DOES CONFLICT BEGET CONFLICT? Explaining Recurring Civil War

DOES CONFLICT BEGET CONFLICT? Explaining Recurring Civil War Barbara F. Walter Graduate School of International Relations and Pacific Studies University of California, San Diego 9500 Gilman Drive La Jolla, CA 92093 (858) 822-0775 bfwalter@ucsd.edu June 2002 Word Count: 8,632 1

Abstract This article seeks to explain why some countries experience recurring civil war while others do not. It argues that renewed war has less to do with the attributes of a previous war, as many people have argued, and more to do with the incentives individual citizens have to join rebel groups at any given point in time. Civil wars will have little chance to get off the ground unless individual farmers, shopkeepers, and workers choose to enlist in the rebel armies that are necessary to pursue a war, and enlistment is likely to be attractive when two conditions hold: the status quo for the average citizen is perceived to be worse than the possibility of death in combat, and there is no non-violent outlet for political change. An analysis of all civil wars ending between 1945 and 1996 suggests that improvements in basic living conditions and in the average person s access to political participation have a significant negative effect on the likelihood of renewed war. Countries that are able to increase the economic well-being of their citizenry, and create a more open political system are less likely to experience multiple civil wars regardless of what happened in a previous conflict. 2

Civil wars create what has been called a conflict trap. 1 Societies that have experienced one civil war are significantly more likely to experience a second or third war than are societies with no prior history of violence. Indonesia, Iraq, Burundi, Rwanda, Sri Lanka, and Iraq, for example, have all experienced recurring civil wars where violence broke out not once, but repeatedly over time. 2 Empirically, however, we know that most civil wars are not destined to repeat themselves. Between 1945 and 1996 only thirty-six percent of civil wars were followed by an additional war. 3 Single civil wars, like those that broke out in Argentina, Greece, and Costa Rica, are more the norm. Given this variation, what explains why some countries experience recurring civil war while others do not? What little has been written about civil war recurrence has tended to focus on the characteristics or attributes of the previous war to explain why a second or third war might occur. 4 In this view, earlier wars set the stage for conflicts that occur thereafter, either because the original grievances failed to be resolved, or because war itself aggravates the conditions that made violence break out in the first place. Wars fought over difficult-to-resolve issues such as ethnic identity, or indivisible stakes such as the complete overthrow of a government are viewed 1 See Paul Collier and Nicholas Sambanis, Understanding Civil War: A New Agenda, Journal of Conflict Resolution, Vol. 46, No. 1 (February 2002) p. 5. 2 Across cases, there was no consistent pattern in the timing of renewed war. Countries experienced renewed civil war as soon as two years after one war ended, as was the case in Iran, and as long as thirty-three years later, as was the case in Iraq. The median duration of peace in conflicts that occurred between 1945 and 1996 was fourteen years. Thus, even a decade of piece was no guarantee that a country would not experience civil war again. 3 Of the fifty-eight cases of civil war that ended between 1945 and 1996, twenty-one experienced renewed war. 4 Almost all of these studies, however, look at the narrower question of negotiated peace settlements and their effect on peace duration. See for example Page Fortna, A Peace the Lasts: Agreements and the Durability of Peace, Ph.D. dissertation, Harvard University, 1998; Suzanne Werner, The Precarious Nature of Peace: Resolving the Issues, Enforcing the Settlement, and Renegotiating the Terms, American Journal of Political Science, Vol. 43, No. 3 (July 1999); Caroline Hartzell, Matthew Hoddie, and Donald Rothchild, Stabilizing the Peace after Civil War, International Organization. Vol. 55, No 1 (Winter 2001); and Amitabh Dubey, Domestic Institutions and the Duration of Civil War Settlements, unpublished paper, Columbia University 2002. 3

as more likely to resurface, as are particularly brutal wars that tend to exacerbate grievances and hatreds. In what follows, I argue that renewed war has less to do with the attributes of a previous war, and more to do with the incentives individual citizens have to join rebel groups at any given point in time. The fact that rebel armies are almost entirely dependent on the willingness of individuals to volunteer for battle, places the onus for war on ordinary people and the trade-offs they must make for going to war or staying at peace. Civil wars will have little chance to get off the ground unless individual farmers, shopkeepers, and workers voluntarily choose to man the rebel armies that are necessary to pursue war, and it is the underlying political and economic conditions that make enlistment attractive that this article attempts to uncover. Only if we identify these micro-level motives for enlistment can we begin to explain why civil wars arise in some countries and not others. 5 An analysis of all civil wars that ended between 1945 and 1996 suggests that changes in basic living conditions and in the average person s access to political participation have a significant effect on the likelihood of renewed war. Countries that are able to increase the economic well-being of their citizenry, and create a more open political system are significantly less likely to experience multiple civil wars regardless of what happened in the previous conflict. In contrast, the issues and grievances that started an earlier war, the degree to which these grievances are resolved, and the cost of that war, have little effect on whether another war 5 I am not the first person to identify rebel recruitment as an important factor in civil war. Both Elbadawi (2001) and Collier and Hoeffler have identified the differential ability to recruit as a possible cause of war, and Gates (2001) has argued that rebel retention is a key variable in the duration of civil war. For a classic argument on the importance of rebel recruitment see Mao Tse-tung, On Protracted War, Selected Works of Mao Tse-tung, Vol. II, Peking: Foreign Languages Press, 1967, pp. 113-194. For arguments that focus on micro-level motives more generally, see work by Kalyvas, Keen, Peterson, Lohmann, Kuran, and Granovetter. 4

