62026 WORLD DEVELOPMENT REPORT 2011 BACKGROUND PAPER CONFLICT RELAPSE AND THE SUSTAINABILITY OF POST-CONFLICT PEACE Barbara F. Walter Graduate School of International Relations and Pacific Studies University of California, San Diego September 13, 2010 The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the World Development Report 2011 team, the World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent Conflict Relapse and the Sustainability of Peace Three disturbing patterns exist regarding civil wars and their recurrence. First, civil wars have a surprisingly high recidivism rate. Of the 103 countries that experienced some form of civil war between 1945-2009 (from minor to major conflict), only 44 avoided a subsequent return to civil war.1 That means that 57 percent of all countries that suffered from one civil war during this time period experienced at least one conflict thereafter. This confirms what Collier and Sambanis (2002) have called the “conflict trap;” once a country experiences one civil war, it is significantly more likely to experience additional episodes of violence.2 Second, recurring civil wars have become the dominant form of armed conflict in the world today. In fact, since 2003 every civil war that has started has been a continuation of a previous civil war.3 Table 1 outlines the pattern of war occurrence by decade.4 It reveals that 57 percent of all conflicts initiated in the 1960s were the first conflict in their country.5 That number falls significantly each decade, to the point where 90 percent of conflicts initiated in the 21st century were in countries that had already experienced a civil war. This trend suggests that 1 Source: UCDP/PRIO Armed Conflict Dataset, vol. 4, 2009. 2 Of the 59 countries that experienced renewed conflict, only 39 (or 66%) experienced more than one subsequent conflict. And in the severest cases, repeated cycles of civil war and civil conflict were commonplace. The three worst cases of Ethiopia, Myanmar, and India experienced 12, 13, and 20 new episodes of civil conflict. Renewed conflict can take the form of the same combatants fighting over the same issues, or new combatants fighting over new issues. 3 The last country to experience its first civil war was the Ivory Coast in 2002. Source: UCDP/PRIO Armed Conflict Dataset (see Gleditsch et al. 2002). For excellent discussions of the trends in civil war onset and termination see Hewitt, Wilkenfeld, and Gurr (2010); Elbadawi, Hegre and Milante (2008), Collier et al. (2003). 4 The other noteworthy trend in Table One is the increasing frequency of civil conflicts over time. The number of conflicts initiated each decade has grown over time and reached a peak of 81 in the 1990s. The last decade has, fortunately, witnessed a sharp decline in civil conflict onsets. 5 Since the earliest wars in the data set are by definition coded as new wars, the table drops data from the 1940s and 1950s to try to reduce the skewed nature of the results in Table One. 1 the problem of civil war is not a problem of preventing new conflicts from arising, but of permanently ending the ones that have already started.6 Table One. Civil War Onset by Decade Decade Onset in a country with Onset in a country N no previous conflict with a previous conflict 1960s 57% 43% 35 1970s 43% 57% 44 1980s 38% 62% 39 1990s 33% 67% 81 2000s 10% 90% 39 Third, civil wars are increasingly concentrated in a few regions of the world. Prior to the end of the Cold War, civil wars were spread over almost every continent, in countries as diverse as Bolivia, Greece, Indonesia, Lebanon and Nicaragua. The end of the Cold War, however, brought an end to many of these conflicts, especially those in Central America and Southeast Asia.7 The result is a greater number of civil wars concentrated in sub-Saharan Africa. For renewed conflict, the concentration of civil wars is even more pronounced. Sub-Saharan Africa accounted for only 13% of all the countries experiencing renewed civil war in the 1960s. That figure increases steadily so that by first decade of the 21st century, fully 35% of all recurring civil war occurs in Sub-Saharan Africa. This suggests that civil wars are increasingly being ghettoized in the world’s poorest and weakest states.8 6 Still, it is possible that the increasing rate of repeat civil wars is simply due to the spread of civil wars across countries over time. Since a civil war cannot repeat itself until it has broken out in the first place, the number of repeat civil wars, by definition, will increase over time. But something more significant is going on. Recurring civil war is not happening equally across all countries that have experienced civil war, but only in particular countries. 7 The removal of Cold War related funding, and the willingness of the UN to intervene created more conducive conditions for negotiated solutions to these conflicts (Mack 2008). 8 Prior to 1990, southern Africa accounted for 21 percent of civil war renewals. Since then, southern Africa has accounted for 38 percent of all civil war renewals. 2 The goal of this paper is to identify what it is about certain countries that makes them particularly susceptible to repeat civil war. Why are these countries unable to permanently end violence? And what is it about certain conflicts that makes them so hard to end? This paper argues that existing explanations that focus on grievances and opportunity are not sufficient to explain why conflicts recur or why they tend to recur in some countries but not others. To explain civil war recurrence, one must first explain why conflicts that repeat themselves resist settlement. That requires a focus on the political alternatives open to disaffected groups and the opportunities they have to negotiate long-term settlements with their governments. This paper argues that political institutions are the key to explaining why some countries can escape the conflict trap while others do not. The argument is fairly simple. Civil wars tend to recur in countries where the government can neither defeat a rebel movement nor credibly commit to a peace plan. If a government was strong enough to defeat the rebels, or trustworthy enough to negotiate a peace settlement, it would eventually do so and war would end. As long as the government can neither defeat the rebels, nor negotiate a settlement, the only remaining option is continued conflict. In what follows, I present analyses that reveal the underlying conditions that affect the risk of civil war relapse. We see that elements related to the credibility of government commitments to share power are critical in shaping the post-conflict peace. Governments that are beholden to a formal constitution, that follow the rule of law, and that do not torture and repress their citizens are much less likely to face renewed violence in any form. This suggests that the current heavy focus by policymakers on economic development as the best way to prevent recurring civil war may be wrong. This study strongly suggests that the way to end 3 repeat civil wars is to either help one side militarily defeat the other so as to permanently end the war or to build stronger political institutions and more credible governments so that negotiated settlements can be reached and implemented. Military and political interventions, rather than economic interventions, are the key to reducing the incidence of repeat violence. What follows is broken down into four sections. In section one I review existing arguments for why civil wars are likely to recur and explain why these explanations are incomplete. I then argue that commitment problems rather than poverty and state capacity are behind most of the recurring conflicts we see in the world today. This does not mean that poverty and state capacity are not contributing factors to civil war. Numerous studies have shown that they are. These two factors, however, are not sufficient to explain why combatants in so many of these recurring conflicts are not able to negotiate their way out of war. In section two, I introduce a dataset of all countries experiencing at least one civil war and present basic findings on how different economic and social factors affect conflict recurrence. Section three then considers whether different political factors related to the credibility of government commitments to a peace deal affect recurring conflict. Section four briefly presents findings on the effect of a number of different interventions, including development aid, humanitarian aid, remittances, and peacekeeping. Section five concludes. Why Civil Wars Recur: Current Theories Most studies of renewed civil war fall into one of two camps: those that argue that civil wars resume because grievances have not been resolved or have been aggravated by the previous war, and those that argue that wars resume because the opportunity costs once again favor war. 4 Renewed War as a Result of Unresolved Grievances Fighting one civil war can have the effect of aggravating the very conditions that encourage groups to rebel in the first place, making additional war more likely. Two sets of factors have been highlighted in the literature: economic under-development, and ethnic differences. Economic Under-development: One of the most frequent findings in the limited literature on civil war recurrence is the effect of economic development on recurrence (Doyle and Sambanis 2000; Walter 2004; Quinn, Mason and Gurses 2007; Collier, Hoeffler and