I. Introduction
The twentieth century is sometimes called the “age” or “century of genocide” or a “century of mass slaughter.”Footnote 1 A recent book on genocide catalogs a ghastly list of cases and concludes that some 60 million civilians died in the twentieth century from mass murders and genocide, more than the number killed on battlefields.Footnote 2 Daniel Goldhagen puts the civilian casualty figure much higher, at as many as 175 million, and concludes that mass murder and elimination are “worse than war.”Footnote 3
In many respects, such characterizations, in which genocide and similar forms of systematic large-scale campaigns of violence against civilians are presented as ubiquitous, are misleading. Rather than being common, genocide and similar forms of mass organized violence against civilians are rare political phenomena.Footnote 4 If one considers the number of situations in which such violence could happen, the frequency of such events is quite limited. Non-genocide, which includes a range of outcomes—from combatant-on-combatant warfare, to repressive violence, to negotiation and accommodation—is the dominant form of conflictual interaction and politics. In other words, genocide is but one of many possible political outcomes (and an infrequent one) of conflict.
Therein lies a more general methodological problem facing scholars of extreme events: what is the comparative reference point? Genocide scholars typically compare genocides to genocides.Footnote 5 One result is a frequency mismatch: most theory posits causal factors that are much more common than the outcome. This is true for the first comparative studies, which highlighted deep social divisions,Footnote 6 authoritarian regimes,Footnote 7 and deprivation.Footnote 8 And it is also true for more recent studies that emphasize forms of warfare,Footnote 9 loss in wartime,Footnote 10 organic nationalism,Footnote 11 colonialism,Footnote 12 and upheaval and instability.Footnote 13 These factors are all much more frequent than the outcome they purport to explain. Thus, a key question is: why does genocide not happen when it could? To answer this question, a spectrum of negative, non-genocide cases needs to be part of the comparison set. (This is a point equally applicable to studies of other rare political outcomes and other forms of political violence.)Footnote 14
Detailed case analysis points to another reason to broaden the comparative lens: change over time. Studies consistently find that genocide is not usually the initial policy choice. Rather, genocide is the outcome of a process of fluid decision-making in which events, interactions, interests, ideology, and actors shape the trajectory of violence. Genocide is also a phase within a longer, broader pattern of majority-minority, state-opposition, or inter-group conflict. This over-time, dynamic dimension is a clear finding in studies of the Armenian genocide,Footnote 15 the Holocaust,Footnote 16 and Rwanda.Footnote 17 Qualitative comparative research produces the same findings. Michael Mann, for example, argues that murderous cleansing is a “Plan C” that emerges after other plans fail.Footnote 18 Therein lies a powerful theoretical lesson that has implications for the study of violence and other rare events. The outcome in question should not be modeled as a two-stage outcome of policy conception and implementation, but rather as a multi-stage, dynamic process subject to conditions that could cause escalation, de-escalation, or non-escalation.Footnote 19
We are thus left with two central puzzles: (1) What explains variation in outcomes (why is mass violence the outcome in some cases of conflict, but not in others)?, and (2) What explains the over-time process of escalation? If these issues are taken seriously, I argue, we arrive at a third general problem: the relative theoretical invisibility of factors that cause an outcome of interest not to occur. In genocide studies, as in the general study of violence, the dominant question is to ask what causes organizations, ordinary perpetrators, and leaders to commit violence. The dominant answer is to identify sources of escalation—from hatred, to hardship, to war.Footnote 20 However, in this paper, I propose to conceptualize the problem differently, arguing that factors of escalation are only one side of the equation. The other side concerns factors of restraint—ideas, interactions, and institutions that prompt leaders and/or citizens to abstain from or moderate the use of extensive violence against civilians. In the main, I propose to reconceptualize violence as an outcome of both factors of escalation and restraint. All other things equal, violence will be more likely and at a higher level when sources of escalation are strong and sources of restraint are weak.
There are multiple implications to this reconceptualization of the process of violence. One is to open up an explicit terrain of investigation for the study of violence, namely what explains non-violence or low violence. At present, restraint is a largely missing concept in genocide studiesFootnote 21 and an embedded one in the violence literature, where it has not received focused and systematic exposition.Footnote 22 Another implication is observational: because restraint always exists (though at varying strength), for violence to succeed sources and voices of restraint must be marginalized, overwhelmed, or destroyed. Overcoming restraint is thus an integral part—and indicator—of the process of violence. A third implication is ethical and policy-oriented. The idea that restraint matters strikes a hopeful note for what is a decidedly gloomy topic. Put simply, societies are not hardwired only for violence, and outsiders can make a difference. In many societies, there are incentives and reasons to avoid violence, and there are leaders who fashion ideas that diminish, rather than intensify, violence. Insiders and outsiders can in turn strengthen these sources of restraint. In short, this article provides theoretical support for the normative assumption that there exists a margin for maneuver in building a less violent world.
I elaborate these points in the first two sections. In the first, I develop the escalation-and-restraint analytical framework just discussed, highlighting the importance of negative cases and change over time. I also introduce a diagram to illustrate the claims. In the second, I reread the literature on violence to explicate sources of restraint. A careful reading of existing studies turns up a number of embedded arguments. I mine these, rendering them explicit in some cases, and divide them into micro, meso, and macro levels. I further argue macro-level factors are most consequential for shaping the trajectory of genocide, but meso-level sources also matter in some circumstances.
I then outline a provisional theory of genocide, thereby showing how the escalation-and-restraint framework addresses the two puzzles identified here. The theory builds upon the most consistent empirical finding in the literature, which is that genocide and similar forms of mass violence typically take place during an acute crisis, in particular war. While the finding is major—it separates out a lot of cases—it is insufficient. Upheaval in the form of sharp economic crises or political instability is common, and most wars do not result in genocide. To understand which wars escalate to mass violence, I turn to ideological and economic factors. Existing studies typically highlight how these factors are sources of escalation.Footnote 23 I endorse that position, but I also claim that ideology and economic interests can create powerful sources of restraint.
Finally, I illustrate the argument with reference to cases of genocide and non-genocide in sub-Saharan Africa. The empirical section is not designed as a rigorous test of the argument, but rather as an illustration of how the reconceptualization of the process of violence changes how we approach the problem. I first show that almost every case of mass violence against civilians at a threshold of 10,000 civilian deaths took place in the context of armed conflict. I then select two comparable cases, Rwanda and Côte d'Ivoire. The former is a case of genocide. The latter, following the “possibility principle” in selecting negative cases, is a case that had, according to existing theory and the fears of many credible observers, a strong risk of escalating to genocide but did not so escalate.Footnote 24 I present evidence to show that while both countries had similar factors of escalation, the factors of restraint were more powerful in Côte d'Ivoire than in Rwanda. The case studies are designed to be indicative rather than dispositive. My purpose is to show empirically how the dynamics of escalation and restraint can be seen to work in tandem, as a way of identifying an alternative approach to the comparative study of violence.
II. Rethinking the Process of Violence
The choice by individuals and organizations to employ violence against domestic civilians is typically deliberate and instrumental.Footnote 25 For incumbents, who generally have access to the state's means to employ large-scale violence, the main objective is generally to retain power and to protect their interests. Yet across time in states and across states, political and military leaders choose a variety of strategies to negotiate differences, manage instability, counter perceived threats, and keep power. Only rarely do leaders choose practices of extreme violence, even in the face of an acute crisis. The key question is what prompts them to pursue policies that involve mass murder of civilians.
