One of the most influential accounts of civil war onset proposes that the exclusion of ethnic groups from state power severely increases the risk of conflict (see Cederman and Vogt Reference Cederman and Vogt2017; United Nations/World Bank 2018, chap. 4). The theory outlines a causal chain in which political exclusion leads to grievances, understood as collective perceptions of injustice, which drive group mobilization. Mobilized groups make claims against the state, but when these are ignored, rejected, or met with repression, tensions escalate, increasing the likelihood of civil war (Cederman, Gleditsch, and Buhaug Reference Cederman, Gleditsch and Buhaug2013, chap. 3). Several quantitative studies have found strong empirical support for this idea, showing a robust association between ethnopolitical exclusion—from which grievances are likely to emerge—and conflict onset (e.g., Buhaug, Cederman, and Gleditsch Reference Buhaug, Cederman and Gleditsch2014; Cederman, Gleditsch, and Buhaug Reference Cederman, Gleditsch and Buhaug2013; Cederman, Weidmann, and Gleditsch Reference Cederman, Weidmann and Gleditsch2011; Wimmer, Cederman, and Min Reference Wimmer, Cederman and Min2009).
At this stage, it is important to go beyond average statistical associations and examine whether and to what extent this relationship occurs according to the proposed theoretical linkage. To stop still-prevailing ethnic conflicts from escalating, it is not enough to know that political exclusion is associated with a higher risk of civil war; we also need to know how exclusion causes civil war. Despite strong correlational evidence, direct assessments of how grievances over political exclusion lead to civil war remain underdeveloped. Large-N statistical studies cannot directly trace whether and how civil wars occur through the theorized sequence of grievance formation, mobilization, and state responses. Moreover, key challenges for large-N studies include reverse causality—political exclusion may result from conflict, rather than cause it—and confounding factors that could drive the observed correlations (Hillesund et al. Reference Hillesund, Bahgat, Barrett, Dupuy, Gates, Nygård, Rustad, Strand, Urdal and Østby2018, 474–76). This means there is a risk that we either would overestimate or underestimate its explanatory power. Measurement validity also remains a concern. Because collective grievances are difficult to quantify, researchers often rely on proxy indicators instead of directly observing how excluded groups mobilize and escalate conflicts (Cederman, Gleditsch, and Buhaug Reference Cederman, Gleditsch and Buhaug2013, 30; Hillesund et al. Reference Hillesund, Bahgat, Barrett, Dupuy, Gates, Nygård, Rustad, Strand, Urdal and Østby2018, 472–74).
Single-case studies have provided valuable insights into these processes (e.g., Aboseadra, Fakih, and Haimoun 2021; Caumartin, Molina and Thorp Reference Caumartin, Molina and Thorp2008; Langer Reference Langer2008; Stewart Reference Stewart2008; Tezcür and Gurses Reference Tezcür and Gurses2017; Ukiwo Reference Ukiwo and Steward2008). However, although they may be rich in context and detail, these studies tend to be narratively oriented and do not directly trace the theory’s empirically observable implications. Even when they provide direct evidence of mechanisms within cases, they do not establish whether those mechanisms operate across a broader population, limiting causal generalization.
Against this backdrop, this article offers detailed evidence about how well the proposed theoretical dynamics—including the formation of grievances, collective mobilization, and state responses—match the empirical reality in a broader set of cases. To address the described methodological challenges, we apply Large-N Qualitative Analysis (LNQA), a method that integrates within- and cross-case inferences by studying specified causal mechanisms across multiple cases (Goertz and Haggard Reference Goertz and Haggard2023; see also Bartusevičius Reference Bartusevičius2019; Ross Reference Ross2004). LNQA offers two key advantages: (1) it allows for the direct tracing of causal mechanisms, rather than inferring them from statistical associations, and (2) it ensures comprehensive case coverage, mitigating the selection bias affecting small-N studies (Goertz and Haggard Reference Goertz and Haggard2023, 1221). Moreover, LNQA also contributes to theory-building by systematically identifying cross-case variation in causal mechanisms, which can refine grievance-based theories by specifying conditions under which exclusion triggers conflict.
Our study examines 15 cases of civil war in which ethnic groups excluded from state power were represented by rebel organizations involved in major civil war outbreaks after the end of the Cold War.Footnote 1 Although Cederman, Gleditsch, and Buhaug (Reference Cederman, Gleditsch and Buhaug2013) examine civil wars globally since 1945, we focus on the post–Cold War period to ensure a feasible number of detailed case studies. This choice also offers the advantage of greater comparability, because conflicts before and after 1991 likely differ, especially in their international context. By applying LNQA to all cases within our scope—post–Cold War major civil war onsets—we can systematically assess whether the theorized mechanisms hold across this full sample of contemporary cases.
The analysis yields two main findings. First, in nine of the analyzed cases, the evidence shows that the onset of civil war followed directly from a mobilization process in which members of excluded ethnic groups sought to redress their grievances about political exclusion. For these theory-corroborating cases, we identify two pathways that differ in how governments respond to excluded groups’ grievance-based mobilization. One pathway is characterized by no substantial attempts of accommodation accompanied by the indiscriminate use of state repression (as opposed to selective targeting of actual dissidents) that escalates conflicts by increasing support for armed resistance tactics. The other pathway is characterized by inconsistent government reactions (involving both accommodative and non-accommodative policies), which increase the groups’ mobilization potential without reducing their political demands (see also Lindemann and Wimmer Reference Lindemann and Wimmer2018). Both counterproductive strategies tend to be adopted by states without sufficient coercive capacity to selectively target dissidents or to repress or accommodate their challengers consistently.
The second main finding is that we see signs of reverse causality in the remaining six cases, where the relevant ethnic groups suffered from disrupted public order and tended to be excluded from state power because of armed conflict. In these cases, rather than political exclusion leading to war, preexisting armed conflict itself generated the exclusion that later became a mobilizing grievance.
Our study offers four key contributions. First, our within-case analysis of major civil war onsets in the post–Cold War era provides direct empirical support for grievance-based explanations of civil conflict. By examining the proposed mechanism across multiple cases, we reaffirm that grievances stemming from political exclusion were key drivers of armed rebellion in the majority of analyzed cases.
Second, our study refines grievance theory by showing that how states respond to excluded groups’ mobilization shapes the pathway to armed rebellion (see Bormann, Cederman, and Vogt Reference Bormann, Cederman and Vogt2017; Cederman, Gleditsch, and Buhaug Reference Cederman, Gleditsch and Buhaug2013, 48–51; Lindemann and Wimmer Reference Lindemann and Wimmer2018). In the cases we analyze, governments responded either by rejecting demands and using indiscriminate repression or by combining accommodation with rejection, which fueled escalation rather than containing dissent. This aligns with broader findings that states with low coercive capacity often resort to inefficient strategies (see Goodwin Reference Goodwin2001; Lichbach Reference Lichbach1987; Mason and Krane Reference Mason and Krane1989; Roessler Reference Roessler2011; Winward Reference Winward2021), suggesting that grievance-based civil war onsets are more likely when the state security apparatus has limited strength and skills to repress (see Fjelde and de Soysa Reference Fjelde and de Soysa2009).Footnote 2
Third, our identification of several instances of reverse causality suggests an additional refinement to grievance theory. Specifically, the assumption that group mobilization always begins with ethnopolitical exclusion should be relaxed. This adjustment allows for an explanation of how grievances over exclusion can shape conflict—not necessarily as the root cause, but as a driver of sustained or intensified fighting.
Fourth, we contribute to methodological debates by showing how within-case causal evidence from a relatively high number of cases can address concerns about endogeneity and measurement validity in large-N research, as well as issues related to causal generalization in small-N research (see Goertz and Haggard Reference Goertz and Haggard2023). This methodological approach allows us to bridge the gap between broad empirical patterns and the nuanced case-specific processes that drive civil conflict.
