Mobilizing Through Terror - Department of Political Science

Mobilizing Through Terror:
The Problematic Relationship Between Terrorism and
Ethnicity in Conflict Situations
Nora J. Keller∗
April 20, 2012
Abstract
How does ethnicity affect terrorism in the context of conflicts? Much of the current
literature on terrorism maintains a strong positive link between ethnic divides and the
use of terrorism, although this relationship remains under-theorized and lacks empirical
support. I argue that although ethnic conflicts experience more terrorism across the
board, the correlation between terrorism and ethnicity depends on conflict intensity.
At a low conflict intensity level, the logic for popular mobilization is similar for ethnic
and non-ethnic rebel groups. However, increased conflict intensity affects ethnic rebel
groups differently. Because shared ethnicity with a rebel group makes the in-group
population an automatic target of government reprisals, making the use of terrorism
costlier to ethnic than ideological rebel groups. At higher conflict levels, the ethnic
in-group population will have incentives to rally behind the rebels. I further examine
whether secessionist rebel aims influence the use of terrorism and might explain the
variation in use of terrorism for ethnic rebels across conflict intensity levels. I support
my theory using a negative binomial count model that measures the number of terrorist
attacks per conflict year.
∗
Thanks to the students in Political Science G8839, Professors Page Fortna, Tanisha Fazal and Shigeo
Hirano, Pierce O’Reilly, Nicholas Lotito and Michael Rubin for helpful comments; and to Michael Findley
and Joseph Young for sharing data.
1
1
Introduction and Research Question
How does ethnicity influence the strategic use of terrorism during intrastate conflicts? The
terrorism and civilian targeting literature is rife with arguments about the underlying ethnic
determinants of terrorism; ethnicity is often given a special status both in terms of perceived
roots, motivations, and group membership of terrorism. Byman (1998) in the to-date most
in-depth theoretical treatment of ethnic terrorism draws out the differences in membership
as well as goals of “ethnic” and “ideological” terrorists. The Liberation Tigers of Tamil
Eelam (LTTE) are an example of a terrorist (rebel) group with ethnically motivated goals
(independence from Sri Lanka). They also mobilized their supporters exclusively from the
Tamil minority, who felt increasing sympathy for the LTTE’s cause as government retribution
became directed at the minority population at large. On the other hand, any supporter of
Peru‘s Shining Path (Sendero Luminoso) communist ideology can join the organization and
work towards its goal of a new Peruvian governmental ideology.
The terrorist literature to date largely ignores how contextual factors affect how terrorism
is used strategically. I explore how shifts in conflict intensity affect the strategic use of
terrorism differently for rebel groups that mobilize along ethnic lines than ideological rebel
groups.
I will argue that terrorism is a strategy chosen by rebel groups to achieve a particular
end. Given the strategies that are commonly associated with terrorism, I will argue that the
decision to use terrorism is linked to mobilization needs of rebel groups as well as the potential cost the group incurs by using terrorism. At low levels of conflict intensity, ethnic rebel
groups will use terrorism more than ideological rebel groups, as they can more easily identify
and target out-groups members, thereby avoiding backlash from their own population. Mobilization needs do not fundamentally differ for ethnic and ideological rebels at low conflict
intensity levels. However, as conflict intensity increases, the ethnic in-group of a rebel group
will flock to their own ethnic group in a logic that follows the ethnic security dilemma (Posen
1993). Furthermore, using terrorism at higher conflict intensity levels brings more costs to
1
the popular base of ethnically mobilized rebel groups, as ethnic in-group members can be
easily identified by the government for retribution.
Secessionist rebels, which are often associated (and more often than not synonymous
with) ethnic rebels, face different costs from non-secessionist rebels that reduce the likelihood
they would employ traditionally-defined terrorist methods. Because they desire to set up
their own independent state, they will feel the need to appeal to the international community
of states (Fazal forthcoming). Furthermore, because of a likely high ethnic heterogeneity in
peripheral, self-contained territories, it may be more difficult for secessionist rebels to carry
out effective terrorist attacks. Finally, the likely geographic segregation and homogeneity
of the (likely ethnic) group that wishes to secede makes the identification of the in-group
mobilization base relatively easy. This, in turn, will make the ethnic secessionist rebels’
population an even more appealing target for government reprisals, and since the rebel
group wishes one day to govern this population, they will not want to offer up their future
political base for bait. I argue that secessionist rebels use less terrorism across the board.
I have chosen to focus in on conflict contexts for several reasons. While I could evaluate
a larger universe of terrorist attacks (not restricted to civil conflict situations), it would be
much more difficult to determine when terrorism is ethnically motivated, and when it is not.
It would then also be difficult to gain sufficient variation on the independent variable how
could I build a counterfactual by only looking at cases where terrorism was used? This will
allow me to more precisely parse out the mechanism that determines the strategic use of
terrorism. sing ethnically motivated civil and regular civil wars allows me to address both
of these issues, using a strategy similar to Fortna‘s (forthcoming), who justified her choice
of civil wars as a universe of cases by saying that she caught all the cases where “a group
has a serious enough perceived grievance against the state to launch a violent rebellion, in
which some groups choose to use terrorism as part of their repertoire of tactics while others
do not (5).
This paper will proceed in the following way: I will first introduce the definitions of
2
contentious key terms (such as “terrorism” and “ethnic group”). Then, I will present the
current state of the literature on terrorism and ethnicity, as well as the literature on ethnicity
and civil war (briefly). I will lay out my theory on ethnic mobilization and terrorism, as
well as secession and terrorism, and introduce testable hypotheses. I will finally test these
hypotheses using a negative binomial count model.
2
Definitions and Background
“Terrorism” is notoriously hard to define and conceptualize. Hoffman argues that “most people have a vague idea or impression of what terrorism is but lack a more precise, concrete,
and truly explanatory definition of the word” (2006, 2). To further complicate the definitional quagmire, “terrorism” often has negative, emotionally loaded connotations. There
is little agreement on what the definition should include or exclude (Schmid and Jongman
(1993) collected 109 different academic and official definitions of the term). There is a lot
of disagreement as to who should be called a “terrorist” (only non-state actors or also state
actors?) or who the target should be (should only civilian targets be included?).
