Votes and violence

Abstract

Many of the world’s most infamous terrorist organizations demonstrate clear political aptitude, maintaining highly successful political parties while simultaneously carrying out terrorist attacks. Yet the relationship between terrorism and a group’s political fortune is unclear. Groups like Hamas and Hezbollah appear to have gained significant support as a consequence of certain attacks, most notably those against US and Israeli targets. Other organizations fight for their political life after certain attacks. The Provisional Irish Republican Army (IRA) and its political wing, Sinn Fein, scrambled to restore its public image after bombs in Enniskillen, Northern Ireland, killed 11 Protestant civilians. In this article I examine the relationship between violence and political participation. I show that rebel groups are less likely to attack civilians when they simultaneously participate in democratic elections. I argue that attacking civilians is not good for political business. Not only can it distinguish the group as a terrorist organization and alienate supporters as a result, but attacking civilians also imposes high costs on the group’s own civilian support base. For these reasons, civilians frequently withdraw political support for rebel groups after they target civilians, which can be profoundly harmful to rebels. I analyze the violent and political behavior of non-state violent organizations from the Middle East and North Africa from 1980 to 2004. I also examine the IRA as a means of describing the causal mechanism advanced here.

Keywords

elections governance terrorism

Successful guerrilla operations involve the people. It is the quality of their resistance to the enemy and support for the guerrillas which in the end will be the decisive factor […] In fact, a guerrilla force will be unable to operate in an area where the people are hostile to its aims. And it must be remembered always that it is the people who will bear the brunt of the enemy’s retaliatory measures.

Irish Republican Army (1985: 17)

Why do some violent groups target civilians while others do not? Although terrorism, defined here as attacks by politically motivated violent non-state groups against non-combatant civilian targets, is a frequent news headline, its use by rebel organizations is not as consistent. For instance, contemporary rebel groups Hamas, the FARC (Colombia), and the New People’s Army (Philippines) employ terrorism as a proportion of their overall violence very differently – specifically, 45%, 27%, and 12% of the time, respectively.¹ Despite this variation, we are only beginning to understand what encourages or discourages terrorist tactics. Several studies suggest a strong willingness on the part of non-state actors to attack civilians in democracies because democratic leaders are particularly vulnerable to terrorist tactics (Chenoweth, 2010; Heger, 2010; Pape, 2005; Li, 2005). Yet, prominent exceptions compound the curiosity of this issue; many violent groups, including ETA (Spain) and the IRA (Northern Ireland), operate solely in democracies and rarely target civilians. Others, including Hamas, use terrorist tactics inconsistently over time.

This article explores how political participation shapes rebel incentives to use terrorism. I argue that groups actively participating in electoral politics are less likely to target civilians. My argument rests on two ideas: (1) rebels care about generating local-level support and (2) civilian-targeted attacks counter these ambitions. Yet while rebels have good reasons to pay attention to the preferences of non-member supporters, unless a mechanism exists for locals to sanction rebel groups, those preferences are unlikely to affect rebels’ behavior. Electoral participation, however, is one such mechanism. Thus, groups that participate in politics are more constrained by their constituents’ preferences. And, as I argue below, there are good reasons to believe that civilians disapprove of terrorist tactics, which translates into less terrorism when rebel groups participate in elections.

Argument

Rebels seek civilian support for a number of reasons including a sincere interest in reflecting a community’s preferences against an oppressive regime. Additionally, rebels can be strategically motivated to care about civilian preferences especially if motivated to contain information leakages. Civilians often know quite a bit about rebel organizations, including the group’s membership, when and where it holds meetings, and how it is financed. For clandestine rebels, protecting this information is paramount to survival. The same can be said of gangs. Akerlof & Yellen (1994) note that individual members of a community can be highly significant to a gang's success or failure because community members often have vital information about the gang. Berman et al. (2011) discuss how a bad economy may lead more community members to share information damaging to insurgents, particularly in the face of economic incentives. Rebels rely on communities to keep organizational information from authorities; it is crucial for survival. A constable stationed in South Armagh, Northern Ireland, fighting the Irish Republican Army (IRA) noted that:

[Government forces] can work without the people, whereas [an IRA unit] can’t […] they have to have the people on their side for intelligence, safe houses, safe place for weapons, cover if they were caught so someone could say: ‘Yes, he was with us all night’. (Harnden, 1999: 82–83)

Rebels also frequently rely on civilians to play an active role in operations. For example, Levitt (2004) notes Hamas’s smuggling operations benefit from civilian involvement:

Hamas is known to use hospitals it supports to secure recruits, medical supplies, and chemicals. In one case, Hamas recruited Mustafa Amjad, a doctor at al-Ghazi Hospital in Jenin, to help infiltrate suicide bombers into Israel from the Jenin area. After his arrest in June 2002, Amjad confessed to helping Hamas terrorists enter Israel while delivering medicines in his professional capacity.

Beyond logistics, rebels likely view communities as strategic political allies. Groups with more political support are more likely to successfully coerce concessions from governments. This specific pressure may be acute in democracies where governments are particularly beholden to voters.

Given the complexities surrounding when and how much support groups need, there is no reason to suspect that all groups rely equally on local populations. Reliance is multifaceted and shaped by many factors such as a group’s size, its resource base, and the specifics of its political agenda. For those less reliant groups, community bonds are less constraining. For the heavily reliant, community bonds are especially constraining. My goal here is to explain the effect of community bonds on target selection, specifically the decision to target civilians. To do that, I assume reliance exists at some minimal level for all groups.

Civilian preferences and rebel behavior

Rebels’ violent choices can be altered by the non-member civilian preferences when two conditions exist. First, rebel preferences must be different from those of their civilian supporters. Even if rebels and their civilian supporters share strong ideological bonds and a common understanding of the internecine conflict, there are reasons to suspect that their preferences diverge at a tactical level. For instance, rebel organizations attract recruits with relatively extreme views and willingness to use extreme tactics.² As a result, ‘moderate rebels’ may in fact be quite different from the civilian community around them. Kalyvas & Sánchez-Cuenca (2005: 219) note that,

In either case, a trade-off exists between the intensity of killing or the selectivity of violence on the one hand and popular support on the other. Supporters may share the ends sought by the insurgent group but not support all the means it uses to achieve those ends.

Brutal attacks often bring this trade-off into the public eye. For instance, on 8 November 1987, an IRA bomb killed 11 civilians in the town of Enniskillen, Northern Ireland.³ The attack left many aghast. Those dead were among many gathered for a remembrance ceremony for servicemen and women killed in conflict. The IRA claimed the bomb had been intended for soldiers and police and that the timing mechanism had failed. This excuse had little sway as communities reacted to the horrifying events. One senior IRA figure noted that, ‘after disasters like Enniskillen, houses were shut to the IRA’ (Moloney, 2002: 324). Another senior IRA member described Enniskillen as a ‘major setback’ and ‘totally devastating’ to many subsets of non-member supporters (McKittrick, 1998).

