Ethnicity, the State, and the Duration of Civil Wars Julian Wucherpfennig Center for Comparative and International Studies ETH Zürich 8092 Zürich, Switzerland [email protected] Kristian Skrede Gleditsch Department of Government University of Essex Colchester CO4 3SQ United Kingdom [email protected] Nils Metternich Department of Government University of Essex Colchester CO4 3SQ United Kingdom [email protected] Lars-Erik Cederman Center for Comparative and International Studies ETH Zürich 8092 Zürich, Switzerland [email protected] Preliminary Draft. Please do not cite without the authors’ permission. Paper prepared for the GROW-Workshop at the Peace Research Institute Oslo 18 June 2010 Abstract Whereas previous research has focused primarily on how ethnicity may trigger civil war, its effect on conflict duration remains disputed. We argue that ethnicity as such does not affect civil war duration. However, if states implement discriminatory policies based on ethnicity, members of politically excluded ethnic groups develop grievances that can make rebel organizations more durable. This argument challenges the view that only cognitive aspects of ethnic identities, e.g. information, trust, or common language, help to overcome collective action problems. We stress that state-induced grievances motivate members of excluded ethnic groups to participate in enduring rebel organizations. In line with our theoretical argument, we find that conflicts last longer if rebel organizations operate on behalf of politically excluded ethnic groups. Contrary to what is assumed in the literature, state-induced grievances, and not rigid identities, prolong civil wars. 1 Introduction Ethnicity remains a contested factor in the study of civil war. Whereas previous research has focused primarily on the role of ethnicity for the onset of civil war, its effect on the duration of such conflicts remains both theoretically and empirically understudied. On a most general level, the discussion revolves around whether ethnic wars systematically differ from non-ethnic civil wars in regard to their characteristics, for example the factors that cause their onset or how they are conducted. Ethnic wars are shaped by an ethnic rather than ideological or economic agenda that characterizes the fighting. It is argued that ethnic identities cannot be transcended and create a unique sense of belonging that facilitates collective action and is reinforced through fighting. Proponents of the ethnic versus non-ethnic conflict differentiation postulate that due to the rigid nature of ethnic identities, ethnic conflicts quickly become intractable, are inherently difficult to resolve, and thus last longer (Kaufmann 1996, 1998; Kaufman 2001, 2006; Horowitz 1985). The opposite skeptical view implies that ethnic identities are merely a cover story concealing underlying economic or private interests and do not affect the conflict dynamics (Mueller 2004). Moreover, other scholars highlight that ethnic identities can be – and frequently are –transcended in civil wars (Kalyvas 2006). From this perspective, ethnicity as such should not have a clear or consistent effect on the dynamics of fighting. This discussion has far-reaching repercussions since it ultimately questions the distinction between ethnic and non-ethnic conflicts. It also challenges the conceptualization of ethnic groups as meaningful actors in civil wars. In this paper we argue that many of the 2 ambiguities and discussions in the literature stem from the failure to distinguish between two conceptually different dimensions. First, we address the issue of agency in civil wars by making a clear distinction between ethnic groups and rebel organizations, while retaining both actor types in our model. Second, we reassess the role of the state by highlighting the effect of exclusionary state policies along ethnic lines. Our main argument is that ethnicity as such does not affect civil war duration. However, if the state implements discriminatory policies based on ethnicity, members of politically excluded ethnic groups develop grievances that facilitate the durability of rebel organizations. This argument challenges the view that only cognitive aspects of ethnic identities, e.g. information, trust, or common language, help to overcome collective action problems. We stress that state-induced grievances motivate members of excluded ethnic groups to participate in enduring rebel organizations. In contrast to previous research, our explanation for conflict duration contains a strong political element that stresses the ethno-political context of civil wars. In short, by focusing on state-induced ethnic policies, we add causal depth to the linkage between ethnicity and conflict duration. In the remainder of the paper we put these theoretical considerations to an empirical test. Drawing on a new dataset that systematically links rebel organizations to politically relevant ethnic groups, we demonstrate that ethnicity indeed can prolong civil wars. However, this effect is entirely driven by the subset of conflicts with state-induced ethnic grievances. These grievances stem from state actions that systematically exclude specific 3 ethnic groups from access to state power. Thus, contrary to what is assumed by many scholars, we find no evidence for an inherent effect of ethnicity. Rather, ethno-nationalist state-policies shape conflict dynamics. Civil War Duration and Ethnicity By definition, fighting in civil wars occurs between the government and one or several rebel organizations challenging the state through violent means with aim of overtaking the state’s monopoly of force, either fully or for some limited territory (see Sambanis 2004 for differences between various operationalizations). What differentiates civil wars from other types of violence, for example communal violence, is that the state must be involved as an actor in the conflict. Yet, as highlighted by this literature review, many studies simply assume the state away. Furthermore, the operational definition leaves some ambiguity concerning the “other” actor challenging the state. As we point out in this section, depending on the theoretical approach, the common choices are (1) ethnic groups and (2) rebel organizations. Against the background of the existing literature, we identify the necessity to specify the role of the state and theorize the linkage between ethnic groups and rebel organizations. Previous explanations for the dynamics of violence in civil wars can be classified by a simple 2x2 typology given in Table 1. It differentiates between whether the government’s challenger is defined as an ethnic group or a rebel organization, and whether the state is 4 conceptualized as an active or passive actor. The conceptual questions are: (1) who is the non-state actor, and (2) what is the role of the state? We address both questions below. Table 1: Approaches to Civil War Duration The State Passive Active Non-State Actor Ethnic Groups Rebel Organizations (a) Security Dilemma (b) Political Economy (c) Sons of the Soil (d) Ethnic Defection Ethnic Groups or Rebel Organizations? Seeking to understand the surge of civil wars during the 1990s many scholars