A definition of terrorism should be narrow enough to distinguish it from other forms
of politically motivated violence, such as guerrilla warfare or even conventional warfare.
However, terrorism and insurgency are not distinct categories, but rather overlap. In keeping
with the literature on insurgents’ use of terrorism, I consider terrorism as a distinct strategy
of insurgency (see, for example, Merari (1993) for a discussion of this question). How, then,
is terrorism different from other strategies of insurgency? Terrorism remains in the zone
of psychological influence and lacks the direct material goals often associated with guerrilla
warfare, such as obtaining resources or controlling specific territory (Merari 1993). While
conventional warfare and most guerrilla tactics are aimed at physically wearing down the
enemy’s ability or willingness to resist, terrorism aims to have a psychological impact. Often,
terrorism is therefore described as “propaganda by the deed.”
3
While terrorist rebel group have as their ultimate aim to coerce the government into
making concessions, the immediate aim of terrorism is to influence a wider audience. There
are a host of possible audiences: The government, the population that the rebel group is
trying to get on their side, other states‘ governments, or the international community at large
(the media, NGOs, international organizations, etc.). Sometimes, scholars also include other
distinctions, such as group size or types of weapons. For example, Merari maintains that
terrorists operate in small groups, and the weapons used do not resemble ordinary military
weaponry, as is often used by guerillas (Merari 1993).
In this paper, I use the definition of terrorism also used in the Global Terrorism Database,
which is also where I get my terrorist attack counts. This definition enjoys, if such a thing
is even possible in the terrorism literature, some degree of consensus. According to this
definition, terrorism is the threatened or actual use of illegal force and violence by a nonstate actor to attain a political, economic, religious or social goal through fear, coercion or
intimidation (LaFree and Dugan 2007).
Second, I require a definition of “ethnicity” and “ethnic conflict.” My basic definition of
ethnicity comes from Kaufman (1996):
An ethnic group (or nation) is . . . a body of individuals who purportedly share
cultural or racial characteristics, especially common ancestry or territorial origin,
which distinguish them from members of other groups. (Kaufman 138)
Kaufman derives from this a differentiation between ethnic and ideological civil wars. He
argues that ethnic conflicts are waged between groups that claim distinct heritages over power
relations between communities, whereas civil wars are contests between different factions
of the same community who disagree over how to best govern the community as a whole
(Kaufman 1996).
My working definition of “ethnic war is in accordance with Eck‘s (2009), who distinguishes
between ethnic mobilization and other aspects of ethnicity, such as the ethnic composition
of a country. This differs from an ethnic grievances approach, which holds that a conflict
4
can be deemed ostensibly ethnic in nature due to the grievances expressed by one or more
of the conflict parties. Rather, Eck‘s (and my) focus is on how membership in a particular
ethnic group can be important for rebel mobilization patterns. Eck therefore circumvents
some of the contentious debates about the merits of labeling conflicts as “ethnic” (see below).
Because of this exclusive focus on mobilization, Eck’s operationalization includes cases on
the borderline between “ethnic” and “ideological” for other scholars of ethnic conflict. For
example, UNITA (Uniao Nacional para a Indepencia Total de Angola) mobilized along ethnic
(Ouimbundu) lines but had a distinct Maoist ideology.
Since my theory focuses on mobilization, Eck‘s (2009) definition of ethnic war is quite
ideal. Eck‘s operationalization is concerned with patterns of rebel recruitment; therefore,
this operationalization may label more conflicts as ethnic than other approaches. Ethnic
mobilization, of course, is closely linked to ethnic grievances and the ultimate cause for
the mobilization in the first place. However, according to Eck, “the contested nature of
determining exactly what a conflict is about and whether the key issues are indeed overtly
ethnic in nature” are avoided by focusing on more observable patterns of rebel organization
(372). Ethnicity may serve as a factor of organization even in conflicts that are not centered
around an ethnic-group-specific grievance. Ethnicity in such circumstances might be a useful
organizational factor due to geographic reasons (if a country is geographically divided along
ethnic lines), as leaders may want to recruit locally, for example. Existing social and political
networks may also be grounded in ethnicity and can therefore be harnessed by rebel leaders.
3
3.1
Where Are We?
Ethnicity and Terrorism
The terrorism and civilian targeting literature often invokes ethnicity as a driver of terrorism,
though most of these arguments are not backed up with empirical evidence and lack a solid
theoretical framework. How is it that ethnicity affects the use of terrorism? And how can
5
we know whether this is true?
Byman (1998) argues that “ethnic terrorism is a distinct phenomenon, different from
other types of terrorism, such as terrorism carried out for ideological, religious, or financial
motives” (149). Byman compares the use of terrorism by ethnic terrorists rather than nonethnic terrorists, and does not focus on the use of terrorism by groups mobilizing along
ethnic rather than non-ethnic lines who might choose to use terrorism or not. In the way
Byman has set up his comparison, it would be very difficult to get enough variation in his
independent variables to draw meaningful, empirically motivated conclusions regarding his
theory.
According to Byman, the main difference between ethnic and non-ethnic terrorists is in
their audience: Byman claims that ethnic terrorists have a distinct goal of also influencing
their own population in order to foster ethnic mobilization; governments can more easily be
goaded into retribution against ethnically fractionalized populations rather than populations
divided along ideological lines. Byman does not consider why ideological groups‘ mobilization
needs might be different than ethnic groups‘.
Bloom (2005) also makes an argument about ethnicity and the ease of out-group targeting. She argues that ethnic (suicide) terrorists can more easily target adversarial populations.
