From doctrine to detonation: Ideology, competition, and terrorism in campaigns of mass resistance

This study proposes an organizational mechanism that links ideology to the use of terrorism in mass dissident campaigns. Ideology affects the level of competition among factions within mass dissident campaigns by shaping whether actors see their interactions as a positive- or zero-sum game. We identify ideological diversity within a campaign and the degree to which ideologies embrace the principle of pluralism as key factors affecting the intensity of factional competition and, consequently, the occurrence of terrorism. We introduce new data on the ideologies of campaigns from the Nonviolent and Violent Campaigns and Outcomes 2.0 dataset and use causal mediation analysis to test our proposed mechanism. We find that greater ideological diversity within a campaign increases the likelihood of terrorism by increasing factional competition. We also find that the presence of a pluralist ideology is associated with a lower likelihood of terrorism by the lowering of factional competition. By shedding light on the mechanisms that link ideology to terrorism, this study helps advance our understanding of why dissident groups might decide to use terrorism tactics within the context of a campaign of mass resistance.


Introduction
How does ideology affect the occurrence of terrorist attacks in mass dissident campaigns?Previous studies have highlighted several elements of ideology that motivate armed groups to target civilians with coercive violence. 1Ideologies can make direct pronouncements that encourage such violence (Drake, 1998;2003;Hoffman, 1998;Juergensmeyer, 1997;Laqueur, 1999;Piazza, 2009;Simon & Benjamin, 2002), they can strengthen a sense of 'othering' that allows certain social groups to be seen as permissible targets of violence (Asal & Rethemeyer, 2008;Polo & Gleditsch, 2016), and they can determine a group's potential pool of supporters, thus altering the cost-benefit calculus of targeting civilians (Nemeth, 2014;Stanton, 2013).
Most of these ideological theories focus on either the behavior of single 'terrorist' organizations or the dyadic interactions between challengers and the state.On the other hand, dissident campaigns are often composed of multiple factions that may espouse diverse ideologies. 2  An alternative 'organizational' tradition of terrorism research highlights how factional competition can drive actors to terrorism by creating incentives for spoiling deals, outbidding and fratricide (Bloom, 2005;Chenoweth, 2010;Conrad & Greene, 2015;Crenshaw, 1987;Kydd & Walter, 2006;Oots, 1986).However, ideology has rarely played a central role in these analyses (Crenshaw, 1987: 13). 3his study melds these two traditions.It develops a new theory of the way in which ideology shapes the intensity of factional competition within campaigns and, consequently, the incentives for using terrorism.We identify two factors that are crucial in this regard: first, the level of ideological diversity within a campaign; and second, whether ideologies present in the campaign embrace the principle of pluralism.To test our argument, we introduce new data on the presence of different types of ideologies in mass dissident campaigns from the Nonviolent and Violent Campaigns and Outcomes (NAVCO) 2.0 dataset (Chenoweth & Lewis, 2013).We use causal mediation analysis to assess whether and to what extent ideology affects the occurrence of terrorism through shaping the levels of competition within mass dissident campaigns.
Our study promises several contributions to research on ideology, terrorism, and conflict.First, as described above, we synthesize the ideological and organizational traditions in terrorism research to develop novel hypotheses about the relationship between ideology and terrorism through factional competition.Our theory emphasizes aspects of ideology that have previously been overlooked.Previous work has emphasized concepts such as audiences (Polo and Gleditsch, 2016), othering (Asal & Rethemeyer, 2008), and extremism (Hafez, 2020;Walter, 2017).Instead, we highlight the importance of diversity and pluralism.
Second, we offer new data on ideologies within mass dissident campaigns.These data open the door to analyses that were not previously possible.By employing campaigns as the units of analysis as opposed to single terrorist groups, we are able to examine how the ideological landscape of a campaign -both the number of different ideologies present in a campaign as well as characteristics of those ideologies -shapes the likelihood of terrorism occurrence.
Furthermore, the data allow us to include cases of 'civil resistance' campaigns that feature significant popular mobilization but levels of violence that fall short of traditional civil war thresholds.Such campaigns have been the most common form of mass dissent since 1980 (Chenoweth, 2020: 71), and 15% of them involved the systematic use of terrorist attacks (Belgioioso, 2018).By examining all types of mass dissident campaigns, we explore an important new set of conflict cases in which terrorism is common.We are also able to avoid bias that may result from looking only at conflict cases in which armed violence has already become the primary strategy of dissent.
Finally, the use of causal mediation analysis allows us to estimate the extent to which observed relationships between ideology and terrorism can be explained by our proposed organizational mechanism as opposed to alternative mechanisms.Understanding the mechanisms through which ideology influences the likelihood of terrorism is important both for improving our theoretical understanding of why terrorism occurs as well as for designing policy interventions that try to prevent it.