Soderbom 2008; Kreutz 2010). Countries that score low on a host of factors associated with income and development are at a significantly higher risk for recurring civil war. Gurr (1971; 2000), for example, has argued that groups are more likely to rebel when they feel disadvantaged vis-à-vis groups in society. Others cite poverty, poor public health, or other features related to low levels of human development that can create anger and resentment against the state.9 Civil war is likely to exacerbate each of these conditions, where a country that was poor to begin with becomes even poorer as a result. If these conditions do not improve, or even deteriorate over time, a second civil war is likely. Ethnic and Religious Differences: Civil wars, however, are likely to exacerbate other factors as well. Civil wars that are fought between competing identity groups are believed to be particularly intractable. According to Gurr “cultural identities – those based on common descent, experience, language, and belief – tend to be stronger and more enduring than most civic and associational identities.” (Gurr 2000:66). Once violence breaks out, ethnic identities 9 Note that there is disagreement in the literature about how development levels affect rebellion. It can affect rebellion by creating grievances against the state, by lowering the opportunity costs for turning to violence (Collier and Hoeffler 2004), or by weakening the state’s repressive capacity (Fearon and Laitin 2003). Statistical analysis is unable to determine which of these interpretations is a more accurate account of individual motivations. 5 may become cemented in ways that make cooperation between groups even more difficult. The recurring conflict in the Balkans, as well as the repeated violence between groups such as the Hutus and Tutsis, Turks and Armenians, Jew and Arabs are often identified as examples of this. Renewed War as a Result of Opportunity Costs Theories that focus on underlying grievances or motives are often complemented by those that focus on the ability of groups to mobilize and sustain a rebel organization over time. A group may have grievances against the state, but unless it is also able to organize and recruit individual soldiers to fight for a particular cause, and then supply them over time, no challenge is likely to emerge and be sustained. The existing literature on civil war has focused on three sets of factors that create the opportunity for rebellion: (1) rebel recruitment, (2) money and supplies, and (3) the repressive capacity of the state. Rebel Recruitment: A number of scholars have argued that factors related to rebel recruitment are important in determining where civil war will occur (Collier and Hoeffler 2004; Gates 2002; Walter 2004a). Again, factors associated with economic development have been purported to matter, but for different reasons. Collier and Hoeffler (2004) have argued that low secondary schooling and per capita income make becoming a soldier more attractive. Similarly, Walter (2004a) has argued that enlistment is likely to become attractive when high levels of individual hardship exist, such as high infant mortality rates and low life expectancy. Money and Supplies: The ability to recruit soldiers, however, is unlikely to be sufficient to sustain a rebellion over time. Money and supplies will also be needed to build an organization, pay soldiers, and pay for the materials needed during active resistance. Collier and Hoeffler (2004) have argued that a group’s access to finance in the form of extractable natural 6 resources and donations from diasporas make rebellion feasible. Stedman (2001) has similarly claimed that access to disposable resources such as gems, minerals, or timber make the implementation of any peace settlement more difficult because it provides armies with a means to continue fighting. Civil wars, therefore, may be more likely to recur in countries where groups have access to such financing. Groups may temporarily stop the fighting because one type of financing has been exhausted, a patron has withdrawn, or the market for a particular good has declined. But as long as additional financing can be found, renewed civil war is possible. State Capacity: The final set of factors has to do with state capacity and a government’s ability to police and control areas where rebels may seek to mobilize. Fearon and Laitin (2003) have argued that countries with rough terrain, large populations, or weak central governments are more likely to experience rebellion because groups are better able to evade government repression over time. This has the effect of lowering the opportunity costs a group pays to organize and fight against the government, making rebellion more attractive (see also Collier and Hoeffler 2004). Similarly, states that lack functioning political institutions or are so weak that they have little control over their own borders are more apt to harbor spoilers capable of sabotaging peace agreements. (Stedman 2001) If these features do not change over time, or become worse as the result of one war, renewed war is more likely. What’s Missing Existing explanations are limited in two important ways. Theoretically, they cannot explain why governments in at-risk countries do not sign more negotiated settlements, either before a civil war breaks out, or once it has broken out. Governments, aware that low economic growth and state weakness place them at higher risk of civil war should take these conditions into account and offer groups deals that reduce the possibility of violence. This is especially true 7 in cases where governments know they face a group that is likely to be particularly persistent over time.10 Grievances and opportunity, therefore, should not cause civil wars to break out and recur, so much as they should cause governments to compensate a particularly dissatisfied group in some other way. Existing studies also cannot explain variation in the outbreak of violence across countries that are at similar risk of recurring civil war. Nicaragua, El Salvador, and Guatemala, for example, were all devastated by civil war, yet all have managed to avoid repeat civil wars. This suggests that factors in addition to grievances and opportunity explain why some countries escape the conflict trap while others do not. In order to explain why violence breaks out repeatedly in some countries one has to view recurring violence as part of a larger bargaining process. Conflicts that produce multiple bouts of violence are, by definition, conflicts that are particularly difficult to resolve: they continue because combatants are unable to reach either a decisive outcome or a negotiated settlement. Violence may stop temporarily because the combatants are able to agree on a ceasefire, one side suffers a temporary set-back, or one or both sides becomes fatigued. But none of these outcomes is a permanent solution to war. Bargaining Problems and Recurring Civil War I argue that one of the main reasons some countries experience recurring civil war is because their governments are too weak militarily to permanently defeat the rebels, and too institutionally weak to credibly commit to a settlement. In particular, I focus on specific characteristics of a country’s governance structure that make it difficult for government leaders to convince rebels to sign and implement a deal. 10 The fact that more negotiated settlements aren’t used to end civil wars is especially puzzling given the high costs of these wars, especially recurring wars. For a discussion of some long-term costs of civil wars, see Ghobarah et al. 2003. 8 Certain countries have political and legal structures that place few constraints on those in power. Leaders of these countries can promise that they will reform in response to the demands of a rebel organization – they can promise to liberalize the political system, institute majority rule, grant greater political autonomy to a group – but unless the rebels have the ability to enforce compliance over time, these promises are likely to be viewed as meaningless. Rebels have some ability to enforce compliance in countries where it is easy for them to organize resistance, where repression is not severe, where individual rights and freedom exist, and where the rule of law dominates. But they have little ability to influence government behavior in countries with few constraints on executive power and where the central values of the rule of law, transparency and accountability do not exist. This suggests that countries with weak political and legal institutions, and those that offer competitors few civil liberties or political rights will not be attractive partners in a negotiated settlement. In these situations, rebels are likely to be better off keeping their arms and continuing to fight to a decisive victory rather than signing a settlement they have little faith with be fulfilled. In what follows, I present analyses that attempt to reveal the underlying conditions that affect the risk of civil war relapse. We see two things. First, the main factors that cause civil wars to break out in the first place are also significantly related to the likelihood that a conflict recurs. As past research on civil wars has found, both grievances and opportunity are important in re-igniting conflict. Second, we also see that elements related to the credibility of the government are critical in shaping the trajectory of post-conflict states. Governments