An analysis of the theoretical and case-specific literature on this topic indicates three reasons civilians are critical of terrorist tactics. First, international norms have developed around the idea that once groups are perceived as terrorists they are viewed as ‘callous, brutal murderers of the innocent’ (Tony Blair, in the Guardian, 2003). The idea of avoiding civilians during war is pervasive throughout modern, mostly Western, democracies and is the cornerstone of contemporary interpretations of the Geneva Conventions. Evidence from Muslim countries indicates similar views (Pew Research Center, 2013). In a 2009 article, Mia Bloom described the consequences of violating these norms for the Tamil Tigers (LTTE) after they killed civilians in a lethal attack on the Colombo Central Bank.

After the deadly 1996 Colombo Central Bank and World Trade Center attack, which killed 91 civilians and injured more than 1,400, resulting in a backlash even among their staunchest supporters, they decided to choose their targets more wisely, focusing on the military, the police and the government. (Bloom, 2009)

Second, when rebels use terrorism their civilian supporters frequently suffer a direct in-kind response by government agents. To the extent that a rebel group, a self-proclaimed representative of the minority, targets a government’s supporters, the state may be more inclined to repress entire areas associated with the rebel movement. Although this may occur in the aftermath of any type of violent attack, pressure to repress and respond may be highest after rebels target civilians.

Alternatively, governments often support a paramilitary response, a strategy that might be more likely in the face of terrorism. Responding to a dramatic uptick in civilian deaths caused by ETA operatives, the Spanish government established Grupos Antiterroristas de Liberacion (GAL) to target ETA members and sympathizers (Sánchez-Cuenca, 2009). GAL death squads operated ‘when killings by the Basque separatist group ETA were at their height’ and ‘more than a third of the people killed by the GAL death squads had no connection to ETA’ (BBC, 1998). This response to ETA’s intense campaign in the early 1980s proved extraordinarily costly for ETA’s civilian supporters, including those with no formal connection to the organization.⁴ Thus, beginning after the mid-1980s ETA’s attacks on civilians dropped while its attacks on other targets did not follow suit.⁵

Terrorists' civilian supporters also suffer when civilian militias organize to counter terrorism by attacking civilians from opposing populations. Combs (2003: 26) notes that violence against innocent persons creates ‘a cycle of violence with those against whom the terror-violence is first carried out becoming so angered that they resort to terrorism in response, directed against the people or institutions regarded as responsible for the initial terrorist acts’. Loyalist militias throughout Northern Ireland wreaked havoc in Catholic neighborhoods following IRA attacks on civilians. Moloney (2002: 145) argues that the Loyalists and the IRA ‘vied with each other in an often indiscriminate sectarian killing game’, resulting in scores of tit-for-tat attacks against civilian targets. The logic behind loyalist killings hinged on undermining the IRA’s Catholic support base. In an author interview, a civilian living in Republican Belfast described this dynamic clearly. ‘Loyalists targeted catholic citizens. They wanted to turn the community against [the IRA] which brought the realization that nobody would win the conflict.’⁶

In this analysis I assume civilians are adverse to terrorist tactics. Yet there are a few instances where evidence suggests this assumption may be incorrect. For instance, opinion poll data suggest at times Palestinians may be relatively less likely to oppose attacks on civilian targets. In 2002, nearly 46% of Palestinians expressed opposition to attacks on civilian targets. By 2005 that number climbed to 58%. By 2008 this number fell to 31%. In 2013 it had grown again, as approximately 50% opposed attacks on Israeli civilians.⁷

These trends show that, at least in the Palestinian case, preferences over the use of terrorism fluctuate. Widespread Palestinian acceptance of terrorism is problematic for this analysis if in fact Palestinians are not an anomaly but, instead, reflect an increasingly acceptant attitude about terrorism. If this were the case it would challenge a key assumption underlying the argument here. Yet neither the literature above nor the analysis below suggests this is true; rather, the data indicate fairly strong evidence that civilians reject terror tactics. However, the Palestinian case highlights two additional points. The first is a caveat to the argument. Where particular rebel behaviors are supported by a group’s civilian constituency, we might find that these behaviors proliferate either for any given group’s strategy or among competing groups in the area (this could account for Palestinian violent groups’ infamy as some of the most callous terrorist organizations). Additionally, the temporal variance in the polling data ought to motivate longitudinal analyses on the use of particular battlefield tactics and civilian preference. The lack of opinion data is a significant challenge for cross-national analysis such as this, but coupling data from public opinion polls with corresponding rebel group behaviors is an exciting prospect for future empirical case research.

Recall earlier I argued two conditions are necessary for rebel behavior to be altered by the preferences of non-member civilian supporters. Regarding the first, divergence in preferences, evidence indicates civilians are relatively less tolerant of terrorism. For rebels to change their behavior in response to civilian preferences, however, civilians must also have some degree of leverage over the rebel organization. If civilians are coerced into providing resources to the group and/or scared into not talking with authorities, there are few reasons to suspect that the organization will be influenced by civilian preferences. When no mechanism for sanctioning exists or when rebels successfully use coercive force to compel community support, they are less prone to be concerned with the effects of their violence on the local population and more likely to act with impunity. When communities can punish terrorists, however, rebels will avoid civilian targets in order to avoid sanctioning. In one of the most comprehensive books on terrorist group organization, Shapiro (2013: 266) concludes similarly that governments can induce rebels to use violence discriminately ‘by highlighting the human costs of terrorism among groups’ supporting populations’.

What are some of the ways civilians punish terrorists? Individuals can choose not to cooperate with violent groups by closing their homes as safe havens or informing police or government authorities about terrorists’ activities. They can withdraw support from a rebel group and instead support a group’s competitor. While there are many ways in which civilians impose costs on rebels, quantifying these exchanges is difficult. Further, to understand the effect of civilian punishment, it is essential to identify a mechanism not subject to rebels’ coercive intervention. To this end, voting in elections is a plausible mechanism.

Violent organizations frequently participate in elections. The IRA, ETA, Hamas, Hezbollah, and al-Fatah are prominent examples, but so do many smaller and lesser known groups. The more costs groups impose on communities by using terrorist tactics, the less likely communities are to elect them to office (de la Calle & Sánchez-Cuenca, 2012). Terrorism, by virtue of its associated costs and ethical conundrums, is an unlikely strategy for groups that compete for electoral office.

H1: Ceteris paribus, groups with political wings that participate in elections are less likely to attack civilians than groups with non-political wings.

For some violent groups, active engagement in electoral politics is something that begins after some time. For others, participating in elections precedes violence. Temporal changes in violent patterns should reflect engagement in electoral politics. Thus, when a group is competing in elections and pursuing violence, its violence against civilians ought to be tempered. Conversely when a group stops competing in elections, its violence will be less constrained.

