t1f - 26121 May 2003 H ~~ / tW w lF- :~- 1/ 9''s< -andDevelopment I --~~~~~~~~~~I x, b ,ohw ' AV: Breaking the Conflict Trap Civil War and Development Policy A World Bank Policy Research Report Breaking the Conflict Trap Civil War and Development Policy Paul Collier V. L. Elliott Havard Hegre Anke Hoeffler Marta Reynal-Querol Nicholas Sambanis A copublication of the World Bank and Oxford University Press C) 2003 The International Bank for Reconstruction and Development / The World Bank 1818 H Street, NW Washington, DC 20433 Telephone 202-473-1000 Internet www.worldbank org E-mail feedback@worldbank.org All rights reserved. 1 2 3 4 06 05 04 03 A copublication of the World Bank and Oxford University Press. This volume is a product of the staff of the World Bank. The findings, interpretations, and conclusions expressed herein do not necessarily reflect the views of the Board of Executive Directors of the World Bank or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgment on the part of the World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. Rights and Permissions The material in this work is copyrighted. Copying and/or transmitting portions or all of this work without permission may be a violation of applicable law. The World Bank encourages dissemination of its work and will normally grant permission promptly. For permission to photocopy or reprint any part of this work, please send a request with complete information to the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA, telephone 978-750-8400, fax 978-750-4470, www copyright.com. All other queries on rights and licenses, including subsidiary rights, should be ad- dressed to the Office of the Publisher, World Bank, 1818 H Street NW, Washington, DC 20433, fax 202-522-2422, e-mail pubrights@worldbank.org. Library of Congress Cataloging-in-Publication Data has been appliedfor ISBN 0-8213-5481-7 Cover photo. C) Peter Turnley/CORBIS Cointents Foreword ix The Report Team xiii Acronyms and Abbreviations xv Overview 1 Let The m Fight It Out among Themselves? 1 What Can We Do about Ancestral Hatreds? 3 The Conflict Trap 4 The Rising Global Incidence of Conflict 5 Nothing Can Be Done 6 PART I. CRY HAVOC: W:HY CIVIL WAR MATTERS 11 1. Civil WGar as Development in Reverse 13 Costs during Conflict 13 Legacy Effects of Civil War 19 Conclusion 3 1 2. Let Them Fight It Out among Themselves? 33 Neighborhood Effects of Civil War 33 Global Effects of Civil War 41 Conclusion 48 PART II. WVHAT FUELS CIVIL WAR? 51 3. What iMakes a Country Prone to Civil War? 53 Understaniding Rebellion 55 The Conflict Trap 79 Conclusion 88 v CONTENTS 4. Why ls Civil War So Comnmaon? 93 Changes in the Global Pattern of CIvlI War 93 Changes in the Incidence of Civil War 98 Unpacking the Global Incidence of Civil War 100 Conclusion: Poverty and the Conflict Trap 117 PART 1[111. POLIICRIES FOR PEACE 119 5. What Works Where? 212 Conflict Prevention in the Successful Developers 122 Marginalized Countries at Peace 125 Ending Conflicts 140 Reducing Postconflict Risks 150 Conclusion 171 6. An Agenda for Interational Action 173 Precedents for International Action 174 International Policies for Peace 175 Conclusion: A New Goal for 2015? 186 Appendix 1. Methods nd D)ata 189 Data Set and Model 189 Data Sources 193 Appendix 2. A SeRected Bibliography of Studies of Civil War and Reellion 1197 Economic Factors 197 Role of Ethnicity and Nationalism 198 Anatomy of Rebellion 199 Role of the State 199 Negotiation and Implementation of Peace 200 Bibliography 200 References 21 1 Boxes 1.1 Violent conflict and the transformation of social capital 16 1.2 Refugees and IDPs in Liberia and Sudan 20 1.3 Angola 28 1.4 Psychological trauma 29 1.5 Landmines: A bitter legacy for Cambodians 31 2.1 Regional arms races 34 2.2 Eritrea 40 3.1 Modeling the risk of civil war 58 3.2 Oil and demands for secession in Nigeria 61 3.3 Inefficient counterinsurgency measures in Indonesia 73 3.4 Financing the Chechen rebellion 78 vi CONTENTS 3.5 Modeling the duration of civil war 80 4.1 Recurrent conflicts example 1: Afghanistan 104 4.2 Recurrent conflicts example 2: Angola 105 5.1 A cornparison of Botswana and Sierra Leone 127 5 2 Transparency of oil revenues in Chad 131 5.3 The iough diamond trade and the Kimberley process 143 5.4 The ]Chmer Rouge and the logs of war 145 Figures 1.1 GDP per capita before and after civil war 14 1.2 Total number of refugees, 1962-2002, 18 1.3 Increase in mortality rates due to civil war 24 2.1 The flow and stock of refugees, 1951-2002 36 2.2 The stock of refugees and civil wars, 1951-2001 37 2.3 Refugees and cases of malaria, 1962-97 38 2.4 Opium production, 1986-2001 43 2.5 Cocaine production, 1986-2001 43 2.6 Opium production in Afghanistan and heroin seizures in Europe, 1980-2001 45 2.7 Estimates of annual opiate and cocaine use in the late 1990s 45 3.1 Ethnic fractionalization and the risk of civil war 59 3.2 Risk of civil war for the typical low-income country with and without ethnic dominance during a five-year period 59 3.3 Risk of civil wars from natural resources endowment 61 3.4 The i isk of civil war in democracies and nondemocracies at different levels of income 65 3.5 Improved economic performance and the risk of civil war 67 3.6 Military expenditures and the risk of civil war 72 3.7 Natural resources and the risk of civil war for low-income counltries 76 3.8 How chances of peace evolve worldwide 81 3.9 Duration of civil wars over time 82 3.10 The iisk of civil war for a typical civil war country, just before and just after war 83 3.11 Diasporas and postconflict risk 85 3.12 Military spending and the risk of renewed conflict in postconflict countries 86 4.1 The global incidence of civil warfare, 1950-2001 94 4.2 Simulating the effects of the waves of decolonization, 1950-2020 95 4.3 Proportion of civil wars that end each year 96 4.4 The global self-sustaining incidence of civil war, by decades 97 4.5 Factors changinig the global risk of conflict 99 4.6 The changing rates of conflict termination 100 4.7 Divergent risks: marginalized countries relative to successful developers 102 vii CONTENTS 4.8 Development of risk of civil war for the marginalized and successful developers, 2000-2020 103 4.9 The conflict trap: risk of civil war relative to a country with no recent war 106 4.10 The conflict trap by type of country 107 4.11 Risk components for marginalized countries in the conflict trap, relative to the same countries preconflict 107 4.12 The conflict trap in 2000: annual flows into and out of conflict 109 4.13a The conflict trap in 2020: annual flows into and out of conflict 111 4.13b The conflict trap in 2050: annual flows into and out of conflict 111 4.14 The incidence of civil war in South and East Asia and in Oceania, 1950-2001 112 4.15 The incidence of civil war in Latin America and the Caribbean, 1950-2001 113 4.16 The incidence of civil war in Eastern Europe and Central Asia, 1950-2001 113 4.17 The incidence of civil war in the Middle East and North Africa, 1950-2001 114 4.18 The incidence of civil war in Sub-Saharan Africa, 1950-2001 114 4.19 The incidence of civil war in Africa and other developing countries, 1950-2001 115 5.1 The contribution to peace of faster growth in the successful developers 124 5.2 The contribution to peace of faster growth in the marginalized countries 135 5.3 The contribution to peace of shortening conflicts 141 5.4 The contribution to peace of successful postconflict policies 151 6.1 The contribution of the policy package to peace 187 TabRes 1.1 Major refugee and IDP populations, 2001 19 1.2 Mortality rates among children under five in refugee and IDP camps, selected conflicts 25 1.3 Effects of civil war on public health 26 1.4 HIV prevalence in the military, selected countries and years 27 2.1 Production of opium and coca, selected countries and years, 1990-2001 42 2.2 Prevalence estimates of opiate and cocaine use, selected industrial countries and years 46 3.1 Size of rebel organizations, selected countries and years 55 VIII Foreword W T7'HY SHOULD THE WORLD BANK FOCUS ON CIVIL WARP Basically, there are two reasons. First, civil war usually has devastating consequences: it is development in reverse. As civil wars have accumulated and persisted, they have generated or in- tensified a significant part of the global poverty problem that is the World Bank's core mission to confront. Part of the purpose of this report is to alert the international community to the adverse conse- quences of civil war for development. These consequences are suffered mostly by civilians, often by children and by those in neighboring countries. Those who make the decisions to start or to sustain wars are often relatively immune to their adverse effects. The international com- munity thc refore has a legitimate role as an advocate for those who are victims. The second reason why the World Bank should focus on civil war is that development can be an effective instrument for conflict pre- vention. T[e risk of civil war is much higher in low-income countries than in middle-income countries. Civil war thus reflects not just a problem fir development, but a failure of development. The core of this report sets out the evidence on the efficacy of development for con- flict prevention and proposes a practical agenda for action. The World Bank and its partner development agencies can undertake parts of this agenda, but other parts depend on international collective action by the governments of industrial countries. One important forum for such ac- tion is the Group of Eight. Our research yields three main findings. First, civil wars have highly adverse ripple effects that those who determine whether they start or end obviously do not take into account. The first ripple is within the country: most of the victims are children and other noncombatants. The second ripple is the region: neighboring countries suffer reduced ix FOREWORD incomes and increased disease. The third ripple is global: civil war gen- erates territory outside the control of any recognized government, and such territories have become the epicenters of crime and disease. Many of these adverse consequences persist long after the civil war has ended, so that much of the costs of a war occur after it is over. The second finding is that the risks of civil war differ massively ac- cording to a country's characteristics, including its economic character- istics. As a result, civil war is becoming increasingly concentrated in rel- atively few developing countries. Two groups of countries are at the highest risk. One we refer to as the marginalized developing countries, that is, those low-income countries that have to date failed to sustain the policies, governance, and institutions that might give them a chance of achieving reasonable growth and diversifying out of dependence on pri- mary commodities. On average, during the 1990s these countries actu- ally had declining per capita incomes. Such countries are facing a Rus- sian roulette of conflict risk. Even countries that have had long periods of peace do not seem to be safe, as shown by recent conflicts in Cote d'Ivoire and Nepal. It is imperative that such countries are brought into the mainstream of development. The other high-risk group is countries caught in the conflict trap. Once a country has had a conflict it is in far greater danger of further conflict: commonly, the chief legacy of a civil war is another war. For this group of countries the core development challenge is to design international interventions that are effective in stabilizing the society during the first postconflict decade. The third finding is that feasible international actions could sub- stantially reduce the global incidence of civil war. Although our pro- posals are wide-ranging, including aid and policy reform, we place par- ticular emphasis on improving the international governance of natural resources. Diamonds were critical to the tremendous economic success of Botswana, but also to the social catastrophe that engulfed Sierra Leone. The Kimberley process of tracking diamonds is intended to cur- tail rebel organizations' access to diamond revenues. The "publish what you pay" initiative, launched by the nongovernmental organization Global Witness, is intended to increase the transparency of natural re- source revenues to governments. Transparency is, in turn, an input into enhanced domestic scrutiny of how such revenues are used. If rebel fi- nances can be curbed and citizens come to believe that resources are being well used, civil war will be less likely. A third element in a pack- age of improved international governance of natural resources is to x FOREWORD cushion the price shocks that exporters commonly face. Price crashes have been associated with severe recessions that directly increase the risk of civil war and have sometimes destabilized economic manage- ment for long periods. At present the international community has no effective instrument to compensate for these shocks. International collective action has seldom looked so difficult, but the cost of failure will be measured in violence and poverty. Nicholas Stern Senior Vice-President and Chief Economist The World Bank April 2003 xli The Report Team T HIS REPORT WAS PREPARED UNDER THE SUPERVISION OF Nicholas Stern, chief economist and senior vice-president. It was written by a team led by Paul Collier (director, Develop- ment Research Group) and consisting of V. L. Elliott, Havard Hegre, Anke Hoeffler, Marta Reynal-Querol, and Nicholas Sambanis. The report builds on research by the Economics of Civil War, Crime, and Violence project in the World Bank Development Research Group. The project was initiated and directed by Paul Collier and has been ongoing since 1999. Ibrahim Elbadawi, Havard Hegre, Marta Reynal-Querol, and Nicholas Sambanis were the project's core staff. In addition, the project has commissioned a large number of studies from researchers outside the World Bank. The project received funding from the Norwegian, Swiss, and Greek governments; the World Bank Post- Conflict Fund; and the World Bank Research Committee. The project has been collaborating with United Nations Studies at Yale, the International Peace Research Institute in Oslo, the Agence francaise de developpement, the Economic Commission for Africa, and the AA-rican Economic Research Consortium. Conferences have been held in Addis Ababa, Irvine, Kampala, New Haven, Oslo, Paris, Princeton, and Washington, D.C. Many of the project papers are referred to in the text of the report, and most of them are posted on the project's web site: http://econ. worldbank.org/programs/conflict. Selections of the papers have also been published as special issues of the Journal of Conflict Resolution, Defence and Peace Economics, the Journal of Peace Research, and the Journal of African Economies. A set of country studies is being prepared for publication under the supervision of Nicholas Sambanis. . i. THE REPORT TEAM We thank for their excellent work Polly Means, who did the graph- ics; Audrey Kitson-Walters, who processed the report; Alice Faintich, who edited it; and Susan Graham, who was in charge of production. The judgments in this policy research report do not necessarily re- flect the views of the World Bank's Board of Directors or the govern- ments they represent. xiv Acr onyms and Abbreviations CPIA country policy and institutional assessment DALY disability-adjusted life year DDR disarmament, demobilization, and reintegration ELN Ejrcito popular de liberaci6n (Colombia) ETA Euskadi ta azkatasuna (Spain) EU European Union FARC Fuerzas armadas revolucionarias colombianas (Colombia) GAM Gerakan Aceh Merdeka (Indonesia) GDP gross domestic product IDP internally displaced person IMF International Monetary Fund IRA Irish Republican Army LICUS low-income countries under stress MIGA Multilateral Investment Guarantee Agency MNC multinational corporation NATO North Atlantic Treaty Organization NGO nongovernmental organization OECD Organisation for Economic Co-operation and Development RENAMCO Resistencia nacional Mo,cambicana (Mozambique) RUF Revolutionary United Front (Sierra Leone) STD sexually transmitted disease UN United Nations UNHCR United Nations High Commission for Refugees UNITA Uniao Nacional para a Independencia Total de Angola (Angola) UXO unexploded ordinance WHO World Health Organization xv Ovrerview OST WARS ARE NOW CIVIL WARS. EVEN though international wars attract enormous global attention, they have become infrequent M S | and brief. Civil wars usually attract less atten- tion, but they have become increasingly com- mon and typically go on for years. This report argues that civil war is now an important issue for development. War retards development, but conversely, development retards war. This double causation gives rise to virtuous and vicious circles. Where de- velopmernt succeeds, countries become progressively safer from violent conflict, inaking subsequent development easier. Where development fails, countries are at high risk of becoming caught in a conflict trap in which war wrecks the economy and increases the risk of further war. The global incidence of civil war is high because the international community has done little to avert it. Inertia is rooted in two beliefs: that we can safely "let them fight it out among themselves" and that "nothing can be done" because civil war is driven by ancestral ethnic and religious hatreds. The purpose of this report is to challenge these beliefs. Let Them Fight It Out among Themselves? P ART I INVESTIGATES THE ECONOMIC AND SOCIAL COSTS OF civil war. The costs the active participants in combat bear ac- count for only a trivial part of the overall suffering. The damage from a war ripples out in three rings. The inner ring is the displace- BREAKING THE CONFLICT TRAP CIVIL WAR AND DEVELOPMENT POLICY ment, mortality, and poverty inflicted on noncombatants within the country, and this is the subject of chapter 1. The United Nations High Commission for Refugees is currently assisting more than 5 million in- ternally displaced persons as a result of civil war. Many of these dis- placed people are forced to move to areas where the partial immunity they have acquired to malaria is no longer effective, and so their mor- tality rate rises. By the end of the typical civil war incomes are around 15 percent lower than they would otherwise have been, implying that about 30 percent more people are living in absolute poverty. However, the end of a civil war does not end the costs arising from it. Many of the economic costs, such as high military expenditure and capital flight, persist for years after the conflict. So too do heightened mortality and morbidity rates. Approximately half of the loss of disability-adjusted years of life expectancy due to a conflict arise after it is over. These eco- nomic and health costs of conflict are not usually compensated by any postconflict improvements in economic policy, democratic institutions, or political freedom. On the contrary, all three usually deteriorate. The typical civil war starts a prolonged process of development in reverse. Chapter 2 focuses on civil wars' spillover effects beyond the country. The second ring of suffering affects neighboring countries. Refugees stream across borders carrying and spreading the infections to which they have been exposed; for example, for every 1,000 international refugees the host country sees around 1,400 additional cases of malaria. Neighboring economies also suffer in other ways: growth rates are sig- nificantly reduced and neighbors increase their military expenditure in a chain reaction of local arms races. Often the costs of a civil war to the combined neighboring countries are of the same order of magnitude as the costs to the country itself. Through all these routes civil war is a re- gional public bad. The outer ring of suffering is global. Civil war creates territory out- side the control of any recognized government. One major use for this territory is to produce and transport drugs: 95 percent of the global production of hard drugs occurs in countries with civil wars and the major supply routes run through conflict territories. A more speculative possible global cost of civil war is the current AIDS pandemic. Some evidence suggests that this was triggered by the rapid spread of a highly localized infection caused by mass rape during a civil war. A further global shock to which civil war has contributed is Al Qaeda. When in- ternational terrorism is conducted on a large scale, the organization 2 OVERVIEW needs a safe haven that can probably only be provided in territory out- side the control of any recognized government. Al Qaeda chose to lo- cate in Taliban-held territory in Afghanistan, even though most of its recruits were not Afghans. It also used the war in Sierra Leone to gen- erate profits from the trade in conflict diamonds and to store its wealth. The global mortality caused by hard drugs and international terrorism is a significant toll, but the wider social costs are immense. The World Bank estimates that the September 11 attacks alone may have increased global poverty by 10 million people. We have no reason to think that those who decide to embark on civil war-the active participants, especially a few leaders-take all this suf- fering of Dthers into account. Furthermore, many of these adverse ef- fects are highly persistent. The typical civil war lasts long enough, around seven years, but the damage persists well beyond the end of the conflict. Once disease has set in, a country may need many years of peace to revert to its preconflict morbidity and mortality rates. Simi- larly, once an economy has experienced a wave of capital flight and em- igration, this tends to continue even when the conflict is over. In addi- tion, the regional escalation in military expenditure can persist because of insufficient coordination to reduce it. In many cases most of the costs of a civil war occur only once it is over. Again, those who have the power of decision are unlikely to take these consequences into account. Thus, in practice, the attitude let them fight it out among themselves gives license to a few thousand combatants and a few dozen of their leaders to inflict widespread misery on millions of others. What Can We Do about Ancestral Hatreds? C AN THE INTERNATIONAL COMMUNITY DO ANYTHING TO reduce the global incidence of civil war? If violence is simply determined by ancestral ethnic and religious hatreds, outsiders can probably do little. Part II turns to the underlying factors that de- termine thie global incidence of civil war. Chapter 3 discusses what makes some countries prone to civil war. Of course, each civil war is different and has its own distinctive, idiosyn- cratic triggers, be they a charismatic rebel leader or a provocative gov- ernment action, but beneath these chance circumstances patterns are apparent. Some social, political, and economic characteristics systemat- 3 BREAKING THE CONFLICT TRAP CIVIL WAR AND DEVELOPMENT POLICY ically increase the incidence of civil war, and we show that ethnicity and religion are much less important than is commonly believed. Indeed, so- cieties that are highly diverse mixtures of many ethnic and religious groups are usuaLly safer than more homogenous societies. By contrast, economic characteristics matter more than has usually been recognized. If a country is in economic decline, is dependent on primary commod- ity exports, and has a low per capita income and that income is un- equally distributed, it is at high risk of civil war. This cocktail is so lethal for several reasons. Low and declining incomes, badly distributed, cre- ate a pool of impoverished and disaffected young men who can be cheaply recruited by "entrepreneurs of violence." In such conditions the state is also likely to be weak, nondemocratic, and incompetent, offer- ing little impediment to the escalation of rebel violence, and maybe even inadvertently provoking it. Natural resource wealth provides a source of finance for the rebel organization and encourages the local population to support political demands for secession. It is also commonly associ- ated with poor governance. Disputes often fall along ethnic and reli- gious divisions, but they are much more likely to turn violent in coun- tries with low and declining incomes. O NCE SUCH A COUNTRY STUMBLES INTO CIVL WAR, ITS RISK of further conflict soars. Conflict weakens the economy and leaves a legacy of atrocities. It also creates leaders and organi- zations that have invested in skills and equipment that are only useful for violence. Disturbingly, while the overwhelming majority of the population in a country affected by civil war suffers from it, the lead- ers of military organizations that are actually perpetrating the violence often do well out of it. The prospect of financial gain is seldom the pri- mary motivation for rebellion, but for some it can become a satisfac- tory way of life. This is a further reason why the participants in a civil war should not be left to fight it out among themselves. Some evidence suggests that decade by decade, civil wars have been getting longer. While this may be due to circumstances in individual countries, it more likely reflects global changes that have made civil wars easier to sustain by allowing rebel groups to raise finance and acquire armaments more easily. OVERVIEW The Ri4sPing Global Incidence of Conflict P -H E INCIDENCE OF CIVIL WAR HAS INCREASED SUBSTANTIALLY over the past 40 years. As this has been a period of unprece- _AFL dented global economic development, it might appear evident that development has not been an effective remedy for violent civil con- flict, but to make sense of the patterns we need to distinguish between different groups of countries This is the subject of chapter 4. Many developing countries have either already reached middle- income status or have policy and institutional environments that should put them on track to do so. Around 4 billion people live in such coun- tries. Currently, as a group, they face a risk of civil war four times as high as the negligible risk societies in countries of the Organisation for Economic Co-operation and Development (OECD) face; however, 30 years ago their risk was five times as high, so they are converging with the group of countries already in secure peace. Nevertheless, more than a billion people live in low-income coun- tries that have been unable to adopt and sustain policies and institu- tions conducive to development. On average, these countries have been in economic decline and have remained dependent on natural resources or other primary commodities. This group of countries face far higher risks: typically around 15 times as high as OECD societies. Indeed, these risks have been rising as economies have deteriorated. Forty years ago there were many fewer independent, low-income countries. Most low-income countries were under the imposed peace of colonialism or were fighiing liberation wars. As countries gained independence they started, in effect, to play Russian roulette with civil war risk. Many of them stumbled into conflict, and where this happened the conflict trap implied an even higher risk of further conflict. This is the group that increasingly accounts for the global incidence of conflict. Thus the overall trend in the global incidence of conflict is made up of two radically divergent components. For most of the world's popu- lation development has been significantly reducing risks, but a signifi- cant minority of people live in low-income countries that have not shared in development. For them the risks have been increasing. If these two opposing forces persist, the global incidence of conflict will not continue to rise indefinitely, but neither will development se- cure global peace. The world will find itself stuck with a self-sustaining incidence of civil war, determined predominantly by the large and per- 5 BREAKING THE CONFLICT TRAP. CIVIL WAR AND DEVELOPMENT POLICY sistent pool of nondeveloping, low-income countries. These countries will account for a small and diminishing share of global income, but they will be responsible for a high share of the regional and global spillovers from civil war. P >wART III TURNS TO THE POLICIES THAT MIGHT BE EFFECTIVE IN reducing the global incidence of conflict. Some of these require action at the national level and others at the global level. Until recently, superpower rivalries made an international policy toward civil war unrealistic. Developing country governments lined up on one side or the other, and many rebel movements could count on some degree of cover from the opposing superpower. Therefore the question of what international responses were appropriate has only been worth posing in the past decade. Because asking the question had made little sense, the analysis to guide post-Cold War responses was not in place. Relative to many other questions the analysis is still seriously incomplete, but we are no longer completely in the dark. We now know enough for a rea- sonable basis for action. Economic development is central to reducing the global incidence of conflict; however, this does not mean that the standard elements of de- velopment strategy-market access, policy reform, and aid-are suffi- cient, or even appropriate, to address the problem. At the most basic level, development has to reach countries that it has so far missed. Be- yond this, development strategies should look different in countries facing a high risk of conflict, where the problems and priorities are dis- tinctive. In addition, some policies that are not normally part of devel- opment strategy affect the risk of conflict, such as the presence of external peacekeeping forces, the tendency toward domestic military expendi- tures, and the design of political institutions. In designing a strategy for risk reduction a useful approach is to view all the interventions that sig- nificantly affect risk in an integrated way. For example, different inter- ventions are most effective at different phases, and so may best be se- quenced. Because different actors who are not used to working together determine the interventions, to date this has not been common practice. The global incidence of conflict is made up of four very different components, each of which needs a distinctive approach. This is the subject of chapter 5. The first is the relatively low risk of conflict that is 6 OVERVIEW faced by a large group of middle-income countries and by some low- income countries that are on track to becoming middle-income because of good policies and rapid growth. For this class of countries the main risk probably comes from sudden economic crashes, such as that Indonesia experience-d in the late 1990s. These crashes are in any case disastrous, and the heightened risk of conflict simply adds a further reason why both national and international action needs to be taken to avoid such shocks and to cushion them when they happen. This is the group of countries that has already participated in global growth. Accelerating their growth would make a modest, but significant, contribution to global peace. The second component of the global incidence of conflict is the much higher risk stagnant or declining low-income countries face. This group has basically been missed by development to date, and is in ef- fect locked in a game of Russian roulette in which the probability of war is dangerously high. Igniting development in this