that follow good governance practices – however measured – are much less likely to face renewed violence in any form. Exactly why governance is important is less clear. Governments beholden to a formal constitution, that follow the rule of law and that do not torture and repress their citizens 9 may create fewer grievances or may be better partners with whom to negotiate a compromise settlement. What is clear is that efforts to reform government practices in nations that have experienced civil conflict could be richly rewarded with reductions in violence. Given that renewed civil war is the main and maybe only source of emerging civil conflict, this is a critical area of potential policy reform that demands attention. Data The conflict data come from the UCDP/PRIO Armed Conflict Database v. 4 (ACD) which is the most comprehensive and accurate available data set on civil war conflict. The ACD codes for each country and year since 1945 whether a violent conflict occurred between a named, non-state armed group and government forces that directly killed at least 25 people. The current version of the data set was prepared for the 2011 World Development Report. The database has a rough estimate of the annual battle deaths for each conflict. It does not, however, provide any particular scheme for identifying episodes of civil war. In line with Fearon (from this project), I identify episodes by including those years of conflict that have over 25 battle deaths in the ACD data. When the conflict has had two years of fewer than 25 battle deaths, the episode has ended. The two year period of fewer than 25 deaths is critical as it helps to ensure that I do not code an ongoing war with short lapses in fighting or small fluctuations in deaths as multiple separate episodes of conflict. Any increase in annual deaths over 25 after this two year interval is counted as renewed conflict. Also, as with Fearon (from this project), the magnitude or intensity of each episode is coded by calculating the average annual battle deaths for each episode. Intensity is coded in the following manner: 1) major war = over 1,000 annual deaths, 2) medium war = 500-999 annual deaths, 3) small war = 250-499 annual deaths, and 4) minor war = 25-249 annual deaths. By 10 coding the average annual death rate over the entire episode, one avoids coding as new wars wars that are essentially ongoing but fall below a threshold of battle deaths in one year only to exceed that threshold in the subsequent year. The alternative generally employed is to code civil war magnitude based on the year with the peak number of deaths. I believe that the intensity coding enacted here offers a more accurate means of gauging the magnitude of the conflict since it does not focus on a single year which may or may not represent the larger pattern of the war (i.e., a war that crosses a high threshold of deaths in one year but then lingers with relatively few deaths for an extended period of time) and takes into consideration both the total number of deaths and the length of time over which those deaths occur. A critical distinction between this analysis and others in this project is that I focus exclusively on countries that have already experienced a civil war. That means that I not only drop from the analysis all countries that have never experienced a civil war but also all years of data prior to the first war in each civil war country. The reason for this is fairly straightforward. The main aim is to understand post-conflict experiences and the two primary empirical goals are 1) to determine the likelihood that a country will return to civil war after experiencing an episode of conflict and 2) to determine the factors that reduce the likelihood of renewed civil war. A country cannot have a subsequent episode of conflict until it has initiated its first conflict. Moreover, the only way to uncover the factors that distinguish countries that have experienced repeated deadly episodes of conflict from those that are able to emerge from a single conflict with a protracted period peace is to compare outcomes in countries after they have experienced their first episode of war. 11 To determine the factors that are related to civil war recurrence, I employ a cross-section times series data set with post-conflict years as the unit of analysis.11 The critical variable in this analysis is war onset. As noted above, an onset is the initiation of conflict that has over 25 battle deaths in a year after a period of two years without 25 battle deaths in either year. Since more than one conflict can emerge in a country in a given year, onset is coded as 1 if one conflict starts in that year, 2 if two conflicts start in that year, and so on. Any year without an onset is coded as 0. I analyze several different onset variables that distinguish between the intensity of the war that follows. Major Onset marks the first year of a conflict that ultimately results in over 1000 battle deaths a year. Medium Onset is the initiation of a conflict that results in 500-999 annual battle deaths. Small Onset is followed by wars with 250-499 annual battle deaths and Minor Onset is a war that ultimately ends with between 25 and 249 battle deaths. Years in which each threshold is not passed are coded as 0. Again since more than one conflict can be initiated in a country in a given year, there can be more than one onset in a given year. In addition, to incorporate all kinds of wars in one summary analysis, the main variable Onset is given a score of one for each type of conflict that is initiated in that country in a given year. Thus, if there is a small, medium, and major onset in a given country in a given year, Onset is coded as 3. Multiple onsets are, however, relative rare so onset is effectively a binary variable. Again, years that do not see the initiation of new conflicts are coded as 0. One concern with all of this coding is that it does not distinguish between renewed civil conflict between the same parties that fought the previous war and renewed civil conflict between new combatants. One might theorize that the factors that lead the same set of 11 Rather than wait until the first episode ends, all country years after the first episode in a country has been initiated are included. This is because the first episode does not have to end before a second conflict between other parties in the same country can emerge. 12 combatants to return to war are different than the factors that lead to the initiation of a new conflict between previously peaceful actors. Also, doing the analysis primarily at the country level rather than primarily at the conflict level means that one cannot incorporate detailed characteristics pertaining to the last conflict. Factors related to how a previous conflict was fought (battle deaths, duration, and the nature of the disagreement, for example), how the conflict ended (in a decisive victory for the government or the rebels or in a negotiated settlement), and how the ensuing peace was or was not monitored (i.e., the presence of peace keepers) all should affect the likelihood that a given conflict will reemerge. In light of these concerns, I plan to conduct further analysis that will focus on a conflict as the basic unit of analysis rather than country. Independent variables are taken primarily from the World Bank data set or from Fearon’s expanded and revised data set for this project. What Makes Countries More Likely to Suffer From Repeated Civil War? Ideally, to answer this question, one would incorporate the entire range of potential factors into a single model that predicts civil war renewal in those countries with past civil war experience. In the analysis that follows, I will try to get as close to that model as possible but there is a significant empirical obstacle – a lack of key data on many of these countries. Several core measures are available for all years and all countries and those are included in the main models that follow. But most of the other variables are not available for at least some of the countries and some of the years. Many of the variables are only available for a small subset of cases. Adding to the difficulty is the fact that different cases are missing from different variables. Data issues are exacerbated by the fact that my analysis has to focus on a small subset of countries and years. To understand what leads to civil war renewal, I have to exclude 13 countries that have not had a civil war and I have to exclude all years prior to the first civil war in a country. Given that the universe of countries that have experienced a civil war in the last half century is relatively small, the total number of cases for this analysis is also relatively small. Thus, a complete model with an extensive list of variables is not possible. As an alternative strategy, I focus first on a few core indicators of each of the main theoretical perspectives that have driven past research. For each of the main theoretical perspectives, I then add a range of alternative measures of each theory to see if they are related to civil war onset. These alternate measures are not included in the main model almost exclusively because they have high rates of missing cases. Finally, I supplement key measures from existing accounts of grievances and opportunities with newer measures of governance and outside intervention into a larger but still far from full model of civil war renewal. Grievances To understand