surfaces. This suggests that micro-level motives do come into play in the outbreak and recurrence of civil wars. Explaining Recurring Civil War Most explanations for repeat war fall roughly into one of three camps: (1) those that focus on why the original war began, (2) those that focus on how the original war was fought, and (3) those that emphasize how the original war ended. Why the Original War Began Scholars of civil war have long argued that some types of disputes are more difficult to resolve than others and thus more likely to lead to repeat confrontations. Civil wars that are fought between competing identity groups are believed to be particularly intractable since, as Ted Gurr has observed, cultural identities those based on common descent, experience, language, and belief tend to be stronger and more enduring than most civic and associational identities. 6 Once war breaks out, ethnic identities and hatreds tend to become cemented in ways that make a return to a peaceful co-existence difficult, and these are the wars that are likely to recur over time. 7 Rebel demands are also thought to affect the likelihood of long-terms solutions to war. Rebels fighting for total goals such as the elimination of a rival, or the revolutionary overthrow of an entrenched political, economic, or social system are less apt to find a stable long-term solution than rebels fighting for more limited aims. As Stephen Stedman has pointed out, some leaders will settle for nothing less than complete control over the country in question. 8 If this is 6 Ted Robert Gurr, Peoples Versus States: Minorities at Risk in the New Century. (Washington D.C.: U.S. Institute of Peace, 2000) p. 66. 7 See also Donald Rothchild and Alexander J. Groth, Pathological Dimensions of Domestic and International Ethnicity, Political Science Quarterly, Vol. 110, No. 1 (Spring 1995); and Chaim Kaufmann, Possible and Impossible Solutions to Ethnic Civil Wars, International Security. Vol. 20, No. 4 (1996). 8 Stephen John Stedman, Negotiation and Mediation in Internal Conflict, in Michael E. Brown ed., The International Dimensions of Internal Conflict, (Cambridge: MIT Press, 1996) p. 345. 5

the case, then wars might end temporarily, but will re-emerge as soon as one or both sides have regrouped sufficiently to seek a more satisfying solution. How the Original War Was Fought A second set of arguments focuses on the costs of a previous war to determine whether combatants are likely to return to the battlefield. Three mechanisms in particular are purported to make war more or less likely in the aftermath of especially costly fighting: the desire for retribution, increased certainty about combatant capabilities, and combat weariness. In a preliminary study of violence in civil wars Stathis Kalyvas found that personal vengeance was a recurrent motive for participation in war. 9 Intense and long-lasting wars, therefore, could exacerbate the animosity between combatants and create a strong need for retribution even after that war ends. 10 If this is true, then combatants involved in more costly wars should be more likely to resume their war than those who inflicted fewer costs and sustained less suffering. Costly wars, however, could have the opposite effect. Countries that have experienced a particularly devastating civil war may actually be less vulnerable to a second or third war because supplies have been exhausted, soldiers fatigued, and popular support used up. 11 Former United States senator George Mitchell, for example, has repeatedly stated that war weariness will eventually bring a solution to the Middle East conflict, just as it did in Northern Ireland. According to Mitchell, people in Northern Ireland had been especially sick of the moving and 9 Quote taken by Kalyvas (2000) from Benjamin D. Paul and William J. Demarest. The Operation of a Death Squad in San Pedro la Laguna, in Harvest of Violence: The Maya Indians and the Guatemalan Crisis, ed. Robert M. Carmack. (Norman: University of Oklahoma Press, 1988). 10 These more violent wars may also indicate a greater willingness (or tolerance) on both sides to pay the costs of war. Divisions in these wars may be so intense that they are unlikely to subside even well into the future. 11 See Steven Rosen, War Power and the Willingness to Suffer, in Peace, War, and Numbers, Bruce Russett ed. (Beverly Hills, CA: Sage Publications, 1972). 6