This article proceeds as follows. The next section reviews the literature on ethnopolitical exclusion and civil conflict. We then outline the theoretical framework and research design, detailing our LNQA approach. The empirical section presents our findings, followed by a discussion of their implications and avenues for further research.
What Do We Know about Ethnopolitical Exclusion, Grievances, and Civil Conflict?
Several quantitative studies have found robust empirical support for the proposition that the exclusion of ethnic groups from state power severely increases the risk of conflict (see Hillesund et al. Reference Hillesund, Bahgat, Barrett, Dupuy, Gates, Nygård, Rustad, Strand, Urdal and Østby2018 for a comprehensive review). Early empirical studies relied on the Minorities at Risk (MAR) dataset, which measured political discrimination and group differentials (Gurr Reference Gurr1993a, Reference Gurr1993b). However, these studies faced criticisms of selection bias and coder subjectivity (Regan and Norton Reference Regan and Norton2005; Vadlamannati Reference Vadlamannati2011). In response, the Ethnic Power Relations (EPR) dataset was developed to more systematically capture ethnic groups’ access to executive power. Analyses using EPR—both at the group and country level—have consistently found that political exclusion and recent losses of executive power increase the risk of armed conflict (Buhaug, Cederman, and Gleditsch Reference Buhaug, Cederman and Gleditsch2014; Cederman, Gleditsch, and Buhaug Reference Cederman, Gleditsch and Buhaug2013; Cederman, Weidmann, and Bormann Reference Cederman, Weidmann and Bormann2015; Cederman, Weidmann, and Gleditsch Reference Cederman, Weidmann and Gleditsch2011; Cederman, Wimmer, and Min Reference Cederman, Wimmer and Min2010; Wimmer, Cederman, and Min Reference Wimmer, Cederman and Min2009).
An important question is why political exclusion and grievances lead to large-scale violent mobilization when they could also result in nonviolent forms of mobilization, such as demonstrations or strikes (see, e.g., Chenoweth and Stephan Reference Chenoweth and Stephan2011; Pischedda Reference Pischedda2020; Thurber Reference Thurber2021; Vogt Reference Vogt2019) that could be accommodated (see Cederman, Gleditsch, and Buhaug Reference Cederman, Gleditsch and Buhaug2013, 49). Hillesund and Østby (Reference Hillesund and Østby2022) review whether horizontal inequalities drive all forms of mobilization or specifically violent conflict. Although they find the link to violent conflict to be robust, evidence on nonviolent mobilization is mixed. Examples of studies include Dahl et al. (Reference Dahl, Gates, Gonzalez and Gleditsch2021), who show that mobilization capacity plays a crucial role in shaping tactical choices, with nonviolent resistance being more viable when groups can mobilize large numbers. Similarly, Vogt (Reference Vogt2019) explains why ethnic mobilization turns violent or remains nonviolent by pointing to colonial legacies and historical state structures, which shape political exclusion and group capacity for rebellion.
These questions also speak to a broader discussion on how conflicts are particularly likely to escalate when governments attempt to demobilize excluded groups by means of violence (Cederman, Gleditsch, and Buhaug Reference Cederman, Gleditsch and Buhaug2013, 50; see Sutton, Butcher, and Svensson Reference Sutton, Butcher and Svensson2014). For instance, Rørbæk (Reference Rørbæk2019) finds that politically excluded groups facing repression are more likely to resort to violence, rather than nonviolent protest, because repression reduces the viability of civil resistance (see also Lindemann and Wimmer Reference Lindemann and Wimmer2018).
Overall, quantitative studies have been crucial in establishing a robust empirical connection between ethnopolitical exclusion and civil war, and this theoretical perspective is considered among the most prominent accounts for civil conflict (Cederman and Vogt Reference Cederman and Vogt2017). However, important methodological caveats remain (Hillesund et al. Reference Hillesund, Bahgat, Barrett, Dupuy, Gates, Nygård, Rustad, Strand, Urdal and Østby2018, 472–76). First, because grievances are difficult to measure across many cases, quantitative studies take an indirect approach, showing that ethnic exclusion—a condition likely to generate grievances—is statistically associated with armed conflict (Cederman, Gleditsch, and Buhaug Reference Cederman, Gleditsch and Buhaug2013, 30). This reliance on indirect measures raises concerns about measurement validity, because researchers often use proxy indicators rather than directly capturing how excluded groups mobilize and escalate conflicts (see also Hillesund et al. Reference Hillesund, Bahgat, Barrett, Dupuy, Gates, Nygård, Rustad, Strand, Urdal and Østby2018, 472–74). Second, large-N studies do not conduct direct assessments of whether civil wars occur through the theorized process of grievance formation, mobilization, and state responses.
Third, endogeneity poses a key challenge. As with all cross-national statistical studies that rely on selection on observables, unmeasured confounders may drive the observed correlations. Moreover, reverse causality is a relevant issue: political exclusion may result from conflict, rather than cause it (Hillesund et al. Reference Hillesund, Bahgat, Barrett, Dupuy, Gates, Nygård, Rustad, Strand, Urdal and Østby2018, 474–76). Thus, reverse causality could put some basic assumptions of the standard grievance-based explanation into question. It is not far-fetched to suggest that governments account for whether specific groups are prone to conflict when they determine their exclusionary policies. One possibility is that rulers tend to include groups politically that they anticipate would pose a threat to the political survival of the regime if excluded. This tendency would lead quantitative studies that do not properly account for endogeneity to underestimate the average causal effect of ethnopolitical exclusion on civil war onset. However, the opposite scenario is also plausible. As shown by Roessler (Reference Roessler2011, 303), rulers in sub-Saharan Africa are more likely to exclude ethnic groups that represent a credible threat, because they calculate that a “distant threat from society is preferable to an existential one from inside their regime.” If this finding is valid and applies to other regions, previous studies may have overestimated the average effect of ethnopolitical exclusion (see also Fearon Reference Fearon2010). This line of reasoning is not dissimilar to that of Cederman, Gleditsch, and Buhaug (Reference Cederman, Gleditsch and Buhaug2013, 52), who are aware of the risk of reverse causality. However, their attempts to deal with the issue are limited.Footnote 3
Grievance explanations have also been examined in single-case (or small-N) studies. For example, Tezcür and Gurses (Reference Tezcür and Gurses2017) show that political inclusion can reduce support for ethnic insurgency in Turkey. Likewise, Abosedra, Fakih, and Haimoun (Reference Abosedra, Fakih and Haimoun2021) find that a lack of power sharing was a significant factor in the onset of the Syrian civil war. Additional studies of Cote d’Ivoire and Ghana (Langer Reference Langer2008), Nigeria (Ukiwo Reference Ukiwo and Steward2008), and Bolivia, Guatemala, and Peru (Caumartin, Molina, and Thorp Reference Caumartin, Molina and Thorp2008) illustrate how group disparities generate grievances that have contributed to violent conflict. These case studies collectively reinforce the argument that grievances stemming from horizontal inequalities play a crucial role in conflict dynamics.Footnote 4 By focusing on within-case and small-N comparisons, these studies illustrate key mechanisms that drive mobilization and violence. In addition, the microlevel implications of grievance-based explanations have also been tested using survey data focusing on specific conflicts (Hillesund Reference Hillesund2015) and regions (Hillesund Reference Hillesund2022). Although small-N studies provide clearer insight into the mechanisms linking grievances and civil war, they also have important limitations. For causal generalization, selection bias is a key concern, because a single case may not be representative. Strictly speaking, demonstrating the mechanism in one case is insufficient; it must be shown across all cases within the scope conditions.