the factors that drive renewed civil conflict, I begin with a model that focuses on underlying grievances. The first model in Table 1 illustrates the relationship between a number of the core sources of grievances highlighted in the existing literature and Onset. Recall that Onset is coded as the total number of new conflicts in a given year for all years after the initiation of the first conflict in a country. The core grievance measures are: - GDP lagged - - per capita income in the previous country year, in 2005 U.S. dollars, using Penn World Tables, - War lagged – a dummy variable indicating whether there was an on-going conflict in the previous country year - Ethnic Fractionalization – a measure of the ethnic fractionalization of the country based on estimates of ethnic group populations (Fearon and Laitin 2003). - Religious Fractionalization – a similar measure of religious diversity (Fearon and Laitin 2003) 14 The results in the first column of Table 1 confirm the critical role that grievances can play in inciting renewed conflict. Those countries with greater fractionalization are significantly more likely to face a subsequent conflict. All else equal, the risk of facing a civil war onset in any given year in a less fractionalized country (the 25th percentile) is 1.93 times lower than the risk in a more fractionalized country (the 75th percentile). (See Table 2) Alternate tests suggest that ethnicity need not be exclusively the area of discord. When ethnic fractionalization is dropped from the model religious fractionalization is also significantly and positively related to civil war renewal. Also, when other measures of ethnicity like ethnolinguistic fractionalization are substituted into the regression model, they prove to be related to the reemergence of conflict. Although Table 1 does not illustrate it, economic measures of grievance also seem to influence the trajectory of civil wars in a country. GDP is not quite significantly related to war onset in this model but two findings suggests that economic grievances matter. First, when onset excludes smaller wars (those with less than 500 annual deaths), there is a significant and negative relationship between lagged GDP and war recurrence. Second, when other types of measures of economic stress like the poverty gap, infant mortality, and life expectancy are added to the model, they are all related to war onset. In each case, renewed war increases when economic conditions decline.12 At the same time, subsequent tests also reveal that other measures of economic conditions are unrelated to civil war renewal. A series of measures of human capital investment and horizontal inequality show no evident link to conflict renewal when added to the analysis. Finally, Table 1 indicates that the presence of a conflict in the country in the previous year markedly increases the probability of a new conflict erupting in the current year. The 12 These other variables are not included in the main model as they are missing for the vast majority of country years. 15 relative risk of civil war onset is 1.26 times higher in countries with a war that was ongoing in the previous year. This relationship fades once I control for a range of other factors but on some level even this basic finding suggest that experience with war may exacerbate grievances leading to more war in the future. The list of grievance measures in Table 1 is far from complete. A comprehensive test of grievances would include a range of factors related to discrimination and rights abuses against individual groups. Groups that suffer from the most egregious abuses and those that are most severely discriminated against by their government should be the most likely to initiate violence against the state. Unfortunately, it is difficult with country level data to incorporate these group level variables. Which groups should be considered when assessing abuse or discrimination? And how should parameters for different groups get aggregated at the national level? Two steps can, however, be taken to begin to address questions related to discrimination, abuse, and the treatment of distinct minority groups in society. First, I incorporate several measures of governance that provide overall assessments of civil liberties and government repression (I do this later in the analysis). Although not ideal for measuring group level grievances, they likely offer some insight into this question. Second, in a follow-up to this study, I plan to conduct analysis of conflict-level data that will more fully incorporate the treatment of individual groups by their governments. Opportunity Opportunity is a second critical factor purported to drive organized outbreaks of violence. In the second column of Table 3, we see that it also plays a significant role in determining whether a country that experiences armed conflict once is likely to experience conflict again. The model 16 adds three basic measures of opportunity highlighted by Fearon (this project) and other past research on civil war: - Political instability – a dummy variable that indicates whether in the previous country year (t-2 to t-1) there was any change in the Polity 2 score is a measure of democracy that runs from -10 (extreme autocracy) to 10 (full democracy) - Log of the country population - Log of the percentage of mountainous terrain in the country (plus one) as judged by geographer A.G. Gerard - Non-Contiguous Land Mass As the second column illustrates, larger populations and more mountainous terrain are associated with renewed civil conflict. Presumably the ability to organize and execute a rebellion is enhanced by the ability to recruit from a larger population, or a larger population is more difficult to police making mobilization easier. Likewise the ability to retreat to mountainous terrain appears to enhance the prospects of rebellion since it can serve both as a staging ground for war and as a home to avoid detection from government forces. Moreover, alternate tests reveal links between other measures of opportunity and renewed conflict. As Fearon and Laitin (2003) have demonstrated for civil war more broadly, I also find that the presence of large oil reserves is positively related to civil war onset. Those countries where rebels can exploit (or hope to exploit) lucrative oil deposits are significantly more apt to return to civil war. Similarly, when a measure of the percent of the population that is urban is added to the model, it is significantly and negatively correlated with war reemerging. It may be that rural populations that are further afield from the capital are easier to recruit. Finally, there is at least some sign that school enrollment can impact civil war renewal. Higher enrollment, by some measures, is marginally negatively related to the reemergence of civil war in a few alternate specifications. 17 Other measures of opportunity, however, are not demonstrably linked to civil war renewal. Neither the overall rate of unemployment nor the unemployment rate of young adults in the country is significantly related to onset when added to Model 2.13 Countries with more young people or more unemployed adults are no more or less likely to experience recurring war. Nevertheless, the preponderance of findings thus suggests that opportunity is a second, important variable along the pathway to renewed violence. The findings regarding government military strength are less clear. A larger armed force, greater arms imports, and police size are all positively and significantly related to war onset if added to Model 2 in Table One. This is, however, exactly the opposite relationship that one would expect to see if larger government forces were deterring rebellion. Likely, these three measures, even when lagged, are picking up government efforts to stockpile an army for a civil conflict that is likely to occur in the near future. At a minimum, the positive relationship that I do find suggests that government spending on the military or police is not serving as an effective deterrent to renewed civil war. Constraints on Government Measures that are believed to be related to grievances and opportunities clearly impact renewed conflict. But a focus on these two mechanisms ignores the potentially central role that credible commitments can have on conflict. If a government is unable to credibly commit to honoring political concessions in a negotiated settlement, rebels are likely to prefer continued fighting (or at the very least intermittent fighting) to any sort of deal. The ability to credibly commit to a peace deal is not the only mechanism by which government institutions can have an effect on renewed civil war. It is also possible that a closed, 13 Given the difficulty in acquiring comparable unemployment figures across countries and the small N for the unemployment data, this conclusion should be read with some caution. 