group would make a far more substantial contribution to global peace, but is diffi- cult because it has not been achieved to date. A particularly helpful as- pect of development for these countries would be to help them diver- sify out of dependence on primary commodity exports. The third component of the global incidence of conflict is the coun- tries currently in conflict. If the typical conflict could be shortened, then the global incidence of conflict would decrease significantly. Past internation-al interventions to shorten conflict have not been systemat- ically effective; however, some evidence suggests that conflicts can be shortened by squeezing rebel organizations of their sources of external finance. Yet in the absence of other interventions shortening conflicts is not particularly effective: countries in the conflict trap simply pass in and out of war more frequently. The final component of the global incidence of conflict is those countries rhat are in the first decade of postconflict peace. For this group the risks of further conflict are exceptionally high: approximately half will fall back into conflict within the decade. This is the area that probably has the most scope for effective international interventions to reduce the incidence of conflict. What is most likely required is a coor- dination of external military peacekeeping for the first few years with a buildup of large aid programs during the middle of the decade. Both military peacekeeping and aid could be made conditional on the rapid reform of government policies and institutions, so that by the end of the decade the sociery is reasonably safe from further conflict. We show that an integrated approach involving external military support, aid, 7 BREAKING THE CONFLICT TRAP CIVIL WAR AND DEVELOPMENT POLICY and policy reform could, over the course of two decades, take postcon- flict countries well out of the zone of high risk and reduce their risk of conflict to only a quarter of its initial level. No single intervention is decisive in reducing the global incidence of conflict; however, different measures complement each other and cu- mulate. Our simulation of a package of development measures suggests how the global incidence of conflict could be reduced to less than half of its present level. Some of the actions needed for such an improve- ment come from the governments of developing countries, and some require action at an international level. Chapter 6 sets out a specific agenda for international action. Because those who decide whether to initiate civil wars and whether to accept a settlement ignore the large and adverse spillovers from civil war, this gives the international community both the moral right and the prac- tical duty to intervene to prevent and shorten conflicts. International interventions have recently had some important suc- cesses, such as the launch of the Kimberley process to regulate trade in diamonds and the international ban on antipersonnel mines. We con- sider three further sets of interventions: aid, the governance of natural resources, and military peacekeeping. Aid has substantial potential for conflict prevention, particularly in postconflict settings, and in the past donors have probably not got their aid policies right. Aid has usually flooded in during the immediate postconflict period, when the country is prominent in the international media, and then rapidly tapered out. Based on our analysis, overall aid should have been larger during the first postconflict decade, but it should have gradually tapered in during the decade. There is also con- siderable scope to retarget aid toward low-income countries: the inter- national community has provided much aid to middle-income coun- tries where conflict risks are usually quite low. Natural resource endowments have the potential for poverty reduc- tion, but historically have often been associated with conflict, poor gov- ernance, and economic decline. Because the adverse effects of natural resources work through a number of routes, several distinct interven- tions could be helpful. One global objective might be to make securing finance more diffi- cult for rebel organizations. The Kimberley process has this objective, and it needs to be monitored. If it is successful, it could be replicated for some other commodities. If it is unsuccessful, the present voluntary agreement may need to be strengthened by legislation. There is also 8 OVERVIEW scope to supplement the tracking of commodities with the tracking of the financial flows that are their counterparts, and the international banking system is now rightly coming under pressure to provide more effective scrutiny of the transactions it administers. A further source of rebel finance is from ransoms and extortion. Obviously such activities are already illegal, but the scale of payments can probably be reduced by government action in the OECD countries in which targeted compa- nies are based. For example, the recent emergence of a market for ran- som insurance is probably undesirable in that it escalates payments. A final source of rebel finance is from illicit primary commodities, notably coca and opium. The current OECD regulatory environment makes territory outside the control of a recognized government extremely valu- able, and this clearly facilitates conflict. Many options for redesigning drug policy are available that would moderate this dangerous effect. A distinct reason why countries dependent on natural resources face problems is their exposure to price shocks. OECD governments, and indeed charities, have been good at responding to such photogenic shocks as earthquakes and hurricanes, but have utterly failed to respond to the much more severe shocks caused by price crashes. There is con- siderable scope for both the international financial institutions and bi- lateral donors to provide better cushioning of these shocks and to con- duct their commercial policies in such a way as to reduce price shocks in the firsi: place. A final reason why countries dependent on natural resources face problems is that their revenues are often used inefficiently or corruptly. The Monterrey consensus emphasized that both industrial and devel- oping country governments have responsibilities in this context. There is a case for a template of governance of natural resource revenues to which governments could choose to adhere. Such a template would in- clude transparency and effective scrutiny. It could potentially be used as a signal of reduced exposure to political risk, and so help to attract more reputable resource extraction companies to low-income environ- ments. The international financial institutions have a potential role here in aggregating revenues from the individual accounts of resource extraction companies and publishing the resulting estimates of revenue in a way that integrates the information with budgetary data. Especially in postconflict situations, government military spending tends to be excessive. High spending tends to increase risks rather than to contain them. Through powerful regional arms races, this high spending becomes a regional public bad. There is scope for regional po- 9 BREAKING THE CONFLICT TRAP CIVIL WAR AND DEVELOPMENT POLICY litical organizations to negotiate mutual reductions in spending. The international financial institutions may have a role here as honest bro- kers monitoring that countries actually implement agreed reductions in spending. Finally, and more speculatively, we consider the coordination of ex- ternal military interventions with aid and policy reform. We suggest that in many postconflict environments neither aid, nor policy reform, nor even new democratic political institutions can realistically secure peace during the first few years. External military intervention may be the only practical guarantor of peace. An effective sequence might be that large aid inflows are phased in during the middle of the postcon- flict decade, generating a growth spurt that may enable a substantial re- duction of the military presence. As the conflict-related aid program ta- pers out at the end of the decade, if the government has used the decade well to accelerate reforms, it should be in a position to sustain the rapid growth that can make the society safer. In securing a safer world, no single intervention is likely to be deci- sive. Conflict risk works through multiple channels, and so calls for a package of complementary solutions. Furthermore, most interventions take time to work. However, our simulations suggest that if action is taken now, by 2015-the timetable for the attainment of the Millen- nium Development Goals-more than halving the global incidence of civil war would be feasible. At present, reducing the global incidence of civil war is not included as a Millennium Development Goal. Yet both because war is so power- fully development in reverse and because peace is a fundamental good in its own right, it is surely appropriate as a core development objec- tive. It is also much more readily monitorable than any of the other goals and, indeed, is already monitored by the authoritative Swedish International Peace Research Institute. The case for treating the halving of the incidence of civil war as a Millennium Development Goal is the same as that for the current goals: explicit commitments help the in- ternational community to sustain collective action. Because the risk of war is so heavily concentrated in the minority of developing countries we have referred to as "marginalized," attaining the overarching goal of halving world poverty without having much impact on the incidence of conflict would unfortunately be entirely possible. The goal of halv- ing the incidence of civil war would help to focus efforts on those coun- tries and people who are at the bottom of the heap. 10 PART I CRY HAVO(C° WHY (CWYL `.AR MATTERS Cry havoc and let slip the dogs of war. Shakespeare, Henry IV CIVIL WAR DIFFERS RADICALLY FROM BOTH IN- 1,000 combat-related deaths, with at least 5 per- ternational war and communal violence. Unlike cent on each side. There are many other forms international war, it is fought outside any struc- of group violence, such as protests, riots, and ture of rules and entirely within the territory of pogroms, but we do not consider them here. the society. Unlike communal violence, it im- The perpetrators of civil war usually adopt plies a r ebel organization equipped with arma- the rhetoric that the war is a necessary catalyst ments and staffed with full-time recruits. Such for social progress. Occasionally this is right, but rebel armies usually have little option but to live more typically war is an economic and social off the land. These features typically escalate the disaster for the affected country. Therefore, for social costs of civil war above the costs of either those who care about development, civil war is international war or communal violence. For a major problem. This is the focus of chapter 1: example, the same conflict between Eritrea and a theme of the chapter is that civilians, not the Ethiopia generated both a civil war and, follow- active combatants, suffer the main adverse con- ing Eritrean independence, an international war. sequences of civil war, and that many of these As a civil war the conflict lasted for 30 years and consequences accrue long after the war is over. was ended only by military victory. As an inter- Hence the people who determine whether war national war the conflict was subject to the full occurs are likely to ignore much of its adverse panoply of international mediation and ended consequences. Furthermore, civil war has severe swiftly in a negotiated settlement. To analyze consequences that spill over regionally and glob- civil war we need to know what we mean by it. ally, and civil war is not just a problem for the We ado pt a precise but conventional definition: countries directly affected. Thus the attitude civil war occurs when an identifiable rebel or- "let them fight it out among themselves" is not ganizati,on challenges the government militarily just heartless, it is foolish. This is the subject of and the resulting violence results in more than chapter 2. During a civil war a society diverts some of its Most of the costs of civil war accrue from these resources from productive activities to violence. destructive activities. The power of the gun dis- As a result, the society loses twice over. The di- places civil rights. Men with guns, from both verted resources are lost to productive activity, rebel and government forces, can steal, rape, and analogous to the loss from what economists call murder with impunity. Behind this veil of havoc, rent-seeking. Because much of the increase in the localized collapse of order extends impunity military spending is on government forces paid to criminal anid other antisocial behavior. The for out of the government budget, resources are primary response to the fear of theft, rape, and disproportionately diverted from government pro- murder is flight. People try to shift their assets vision of useful public goods, such as health care to safety, and they themselves flee. This flight and policing. However, whereas rent-seeking ac- in turn creates massive problems, especially for tivities are simply unproductive, the increase in health, as people are pushed into areas where they violence is harmful. One part of society is pro- lack immunity to disease. They then carry these ducing while another part is destroying. diseases with them, infecting host populations. CHAPTER ONE Ciril War as Development in Reverse , _ _ HIS CHAPTER FOCUSES ONLY ON THE EFFECTS OF civil war within the country that is directly affected. War has economic and social costs. Some of these accrue to the combatants, but many affect people who have no part in the decisions that create and ______ sustain the conflict. Furthermore, many of the ad- verse consequences of a conflict occur only once it is over and are prob- ably ignoted in combatants' decisions. We begin with the costs that arise during conflict and then turn to the legacy effects. Costs dluring Conflict T1 tHIS SECTION DISTINGUISHES BETWEEN THE ECONOMIC and social costs of conflict. Economic Costs During a civil war a society diverts some of its resources from produc- tive activities to destruction. This causes a double loss: the loss from what the resources were previously contributing and the loss from the damage that they now inflict (figure 1.1). The first loss can to some extent be quantified, as governments in- crease their military expenditure during civil war, and this directly re- duces economic growth. During peacetime the average developing country, oefined as a country with less than US$3,000 per capita gross 13 BREAKING THE CONFLICT TRAP CIVIL WAR AND DEVELOPMENT POLICY U.S. dollars 3,000 - Before civil war El After civil war 2,000 X," K ---1S_ Pen El Salvador Nicaragua Ango0a [em. Rep. Bunrndi of Congo Source Sambanis (2003) domestic product (GDP) in 1995, spends about 2.8 percent of GDP on the military. During civil war, on average, this increases to 5 percent. This is likely to cause a decrease in other public expenditures such as those on infrastructure and health. The decrease in the supply of such public goods has consequences for incomes and social indicators, and here we focus on the effects on income. Before taking any of the de- structive effects of military activity into account, we can estimate its consequences for crowding out productive expenditures. Knight, Loayza, and Villanueva (1996) quantify the costs to growth of military spend- ing during peacetime. Their simulations suggest that the additional 2.2 percent of GDP spent on the military, sustained over the seven years that is the length of the typical conflict, would lead to a permanent loss of around 2 percent of GDP. Of course, the increase in government military spending is only part of the diversion of resources into vio- lence. The resources rebel groups control are also a diversion from pro- ductive activities. However, the main economic losses from civil war arise not from the waste constituted by diverting resources from production, but from the damage that the diverted resources do when they are used for violence. The most obvious cost arises from the direct destruction of infrastruc- ture. During the war rebel forces target physical infrastructure as part of their strategy. The main targets are the enemy's communication and support lines, such as telecommunications, airports, ports, roads, and 14 CIVIL WAR AS DEVELOPMENT IN REVERSE bridges. In addition to this strategic destruction of key infrastructure, rebels and government soldiers loot and destroy housing, schools, and health facilities. An example is Mozambique (Briick 2001), where about 40 percent of immobile capital in the agriculture, communica- tions, and administrative sectors was destroyed. The prewar transport system had been a large foreign income earner, as goods were trans- ported from and to the neighboring states of Malawi, South Africa, Swaziland, and Zimbabwe, but 208 out of 222 units of rolling stock were lost or badly damaged between 1982 and 1989. Similarly, during the war in Liberia in the mid- 1990s all major infrastructures were dam- aged and looted. Monrovia, the largest port, suffered major damage during the first few months of the war, most of the electricity generat- ing capacity of the Liberian Electricity Corporation was destroyed, and looting rernoved much of the distribution and transmission systems. Infrastructure is an important determinant of economic growth (Can- ning 1998), and so destruction of infrastructure on such a scale is bound to reduce incomes. Probablv a more substantial cost arises from the fear that violence in- evitably generates. Frightened people flee from their homes. They also tend to lose the few assets they possess. For example, in a survey of households in Uganda, Matovu and Stewart (2001) found that two- thirds of respondents had lost all their assets. Their houses were bombed or unroofed; their household belongings, such as bicycles and furniture, were lootecL; and their cattle were stolen by soldiers. In Mozambique less than a fifth of the recorded 1980 cattle stock remained by 1992. Cattle were lost because of direct rebel activity, that is, rebels stole them to feed their troop:s and killed them to spread terror, and because of indirect ef- fects of warfare, namely, a lack of feed and veterinary attention during the war. Faced with the prospect of such losses, people try to protect their assets by shifting wealth abroad (Collier, Hoeffler, and Pattillo 2002). Prior to conflict the typical civil war country held 9 percent of its private wealth abroad. By the end of the civil war this had risen to an astonishing 20 percent, so that more than a 10th of the private cap- ital stock had been shifted abroad. Even this probably underestimates the extent of overall capital flight, for example, cattle may be moved to neighboring countries and sold. The disruption of civil war shortens time horizons and the displace- ment severs family and community links. Both weaken the constraints on opportunistic and criminal behavior. For example, during the Rus- 15 BREAKING THE CONFLICT TRAP CIVIL WAR AND DEVELOPMENT POLICY Cambodia Rwanda Thirty years of warfare all but destroyed most forms During the genocide, social capital atrophied as of social capital in Cambodia. During the Lon Nol the country, communities, and faiiilies fell prey to regime, traditional sources of social capital were hatred and violence. Yet integrative forms of social severely eroded throughout Cambodia. Many vil- capital increased within families fighting for sur- lages were forced to reallocate or were split as a vival; among individuals attempting to save or res- result of warfare, bombing, and Lon Nol recruit- cue Tutsi; and in the small Muslim community ment. Within villages exchange slowed, and soli- within Rwanda, which never took part in the geno- darity around the temple dissolved. cide. Strong, exclusionary social capital also emerged The Khmer Rouge ushered in another era of or- within Hutu extremism, with extremely negative ganized violence that included systematic attacks on ramifications for those excluded, showing that vio- traditional Cambodian society norms, culture, reli- lence can coexist with, or be the result of, strong gion, organizations, networks, and even the family. bonding social capital among its perpetrators. Community and family members were encouraged Once the killing began, Hutu killed not only to spy on and report on each other, which destroyed Tutsi unknown to them, but also their neighbors trust and planted the seeds of deeply rooted fear. A and, in some cases, even family members. These in- war against class distinctions was waged, as at- discriminate yet intimate killings led to the disinte- tempts to level economic status were instituted by gration of communes and families and fragmented making everyone an unpaid agricultural laborer. By social cohesion in general. High levels of social cap- destroying all social, political, and economic in- ital existed both vertically and horizontaLly among stitutions in this extreme communistic experiment, Hutu ranks, while bridging social capital that linked the brutal Khmer Rougc regime transformed and Hutu with Tutsi was all but eliminated. depleted what little social capital had remained from the Lon Nol period. Source CoUetta and Cullen (2000). sian civil war of 1920 the town of Nikolaev was in limbo between White and Red occupation for two days. During those two days local crooks chopped down all the trees lining the main avenue and stole the wood (Figes 1996). During the Rwandan genocide of 1994, those with assets faced a greater risk of being murdered (Andre and Platteau 1998). Colletta and Cullen (2000) analyze the relationship between vi- olent conflict and the transformation of social capital using four case studies: Cambodia, Guatemala, Rwanda, and Somalia (see box 1.1). In response to heightened opportunism and uncertainty, people invest less and retreat into those subsistence activities that are less vulnerable. For example, in Uganda during the long period of social chaos the share of the subsistence sector increased from 20 percent of GDP to 36 percent. 16 CIVIL WAR AS DEVELOPMENT IN REVERSE Investigators have used both econometrics and case studies to esti- mate the overall effect of civil war on the economy. An econometric study finds that during civil war countries tend to grow around 2.2 per- centage pomts more slowly than during peace (Collier 1999). Hence after a typical civil war of seven years duration, incomes would be around 15 percent lower than had the war not happened, implying an approximately 30 percent increase in the incidence of absolute poverty. The cumulative loss of income during the war would be equal to around 60 percent of a year's GDP. Note that this is much larger than the loss directly catised by the resources wasted on extra government military spending, which suggests that most of the costs of war are due to the ad- verse effects of violence rather than simply to the waste of resources. Stewart, Huang, and Wang (2001) survey data from about 18 countries affected by civil war. For the 14 countries whose average growth rates of gross national product per capita could be calculated, the average annual growth rate was negative, at -3.3 percent. Furthermore, they found that a wide range of macroeconomic indicators worsened during the conflict: in 15 countries per capita income fell, in 13 countries food production dropped, in all 18 economies their external debt increased as a percent- age of GDI' and in 12 countries export growth declined. Social Costs The most direct human effects of civil war are fatalities and population displacements. In the modern civil war the composition of victims dif- fers radically from the wars of the early 20th century, in that the impact has shifted from military personnel to civilians. At the beginning of the 20th century about 90 percent of the victims were soldiers, but by the 1990s nearly 90 percent of the casualties resulting from armed conflict were civilian (Cairns 1997). To some extent the rise in civilian casualties is a consequence of new military practices. Rebel recruitment strategies are now commonly co- ercive, so people flee to avoid recruitment. For example, in response to a recent rebel attack in rural Nepal, "About 35,000 people (out of a population of 75,000) have left the district, mainly young men mov- ing to India to avoid being forcibly recruited by the Maoists" (Holt 2003, p. 23). Furthermore, the military sometimes deliberately targets civilians to create forced migration. Azam and Hoeffler (2002) analyze 17 BREAKING THE CONFLICT TRAP CIVIL WAR AND DEVELOPMENT POLICY Millions of people 20 Total refugees 10 5 Internally displaced person, retumees, asylum seekers 0- 1 1 I 1M92 1967 2972 1977 1902 n107 1992 1997 Source UNHCR (2002). the different motives for targeting civilians in internal wars. On the one hand, soldiers may terrorize civilians because they need loot to augment their resources. An alternative hypothesis suggests that terrorizing the civilian population plays a direct military role. Using cross-country data from Sub-Saharan Africa they find support for the latter hypothe- sis. Civilians are targeted mainly because the displacement of large frac- tions of the civilian population reduces the fighting efficiency of the enemy, as they cannot hide and obtain support as easily. Forced migration broadly consists of two groups: refugees and inter- nally displaced persons (IDPs). The United Nations High Commission for Refugees (UNHCR) provides data on "people of concern," that is, people who received assistance from the organization. Approximately 86 percent of people of concern are refugees and IDPs. In 2001 the UNHCR assisted about 12 million refugees and about 5.3 million IDPs worldwide (figure 1.2). Table 1.1 lists the world's major refugee populations by country of origin and the countries with the largest numbers of IDPs. Afghanistan is one of the countries with the largest percentage of displaced popula- tions in the world. During the 1990s almost 40 percent of the Afghan population was living in refugee carmps in asylum countries, mostly in Iran and Pakistan. Liberia and Sudan also have exceptionally large per- centages of their population either living as refugees in asylumn coun- tries or internally displaced (see box 1.2). Such displacements have huge consequences for these individuals and their societies. 18 CIVIL WAR AS DEVELOPMENT IN REVERSE Refugees IDPs Country of origin Years of war Main asylum countries Total in millions Total in millions Afghanistan 1978-2002 Iran, Pakistan 3.80 1.200 Burundi 1991-ongoing Tanzania 0.55 Iraq 1985-92 Iran 0.53 Sudan 1983-ongoing Central African Republic, Democratic Republic of Congo, Ethiopia, Kenya, Uganda 0.49 Angola 1992-2002 Zambia, Democratic Republic of Congo, Namibia 0 47 0.202 Somalia 1988-92 Ethiopia, Kenya, United Kingdom, United States, Yemen 0.44 Bosnia-Herzegovina 1992-95 Serbia and Montenegro, United States, Sweden, Denmark, Netherlands 0.43 0.438 Democratic Republic of Congo 1997-99 Burundi, Republic of Congo, Rwanda, Tanzania, Zambia 0.39 Vietnam 1960-75 China, United States 0.35 Eritrea r 1998-2001 Sudan 0.33 Colombia 1984-ongoinig n a - 0 720 Sii Lanka 1983-2002 n.a. - 0.683 Azerbai)an 1991-ongoing n a. - 0.573 Russia 1999-ongoing n a - 0.443 Georgia 1991-ongoing n a. - 0.264 Serbia and Montenegro (former FR Yugoslavia) 1991-99 n.a. - 0 263 Liberia 1992-1996 n.a. - 0.196 n a Not applicable Source UNHCR (2002) Legac UEses a CUA0 MiEl T O0 THE EXTENT THAT CIVIL WAR HAS A POLITICAL RATIONALE it is as a catalyst for social progress. A rebel leader might hon- orably accept the terrible costs incurred during war as a high but necessary price to pay for future improvements, but far from being 19 BREAKING THE CONFLICT TRAP CIVIL WAR AND DEVELOPMENT POLICY Liberia: A Nation Displaced pendent on aid. Violence has visited every one of While official estimates for IDPs and refugees peak at Liberia's 15 counties and territories. The official 70 percent of the population, it is hard to see how a numbers therefore disguise the fact that all Liberi- single Liberian family has not been displaced at some ans, from wealthy merchants in Monrovia to peas- stage by Liberia's civil war. Displacement in Liberia ant farmers up-country, have had their way of life has been driven by the conflict, with all its devastat- severely disrupted and, more often, destroyed. Most ing political, economic, and social consequences. of the refugees are in Cote d'Ivoire. The human costs of displacement are apparent, but hard data have been difficult to collect, and of- Sudan ficial statistics do nor tell the whole story. The last "Civil war, primarily between northern and southern rehable census took place in 1974, with a subse- Sudanese from 1955 through 1972 and from 1983 quent exercise in 1984 remaining unfinished. The to the present, has left more than 1.5 million south- baseline population figure most often used of 2.6 ern Sudanese dead and a majority of the remaining million at the outset of fighting is an extrapolation southern Sudanese population uprooted. The mas- based on previous birth and mortality rates. Of this sive level of often deliberate death and displacernent estimated population, at least 750,000 have fled as has been one of the century's largest, yet least-recog- refugees to neighboring or distant countries, an ad- nized, tragedies.... The vast majority of the south- ditional million have been displaced internally, and ern Sudanese forced from their homes are internally an estimated 100,000 to 150,000 have died or been displaced. At the end of 1996, USCR reported that killed. The United Nations reports that 1.8 million, as many as 4 million Sudanese, mostly southerners, virtually the entire remaining population, are de- were internally displaced throughout Sudan." Source" Liberta Scott (1998); Sudan Ruiz (1998, pp 139, 141). a catalyst for beneficial change, civil war typically leaves a persisting legacy of poverty and misery. IE¢GU0n0un and[o[IMC L5u& egacy Several of the adverse economic effects of civil war are highly persistent. Recall that during civil war military expenditure rises as a percentage of GDP from 2.8 to 5.0 percent; however, once the war has ended, mili- tary expenditure does not return to its former level. During the first postconflict decade the average country spends 4.5 percent of GDP on the military. The government often presents the modest reduction in military spending from its wartime level as a peace dividend, but a more accurate way of viewing postconflict military spending is to see it 20 CIVIL WAR AS DEVELOPMENT IN REVERSE as a major hidden cost of conflict, hidden because abnormally inflated military spending persists long after the conflict is over. Cumulatively over the first decade of peace some 17 percent of a year's GDP is lost in increased military spending. This is far from being the only postconflict cost of war, but alone it is substantial: during the typical conflict the total income loss cumulates to around 60 percent of a year's GDP. A second cost during conflict is capital flight. Recall that during war capital flight increases from 9 percent of private wealth to 20 percent. By the end of the first decade of postconflict