional hurdle model for round 4 . . . . . . . . . . . . . . . . 257

    10.5.7 Modeling survey risk . . . . . . . . . . . . . . . . . . . . . . . . . 259

    10.6 Maps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261

    11 Appendix to Chapter 6 265

    11.1 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 265

    11.2 Variable descriptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 266

    11.2.1 Dependent variables from Afrobarometer . . . . . . . . . . . . . . 266

    11.3 Descriptives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267

    11.4 Excluding random intercepts . . . . . . . . . . . . . . . . . . . . . . . . . 268

    11.5 Lag length . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271

    11.6 No post-treatment variables . . . . . . . . . . . . . . . . . . . . . . . . . 274

    11.7 Casualties instead of event counts . . . . . . . . . . . . . . . . . . . . . . 277

    11.8 Cumulative conflict . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 280

    11.9 Conditional effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 281

    11.10Region dummies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283

    11.11Countries included . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283

    12 Bibliography 285

    7

  • List of Figures

    1.1 Murdock’s ethnographic map of pre-colonial polities in Africa . . . . . . 17

    1.2 Ghana (red border) and contemporary Ashantiland (green border) . . . . 20

    1.3 GED conflict events (1989-2010) overlaid on country borders . . . . . . . 39

    2.1 In-sample ROC and Precision Recall curves for a standard civil-war onset

    model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

    2.2 In-sample AUROC and AUPR . . . . . . . . . . . . . . . . . . . . . . . . 71

    2.3 Out-of-sample AUROC and AUPR (cross-validation runs=1000) . . . . . 73

    2.4 Out-of-sample AUROC and AUPR (cross-validation runs=1000) . . . . . 74

    2.5 Out-of-sample AUROC and AUPR with different regions as test sets . . 76

    2.6 Out-of-sample AUROC and AUPR with young regimes as test sets . . . 79

    2.7 Out-of-sample AUROC and AUPR with large countries as test sets . . . 80

    2.8 Out-of-sample AUROC and AUPR comparisons to baseline with income

    removed (cross-validation runs=1000) . . . . . . . . . . . . . . . . . . . . 82

    3.1 Murdock‘s map of pre-colonial ethnic institutions . . . . . . . . . . . . . 95

    3.2 Predicted probabilities of ethnic armed conflict onset when Pre-colonial

    CentralizationMEG increases . . . . . . . . . . . . . . . . . . . . . . . . . 101

    4.1 Descriptives for FCI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123

    4.2 Simulated probability of dyad-level communal conflict for different levels

    of FCI, 1989-2013 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128

    4.3 Distribution of coefficient estimates from Bayesian model averaging . . . 132

    4.4 In-sample predictive power for models with and without institutional vari-

    ables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133

    5.1 Map showing the distribution of Local Institutional Quality (district level)

    in Afrobarometer rounds 3 and 4 . . . . . . . . . . . . . . . . . . . . . . 148

  • 5.2 Institutional quality at national level (World Bank) and national means of

    Local Institutional Quality . . . . . . . . . . . . . . . . . . . . . . . . . . 150

    5.3 Map showing the distribution of Local Institutional Quality (administrative-

    district level) in Uganda and Nigeria, round 3, and GED conflict events . 151

    5.4 Map showing the continental distribution of Local Institutional Quality

    (district level), round 3, and GED conflict events . . . . . . . . . . . . . 152

    5.5 Expected count of GED-events in a district in the post-survey period . . 156

    6.1 All survey clusters (2005,2008 and 2012) and all conflict events (1989-2010) 172

    6.2 Looking for threshold effects: Experienced and perceived corruption . . . 185

    6.3 Looking for threshold effects: Trust in local institutions . . . . . . . . . . 186

    7.1 Out-of-sample AUROC and AUPR comparisons to baseline with added

    control variables (cross-validation runs=1000) . . . . . . . . . . . . . . . 194

    7.2 Out-of-sample AUROC and AUPR comparisons to baseline when Democ-

    racy is endogenized using WAVE (cross-validation runs=1000) . . . . . . 195

    8.1 In-sample ROC plots showing improvements in predictive accuracy . . . 212

    8.2 Distributions of out-of-sample AUC values from 100 cross-validation runs

    without and with Pre-colonial CentralizationMEG . . . . . . . . . . . . . 213

    10.1 Scatterplots of Local Institutional Quality . . . . . . . . . . . . . . . . . 239

    10.2 Distribution of Local Institutional Quality (Afrobarometer rounds 3 and 4) 245

    10.3 Administrative districts, round 3 . . . . . . . . . . . . . . . . . . . . . . 262

    10.4 Administrative districts, round 4 . . . . . . . . . . . . . . . . . . . . . . 263

    10

  • List of Tables

    1.1 Survey items capturing perceptions of local institutional quality from the

    Afrobarometer survey (round 4) . . . . . . . . . . . . . . . . . . . . . . . 18

    1.2 Communal conflicts by world region, 1989-2013 . . . . . . . . . . . . . . 37

    2.1 Concepts, variables and factor loadings on common dimension . . . . . . 67

    2.2 Logit models of conflict onset . . . . . . . . . . . . . . . . . . . . . . . . 68

    3.1 Bivariate correlations between pre-colonial centralization and various group-

    level traits (group is the unit of analysis, N=243) . . . . . . . . . . . . . 98

    3.2 Logit Models of the probability of ethnic armed conflict onset in EGIP-

    MEG dyads . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100

    3.3 Cross-sectional logit and OLS models of ethnic armed conflict onsets (bi-

    nary and count) in EGIP-MEG dyads . . . . . . . . . . . . . . . . . . . . 103

    3.4 Logit models of the probability of ethnic armed conflict onset in EGIP-

    MEG dyads . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105

    4.1 Correlations between customary institutions . . . . . . . . . . . . . . . . 122

    4.2 Descriptive statistics main independent variables . . . . . . . . . . . . . . 125

    4.3 Logit model of communal conflict . . . . . . . . . . . . . . . . . . . . . . 127

    4.4 Linear models of log(communal conflicts+.001) . . . . . . . . . . . . . . . 128

    5.1 Negative binomial count models regressing GED events on the quality of

    local institutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155

    5.2 Negative binomial count models of GED events when observations are

    matched on conflict history . . . . . . . . . . . . . . . . . . . . . . . . . . 158

    5.3 Hurdle model (rounds 3 and 4 combined) . . . . . . . . . . . . . . . . . . 159

    6.1 Overview of hypotheses linking conflict to institutional quality, and related

    mechanisms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170

  • 6.2 Items used to tap perceptions of institutional quality . . . . . . . . . . . 173

    6.3 Sample means of outcome variables in conflict vs. non-conflict areas . . . 178

    6.4 Random-intercept GLS models of corruption experiences . . . . . . . . . 179

    6.5 Random-intercept GLS models of corruption perceptions . . . . . . . . . 180

    6.6 Random-intercept GLS models of trust in police and politicians . . . . . 182

    6.7 Summary: Estimated relationships for items capturing perceptions of in-

    stitutional quality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183

    6.8 Random-intercept GLS models investigating effect of long-term conflict on

    institutional quality items . . . . . . . . . . . . . . . . . . . . . . . . . . 187

    7.1 Fist stage OLS regression where Democracy is regressed on WAVE and

    baseline covariates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196

    7.2 Comparing predictions (for onset cases) from baseline and liberal institu-

    tions models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197

    7.3 Continued: Comparing predictions (for onset cases) from baseline and lib-

    eral institutions models . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198

    7.4 Continued: Comparing predictions (for onset cases) from baseline and lib-

    eral institutions models . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199