In such a way, “[t]argeting the other side is easier when its members are of a different race,
ethnicity, religion, or nation” (79). She calls upon Kaufman (1996), who claims that ethnic
wars experience more violence and atrocities directed against civilians because the key issues
revolve around control of territory rather than convincing your audience that your cause is
just.
Stanton (forthcoming) tests the relationship between ethnicity and terrorism in conflict
situations empirically. She finds that exclusive groups are more likely to use terrorism.
She introduces the useful idea that rebel groups, in their decision whether to use violence,
consider the costs of violence, which depend on their own goals and their respective civilian
constituency. Along these lines, if rebel groups depend on a large civilian support base,
6
they cannot afford to use violence in ways that might lead to significant popular backlash.
Stanton finds that exclusive rebel groups (ones that have exclusionary political objectives,
such as seeking to achieve gains for one particular religious or ethnic group) are more likely
to use terrorism than more inclusive groups that rely on broad-based support, in order to
minimize backlash. However, she does not explore how this relates to a mobilization logic
that might vary along levels of conflict intensity.
Pape’s (2005) work on suicide terrorism can also be read as an argument about secessionist groups’ use of terrorism. He argues that modern suicide terrorism is best understood
as a strategy of national liberation, where an aggrieved group attempts to expel a foreign
occupier from a territory they consider as their homeland. Secessionist rebels can be seen
to follow a similar logic. Stanton (forthcoming) argues that wars of secession are likely to
experience more terrorism because rebels do not have to worry about later governing the
population they are now alienating. Fortna (forthcoming) argues contrary to Pape, and calls
upon ’s (forthcoming) work on rebel adherence to the law of war, and argues that secessionist
rebels are more likely to not target civilians because they want to appeal to the international
community. Fortna also confirms this empirically.
3.2
Ethnicity and Civil War
The question of whether terrorist strategies are used differently in ethnic rather than nonethnic intrastate conflicts speaks to the question of whether ethnic and ideological conflicts
are fundamentally different. Gurr (2000) and Horowitz (1985) argue that specific factors
predispose ethnically divided societies to violence along ethnic fault lines. Furthermore,
they maintain that the issues of contention in ethnic wars are particularly deeply entrenched
and persistent. Easterly (2000) points to ethnic fractionalization as an important driver of
recurring bloodshed in Africa.
Eck (2009) argues that ethnic conflicts are more violent specifically because of the mobilization patterns experienced in ethnic conflict; because of the clearly identifiable potential
7
mobilization base, ethnic conflicts are more intense. Because of the increased potential for
mobilization, a stronger fighting force can be assembled, more fighters can be killed, and
the amplified threat to government power created in this way will facilitate high levels of
violence. However, I argue that while this may be true for violence in general, this is not true
for terrorism in particular. She finds that wars that include rebel groups mobilized along
ethnic lines incur a 92% greater risk for intensification to war than non ethnically mobilized
conflicts.
There are, however, severe criticisms of such a point of view. Mueller (2000) argues that
explanations centered on the notion of ancient hatreds or top-down manipulations of the
populace in civil wars are often misleading. Rather, ethnic war is often
A condition in which a mass of essentially mild, ordinary people can unwillingly
and in considerable bewilderment come under the vicious and arbitrary control
of small groups of armed thugs, bands of opportunistic marauders recruited by
political leaders and operating under their general guidance. (Mueller 2000, 42)
Along similar lines, Fearon and Laitin (2004) take issue with the “conventional wisdom” that
civil wars are at their root caused by religious or ethnic antagonism. Rather, they argue,
countries are at risk for civil war if conditions favor the emergence of insurgency. States
likely to experience internal violence were financially, bureaucratically, and militarily weak
due to decolonization between the 1950s and the 1970s.
This, however, does not mean that ethnicity cannot be used in a distinct way during the
mobilization process than ideology, which is the main factor I am interested in. Mueller’s
argument therefore does not touch my theory. There is, though, the question whether ethnicity can be measured at all, and whether an ethnic framework is ever suitable for the study
of politics. Gilley (2004) fairly points out that while ethnicity might be a useful mobilization
tool, there are usually underlying structural issues that necessitate this mobilization in the
first place, which must be considered distinct from the ethnic fault lines. However, such
considerations are beyond the scope of this paper.
8
4
Terrorism: To What End?
If terrorism is a strategic choice by insurgent groups, to what end is it used? The possibilities
must be established before considering how circumstances would affect terrorism. I will
introduce the most commonly assumed roles of terrorism, as well as the audiences it is aimed
at influencing. I will introduce the use of terrorism as an attrition strategy, intimidation
strategy, provocation strategy, spoiling strategy, outbidding strategy, as well as propaganda
of the deed. In this discussion, I will consider whether the use of terrorism by insurgents
affects the use of these strategies. In the next section, I will consider how the nature of
circumstances (ethnic versus ideological) affects the strategic use of terrorism, and whether
rebels’ mode of mobilization is likely to affect the use of terrorism.
The strategies of terrorism are intended to influence a particular audience. Fortna (forthcoming) identifies three main audiences: First, there is the population on whose behalf the
rebel group claims to fight. Second, there is the government and the civilians who support
the government. This group includes two groups: It includes both “complicitous civilians1
(who benefit from the state and support it) “fence-sitters who are neither members of the aggrieved group nor active supporters of the states policies. There is, finally, the international
public opinion and support (either in the form of other states, international organizations,
or NGOs). In order to consider whether “ethnicity should make a difference in a conflict
context, we need to consider both terrorist rebel group tactics and their likely audiences.