peace capital flight has risen furthei to 26.1 percent. Far from realizing a peace dividend here, the country experiences a war overhang effect. A possible reason for this is that asset portfolios can only be adjusted gradually, so that even by the end of a war the typical portfolio may not have fully adjusted to the political uncertainty created by the war. Once a country has experi- enced a civil war it is much more likely to see further conflict, so that even though peace is an improvement, risk levels do not return to their preconflict level. Thus even once peace has returned, people may still wish to move more of their assets abroad. Capital repatriation requires more than just peace. The same is true, only much more powerfully, for human flight. Civil war gives a big impetus to emigration, but some of these emigrants, especially those in industrial countries, then provide a postconflict channel for further emigration. A third persistent adverse legacy is the loss of social capital. Civil war can have the effect of switching behavior from an equilibrium in which there is an expectation of honesty to one in which there is an expecta- tion of corruption. Once a reputation for honesty has been lost, the in- centive for honest behavior in the future is greatly weakened. Clearly civil war is not the only way in which a society can become corrupted, but the point is that the costs inflicted by corruption are likely to per- sist long after the conflict is over. For civil war to have some redeeming features, the most hopeful areas would be policies, political institutions, and human rights. The impact of civil war on each of these can, to an extent, be measured. With respect to policy we use a measure adopted by the World Bank, the country policy and institutional assessment (CPIA). The CPIA is an assessment on a 5-point scale of economic policy in four areas-macro- economic, structural, social, and public sector management-with a higher score indicating better policies. While what constitutes "good" policies can be controversial, consensus on the recognition of bad poli- 21 BREAKING THE CONFLICT TRAP. CIVIL WAR AND DEVELOPMENT POLICY cies is wider, and, unfortunately, civil war countries tend to be at this end of the spectrum. Those low-income countries that are neither at war nor in the first decade of postwar peace have, on average, a CPIA score of 2.75. For those countries that have had a civil war and re- established peace, we can track whether the war served as a catalyst for improvement. On average, during the last five years prior to war the CPIA for these countries was 2.56. During the first postconflict decade it averaged only 2.29. Although the numbers are close together, they actually reflect a substantial deterioration in policies. All four policy areas are worse in postconflict societies: their macroeconomies are less stable, their structural policies such as trade and infrastructure are less conducive to growth, their social policies are less inclusive, and their public sectors are less well managed. Civil war is thus not normally a catalyst for policy improvement, but rather for policy deterioration. With respect to the extent to which political institutions are demo- cratic, we use the standard political science index, polity IV. This is a 10- point scale, and as with the CPIA the bottom end of the range is proba- bly more clear-cut than the top. The typical low-income country that is neither at war nor in postwar peace has a score of 2.11, while countries in the first decade of postwar peace average a score of only 1.49. Thus again, on average, civil war leads to a deterioration rather than an im- provement in political institutions. A related measure is an index of po- litical freedoms compiled by Freedom House. This is a 7-point scale in which, unlike the other indexes, a low score is better than a high score. The comparable numbers are 4.79 prior to conflict and 5.66 postcon- flict. Hence civil war again leaves a legacy of reduced freedom rather than increased freedom. A further new measure combines a democracy score and an autocracy score and ranges from 0 to 20. Countries are typically democratic if they score 15 or above. Five years after the end of the civil war the average score on this index is only 8.1 (Doyle and Sambanis 2003). Some evidence suggests that postconflicr, countries tend to revert approximately to their preconflict political conditions (Sambanis 2000). As chapter 3 shows, in reality the political legacy of civil war is far worse than these indicators imply. Once a country has had a civil war it is far more at risk of further war. This is partly because war leaves the society divided and embittered, and partly because war creates interests that favor continued violence and criminality. As a result, people's fears of a relapse into further conflict may dominate the postconflict eco- nomic landscape. 22 CIVIL WAR AS DEVELOI'MENT IN REVERSE The overall economic and political legacy from civil war is thus suf- ficiently adverse that rapid recovery is unlikely. Collier (1999) finds some evidence for a war overhang effect, whereby after short wars the economy continues to have exceptionally low growth. This is consistent with the capital flight story, in that a short war may not give people enough time to shift their assets abroad, so they continue with capital flight even after the war is over. Chapter 5 discusses the postconflict economic recovery in detail with a focus on national and global poli- cies. The pace of postconflict recovery is highly dependent on national policy choices and the scale and nature of international support. Re- covery is not an automatic process of bouncing back. Even in success- ful recoveries the process is slow. Consider, for example, Uganda, where recovery was unusually rapid, yet even by the late 1990s, 10 years after the end of the civil war, per capita income had barely regained its level of the early 1970s and the retreat into subsistence had barely been re- versed. At the household level, even though most respondents had been able to replace some of their assets, when interviewed 60 percent indi- cated that they were still worse off than before the war (Matovu and Stewart 2001). Social Legacy Civil War Increases Mortality Rates. Mortality rates only capture one dimension of the human consequences of conflict; however, they are a useful summary measure of the crisis and its impact. Mortality es- timates can be highly inaccurate, but they are often better and more easily captuced than other health indicators, which may be subject to different definitions and cultural interpretations (Keely, Reed, and Waldman 2000). Other human damage as a consequence of conflict in- cludes morbidity and psychological effects, but mortality rates have been one of the most easily and accurately measured indicators in emer- gency settings. The long-term effect of civil war on mortality can be investigated using both econometrics and case studies. A new econometric study in- vestigates the effect on infant mortality (Hoeffler and Reynal-Querol 2003). Unsurprisingly, the mortality effect depends on the duration of the conflict. Considering a rypical five-year war, the study finds that in- fant mortality increases by 13 percent during such a war; however, this 23 BREAKING THE CONFLICT TRAP: CIVIL WAR AND DEVELOPMEN'I POLICY rgFgn2. Dez D S bm5dn LSse tur Libenia Somalia Sudan Afghanistan Angola TDem. Rep. of Congo Ethiopia D El Infant mortality Uganda Cl Due to combatant's deaths E Adult mortality Nepal Iraq I _ 1- -20 0 20 40 so s0 Change in mortality rates, percentage points Source Guha-Sapir and Van Panhuis (2002). effect is persistent, and in the first five years of postconflict peace the infant mortality rate remains 11 percent higher than the baseline. Guha-Sapir and Van Panhuis (2002) collected intensive case study data on mortality rates following civil conflicts. They find that the im- pact on adult mortality is generally even worse than that on infant mor- tality (figure 1.3). The numbers indicate the percentage change in deaths per month from before the war for children under five years old and for the rest of the population. The figure compares the mortality rates of refugees and IDPs with the mortality rate of the country in the year before the conflict started (the baseline year). Among the cases listed in the table, 60 percent of the cases refer to refugees, 20 percent to IDPs, and 20 percent to residents of the country. Mortality rates were higher after conflict than before. While the rise in adult mortality might be expected to have occurred because of adults' greater exposure to the risk of combat death, few of these adult deaths are directly com- bat related. A comparison of these increases in mortality with the esti- mates of deaths as a direct result of combat reveals that the death of 24 CIVIL WAR AS DEVELOPMENT IN REVERSE Table L2 llKoali, rates among chRihhrn undeir five On refugee and ODP camps, seIee0id conficts Percentage of deaths Population sample and year Disease Baseline Conflict IDPs in Somalia, 1992 Measles 10 1 36.5 Diarrhea 20.0 39.0 Kurdish refugees in Iraq, 1991 Diarrhea 22.9 74 0 Sudaniese refugees in northern Uganda, 1994 Meningitis 0 6 0.2 Rwandan refugees in Zaire, 1994 Diarrhea 20 0 87.0 Bhutanese refugees in Nepal, 1992-93 Respiratory infections 26.2 41 4 Diarrhea 22.9 22.9 Residents in eastern Democratic Republic of Congo, 2000 Malaria 15 5 26.0 Diarrhiea 20.0 11.0 Soutrce Centre for Research on the Epidemiology of Disasters (2001) combatants is only a minor component of the overall rise in mortality. These numbers suggest that civil wars kill far more civilians even after the conflict is over than they kill combatants during the conflict. Table 1.2 compares the preconflict baseline for mortality rates among children under five years old with postconflict rates for selected diseases and conflicts. The numbers indicate the highly mortality rates caused by infectious diseases in refugee and IDP camps after war. Moving beyond mortality, useful summary measures are disability- adjusted life expectancy and disability-adjusted life-years (DALYs) com- piled by the World Health Organization (WHO). These measures take into account both years of life lost because of disease and injury and years of healthy life lost to long-term disability. Ghobarah, Huth, and Russett (2003) find that civil wars significantly reduce these aggregate measures of national health performance. They use information on the major 23 diseases for categories of the population distinguished by gen- der and five different age groups. They find important effects of civil war in increasing the postconflict incidence of death and disability caused by particular infectious diseases and conditions among different population subgroups. As an example, in 1999 WHO (2000) estimates that 8.44 million DALYs were lost as a direct effect of all wars that were ongoing at that time. However, that same year a further 8.01 million DALYs were lost as a result of civil wars that had ended during 1991-97, but had increased the incidence of persistent infectious diseases. Thus the 25 BREAKING THE CONFLICT TRAP. CIVIL WAR AND DEVELOPMENT POLICY legacy effect of civil wars on DALYs was approximately as large as the ef- fect during conflict (Ghobarah, Huth, and Russett 2003). Why are these health effects of civil wars so persistent? They affect people through the following two main channels (Ghobarah, Huth, and Russett 2003): o Channel 1: "technical regress," that is, changes in living conditions make staying healthy more difficult. Civil wars raise the exposure of the civilian population to conditions that increase the risk of dis- ease, injury, and death. o Channel 2: the government has less money in the budget to spend on public health. Civil wars produce longer-term negative conse- quences for public health by reducing the pool of available finan- cial resources for expenditures on the health care system. Table 1.3 summarizes how each of these channels affects health condi- tions during and after civil war ends. Time Channel 1: Technical regress Channel 2: Budget reduction During civil war Civil wars destroy the infrastructure needed to Civil wars reduce economic growth and divert maintain health care programs public expenditures from health care to military needs. Prolonged civil wars displace large populations, either internally or as refugees. Epidemic dis- eases are likely to emerge from crowding, bad water and poor sanitation in camps, while mal- nutrition and stress compromise people's im- mune systems. After civil war Civil wars reduce the efficient use of resources The economic legacy of civil war is to reduce that are allocated to public health, and these re- the level of income for a considerable period. ductions in efficiency extend into the post-civil This squeezes all forms of public expenditure. war period. Moreover, often refugees and inter- In addition, postconflict governments typically nally displaced people do not return to their maintain far higher levels of military expendi- original homes after the war ends, but remain ture than prior to the conflict, so that expendi- in makeshift camps for years. The population ture on health care continues to be accorded a continues to be exposed to conditions that in- lower priority than it would otherwise have crease the risk of infections had. Source Ghobarah, HRLth, and Russett (2003) 26 CIVIL WAR AS DEVELOPMENT IN REVERSE Table 1.4 HIV prevalence in i!ihry selected countries-and years Countu and 'ear H R'.f iaIenLe Caffir,bl''.Ii , i I Ugtindi I '' 1' ' - r Ghobarah, Huth, and Russett (2003) find that infectious diseases are the most important cause of the indirect deaths of civil war. Of these malaria is the most important, and the evidence suggests that all the age groups under 60 are affected by malaria. As AIDS is now such a common cause of death we examine the re- lationship between civil war and HIV/AIDS in more detail. Military re- cruits are typically young, sexually active men, often unmarried. Mili- tary personnel tend to have high rates of sexually transmitted diseases (STDs), including HIV: estimates indicate that rates among military personnel aie two to five times higher than among the general popula- tion, even during peacetime. When military personnel are stationed away from hiome, social controls in relation to engaging in sexual rela- tionships are lower and the risk of HIV infection is likely to be higher. Prostitution around army bases also increases the spread of infection. In addition, in times of war the risk of contracting HIV or other STDs may seem low relative to the risk of death in combat. Some figures for HIV prevalence in the military are available (table 1.4). No reliable figures are available for rebel forces, but they are likely to be at least as high as for the regular armed forces. Studies also find that the level of militarization increases the prevalence of HIV. One study finds that halving the number of men in the armed forces is associated with a reduction in the rate of seroprevalence among low-risk adults by about a quarter (Over 2003). However, HIV is not only spread through consensual intercourse, but also through gender-based violence. Regular soldiers and rebels force women to give sexual favors in exchange for "protection." Also the incidence of rape increases, often dramatically, during war, with refugees and displaced women and girls being particu- 27 BREAKING THE CONFLICT TRAP CIVIL WAR AND DEVELOPMENI POLICY "THE MATERNAL AND INFANT MORTALITY RATES urban areas of Matala, Chibia, Lubango, Lobito, are the worst in Africa, estimated in 1998 at 1,854/ Baia Farta and Benguela indicated that there is: (1) 100,000 and 166/ 1,000 live births, respectively. The very poor attendance of pregnant women; (2) a lack estimated national contraceptive prevalence rate is of knowledge about child spacing and sexuality is- very low (3 percent), and only 19 percent of women sues, among men and women; (3) little use of fam- have assisted deliveries. IDP women are known to ily planning methods; (4) little knowledge about be at higher risk of dying from pregnancy related STDs/AIDS; and, (5) an overall expectation of large causes due to lack of access to health services and family size. With regard to questions about forced life in stressful conditions. A survey conducted by sex, 19 percent of women indicated they knew of UNFPA and the implementing agencies in 1999 women who were forced to have sex." (UN 1999, with 710 men and women in IDP camps and peri- pp. 42, 50). larly vulnerable. Carballo and Solby (2001) estimate that more than 200,000 women refugees were raped during the Rwandan war. Diseases have long been used as weapons of war, and AIDS is no ex- ception. HIV-infected soldiers made widespread use of rape as a system- atic tool of warfare in conflicts in Liberia, Mozambique, Rwanda, and Sierra Leone. According to Elbe (2002, p. 168): "There is documented testimony from female survivors of rape in Rwanda that the transmission of HIV was a deliberate act. According to some accounts, HIV-positive Hutu men would tell women they were raping that they would eventu- ally suffer an agonizing death from AIDS ... some of the rapists al- legedly said 'We are not killing you. We are giving you something worse. You will die a slow death.'" After war, the reintegration of ex-combatants into civil society poses a health problem because of their comparatively high levels of HIV prevalence (Carballo and Solby 2001). For example, a study of mass de- mobilization of Ugandan troops revealed devastating results for the rural areas where the demobilized HIV-positive soldiers retired. About 50,000 Uganda People's Defence Forces soldiers have been demobi- lized. Many of the soldiers with AIDS were retrenched to their home villages, AIDS being a major criterion for demobilization. However, hardly any counseling was undertaken to prepare such soldiers to avoid risky behavior as they reintegrated into civilian society. The destruction of social and physical infrastructure during wartime also contributes to the spread of HIV (see box 1.3). The health system 28 CIVIL WAR AS DEVELOPMENT IN REVERSE is less likely to detect the diseases associated with HIV/AIDS infection or to screen blood supplies. War also destroys the education system, which makes teaching about prevention more difficult. Finally, in most war or post:conflict situations women do not have a choice regarding breastfeeding their babies, thereby increasing the risk of infecting the next generation (Machel 2000). Psychological Damage of Civil War. Quantitative research on the ef- fect of civil war on mortality is feasible because mortality is easy to mea- sure. At the other end of the spectrum of measurability is the psycho- logical daniage done by civil war. Mental health services are typically highly inacdequate during conflict and postconflict situations, and so the evidence is much more fragmentary; however, the evidence that is available suggests, unsurprisingly, that the psychological effects of civil war are large and highly persistent (see box 1.4). l Bosnia major depression and 37% showed symptoms asso- "It is estimated that the recent war in Bosnia and ciated with the diagnosis of posttraumatic stress dis- , Herzegovina caused more than 250,000 deaths, order (PTSD). These results were anticipated by created more than 2 million refugees and internally clinical studies of Cambodian refugees in the United displaced persons, and wounded 200,000 in Bosnia States that revealed high rates of depression and and Herzegovina alone. Recent epidemiologic stud- PTSD." ies have revealed that the psychiatric morbidity as- sociate(d with mass violence in civilian and refugee Sierra Leone populations is elevated when compared with non- "One worker from Doctors without Borders ... in traumatized communities.... Clinical reports of Sierra Leone asserted that the severe 'psychosocial Bosnian refugees in treatment show rates of depres- problems ... may ultimately threaten the prospects sive symnptoms ranging from 14 to 21% and post- for long-term stability in society.' While an Inter- traumatic stress disorder (PTSD) symptoms rates agency Appeal for the CIS region claims that psy- ranging from 18% to 53%." chological trauma is 'deep' and will 'probably lead to irreversible psychological consequences.' Cambodia "Approximately 68% of the Cambodian refugees living on the Thai border displayed symptoms of I Sour-e Bosnia and Cambodia- Mollica and others (1999, p. 38), Sierra Leone McDonald (2002, p 6) 29 BREAKING THE CONFLICT TRAP. CIVIL WAR AND DEVELOPMENT POLICY Civil war survivors have lost family members, friends, livelihoods, and identity. Many of them are living in refugee camps. This experience of trauma suffered on a wide scale has psychological consequences: "In- timate exposure to brutality and subsequent displacement and civil dis- order leave individuals psychologically scarred and the intricate net- work of social interaction deeply torn" (McDonald 2002, p. 4). The experience of trauma continues after war. Moreover, living in a refugee camp or transitory settlement can constitute a "secondary wound." Most individuals will experience low-grade but long-lasting mental health problems (McDonald 2002). During the period following displacement the threat of violence is high, as are mortality and morbidity rates. These features, together with having to live in camps, contribute to the development of a prevailing sense of hopelessness that increases the traumatic experience. Traumas are of two types, single event traumas and ongoing traumas. Life in a refugee camp is an ongoing trauma (McDonald 2002). Clinical condi- tions such as depression and schizophrenia are linked to premature death in refugee populations. Ghobarah, Huth, and Russett (2003) find an indirect effect of civil wars on suicides of women of childbearing age. This probably reflects the trauma of rape. Longitudinal studies of survivors of terror, such as Holocaust survivors and survivors of the Pol Pot regime in Cambodia, show the existence of an intergenerational transmission of trauma. "The effects of massive trauma, tragically ... do not end with deaths of survivors and may continue into the lives of their children" (box 1.4). Lauoidmines. A final legacy of civil war, landmines, affects both eco- nomic activity and health (this section is based on ICBL 2002). Land- mines are, in effect, a negative capital stock that the society accumulates during conflict. They continue to kill and maim people long after the actual fighting has stopped. For 2001 the International Campaign to Ban Landmines recorded 7,987 landmine casualties in 70 countries, of which about 70 percent were civilians; however, as reporting is incom- plete, the campaign estimates that the total number is more likely to be between 15,000 and 20,000. In comparison with previous years, when the number of casualties was estimated at around 26,000 per year, this is a considerable improvement. The decrease in the number of landmine victims is due to the international ban of antipersonnel mines in 1997, 30 CIVIL WAR AS DEVELOPMENT IN REVERSE CAMBCODIA IS ONE OF THE MOST HEAVILY LAND- villages. Thus the threat of UXO and mines impedes mine and unexploded ordnance (UXO) contami- mobility, security, economic activity, and develop- nated countries in the world. Although the actual ment in several provinces, particularly in the north fighting stopped more than a decade ago in 1991, and northwest. With more than 80 percent of the on average more than two people are injured or country's population residing in rural areas, and killed by landmines in Cambodia every day. Even with 40 percent of these estimated to be living below though 166 million square meters of land were the poverty, line, mine action programs continue to cleared from 1992 through 2001 and a total of be of the highest priority in Cambodia's poverty re- 313,586 antipersonnel mines were found and de- duction policy. Official reports show that 173 pco- stroyed, all 24 provinces still have areas contami- ple were killed and 640 were injured in mine or nated by mines and UXO. In 2001, 6,422 vlltages, UXO incidents during 2001. The proportion of or 46 percent of Cambodian villages, had areas civilian casualties was about 95 percent, with chil- contaminated with mines and/or UTXO. Mine and dren particularly at risk (232, or 28 percent, of the UXO contamination restricts access to homes, agri- casualties were children). Clearing mines, caring for cultural land, pastures, water sources, forests, the injured and disabled, and providing mine edu- schools, dams, canals, markets, business activities, cation programs place a severe strain on Cambodian health centers, pagodas, bridges, and neighboring public expenditure. Souirce ICBL (2002) which resulted in the destruction of stockpiles as well as a drastic de- crease in the production and trade of landmines. In addition, mine- sweeping operations have been extremely successful in detecting and de- stroying mines in many countries. Yet as the example of Cambodia shows (see box 1.5), landmines continue to severely disrupt normal daily activihies and therefore constitute a serious obstacle to economic and social recovery. T tHIS CHAPTER HAS FOCUSED ONLY ON THE EFFECTS OF CIVIL war within the affected country and has clearly shown that most of the suffering inflicted by civil war accrues to noncom- batants who typically have no say in either whether the conflict is ini- tiated or whether it is settled. During the war income losses are severe and mortality and morbid- ity exhibit large increases. Even if a war is viewed as a costly investment 31 BREAKING THE CONFLICT TRAP: CIVIL WAR AND DEVELOPMENT POLICY for subsequent social progress, the costs during the conflict are typically so high that postconflict progress would need to be dramatic for subse- quent benefits to outweigh these costs. Yet the legacy effects of civil war are usually so adverse that they cannot reasonably be viewed as social progress. Many of the costs of the war continue to accrue long after it is over. For example, the country tends to get locked into persistently high levels of military expenditure, sees capital continuing to flow out of the country at an unusually high rate, and faces a much higher incidence of infectious disease. Even economic policies, political institutions, and po- litical freedom appear to deteriorate. Hence most modern civil wars are not remotely like the 19th century American civil war that ended slav- ery. Of course, finding some modern civil wars that can reasonably be seen as ushering in social progress is always possible, but these are surely the exceptions. On average, modern civil war has not been a useful force for social change, but has been development in reverse. 32 CHAPTER TWO Let; Them Fight It Out Amrong Themselves? HAPTER I SHOWED THAT FOR THE COUNTRIES directly affected, civil war is development in reverse, therefore preventing civil war is important for those C L concerned about the development of low-income tg Xcountries. However, the constituency for action to prevent war is potentially much larger than this, be- cause civil war has spillover effects for both neighboring countries and the entire international community. This chapter first considers the neighborhood effects and then turns to the global effects. Meighbiorhood Effects of Civil War P7=- EACEFUL COUNTRIES THAT ARE ADJACENT TO COUNTRIES EN- | egaged in civil war suffer from direct and long-term effects caused by the civil wars of their neighbors. Economic Spillovers Civil war, are not only costly for the countries in which they are fought, but for the entire region. Neighboring countries must usually accommodate large numbers of refugees, because the victims of war do not usually have the means to travel to countries further away from their home country, and, in any case, arrive on foot. For example, Pak- istan's buiden of accommodating more than 2 million refugees from Afghanistan is considerable. However, this direct burden is probably 33 BREAKING THE CONFLICT TRAP CIVIL WAR AND DEVELOPMENT POLICY A RECENT STUDY ATTEMPTS TO MODEL WHAT Both a past history of international war and the level of military spending governments choose as a military expenditure of neighboring countries are share of GDP. On average, governments spend 3.4 highly significant. Because of these neighborhood percentage points of GDP on the military, but this effects military spending is, in effect, a regional average varies widely and predictably. The largest in- public bad. Controlling for all these risks, miliLary crease is if a country is engaged in international war- governments stlfl spend more on the military, pre- fare, when spending rises by 2.5 percentage points. sumably because they are more susceptible to the Civil war raises spending almost as much, by 1.8 military lobby. A wave of democratization, such as percentage points, that is, the military budget in- occurred in the early 1990s, is thus a regional pub- creases by 50 percent. The risk of war also matters. Iic good. Finally, the end of the Cold War gradually, Each 10.0 percentage points on the risk of civil war but substantially, yielded a peace dividend, as mlli- raises military spending by 0.4 percentage points. tary spending fell by around 0.7 percentage points. Source Collier and Hoeffler (2002d) not the most important regional economic spillover. Other effects are on the military budget, the costs of transport, and the reputation of the region in relation to investors. An important route by which civil war affects neighbors is through regional arms races (see box 2.1). Both in response to the risk of a civil war, and especially once it has started, a government tends to increase its military expenditure sharply, rypically by around two percentage points of GDP. Unfortunately, one of the strongest influences on the level of military expenditure a government chooses is the level its neigh- bors have chosen (Collier and Hoeffler 2002d). This may be partly be- cause of a perceived threat, and partly because of norm setting and the emulation and rivalries of military leaderships. On average, if civil war leads a government with two neighbors to increase its military expen- diture by 2.0 percentage points of GDP, by the time the arms race is back to equilibrium, the neighboring countries will each have increased their spending by around 0.7 percentage points. In some situations war in one country directly increases the risk of war in neighboring countries. Chapter 3 discusses how the supply of armaments has sometimes spilled across borders and how rival govern- ments can find themselves financing each other's rebel movements. Such regional increases in conflict risk are compounded by their effects 34 LET TFIEM FIGHT IT OUT AMONG THEMSELVES' on regional military expenditure. A simulation of a regionwide increase of 10 percentage points in the risk of civil war predicts that regional military expenditure would end up rising by around 1 percentage point of GDP after arms race effects. A couni ry's neighbors' military spending has an adverse effect on the economic ,rowth rate. For each additional 1.0 percentage point of GDP that neighbors spend on the military, the growth rate is reduced by 0.1 percentage point.1 Recall that during and after a civil war the govern- ment directly affected raises its military spending by around two per- centage points and that this is liable to trigger a regional arms race that can persist long after the conflict is over. This alone can produce a small but widespread reduction in growth across an entire neighborhood. Some studies have attempted to estimate the overall effect of a neighboring conflict on growth. Having a neighbor at war reduces the annual growth rate by around 0.5 percentage points.2 Murdoch and Sandler (2002) show that civil war reduces not only the country's own growth rate, but also growth across an entire region. As most countries have several neighbors, this is a major multiplier effect of the economic cost of conflict. Recalling that the growth cost for the country itself is around 2.2 percent, a country with four neighbors is likely to inflict approximately as much economic damage on its neighbors during con- flict as it ctoes on itself. Neighbors' growth rates may be reduced for a number of reasons. In addition to the direct burden the refugee population poses and the ef- fect on military spending, trade is also disrupted, and this is a particu- larly severe problem for landlocked countries. For example, the war in Mozambiq1ue doubled Malawi's international transport costs and trig- gered an economic decline. Similarly, the war in the Democratic Re- public of C(ongo closed the river route to the sea for the landlocked Cen- tral African Republic. A further effect is that the entire region is regarded as riskier, which results in a negative reputation effect with investors. Social Spillovers The most immediate effect of civil war on neighboring countries is the arrival of thousands of refugees and their consequences for the popula- tion of the asylum countries. As refugees stay in asylum countries for 35 BREAKING THE CONFLICT TRAP: CIVIL WAR AND DEVELOPMENT POLICY long periods after the civil war ends, the social effects of civil war on asylum countries are also persistent. Among all the long-run, indirect effects of civil war, it causes most deaths in neighboring populations through infectious diseases, especially malaria. Large-scale refugee flows put people into crowded conditions in the asylum countries without access to clean water and food, making the camps a perfect environ- ment for the spread of infectious diseases. I3f[Qugees and 2Wi1aMa& A global effort to eradicate malaria was un- dertaken in the 1950s and 1960s. By the end of the 1960s these at- tempts to control the disease had faded in the face of the internal prob- lems of the countries where the incidence of malaria was the highest. The most relevant internal problem was civil war. Civil war has been a basic reason behind the observed increase in the incidence of malaria. Conflict affects the incidence of malaria both directly, when nonim- mune refugees come into contact with infected individuals when they flee through rural and rainforest areas to reach a foreign country, and indirectly, when conflict impairs active control measures (Montalvo and Reynal-Querol 2002). Figure 2.1 shows the flow and stock of refugees. Refugees stay in camps for a long time after civil wars end. Figure 2.2 shows the rela- tionship between the stock of refugees and ongoing civil wars. ~lgnre 2.l Tre aWas s3 t Do itNgec Re 2-@@202 Flow of refugees (millions) SSck of refugees (miNlions) 3 20 2 Flow of refugees 0- I- 10 -1 5 -2 - Stock of refugees -3 0 . I I I . I I 1 1951 1955 1959 1963 1967 1972 1975 1979 1983 2987 2991 1995 1999 Source UNHCR (2002). 36 LET TH{EM FIGHT IT OUT AMONG THEMSELVES? Figure 2.2 The stock of refugees and civil wars, 1951-2001 Civil wars oitgoing Stock of refugees (millions) 30- - 20 - 15 20 - Civil wars ongoing 10 10 5 - -~~~~~~Stc of refugees 0* i I iI 1951 1955 1959 1963 1967 1971 1975 1979 1983 1987 1991 1995 1999 Source Gled,isch and others (2002), UNHCR (2002) The number of countries reporting cases of malaria varies over time. In particular, China and India have a critical influence on the number of cases. China started to report officially to WHO in 1977. Initially it reported close to 4 million cases, but then a rapid decrease took place. Meanwhile India drove the growth of cases during the 1974-77 epi- demic period, when it accounted for close to 20 percent of the total cases of malaria in the world. For this reason we exclude China and India. A further reporting problem is Africa, where reporting is irregu- lar. We use last available data before a missing period and first available figures once reporting has resumed. Figure 2.3 shows the resulting series for cases of malaria compared with the number of refugees worldwide. The high correlation suggests that the increase in the incidence of malaria has been strongly affected by the rise in war refugees. In the Bonga refugee camp in Ethiopia in the mid- I 990s: "Malaria remains clearly the main cause of morbidity accounting for 17 percent of total caseload.... The profile argues for an active malaria control campaign in the camps to reduce morbidity" (Guha-Sapir and Forcella 2001, p. 34). Why might there be such a strong connection between refugees and the incidence of malaria? War leads to the movement of people. In gen- eral the anarchic situation caused by this social unrest and the military importance of paved roads force people to walk through unfamiliar rural areas and forests to avoid areas of military operations. If the civil war is 37 BREAKING THE CONFLICT TRAP CIVIL WAR AND DEVELOPMENT POLICY RgnIwe 2.3 Refgnaees aed cases of unanPls, 2R2-07 Onflerpolated number of cases (millions, not including China or India) Refugees (millioins) s6 40 12 30 20 - 10 Cases of malaria Refugees | 4 1962 1967 1972 2977 1082 1987 1992 1997 Source UNHCR (2002), WHO (1983, 1999) extended, this movement will end up in migration to a contiguous country as war refugees. Population movement caused by political con- flicts, rural population migrations, or natural disasters is potentially the most important factor in the transmission of malaria (conditional on the dynamics between the vector, the parasite, and the environment). While refugees move from cities to the borders, if the country has endemic malaria their probability of becoming infected by the malaria parasite increases as a result of their contact with locally immune rural populations and movement through remote areas where the vector is still predominant. The importance of contact with immune individu- als is critical. Repeated infection among individuals of rural endemic areas generates an immune response in the host that controls the infec- tion. This implies that the prevalence of malaria could be extremely high among rural populations despite a small number of reported cases. If the migrants have contracted malaria they will probably not be di- agnosed until they have arrived in the host country. These cases will therefore be counted as cases of malaria in the asylum country. The ex- istence of many migrants infected by the malaria parasite in the asylum country increases malaria transmission to citizens of the asylum coun- try and the contagion effect among the refugees themselves. This will happen if the asylum country has the vector, even though it may not originally be malaria endemic. In general, the concentration of refugees in camps where nonimmune and infected individuals live together in- 38 LET TliEM FIGHT IT OUT AMONG THEMSELVES? creases the risk of transmission conditional on the existence of the Anopheles mosquito. Research using annual data for 135 countries between 1960 and 1999 studies the effects of refugee flows from tropical countries with civil war to neighboring tropical countries (Montalvo and Reynal- Querol 2002). It finds that for each 1,000 refugees the asylum countries see 1,406 new cases of malaria. The size of the refugee population com- ing from tropical countries with a civil war thus has an important im- pact on malaria in the asylum countries. Preventing civil wars, especially in tropical countries, is therefore important for controlling malaria. The eflect of war-driven refugees on malaria is qualitatively similar to that of other refugees, an effect that is better known but is quanti- tatively more important. Refugees fleeing from droughts and famines do not have such a significant effect on the incidence of malaria for two reasons. First, the mechanism whereby refugees escaping from war be- come infected is because civil wars force people to walk through unfa- miliar rural areas and forests to avoid areas of military operations, but people displaced by famines and droughts do not have to avoid paved roads, so they are less likely to be exposed to the mosquito. Second, refugees fi-om war stay in asylum camps for long periods after the war ends, whereas once droughts and famines end refugees can quickly re- turn home. Montalvo and Reynal-Querol (2002) find that refugees from drought who come to tropical asylum countries have no signifi- cant effect on the incidence of malaria in the asylum country. Refugees and the Spread of HIV/AIDS. Refugees and other displaced populations are at increased risk of contracting HIV/AIDS during and after displacement because of poverty; disruption of family and social structures and of health services; increased sexual violence; and in- creased socioeconomic vulnerability, particularly among women and youth. Data on HIV prevalence in refugee camps are scarce; however, some examnples described in box 2.2 suggest the extent of HIV infec- tion in refugee camps in asylum countries. Ghobarah, Huth and Rus- sett (2003) find that the most important effect of civil war on neigh- boring countries is caused by HIV/AIDS, with the groups that are most affected being young children (who are infected by their mothers) and young and middle-aged adults. The average loss of healthy life for these groups ranges from roughly 2 to 10 years. 39 BREAKING THE CONFLICT TRAP CIVIL WAR AND DEVELOPMENT POLICY a. a 51 UE';X22 "THERE ARE CONCERNS ABOUT THE INCREASE IN rent drought have resulted in large-scale population the spread of HIV/AIDS, which rose from 8 in movements that included internal displacement, in- 1988 to over 13,500 cases in 2001. It is estimated flux of deportees from Ethuopia and returnees from that approximately 60,000-70,000 Eritreans are Sudan. Mobilisation of young men and women to currently infected with HIV, which could increase the military has also contributed by creating a social significantly ... with demobilisation and increased and economic environmcnt that is conducive for cross border movement" (UN 2002, pp. 10-11). the spread of HIV/AIDS. Intertwined with this is "The HIV/AIDS epidemic is perhaps the gravest violence against women, including rape and other health threat faced by Eritrea. The recent border physical trauma that can be experienced during conflict with Ethiopia (1998-2002) and the recur- conflict and displacement" (UN 2002, p. 27). C viP 9a2r SplDSoveir Civil wars are clustered in particular regions. In the 1980s there were several wars in Central America. In the 1990s there were several civil wars in, for instance, the African Great Lakes area, in Central Asia, and in the Balkans. Civil wars cluster for several reasons. They may share the same historical background: the former Yugoslavia's wars in Croatia in 1991, Bosnia in 1992-95, Croatia again in 1995, and Kosovo in 1998-99 all shared similar characteristics and were influenced by the ideology of greater Serbia and greater Croatia (Kalyvas and Sainbanis 2003). In the former Soviet republics wars clustered around the Cau- casus in the early 1990s, taking advantage of war- and region-specific physical and human capital (Zurcher, Kohler, and Baev 2002). Direct contagion may OCCUL The civil wars in the African Great Lakes region are examples of this, as recurrent wars in Burundi and Rwanda spilled over their borders in both directions and into the Democratic Re- public of Congo. The latter war also provoked interventions by Uganda and Zimbabwe. In all these wars Hutu-Tutsi antagonism was predomi- nant (Ngaruko and Nkurunziza 2002; Prunier 1995). This recurrent ethnic conflict crossed borders and lasted over time, being at the core of around seven episodes of civil war in the two countries. Countries em- broiled in civil war also often provide a safe haven for rebel groups of other countries. The wars in Liberia and Sierra Leone alternately served these purposes for the other country's rebel groups (Davies and Fofana 2002). 40 LET THEM FIGHT IT OUT AMONG THEMSELVES? Refugee flows caused by civil wars may also be destabilizing to the host country. During the war in the 1990s, Burundian rebels sought refuge in neighboring Tanzania and the Democratic Republic of Congo and recruited among the Burundi refugee population in Tanzania. The provinces in the Democratic Republic of Congo neighboring those two countries had the highest incidence of fighting and displacements of people (Ngaruko and Nkurunziza 2002). The economic spillover also increases the risk of civil war in neigh- boring countries (see chapters 3 and 4). MobA IEfects Of CIA W r C IVIL WAR IS NOT JUST BAD FOR THE NEIGHBORHOOD DUR- ing the past 30 years three major global social evils can reason- (L-Jably be ascribed in substantial part to side effects of civil war. The global cost of these social evils has already been astronomical, and they are proving highly persistent. They are hard drugs, AIDS, and in- ternational terrorism. The link from civil war to hard drugs is through both production and distribution. The cultivation of hard drugs, coca and opium, nowa- days predominantly requires territory that is outside the control of any recognizecl government. Where territory is under the control of an in- ternationally recognized government it can generally be prevailed upon to enforce anticultivation policies reasonably effectively. A by-product of civil war is that large rural areas cease to be under government con- trol. Currently some 95 percent of the global production of opium is in civil war countries. Not only is production concentrated in civil war territory, but distribution and storage channels rely on the lawlessness civil war generates. The link between civil war and the spread of AIDS within a nation and a region has already been discussed; however, the most far-reaching claim is that the origin of the global pandemic is a consequence of a par- ticular civLl war. The hypothesis, for which credible evidence exists, but which is fiar from proven, is that the conditions of war enabled what would otherwise have been a routine, localized outbreak to spiral out of control. Even if we attach only a small likelihood to this explanation being correct-say 10 percent-then one-tenth of the global cost of the AIDS pandemic should be added to the estimated global cost of civil war. 41 BREAKING THE CONFLICT TRAP- CIVIL WAR AND DEVELOPMENT POLICY The link between civil war and international terrorism has only re- cently become evident. Civil war provides territory that serves as a safe haven for terrorists, and the illegal products of conflict, notably dia- monds, are used both as a source of revenue and as a store of value. ( 3H Wair adill )Dmg PmalEllducM anDdl YvaMhMD g Table 2.1 shows the production of opium and coca by country from 1990 until 2001. Figures 2.4 and 2.5 group this information according to whether countries are in conflict, are postconflict, or are at peace. As the figures show, virtually all production throughout the period has been in conflict or postconflict countries. When civil war ended in Peru and intensified in Colombia production trends changed. While production rises sharply during conflict, it is not completely eliminated in postconflict situations. This accords with the discussion in 1-' -@.. s. (metric tons) Country 1990 1995 2000 2001 Opium Afghanistan 1,570 2,335 3,276 185a Colombia 71 88 88 Lao PDR 202 128 167 134 Mexico 62 53 21 71 Myanmar 1,621 1,664 1,087 1,097 Other Asian countries 45 78 38 40 Pakistan 150 112 8 5 Thailand 20 2 6 6 Vietnam 90 9 - - Total 3,760 4,452 4,691 1,626 Coca Bolivia 77,000 85,000 13,400 20,200 Colombia 45,313 80,931 266,161 236,035 Peru 196,900 183,600 46,258 49,260 Total 319,213 349,531 325,809 305,495 - Not available. a. Opium production in Afghanistan was reported to havc dropped by 95 percent from 2000 to 2001, but the UNODCCP (2003) reports that in 2002 opium production sur- passed its 2000 level (3,422 metric tons). Source UNODCCP (2002). 42 LET THEM FIGHT IT OUT AMONG THEMSELVES' Figure 2.4 Opium production, 1986-2001 Percentage of opium production Total of conflict and postconflict countries producing opium 100- 75 - ct countries producing opium 50 - 25 - Postconflict countries producig opium 0- ~ ~ ~ ~ ~ ~ ~ ~ ~~~~~~~~~~~~~I l l I I 1986 1988 1990 1992 1994 1996 1998 2000 Source UNOI)CCP (2002), Gleditsch and others (2002) Figure 2.5 Cocaine production, 1986-2001 Percentage of cocaine production 100 - Total of conflict and postconflict countries producing cocaine 75 - 50- 25 - Postconflict countries producing cocaine 0 4- I I l I l I 1986 1988 1990 1992 1994 1996 1998 2000 Souirce UNOIDCCP (2002), Gleditsch and others (2002) chapter 1 on the persistence of the loss of social capital and the criminal- ization of society. Production of drugs prevails long after civil war ends. Civil war does not only affect production. The routes traffickers fol- low from the origin country to American, Australian, and European markets also go through conflict and posrconflict countries. "Between 70% and 90% of the heroin found in European markets (both West 43 BREAKING THE CONFLICT TRAP CIVIL WAR AND DEVELOPMENT POLICY and East Europe), which has traditionally been trafficked along the so called 'Balkan' route (Afghanistan - the I.R. of Iran -Turkey - Balkan countries - West Europe) with indications in recent years of the de- velopment of an alternative route through Central Asia and Russia" (UNODCCP 2002, p. 11). Colombian trafficking organizations control the worldwide supply of cocaine, for which North America remains the principal destination. Africa, in particular, the conflict countries of West and Southern Africa, is increasingly used as a transit area for cocaine trafficking from South America to Europe. The production of hard drugs is concentrated in civil war countries for two main reasons. First, civil war creates territory out- side the control of a recognized government on which drugs can be cul- tivated. It also creates an environment in which many people can behave opportunistically with no cost, because the normal policing institutions are weakened and are unable to control illegal activities. Second, during civil war conventional economic opportunities are severely reduced. In- ternational crime, of which drug production and trafficking are the prime example, provides a rare instance of a new economic opportunity. Drug production affects the industrial world through two channels. First, the production of drugs in civil war countries is intimately related to their consumption in industrial countries. Not surprisingly, produc- tion and consumption trends follow the same pattern. Figure 2.6 shows the trends for opium production in Afghanistan and heroin seizures in Europe during 1980-2001. The consumption of illegal drugs results in thousands of deaths among young people in Australia, the United States, and Europe. Second, crime in the industrial world is intimately related to drugs. The supply of drugs has social consequences for the societies of the countries where it arrives that can be divided into three groups: drug use; drug-related crime; and indirect, adverse effects of use (Reuter 2001). Drug use directly results in dependency and such risky behavior as needle sharing. Drug-related crime includes both the violation of drug laws themselves and theft and violence. The indirect, adverse ef- fects of drug use include overdoses, suicides, abuse and discord within families, and poor school or work productivity. Moreover the govern- ment faces costs in terms of law enforcement and health expenditures. The per capita consumption of hard drugs is highest in industrial countries (figure 2.7, table 2.2). The pattern of consumption by conti- nent reflects production. Crossing the Atlantic is expensive and diffi- 44 LET THEM FIGHT IT OUT AMONG THEMSELVES? Figure 2.6 Opium production in Afghanistan and heroin seizures in Europe, 1980-2001 Opium produ(ction, metric tons Heroin seizures, metric tons 4,000 - 25 3,000 - - 20 3,000 - Opium production Heroin seizures - 15 2,000- 10 1,000 0 l- -0 19801 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 199i 9719 199 000 2001 'Soviet' period 'Warlord' period 'Taliban' period Source UNODCCP (2002) Figure 2.7 Estimates of annual opiate and cocaine use in the late 1990s Percentage (if population age 15 and above (log) 20 Opium 2 Cocaine 0u I Oceainiia Europe North South Asia Aftica Global America America Source UNODCCP (2002). cult, so the main market for opium is Asia and Europe. By contrast, North America is the region that has the highest percentage of cocaine consumers (UNODCCP 2002). Goldstein (1985) propoSes three models to explain the connection between dr-ugs and crime. The first, the psychopharmacological link, 45 BREAKING THE CONFLICT TRAP CIVIL WAR AND DEVELOPMENT POLICY (percentage of the population age 15 and above) Country and year Opiate use Cocaine use Australia (1998) 0.8 1.4 Belgium (1998/2000) 0 2 0 8 France (1999) 0 4 0 2 Germany (2000) 0.2 0.9 Italy (1998) 0 6 Italy (1998/99) 0.6 0.8 New Zealand (1999) 0.7 0.4 Spain (1998/99) 0.4 1 5 United Kingdom (2000) 0 6 1 7 United States (2000) 0.5 2 6 - Not available. Source UNODCCP (2002) argues that crime is linked to the psychopharmacological effects of cer- tain drugs. The second model, the economic-compulsive link, argues that drug users commit crimes to obtain money to buy drugs. The third model, the systematic violence link, suggests that crime among illegal drug users is linked to the drug market. Little evidence supports the existence of a psychopharmacological link. The second explanation of the relationship between drugs and crime, the economic-compulsive link, is well supported by the litera- ture. Statistical studies show that the rate of use of illegal drugs is much higher among people who have been in contact with the criminal jus- tice system than among the general population (Casavant and Collin (2001). Data on drug-related crimes are scarce. The U.S. Bureau of Jus- tice and Statistics (2002) finds that in 1997, 73 percent of federal pris- oners and 83 percent of state prisoners reported prior drug use. The third model suggests that the relationship between drugs and crime goes through the illegal drug distribution market. Violence is part of this market, basically because the drug market affords no legal way of obtaining justice when rules are violated. According to Casavant and Collin (2001, p. 14): "Crime in the drug world is often caused by ri- valries among individuals attempting to corner the market. This vio- lence may involve various players-including traffickers, importers, merchants or dealers-and may be intended to control various territo- 46 LET THEM FIGHT IT OUT AMONG THEMSELVES' ries, such as a neighbourhood, streets or school. Violence is then used as an organizational management strategy." This evidence gives an idea of the impact of drugs on crime. The high proportion of convicts who are drug users indicates the potential for crime reduction in the industrial world if illegal drugs were less eas- ily available. Civil War as an Explanation of the Origins of the AIDS Pandemic Epidemiological research on the spread of HIV/AIDS points out that the initial spread of HIV is closely associated with the war in Uganda in 1979. Smallman-Raynor and Cliff (1991, p. 78), geographers at Cambridge University, conclude that "the apparent geographical pat- tern of clinical AIDS in Uganda partially reflects the diffusion of HIV associated with civil war during the first six years of the post-Amin pe- riod." Using regression analysis they find a significant and positive cor- relation between the spread of HIV infections in the 1980s and 1990s and the ethnic patterns of recruitment into the Ugandan National Lib- eration Army. Their research supports the following hypothesis. Many rapes occuired along the borders of Tanzania and Uganda in 1979. HIV was in this region, but before 1979 contagion was sporadic and there was still no epidemic. However, because of the continuous rape, promiscuity, and dislocation during and after the war, HIV started to become an epidemic infection. The spread of AIDS from the south to the north of Uganda exhibited the same route as the one Idi Amin's sol- diers folloxAed after the war in 1979. Civil War and International Terrorism The link between civil war and Al Qaeda is well established. The main activists in the organization were not Afghani, yet they chose to locate in Afghanistan because it provided territory outside the control of a rec- ognized government and under the control of the Taliban, a recently successful rebel organization. Small-scale international terrorism can hide and survive in most societies. What was distinctive about Al Qaeda when compared with other terrorist organizations was its scale. The large scale of Al Qaeda operations, such as training camps for 47 BREAKING THE CONFLICT TRAP. CIVIL WAR AND DEVELOPMENT POLICY thousands of recruits, would have been infeasible except in territory outside the control of a recognized government. Hence the safe havens produced by civil war are not just convenient for large-scale global ter- rorism, they are likely to be essential. Widespread civil war offers such organizations a choice of location and relocation. For example, there has been speculation that Al Qaeda might relocate in Somalia, another civil war territory with no recognized government. Evidence indicates that Al Qaeda acquired substantial revenues from trafficking in West African conflict diamonds (Farah 2002). Recent ev- idence also suggests that in response to greater scrutiny in the interna- tional banking system it has shifted its assets into conflict diamonds. As with AIDS, claiming that without civil war large-scale interna- tional terrorism would have been impossible is unnecessary. There is sufficient evidence for a reasonable inference that civil war facilitates such terrorism. If we attribute to civil war only 10 percent of the con- tributing factors to the September 11 attack, the cost remains enor- mous. The World Bank has estimated that as a result of September 11 global GDP is currendy around 0.8 percent lower than it would have been (World Bank 2002a) and that about 10 more million people worldwide live in poverty than would otherwise have been the case. S OMETIMES THE INTERNATIONAL COMMUNITY EXHIBITS AN understandable impatience with civil war, taking an it's-not-our- problem attitude and remarking that the participants should be left to fight it out among themselves. This chapter has tried to show why such an attitude is mistaken. Many of the costs of civil war, indeed, probably most of them, accrue outside the affected country. The active participants in conflict can be presumed to ignore these costs, as they neither bear them nor are even aware of them. The costs of civil war can be though of as forming three ripples be- yond the direct effects on combatants. The inner ripple, discussed in chapter 1, is the effect on the civilian population: the loss of income and the severe deterioration in health. Many of these losses accrue after the conflict is over, so that even if the active participants care about the effects on civilians, they are likely to be unaware of them. The second ripple, discussed in the first part of this chapter, is across the neighbor- 48 LET THEM FIGHT IT OUT AMONG THEMSELVES? hood. The economic costs of conflict suffered elsewhere in the region may be approximately as large as those suffered within the country, and severe health spillovers also occur, predominantly through refugees. The outer ripple is the global costs. In the past 30 years three devastat- ing social Shocks have been facilitated by civil war. We have not at- tempted to quantify the costs of these shocks, but they are clearly huge. Our point is not to emphasize or inflate the importance of civil war among the world's problems. Rather, we wish to make the simpler point that decisions about these conflicts should not )ust be left to the participants. The participants bear such a small share of the costs of their actions that they will systematically indulge in civil war far beyond its likely social value. Notes 1. Results are available on request. 2. Results are available on request. 49 PART I WHAT FUELS C:HVL WARD PART I SHOWED THAT CIVIL WAR IS A MAJOR the most important factors affecting proneness impediment to development and has spillovers to conflict turn out to be closely associated with that make it a problem of global concern. We economic development: risks are much higher now turn to what fuels civil war. An under- for the poorest countries. Furthermore, far from standing of the factors that make civil war more war resolving political struggle, countries are at or less likely is a helpful input into the formula- risk of falling into a conflict trap whereby one tion of policy responses, which is the subject of civil war tends to lead to another. Chapter 4 part III. scales this analysis up to the global level, trying Chapter 3 analyzes what makes a country to understand what has determined the global more or less prone to civil war and considers incidence of conflict and how it might change. both the risk that a war will ignite and the fac- The main statistical techniques that we use and tors that tend to sustain it once it has started. a selective bibliography of the broader literature Although civil war is intensely political, some of are set out in appendixes 1 and 2, respectively. CHAPTER THREE WhIat Makes a Country Prcone to Civil War? D _gp- IVIL WAR IS FUELED PARTLY BY THE CIRCUM- stances that account for the initial resort to large- scale organized violence, and partly by forces gen- erated once violence has started and that tend to perpetuate it. We refer to the initial circumstances as - the root causes and to the perpetuating forces as the conflict trap. Most people think that they already know the root causes of civil war. Those on the political right tend to assume that it is due to long- standing ethnic and religious hatreds, those in the political center tend to assume that it is due to a lack of democracy and that violence occurs where opportunities for the peaceful resolutioni of political disputes are lacking, and those on the political left tend to assume that it is due to economic inequalities or to a deep-rooted legacy of colonialism. None of these explanations sits comfortably with the statistical evidence. Em- pirically, the most striking pattern is that civil war is heavily concen- trated in the poorest countries. War causes poverty, but the more im- portant reason for the concentration is that poverty increases the likelihood of civil war. Thus our central argument can be stated briefly: the key root cause of conflict is the failure of economic development. Countries with low, stagnant, and unequally distributed per capita in- comes that have remained dependent on primary commodities for their exports face dangerously high risks of prolonged conflict. In the absence of economic development neither good political institutions, nor ethnic and religious homogeneity, nor high military spending pro- vide significant defenses against large-scale violence. Once a country 53 BREAKING THE CONFLICT TRAP CIVIL WAR AND DEVELOPMENT POLICY has stumbled into conflict powerful forces-the conflict trap-tend to lock it into a syndrome of further conflict. Each war is distinctive, with its own particular personalities, events, and antecedents. Any all-embracing, general theory of civil war would therefore be patently ridiculous, and sensibly enough most analyses are country-specific, historical accounts. However, when we pan back from the particular patterns emerge, some of them surprisingly strong, which suggests that some characteristics tend to make a country more or less prone to civil war. This chapter summarizes the evidence on these statistical patterns based on global experience since the 1960s. We abstract from triggering events: the day by day political and mili- tary changes that usher in war. Our focus is on a country's longer-term social, economic, and institutional features. Recall that we are using a precise definition of civil war that excludes several other forms of violence: civil war occurs when an identifiable rebel organization chal- lenges the government militarily and the resulting violence results in more than 1,000 combat-related deaths, with at least 5 percent on each side. Statistical patterns are useful in that they can suggest policies that might typically work in particular situations. They can also defend us from the temptation to overgeneralize from particular conflicts and from the tendency to pick out from the multiplicity of possible causes that which conforms with the beliefs of the researcher. We will see that the large differences in proneness to conflict reflect the conjunction of several risk factors. In this sense, a conflict will usually have multiple causes. Patterns, however, are only a supplement to analysis, not a substitute for it. Patterns come about because of behavior. Civil war occurs if a group of people forms a private military organization that attacks gov- ernment forces and ordinary civilians on a large scale and with a degree of persistence. The typical such organization has between 500 and 5,000 members, although a few, such as the Sudanese People's Libera- tion Army, range up to 150,000 (table 3.1). Globally, such organiza- tions are rare, but they are relatively common in extremely poor coun- tries. To understand the root causes of civil war we need to understand the formation of these private military organizations. Why are such groups formed, that is, what are their motives? How are they formed, that is, what are their opportunities? 54 WHAT MAKES A COUNTRY PRONE TO CIVIL WARz Table 3.1. Size of rebel orgaunizations, selected countries and years Country Rebel organization Size of group and date Azerbaijan Republic of Nagorno-Karabakh 1,000 in 1988, 21,000 in 1992-94 Burunidi San Echec and San Defaite A few hundred in the mid-1990s National Council for the Defense of Democracy 1,000 in the mid-1990s Forces pour la defense de la democratie 10,000 in the mid- 1990s Forces nationales pour la hb6ration 2,000-3,000 in the mid-1990s Colombia Fuerzas armadas revolucionarias colombianas 850 in 1978, 6,000 in 1987; 16,000 in 2000 (FARC) Ejtrcito popular de liberaci6n (ELN) 30 in 1965; 270 in 1973; 350 in 1984; 4,500 in 2000 Movimnienro 19 de Abril 1,500 in 1987; disbanded in 1991 to become a po- litical party United Self-Defense of Colombia 10,000 in the 1990s Indonesia Gerekan Aceh Merdeka (GAM) 24 to 200 in 1976-79; almost disappeared by the early 1980s, 200 in 1986-87, 200 to 750 in 1989-91; 800 in July 1999, 2,000 to 3,000 and 24,000 militia in 2001; 15,000 to 27,000 1 irregulars in 2001-02 Laskar Jihad 2,000 in May 2000 Mali Mouvement populaire de la libration 7,000 to 8,000 in 1992 de l'Azaouad Mozambique Resist8ncia nacional Mocambicana (RENAMO) 200 to 400 in 1976-77; 2,000 to 2,500 in 1978-79; 6,000 to 10,000 in 1980-81; 20,000 in 1984-85 Russia Chechen fighters 1,000 in 1994; 7,000 in 1995; 9,000 in 1999, 7,000 in 2000; 4,000 in 2001 Senegal Maquis 3,000 at the end of 1990 Mouvement de forces d6mocratiques de 2,000 to 4,000 in the late 1990s Casamance Source Sambanis (2003). Udesntndhg Re[[behDDn R EBEL LEADERS USUALLY PROCLAIM SOME NARRATIVE OF grievances against the government, that is, they are usually at least in part leaders of political organizations pursuing objec- tives of political change. While this is evidently an element in their for- 55 BREAKING THE CONFLICT TRAP CIVIL WAR AND DEVELOPMENT POLICY mation, political opposition to governments is not usually conducted through military organizations. The normal vehicles for political oppo- sition are political parties and protest movements. These are quite dif- ferently structured from a private military organization. Most political opposition is somewhat democratic and participatory, whether structured political parties, such as the African National Con- gress during the apartheid era in South Africa and the Movement for Democratic Change in present-day Zimbabwe, or unstructured, non- hierarchical protest movements, such as the revolutions that overthrew the communist dictatorships in Eastern Europe. By contrast, a private military organization is typically small and highly hierarchical, with power concentrated at the top of the organization, often in a single charismatic leader, with a high degree of discipline and severe punish- ment for dissent. Furthermore, most political opposition does not require substantial finance for the organization to be effective. Most participation is vol- untary and part-time, and activities do not require a lot of expensive in- puts. By contrast, a private military organization is a costly operation. It must meet a payroll, because most members are full-time and there- fore dependent on the organization for their material needs, and it must be able to purchase a good deal of imported military equipment. Thus as well as being a political organization, a private military or- ganization is an army and a business. Those analyzing rebel groups must always keep this triple feature-political organization, military organization, and business organization-in mind. Rebellions occur predominantly in countries where circumstances are conducive to all three features. So what are the features conducive to each aspect of a successful rebel organization? Rebeg Gouips as lOki2icaE ftaniza&ns Like all political organizations, a rebellion thrives on group grievances; however, political organizations opposing the government are found in virtually all societies. Even in societies where group grievances are rela- tively modest, as in the high-income societies where income is equally distributed, vigorous mass opposition parties exist. Political grievances and the political conflict they generate are universal. If the main impe- tus for rebel groups is the representation of political grievances, then 56 WHAT MAKES A COUNTRY PRONE TO CIVIL WAR? the obvious question is why does political organization take the un- usual form of small, hierarchical, violent rebellion rather than the more conventional forms of mass parties or mass protest? Why Are so Many Rebellions Ethnic? Many rebellions have an ethnic or religiouS dimension. This accords with an explanation of conflict common on the political right that ethnic and religious hatreds are the root cause of many wars. However, the statistical patterns are quite sur- prising. Here we use Collier and Hoeffler's (2002c) analysis (see box 3.1). Substantial ethnic and religious diversity significantly reduces the risk of civil war. Controlling for other characteristics, a society is safer if is composed of many such groups than if everyone has the same eth- nicity and religion. Obviously such diverse societies are likely to be less harmonious than homogenous societies, in that people identify more with their own ethnic or religious group and less with the society as a whole, and they frequently dislike other groups, but evidently a major gulf exists between such disharmony and the resort to rebellion. An unresolved dispute in political science concerns whether such societies are better suited to proportional representation electoral systems, with each group represented by its own party, or by winner take all systems, which encourage the formation of two large, multi-ethnic parties. Overall, however, the basic circumstances of diversity may be much less dangerous than has popularly been thought (figure 3.1). Although eth- nically diverse societies are commonly seen as fragmented, ethnicity provides an effective basis for social networks. Such societies might therefore L,e less atomistic than homogenous societies. Some evidence indicates that ethnically diverse societies find nationwide collective ac- tion more difficult, but have an offsetting advantage in private sector activity that can benefit from ethnic networks (Collier 2001). More limited ethnic differentiation can, however, be a problem. If the largest ethnic group in a multi-ethnic society forms an absolute ma- jority, the risk of rebellion is increased by approximately 50 percent (figure 3.2). Around half of developing societies have this characteristic of ethnic dominance. Presumably, in such societies minorities may rea- sonably fear that even a democratic political process will lead to their permanent exclusion from influence regardless of the electoral system. Ethiopia and Sri Lanka are examples of ethnically dominant societies with civil vars. 57 BREAKING THE CONFLICT TRAP CIVIL WAR AND DEVELOPMENT POLICY EACH CIVIL WAR IS UNIQUE AND NEEDS TO BE ports. Doubling per capita income approximately studied accordingly, but investigating whether any halves the risk of rebellion; each additional percent- patterns are common to many such wars is also age point on the growth rate reduces the risk by ap- useful. One such approach is that of Collier and proximately one percentage point; and the effect of Hoeffler (2002c). They adopt the conventional po- primary commodity dependence is nonlinear, peak- litical science definition of a civil war, the same def- ing when such exports are around 30 percent of inition explained earlier in the chapter. Investlgating GDP. A country that is otherwise typical but has all such wars that took place during 1960-99 they this high level of primary commodity exports has a focus on 52 for which sufficient data are available 33 percent risk of conflicr, whereas when such ex- to include in subsequent analysis. They then take ports are only 10 percent of GDP the risk falls to 11 all the countries in the world and divide the period percent. Ethnic and religious composition also mat- 1960-99 into eight five-year subperiods. During ters. Societies in which the largest ethnic group con- each subperiod each country could potentially expe- stitutes between 45 and 90 percent of the popula- rience an outbreak of civil war, and the statistical tion-which Collier and Hoeffler refer to as ethnic challenge is to explain why this happened in the 52 dominance-have a risk of rebellion that is about cases but not in the others using only characteristics 50 percent higher; however, other than this, ethnic at the start of each subperiod. The typical develop- and religious diversity actually reduces the risk of re- ing country faced a risk of around 17 percent that bellion. Once a society has had a civil war its risk of rebellion would occur in each subperiod. rebellion goes up sharply, although this risk fades at Collier and Hoeffler adopt an agnostic empirical about one percentage poinit a year. approach in which, in principle, a wide range of Several other statistical models of the initiation characteristics-political, historical, geographic, of rebellion are available (see, for example, Elbadawi economic, and social-could be significant and are and Sambanis 2002; Fearon and Laitin 2003; Hegre introduced into a logit regression. Factors that are and others 2001; Reynal-Querol 2002a). Approaches, insignificant are gradually eliminated, and the re- and consequently results, can legitimately differ be- sulting model is then tested for robustness. Three cause of choices of statistical specification and of economic factors are significant: the level of per data, as this work often requires difficult judgment capita income, its rate of growth, and its structure, calls; however, all the studies agree that a link exists namely, the dependence on primary commodity ex- between poverty and civil war. Just as dominance can cause problems, so too can polarization. Dominance occurs when one group is larger than others, polarization occurs when the society is split into two fairly equal groups. A com- pletely polarized society, divided into two equal groups, has a risk of civil war around six times higher than a homogenous society (Montalvo and Reynal-Querol 2002). The risks polarized societies face depend on the political leadership. In normal circumstances each group tends to police its behavior toward the other group, maintaining nonviolent re- lations (Fearon and Laitin 1996). However, ethnicity is more easily 58 WHAT MAKES A COUNTRY PRONE T'O CIVIL WAR' Figure 3.1 Ethnic fractionalization and the risk of civil war Risk of civil iNar (percent) 30 30 *#.*.*$ * 20 rt 4 $;4 10 - . *.:x^+* 0 20 40 60 80 100 Ethnic fractionalization (percent) Source Collier and Hoeffler (2002c) Figure 3.2 Risk of civil war for the typical low-income country with and without ethnic dominance during a five-year period Risk of civil var (percent) 20 9% 20 - 14.1% 10 0 No ethnic dominance Ethnic dominance Source Collier and Hoeffler (2002c) manipulated by politicians than other bases for identity (Horowitz 1998). Eliies can capitalize on ethnic networks to coordinate violence (Brass 1997; Gurr 2000; Hardin 1995). Thus while ethnic polarization and dominance are probably not inherently conflictual, populist poli- tics may become disproportionately dangerous. Nationalism has often been used to counter ethnic particularity: this was how several Euro- 59 BREAKING THE CONFLICT TRAP: CIVIL WAR AND DEVELOPMENT POLICY pean states were build in the 19th century (Hechter 2001). However, even nationalism can be manipulated for internal division. Irredentist nationalism attempts to extend the boundaries of a state by incorpo- rating adjacent territory occupied by those of the same ethnicity. An important circumstance in which ethnic differentiation can ap- pear to be the cause of rebellion is if a country discovers a valuable nat- ural resource such as oil. Natural resources are seldom found uniformly distributed over the entire country, but are usually concentrated in a particular part of it. The issue then arises as to who owns the resources, the whole nation or the lucky locality. The inhabitants of the lucky lo- cality have an obvious interest in seceding from the rest of the nation and keeping the wealth for themselves. In all societies locality is one as- pect of people's identity, and in ethnically differentiated societies eth- nicity can be used to reinforce this sense of local identity. In most soci- eties, wherever valuable resources are discovered some particular ethnic group is likely to be living on top of them that then has an incentive to assert its rights to secede. All ethnically differentiated societies have a few ethnic romantics who dream of creating an ethnically "pure" political entity, but resource discoveries have the potential to shift such move- ments from the margin of romanticism to the core agenda of economic self-interest. Take, for example, the politics of oil in the United King- dom. Oil was discovered off the shores of Scotland during the 1960s, but it first became really valuable in 1973 when its price quadrupled. The following year the tiny Scottish Nationalist Party, which had only one seat in parliament, launched the "it's Scodand's oil" campaign, and gained 30 percent of the Scottish vote (Collier and Hoeffler 2003). Statistically, secessionist rebellions are considerably more likely if the country has valuable natural resources, with oil being particularly po- tent (figure 3.3). Examples of this sort of secessionist movement are Cabinda in Angola, Katanga in the then Congo, Aceh and West Papua in Indonesia, and Biafra in Nigeria (see box 3.2). Some evidence sug- gests that rebel leaders massively exaggerate the likely gains from cap- turing ownership of the resources. Partly this exaggeration is strategic: the leaders of secessionist movements are often ethnic romantics who simply use the resource issue opportunistically to reinforce their sup- port. Party leaders may themselves succumb to the glamour of natural resources and overestimate the likely gains. For example, leaders of the Gerekan Aceh Merdeka (the Aceh Freedom Movement or GAM) re- bellion in Aceh told the local population that secession would raise 60 WHAT MAKES A COUNTRY PRONE TO CIVIL WAR> Figure 3.3 Risk of civil wars from natuiral resources endowment (a) additional risk when the natural resource (bh the risk that the war is secessionist endowment is double the average Risk of civil wvar (percent) Percent 10 - 100- 8 2% 8 75 - ~~~~~~~~~~67 6% 6- 50 - 4 - 3 1% 2 ~~~~~~~~~~~25 - : Risk of an Risk of a 0- Without oil With oil ideological war secessionist war Source Coltier and Hoeffler (2002c) NIGERIA INHERITED A FEDERAL SYSTEM FROM ITs manded that oil revenue be paid to the regional British colonial rulers in 1960. Upon independence, treasury, and the demand for independence grew a British-style parliamentary democracy was cre- when oil reserves were discovered. A history of po- ated, w.ith three semi-autonomous regions (North, litical instability presaged the war: ethno-regional East, and West). With intensifying competition over conflict over civil service appointments, electoral the distribution of revenues by the central govern- fraud allegations, a coup in 1966 followed by mas- ment, and after the withdrawal of the British, ethno- sacres of the Ibos, and a countercoup. Triggering the regiono l conflict escalated into the Biafran war of in- escalation in violence was the central government's dependence in 1967 after the discovery of oil in the decision to renege on regional autonomy arrange- East. C)jukwu, the governor of the East region, de- ments after the 1967 Aburi Agreements. Sour.e Zinn (2002) their incornes to the level of Brunei's, a more than 1 0-fold exaggeration. Although such natural resource secessions are ethnically patterned and deploy the language of historic ethnic grievances, regarding their root cause as ethnicity is surely naive (see Ross 2002b for a detailed discus- sion of the' civil war in Indonesia). 61 BREAKING THE CONFLICT TRAP CIVIL WAR AND DEVELOPMENT POLICY In many developing countries the government is unwilling to meet such demands for secession, even if a majority of the locality supports it. Indeed, strong ethical arguments can be made against secession. For example, the influential theory of justice proposed by Rawls (1971) asks us to imagine making our choices behind a veil of ignorance: would the secession still be as well supported if the local population did not know in what region of the country the resources were located? The government has a legitimate interest in retaining these resources for use by the poorer majority rather than permitting them to be expropriated to create a small, rich group. The local demand may well be rational, but were such demands met, the world would become more unequal. A more legitimate demand would be that the resources should indeed be used for the poor majority rather than for a small elite. In many countries natural resources have been associated with elite corruption. For example, the International Monetary Fund (IMF) has recently re- ported that more than US$1 billion per year of Angolan oil revenues have been misappropriated, with large sums being paid directly into offshore bank accounts. Where a region sees a corrupt national elite stealing "its" resources, secessionist pressures are surely more likely. Another reason why rebel leaders promote ethnic grievances so prominently is that they are a plausible and legitimate smokescreen for less reputable agendas. The discourse of grievances articulated by rebel groups cannot necessarily be trusted. As with aHl political movements, the rebel organization needs to emphasize grievances, and if necessary it will attempt to exaggerate them or to disguise its true interests in terms of more populist ones. For example, a violent attempted coup d'etat in Fiji appeared at first sight to be motivated by the interests of the in- digenous ethnic group. It turned out, however, that the leader of the coup attempt was a businessman who had been seeking a timber con- cession for the private American company he was representing. When the government awarded the contract to a public agency instead, he launched the coup. The coup's rallying cry of "power to indigenous peo- ple" was undoubtedly more appealing, but perhaps less accurate, than had it been "give the timber contract to the Americans." Similarly, the litany of grievances proclaimed by the Revolutionary United Front (RUF) in Sierra Leone eventually led to the offer of a settlement by the government in which the rebel leader, Foday Sankoh, would become vice president of the country. Sankoh refused this offer and instead de- manded political control of the diamond trade. When he was offered 62 WHAr MAKES A COUNTI RY PRONE T'O CIVIL WAR? this he accepted the peace settlement. As with most conflicts, that in Sierra Leone had multiple causes, including a history of clientalist pol- itics. Natural resources are seldom the entire story behind a conflict, but they have the potential to compound other problems and make them unmanageable. Rebel leaders often use some of the military force at their command to weaken the normal political movements whose objective is to ad- vance the cause that the rebel group ostensibly supports. A common strategy is for a rebel organization to assassinate the moderate political leaders of the interests it purports to represent. If some of these politi- cal organizations are provoked into a military capabiliry as a survival strategy, then one dimension of violent political conflict might be among rebel organizations with apparently similar political objectives. Civil war between rebel groups ostensibly representing the same cause or group is indeed common, for example, in Sudan (see Elbadawi, Ali, and Al Battahani 2002). Thus while the leadership may rely on a dis- course of ethnic grievance and ethnic solidarity, its main energies may be devoted to a power struggle within the ethnic group. Is the Motive Usually Greed? While political scientists and anthro- pologists have tended to focus on political and ethnic agendas, respec- tively, as the motive for civil wars, economic theorists writing on con- flict have treated the motivation quite differently. Grossman's (1991, 1999) model does not distinguish between rebels or revolutionaries and bandits or pirates. Hirshleifer (2001), probably the leading economic theorist of conflict, analyzes rebellion as the use of resources to exploit others for an economic gain. The natural resource secessions discussed earlier broadly fit this economic model: political and ethnic agendas piggyback onto what is basically an attempt to expropriate resources. Is this the norm for rebellion? Sometimes lucrative resources cannot be captured by secession, but require the capture of the state. The most obvious case of this is where the resource is foreign aid: the aid accrues to the recognized government and a rebel group can only acquire it if it overthrows and replaces that government. Grossman (1992) applies his model to aid and predicts that it will increase the risk of rebellion. For many low-income coun- tries aid is certainly a substantial part of the government budget, and so indirectly finances many public sector jobs and contracts that are keenly 63 BREAKING THE CONFLICT TRAP CIVIL WAR AND DEVELOPMENT POLICY contested politically. Hence a large aid inflow makes a state more at- tractive to capture. An empirical test of whether aid increases the risk of rebellion is thus, to an extent, a test of whether greed is an important underlying motivation for conflict. Contrary to the assumption econo- mists commonly make, aid does not appear to increase the risk of re- bellion (Collier and Hoeffler 2002b). Indirectly, as discussed later, aid affects the risk of conflict through its effects on growth, but controlling for this it has no direct effect. While the prevalence of natural resource secessions suggests that greed cannot be entirely discounted, it does not appear to be the powerful force behind rebellion that economic theo- rists have assumed. km RebeODlonis IResponses to [oMcAl RepIession? While the politi- cal right tends to focus on ethnic and religious differences as explana- tions for rebellion, the political center tends to focus on the absence of political rights, maintaining that rebellion occurs where other forms of political organization are not permitted, so the big driver must be polit- ical repression or the lack of political opportunities. Surprisingly, this is not supported empirically. The evidence is muddled, but autocracies are approximately as safe as full democracies, with partial democracies hav- ing a somewhat higher risk than either (Esty and others 1998; Fearon and Laitin 2003; Hegre and others 2001). This is partly because partial democracies allow some political opposition, but do not give the oppo- sition real influence. However, the association between partial democ- racy and civil war may be spurious, because partial democracies have other characteristics such as low income that increase the risk of conflict. A much clearer empirical association is apparent between a change in political institutions and subsequent civil war: stability increases safety (Hegre and others 2001). So how does democracy affect the chances that political institutions will be stable? Unfortunately, this ap- pears to be critically dependent on the level of economic development (Hegre 2003; figure 3.4). At low levels of per capita income, political institutions tend to be less stable in democracies than in autocracies. The average duration of a democratic political system in a low-income country is only nine years. The first four or five years are the most crit- ical: only half survive beyond the first election (figures are calculated using the dataset developed by Gates and others 2003). As per capita income rises, democracies gradually become more stable, whereas the 64 WHAT MAKES A COUNTRY PRONE TO CIVIL WAR? Figure 3.4 The risk of civil war in democracies and nondemocracies at different levels of income Relative risk of armed conflict (log) 2- Democracies 0 Nondemocracies -2 - -3 11 - 7 I I I I II I I I I I I I I 7 7 148 243 401 661 1,090 1,797 2,964 4,886 8,056 13,282 21,898 GNP per capita (1995 U.S. dollars) Source Hegre (2003) stability of autocracies is unaffected. At some point, typically around US$750 annual per capita income, democracies start to become more stable than autocracies, and at high levels of income their political in- stitutions are extremely robust (Gates and others 2003; Lipset 1959; Przeworski and others 2000).1 Thus at higher income levels democracy indeed reduces the risk of civil war, but "one size fits all" simply is not applicable. At low income levels democracy may well be highly desir- able for many reasons, but it cannot honestly be promoted as the road to peace. listorically, political institutions in low-income democracies are characterized by relatively high levels of instability, and this has probably tended to increase their risk of civil war. While exceptions doubtless exist, in low-income countries, where re- bellion is concentrated, no general tendency is apparent for it to be a strategy of Last resort where other means of political expression are denied. Are Rebellions Responses to Acute Grievances? The interpretations of civil war popular with the political left are economic inequality and colonial legacies. In his analysis of the "paradox of power" Hirshleifer (2001) argues that poor people have more to gain from resorting to co- ercion than the rich. All rebel groups provide a litany of severe griev- 65 BREAKING THE CONFLICT TRAP- CIVIL WAR AND DEVELOPMENT POLICY ances, many of which are undoubtedly genuine; however, for such grievances to explain rebellion they should be significantly worse than those of groups in other societies that resort to less violent political processes. Obtaining good objective measures of the intensity of griev- ances is difficult. Two measures that researchers have investigated are inequality of household incomes and inequality in the ownership of land. Collier and Hoeffler (2002c) find no effect of either income or land inequality on the risk of conflict, but do find that once a conflict has started it will tend to last much longer if income is unequal (Col- lier, Hoeffler and Soderbom 2003). In relation to the colonial legacy, Acemoglu, Johnson, and Robinson (2001) find that colonial institutions can have long-lasting effects for good or ill. Where settlers' mortality rates were low, colonial institutions were designed for long-term growth, whereas where their mortality rates were high, colonial institutions were designed for exploitation. This legacy of institutions, as proxied by mortality rates among settlers, is highly significant in accounting for differences in recent growth per- formance, but turns out to have no significant explanatory power in re- lation to either the risk or the duration of conflict. While the colonial legacy presumably affects the risk of conflict to some degree, the connec- tion appears to be weaker than the influence on economic performance.2 Whether or not acute grievances are an important driver of conflict, the evidence reviewed in chapter 1 shows clearly that civil war is a highly unreliable route to social progress. Even where the objective of correcting serious injustices motivates rebel organizations, unfortu- nately, the usual legacy of war is to intensify social problems. What kwe the liioioves fonr 1RebaMouD? The analysis of motives for re- bellion has not led us to any definitive conclusions. Although most rebel groups have public political agendas that appear reasonable, their actual agendas may be somewhat different, and in any case, similar agendas are normally promoted by mass political action rather than by rebellion. Viewed prior to a conflict, predicting which, if any, of the multiplicity of political disputes, grievances, and organizations will turn violent unless addressed is hard. To the extent that political objec- tives determine rebellions, the key drivers are more likely to be either a fear of the potential consequences of structural exclusion or the lure of imagined wealth, rather than the realistic prospect of rectifying acute 66 WHAT MAKES A COUNTRY PRONE TO CIVIL WAR' Figure 3.5 Improved economic perfornance and the risk of civil war Risk of civil war (percent) 17 1% 16 - 14.5% 12.3% 12 - 8~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 4- 0 Typical low-income Country with additional Country with additional country 2 percent growth, 2 percent growth, short-term effect sustained for 10 years Source Collier and Hoeffler (2002c) grievances in the context of severe repression. This is not to deny that rebel groups have specific grievances, but rather to recognize that griev- ances are common, whereas private, illegal, military organizations are rare forms of political opposition. The motive for rebellion need not be a group-specific grievance, in that rebels could be concerned about improving conditions across soci- ety. Indeed, the risk of rebellion increases substantially if average in- comes are low and if the economy is in decline (figure 3.5). However, group-specific issues are more likely to motivate rebellions because the collective action problem is less acute: if rebellion is promised to im- prove conditions for everyone, then no one in particular has much of an incentive to fight. Generalized discontent is perhaps more likely to lead to mass protest movements than to small rebel groups. As dis- cussed later, the association of rebellion with low incomes and eco- nomic decline may reflect other causes of rebellion. Rebel Groups as Military Organizations Regardless of its political agenda, a rebel group is a military organiza- tion. As such it faces problems of recruitment, cohesion, equipment, and survival. 67 BREAKING THE CONFLICT TRAP: CIVIL WAR AND DEVELOPMENT POLICY Re¢cruNing a Privafte &my In terms of recruitment rebel groups usu- ally look much more like an army than a political movement. First, the actual numbers of people involved in rebel activities are usually only a tiny proportion of the society. "Given the right environmental condi- tions, insurgencies can thrive on the basis of small numbers of rebels without strong, widespread, freely-granted, popular support rooted in grievances and, hence, even in democracies" (Fearon and Laitin 2003, p. 81). Even a relatively large rebel group such as the Fuerzas armadas revolucionarias colombianas (the People's Army or FARC) in Colom- bia is recruiting less than 1 Colombian in 2,000. Second, the people who join rebel groups are overwhelmingly young, uneducated males. For this group objectively observed grievances might count for relatively little. Rather, they may be disproportionately drawn from those easily manipulated by propaganda and who find the power that comes from the possession and use of a gun alluring. Social psychologists find that around 3 percent of the population has psycho- pathic tendencies and actually enjoys violence against others (Pinker 2002), and this is more than is needed to equip a rebel group with re- cruits.3 In Nigeria's Maitatsine region, a rebel movement was created in the 1980s by a "prophetic" leader, Marwa, who recruited 8,000 to 12,000 members. Ideological indoctrination and religious teaching were targeted on the homeless and refugees. Their insurgency caused around 5,000 deaths (Zinn 2002). Third, as chapter 1 noted, a seemingly paradoxical, yet common, motivation for recruitment is safety. Compared with the starvation and disease facing the thousands of people displaced from their homes, the organized facilities of a rebel group provide a haven. Fourth, many rebel movement "recruits" do not volunteer; for example, around 80 percent of Resistencia Nacional Mocambicana (RENAMO) recruits were coerced. One standard technique is to kid- nap recruits and then force them to commit atrocities in their home areas, thereby reducing their incentive to escape. Another technique, which the RUF in Sierra Leone adopted, was to target drug addicts on the grounds that such recruits would be easier to control. A further widespread technique is to recruit children. Children are attractive to rebel groups because they are cheap and have little regard for their own safety. For example, in Burundi rebel groups recruited children by force, purchasing Kenyan street children at the price of US$500 for 68 WHAT MAKES A COUNTRY PRONE TO CIVIL WAR' 150 boys (Ngaruko and Nkurunziza 2002). Obviously children do not join rebellions because of objective social grievances. Even where rebel groups do rely upon grievances for recruitment, they sometimes exploit them. A technique common to several groups is to target people whose parents were victims of previous government atrocities. The recruiter pretends to know who on the government side committed the atrocity and offers the opportunity for revenge (Ross, 2002b). Recruits frequently desert. In the largest civil war of the 20th cen- tury, Russia in 1919-21, around 4 million men deserted from the Red and White Armies. The desertion rate was 10 times greater in summer than in winter, because most recruits were peasants whose time was much more valuable during the harvest season (Figes 1996). Using Ethrnicity for Cohesion Rebel military organizations face severe difficulties of maintaining cohesion. As they operate outside the law they do not have recourse to normal contract enforcement techniques. Governments can divide a rebel movement by buying off local com- manders, a technique used against the Khmer Rouge. One technique for maintaining cohesion is to have a hierarchical, dictatorial decision structure, with most power vested in a charismatic leader. A measure of this is that if such leaders are removed the rebel organization tends to collapse rapidly, examples being the eclipse of the Shining Path in Peru once Abimael Guzman had been imprisoned and the surrender of the massive Uniao Nacional para a Independencia Total de Angola (UNITA) forces in Angola following the death of Jonas Savimbi. An- other common technique rebel organizations use to increase cohesion is to confine recruitment to a single ethnic group with leaders drawn from the same clan (Gates 2002). The rebellion thus uses existing eth- nic "social capital." In this they resemble the solutions successful dicta- tors favor, a spectacular example being Saddam Hussein's reliance on the Tikritr clan. In the Democratic Republic of Congo all the rebellions drew their support predominantly from particular ethnic groups, even if the conflict was resource driven. For example, the Katanga secession and the Shaba wars were led by the Lunda, Ndembu, and Yeke ethnic groups. Sinilarly, the Kwilu rebellion involved the Mbunda and Pende ethnic groups, while the 1996-97 rebellion led by Laurent Kabila drew 69 BREAKING THE CONFLICT TRAP CIVIL WAR AND DEVELOPMENT POLICY its initial combatant force from among the Banyamulenge (Ndikumana and Emizet 2002). Where the society is divided into a few large ethnic groups civil wars tend to last much longer. This is probably an indication that ethnicity is being used to maintain rebel cohesion. RecaLl that where societies are highly fragmented in ethnic and religious terms the risk of rebellion is actually lower than in homogenous societies, and when conflicts do occur they tend to be brief. A possible explanation for this is that in such societies large rebel groups will usually need to be multi-ethnic, but multi-ethnic groups cannot maintain cohesion. An example of a society with high ethnic fragmentation and correspondingly limited large-scale violence is Papua New Guinea. At the other end of the spec- trum, Somalia is one of the most ethnically homogenous societies in Africa. Because rebel leaders actively use ethnicity to encourage cohe- sion, this is another reason why ethnicity is so prominent in the rebel discourse and appears to be an important root cause of conflict. lEqupping a PrDiaR L\MY A private military organization needs to acquire armaments and ammunition. This is normally extremely diffi- cult: even criminals seldom have access to armaments more powerful than handguns. Access to armaments varies enormously between coun- tries and over time. Where rebels face large but poorly run government forces, they have occasionally been able to equip themselves by captur- ing government equipment, a classic case being the Eritrean People's Liberation Front versus the Derg government of Ethiopia. Similarly, in Albania and Somalia brief episodes of social disorder enabled local gangs to raid government arsenals. In Albania the looted armaments were taken across the border and became the basis for arming rebellion in the Balkans. In Somalia this set off a chain of gang militarization, de- stroying the possibility of central government on a long-term basis. The breakup of the former Soviet Union established some new gov- ernments that faced acute shortages of revenue, but had huge stockpiles of armaments for which they had no use. Major illegal businesses de- veloped, often run by former soldiers such as the Russian Victor Bout, in which stocks were air freighted to conflict zones in return for natu- ral resource wealth. Thus the availability of military equipment for rebel groups expanded enormously during the 1990s, and its cost col- lapsed. AK-47s now sell for as little as US$6 in some African countries 70 WHAT MAKES A COUNTRY PRONE TO CIVIL WAR' (Graduate Institute of International Studies 2001; U.S. Department of State 1999). Surviving and Military Viability The sheer military viability of rebel- lion will differ greatly between societies, and so will influence the risk of conflict. One simple factor influencing military viability is the terrain. It is easier for large rebel groups to conceal themselves in rural areas with a low population density than in urban areas. Countries in which the population is concentrated in urban areas, but with large, scarcely pop- ulated hinierlands, are statistically more at risk of rebellion. Some evi- dence also suggests that rebellions are more likely to be launched in countries with extensive mountainous terrain. For example, the Eritrean People's Liberation Front was able to rely on safe havens in mountain retreats, arid Nepal, one of the most mountainous countries in the world, has seen a substantial war. Countries such as Colombia, with both mountains and forests, may be geographically more prone to re- bellion than countries such as Saudi Arabia (see Buhaug and Gates 2002 for an empirical study of geography and war). A second factor influencing military viability is the capability of the government. Both good policing and military counterinsurgency oper- ations are organizationally demanding and are much more difficult, for example, than providing basic social services. Deterring rebellion in its early stages requires an effective local pres- ence of government and a willingness to share information on the part of the population. Typically rebel groups kill people they suspect of being infoimers, and so for people to give the government information they must trust it to be effective. Local populations many neither ap- preciate nor trust weak states, which therefore lack the information to contain rebellion. Even highly effective governments find containing rebellion to be an arduous and complex process, although France, Ger- many, Spain, and the United Kingdom were eventually fairly successful in containing violent actions by, respectively, the Front de Liberation Nationale de la Corse, Baader-Meinhof and its later manifestations such as the Red Army Faction, Euskadi ta azkatasuna (ETA), and the Irish Republican Army (IRA). Less effective governments commonly attempt to prevent rebellion by substantially raising conventional military expenditure. For exam- ple, when Ihe objective risk of rebellion is proxied by the risk estimated 71 BREAKING THE CONFLICT TRAP CIVIL WAR AND DEVELOPMENT POLICY FigInv 3.6 RliMU7y es mresns a 1 ris 0 of cHi wta (a) how governments respond to risk (b) how rish responds to military expenditures lMilitary expenditures as a percent of GOP Risk of civil war (percent) 5.1% 60- 5 - 4.3% 5s - 4 - 3 5% ' 1, 40 - 3 10 0 ~ ~ ~ ~ ~ ~ ~ ~~~3 10 percent 30 percent 50 percent 0 5 10 215 20 25 30 risk of war risk of war rish of war JiIitary expenditures as a percent of GOP Source Collier and Hoeffler (2002d) by the Collier-Hoeffler model (Collier and Hoeffler 2002d), an addi- tional 10 percent risk of rebellion increases the government's preemp- tive military budget by more than 10 percent (figure 3.6). Yet such mil- itary expenditure is usually ineffective in deterring rebellion (see box 3.3). Controlling for this tendency of expenditure to be higher where risks are higher, high military spending has no significant deterrence effect on the risk of rebellion. Rebellion is expensive. Typically several thousand people will be full- time workers for the organization for several years. These people and their dependents must be fed, clothed, and housed. They must also be equipped. Depending on its sophistication, military equipment and am- munition can be extremely expensive and in combat conditions needs to be replaced frequently. The rebel organization faces all these costs, yet its military activities do not directly generate any revenue. As a business or- ganization a rebellion therefore faces an acute financing problem. If it cannot overcome this financing problem the rebel group will be unvi- able. This is perhaps the fate of many would-be rebel movements. 72 WHAT MAKES A COUNTRY I'RONE TO CIVIL WAR' INDONESIA HAS KNOWN MUCH POLITICAL V[O- expectations of revenue windfalls that could result lence in its history, including civil wars, self- from ruling a resource-rich independent state of determination movements, ethnic clashes, coups, Aceh. But what gave GAM greater legitimacy and and strote-sponsored massacres. A civil war took access to a larger pool of potential recruits than in place in the resource-rich province of Aceh in the previous years was a negative reaction by the public early 1990s and again in 1999-2002. The war was to the government's counterinsurgency measures fought berween the government and GAM, an or- of the 1980s. These actions intensified after GAM's ganizationi that had pursued autonomy since the reappearance in 1990-91. Following a period of early 1970s. For more than two decades GAM was dormancy, GAM emerged stronger in 1999 at least poorly funded, had little military equipment, and in part because of increased support by Acehnese few reci uits. Part of the reason for GAM's growth in public opinion, possibly resulting from the public the 1990s was the demonstration effect in neighbor- outrage against alleged human rights abuses com- ing East Timor, which encouraged the Acehnese to mitted by Indonesian security forces between 1990 demand independence as well. Also relevant were and 1998. <__ re Ross(2002b) l All the rebel groups that succeed in escalating violence to the scale of civil war must therefore in part be business organizations. This does not imply that personal wealth, or indeed any other economic ambition, is the motivation for the rebellion. Rebel organizations have to be busi- nesses because they have to cover their costs, but most are probably not run for profit. Much of the economics literature on rebellion assumes that the rebel group has economic objectives, whereas much of the po- litical literature neglects to consider finance as a constraint, yet finance can be critical in explaining rebellion, even though it is not motivating. Rebel groups have three broad options in raising finance: they can be initiated by someone who is already wealthy, they can seek dona- tions, and they can operate commercial businesses. The super-rich oc- casionally launch their own political parties, for instance, James Gold- smith in France and the United Kingdom and Ross Perot in the United States, and occasionally they also launch rebellions. Osama Bin Laden is a spectacular current example, and when Jonas Savimbi relaunched the war in Angola in 1994 he was among the richest people in the world. As the numbers of super-rich increase, this may become more common, but historically rebel groups have usually been funded by donations or by their own commercial enterprises. 73 BREAKING THE CONFLICT TRAP CIVIL WAR AND DEVELOPMENT POLICY ho IDonates DIeh? In seeking donations, rebel groups typically do not rely heavily on voluntary contributions from the local group whose interests they promote. In this they differ markedly from normal polit- ical movements. Their main sources of donations are from foreign gov- ernments hostile to the government they oppose and from diasporas living in rich countries. Hostile governments see several advantages in this type of military intervention. It is covert, and so avoids the normal pressures of inter- national dispute settlement. It is containable, and does not result in domestic casualties. Until the end of the Cold War the chief sources of government finance for rebel movements were probably the two super- powers. Since the end of the Cold War regional conflicts have become more feasible, and so neighboring governments may have increased their funding of rebel groups. Obviously, obtaining clear evidence of the importance of government funding for rebel groups is difficult. One such case was the role of the government of Southern Rhodesia in funding and training RENAMO during the 1970s. Once this govern- ment collapsed, RENAMO collapsed. It was then restarted in the early 1980s by the government of South Africa. The United Nations (UN) has documented how several African governments supported UNITA. Similarly, clear evidence points to the involvement of the governments of Rwanda, Uganda, and Zimbabwe in the Democratic Republic of Congo and of the government of Liberia in Sierra Leone. Sometimes involvement is reciprocal, so that the conflict is, in effect, an interna- tional war. For example, at one stage the government of Sudan was sup- porting the Lord's Resistance Army fighting in northern Uganda and the government of Uganda was supporting the Sudan People's Libera- tion Movement fighting in southern Sudan. The rebel group probably gets significant support from a foreign government in most civil wars. The other major source of donations for rebel groups is from dias- poras in rich countries. Diasporas do not suffer the consequences of violence, nor are they in day-to-day contact and accommodation with "the enemy." Case studies suggest that diasporas tend to be more ex- treme than the population remaining in the country of origin: sup- porting extremism is a simple way of asserting continued identity with the place that has been left. A spectacular example of such financing was for the Eritrean People's Liberation Front, which levied an informal income tax on its huge diaspora. Other examples are support from the 74 WHAT MAKES A COUNTRY PRONE TO CIVIL WAR? Tamil diaspora in North America for the Tamil Tigers and support from the Albanian diaspora in Europe for the Kosovo Liberation Army. Unlike the other sources of finance for rebellion, diaspora contribu- tions are sensitive to the media image of the rebel group. Hence a shrewd rebel group will attempt to manage its image, playing on the concerns and memories of the relevant diaspora. After September 11 the American population became more aware of the true consequences of financing political violence, and donations to rebel groups have re- putedly declined sharply. Following September 11 two rebel organiza- tions highly dependent on diaspora contributions from North America, the IRA and the Tamil Tigers, both took unprecedented steps toward peace, with the IRA accepting "decommissioning" of its weapons and the Tamil lTigers withdrawing their demand for independence. What Sorts, of Commercial Enterprises Do Rebel Groups Engage in? Most successful rebel organizations now rely substantially on generat- ing finance by running businesses alongside their military and political activities. The question then becomes in what types of business activi- ties are rebel organizations likely to be competitive? Unfortunately, the obvious answer is that rebel groups have only one competitive advan- tage, namely, their possession of an usually large capacity for violence. Thus the business activities to which they are well suited are various forms of extortion rackets or activities that only require military con- trol over a limited territory. These business activities are most com- monly associated with the extraction of natural resources, and civil wars occur disproportionally in countries with extensive dependence on nat- ural resources (figure 3.7). Recall that for military reasons rebel groups will tend to locate in rural areas. Most rural areas are poor. Obviously extortion rackets only work if there is something to extort, and this constitutes a major limi- tation on rebel activity: extremely poor areas are not well suited to ex- tortion, and so tend to be unsuited for rebellion.4 However, a minority of rural areas are well suited to extortion, namely, if they are producing primary cornmodities with high economic rents. Such commodities are generally for export, and the largest rents are usually from the extrac- tion of natural resource wealth. Where such activities are under way, for rebel groups to run an extortion racket that involves charging produc- 75 BREAKING THE CONFLICT TRAP CIVIL WAR AND DEVELOPMENT POLICY Mgum 3.7 N&XmD um=¢s3 &es RmgcoU s a3 CM ¢-mD ft: i- Pvvv4ncjs Risk of civil war (percent) 30 - 29.5% 23.6% 20 1 16.8% 10.5% 10 n 5.7% 0-E 5% 10% 15% 20% 25% Primary commodity exporKs as a shnare of GDP Source Collier and Hoeffler (2002c). ers for protection is a relatively simple matter. The best known exam- ple is diamonds in Angola and Sierra Leone. Alluvial diamonds are par- ticularly well suited to rebel groups because the technology is so simple that the organization can directly enter the extraction process. Simi- larly, timber felling is a simple technology. However, high-value agricultural exports are also sometimes a target for rebel extortion. Here the rebel group does not produce the crop it- self, but levies informal taxes on production. The most spectacular ex- ample is illegal drugs, which because of their illegality are extremely valuable. Current global policy on drugs implies that drugs can only be grown on territory outside the control of a recognized government. Those rebel groups that control territory on which drugs can be grown can therefore charge large rents to producers. For example, when the U.S. government ceased to fund the mujahideen in Afghanistan, the group shifted into drug production. Similarly, estimates indicate that FARC in Colombia generates around US$500 million per year from its control of drug cultivation. Even lower-value export crops are some- times the target of rebel extortion rackets. For example, the RUF in Sierra Leone started by levying informal taxes on coffee, and only shifted its activities to the diamond areas once it had become established. Some extractive industries require technology that is too sophisti- cated for rebel groups and requires multinational corporations (MNCs), 76 WHAT MAKES A COUNTRY PRONE TO CIVIL WAR) but this does not prevent extortion. Rebel groups can target MNCs by threatening expensive infrastructure. The classic infrastructure target is a pipeline: typically oil companies pay protection money to "violence entrepreneurs" in local communities. Such entrepreneurs sometimes fight among themselves for the right to extort. For example, in the delta region of Nigeria violence entrepreneurs from rival villages on either side of a new Shell pumping station recently fought it out for the extortion rights, resulting in 75 deaths. Violence in the Nigerian delta began in the mid-1990s at a modest level. It was essentially political, being di- rected against a military government. Despite democratization, the vio- lence has escalated sharply, but has been transformed into something more akin to American gangland fights for control of the drug trade. A particularly remarkable recent development is for rebel groups to raise finance by selling the advance rights to the extraction of minerals that they currently do not control, but which they propose to control by purchasing armaments financed through the sale of the extraction rights. Kabila, subsequently president of the Democratic Republic of Congo, reportedly raised several million dollars from Zimbabwean com- mercial interests in return for extraction contracts before launching his successful assault on Kinshasa (Graduate Institute of International Stud- ies 2001). Similarly Denis Sassou-Nguesso, subsequently president of the Republic of Congo, reportedly sold extraction rights to help finance his military bid for power. An alternative technique for extortion against MNCs is kidnapping followed by ransom demands. FARC generates around US$200 million annually from ransoms, disproportionately from kidnapping the em- ployees of MNCs. Oil companies are common targets for kidnapping, and in some regions companies now suffer kidnaps as a daily occurrence. Pax Christi Netherlands (2001) estimates that during the 1990s Euro- pean companies' ransoms to rebel movements amounted to US$1.2 bil- lion, a sum that far exceeds official European aid flows to the affected governments. The Colombian rebel group Ej&cito de Liberaci6n Na- cional (ELNi) reputedly got US$20 million in ransom from the German company Mannesmann, money that was critical for the group's purchase of sophisticated military equipment and its subsequent expansion. Rebel groups also target foreign tourists for kidnap. For example, a small rebel group in the Philippines recently ransomed a party of European tourists via Libya for US$1 million per person. Following each successful kid- napping rebel recruitment soars, presumably as young men anticipate 77 BREAKING THE CONFLICT TRAP CIVIL WAR AND DEVELOPMENT POLICY BETWEEN 1991 AND 1993 BREAKAWAY CHECHNYA tremely lucrative. Consumer goods were imported controlled more than 300 kilometers of the Russian duty free via Chechnya, while natural resources and border. During this time Chechnya became an enor- weapons were exported to world markets without mously profitable, illegal but tolerated, free trade any regulation. The financial flows, which financed zone that ensured its owners a fortune of millions in Dzokhar Dudaev's regime and later the war, origi- hard currency. In practice independent after 1991, nated in the shadow economy. Not surprisingly, Chechnya possessed an international airport and in- Dudaev's independent Chechnya was supported ternational borders with Georgia, but was still fully and used by entrepreneurs in the shadow economy, integrated into the Russian economic space. This who exploited the "free trade zone" of Chechnya meant, first, that Chechnya had access to cheap and for their business. These so-called patriotic busi- exportable Russian natural resources; and, second, nessmen were interested in an independent Chech- that it had access to the Russian consumer market, nya, out of reach of the Russian state, but with ac- which was eager for all sorts of consumer goods. cess to the Russian space of opportunity and to the This made Chechnya a bonanza for the shadow world market. They also had an interest in ensuring economy, and its position as a hub between world state weakness in Chechnya to maintain their free- markets and the Russian markets proved to be ex- dom of activity. Soturce Zurcher, Koehier, and Baev (2002) large payoffs. In Colombia rebel groups have combined with urban- based criminals to create a market in kidnapped people. Criminals un- dertake the kidnap, selling the victim on to the rebel group, which then demands a ransom. Just as markets in the victims of kidnap are arising in developing countries, so markets in ransom insurance are arising in industrial countries. Perversely, the eventual effects of kidnap insurance are to reduce the incentive to protect workers from kidnap and to in- crease the size of ransom payments. Although natural resources are probably the most common target for rebel extortion in rural areas, another valuable attribute is if the area includes an international border. Physical control over a border can be valuable because of the potential for smuggling. A post-Soviet aphorism states that control over a kilometer of the Russian border sufficed to be- come a millionaire (see box 3.4). The potential for exploiting a border depends on the trade policies the country and its neighbors have adopted. As Russia was highly protectionist, control of the border en- abled goods to be smuggled into the country. Sometimes the smuggling can go in the other direction. For example, Afghanistan is bordered by 78 WHIAT MAKES A COUNTRY PRONE TO CIVIL WAR' countries that have usually been highly protectionist; thus control of frontier areas in Afghanistan has enabled goods to be flown in at world prices and smuggled into these neighboring countries where they are far more valuable. Finally, some rebel groups have used their comparative advantage in violence to capture some niche markets in extortion in industrial coun- tries. For example, the Albanian mafia associated with the Kosovo Lib- eration Army reputedly now controls around 80 percent of the prosti- tution trade in central London (The Observer 2002). So Is the RMot the Loot? We have already argued against a greed-based interpretation of rebellion. Most entrepreneurs of violence have essen- tially political objectives, and presumably initially undertake criminal ac- tivities only as a grim necessity to raise finance. However, over time the daily tasks involved in running a criminal business may tend inadver- tently to develop a momentum of their own. The organization starts to attract more criminal types and fewer idealists, so that it may gradually change its character. Some rebel leaderships tend to do well out of war and may be quite reluctant to see it end. In some cases, such as RUF's movement from Sierra Leone to Guinea, a rebel group that finds its criminal activities thwarted in one country relocates in another country. At this point any political agenda has withered away, leaving a "roving bandit" that classic analysis tells us is the most destructive form of power (Olson 199.3). Loot is not usually the root motivation for conflict, but it may become critical to its perpetuation, giving rise to the conflict trap. The Conflict Trap NCE A REBELLION HAS STARTED IT APPEARS TO DEVELOP A momentum of its own. Getting back to peace is hard, and even when peace is re-established, it is often fragile. Getting Back to Peace The best predictor of whether a country will be in civil war next year is whether it is at civil war now (see box 3.5). Wars are highly persistent: 79 BREAKING THE CONFLICT TRAP CIVIL WAR AND DEVELOPMEN T POLICY COLLIER, HOEFFLER, AND SODERBOM (2003) USE A rebellions and those without. Explaining the dura- hazard regression to investigate why some wars last tion of rebellions depends on the much more lim- much longer than others. Investigating the duration ited variation between countries with wars. of civil war is more demanding statistically than Other empirical studies of civil war duration in- studying its onset, so the results vary considerably. clude Balch-Lindsay and Enterline (2000); Buhaug, Explaining the onset can use a large number of ob- Gates, and Lujala (2002); DeRouen 2003; Elba- servations with a wide variation in characteristics, dawi and Sambanis (2000); Fearon (2002); and because the comparison is between countries with Regan (2002). the typical civil war lasts around seven years. As part I indicated, the costs of such wars are astronomical, and thus they are seldom forces for successful transformation. Here we are concerned with why they last so long. Superficially, given that conflicts are so costly finding mutually ben- eficial agreements that end them might seem to be easy. However, con- sider the radical difference between rebellions against governments and strikes by workers against a company. Few unions and companies can prevent strikes altogether, but once they occur they are generally settled within days or weeks: ending strikes quickly is often easier than pre- venting them altogether. With rebellions it is the other way round: most governments never face a rebellion, but once one has occurred ending it is difficult. Why is rebellion so persistent? Even where the population has significant grievances, governments are understandably reluctant to concede to violence what they have not conceded to nonviolent pressure. Clearly governments cannot afford to signal that violence is an effective political strategy, given that all soci- eties have many groups that are willing to resort to violence to achieve their goals, so the potential is limitless. A further problem is that even if governments are willing to concede to rebels' demands, they might have no credible means of committing to the agreement, and thus the rebel group might fear that once it loses its fighting capability the gov- einment will renege, a problem known as time inconsistency. Conced- ing to all rebel demands may even be logically impossible. The circum- stances under which one rebel group is able to thrive often also enable other groups to thrive, and sometimes these groups have opposing ob- 80 WHAT MAKES A COUNTRY PRONE TO CIVIL WAR' Figure 3.8 How chances of peace evolve worldwide Probability oF peace during the year (percent) 2- IO - I I II I I I 1 2 3 4 5 6 7 8 9 10 Duration (years) Source Collier, Hoeffler, and Soderbom (2003) jectives. For example, in Colombia to the extent that the rebel groups have discernable political agendas, they are contradictory. Yet significant patterns are apparent. Wars are particularly lengthy if a society has extremely unequally distributed income and a very low av- erage incorne, possibly because the cost of sustaining rebellion is low if a country has many destitute people, and possibly because the govern- ments of such countries are typically weak. Wars are particularly lengthy if the socicty is composed of two or three ethnic groups, perhaps be- cause this makes creating distinct identities of support easier for both rebels and government. Over the first four years of war the chances of peace gradually dete- riorate. Presumably the conflict intensifies hatreds, and it may also gradually shift the balance of influential interests in favor of continued conflict. Criminal entrepreneurs do well out of war at the expense of other interests, and so in these early stages of conflict the criminals thrive while the honest decline. Beyond four years the chances of peace gradually improve again, perhaps reflecting the declining opportunities for extortion as the economy goes into retreat (figure 3.8). Wars also appear to have been getting longer (figure 3.9). Note that the modest shortening of wars in the 1990s may well be temporary. As discussed in chapter 4, the end of the Cold War saw a surge in peace 81 BREAKING THE CONFLICT TRAP CIVIL WAR AND DEVELOPMENT POLICY F'Dgure 3.9) [OMTEON i CM3 ta7m$ (ID5Y(D &IS( Duration of twar in months 125- 75- 50- 25- 0- !i2Ss 1970S iD98S n!eos Source Collier, Hoeffler, and Soderbom (2003) settlements, but this was not sustained beyond the mid-1990s. The ex- pected duration of conflict is now more than double that of conflicts that started prior to 1980 (Collier, Hoeffler, and Soderbom 2003). One possible explanation is that sustaining a conflict is easier than it used to be, because even without support from a superpower or a neighboring government, rebel groups can generate revenues and purchase arma- ments. Another possibility is that rebellions have gradually changed their character, becoming less political and more commercial. Violence entrepreneurs, whether primarily political or primarily commercial, may gain from war to such an extent that they cannot credibly be compen- sated sufficiently to accept peace. Those who see themselves as political leaders benefit from war because they can run their organizations in a hierarchical, military style with power concentrated in their own hands, something much more difficult to justify in peacetime. Those who see themselves as extortionists benefit from the absence of the rule of law in the areas they control. However leaders see themselves they will have invested in expensive military equipment that will become redundant once they agree to peace. Asking a rebel leader to accept peace may be a little like asking a champion swimmer to empty the pool. The international community has made many efforts to shorten civil wars by means of diplomatic, economic, and military interventions. Our analysis suggests that none of these types of interventions has been sys- tematically successful. Particular interventions might have worked, but no 82 WHAT MAKES A COUNTRY PRONE TO CIVIL WAR) Figure 3.10 The risk of civil war for a typical civil war country, just be- fore and just after war Risk of civil war (percent) 50 - 43.6% 40 - 30 - 24.8% 20 - 10 - 0 Eve of war Start of peace Source Collier and Hoeffler (2002c) general significant effect is apparent. Hence once a rebellion has started, there appears to be something of a trap: powerful forces keep a conflict going, while the international community appears almost impotent to stop it. Unfortunately this continues even once peace has been reached. Reverting to War The typical country reaching the end of a civil war faces around a 44 percent risk of returning to conflict within five years (figure 3.10). One reason for this high risk is that the same factors that caused the initial war are usually still present. If before a war a country had low average income, rural areas well endowed with natural re- sources, a hostile neighbor, and a large diaspora, after the war it is still likely to have these characteristics. Some countries are intrinsically prone to civil war by virtue of their geography and economic structure, so that as the government settles with one rebel group another is likely to emerge. We would expect a country such as Colombia, with moun- tains, forests, and a lot of sparsely populated territory, to have a per- sistently higher incidence of civil war than, say, the Netherlands. This is indeed part of the explanation for the persistence of civil war. For example, countries that go into civil war tend to have much lower incomes than other countries. This low income tends to make the con- flict last a long time and to make the country more likely to have a fur- 83 BREAKING THE CONFLICT TRAP CIVIL WAR AND DEVELOPMENT POLICY ther conflict once it has reached peace. However, another possibility is that a high degree of conflict persistence arises because of a vicious cir- cle of civil war. We now explore various ways in which conflict in one period may increase the risks of subsequent conflict. War Reverses DeveopueIuQ The most obvious way in which conflict has a feedback loop is that civil war interrupts, and indeed reverses, eco- nomic development. As chapter 1 showed, during a civil war a country loses, on average, around 2.2 percentage points off its normal annual growth rate. Because the average civil war lasts around seven years, by the end of the war per capita income is around 15 percent lower than it would otherwise have been. Our previous analysis indicates that this will raise the long-term incidence of conflict for the country both by increasing its risk of further rebellion and by increasing the duration of rebellion should one occur. For the typical country experiencing a civil war, this effect of the war would increase the risk by 13.5 percent and the duration by 5.9 percent, so that the long-term incidence would rise by 16.9 percent.5 A related feedback loop works through the effect of conflict on the structure of the economy. Natural resource exports are relatively robust in the face of conflict, because of the high rents normally involved in their production and their relative independence of inputs from the rest of the economy. By contrast, more sophisticated exports are typically low-margin and dependent on a fragile network of business interdepen- dencies, and these tend to get severely disrupted by the war. Further- more, economic policy and institutions deteriorate significantly during civil war, and this takes time to put right. Studies show that diversifica- tion out of primary commodity dependence is influenced both by the level of income and by policies and institutions (Collier and Hoeffler 2002b). Thus as policies, institutions, and income all deteriorate during war and take a long time to rectify, for a much longer period than the war itself the country will find itself trapped into dependence on pri- mary commodities. This in turn will increase the risk of further conflict. War Thggers IEOigIreiOu and iasoporas A further feedback loop is through emigration of the work force. Civil war triggers an exodus of people: some are refugees to neighboring countries, others are asylum 84 WHAT MAKES A COUNTRY PRONE TO CIVIL WAR' Figure 3.11. Diasporas and post-conflict risk Risk of relapse into civil war (percent) 30 ~~~~~~~29 4% 30 - 20 - 15.1% 10 - Small diasporas Large diasporas Note Small diasporas are those of similar size to that of the United States, large as involving 10 times larger relative to the population Source Collier and Hoefficr (2002c) seekers in rich countries, and others are simply economic migrants in- duced to emigrate by the collapse in economic opportunities at home (Collier, Hoeffler, and Pattillo 2002). For different reasons, this emi- gration is also highly persistent, in that when one group of people has migrated, it tends to assist others to follow. Thus even once peace has been reached the society might continue to experience rapid emigration of workers, thereby further depressing economic growth. Emigration not only deprives the economy of its labor force, its cre- ates a large diaspora living in rich countries. Statistically, such diasporas increase the risk of a return to violence (Collier and Hoeffler 2002c). A potential problem is involved in interpreting this statistical association causally: to the extent that diasporas are the result of civil war, a large diaspora might simply be proxying a particularly severe war; however, when this is controlled for, the adverse effect of diasporas remains. Fig- ure 3.11 illustrates the risk of conflict for a country with an average size diaspora in the United States versus one with a diaspora that is 10 times larger relative to the home country population. The most likely route by which diasporas increase the risk of repeat conflict is through their tendency to finance extremist organizations. To give an example, detec- tive work has established that the massive bomb that killed 86 civilians and injured more than 1,400 in Colombo, Sri Lanka, in 1996 used 60 tons of East European explosives purchased using funds from a Singa- 85 BREAKING THE CONFLICT TRAP CIVIL WAR AND DEVELOPMENT POLICY pore bank account opened by a Canadian of Sri Lankan origin (Bell 2000). As noted earlier, diasporas tend to be more extreme than the populations they have left behind. War Leaves a I?ersWsiert anud Damagiung MiDRauy Lobby During wars military spending obviously rises, and during the typical civil war the military budget increases by nearly 50 percent. Reducing this spending in the early postconflict period is not easy. There is often a widespread awareness of continued risks of conflict, and as with any powerful lobby, the military will be reluctant to see its budget cut. Furthermore, the government sometimes needs to integrate rebel forces into the army, which creates pressures for expansion. Military spending reduces growth (Gleditsch and others 1996; Knight, Loayza, and Villanueva 1996); therefore both during and after a civil war such high military spending will be a drag on growth. How- ever, the adverse effects of high military spending in postconflict situa- tions can be even more serious. We have already noted that government military spending is normally ineffective as a deterrent of rebellion. Fig- ure 3.12 shows that in postconflict situations it is actually significantly counterproductive. Statistical analysis indicates a potential problem of Ifginev 3.2 eeuili>iw ZIU-,ILl. 3Th enX W$e ~i e23 J@ ¢cenu em Risk of civil war (percent) 50 - 39 7% 40 30- 264% 20- 10 0- Average military spending Twice-average military (2.87% of GDP) spending (5.74% of GDP) Source Authors' calculations. 86 WHAT MAKES A COUNTRY PRONE TO CIVIL WAR' bogus causality: high spending will sometimes reflect a correct percep- tion of an unusually high risk of further conflict, and so will simply be proxying the risk, but when this is controlled for the effect remains. Why should high military spending in postconflict situations be so dysfunctional? A possible reason is that military spending may inadver- tently be a signal of government ill-intent. Recall that one obstacle to a settlement is the low credibility of an agreement, that is, the govern- ment has more interest in promising generous peace terms than in ac- tually delivering on them. High military spending might thus be seen as an indication that the government is likely to renege. Some indirect support for this interpretation comes from an analysis of which policies are most conducive to growth in postconflict situations. On average, countries emerge from conflict with poor policies across the board: macroeconomic, structural, and social. Collier a,nd Hoeffler (2002a) investigate how policy priorities should differ in such a country from one that has the same poor policies but is not postconflict. They find that simply on the criterion of maximizing short-run growth, if the country is postconflict it should give greater priority to such inclusive social policies as widening access to education and health care. Although education and health care eventually con- tribute to growth, they do so with long lags, so that the unusually strong effect of social policies is unlikely to be due to their direct contribution to growth. An alternative route may be that prioritizing inclusive social policies signals to the population that the government is committed to a peace settlement. On this interpretation, postconflict governments that prioritize military spending are inadvertently signaling that they will renege on the peace settlement and those that prioritize social spending are signaling that they will adhere to it. The former signal in- creases the risk of conflict, while the latter builds private sector confi- dence and thereby accelerates growth. If this interpretation is right it suggests that governments are not impotent: their policy choices can alter the risks they face. We return to this in more detail in chapter 5. War Changes the Balance of Interests and Intensifies Hatreds All the foregoing feedback loops work through factors that investigators have incorporated into models of conflict risk: the level, growth, and structure of income; military spending; and diasporas. However, the risk of a reversion to conflict is much higher than is accounted for by 87 BREAKING THE CONFLICT TRAP CIVIL WAR AND DEVELOPMENT POLICY these effects. On average, only about half of the 44 percent risk of re- peat conflict is due to characteristics either present before the conflict started or explicitly modeled as deteriorating as a result of conflict. The other half of the risk is due to things that happen during the conflict but are not included in the analysis. By definition, as these factors are omitted from the modeling analysis, it cannot guide us as to what they are, but we can speculate on some likely ways in which conflict in- creases the risk of further conflict. One likely feedback mechanism is that violent conflict changes the balance of assets in the society, reducing the value of those that are use- ful during peacetime and increasing the value of those that are useful only for violence. The violence-specific assets are partly physical, such as armaments; partly human, such as the skills to use weapons and the re- duced regard for human life and dignity; and partly organizational, such as the hierarchical rebel management structure and established com- mercial ties with arms suppliers and natural resource traders. The own- ers of these assets are unlikely to sit on the sidelines while their value col- lapses. They do well out of war and would like to get back to it. Another likely feedback mechanism is that violent conflict leaves a legacy of atrocities. As a result, hatreds build up during periods of vio- lence, leaving the society polarized. People of both sides want vengeance for atrocities committed during the conflict and these may supplant any prior grievances. We have already noted how rebel recruitment some- times capitalizes on such grievances. HE INTERPRETATIONS OF CIVIL WAR THAT HAVE BEEN MOST common in industrial countries either treat them as wholly an outcome of primordial ethnic and religious hatreds or force them into the familiar framework of Western politics. Rebel leaders have learnt to play up to these images of their organizations, raising money from ethnic diasporas while styling themselves as heroic politi- cal leaders. Another tempting framework, favored by economists, is to see rebel leaders as being at the apex of organized crime, enriching themselves from massive protection rackets at the expense of the wider society. The recent prominence of so-called "conflict diamonds" has in- 88 WHAT MAKES A COUNTRY PRONE TO CIVIL WAR' creased popular awareness of this darker side of rebellion. Both these interpretations miss the reality of many rebellions; that is, even though rebel leaders are indeed violence entrepreneurs heading private military organizations that run protection rackets, they usually have some po- litical agenda. However, they are not conventional political leaders in that they have chosen not to lead normal political movements. Motivations-grievances and greed-are obviously part of the expla- nation for r ebellion, but if we focus exclusively on motivation we rapidly encounter a paradox. In many situations of the most grievous injustice, both currently and historically, rebellion does not occur. Highly repres- sive societies often fail to trigger civil war, such as Iraq and the Dem- ocratic People's Republic of Korea. Highly unequal societies often fail to trigger civil war, such as Chile and Kenya. Extreme cases of ethnic abuses of power have often failed to trigger civil war, such as white domination in South Africa, and, delving back into history, Norman domination in England, although some forms of ethnic political exclusion do appear to increase the risk of war. Greed perhaps fares a little better as an explana- tion, as secessionist rebellions seem to be linked to the desire to appro- priate valuable resources and some rebel leaders appear more committed to a personal than to a social agenda; however, even greed does not seem to get us very far, because states with large aid inflows are much more at- tractive to capture, but they do not face any greater risk of rebellion. While the literature that tries to explain civil war has focused over- whelmingly on motivation, we also need to note that the circumstances in which rebel groups are militarily and financially viable are relatively rare. Hirshleifer (2001) has put forward a depressing proposition, the Machiavelli theorem, whereby no advantageous opportunity to exploit someone will be missed. Even though many rebellions are not moti- vated by the desire to exploit someone, a closely analogous proposition may be fairly accurate: no militarily and financially viable opportunity to promot, a political agenda by rebellion will be missed. If a neigh- boring government is sufficiently hostile and the circumstances are pro- pitious, it will seek out and promote a local violence entrepreneur. If resource-extracting MNCs offer sufficiently easy pickings in unpro- tected rural areas, local violence entrepreneurs will set up rudimentary protection rackets loosely linked to political demands. In such circum- stances the ostensible grievance might be any of a wide range of things: grievances are not in short supply. 89 BREAKING THE CONFLICT TRAP CIVIL WAR AND DEVELOPMENT POLICY Globally, one of the largest mass political protests of recent years, which brought more than 400,000 people onto the streets of London, was to defend the right to hunt foxes. The typical rebel group does not need a cause that attracts anything like this level of support: a few hun- dred or a few thousand people will suffice to reach the level of violence that constitutes civil war. Thus most societies probably have several issues on which it is possible to find a small core of people who feel passionate and who are not averse to violence. Identifiable political groups have perpetrated violence in France (Breton separatists), the United Kingdom (animal rights activists), and the United States (anti- abortion activists), and political assassinations have occurred in Italy, the Netherlands, and Sweden. Hence most societies have the political potential for violence. Whether such violence remains peripheral, as in the foregoing examples, or becomes large enough to generate wide- spread death and destruction, may depend as much upon whether an illegal, private, military organization is militarily and financially viable as upon the political issue itself. Obviously governments should address justified grievances, whether or not they are likely to lead to large-scale violence. A government that is considerate and inclusive is surely less likely to face rebellion, and, in any case, it will be a better government. However, we should be wary of vilifying those governments of low-income, natural resource- dependent countries that face rebellion. Rebellion need not be a symp- tom that they are markedly worse than other governments. Instead, they may be in an economic and geographic environment where rebel- lion is particularly easy, and perhaps even particularly attractive. A jour- nalist interviewed Kabila when he was marching on Kinshasa. He re- portedly explained that in Zaire rebellion was easy-all that was needed was ten thousand dollars and a satellite phone. The dollars were to re- cruit a small army, cheap because the population of Zaire was among the poorest in the world. Recall that even in Zaire the quote was an ex- aggeration. Kabila had received several million dollars and the support of foreign armies to launch his assault. The satellite phone was to make deals with foreign businesses in extractive industries. Although occasionally rebellion leads to an improvement in govern- ment, more often it leads to spectacular deterioration, and therefore the presumption that rebellion should be avoided is reasonable. Partly this is a matter for governments to make greater efforts to redress reasonable 90 WHAT MAKES A COUNTRY PRONE TO CIVIL WARP grievances, but it is also a matter of making rebellion less easy. Many of the things that would make rebellion more difficult require action at the regional or global level, and the international community can ac- tively discourage rebellion without taking sides in political disputes. This is the subject of part III. Although political conflict Is common to all societies, civil war is con- centrated in the lowest-income countries. In a sense this is hopeful. It is an indication that peace does not depend on resolving all political con- flict and that such conflict is normal. Rather, economic development is the critical instrument in preventing rebellion and in building the con- ditions in which groups engage in their conflicts through normal polit- ical means Economic development in the lowest-income countries is not easy, but neither is it unprecedented, incredibly complex, or wildly expensive. Once a rebellion has started, a society risks being caught in a con- flict trap. Elnding the conflict is difficult, and even if it ends, the risk that it will start again is high. Strong global actions can be targeted to- ward conflict prevention in these high-risk environments. Building a peaceful world is not just a matter of encouraging tolerance and con- sensus. It involves a practical agenda for economic development and the effective regulation of those markets that have come to facilitate re- bellion and corrupt governance. Notes 1. Heavy dependence on natural resources also tends fertile land in arid areas. Homer-Dixon (1991) has em- to make autocracies stable and democracies unstable phasized this category of conflict, but see the discussion (Ross 2000). by Gleclitsch (1998). 2. Results available on request. 5. The change in the long-term or self-sustaining in- cidence is calculated using the method explained in ap- 3. Mueller (2000) analyzes the wars in Bosnia and pendix 1. We assumed the initial probability of war mi- Rwanda and concludes that the number of rebels com- tiation is 0.016 and that of termination is 0 123. We mitting the atrocities was relatively small. He estimates multiply the initial w probability with 1.135 (corre- that the genocide in Rwanda was carried out by approx- sponding to a 13.5 percent increase) and the initial v imately 2 percent of the male adult population probability with 0.9405 and recalculate the self-sustain- ing incidence. 4. A possible exception is where resources are valu- able because they are locally scarce, such as water and 91 CHAPTER FOUR Whzy Is Civil War So Common? HIS CHAPTER TURNS FROM A MICRO-LEVEL ANALY- sis of what circumstances are conducive to rebellion to a macro-level analysis of what determines the T global incidence of civil war. It looks at how the in- cidence of civil war has changed over time and space and attempts to account for these changes in terms of the uncLerlying causes of civil war identified in chapter 3. Civil war is increasingly concentrated in relatively few conflict-prone countries, many of them in Africa. We use the macro-level analysis to investigate how economic development is changing the overall incidence of civil war. Development has not been reaching those countries most prone to civil war. As a result, if past trends continue, the world will evolve into a two-class system, with the majority virtually conflict free and a minority trapped in a cycle of long internal wars interspersed with brief, unstable periods of peace. The minority of countries caught in the conflict trap will increasingly dominate the global incidence. Changes in the Global Pattern of Civil War A CTfVE WARFARE HAS CHANGED ITS CHARACTER OVER THE past 50 years in that international wars have become rare, vvhereas civil wars have become more common. In 2001 all but one or the world's wars were civil wars. Furthermore, when inter- national wars do occur, they tend to be short: most last less than six months (13ennett and Stam 1996). By contrast, civil wars last a long time, on average about seven years, and their duration has tended to increase. 93 BREAKING THE CONFLICT TRAP CIVIL WAR AND DEVELOPMENT POLICY RpmB 4A Me gOtDb e¢cd(ne GI du' tieven e W& n9 s$g208 Percent 20 15 Ongoing wars O~ New wars \ 10 _- .- A - - -- =~ _ _ _ 1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 Note Proportion of countries in civil war by year Figure shows only conflicts with at least 1,000 battle deaths over the course of the conflicts. Source Gleditsch and others (2002). Figure 4.1 shows the incidence of civil war, that is, the proportion of countries that are at civil war at any one time. Between 1950 and 2000 the overall incidence rose, but this has not been a steady process: the global incidence of civil war peaked around 1990. The global incidence of civil war at a particular time is determined by the average risk that a rebellion will ignite and by the average duration of a war once it has started. If both the risk of ignition and the duration of war were constant over a long period, the global incidence of conflict would reach a self-sustaining level: the number of wars starting would be balanced by the number of wars ending, so that the stock of active civil wars would stay constant. Throughout 1950-2001 the average an- nual risk that a rebellion would ignite was around 1.6 percent, while the average annual probability that an ongoing war would end was 12.0 percent, corresponding to a median duration of wars of 5.5 years. If both these probabilities turned out to be persistent, then the global in- cidence of conflict would eventually settle at around 12 percent, which is roughly the global incidence of conflict in the last eight years. Fifty years ago the global incidence of civil war was clearly lower than 12 percent. This relatively peaceful period may have ended be- cause of fundamental changes in the underlying factors that cause civil war during the 1950s and 1960s. However, in the 1950s many low- 94 WHY IS CIVIL WAR SO COMMON' income countries were still colonies and colonialism suppressed the possibility of civil war. Independence has been bunched in two big waves, the British and French decolonizations of Africa in 1960-62 and the Russian decolonization of the early 1990s. If countries tend to be at peace cturing their first year of independence, there will be a long phase of adjustment after large numbers of countries have become in- dependent. Thus for much of the period the world has had an un- sustainably low incidence of civil war, and at least part of the rising incidence of civil war has been due to a movement toward the self- sustaining level. Note that a self-sustaining level need not be a desirable condition, but simply indicates the global incidence of conflict that the international community will eventually have to cope with unless it can reduce thc risk of rebellion and its duration. The observed rise in the global incidence of civil war from the 1950s to the 1970s need not of itself reflect a deterioration in the factors that cause and prolong conflict, but may simply reflect the existence of many moi-e independent, low-income countries. To illustrate this, fig- ure 4.2 sirnulates what would have happened to the global incidence of conflict since 1950 as a result of newly independent countries entering the system had all countries faced the actual average risk and duration of conflict: during the period. We assume for the time being that all countries have the same risks of conflict ignition and termination and Figure 4.2 Simulating the effects of the waves of decolonization, 1950-2020 Countries al war worldwide (percent) 20 - 15 - 10 - 5 - 1950 1960 1970 1980 1990 2000 2010 2020 Source Authors' calculations (see appendix 1) 95 BREAKING THE CONFLICT TRAP CIVIL WAR AND DEVELOPMENT POLICY FOgre 4.3 PmpofLon i cMg wiarz Q£ and ezch yeaw Percent 40- 30- 20- 10 1950 1955 1960 1965 1970 1975 1900 1985 1900 1995 2000 Source Gleditsch and others (2002). that all countries are at peace in their first year of independence.1 The initial distribution is the actual observed incidence in 1950, around 7.5 percent. In the simulation, waves of decolonization gradually push up the global incidence of conflict to a self-sustaining level of nearly 12 percent by 2020. The proportion of countries with new wars is shown with a darker shade in figure 4.1. No strong trend in the risk of new wars is appar- ent. Rather, the figure shows how wars have been steadily accumulat- ing, as the idea of the self-sustaining level implies. However, the rate at which wars end exhibits a disturbing trend. Figure 4.3 shows the rate of war termination during 1950-2001. From 1950 to the late 1980s conflicts became steadily less likely to end. This is why we observe a peak in the incidence of conflicts around 1990, as a surge of peace set- tlements took place in the first half of the 1990s, but unfortunately this seems to have been a temporary phenomenon. The most likely explanation for this surge in peace settlements is the end of the Cold War: many conflicts ended as that source of finance dried up, for example, in Mozambique. The end of the Cold War also allowed peacekeeping operations on an unprecedented scale. By contrast, other wars were made sustainable because of the inflow of weapons from the former Soviet republics (see chapter 3). This problem became important later in the 1990s, and may explain the reduction in effec- 96 WHY IS CIVIL WAR SO COMMON) Figure 4.4 The global self-sustaining incidence of civil war, by decades Percent 30 * New wars O Continuing wars 20- 10 1950-60 1961-70 1971-80 1981-90 1991-2001 Note Incidence of conflicts of the five decades decomposed into the share of years with wars that were new and with those that were ongoing Source Based on Gledirsch and others (2002) tive war terminations during the past five years. Overall, the net posi- tive effect of the end of the Cold War on war duration seems to have been modest and transient. The declining global risk of rebellion ignition and the lengthening duration of rebellion have together changed the self-sustaining global icidence of civil war. Had the risk and duration prevailing in 1971-80 persisted, the self-sustaining incidence would have been 11.5 percent, whereas had the risk and duration prevailing in 1990-2001 persisted, it would have fallen to 10.6 percent. Figure 4.4 shows the self-sustain- ing incidence based on the risks and termination rates for each decade. The 1980s stand out. If wars had continued to end at the same rate as in the 1980s, the incidence of war would have reached even higher lev- els than observed during that period. Fortunately, the improved success in ending conflicts in the 1990s prevented such a rise. Thus while the actual global incidence of civil war has risen over the past 40 years, the underlying self-sustaining incidence may have fallen slightly. The contradictory forces have been the large increase in the number of independent, low-income countries that find themselves playing the Russian roulette of conflict risk, versus the spread of eco- nomic development that has been making the world a safer place. 97 BREAKING THE CONFLICT TRAP CIVIL WAR AND DEVELOPMENT POLICY HIS DISCUSSION OF HOW THE UNDERLYING INCIDENCE OF civil war might itself have been changing starts by looking at changes in the risk that a rebellion will ignite. The models dis- cussed in chapter 3 attempt a systematic, empirical analysis of the factors that underlie this risk. Here we use Collier and Hoeffler's (2002c) model. The Collier and Hoeffler model obviously omits many important things, but tests for a pure time trend find that in aggregate, these things have not tended to get significantly worse or better over time. Changes in the risk of rebellion are therefore due to changes in the variables included in the model. Whereas the end of the Cold War clearly created a surge of peace settlements, it does not appear to have had a net effect on the risk of new rebellions. Controlling for 17 new low- and middle-income states, the risk of rebellion seems to have neither increased nor decreased. The downfall of the Soviet Union definitely let loose a few civil wars that had previously been repressed,2 but the end of the Cold War also cut off a source of finance for an unknown number of potential wars. Hence to understand the global changes over time we need to turn to the explanatory variables included in the model. Many of these vari- ables change only slowly or not at all, such as the ethnic and religious composition of a society and its geographic characteristics. The main factors that can change relatively rapidly are the economic variables. Recall that the three big economic drivers of rebellion are the level, growth, and structure of income. In addition, newly independent countries have a much higher risk of conflict than other countries. The very fact that they are new countries with weak institutions and often with a legacy of decolonization wars makes them five times more war prone in their first year of indepen- dence than comparable but older countries (Hegre and others 2001). If these new countries are able to sustain peace, this history of stability it- self gradually makes them safer. Moreover, most new countries are low- income, developing countries, with average income approximately half that of older countries. In sum, these two factors mean that newly inde- pendent countries face a risk that is 10 times higher than other countries. Globally, if we compare the 1960s with the 1990s these characteris- tics were very different. The countries that were independent in the 1960s typically had considerably higher per capita income by the 1990s, and this tended to reduce their risk of rebellion. Also working favorably 98 WHY IS CIVIL WAR SO COMMON) Figure 4.5 Factors changing the global risk of conflict Change in risk of civil war (log) 10 - Clhange in Change in Change in Consolidation primary income growth rate of peace commodity 0.1 - delpendence Note ContribLItions to the chanige in risk of civil war from 1965 to 1995 Source Based oni Collier and Hoeffler (2002c) was a decline in the average extent of dependence on primary com- modities. O)ffsetting this, growth rates were lower and new low-income countries had become independent. The Collier and Hoeffler model can be used to compare the typical risk of rebellion facing countries in 1965 and in 1995. It suggests that the typical risk declined from 9.2 percent to 6.8 percent in 1995. The main reason for this improvement was global economic development and the consolidation of new states. Figure 4.5 shows the overall reduction in risk and its constituent components. The growth in average per capita income and reduction in primary commodity dependence reduced the global average risk of conflict by something like 30 percent from 1965 to the mid-1990s. This reduction was offset by the lower growth rates relative to those of the late 1960s. The increase in the average duration of postindepen- dence peace is the factor that has made the strongest impact. This has lead to a 50 percent decrease in risk. What explains the trend in the duration of war? To look at this we use a model designed to study the duration of civil war (Collier, Hoef- fler, and S6derbom 2003). As with the risk of rebellion, the duration of conflict may have changed over the past 40 years either because of changes in the variables included in the model or because of changes in factors that are important but are omitted. Whereas no significant time trend in the risk of conflict ignition was apparent, its duration shows a substantial time trend: after controlling for the explanatory variables, 99 BREAKING THE CONFLICT TRAP CIVIL WAR AND DEVELOPMENT POLICY Fogare 4.6 Thae cDhamgoIg mfts d coniMcK f¶mI1oflaf Change in rate of termination (log) 10 - Change in income since the 1960s 0 Due to changes in unobserved factors Net effect 1- 1SS6s 1970s 1930s 199as Source Collier, Hoeffler, and Soderbom (2003) conflicts were harder to end in the 1980s than in earlier periods. They may also have been more persistent in the 1990s. By definition, the model cannot tell us why this has happened: it is due to factors not in- cluded in the model. In addition, some of the variables included in the model have changed. The higher per capita income is, the shorter the civil war. Re- call that this might be expected for various reasons, namely, civil war is costlier at higher income levels, and thus the incentive to reach a set- tlement is stronger. Whatever the explanation, the strong rise in global per capita incomes has tended to shorten the duration of wars. The overall change in the termination rate of conflict is thus the net effect of the unexplained lengthening of conflict, decade by decade, and the shortening of conflict resulting from global economic growth. Fig- ure 4.6 shows the net effect, decade by decade. Overall, the unexplained effect has more than offset the favorable effect of global growth, there- fore the duration of conflict has increased. UIMPmUI E Mell Qobd @1 ndttU e@(pmce (m MA1 Muq S O FAR WE HAVE FOCUSED ON GLOBAL AVERAGES. AVERAGES often conceal wide dispersions, and sometimes they also conceal important structural differences. This is the case with conflict. 100 WHY IS CIVII WAR SO COMMON; Divergent Development Trends For the past 20 years global growth has been raising incomes in much of the developing world and reducing the incidence of poverty. Much of the world's population now lives in middle-income countries, defined as those with per capita annual income above US$745. The structure of developing countries' exports has also changed dramatically. Whereas in 1980 primary commodities still accounted for three-quarters of exports, they now account for only 20 percent. Some low-income countries, in- cluding the largest, have succeeded in implementing and sustaining pol- icy reforms conducive to rapid growth and integration into global mar- kets. While currently they are still low-income countries, they are on track to joining China in becoming middle-income countries. Recall from chapter 3 that fast growth is not just a route to the eventual low risk that goes along with higher income levels, but also contributes di- rectly to risk reduction. We therefore aggregate those low-income coun- tries that have sustained reasonably good economic policies with the middle-income countries and term the combined group "successful de- velopers." Specifically, we include all those low-income countries that over the 1.990s averaged CPIA scores of 3.5 or better.3 Some of the suc- cessful developers are still at high risk of conflict, but as a group they are already muich safer than other developing countries and are on course for continuing reductions in risk. Many developing countries have not, however, participated in these favorable trends. They have either been unable to implement reform or their reforms have not been sustained and they remain stuck in undi- versified primary commodity exports. We refer to this group as the "marginalized" low-income countries. The growth rates of per capita in- come in the two groups of countries were dramatically different in the 1990s, negative at -1.0 percent for the marginalized countries and pos- itive at 2.0 percent for the successful developers. The average level of in- come in the marginalized countries was less than a third that of the suc- cessful developers when measured on a purchasing power parity basis. Thus in aggregate, the marginalized countries are the one group that has all three of the economic characteristics that appear to increase proneness to conflict: low income, economic decline, and dependence on primary commodities. The following section compares the risks and incidence of a civil war for a typical marginalized country with that for the typical successful developer. Figure 4.7 shows the predicted risk for 101 BREAKING THE CONFLICT TRAIP CIVIL WAR AND DEVELOPMENT POLICY [Figure 4.7 oovergenQ rMshs: Gr'Eft:3> cos Fe b SE:ZG563 deve0opers Risk of cMI war (log) 100 10 Pimary Oncome Growth Consolidation Population Total commodriy rate of peace size difference dependence Source Based on Collier and Hoeffler (2002c). the typical marginalized country relative to the typical successful devel- oper and the contribution of some important risk factors. The pre- dicted risk is more than 10 times higher for the marginalized country. Low income has the largest impact, accounting for half the difference. To date global development has largely missed the marginalized coun- tries; thus, while global growth is indeed reducing the global incidence of conflict, it is doing so dramatically unevenly and cannot be relied on to secure a peaceful world. If the trends of the past 20 years continue, the successful developers will evolve into low-risk societies while the mar- ginalized countries will face increasing risks as their per capita incomes dedine. Figure 4.8 shows how the predicted risk of civil war ignition evolves for the marginalized countries and successful developers relative to the high income countries if recent growth patterns persist. Global growth is part of the process of reducing the incidence of civil war, but unless it reaches the currently marginalized countries it will progressively become less effective as a force for peace. As the successful developers evolve into a group with lower risks of rebellion, the increment to peace achieved by further growth and diversification in income becomes smaller and smaller. Global growth is not sweeping the world into peace at an accelerating rate. If present trends continue its contribution to peace will fizzle out well before global peace has been achieved. 102 WHY IS CIVIL WAR SO COMMON' Figure 4.8 Development of risk of civil war for the marginalized and successful developers, 2000-2020 Risk of civil war (log) 25 - 20- MIarginalized countries 10 5- Successful developers 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018 2020 Note The contribution from growth to income per capita to the odds of war onset relative to high-income countries, with growth projections of 1 5 percent (high income), -1.0 percent (marginalized), and 2.0 percent (successful developers) Source Calculated from Collier and Hoeffler (2002c). The radically different risks the successful developers and marginal- ized countries face imply different incidences of civil war for the two groups in the long term. As long as they remain stagnant, the margin- alized countries will remain at the incidence experienced during 1990-2001, whereas the successful developers will slowly but surely re- duce their incidence from their current somewhat lower level. Changes in the global incidence depend both on these two divergent trends and on the relative size of the two groups. The successful developer group is largest in terms of both number of countries and population-71 countries with around 4 billion people-versus 52 marginalized coun- tries with around 1.1 billion people. Nevertheless, the global incidence of civil war will increasingly come to be dominated by wars in the group of poor, declining, primary-commodity exporting countries as the incidence of war in the successful developer group decreases. Implicationis of the Conflict Trap Chapter 3 introduced the concept of the conflict trap. Through various routes, once a conflict has started a society faces a greatly increased risk 103 BREAKING THE CONFLICT TRAP CIVIL WAR AND DEVELOPMEN r l'OLICY of further wars. Conflicts are hard to stop, and what happens during conflict increases both the risk and duration of subsequent conflict. Countries that have had a war have a two to four times higher risk of a subsequent war, even when controlling for the factors we identified ear- lier. Boxes 4.1 and 4.2 describe two recurrent conflicts. Figure 4.9 indicates how the risk of war ignition is altered after a civil war compared with before a war. The risk depends on how long the country has been independent and at peace. In the first month of postindependence peace the risk of war is more than four times as high as after a decade of peace. After the first decade of consolidation, the risk does not change much as time goes by; however, if a civil war breaks out the gain from this consolidation is lost. After the war, the risk of war re-igniting is two to four times higher than the risk facing new states. This is the conflict trap: a country that first falls into the trap may have a risk of new war that is 10 times higher just after that war has ended than before the war started. If the country succeeds in maintaining postconflict peace for 10 years or so, the risk is consider- ably reduced, but remains at a higher level than before the conflict. This legacy of war seems to take a long time-a generation or two- before withering away (Hegre and others 2001). THE WAR IN AFGHANISTAN STARTED IN 1978 WHEN From 1992 until 1996 the war was waged by the members of the Marxist-Leninist People's Democra- Pashtun-dominated Taliban seeking to overthrow tic Party of Afghanistan captured the state; assassi- Burhanuddin Rabbani and his Tajik-dominated nated political, ethnic, and religious elites; and in- Jam'iyat-i-Islam party. After a Taliban victory in cited uprisings (Asia Watch 1991). After the Soviet 1996 a new war started in which the Tajiks, Uzbeks, invasion of December 1979 and the assassination of and others became the insurgents against the new Afghan president Hafizullah Amin, the war contin- "government" (Gurr, Marshall, and Khosla 2001). ued with mujahideen fighting against the Soviet- In 1997 the Taliban proclaimed the Islamic Emirate installed Afghan government of Mohammad Na- of Afghanistan, which was recognized by Pakistan, jibullah. In 1992 the mujahideen captured power Saudi Arabia, and the United Arab Emirates. The and the state changed hands, but peace negotiations Taliban never controlled all the territory of Af- among Afghanistan's many resistance factions ex- ghanistan, and about 5 to 10 percent of that terrn- cluded key parties. One such group was Gulbuddin tory was controlled by the alliance known as the Hikmatyar's Hizb-i Islami, which rejected the result- United Front, formed in 1996 by non-Pashtun ing agreement and began a series of rocket attacks on groups opposed to the Taliban and led by Rabbani's Kabul that continued into 1995 (Hiltermann 2002). former defense minister, Ahmad Shah Massoud. 104 WHY IS CIVIL WAR SO COMMON' IOz 4.2 RecMrent cocQ$ nuz mpu 2b A\nWgoga A PATTERN OF FAILED PEACE AGREEMENTS IN than 300 areas previously returned to the gov- Angola has checkered a history of civil war that has ernment, but by the end of 1999 the government, been ongoing since the country's independence in with the support of Namibian government forces, 1975. The war against UNITA over control of the had overrun UNITA's former headquarters (Parker, central government from 1975 until 1994 caused Heindel, and Branch 2000). Thereafter, UNITA's approxtmately 345,000 deaths and ended in a stale- military position continued to deteriorate because mate that led to the Lusaka Accord and the deploy- of a double squeeze. The government used the ment of a UN peacekeeping force. Failure to imple- opportunity of high oil prices to increase military ment the agreement led to a renewal of war in late spending. At the same time the Fowler Report of 1997. 'he U.S. State Department noted more than the UN exposed the routes by which UNITA had 100 ceasefire violations in a three-month period in been financed and supplied, as a result of which 1996. Despite that instability, the period 1996-97 it was closed off. In February 2002 Savimbi was was one in which UNITA officials were becoming cornered and killed and UNITA accepted a peace increasingly integrated into the government, and settlement. The Angolan government was able to the annual death toll during this time was probably negotiate from strength. More than 10,000 people "only" in the low hundreds. Fighting resumed in were killed in the new round of fighting, and ac- March 1998 despite an agreement reached on Jan- cording to the United Nations Children's Fund, uary 9, 1998, for resolution of the remaining issues nearly 75,000 people died of starvation in 1999 and under the Lusaka Accord UNITA leader Jonas at least 1.5 million people were displaced as of Jan- Savimbi refused to move to the capital and )oin the uary 2000 because of the war (Parker, Heindel, and government. UNITA forces quickly retook more Branch 2000). This increase in risk is before we account for the changes in the ob- servable risk factors caused by the war itself. In particular, the impact of the civil war on the economy is extremely damaging (see chapter 1). Growth of GDP per capita is reduced by around 2.2 percentage points per year during war. Moreover, the effects of the war linger on after the conflict, so that the country's economic performance is hampered for several years after the conflict has ended. Only after extremely long conflicts, for example, in Mozambique, where disruption is so complete that the mere fact that large numbers of people return to work shows up as a significant improvement, will a peace agreement mean an im- mediate improvement in growth performance. Hence the typical con- flict reduces income by some 10 to 15 percent. Such losses in income are also often associated with an increase in primary commodity de- pendence of roughly two percentage points (Collier and Hoeffler 2002b). These two changes imply an increase in the risk of war onset of an additional 5 percent. 105 BREAKING THE CONFLICT TRAP CIVIL WAR AND DEVELOPMEN T POLICY Hlgluije 4.O TDe a!! tiraip: reis i o c,r tuair lwralve Qo a ccmD w Risk of cMI war (logl Nust afer One decade after lTwo decades after independence independence, independence, one just after wiar decade after twar Source Hegre and others (2001); Collier and Hoeffler (2002c). Some risks also arise from neighbors in conflict, so to some extent the conflict trap operates at the level of a neighborhood, not just of a single country. Quantitative studies of civil war onset find no evidence that civil wars are more frequent in countries bordering on conflict countries, controlling for the explanatory variables (see, for example, Hegre and others 2001); however, civil wars spill over indirectly through their ef- fects on the explanatory variables such as income (Murdoch and Sandler 2002). Reduced income in neighboring countries indirectly increases their risk of conflict, and as most countries have several neighbors, in aggregate, such small increases in risk can have significant effects. The conflict trap is a tendency, not an iron law. Middle-income countries have a lower probability of falling into it. A previous conflict seems to increase the risk for middle-income countries by the same fac- tor as for low-income countries, but as they have a lower general risk, they have better chances of maintaining peace beyond the first post- conflict decade. Figure 4.10 summarizes the predicted risks of war ignition and re-ignition for the typical country in each group. Figure 4.11 decomposes the effect of the conflict trap into the eco- nomic factors that change as a consequence of the conflict and other unobserved factors that change during the war. Such other factors are the accumulation of weapons and military organizations and less tan- 106 WHY IS CIVIL WAR SO COMMON' Figure 4.10 The conflict trap by type of country Annual probability of outbreak of war (percent) 10 - _3 Postconflhct penods E0 Peace penods 5- 0 , , , , ,,- ,,, l~I Marginalized countries Successful developers Source Based csn a revised version of Collier and Hoeffler (2002c) Figure 4.11 Risk components for marginalized countries in the conflict trap, relative to the same countries preconflict Risk of civil war (log) 10 - 1 ~_ _ I Priniary Income Growth Previous Population Total commiodity rate conflict size difference dependence Source Based on a revised version of Collier and Hoeffler (2002c) gible effects of war, such as the breakdown of institutions and social po- larization. The figure compares the risk of the typical marginalized country that has not had a war for 10 years and the typical postconflict marginalized country. The postconflict country has a risk of conflict that is five times greater. Around half of that increased risk is due to 107 BREAKING THE CONFLICT TRAP CIVIL WAR AND DEVELOPMENT POLICY negative changes in primary commodity dependence and reduced in- come and growth. The other half of the increased risk is unexplained and will in part be due to the selection problem: conflict countries al- ready had unobserved characteristics that increase the risk of conflict. The conflict trap has implications for the global incidence of con- flict. The countries most prone to the trap are the marginalized low- income countries. Although poor, peaceful, stagnant economies look as if they are stuck in an equilibrium, they are, in effect, playing Russian roulette. A low-income, stagnant country that starts its independence at peace does not have a very long expected duration of that peace, al- though some countries, even though economically stagnant, have to date preserved peace, whether through prudent policies or good for- tune, for example, Malawi, Tanzania, and Zambia. However, even long periods of peace are no guarantee of safety. C6te d'Ivoire and Nepal are recent instances of moderately democratic low-income countries with long histories of peace collapsing into civil war. The marginalized stagnant but peaceful countries are thus living dangerously. Not only are they prone to civil war, more important, once a war has started they also face a permanently changed risk of con- flict, that is, they are stuck in the conflict trap. The poor but peaceful category of countries, although currently numerous, is thus not likely to be so numerous in a global self-sustaining level. We would expect these countries either to develop, joining the successful developers, or at some stage succumb to civil war, with many then becoming trapped in conflict. In the long run poor but peaceful is not an option. The world is therefore evolving into a state in which most countries are per- manently conflict free while a minority are trapped in a cycle of lengthy war, uneasy peace, and reversion to lengthy war. This leads to a different view of the self-sustaining incidence of con- flict, with radically different risks for different groups. The high-income countries have a negligible risk of civil war. A second group of coun- tries, a majority, will be in a virtuous circle of peace, with income ris- ing strongly and diversifying out of dependence on primary commodi- ties. These countries will face a low and declining risk of conflict. The few civil wars that occur in this group may be long, but they will tend not to trap countries into a cycle of conflict. A third smaller group of countries will be stuck in a conflict trap. Although they may periodi- cally reach peace, the legacy of the conflict is such that peace is not sus- tained. Occasionally countries will switch between these groups. Once 108 WHY IS CIVIL WAR SO COMMON? in a while a peaceful and prosperous country might collapse into civil war and find itself subsequently stuck in a conflict trap. Similarly, once in a while a country that has been mired in repeat conflict will climb out of it. A. fourth group, the poor but peaceful, will hover in between development and the conflict trap. Figure 4.12 illustrates the trap.4 We grouped 156 countries with ad- equate data coverage into the four groups of countries. Here we regard countries as in a postconflict state during the first decade after a war has ended, and as at peace if they have not had a war in the past 10 years. We estimated the predicted risks of war for a typical country in each group. The risk is a function of levels of income, primary commodity dependence, growth, and the other characteristics found to be pertinent in chapter 3. The risk changes over time after independence or war. The model was estimated for the 1960-99 period. For the typical low-income country, the predicted probability of going to vvar from a state of peace is 2 percent per year, whereas the Figure 4.12 The conflict trap in 2000: annual flows into and out of conflict High income, at peace (32 countriesl ,sr-.i ' . - z 0 05 SMucseful,- 0.3 0.7 odiope is- => A ctive conflict Marginalized OR peact"'4 17, countriesl countries 164.countcresl - at peace '- ' ' r'!-F,_s,8-, A_ 0 (32 countries} *X < 2 0 \/ 7S1 0 0.~3 0.7 Postconflict 112;countries) Strongly | developing ISagnant Note Numbers next to the arrows indicate the number of countries per year that move between thie different st,ites of conflict in the self-sustaining state Numbers in boxes indicate the self- sustaining ntmiber of countries in each conflict state See appendix I for fuller coverage Source Based on a revised version of Collier and Hoeffler (2002c) 109 BREAKING THE CONFLICT TRAP CIVIL WAR AND DEVELOPMENT POLICY probability of war from a postconflict situation is about 10 percent per year. We adopt a median duration of wars of about five years. Just as we computed a self-sustaining incidence of war from the probabilities of starting and ending wars, we can compute the self-sustaining distribu- tion of peace, war, and the postconflict state for the typical low-income country: it is predicted to be at war 24 percent of the time, in a post- conflict state 15 percent of the time, and at peace 61 percent of the time. The corresponding predicted distribution for a typical middle- income country is 5 percent, 5 percent, and 90 percent, respectively. Figure 4.12 simulates how this self-sustaining distribution will be re- flected in global numbers of conflict onsets and fall-backs. Seventeen countries are predicted to be involved in a civil war, of which 15 are on- going wars. Half of the conflict onsets will be from the group of post- conflict countries. In the simulation, there is one re-entry into war every year. Each year 0.7 low-income countries will go from a state of estab- lished peace to war, whereas only 0.3 peaceful middle-income countries will do so. As this pattern is self-sustaining, two wars end every year and one country leaves the postconflict period in peace. The simulation is only an approximation, for example, it abstracts from differences within each group. However, we would expect that continued divergence in growth rates between the successful developers and the marginalized countries would gradually alter the structure of global risks. Figure 4.13 shows how given this scenario of stagnation for some and growth for others the global incidence of civil war would evolve by 2020 and by 2050. If these projections are broadly correct they carry a disturbing mes- sage. A further 50 years of development along past trends will have lit- tle impact on the global incidence of civil war: the number of civil wars declines from around 17 to around 13. This disappointing outcome is because the outbreak of war becomes increasingly concentrated in the marginalized and postconflhct countries, with their combined share of global conflict rising from 82 percent in 2000 to 94 percent by 2050. The Chaniiging egiOuGA EaeNM The incidence of civil war has differed dramatically across regions. In part, this is because the countries in a region tend to have many features in common and some of these features affect the risk of conflict. In 110 WHY IS CIVIL WAR SO COMMON' Figure 4.13a The conflict trap in 2020: annual flows into and out of conflict High income, at peace (32 countries) 0 05 4SUccEMiI 0 2 0.7 ' odevpelop - t = 4Ative conflict Marginalized ('15 countriesl countries '166,countres at peace (32 countries) :X;g,'- ~~~- < t.9\/7 0. 7 0.2 - Lj0.2 0.7 -Postconflict 11 countries) developing I Stagnant Figure 4.13b The conflict trap in 2050: annual flows into and out of conflict High income, at peace (32 countries) '0.05 * d:eIoperS ,=r==k> ;Aetive conflict Marginalized :.'iie~ii~' 1 countries) This document did not complete OCR process. <==**"