Linking individual and group motives for violent conflict

Abstract

Research on the role of grievances in civil conflict is surprisingly inconclusive, with well-cited studies disagreeing on the relationship between perceived deprivation and violence. I argue that the role of grievances depends on an interaction between individual and group-level incentives. Individuals who perceive themselves as personally deprived are more likely to support or participate in anti-regime violence, but only if a successful rebellion would enhance their group’s power relative to the status quo. I test this argument in the context of Iraq’s sectarian civil war using data from a 2016 survey of 800 Baghdad residents. Using a list experiment to measure individuals’ willingness to consider violence against a government they feel is ignoring their needs, I find that minority Sunnis who are economically dissatisfied are significantly more willing to consider violence than similarly aggrieved Shias. However, as economic satisfaction increases, Sunnis’ propensity for violence decreases until it becomes indistinguishable from Shias’ propensity. These results clarify the joint impacts of vertical and horizontal grievances. Group inequality and individual deprivation are each necessary but not sufficient to fully explain individuals’ propensities for anti-state violence.

Empirical research on violent conflict disagrees on whether inequities matter vertically, i.e. between better- and worse-off individuals (Gurr, 1970; Muller, 1985), horizontally, that is, between better- and worse-off groups (Buhaug et al., 2014; Cederman et al., 2011; Østby, 2008), or little at all (Collier and Hoeffler, 2004; Fearon and Laitin, 2003). This article points to an interactive effect of grievances at the individual and group levels. I propose that grievances matter vertically, but primarily within the aggrieved group. Grievances also matter horizontally, but only among the most deprived individuals. Failure to account for the interaction between individual- and group-level characteristics can therefore obscure grievances’ effects on individuals’ propensities for violence. Some research alludes to the differential impacts of grievances depending on an individual’s group membership (Mazur, 2019), perceived opportunity to affect change (Dyrstad and Hillesund, 2020), or preferred conflict outcome (Humphreys and Weinstein, 2008). However, these studies stop short of empirically demonstrating the interactive effects of individual- and group-level grievances.

This study bridges this gap between theory and evidence using a survey of 800 Baghdad residents and a list experiment which measures individuals’ willingness to consider anti-government violence. I find that minority Sunnis who are economically dissatisfied are significantly more willing to consider violence than similarly aggrieved Shias. As economic satisfaction increases, Sunnis’ propensity for anti-state violence decreases until it becomes indistinguishable from Shias’ propensity. These results speak to the debate over grievances, and imply that group inequality and individual deprivation are each necessary but not sufficient to explain individuals’ propensities for anti-state violence.

1. Theory

Where civil conflict is feasible, each potential combatant considers his expected costs and rewards of participating in or supporting rebellion. Costs stemming from the risk of physical harm, arrest, and lost wages accrue at the individual level. While some rewards like payoffs from looting accrue to individuals, many depend on whether political change would make one’s group better or worse off. Political competition and the distribution of goods often occur along an identifiable group cleavage like ethnicity or class. Individuals whose group is out of power have the most to gain from political change, while those whose group is in power have the most to lose.

Research on “horizontal grievances” uses this logic to argue that the more deprived a group is, the more likely its members will rebel. A successful rebellion addresses deprivation by providing members of the previously deprived group with access to political power and a more favorable distribution of economic resources. However, in anything short of total war, not all members of a group fight. Those who are personally worse-off should be more willing to incur personal risk and fight, underlying theories of “vertical grievances.”

Thus, to fully predict people’s incentives to fight, we must account for both their perceived well-being and their position in society. Individuals who perceive themselves as particularly deprived should be more willing to incur personal risks to fight, but only to the extent that they expect that a successful rebellion will address their grievances. In divided societies, membership in a deprived group implies that a successful rebellion may address one’s grievances, while membership in an empowered group implies that a change in group power is unlikely to address one’s personal deprivation.