Ideology, factional competition, and terrorism
Research on the organizational approach to conflict demonstrates that dissident groups are not focused solely on the goal of victory against the state.They must also pay attention to their position and power vis-a `-vis other dissident factions (Crenshaw, 1987;Oots 1986).If a dissident campaign achieves victory or reaches a settlement with the state while another faction is in a subordinate position of power within that campaign, the subordinate faction is unlikely to see its interests realized in a post-conflict political order.Therefore, dissident actors are engaged simultaneously in a dual struggle: the 'wars of movement' against the state, and the 'wars of position' among rival dissident factions (Gramsci, 1971;Krause, 2017).Although factional competition has been extensively examined in the context of armed civil conflicts, studies have shown parallel dynamics of competition that create similar incentives for the use of violence and even terrorism in civil resistance campaigns (Belgioioso, 2018;Pearlman, 2011).
The use of terrorist attacks can serve three main purposes for factions seeking to strengthen their position visa `-vis rival factions.First, terrorism can be used to 'spoil' efforts directed at a negotiated settlement between the state and a rival faction (Kydd & Walter, 2002;Pearlman, 2009;Stedman, 1997).By targeting civilians with terrorist violence in the midst of such negotiations, a dissident group can undermine the state's trust that the rival can fulfill commitments to ensure peace.The faction hopes that the use of terrorism will prolong the conflict, allowing it more time to enhance its own position vis-a `-vis its rival and to force the state to negotiate with it rather than the rival faction.
Second, a dissident faction may use terrorism in an effort to 'outbid' rivals for attention and support (Bloom, 2005;Kydd & Walter, 2006).By engaging in what is often perceived to be the most extreme form of violence, the faction attempts to win support and notoriety by drawing attention to itself and framing itself as the group most committed to the cause.
Third, dissident factions may engage in 'fratricide', in which they directly attack the members or supporters of other factions (Gade, Hafez, & Gabbay, 2019;Hafez, 2020;Mendelsohn, 2021;Pischedda, 2018).In doing so, they not only physically weaken their rivals, they also intimidate others from joining or offering support to those rivals. 4owever, engaging in factional competition comes at a cost.It increases the risk of losing the conflict against the state by weakening other dissidents and diverting resources.Therefore, dissident actors need to weigh expected gains from fighting wars of position against potential losses in the war of movement.What factors influence this calculus?Previous work has emphasized the availability of resources (Fjelde & Nilsson, 2012), the overall number of dissident organizations (Conrad & Greene, 2015), and the relative distribution of power among those organizations (Krause, 2017).
We claim that ideology plays an important role in the trade-off between wars of movement and wars of position.Factions within a campaign are likely to avoid intracampaign conflict when they can conceive their interactions with other factions as a positive-sum game.That is, when actors perceive other factions' success as compatible with the advancement of their own goals and interests, they are more likely to cooperate.However, when they perceive dynamics to be zero-sum, in which others' success comes at their expense, they will probably turn to competition.Two elements of ideology are likely to influence these positive-vs.zero-sum dynamics: the extent of ideological diversity within the campaign; and whether ideologies within the campaign either embrace or reject pluralism.

Conceptualizing ideology
For the purposes of this study, we follow Gutie ´rrez Sanı ´n & Wood (2014: 214) in defining ideology 'as a set of more or less systematic ideas that identify a constituency, the objectives pursued on behalf of that group, and a program of action'.We also follow several previous studies in considering ideology in terms of broad general classes, sometimes referred to as the 'big-isms' (Asal & Rethemeyer, 2008;Drake, 1998;Piazza 2009;Polo & Gleditsch 2016).Although there is certainly potential for variation in the specific ideological vision within each of these categories, we believe that these broad classes sufficiently capture key attributes of ideology most relevant to our theory.
Our ideological schema is presented in Table I.The ideological categories are not mutually exclusive; we code Liberal democratic Advocates a political order based on electoral competition or greater adherence to principles of free and fair elections and accountability.

Yes
Religious Advocates a political order in which a particular religious interpretation is to be used as the basis for governance and law.

No
Right-wing Advocates the restoration or maintenance of a hierarchical and authoritarian political and social order.Ethnonationalist Advocates a political order tied to a particular identity group (ethnic, religious, racial, tribal, national, or other).

Leftist
Advocates the rearrangement of class order and redistribution of wealth.Usually explicitly invokes the Marxist, Leninist, and/or Maoist label.

Unknown
Non-ideological Advocates a change in state leadership to another set of leaders with no claims of a different kind of political order.
multiple ideologies as present simultaneously within a campaign when appropriate.We designate each ideological category as promoting pluralism, not promoting pluralism, or as 'unknown' when an ideology's commitment to pluralism is either unclear or varied.In the theoretical discussion that follows, we elaborate on our definitions of these ideologies as well as how diversity and pluralism shape competitive dynamics.

Ideological diversity
Competing visions of the future political order are likely to exacerbate conflict within a dissident campaign.The basic premise that ideological divergence increases the potential for factional competition is frequently suggested in the literature (Maynard, 2019: 638;Zald & McCarthy, 1980: 12).This is not to say that ideologically homogenous campaigns will be free from in-fighting -far from it.Personality conflicts, concerns about the distribution of resources, and even political disagreements under the umbrella of a single ideological category can motivate factional competition (Ron, 2001).However, ideological difference is likely to make such competition even more intense.When a mass dissident campaign includes factions that embrace different ideologies, dissident groups are concerned that success for rival factions -in terms of garnering media attention, tactical victories, or coercing concessions from the state -will stop the progression of their own political goals.Under such circumstances, 'ideological opposition becomes especially salient [ . . .] because the incompatible interests of the respective groups cause competition for public influence to be a zero-sum game' (Chenoweth 2010: 20).Actors struggle to make credible commitments to each other because they have rallied their own supporters around ideas and goals that conflict with those of other factions (Gade, Hafez & Gabbay, 2019: 324).
Competition may be most visible when a campaign consists of a coalition of distinct named organizations that espouse different ideologies.However, it may also occur within single organizations that embrace multiple ideologies.In these cases, disputes can arise between factions of the same organization that prioritize one ideology over the other.For example, in both the Irish Republican Army and Euskadi Ta Askatasuna, debates about the relative importance of socialism vs. nationalism led to the formation of internal factions and, eventually, formal organizational fragmentation (Carthew, 1971;Perkoski, 2019: 882;Sullivan, 2015: 129).Ideological diversity within a campaign can thus create incentives for competition irrespective of how competing factions map onto formal organizational structures.
These incentives for competition are likely to increase in line with the total number of different ideologies present within the campaign.The more ideologies present, the smaller the possibility that a particular political vision will be implemented if the state decides to pursue negotiations.More factions will feel at risk of losing out on their priorities even if the campaign as a whole is successful, increasing the importance of relative positioning within the campaign.
With such uncertainty that victory for the campaign will result in the realization of a particular faction's ideological goals, winning the war of position vis-a `-vis other factions becomes increasingly important for all groups.Hindering the campaign against the state may even be beneficial in order to buy more time to obtain a position of primacy within the campaign.Targeting civilians with coercive violence may be an effective tactic toward this end.A faction might use terrorism to try to spoil an emerging deal between a rival faction and the state that would sideline its ideological platform.It could engage in fratricidal terrorism, targeting the rival's supporters in an effort to intimidate others and stop them from associating with that group.Or, it could use terrorism in an effort to outbid rivals and build support for its own ideology by building a reputation as the most aggressive entity confronting a broadly despised adversary.
Ideological differences have featured prominently in many case studies of factional rivalry leading to terrorism.For example, the use of suicide bombings by Islamist factions in Palestine is frequently explained as an effort to both derail peace negotiations with Israel and gain popular support at the expense of the secular nationalist Fatah (Bloom, 2004;Kydd & Walter, 2002;Pearlman, 2009).Boyle (2009) also cites competition and outbidding among nationalist, separatist, and religious factions in Iraq's civil war as a driver of the especially high levels of terrorism in that conflict.
By contrast, when all factions share a common general ideology, interactions are less likely to be perceived as zero-sum.Other factions' successes can benefit the campaign as a whole as they advance a common ideological vision.As Gade, Hafez & Gabbay (2019: 324) write, 'groups with shared conceptions of the ideal polity corroborate each other's core political preferences and thus can readily signal to their ideological kin their intentions to share power in the post-conflict political order'.In such circumstances, defeating the state can take priority over competing with rivals, making outbidding, spoiling, and fratricide counterproductive.
To our knowledge, the basic premise that ideological diversity increases factional competition and, therefore, makes terrorism more likely has not yet been systematically tested across cases.We present this as our first hypothesis.
Hypothesis 1: Greater ideological diversity within a campaign increases the likelihood of terrorism occurrence by increasing levels of factional competition.
Similarly, we believe our hypothesis about the relationship between ideological diversity, competition, and terrorism is likely to apply to both campaigns that employ primarily armed and unarmed strategies.It could be argued that civil resistance campaigns could be better able to handle the presence of multiple ideologies than those that utilize primarily violent methods because the former foster a general culture of democratic exchange and pluralism (Chenoweth & Stephan, 2011: 207).In our view, this is not an inherent feature of campaign strategy, but is instead tied to specific ideologies, which may be present or absent in both types of campaign.We develop this argument further in the next section.

Ideological pluralism
Beyond the total number of ideologies present in a dissident campaign, the presence of specific ideologies may exacerbate or mitigate factional competition.Specifically, we argue that the presence of an ideology that embraces the concept of pluralism can mitigate competition among factions by providing a framework for positive-sum interactions.In using the term pluralism, we mean specifically that the ideology recognizes the existence and legitimacy of multiple political values (Blattberg, 2001: 198).Therefore, a pluralist ideology emphasizes 'procedural justice', prescribing institutions and processes rather than specific outcomes (Hampshire, 2000: 37).In doing so, it opens up bargaining space for conflict resolution between dissident groups.The pluralist approach allows different factions to keep a clear focus on their interests, rather than positions (Fisher & Ury, 1981).Furthermore, it offers institutional mechanisms for arbitrating outcomes of interest through an objective criterion such as elections.This allows all factions -even those espousing ideologies that are not pluralist -to focus on the common interest war of movement.
Liberal democratic ideology is unique, because it is the only ideological category in our conceptual scheme that broadly and consistently promotes pluralism.The goal of groups with liberal democratic ideologies is precisely to put in place governance structures such as elections, proportional representation, and rule of law that create institutionalized mechanisms to resolve disagreements about policy outcomes.This vision for a postconflict political order has ramifications for factional negotiations during the heat of conflict.Because democratic institutions allow for the possibility of most factions achieving their substantive goals in terms of outcomes, this creates a positive-sum environment, which reduces the importance of relative position among factions with different ideologies.
Being able to maintain cooperation among campaign factions is especially likely if all share a commitment to liberal democratic ideology.However, the presence of liberal democratic ideology is likely to have an overall cooperation-inducing effect even when other ideologies are present.This is because it offers a mechanism for disagreements about future political outcomes to be resolved institutionally at a later point in time.For example, in 2006, prodemocratic political parties in Nepal formed an alliance with a Maoist insurgent group to overthrow the monarchy.The Maoists, who had previously used terrorism in their campaign against the state, agreed to abstain from terrorist violence during the joint campaign and to use future elections as a means of determining which faction would hold power (Adhikari, 2014;Thurber, 2019Thurber, , 2021)).
Religious, right-wing, and ethnonationalist ideologies do not similarly promote pluralism.They advance political visions that offer little space for the inclusion of alternatives.Therefore, they must prioritize achieving primacy within the campaign, because a failure to do so imperils their ability to enact their political vision in the aftermath of conflict.Campaigns that include only these ideologies are more likely to experience factional competition that leads to terrorism.
We define religious ideology as advocating a political order in which a particular religious interpretation is to be used at the basis for governance and law.Brubaker (2015: 5) argues that conflicts in which religious ideologies are present differ from others because the stakes involve 'distinctly religious understandings of right order that are held to be binding for all in the wider society and polity'.This focus on the substantive and static regulation of public life according to a specific reading of holy texts and the need for this to be applied universally, offers little room for consensus with other factions.
Similarly, we consider that right-wing ideology does not promote pluralism.Although definitions of rightwing ideology have often included many different 338 journal of Peace Research 61(3) components (Mudde, 1995), we follow Carter (2018) in using a minimal definition centered on the desire to either bring back or maintain an authoritarian political system and a traditional social hierarchy.Thus, rightwing ideology is antithetical to pluralism in its explicit opposition to the opening up of the polity to greater participation and viewpoints.This offers little room for compromise with any factions that espouse a different vision of the future social and political system.
Finally, we also categorize ethnonationalist ideologies as not pluralist.We define this ideology as one that seeks to define the state according to a particular ethnic group or set of groups at the exclusion of some others (Asal & Rethemeyer;2008).Ethnonationalist ideology rejects pluralism in that it focuses on rights and privileges that are conferred on a single group on the basis of ethnic, national, religious, or some other form of identity.
In the context of ideologically diverse campaigns, competition will be more likely in the presence of factions that do not embrace pluralist ideologies, exacerbating the dynamics anticipated by our first hypothesis.Even if achieving hegemony within the campaign is not possible, using terrorism during the course of the campaign allows these groups to signal to other factions how difficult it would be to govern in the future.The arrival of Jihadists in Chechnya, for example, sparked competitive pressures that drove both religious and nationalist factions to employ terrorist tactics (Bakke, 2014).
However, even when factions share a common nonpluralist ideology, competition is likely to emerge with regard to the proper interpretation and implementation of that ideology.For example, when imprisoned al-Gama'a al-Islamiyaa leaders negotiated a 'nonviolence initiative' with the Egyptian government in July, others within the group declared the agreement to be a betrayal of core ideological tenets.In response, they organized a terrorist attack on tourists in Luxor in an attempt to scuttle the deal and to position themselves as the faction most committed to the Islamist cause (Nemeth, 2014: 341).Similarly, factions espousing an ethnonationalist ideology are likely to come into conflict both with other ethnonationalist factions representing other ethnic groups, as well as with factions favoring other ideologies who seek to advance a political vision that does not grant special status or autonomy to members of a particular group (Pischedda, 2018).Consistent with this logic, Phillips (2019) finds that ethnonationalist groups are more likely to engage in violent rivalry with other factions.
In summary, we claim that the presence of a pluralist ideology within a campaign mitigates competition among all factions, including those that do not espouse a pluralist ideology.This is because by focusing on producing institutions and processes rather than specific outcomes, an ideology that embraces the ideal of pluralism provides a framework for positive-sum interactions that allow all other ideological factions to focus on the common war of movement while delaying disputes over the nature of a future political order until after victory has been achieved.
Hypothesis 2: The presence of pluralist ideology within a campaign decreases the likelihood of terrorism occurrence by lowering factional competition.
Our schema leaves two ideological categories -leftist and non-ideological -as unknown in terms of their promotion of pluralism.By this, we mean simply that we are unable to determine any obvious commitment to pluralism within the entire ideological category.For example, some leftist groups have cooperated with religious or ethnic-based dissident organizations, proposing communist political institutions as a mechanism for empowering historically marginalized ethnic groups.However, other leftist factions, such as the Sandinistas in Nicaragua, have viewed the embracing of social identities other than class as incompatible with leftism (Cleary, 2000(Cleary, : 1139)).
Similarly, non-ideological campaigns, in which no larger ideological doctrine appears to be present other than a practical stated goal of regime change, cannot be deemed to be broadly either for or against pluralism.One way of viewing this is as a missing data problem; although it might be possible to assess the commitment to pluralism on a case-by-case basis, we are unable to assess a commitment to pluralism in leftist or nonideological categories generally.As such, we do not make specific theoretical predictions about campaigns in which only these ideologies are present.

