16-conrad

Competition, Differentiation, and the Severity of Terrorist Attacks
Author(s): Justin Conrad and Kevin Greene

Source: The Journal of Politics, Vol. 77, No. 2 (April 2015), pp. 546-561
Published by: The University of Chicago Press on behalf of the Southern Political Science
Association
Stable URL: http://www.jstor.org/stable/10.1086/680262 .
Accessed: 01/10/2015 03:30
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp
.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact [email protected].
The University of Chicago Press and Southern Political Science Association are collaborating with JSTOR to
digitize, preserve and extend access to The Journal of Politics.
http://www.jstor.org
This content downloaded from 132.203.227.61 on Thu, 1 Oct 2015 03:30:02 AM

All use subject to JSTOR Terms and Conditions
Competition, Differentiation, and the Severity
of Terrorist Attacks
Justin Conrad, University of North Carolina at Charlotte
Kevin Greene, Michigan State University
A wide range of literature on ethnic conflict and terrorism has argued that domestic competition increases the likelihood
that a political organization will use violence in an effort to distinguish itself. Known as “outbidding,” empirical evidence
for such a phenomenon has thus far been limited. The bulk of the empirical analysis, however, has focused on the effect
of domestic competition on the quantity of violence. This study instead argues that competition should have an ob-
servable effect on the quality of violence, as organizations seek to differentiate their “brand” from others. Using in-
formation on the tactics and targets of terrorist attacks, the results suggest that an increase in the competitiveness of a
political market leads to more severe or “shocking” types of attacks.
W hen explaining the behavior of terrorist orga-

nizations, scholars frequently cite a strategy of
“outbidding” as a motivating factor in the deci-
sion to launch attacks. When a terrorist organization ex-
periences an increase in domestic political competition, they
tics because it distinguishes their “brand” from competi-
tors, ultimately making the organization less sensitive to
competition.
While previous treatments of outbidding have implicitly
acknowledged the role of competition in the tactical choices
may respond by increasing their level of violence in order to of organizations, almost all have focused on a single tactic:
“outbid” their competition in garnering public support (e.g., suicide terrorism (e.g., Bloom 2005). Although limited evi-
Bloom 2005; Crenshaw 1985, 1987; Kydd and Walter 2006; dence of a relationship between domestic competition and
Oots 1989).1 Hamas, which used unprecedented levels of the use of suicide terrorism exists, we argue that suicide at-
violence during the First Intifada in an effort to outbid other tacks are simply one of many tactical choices that an orga-
Palestinian groups, is often cited as the classic example of a nization can make to distinguish itself. We therefore analyze
terrorist organization engaging in such behavior. Despite how domestic political competition affects the likelihood
the intuitiveness of this strategy, however, empirical tests of of a wide range of attacks based on target types and the
outbidding (e.g., Chenoweth 2010; Findley and Young 2012; methods used during the attack. All terrorist attacks are not
Nemeth 2013) have found limited support for the theory. created equal, and some have a larger impact than others,
This article argues that previous literature has not fully ex- either because of who is being targeted or the kind of attack
amined the hypothetical implications of outbidding theory. that is carried out. Attacks on civilians, for instance, are far
Most of these studies only analyze the effect of competition more shocking than attacks on rival political groups, and
on the quantity of violence used by terrorist organizations, bombings are more shocking than unarmed assaults. The
while often overlooking that domestic competition should nightmarish “human lawnmower” proposed by the former
have a more discernible effect on the quality of violence. In al Qaeda in Iraq involved attaching moving blades to the
particularly competitive environments, violent groups are front of a pickup truck and driving it through a crowded
incentivized to engage in more shocking or innovative tac- public venue, providing anecdotal support that terrorist or-
Justin Conrad is an assistant professor in the Department of Political Science and Public Administration at the University of North Carolina at Charlotte,
Charlotte, NC 28223. Kevin Greene is a Ph.D. student in the Department of Political Science at Michigan State University, East Lansing, MI 48824.
1. Data and supporting materials necessary to reproduce the numerical results in the article will be made available at the authors’ professional websites.