H2: For any given group, violence against civilians will be lower if the group simultaneously competes in elections and higher when it abstains from elections.

Why would groups run for office if it constrains their violent options? Assuming groups see some value in violence, limiting it in a manner consistent with the argument above requires explanation. In reality, the costs associated with political participation likely go much further. Losing at the polls implies difficulty for a group in (a) convincing a government that it should concede demands, (b) fundraising, (c) recruiting for a cause publicly revealed to be unpopular, and (d) fighting off challenges from groups with complementary political agendas.

Why do violent groups subject themselves to these risks (i.e. run for office)? Elections are one way groups legitimize themselves. By demonstrating legitimacy through electoral success, terrorist groups force governments to take their demands and, to the extent that electoral participation makes terrorism a costly signal, their threats seriously. Political success is also another way to appeal to potential supporters, particularly more moderate supporters who are opposed to all forms of violence. Political participation enlarges a group’s reach, enabling it to engage in public debates, hold fund-raising events, and appeal to a broad international audience that only a political arm can openly embrace. Thus, while running for office has potential pitfalls, there are clear benefits.

Methods and analysis

To test my hypotheses, I use data on non-state actors from the Minorities at Risk Organizational Behavior (MAROB) database (Asal, Pate & Wilkenfeld, 2008). As a subsidiary of the Minorities at Risk (MAR) project, MAROB reports on non-state actor behavior from 1980 through 2004. It focuses on the behavior of organizations that represent ethnopolitical groups from the Middle East and North Africa. The unit of analysis is the organization-year. It covers 112 organizations in 12 countries. Groups are most frequent in Iraq, Israel, and Lebanon. The four ethnic groups most commonly represented are Palestinians, Shi’is, Sunnis, and Kurds.

I use MAROB for a number of reasons. First, inclusion in the MAROB universe of non-state organizations is not limited to exclude groups based on a behavioral trait. MAROB groups use many forms of violent and nonviolent activities. Some datasets cover only groups that use terror tactics. For research questions about the nature of civilian targeting, this limitation may not be problematic. However, this is wholly inappropriate here because this analysis attempts to quantify when groups are more or less likely to use terrorism.

Second, and most importantly, to the best of my knowledge MAROB is the largest and most inclusive database tracking organizational behavior. Other data sources have limited coverage and no other source offers the type of organizational data necessary to test the hypotheses presented here. For instance, the Terrorism in Western Europe (TWEED) dataset (Engene, 2007) covers attacks from 1950 to 2004, but uses attacks as the unit of observation, is limited to Western Europe, and offers very little information about organizational traits.⁸ The Global Terrorism Database (GTD) (National Consortium for the Study of Terrorism and Responses to Terrorism (START), 2012), also an attack-level database, probably covers the widest range of groups and attacks, but offers no information on group-level attributes.

That said, there are some limitations to MAROB. Geographically, the data cover only the Middle East and North Africa. Given the focus of today’s most pressing threats, this coverage is germane. Nonetheless, readers should bear in mind that lessons and findings below reflect relations between organizations and their supporters in MENA region countries. Additionally, MAROB’s regional emphasis may underrepresent democracies and, therefore, the opportunities for rebels to pursue elected office. This does not appear to be a significant issue. Although groups from MAROB democracies Israel, Cyprus, and Turkey represent nearly 25% of the observations, approximately 34% of groups in the dataset are coded as participating in elections. To the extent that being in or out of a democracy matters, below I control for level of democracy.

Another caveat of the data set is that it includes both violent and nonviolent organizations. However, all violent and nonviolent agitation methods are strategically determined.⁹ Thus, the decision to use terrorism is more aptly framed as one in which the alternatives are nonviolence or non-terrorist violence. Theoretically, I see few challenges to this idea. However, I assess the data using a model to account for the possibility that (non)violent groups are unique in ways that might also shape their relationship to civilians (they may be, for instance, more willing to use coercion and less concerned about civilian preferences).

One sincere limitation of MAROB stems from its ethnopolitical nature. To the extent that ethnically determined groups are distinct from groups united by shared class (e.g. leftist) or ideational (e.g. environmental) bonds, the findings below are less revealing. A related criticism of MAR (and, by extension, MAROB) is that its ethnopolitical basis results in the overselection of groups that suffer discrimination. This is especially problematic for studies using MAR groups to determine when groups use violence; presumably, there is a positive relationship between discrimination and violence. However, MAR groups are selected based on discrimination and/or mobilization. Birnir et al. (2011) show that MAR tends to select on mobilization and violence more than discrimination, which is unproblematic for this analysis given its focus on behavior of violent (i.e. mobilized) groups.¹⁰

Terrorism

My dependent variable, terrorism, comes from MAROB and specifically indicates whether in the year coded the organization attacked domestic non-combatants. I generated this binary indicator using MAROB’s DOMORGVIOL variable which tracks the degree to which groups attack particular domestic targets. Observations are coded as having used terrorist tactics when DOMORGVIOL is a 3 or above indicating the group used violence against domestic non-combatant and civilian targets. The location of the target is important to note. It is unclear what, if any, reprisals groups’ supporters will face for terrorist attacks on civilians outside the state. Thus, here I assess domestic terrorism and political participation as the most relevant variables.¹¹ Terrorism is used as a strategy in approximately 11% of the cases observed. Of these, most took place in Iraq, Israel, and Lebanon. The distribution of terror attacks across time in this population is relatively stable, with only one slight uptick in the mid-1990s.

Electoral politics

The primary independent variable in this study is participation in electoral politics. The data for electoral participation come from MAROB. It is a trichotomous indicator where 0 indicates no participation, 1 indicates groups with members in office or that members will run for office outside of an election year, and 2 indicates groups with members running for office during an election year, and I expect that the limiting effects of elections on terrorism to be most substantial in these cases. Not surprisingly, there are fewer cases of groups running for office during an election year. Only 12% of groups fit this classification. Nearly 22% receive a score of 1 and the rest, 66%, do not participate.

Controls

I control for organizational and state-specific variables. At the organization level, I use a binary indicator for nationalist group identity (from MAROB), defined as groups that advocate for autonomy or independence.¹² Groups with nationalist goals, 37% of the sample, tend to distinguish their constituent populations. This is important insofar as this distinction might correlate with the amount of terror against ‘the other’ (Drake, 1998). An easily distinguishable out-group could, however, be a mixed blessing. On the one hand, terrorism should increase if rebels convince their supporters that all members of the out-group are threatening. This may be more likely in societies with tightly knit ethnic bonds where group members behave more or less alike one another. As victims of an oppressive and easily identifiable group, thus, rebel supporters may be more willing to give rebels carte blanche when it comes to violent tactics. On the other hand, if the out-group is easily identifiable because a country’s ethnic groups are segregated, then the rebel group’s supporters are likely also easily identifiable. This could make terrorism more costly by making it easier for governments and/or paramilitaries to launch reprisals.