suggested that conflicts are fought between ethnic groups. Much of this work referred to the conflicts in Yugoslavia, the former Soviet Union, Burundi, and Rwanda. Drawing on international relations theory, Posen (1993) explained these conflicts as the result of an ethnic security dilemma caused by state breakdown. According to this approach, ethnic groups engage in preemptive violence because they fear for their own survival. The particular history of intergroup relations, but also the physical and ethnic geography are argued to be amongst the core determinants of the security dilemma (Posen 1993; Toft 2003). The hypothesized dynamics of this approach have important implications for the duration of such conflicts. While not all civil wars are ethnic, civil wars fought between ethnic groups quickly become protracted and therefore endure longer than non-ethnic civil wars (Kaufmann 1996; 1998; Rose 2000; Kaufman 2001, 2006; Horowitz 1985; van Evera 2001). This perspective is grounded in the assertion that ethnicity is fixed thus distinguishing it from other identity concepts. As Kaufmann explains: 5 Ethnic conflicts are disputes between communities which see themselves as having distinct heritages, over the power relationship between the communities, while ideological civil wars are contests between factions within the same community over how that community should be governed. The key difference is the flexibility of individual loyalties, which are quite fluid in ideological conflicts, but almost completely rigid in ethnic wars (Kaufman 1996: 138). Moreover, battles, massacres, and other forms of violence harden these identities to the point where compromises become delegitimized (Kaufman 2006: 205). As a consequence, continued fighting is inevitable, leading to a “spiral of escalation” that renders these conflicts extremely difficult, or entirely impossible, to resolve without third party intervention (Kaufman 2006). Indeed, proponents of this view go as far as claiming that partition is the only possible solution to such conflicts (Kaufmann 1996). While this line of research claims that ethnic wars last longer, little systematic evidence has been provided to support this. Indeed, empirical insights from case studies are challenged by quantitative research revealing no significant effect of ethnic indicators such as ethno-linguistic fractionalization or ethnic polarization (Collier 2000; Fearon 2004; Cunningham 2006, 2010; Brandt et al. 2008; Cunningham et al. 2009; although see Collier, Hoeffler and Söderbom 2004; Montalvo and Reynal Querol 2010). Thus, our theoretical and empirical approach is motivated by the need to reconcile these qualitative and quantitative findings. 6 The opposing view to ethnic group based approaches questions the relevance of ethnic groups as theoretically meaningful actors in civil wars. From this perspective, rebel organizations, rather than ethnic groups, are the appropriate unit of analysis (Sinno 2008; Kalyvas 2008). An important argument in favor of this view is that rebel organizations are always present in civil wars, ethnic or not (Cunningham et al. 2009). According to one strand of this approach, fighting in civil war is driven by incentives to secure private gains. Such insurgencies are most likely to emerge and endure under conditions of state weakness allowing rebel organizations to sustain successful operations (c.f. Fearon and Laitin 2003; Collier and Hoeffler 2004). More generally, while some scholars opt for the more neutral label of ‘rebel organizations’ (e.g. Kalyvas 2006), others are more subjective when referring to the rebels as ‘greedy bandits’ (e.g. Collier and Hoeffler 2004), or equating them with ‘criminals’, ‘thugs’, or ‘warlord gangs’ (Mueller 2003). The tacit nature of rebel organizations has important theoretical implications for how quickly conflicts can be resolved. For example, Collier, Hoeffler and Söderbom (2004) argue that the presence of looting opportunities prolongs conflicts, since rebel revenues are high and recruitment costs are low. Presumably these conditions for “doing well out of war” (Collier 2000) are found in the presence of lootable resources, such as narcotics, gemstones or oil, and where effective state control is largely absent (see also Buhaug et al. 2009; Fearon 2004). However, Collier, Hoeffler and Söderbom (2004) also find that a moderate degree of ethnic diversity is associated with long durations, which they interpret as facilitating cohesion on the side of the rebels. What is not supported by the empirical data, the 7 authors argue, are effects of altruistic post-conflict payoffs related to its outcome, with an explicit mention of political repression as an example. Accordingly, fighting occurs for private rather than for collective benefits. Similarly, Mueller (2004) argues that at the micro-level beliefs and emotions, such as hatred, are not stable enough to motivate long fighting durations. Rather than the cause, they are the result of violence, or merely its excuse. This view is shared by Brubaker who advances an analytical focus on organizations rather than groups since ethnic identities (“groupness”) become relevant merely as a result of conflict: “Although […] perceived groupness does not necessarily reflect what is felt and experienced by participants in an event, a compelling ex-post framing can exercise a powerful feedback effect, shaping subsequent experience and increasing levels of groupness” (Brubaker 2004: 16). In line with most political economy approaches, Brubaker claims that ethnic identities are at best endogenous to conflict: “What is represented as ethnic conflict or ethnic war—such as violence in former Yugoslavia— may have as much or more to do with thuggery, warlordship, opportunistic looting, and black-market profiteering than with ethnicity” (Brubaker 2004: 19). Thus, “ethnic” warfare is simply a cover story for criminal violence and predation, an interpretation that is in line with the dominant political explanations for the onset of such conflicts (e.g. Fearon and Laitin 2003; Collier and Hoeffler 2004). Others do not dismiss the effect of ethnicity in regard to the onset of conflict, but argue that it does not clearly impact on the dynamics of fighting. For example, Kalyvas claims 8 that ethnic identities are unstable over the course of a conflict: “Even when ethnic divisions cause the eruption of civil war in the first place, these identities do not always remain stable and fixed during the conflict; if they do change, they may soften rather than only harden” (Kalyvas 2008: 1045). Thus, collective and/or individual preferences and identities are continuously reshaped during the course of a civil war, while master cleavages are frequently undermined by local ones (Kalyvas 2006, 2007). This argument runs directly counter the claim that ethnic identities are fixed since it implies that they can be—and frequently are—transcended through the mechanism of ethnic defection, i.e. fighting for ethnic “others” against members of one’s own group. Thus, while Kaufmann (2005: 183) bolsters his key argument with the conjecture that “cross-ethnic recruitment or defection is rare,” Kalyvas (2008) attempts