Measures and descriptive patterns
To test our hypotheses, we gather data on ideologies, competition, and terrorism within mass dissident campaigns contained in the NAVCO 2.0 dataset (Chenoweth & Lewis, 2013).NAVCO 2.0 includes year-level observations of mass dissident campaigns seeking maximalist goals such as regime change, institutional reform, major policy change, territorial secession, or greater autonomy and the removal of occupying forces in independent states.To be included, a campaign must achieve a threshold of 1,000 observed participants on at least two occasions and use tactics that are outside the bounds of institutionalized politics.The campaign is characterized as primarily 'nonviolent' when the dissidents are unarmed civilians, and primarily 'violent' if the campaign is characterized by the overt participation of armed dissident groups (Chenoweth & Lewis, 2013: 418).Crucially, the coding of a campaign as primarily nonviolent does not require the complete absence of lower-scale violent tactics, including terrorist attacks.We maintain the original campaign-year unit of analysis.By focusing on the campaign (as opposed to specific organizations), it is possible to analyze the presence of multiple ideologies and the intensity of interorganizational competition, both of which are central to our theory.

Dependent variable
Our dependent variable, Terrorism occurrence, is a dichotomous variable drawn from Belgioioso (2018) that indicates the presence or absence of systematic terrorist attacks within a mass dissident campaign-year.Belgioioso's operationalization of terrorism is adopted from the Global Terrorism Database (GTD) (START, 2012), which uses three basic coding rules and three additional criteria to identify terrorist attacks: attacks must be intentional; they must entail the use of violence or the threat of violence; they must be aimed at political, economic, or social goals; they must have the intention to coerce or intimidate, or to transmit a message to a larger audience than the immediate victims; they must violate international humanitarian law in that they target civilians or non-combatants; and they must be perpetrated by non-state actors. 5 GTD terrorist attacks are attributed to mass dissident campaign-years when they are perpetrated 'by actors engaging in mass dissent and which share the dissident campaigns' broad political goals' (Belgioioso, 2018: 647).Participation in mass civil resistance is established when groups (1) contributed to coordinating the emergence of nonviolent mass movements and/or (2) took part in the broader coalition waging mass civil unrest.Furthermore, for the terrorism occurrence variable to take a value of 1, the use of terrorism must be systematic.That is, at least three terrorist attacks must occur within a year from the first attack.Systematic terrorism occurs in almost 20% of the observed years of mass dissident campaigns.We believe that a dichotomous indicator is more appropriate than count events, because we are not interested in the intensity of terrorism, and the severity of attacks in any event is not measured properly by the number of attacks.

Core explanatory variables
Our explanatory variables come from our own original coding of the ideologies present in NAVCO 2.0 campaigns.For each campaign-year, we first code for the presence of any of the seven ideologies in our conceptual schema as listed and defined in Table I.We based the coding of ideologies on articulated claims and public statements, and identify the presence of an ideology when at least two sources report evidence that such an ideology was embraced by an actor participating in the campaign.We count claims and statements that are linked either to the campaign as a whole or to any faction that is identified as a participant in the campaign.Ideologies are not mutually exclusive and, therefore, multiple ideologies might be observed simultaneously within a campaign.Furthermore, by coding at the campaign-year level, we account for changes over time when groups embracing a specific ideology join or exit a campaign.Such changes, however, are relatively infrequent.
After coding for the presence or absence of each ideology within each campaign-year, we create core explanatory variables that operationalize the concepts articulated in our hypotheses.To test the first hypothesis, Ideological diversity is a count of the total number of different ideologies present within the campaign in that year.To test the second hypothesis related to pluralism, we create a set of mutually exclusive dummy variables.The first dummy, Pluralist ideology, takes the value of 1 when we observe the presence of liberal democratic ideology within a campaign, and the value of 0 when liberal democratic ideology is not present (i.e.all ideologies in the campaign are either nonpluralist or undetermined).
Our second dummy, Unknown, takes the value of 1 when the only ideologies observed are ones with indeterminate levels of pluralism (i.e.leftist or non-ideological), and the value of 0 when any pluralist or nonpluralist ideology is present in the campaign.As explained above, we do not expect any specific effect of 5 Belgioioso's data cover the years between 1948 and 2006.Information on terrorist attacks from 1970 onward was drawn directly from the GTD.Prior to 1970, Belgioioso's data come from Lexis Nexis news reports, using parallel inclusion criteria.Importantly, '[b]attle-related deaths accounted for in violent dissident campaigns do not include deaths due to attacks deliberately directed against civilians' (Belgioioso, 2018: 646-47) and, therefore, problems in relation to the double counting of casualties are absent.
leftist ideology or non-ideological factions on the competition within campaigns, and so we effectively separate from our analysis those campaigns in which only these ideologies are present.
Finally, No pluralist ideology takes the value of 1 when we observe the presence of religious, identity-exclusive, or right-wing factions in the absence of a pluralist ideology, and the value of 0 in all other cases, in other words, when either of the other dummies is equal to 1.Our expectation, in line with Hypothesis 2, is that campaigns in which a Pluralist ideology is present will experience a lesser likelihood of terrorism occurrence than those with No pluralist ideology.
Table II shows the frequencies with which Pluralist ideology, No pluralist ideology, and Unknown were present among 1,485 campaign-years, and the frequency of terrorism occurrence according to the presence of each given ideology.Consistent with our expectations, campaigns that included a Pluralist ideology experienced terrorism in only about 8% of campaign-years, whereas terrorism occurred in about 22% of campaign-years with No pluralist ideology.In the analysis that follows, we will attempt to measure how much of this difference can be attributed to factional competition.