Second, I include a variable indicating whether the group provides social services within communities. Heger, Jung & Wong (2012) find that social services are significant predictors of targeting and lethality. MAROB’s data on service provision is a trichotomous variable where 0 indicates no provision. Third, I include a binary control for whether the group has a conflict with another organization (including those that represent the same or dissimilar ethnic groups). This control, again from MAROB, is intended to capture any incentives groups may have to target civilians

Table I.

Descriptive statistics

Variable	No. obs.	Mean	Std. dev.	Min.	Max.
Civilian attacks	1,789	0.10	0.31	0	1
Electoral participation	1,789	0.45	0.69	0	2
Service provision	1,707	0.29	0.64	0	2
Palestinian	1,789	0.30	0.45	0	1
Interorganization conflict	1,785	0.21	0.41	0	1
Nationalist organization	1,764	0.36	0.48	0	1
Foreign support	1,572	0.37	0.48	0	1
Suicide tactics	1,788	0.03	0.17	0	1
Group proportion	1,779	0.25	0.16	0.0072	0.87
Political discrimination index	1,209	2.10	1.77	0	4
Democracy	1,789	0.27	0.44	0	1
Ethnic fractionalization	1,789	0.25	0.15	0.04	0.75
Log mountain terrain	1,789	2.66	1.24	0	4.28
Log population	1,789	8.76	1.24	5.81	11.17
Log real GDP	1,789	8.60	0.85	6.83	10.34

due to existing conflicts with other non-state actors. Fourth, to the extent that rebels place inherent value on extreme tactics (e.g. attacks on civilians) I include a MAROB control indicating whether the group uses suicide tactics. For two reasons I also include a MAROB indicator designed to capture whether a group receives support from an outside state. First, Weinstein (2006) argues that groups that receive external support tend to target civilians more indiscriminately. Second, discontinuing attacks on civilians in the face of elections may be the result of appeals to international donors who care about the legitimacy lost during elections when groups are perceived by outsiders as murderers and terrorists. Siddiqui (2013), for example, discusses how Khaled Mashal strategically discontinued suicide attacks in 2005 in anticipation of donors’ reactions during Hamas’s run in the 2006 Palestinian elections. Additionally, I include a control for the Israeli–Palestinian conflict. Some evidence suggests that the competitive atmosphere created by numerous Palestinian groups generates incentives for terrorism as a form of outbidding (Bloom, 2004; Clauset et al., 2010).

Finally, I include two variables from the MAR dataset that relate to the MAR group each MAROB organization represents. The first is group proportion (GPRO) which indicates the percentage of the country’s population that belongs to the ethnic minority. GPRO is available in the MAR dataset for years 1990, 1995, and 1998. I averaged across these years which results in a time-invariant indicator for each organization.¹³ By controlling for the relative size of a group’s constituency, the analysis may better model incentives small (or large) groups have to decrease terrorism related with electoral participation. Small groups that do not stand to win much may be less constrained by constituent outcry. Alternatively, big groups might be more likely to use terrorism more discriminately against smaller, tight-knit minorities with fewer electoral consequences. The second variable included here attempts to quantify the extent of political discrimination a MAR group faces. The variable, POLDIS, is an ordinal indicator where 0 means the group suffers no discrimination while a score of 4 indicates the group suffers substantially restricted political participation. Groups that face extreme discrimination may be more likely to use terrorism in the face of already harsh conditions. One caveat to both of these MAR variables is that their coverage is substantially limited. POLDIS, for instance, is mostly limited to observations post-1989.

At the country level I include controls for democracy as a binary indicator using POLITY (Marshall & Jaggers, 2006) scores of 7 or above, ethnic fractionalization (Fearon & Laitin, 2003), mountainous territory (Fearon & Laitin, 2003), log of population, and the log of real GDP (log of gle_pop and gle_rgdp from Gleditsch, 2002, in Teorell et al., 2011). All of these are standard in the conflict literature. Descriptive statistics for all variables are in Table I.

Table II.

Logit models assessing civilian attacks

Variable	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6
Electoral participation	–0.55* (0.28)	–0.43 (0.28)	–0.67 (0.35)	–0.77* (0.31)	–0.88* (0.36)	–0.84* (0.39)
Service provision		–0.44 (0.31)	–0.35 (0.32)		–0.05 (0.32)	0.19 (0.30)
Palestinian		0.003 (0.60)	1.02 (0.83)		–0.53 (0.90)	1.85 (0.94)
Interorganization conflict		1.52** (0.39)	2.04** (0.38)		1.96** (0.38)	2.28** (0.41)
Nationalist organization		0.89 (0.53)	1.11* (0.55)		–0.21 (0.55)	–1.27* (0.58)
Foreign support		1.02* (0.45)	1.04* (0.51)		1.47** (0.45)	1.76** (0.45)
Suicide tactics		2.37** (0.68)	3.28** (0.89)		1.86** (0.62)	2.55*** (0.75)
Group proportion			3.95** (1.19)			3.91* (1.64)
Political discrimination index			–0.33 (0.20)			–0.83* (0.36)
Democracy				2.40** (0.61)	2.76** (0.68)	3.38** (0.96)
Ethnic fractionalization				–4.36** (1.31)	–5.52* (2.60)	–2.34 (2.72)
Log mountain terrain				–0.36 (0.22)	–0.62* (0.17)	–0.002 (0.23)
Log population				0.26 (0.21)	0.45 (0.25)	1.25** (0.38)
Log real GDP				–0.92** (0.24)	–0.94** (0.26)	3.91 (1.64)
Years included?	Yes	Yes	Yes	Yes	Yes	Yes
Constant	–2.05 (0.42)	–3.35 (0.82)	–4.00 (0.92)	5.30 (2.78)	–3.47 (3.13)	–16.19 (6.38)
Number obs.	1,789	1,490	991	1,789	1,490	991
Clusters	112	106	102	112	106	102
Log pseudolikelihood	–594.32	–409.03	–224.03	–525.14	–353.35	–192.36
Pseudo R2	0.02	0.24	0.37	0.14	0.34	0.46

**p =< 0.01, *p =< 0.05.

Analysis

The analysis uses a series of logit models. Table II presents six core models. The first analyzes the singular relationship between electoral activism and attacks on civilians. The second, third, and fourth models incorporate the group-specific controls and the country-specific controls, respectively. The fifth model includes all controls with the exception of group proportion and political discrimination. Inclusion of these variables decreases the number of observations by roughly 35%; the sixth model includes these controls for comparison. Additionally, in each instance I include dummy variables for all years and, because there are good reasons to think that attacks across the same group are not independent, all models are clustered at the organizational level.