The Journal of Politics, volume 77, number 2. Published online February 6, 2015. http://dx.doi.org/10.1086/680262
q 2015 by the Southern Political Science Association. All rights reserved. 0022-3816/2015/7702-0018 $10.00 546

Volume 77 Number 2 2015 / 547
ganizations in competitive marketplaces (such as Iraq dur- (2002, 17–19) notes that terrorism offers the hope of shift-
ing the height of the war there) have actively tried to launch ing the balance of power between the terrorist organization
more “shocking” attacks to distinguish themselves.2 and the state, giving the organization better bargaining power
The wide variation in the severity of attacks and their in the future. The terrorists’ best chance of shifting the bal-
anticipated psychological effects must be accounted for in ance of power is by convincing more and more people in the
order to properly test the logic of outbidding. To do this, we target audience to join, or at least sympathize with, their cause
have categorized domestic terrorist attacks by their relative (Kydd and Walter 2006). To that end, terrorist organizations
level of severity or “shock value.”3 We measure severity in allocate resources and make tactical choices in an effort to
two ways: the severity level based on the target type and maximize their utility (Caplan 2006, 94).
the severity level of the methods used in the attack. The Organizational theories, including those focusing on ter-
severity level therefore captures the likely impact of an act rorist groups, argue that political organizations must balance
based on the identity of the victim and how the attack was their strategic goals with the need to sustain the organiza-
carried out. tion (Crenshaw 1985, 1987; Wilson 1995). Self-preservation,
In the following sections, we lay out how past studies therefore, is a fundamental concern of all political organi-
have analyzed the relationship between domestic compe- zations. Crenshaw (1985) notes that while terrorist organi-
tition and terrorism and then incorporate the economic zations and their leaders, in particular, have long-term goals,
concept of differentiation to explain how terrorist organi- a primary focus of most organizations is survival. Since the
zations may derive greater utility by using more extreme membership size of terrorist organizations and the level
or shocking attacks. By applying the logic of traditional of popular support they enjoy seem to be correlated with
firms in competitive markets, we are able to better under- their longevity (Blomberg, Gaibulloev, and Sandler 2011;
stand why terrorist organizations might use different types Cronin 2006, 2009), increasing recruitment and/or public
of violence when faced with domestic competition. This in- support may lead to greater chances of survival and, perhaps,
sight also helps disentangle the strategic and organizational greater chances of achieving long-term strategic goals. In
processes of terrorist organizations; we argue that the use environments where there are multiple competing organi-
of extreme violence primarily benefits the organization in zations, however, terrorist organizations must vie for ac-
short-term recruitment and “shoring up” of support, de- cess to a limited pool of resources (e.g., media attention, re-
spite the fact that it may harm their long-term strategic cruits, public support, etc.). Since competition directly and
goals. Analyzing patterns of violence at both the state and indirectly threatens the resource base necessary to sustain
organization levels, the results of the study suggest that, on the organization and ensure its effectiveness, it follows that
average, greater domestic political competition leads to an terrorist organizations should make tactical choices in an
increase in the severity of attacks. effort to increase their share of resources within a compet-
itive environment.
RATIONALISM AND OUTBIDDING The theory of outbidding (e.g., Bloom 2005) is based on
Much of the literature examining the relationship between such logic, arguing that terrorist organizations seek to gain
political competition and terrorism assumes that terrorist the upper hand in media coverage, recruiting, and public
organizations are rational actors that select strategies and support when there are multiple competing parties. Bloom
tactics offering them the highest expected utility. The very argues that “where there are multiple groups, violence is a
choice to engage in terrorism is thought to be indicative technique to gain credibility and win the public relation
of a group which expects a lower utility from conventional competition” (2005, 95). Further, outbidding is more likely
tactics (e.g., Crenshaw 2002; Fromkin 1975; Lake 2002). when there is a greater number of actors competing for the
While terrorism rarely seems to achieve the stated political same resource base (Bloom 2005). Outbidding is expected
goals of most organizations (Abrahms 2006, 2012), Lake to occur, therefore, when two key conditions hold: two or
more domestic parties are competing for the leadership role
2. J. Warrick, “Bin Laden’s Last Stand,” Washingington Post, April 30,
of a particular cause, and the general population is uncer-
2012. http://articles.washingtonpost.com/2012-04-30/world/35452658_1 tain about which of the groups best represents their inter-
_qaeda-osama-bin-laden. ests (Kydd and Walter 2006). With incomplete informa-
3. Throughout the study, we use terms such as “severity,” “more ex- tion, the public cannot be sure which group is the most
treme,” and “shock value” interchangeably. The concept which we are
capturing, however, is the same: some attacks have greater psychological
committed to the cause. The public (from which recruits are
impacts, garner greater media coverage, and are likely to influence a wider also drawn) may view the organization as either “a strong
audience than others. and resolute defender of the cause (zealots) or weak and in-

548 / Differentiation and Severity Justin Conrad and Kevin Greene
effective stooges of the enemy (sellouts)” (Kydd and Walter rivals (Bloom 2005). As Enders and Sandler (2006) note,
2006, 76). Chenoweth (2010) adds that increased competi- changes in the costs and benefits “of one type of terrorist
tion creates conflict by default, even when two or more po- activity will cause the terrorist group to substitute” another
litical organizations are ostensibly fighting for the same cause. activity which provides a greater expected utility. From this
Even in nonviolent political interactions, the public is likely perspective, if it becomes more costly to use a particular
to favor more extremist representatives, who are seen as pro- tactic, or if the benefits of that tactic decline, terrorist or-
viding a stronger bulwark against the enemy (Canes-Wrone ganizations should seek alternative tactics. And increased
and Shotts 2007). competition among terrorist organizations reduces the ben-
In competitive environments, therefore, organizations efits derived from tactics that have already been used. In such
have additional incentives to exaggerate their strength and environments, organizations should look to differentiate
commitment to stand out from rivals. By engaging in par- themselves with new and innovative tactics.
ticularly noteworthy violence, groups can signal resolve, cre-
ativity, and credibility. Such signaling, in turn, can indicate DIFFERENTIATION OF TERRORIST ATTACKS
to the public that their group has the best odds of revising the Why does the quality of terrorist attacks matter? The logic
status quo. This is significant for terrorist organizations be- of differentiation in traditional economic markets has dem-
cause, while other political organizations might campaign onstrated how quality helps firms and brands distinguish
for support based on past victories or accomplishments, ter- themselves from competitors (Chamberlin 1933; Robinson
rorists must frequently campaign on future achievements. 1933; Smith 1956). Differentiation was traditionally thought
This increases a terrorist organization’s need to signal that it of as the ability of a firm to charge a higher price for its prod-
has a legitimate chance of achieving its goals over time and uct or make a larger profit (see Sharp and Dawes 2001). Sharp
that it can impose costs on those who stand against it. In this and Dawes argue that differentiation occurs “when a firm/
respect, the literature on outbidding has focused largely on brand outperforms rival brands in the provision of a fea-
the phenomenon of suicide terrorism, with Bloom (2005) ture(s) such that it faces reduced sensitivity for other fea-
and others arguing that this particular tactic is a function of tures” (2001, 739). Baker (1996), Mercer (1992), Dickson
competitive political environments. The willingness of an (1997), and Powers (1991) all note that firms differentiate
organization to sacrifice one of its own members (not to their goods or services in order to achieve a reduced price
mention the willingness of an individual to sacrifice herself ) sensitivity or justify a higher price. In other words, differen-
in order to further a cause seems to signal the deepest level tiation can be thought of as anything that allows a firm and/
of commitment. or its products to “stand out” in a given marketplace (Guil-
Importantly, this argument about how competition leads tinan and Paul 1991; Kotler et al. 1996; Saunders 1995). As a
to suicide terrorism is an argument not about the frequency firm’s products become more differentiated relative to com-
of violence, but rather the type of violence that is chosen. peting products, the firm experiences reduced sensitivity to
Especially in environments where groups are already using other factors. This reduced sensitivity, in turn, benefits a firm
violence, the quality of violence may be the most effective or organization by diminishing the threat from direct com-
way to distinguish themselves from competitors. Hamas’ use petition and allowing the firm to capture a greater market
of suicide terrorism in the early 1990s was seen as a means share (Sharp and Dawes 2001). And successful differentia-
of distinguishing itself from Fatah, which had already used tion can increase brand loyalty, resulting in long-term sup-
more “traditional” types of terrorist attacks. In other words, port for a product even in the face of short-term setbacks
Hamas likely would not derive much purchase from “more (Pinson and Brosdahl 2014).
of the same” but instead generated support by choosing a Differentiation of organizations, therefore, leads to a
relatively innovative tactic.4 The Liberation Tigers of Tamil higher “premium” associated with their activities and, ulti-
Eelam (LTTE), one of the first organizations to use suicide mately, a decreased sensitivity to other potential costs cre-
attacks, also used suicide terrorism when the political envi- ated by competition. While differentiation has generally been
ronment in Sri Lanka was very competitive, but they reduced applied to economic markets, the same logic should apply to
their use of the tactic after they had defeated their political any organization trying to stand out in its respective “mar-
ketplace.” In this case, the “firms” are terrorist organizations,
whose “products” are the attacks they carry out. Much like
4. The tactic of suicide terrorism was not innovative within Israel and
the Palestinian territories per se. It was arguably pioneered by Hezbollah
traditional firms, terrorist organizations are forced to com-
in Lebanon some 10 years earlier, but the First Intifada in Israel marks the pete directly with other firms (nonviolent or violent political
beginning of its widespread use in the Israeli-Palestinian conflict. groups). Also like traditional firms, terrorist organizations

seek to control a larger market share and to overcome their

competitors.
The overall efficacy of terrorist organizations relies
largely on how the public perceives them (Lake 2002). While
committing more shocking attacks may not indicate greater
strength or help the group achieve long-term goals, the
spectacle of more severe attacks might generate the per-
ception that the organization is more effective than it ac-
tually is. As such, terrorist organizations may expect to de-
rive greater utility by focusing on a few high-yield attacks
rather than carrying out a large number of attacks with
smaller impacts. It is reasonable to argue that combined, all
of al Qaeda’s attacks prior to, and following 9/11, did not
have the same impact as that single attack. By executing an
innovative or shocking method of attack, or by attacking a
high-profile target, terrorist organizations should be better
able to distinguish themselves in a given marketplace, par-
ticularly with respect to their competitors. For terrorist or- Figure 2. Differentiated organization.
ganizations, more innovative and shocking tactics make the
organization less sensitive to ineffectiveness in achieving
their strategic goals. Differentiation therefore allows an or- public support and recruitment (demand) will be less sen-
ganization to steepen its demand curve, making it less sensitive to a group’s achievement of its goals (supply) when a
sitive to changes in the valuation of its product. group successfully differentiates itself. The graphs are iden-
The modified supply and demand curves in Figures 1 tical, except that the graph in Figure 2 has a steeper demand
and 2 illustrate this relationship between a terrorist orga- curve than the graph in Figure 1. The steepness of the de-
nization’s level of public support and the perception of the mand curve is determined by elasticity of demand, or how
group’s chance of achieving their stated political goals. The sensitive one variable is to a change in another. In tradi-
x-axis in each figure represents the organization’s level of tional economic studies, elasticity determines how respon-
public support, while the y-axis represents the public per- sive customers are to a change in price. In our model, elas-
ception of the group’s probability of success. We argue that ticity measures how sensitive a terrorist organization’s public
support is to changes in the perception of their chances of
success. The steeper demand curve for a differentiated or-
ganization indicates that it will experience a lower rate of
decline in public support when its perceived chances of suc-
cess fall, relative to an undifferentiated organization. This is
represented by the large gap between s1 and s2 in Figure 1 and
the smaller gap between s1 and s2 in Figure 2.