18 repressive and corrupt system could work in conjunction with economic grievances to encourage armed violence. In Models 3 and 4 of Table 1, I test these relationships. I begin the investigation by examining the link between broad measures of overall governance and renewed conflict. Specifically, I incorporate the World Bank’s Country and Policy Institutional Assessment (CPIA), and two Polity measures that assess the level of democracy and autocracy in the country. The CPIA offers a telling overall assessment of the impact of governance that incorporates four equally weighted clusters described as “public sector management and institutions, “ “policies for social inclusion/equity,” “structural policies,” and “economic management.”14 The two Polity scores for Democracy and Autocracy are the standard measures of the openness of political representation in a nation. As the third column of the Table 5indicates, governance greatly affects the chances of falling back into war. Countries that score higher on the CPIA index are significantly less likely to experience renewed conflict. All else equal, a higher CPIA score (75th percentile) is associated with a 32 percent reduction in the likelihood that a country faces a civil war onset in any given year over when compared to a country with a low CPIA score (25th percentile). (See Table 2) It is also important to note, that more standard measures of democracy are largely unrelated to the pattern of civil war conflict in a nation over time. Neither the democracy measure nor the autocracy measure that is produced by Polity is significantly related to renewed conflict in the third model in Table 5. Moreover, even when the CPIA score is removed from the 14 For each country World Bank personnel code on 16 or more dimensions concerning the quality of policies and institutions. The aggregate index varies from 1 to 6 with higher values indicating better governance. The key drawback of the CPIA index is that it is only produced for countries that receive aid from the bank so it is limited in its reach (85 countries) and includes a biased sample of countries that is skewed toward less developed nations. 19 model, democracy and autocracy remain insignificant. It is not the overall level of democracy that matters but rather something to do with how the regime actually governs. Unfortunately, although this analysis strongly suggests that governance significantly shapes the conflict path that countries take, the multi-faceted nature of the CPIA index makes it hard to determine which element of governance is most critical in determining whether a countries plunges back into civil war. Does repression lead to grievances that then motivate groups to organize? Or does a lack of formal institutions and a lack of the rule of law indicate that government promises are not likely to be kept? Model 4 of Table 1 tests these relationships by adding two measures of government repression and one measure of formal institutional structure. The two measures of government repression are EXTRAJUDICIAL KILLINGS, POLITICAL PRISONERS. The presence of a formalized democracy is measured by a dummy variable indicating whether or not a formal constitution is in place. The fourth column of Table 1 suggests that both of these factors are related to renewed civil war onset. Model 4 reveals a highly significant relationship between government repression and the reemergence of civil conflict. After controlling for other factors, the presence of extra- judicial killings and large numbers of political prisoners makes a civil war onset in any given year 2.86 times and 1.91 times more likely (respectively). (See Table 2). Alternate tests indicate that other measures of government repression are equally significant predictors of renewed conflict. When I substituted in two political terror scales that are derived from reports of government abuses from Amnesty International and the U.S. State Department, both were strongly and positively related to subsequent conflict (Gibney, Cornett, and Wood 2008). Similarly, an overall index of civil liberties developed by Freedom House which measures the 20 extent to which governments infringe upon basic human rights was also positively and significantly related to renewed civil war onset when substituted into the model. A reasonable interpretation of these results is that greater repression and abuse by a government creates both grievances and signals that those governments are not dependable negotiating partners; suggesting that less coercive and more accountable approaches significantly decrease the risk of civil conflict.15 Also, it is evident from Table 1 that formal rules and institutions can play a role in preventing repeated conflict. As Model 4 indicates, there is a strong, negative relationship between the presence of a formalized constitutional democracy and renewed conflict.16 A formal constitution reduces the odds of conflict renewal by 64 percent. (See Table 2). This suggests that countries where the rule of law is not sufficiently formalized cannot credibly promise to abide by the law and commit to agreements with past combatants or potential future rebels. Other measures of the institutionalization of the rule serve equally well to predict outcomes. When it is substituted into Model 4, the “rule of law” dimension of the World Governance Indicators (WGI) is negatively related to civil war renewal Kaufmann, Kraay, and Mastruzzi (2009). The WGI scale is designed to capture perceptions of the extent to which agents have confidence in and abide by rules of society, and in particular the quality of contract enforcement, property rights, the police, and the courts. Thus, countries that score low on the scale are unlikely to be able to credible commit to agreements with potential rebels. To try to further get at which specific features of governance affect the likelihood of conflict renewal, I also considered several measures of government competence or effectiveness. 15 Tests regarding the influence of torture were less conclusive. When substituted into Model 5, TORTURE was positively but not quite significantly related to civil war onset. An estimate of government corruption and a count of disappearances both showed no obvious link to conflict renewal. 16 Other similar measures like whether the constitution is in effect and whether the constitution is written are negatively but not quite significantly related to civil war onset. 21 The results of several alternate tests are not conclusive but do hint at the relevance of competence and effectiveness. When I substituted a regulatory quality scale produced by the WGI project into Model 4, it was negatively and significantly related to renewed civil war. A slightly different WGI measure, the “government effectiveness” scale, which captures perceptions of the quality of public services and the quality of the civil service and its degree of independence from political pressures, and the quality of policy formulation and implementation was negatively but not quite significantly tied to repeat civil war onset. It seems more than plausible that when a government is incapable of responding to group demands in an effective manner, those groups will feel less inclined to have confidence in the government and its ability to follow through on its commitments and will turn to other alternatives. These conclusions about role of governance and the ways through which governance may impact renewed conflict should, however, be read with some caution. One reason has to do with the vagueness and subjectivity of many of these measures. Most of the governance measures are based on perceptions rather than quantifiable actions and are thus prone to any number of biases. Also, several of the scores include a variety of different elements which makes it hard to determine exactly what is being measured. Also, since each of the different measures of governance is available for a limited number of cases and most are relatively highly intercorrelated, it is difficult, if not impossible, to definitively determine the relative contribution of different governance features on the pathway to civil war reemergence. Governance appears to be vital to understanding why some countries get caught in repeated episodes of civil war while others are not, but exactly which aspects of governance dominate this relationship is much less clear. 22 Another very real concern is the possibility of endogeneity. Experts who are scoring these nations on a number of different governance measures may rightly perceive government to be repressive, ineffective, or both. But these experts may also be basing their evaluations, in part, on the expressed grievances of groups that are at or near the point of open rebellion. This endogeneity problem can be dealt with partly by lagging the independent variables as I have done here. Nevertheless, correlation is not causation and questions about endogeneity will remain. Outside Intervention One last set of factors to consider is outside intervention. Here there are at least two likely suspects. The first is aid. If poor local conditions drive repeated rebellion as my earlier results and the existing literature suggests, than the efforts of foreign actors to provide aid may play a role in improving conditions and diminishing the drive toward renewed violence. The second potentially relevant form of outside intervention is peacekeeping. If, as I have suggested, credible commitments are an important barrier to resolving civil war conflicts over the long term, then outside actors could influence events by supplying peacekeepers. Large peacekeepers forces could check the actions of countries with low cpia scores and offer a solution to the credible commitment problem. To test these two possibilities, I incorporated two measures into Model 5 in Table One. Per capita foreign aid is a composite measure that includes total development assistance and official aid that is provided by the Development Assistance Committee of the OECD. United