The results for the central variable, electoral participation, are significant and in the assumed direction. Groups with an active political party are significantly less likely to target civilians. This relationship is negative and significant in all but two models where it just nearly misses the mark for standard levels of significance. Using Model 5 specifications, the predicted probability that a group will use terrorism when it does not participate is approximately 6%. When groups do participate, however, the predicted probability drops to roughly 1%. The findings strongly support H1 and imply support for H2; participating in elections has a significant negative effect on the likelihood that any given group will employ terrorist tactics.

Results for several of the control variables are striking. Interorganizational conflict is positive and significant, indicating groups in conflict with other groups are more likely to use terrorism. Further, the effect is not trivial. The predicted probability that a group involved in interorganizational conflict will use terrorism is approximately 17%. When groups are not involved in interorganizational conflict, the odds of terrorism fall dramatically to approximately 3%. This is a dramatic change and suggests that the dynamics between groups may be as important when considering terrorism as the dynamics between antagonistic governments and non-state groups.

Foreign support is also significant and positively related to attacks on civilian targets. Groups that get outside support are more likely to employ more extreme tactics, consistent with Weinstein (2006). This suggests that, at least to a certain extent, foreign support may be sustaining these organizations as ‘roving bandits’ that care less about consolidating legitimate authority in the areas where they hold power. At the same time, these results call into question the idea that foreign support might be leveraged to moderate attacks on civilians. Groups that use suicide attacks are also more likely to target civilians. Although suicide missions are practical against hardened targets (Berman & Laitin, 2005), this finding is also consistent with many of the most infamous suicide missions that take place in public areas and seem intentionally designed to inflict maximum damage against civilian targets.

Group proportion is significant and positive, indicating that the bigger the MAR constituent group, the more likely the organization will use terrorism. Groups that are quite sizable may be able to better target their attacks against ‘other’ civilians and repress enemy retaliation more effectively, thus dampening the effect of terrorism on their own constituents. Terrorism may also be a form of violent tyranny against minority groups intended to increase the costs of minority group mobilization. Though not as robust as group proportion, results for political discrimination indicate that repression decreases terrorism. This seems counter-intuitive because repression ought to increase a group’s grievances. However, it could also make it much more costly for a group to pursue violent attacks. At least in the case of terrorist tactics, the findings here indicate the latter may be a more significant concern.

Both relatively higher GDP and mountains tend to discourage terrorism. The latter seems consistent with common descriptions of terrorism as urban while the former is consistent with accounts of terrorism in relatively weak states. Two additional state-specific characteristics are relevant. First, democracy positively correlates with attacks on civilian targets. This is consistent with the findings of Pape (2005), Heger (2010), and Chenoweth (2010). Also, ethnic fractionalization is negatively associated with civilian targeting. The finding may reflect a sort of stalemated dynamic in which groups are dissuaded from attacking civilians in other ethnic groups because reprisals are deadly and easy to orchestrate in the presence of high ethnic fractionalization. Alternatively, this finding may also indicate that societies with high ethnic diversity have the highest incentives to establish transparent and inclusive political institutions that, in turn, decrease the odds of terrorism.

Equally interesting are some of the non-significant findings. The fact that service provision is not statistically related to terrorism is surprising, although the coding for this variable may not fully capture the impact of social service provision, especially if not every service is equal (Heger & Jung, 2013). Some might be more or less imposed by the group (e.g. security and policing) on a population while others (e.g. schools and welfare) may be provided non-coercively to address a dire need within a community. Both instances suggest different community/group dynamics and, by extension, differences in communities’ willingness to sanction organizations. Also interesting is the fact that the coefficient for nationalist group type is neither consistent nor robust. This may indicate that nationalist conflicts are not tactically unique. Alternatively, in these conflicts the incentives to attack civilians do indeed run both ways; civilians in the ‘out-group’ are easier to attack but the repercussions for rebels’ civilian constituencies might also be more acute. In the event that group identification might be different when groups are religious, I re-ran Models 2, 3, and 5 substituting ‘religious group’ for nationalist.¹⁴ The substantive results were unchanged. This leaves open some interesting questions. Namely, are nationalist conflicts unique and, by extension, is group distinguishability an important factor?

Table III shows results pertinent for a series of robustness checks.¹⁵ Recall I argued above that when

Table III.

Robustness checks for civilian attacks

Variable	Model 7 international	Model 8 domestic	Model 9 militant
Electoral participation	–0.18 (0.30)	–1.07* (0.42)	–0.58* (0.27)
Service provision	0.35 (0.26)	–0.30 (0.35)	–0.19 (0.26)
Palestinian	–1.08 (0.68)	–0.59 (0.99)	–0.99 (0.93)
Interorganization conflict	1.34** (0.40)	2.87** (0.54)	1.79* (0.36)
Nationalist organization	–1.92** (0.58)	–1.52 (1.01)	–0.54 (0.69)
Foreign support	1.60** (0.53)	0.50 (0.80)	0.75 (0.40)
Suicide tactics	2.45** (0.74)	–0.63 (1.35)	1.51* (0.62)
Democracy	6.55** (0.91)	0.60 (1.29)	4.60** (0.96)
Ethnic fractionalization	–0.49 (1.95)	–1.55 (3.30)	–2.96 (1.86)
Log mountain terrain	–0.76** (0.21)	–0.50 (0.27)	–0.31 (0.17)
Log population	0.51 (0.29)	–0.48 (0.28)	0.14 (0.36)
Log real GDP	–1.01** (0.31)	–1.70** (0.61)	–1.07** (0.30)
Years included?	Yes	Yes	Yes
Constant	3.09 (4.47)	18.03 (5.50)	6.65 (4.19)
Number obs.	779	711	900
Clusters	71	66	60
Log pseudolikelihood	–161.76	–105.46	–293.95
Pseudo R2	0.51	0.44	0.33

**p =< 0.01, *p =< 0.05.

groups attack civilians, their supporters pay certain costs related to the specific response to terrorist activities. In instances when rebels operate across borders and holding all else equal, I expect the costs imposed on a rebel groups’ supporters will be lowered because of the legal and normative barriers restricting a state’s response against a civilian population in another state. Because civilians are willing to sanction as a result of these costs, I anticipate that electoral participation will have a less pronounced effect on terrorism compared with groups that exist and operate in the same state. Below I examine the relationship between elections and terrorism for subpopulations distinguished by the extent of groups’ domestic or international activity. To distinguish these organizations, I employed two variables from MAROB: where is the organization using violence as a strategy and where is the organization active nonviolently. When groups’ activities are split between domestic and international environments or when group activities take place entirely internationally, I consider them ‘international’ (53.55% of the sample versus domestic groups which make up 46.45% of the sample). Models 7 and 8 below show the results for each subpopulation. Consistent with my expectations, electoral participation is more constraining for domestic groups than international groups where there is no apparent statistical relationship. Another interesting finding is that for those groups that are primarily domestic democracy is not significant (though the policy scores for each subsample are relatively similar). International borders might accentuate the incentives groups have to attack civilians in democracies if they do indeed reduce the costs associated with terrorism.