Actors pursue an attrition strategy through the use of terrorism that the group is strong
enough to inflict serious costs and resolute enough to take into account civilian casualties
(Kydd and Walters 2006). The intended audience is therefore the enemy government. This
is intended to make the enemy yield to terrorist demands. This is, according to Kydd and
Walters, predicated on an assumption that the terrorist (insurgent) group is considerably
weaker than the government; an attrition strategy is considered the most cost-effective way
to signal resolve. Fortna (forthcoming) questions the usefulness of terrorism as an attrition
1
Term from Bueno de Mesquita and Dickson 2007, p. 369
9
strategy in a situation of insurgency, as rebels could also use guerrilla tactics against military
targets, such as hit-and-run attacks. She further argues that the use of terrorism as a costly
signal of resolve is unclear at best. What should be taken into consideration, though, is
whether terrorist attacks against a suitable part of the civilian population are easy to carry
out and whether they will be viewed positively or negatively by the insurgent terrorist group’s
support population. In this manner, we have to me cognizant that the strategies of terrorism
are not in isolation from each other.
Provocation strategies, outbidding strategies, intimidation strategies and propaganda of
the deed strategies are aimed at the population as an audience. Intimidation strategies are
pursued in order to demonstrate that the government is powerless to stop the terrorist group,
and that they have the power to punish whoever wishes to disobey them. Government backers
are to be convinced that backing the government will be costly. Kydd and Walter conclude
from this that intimidation is likely to be used against civilians on the fault lines of rebel
and government control, to deter individuals form supporting the government. Provocation
strategies are meant to work in the opposite way: Through provoking the government into
retaliating against the population in order to clamp down on terrorism, the population is to
be persuaded to consider the government as evil and untrustworthy. Outbidding strategies
are pursued when there are two or more groups vying for the support of a population that
is uncertain about which of these groups best represents their interests. According to Kydd
and Walter, citizens are likely to prefer being represented by groups that are more hard-line
than them.
Terrorist groups pursue a provocation strategy in pursuit of either regime or territorial
change, which is to be attained by provoking the government to react to the attack with
violence against the population at large. Kydd and Walter argue that in this way, groups
hope to persuade the population to think of the government as evil as they think of the
government.
Another strategy is spoiling, by which terrorists attempt to ensure that peace overtures
10
between the government and more moderate factions on the terrorists side are not fruitful.
Terrorists are alarmed at the prospect of peace overtures because they risk losing their
popular support to the prospect of peace. Through spoiling, terrorists attempt to paint the
moderates as untrustworthy to the opposition (government). This is not truly relevant until
a conflict draws towards an end, however.
A final strategy of terrorism that is particularly salient for intrastate wars is “advertising
the cause,” which assumes that terrorism can be used as a more direct mobilization tool
than acknowledged by Kydd and Walter. In such a manner, as convincingly argued by
Crenshaw (2011), terrorism publicizes grievances and thereby puts the cause on the political
agenda (groups can garner support through more press attention or word-of-mouth rather
than only depend on the government’s reaction in order to influence the public). In a
similar vein, Young and Findley (forthcoming) argue that whether insurgent groups are
active combatants or only at the stage of garnering support in order to shift from a popular
movement to an organized fighting force, the opposition group(s) seek more popular support,
as strength is (arguably) increased through more popular support. There is disagreement
over the effectiveness of using terrorism for propaganda of the deed purposes in civil wars.
Fortna (forthcoming) points out that “publicizing grievances is probably less of an important
role for terrorism in civil wars than in lower-level conflicts (17). Fortna further points out that
such a strategy is likely to repulse rather than attract fence-sitters to the cause of the (likely
more extremist) rebels. However, she does not consider how this might change depending
on how easy it is for terrorist rebel groups to target out-groups rather than in-groups.
5
5.1
Terrorism, Ethnicity and Conflict: A Theory
Terrorism, Mobilization and Ethnic Conflict
What are the common arguments as to why terrorism would be more frequently used in ethnic
than ideological conflicts? The most obvious argument in favor of this view is instrumental:
11
As Bloom (2005) argues when discussing suicide terrorism, “[t]argeting the other side is
easier when its members are of a different race, ethnicity, religion, or nation (79). Not only
is it easier to target your attacks at the correct target (discrimination), but clearly targeting
an out-group can, under circumstances of tension, directly generate support for the rebel
group and their cause among the ethnic in-group. In this manner, terrorism is at work as a
propaganda of the deed, provocation, or even outbidding strategy.
Ethnic rebel groups might also be more likely to use terrorism because of their relative
weakness or because of the seeming hopelessness of their cause. If terrorism is a weapon of
the weak, which is subject to debate (see Fortna (forthcoming)), then ethnic rebel groups
could also be expected to use more terrorism.
Only considering these factors, then we should expect the following hypothesis to bear
out:
H1: Terrorism is more likely in ethnic than ideological intrastate conflicts, no matter the
intensity level.
Terrorism has an expected mobilization effect through most strategies of terrorism (provocation, intimidation, outbidding, and propaganda of the deed ). Mobilization needs do not
necessarily have to vary for ethnic and ideological rebel groups. Both ethnic and ideological
rebel groups have needs for popular mobilization. Mobilization is crucial for both ethnic
and ideological rebel groups, as “purely military successes are often indecisive as long as
the enemy’s base of political support is undamaged” (Kaufman 1998, 140). In such a way,
territorial gains only have value if they go along with a supportive population, be it that the
rebel group‘s overall goal is ideological change (in which case the population would at some
point be governed by the rebels or rebel sympathizers), or whether it is closer associated
with the needs of a particular ethnic group.
However, there are key differences between ethnic and ideological mobilization. According
to Kaufman (1998), “ideological loyalties are changeable and difficult to assess, and the same
population serves as the shared mobilization base for both sides” (140). The same is not true
12
for the popular base of ethnically mobilized rebel groups: the ascriptive nature of ethnicity
will lead to a more accommodating recruitment environment, given that shared ethnicity
is an important marker that identifies potential popular loyalties. Ethnic mobilization can
also take advantage of the organizational links constituted by ethnicity in divided societies
(Horowitz 1985); according to Chandra (2006), “information about an individuals ethnic
identity categoriescan be obtained through supercial observation (399).