Mediator variable
To measure the Degree of competition within mass dissident campaigns, we use NAVCO 2.0's 'camp_conf_intensity' variable.This is an ordinal variable that ranges from 0 to 3 based on whether factions within a campaign are seemingly united, exhibit moderate disunity (i.e. policy disagreements), engage in verbal or active competition short of physical violence, or target each other with open violence (Chenoweth & Lewis, 2013: 6).It is important to note that the open violence coded in 'camp_conf_intensity' differs from terrorist violence captured by our dependent variable, that is, terrorism occurrence.Violence between factions, as measured by the degree of competition variable, includes actions that aim to openly and directly damage the resources of the opponents.Terrorism, by contrast, entails underground attacks against civilian targets. 6Table III summarizes how the occurrence of terrorism varies according to degree of competition and shows that, as expected, terrorism occurs more frequently in campaign-years that experience higher levels of factional competition.

Potential confounders
Because both dissident ideology and the use of terrorism may be correlated with many structural factors and campaign-specific characteristics, we control for several potential confounders.We control for the primary mode 6 As an alternative measure of dynamics of competition within a campaign we generate the dichotomous variable Unity, which is equal to 1 when the campaign is seemingly united or characterized by disagreement about policies and equal to 0 when evidence of competition between groups emerges through statements or open violence.We report the results in the Online appendix, Figures 1-3, which are nearly identical to our main findings.
of dissent, that is, Primary resistance method.This is to ensure that the effects of ideology on terrorism are not simply an artifact of whether the campaign is primarily violent or nonviolent.As explained above, mass dissident campaigns are considered primarily nonviolent '1' based on the primacy of nonviolent resistance methods and primarily violent '0' when participants use primarily physical force through the use of arms (Chenoweth & Lewis, 2013: 418).
A second potential confounder is the goal of mass dissident campaigns.Ideology, in our conceptualization, is distinct from 'goals' defined in the NAVCO 2.0 project as the broad objectives of campaigns (i.e.regime change or secession).Rather, ideology captures the expressed ideal justification or objective articulating why that goal is necessary and how the new political order might differ in the event of that goal being achieved.Certain ideologies might be closely correlated with certain goals, but it is possible for campaigns with a common ideology to pursue different goals in different contexts.For example, the vast majority of campaigns with secessionist goals might have an antipluralist ideology, but there are 21 campaigns that included an antipluralist ideology and that sought the goal of regime change. 7We extract the goals of mass dissident campaigns using the variable 'camp_goals' included in NAVCO 2.0.We make one alteration to the original coding of Chenoweth and Lewis in that we collapse Significant institutional reforms and Regime change into a single category due to the paucity of cases in the former category and conceptual similarity between the two.
We include a measure of the number of new organizations within mass dissident campaigns extracted from NAVCO 2.0 using the variable 'camp_orgs'.In bigger, fast-growing campaigns, dynamics of competition and outbidding might be exacerbated, in turn increasing the risk of terrorism.By controlling for the number of new organizations, we ensure that any relationship between ideology, competition, and terrorism is not simply the result of an increasing density of dissident organizations.
We control for the Duration of campaigns, obtaining this measure by generating a count of the years elapsed since the start of the individual mass dissident campaign.
Protracted mass dissent might exacerbate internal organizational pressure, and this is likely to motivate leaders to initiate terrorist campaigns to secure organizational survival (Belgioioso, 2018).
We also account for Repression and we extract this from NAVCO 2.0.Repression against mass tactics may push the dissent underground and spur organized violence such as terrorism (della Porta, 1995).The level of internal repression against mass movements might also vary, depending on governments' external dependency and expectation of international sanctions (Davenport, 1995).In turn, external support might vary depending on the ideology of campaigns, and this was especially true during the Cold War.
Research on the structure of political opportunity and resource mobilization suggests that democracies and anocracies are a favorable environment for the use of terrorist campaigns (Eubank & Weinberg, 1994;Li, 2005;Schmid, 1992).Democracies may be better equipped to absorb challenging extra-institutional political demands connected to democratic reforms, thereby reducing the incidence of these type of campaigns in the first place.We thus control for Autocracy via a dichotomous item for countries with values of -5 or below on the polity2 score from the Polity IV Project (Marshall, Gurr & Jaggers, 2014).As an alternative stricter measure for Democracy, we also control for another dichotomous item equal to 1 for countries with values of 1 or above on the polity2 score.
We include a measure of total population (logged) from Gleditsch (2002).A larger population is associated with higher use of domestic terrorism (Savun & Phillips, 2009).States with a larger population might be more likely to experience the emergence of antipluralist ideology.For example, states with larger populations might be more likely to experience the emergence of identity-exclusive minority groups expecting to be able to gain independence from the central state.Widespread poverty creates grievances that might cause people to resort to terrorism (Crenshaw, 1981).In addition, poverty tends to affect particularly minority groups that might be more likely to endorse an antipluralist ideology when mobilizing mass resistance campaigns.Thus, we control for a country's logged GDP per capita using data from Gleditsch (2002).

Causal mediation analysis
We seek to demonstrate not only that ideology is correlated with terrorism, but that its effect passes through dynamics of factional competition occurring within campaigns.To this end, we employ causal mediation analysis (Imai et al., 2011), which moves beyond the calculation of total average treatment effects and instead seeks to quantify the effect of a treatment that operates through a particular mechanism.In other words, it allows us to estimate how much of any relationship between our ideology variables (T) and terrorism occurrence (Y) is explained through the intervening variable degree of competition (M) (see Figure 1).
We use Hicks & Tingley's (2011) 'mediation' package to perform the analysis.This package implements the potential outcome framework, which has two important advantages over the more traditional structural equation modeling.First, it allows for the estimation of the average causal mediation effect (ACME) and the average direct effect by using a non-parametric identification strategy.Second, it allows us to formally evaluate the robustness of our findings in relation to potential violations of underlying assumptions.This is particularly important because the additional assumption needed for the ACME to be unbiased, that is, the sequential ignorability assumption, might be violated both with experimental and observational designs (Imai et al. 2011).The package first estimates a regression model describing the relationship between the treatment and mediator variables.Next, it estimates a regression model for the outcome variable including both the treatment and mediator.Then, it generates two sets of predictions for the mediator, one under the treatment and the other under the control.First, the outcome is predicted under the treatment using the value of the mediator predicted in the treatment condition.Second, the outcome is predicted under the treatment condition using the mediator prediction from the control condition.The ACME is then computed as the average difference between the outcome predictions using the two different values.Finally, a bootstrap approximation based on the asymptotic sampling distribution (King, Tomz & Wittenberg, 2000) is used to compute statistical uncertainty. 8s presented in Table IV, each exposure-mediator model is an ordinary least squares (OLS) and shows the relationship between our core explanatory variables and the continuous measure of degree of competition, accounting for the average effect of all relevant control variables.The mediator-outcome models are probit models in which terrorism occurrence is the outcome, and competition along with the ideology variables is included on the right-hand side of the equation together with the full set of control variables.Including the ideology variables in the mediatoroutcome models allows us to distinguish between direct effects of these ideologies on terrorism and the mediated effect that is the result of ideology's impact on competition.It also ensures that our estimates for ideological diversity control for the presence of specific ideologies.The three pairs of models presented in Table IV vary only according to which of our three dummy variables related to pluralism is omitted to serve as the reference.
If our hypotheses are correct, we would expect to see statistically significant correlations between ideology and competition in the exposure-mediator models, and between competition and terrorism in the mediator-outcome models.Relationships between ideology and terrorism in the mediator-outcome model reflect effects of ideology on terrorism through mechanisms other than factional competition.Our theory can be evaluated even more clearly by viewing the estimated average mediated effects as presented in Figure 2. Here, we expect the mediated effect to be in the direction anticipated by our hypotheses and with confidence bounds that do not cross zero.Both the coefficient and marginal effects estimates reveal findings consistent with our theoretical expectations.
Figure 2 shows that the mediated effect of ideological diversity on the likelihood of terrorism occurrence through competition is positive and significant, providing evidence for our proposed mechanism. 9Tabular results confirm that ideological diversity significantly increases the degree of competition within dissident campaigns (p ¼ 0.000), which in turn increases the likelihood of terrorism occurrence (p ¼ 0.016).Consistent with our second hypothesis, Figure 2 shows that campaigns in which a Pluralist ideology is present are associated with a negative and statistically significant mediated effect on terrorism related to factional competition, compared with campaigns that have No pluralist ideology.The models in Table IV show that pluralist ideology significantly decreases the degree of competition within mass dissident campaigns (Model 2a, p ¼ 0.001) and that competition significantly increases the likelihood of terrorism occurrence (Model 2b, p ¼ 0.016).Models 3a and 3b show the parallel inverse relationship for non-pluralist campaigns.Figure 2 also shows a statistically significant direct effect for the pluralism variables.This represents the effect of ideology on terrorism unrelated to factional competition.This finding is consistent with existing theories that focus on audiences, norms, or othering (e.g.Asal & Rethemeyer, 2008;Polo & Gleditsch, 2016).However, our finding of a significant mediated effect shows that ideology has an impact on the likelihood of terrorism through dynamics of factional competition as well.

Disaggregating pluralist and non-pluralist campaigns
We offer a harder test of our second hypothesis in the Online appendix (Table 3, Figure 4), in which we disaggregate the effect of pluralist ideology, separating cases in which pluralist ideology occurs in absence of no pluralist ideology (Pure pluralist ideology) from cases in which pluralist ideology occurs contemporaneously with at least one non-pluralist ideology (Mixed ideology).Our theory predicted that by providing a framework for positivesum interactions across factions, the presence of a pluralist ideology could reduce terrorism even in the presence of other non-pluralist ideologies.Our findings show this to be the case, with Mixed ideology and Pure pluralist ideology both yielding negative and statistically significant mediated effects compared with No pluralist ideology campaigns.
To assess whether the observed patterns are driven by one particular non-pluralist ideology, we run models estimating the effects of each individually.The results are presented in Figure 3.Although the direct effects of each of these vary, the mediated effect on terrorism through competition relevant to our theory is positive and significant for religious, right-wing, and ethnonationalist ideology.For each treatment of interest, the circle represents the average causal mediation effect of a given treatment on the likelihood of terrorism mediated by the degree of competition within mass dissident campaigns.The square represents the average direct effect of a given treatment on the likelihood of terrorism keeping the mediator at its average.The horizontal lines represent 95% confidence intervals.The effects are estimated using Imai, Keele & Tingley (2010) with a linear regression for the mediator models and with probit for the outcome models.