emotional sight of those small white coffins holding innocent children being buried I think the same thing s going to happen in the Middle East. 12 The costs of conflict, however, may also provide important information to future combatants about the government s relative strength and their own chances of winning a war, making future encounters less likely. Alastair Smith and Allan Stam, for example, have argued that wars that end quickly, leaving one or both sides still quite uncertain about the true balance of power, are much more likely to reopen than those wars that end not only with agreement about the balance of power but also little doubt about the certainty of that agreement. 13 In this case, long and intense wars act as an important information source, with more accurate information helping to prevent renewed conflict in the future. How the Original War Ended Other scholars point to how the original war ended to predict whether it will start again. Two arguments in particular are made. The first focuses on the degree to which combatants are able to resolve the grievances driving the war. Governments that are willing to specifically address and settle key issues underlying a conflict are believed to have a higher chance of preventing renewed wars than governments that leave important issues unresolved. 14 The second argument points to the balance of power that exists between combatants at the end of the war to determine whether war will recur. 15 Large power imbalances (such as those that would result if one side decisively beat the other) create few opportunities for the weaker side to launch a renewed attack, and affect expectations about the likely outcome of future wars. In these cases, 12 Reuters, Mitchell: Weariness will help end Middle East conflict, Jerusalem Post. October 29, 2001, p. 2. 13 Alastair Smith and Allan Stam, Bargaining and the Nature of War, Unpublished Paper, Yale University, January 2002, p. 7. 14 For arguments that focus on the specific issues under dispute see Hensel 1996; Bennett 1997; Goertz and Diehl 1995, 1998; and Diehl and Goertz 2000. 15 See especially Zartman, 1989, 1995; and Wagner 1993, 1994. 7

the losing side has no hope of eventually defeating its opponent and is likely to remain quiet as a result. In contrast, undefeated combatants retain the capabilities and perhaps the hope to seek better terms in the future, especially if dissatisfied with any new political system that emerges. Argument In what follows, I explore a different story for why civil wars might recur in some countries and not others. It rests on the simple observation that for civil wars to occur, hundreds or thousands of individual citizens must voluntarily enlist in a rebel army. 16 This means that renewed civil war is likely to have less to do with factors associated with the last war (such as unfulfilled demands), and more to do with the direct and immediate incentives ordinary citizens have to pick up a gun and begin to fight. The basic story is this. Unlike standing armies that exist in independent countries to fight at a government s discretion, rebel armies must be recruited and re-mobilized for each individual civil war. Rebel leaders do not have the luxury to call on a standing army or to forcibly conscript citizens should grievances or opportunities for rebellion arise. Instead, civil wars require the cooperation of a sizeable number of citizens, almost all of whom are likely to base their decision to enlist (or not enlist) on very personal cost calculations. Enlistment, therefore, is likely to become attractive when two conditions hold. The first and most important is a situation of individual hardship or severe dissatisfaction with one s current situation. The status quo must be perceived to be worse than the possibility of death in combat, a condition I call misery. 17 The 16 This is not to say that all rebel soldiers join voluntarily. There are well-documented cases where individuals and children are forced into military service by rebel leaders using threats and coercion. It also does not mean that individual incentives for rebels to enlist are the only factors necessary for civil wars to break out. For civil wars to occur, leaders must emerge to coordinate and manage recruitment, and resources and supplies must be available to support the movement over time. For discussions on other issues necessary to recruit and sustain rebel organizations see especially Gates (2002); and Collier and Hoeffler (2001). 17 Not all individuals will have the same tolerance for the costs and risks of war. Some citizens will choose to enlist earlier than others, and some not at all. For a fascinating literature on the distribution of cost and risk thresholds and 8

second is the absence of any non-violent outlets for change. Violence must be perceived as the only available tool for the average citizen to improve the status quo, a condition which can be termed lack of voice. Because rebel recruitment is vital to the emergence of war, countries with increasing levels of individual hardship, and decreasing outlets for non-violent change should be more likely to experience recurring civil war, regardless of the characteristics of past conflicts. States with improving standards of living and increasing outlets for non-violent change should not. 18 Research Methodology The aim of this paper is to present and test a theory of civil war recurrence that highlights the importance of individual incentives to fight. It is, in other words, an attempt to emphasize the importance of the average person in our understanding of civil wars. Competing theories were tested against every civil war that ended between 1945 and 1996 as based on the coding criteria proposed by the University of Michigan s Correlates of War (COW) project. 19 Measuring the Dependent Variables In order to see which of the theories presented above help predict the conditions under which countries will experience repeated civil war, I created a binary variable coded 1 for renewed war, and 0 otherwise. 20 A war was coded as a Renewed War if it was fought by the same combatants for the same goals as the original war, and met all the coding criteria outlined on the factors that trigger the timing of certain types of protest and violent behavior see works by Kuran, Granovetter, and Lohmann. 18 The observation that declining living standards and a lack of openness are likely to lead to civil war is not new. A number of large studies have found that different measures of quality of life are significantly related to the outbreak of war. What is new is the observation that these factors are important because civil wars require volunteers for their very existence. 19 The Correlates of War dataset is on deposit at the Inter-University Consortium for Political and Social Research, University of Michigan and includes data on all interstate and civil wars fought between 1816 and 1992. For a detailed explanation of the rationale and procedures used in collecting and coding these data see Melvin Small and J. David Singer. Resort to Arms: International and Civil Wars, 1816-1980. (Beverly Hills: Sage Publications, 1982). See Appendix A for a list of the fifty-eight conflicts that met these criteria. 20 See Appendix B for a list of data sources for each of the variables included in the study. 9