To address the objection that violent groups are different from nonviolent groups, Model 9 examines the relationship between electoral participation and terrorism for only those groups that MAROB specifies as being militant. Approximately 63% of the observations are coded in this manner. The results are largely unchanged. Among violent groups, elections are a moderating influence on groups’ attacks against civilian targets.

Identifying causality and the IRA

Thus far the evidence has largely been supportive of this article’s central claim: electoral participation correlates negatively with attacks on civilian targets. I propose this link is the result of potential losses at the polls violent groups would suffer if they chose to engage civilian targets. Yet two issues related to causality warrant further discussion. First, a central concern is that the negative relationship captured empirically here is actually the result of some other environmental variable not currently modeled that is causing groups to choose to participate in elections and also target civilians proportionally less frequently. Thus, the robust findings in the regressions above reflect the fact that these two traits may co-vary but are actually driven by some other factor. For instance, domestic cultural or international influences that emphasize human rights may simultaneously advocate more democracy and better human security as part of a larger human rights agenda. Thus, one might expect the same relationship between democratic participation and terrorism but for entirely different reasons. I anticipate many of these influences are captured by controlling for years. Going forward, the International Criminal Court (ICC) might have this sort of influence. The ICC was only two years old when this dataset terminates (2004) and no cases had been fully investigated.

Another challenge is the ability of this analysis to speak to the proposed causal mechanism. Recall, above, I argued that because elections allow civilians to sanction a group for unfavorable behavior, groups’ targeting choices will more closely reflect civilian preferences against deliberately targeted civilian attacks. The negative correlations above support this conclusion but perhaps another intra-organizational mechanism is at work. Running a political wing may be the result of more moderates in an organization who are simultaneously unlikely to attack civilians and likely to run for office. This is problematic. Do groups standing for election choose to attack fewer civilians because of moderates’ opinions/influence or because communities can use elections to sanction the group’s behavior? In reality, it could be both. A group’s actions could be predicated on the existence of both the punishment mechanism and its moderate members.

A compelling method to assess these alternatives would utilize detailed crucial case design (Levy, 2008), something space requirements do not allow here. However, to shed light on the issue I briefly present dialog during the IRA’s formative political years in the 1980s. The IRA, established as the Provisionals in 1969, was the primary organization opposing British influence in Northern Ireland. The battle between British authorities and the IRA resulted in one of the longest bouts of low-intensity conflict in any contemporary democracy. The Troubles, as the conflict came to be named, ended in 1998 with the signing of the Good Friday Agreement. In 2005 the IRA officially laid down arms.

The IRA’s experience with politics and violence is informative for several reasons. First, as a method to demonstrate plausibility, the case serves nicely to sharpen the theoretical specificity of the arguments above. Second, it adds diversity to the sample of cases under investigation (recall MAROB’s focus on MENA region countries), which serves to strengthen the external validity of the results. Third, the IRA was violent first and then politically active, a sequence that makes it possible to evaluate the extent to which political participation may have altered established patterns of violence. This is, of course, one possible group trajectory. Other groups may alternate sequencing between politics and violence. However, the IRA’s relatively linear path maximizes our ability to focus on the theory’s plausibility, therefore providing a more concrete illustrative example.

In dealing with its transformation into politics, IRA and Sinn Fein (the political party closely associated with the IRA and its Republican support base) leadership began to consider and eventually advocate for a more professional violent strategy. Underpinning the group’s transformation were two central pressures. First, the Republican leadership had to convince the movement’s base that violence would remain an important, if not the most important tactic in the struggle. This was important because many hardliners who fundamentally opposed the move to politics were primarily concerned that the IRA would take a back seat to Sinn Fein. In order to reassure those uncomfortable with the move into politics, Republican leaders regularly described and justified the political struggle as a complement to the violent struggle. In a now famous speech at a Sinn Fein convention in November 1981, party leader Danny Morrison asked delegates, ‘Who here really believes we can win the war through the ballot box? But will anyone here object if, with the ballot paper in one hand and the Armalite in the other, we take power in Ireland?’ (Rafter, 2005: 114).¹⁶ In 1982 at a Sinn Fein convention, a motion passed indicating that ‘all candidates in national and local elections, and all campaign material, be unambivalent in support of the armed struggle’ (Rafter, 2005: 119). Even as late as 1984 Sinn Fein leader Gerry Adams noted that, ‘armed struggle is a necessary and morally correct form of resistance in the six counties [Northern Ireland] against a government whose presence is rejected by the vast majority of Irish people’ (English, 2003: 244). The move into politics was not intended to end or diminish the violent struggle. In fact, just the opposite occurred. The move into politics forced Republican leaders to reaffirm their commitment to violence.

In isolation, it is quite plausible that the IRA’s more extreme members would have pushed the group toward more violent tactics. However, a second pressure resonating through the Republican ranks during the political transformation was the acknowledgment that political participation demanded strategic reforms to the group’s violent approach. Danny Morrison, in an interview after coining the ‘Ballot Box and Armalite’ strategy, expressed a tension between politics and violence. ‘I wanted to reassure people that it was possible to support the waging of an armed struggle and simultaneously take part in electoral politics – even though deep down I knew there were contradictions. I knew there was a ceiling to how far you could go’ (English, 2003: 225). Through its formative years the political leadership continued to reaffirm that that there were consequences imposed by IRA attacks on civilian targets. Shapiro (2013: 187) notes:

By mid-1985, PIRA leaders were seeing their political prospects fall and sought to limit attacks to military targets or to individuals who would be seen as legitimate high-visibility political targets. In the spring of 1986, Northern Command sought and received permission from the Army Council to vet most tactical operations in order to prevent electoral damage.

In response to a 1988 IRA attack that killed two Catholics, Republican leader Martin McGuinness said that ‘while the freedom struggle goes on the IRA has a responsibility to ensure as much as humanly possible that civilians are not endangered’ (English, 2003: 259). A year later, the group’s leadership reaffirmed McGuinness’s statement and indicated an acute awareness that attacks on civilians were bad for political business. In reporting on a closed session of a 1989 Sinn Fein conference in which the IRA Army Council delivered a message, the Belfast Telegraph reported that ‘the IRA is to take greater care to avoid civilian casualties in its terror campaign in response to concern expressed by Sinn Fein over the effect of its operations on the party’s political prospects’. At that same conference, in his Presidential address Gerry Adams warned members about the ‘detrimental effect of the IRA killings on civilians’. He went on to say that, ‘the wider base of the republican movement had the right to question and did question those IRA operations in which civilians were killed or injured’. Then vice-president McGuinness reiterated Adams’s sentiments by noting that the ‘killings of “non-combatants” had a damaging effect on morale’.¹⁷ Thus, while the group did not turn away from violence entirely, a move that might have alienated many of the group’s members, it did reform its tactics out of concern for electoral outcomes.