Violent vs. nonviolent campaigns
Our theory predicts that the dynamics through which ideology shapes factional competition and, consequently, terrorism, will hold irrespective of whether a campaign is otherwise employing primarily armed or primarily unarmed methods of resistance.This is consistent with previous scholarship that has shown factional competition to be a driver of violence and terrorism even within campaigns that were largely nonviolent (Belgioioso, 2018;Pearlman, 2011).This is not to say that terrorism is equally likely across campaign types, although, certainly, it is plausible that terrorism is used with higher intensity in campaigns that are already using violent means once they have been adopted a first time.This is borne out by our results in Table IV.However, our assumption is that when wars of position become salient, factional competition is more likely to lead to the occurrence of terrorist attacks independently of whether a campaign is primarily nonviolent or primarily violent.
Our data give us the opportunity to test this assumption.To do so, we estimate nested models for the mediator-outcome models with interaction terms between primary method of mass dissent and degree of competition (See Online appendix, Table 23).Figure 4 shows the estimated relationship between competition and terrorism, disaggregated by the campaign's primary method of resistance.For both violent and nonviolent campaign types, we see a positive relationship between factional competition and the likelihood of terrorism.The relationship appears flatter for primarily nonviolent campaigns, although a test for difference in slopes reveals that the difference falls just short of traditional levels of statistical significance (chi square ¼ 2.60; p ¼ 0.106).

Robustness and sensitivity
We run a number of additional models, all reported in the Online appendix, to test the robustness of the findings in relation to changes in model specification and functional form.First, we run two new mediation models to test whether the effects of the presence of an antipluralist ideology depend on what specific baseline is used, that is, pure pluralist ideology or mixed ideology (Table 4, and Online appendix, Table 5, Figure 5).The main results for antipluralist ideology do not change depending on what baseline we use for the occurrence of pluralist ideology.We also rerun the whole causal mediation analysis with an individual measure of ideology on models, including controls exclusively for the main predictors (Online appendix, Tables 6-9, Figure 6).We then replicate this analysis using a simplified dichotomous variable for campaign Unity as an alternative mediator (Online appendix, Tables 10-18, Figures 7-9).We also use the simple sequential test proposed by Baron & Kenny (1986), which allows us to check the robustness of our hypothesized mediation effects on the alternative model specifications while introducing standard errors clustering by country.The results of these additional analyses are virtually unchanged from the results presented in the main text (Online appendix, Tables 19-22).
We then conduct sensitivity analyses to investigate the extent to which our conclusions are robust in relation to unobserved pre-treatment confounders using the 'medsens' command (Hicks & Tingley, 2011).Across the full models presenting significant mediation effects, an omitted variable confounder would have to explain the 10% of the total variation not yet explained by the observed predictors (Online appendix, Figure 10).
Finally, we test for potential reverse causality, specifically, whether the use of terrorism causes new factions to emerge, thereby shaping the level of ideological diversity.The results of this analysis (Online appendix, Table 24) show no evidence that the occurrence of terrorist attacks increases the number of new organizations in mass dissident campaigns or that new organizations in mass dissident campaigns have an effect on the degree of ideological diversity.

Conclusion
This article has drawn on the ideological and organizational traditions in conflict studies to propose a novel theory on how ideology has an impact on the likelihood of terrorism by shaping the dynamics of factional competition within a campaign.It introduced new data to test this theory across a broader set of conflicts than was previously possible and used causal mediation analysis to test the validity of this mechanism against alternative explanations.
Disentangling the mechanisms through which ideology influences the use of terrorism in conflicts is important for scholarly and policy purposes.In scholarly debates, ideological explanations for terrorism are frequently posited as competing with strategic and organizational approaches (Chenoweth & Moore, 2018).On the contrary, we find that they often work in conjunction with each other.Crucially, from a policy perspective, robust findings showing that ideology and terrorism are connected through dynamics of competition among factions suggest a potential for mitigating terrorism through peacebuilding and conflict resolution approaches, as opposed to efforts focused on countering extremist narratives that attempt the more ambitious task of changing individuals' or groups' ideologies altogether.
Our analysis of ideology at the level of the mass dissident campaign presents some limitations.It offers a unique opportunity for understanding how the constellation of ideologies within a campaign leads to overall levels of competition and terrorism, but it does not allow us to see which actors, espousing which ideologies, actually engaged in acts of terrorism.To be able to do so would require more precise data.Synthesizing an organization-level approach with campaign-level data on the presence of other ideologies would be a promising next step for further research.

Figure 1 .
Figure 1.Diagram representing the casual mechanism linking ideology to terrorism 10

Figure 2 .
Figure 2.Estimated average mediation effects and direct effects of treatments of interest on the likelihood of terrorism occurrence

Figure 3 .
Figure 3.Estimated effects of disaggregated No pluralist ideologies

Table I .
Ideological categories and definitions

Table II .
Frequencies of ideologies and terrorism occurrence observed in NAVCO 2.0 campaign-year data

Table III .
Terrorism occurrence by degree of competition within campaign