Terrorism may therefore have a higher mobilization pay-off for ethnic rebel groups than
ideological rebel groups. Lichbach (1995) argues that shared ethnicity decreases the freerider problem, as ethnicity provides a marker to identify potential recruits that try to evade
rebel co-optation and try to hide in the population at large. Once the ethnic population
is aware of the issue, intra-group pressure may induce more members to join the cause.
Shared ethnicity therefore diminishes the coordination costs and decreases the collective
action problem. Since we made the assumption that terrorism is a strategic choice, then, it
seems that at a low level of conflict intensity, terrorism would be the more likely chosen in
an ethnic rather than ideological conflict, simply because the marginal mobilization pay-off
is higher.
At low levels of conflict intensity, both ideological and ethnic rebel groups will have a
need to advertise their cause or plight. This is especially true in the case of extreme military
imbalance, where, as Kaufman (1998) claims, mobilization may be more difficult due to a
cause’s seeming hopelessness. If an ethnic group is particularly weak, then this cause would
appear particularly hopeless, making the need for increasing mobilization efforts particularly
dire.
However, as conflict intensity increases, the logic changes. On the one hand, increased
ethnic conflict intensity will make mobilization efforts on the part of the rebel group less
necessary. These shared goals and interests take on a more crucial nature in a conflict situation, as has been convincingly argued (though not empirically shown) by proponents of the
security dilemma in ethnic conflict, most famously associated with Posen (1993). According
13
to Posen (1993), the ethnic groups of a collapsed imperial regime face a situation of anarchy
and conditions that make offensive and defensive capabilities and intentions indistinguishable, and windows of vulnerability and opportunity that can lead to conflict. In this process,
ethnic group cohesion is seen as offensive, since military capability is often determined by
group cohesion. Posen further surmises that geography (not technology) matters most in
determining the offense-defense balance in ethnic conflict some groups will have greater
offensive capabilities because they will effectively surround some or all of the other groups
(ethnic “islands). Kaufman argues along similar lines, and maintains that by the time a
full-fledged ethnic conflict breaks out, all members of the ethnic group must (or will) be mobilized, because other ethnic groups will inevitably identify them as enemies. By supporting
rebel groups, ethnic group members can at least gain a modicum of security.
Therefore, the use of terrorism as a mobilization device of the co-ethnic population loses
strategic value at higher levels of conflict intensity. Ethnically mobilized rebel groups also
have less to worry about in terms of possible member cooptation. Once mobilized for the rebel
cause, a host of informal ethnic institutions and networks promotes the flow of information
(Ahn and Ostrom 2001) about people’s loyalty, and makes defection harder. Therefore,
rebels have less to worry about regarding potential member defection. Both because of
ethnic in-group members’ automatic closer adherence to the ethnic rebel cause at higher
conflict intensity and because of the smaller risk for defection, ethnic rebels can channel
their likely constrained resources into other military tactics with a higher pay-off.
At a high level of conflict intensity, the use of terrorism is also costlier for ethnic rebels.
The government will have more incentive to stifle opposition, and will likely target the ethnic
group as a whole in an attempt at retribution; given ethnicity‘s ascriptive nature, targeting
is easier in this case.
None of these factors apply for ideologically mobilized rebel groups. In fact, as conflict
intensity increases, so does the need for increased mobilization, which implies strength.
Because potential members cannot be identified based on ethnic characteristics, and because
14
the government might still be hesitant to indiscriminately target the population at large, the
use of terrorism should increase for non-ethnic rebel groups as conflict intensity increases.
H2a: The use of terrorism in ethnic conflicts is likely to decrease with increasing conflict
intensity.
And, concurrently:
H2b: The use of terrorism in ideological conflicts is unlikely to decrease with increasing
conflict intensity.
5.2
Terrorism and Secession
As there is a commonly perceived high correlation between “ethnic” and “secessionist” conflicts (though with this particular operationalization of “ethnic conflict”, it is only .64), it
is necessary to consider whether any findings for ethnic conflicts are due to the goal of a
specific conflict (to secede) rather than due to a mobilization logic.
Fazal (forthcoming) argues that secessionist rebels are more likely to comply with the
laws of war because they want to appeal to the international community of states. In this
way, rebel groups want to signal both resolve (to become viable members of the international
community) and capability (to restrain group members’ use of violence against civilians).
This relationship was tested by Fortna (forthcoming), who found that secessionist rebels are
indeed less likely to use terrorism. This logic goes along with Bapat‘s (2006) argument that
states do not like to negotiate with terrorists because of inherent credible commitment problems on the part of terrorist groups. Therefore, rebels might want to avoid using terrorism
in order to remain desirable future negotiation partners. Therefore, I propose the following
hypothesis:
H3a: Terrorism is less likely in secessionist rather than non-secessionist conflicts.
If secessionist rebel groups mobilize along ethnic lines, they are likely to experience the
same incentives and constraints to mobilization as non-secessionist ethnic rebel groups. However, there might be other reasons why rebel groups that are also secessionist will experience
15
less terrorism. Secessionist areas tend to be peripheral, and are likely relatively homogeneous,
which Eck (2009) argues will influence rebel mobilization (make it easier). The peripheral
location of the secessionist territory might make it difficult to carry out terrorist attacks,
as the likely homogeneous, self-contained and peripheral nature of the secessionist territory
would make it more difficult to target out-groups, since they will be out of physical reach.
This is in accordance with Fazal’s (forthcoming) argument that secessionist rebel groups are
“unlikely to target their own people (unless to eliminate collaborators) or cultural property,
and also unlikely to have a military reach much beyond their region” (47).
At the same time, the contained nature of the secessionist territory makes it particularly
easy for the government to retaliate against the secessionist population for rebel actions.
The secessionist rebel group, both out of a desire to appeal to the international community
but also to not lose the support of the population they one day wish to govern, will want to
avoid being blamed for provoking government violence.
For these reasons, I propose:
H3b: Secessionist groups that mobilize along ethnic lines are less likely to use terrorism
than secessionist groups that do not mobilize along ethnic lines.