Nations Peacekeeping is a dummy variable indicating whether or not UN Peacekeepers are on the ground in each country year. 23 The results presented here suggest that neither variable matters in preventing civil war renewal. The results on aid all point in the same direction. More aid, whether measured as overall aid per capita, as is done in this model, or by any of the other available measures of aid is insignificant. Alternate tests show that a range of measures of aid including a more focused measure of humanitarian aid, an account of total debt relief, a set of measures of development assistance, and an estimate of all remittances to the country are all unrelated to civil war renewal. Also when aid is broken down by sector (i.e., health versus education) by mode (i.e., technical assistance versus direct aid) or by time-horizon (i.e., short-term versus long-term aid), there is no indication that increased aid significantly reduces the chances of renewed conflict. Importantly, none of these measures of foreign aid is significant even if I drop controls for grievances and opportunity. At least as modeled here, aid does not appear to work to limit renewed conflict either by alleviating the conditions that drive grievances or by altering the relative attractiveness of other non-conflict opportunities. Firm conclusions about the value of aid are, however, not warranted at this point. The underlying problem with these results is that aid is very likely to be endogenous. Aid tends to be offered to countries that are particularly susceptible to renewed civil war. As one would expect, subsequent analysis shows that aid is higher in poorer countries. Countries with lower gdp, higher rates of infant mortality, lower life expectancy, and high rates of unemployment are substantially more likely to receive aid. This endogeneity means that the true relationship between aid and renewed war is extremely difficult to assess. The absence of a correlation between aid and conflict could be a sign of the absence of a relationship or simply a reflection of the fact that aid funnels into exactly the countries that need it – those that are already very likely to return to war. 24 Similar concerns also need to be raised about the UN peacekeeping results. Although the UN peacekeeping dummy variable is insignificant in Model 5, firm conclusions about the role of peacekeeping require better and more in-depth analysis of outside intervention.17 One real problem with the peacekeeping measure being utilized here is that we do not know which conflict the peacekeepers are assigned to help end, or which negotiated settlement to which conflict they are directed to help implement. In many of the countries with multiple civil wars there are a range of different groups considering violence. The current analysis, therefore, does not allow us to determine if the peacekeepers that are on the ground are assigned to deal with the conflict that actually breaks out in a given year or if they are being employed to prevent the renewal of a different conflict. The peacekeeping variable also does not take into account whether the peacekeepers were essentially imposed on the nation by outside actors or whether peacekeepers emerged out of a comprehensive peace agreement that was supported by the combatants in the conflict. One would certainly expect peacekeepers to be more effective at reducing the likelihood of renewed war when the combatants have signed a mutually satisfactory peace agreement and only need to overcome the remaining problem of enforcement, than in other cases where one or both sides is unhappy with the current outcome of the war. Thus, future research that incorporates the context of UN intervention is necessary. The other concern is that peacekeeping, just like aid, is endogenous. Even among the dubious set of countries that have already experienced a civil war, the United Nations is more likely to step in when conditions on the ground are bleak. Subsequent analysis indicates that peacekeepers are especially apt to be present in poorer countries with lower life-expectancy, higher infant mortality, and lower literacy. The UN also more readily offers peacekeeping 17 Annual peacekeeping expenditure data and personnel data are also not significantly related to civil war reemergence in this model. 25 assistance in countries with relatively low levels of democracy and high rates of religious and ethnic fractionalization. Even if UN involvement was reducing the likelihood of renewed conflict, we might not see it here. Robustness Checks To test the robustness of these results, I performed a series of additional analyses that, in different ways, examined civil war onset under different levels of war intensity. In the first set of tests which are reported in Table 3, I repeated the main analysis but focused only the most intense or deadly war. The first model in the table drops minor civil wars (those with average annual deaths of under 25 people) out of the category of civil wars. The second model excludes small civil wars (those with average annual deaths under 250 people). The pattern of results in both models in Table 3 is generally similar to what we saw in Table 1. Some of the marginally significant findings now fall just below conventional levels of statistical significance but the overall conclusion that civil war renewal is a function of grievances, opportunity, and political institutions largely remains intact. In alternate tests, I also repeated the analysis focusing separately on each level of intensity. In other words, the dependent variables were Major Onsets, Medium Onsets, Small Onsets, and finally Minor Onsets. The concerns with this type of analysis are two-fold. First, when we focus on only one level of intensity, the dependent variable ends up with a relatively small number of cases of conflict and that makes the modeling process somewhat more difficult. Second, and perhaps more importantly, it is not clear theoretically what it means to isolate one level of intensity from another. For example, by focusing on medium wars, the modeling process is implicitly contrasting cases with medium wars both to cases with no wars and to cases 26 with major, minor or small wars. Is there anything about any particular level of intensity that we think should make that kind of war unique and subject to different explanatory factors.18 Nevertheless, one might want to single out Major Onsets as being distinct from smaller wars and also single out Minor Onsets as being different from larger wars. I undertake this latter analysis in Table 4. When I isolate Major Onsets some of the significance levels fall but the overall pattern of results is not dramatically different. Major Wars do not appear all that different from smaller wars. Similarly, when I focus exclusively on Minor Onsets significance falls but the overall pattern is similar. Civil war recurrence –whether for low level intensity or for major conflicts – appears to be driven by a similar set of factors. The other interesting finding that emerges out of Table 4 concerns change over time. The result suggest that all else equal major civil war onsets are becoming less common over time while minor war onsets are becoming more common. I also attempted to determine if the factors driving civil war renewal varied depending on whether a country had already experienced a relatively deadly and intense civil war or if it had just experienced a relatively minor civil conflict. Unfortunately, due to the small number of cases at each level of intensity, the results from this kind of analysis are suspect. With only a few countries included and only a few hundred country-years, the results are highly variable and likely highly idiosyncratic and are thus not presented here. Finally, I sought to determine if the factors that lead to renewed conflict over territory were substantially different from the factors that lead to renewed conflict over governmental reform. With fewer cases of renewal, significance levels tended to fall for each subset of renewal but the overall pattern of results was similar for both subsets. Governance, grievance, 18 Analysis isolating Medium Onsets from all other levels of intensity generally revealed no significant patterns of results. The same is true when I attempted to isolate Small Onsets from all other cases. 