to demonstrate that this view is mistaken. We return to this argument below. While many authors concentrate on either natural resources or ethnicity as an explanation for civil war, others try to specify the conditions under which one or the other matters. Weinstein (2007) presents a theory of rebel behavior in which ethnicity represents an endowment from which rebel leaders choose to draw in the absence of more favorable economic resources, such as foreign sponsors or natural resources. Thus, rebel organizations are regarded as political entrepreneurs, whereas the presence of ethnic identities—just like lootable resources—represent an exogenously given endowment that has an effect on the organizational structure of a rebel organization (see also Beardsley and McQuinn 2009). However, social and economic endowments have a differential effect on the (self-)selection of fighters. In contrast to recruits motivated by personal 9 gains, fighters recruited through ethnic identities are more committed. By implication, they should therefore be expected to fight longer. In sum, most political economy approaches argue that ethnic identities have no predictable or uniform effect on the dynamics of warfare, and thus on its duration. Accordingly, it is not clear whether fighting hardens or softens identities. Since it is not ethnic groups that fight, but simply rebel organizations. Indeed, due to the assumed endogenous nature of ethnic identities, groups are held to be entirely irrelevant since their presence does not yield any systematic prediction with regard to the process of warfare. The Role of the State. One common problem pertaining to both approaches is the alleged passiveness or absence of the state as an actor (we discuss notable exceptions below). This realization brings us to the lower part of Table 1. Political economy approaches focusing on rebel organizations typically argue that this type of actor engages in civil violence when state weakness makes self-enrichment feasible. Thus, the inability of the state to prevent violence creates a sufficient opportunity structure for rebel organizations. Similarly, group-based approaches frequently explain the occurrence of violence with a security dilemma that arises when the state cannot credibly provide security to groups due to structural weakness (Posen 1996). According to this explanation, by assumption the state is largely absent (Cederman, Wimmer and Min 2010: 88). Indeed it is assumed that state failure provides an arena of ‘emerging anarchy’ in which such ethnic conflicts can occur. 10 As notable exceptions, Fearon (2004) and Fearon and Laitin (2010), drawing heavily on Weiner (1978), analyze particular conflict dynamics pertaining to a subset of ethnic conflicts: sons of the soil conflicts (quadrant (c) in Table 1). These are described as conflicts between a peripheral but geographically concentrated ethnic minority, the sons of the soil, that is confronted with state-initiated migration to the minority’s perceived homelands by a dominant ethnic group from the center. Scarce resources, such as land or jobs, result in strong grievances and local struggles. Importantly, these grievances are state-induced, since the migration is assumed to be at least partly orchestrated by the government. They are likely to escalate where the state sides with migrants to appease their support base. Because migration is path dependent and the government has a selfinterest in maintaining such policies it is unable to credibly commit to a peace agreement. As argued by Fearon (2004), this renders the sons of the soil conflicts especially difficult to end, thus prolonging the armed struggle. Another rare example of active state behavior in the civil war literature is found in Kalyvas’s (2008) account of ethnic defection, which is held as the prime argument as to why ethnic identities should be treated as fluid, rather than fixed. It is described as follows: “(a) the incumbent state is willing and able to recruit members of the rebellious ethnic minority, (b) a substantial number of individuals collaborate with the a political actor explicitly opposed to their own ethnic group, and (c) fighters and sympathizers switch sides from ethnic rebels to the state” (Kalyvas 2008: 1050). 11 But although the theory postulates the effect of ethnicity to be empirically variable, this variation is not random. Rather, it is the result of systematic state action. As Kalyvas (2008: 1045) explains, “the behavioral potential of ethnicity is empirically variable … a key determinant … is the willingness of incumbent states facing ethnic rebellion to recruit ethnic defectors, which in turn depends on their resources.” Put differently, state strength, in particular territorial control, is regarded as a systematic modifier of ethnic identities; hence we locate this approach in quadrant (d) in Table 1. In sum, the theoretical effect of ethnicity is disputed and the the subject of an ongoing debate. Our survey of the literature shows that ethnic identities are either treated as exogenously fixed or endogenous to conflict. Surprisingly, however, neither approach has been tested systematically using large-n data. A crucial question thus concerns the concrete causal mechanisms through which ethnicity may operate. This is what we address in the next section by laying out a critique of the assumed nature of ethnic identities. Cognition or Motivation? Ethnicity and Nationalism: Ethnic defection represents an interesting case in point. The defection argument rests largely on anecdotal or case-study evidence, with systematic evidence across a large number of cases missing. Moreover, even if ethnic defection is not as rare as argued by others (e.g. Kaufmann 2005), its empirical rate vis-à-vis non-ethnic defection from rebel organizations remains unclear, as does the net effect of ethnicity in the presence of defection. Thus, even if defection occurs across both ethnic and non-ethnic lines, the relative frequencies are based on 12 conjectures. This net effect is important in light of another point emphasized by Kalyvas, namely that ethnicity is argued to have an effect on hindering co-ethnics from free-riding in civil wars: “If an individual’s chances of being victimized depend on a profile rather than on his or her behavior, then shunning participation in the rebellion and free riding may actually prove deadlier than joining it” (Kalyvas 2007: 186).1 Thus, under the very conditions that make free-riding costly and risky (e.g. a profile), defection should also be more difficult.2 Accepting both arguments at face value then leads to the prediction of a positive net effect of ethnicity instead of the hypothesized non-effect. The theoretical argument about ethnic profiles hindering free riding due to straightforward identification by members of their own group implies that ethnic identities are viewed not so much as identities, but more as traits, such as one’s physique or language, which in turn facilitate the cognitive act of identifying group membership. Such arguments are commonplace in the literature; ethnic groups are commonly conceived of as social networks that facilitate cooperation through the threat of social sanction (Axelrod 1984; Fearon and Laitin 1996; Habyarimiana et al. 2007). Thus, it is the social context within which rebellion takes place that affects micro-level cooperation by providing cognitive cues (see Lichbach 1998). However, this fails to explain where ethnic preferences come from in the first place (Varshney 2003). As a consequence, we conclude that much of the current debate 1 The argument is also not free from problems; for free-riding to become costly in the first place, the organization must have overcome at least an initial collective action problem as to become powerful enough to be able to induce costs upon defectors. 2 Lyall (2010) makes a related argument, aiming to demonstrate that co-ethnics are superior counterinsurgents, potentially because of better “reachability”. 13 revolves around the very nature of identities. Where ethnicity is treated as fixed, cognitive profiles are central, but preferences merely assumed as given. Where ethnicity is treated as fluid, motivations are assumed to be endogenous to conflict dynamics. However, this latter view tends to disregard the role of the state as the key actor shaping ethnic identities and preferences through ethno-nationalist politics. The principle of nationalism demands the unit of governance and the nation to be congruent (Gellner, 1983). Whereas the former is commonly conceived of as the state, the latter refers to a community of common origin and shared historical destiny (Anderson 1991). Accordingly, groups struggle over inclusion and exclusion from state power. The argument is crucial. Whereas ethnic profiles can hardly be changed, ethnic identities can be charged with grievances, but not necessarily so. Ethno-nationalism offers a systematic origin that explains ethnic grievances as motivations to attain ethno-nationalist representation wherever groups in power hinder other groups from access to executive power. Thus, a more complete argument of ethnicity’s relevance for war-fighting has to factor in the state as an explicit actor. We present such a theory below. Theory: The Macro-Context of Micro-Recruitment Our theory comprises all three types of actors described above: the state, ethnic groups and rebel organizations. A basic visualization of our approach is given in Figure 1. 14 Broadly speaking, we argue that state induced ethno-nationalist policies lead to a particular type of grievances amongst members of the affected groups. Previous research has already demonstrated that such conditions are more likely to cause conflict onset (Cederman et al. 2010). Our theoretical approach laid out in this section stipulates that rebel organizations recruiting from excluded groups are staffed with highly committed fighters that are willing to endure long periods of fighting. Thus, by focusing on the relation between ethnic groups and rebel organizations, our theory highlights how ethnonationalist policies of exclusion affect conflict duration. We challenge previous research which suggests that ethnic identities, understood as fixed and rigid, can help to overcome collective action problems, since ethnic identities increase “social cohesion on the rebel side” (Collier et al. 2004: 263) and provide strong networks and mechanisms that sanction potential defectors (Fearon and Laitin 1996, Habyarimana et al. 2007). Likewise, Weinstein (2007) has argued that rebel leaders can resort to mobilizing on the basis of identity, rather than economic endowments. This line of research would suggest that rebel organizations relying on ethnic linkages have advantages in sustaining their fighting efforts. This implies that ethnic linkages as such lead to longer fighting durations. H1: Rebel organizations linked to ethnic groups fight longer. 15 Figure 1: A Theory of Fighting in Civil Wars State Groups nationalism opportunity cost exclusion Organizations political entrepreneurs Violence duration recruit, delegate Our theoretical point of departure is a concern brought forward by Kalyvas (2008: 1063) who argues that the “relation [of rebel organizations] to underlying populations must be the object of systematic theoretical and empirical investigation as opposed to mere assumption” (see also Brubaker 2004). Whereas these authors base their argument about the putative ex-ante irrelevance of ethnicity on the varying connection to ethnicity in rebel organizations, we instead take their concern seriously by focusing systematically on variation in nexus between rebel organizations and ethnic groups and conflict duration. We begin with the assumption that rebel organizations are political entrepreneurs seeking to mobilize or sustain forces with sufficient capacity to compete against the government by violent means. Their goal is to challenge the government’s force monopoly, either in the entire country or locally in some limited territory. Successfully sustaining forces is therefore a key mechanism that differentiates short from long civil wars. The key challenge for rebel organizations is the recruitment of sufficient people to challenge the government. But initial recruitment is insufficient to sustain armed conflict; rebel organizations in equal measure need to provide incentives to retain recruits over 16 longer periods of time. In other words, rebel organizations need to create incentives for the fighters not to abandon the rebellion. This can only be achieved by rewarding fighters through different kinds of compensation. We argue that variation in the reward structure– time horizons in particular–explain why some organizations endure longer than others (cf. Weinstein 2007). Here, we extend the literature by emphasizing that this reward structure is not only shaped by economic opportunities, but also by the state imposed ethno-political power configuration. This approach increases the causal depth of how ethnicity matters and has specific empirical implications. Two main strategies can be pursued to attract and sustain a sufficient number of fighters. First, an organization may recruit opportunistic fighters, who seek short-run economic benefits. Such recruits (“greedy bandits'') need to be paid immediately (short-run payments), for example through rents generated by lootable resources. The availability of short-run payments is therefore beneficial to the recruitment process, because the rebel organization does not need to credibly promise future rewards. Even though short-run incentives might attract more opportunistic fighters that are more likely to attack civilians and harder to control (Weinstein, 2004), we argue that this does not lead to shorter conflict durations (see also Buhaug et al. 2009; Collier et al. 2004; Fearon 2004; etc.). We hypothesize that the availability of short-run incentives actually sustains the fighting efforts of rebel organizations. H2: Rebel organizations fighting in the presence of lootable resources exhibit longer conflict durations. 