In the case of where secessionist groups mobilize along ethnic lines, we could also expect
a differentation along different levels of conflict intensity. Byman (1998) argues that ethnic
groups (or, presumably, ethnically mobilized secessionist rebel groups) have a need to instill
fear in rival groups in order to provoke “voluntary” emigration and ethnic homogenization.
This is likely more necessary at lower levels of conflict intensity than higher ones, where the
out-group population will likely have fled the conflict zone. In this iteration of my project, I
do not test for the triple interaction term between secession, ethnic mobilization, and war.
16
6
Research Design
6.1
Data and Variables
In order to parse out whether the ethnic characteristics of intrastate wars affect the likelihood
of terrorism, I conduct a global analysis of terrorism from 1970 to 2003. I have combined
and recoded data from several existing datasets into a new dataset, and analyzed the data
using a negative binomial count model.
My independent variable is terrorist attacks per conflict-year (“terrattacks”). The data
comes in part from the Global Terrorist Database (LaFree and Dugan 2007) and was amended
using data from Young and Findley (forthcoming), who recoded some of the GTD, taking
into account the possibility of several intrastate conflicts taking place in a specific country
at the same time, as well as intrastate conflicts being restricted to a specific territory, which
the GTD does not do. Data from 1993 is missing due to data loss within GTD. Incidents
in GTD are coded as terrorist attacks if they substantially concur with the definition of
terrorism also used in this paper (LaFree and Dugan 2007). It should be noted that the data
in GTD was coded (slightly) differently from 1998 to 2004, which in this iteration of my
project I was not able to account for, though it should not bias my findings significantly, as
it is unlikely that this difference affects ethnic conflicts differently from ideological conflicts.
GTD is a particularly useful dataset for this analysis, as it includes not only transnational
terrorist attacks but also domestic terrorist attacks, which is necessary in order to study
terrorism used in the context of intrastate conflicts.
Data on conflict-years is taken from Gleditsch et. al. (2002). My operationalization of
intrastate conflict or war (“civil war”) follows the definition used by Gleditsch et al (2002).
They define an intrastate armed conflict as a contested incompatibility either concerning
government or territory between a states government and a non-governmental party that
results in at least 25 battle-related deaths in one calendar year. I am examining 1172 conflict
years.
17
The use of conflict-year level data requires some justification. Since my theory is one
primarily based on different mobilization needs during ethnic rather than ideological conflicts,
it would be more useful to have precise data on which rebel groups utilized terrorism and
which did not. The issue is with data availability: Within GTD, the majority of terrorist
perpetrators is unknown, which would limit my universe of cases significantly. It is also well
possible that these data omissions are not randomly distributed, but that there is rather
more missing data for more intense conflict, where data gathering is difficult, or in cases of
severe repression (as in the case of very weak rebel groups), where a government has power
to suppress the dispersion of information about terrorist attack perpetrators, as to not give
terrorists additional propaganda. My choice of observational unit therefore circumvents the
attribution problem.
How much does it matter whether the data is at the rebel-group or conflict-year level?
As argued by Young and Findley (forthcoming), it may even be justifiable on purely theoretical grounds to use conflict-year data. In addition to insurgents actively involved in armed
conflict, other (weaker) movements might use terror as a way to increase their support and
participate more fully in a conflict. These groups are likely to be motivated by factors occurring within the conflict zone. By using “ethnic mobilization” as my primary independent
variable, I include most grievance-based measures of ethnic conflict as well, as well as some
in which the expressed grievances of the warring party do not take ethnic overtones. The
original coding of “ethnic” based on “ethnic mobilization” comes from Fearon and Laitin
(2003) and Kreutz (2006), and was combined by Eck (2009). However, it also includes some
cases where mobilization may be along ethnic lines for reasons of geography (if ethnic groups
are fairly segregated) and convenience, even though the conflict may at heart not be of an
ethnic nature. However, if a rebel group’s base is easily identifiable to the rebels, it is also
easily identifiable for government retribution. In future iterations of this project, I hope to
include an alternative operationalization of “ethnic conflict.”
The operationalization of “secession” follows Gleditsch et. al. (2002), simply measures
18
whether the incompatibility between government and rebel group is concerns the status of
a particular territory, such as secession or autonomy.
My operationalization of conflict intensity follows the basic conceptualization of conflict
intensity from Gleditsch et. al. (2002). Conflict intensity, as persuasively argued by Lindberg
(2008), can include both direct battle-related fatalities and indirect deaths due to disease
as well as the destruction of infrastructure. However, we can infer that a battle-death
proxy for conflict intensity is valid in that battle deaths are likely to be highly correlated
to indirect fatalities and infrastructure destruction. I considered two different specifications
for conflict intensity: the move from low-scale armed conflict to high-fatality violence of
war (as used by Eck and taken from Gleditsch et. al. (2002)) and the actual number of
battle deaths. However, the precise number of battle deaths is only available for two thirds
of my observations; I furthermore suspect that the missing observations are not randomly
distributed, but rather are skewed towards a high conflict intensity. I have run my model
(without intensity specification) including all of the observations and then excluding the
missing observations for the number of battle deaths, and found that the results were very
different, which implies a selection issue. I have therefore decided to use the more simple but
also more conventional specification of 0 and 1 for conflict intensity (“war”), where conflicts
experiencing between 25 and 999 battle deaths in a given conflict year are coded 0 and
conflicts that experience more than 1000 are coded 1.
I also use a number of control variables: I use measures for democracy (democracies
are often associated with higher levels of terrorism use), GDP (often associated with less
terrorism), as well as pluralism (a measure for the proportion of the population belonging
to the largest ethnic group) (data from Eck (2009), who adapted it from Fearon and Laitin
(2003)).
Finally, I have constructed interaction terms for secession and ethnic conflicts, ethnic
wars and conflict intensity, and secession and conflict intensity. This allows me to see how
conflict intensity affects different types of conflict differently.