Some perspectives have chosen a methodological middle ground, focusing to a larger extent on multiple cases but without reducing them to statistical correlations. Based on the assumption that ethnic mobilization is caused by grievance, Lindemann and Wimmer (Reference Lindemann and Wimmer2018) use qualitative comparative analysis (QCA) in 58 cases to show that indiscriminate state violence is an important condition for conflict escalation. Lieberman and Singh (Reference Lieberman and Singh2012) argue that state-driven ethnic categorization fosters competition and mobilization, making identity a salient political cleavage and laying the groundwork for both exclusion and violence. This perspective complements grievance-based theories of conflict by showing how ethnic divisions emerge even before exclusion occurs. They use a comparative-historical analysis of 11 southern African cases, providing a detailed examination of all countries in the region from 1945 (or independence). As such, their empirical approach—an in-depth examination of multiple cases within a theoretically defined scope (here, proxied by region)—is very similar to LNQA.
There are also broader conceptual and theoretical critiques of the ethnic grievance-based accounts. Rival theoretical perspectives emphasize opportunities for conflict (e.g., Collier and Hoeffler Reference Collier and Hoeffler2004) or at least a combination of perspectives on motives (grievances) and opportunity structures (see, e.g., Cunningham et al. Reference Cunningham, Gleditsch, Gonzalez, Vidovic and White2017).Footnote 5 Moreover, some studies challenge conventional understandings of the ethnic component in civil wars. A number of studies have demonstrated that civil wars often labeled as “ethnic” can involve organizations or factions turning against their own ethnic group (Christia Reference Christia2008; Kalyvas Reference Kalyvas2008; Staniland Reference Staniland2012); others suggest that interethnic violence in civil wars often arises from local dynamics and strategic incentives, rather than from preexisting, fixed divisions (Kalyvas Reference Kalyvas2003; Reference Kalyvas2006). Ethnic identities can be activated, instrumentalized, or intensified during conflict, with their salience emerging or deepening as a result of war. These perspectives help highlight how ethnopolitical exclusion-related grievances may not always be central to the logic of a civil war.
Recent studies offer nuanced perspectives on the causal sequence between political exclusion and conflict. Basedau et al. (Reference Basedau, Fox, Pierskalla, Strüver and Vüllers2017), for instance, find that although discrimination against religious minorities generates grievances, these do not automatically translate into violence, indicating that additional factors such as organizational capacity and opportunity structures are crucial in the escalation process. Similarly, Vogt, Gleditsch, and Cederman (Reference Vogt, Gleditsch and Cederman2021) emphasize the role of internal dynamics within ethnic movements, showing that fragmentation can lead to radicalized demands as organizations vie for dominance. These radical demands, in turn, increase the likelihood of conflict escalation by provoking governmental repression. Thus, the pathway from ethnic exclusion to civil war might be mediated by a complex interplay of grievances, organizational strategies, and state responses, warranting a more comprehensive analytical approach to examining the causes of civil conflict.
Taken together, there is relatively strong empirical evidence linking political exclusion to civil war. However, conceptual and methodological challenges remain, and the perspective is not without its critics. To further scrutinize grievance theory, we conduct a direct empirical examination across all theoretically relevant cases using LNQA, which mitigates several limitations of both quantitative and small-N approaches by disentangling the sequences and key factors leading from political exclusion to civil war onset in multiple cases. In the following section, we outline the main tenets of grievance theory and derive observable implications.
Deriving Observable Implications from Grievance Theory
How does ethnopolitical exclusion stimulate civil war onset? Quantitative studies agree that grievances about political exclusion—that is, collective perceptions of injustice about being denied access to state power—trigger violent collective action through a process of group mobilization (Cederman, Weidmann, and Gleditsch Reference Cederman, Weidmann and Gleditsch2011; see also Buhaug, Cederman, and Gleditsch Reference Buhaug, Cederman and Gleditsch2014; Gurr Reference Gurr2000). According to Cederman, Wimmer, and Min (Reference Cederman, Wimmer and Min2010, 95), these grievances make excluded ethnic groups “fertile breeding grounds for organizations that challenge the government.” In related work, they argue that struggles between ethnic groups over the state “may lead to a process of political mobilization, counter-mobilization, and escalation,” but they add that their “theory does not explicitly address the logic of this escalation process” (Wimmer, Cederman, and Min Reference Wimmer, Cederman and Min2009, 321). Similarly, Bormann, Cederman, and Vogt (Reference Bormann, Cederman and Vogt2017) consider ethnic civil war to be the outcome of a process involving collective grievances, mobilization for collective action, and government reactions.
The most comprehensive attempt in quantitative research to theorize a grievance-based relationship was undertaken by Cederman, Gleditsch, and Buhaug (Reference Cederman, Gleditsch and Buhaug2013, chap. 3), who outline a detailed causal chain. They argue that political (and/or socioeconomic) exclusion fosters group grievances through a process of group identification, intergroup comparison, perceptions of injustice, and the framing and attribution of blame to the state. These grievances fuel mobilization, which may escalate into civil war if governments are unable or unwilling to accommodate the demands of excluded groups by failing to introduce substantial reforms, neglecting efforts to address discriminatory practices, or responding with violent repression. Figure 1 presents a stylized summary of this argument. For the sake of simplicity and comprehension, it does not capture all the nuances of the original argument.
Overview of the Theoretical Link between Ethnopolitical Exclusion and Civil War
The argument is formulated in probabilistic terms, meaning that not all exclusion leads to grievance-based mobilization and not all mobilization triggers civil war. At the same time, Cederman, Gleditsch, and Buhaug (Reference Cederman, Gleditsch and Buhaug2013, 37) hold that the links of their causal chain can be seen “as necessary conditions rather than as contributing causes.” Accordingly, if excluded ethnic groups engage in armed rebellion because of grievances about political exclusion, we should be able to observe that (a) group members mobilize against the government to redress their political grievances and (b) governments reject the groups’ political demands rather than accommodate them before the outbreak of rebellion. If either of these necessary conditions cannot be observed, this would strongly indicate that ethnopolitical exclusion did not cause civil war onset via grievances in that case. This is an important point of departure for our investigation.
Grievance-based Mobilization
The next step is to derive observable implications of the two necessary conditions: grievance-based mobilization and the government’s rejection of demands. Ethnic groups are considered excluded from state power when they have no de facto influence over the most important executive-level institutions of a country, typically the presidency, cabinet, and senior posts in the administration and the military (Wimmer, Cederman, and Min Reference Wimmer, Cederman and Min2009, 326). For political exclusion to morph into grievances and, ultimately, collective action, members of a self-conscious group need to identify their group’s inferior political status in comparison with other groups. At the same time, they must evaluate this situation as unjust and reach the understanding that representatives of the state are responsible for this perceived injustice (Cederman, Gleditsch, and Buhaug Reference Cederman, Gleditsch and Buhaug2013, 37–44). As argued by Cederman, Wimmer, and Min (Reference Cederman, Wimmer and Min2010, 94–95), the motive for redressing grievances about political exclusion is simultaneously material, political, and symbolic: “‘Adequate’ or ‘just’ representation in a central government offers material advantages such as access to government jobs and services, legal advantages such as the benefits of full citizenship rights, a fair trial, and protection from arbitrary violence, and symbolic advantages such as the prestige of belonging to a ‘state-owning’ ethnic group.”
Status reversal is a particularly strong subcase of political exclusion, because it provokes heightened discontent and conflict risk. Cederman, Gleditsch, and Buhaug (Reference Cederman, Gleditsch and Buhaug2013, 41, 62) argue that groups experiencing a downgrade in power status (especially those that were previously dominant) react strongly due to a sense of loss and entitlement, which often leads to violence.
To understand when exclusion escalates into civil war, it is essential to consider how elites actively shape and amplify grievances, given that exclusion alone rarely provokes civil war. As emphasized by Kaufman (Reference Kaufman1996), conflict ignites when belligerent elites engage in ethnic outbidding, deliberately magnifying grievances and recasting moderate demands as existential threats. By stoking fear of “group extinction” and competing to voice ever more extreme claims, ethnic entrepreneurs often supply the emotional heat and organizational drive that turn latent discontent into collective mobilization and a self-reinforcing security dilemma.
How, then, do we observe grievance-based mobilization? A significant share of group members must identify with their group, find the group to be unjustly disadvantaged, and explicitly and collectively blame the government. Members of excluded ethnic groups may engage in protests, demonstrations, boycotts, and strikes, as well as other forms of antigovernment collective action (Cederman, Gleditsch, and Buhaug Reference Cederman, Gleditsch and Buhaug2013, 44–48). Although the expression of grievances might take the form of mass mobilization, it does not need to be the case. What is crucial is that claims focusing on redressing grievances about political exclusion are spread below the elite level and are expressed in a collective fashion. Against this backdrop, we can formulate the first observable implication of the proposed mechanism:
Observable implication of grievance-based mobilization: Members of excluded ethnic groups are involved in antigovernment collective action and claims-making directed at representatives of the state with the intention to redress grievances with explicit reference to political exclusion.
More concretely, we should expect to observe demands for political reforms, regime change, or secession on behalf of the excluded ethnic group. These demands can be articulated by crowds of protesters or by group leaders at meetings, on radio, or in newspapers, pamphlets, manifestos, and petitions. Likely articulated grievances include limited political representation, lack of self-determination/autonomy, unjust citizenship requirements, and unequal transfers of public goods, including government jobs. Overall, mass mobilization serves as a strong indicator for grievance diffusion, but it is not necessary to fulfill the criterion. Mobilization occurring on a smaller scale or through specific factions within a group also counts if explicit reference is made to grievances about political exclusion.
It follows that the absence of grievance-based mobilization before the outbreak of rebellion would not be in line with the grievance-based explanation. Observing that perceptions of unfair political status do not resonate broadly among group members, or that mobilization efforts are instead motivated by other factors, provides evidence against the theory. For example, absolute poverty, economic crisis, environmental exploitation, or natural resource abundance may offer alternative explanations for why excluded ethnic groups take up arms (e.g., Collier and Hoeffler Reference Collier and Hoeffler2004; Fearon and Laitin Reference Fearon and Laitin2003; Ross Reference Ross2004).
Government Rejection of Demands
Even in the presence of substantial grievance-based mobilization, we would only expect escalation into civil war when the government decides not to accommodate the challengers’ demands (Bormann, Cederman, and Vogt Reference Bormann, Cederman and Vogt2017, 749). We understand accommodation as the extent to which a state meets the political, economic, or territorial demands of a group. Rejection, by contrast, occurs when the government does not seriously attempt to alter policies in the direction requested by aggrieved groups, instead relying on vague or insincere offers alongside targeted or indiscriminate repression. Excluded groups are unlikely to take up arms if they manage to attain some access to the state or if governments manage to convince them that access to the state can be obtained over time without engaging in costly warfare (Cederman, Gleditsch, and Buhaug Reference Cederman, Gleditsch and Buhaug2013, 48–51). We can now derive the second observable implication for the proposed mechanism:
Observable implication of rejection: The government and agents of the state do not consistently accommodate the demands stated by excluded ethnic groups but continue discriminatory policies and practices and engage in repression.
Observing that excluded groups are brought to the negotiation table and are offered concrete and significant changes in the form of increased policy influence, state spoils, or larger measures of autonomy would indicate that their political grievances are being accommodated. Such strategies can take the form of granting autonomy, power-sharing arrangements, economic concessions, and the like. As long as they are consistent, sincere, and introduce significant policy changes, they are considered to be instances of accommodation.
By contrast, governments that consistently reject excluded groups’ demands for policy change consequently block peaceful avenues for negotiation and reforms and thus increase the risk of conflict escalation. We observe this lack of accommodation when, for instance, government affiliates express their condemnation of demands for larger measures of ethnic autonomy or when governments use violent repression to put an end to mobilization. A particularly strong indicator of rejection is indiscriminate repression, which includes pogroms, mass killings, and campaigns of state terror and serious harassment where the target is the excluded ethnic group more generally and not only those group members engaged in dissident activities.
As indicated earlier, rejection of the excluded group’s demands is a necessary condition for the grievance-based explanation, because such a group that is accommodated but nonetheless engages in armed rebellion must have more important motivations than grievances about political exclusion. That is, observations that groups engage in armed rebellion despite being offered significant policy concessions in the prelude to civil war serves as evidence against the proposed mechanism. In practice, however, it is likely to be difficult to assess whether such government offers are sincere, and unless the excluded groups gain representation in a de facto power-sharing government before warfare first occurs, we are unlikely to disconfirm the grievance-based explanation based on observations of accommodation.
Note that there may also be an alternative pathway. Although complete and consistent rejection of demands tends to spur civil war onset, mixed responses that combine accommodation with repression may likewise trigger escalation. Such situations, where the state sends contradictory signals rather than pursuing a clear strategy, can both enable and motivate excluded ethnic groups to engage in armed rebellion.
Research Design
We apply LNQA to systematically evaluate whether the theorized causal processes—exclusion, mobilization, and rejection—consistently operate within a defined population of relevant cases. As Goertz and Haggard (Reference Goertz and Haggard2023) argue, multi-case qualitative designs are particularly well suited for rare events with complex strategic interactions, such as civil war onsets, in which causal mechanisms unfold sequentially over time (see Ross Reference Ross2004 and Bartusevičius Reference Bartusevičius2019 for similar approaches to study civil wars).
The LNQA approach mitigates typical concerns that characterize cross-national statistical studies (endogeneity, measurement validity, and lack of attention to mechanisms) and single-case studies (generalization). Regarding endogeneity, we use LNQA to examine the sequence of events within cases, clarifying whether exclusion drives conflict or vice versa. Moreover, we can minimize omitted variable bias by directly examining the causal mechanisms across all relevant cases. In addition, we can circumvent measurement validity problems by providing a direct assessment of grievances and mobilization, variables that are difficult to quantify in large-N datasets.Footnote 6 If the within-case evidence supports the mechanism, it strengthens confidence that the statistical associations previously established are causal. Conversely, if the proposed mechanism fails to appear in cases where it should, this challenges the interpretation of statistical findings (Goertz and Haggard Reference Goertz and Haggard2023, 1234).
Moreover, the LNQA approach ensures that causal generalizations do not rest on isolated instances or case selection bias, as may be the case in single-case studies, but instead emerge from systematically verifying whether the theorized mechanisms operate consistently across all relevant cases within our scope. More specifically, the population of relevant cases consists of ethnic groups excluded from state power that were represented by a rebel organization in a civil war onset. These “positive cases” are precisely where we should expect to observe the theorized mechanisms if the grievance model is correct (Beach and Pedersen Reference Beach and Pedersen2016; Gerring Reference Gerring2007, chaps. 4 and 5; Goertz and Mahoney Reference Goertz, Mahoney, Byrne and Ragin2009). Although this approach selects on the dependent variable, it does not pose an inferential problem, because our primary goal is to evaluate causal processes within cases, rather than to estimate cross-case correlations (Goertz Reference Goertz2017).
We identify the population of cases based on the Ethnic Power Relation dataset (group-level EPR-ETH v.2.0 and a recent update with coverage until 2021). This dataset codes all politically relevant ethnic groups and their access to state power (executive-level institutions) in countries with a population greater than 500,000 for the period 1946–2021. Ethnic groups are defined broadly as linguistic, racial, religious, regional, or other social groups that have a shared culture and believe in common ancestry (see Wimmer, Cederman, and Min Reference Wimmer, Cederman and Min2009).
Ethnic groups are considered represented in a civil war if a rebel organization fighting in that war “expresses an aim to support the ethnic group and members of the group in question participate in combat” (Cederman, Gleditsch, and Buhaug Reference Cederman, Gleditsch and Buhaug2013, 69).Footnote 7 Civil war onset is commonly defined as the first year in an internal armed conflict that produces as least 1,000 battle-related deaths (e.g., Collier and Hoeffler Reference Collier and Hoeffler2004; Fearon and Laitin Reference Fearon and Laitin2003). Using the UCDP/PRIO Armed Conflict Dataset (Gleditsch et al. Reference Gleditsch, Wallensteen, Eriksson, Sollenberg and Strand2002), Cederman, Gleditsch, and Buhaug (Reference Cederman, Gleditsch and Buhaug2013, chap. 4) show that the effect of ethnopolitical exclusion on civil war onset is substantively similar when operationalizing civil war as either low-intensity armed conflict (25 annual battle deaths) or high-intensity armed conflict (25 annual battle deaths and a minimum of 1,000 casualties during the overall course of the conflict). We select cases based on this latter operationalization to meet the standard definition of civil war in the literature (see Ross Reference Ross2004; Sambanis Reference Sambanis2004) and because, from a pragmatic point of view, the number of cases for our qualitative analysis becomes much more manageable. However, this choice comes with the cost that we do not have sufficient background to distinguish between mechanisms that explain a switch to violence from those explaining spiraling intensity, because we only cover cases where the absence of conflicts or low-intensity conflicts escalate to full-scale civil wars.
Cederman, Gleditsch, and Buhaug’s (Reference Cederman, Gleditsch and Buhaug2013) book deals with civil wars after 1945, which provides the potential temporal scope of our theoretical argument. However, like Ross (Reference Ross2004) and Bartusevičius (Reference Bartusevičius2019), we limit our focus to the post–Cold War period. This restriction helps address concerns about causal heterogeneity, given that conflicts before and after 1991 occurred in different international and domestic contexts. The end of the Cold War marked a shift in external interventions, the role of international organizations, and the global economic environment, all of which likely shaped conflict dynamics in ways that are not necessarily comparable across periods. Focusing on post–Cold War cases ensures that the structural context within which conflicts occur remains more comparable.
Furthermore, this restriction improves case selection feasibility in a medium-N (LNQA) study. A more recent timeframe allows for a thorough examination of each case while maintaining meaningful cross-case comparisons. Including older cases might expand the dataset but would come at the cost of access to systematic documentation and detailed primary sources, which are crucial for within-case process tracing.
At the same time, we recognize that this focus means our findings are not directly generalizable beyond the post–Cold War period. Although our approach allows us to probe the causal mechanisms underlying grievance-based explanations within our sample, we do not assume that these mechanisms operated in the same way in earlier conflicts. Instead, our study should be understood as testing grievance theory in the context of contemporary conflicts, rather than as a claim about its validity across all historical periods.
Moreover, we focus on cases where excluded ethnic groups were involved in a civil war onset in the period 1991–2021 and where the relevant excluded group only became involved in armed struggle after 1990. Relapses of pre-1991 cases, such as the renewed fighting between Moro rebels and the Philippine government in 1993, are not included. For such cases, the mechanisms need to be traced back to the period before the initial civil war onset, specifically the late 1960s in the case of the Moro conflict. Following the same logic, post–Cold War cases that relapse are only included once, such as the Chechen conflict in which the initial civil war onset in 1994 was followed by a second onset in 1999.
The sample thus comprises excluded groups that were not previously represented by a rebel organization in a civil war onset. Arguably, the likelihood of observing reverse causality would be larger if war relapses were included in the sample. However, current theory focuses on conflict-generating processes rather than on explaining recurring outbreaks of war, and Cederman, Gleditsch, and Buhaug (Reference Cederman, Gleditsch and Buhaug2013, chap. 4) show that the effect of ethnic exclusion on civil war onset remains when only initial civil war onsets are included (Model A4.11 in their online appendix). In short, although prior conflict is an important determinant of civil war (e.g., Cederman, Gleditsch, and Buhaug Reference Cederman, Gleditsch and Buhaug2013, 73), we seek to examine the mechanism in its “clean” form, unaffected by previous iterations of conflict and exclusion.
Based on these criteria, we identify 15 cases (see table 1). The rule that conflicts need to be new onsets leads to the exclusion of four cases; we therefore cover 78% of major civil wars from the Cederman, Gleditsch, and Buhaug (Reference Cederman, Gleditsch and Buhaug2013) dataset for the post-1990 period.Footnote 8 In appendix B, we present additional details about sources and procedures for case selection, including a table showing excluded cases.The sample represents a diverse set of cases: geographically, Africa, Asia, Europe, and the Middle East are represented, and the sample includes both governmental and territorial conflicts. That said, the post–Cold War selection criterion means that the sample is not representative of the larger population of historical cases. However, by analyzing all post–Cold War cases, we at least avoid potential problems associated with further criteria for case selection, which could generate bias (not to mention cherry-picking).Footnote 9
Summary of Findings
Note: Checkmarks indicate that clear evidence of the respective observable implications was found. Marks placed in brackets indicate inconsistent response by the government.
To systematically assess these cases, we rely on a range of secondary sources that provide insights into conflict dynamics, group mobilization, and state responses. We prioritized academic sources, accounts by recognized news media, and well-documented NGO and human rights reports. Although source availability and quality vary across cases, the source base is generally comprehensive and triangulated to minimize bias (see appendix A). To enhance transparency, we provide explicit citations supporting each case evaluation, allowing critical readers to assess the empirical basis of our findings. Where evidence is ambiguous, we mention it in the appendices, ensuring that the reasoning behind our evaluations remains transparent. Ambiguities include difficulties in distinguishing elite strategies from mass-level grievance diffusion, uncertainty over whether government policies constituted consistent rejection, and source disagreements.
Findings
The results from the analysis are summarized in table 1. Checkmarks indicate whether evidence of the two observable implications was found in each of the 15 analyzed cases. Given space constraints, elaborate case descriptions are available in the online appendix, and the discussion here highlights common patterns and interprets the results.
The grievance-based explanation finds support in 9 of the 15 cases. We found clear evidence of the grievance mechanism in the form of grievance-based mobilization in all the corroborating cases. This included collective claims directed at the state that often demanded inclusion or power sharing and extended beyond elite factions to wider ethnic constituencies. The form and intensity of this mobilization varied, but it consistently referenced perceived exclusion and political injustice.
However, the corroborating cases differ in terms of how governments’ reactions to excluded groups’ claims-making triggered conflict escalation. In the first parts of the following discussion, we focus on this variation, which has not gained much attention in previous research, and identify two specific causal pathways by which ethnopolitical exclusion causes civil war onset via grievances: indiscriminate or inconsistent state repression.
Second, in the remaining six cases, the qualitative evidence suggests the presence of reverse causality vis-à-vis the grievance-based explanation. In each case, the relevant ethnic group is excluded because it is engaged in armed conflict, which at the same time serves as a key motivation for the groups’ mobilization for violent collective action.
Pathway 1: Grievance-Based Mobilization—Rejection
The cases that follow the first pathway with group mobilization combined with no accommodation are Adibasi Janajati in Nepal, 1996; Albanians in Serbia-Montenegro, 1998; Armenians in Azerbaijan, 1991; Fur in Sudan, 2003; Tutsi-Banyamulenge in Zaire, 1996; and Murle in South Sudan, 2013. In each case, the government did not make any concessions, systematically repressing the excluded groups’ claims-making, and this repression directly influenced the outbreak of civil war. State repression was indiscriminately targeted against excluded groups before the outbreak of civil war.
How did the lack of accommodation, including indiscriminate state repression, induce armed rebellion in these cases? The analysis shows that indiscriminate campaigns of repression serve as “formative events” that radicalize members of the excluded, mobilized ethnic groups and cause violent backlashes in the short- or medium-term. To illustrate, in November 1995, the Nepali government launched Operation Romeo, a police operation intended to eliminate left-wing radicals in Adibasi Janajati (i.e., Indigenous ethnic groups) districts. As reported by Thapa and Sijipati (Reference Thapa and Sijipati2004, 48), Operation Romeo “was characterized by random arrests, torture, rape, and extra-judicial killings, leading to complete alienation of large parts of the local population.” Because of this indiscriminate nature, it had the unintended effect of boosting Maoist support and participation among Adibasi Janajati, thereby providing “a perfect setting to initiate the people’s war” (48; see also appendix D).
Similarly, the lack of willingness to reform was accompanied by heavy-handed repression by the South Sudanese Army that severely affected civilians and boosted recruitment for the Cobra Faction rebels from the Murle ethnic group (see appendix K). The harsh rejection of demands thus becomes counterproductive because it draws group members into the conflict who would otherwise have remained idle bystanders. Instead, these would-be-moderates join rebel organizations to get revenge or to protect themselves against state violence (see Goodwin Reference Goodwin2001; Mason and Krane Reference Mason and Krane1989; Roessler Reference Roessler2011).
The Serbian governments’ fierce and ham-handed reaction to the 1981 mass protests in Pristina likewise became formative for the Kosovo Albanians’ armed rebellion. At least 57 people were killed in the immediate government reaction, and more than 500,000 people—half of the adult Albanian population—were arrested or interrogated at least once in the eight years following the demonstrations. As argued by Judah (Reference Judah2008, 58), this reign of state terror was an important cause of the 1998 civil war, because “many of the key people who would later set up the Kosovo Liberation Army were imprisoned in or after 1981 [and this] was hugely important in radicalizing them against Yugoslavia and Serbia” (see also appendix E).
The case-specific evidence also shows that fear for group survival induced by indiscriminate violence motivates armed resistance. In Azerbaijan, the anti-Armenian pogroms in Sumgait 1988 were perceived by Armenians as the beginning of a new genocide (or as a recurrence of the 1915 Armenian genocide committed by the Ottoman government). The result was a “wide consensus among Armenians … that, first, the security of the nation required that Karabakh was no longer part of Azerbaijan but united with Armenia, and, second, there were no grounds for Azeris and Armenians to live together” (Zürcher Reference Zürcher2007, 156; see also appendix F). The rejection strategies implemented by state authorities in the Fur and Tutsi-Banyamulenge cases, involving indiscriminate violence, also resulted in “no other way out” reasoning (Goodwin Reference Goodwin2001). For example, one rebel commander fighting in Darfur said, “We had no choice but to organize. We were fighting for our lives” (Flint and de Waal Reference Flint and de Waal2009, 74; see also appendix H).
As is clear from these descriptions and the more detailed accounts in the appendices, the individual conflict processes are, of course, more complex than a stylized account of aggrieved ethnic groups that take up arms against their rulers. For instance, foreign actors often play a pivotal role, such as Rwanda’s strong influence on the Tutsi-Banyamulenge rebellion, which, according to several sources, was orchestrated by Paul Kagame’s RPF regime (see appendix P). In some cases, the role played by ethnic identities is also less clear-cut than indicated in the theory. For instance, although Adibasi Janajati participation in the Maoist insurgency was very high, the Maoist leaders belonged to Nepal’s dominant Hindu groups (see appendix D). Yet, such complexity should not overshadow the findings that grievances about political exclusion do play an important role in mobilizing members of ethnic groups against their governments and that the absence of accommodation does increase support for armed rebellion.
Why would governments ever engage in full-blown rejection of demands that include indiscriminate state repression if it is in fact a counterproductive strategy? Mason and Krane (Reference Mason and Krane1989) suggest that governments do so, not because they expect a high probability of success but because they do not possess the “institutional machinery” to selectively target dissidents. The case studies support this argument. In Nepal, for example, two factors can help explain the government’s inability to use selective repression. First, it relied on poorly equipped police forces that had no counterinsurgency training. Second, the Adibasi Janajati districts, the Maoist stronghold, were characterized by relatively autonomous political structures that impeded the government’s societal and territorial control, thereby reducing its ability to acquire accurate information about dissidents (appendix D; see also Roessler Reference Roessler2011, 316).
The case studies further suggest that the use of indiscriminate repression may be a result of governments’ lack of control over their repressive agents. The Sudanese government countered Fur mobilization with what Flint and de Waal (Reference Flint and de Waal2009, 23) term “counterinsurgency on the cheap”: government-sponsored Arab militias that enjoyed complete impunity and were strongly influenced by racist Arab ideology. This lethal cocktail resulted in genocidal violence. The bulk of the violence committed against the Tutsi-Banyamulenge in Zaire was likewise committed by militias from rival ethnic groups. For example, in the months leading up to the Tutsi-Banyamulenge rebellion, “armed Babembe and Barega thugs sponsored by local Zairian authorities had been killing and looting Banyamulenge at random” (Prunier Reference Prunier2009, 70; appendix P). In both the Fur and Tutsi-Banyamulenge cases, it is unclear what the governments’ preferred repression strategies were. Nonetheless, it seems fair to conclude that contracting violence to militias with their own motives encouraged extreme, violent reactions.
Taken together, the clear rejection of demands expressed by mobilized groups increased support for armed rebellion. The use of indiscriminate violence especially radicalized would-be-moderate group members and inflicted fear for group survival. This finding is in line with previous research arguing that conflict escalation is particularly likely when indiscriminate group repression has occurred: “states that resort to indiscriminate campaigns of repression targeting their challengers leave the movement with little choice but to respond with violent means, thereby triggered revolutionary civil war” (Cederman, Gleditsch, and Buhaug Reference Cederman, Gleditsch and Buhaug2013, 50). This notion is supported by Lindemann and Wimmer (Reference Lindemann and Wimmer2018), who find that excluded ethnic groups only engage in armed conflict when their members have been subjected to indiscriminate acts of state violence.
Moreover, our case studies also suggest that indiscriminate repression was mainly used by states with low coercive capacities that did not have the policing capability to selectively target actual dissidents or that relied on ill-disciplined repressive agents.
Pathway 2: Grievance-Based Mobilization—Inconsistent Reactions
Grievance-based mobilization also preceded civil war in the cases of Abkhazians in Georgia in 1992, Chechens in Russia in 1994, and Serbs in Bosnia in 1992 (appendices C, G, and L, respectively). What distinguishes these cases from the ones ascribed to the first pathway is the inconsistency by which governments reacted to their challengers. The case studies show that the excluded groups neither faced consistent rejection nor consistent accommodation prior to the onset of civil war. The governments’ lack of will and ability to handle ethnopolitical demands in a coherent manner both enabled and motivated excluded ethnic groups to engage in armed rebellion.
When the Abkhazians proclaimed their independence from Georgia in 1990, the Georgian reaction was mixed. On the one hand, the Georgian leadership immediately declared the proclamation invalid, but on the other hand, it struck a power-sharing deal that allowed the Abkhazians (who comprised less than 20% of the population living in Abkhazia) to control the majority of seats in the Abkhazian parliament. However, this deal fell apart because of internal Georgian strife marked by nationalist outbidding, and “the Abkhazian leadership saw a window of opportunity in the breakdown of authority and legitimacy in Georgia” to take full control over Abkhazia (Zürcher Reference Zürcher2007, 130). The civil war broke out in August 1992, when Georgian troops invaded Abkhazia under the pretext of the abduction of a Georgian minister (appendix C).
As Chechen separatism began to accelerate in 1991, the Russian/Soviet reaction was also inconsistent, to say the least. The Russian government under Yeltsin initially supported the Chechen separatist leader Dzokhar Dudaev, who was seen as a natural ally in the struggle against the old Soviet nomenklatura. However, when Dudaev declared Chechnya a fully independent state in November 1991, Yeltsin proclaimed martial law in Chechnya and had paratroopers airlifted in from Russia. The Russian troops then surrendered to Dudaev’s forces without a fight, because then-Soviet president Gorbachev, who fiercely opposed a bloody military operation in Chechnya, “ordered the army to remain strictly neutral and take no part whatsoever in enforcing martial law” (Dunlop Reference Dunlop1998, 120). Chechnya was de facto independent from December 1991 to the autumn of 1994, when Russia invaded in an (unsuccessful) attempt to regain control of the breakaway republic (appendix G).
In line with the Abkhazian and Chechen cases, the Bosnian government was neither willing to accommodate the Serb minority into peace nor able to repress it into submission. When Presidency Chairman Alija Izetbegovic took steps toward an independent, Muslim-dominated Bosnia in 1991, the Bosnian Serbs formed their own parliament and later declared their own Republika Srpska, which was to remain within the Yugoslav Federation. Izetbegovic rejected this attempt to partition Bosnia and accused the Serbs of trying to destabilize the country. However, in early 1992, he decided to accommodate the Serbs and agreed to devolve autonomous power to the country’s ethnic communities by forming national territorial units within Bosnia—an agreement that he would soon renege. Izetbegovic’s capriciousness reflected his desire to establish a unitary Bosnian state without triggering ethnic violence that would draw the Serb-sympathetic Yugoslav Army into the conflict. He was thus walking a tightrope, because “the Serbs were deliberately provoking Bosnian security forces into retaliation, in the hope that [the Yugoslav Army] would intervene on their behalf” (O’Ballance Reference O’Ballance1995, 11). When civil war erupted in April 1992, it was the Bosnian Serbs who, supported by Serb militias and the Yugoslav Army, sought to use their first-strike advantage (appendix M).
In all three cases, inconsistent government reactions help explain conflict escalation. Accommodative policies facilitated the excluded groups’ mobilization efforts, and repressive policies convinced them that violent collective action was needed to redress their political grievances; that is, because the excluded groups were not fully accommodated, they maintained their goal of bringing about a new political order. As in the first pathway, low coercive capacity played a crucial role in the inefficient government reactions in these cases. Specifically, political instability at the center exhausted the governments’ repressive capabilities. In addition to the “Abkhazian problem,” successive Georgian governments were battling conspiracies from within and separatism in South Ossetia (appendix C). Chechen separatism was aided by the division of authority between the Russian and Soviet leadership and the power struggle in Moscow, which culminated in the unsuccessful coup d’état in August 1991 (appendix G). The newly independent Bosnian state, which had little time to expand its military capabilities, relied on lightly armed police forces that were no match for the federally controlled Yugoslav Army (appendix M).
Notably, the three instances all took place in the context of disintegrating communist regimes. However, several Cold War cases suggest that this pathway can be observed more generally. For example, ethnonationalist claims that met with inconsistent government reactions preceded armed rebellions in early post-independence Myanmar and the 1961 civil war in Iraqi Kurdistan (e.g., McDowall Reference McDowall2005, chap 15). It is also worth noting that all these cases were territorial conflicts. Although this suggests that inconsistency in government reactions could be particularly likely to trigger territorial civil war, historical cases again suggest that the mechanism works more generally. As Lichbach (Reference Lichbach1987, 287) reminds us, the revolutions in France (1789), Mexico (1912), and Russia (1919) were all characterized by prerevolutionary regimes that incoherently mixed accommodation and violent rejection, thereby facilitating mass mobilization without satisfying popular demands.
The second pathway, in which inconsistent government reactions have an escalating effect on ethnopolitical struggles, has received less attention in previous studies. Our analysis shows that state responses are not always consistently seeking to accommodate or reject demands from excluded groups. Hence, it is important for an updated version of grievance theory to explicitly incorporate how an inconsistent mix of accommodation and rejection favors excluded groups’ mobilization efforts.
Disconfirmatory Evidence: Reverse Relationship
The qualitative evidence from the remaining six case studies are examples of reverse causality, in which armed conflict was responsible for political exclusion and at the same time motivated the excluded groups’ rebellion. We label these instances as a reverse relationship because the causal chain did not begin with ethnopolitical exclusion, which led to civil war (see figure 1): instead, an ongoing conflict led to exclusion. Exclusion dynamics or grievances may still have explanatory power in these cases but not in the linear fashion where they constitute the root cause. For instance, exclusion because of an ongoing conflict may motivate group members to sustain their armed efforts. We return to this issue in the discussion.
In Afghanistan, the ongoing civil war led to Pashtun exclusion when mujahideen groups overthrew the Najibullah regime in 1992 and Tajik and Uzbek warlords took control over Kabul. The qualitative evidence strongly indicates that the ongoing war in Afghanistan was the main motivation for the Pashtun-dominated Taliban movement, which emerged in the Kandahar province in 1994 (appendix L). The emergence of the Taliban was a local response to a lack of order and moral deprivation (Nojumi Reference Nojumi, Crews and Tarzi2008, 101; Sinno Reference Sinno2008, 78). The two other Afghan cases in our sample, Tajik in 1996 and Uzbek in 1996, also exhibit evidence of reverse causation. A Tajik–Uzbek alliance had conquered Kabul in 1992 and subsequently fought internally over Afghan state power. However, when the Taliban attacked Kabul in 1996, the Tajik and Uzbek forces retreated north of Kabul, once again allied, to resist further Taliban advancement. As argued by Goodson (Reference Goodson2001, 80), the Taliban “constituted the de facto government of Afghanistan since their capture of Kabul in September 1996.” Yet, their takeover did not end the civil war; it only reversed the roles of who were “government” and who were “rebels.” Accordingly, the Tajik and Uzbek did not rebel in 1996 because of their exclusion from state power; rather, they lost state power because of their involvement in an ongoing civil war (appendices N and P).
The Lari/Bakongo case in Congo-Brazzaville in 1998 closely resembles these Afghan cases. The Lari/Bakongo were part of President Lissouba’s government when armed conflict broke out in the capital of Brazzaville in 1997 between militias supporting Lissouba—including Lari/Bakongo Ninjas—and militias supporting former president Sassou-Nguesso. The Lari/Bakongo became politically excluded when Sassou-Nguesso’s militias, supported by Angolan forces, captured Brazzaville in October 1997. The civil war continued in the Pool region in 1998, where the now-excluded Lari/Bakongo fought the government forces as members of Fréderic Bitsangou’s Ntsiloulous, “the majority of whom were former Ninjas militiamen” (Demart and Tonda Reference Demart and Tonda2016, 203; appendix J). Thus, the Lari/Bakongo lost their access to state power because of a defeat in an ongoing civil war, and most of those representing the Lari/Bakongo in armed combat in 1998 were already fighting before the group became politically excluded. This circumstance strongly indicates that grievances about political exclusion do not explain the Lari/Bakongo mobilization for violent collective action.
The Hutu in Rwanda were also politically included when civil war broke out in 1990, and they lost state power when Paul Kagame’s Tutsi army conquered the country in the wake of the 1994 genocide. More than one million Hutu took refuge in eastern Zaire in 1994, including the notorious Interahamwe militia and the former Rwandan Armed Forces (FAR), “complete with its heavy and light weapons, its officer corps, and its transport echelon” (Prunier Reference Prunier2009, 25). The retreat from Rwanda merely transformed the Hutu army into a rebel force that regrouped and prepared to strike back at Kagame. The Hutu’s first armed infiltrations back into Rwanda occurred in late 1994, only three months after the end of the genocide, and the armed conflict continued at a low level of intensity until Kagame reescalated it by attacking the Hutu refugee camps in late 1996 (appendix I). Thus, we find indications that the Hutu exclusion was caused by civil war, and the 1997 reescalation was a direct continuation of that same civil war, involving the same, never-demobilized parties. Again, this evidence indicates that prior armed conflict, rather than grievances about political exclusion, was the original cause of the mobilization for violent collective action.
Finally, armed conflict also preceded ethnopolitical exclusion in Iraq. The first manifestations of Sunni exclusion after the US-led invasion came in late 2004, when the dominant Sunni Arab party resigned from the cabinet and called for a boycott of the 2005 election because of the regime’s harsh crackdown on Falluja. Falluja was the center of the Sunni insurgency, which had emerged as early as mid-2003 (appendix N). Ethnopolitical exclusion was thus a result of armed conflict, rather than the other way around, and armed conflict—namely the US-led invasion—was another key motivation behind the Sunni’s rebellion. As argued by Hashim (Reference Hashim2009, 14), “the most fundamental cause of the Sunni insurgency was resistance to foreign occupation.”
The Iraqi case begs an important question: If the US-led invasion in 2003 caused the Sunni population to anticipate exclusion from central state power, might this explain their mobilization for violent collective action? This logic, which is in line with Gurr’s (Reference Gurr2000) notion of “progressive deprivation,” can certainly be observed in the Iraqi case, where fear of future Shia domination was a contributing motivation for the Sunni insurgency (appendix N). However, as currently formulated, the grievance-based explanation is about “actual and objectively measurable differences in group status” (Cederman, Gleditsch, and Buhaug Reference Cederman, Gleditsch and Buhaug2013, 39).Footnote 10
To assess the extent to which reverse causality challenges conventional grievance theory, it is important to consider the nature of the case selection. The sample is not random; it consists solely of excluded groups represented by rebel organizations in civil wars that began after 1990. This means that these cases are precisely those in which the theorized grievance-to-war pathway should be observable. The fact that 40% of these cases exhibit evidence of reverse causality highlights the need to move beyond strictly unidirectional grievance-based explanations, as we elaborate next.
Discussion
Prior research has established a strong correlation between ethnopolitical exclusion and civil war, but the causal mechanisms driving this relationship have not been systematically scrutinized across multiple cases. This study has applied LNQA to assess whether the theorized grievance-based pathway—linking political exclusion to armed conflict via mobilization and state repression—holds empirically across 15 cases of politically excluded ethnic groups engaged in major civil war onsets after 1990.
Our results indicate that in 60% of the cases, civil war followed a pattern broadly consistent with grievance-based theories. In these nine cases, political exclusion functioned as a catalyst for grievance-based mobilization, which escalated into civil war in response to clear state rejection of demands, including state repression. This finding offers important empirical support for the grievance theory by moving beyond correlation and demonstrating that the theorized causal process can be traced across diverse cases. Given the complexity of civil war onset, this level of corroboration is noteworthy.
However, our results also pointed to important refinement of the relationship. A key insight is that the full grievance sequence typically unfolds under two specific patterns of state response. In one pathway, excluded groups escalate after facing indiscriminate repression that leaves no room for accommodation, often in contexts of weak state capacity. In the other pathway, escalation results from inconsistent or incoherent state responses that oscillate between concessions and coercion and create uncertainty and foster radicalization. These two pathways suggest that it is not just exclusion but also how states respond to dissent that determine whether conflict escalates.
Moreover, six cases deviated from the grievance model. In these, exclusion followed rather than preceded violent mobilization. Rather than treating these as anomalies, we interpret them as evidence of a feedback loop: exclusion and civil war can mutually reinforce each other, with conflict triggering exclusion that in turn hardens grievances and leads to further escalation. Notably, our decision to focus exclusively on the first onset of major civil wars should, in principle, reduce the risk of reverse causality. The fact that we still observe several such cases reinforces the conclusion that recursive dynamics between exclusion and violence are more common and more theoretically consequential than often acknowledged.
Figure 2 reflects these theoretical insights by integrating both the original grievance pathway and a newly identified recursive dynamic. Solid arrows represent the standard exclusion-to-conflict mechanism; the dotted arrow captures the feedback loop observed in the divergent cases. Across all cases supportive of the grievance mechanism, rejection of demands through indiscriminate or incoherent repression was present. This underscores that exclusion alone is rarely sufficient because escalation depends on patterns in the state’s response.
Revised Theoretical Link between Ethnopolitical Exclusion and Civil War
Note: The dotted line refers to a potential feedback loop.
Our findings refine existing theory in two important ways. First, they underline the central role of state behavior, particularly rejection and repression, as a trigger of conflict escalation. Second, they highlight the need to allow for endogeneity in theoretical models because exclusion can be both a cause and a consequence of conflict. Theorizing civil war onset thus requires attention to recursive dynamics and sequencing.
In methodological terms, our research design has limitations. We did not analyze “negative” cases—excluded groups that did not resort to violence—and hence, we cannot draw firm conclusions about the relative effects of different repression strategies. Yet the study demonstrates the value of LNQA for investigating causal mechanisms across multiple cases. It complements large-N studies by adding process-level evidence that helps assess the plausibility of hypothesized pathways. This approach also guards against overgeneralization by showing when and how a mechanism breaks down. Moreover, our analysis has demonstrated that the temporal ordering between exclusion and violence is not always clear-cut, especially in state collapse scenarios. This situation raises questions about how conflict spells are treated in well-known conflict datasets, such as those provided by the UCDP. The operational definitions of conflict onset, duration, and recurrence can obscure dynamic processes in which exclusion emerges as a consequence of prior conflict, rather than as a precondition for it.
Our findings also suggest that additional care is needed when replicating or interpreting quantitative studies. Many large-N analyses treat exclusion in the year prior to conflict onset as exogenous. Our evidence indicates that in several cases, exclusion follows or co-evolves with violence. Although such studies partly take this risk into account by controlling for a country’s prior conflict history, this does not fully resolve the issue of endogeneity in the timing and coding of exclusion. In several cases, past or ongoing low-intensity violence appears to have shaped the classification of groups as excluded, raising questions about the direction of causality. Nonetheless, when replicating Cederman, Gleditsch, and Buhaug’s (Reference Cederman, Gleditsch and Buhaug2013) model excluding the ambiguous cases, we find that the relationship between exclusion and civil war remains positive and significant but becomes a bit weaker. Hence, the average effect tends to be robust, but more caution is called for when interpreting statistical associations as evidence of causal mechanisms.
In sum, this study has sharpened our understanding of when and how ethnopolitical exclusion leads to civil war. Grievance-based mechanisms are clearly at work in many cases, but only under specific conditions of state response. Conflict onset is not an automatic consequence of exclusion but depends on how states can and want to handle dissent. Moreover, exclusion can arise as a response to preexisting conflict, creating a dangerous cycle of marginalization and violence. Preventing ethnic civil wars requires not just ending exclusion but also crafting credible, consistent, and inclusive responses to emerging claims.
Supplementary material
To view supplementary material for this article, please visit http://doi.org/10.1017/S1537592725103101.
Acknowledgments
We thank Henrikas Bartusevičius, David Andersen, Lasse Lindekilde, Kristian Gleditsch, Christian Davenport, Jakob Tolstrup, and Ragnhild Nordås, as well as the participants at the NOPSA 2024 workshop on political regimes, especially Jonas Schmid. We are also grateful to the editors and the four anonymous reviewers for their thorough and constructive comments.
Open access