A related argument refers to the interactive effects between grievances and opportunity to affect change. Dyrstad and Hillesund (2020) find that the impact of grievances on violence is moderated by individuals’ perceptions that they can affect change through peaceful methods. In divided societies, discrimination often limits access to such opportunities, for example through voting rules which disenfranchise members of certain groups. Even where access is equitable, however, violent and non-violent methods of regime opposition differ in the likelihood that they will address each group’s grievances. Iraqi Sunnis and Shias have equal opportunity to vote against the incumbent regime. However, deeply entrenched sectarianism and Shias’ demographic plurality ensure that any elected government will be led by a Shia party. To the extent that Sunnis attribute their economic deprivation to sectarianism, replacing the Shia-led government with a different Shia-led government is unlikely to solve their grievances. Only a wholesale change in group power is likely to address their underlying motives. On the other hand, Shias who perceive themselves as personally deprived are less likely to attribute this deprivation to the fact that a Shia government is in power, and therefore may be more willing to pursue non-violent methods to address their grievances.

Relevant activities include direct participation in fighting, as well as indirect support like providing shelter for rebels or withholding information from the authorities (Berman and Matanock, 2015). This individual-level propensity for violence influences the realization and intensity of civil conflict, alongside factors like collective action potential (Esteban and Ray, 2011) and physical geography (Fearon and Laitin, 2003).

Perceived individual deprivation and group membership should interact anywhere that political power is distributed across groups with relatively static membership. Divisions across ascriptive identity like race or ethnicity are likely to qualify. In terms of class conflict (Skocpol, 1979), in a capitalist society workers should become more likely to rebel during a market crisis as their wages decrease, but factory owners will not call for rebellion even if the same crisis erodes their personal economic well-being for fear of upending the system from which they benefit.

Accounting for the interaction between group and individual incentives helps clarify theoretical expectations and leads to a testable hypothesis about the propensity for anti-state violence:

Hypothesis

Grievances over perceived personal deprivation are positively associated with individuals’ propensities for anti-state violence, but only among individuals from the out-of-power group.

2. Empirical context

From 2003 to 2018, fighting in Iraq caused about 200,000 civilian deaths, ¹ making it one of the deadliest conflicts since World War II. Following the US-led invasion, fighting coalesced into civil war along sectarian lines. Iraqis hold identities on a variety of salient dimensions, but sectarian cleavages are especially relevant. Shia Arabs make up a plurality, and Sunni Arabs constitute the next largest group. Before the invasion, Sunnis dominated Iraqi politics for most of several centuries. When Iraq transitioned to democracy in 2004, Shias’ demographic advantage gained them control of most high-level government positions.

After the Baath regime’s fall, Iraqis increasingly turned to sectarian organizations for services and security (Robinson, 2009). As fighting transitioned from disorganized looting to anti-state insurgency, sectarian groups fought against each other and against state forces. Violence occurred on several fronts, including Sunni and Shia militias fighting against each other and against government forces dominated by Shia partisans. Minorities feared that democracy would lock them out of government, that the Shia-dominated system would fail to serve to their communities, and that Shias would retaliate for past governments’ actions (Haddad, 2011). In Sunni neighborhoods of Baghdad, police were absent at best, or directly complicit in sectarian murders at worst (Nanes, 2019). Reports emerged of Shia doctors refusing to treat Sunni patients in state hospitals (Robinson, 2009, 154). When the Islamic State invaded predominantly Sunni provinces in 2014, government security forces failed to protect the population (Liz and Ramadan, 2014).

Under these conditions, the expected implications of a change in government depend on both the method of change and one’s group membership. Iraq’s democratic institutions and entrenched sectarianism in party politics make the election of a Shia-dominated government a near certainty, with key cabinet positions distributed to Shia parties. In contrast, non-democratic government under the Ottoman empire, British colonialism, and Baathist dictatorship yielded centuries of Sunni dominance. Thus, a Shia seeking to replace the current government with other leaders from her group could reasonably expect to do so through elections, while a Sunni seeking to install a coethnic government may turn to extra-systemic means, resulting not just in a reshuffling of representatives but a wholesale status reversal for her group.

Iraq provides a useful setting for research in part because of its salience to global politics: dozens of countries committed thousands of lives and trillions of dollars to building appropriate government structures for a divided society, only to watch these attempts collapse into sectarian civil war. Yet, the relationship between grievances and support for violence should function similarly in all divided societies prone to violence. The key features of Iraq—deep-seated social and political cleavages based on group identity, and context which makes violent insurgency feasible—are common to Lebanon, Yemen, Afghanistan, Nigeria, parts of the Philippines, and many other locations worldwide.

3. Research design and data collection

Within Iraq, I leverage within-unit variation by analyzing responses to a 2016 individual-level survey of adults living in Baghdad. Conducting the analysis at the individual respondent level is necessary to identify the hypothesized interactive effects between group and individual characteristics. Due to sectarian violence and migration during the 2000s, Baghdad remained one of the few governorates with large populations of both Sunnis and Shias. A local research firm implemented the survey using local enumerators. I discuss safety and ethical considerations in the supporting information. All Arabs ages 18 and over living in Baghdad were eligible for inclusion. A stratified random design blocked on sect yielded a sample of 400 Sunnis and 400 Shias, each representative of the group’s adult population in Baghdad. I measure economic grievances through self-reported economic satisfaction:

“How satisfied are you with your economic situation?” (1–5 scale)

I focus on economic grievances for several reasons. First, foundational studies of civil conflict specify economic deprivation in their theories and operationalizations linking grievances with violent conflict (Gurr, 1970; Stewart, 2011; Cederman et al., 2011). Thus, it is appropriate to focus on economic grievances in this attempt to clarify the relationship between individual- and group-level deprivation. Second, economic well-being has broad implications for class conflict as well as for ethno-sectarian conflict. Finally, it is highly likely that economic conditions stem at least partially from, and therefore correlate with, sources of political grievances like unequal delivery of government goods and services.

Attitudinal outcomes appropriately measure grievances, as individuals respond to their perceptions. Even if individuals’ evaluations of their relative well-being are not accurate, grievances depend on the extent to which they perceive the distribution as unfair, making perceptions the most relevant construct. In the supporting information, I analyze a second measure of grievances, individuals’ education, as a direct measure of attainment.

I measure propensity for anti-state violence using the same survey, allowing me to test individual-level hypotheses while holding constant local-level structural conditions. Geolocated data on violence can test which factors correlate with incidences of violence, but cannot be used to test causal theories about why individuals participate in violence. People may commit violence away from their home, making it difficult to associate personal conditions with conflict behavior based on location-level data. Of course, survey measures have limitations. Expressing willingness to consider violence on a survey costs little relative to actual participation in anti-state violence, and may not reflect individuals’ true likelihood of turning to violence. However, even low-cost methods of anti-state behavior like providing shelter to insurgents or withholding information from the authorities can facilitate rebellion (Berman and Matanock, 2015; Dyrstad and Hillesund, 2020). Thus, self-reported willingness to consider anti-state violence is consequential even though it is unlikely to map perfectly onto violent action.

I measure respondents’ willingness to “consider using non-peaceful methods” against the government, which could encompass a range of activities including participation in riots, occupying government buildings, and other activities which contribute to the regime’s destabilization. Asking respondents whether they would “consider” violence reduces sensitivity. I use a list experiment to further address sensitivity bias.

I am going to read you a list of [4 / 5] strategies that citizens sometimes use when the government does not seem to be listening to them. Please tell me how many of these strategies you would consider using if you felt the government was ignoring your needs. Remember, I don’t need to know which ones you would use, only how many of these [4/5] you would consider. ³

Subjects in the “short list” group received a list of four non-controversial items: voting against the government, writing letters to the government, writing letters to an international organization, and protesting peacefully. The “long list” group received the same four items plus a fifth, using non-peaceful methods. Since respondents were randomly assigned across groups, the respondents in the short and long list groups should have the same characteristics on average, and therefore select the same number out of the four non-controversial items on average. Thus, any increase in the number of items selected between the two groups is attributable to the presence of the sensitive item (willingness to use non-peaceful methods).

The mean response to the long list is .30 higher than the mean response to the short list (p < .01), indicating that on average about 30% of Iraqis would consider using non-peaceful methods against the government, including 24% of Shias and 36% of Sunnis (both estimates p < .01. See Supplementary Table S3). In the supporting information, I show that respondents are balanced across lists on all observed characteristics (Supplementary Table S1). Supplementary Table S2 rejects the presence of design or floor effects. Modest evidence of a ceiling effect suggests that this estimate of propensity for violence may be a conservative lower bound, but is unlikely to bias the analyses described below on the differences in propensities between groups.

4. Results

I use multivariate regression to test the conditional relationship between economic satisfaction and propensity for anti-state violence. I estimate the relationship between each independent variable and propensity for violence for each individual i in district n. Each model contains the measure of economic grievances (θ), a dummy variable for Sunni (ρ), and the interaction between grievances and Sunni. I control for the respondent’s age and gender (vector X), and district fixed effects account for structural differences across locations (ϕ). The fixed effects isolate individual motives from locally determined costs of fighting. For example, individuals in districts with narrow streets might find it easier to place IEDs without being detected. Standard errors are clustered by neighborhood, the primary sampling unit. ² Because the dependent variable comes from a list experiment, independent variables are interacted with the respondent’s treatment status (π, 1 if long list, 0 if short list). The resulting coefficient is the difference in the effect of each independent variable between respondents assigned to short versus long lists, that is, the impact of the presence of the sensitive item (propensity for violence) on the number of list options a respondent selected.

$y_i={\beta }_0+{\beta }_1{\pi }_{in}+{\beta }_2{\rho }_{in}+{\beta }_3{\pi }_i{\rho }_{in}+{\beta }_4{\theta }_{in}+{\beta }_5{\theta }_{in}{\rho }_{in}+{\beta }_6{\theta }_{in}{\pi }_{in}+{\beta }_7{\pi }_{in}{\rho }_{in}{\theta }_{in}+{\sigma }^{\prime }{X}_{in}+{\varsigma }^{\prime }{X}_{in}{\pi }_{in}+{\delta }^{\prime }{\pi }_i{\varphi }_n+{\epsilon }_{in}$

(1)

Figure 1 shows the marginal effect of being in the list experiment treatment group (functionally, the estimated proportion of individuals who would use violence) across the range of economic satisfaction. A shift of 1 point on the y axis corresponds to a 100% change in expected willingness to use violence. Economic grievances exhibit a striking impact. Economically dissatisfied Sunnis are more likely than similarly-dissatisfied Shias to support anti-government violence. As economic satisfaction increases, the gap between Sunnis’ and Shias’ propensities for anti-state violence decreases to a statistical zero. Failing to account for the interactive effect would have provided only moderate evidence at best for either vertical (individual) or horizontal (group) grievances in the pooled population. Among Sunnis, propensity for violence decreases as economic satisfaction increases, although this relationship just misses statistical significance at conventional levels.

Figure 1.

Impact of individual characteristics on willingness to use violence.

Shias appear to exhibit the opposite trend, with violence increasing alongside economic satisfaction. However, the relationship is not significant on its own. Binned marginal effects (Supplementary Figure S8) as recommended by Hainmueller et al. (2019) show similar trends as in Figure 1 among Sunnis, but suggest a relatively flat and imprecisely estimated slope among Shias. Whereas, the substantive impacts of economic grievances are meaningful among Sunnis, there is no evidence of a corresponding relationship among Shias. This null finding for Shias is consistent with the theory that individual grievances should only instigate anti-state violence within the group which would expect violent regime change to address the source of their grievances. In the context of Iraq’s sectarian politics, aggrieved Shias have little reason to hold such expectations. In the supporting information, I find largely similar trends when measuring grievances using education: among Sunnis, those with higher levels of education (interpreted as unrealized investment in job prospects, given Baghdad’s tattered economy) become increasingly prone to anti-state violence, whereas there is no evidence of a corresponding relationship among Shias.

In the supporting information, I engage with the related explanation that opportunity to influence politics non-violently moderates individuals’ propensities for violence. I evaluate whether Shias are more likely than Sunnis to use non-violent methods to oppose a regime “which is ignoring [their] needs” using the same list experiment. I find no evidence of sectarian differences in the use of non-violent methods across values of economic satisfaction or education (Supplementary Figure S9).

The regression framework rules out several potential sources of endogeneity. Many likely omitted predictors of both grievances and violence are near-constant within location and accounted for by district fixed effects. For example, citizens exposed to more violence might be desensitized and therefore more likely to consider violence, and also have fewer economic and political opportunities. In the supporting information, I show that including fixed effects at the smaller neighborhood level produces similar coefficients, albeit with larger standard errors (Supplementary Figure S7). The analysis remains susceptible to omitted variable bias at the individual level. For instance, politically active Sunnis may be discriminated against in the labor market. While plausible, the data do not bear out such a relationship: the correlation with the number of items of non-violent political participation selected by short list respondents is −0.07 for education and −0.06 for economic satisfaction, meaning that individual deprivation is unrelated to general levels of political activity in this sample.

Reverse causality could occur if people predisposed to anti-state violence express negative attitudes about their situation on the survey in order to reduce cognitive dissonance. However, the measure of propensity for violence came at the end of the survey, after the measures of grievances. Furthermore, the list experiment eliminates externally-visible dissonance. Finally, while reverse causality might explain an overall relationship between the perception-based measures and responses to the list experiment, it would not explain the mediating effects of sect.

5. Discussion

I show that in Baghdad, the effects of perceived economic deprivation depend on group membership. These results are consistent with a conceptualization of vertical grievances in which individuals compare their personal well-being against society, but with the caveat that only those who reasonably expect to benefit from a change in group power seem inclined to turn to violence. Perceived economic deprivation is associated with increased propensity for violence only among Sunnis, the out-of-power group. As economic satisfaction increases, propensities for anti-state violence become indistinguishable across groups. This caveat is empirically consistent with work on horizontal grievances. Violence that appears in country-level data to be driven by inequality across groups may reflect an increasing proportion of economically and politically underserved individuals in the marginalized group who view rebellion as potentially beneficial to their personal situations. These results are also closely related to work which shows that individual-level opportunity to affect change non-violently moderates the impact of grievances on propensity for violence.

The evidence in this study implies that people use violence to affect change in the political balance of power, rather than out of an inherent opposition to unfair conditions for themselves or their group. Of course, this evidence comes from a snapshot in time in a single city, but it suggests a need to revisit the theoretical foundations of grievances. Empirically, it shows that we cannot dismiss the role of grievances in conflict based only on analysis of data at the individual, group, or society level, as interactive effects across levels may obscure grievances’ effects.

The impact that this interactive effect has on the overall level of violence depends on several additional factors, complicating the link between individuals’ propensities for rebellion and civil conflict. One such factor is the propensity of dominant group members to take up arms to protect the state. I speculate that their decision to do so should mirror patterns of marginalized group members’ willingness to rebel. Dominant group members who are relatively better off under the status quo should take greater risks to preserve it. Because my survey asks specifically about anti-state violence, additional research is necessary to test this hypothesis. The occurrence and intensity of civil conflict also depends on collective action. Better resourced elites from the marginalized group may be better able to target club goods to induce participation. By virtue of being well-resourced, these individuals are less likely to hold individual-level grievances. This elite-level mechanism biases against my finding on economic satisfaction, and likely applies only to a small number of individuals. On the whole, while the realization of conflict depends on myriad factors not captured here, the results in this article speak to important individual- and group-level dynamics which induce ordinary individuals to support violence.

The dynamics in this article should apply to all societies where conflict occurs between groups whose membership is largely static. “Divided societies” where political competition occurs between religious, ethnic, racial, or linguistic groups are likely to experience these patterns. Societies undergoing modernization and the resulting inequality may also experience these patterns if individuals identify strongly enough with a particular economic class and between-class mobility is sufficiently difficult. On the other hand, political partisanship is unlikely to be sufficiently rigid to experience these dynamics. Ultimately, this article points to the importance of individual motives, group motives, and their interaction for predicting individuals’ propensities for rebellion. Empirical studies of civil war must consider factors at both levels to explain the available supply of combatants in any given conflict.

Supplemental Material

sj-pdf-1-rap-10.1177_20531680211061056 – Supplemental Material for Linking individual and group motives for violent conflict

Supplemental Material, sj-pdf-1-rap-10.1177_20531680211061056 for Linking individual and group motives for violent conflict by Matthew Nanes in Research & Politics

Supplemental Material

sj-pdf-2-rap-10.1177_20531680211061056 – Supplemental Material for Linking individual and group motives for violent conflict

Supplemental Material, sj-pdf-2-rap-10.1177_20531680211061056 for Linking individual and group motives for violent conflict by Matthew Nanes in Research & Politics

Supplemental Material

sj-pdf-3-rap-10.1177_20531680211061056 – Supplemental Material for Linking individual and group motives for violent conflict

Supplemental Material, sj-pdf-3-rap-10.1177_20531680211061056 for Linking individual and group motives for violent conflict by Matthew Nanes in Research & Politics

Supplemental Material

sj-pdf-4-rap-10.1177_20531680211061056 – Supplemental Material for Linking individual and group motives for violent conflict

Supplemental Material, sj-pdf-4-rap-10.1177_20531680211061056 for Linking individual and group motives for violent conflict by Matthew Nanes in Research & Politics

Supplemental Material

sj-pdf-5-rap-10.1177_20531680211061056 – Supplemental Material for Linking individual and group motives for violent conflict

Supplemental Material, sj-pdf-5-rap-10.1177_20531680211061056 for Linking individual and group motives for violent conflict by Matthew Nanes in Research & Politics

Supplemental Material

sj-pdf-6-rap-10.1177_20531680211061056 – Supplemental Material for Linking individual and group motives for violent conflict

Supplemental Material, sj-pdf-6-rap-10.1177_20531680211061056 for Linking individual and group motives for violent conflict by Matthew Nanes in Research & Politics

Supplemental Material

sj-pdf-7-rap-10.1177_20531680211061056 – Supplemental Material for Linking individual and group motives for violent conflict

Supplemental Material, sj-pdf-7-rap-10.1177_20531680211061056 for Linking individual and group motives for violent conflict by Matthew Nanes in Research & Politics

Supplemental Material

sj-pdf-8-rap-10.1177_20531680211061056 – Supplemental Material for Linking individual and group motives for violent conflict

Supplemental Material, sj-pdf-8-rap-10.1177_20531680211061056 for Linking individual and group motives for violent conflict by Matthew Nanes in Research & Politics

Supplemental Material

sj-pdf-9-rap-10.1177_20531680211061056 – Supplemental Material for Linking individual and group motives for violent conflict

Supplemental Material, sj-pdf-9-rap-10.1177_20531680211061056 for Linking individual and group motives for violent conflict by Matthew Nanes in Research & Politics

Supplemental Material

sj-pdf-10-rap-10.1177_20531680211061056 – Supplemental Material for Linking individual and group motives for violent conflict

Supplemental Material, sj-pdf-10-rap-10.1177_20531680211061056 for Linking individual and group motives for violent conflict by Matthew Nanes in Research & Politics

Supplemental Material

sj-pdf-11-rap-10.1177_20531680211061056 – Supplemental Material for Linking individual and group motives for violent conflict

Supplemental Material, sj-pdf-11-rap-10.1177_20531680211061056 for Linking individual and group motives for violent conflict by Matthew Nanes in Research & Politics

Supplemental Material

sj-pdf-12-rap-10.1177_20531680211061056 – Supplemental Material for Linking individual and group motives for violent conflict

Supplemental Material, sj-pdf-12-rap-10.1177_20531680211061056 for Linking individual and group motives for violent conflict by Matthew Nanes in Research & Politics

Supplemental Material

sj-pdf-13-rap-10.1177_20531680211061056 – Supplemental Material for Linking individual and group motives for violent conflict

Supplemental Material, sj-pdf-13-rap-10.1177_20531680211061056 for Linking individual and group motives for violent conflict by Matthew Nanes in Research & Politics