    7.5 Continued: Comparing predictions (for onset cases) from baseline and lib-

    eral institutions models . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200

    7.6 Continued: Comparing predictions (for onset cases) from baseline and lib-

    eral institutions models . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201

    7.7 Continued: Comparing predictions (for onset cases) from baseline and lib-

    eral institutions models . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202

    8.1 Matching EPR to Ethnographic Atlas: Matched cases by matching criteria 205

    8.2 OLS estimates of associations between Pre-colonial Centralization of re-

    spondents ethnic group and affirmation of authority of traditional rulers . 207

    8.3 Summary stats of main variables used in the paper and in the appendix . 210

    8.4 Logit models of ethnic armed conflict onset, controlling for additional group

    traits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215

    8.5 Comparing models run on imputed and unimputed data . . . . . . . . . 216

    8.6 Models with influential observations removed . . . . . . . . . . . . . . . . 217

    8.7 First-stage reduced form estimates: Pre-colonial CentralizationMEG re-

    gressed on Ecological DiversityMEG and covariates . . . . . . . . . . . . . 219

    8.8 IV probit models where Pre-colonial CentralizationMEG is instrumented

    by Ecological DiversityMEG . . . . . . . . . . . . . . . . . . . . . . . . . 220

    8.9 Logit models of ethnic armed conflict onset with group-year as unit of

    analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222

    8.10 Logit models of ethnic armed conflict onset . . . . . . . . . . . . . . . . . 224

    8.11 Random intercept logit models of ethnic armed conflict onset with dyad-

    year as unit of analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225

    12

  • 9.1 Snippet of dataset matching communal conflicts to EPR groups . . . . . 229

    9.2 Logit models of communal conflict using FCI-branches instead of FCI . . 230

    9.3 Poisson and negative binomial count models of communal conflict onsets,

    1989-2012 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231

    9.4 Regions and countries . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233

    9.5 List of EPR groups involved in communal conflict . . . . . . . . . . . . . 234

    9.6 Continued: List of EPR groups involved in communal conflict . . . . . . 235

    10.1 Factor loadings from a factor analysis with 4 factors . . . . . . . . . . . . 242

    10.2 Descriptive statistics (Afrobarometer round 3) . . . . . . . . . . . . . . . 243

    10.3 Descriptive statistics (Afrobarometer round 4) . . . . . . . . . . . . . . . 244

    10.4 Covariate balance (pre- and post matching), when matching on conflict

    history (Afrobarometer round 3) . . . . . . . . . . . . . . . . . . . . . . . 246

    10.5 Covariate balance (pre- and post matching), when matching on conflict

    history (Afrobarometer round 4) . . . . . . . . . . . . . . . . . . . . . . . 246

    10.6 Table 1 above replicated with Poisson models . . . . . . . . . . . . . . . 248

    10.7 Dropping the most intense conflict areas and re-estimating the core models 250

    10.8 Dropping low-respondent districts . . . . . . . . . . . . . . . . . . . . . . 252

    10.9 Additional control variables . . . . . . . . . . . . . . . . . . . . . . . . . 254

    10.10Parsimonious models, round 3 . . . . . . . . . . . . . . . . . . . . . . . . 255

    10.11Parsimonious models, round 4 . . . . . . . . . . . . . . . . . . . . . . . . 256

    10.12Hurdle model, round 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 258

    10.13Logit models of survey-location (unit-of-analysis=PRIO-GRID cells) . . . 259

    10.14Core models, with survey-risk as a covariate . . . . . . . . . . . . . . . . 261

    11.1 GLS and 2SLS models of corruption experience without random intercepts 268

    11.2 GLS and 2SLS models of corruption experience without random intercepts 269

    11.3 GLS and 2SLS models of corruption experience without random intercepts 270

    11.4 GLS and 2SLS models of corruption experience with different lag lengths 271

    11.5 GLS and 2SLSmodels of corruption perceptions with different lag lengths 272

    11.6 GLS and 2SLS models of trust in institutions with different lag lengths . 273

    11.7 GLS and 2SLS models of corruption experience w/post-treatment bias can-

    didates removed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 274

    11.8 GLS and 2SLS models of corruption perceptions w/post-treatment bias

    candidates removed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275

    11.9 GLS and 2SLS models of trust in institutions w/post-treatment bias can-

    didates removed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276

    11.10GLS and 2SLS models of experienced corruption, with casualties instead

    of event counts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 277

    11.11GLS and 2SLS models of corruption perceptions, with casualties instead

    of event counts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 278

    11.12GLS and 2SLS models of trust in institutions, with casualties instead of

    event counts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 279

    13

  • 11.13GLS models investigating effect of long-term conflict on institutional qual-

    ity items . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 280

    11.14Random-intercept GLS models investigating conditional effects: Interac-

    tions with homeland of excluded ethnic group . . . . . . . . . . . . . . . 281

    11.15Random-intercept GLS models investigating conditional effects: Interac-

    tions with log distance to capital . . . . . . . . . . . . . . . . . . . . . . 282

    11.16GLS models with region-fixed effects . . . . . . . . . . . . . . . . . . . . 283

    11.17Countries included in the surveys and number of respondents . . . . . . . 284

    14

  • Tore Wig 1

    1University of Oslo and Peace Research Institute Oslo, PRIO, email: [email protected]

  • Acknowledgements

    If this thesis has a bumper sticker, it must surely be “institutions matter”. When this

    document is printed I will have spent close to ten years of my life in the institution that

    is the University of Oslo. To say that this institution has mattered to me would be a

    huge understatement. I will therefore start by acknowledging the University of Oslo, and

    the Department of Political Science, for shaping me into a walking piece of evidence for

    beneficial institutional effects.

    A long list of people have contributed directly or indirectly to this work, and deserve

    to be mentioned.

    First and foremost, I would like to thank my thesis advisers, H̊avard and Carl Henrik,

    who have been the most important contributors to my development as a political scientist

    and to the completion of this thesis. H̊avard was the one who first trusted my abilities

    as a researcher, and hired me at PRIO. He was the first one to encourage me to do a

    PhD, and he has supported me ever since. Without him, I would never have gotten into

    conflict research, or gone the (informal) “PRIO school” that has been so crucial to my

    development as a researcher. Carl Henrik has been my indispensable go-to guy at UiO.

    If it had not been for him, I would never have learned how to get stuff published, or

    thrived as well as I have at the department. You have both provided me with excellent

    guidance, and – perhaps more importantly – served as role models and facilitators. You

    have introduced me to the people, high standards, methods, and mindsets of the world

    class political science community you are both important contributors to. You have both

    been great supervisors, friends, and collaborators. I owe you a great debt!

    Two institutions have served as my academic homes. My primary home has been the

    Department of Political Science at UiO, and my second has been the Peace Research Insti-

    tute Oslo, PRIO. From the UiO, I would like to mention the long list of senior colleagues

    who have provided me with inspiration and guidance throughout these years; Bjørn Høy-

    land, Jostein Askim, Raino Malnes, Elin Allern, Øyvind Østerud, Øivind Bratberg, Jon

    Hovi, Dag Einar Thorsen, Anders Jupsk̊as, Arild Underdal, H̊avard Strand, Helge Holter-

    mann, Kim Angell, and Robert Huseby. Credit also goes to the excellent group of fellow

  • Phd-students and friends that I have studied with. You have all contributed to a great

    working environment.