17 Second, as an alternative strategy, rebel organizations may promise rewards conditional on conflict outcomes. Initially, fighters may be attracted by the prospect of enjoying political representation or better economic access once the government is defeated (longrun payments). However, the inherent difficulty of such a reward structure is that is associated with greater uncertainty for at least two reasons. First, the rewards are conditional on the future success of the rebel organization. Second, the rebel organizations face a commitment problem if they rely on long-term rewards, because rebel leaders have incentives to renege on reward promises once the government is defeated. If recruitment is intricately linked to uncertain future rewards, long fighting durations should be most likely when individual opportunity costs are low. In other words, claims about future benefits resonate particularly well in the presence of grievances. While there are multiple scenarios under which such grievances can arise, we focus on ethnonationalist policies of the state. In such a model, challengers seek to avoid the rule dominant ethnic groups by gaining access to the state or leaving it in favor of a new polity or an already existing kin state (see Cederman, Wimmer and Min 2010). Ethnic groups excluded from state power are deprived of political representation and likely to be disadvantaged in accessing government provided goods. This implies that their status quo is highly unsatisfactory. Therefore, we argue that individuals excluded from economic goods or political representation are more likely to continue fighting for long periods of time, even in the absence of short-term economic rewards, precisely because they have less to lose to begin with. 18 More specifically, we argue that the socio-economic and ethno-political base of recruits shapes their incentives to fight. Ethnic exclusion manifests itself in everyday life; indeed members of excluded groups are subject to humiliation and are treated like second-class citizens compared to other groups. Thus, claims by rebel organizations are likely to resonate with members of excluded ethnic groups since the prospect of future benefits outweighs the costs of fighting. This is because the members’ opportunity costs are relatively low as they have little or no access to the political process and are disadvantaged or even actively discriminated when it comes to redistributive policies of the state. Therefore, members of excluded ethnic groups are more likely to continue fighting in enduring rebel organizations. By contrast, members of groups included in the political process enjoy political rights and re-distributional benefits. Included group members should therefore be more satisfied with the status quo than excluded group members. As a consequence, dissatisfied group members are particularly likely to bear long fighting durations as their opportunity costs are low due to their exclusion status, whereas included group members are more likely defect, resulting in shorter fighting durations. Hence, we state our core hypothesis: H3: Rebel organizations recruiting from and operating on behalf of excluded ethnic groups are associated with longer fighting durations. 19 Data and Empirical Strategy Data Coding the group-organization nexus. We draw on a new data project that systematically codes the link between rebel organization and ethnic groups. This dataset relates two existing datasets: a dataset on rebel organizations (NSA) in civil wars by Cunningham et al. (2009), and a dataset on politically relevant ethnic groups worldwide (EPR) by Cederman et al. (2010). Thus, rather than determining whether a given conflict is ethnic or not, we focus on the explicit linkage between rebel organizations and ethnic groups. The logic of the dataset, which we label NSA2EPR, is depicted in Figure 2 and discussed further below. Figure 2: NSA2EPR. Relationship between Ethnic Groups and Rebel Organizations 20 Focusing on rebel organizations as the unit of analysis, rather than ethnic groups, has the advantage of seamlessly relating our theoretical considerations to the empirical analysis. Moreover, since many conflicts involve more than one rebel organization (Cunningham 2006; Cunningham et al. 2009; Metternich and Wucherpfennig 2010) we can differentiate between organizations that do and do not pursue an ethnic agenda. Drawing on information included in EPR, we further differentiate whether the ethnic groups within a rebel organization are subject to state-induced exclusionary policies. We can therefore establish the ethno-political context within which particular rebel organizations are situated. Finally, note that the mapping is many-to-many, i.e. a single rebel organization can share linkages with multiple ethnic groups. Similarly, a single ethnic group can be connected to multiple rebel organizations. Regarding the linkage between ethnic groups and rebel organizations, we focus on two criteria. The first criterion assesses the ethnicity of the fighters. Put differently, we coded from which ethnic groups, if any, a particular rebel organization recruits their fighters. For such a linkage to exist, we require a significant number of the group members to actively participate in the organization’s combat operations. While not all rebel organizations recruit along ethnic lines, recruitment alone is insufficient because it may be merely the result of local availability. In other words, depending on where a rebel organization is active, agency may not be deliberate; ethnic recruitment may simply be the result of local availability, and may have to do little with the organization’s actual agenda. Therefore our second criterion is whether a given rebel organization publicly announces to operate on behalf of the relevant ethnic group, i.e. pursues an interest that is 21 linked to the group’s fate. If recruitment and claim occur jointly, we code a rebel organization as “ethnic”. Coding Grievances: the Ethno-Political Context. We argue that much of the current literature equates ethnicity in civil wars with rigid identities, but tends to disregard variability in actual grievances. We also argue that ethno-nationalist policies by the state that systematically exclude parts of the population from political representation on the grounds of ethnicity are likely to induce such grievances (see Cederman et al. 2010). Members of ethnic groups excluded from political representation are more likely to exhibit strong grievances compared to members of groups that are part of the political process at the center. It is this macro-political context within which rebel organizations fight on behalf of ethnic groups, and which alters the individual level opportunity costs, and thus micro-motives, of the fighters. Fortunately, EPR contains information about the ethno-political power status of ethnic groups. Through the linkage created by NSA2EPR we are able to integrate this information at the level of rebel organizations. Thus, where a linkage between an ethnic group and a rebel organization exists, we assess whether the group was excluded or included from state power. This allows us to differentiate between ethnic rebel organizations with and without ethno-nationalist motivated fighters against a common baseline of rebel organizations with no ethnic linkage. 