above. 21 A number of countries such as Burma and Iraq, however, experienced multiple unrelated wars. Since it is possible that the factors that lead the same set of combatants to renew a conflict may be very different from the factors that lead a new set of combatants to initiate a conflict, I created a second dependent variable, Subsequent War, which was coded 1 for any recurring war regardless of who the combatants were and what their goals were and 0 otherwise. Finally, since it is important not only to understand if a renewed war occurs but when it is likely to happen, I undertook some basic analysis of peace duration. In this case, I simply recorded the number of years that peace lasted in each country and used that as the dependent variable. Measuring the Independent Variables Several measures were developed to test each of the four sets of theories outlined above. To determine if the issues under dispute affected the likelihood of war repeating itself, two different variables were included in the analysis. To see whether wars fought over identity, and in particular ethnicity, were likely to spark repeat conflict, each of the wars was classified based on whether a clear ethnic division existed between the combatants. If the combatants broke down along ethnic lines, or that faction defined itself as a separate ethnic group, it was coded as an Ethnic Civil War; all other wars were coded as non-ethnic. 22 In addition, a measure of the Ethnic Heterogeneity of each country was included in the model to see if countries with more diverse populations were prone to a greater number of conflicts. 21 Most of the renewed wars were fought between the same set of combatants: thirteen of the twenty-one renewed wars were essentially mirror images of the first war. In these thirteen cases, combatants stopped fighting and often signed detailed agreements that outlined the transition to peace and the new form of the government only to go back to war sometime in the future. 22 A different measure of identity - an Ethnolinguistic Fractionalization Index was included as an additional control. The results remained the same. 10

The demands of the rebels were also recorded to determine what if any effect total goals have on the likelihood of repeat wars. If the rebels initiated the war to obtain anything less than total control over the government (i.e., political reform, regional autonomy, territorial secession), the war was coded as involving non-total goals. 23 These conflicts were expected to have a lower rate of recurrence than those fought in pursuit of more absolute goals. All other wars were coded as having Total Goals. 24 The second set of independent variables focused on the impact of war costs on renewed war. High costs were expected either to incite greater hatred, or have the opposite effect and deter citizens from returning to war. To engage both of these possibilities, three variables were included in the analysis. The first, the number of War-Related Deaths that took place during a war, was measured as a continuous variable that ranged from a low of one thousand deaths, to a high of more than a million and a quarter deaths. 25 The second, the Duration of War, was measured as a continuous variable which varied from a low of one week to a high of 396 months. The third criterion, Displaced People, measured the number of people displaced both internally and externally due to the war. The final set of variables focused on factors associated with the resolution or outcome of war. One logic suggests that if grievances surrounding the first war are resolved, renewed war should be less likely. Two measures were used in an attempt to isolate the effect of Grievance 23 Cases were coded based on the stated aims of the rebels at the beginning of the conflict rather than the stated goals of the government since it is the rebels who almost always initiate a war and are therefore likely to define its parameters. 24 In addition to this dummy variable, two additional measures of goals that differentiated between different levels of rebel aims were analyzed. These measures included an indicator of territorial goals (coded 1 if the rebels sought greater territorial autonomy /separation, 0 otherwise), and a categorical indicator of goals (coded 1 if the war was fought over political reform, 2 if over territorial autonomy or separation, 3 if over full control of the government, and 4 for social revolution). No significant differences to the substantive results emerged. 25 It should be emphasized that accurate counts of war-related deaths are notoriously difficult to obtain. Many deaths go unreported, while others are either under or over-reported for strategic reasons. For a detailed discussion of how the COW project attempted to correct this problem, see Small and Singer pp. 70-77. 11

Resolution on war recurrence. The first was whether the terms of a treaty addressed and resolved the main rebel grievances that were stated at the beginning of the war. Grievances were considered settled if a final peace treaty included provisions that specifically addressed these issues, or if the rebels won a decisive military victory and were therefore able to unilaterally implement these changes on their own. The second measure attempted to address the notion that settlements based on some form of Partition between or among the combatants would be more likely to bring peace. Here I used the definition first presented by Nicholas Sambanis where partition is defined as a war outcome that involves both border adjustment and demographic changes. Wars are coded 1 if an event of partition is observed, 0 otherwise. 26 Another potentially relevant factor associated with the outcome of the original war is the balance of power. Wars that ended with one side dominating the other were presumed to leave little opportunity for the weaker side to renew the war. Thus, each regression includes a dummy variable indicating whether the war ended in a Decisive Victory (for either side) rather than in a peace treaty, truce, or not at all. 27 In contrast to existing theories which highlight conditions surrounding the previous war, I argue that the key to understanding the emergence of new conflicts are the economic and political conditions on the ground after the previous war ends. Two sets of factors are likely to matter. The first are quality of life issues. As misery declines, the motivation for ordinary citizens to pick up arms should correspondingly decline and the odds of returning to war diminish. To see whether declining living standards affect the likelihood of renewed war, I 26 See Nicholas Sambanis, Partition as a Solution to Ethnic War: An Empirical Critique of the Theoretical Literature, World Politics, Vol. 52, (July 2000) p. 445. 27 In an alternate test, decisive victories by the government were distinguished from decisive victories for the rebels. These tests indicate that the effects of a decisive victory are the same regardless of who wins. Source: Sambanis Civil Wars dataset, variable OUTCOME2. 12