    At PRIO, I would like to thank all the brilliant researchers that have been such an

    inspiration to me, and without which I would never have learned how to do conflict

    research properly: Gudrun Østby, Ragnhild Nord̊as, Helga Malmin Binningsbø, Halvard

    Buhaug, Scott Gates, Siri Aas Rustad, Øysten Rolandsen, Marianne Dahl, Jeff Checkel,

    H̊avard Mokleiv Nyg̊ard, Nils Petter Gleditsch, Jonas Nordkvelle, Kristian Gleditsch,

    Andreas Tollefsen, Ida Rudolfsen, Idunn Kristiansen, and Henrik Urdal. I would also like

    to thank Cathrine Bye, for making me feel at home at PRIO.

    In this project, and others, I have collaborated with a number of excellent scholars

    and colleagues. In this respect, I would like to thank Espen Geelmuyden Rød, Sirianne

    Dahlum, H̊avard Nyg̊ard, Carl Henrik, Marianne Dahl, Daniela Kromrey, Andreas Forø

    Tollefsen, Øyvind Stiansen, Andreas Kotsadam, Eivind Hammersmark Olsen, Pat Regan,

    John Gerring, Magnus Rasmussen, and Øyvind Skorge, for collaborating with me on

    various projects.

    Several people have been important to this project through providing comments at

    presentations and after reading through parts of the manuscript. I would especially like

    to thank the PRIOites mentioned above, all of whom have provided comments on various

    papers I have presented at PRIO brownbags. I would also like to thank Øyvind Østerud,

    Bjørn Høyland, H̊avard Strand, Olav Schram Stokke, Andreas Hvidsten, Silje Lyder

    Hermannsen, Vibeke Wøien Hansen, Tatjana Stankovic, Kacper Szulecki, Nils Weidmann,

    Hanne Fjelde, Nynke Salverda, Henning Finseraas, Cristina Bucur, Helge Holtermann,

    Sabine Otto, Espen Geelmuyden Rød, Magnus Rasmussen, Nina Von Uexkull, Kristian

    Gleditsch, Kristin Bakke, Svend Erik Skaaning, Agnes Cornell, Jørgen Møller, Merethe

    Bech Seeberg, Jose Antonio Cheibub, Fenja Søndergaard Møller, Jacob Tolstrup, and

    Lasse Lykke Rørbæk for commenting on one or more of the different articles comprising

    this thesis on various occasions. I would especially like to thank my good friends Emil

    Aas Stoltenberg, Magnus Rasmussen, Øyvind Skorge and Sirianne Dahlum for reading

    and providing detailed comments on parts of this manuscript at different stages.

    I would never have survived 10 years at UiO without intellectual companions and close

    friends. In this respect, I want to thank my great political science classmates; Aksel, Emil,

    Per Anders, Anders J., Øyvind, Magnus, Lars Petter, Ørjan, Alf, and Rune. Although

    we are no longer formally students, we will never stop studying together. The same goes

    for fellow Blindern-travelers and close childhood friends, André Anundsen and Anders

    Solli Sal, who got bitten by the academia bug together with me when we started at the

    University (and are still affected). I would also like to thank all of my close friends from

    Grenland who have stuck with me and kept it real from high-school through university.

    You know who you are.

    Importantly, I want to thank my closest family; St̊ale (brother), Grethe (mom), and

    Bjarne (dad), who have all encouraged me and supported me in all of my pursuits. None

    of this would have been possible without you! Finally, I owe a debt to my harshest critic,

    fellow researcher, closest friend, biggest fan (?), traveling buddy, and girlfriend, Sirianne,

    4

  • who is more important to this project and me than I can do justice to in these pages.

    Working on this thesis would have been lonely without you.

    5

  • 1 Institutions and Civil Peace in Weak States

    1.1 Introduction

    Somewhere close to a million people have been killed in internal armed conflicts worldwide

    since the end of the cold war (Petterson, 2014a). The indirect costs are equally severe,

    as internal conflicts have large detrimental effects on human development (Gates et al.,

    2012).1 Even if organized violence has been decreasing worldwide (Pinker, 2011), it is

    still one of the greatest global threats to human well-being.

    One often proposed cure to this global burden is the development of political institu-

    tions that promote peace through accountability and good governance. According to this

    argument, currently conflict-prone countries such as Nigeria, Syria, Iraq and Afghanistan,

    could lower their risk of civil war if they became sufficiently well governed and demo-

    cratic, thereby entering a “civil peace” induced by pacifying national institutions (Hegre

    and Nyg̊ard, 2014; Hegre et al., 2001; Walter, 2014). This dissertation shows that this

    institutional civil-peace argument needs updating. The argument is partly right in that

    political institutions matter for conflict. However, in many of the world’s most conflict-

    prone developing nations, state institutions have a limited geographical and substantive

    reach, making their descriptive features unrepresentative of institutions on the ground.

    The situation in many conflict-ridden countries is akin to that described by a senior gov-

    ernment official in the Central African Republic, remarking that “the State stops at PK

    12, twelve kilometers from the capital, Bangui” (Herbst, 2000, 139).

    I will show that the current focus on national institutions as solutions to the problem

    of armed conflict misses much of the subnational institutional variation that matters.

    In the following chapters, I demonstrate that the quality, accountability, strengths, and

    capacities of subnational institutions contribute significantly to civil peace, drawing on

    evidence from Africa. Consequently, I propose that in neglecting the subnational level, the

    1A country suffering from a medium-sized conflict, causing around 2500 battle deaths, is estimated toexperience an increase in infant mortality rate of 10%, and a 3.3% reduction in average life expectancyas a result (Gates et al., 2012).

  • state-centric orientation of the institutional civil-peace argument overlooks the political

    institutions most central to peace in the developing world.

    A state of civil peace is a situation where societal actors with conflict potential reach

    non-violent bargained solutions to differing interests rather than engage in fighting. The

    thesis highlights how two kinds of subnational political institutions contribute to such

    a state. One part of the thesis considers the customary institutions that often organize

    ethnic groups in Africa. A second part studies the role of formal local government insti-

    tutions, like municipalities, regional governments, and local bureaucracies. Formal local

    government institutions are the local representatives of the modern state, while custom-

    ary institutions (often) are the contemporary remnants of pre-colonial states and empires.

    The main claim of the thesis is that the quality and strength of these two kinds of subna-

    tional institutions contribute significantly to local civil peace in Africa. Building on this,

    I argue that subnational institutions should matter more in weak-state contexts such as

    in Africa where state institutions often have a poor reach (Herbst, 2000). To investigate

    these claims, I utilize two datasets on customary institutions compiled from ethnographic

    data and expert surveys, and two datasets using georeferenced survey data to measure the

    quality of formal local government institutions. These datasets have never before been

    used to study internal conflict. Consequently, I present one of the first comprehensive

    and systematic attempts to study subnational customary and formal local government

    institutions in relation to internal conflict.

    Each part of the thesis provides an independent supporting argument for the general

    claim that subnational institutions contribute to civil peace, and that these are particu-

    larly important in states with weak institutions such as in most African countries. First, I

    demonstrate the limits of the institutional civil peace argument when it comes to predict-

    ing civil conflict at the national level. This happens in Chapter 2, where I present a novel

    take on empirically evaluating the links between national political institutions and civil

    peace. The chapter investigates whether liberal institutional attributes like democracy,

    quality of government and rule of law, are able to predict civil conflict out of sample. Cru-

    cially, I examine which countries the institutional-civil-peace model can predict well and

    which countries it has trouble with. This analysis shows that the democracy-autocracy

    dimension and institutional-quality aspects perform particularly poorly when it comes to

    predicting conflict in states with weakly consolidated institutions. In doing so, I outline

    some of the limitations of the national-level approach to the institutions-conflict link,

    making the case that this might reflect the low saliency of national institutions in de-

    veloping countries with weakly consolidated national institutions. This chapter sets the

    stage for the subsequent chapters, motivating a more detailed exploration of the role of

    subnational institutions in weak states.

    The rest of the thesis takes up this challenge. Chapters 3-4 investigate the role of cus-

    tomary institutions in African conflicts, making the claim that historically centralized and

    highly institutionalized customary institutions reduce conflict by facilitating bargaining

    between ethnic groups. Chapter 3 focuses on the historical institutions of ethnic groups,

    demonstrating that groups with historically centralized customary institutions are less

    8

  • prone to ethnic civil wars. Chapter 4 uses contemporary data on the customary insti-

    tutions of ethnic groups to show how the presence of formalized customary institutions

    at the local level, like customary courts, houses of chiefs and legislative councils, reduces

    the involvement of ethnic groups in communal conflicts.

    Although this part of the thesis presents robust evidence that customary institutions

    shape the conditions for ethnic conflict, these customary institutions exist alongside for-

    mal local government institutions whose quality and organization also matter for civil

    peace in Africa. This brings us to chapters 5-6, which exclusively look at the role of

    formal local government institutions in African conflicts. Chapter 5 investigates the rela-

    tionship between subnational institutional quality and local conflict violence, presenting

    evidence for a “local civil peace” whereby administrative districts with high-quality local

    institutions are less prone to conflict. In this context, a clear threat to inferences attribut-

    ing a causal effect of institutions is the possibility of reverse causality, which would be

    the case if conflict affected institutional development. Chapter 6 investigates the impor-

    tance of this reverse-causality threat as it relates to formal local government institutions

    and conflict. This chapter demonstrates that the effects of conflict on local institutional

    development are less grave than expected. While exposure to local conflict violence has

    marginal effects on citizens’ trust in institutions, it does not have any detrimental effects

    on more objective institutional parameters like experiences with local corruption. This

    suggests that the root causes of local institutional quality are much deeper than what can

    be explained by contemporary conflict levels. Together, Chapters 2-6 form a unified pic-

    ture supporting the contention that the quality and strength of subnational institutions,

    both customary and formal, matter for creating civil peace at the local level in Africa,

    and perhaps even more so than national-level institutions.

    The bulk of the chapters (3 through 6) study conflict in Africa. This is partly due to

    data availability. The African continent is where the subnational institutional variety of

    the kind mentioned above is arguably the most prevalent. It is also where we find most of

    the available sources for constructing datasets to capture this variety. Furthermore, the

    African continent represents a cluster of countries with weakly consolidated institutions

    with a short geographical reach (Michalopoulos and Papaioannou, 2014; Herbst, 2000),

    where the subnational variation in institutions should be expected to matter more. In

    states recovering from colonization where state-building has not yet led to institutional

    consolidation is where we would expect to find a) a myriad of surviving political entities

    with their own institutions existing alongside central governments, and b) the largest

    subnational variation in attributes of local government institutions.

    This contrasts with regions such as Europe and North America, where the process of

    state building has taken its course. In such settings, the attributes of nation-states should

    play a larger role in conflict reduction than in weakly consolidated states. For example,

    we would not expect the successors to historical polities with roots prior to the formation

    of modern states to matter for (the absence of) civil conflict in Europe today. While pre-

    state institutions, like the various kingdoms populating the Holy Roman Empire until its

    dissolution in 1806, played a large role in European conflicts prior to the rise of territorial

    9

  • states (Spruyt, 1996), they are no longer as consequential. However, in parts of the world

    where the state-building process is more recent, mainly due to the crippling impacts of

    colonialism, remnants of pre-colonial institutions are still present and play a role in civil

    conflict. Hence, studying subnational institutions in Africa is necessary, both to assess

    the institutional causes of peace in these countries, and because these are the only settings

    in which we will observe this degree of subnational institutional variation in institutions.

    This makes Africa a crucial case for the argument proposed in this thesis.

    Several policy implications flow from the findings and arguments presented below.

    Primarily, in contrast to many arguments made in the institutions-conflict literature (e.g.,

    Walter, 2014; Hegre and Nyg̊ard, 2014; Hegre et al., 2001), the studies indicate that simply

    trying to “fix” national institutions, for example by improving the quality of national

    governance, or introducing free and fair elections, might not be the best institutional

    cure for civil conflict in places where national institutions are weakly consolidated and

    subnational institutions might matter more. In contrast to much literature highlighting

    the negative role played by local authorities in African politics (Mamdani, 1996), and to

    those highlighting that rule by customary rulers could increase conflict risk (Eck, 2014),

    this thesis emphasizes that these institutions are central to sustaining local civil peace

    in Africa. Consequently, a state-building process that undermines local authorities, like

    local customary institutions, might be dangerous precisely because it removes some of

    the institutional frameworks that contribute to local civil peace.

    1.1.1 Crucial definitions

    In section 1.3, I discuss the conceptual framework of the thesis in detail. However, I

    provide some brief definitions of key concepts here, to make the subsequent discussions

    more precise. First, by institutions, I mean “the patterns of interactions that govern and

    constrain the relationships of individuals” (North, Wallis and Weingast, 2009, 15), with

    emphasis on the formal and informal rules governing political interactions.

    Second, in defining customary institutions, I emphasize that these are institutional

    structures that organize ethnic groups. These structures often descend from, and are

    similar to, pre-colonial political systems, whose “legitimacy is rooted in history – either

    real or invented – and culture” Ubink (2008a, 8). Hence, customary institutions are

    founded on traditional authority with reference to an established past. Furthermore,

    what distinguishes customary institutions from other traditional institutions, such as

    for example the Catholic church, is that they are institutions that function as political

    systems for the ethnic groups organized in these institutions, where a political system is

    conceived as “any persistent pattern of human relationship that involves, to a significant

    extent power, rule or authority” (Dahl, 1964, 6).

    I also focus on formal local government institutions, that are here defined as public

    and formal political institutions representing the modern state at the local level. These

    will typically be the institutions governing formal administrative units below the nation

    state, such as municipalities, regional governments and city councils. Local government

    10

  • institutions are the local manifestations of the institutions of the modern state, while cus-

    tomary institutions commonly have independent pre-colonial roots. I shall therefore treat

    local government institutions and customary institutions as separate, while recognizing

    that they often will matter for the same constituents. However, it is important to note

    that the lines between customary authorities and formal local government institutions

    are often quite fluid. In many parts of Africa, as will be discussed below, the power that

    is commonly vested in local formal government institutions is delegated to traditional

    rulers. Moreover, customary leaders will in many cases blend in with formal government

    authorities, for example when customary chiefs are also elected councilors in local formal

    government institutions.

    Finally, the outcomes studied below all sort under the general heading of internal

    armed conflict. I will here loosely follow the standard UCDP-PRIO definition (Gleditsch

    et al., 2002), and define internal armed conflict very broadly, as a militarized violent

    confrontation between two organized actors fighting over some contested incompatibil-

    ity. Although I focus on different manifestations of civil conflict throughout the thesis,

    from classic instances of rebel vs. government civil war, to non-state communal conflicts

    between ethnic groups, these distinctions between conflict types are not of fundamental

    importance to the arguments in the thesis (for further discussion, see section 1.5).

    The structure of this chapter is as follows. In the next section (1.2), I place my

    contribution in the broader literature and make the case for the subnational approach

    advocated throughout the thesis. I will here argue for the special role of subnational

    institutions in weak states, discuss the African case in light of this, and present two case

    examples that illustrate the main arguments. Section 1.3 presents the conceptual frame-

    work of the thesis, discussing institutions, their attributes, and other concepts relevant

    to the thesis in greater detail. Section 1.4 presents the explanatory framework of the

    thesis, outlining the role that subnational institutions can play in reducing the risk of

    conflict. Section 1.5 discusses my approach to measuring subnational institutions and

    conflict. In section 1.6, I discuss how we can draw inferences regarding the pacifying role

    of subnational institutions, with a focus on the philosophical approach to explanation and

    inference taken in the thesis, as well as the general methodological approach advocated.

    Finally, I present the structure of the thesis, in section 1.7, before discussing implications

    and policy recommendations in section 1.8.

    1.2 The institutional causes of conflict: The case for

    a subnational approach

    The questions asked in this thesis tie into a larger debate concerning which political in-

    stitutional structures are conducive to internal peace. This debate has deep historical

    roots. In Politics, Aristotle mused over the causes of revolution under different forms

    of government, suggesting that both Oligarchies, Democracies, Aristocracies and Tyran-

    nies were likely to experience internal upheavals whenever a mismatch arose between the

    11

  • distribution of political power and the distribution of economic power (Aristotle, 2000).2

    The institutional cure for conflict is also a prominent question in post-classical political

    theory. Thomas Hobbes (in)famously proposed a strongly centralized Leviathan as a so-

    lution to the problem of internal violence (Hobbes, 1651/1968), while Locke prescribed

    a more limited form of constitutional rule (Locke, 1689/1988). These debates have con-

    tinued in contemporary political science, where the aim is to evaluate the institutional

    causes of internal conflict using the methods and toolbox of modern science.

    As in Aristotle’s theory of revolution, the primary focus in the contemporary literature

    on institutions and conflict has been on regime type, and how different forms of government

    affect the risk of civil war. An important strand of this discussion revolves around whether

    democracy pacifies. Some arguments imply that more inclusive political regimes should

    be less prone to conflict than others, for example through creating political legitimacy

    that reduces popular grievances (Gurr, 1970), or by providing institutional guarantees

    for the redistribution of wealth (Acemoglu and Robinson, 2006). These views suggest

    a monotonic relationship between democracy and civil war, whereby higher levels of

    democracy yield less conflict.

    The most popular take on the democracy-civil conflict relationship, however, is the so

    called “inverted-u curve” expectation. This proposes that regimes that are “inconsistently

    democratic”, like present-day Nigeria, which combines competitive elections and substan-

    tial restrictions on both civil liberties and political rights (see e.g., Freedom House, 2015),

    are most likely to experience civil war. Meanwhile, consolidated democracies, like Den-

    mark, and consolidated autocracies, like Saudi Arabia, should experience fewer internal

    conflicts (Hegre et al., 2001; Muller and Weede, 1990). What gives rise to this expecta-

    tion is the notion that harshly authoritarian regimes should have the ability to repress

    potential dissidents enough to prevent an uprising, while consolidated democracies should

    allow opposition groups to peacefully voice and negotiate demands through democratic

    channels. In short, violence is too costly in the autocratic end of the spectrum, while

    non-violent opposition is cheap and effective in the democratic end. Although the ini-

    tial evidence for the inverted-u association was strong (Hegre et al., 2001), it has been

    criticized for hinging on biases in the democracy measure used (Vreeland, 2008), and

    insufficient attention to measurement error (Treier and Jackman, 2008).3 However, the

    association remains in analyses where the criticism of coding bias is accounted for (e.g.,

    Gleditsch, Hegre and Strand, 2009; Regan and Bell, 2010).

    While arguments concerning the inverted-u curve are mostly about how levels of

    democracy relate to conflict, some argue that democratic transitions are dangerous. Some

    attribute this to the sudden increase in political participation in settings where institu-

    tional prerequisites for peaceful political competition – like the rule of law or an impartial

    bureaucracy – are lacking (Huntington, 1968; Linz and Stepan, 1996). Others argue that

    2For a precise statement of this interpretation of Aristotle‘s theory of revolution, see Kort (1952).3Another important caveat to the u-curve is presented in Gleditsch and Ruggeri (2010), who find that

    once controlling for opportunity structures – proxied by recent irregular leader removal – the monotonicrelationship between democracy and civil conflict reappears.

    12

  • democratization is conflict-inducing because of commitment problems and strategic un-

    certainty (Przeworski, 1991; Kalyvas, 2000), while many highlight that transition periods

    are particularly vulnerable to ethnic-mobilization dynamics that lead to ethnic conflict

    (Snyder, 2000; Mansfield and Snyder, 2005).

    A significant number of contributions have looked at other institutional aspects than

    democracy. Some studies have focused on the presence of “consociational” power-sharing

    institutions that protect ethnic minorities (e.g., Hartzell and Hoddie, 2007; Gates et al.,

    2013), primarily building on the work of Lijphart (1999, 1977, 1969). More recent studies

    have emphasized aspects related to quality of government, like a well-functioning bu-

    reaucracy and the rule of law (Hegre and Nyg̊ard, 2014; Fearon, 2011; Taydas, Peksen

    and James, 2010; Walter, 2014). Scholars focusing on quality of government argue that it

    reduces conflict-inducing grievances (Hegre and Nyg̊ard, 2014), and diminishes opportuni-

    ties for conflict through improving state capacity (e.g., Taydas, Peksen and James, 2010).

    It has also been hypothesized that quality of government matters by solving bargaining

    problems (Walter, 2014).

    The broad strokes of this literature reveals the contours of what I call an institutional

    civil-peace argument, where states that are consistently democratic and well-governed

    should be the least prone to conflict. This argument is commonly supported by either

    of three general “meta-theories” concerning how institutions cause conflict, that each ap-

    pear in the works surveyed above. Although these will be discussed further in section

    1.4 below, they deserve brief mention here. One perspective emphasizes that political

    institutions can cause conflict in so far as they create political grievances and feelings of

    injustice that motivate rebellion (Gurr, 1970; Cederman, Gleditsch and Buhaug, 2013).

    The “opportunity” perspective claims that institutions should cause violence if they en-

    gender opportunities for rebel activity (Fearon and Laitin, 2003), exemplified by most of

    the arguments for the“inverse u-curve”(e.g., Hegre et al., 2001). Finally, others emphasize

    that institutions are devices for solving commitment problems, highlighting that institu-

    tions that enshrine power sharing (Walter, 2002), high quality of government (Walter,

    2014), or democracy (Acemoglu and Robinson, 2006) will allow actors to more credibly

    distribute gains and losses in ways that would avoid war.

    Although the studies and arguments surveyed here represent very important contribu-

    tions to the study of which institutional arrangements increase the risk of internal conflict,

    the field seems to have reached a plateau from which progress appears to be much slower.4

    In spite of the high explanatory power of the argument that inclusive political institu-

    tions and quality of government promotes peace, several key questions are insufficiently

    explored. First, and importantly for this thesis, we do not know whether the statistical

    associations identified, for example the inverse u-curve association, are general and robust

    relationships that allow us to predict conflict in diverse settings. If the institutional civil

    peace thesis is a robust general explanation for civil conflict, it should allow us to pre-

    4The studies mentioned use many of the same data sets, and many of the investigations resemble eachother. Hence, there is only so much more we can learn from the next cross-country time-series study ofcivil conflict, using the same measures of institutional characteristics like regime-type and power sharing.

    13

  • dict civil conflict across a range of spatio-temporal contexts, and, crucially, it should be

    able to predict “unseen” cases that have not been used for establishing the relationships

    asserted by the theory.5 Furthermore, we do not know whether the institutional argu-

    ments for conflict can “travel” to incorporate other institutions than those at the national

    level. For example, whether these arguments yield implications that are validated at the

    subnational level. Finally, the possibility that the effects of national institutions might

    be conditional on the reach of these national institutions, which is suggested in studies

    of economic growth (Michalopoulos and Papaioannou, 2014), is rarely entertained.

    These considerations hint at a more fundamental gap in the literature on political

    institutions and civil peace, and an unexplored alternative to the institutional civil peace

    argument. Namely, that too little attention has been devoted to other institutions than

    national ones in contexts where national institutions are weak and unrepresentative of

    the institutional variation on the ground in developing countries. I will now go on to

    develop this argument, making the case for a novel civil-peace research agenda focusing on

    subnational institutions in weak states. In doing so, I will present Africa as a particularly

    suited case for the implementation of such a project.

    1.2.1 Subnational institutions in weak states: Beyond method-

    ological statism

    While many research areas in conflict research have moved towards disaggregation from

    the national to the subnational level (for an overview, see Gleditsch, Metternich and Rug-

    geri, 2014), the brand of conflict research looking at the effects of political institutions has

    not.6 It is still almost exclusively focused on the institutions of nation-states. There are

    several reasons why this focus should be reoriented towards investigating the institutional

    causes of peace at the subnational level.

    First, there are examples from related sub-fields within conflict research indicating

    the potential benefits of disaggregation. Disaggregating to subnational units has yielded

    significant returns, for example in investigations of the economic correlates of civil conflicts

    (e.g., Buhaug et al., 2011; Hegre, Østby and Raleigh, 2009).7 This is due to a realization

    of the dictum that “all politics is local” (Rustad et al., 2011). Indeed, civil war, like many

    other phenomena has origins and dynamics that are essentially local in origin (Rustad

    et al., 2011). For example, even if national GDP per capita indicates a low risk of conflict,

    a country might have particularly poor regions whose income levels dispose for conflict

    in contrast to the national average (e.g., Buhaug et al., 2011). Furthermore, it might

    be precisely those groups residing in regions that are richer or poorer than the country

    5There is some indication that the democracy-autocracy dimension predicts relatively poorly out ofsample (Ward, Greenhill and Bakke, 2010), which might suggest that it describes a specific group ofcountries and not others.

    6See also the special issue of Journal of Conflict Resolution on“Disaggregating Civil War”’ (Cedermanand Gleditsch, 2009).

    7This has partly been a response to the very large number of studies having investigated this associ-ation at the national level (Collier and Hoeffler, 2004; Miguel, Satyanath and Serengeti, 2004)

    14

  • average that will resort to violence (Cederman, Weidmann and Gleditsch, 2011). In these

    cases, exclusively looking at national GDP increases the threat of ecological fallacies.

    In light of this, disaggregation has resulted in much firmer conclusions regarding, for

    example, the relationship between poverty and conflict (e.g., Buhaug et al., 2011).

    Given that subnational disaggregation has been a largely successful endeavor, why

    has it not taken place in the study of the institutions-conflict link, and why does this

    matter?8 First, one obvious answer is that many institutions by definition only exist at

    national levels. It makes little sense, for example, to investigate the effects of democratic

    regime-type on the risk of civil war onset using subnational administrative districts as

    units of analyses. This is because democratic regime types by definition exist at the

    national level.9 A second reason is a lack of data describing subnational institutional

    variation.

    In spite of these caveats, there are several reasons why disaggregation should nev-

    ertheless be pursued in studies of political institutions and conflict. The main reason

    highlighted here is substantive. Disaggregation it allows us to study the role of subna-

    tional institutions in the contexts where attributes of national institutions might matter

    less for conflict. There are several reasons why we should expect this to be the case. In

    states with weakly consolidated national institutions, attributes of these institutions will

    not describe areas where national level institutions have no reach. In Africa, coloniz-

    ers primarily focused on ruling their capitals, commonly located at the coast, governing

    through local customary authorities in the periphery (Mamdani, 1996). As emphasized

    by Africa scholars, this lack of centralization has persisted, partly due to the colonial

    heritage, but also due to geographies that make political centralization difficult (Herbst,

    2000; Mamdani, 1996). Recent large-n studies support the contention that national insti-

    tutions in Africa have a short reach. In a recent study, Michalopoulos and Papaioannou

    (2014) use satellite night-time light density as a proxy for local economic growth, find-

    ing that the effect of national institutional quality on regional development decays with

    distance from the capital. They find that for approximately 60% of their sample (of

    ethnic-group homelands), features of national institutions, measured by several World

    Governance Indicators (Kaufmann, Kraay and Mastruzzi, 2009), essentially have no ef-

    fect on local economic development. This is evidence suggesting the poor reach of national

    institutions in Africa. The statement by the bureaucrat in CAR (quoted above), that ”the

    State stops at PK 12”, might apply to more countries than the Central African Republic.

    If national institutions have a poor reach in weak states, like those found in Africa,

    we must move beyond studying how national institutional attributes shape conflict risk.

    8An exception is the emerging literature looking at the institutional consequences of conflict (e.g.,Bellows and Miguel, 2009; Voors and Bulte, 2014; Gilligan, Pasquale and Samii, 2014; Voors et al.,2012), which will be the subject of chapter 6. These studies are not extensively reviewed here, since theyare not about how institutions condition conflict, but the reverse causal relationship.

    9Although one could investigate the more abstract question of whether inclusive political structures ingeneral breed less violence by looking at subnational variations in the inclusiveness of local government,the fact that many political institutions, in certain respects, only exist (by definition) at the nationallevel has probably halted disaggregation in this part of the literature.

    15

  • Crucially, we must understand the role of local institutions. But what kind of local

    institutions? In the following, I discuss local-level institutional variation in Africa as a

    particularly illustrative example of the relationship between a weak state consolidation

    and subnational institutions.

    Variations in subnational institutions: The case of Africa

    One dimension of subnational institutional variation that this thesis considers is the

    customary institutions of ethnic groups, descending from pre-colonial political systems.

    During colonialism, African states were constructed on top of a plethora of pre-existing

    polities, a majority of which have survived in various forms up until the present day.

    pre-colonial kingdoms like the Zulu kingdom, Ashanti empire or the state of Buganda

    are notable examples. Unlike European nations, which have grown endogenously, and

    whose borders have been chiseled out by millennia of warfare and migration (Kitamura

    and Lagerlöf, 2015), most contemporary African states are modern constructs, layered

    on top of “preexisting political institutions, underlying norms of political behavior, and

    customary sources of political authority” (Englebert, 2000, 5).

    These pre-colonial political systems, described as customary or pre-colonial institu-

    tions in this thesis, have been shown to play a significant role in contemporary African

    politics. Some highlight their detrimental effects. Eck (2014) argues that customary

    political systems within states create competing legal jurisdictions that increase the risk

    of communal conflicts. Furthermore, Englebert (2000) argues that the existence of pre-

    colonial states within artificially constructed borders have caused a legitimacy deficit for

    national institutions, resulting in poor institutional quality and low economic develop-

    ment. In this account, poor state capacity in Africa is partly endogenous to the existence

    of pre-colonial political structures within artificially constructed borders.

    Regardless of the role of pre-colonial polities in creating state weakness, the fact that

    African national institutions are exceptionally weak (see e.g., Herbst, 2000) has allowed

    these customary political systems to play a major role as institutional structures beneath

    the state (see also Migdal, 1988). Accordingly, much recent work has documented the

    significant impact that pre-colonial states have on contemporary economic growth in their

    traditional homelands (Michalopoulos and Papaioannou, 2015, 2013, 2014). Others have

    highlighted the role of customary chiefs in local public goods provision (Baldwin, 2013),

    and their role as arbiters in resource- and land management (Boone, 2003, 2014). The

    resilience of these institutions is attested by the significant degree of popular support that

    they still enjoy (Logan, 2013).

    A representation of the variety in pre-colonial polities in Africa can be found in fig-

    ure 1.1 which shows the ethnographic map of Murdock (1967) – discussed at length in

    Chapter 3 – where each polygon represents a pre-colonial polity, as these are described

    in the comparative ethnographic data collected by Murdock (1967). This map, although

    arguably subject to some measurement error (Michalopoulos and Papaioannou, 2015),

    shows the immense variety in pre-colonial polities existing before the creation of modern

    16

  • African states. Many of these polities have survived in the form of customary authorities

    at the local level. Visible manifestations of the persistence of customary political systems

    can be found in the traditional legislatures, courts, kings and houses of chiefs that mark

    the institutional landscape in Africa. This thesis highlights the importance of studying

    the role of these customary institutions in civil conflicts, which will be the subject of

    chapters 3 and 4.

    Figure 1.1: Murdock’s ethnographic map of pre-colonial polities in Africa

    Another source of subnational variation in Africa is the variation in formal local

    government institutions. As is frequently noted (e.g., Rustad et al., 2011), conflicts in

    the developing world are often distinctly local phenomena, that haunt particular regions

    rather than entire countries. Northern Nigeria, northern Uganda, or the Kivu region in the

    DRC, are all contemporary examples of regions that are significantly more conflict prone

    than the countries they are part of, hosting the Boko Haram, Lord‘s Resistance Army,

    and the M23 insurgencies respectively. There is also significant subnational variation

    in attributes, and particularly in the quality, of formal local government institutions in

    Africa.

    An illustrative example of this pattern can be seen in the Afrobarometer surveys (Afro-

    barometer, 2008). Chapter 4 maps survey responses in the Afrobarometer to administra-

    tive districts, using aggregate responses (at the district level) as proxies for institutional

    17

  • Table 1.1: Survey items capturing perceptions of local institutional quality from theAfrobarometer survey (round 4)

    Survey question R2 attributed tocountry-specific effects (Afrobarometer round 4)

    Experienced corruption .23Police corrupt .51Local authorities corrupt .43Trust police .37Trust local authorities .38Trust courts .39

    The items used will be described in greater detail in chapters 4 and 5, as well as in the appendices tothese chapters.

    quality. Using these data, table 1.1 (below) considers responses tapping institutional

    quality from this dataset, using items from round 4 of the Afrobarometer survey. The

    table shows how much of the variance in survey items (tapping institutional quality) that

    is accounted for by country level factors (a set of country dummies). The items register

    experiences with corruption, perceptions of corruption among the police and other local

    authorities, trust in police and other local authorities, and trust in courts. Table 1.1

    displays the R2 yielded by country dummies when modeling these items. We see that

    that even if national-level factors explain a substantial amount, the highest R2 being .51,

    country-level factors commonly account for less than half of the district-level variation in

    perceived and experienced institutional quality. This indicates the importance of looking

    at variation in subnational institutions in Africa.

    Drawing inference based on African cases

    As noted, there are substantive reasons for investigating subnational institutions in the

    context of weak states, and Africa is the best candidate for doing this. However, there are

    several external validity issues related to generalizing from such a study. First, a clear

    limitation is that Africa-based inferences might only be valid for contexts where there

    are weak national-level institutions, or, in the worst case, only for Africa. It should be

    explicitly noted that I do not expect subnational institutions to matter as much in fully

    consolidated states, such as those found in most OECD countries.

    A more subtle concern is that studying these institutions in weak states can be seen as

    selection on the dependent variable. After all, state weakness and lack of state building

    is probably both an outcome and a cause of conflict, meaning that the subnational level

    is partly relevant because of factors that yield a high risk of conflict. In other words, the

    very fact that these institutions matter is because these societies are particularly conflict

    prone.

    Nevertheless, I believe there are important theoretical lessons to be learned from

    disaggregation. Even if many of the findings are not directly translatable to other contexts

    18

  • than the weak-state situation studied, or perhaps not even translatable to contexts outside

    of Africa, the studies presented below yield many general insights that are of theoretical

    significance, and that, importantly, add to the general empirical content of the theories

    applied. For example, Chapter 5 investigates implications of general theories relating

    to how institutional quality should promote civil peace – commonly tested on country-

    year data (e.g., Hegre and Nyg̊ard, 2014) – on subnational data. Since the arguments

    relating institutional quality to civil peace at the country level also imply that local

    institutional quality should reduce the prevalence of local conflict (see Chapter 5 for a

    survey), investigating this proposition yields an additional piece of evidence in favor of

    those theories.

    Another example can be found in chapters 3 and 4, where I draw on bargaining theory

    (Fearon, 1995) to generate many of the expectations. In these cases, empirical support

    for these hypotheses add to the empirical support for bargaining theory more generally.

    Moreover, in Chapter 4 I explicitly set out to test general institutional explanations for

    war developed for intra- and inter-state conflicts, on the customary institutions of ethnic

    groups in Africa and their involvement in communal conflicts. In this case, the specific

    inferences relating to the customary institutions of ethnic groups have low external valid-

    ity because of the peculiarities of African customary institutions. However, the findings

    add to the strength of the overarching theories tested, since the fact that implications of a

    general theory hold in a novel context yields additional empirical support. The fact that

    most theories applied in this thesis is developed in other contexts, like theories relating

    to institutions and international conflict, adds to the strength of the general theoretical

    inferences drawn from each study.10

    To recap, I maintain that conflict researchers interested in how institutions shape

    conflict outcomes in weak states need to study subnational institutions. Before discussing

    the conceptual, empirical and methodological tools I will use to put this appeal into action,

    I briefly discuss two concrete cases that illustrate the arguments made in the thesis.11

    1.2.2 Customary institutions: Ashanti in Ghana

    One example of a pre-colonial polity that has persevered into modern times is the Ashanti

    kingdom in Ghana. Prior to colonization Ashanti had developed into an empire, with

    a centralized government, a judiciary, a bureaucracy, police force and diplomatic corps

    (Edgerton, 2010). Indeed, Ashanti resembled many of the young European states of

    the eighteenth and nineteenth century (Morisson, 1983, 450). Many of the structures of

    10One of the virtues of testing social scientific theories is to maximize the number of implications ofthe theory under investigation, and to test as many of these implications as possible (King, Keohaneand Verba, 1994, 24,30). If a theory can predict and test entirely novel implications, then this adds tothe truth-content of the theory (for a general statement of this view, see Lakatos, 1980). This points toanother benefit to disaggregation, namely that it presents us with a promising arena for testing additionalimplications from theories previously tested on country-level data. Several parts of the thesis show thisphilosophy in action.

    11These cases are included for illustrative purposes, and should not be understood as case studies intheir own right.

    19

  • the Ashanti kingdom have survived. After initially having resisted British colonization

    Ashanti became a protectorate of Britain in 1896, and was subsequently incorporated

    as a quasi-sovereign state in a state union with Ghana after independence. Although

    the Ashanti kingdom has lost many of its formal and de-facto powers in relation to the

    Ghanaian state, its key institutions, such as the kingdom, the institution of chieftaincy

    and its traditional house of chiefs, all persist and play important roles in local politics in

    Ashantiland. For example, traditional Ashanti authorities are instrumental in land man-

    agement, conflict mediation, and the adjudication of general disputes and land claims

    (e.g., Boone, 2014, 2003; Ubink, 2008b,c; Owusu-Mensah, 2014). In this thesis, I inves-

    tigate three expectations regarding traditional institutions and their relation to ethnic

    conflict. I will now proceed to illustrate each of them by applying them to the Ashanti

    case.