19
6.2
6.2.1
Model and Results
Model
I am using a negative binomial count model to measure the effect ethnic mobilization, secession, and conflict intensity on the use of terrorist attacks. There are several justifications
for this choice. As King (1988) argued, ordinary least squares analysis of event counts is
inefficient, have inconsistent standard errors, and may produce unreasonable predictions of
negative number of events. An event count model is more appropriate. While the Poisson
model is the simplest event count model, it assumes that the probability of an event occurring at any given time is constant within a period and independent of all previous events
during that period. In the case of terrorism, this may be an overly optimistic assumption,
as we could well assume that a group will continue to use terrorism once they have done
so, or that several groups will revert to the use of terrorism in an attempt to outbid one
another (see Young and Findley forthcoming). If this assumption of homogeneity does not
hold, estimators based on the Poisson regression, while consistent, will be inefficient, and
standard errors will be inconsistent (King 1988). Terrorist attacks may therefore be “contagious.” We can also have contagion across observations. Contagious processes may lead
to either over-dispersion (the variance exceeds the mean) or under-dispersion (the variance
becomes smaller than the mean), which would violate the homogeneity assumption. In the
case of over-dispersion, a negative binomial model may be more appropriate. My variance is
a lot larger than my standard deviation, which indeed indicates over-dispersion. I therefore
ran my model using a negative binomial count model.
6.2.2
Results
I first calculate the mean number of terrorist attacks per conflict year for ethnic and nonethnic conflicts for both low and high conflict intensity in Table 1. These results do not
take into account other factors, though, and do not report statistical significance, but they
20
do illustrate an overall trend. While the average number of attacks per conflict year does
not change much for ethnic conflicts, it increases a lot for non-ethnic wars, which overall
supports my theory. 1.
Table 1: Summary of terrorist attacks, ethnic
War = 0
War = 1
23.18
43.74
32.39
136
Ethnic
Not Ethnic
A similar comparison is available for secessionist conflicts based on conflict intensity in
Table 2. This table indicates that there is more terrorism for non-secessionist conflicts rather
than secessionist ones across the board, and that the use of terrorism for secessionist conflicts
does not increase to the degree it does for non-secessionist to the same degree.
Table 2: Summary of terrorist attacks, secession
War = 0
War = 1
22.33
37.86
47.90
82.96
Secessionist
Not Secessionist
In order to explore the statistical significance of these results, I provide several versions
of my negative binomial count model in Table 3. The results indicate that, overall, ethnic
mobilization does increase the use of terrorism overall. Secessionist conflicts are likely to
experience less terrorism, though this value is only significant if there is no interaction term
included for ethnic and secession, which indicates that ethnic secessionist conflicts experience
less terrorism. On its own, “secession” is associated with a p-value of .132, which for some
might approach an acceptable level of statistical significance.
Once we look at “ethnic” interacted with “war”, the coefficient is negative and statistically significant, which provides strong evidence that higher conflict intensity does decrease
the likelihood of terrorism being used. Secession and war, though, does not seem to have
such an effect.
21
Table 3: Negative binomial regression - dependent variable: terrorist attacks per conflict
year
Model 1
ethnic
secession
war
plural
gdp1
demo
ethnicXwar
ethnicXsecession
secessionXwar
Constant
Model 2
Model 3
0.590**
0.287
0.190
(2.83)
(1.75)
(1.20)
-0.343 -1.366*** -1.274***
(-0.73)
(-9.21)
(-8.96)
1.308*** 1.278*** 1.129***
(3.85)
(3.73)
(5.35)
2.299*** 1.877*** 1.892***
(7.22)
(7.11)
(7.20)
0.000*** 0.000*** 0.000***
(5.65)
(6.59)
(6.63)
1.728*** 1.591*** 1.583***
(12.39)
(12.45)
(12.38)
-1.150*
-1.095*
(-2.25)
(-2.23)
-1.282*
(-2.40)
1.158*
1.141*
(2.25)
(2.30)
0.970*** 1.293*** 1.300***
(3.92)
(6.08)
(6.13)
lnalpha
1.167***
(25.24)
22
1.174***
(25.41)
1.181***
(25.60)
The size of coefficients cannot be directly interpreted in a negative binomial count model.
In order to more easily interpret my results, I have calculated the expected change in the
number of terrorist attacks per conflict year. If war is 0 (low conflict intensity) the effect of
“ethnic” going from 0 to 1 is 8.88, or increases the number of terrorist attacks per conflictyear by 8.88 (holding all other values at their mean). However, if war is 1, then the effect
of ethnic on terrorist attacks is -3.16. Similarly, the change in number of terrorist attacks
if “secession” goes from 0 to 1 is -5.84 at war = 0. If war = 1, the number of expected
terrorist attacks is actually expected to increase by 29.5. This, however, is still less than the
increase if war goes from 0 to 1 for conflicts that are neither secessionist nor ethnic, where
the expected number of terrorist attacks increases by 40.84.
What do these results imply about my hypotheses? Ethnic mobilization affects the
strategic use of terrorism differently depending on the level of conflict intensity. While a low
level of conflict increases the use of terrorism by ethnically mobilized rebel groups (probably
because of the ease of targeting and possibly because of the higher mobilization pay-off),
all else held equal ethnically mobilized rebel groups are less likely to use terrorism as level
of conflict intensity increases. This is in stark contrast with the effect of conflict intensity
increase for ideological rebel groups, who are significantly more likely to use terrorism as
conflict intensity levels increase. Secessionist conflicts associated with an overall lower rate
of terrorism than non-secessionist conflicts, though this result only is statistically significant
for secessionist conflicts that are also ethnically mobilized. In short, all of my hypotheses
find (at least limited) support.