included a measure of the Infant Mortality Rate (deaths per 1,000) in each country. 28 In additional tests, I included a measure of Real GDP per capita and a measure of Income Inequality to see whether the reemergence of war was more closely linked to misery, overall wealth, or relative inequality. Analysis of all of these measures of deprivation, however, is complicated by the fact that living standards have generally improved over time across the globe. To ensure that this general improvement did not affect the results in any way, I included the rate of change in living conditions in each country rather than the absolute living conditions in a country at any given point in time. Thus, the actual measure presented in the analysis is the average yearly change in infant mortality for a country in the years between the original war and the following war (or 1999 if no war followed). 29 The second factor that should affect individual citizens decision to take up arms and thus the likelihood of renewed war is the openness of the political system. Countries with few nonviolent outlets for the population to seek governmental reform should be more susceptible to renewed war. Three measures were tested to see if voice, or lack therefore, had any effect on renewed war: (1) an overall Democracy-Autocracy scale, (2) a measure of political constraints on a government s executive branch (Executive Constraints), and (3) a measure of Political Openness published by Freedom House (also known as the Gastil Index). Since all three measures had nearly identical effects and the Gastil Index most closely approximated the concept of individual participation in the political process, only the Gastil Index was used in the final 28 To test for the robustness of this theory, in alternate analyses I also included a measure of Life Expectancy, measured at birth, and the percent of Adult Illiteracy in a country. These measures of misery had nearly identical effects on renewed war. 29 While difficult to assess empirically, there is the possibility that absolute conditions rather than changes in conditions are more closely related to the onset of new wars. Given the small number of cases and the reasonably high correlation between measures of change over time and absolute conditions, it is difficult to say with certainty which matters more to renewed war. When both types of measures were included simultaneously, change tended to be significantly related to renewed war, while absolute conditions were less so. 13

analysis. Attempts were made to distinguish between the effects of change in openness of the political system and absolute levels of democracy. While problems with the limited number of cases and collinearity prevented any systematic testing of this distinction, alternate tests with different measures included in the analysis suggest that once again change over time rather than absolute conditions are more important to war renewal. Findings Why do some countries experience recurring civil war while others do not? Table 1 shows the results of a poisson regression with renewed war as the dependent variable. A poisson regression was chosen to help deal with the time dependence problem involved in predicting subsequent wars. Specifically, wars that ended many years ago have a much greater chance of experiencing a subsequent war than wars that ended recently. To address this problem, the poisson regression explicitly takes into account the relative exposure to peace for each country (i.e., the number of years between the onset of peace and 1999, the last year that data were recorded). As an additional correction, I excluded all cases where peace began less than five years ago. 30 Two conclusions can be drawn from this analysis. First, civil war recurrence has less to do with the characteristics of a previous war and much more to do with issues directly affecting the misery and voice of the population after the previous war ends. Despite the fact that the analysis specifically seeks to predict new wars fought between the same set of combatants over the same set of issues, the goals of the original combatants, their ethnicity, the costs of war, and the degree to which rebel grievances were settled appear to have no effect on the re-emergence of war. Wars fought for total goals, or between different ethnic groups were no more likely to 30 Standard errors are also corrected to take into account the non-independence of multiple cases from the same country. 14

recur than wars fought over less demanding issues, or between the same ethnic group. 31 Longer, more bloody civil wars were also no more or less likely to re-occur than other types of wars. 32 And wars in which the main grievances remained unresolved were also not more prone to repeat themselves than those that addressed these underlying concerns. Finally, whether a previous war ended with a decisive victory or a truce/treaty had little impact on the likelihood of renewed violence. 33 In short, the attributes of the previous war do not appear to consign a country to a history of civil war. The results presented in Table 1 show, however, that a number of factors do predict whether war will emerge in the future. As discussed above, the economic and political environment individual citizens face did significantly affect the likelihood of renewed war. 34 Renewed war is in part a function of the basic well-being of the country s population. The less quickly a country was able to reduce infant mortality rates after the end of the first war, the more likely it was to return to war. Alternate tests revealed similar relationships between life expectancy and civil war renewal, and adult literacy and civil war renewal, although multicollinearity prevents the simultaneous inclusion of these variables in a single model. 35 This strongly suggests that when citizens are experiencing real improvement in social welfare, they are less apt to turn to violence regardless of what happened in a previous conflict. 31 This finding supports results by Hartzell, Hoddie, and Rothchild, as well as Dubey (using different datasets) who found the identity wars are no more likely to resume than non-identity wars, but contradicts results by Licklider, and Doyle & Sambanis. 32 Hartzell, Hoddie, and Rothchild, however, found that longer wars were more likely to lead to longer periods of peace if a war ended in a negotiated settlement. Doyle and Sambanis, however, found only weak support for this hypothesis, and Dubey 2002 found no relationship at all. 33 Dubey found the same. I should note, however, that empirical findings on the effect of decisive victory and of negotiated settlements on the duration of peace are quite contradictory. Doyle and Sambanis, for example, found that the presence of a peace treaty is positively correlated with successful peace-building after a civil war, while Licklider found just the opposite, that negotiated settlements are much more likely to lead to renewed war. 34 Elbadawi and Sambanis had similar findings. Using a cross-sectional time-series data set with five-year frequency covering the period 1960-1999, they found that greater political liberalization and economic development both reduces the risk of civil war. See Elbadawi and Sambanis 2002. 35 In contrast, additional tests suggested that wealth, or more specifically the rate of growth in real GDP, did not affect the likelihood of renewed war. 15