The fact that IRA leaders were discussing the disutility of civilian deaths years after the group’s first electoral foray begs an explanation. Why did the IRA take so long to reform its violent tactics? One explanation is that information within the organization was not internally vetted or clearly communicated in such a way to prevent damaging attacks. In 1986 the IRA leadership published a list of ‘legitimate targets’, which suggests the presence of such an information problem within the organization. Another possibility is that although the group was suffering a loss of support, its ability to do anything about it was severely curtailed by inadequate strategic planning. Shapiro (2013: 188) describes an intra-organizational strategic change that took place in 1988:

By the fall of 1988 even the PIRA policy for how to respond to Protestant killings was changed in a way that reflected the increasing importance of discrimination to the leadership. Prior to that point the policy had been to respond quickly with an equally violent attack against Protestant civilians. After that, the policy was to only hit back if a named target could be shown to be involved in the killings. Moreover, local commanders had to vet their retaliation plans through Northern Command leadership.

The evidence above strongly suggests that the IRA was forced to modify its violent tactics in the face of decreasing electoral support for Sinn Fein. To that end, the case provides compelling evidence that violent groups running for political office are less likely to use terrorist tactics as a result of political pressure to avoid civilian casualties. Additionally, the case indicates that organizational structure and disagreement among group members are also significant for tactical decisions. Because the IRA had members unwilling to abandon violence entirely, once it began participating in elections the leadership became more concerned that attacks continued but that the targets were not civilians. Enforcing this new mandate required structural changes. Electoral participation, thus, may be the starting point for tactical reform. Groups with fewer internal preference conflicts and clearer lines of hierarchy should, all else being equal, reform their tactics more quickly in response to electoral outcomes.

Conclusion

This analysis demonstrates how studying violent outcomes in relation to political activities can yield insightful results. Target choice is simply one of a myriad of strategic decisions rebels make. There is enormous potential for research projects to explore how violent strategies are shaped by factors outside the battlefield such as rebels’ political activities. The reverse may be true as well. Battlefield tactics could play a big role in conditioning political participation. To these ends several specific areas stand out as areas for further inquiry. The first has to do with the interplay and impact of internal and external constraints on a group’s decisions. The theory above outlines one significant external constraint for organizations: civilian supporters. However the influence of donors and other supporters may be equally if not more influential and is an issue that needs further exploration. Additionally, a different strand of literature suggests that violence against civilians is contingent on the degree of support or control that either the government or the terrorist group has in any particular territory (Kocher, Pepinsky & Kalyvas, 2011; Balcells, 2010; Kalyvas & Kocher, 2009). Additional cross-national and temporal analyses of these theories would be extraordinarily beneficial. Finally, the finding and mechanism identified in this analysis are based on a form of electoral governance that is not universal. One path for future scholarship might identify how non-democratic institutional features bind rebels to their supporters.

The results here demonstrate a link between how groups manage civilian support and the character of their violence. Perhaps inherent to all organizations with political arms, the results indicate that electorally active rebels might have more to lose when it comes to the decision about whom or what to attack. The work of de la Calle & Sánchez-Cuenca (2012) and Foster, Braithwaite & Sobek (2013) suggests an even more nuanced approach may be in order. In the case of ETA, de la Calle & Sánchez-Cuenca (2012) find significant differences even within the category of ‘civilians’ such that examining how different types of civilian attacks (e.g. non-state security personnel, sectarian civilian killings, and informants) may elicit very different electoral responses. Foster, Braithwaite & Sobek (2013) suggest that electoral rules can generate perverse outcomes for groups to actually gravitate towards violence. Thus, democratic participation as the simple answer to decrease terrorism may indeed be predicated on a much more complex set of environmental conditions.

Footnotes

Replication data

All analyses were conducted using STATA 13. The dataset, codebook, and command files for the empirical analysis can be found at .

Acknowledgements

These ideas benefitted tremendously from the feedback of Barbara F Walter, Eli Berman, Miles Kahler, David Lake, Wendy Wong, and Danielle Jung. I wish to acknowledge support from the One Earth Future Foundation. Additionally, this material is based upon work supported by the Air Force Office of Scientific Research (AFOSR) under Award No. FA9550-09-1-0314. Any opinions, findings, and conclusions or recommendations expressed in this publication are those of the author and do not necessarily reflect the views of AFOSR.

Notes

References

Akerlof

George

Yellen

Janet

(1994) Gang behavior, law enforcement and community values. In: Aaron

Henry

Mann

Thomas

Taylor

Timothy

(eds) Values and Public Policy. Washington, DC: Brookings Institution, 173–209.

Asal

Victor

Karl Rethemeyer

(2008) The nature of the beast: Terrorist organizational characteristics and organizational lethality. Journal of Politics 70 (2): 437–449.

Asal

Victor

Pate

Amy

Wilkenfeld

Jonathan

(2008) Minorities at Risk Organizational Behavior. Data and codebook, version 9/2008 (http://www.cidcm.umd.edu/mar/data.asp).

Balcells

Laia

(2010) Rivalry and revenge: Violence against civilians in conventional civil wars. International Studies Quarterly 54 (2): 291–313.

BBC (1998) Spain’s state-sponsored death squads. 29 July (http://news.bbc.co.uk/2/hi/europe/141720.stm).

Berman

Eli

(2009) Radical, Religious, and Violent: The New Economics of Terrorism. Cambridge, MA: MIT Press.

Berman

Eli

Laitin

David

(2005) Hard targets: Theory and evidence on suicide attacks. NBER working paper no. 11740 (http://www.nber.org/papers/w11740).

Berman

Eli

Shapiro

Jacob N

Felter

Joseph

Callen

Michael

(2011) Do working men rebel? Insurgency and unemployment in Afghanistan, Iraq, and the Philippines. Journal of Conflict Resolution 55 (4): 496–528.

Birnir

Jóhanna

Wilkenfeld

Jonathan

Brancati

Dawn

Fearon

James

Gurr

Ted

Laitin

David

Pate

Amy

Saideman

Stephen

(2011) A-MAR (All-Minorities at Risk): Addressing the selection bias issue (http://www.academia.edu/3604631/A-MAR_All-Minorities_at_Risk_addressing_the_selection_bias_issue).

10.

Bloom

Mia

(2004) Palestinian suicide bombing: Public support, market share and outbidding. Political Science Quarterly 119 (1): 61–88.

11.

Bloom

Mia

(2009) What the tigers taught Al-Qaeda. Washington Post 24 May (http://articles.washingtonpost.com/2009-05-24/opinions/36908404_1_female-suicide-bomber-terrorist-groups-ltte).

12.

Chenoweth

Erica

(2010) Democratic competition and terrorist activity. Journal of Politics 63 (3): 16–30.

13.

Chenoweth

Erica

Stephan

Maria J

(2012) Why Civil Resistance Works: The Strategic Logic of Nonviolent Conflict. New York: Columbia University Press.