22 Coding Resources. Much of the literature on civil war duration emphasizes the importance of lootable resources in the conflict zone as a means to finance fighting. We rely on data by Buhaug et al. (2009) that indicate the presence of three types of resources in the area where the conflict takes places. These resources are gemstones, petroleum and drugs, all coded as dummy variables. Control Variables. We also employ a set of control variables. The NSA data set provides a series of variables on rebel organizations (Cunningham et al., 2009). The variable Territorial Control indicates whether rebel organizations have actual territorial control over some area in the country. Strong Central Command is a variable that indicates whether a rebel organization has a coherent command structure. Additionally, we include a dummy variable Legal Political Wing providing information about a potential rebel organization’s political wing. On the country level we include a Democracy dummy taking the value of one if a country has a Polity score equal or greater than 6. Furthermore, we control for the GDP per capita and Population of a country (Gleditsch, 2004). Finally, we code whether the particular dyad exhibited Sons of the Soil dynamics relying on the definition provided by Fearon (2004) and Fearon and Laitin (2010). Method To test our main hypotheses we estimate a series of semi-parametric hazards models in which the dependent variable measures the fighting durations of rebel organizations. The dyadic data set includes 290 rebel organizations in 198 conflicts between 1946 and 2005. The unit of analysis is a dyad-day. The data includes 637,585 dyad-days aggregated to 23 1941 spells. A change in any of the time varying independent variables requires a new spell. Out of the 290 rebel-government dyads 18 are right censored. The average fighting duration of a rebel organization is 2207 days (about 6 years) with the median duration being 758 days (about 2.1 years). Since our theoretical considerations do not predict any particular functional form of the underlying baseline hazard, we estimate Cox proportional hazards models. This leaves the duration dependency unspecified and focuses the empirical analysis on how the covariates shift the baseline hazard. We estimate all models using clustered standard errors to account for possible interdependence between dyads within one conflict. Preliminary analysis revealed that coups differ in their baseline hazard to other conflicts. Therefore, to allow for different underlying baseline hazards we stratify all models by coups. We tested all models and the included variables for a possible violation of the proportionality assumption. The tests suggest that the proportionality assumption is not violated in any of our models. Results. Cox proportional hazards models were estimated to assess the hypotheses. The estimation results can be found in Table 2. The coefficients show the impact of the explanatory variables on the underlying baseline hazards. Positive coefficients imply an increase in the hazard of a conflict dyad ending and thus shorter fighting durations. We begin with a baseline model that includes characteristics of the rebel organization, resources in the conflict area, and country level controls. Model 1 suggests that rebel organizations gaining territorial control are able to fight longer, while strong central 24 command structures and a legal political wing are associated with shorter conflict durations. Natural resources in the conflict area have a conflict-prolonging effect. Before turning to the main results, we shortly discuss the findings related to the country level controls. Democratic countries are prone to very long lasting rebel organizations, even though this result is possibly driven by India and Indonesia, while GDP per capita and the population size seem not to significantly impact conflict dyad duration. This insight holds across all estimated models. Turning to the main results, we include our first main independent variable, which indicates whether a rebel organization claims to fight on behalf of a political relevant ethnic group and recruits from their members. Therefore, Model 2 investigates whether ethnic linkages as such impact on conflict duration. Model 2 indicates that rebel organizations linked to at least one political relevant ethnic group fight significantly longer than rebel organization without such a link. The hazard of a conflict dyad ending decreases by 23 percent if a rebel organization is linked to a politically relevant ethnic group. In line with most of the literature this general ethnic effect is only marginally significant. However, our theoretical argument suggested that it is not ethnicity per se, but the state induced political exclusion of ethnic groups that leaves their members willing to engage in prolonged fighting (Hypothesis 3). 25 Table 2: Main Results: Cox proportional hazards estimates Model 1 Model 2 Model 3 Model 4 Model 5 VARIABLES Ethnic Linkage -0.249* (0.131) Ethnic Linkage with Excluded Group Ethnic Linkage with Included Group Territiorial Control Strong Central Command Legal Political Wing Democracy GDP per capita (log) Population (log) Natural Resources in Conflict Zone -0.313** (0.151) 0.344** (0.136) 0.364* (0.204) -0.830*** (0.193) 0.094 (0.076) -0.049 (0.049) -0.370** (0.163) -0.306** (0.155) 0.369*** (0.138) 0.366* (0.192) -0.816*** (0.193) 0.080 (0.078) -0.036 (0.049) -0.390** (0.159) -0.320** (0.155) -0.377*** (0.145) 0.272 (0.207) -0.317** (0.157) 0.417*** (0.143) 0.362** (0.179) -0.824*** (0.194) 0.089 (0.077) -0.037 (0.049) -0.364** (0.155) Sons of the Soil -0.364** (0.152) 0.285 (0.208) -0.319** (0.157) 0.408*** (0.147) 0.358** (0.179) -0.811*** (0.201) 0.094 (0.081) -0.037 (0.049) -0.360** (0.155) -0.085 (0.249) Ethnic Linkage X Territorial Control Dyads Failed Dyads Spells Days at Risk Log-likelihood Chi2 290 290 290 272 272 272 1941 1941 1941 637585 637585 637585 -1161 -1159 -1156 54.47 66.17 73.34 Clustered standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 -0.411* (0.226) 0.363*** (0.137) 0.388** (0.191) -0.802*** (0.192) 0.086 (0.077) -0.037 (0.049) -0.394** (0.161) 0.218 (0.279) 290 272 1941 637585 -1156 76.37 290 272 1941 637585 -1159 65.92 To test Hypothesis 3 we empirically differentiate between organizations that are linked to excluded and non-excluded political relevant ethnic groups. The baseline category contains rebel organizations that do not have an ethnic link. Model 3 suggests strong 26 support for our main hypothesis. While rebel organizations related to non-excluded ethnic groups are not associated with longer conflict durations, we find that rebel organizations that claim to fight on behalf of political excluded ethnic groups, and recruit from their members, fight significantly longer. Compared to the residual category, these rebel organizations shift the underlying baseline hazard by 32 percent downwards, thus significantly prolonging conflict dyad duration. The difference between rebel organizations with an excluded and non-excluded ethnic link is roughly 60 percent. Again this difference is highly statistically significant. Comparing our main explanatory variable to the natural resource dummy highlights another result; While most of the literature finds that resources and economic factors overtrump ethnic factors, our results suggest that both explanations are important and that their substantive effects are similar.3 Alternative Explanations. An important question is whether we are merely capturing effects that can be attributed to alternative explanations. Most importantly, our empirical explanations could be driven by what Fearon (2004) calls the “sons of the soils”. Model 4 suggests that sons of the soil dynamics (SoS) do not change our main results even though the standard errors are slightly larger. Moreover, in our models, the SoS variable itself does not significantly impact rebel organization duration. Given the theoretical definition of sons of the soil wars as conflicts involving “disadvantaged minorities” we interpret this finding as evidence that SoS wars are merely a subset of ethnic conflicts involving ethnic groups excluded form state power. Indeed a cross-tabulation reveals that all 37 dyads coded as SoS are linked to ethnic groups excluded from state power. 