The analysis to this point leaves open the question of whether citizens respond most strongly to the rate at which basic living standards improve, the absolute level of misery, or their status relative to other members of their own society. Given the small number of cases and a fair degree of collinearity between many of these different measures, it is almost impossible to definitively adjudicate between the three different accounts. However, of the three possible relationships, relative status or levels of inequality appears to be the least relevant. A gini coefficient of income inequality for each country was never significantly related to war renewal in any of the possible specifications that were attempted. Whether the rate of change in living conditions or absolute living conditions is more important is less clear. In the specifications tested here, changes in basic conditions tended to be significantly related to war renewal more regularly than absolute conditions, although there are models in which levels of absolute living conditions were also significantly tied to war renewal even after controls for the rate of change were included. Beyond basic living conditions, greater political openness was also significantly related to the onset of repeat civil wars. In countries where there were increasingly fewer opportunities for citizens to influence the actions of government, people were more likely to initiate violent challenges. This finding remained robust when alternate measures of openness, such as a scale adding democratic and autocratic tendencies, and a measure of executive constraints were substituted in the analysis. 36 Openness or voice, no matter how it was measured, appears critical in helping deter renewed war. 37 All of this suggests that individual citizens base their decision to 36 This finding fits well with work by Gurr in which he finds a strong negative relationship between political barriers to democratic participation and the decision by minority groups to turn to rebellion. See Gurr 2000. 37 Once again, it is not clear whether changes in the ability to voice one s objections democratically or absolute levels of democratic openness are more critical. Models which included both sets of measures tended to show that changes in democratic conditions were more regularly related to renewed war than were absolute conditions but the analysis was far from definitive. 16

fight not only on concerns about their immediate quality of life, but also on their ability to change it over time. 38 Table 1: Predicting Renewed Civil War Dependent Variable: Did Civil War Between the Same Combatants Recur? Independent Variable Coefficient Standard Error CHARACTERISTICS OF PREVIOUS WAR UNDERLYING ISSUES: Ethnic Conflict.76.87 Ethnic Heterogeneity -.002.01 Total Goals -.75.89 COSTS OF WAR Battle Deaths/1000.000.001 Duration of War.003.004 Displaced Persons.000*.000 WAR RESOLUTION Decisive Victory (1-yes).005.73 Grievances Settled (1-yes).14.12 Partition -15.52** 1.22 LIVING CONDITIONS Infant Mortality 1.85*.41 Democratic Openness 1-6.34** 1.70 Constant -4.52** 1.30 Chi square 398.51 N 50 Figures are poisson regression coefficients with robust standard errors in parentheses. * p<.05 ** p<.01 1 Yearly change during the years after the original war. But misery and voice were not the only factors affecting the onset of repeat wars. Interestingly, two factors (one of which focused on the costs of the previous war, and one which was related to the manner in which the previous war was settled) had an effect on war recurrence. As one might expect, wars which ended in the physical separation of the combatants were much less likely to repeat themselves. This indicates that partition is a powerful tool to prevent groups from reengaging each other in endless conflict. Unfortunately, as we will see later, the effects of partition are less promising when one looks at the probability that a different 38 Elbadawi found similar evidence for poverty and the risk of war, and political rights and the risk of war. See Elbadawi 2001. 17

set of combatants will start a war. While partition is less likely to trigger a repeat war between the same set of combatants, it appears to be more likely to trigger renewed violence from other parties. The only other measure that affected the likelihood that the same combatants would return to war was the number of persons displaced by previous violence. The greater the number of individuals displaced in a given war, the more likely that war was to recur. One explanation for this is that displaced persons are some of the most dissatisfied people in the world; they suffer disproportionately from poor living conditions, and have no voice in any political decisions. Left with few options, these persons are likely to consider violence as the only means to improve their plight. This variable, therefore, could justifiably be viewed as a proxy for both misery and voice. Since it is difficult to interpret the magnitude of these effects from the poisson regression coefficients shown above, Table 2 converts the coefficients into probabilities that a country will return to civil war given particular values for infant mortality, political openness, displaced persons, and partition. In each case, all of the other variables are held constant at values for a hypothetical median case. 39 39 The values for this median case are: ethnicity (0), extent of goals (.66), battle-deaths per 1000 (88), duration (51 months), and decisive victory (1). These probabilities and their significance levels are calculated using a simulation procedure developed by King, Tomz, and Wittenberg (1998). 18