27 and opportunity mattered whether a new civil war was fought over territory or over governmental reform. Repeat Civil War: Incorporating the Characteristics of Each Conflict So far, all of the analysis has focused on countries and the characteristics that place these countries at higher risk of renewed civil war. Focusing solely on the country level, however, has two drawbacks. First, we do not know if a subsequent conflict in a country is essentially a repetition of an old war that was fought between the same two sets of combatants or a new and potentially unrelated war in which new rebel groups fight the government over a distinct set of issues or grievances. Distinguishing between new civil wars and repeat civil wars is important because it is possible that the factors that generate repeated violence between the same combatants over the same issues are different from the factors that lead particular countries to face recurring conflict with a range of different internal groups. Second, by focusing on the country level, the previous analysis necessarily ignores a number of conflict specific features that may make a particular civil war more likely to restart. Factors associated with the earlier outbreaks of violence may affect whether a second outbreak of violence occurs. In particular, factors such as the number of deaths in the previous war, the duration of the previous war, whether the war ended in a decisive victory or a negotiated settlement, and whether the settlement was supported by peacekeepers could all exacerbate commitment problems, incite grievances, or generate greater opportunities for recruitment and ultimately make recurrence more likely. To address these possibilities, I undertook a second set of analyses that focused on recurring civil wars – those fought between the same combatants over the same basic issue (territory or 28 governance reform). The data are derived from the UCDP/PRIO Armed Conflict Dataset. In this second set of analysis I focused solely on major wars. To be included as an episode of civil war a conflict had to lead to a thousand battle deaths in one year. To be coded as a repeat war, the main set of combatants in the current war had to be the same as the combatants in a previous war. A renewed war was thus defined as an episode of civil war with at least a thousand deaths in one year that occurred after a previous episode of civil war between the same two sets of combatants. To avoid coding ongoing wars that temporarily went above or below a threshold of 1000 battle deaths as separate wars, an episode of war did not end until a country experienced two years without 1000 battle deaths and one year with under 25 battle deaths. Renewed war was coded as 1 for the first year in which the subsequent civil war exceeded 1000 battle deaths. Renewed war was coded as 0 for every year after the first civil war episode in a conflict had ended and no subsequent war was ongoing. By this definition, there were a total of 24 episodes of repeated civil war in the data set. These 24 cases of renewed war represent a small subset of the civil war conflicts that were examined earlier. In addition to considering the variables that I looked at earlier, this analysis incorporates a number of different elements associated with the first civil war that could affect whether or not that war restarts: (1) the duration of the previous war, (2) the intensity of the previous war (defined as deaths per year), (3) how the previous war ended, (4) if a settlement was signed, the terms of the settlement, and (5) whether peacekeepers were available to help enforce the settlement. Each of these factors could affect grievances, opportunities, and/or the credibility of any agreement with the government. Data on previous duration is from the UCDP/PRIO Armed Conflict Dataset, data on battle deaths is from the UCDP/PRIO Battle Deaths Dataset, 29 information on how the previous war ended and the presence of UN peace keepers was obtained from the UN Peacemaker website. Table 5 presents the results of this alternate analysis of repeated civil war. The Table follows the earlier pattern in Table 2 where I looked first at key measures of grievances and then in subsequent models added different indicators of opportunity and governance. Models 1 through 3 thus essentially repeat the earlier analysis and focus on country level indicators exclusively. Then, in Models 4 and 5 I incorporate factors specific to the main combatants and in particular factors related to the previous episode of civil war between those combatants. The results presented in Table 5 suggest that the factors that lead to repeat civil war are generally similar to the factors that lead to all types of renewed civil conflict. Just as we saw earlier, grievances and governance play a critical role. Civil war between the same two sets of combatants is more likely to occur in countries where grievances are high and in nations where governance is generally poor. In terms of grievances, the key measures are religious fractionalization and the presence of another war in the country. In some of the models, ethnic fractionalization is significant but it is signed in the wrong direction – less fractionalization associated with a higher chance of repeat war. One possible explanation for this pattern is collinearity between the two measures of fractionalization as ethnic fractionalization becomes insignificant when religious fractionalization is dropped from the model. For governance, once again, renewed violence is associated with signs that a government is unresponsive and untrustworthy. War between the same combatants is especially apt to reignite in countries without a formal constitution and in countries where the government regularly imprisons its opponents. As before, it is difficult to know exactly which element of governance is most important. Alternate models reveal similar relationships between repeat civil war and 30 measures of political terrorism, the extent of the rule of law, the degree to which the government tramples civil liberties, the effectiveness of government programs, and extent of political participation in the nation. As before, there are also concerns with endogeneity. Is poor governance causing repeat civil war or are these measures of governance picking up the efforts of a government that is trying to fend off an impending civil war? Table 5 also tends to corroborate earlier findings regarding post-conflict intervention. Once again, the data reveal no apparent relationship between aid and repeat war or between peacekeeping and repeat war. The main per capita aid measure is insignificant. So too are measures of aid broken down by sector (i.e., education vs health), by type of assistance (i.e., monetary vs technical assistance), and by the time horizon of the aid (short-term vs long-term aid). The main peacekeeping measure is also uncorrelated with repeat war.19 Alternate measures that count up the number of personnel or the costs of the operation are also insignificant. However, the same caveats that applied to aid and peacekeeping in earlier tables apply here as well. Both aid and peacekeeping are endogenous and they tend to be offered in countries that are relatively worse off - even among those countries that have already had one civil war. Thus, it is nearly impossible to determine the true relationship between these two factors and repeat civil war. The one aspect of civil war that is not substantiated by this analysis of repeat civil war relates to opportunity. Contrary to what we found earlier with renewed civil wars, in most of the models in Table 5, repeat civil war is not significantly related to the different measures of opportunity that were included. The one exception is the presence of a non-contiguous land 19 The UN peacekeeper measure is dropped from Model 5 in Table 5 because in this relatively small number of cases it perfectly predicts the absence of repeat civil war. However, when other variables are dropped and all of the cases of the dependent variable are included, there is no apparent correlation between the presence of UN peacekeepers and the likelihood of repeat civil war. 31 mass in the nation. At least in models 3 through 5 in the table, non-contiguous land is significantly and positively related to repeat civil war. If the relationship is true, non-contiguous land areas presumably serve as staging grounds from which rebels can regroup and re-launch a civil war. The other important findings in Table 5 pertain to the characteristics of the previous conflict. In Models 4 and 5, I incorporate a range of measures related to the previous conflict between the same two combatants. Model 4 first adds three measures related to how the previous war was fought: 1) the duration of the previous conflict, 2) the intensity of the previous conflict (defined as deaths per year), and 3) whether the previous conflict was fought over territory or issues of governance. One might expect that longer and more deadly wars would deter combatants from initiating a new war. Similarly, given past findings concerning the intransigence of territorial conflicts, one might predict that territorial conflicts would be more likely to reignite than conflicts over governance. However, at least in these models, none of these three measures is significantly related to repeat civil war. Model 5 then adds two measures that describe how the previous war ended: 1) with a ceasefire, or 2) with a comprehensive peace treaty. The expectation is that these kinds of pacts are likely to break down in the post conflict period when problems related to credible commitments are present. By contrast, wars that end with a decisive victory for either the rebels or the government should be less likely to be renewed. Yet again, I find no apparent relationship between war ending and repeat civil war. Conclusion Civil wars have become the dominant form of conflict in the world today in large part because they are so difficult to end and tend to recur. To date, almost all of the literature on recurring civil war (including my own) has argued that certain countries are at particular risk of 32 civil war relapse because the underlying economic and political conditions make them prime candidates for civil war relapse. But we know that some countries are able to escape the conflict trap despite the fact that war has exacerbated many of these underlying conditions. This study suggests that the conflicts that recur are the ones where the combatants are neither able to decisively beat the other, nor able