Framing the question in that way recognizes the strategic origins of mass violence but broadens the universe of comparable cases away from a cloistered set of high-magnitude violence ones into a larger set in which political authorities manage crises in a variety of ways. The framework also implicitly endorses a dynamic view of violence, but raises the question of what drives leaders towards the use of increased levels of violence against civilians, and what drives them in the other direction.
In the existing literature, the emphasis is on the former. In civil war studies, such factors include more brutal types of modern warfare,Footnote 26 weak state capacity and non-democracy,Footnote 27 territorial control in civil war,Footnote 28 unit cohesion,Footnote 29 emotions such as fear, hatred, and resentment,Footnote 30 and economic endowments,Footnote 31 among many others. In the genocide literature, as discussed, there is a range of emphases, from hatred to deprivation to ideology to armed conflict. Much less developed, though implicit in many studies and explicit in some others, are the conditions that prompt strategies where violence is absent, not escalated, or de-escalated—a point I pick up in the next section. This is puzzling, because strategies of accommodation and low-level violence are empirically much more common than strategies of large-scale violence.
To illustrate the framework, I present a stylized diagram of a “ladder of violence” in which political authorities may select from a variety of violence strategies (see Figure 1).
Ladder of violence
The main outcome of interest is genocide, which I conceptualize as large-scale, organized, group-destructive violence that targets a specific social group in a territory. The other outcomes relate to how an authority that has nominal governance responsibility in a particular territory develops policies towards that social group, in particular during an acute crisis. Such policies include a range of “negative” cases of genocide, from the politics of accommodation and incorporation to the politics of exclusion and persecution to limited repressive violence.
In the diagram, moving down the ladder of violence represents increasing levels of violence against the social group in question. Moving up the ladder represents decreasing levels of violence. Though two-dimensional, the diagram is meant to be dynamic, not necessarily linear, and non-teleological. That is, at any point in a crisis, authorities can move up or down the ladder of violence.
Factors of escalation are listed on the left. These factors create pressure to move down the ladder of violence, i.e. to increase the level of violence. By contrast, factors of de-escalation are on the right. They create pressure to move up the ladder of violence, i.e. to decrease the level of violence. In any given situation, there will always be factors of escalation and factors of de-escalation that vary in strength. The central claim is that the relative strength of both factors of escalation and restraint shapes whether authorities move up or down the ladder of violence and how far they will go.
The diagram highlights three dimensions of violence. First, the diagram draws attention to variation. Group-destructive, mass violence is only one policy—and a rarely-chosen one—out of many that political authorities may select when confronted by a crisis and conflict. Second, the diagram emphasizes that mass violence is the outcome of a dynamic process; how and why practices of violence change should be key questions for scholars.Footnote 32 Third, policies of violence are the product not only of factors of escalation, but also factors of restraint. At any point in a crisis, authorities are pulled in both directions, but the relative strength of each determines the trajectory of violence.
III. Bringing Out Factors of Restraint
Having argued that restraint should matter, in this section I look to the literature on violence and genocide, not to identify what causes violence, but rather what prevents it.
Micro-Level Sources of Restraint
At the individual level, many people have values and morals that lead them not to commit violence. Many find violence against innocents or non-combatants, including their neighbors, abhorrent. Few are good at it, and committing violence runs contrary to “basic mechanisms of emotional entertainment and interactional solidarity.”Footnote 33 Kristen Renwick Monroe explains variation among rescuers, bystanders, and Nazi supporters during World War II as a function of personal, moral choice driven by their self-identity.Footnote 34 Violence can also be detrimental to one's self-interest: violence invites revenge and disrupts inter-personal systems of mutual benefit. These arguments are the inverse of the group-animosity theory prominent in ethnic conflict and genocide studies. Rather than holding sharply negative views of others, many individuals harbor life-affirming values and cherish cooperation, and these personal values in turn can serve as a restraint on violence at the micro level.
These arguments go a long way towards explaining why it is that, if one considers the history of human interaction, non-violence is vastly more common than violence. People do not regularly kill or physically harm others. The frequency of non-violence is surely partly due to institutions, rather than only to values; nonetheless, most people prefer to interact with other people without violence. This dimension must be theorized in explaining how and why violence succeeds, suggesting more specifically that for sustained violence to succeed on a large scale, micro-level sources of restraint must be overcome.
Indeed, a number of studies of violence show precisely how such a process occurs. Randall Collins's research shows that it is in situations of “confrontational fear” that individuals commit violence.Footnote 35 Stanley Milgram's social psychology experiments demonstrate that persuasion from a “legitimate authority” can cause ordinary individuals to seriously harm people like themselves for whom they have no prior hatred.Footnote 36 The Stanford Prison Experiment by Philip Zimbardo similarly shows that it was when students were given excessive power over other, more vulnerable students that they quickly and willingly inflicted harm.Footnote 37 Multiple studies of armed conflict and other highly tense scenarios show how individuals, when they fear for their lives or when they face horizontal peer or vertical coercive pressure, willingly commit harm against other civilians.Footnote 38 In short, micro-level, personal values are real sources of restraint, but in particular situations—notably danger (often in the contexts of fear and war), horizontal or vertical pressure, or unrestricted power and impunity—and with high-level political authorities committed to violence, such values are comparatively weak bulwarks against the escalation of violence. Theories focusing on such micro-level values are thus probably best at explaining variation among individuals in the context of the commission of violence, rather than explaining variation among cases of violence.Footnote 39
Meso-Level Sources of Restraint
At the meso level, an important game-theoretic insight is that when groups interact with other groups repeatedly over time, it is in their interest to develop formal and informal institutions to facilitate inter-communal, mutually-beneficial cooperation.Footnote 40 Daniel Chirot and Clark McCauley emphasize inter-group codes of conflict, exogamous marriage practices, commercial exchanges, and ritualized gift-giving that foster cooperation or limit violence.Footnote 41 Saumitra Jha argues that inter-ethnic commercial complementarities have been a source of tolerance between Hindus and Muslims in some towns in India, where the groups rely on each other in order for their trade and livelihoods and have developed institutions that provide incentives for continued cooperation.Footnote 42
Beyond groups, civil society organizations can foster inter-group dialogue and understanding in crisis periods.Footnote 43 In her analysis of variation in Jewish victimization rates in the Holocaust, Helen Fein argues that where the Catholic and other churches actively opposed the persecution of Jews, the level of violence diminished—especially in places not under direct Nazi or SS control—compared to places where Christian organizations was quiescent or supported such violence.Footnote 44 Timothy Longman similarly shows that in Rwanda, the Catholic and Presbyterian churches, which are highly influential in that country's society and politics, acted to facilitate and legitimize genocide in Rwanda by practicing ethnic politics, promoting subservience to state authorities, and failing to condemn the ethnic violence that had occurred in the years before the 1994 genocide.Footnote 45 Longman also provides evidence of religious dissent and non-cooperation that served to slow and displace genocidal violence in local communities in western Rwanda. Church leaders in these communities did not stop the violence, but their opposition did yield some effect, which Longman interprets as evidence of what could have happened all over Rwanda had the churches not generally condoned the violence.