3 Preliminary results (not shown) suggest an interaction effect of natural resources and ethnic exclusion, where the existence of both factors is associated with even longer conflict. 27 We also assess whether our results are partly capturing ethnic defection. Kalyvas (2008) argues that ethnic defection becomes likely if the government has territorial control vis-àvis the rebel organization. Empirically, this would imply that ethnic defection is unlikely when the rebel organization have territorial control, thus prolonging conflict. Therefore, the effect of ethnicity should be stronger when rebels exercise territorial control, and weaker or absent in conflicts where the rebels so not control territory. To assess this implication we interacted Ethnic Linkage with Excluded Group with Territorial Control (Model 5). The results reveal no significant interaction effect suggesting that we are not merely tapping into ethnic defection. Robustness Checks We estimated a number of alternative models to assess the robustness of our results (Table 3). In Model 6 we include further geographic variables that often are associated with prolonged conflict. Our estimates demonstrate that including a term for whether the conflict area abuts a border and the distance to the capital do not change our main results. Additionally, we estimate a model controlling for whether the conflict takes place after the end of the Cold War (Model 7). While civil wars have become significantly shorter following the end of the Cold War, our main results stay robust to this model specification. 28 Table 3: Robustness Results: Cox proportional hazards estimates Model 6 Model 7 Model 8 -0.348** (0.142) 0.302 (0.210) -0.314* (0.167) 0.403*** (0.140) 0.363** (0.179) -0.804*** (0.191) 0.081 (0.076) -0.048 (0.057) -0.325* (0.167) -0.033 (0.061) -0.255 (0.196) -0.395*** (0.141) 0.062 (0.213) -0.270 (0.165) 0.375*** (0.144) 0.391** (0.173) -0.776*** (0.190) 0.059 (0.076) -0.056 (0.057) -0.372** (0.153) -0.026 (0.064) -0.217 (0.189) 0.564*** (0.146) -0.440*** (0.145) 0.262 (0.211) -0.311* (0.161) 0.429*** (0.139) 0.329* (0.184) -0.805*** (0.197) 0.063 (0.081) -0.041 (0.050) VARIABLES Ethnic Linkage with Excluded Group Ethnic Linkage with Included Group Territiorial Control Strong Central Command Legal Political Wing Democracy GDP per capita (log) Population (log) Natural Resources in Conflict Zone Conflict-Capital Distance (log) Conflict Zone at Border Post-Cold War Drugs in Conflict Zone Gemstones in Conflict Zone Hydrocarbon in Conflict Zone Dyads Failed Dyads Spells Days at Risk Log-likelihood Chi2 290 290 272 272 1941 1941 637585 637585 -1154 -1145 76.43 82.44 Clustered standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 -0.192 (0.235) -0.190 (0.184) -0.185 (0.165) 290 272 1941 637585 -1156 74.67 29 Model 8 investigates whether any particular resource is driving the effect of natural resources in the conflict area and whether the unpacking of natural resources has any effect on our main estimates. The results demonstrate that only the aggregate measure of natural resources has a significant effect and that our estimates stay robust. In addition, we checked whether our results are robust with respect to alternative estimation approaches. Therefore, we estimated several parametric hazards models (Weibull, Lognormal, and Loglogistic), all of which yielded similar and significant results. To assess whether our results are driven by particular conflicts or countries, we re-estimated Model 3 with group-wise jackknifing of our sample by (a) conflicts and (b) countries. Our estimated coefficients do not change and the standard errors stay small. Competing Risks. Finally, we analyze whether the effect of ethnic exclusion of rebel organization members is conditional on the conflict outcome. We therefore estimate five further models that account for different conflict outcomes: agreements, victories (by either party), government victories, rebel victories, and low activity. The results can be found in Table 4. Interestingly, ethnic exclusion is strongly associated with victories and low-activity, but has no effect on the likelihood of agreements. This suggests that ethnic exclusion is significantly related to longer conflicts if parties cannot agree to a negotiated settlement. This supports our theoretical conjecture that rebel organizations fighting on behalf of excluded ethnic groups are generally more willing to accept longer periods of fighting until a decisive outcome is reached. This, in turn, lends support to the proposition that prior exclusion creates a commitment problem on the side of the 30 government, since any agreements are not necessarily credible. In addition, rebel organizations that recruit from political non-excluded groups are associated with quick agreements. Again this supports our theoretical ideas that higher opportunity costs due to less strong grievances are associated with shorter duration. Table 4: Competing Risks Results: Cox proportional hazards estimates Agreement Victory Government Victory Rebel Victory Low activity 0.160 (0.295) 1.067** (0.501) 0.151 (0.317) 0.100 (0.323) -0.065 (0.534) -0.793 (0.539) 0.007 (0.218) -0.313** (0.137) -0.611 (0.384) -0.782*** (0.252) -0.095 (0.274) -0.293 (0.253) 0.990*** (0.197) 0.469** (0.216) -1.056*** (0.344) -0.072 (0.127) -0.101 (0.076) -0.369 (0.229) -0.602* (0.335) -0.026 (0.356) -0.005 (0.318) 0.586** (0.284) 0.512* (0.303) -0.816** (0.401) 0.085 (0.169) -0.024 (0.105) -0.332 (0.327) -1.186** (0.462) -0.134 (0.606) -0.573 (0.385) 1.474*** (0.341) 0.337 (0.347) -1.696** (0.699) -0.332* (0.182) -0.231* (0.123) -0.422 (0.406) -0.481* (0.284) -0.550 (0.646) -0.488 (0.408) -0.282 (0.417) 0.330 (0.319) -0.662* (0.346) 0.024 (0.163) 0.214* (0.113) 0.145 (0.397) 290 290 290 56 115 66 1941 1941 1941 637585 637585 637585 -250.5 -490.9 -282.3 -233.9 -460.5 -272.3 25.55 56.96 17.67 Clustered standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 290 48 1941 637585 -203.3 -174.8 51.68 290 68 1941 637585 -305.6 -297.0 17.17 VARIABLES Ethnic Linkage with Excluded Group Ethnic Linkage with Included Group Territiorial Control Strong Central Command Legal Political Wing Democracy GDP per capita (log) Population (log) Natural Resources in Conflict Zone Dyads Failed Dyads Spells Days at Risk Log-Likelihood (0) Log-Likelihood Chi2 31 Conclusion and Policy Implications In this paper we examined the role of ethnicity in the context of civil war duration. Focusing on actors and agency in civil wars, we presented a theory that links rebel organizations and ethnic groups and also includes the state as an active actor. In particular, we have argued that exclusionary policies enacted by the state are likely to influence the salience of ethno-nationalist grievances. More specifically, members of ethnic groups which are systematically excluded from state power are characterized by low opportunity costs and are therefore more likely to fight longer conflicts. These fighters have less to lose in war and are therefore more willing to accept uncertain payments related to future conflict outcomes. It is these grievances that allow rebel organizations to recruit and fight on behalf of such groups to maintain their fighting base for longer periods of time. Thus, contrary to what is assumed by many scholars, we find that ethnic conflicts last longer not because fixed individual loyalties are non-transendable and thus irresolvable. Rather, our results point directly to institutions and policies that make such conflicts less likely to begin with, but also help them come to an end, namely ethnic power-sharing through inclusion. Whereas the view that fixed and rigid ethnic identities are at the heart of such violence, grievances are by no means fixed. Indeed they are variable as a direct function of ethno-nationalist exclusion. Representation at the political center should 32 therefore be seen as a powerful tool to alleviate grievances, and thereby shorten armed conflicts. Theoretically, we believe that our approach is not only more complete, since it covers a more extensive set of actors, but also offers more causal depth since it goes beyond the current literature by explaining variation in the level of grievances as the result of state action. Our empirical results provide strong support for the argument that ethnicity only impacts on conflict duration if members of ethnic groups are excluded from political power. If rebel organizations are linked to these excluded ethnic groups, they fight significantly longer. Therefore, it is not ethnicity as such but ethno-nationalist policies that drive conflict duration. 33 Bibliography Anderson, Benedict. 1996. Imagined communities: Reflections on the origin and spread of nationalism. Revised ed. London: Verso. Axelrod, R. 1984. The Evolution of Cooperation. New York: Basic Books. Beardsley, Kyle, and Brian McQuinn. 2009. "Rebel Groups as Predatory Organizations: The Political Effects of the 2004 Tsunami in Indonesia and Sri Lanka." Journal of Conflict Resolution 53 (4):624-45. Brandt, Patrick T., T. David Mason, Mehmet Gurses, Nicolai Petrovsky, and Dagmar Radin. 2008. "When and How the Fighting Stops: Explaining the Duration and Outcome of Civil Wars." Defence and Peace Economics 19 (6):415-34. Brubaker, Rogers. 2004. Ethnicity without Groups. Cambridge: Harvard University Press. Buhaug, Halvard, Scott Gates, and Paivi Lujala. 2009. "Geography, Rebel Capability, and the Duration of Civil Conflict." Journal of Conflict Resolution 53 (4):544-69. Cederman, Lars-Erik, Andreas Wimmer, and Brian Min. 2010. "Why Do Ethnic Groups Rebel? New Data and Analysis." World Politics 62 (1):87-119. Collier, P. 2000. "Doing Well Out of War." In Greed and Grievance: Economic Agendas in Civil Wars, ed. M. Berdal and D. M. Malone. Boulder: Lynne Rienner. Collier, Paul, and Anke Hoeffler. 2004. "Greed and Grievance in Civil Wars." Oxford Economic Papers 56:663-95. Collier, Paul, Anke Hoeffler, and Måns Söderbom. 2004. "On the Duration of Civil War." Journal of Peace Research 41 (3):253-73. Cunningham, David E. 2006. "Veto Players and Civil War Duration." American Journal of Political Science 50 (4):875-92. 34 Cunningham, David E., Kristian Skrede Gleditsch, and Idean Salehyan. 2009. "It Takes Two: A Dyadic Analysis of Civil War Duration and Outcome." Journal of Conflict Resolution 53 (4):570-97. Fearon, James D. 2004. "Why Do Some Civil Wars Last So Much Longer Than Others?" Journal of Peace Research 41 (3):275-301. Fearon, James D., and David D. Laitin. 2003. "Ethnicity, Insurgency, and Civil War." American Political Science Review 97 (1):75-89. Horowitz, Donald L. 1985. Ethnic Groups in Conflict. Berkeley: University of California Press. Kalyvas, Stathis N. 2006. The Logic of Violence in Civil War. Cambridge: Cambridge University Press. ———. 2007. "Civil Wars." In The Oxford Handbook of Comparative Politics, ed. C. Boix and S. C. Stokes. Oxford: Oxford University Press. ———. 2008. "Ethnic Defection in Civil War." Comparative Political Studies 41 (8):1043-68. Kaufman, Stuart J. 1996. "Spiraling to Ethnic War: Elites, Masses, and Moscow on Moldova's Civil War." International Security 21 (2):108-38. ———. 2001. Modern Hatreds: The Symbolic Politics of Ethnic War. Ithaca: Cornell University Press. ———. 2006. "Escaping the Symbolic Politics Trap: Reconciliation Initiatives and Conflict Resolution in Ethnic Wars." Journal of Peace Research 43 (2):201-18. ———. 2006. "Symbolic Politics or Rational Choice? Testing Theories of Extreme Violence." International Security 30 (4):45-86. 35 Kaufmann, Chaim. 1996. "Possible and Impossible Solutions to Ethnic Civil Wars." International Security 20 (4):136-75. ———. 2005. "Rational choice and progress in the study of ethnic conflict: A review essay." Security Studies 14 (1):178-207. Lichbach, Mark I. 1998. The rebel's dilemma. Ann Arbor: University of Michigan Press. Metternich, Nils, and Julian Wucherpfennig. 2010. "Rebel Organization Interdependence and Civil War Duration." Paper prepared for presentation at the International Studies Association Annual Convention, New Orleans, 17-20 February. Montalvo, José G, and Marta Reynal-Querol. 2010. "Ethnic polarization and the duration of civil wars." Economics of Governance 11:11:123–43. Mueller, John. 2003. "Policing the remnants of war." Journal of Peace Research 40 (5):507. ———. 2004. The Remnants of War. Ithaca: Cornell University Press. Posen, Barry R. 1993. "The Security Dilemma and Ethnic Conflict." Survival 35 (1):2747. Rose, William. 2000. "The security dilemma and ethnic conflict: Some new hypotheses." Security Studies 9 (4):1-51. Sambanis, Nicholas. 2004. "What Is Civil War?: Conceptual and Empirical Complexities of an Operational Definition." Journal of Conflict Resolution 48 (6):814-58. Sinno, Abadulkader H. 2008. Organizations at War in Afghanistan and beyond. Ithaca: Cornell University Press. Toft, Monica Duffy. 2003. The Geography of Ethnic Violence. Princeton: Princeton University Press. 36 Van Evera, Stephen. 1994. "Hypotheses on nationalism and war." International Security 18 (4):5-39. Varshney, Ashutosh. 2004. "Nationalism, ethnic conflict, and rationality." Perspectives on Politics 1 (01):85-99. Weiner, Myron. 1978. Sons of the Soil. Princeton: Princeton University Press. Weinstein, Jeremy M. 2007. Inside rebellion: The politics of insurgent violence. Cambridge: Cambridge University Press. 37
© Copyright 2026 Paperzz