to reach and implement mutually acceptable settlements. They are cursed with a situation where the parties have no ability to reach any resolution to their conflict on their own. This study reveals that a government’s ability to credibly commit to a peace agreement likely affects its ability to avoid repeat civil war. Governments that are constrained by a formal constitution, and that follow the rule of law are much less likely to face renewed violence in any form. In fact, any measure that limits the government’s ability to act outside the law and unilaterally usurp power, makes the government a more attractive negotiating partner, and offers combatants an alternative way out of war. This suggests that a heavier focus on political institution building rather than economic development may be the most effective way to resolve existing civil wars and could help reduce the rate at which these conflict repeat themselves over time. 33 Table One. The Determinants of Civil War Renewal (All Onsets) Model 1 Model 2 Model 3 Model 4 Model 5 GRIEVANCE Ethnic 1.26 (.33)** 1.12 (.33)** 1.30 (.53)** 1.63 (.63)* 1.42 (.55)** Fractionalization Religious .05 (.42) .23 (.43) -.31 (.62) -.41 (.78) -.14 (.70) Fractionalization GDP -.14 (.09) -.16 (.10) -.47 (.18)* -.36 (.25) -.08 (.18) Ongoing war .24 (.07)** .11 (.07) -.05 (.10) -.09 (.12) -.00 (.11) OPPORTUNITY Population Size .22 (.06)** .29 (.09)** .15 (.12) .15 (.11) Mountainous .15 (.07)* .24 (.10)* .17 (.12) .15 (.11) Terrain Political -.03 (.21) -.41 (.31) -.80 (.41)* -.82 (.41)* Instability Non-Contiguous .19 (.23) 1.03 (.33)** 1.04 (.38)** .72 (.33)* Land Mass GOVERNANCE Democracy -.02 (.02) -.01 (.02) -.01 (.02) Autocracy .02 (.02) .01 (.02) .01 (.02) CPIA -.37 (.15)* -.15 (.18) -.13 (.18) Extra-judicial .53 (.23)* .52 (.23)* Killings Formal -1.38 (.53)** -1.31 (.56)* Constitution Political Prisoners .71 (.26)** .69 (.26)** OUTSIDE INTERVENTION Per Capita Aid -.00 (.01) UN Peacekeeping .07 (.54) Year -.01 (.01) -.01 (.01) .02 (.02) .04 (.02) .04 (.02) Constant 12.0 (10.7) 5.7 (11.1) -40.4 (28.2) -78.9 (40.9) -79.6 (41.2) R squared 0.4 0.6 .11 .18 .18 N 3338 3325 1795 1389 1385 34 Table Two. The Magnitude of the Effects on Civil War Renewal – Models 1-4 Level Percentile Chance of Onset Relative Risk Model 1 Ethnic Fractionalization .16 25 .030 .70 75 .058 1.93 Ongoing War 0 50 .038 1 75 .048 1.26 Model 2 Ethnic Fractionalization .16 25 .029 .70 75 .052 1.79 Population Size 8.41 25 .032 10.24 75 .047 1.46 Mountainous Terrain 1.43 25 .034 3.60 75 .048 1.41 Model 3 Ethnic Fractionalization .16 25 .022 .70 75 .042 1.91 GDP 5.31 25 .112 8.86 75 .022 0.20 Population Size 8.41 25 .024 10.24 75 .040 1.67 Mountainous Terrain 1.43 25 .026 3.60 75 .044 1.22 CPIA 2.89 25 .040 3.96 75 .027 0.68 Model 4 Ethnic Fractionalization .16 25 .014 .70 75 .032 2.28 Political Instability 0 25 .028 1 75 .013 0.46 Non-Contiguous Land Mass 0 25 .022 1 75 .060 2.73 Extra-Judicial Killings 0 25 .014 2 75 .040 2.86 Formal Constitution 0 25 .092 1 75 .023 0.36 Political Prisoners 0 25 .011 1 75 .021 1.91 35 Table Three. The Determinants of Civil War Renewal Onset (Major, Medium and Onset (Major and Medium Small Wars Only) Wars Only) GRIEVANCE Ethnic Fractionalization 1.18 (.66) 1.15 (.98) Religious Fractionalization .13 (.83) .53 (1.24) GDP -.34 (.26) -.81 (.40) Ongoing war .02 (.12) -.08 (.17) OPPORTUNITY Population Size .01 (.14) .07 (.20) Mountainous Terrain .14 (.13) .21 (.19) Political Instability -1.01 (.46)* -1.59 (.78)* Non-Contiguous Land Mass .91 (.41) 1.17 (.63) GOVERNANCE Democracy -.02 (.02) .01 (.03) Autocracy .02 (.02) .03 (.03) CPIA -.10 (.18) -.14 (.25) Extra-judicial Killings .46 (.22)* .62 (.34) Formal Constitution -1.28 (.55)* -.93 (.74) Political Prisoners .83 (.54) .79 (.43) OUTSIDE INTERVENTION Per Capita Aid -.01 (.01) -.00 (.01) UN Peacekeeping .25 (.55) -.02 (.85) Year .03 (.02) -.02 (.03) Constant -65.3 (43.0) 32.6 (59.9) R squared .17 .21 N 1385 1385 36 Table Four. The Determinants of Civil War Renewal Major Onset Only Minor Onset Only GRIEVANCE Ethnic Fractionalization 2.12 (1.38) 1.32 (.87) Religious Fractionalization -.54 (1.64) -.31 (1.04) GDP -.33 (.55) -.21 (.35) Ongoing war -1.18 (.54)* .05 (.15) OPPORTUNITY Population Size .15 (.28) .06 (.19) Mountainous Terrain .17 (.31) .13 (.17) Political Instability -1.28 (1.08) -.63 (.56) Non-Contiguous Land Mass 1.96 (.84)* .86 (.52) GOVERNANCE Democracy .03 (.05) -.03 (.03) Autocracy .01 (.04) .02 (.03) CPIA -.26 (.38) .01 (.27) Extra-judicial Killings 1.13 (.51)* .17 (.30) Formal Constitution -- -1.75 (.73)* Political Prisoners .57 (.55) .74 (.36)* OUTSIDE INTERVENTION Per Capita Aid .01 (.01) -.01 (.01) UN Peacekeeping .15 (1.2) .45 (.68) Year -.09 (.05)* .08 (.03)** Constant 180 (93.4)* -166 (60.6)** R squared .21 .16 N 1338 1385 37 Table Five. The Determinants of Repeated Civil War (Repeat Conflicts Between the Same Combatants) Country Level Predictors Adding Conflict Level Only Predictors Model 1 Model 2 Model 3 Model 4 Model 5 GRIEVANCE Ethnic Fractionalization -2.13 (.90)* -1.64 (1.61) -7.13 (3.76) -7.04 (3.21)* -7.04 (3.19)* Religious 2.81 (1.14)* 2.37 (1.71) 12.4 (3.64)** 12.48 (4.81)** 13.8 (6.07)* Fractionalization GDP -.00 (.00) -.00 (.00) .00 (.00) -.00 (.00) -.00 (.00) Ongoing war 1.52 (.58)** 1.89 (.59)** 2.74 (2.34) 2.62 (2.03) 2.85 (2.29) OPPORTUNITY Population Size -.29 (.35) -.69 (.97) -1.07 (.96) -1.07 (.92) Mountainous Terrain -.26 (.21) -.37 (.67) -.65 (.62) -.84 (.84) Political Instability .12 (.52) .24 (1.32) -.14 (1.04) -.04 (1.06) Non-Contiguous Land .28 (.91) 4.82 (.96)** 5.19 (1.64)** 5.78 (1.80)** Mass GOVERNANCE Democracy .02 (.06) .11 (.08) .09 (.09) Autocracy .06 (.07) -.04 (.08) -.01 (.09) CPIA .21 (.50) .65 (.62) .54 (.56) Extra-judicial Killings .49 (.76) .50 (.83) .51 (.89) Formal Constitution -4.16 (1.28)** -4.83 (1.61)** -5.10 (1.86)** Political Prisoners 2.69 (.83)** 3.02 (1.06)** 3.05 (1.13)** PREVIOUS CONFLICT Intensity .61 (.43) .46 (.45) Duration -.13 (.47) -.21 (.47) Territorial Conflict .63 (.74) .88 (.78) HOW PREVIOUS CONFLICT ENDED Ceasefire -.85 (1.05) Comprehensive Treaty 1.62 (1.49) OUTSIDE INTERVENTION Per Capita Aid .01 (.01) UN Peacekeeping --- Year -.00 (.02) -.00 (.01) .02 (.06) .03 (.06) .03 (.06) Constant -2.8 (31.9) -1.45 (31.3) -35.8 (115) -72.7 (121) -61.1 (118) R squared .08 .11 .29 .31 .60432 N 1199 1188 650 604 38 Works Cited Collier, Paul and Anke Hoeffler. (2004) ‘Greed and Grievance in Civil War’, Oxford Economic Papers. Collier, Paul & Nicholas Sambanis, eds, (2002) ‘Understanding Civil War: A New Agenda’, Journal of Conflict Resolution, 46(1). Collier, Paul, Lani Elliot, Havard Hegre, Anke Hoeffler, Marta Reynal-Querol and Nicholas Sambanis (2003) Breaking the Conflict Trap: Civil War and Development Policy. Oxford: Oxford University Press. Collier, Paul, Hoeffler, Anke., and Soderbom, M. (2008) 'Post-conflict risks', Journal of Peace Research, 45(1): 461–78. Doyle, Michael W. and Sambanis, Nicholas (2006) Making War and Building Peace: United Nations Peace Operations. Princeton: Princeton University Press. Elbadawi, Ibrahim A., Havard Hegre and Gary Milante. (2008) 'The Aftermath of Civil War." Journal of Peace Research, 45(4): 451-459. Fearon, James D. and Laitin, David D. (2003) 'Ethnicity, Insurgency and Civil War', American Political Science Review, 97(1): 75-90. Gleditsch, Nils Petter; Wallensteen, Peter, Eriksson, Mikael, Sollenberg, Margareta and Strand, Håvard (2002) ‘Armed Conflict 1946–2001: A New Dataset’, Journal of Peace Research 39(5): 615–637. Gates, Scott (2002) 'Recruitment and Allegiance: The Microfoundations of Rebellion', Journal of Conflict Resolution, 46(1): 111-130. Ghobarah, Hazem, Huth, Paul and Russett, Bruce (2003) 'Civil Wars Kill and Maim People – Long after the Shooting Stops', American Political Science Review, 97(2):189-202. Gurr, Ted R. (1971) Why Men Rebel. Princeton, NJ: Princeton University Press. Gurr, Ted R. (2000) Peoples Versus States. Washington, DC: United States Institute of Peace Press. Harbom, Lotta and Wallensteen, Peter. (2010) 'Armed Conflicts, 1946-2009' Journal of Peace Research 47(4): 501-509. Hewitt, J.J., Wilkenfeld, Joseph and Gurr, Ted (2010). 'Peace and Conflict 2010,' Center for International Development and Conflict Management of the University of Maryland. Available at Kaufmann, Daniel, Kraay, Aart and Mastruzzi, Massimo (2009) 'Governance Matters VIII: Aggregate and Individual Governance Indicators, 1996-2008,' World Bank Policy Research Working Paper No. 4978. Available at SSRN: http://ssrn.com/abstract=1424591 39 Kreutz, Joakim. (2010). "How and When Armed Conflicts End: Introducing the UCDP Conflict Termination Dataset. Journal of Peace Research 47(2): 243-250. Mack, Andrew. (2008) "Global Political Violence: Explaining the Post-Cold War Decline‟ in Volker Rittberger and Martina Fischer (eds), Strategies for Peace: Contributions of International Organisations, States, and Non-State Actors. Barbara Budrich Publishers. Quinn, J., Mason, T.D. and Gurses, M. (2007) 'Sustaining the Peace: Determinants of Civil War Recurrence', International Interactions, 33(2): 167 – 193. Stedman, Stephen John. (2001) ‘Implementing Peace Agreements in Civil Wars: Lessons and Recommendations for Policymakers’, International Peace Academy and Center for International Security and Cooperation, New York. Available at: Walter, Barbara F. (2004) 'Does Conflict Beget Conflict? Explaining Recurring Civil War', Journal of Peace Research, 41(3): 371-88. 40