The use of more severe attacks may steepen the demand
curve in a number of ways. The spectacle of more extreme
attacks generates more attention and signals that a group is
committed to achieving their goals, regardless of their ac-
tual level of success. Commentators have suggested that al-
Shabaab was at its weakest in late 2013, when their spec-
tacular attack against a Kenyan mall significantly raised their
profile (see Young 2013). Groups that do not effectively dif-
ferentiate themselves will have a more shallow demand
curve. As a result, any perception that the group is weak or
has a low chance of success has a greater influence on their
ability to gain support and recruits (Sharp and Dawes 2001,
Figure 1. Undifferentiated organization. 755). Additionally, the use of more extreme attacks can

create barriers to entry for other terrorist organizations that But differentiation through violence may be the key to
are not willing or able to carry out these types of attacks. increasing recruitment, deterring defection, and maintain-
Aside from creating moral dilemmas for some organiza- ing public support among the subset of the population that
tions, the use of more extreme attacks might simply reduce approves of such violence. As Chai (1993) has demon-
the number of available substitutes. Because of the higher strated, more violent organizations attract more violent re-
shock value of increasingly severe attacks, there may be less cruits. So at least in the short-term, extreme violence serves
room for differentiation by default. the important function of increasing recruitment and sup-
Qualitative differentiation offers additional benefits for a port among such individuals. There is some evidence, for
terrorist organization because tactics can be copied (Sharp example, that many individuals joining the Islamic State of
and Dawes 2001). This possibility raises the incentive for Iraq and the Levant (ISIL) during the summer of 2014 did
each terrorist organization to devise some signature attack so precisely because they were using severe violence.5 These
or innovative tactic that separates them from other groups. types of recruits are, in turn, more willing to commit ex-
Further, it is most efficient for an organization to differen- treme acts of violence that may further assist a group during
tiate on an aspect that is low cost to provide which results in periods of increased competition. While the size of this vi-
decreased sensitivity to other costs (Sharp and Dawes 2001, olent subset of the population is limited, maintaining their
752). For terrorist organizations, some of the most extreme support can often be crucial in ensuring that the organiza-
attacks are also the lowest costs to deliver, such as attacks tion survives from one day to the next. Violence, then, while
against unprotected civilians and targets like shopping malls ostensibly serving a strategic purpose, may be far more im-
and restaurants. And we argue that such attacks are likely to portant in the maintenance of the organization.
lead to decreased sensitivity to the public perception of the And while extreme violence risks public backlash, it may
group’s success, something that is much more difficult to also effectively coerce members of the broader population
deliver. For instance, a group that sufficiently differentiates to support the organization. By targeting civilians and en-
itself from competitors might find that an event like a gov- gaging in shocking displays of violence, a terrorist organi-
ernment crackdown on their operations or the death of a zation can demonstrate that the state is incapable of pro-
key leader—events that signal declining effectiveness—have tecting the population, encouraging more of the population
less of an impact on their ability to garner support and re- to view the organization as a more credible and capable
cruits. A group with a steeper demand curve made possible actor. Heavy-handed responses by the state may also drive
through differentiation will take a smaller hit in terms of members of the public toward support of the terrorist or-
public opinion and recruitment relative to an undifferenti- ganization (Kydd and Walter 2006; Lake 2002). Downes
ated group. In an unusually competitive market, this might (2006), Thomas (2014), and Wood (2014) all find that the
be the edge that a terrorist organization needs to sustain itself use of violence against civilians can be an effective strategy
or to build enough of a support base to overtake a compet- for weak actors.
itor. Successful differentiation, therefore, offers a useful tool Civilian targeting, for instance, signals to the government
to mask an organization’s failure to achieve their political and the broader population that even if an organization
goals. Among all political organizations, then, it can be argued cannot defeat government forces, they are willing to im-
that differentiation is most important to terrorist groups, pose significant costs on the public (Wood 2014). Weak
precisely because they are unlikely to achieve their stated goals groups should be especially likely to target civilians, because
(Abrahms 2012). stronger groups have more resources to offer positive in-
ducements (Wood 2010). In competitive environments all
Organizational Benefits vs. Strategic Risks terrorist organizations are, by default, relatively weak. So
Engaging in such extreme behavior, however, presents a while the use of more severe attacks offers the risk of back-
number of risks. For instance, the use of extreme violence lash from the civilian population (Kalyvas 2006), it also of-
can potentially lead to public backlash against the organi- fers the possibility of increasing recruitment and support
zation (Crenshaw 1991, 2002; Cronin 2009), reducing its among a violent subset of the population while simulta-
ability to achieve its broader strategic goals. And violence neously coercing the broader population.
may actually be counterproductive compared to nonviolent
campaigns (Chenoweth and Stephan 2011). Abrahms (2008,
5. S. Brown, “Young people ‘attracted to Isis by its brutality,’ ” The
2012) specifically finds that terrorist groups that use vio- Independent, Oct. 25, 2014, http://www.independent.co.uk/news/world
lence against citizens are unlikely to achieve their stated /europe/young-people-attracted-to-isis-by-its-brutality-says-german
political goals. -security-chief-9702661.html.

THE QUALITY OF VIOLENCE AND MEDIA COVERAGE wald (2006) note that a terrorist attack must pass an “emo-
The key mechanism by which terrorists are able to commu- tional threshold” in order for the media to cover the event,
nicate their message, and successfully differentiate them- and because homicide (especially one caused by a terrorist
selves from competitors, is through media coverage. Terror- attack) is a relatively rare event, this is generally “news-
ist organizations rely on media coverage to convey their worthy.” The identity of the victim also seems to matter, as
message and spread fear among a broader audience than the Delli Carpini and Williams (1987) find that the amount of
immediate victims of their attacks (Hoffman 2006). One coverage of attacks against members of the military and
study (Keinan, Sadeh, and Rosen 2003) found that indi- government fluctuate widely, while coverage of attacks
viduals who watched media coverage of 9/11 experienced against private citizens are consistently overcovered. Al-
symptoms similar to posttraumatic stress disorder, and the though ISIL has engaged in a variety of violent activities,
same is true for Israelis who viewed coverage of terrorist at- it was the beheading of American journalists James Foley
tacks (Bleich, Gelkopf, and Solomon 2003) and individuals and Steven Sotloff that dominated the news cycle and gave
who watched coverage of the Boston Marathon bombing increased exposure to the organization, which had risen
(Holman, Garfin, and Silver 2014). Without media coverage, out of the especially competitive environment of the Syrian
the impact of a terrorist attack becomes more limited in Civil War.
geographic scope: Hoffman, Shelton, and Cleven note that
media coverage is particularly important when a terrorist ASSESSING THE EVIDENCE
group’s “publicity goals exceed their propagandists’ reach” The original logic of outbidding seems quite intuitive: more
(2013, 899). armed groups create an incentive to distinguish oneself.
If media coverage is the oxygen of terrorism, then ter- Despite this intuitiveness, past studies of outbidding (see
rorist groups in competitive markets gain a significant ad- Findley and Young 2012) find little empirical support for
vantage relative to their competition if they are better able the theory. Nemeth (2013) does find limited support for
to capture the media’s attention. Engaging in more severe outbidding but notes that it may depend on the country’s
types of attacks and against more “newsworthy” targets is acceptance of violence. Nemeth also finds that left-wing
one way to ensure media coverage. Since most terrorist at- organizations reduce their violence when faced with do-
tacks receive no media coverage (Chermak and Gruenewald mestic pressure, while religious and nationalist groups carry
2006; Paletz, Fozzard, and Ayanian 1982; Weimann and out more attacks when faced with domestic competition.
Winn 1994), there is an incentive for terrorists to pull off These studies share a common approach to measuring
attacks that are particularly likely to be covered. Scott notes evidence of outbidding. Findley and Young (2012) analyze
that, “to compete successfully for media attention, terror- the effect of competition on the number of suicide and
ists must be original enough to stage incidents that are a nonsuicide terrorist attacks that occur in a given year, both
departure from past events. Hence, large media returns to domestically and transnationally. Nemeth (2013) examines
terrorism come mostly from the perpetrators’ imaginative the number of domestic and transnational attacks, respec-
abilities” (2001, 126). This aligns with the logic of this ar- tively, that occur in a given year. In other words, all of these
ticle: terrorist organizations may receive at least short-term studies have analyzed the effect of terrorist competition on
benefits from the use of more innovative attack types and the quantity of terrorism. The logic of outbidding, and the
by attacking more “shocking” targets that will influence the points that we have made about differentiation, however,
media, and ultimately, the general public. imply that simply increasing the amount of violence may
Specifically, Weimann and Winn (1994) find that attacks not be the only (or even a useful) way to distinguish an or-
that cause harm to individuals are twice as likely to be ganization. Raw counts of the number of attacks, therefore,
covered as attacks that do not. Paletz, Fozzard, and Ayanian do not capture the fact that the quality of terrorist attacks
(1982) find that attacks that involve violence receive a (in terms of severity) varies widely. For this reason, we ex-
greater amount of coverage compared to nonviolent meth- pect that evidence of differentiation may be more apparent
ods. Nacos (2003) finds that attacks that kill a large number when considering who the terrorist organizations are tar-
of people or cause large amounts of destruction are likely geting and the methods of attack that they use. Attacks
to receive more media attention. Weimann and Winn (1994) against “soft targets” like the civilian population at the
and Delli Carpini and Williams (1987) find that kidnap- 2013 Boston Marathon better differentiate terrorists com-
pings, hijacking, and hostage situations are the most likely pared to attacks against government and military targets.
to receive coverage and that attacks that result in death or Terrorists, therefore, consider the “shock value” of their at-
injury are more likely to be covered. Chermak and Gruene- tacks, and it is reasonable to believe that increasing “shock

value” may be one way in which organizations can stand out in increasingly extreme tactics. As the quantity of attacks
in a crowded marketplace. Existing studies that have ex- increases, the marginal utility of any one attack decreases,
plored this possibility have focused almost exclusively on the thus organizations may engage in attacks with higher “shock
use of suicide terrorism. But Piazza (2008) notes that from value” as a way to stand out in a market already oversatu-
1968 to 2005 only 3.36% of the total number of domestic rated with violence:
terrorist attacks were suicide attacks. If terrorist organiza-
tions do indeed seek to differentiate themselves in competi- Hypothesis 2: States that experience more terrorist
tive environments, and if suicide terrorism is so uncommon, attacks will experience more severe types of terror-
it follows that there must be other methods that organiza- ist attacks than states that experience fewer terrorist
tions use to “stand out.” By only looking at the quantity of attacks.
attacks, or only analyzing one particular type of attack (su-
icide), previous literature has potentially missed significant In the following sections, we describe how we plan to test
evidence of the differentiation process. our hypotheses and then present results from our analysis
Additionally, in an environment where multiple terrorist of the effect of competition on the quality of violence.
organizations are committing terrorist attacks, the utility
of each additional attack will inevitably lead to diminish- RESEARCH DESIGN
ing returns in terms of differentiating the organization. In To improve our understanding of how organizations use
this situation it may become necessary to introduce a new violence to differentiate themselves, we propose a new set of
“product” in order to capture a larger market share. Even in ordinal measures capturing the severity level or “shock
a situation of extreme free riding, where only one terrorist value” of terrorist attacks. We rely on the Global Terrorism
organization carries out attacks and the other terrorist or- Database, which maintains a wide range of information on
ganizations commit no attacks and simply claim credit for all terrorist attacks from 1970 to 2010 (National Consor-
the attacks, there still remains an incentive to innovate be- tium 2011).6 We focus on two pieces of information about
cause even the free riders should receive diminishing re- each attack that allow us to compare the relative severity
turns from the additional attacks. This may also create an level of the violence being employed at any given time: the
environment where the credibility of groups is questioned, type of target that the terrorists attacked and their method
because it is highly unlikely that multiple groups are re- of attack.
sponsible for the same attacks. This provides even further Target Severity. GTD identifies 22 different categories of
incentive to differentiate and create a brand that is harder to targets, from the type of individual (e.g., tourist) to the type
copy. If a terrorist organization is able to effectively differ- of organization (e.g., media). Although the categories pro-
entiate, they will likely reap more utility from each attack. vided are nominal, to create our measure of severity, we
Additionally, organizations may suffer damage to their cred- grouped target types into three ordinal categories, based on
ibility if they launch multiple attacks that are perceived as the “shock value” or likelihood that the attack would be
weak. The preference for most organizations should therefore particularly noteworthy. The variable equals “1” if the at-
be a single, sensational attack that demonstrates their capa- tack is against infrastructure targets, including telecom-
bility and credibility effectively, rather than multiple weak munications, transportation, airports, maritime infrastruc-
attacks that diminish their credibility over time. ture, food or water sources, and utilities. In each of these
This leads us to two testable hypotheses. The first of cases, the attack is against a nonhuman target even though
these is a test of the underlying logic of differentiation, that some attacks may have implications for humans. We con-
more competition will cause terrorist organizations to en- sider this to be the lowest level of target severity. While at-
gage in more extreme forms of terrorism as the groups attacks that disrupt infrastructure and weaken the govern-
tempt to distinguish themselves from their competitors: ment’s ability to provide for its citizens may be strategically
useful, we argue that these attacks are not as poignant as
Hypothesis 1: States with more armed groups will
experience more severe types of terrorist attacks than 6. GTD has recently come under fire by Pape, Ruby, and Bauer (2014)
states with fewer groups. for allegedly inflating the global number of terrorist attacks. Their evi-
dence, however, relies largely on trends in suicide terrorism specifically.
We have argued that analysis of many issues related to terrorism, in-
The second hypothesis captures the argument that more
cluding outbidding behavior, has been limited by a focus on suicide ter-
of the same type of terrorist attack provides diminishing rorism alone. In a rebuttal, Distler et al. (2014) also suggests that criticism
returns and therefore incentivizes organizations to engage by Pape et al. may be unfounded.