Though the mechanisms differ, modern domestic non-governmental organizations, which in turn are plugged into international networks, also serve as buffers against the escalation of violence. Two recent studies from different parts of the globe make this point. In East Timor, a careful study by historian Geoffrey Robinson shows that a robust network of non-governmental organizations, with ties to national and international policymakers, was established in the late 1990s. When East Timor looked to be on the verge of rapidly escalating violence in 1999, the organizations, with the help of international journalists, spread news of the risk of genocide around the world, and served as conduits to policymakers in a position to stop the violence.Footnote 46 As other work on transnational advocacy networks has shown, information exchange was a key mechanism.Footnote 47 Externally-funded NGOs may not have shaped public attitudes on violence in East Timor, but they effectively transmitted information to international actors who could do so, and who did take forceful action to halt the violence. Studying ethnic peace in Eastern Europe in the 1990s, Patrice McMahon reaches similar conclusions. She argues that a network of transnational organizations emphasizing cooperation and information exchange took root in the 1990s that provided a common “message” of ethnic peace, a “motivation” to seek it (transnational actors could credibly offer incentives to states that avoided violence or sanctions to states that did not), and the “means” to sustain it (financial, technical, and moral assistance for dialogue, education, and training).Footnote 48 Eventually, their efforts paid off in changes in policies and social behavior.Footnote 49
Another organizational-level source of restraint is identified in the work of Jeremy Weinstein. His analysis focuses on the different initial endowments of rebel organizations. Dividing rebellions into “activist” and “opportunistic” categories, Weinstein argues that the initial endowments of the organizations shape their institutional makeup and the nature of their interactions with civilians. Because of their common political and social commitments, activist rebel organizations tend to develop shared identities and ideologies, which lead to norms of cooperation with civilians. Crucially, activist rebellions also depend on civilians for provisions, shelter, and recruitment. Thus they have an interest in restraining violence; if they alienate civilians, they risk their own future. By contrast, Weinstein argues, opportunistic rebellions that rely on cash-flowing resources like diamonds or gold, attract recruits who are more interested in making money and gaining power. The institutional structures that opportunistic rebellions build are weaker and less able to police their members; and, crucially, because of their resource endowment, they do not require as much civilian support. Thus opportunistic rebellions tend toward greater use of indiscriminate violence.Footnote 50 Whether or not the empirical world of rebellions can be divided neatly in this way,Footnote 51 Weinstein provides a compelling argument for why organizations whose survival depends on cooperative interaction have incentives for restraint. The implications of Weinstein's argument resonate with those of Jha's—certain political economies can create powerful commercial or organizational incentives for moderation.
A structurally similar argument is found in Steven Wilkinson's analysis of riots in India.Footnote 52 Unlike Jha and Varshney, both of whom locate the sources of restraint at the group level, Wilkinson focuses on political parties and the ways in which different electoral calculi create incentives for violence or non-violence. Wilkinson argues that when regional or national parties rely on the votes of minorities to form minimum winning coalitions, they will act to prevent political violence against those minorities. By contrast, where parties need the support of a dominant ethnic or religious group to win, they may stimulate violence in order to raise the salience of identity and trigger bloc voting. Here again, we see a strategic argument for why restraint might be in the interest of political elites, especially when their own electoral success depends on peaceful, cooperative relations.
This review of meso-level mechanisms suggests a number of factors that can restrain violence: group-level formal and informal mechanisms that foster cooperation (e.g. sanctioning, codes of conduct, and marriage pacts); group-level and organizational-level incentives for moderation based on a recognition of mutual dependence; organizational-level mechanisms that shape public and elite attitudes; and organizational-level mechanisms that supply information to influential actors via transnational networks. While this discussion has focused only on certain types of organizations (democracy-supporting NGOs, churches, rebel groups, and political parties), the theoretical implications also apply to other types of organizations, including the press and business lobbies.
The question, though, is whether and when they work, and here I return to the question of genocide. The answer is of course subject to empirical testing, but even from the above discussion we can generate some hypotheses. The first concerns the model of violence. If genocide takes the form of top-level, state-enforced violence, then group-level and organizational-level mechanisms are likely to be relatively weak bulwarks against the escalation of violence. Fein's work, which shows that where Nazis exerted direct control, churches had less autonomy and power, suggests as much. Similarly, Longman found that the holdout churches in Rwanda did not stop the genocide in their communities; they only changed how and where the violence occurred. Other case-study evidence from Rwanda also shows that group-level sources of moderation—the fact that inter-ethnic ties and inter-marriage were high before the genocide; that there was a great deal of neighborhood interdependence; that farmers cooperated and mutually assisted one another in planting and harvesting their crops—were weak bulwarks against state and paramilitary-backed violence.Footnote 53 Similar points apply to rural Bosnia-Herzegovina and Croatia, where pre-existing intergroup marriage, interaction, and other forms of cooperation were easily overpowered when organized military and state actors promulgated interethnic violence in the context of war.Footnote 54 But a different model of genocide might have different theoretical implications. If genocide occurs in the context of a weaker, non-centralized state, a state must win local cooperation, rather than coerce it. In those cases, meso-level mechanisms should shape the willingness of local actors to foment violence.
A second hypothesis concerns periodization. Informal exchange mechanisms are likely to be weak at the moment when mobilization and coordination to commit mass violence occur—that is, after national authorities have chosen a strategy of mass violence. But group-level and organizational-level sources of restraint will have more bite at earlier stages of escalation, as national and local elites respond to perceived threats. Even if the initial response is one of repression—for example, arresting leaders or violently breaking up protests—strong meso-level mechanisms of restraint create pressure on the authorities to avoid further escalation, or to tone down their response. By contrast, where meso-level mechanisms for restraint are weak, political authorities are more unconstrained in their decision-making as they continue to confront a perceived threat in a crisis. This is one implication of Longman's research; he argues that over decades the churches' embrace of racialist policies, its emphasis on subservience, and its failure to denounce pre-genocide violence accustomed the faithful in Rwanda to accept practices of violence, which in turn facilitated political elites' ultimate escalation. But the opposite should also be true: namely, where there is an active challenge to violent practices, elites have a harder time gaining public acceptance of such policies, and face a higher hurdle to their promulgation.
A third hypothesis is that civil society institutions cannot be assumed to exercise a peaceful, moderating influence. As Fein and Longman argue explicitly, such institutions can just as easily contribute to a genocidal consensus, or at least help form worldviews that are consistent with ethnic violence against a target population. In the Rwanda case, the Catholic Church in particular was deeply entangled in the history of colonial and post-colonial state-building; it was hardly an ideal-typical autonomous institution, independent of the state. But similar arguments are applicable to more independent civil institutions, such as the media. In Nazi Germany, the former Yugoslavia, and in Rwanda some media institutions actively promoted leaders who advocated genocide, or spread pejorative views of minorities that were consistent with genocide. Civil society can act in illiberal ways.Footnote 55
Finally, organizations' ability to restrain violence will depend on their power. Here again, we have another source of variation. In some locations, such as Rwanda before the genocide and Europe during wartime, institutions such as the Church wielded significant social and political influence. But in other locations, or considering other types of organizations in the same places, civil society institutions may be less socially and politically embedded. In McMahon and Robinson's analysis, the power of local NGOs depended on their pre-existing effective placement in transnational networks. But in other locations, human-rights NGOs may have little domestic embeddedness and weak transnational connections. Thus, the presence of peace- or tolerance-promoting civil society institutions is not sufficient to act as a source of de-escalation; one must know the extent of their power.
In short, meso-level mechanisms—from group-level informal mechanisms of cooperation to key civil society organizations—in theory can serve as important constraints against the escalation of violence of persecuted groups. They may be able to reinforce incentives to cooperate, provide information, turn public or elite opinion away from violence, articulate more moderate visions, and leverage international actors. But in assessing the effects of such institutions, scholars should consider the model of violence, the institutional setting, the period, the nature of civil society institutions, and the relative power of such institutions.
Macro-Level Sources of Restraint
The political-economy perspective introduced above points to an often overlooked dimension of extreme violence: namely, that such violence carries high costs. While the genocide studies literature now recognizes the ways in which leaders sometimes view the commission of mass violence as being in the interest of the state, and thus strategically valuable,Footnote 56 there is less attention paid to when and why political and military elites would view the commission of such violence as highly costly to state interests. Some work outside the field recognizes clearly that repressive violence is costly,Footnote 57 but the insight has not, surprisingly, been incorporated very widely into existing models of genocide and political violence.
There are a number of ways in which the practice of genocide could be costly. Most obviously, genocide invites international condemnation and carries significant reputational costs. In the presence of increasingly common judicial mechanisms of accountability, the commission of large-scale human rights abuses will now likely trigger international arrest warrants, as the twenty-first century cases of Sudan and Libya make clear. In situations of armed conflict, genocide has presumable opportunity costs: the resources devoted to violence against civilians are unavailable for broader military campaigns against an opposing military. To be sure, genocidal leaders almost always view killing civilians as integral to a war effort; nonetheless, cooler military heads may well recognize the costs of resource diversion.
But there are other domestic costs, in particular the economic costs of committing large-scale violence. There is some discussion of economic sources of restraint in the existing genocide studies literature. Reminiscent of Jha, one argument is that mutually-beneficial commercial exchange between potentially conflicting groups establishes incentives to limit damage.Footnote 58 Another finding by Barbara Harff is that low trade openness is a significant risk factor for the onset of genocide or politicide. Harff argues that the mechanism is one of international interdependence (as opposed to costs); where more interdependence exists, state elites that would commit genocide are more exposed and sensitive to international condemnation.Footnote 59 The implicit argument is one of socialization and reputational costs; states that are more exposed to and more dependent on international markets will be more sensitive to, and keen to limit, those costs.Footnote 60 A third argument is the inverse of the prominent arguments that economic loss creates frustration and that poverty lowers the opportunity costs for recruitment for violence.Footnote 61 Economic gains and prosperity, on the other hand, could promote a sense of well-being and instill attitudes of generosity; such growth could also increase the cost of recruitment towards violence.
But economic-oriented analysis should go further. I propose two additional economic sources of restraint. The first is a revenue mechanism. Genocide entails large-scale human destruction; it is sustained, committed violence over time and space. Such violence is likely to trigger significant population upheaval, which in turn could significantly disrupt an economy. If a state depends on tax revenue from sectors that would be highly sensitive to such disruption, then the state would have an incentive for restraint. There are likely to be industries, such as manufacturing and agriculture, that are highly sensitive to violence, because they require skilled labor, long planting seasons (in the case of agriculture), or stable space for transporting raw materials and marketing finished goods or produce. By contrast, some industries are likely to be insulated from violence, such as the extraction of (offshore) petroleum, or of high-value minerals and metals such as diamonds and gold. If oil wells or diamond mines are protected geographically from disruption and provide large revenue streams to states, there are fewer incentives to avoid the escalation of violence.Footnote 62
The second is a class mechanism. Where there exists a large middle class and/or a stratum of economic elites whose prosperity depends on a functioning economy, all things being equal they should act as a restraint on the escalation of violence. Middle-class and economic elite actors have an incentive to seek stability, because large-scale disruption and violence can threaten their wealth and property. They also are presumably influential; their opinions will matter to the ruling elite. The importance of middle classes for democratization has received recognition in the existing scholarship.Footnote 63 However, this insight has not really taken hold in the genocide studies and political violence literature, perhaps because the Holocaust is such an influential model; many sectors in German society, middle-class and elite included, benefited from Nazi rule.
Another macro-level source of restraint concerns formal political institutions. An argument prominent in some early comparative research is that democratic institutions establish constraints on executive power; they thereby, theoretically, serve to limit escalation and restrain the execution of highly violent policies.Footnote 64 Harff's study endorses that view, and she also finds strong empirical support for the claim that democratic states protect minority rights. A related argument is that international human-rights laws are more effective in the presence of political competition, independent judiciaries, and private media—all indicators of democratic polities.Footnote 65 These domestic democratic institutions serve as multiplier effects and increase the costs of non-compliance; in their absence, autocratic states may simply ignore previous commitments to international human rights. All of these accounts suggest that democratic institutions constrain escalation.
Ideology could also be a source of constraint. Most scholars of mass violence conceptualize ideology as a factor of escalation, but again the opposite could hold. In crises, national political culture or explicit ideologies that either promote multi-ethnic cooperation or eschew exclusivist conceptions of the national community could serve as a check on the escalation of violence. In one of the few existing studies along these lines, Chirot and McCauley argue that an embrace of individualism (as opposed to ethnic or categorical solidarity), and of modesty and doubt (as opposed to arrogance and certitude), serve to promote non-violence.Footnote 66 These are theoretical claims that clearly deserve more attention.
Finally, at the international level, there are a number of different possible mechanisms of restraint. The imposition of costs through threatened judicial action or sanctions could create incentives for leaders to moderate violence. Peacekeeping could work in the same way. Virginia Page Fortna identifies multiple ways that peacekeeping affects the prospects for a stable peace, including deterrence (by threatening to punish defectors to an agreement), increasing benefits (by signaling to donors the good behavior of belligerents), reducing uncertainty and retaliatory cycles (by monitoring and providing information), and strengthening moderates.Footnote 67
Beyond these, there are a number of other plausible arguments one can make about violence and restraint at this level of analysis. One concerns the dynamics of armed conflict. If threat perception is key to escalation, then low threats from weak armed opponents could prompt more moderate solutions. Another argument concerns capacity to inflict violence. Genocide and mass violence often require perpetrator coordination, target identification selection, and the systematic infliction of violence across time and territory. Genocide and mass violence thus require capacity—institutional capacity, and capacity for organization or alliances. Yet not all states exhibit such capacity, and the absence of such capacity could constitute a restraint.Footnote 68 A similar argument could be made about access to target populations. Genocide requires sustained access to populations that are the victims of violence, but not in all situations do perpetrating authorities have such access.
This review is not exhaustive. Rather, the main point is that existing research points to a number of plausible sources of restraint that explain why strategies of increasing levels of violence are avoided, or tried and abandoned. New research could well point to other sources. The key point is that research on violence should privilege these sources of restraint and the dynamics of non-violence as much as the sources of escalation and the dynamics of violence. Assessing both is crucial to explaining the observed variation in how cases with the potential for violence follow divergent trajectories.
IV. Toward a Theory of Genocide
This critical review of restraint implies a proposition that should be rendered explicit: macro-level factors are most consequential for shaping the process of violence that leads to genocide. Genocide is a form of large-scale, sustained, coordinated, extra-local violence. In genocide, violence is sustained for months, sometimes years; violence occurs in multiple locations in a territory under a perpetrator's control; and the violence requires coordination between institutions, such as the army, police, paramilitary units, and civilian administration. Such violence typically requires the participation of a supreme, national-level organization. Generally, that organization will be the state, but an insurgent or criminal organization that exercised territorial control over a large space could also be the key actor. However, in all cases, given the centrality of a supreme national actor, macro-level factors will be most consequential.
But local actors matter too. Empirically, local actors are essential for the identification and sorting of victims; local actors are also often the key perpetrators of violence. Thus, genocide requires sustained coordination between national and local actors. I argued above is that in a centralized state, once national authorities have chosen mass violence, meso-level actors are weak bulwarks against it. From this, two further propositions can be derived: meso-level factors of restraint will be most consequential (1) for containing the escalation during early stages in the process of violence and (2) in decentralized states where local actors enjoy a degree of autonomy.
Which macro factors matter most? The most consistent finding in the genocide studies literature is the importance of an acute crisis, in particular war. A number of scholars explicitly consider war or a similar type of mass upheaval one of the most important causal factors shaping the trajectory of genocide.Footnote 69 Other scholars, while privileging different factors, still acknowledge the importance of wars or similar types of upheaval.Footnote 70 No other empirical factor enjoys as much support in the genocide studies literature. Almost every major twentieth-century case of genocide—from the Armenian genocide, to the Holocaust, to Bosnia, to Rwanda—took place in the context of war.
The finding makes sense using the escalation-and-restraint framework elaborated above. On the escalation side, several factors prompt leaders toward violence during war. The use of violence is legitimated in war; agencies that specialize in violence, such as armies and paramilitaries, are activated in war; and violence as a protective measure is more justifiable in war. On the restraint side, in wartime, citizens and leaders often experience fear and threat, which lead them to weaken or suspend their inclinations toward moderation. In wartime emergencies, calls to suspend liberal codes of tolerance grow stronger, and hardliners have an easier time drowning out the voices of moderates.
But war is not a sufficient explanation for genocide. Emphasizing war does not resolve a central puzzle for any analysis of mass violence, which is why political and military elites would consider certain categories of citizens to be a dangerous enemy deserving of murderous violence. Moreover, empirically, most wars do not result in genocide. Here the most other consistent finding in genocide studies—that ideology matters—provides insight. Scholars argue that the dominant ideational frameworks in a polity affect how leaders interpret crises and threats, and how they define their goals. Key frameworks include nationalism (how “legitimate” citizens are distinguished from outsiders),Footnote 71 revolutionary or other visions that legitimate violence,Footnote 72 and racialist discourse about purity.Footnote 73
In my provisional theory, I endorse war and forms of exclusionary ideology as the two most important factors of escalation. But that is only half the story. On the restraint side, I argue that alternative ideational frameworks as well as state interests usually grounded in economics constitute the most important sources of moderation. While some ideologies construct categories of citizens as outsiders or enemies of the revolution, other ideational frameworks construct multi-ethnic citizenship or emphasize national values of tolerance, mediation, dialogue, and democratic compromise. Even in the face of nationalist hardliners, such ideological constructs can create a beachhead of restraint that prompt national-level decision-makers to find solutions to crises without recourse to mass violence.
Building from the economic and strategic logic identified in Jha, Weinstein, and Wilkinson, I further hypothesize that certain alignments of incentives can constitute a reserve of restraint. In particular, where leaders perceive their economies and revenue sources to be vulnerable to the escalation of violence, or if they otherwise judge the costs of escalation to be too high (for example, because of international reaction), they will seek alternative solutions that protect state revenue and their interests. No single economic structure is always violence-sensitive or -insensitive. In the aggregate, enclave economies that generate the majority of revenue from insulated industries, such as petroleum or mining, are likely to be less sensitive to the costs of escalating violence than those economies based around manufacturing or agriculture. The latter require domestic stability and often skilled labor, both of which domestic mass violence can severely disrupt. In particular, where manufacturing and agriculture is based upon the skilled labor of the target groups, there exist strong incentives to moderate the violence. The key point is that costs matter.
V. Empirical Illustration
To unpack mechanisms and to illustrate the argument empirically, I choose the scope conditions of post-colonial sub-Saharan Africa. The region is chosen not because the area is disproportionately more violent than other world regions (it is not).Footnote 74 Rather, the idea is to increase the structural similarities among cases in order to isolate key causal processes. At most times, African states pursue strategies of accommodation, incorporation, patronage, and repression. But at other times, leaders choose strategies of mass violence. I proceed with a broad cross-national overview and then delve into two cases.
To start, I construct a dependent variable of 10,000 annualized civilian deaths. Any civilian death cutoff point is arbitrary, but an average of 10,000 deaths indicates a sufficiently high threshold of mass violence as opposed to other thresholds of 1000 or 5000 civilian deaths.Footnote 75 Since 1960, I have identified sixteen cases in sub-Saharan Africa that meet this threshold.Footnote 76 All but two took place during an ongoing armed conflict. The exceptions were anti-Ibo violence in northern Nigeria in 1966, which occurred in the same year as a violent military coup, and Burundi in 1988, during a period when the state feared an insurgency. In short, these cases support the general proposition that mass violence is most likely to occur in the context of armed conflict.
The most common scenario is that a state faces an insurgency with bases of support among marginalized ethnic, religious, or regional groups. It responds with mass collective violence against the civilian populations presumed to support the rebels. The Biafran civil war is a good example, as is the most recent violence in Darfur (Sudan). The high correlation between genocide and armed conflict is consistent with other large-N studies of genocide.Footnote 77 However, in the same historical period (1960–2008), sub-Saharan Africa experienced 79 distinct episodes of armed conflict, according to PRIO data. Thus, most armed conflicts did not result in extensive mass violence.
No other macro-level factor is as robust as armed conflict. With regard to ideology, in some states, such as Rwanda and Sudan, there was evidence of ethnic nationalism. But in most cases there was an ethnic concentration of power rooted in clientelistic politics, which was not atypical for this period in sub-Saharan Africa. Some economies were mineral-based, as in the Democratic Republic of Congo, Nigeria, and Sudan, but most were not. No genocide took place in any full-fledged democracy. Most episodes took place in authoritarian states, but until the early 1990s, most African states were authoritarian. Some episodes took place during transitions to democracy. This mixed finding about regime type replicates the large-N literature, where some scholars find the factor significant and others do not.Footnote 78 We are thus still left with the central puzzle of explaining why some armed conflicts (and crises) escalate to mass violence while (most) others do not.
To probe that question, I turn to more in-depth, qualitative analyses of two cases—Rwanda in the early 1990s and Côte d'Ivoire in the 2000s. The qualitative analysis allows me to pick up on nuances not easily quantifiable across a large number of cases, and it allows me to illustrate some of the causal mechanisms discussed in this paper. While Rwanda and Côte d'Ivoire are of course different, at key points of crisis they both experienced key factors of escalation, such as armed conflict in the context of deep regime instability combined with ethnic nationalism. Both states were also comparatively high capacity states and had seen the creation of parallel militia and militant institutions. Both states experienced French and United Nations military intervention during their crises.
Yet the two cases diverged considerably in the logic of violence that elites pursued. In 1994, Rwandan government forces launched a countrywide campaign of group destruction, which claimed an estimated 500,000 civilian lives in three months.Footnote 79 By contrast, in Côte d'Ivoire, government forces repressed opposition and sanctioned violence against civilians in the war-concentrated areas in the west, but they consistently pulled back from the brink of genocide. The logic of violence in Côte d'Ivoire conformed more to accommodation and repression, rather than mass group-destructive violence, as was the case in Rwanda. The violence levels reflect as much: in contrast to Rwanda, in Côte d'Ivoire during the two most acute periods of crisis, estimates of the number of civilian deaths were generally in the low thousands.Footnote 80 While these numbers speak to terrible violence in the country, they also indicate a strategy of non-genocide.