Table 2: Probability of a Renewed War Variable Infant Mortality Positive Change Negative Change Change in Probability Political Openness Positive Change Negative Change Change in Probability Displaced Persons High Low Change in Probability Partition Yes No Change in Probability ^p<.10 *p<.05 **p<.01 Probability of Renewed War.01.12.11^.03.09.06^.31.02 -.29^.14.00 -.14** 1 Probabilities and their significance levels are calculated varying particular independent from the 25 th to 75 th percentiles while holding other variables at values constant at a hypothetical median case using a simulation procedure developed by King, Tomz, and Wittenberg (1998). Two conclusions can be drawn from Table 2. First, both how well citizens are doing on basic measures of social welfare and how much input they have in the democratic process seem to play an important role in preventing the outbreak of repeated wars. Holding all other characteristics of the war constant at a median level, a move from the bottom quarter (25 th percentile) on the infant mortality scale to the top quarter (75 th percentile) decreases the probability that war will recur from 12 percent to only 1 percent. Similarly, moving from a country that has refused to open the political process at all to one that has increasingly supported democratic participation decreases the probability that a country will return to war from 9 percent to 3 percent. The number of displaced persons also affects renewed war. An average country that has almost 2 million displaced persons has a 31 percent chance of renewed war versus only a two percent chance when there are no displaced person in that country. 19

Second, it is also clear from Table 2 that partition can affect the likelihood that the two sides will renew their conflict in the future. Countries that resolve a civil war by physically separating the main combatants and partitioning the country have almost no chance of having the same combatants go to war, while those that do not separate combatants have a 14% chance of renewal. Resuming Old Wars versus Starting New Wars While it is important to determine the underlying causes of repeat conflict between the same combatants, it is equally important to assess the causes of renewed war between any new set of combatants. Given the very real possibility that the factors that lead the same set of combatants to renew a conflict may be very different from the factors that lead a new set of combatants to initiate a conflict, I undertook a second set of analyses that examined the propensity for any type of civil war to occur in the period after combatants have agreed to end a civil war. Table 3 presents the results of an identical poisson regression as table 1, only this time the dependent variable is simply whether a war broke out in the country regardless of who the combatants were or the issues. 20

Table 3: Predicting a Return to Civil War Dependent Variable: Did Any New Civil War Start? Independent Variable Coefficient Standard Error CHARACTERISTICS OF PREVIOUS WAR UNDERLYING ISSUES: Ethnic Conflict.81.49 Ethnic Heterogeneity.003.006 Total Goals -.21.50 COSTS OF WAR Battle Deaths/1000 -.001.001 Duration of War -.001.004 Displaced Persons.000.000 WAR RESOLUTION Decisive Victory (1-yes) -.14.55 Grievances Settled (1-yes).04.06 Partition.85.78 LIVING CONDITIONS Infant Mortality 1.69**.21 Democratic Openness 1-2.81* 1.20 Constant -4.17**.79 Chi square 41.99 N 50 Figures are poisson regression coefficients with robust standard errors in parentheses. 1 Yearly change during the years after the original war. * p<.05 ** p<.01 Almost the same conclusions apply. As one might expect, the various measures associated with the last war are even less significant here. The previous war s grievances are once again largely unrelated to the onset of a new war. Countries with ethnic divisions or with combatants with more severe demands were no more likely to have subsequent conflicts than other countries. Moreover, the costs associated with a previous war did not significantly affect (either positively or negatively) the likelihood of another war emerging. Longer wars with more deaths did not deter or encourage other combatants to launch their own war against the government. 40 40 Since some of the similarities between Tables 1 and 3 may be driven by the thirteen cases where the same combatants renew war, I undertook alternate analysis in which these thirteen cases were dropped. The increasingly small N creates problems for the poisson regression, but the results that emerge are largely the same. Basic living conditions are central to the likelihood of war. The only real substantially different findings are that the costs of war become significant and partition is now positively related to subsequent wars. Specifically, as the costs of previous 21

What did matter were basic indicators of the well-being of the local population. Citizens who faced rising standards of living, and were given an increasing ability to participate in government decision making, were less likely to turn to war. Specifically, the more quickly infant mortality decreased and political openness improved, the lower the probability of any new war. Similarly, the greater gains in political openness, the less likely war was to recur. In short, governments that failed to improve the lives of their people, and that ruled despotically, tended to face a series of violent challenges regardless of past history. Those that were able to raise the standard of living of their citizens, and create a competitive political process, did not. The Duration of Peace Since it is important to understand when a war is likely to recur and not just if it is likely to recur, I undertook a basic analysis of peace duration. Here the dependent variable is simply the number of years that peace lasted in each country. Table 4 presents the results of a censored weibull survival time model with the same list of independent variables. Cases that had not ended in war by 1999 were treated as censored. war increase, the probability of new and different wars emerging declines suggesting that high costs can serve as a deterrent to future wars. And in cases where partition occurs, the odds of new and different wars emerging increases. This could be part of a signaling game. Governments that are willing to offer one group of rebels their own territory may inadvertently encourage other rebel groups to form to try to gain their own territory. For a paper that deals specifically with the effects of accommodation on the likely outbreak of additional conflict see Walter 2002. 22

Table 4: Determinants of the Duration of Peace Censored Weibull Regression Independent Variable Coefficient Standard Error CHARACTERISTICS OF PREVIOUS WAR UNDERLYING ISSUES: Ethnic Conflict -.83.44 Ethnic Heterogeneity -.01.005 Total Goals.04.39 COSTS OF WAR Battle Deaths/1000 -.000.001 Duration of War.001.004 Displaced Persons -.000.000 WAR RESOLUTION Decisive Victory (1-yes) -.28.51 Grievances Settled (1-yes) -.00.05 Partition -1.40**.42 LIVING CONDITIONS Infant Mortality 1 -.68** Democratic Openness 1 5.14* 1.85 Constant 4.18**.69 Sigma 1.68 (.29) Chi Squared 74.9 ** N 50 Figures are poisson regression coefficients with standard errors in parentheses. * p<.05 ** p<.01 1 Yearly change during the years after the previous war. The results largely mirror the earlier findings. 41 The same two factors that played a role in civil war recurrence also play a role in peace duration. The better the average yearly change in conditions such as infant mortality, the longer peace is likely to last. Similarly, the more quickly the population was given the right to influence political outcomes, the longer peace is likely to last. The most interesting difference between the results presented in Table 4 and those presented in Table 1 is that partition is now negatively related to peace duration; countries that resort to partition in the first war are likely to have shorter periods of peace. The sum of findings 41 Interestingly, sigma is significantly greater than one indicating that the hazard function is decreasing: the likelihood of failure becomes lower over time. Collier, Hoeffler and Soderbom (2001) had a similar finding using a slightly different dataset. They found that a typical post-conflict country faced a fifty percent risk of renewed conflict within the first five years of reaching peace. These risks gradually declined if peace was maintained. 23

coupled with earlier results suggests a complex but clear pattern on the effects of partition on war recurrence. As one might expect separating the old combatants and giving them their own separate territory significantly reduces the likelihood that they will renew their conflict. But at the same time, the government s willingness to give up territory and partition the country sends a signal to other potential challengers about the resolve of government and the potential benefits of launching one s own separatist challenge. As a result, partition leads to a greater likelihood of a new and different war occurring and shortens the duration of peace. Conclusion This paper attempts to explain why some countries are able to escape the conflict trap and others not. It offers an explanation that stresses the role individual citizens play in deciding whether to join a rebel army, thus allowing a civil war to get underway. Citizens whose quality of life fails to improve after a war, and who are given little or no additional access to central decision making are far more likely to re-enlist in a rebel organization than those citizens who are making at least some gains. This runs counter to existing explanations that highlight the characteristics and grievances of previous wars to predict if and when war will resume. The empirical findings presented in the paper strongly support the individual incentive view of civil war recurrence. Across all specifications, renewed war was in large part a function of the basic well-being of the country s population and the accessibility of government decisionmaking to the average citizen. In several cases, war renewal was also closely tied to the number of displaced persons a variable that can easily be related to individual incentives to fight. Clearly, more displaced persons mean more individuals with limited life chances and limited political means to improve their current situation. Outside of individual incentives, only one other factor, partition, was found to be significantly related to war renewal. Partition reduced the 24

odds of the same combatants renewing their conflict but also encouraged potential challengers to begin their own wars with their government. A number of implications follow from this research. If it is true that war is most likely to recur when social welfare and political rights are in decline, then micro-level decisions by the average citizen play a much larger role in the outbreak of civil war than scholars have heretofore acknowledged. Civil wars are certainly affected by an array of other causal factors, but this research strongly suggests that cost calculations of the individual citizen matter a great deal in the reemergence of civil war. Without misery and an absence of voice, individuals would have little reason to volunteer for the armies that are necessary to pursue a war, and studies that fail to take this into account will overlook a key element driving war. This research also speaks to the debate on whether conflict begets conflict. Empirically, we know that once a country has experienced one civil war it is more likely to experience a second or third. Yet the assumption has always been that this occurs because the original war exacerbated the conditions that made violence break out to begin with. As Collier pointed out, the conflict is likely to have caused some of these underlying factors, such as per capita income, to deteriorate. 42 What this research shows is that individuals turn to violence not because they suffer from a particularly low level of social welfare, or an egregious absence of political rights, but because what little comforts and rights they do enjoy are in decline. The average citizen cares less about his or her absolute welfare, or welfare inequalities within a given country, then about improvements over time. 43 This leads to the somewhat counter-intuitive conclusion that countries that are quite poor or highly undemocratic at the end of one war are those that are best 42 Collier 2000 p. 2. 43 This is confirmed by a simple correlation analysis that reveals that countries that end the previous war with the worst levels of openness and highest levels of misery are more likely than other countries to experience rapid positive change. 25