14.

Clauset

Aaron

Heger

Lindsay

Young

Maxwell

Gleditsch

Kristian Skrede

(2010) The strategic calculus of terrorism: Substitution and competition in the Israel–Palestine conflict. Conflict and Cooperation 45 (1): 6–33.

15.

Combs

Cindy C

(2003) Terrorism in the Twenty-First Century. Upper Saddle River, NJ: Prentice Hall.

16.

de la Calle

Luis

Sánchez-Cuenca

Ignacio

(2012) Rebels without a territory: An analysis of nonterritorial conflicts in the world, 1970–1997. Journal of Conflict Resolution 56 (4): 580–603.

17.

Devine

Michael

(1989) IRA ‘will try not to kill civilians’. Belfast Telegraph, 30 January: 17.

18.

Drake

CJM

(1998) Terrorists’ Target Selection. London: Macmillan.

19.

Engene

Jan Oskar

(2007) Five decades of terrorism in Europe: The TWEED dataset. Journal of Peace Research 44 (1): 109–121.

20.

English

Richard

(2003) Armed Struggle: The History of the IRA. London: Pan Macmillan.

21.

Fearon

James D

(1995) Rationalist explanations for war. International Organization 49 (3): 379–414.

22.

Fearon

James D

D Laitin

David

(2003) Ethnicity, insurgency, and civil war. American Political Science Review 97 (1): 75–90.

23.

Foster

Dennis M

Braithwaite

Alex

Sobek

David

(2013) There can be no compromise: Institutional inclusiveness, fractionalization, and domestic terrorism. British Journal of Political Science 43 (3): 541–557.

24.

Gleditsch

Kristian Skrede

(2002) Expanded trade and GDP data. Journal of Conflict Resolution 46 (5): 712–724.

25.

Guardian (2003) Transcript: The Bush/Blair press conference. 20 November. (http://www.guardian.co.uk/politics/2003/nov/20/iraq.iraq).

26.

Harnden

Toby

(1999) Bandit Country. London: Hodder & Stoughton.

27.

Heger

Lindsay

(2010) In the crosshair: Explaining violence against civilians. Doctoral dissertation, University of California, San Diego (http://escholarship.org/uc/item/6705k88s).

28.

Heger

Lindsay

Jung

Danielle F

(2013) Negotiating with rebel governments: The effect of service provision on conflict negotiations. Working paper presented at Annual Peace Sciences Meeting, Knoxville, TN, 26 October.

29.

Heger

Lindsay

Jung

Danielle F

Wong

Wendy H

(2012) Organizing for resistance: How group structure impacts the character of violence. Terrorism and Political Violence 24 (5): 743–768.

30.

Irish Republican Army (1985) Handbook for Volunteers of the Irish Republican Army: Notes on Guerrilla Warfare. Boulder, CO: Paladin.

31.

Kalyvas

Stathis

Adam Kocher

Matthew

(2009) The dynamics of violence in Vietnam: An analysis of the hamlet evaluation system (HES). Journal of Peace Research 46 (3): 335–355.

32.

Kalyvas

Stathis

Sánchez-Cuenca

Ignacio

(2005) Killing without dying: The absence of suicide missions. In: Gambetta

Diego

(ed.) Making Sense of Suicide Missions. New York: Oxford University Press, 209–232.

33.

Kocher

Matthew Adam

Pepinsky

Thomas B

Kalyvas

Stathis

(2011) Aerial bombing and counterinsurgency in the Vietnam War. American Journal of Political Science 55 (2): 201–218.

34.

Levitt

Matthew

(2004) Hamas from cradle to grave. Middle East Quarterly 11(1) (http://www.meforum.org/582/hamas-from-cradle-to-grave).

35.

Levy

Jack S

(2008) Case studies: Types, designs, and logics of inference. Conflict Management and Peace Science 25 (1): 1–18.

36.

Quan

(2005) Does democracy promote or reduce transnational terrorist incidents? Journal of Conflict Resolution 49 (2): 278–297.

37.

Marshall

Monty

Jaggers

Keith

(2006) Polity IV Project: Political regime characteristics and transitions, 1800–2006 (http://www.systemicpeace.org/polity/polity4.htm).

38.

McKittrick

David

(1998) Can good emerge from the evil of the Omagh bomb? Independent 20 August (http://www.independent.co.uk/arts-entertainment/can-good-emerge-from-the-evil-of-the-omagh-bomb-1172787.html).

39.

Moloney

Edmund

(2002) A Secret History of the IRA. New York: Penguin.

40.

National Consortium for the Study of Terrorism and Responses to Terrorism (START) (2012) Global Terrorism Database (http://www.start.umd.edu/gtd).

41.

Palestinian Center for Policy and Survey Research (2002) Palestinian Public Opinion Poll No (5) (http://www.pcpsr.org/en/node/250).

42.

Palestinian Center for Policy and Survey Research (2005) Palestinian Public Opinion Poll No (18) (http://www.pcpsr.org/en/node/237).

43.

Palestinian Center for Policy and Survey Research (2008) Palestinian Public Opinion Poll No (27) (http://www.pcpsr.org/en/node/226).

44.

Palestinian Center for Policy and Survey Research (2013) Palestinian Public Opinion Poll No (48) (http://www.pcpsr.org/sites/default/files/p48e.pdf).

45.

Pape

Robert

(2005) Dying to Win: The Strategic Logic of Suicide Terrorism. New York: Random House.

46.

Pew Research Center (2013) Muslim publics share concerns about extremist groups. 10 September. Global Attitudes Project (http://www.pewglobal.org/2013/09/10/muslim-publics-share-concerns-about-extremist-groups/).

47.

Rafter

Kevin

(2005) Sinn Fein 1905–2005: In the Shadow of Gunmen. Dublin: Gill & Macmillan.

48.

Sánchez-Cuenca

Ignacio

(2009) The persistence of nationalist terrorism: The case of ETA. In: Mulaj

Kledja

(ed.) Violent Non-State Actors in Contemporary World Politics. New York: Columbia University Press, 69–92.

49.

Shapiro

Jacob N

(2013) The Terrorist’s Dilemma. Princeton, NJ: Princeton University Press.

50.

Siddiqui

Muneeb

(2013) Khaled Meshaal & the ascendance of Hamas. 23 May (http://muftah.org/khaled-meshaal-the-ascendance-of-hamas/).

51.

Teorell

Jan

Charron

Nicholas

Dahlberg

Stefan

Holmberg

Sören

Rothstein

Sundin

Petrus

Svensson

Richard

(2011) The Quality of Government Dataset, version 6Apr11. Quality of Government Institute, University of Gothenburg (http://www.qog.pol.gu.se).

52.

Weinstein

Jeremy

(2006) Inside Rebellion: The Politics of Insurgent Violence. Cambridge: Cambridge University Press.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB