Rebel tactics and external public opinion

Rebel groups vary widely in the tactics they use to pursue political change, and they make these strategic choices weighing the pros and cons of specific tactics. Existing studies of contentious politics highlight the violent tactics rebel groups use (Salehyan et al., 2014; Thomas, 2014), as well as the consequences of these tactics (Stewart and Liou, 2017; Cohen, 2013). Different violent tactics can quickly bring new recruits (Stewart and Liou, 2017), increase local compliance (Kalyvas, 2006), and enhance group cohesion (Cohen, 2013). Yet, some violent tactics can decrease local civilian support in certain contexts (Lyall et al., 2013). Nonviolent tactics by rebels can lead to foreign recruitment or sponsorship (Coggins, 2014) or increase local civilian support (Flynn and Stewart, 2018). Many studies of tactical choice reflect an implicit (or at times explicit) belief that particular tactics will be viewed more or less favorably by both people on the ground and external observers (c.f. Chenoweth and Stephan (2011) on the legitimacy advantage of nonviolence), and that this favorability will translate into increased support for the actors using these tactics. Although a central part of arguments related to both violent and nonviolent tactics is how they elicit external perceptions (positive or negative), empirical work on this perceptual link remains limited.

In this article, we employ two survey experiments to map the response of American public opinion to specific rebel group tactics, providing survey respondents with randomized informational text varying the tactics used by the rebel group and then comparing how much support they profess for the group and its leader. This research increases our understanding of tactical choice by focusing on a particular audience (American civilians) and employing experimental design to narrow the focus on tactics of rebel groups rather than other possible confounders. We advance the growing body of experimental work on tactics by testing several unique expectations. First, we examine the expectation that specific nonviolent tactics will lead to increased support for rebel actors, examining four discrete nonviolent tactics across two experiments. Second, we probe the link between violent tactics and external public support shown in other studies (c.f. Flynn and Stewart 2018), but examine an additional lower-intensity tactic—stone throwing. Third, we look more specifically at “democratic” tactics, arguing that rebel use of the democratic process can increase support for the group from an American audience. To our knowledge, this is first experimental analysis that examines internal political process within rebel actors and public perceptions of these rebels.

In the first experiment (Study 1), we recruited a national online sample and examined how various tactics of resistance influence support for rebel groups amongst an American public.¹ We found that nonviolent tactics (such as demonstrations) increase support for rebel groups and the use of terrorism by rebel groups decreases support. We find that rebels’ use of elections internally is also associated with increased support for the rebel group. In the second experiment (Study 2), we focused explicitly on three contentious nonviolent tactics: demonstrations, nonviolent interventions (including blockades and sit ins), and social noncooperation (hunger strikes) to further explore the relationship between nonviolent tactics and support amongst an American public. Study 2 provides additional support for our expectations that nonviolent tactics (specifically hunger strikes and demonstrations) increase support for rebel actors.

Favorable perceptions of rebel groups by powerful external actors are often essential for rebel success. For separatists, international recognition depends largely on major power support (Coggins, 2014) and rebels “market” themselves to powerful outside patrons to obtain this support (Bob, 2005). Furthermore, rebel-controlled areas often require major power patronage to survive over a long period of time (Caspersen, 2009). Although the importance of “marketing” has been identified, it remains unclear exactly how different tactics play a role in the successful or unsuccessful “marketing” to major powers. This is, in part, because observational studies nearly always include a mix of different tactics employed by rebel groups.

Domestic public opinion in major power states can play an important role in rebel groups securing external support. The support of constituent populations within major powers works to shape policy with respect to supporting and recognizing rebels. In democratic states, for example, the public can constrain or reward foreign policy action (Tomz and Weeks, 2013), and can punish leaders retrospectively (Croco, 2011). Even in non-democratic states, public opinion amongst key constituent groups has influence on foreign policy choices that are important to leader survival in office.

Previous observational studies center on the calculated choice rebels make when employing certain tactics (Weinstein, 2006; Stanton, 2013). Rebel violence can generate more complete compliance and establish local order (Kalyvas, 2006), violence against civilians can facilitate resource extraction and recruitment (Stewart and Liou, 2017), the use of sexual violence can improve group cohesion (Cohen, 2013). To varying extents, studies have considered the implications of tactical choice for garnering positive external support. For example, terrorist violence can more credibly signal capability both domestically and internationally (Hoffman and McCormick, 2004). Others have shown violent tactics have domestic and international public support penalties (Lyall et al., 2013, Salehyan et al., 2014). Weinstein (2006) also suggests that selective violence or the use of less violence can lead to an “activist rebellion” with mobilization and recruitment based in social support.

Nonviolent tactics are argued to have more positive reputational and public support effects in general (Chenoweth and Stephan, 2011). In particular, use of nonviolent tactics by rebels can potentially offset the reputational cost of violence (Lyall et al., 2013).

While these studies theorize the costs and benefits that come from distinct tactics and provide observational evidence, only a handful of studies have begun to outline a clear perceptual link between specific tactics and support perceptions from specific audiences. At the domestic level, Lyall et al. (2013) show that harm by international forces (ISAF in particular), increased support for the opposition Taliban forces. Huff and Kruszewska (2016) examine perceptions of an actor based on its use of tactics and find the use of bombings by social movements has a negative impact on domestic public opinion towards them. Flynn and Stewart (2018) center on external public opinion (similar to this study) and show that rebel provision of social service has a favorable impact. Others have focused on perceptions of state tactics, including effective counter-terrorism (Matanock and García-Sánchez, 2017) and foreign development projects (Beath et al., 2011).

Experiments are well suited to bridging the gap in our understanding of specific tactics and public perceptions by establishing a clearer causal link between these tactics and public opinion. Our experiments allow us to test the effect of tactics used by rebel groups in the context of ongoing armed conflict and to see how these tactics influence American public opinion related to the rebel groups. It contributes to the extant experimental work on tactics specifically by considering a range of nonviolent tactics on external public opinion.

Building on the existing observational studies cited above, we expect there are reputational benefits of nonviolent tactics. However, not all nonviolent tactics are equivalent (Cunningham et al., 2017); different tactics (such as demonstrations versus hunger strikes) may resonate differently with publics. As such, we offer a series of hypotheses about discrete tactics. Although we do not explicitly theorize differential effects, we provide unique hypotheses and experimental treatments for each tactic.²

We also examine two hypotheses related to violence. A number of studies have focused on the use of terrorism. However, rebels employ a range of violent tactics, and those imbedded in established social movements also use lower-intensity violence. Stone-throwing has been particularly important in contexts such as the Palestinian/Israeli dispute, and we include this (as well as terrorism) in our study.

Building on insights from the growing literature on rebel governance, we also assess the use of democratic behavior by rebel groups. Cunningham and Sawyer (2019) show that elected rebel leaders are more likely to be negotiated with, arguing that rebel groups with such elections will be seen as more cohesive and viable bargaining partners by the state they face. “Democratic” tactics like election of leaders within the rebel group can impart legitimacy, as elections signal a commitment to their local population (and perhaps reflect support from them) and democratic action is often seen as a condition of group recognition by external states (Caspersen, 2015). Moreover, Matanock and Garbiras-Díaz (2018) have directly linked electoral process to legitimacy perceptions in their experiment on the responses to the FARC peace deal in Colombia.³

Our two experiments are oriented around the tactics and behavior by rebel actors. The actors of interest are engaged in ongoing disputes with the government that have led to casualties and constitute a challenge to the status quo. While often defined by their use of attacks against the state, “rebel” actors use a variety of violent tactics and these actors are defined and conceptualized in varying ways in the literature (c.f. Kalyvas and Balcells, 2010). Yet, rebel groups also engage in many behaviors that are not violent, including nonviolent direct action (Cunningham et al., 2017), providing public goods (Heger and Jung, 2017), and judicial processes (Binningsbø and Loyle, 2018). By framing the experiment about actors already considered “rebels,” we provide a hard test of the effect of tactics and behaviors that may increase positive external public opinion.⁴

We used Amazon’s Mechanical Turk (MTurk) online recruitment platform to conduct our first experiment on public perceptions. This platform is frequently used in social science experimental survey research (Huff and Tingley, 2015), because it offers a cost-effective way to recruit a participant pool that is more representative of the United States (U.S.) population than in-person convenience samples (Berinsky et al., 2012). Although MTurk samples tend to be younger, more educated, and more liberal than the general U.S. population (Huff and Tingley, 2015), they typically show similar effects in direction, significance, and magnitude in comparison to studies using nationally representative samples (Mullinix et al., 2015).

We conducted Study 1 in April 2017 and all participants were at least 18 years old, could read and write in English, lived in the U.S., had a satisfactory completion percentage of 96%+, and had completed more than 500 MTurk tasks. Participants were compensated on completion. We also collected demographic data on age, ideology, gender, education, and income. As expected with random assignment, chi-squared tests show that demographic characteristics were not statistically correlated with any of the experimental groups and the demographics of our participants are similar to other MTurk studies. We present the demographic breakdown and measures by experimental group in Online Appendix A.

We conducted Study 2 using students enrolled in political science courses at the University of Maryland in spring 2018. Although university students offer a less representative sample than our first study, we were better able to standardize the environment in which the participants completed the study.

We embedded both survey experiments in Qualtrics and asked participants to read a short “news story” (constructed by the authors) about potential ceasefire accords between a rebel group and a government. Study 1 used the Donetsk People’s Republic (DPR) and the government of Ukraine; Study 2 focused on the Afar group versus the government of Ethiopia. Varying the groups and regions used allowed us to assess the generalizability of our findings. We constructed the stories to look like BBC news articles to increase believability and external validity. All stories state similar background information on the conflict.⁵ We then experimentally manipulated the types of tactics used by the rebel group to determine how various tactics influence support for the rebel group and the government.

Participants in Study 1 were randomly assigned to one of the following treatments: (a) control (no mention of specific strategies); (b) terrorism, (c) stone throwing at cars, windows, and people; (d) elections carried out by the group; (e) small and large demonstrations; (f) nonviolent interventions, such as sit ins, occupations, or blockades; or (g) social noncooperation tactics, such as hunger strikes and self-immolation.⁶ The “news story” each group received in Study 1 was identical except for the tactic used by the rebel group.

Study 2 focused explicitly on nonviolent tactics, and participants were randomly assigned to one of the following treatments: (a) control; (b) small and large demonstrations; (c) nonviolent interventions (blockades and sit ins); and (d) social noncooperation tactics (hunger strikes). The stimuli in Study 2 were mostly similar to Study 1 stimuli except that we provided specific details about the tactics in these “news stories” to mirror the type of information news outlets include in their coverage of these types of stories. For example, the demonstrations vignette included that supporters “initiated a peace march throughout” the capital city of Addis Ababa “to create awareness” and “chant[ed] slogans and wav[ed] banners in the demonstration” to pressure government leaders. We did this to determine if specific types of nonviolent direct action mattered differently. In contrast to Study 1, we eliminated any mention of self-immolation to isolate the effects of a particular form of social noncooperation (hunger strikes). We did this because there is disagreement as to whether self-immolation counts as a form of nonviolence among nonviolent activists. All news stories are available in Online Appendix B and Online Appendix D. The experimental questions are available in Online Appendix C and Online Appendix E.

In the online experiment, we recruited 1533 participants to complete the study and the number of participants in each treatment ranged from 214 to 221. In the laboratory study, we recruited 459 participants to complete the study, and the number of participants in each treatment ranged from 114 to 115. The number of participants per group is shown in Table 1.

After reading the “news story,” participants answer the dependent variable questions. These questions assess participant attitudes towards the rebel group and its leader, by asking them to rate them on a five-point scale. The group question reads: “What is your overall opinion of the Donetsk People’s Republic?”⁷ and respondents can select: very favorable, favorable, neutral, unfavorable, or very unfavorable. The leader assessment is the same but with the leader named in the article replacing the group name. Question order is randomized to eliminate any potential order effects. In addition to standard demographic questions, we include an attention check question.

We use ordered logistic regression for our analysis because the question responses use a five-point ordered scale. To illustrate substantive effects of these treatments, we estimate and show the marginal effects. We report regression results for our attitude questions from Study 1 in Table 2. In Online Appendix A, we present regressions that include controls for political ideology, gender, age, education, and income.

We find some support for a relationship between nonviolent tactics and increased support of the rebel group and the group’s leader. The demonstrations treatment yielded a positive coefficient for both opinion of the group and the specific leader in the treatment. Social noncooperation (H1b) returned a positive, but not statistically significant, effect on support for the group or leader. The nonviolent intervention treatment produced a positive effect (as expected by H1c), but is not significant at the 0.05 level. As expected, the terrorism treatment returned a negative coefficient for both the group and the leader (H2a), and we find a positive but not statistically significant effect for stone throwing. Consistent with our expectation for H3, the use of internal elections in the rebel group increases positive opinion of the rebel group and its leader.

In Study 1, participants ranked their favorability towards the DPR on a five-point scale ranging from “very unfavorable” to “very favorable.” We asked participants, “What is your overall opinion of the Donetsk People’s Republic?”⁵ This question allowed us to capture overall support for the rebel group with respect to tactics they used. Figure 1 illustrates the differences in effect on the very favorabile outcome, showing the marginal effect of the different treatments. Use of elections in the group had the largest positive marginal effect, whereas terrorism has a negative effect.

We asked participants what they thought of the DPR leader, framed as “What is your overall opinion of Alexander Zakharchenko?” Figure 2 illustrates the differences in marginal effect on the “very favorable” rating across the treatments.

Study 2, conducted in the university laboratory, explored variation in nonviolent tactics. Demonstrations (H1a) have a strong and positive impact on opinion of both the rebel groups (DPR from Study 1 and the Afar in Study 2). We also find support for H1b, in favor of social non-cooperation tactics. The hunger strikes treatment leads to positive and statistically significant increase in opinions of both the group and the leader. This specific treatment did not show a significant effect in Study 1. The difference may be due to the inclusion of self-immolation in Study 1, and its exclusion in Study 2, as participants could view self-immolation as a form of violent action. The blockades and sit-in treatment returned a positive coefficient (as in Study 1), but is not statistically significant. Table 3 reports the coefficients for these models, and Figures 3 and 4 show the predicted effect based on the treatment group.

Similar to Study 1, participants ranked their favorability towards the Afars on a five-point scale ranging from “very unfavorable” to “very favorable.” We asked participants, “What is your overall opinion of the Afar group?” We found that nonviolent demonstrations and social noncooperation increased support for the rebel group but that nonviolent interventions did not change attitudes in a statistically significant way. Figure 3 illustrates the differences in marginal effect on the “very favorable” rating across the treatments for the Afar group.

We also captured attitudes towards the Afar leader by asking participants, “What is your overall opinion of Alimirah Hanfere?” We found the hunger strike treatment increased support for the leader. Demonstrations and nonviolent intervention (blockades and sit ins) both yield positive coefficients, but are not statistically significant at conventional levels. Figure 4 shows the marginal effects.

In sum, there is experimental support for the expectations that nonviolent tactics increase support for rebel groups (particularly demonstrations (H1a) and hunger strikes (H1b)) and that terrorist tactics decrease support (H2a). We also find support for the idea that democratic behavior among rebels is associated with higher favorability (H3). Variation in the effects of specific nonviolent actions suggests room for further research.

Experimental analysis allows us to isolate the effects of rebel behavior and tactics on external public opinion. Select other studies have addressed both local and external public opinion about rebel behavior. This study contributes to that growing research agenda by examining a novel set of tactics, including distinct nonviolent actions and internal democratic process. Our results suggest a strong role for rebel elections in eliciting positive perceptions of the group and the rebel leader. Nonviolent tactics also increase favorability of respondents, but there appear to be nuances across types of nonviolence. Terrorism leads to negative favorability of both rebel groups and their leaders.

Positive public opinion in powerful states can be a key part of rebel group success. Rebel groups court external public opinion (Bob, 2005), and there is increasing evidence that foreign policy may be more firmly rooted in public opinion than previously believed (Kertzer and Zeitzoff, 2017). Demonstrating the link between rebel behavior and external public opinion provides important evidence for the idea that what rebels do beyond violence can play a central role in their success. This is an area rich for future research into other tactics and their impacts on external public opinion, on how closely rebel leadership monitors and responds to these public attitudes, and on the variety of different support patterns between diasporas (and other external publics) and rebel groups.

The authors thank the editors and reviewers for their constructive feedback, as well as Megan Stewart and Sarah Croco. Thanks to Antione Banks for supporting the project through the University of Maryland Government and Politics Research Laboratory.