Rwanda
Rwanda has a complex history, simplified here for brevity.Footnote 81 A small, densely populated, low-income state in Central Africa, Rwanda was first a German and then a Belgian colony. Under colonialism, European rule effectively widened and racialized differences between the two main social identity categories, Hutu and Tutsi, privileging the latter, who were treated as a superior race of “Hamites” from North Africa. In the late 1950s and early 1960s, Rwanda underwent a revolution in which Hutu counter-elites, with Belgian assistance, dislodged the Tutsi aristocracy under the premise of Hutu nationalism—namely, that the Hutu were the overwhelming majority who had been oppressed but who should rule by virtue of their demographic dominance. Rwanda's first and second Republics, while having different regional centers of power, were ruled under this premise. Tutsis were tolerated, but they had minimal representation in the state and military. In the early 1990s, Rwanda experienced a deep political crisis, brought on principally by the introduction of multi-partyism and the onset of a civil war. The former came in the form of a domestic, primarily Hutu-led threat to the ruling regime. The latter came in the form of an insurgency led by Tutsi exiles. It was the escalating response on the part of Rwandan political and military elites to these twin crises in the 1990s that ultimately led to genocide.
Policies of violence toward Tutsis varied over time. In the late 1950s and early 1960s, there was repression, including purges of Tutsis from positions of authority, murders of Tutsis, and attacks on homes. In 1963, following a major attack from Tutsi insurgents, there was mass, group-destructive killing of Tutsi civilians in one main region, Gigonkoro. Following that violence and the suppression of the rebellion, there followed a long period of tolerance combined with formal exclusion. Despite one short period of limited anti-Tutsi violence in 1973, which was linked to massacres in Burundi in 1972 and a looming coup in Rwanda in 1973, volence did not escalate again until the early 1990s. Then, facing a Hutu domestic democratic threat and a Tutsi armed threat, state policies reverted at first to repression and limited violence, including mass arrests of Tutsis, increased racist propaganda, militia formation, and several bouts of short (i.e. not sustained), localized massacres. In 1994, following the assassination of the president and a rebel advance, the Hutu political and military elites who took power launched a campaign of genocide.
By looking briefly at the Rwandan case, we can see how strategies of violence vary over time and are not linear. The initial periods of repressive violence gave way to mass violence with the onset of an armed threat, but thereafter the policies reverted to accommodation and discrimination as the insurgency was defeated and the threat subsided. Political instability in the early 1970s triggered new violence, but the policies again reverted to accommodation and discrimination in the late 1970s and 1980s once Rwanda and the region stabilized. However, with the onset of a new crisis in the 1990s, repression returned, ultimately escalating to a policy of genocide.
What were the major factors of escalation? Political authorities and ordinary perpetrators clearly connect armed conflict and the use of violence against civilians. In 1963, the First Republic was just more than a year old; its army was weak, and the polity internally divided. In December of that year, Tutsi exiles attacked and rapidly advanced toward the capital, Kigali. After an initial panic, Rwandan officials turned to Belgian advisers who helped rally a defeat of the armed attackers. Thereafter President Grégoire Kayibanda called on his ministers to prepare a self-defense plan, and they in turn launched policies against “Tutsi terrorists” that included violence against civilians. Perpetrators at the time referred to the violence as “war,” according to eyewitnesses. In the area where the violence was worst, the leading government authorities held meetings at which citizens were told that self-defense meant paralyzing and killing Tutsi civilians.Footnote 82
In the early 1960s, the state linked Tutsi civilians to Tutsi insurgents, placing them in a single collective category, and used violence against civilians as an effort to contain a perceived threat. That was also the essential logic of violence in the 1990s. After the war started in 1990, state forces arrested Tutsi elites and massacred Tutsi civilians in areas where the rebels were thought to be active. In this period, military and political elites tended to define their enemy as a collective category, as the “Tutsi” (more on this below). The final sharp escalation took place after the plane of Hutu President Juvénal Habyarimana was shot down and after the rebels began advancing. In this period, killing Tutsi civilians was again legitimized as acts of self-defense and war designed to contain the rebel threat. Tutsis were called the “enemy.”
We may never know exactly why political and military elites labeled all Tutsis as enemies, but by examining some key moments in the process of escalation it is reasonable to conclude that ideology was a major factor. The Hutu Social Revolution of the late 1950s and 1960s was predicated on ethnic nationalism, namely the idea that Hutus were the majority who had predated Tutsi “foreign” arrival. As a majority, by this logic, the core ethnic group had the rightful claim to rule. When Tutsi insurgents attacked, Hutu political and military authorities responded by claiming that they were protecting the gains of the Revolution and that Tutsis, as the minority, should not rule. The logic of defending the Revolution led to a collective categorizing of Tutsis.
The mechanism can be observed in one of the most infamous pieces of propaganda during the period, the “Hutu Ten Commandments.” Published in 1990, it exhorted Hutus to control exclusively state institutions and to spread the “Hutu Ideology” of the Revolution, among other things. Another key document was a 1992 Military Commission report, which identified the primary enemy as “extremist Tutsi within the country and abroad who are nostalgic for power and who have NEVER acknowledged and STILL DO NOT acknowledge the realities of the Social Revolution of 1959.”Footnote 83 In these documents, we see an interaction between war and ethnic nationalism, such that elites came to define their objective as protecting the interests of Hutus while collectively categorizing and excluding Tutsis.
By contrast, Rwanda had weak factors of restraint. At the meso level, the strongest civil society institution were the churches, which, as Longman shows, was closely tied to the state and facilitated racist and subservient policies. At the macro level, in ideological terms, there was no strong alternative to Hutu nationalism. While a complete archive of presidential speeches does not yet exist, an examination of 15 available speeches by Kayibanda and Habyarimana shows a consistent emphasis on the threat that Tutsis represented. While some speeches contain calls for peace, the main messages are to promote unity and to trust the ruling party and military to solve Rwanda's problems. At the economic level, the main sources of domestic tax revenue and foreign currency in the early 1990s were coffee and tea exports. While these are violence-sensitive industries, the total percentage of all exports to GDP in Rwanda was on average about 5% in the five years prior to and including 1994, according to African Development Bank indicators. Rwanda also had a tiny middle class. While detailed data do not exist on the early 1990s, a recent African Development Bank report put the figure at 2.6% of the population.Footnote 84 At the regime level, Rwanda was democratizing in the early 1990s, but was far from a consolidated democracy.
Though cursory, the Rwandan case shows how the onset of a military challenge to the state from a marginalized ethnic minority triggered the use of violence against civilians who belonged to the same identity category as the insurgents. War created a pretense for legitimizing violence as self-protection, and war facilitated the rise of hardliners and hardline discourse that called for the victimization of an ethnic category. Ethnic nationalism at least reinforced and at most was responsible for categorizing an identity category as the wartime enemy. The ideology divided the population into the rightful heirs of the state and those who would deny that right, and as the armed crisis heated up, political authorities, military officials, and ideologues drew upon the rhetoric of nationalism to define battle lines, categorize an enemy, and chose a strategy.
On the other side, little in Rwanda restrained the process of escalation. The economic costs of escalation existed, but export agriculture was a fraction of the economy, and the country had only a tiny middle class to make economic arguments to the ruling establishment. Hutu democrats existed in the country in the political opposition, but they had little recourse to an established ideological alternative to ethnic nationalism; they also had limited power. That combination of strong escalation factors coupled with weak restraint factors shaped the trajectory of violence toward an extreme version.
Côte d'Ivoire
Consider now Côte d'Ivoire. Under the country's first president, Félix Houphouët-Boigny, who ruled for more than three decades, the regime rested on economic growth through agricultural exports and political stability through patrimonial patronage. Houphouët encouraged domestic migration and regional immigration to grow cocoa and coffee; so successful were these policies that Côte d'Ivoire became in the 1970s the leading exporter of cocoa in the world and the third largest coffee exporter. At the ideological level, Houphouët preached the values of inter-ethnic cooperation, dialogue, and tolerance. He balanced his government cabinets with representatives from the main ethnic groups in Côte d'Ivoire. While Houphouët engaged in political repression of dissidents, during the first three decades of Ivoirian independence, categorical violence against social identity groups did not occur. The dominant strategy was accommodation and redistribution, punctuated by political repression in times of dissent.Footnote 85
However, Côte d'Ivoire entered into a multifaceted crisis in the late 1980s and early 1990s, due to a number of factors. First, the prices of commodities collapsed, bringing considerably less revenue. Second, the state submitted to a structural adjustment program, which weakened the patrimonial bargain. Third, the country came under domestic and international pressure to democratize. Fourth, when Houphouët died, his death triggered a succession crisis. As part of the succession crisis, two of Houphouët's protégés struggled for control. Henri Konan-Bédié, a Baoulé like the first president, eventually triumphed, and he did so in part by championing “Ivoirité,” an ethnic nationalist philosophy that emphasized local culture and an Ivoirian right to rule. Bédié's target was his rival, Alassane Ouattara, who was born in the center of Côte d'Ivoire, but had one parent from Burkina Faso and had spent a good part of his life outside the country. Ivoirité was also aimed at the more than four million immigrants, mostly Muslim, who had migrated to Côte d'Ivoire to plant coffee, cocoa, and other crops, many of whom had settled in large numbers in the fertile forested areas in the west and southwest.Footnote 86
While president, Bédié implemented a series of exclusionary and discriminatory measures, such as restricting ownership of land to Ivoirians, preventing Ouattara from contesting elections, and restricting the voting rights of many northern, Muslim Ivoirians (who were lumped into a “foreigner” category). In 1999, there was a military coup, ostensibly to reverse some of these trends. In 2000, there were failed elections, in which neither Ouattara nor Bédié were allowed to participate; eventually, another prominent opposition figure, Laurent Gbagbo, came to power. In 2002, a civil war broke out. Led by Northern political officers, the rebellion's stated objective was to reverse the political exclusion of Muslim northerners.
There followed a number of developments. There was a swift intervention, first by France, then by the African Union, and then by the United Nations. The locus of war fighting moved to the west, where there was intense violence until a buffer zone was established. Throughout this period, a number of observers reasonably worried about the risk of genocide, or of another Rwanda-like explosion, given the racist exclusionary practices and nationalism, the deep crisis and civil war, and the use of militias and paramilitaries by Gbagbo's ruling party.Footnote 87 However, the strategy of violence remained repression. There were targeted assassinations, there was in the south widespread suspicion of and discrimination against northerners, and there was suppression of protests. In the western areas, there was a great deal of tit-for-tat violence between indigenous populations and the state-funded militias supporting them, on the one hand, and northerner/foreigner/central populations who sometimes were supported by the rebels, on the other.
In the mid-2000s, several peace agreements were brokered, leading to a series of reconciliation governments. In 2010, presidential elections were held. In a runoff, Ouattara won, but Gbagbo refused to cede power, leading the rebels to start the war again. They advanced from positions in the north, and with French and UN help, captured Abidjan. In that period, again, there were many public warnings about the risk of mass violence.Footnote 88 But the state's strategy of violence remained repression of protestors and persecution of northerners and Muslims.
Like Rwanda, Côte d'Ivoire had a number of key factors of escalation: deep political instability, an explicit exclusionary nationalist ideology, a civil war, a capable state, and the use of militias and paramilitaries. But unlike Rwanda, the political responses were more moderate, ranging from discrimination and exclusion under Bédié to repression and accommodation under Gbagbo. Though the risk of genocide was high, large-scale mass violence and genocide did not emerge as the state strategy.
A key difference between Rwanda and Côte d'Ivoire is the strength of restraint factors in the latter at the macro level. Like Rwanda, Côte d'Ivoire did not have strong democratic institutions, and it was undergoing a very turbulent political transition from single-party rule. But at the ideological and economic levels, Côte d'Ivoire had stronger restraint mechanisms, which served as a reservoir of moderation.
At the ideological level, while Ivoirité tapped into widespread resentment against foreigners in the country, that narrative was counter-balanced by other ideological currents. In particular, for more than thirty years, Houphouët had emphasized the values of dialogue, tolerance, solidarity, and “brassage” (inter-mixing), often at length. An analysis of forty-seven available public speeches on national holidays between 1961 and 1990 reveals that Houphouët's consistent message was that the country had to build national unity, to avoid violence, and to modernize.Footnote 89 In about 50% of the speeches, he stressed values of solidarity or tolerance and peace (or both). He consistently referred to a “spirit of permanent dialogue” in the country. When he identified threats and challenges, the main ones were economic ones, such as terms of trade, and the difficulty of developing a harmonious nation. He did not identify an identity category as a danger or threat. In short, unity, dialogue, economic growth, and peace represent core national values that Houphouët promoted. And, indeed, his political practices reflected these principles. He sought multi-ethnic representation in his cabinets, he encouraged ethnic/religious mobility rather than exclusion and rigidity, and, under him, there was widespread intermarriage. This ideological idiom created, I argue, a strong counter-balance to the exclusionary drift of Ivoirité.
During the Ivoirian civil war, the ideas of Ivoirité did not translate into military strategy. Based on interviews with army officials and a reading of secondary sources, it appears that there was never an ethnic definition of the enemy. While a militant nationalist, President Gbagbo did not turn the war into a fight between ethnic groups. Several weeks after the September 2002 rebel attacks, for example, he said: “There is nothing that can't be agreed through dialogue in our homeland … Foreigners are not our problem.”Footnote 90 In interviews, military officers insisted that the enemy was not defined as an ethnic group and that Côte d'Ivoire was fundamentally a mixed nation, a product of “brassage.”Footnote 91 “In the heart of Côte d'Ivoire, we are brassé,” said one colonel, while a major explained emphatically that Côte d'Ivoire was a country of intermixing and intermarriage. “Ivoirité was a cultural idea used by politicians to win elections,” he said. “It was not in the minds of Ivoirians.”
At the economic level, Côte d'Ivoire had strong incentives for moderation. Côte d'Ivoire's agriculture sector is vast. Besides cocoa and coffee, Côte d'Ivoire is also home to high volumes of traffic in other primary products; their export is the largest source of state revenue. Even as late as the mid-2000s, cocoa and coffee accounted for about 50% of GDP and nearly 60% of export revenue.Footnote 92 Moreover, the main labor supply for primary products in the fertile west and southwest are Ivorian migrants and non-Iviorians. Those laborers are precisely the would-be targets of any state-led categorical mass violence campaign. According to the 1998 census figures, of the 3.2 million people involved in agriculture, 25.6% were non-Ivoirian, 31.2% were Akan, and 27.3% were northerners; these three identity categories represent the migrant/immigrant communities in the most violence-prone areas of the west and southwest.Footnote 93 In short, any mass campaign of violence against foreigners and migrants would negatively impact the main labor supply for the state's key sources of revenue. In addition, Côte d'Ivoire also has a comparatively large middle class in Africa. The country's middle class is 18.9% of the population (compared to Rwanda's 2.6%), and much of the political leadership in Gbagbo's government emerged from a middle-class milieu (including the former president, who was a history professor before entering politics).
There is indirect evidence to support the contention that the country's economic structure encouraged moderation. First, there was a concerted effort to maintain production. According to the Economist Intelligence Unit and the International Cocoa Organization, cocoa production did not fall, but rather expanded, during the course of the conflict. In the six years prior to the 2002 conflict, the average annual export tonnage of cocoa was 1.2 million tons. In the six years after 2002, the annual average was 1.35 million tons of export.Footnote 94 Similarly, if one compares the volume of tonnage imported and exported from the country's two main ports of Abidjan and San Pedro, the volume of freight increased during the course of the crisis from 16 million tons in 2002 to 18 million tons in 2006.Footnote 95 Second, in interviews, high-ranking military officers and government officials claimed that protecting the sources of wealth created in Côte d'Ivoire was a priority. Said one colonel: “What was necessary from the government side is that we had to preserve the essential fabric of the economy … That was the essential strategy.”
The risk of mass violence existed in Côte d'Ivoire, and significant violence did take place. But, in contrast to Rwanda, the sustained escalation of violence faced strong headwinds. The country's first and longest-serving president explicitly sought to instill national values of solidarity, tolerance, and dialogue; he encouraged migration and immigration; and overall he championed a vision of a multi-ethnic society. However flawed his policies were, under him a distinctive vision of the Ivoirian state emerged and became rooted within the minds of military and political elites, such that when a severe crisis arrived and ethnic-nationalist claims began to emerge, there existed a strong counternarrative to defining the enemy as a categorical identity group. Côte d'Ivoire is also a country that was transformed into an agricultural powerhouse, which in turn fed a middle class. That economic structure created elite incentives to seek solutions other than a campaign of extensive violence that would disrupt the key sources of state revenue.
Other factors mattered. Côte d'Ivoire experienced a more robust international military intervention than did Rwanda, and the Ivoirian mission had a stronger civilian protection mandate, for example. But domestic dynamics of restraint in the ideological and economic domains powerfully shaped a different trajectory of violence from that in Rwanda.
VI. Conclusions and Implications
To understand genocide, we need to study negative cases—where the preconditions exist but where genocide does not occur—as well as positive cases. Factors that lead authorities to moderate and de-escalate violence should receive more scholarly attention than they do, with special emphasis on developing an analytical framework by which to conceptualize the process of violence as a balance of factors of escalation and factors of restraint. A provisional theory of genocide might claim that armed conflict and exclusionary ideologies are the most significant factors of escalation, while non-exclusionary ideologies and certain alignments of incentives act as restraints on violence, as even a brief consideration of mass violence (and its absence) in sub-Saharan Africa demonstrates. This might explain other cases around the world where the potential for mass violence existed, but the violence did not materialize, such as MaliFootnote 96 and Kenya,Footnote 97 for examples.
These claims also have theoretical implications for the comparative study of violence. My broad goal is to prompt research on why violence fails, or why political decision-makers choose to deescalate and moderate violence. We have relatively few comparative findings about which restraint factors, and at what level, matter most for the occurrence or non-occurrence of different forms of political violence. There is also little qualitative research on how restraint shapes the process of violence or its absence.Footnote 98 And we have only limited comparative research on how marginalizing or overcoming restraint is an integral part of escalation, in particular at the macro and meso levels. Yet detailed case studies suggest that overcoming restraint is central to how and why violence unfolds on a large scale.Footnote 99 Any theory of large-scale violence must account for how and why it happens.Footnote 100
New empirical and theoretical research on restraint will in turn have potentially important policy implications. The dominant predictive approach in the public realm for identifying likely episodes of mass violence is to catalog risk factors.Footnote 101 The approach is derivative of genocide studies scholarship, which emphasizes factors of escalation.Footnote 102 A more holistic policy approach to forecasting both genocide and other forms of violence should incorporate restraint factors. Moreover, observing the strength of these restraints should serve as an indicator of whether escalation to mass violence is more or less likely. New research should develop a series of restraint indicators that can be monitored.
Finally, the framework has more general normative implications. By claiming that restraint is fundamental to the process of violence, the argument reinforces the policy assumption that external investments in peace can make a difference. In short, because restraint matters to how and why violence escalates, finding ways to build up sources of restraint should diminish the likelihood of future violence.
In the past two decades alone, states, international organizations, non-governmental organizations, and committed individuals have spent much time, energy, and resources on a variety of peacebuilding strategies. These include formal, increasingly complex peacekeeping missions led by the United Nations and regional organizations. They also include supporting a range of accountability mechanisms, from international and domestic trials to national truth commissions and local justice processes, in order to increase social cohesion. At the community level, peacebuilding policies include establishing multi-constituent peace committees to prevent the escalation of local crises or consensus-oriented, community-based committees to disburse development aid after war.Footnote 103 At the national level, transnational activists have encouraged a variety of conflict mediation measures. A good example is the work of Howard Wolpe, the recently-deceased former U.S. Congressman and Special Envoy to the Great Lakes Region of Africa, who helped develop an innovative campaign to train political and military leaders to recognize mutual interests and to avoid the kinds of collective stereotyping that can escalate conflict.Footnote 104
Much remains to be discovered about which interventions work and why, as well as which interventions backfire and why.Footnote 105 Some recent work on peacekeeping describes the conditions and mechanisms that make peacekeeping successful.Footnote 106 For example, Page Fortna argues that peacekeeping strengthens moderates; Lise Morjé Howard and Séverine Autesserre in different ways emphasize organizational learning about local conditions. My arguments about restraint dovetail with these claims. Not only will there always be voices of moderation in any situation—ones that should be strengthened—but there will always be domestic sources of restraint. Successful peacekeeping should investigate and invest in them. An innovative study by James Fearon, Macartan Humphreys, and Jeremy Weinstein found that an externally-funded community development project in Liberia had the intended effect of increasing social cohesion.Footnote 107 More studies of that sort, as well as studies that assess the impact of peace committees, are in order. While increased community-level cohesion and mechanisms for short-circuiting conflict may not deter genocide, my arguments about restraint imply that such developments should impact the likelihood of other forms of violence, such as riots or violence over access to resources.
There is a domestic politics story here as well, one that is consistent with the mediation work of Wolpe and others who seek to change how elites perceive and act in crises. In the violence literature, we gravitate to the Adolf Hitlers, Slobodan Milosevices, Pol Pots, and Muammar Qaddafis of the world—to political leaders who stoke loathing, fear, and racism. But an analytical focus on restraint should prompt us to focus on how other leaders defuse tension, or fashion ideas of social cooperation as being in the national interest. One lesson from the albeit brief comparative case study in this paper is that there exist margins for maneuver and domestic sources of innovation for durably shaping elite and public opinion. How political leaders craft national narratives at key moments in history has long-term implications for the probability of mass violence.Footnote 108 In Côte d'Ivoire, Houphouët's emphasis on dialogue and a multi-ethnic political community established a reservoir of moderation, whereas in Rwanda the post-colonial ideological emphasis on ethnic majoritarianism had the opposite effect. In short, how political leaders retreat from the brink and how they foster unity matters, and deserves scholarly recognition and study. Through such study, we political scientists can both develop better theoretical understandings of the causes of mass violence, and perhaps contribute to crafting interventions that will lead to more peaceful forms of politics.