7
Conclusions
I have shown that the use of terrorism by ethnically versus ideologically mobilized rebel
groups significantly depends on the level of conflict intensity. I have theorized that this is
due to differences in mobilization logics, which change for ethnically mobilizing rebel groups
23
along conflict intensity changes, but remain the same for ideological rebel groups. In fact,
the use of terrorism is more likely at a higher conflict level for ideological rebel groups. I
have further found support for the effect of secessionist aims on the use of terrorism, and
also shown that this is only significant for secessionist groups that also mobilize along ethnic
lines.
In further work, it would be interesting to include a measure for group grievance; terrorism
as a mobilization tool can be expected to have a particularly strong effect if the group‘s goals
seem out of reach or hopeless. An interesting measure would be Wimmer et. al. (2009)’s
measure of ethnic group’s exclusion from power. One weak proxy for this is the “pluralism”
measure included in the regression, which measures power imbalances between dominant and
minority ethnic groups. It seems to increase the use of terrorism significantly, but it would
be interesting to measure how it interacts with some of the other variables, though this also
would be more fruitful with a more precise measure of exclusion from power.
These results could be further strengthened through including a measure for ethnic segregation within a state; this could get at the potential difficulties groups have to actually
target out-group members, and at the relative ease with which the government could target members of the ethnic population in retribution against terrorism. Further, it could be
interesting to look at the effect of religious, and not only ethnic, fractionalization or power
relations.
Nonetheless, I have shown that ethnicity has a more complex effect on the use of terrorism
in the context of intrastate conflict than is generally acknowledged.
24
References
Ahn, T. K. and E. Ostrom. 2001. “Social capital and the second-generation theories of
collective action: An analytical approach to the forms of social capital.” Annual Meeting
of the American Political Science Association, Boston, MA .
Bapat, N.A. 2006. “State bargaining with transnational terrorist groups.” International
Studies Quarterly 50(1):213–230.
Bloom, Mia. 2007. Dying to Kill: The Allure of Suicide Terror. Capstone.
Bueno de Mesquita, Ethan and Eric S Dickson. 2007. “The Propaganda of the Deed:
Terrorism, Counterterrorism, and Mobilization.” American Journal of Political Science
51(2):364–381.
Byman, D. 1998. “The logic of ethnic terrorism.” Studies in Conflict & Terrorism 21(2):149–
169.
Chandra, K. 2006. “What is ethnic identity and does it matter?” Annu. Rev. Polit. Sci.
9:397–424.
Crenshaw, M. 2011. Explaining Terrorism: Causes, Processes and Consequences. Routledge.
Easterly, W. 2000. Can institutions resolve ethnic conflict? Vol. 2482 World Bank Publications.
Eck, K. 2009. “From Armed Conflict to War: Ethnic Mobilization and Conflict Intensification.” International Studies Quarterly 53(2):369–388.
Fazal, Tanisha. Forthcoming. Declaring War and Peace. Book Manuscript.
Fearon, J.D. and D.D. Laitin. 2003. “Ethnicity, insurgency, and civil war.” American Political
Science Review 97(1):75–90.
25
Fortna, V. Forthcoming. “Do Terrorists Win? Rebels’ Use of Terrorism and Civil War
Outcomes.”.
Gilley, B. 2004. “Against the concept of ethnic conflict.” Third World Quarterly 25(6):1155–
1166.
Gleditsch, N. P, P. Wallensteen, M. Eriksson, M. Sollenberg and H. Strand. 2002. “Armed
conflict 1946-2001: A new dataset.” Journal of Peace Research 39(5):615–637.
Gurr, T. R. 2000. Peoples versus states: Minorities at risk in the new century. United States
Inst of Peace Pr.
Horowitz, Donald. 1985. Ethnic Groups in Conflict. Berkeley: University of California Press.
Kaufman, C. 1996. “Possible and impossible solutions to ethnic civil wars.” International
security 20(4):136–175.
King, G. 1988. “Statistical models for political science event counts: Bias in conventional
procedures and evidence for the exponential Poisson regression model.” American Journal
of Political Science pp. 838–863.
Kreutz, J. 2006. “Once Again: A New Test of Recurring Intrastate Conflicts, 1946-2001.”
Paper presented at the 47th International Studies Association Convention, San Diego, CA
.
Kydd, A. H and B. F Walter. 2006. “The strategies of terrorism.” International Security
31(1):49–80.
LaFree, G. and L. Dugan. 2007. “Introducing the global terrorism database.” Terrorism and
Political Violence 19(2).
Lichbach, M. I. 1996. The cooperator’s dilemma. Univ of Michigan Pr.
26
Lindberg, J. E. 2008. “Running on Faith? A Quantitative Analysis of the Effect of Religious
Cleavages on the Intensity and Duration of Internal Conflicts.” Master’s Thesis, University
of Oslo .
Merari, A. 1993. “Terrorism as a Strategy of Insurgency.” Terrorism and P 5(4).
Mueller, J. 2000. “The banality of “ethnic war”.” International Security 25(1):42–70.
Pape, R. A. 2005. Dying to win: The logic of suicide terrorism. New York: Random House.
Posen, B. R. 1993. “The security dilemma and ethnic conflict.” Survival 35(1):27–47.
Sambanis, Nicholas. 2000. “Partition as a Solution to Ethnic War: An Empirical Critique
of the Theoretical Literature.” World Politics 52(4):437–483.
Schmid, A. P. and A. J. Jongman. 2005. Political Terrorism: a new guide to actors, authors,
concepts, data bases, theories, & literature. Transaction Publishing.
Stanton, Jessica. Forthcoming. “Terrorism in the Context of Civil War.”.
Wimmer, A, L.E. Cederman and B. Min. 2009. “Ethnic politics and armed conflict: a
configurational analysis of a new global data set.” American Sociological Review 74(2).
Wood, R. 2011. “The Strength to Kill or the Power to Protect?: Power Resources and
Violence against Civilians in Civil Conflict.”.
Young, Joseph K. and Michael G. Findley. Forthcoming. “More Combatant Groups, More
Terror?: Empirical Tests on an Outbidding Logic.” Terrorism and Political Violence .
27