    Figure 1.2: Ghana (red border) and contemporary Ashantiland (green border)

    ABE ADANGME

    ADELE

    AJUKRU

    AKPOSO

    AKYEM

    ANA

    ANYI

    ARI

    ASHANTI

    ASSINI

    ATTIE

    ATYUTI

    AVATIME

    AVIKAMBAKWE

    BARGU

    BASARI BASILA

    BAULE

    BETE

    BIRIFON

    BOBO

    BRONG

    BUEM

    BUILSA BUSA

    BUSANSI

    CHAKOSSI

    DAGARI

    DAGOMBA

    DENDI

    DIAN

    DIDA

    DIULA

    DOROSIE

    EBRIE

    EGBA

    EWE

    FANTI

    FON

    GA

    GAGU

    GAN

    GREBO

    GRUNSHI

    GUANG

    GUIN

    GUN

    GURENSI

    GURMA

    GURO IFE

    IJEBU

    KABRE

    KARABORO

    KEBU

    KILINGAKOMONO KONKOMBA

    KRACHI

    KULANGOLIGBI, DEGHA (SE)

    LILSE

    LOBI

    MAMPRUSI

    MEKYIBO

    MINIANKA

    MOBA

    NAFANA

    NAUDEBA

    NUNUMA

    POPO

    RESHE

    SENUFO

    SIA

    SOMBA

    TEM

    TIENGA

    TRIBU

    TUSYAN

    VAGALA

    WABA

    WARA

    WOBE

    YORUBA

    First, I argue that customary authorities with a tradition of political centralization,

    such as Ashanti, are less likely to experience conflicts with central governments because

    they have been able to rely on their historically centralized institutions to reach credible

    non-violent bargains with the state (Chapter 3). Although one can not observe a “dog

    that did not bark”, meaning a counterfactual conflict that occurred in the absence of

    Ashanti institutions, we can observe successful bargaining between Ashanti leaders and

    the central Ghanean government, as well as the absence of major conflict between the

    two.

    20

  • For example, under British colonial administration the Ashanti were given formal

    authority over many aspects of land management, a deal described as “the linchpin of

    British indirect rule in Ghana” (Boone, 2014, 215). This endowed power-sharing sta-

    tus continued after the end of British rule. For example, the first Chieftaincy act of

    1961 assigned a privileged status to Ashanti chiefs, which occurred in spite of president

    Kwameh Nkrumah‘s attempts to undermine traditional leaders in Ashantiland (Boone,

    2003, 159-163).12 The powers of Ashanti chiefs were further expanded and codified in the

    constitution of 1969, the chieftaincy act of 1971, the constitution of 1992 and the recent

    chieftaincy act of 2008.

    These examples illustrate how successive Ghanean governments have devolved powers

    to Ashanti chiefs, mostly concerning land rights (Owusu-Mensah, 2014). A key reason

    was that they were seen as credible partners that could command loyalty from the popu-

    lations that identified with them (e.g., Boone, 2003). Furthermore, although there have

    been many incompatibilities between Ashanti leaders and the Ghanean government, for

    example over the control of cocoa production (Boone, 2003, 159-174), there has been no

    organized military conflict between the Ashanti ethnic group and Ghana to date.

    In Chapter 4, I argue that formal customary institutions, like customary courts and

    legislatures, can reduce the likelihood of inter- or intra-group conflict between or within

    ethnic groups. I argue that this is because these institutions facilitate credible bargaining

    and conflict mediation, between tribes belonging to the same ethnic group, and with out-

    side groups. This mechanism is also evident in the case of Ashanti. In a description of the

    role of Ashanti chiefs in managing conflicts over land, Boone writes that“chiefs are pivotal

    in dispute adjudication” (Boone, 2014, 217). Based on survey data on citizen attitudes

    to chiefs in the Ashanti region, Ubink (2008b, 152–154) finds that 78% of respondents

    noted “dispute resolution” as a main task of traditional chiefs, and 53% listed “ensuring

    peace” as a main task. An example of dispute resolution by Ashanti institutions is given

    by Alden Wily (2003, 66), who describes how the Ashanti King intervened as a response

    to frustration over a huge backlog of land-dispute cases that had yet to be brought in

    front of government courts:13

    [T]he King of the Ashanti, who was so frustrated with the failure of the

    courts to deal with disputes that he ordered his subjects to remove their cases

    from courts and have them resolved by his local chiefs or, failing resolution,

    bring them to his central Customary Court. By 2000, all but ten cases had

    been resolved.

    These anecdotes shed some light on the more general patterns investigated in Chapters

    3 and 4, illustrating how Ashanti rulers are seen as credible bargaining partners by the

    central government, and how their institutions contribute to dispute adjudication and

    conflict resolution locally.

    12It is important to note that Ashanti chiefs were given much fewer powers under post-colonial gov-ernments, especially Nkrumah‘s, than under British colonial rule (Boone, 2003, 144-174)

    13This is also described in Ubink (2008b, 154)

    21

  • 1.2.3 Formal local government institutions: The Boko Haram

    insurgency in Nigeria

    In this thesis, I claim that areas with poor local institutional quality will be more likely to

    experience local conflict violence, because low-quality local government generates political

    grievances and low state capacity (Chapter 5). Nigeria is one illustrative example, where

    reports suggest a relationship between poor local institutional quality and local conflict.

    Some regions are notoriously poorly governed, like the Northern states of Nigeria, which

    have become the core of the Boko Haram insurgency. Up until, and during, the Boko

    Haram insurgency, northern authorities have been plagued by corruption charges. For

    example, in 2010, the head Police Commisioner in the northern state of Borno was caught

    embezzling over 100 000 dollars in government money (Human Rights Watch, 2010a, 97),

    and the region has been continuously plagued by widespread corruption and extrajudicial

    killings (Human Rights Watch, 2012). Many analysts have traced the origins of the Boko

    Haram insurgency to the corruption and general incompetence that has characterized

    local authorities in the northern states. A Nigerian journalist that has interviewed Boko

    Haram leaders, is here quoted in a report by Human Rights Watch:

    Corruption became the catalyst for Boko Haram. [Mohammed] Yusuf [the

    group‘s first leader] would have found it difficult to gain a lot of these people

    if he was operating in a functional state. But his teaching was easily accepted

    because the environment, the frustrations, the corruption, [and] the injustice

    made it fertile for his ideology to grow fast, very fast, like wildfire. (Human

    Rights Watch, 2012, 24)

    Boko Haram has both capitalized on the weaknesses of the security forces in the north,

    and on citizen’s frustrations with poor governance. The group’s promise of security and

    uncorrupt administration based on Sharia law has found fertile ground in such a setting.

    Meanwhile, other parts of Nigeria, like Lagos, are becoming increasingly well-governed.

    In a recent New York Times op-ed, Lagos is described as“A model city for the world”, and

    much of its recent economic boom is attributed to competitive local elections in which

    candidates have increasingly been forced“to show pragmatism and competence”’ (Kaplan,

    2014). The point of this comparison is to illustrate how regional variation in institutional

    quality might matter for where in a country we will see violent conflict; the state of Borno,

    for example, has poor institutions and is highly affected by the insurgency, while Lagos

    is comparatively better run and has remained much less touched by the violence.14 A

    systemati