attacks aimed at citizens. Moving up in severity, we code government and military personnel are less likely to gener-
an attack as a “2” if it is against any of the following target ate media coverage than attacks against civilians (Delli Car-
types: police, military, government personnel, other terror- pini and Williams 1987). Our broad categorization decisions
ists, or violent political parties. In each of these cases, the also accord with those in the Minorities at Risk Organiza-
target represents either a type of combatant or potential rival tional Behavior (MAROB) dataset, which we use below to
to the terrorist organization and captures attacks against test our hypotheses at the organization level. MAROB’s
individuals, rather than infrastructure. It can be argued that coding of “severity,” in turn, is based on precedents in in-
directly contesting the government is a greater show of ternational law (including the United Nations and Geneva
strength for the organization, but overall, groups that use conventions) about the use of violence. Our measures, while
terrorism are not in a position to effectively change the subjective, are therefore not arbitrary and provide a rea-
government directly. It might be more effective in the short- sonable ranking of the severity of terrorist attacks.
run for groups to target relatively softer targets (citizens), Attack Type. The second way in which we categorize the
which still serves to undermine the government’s credibil- severity level of attacks is by the method of attack itself. In
ity. At the highest level of severity, level “3,” we include this case, GTD categorizes the tactic used in the attack by
attacks against all civilians (nongovernmental, noncombat- classifying it as one of nine types of attacks. Once we again,
ants), including private citizens, tourists, educational person- we collapse the nominal categories into an ordinal scale in
nel, nongovernmental organizations (NGOs), private busi- which the lowest value (“1”) indicates attacks in which hu-
nesses, and the media. Such targets are likely to be more man beings are not in dire physical danger. These include
“shocking” and therefore provide greater differentiation com- attacks against infrastructure and unarmed assaults. These
pared to infrastructure and government/combatant targets. attacks are all characterized by low levels of violence and/or
Finally, all attacks listed as “Other” in the GTD data are an absence of human injury. Attacks are coded as more se-
coded as missing.7 vere (a value of “2”) if they involve hostage takings or
While the design of these measures rely on a set of sub- hijackings, attacks which pose an explicit threat to human
jective decisions, past research on media coverage and at- life. The third and final category, coded as a “3,” captures the
tack types/targets strongly support our classification scheme. most violent attacks. Assassinations, armed assaults, and
Specifically, we believe that our conceptual focus on attacks bombings are all included here.8 Once again, creating such
targeting civilians or involving potential injury to civilians a scale involves subjective decision making, but we believe
as the highest levels of severity is well founded (Delli Carpini that this is a reasonable way to order severity by attack
and Willliams 1987; Fozzard, and Ayanian 1982). In addi- method. Although Lee (2013) makes the case that hostage
tion, the principle of distinction (discrimination) in inter- situations draw considerable media coverage, past literature
national law requires that actors distinguish civilians from has found that the most covered terrorist events are those
military targets and only attack military targets (Blank and that result in actual death or injury to the targets (Delli
Noone 2013). This principle is a fundamental norm of war Carpini and Williams 1987; Weimann and Winn 1994). The
that has existed for thousands of years. The underlying recent beheadings of Western journalists by members of
moral arguments are seen in a variety of religious texts and ISIL demonstrates that this violent act is far more shocking
were further developed and codified during the Enlighten- and has received far more attention than the initial kid-
ment to form the basis for current international law and napping itself. Finally, either intentionally or unintention-
norms (Blank and Noone 2013). The fact that these norms ally, kidnappings often do not make the news at all and are
are so imbedded in the human understanding of warfare frequently resolved in private (Briggs 2001). Because of this,
reveals why their violation is so shocking. Additionally, ex- we argue that actual violent attacks involving assassinations
isting research supports our contention that attacks against
8. As with the target severity variable, we developed alternative coding

7. Two other miscellaneous categories were coded as missing, “reli- schemes for robustness purposes. We analyzed our attack severity models
gious” and “abortion,” because the categorization of these two was less using a two-category version, coded as “1” for attacks in which violence or
straightforward. However, in robustness checks, we included both target potential violence against humans is not central to the attack and “2” for
types in the highest severity level (“3”), and the conclusions that we draw attacks where violence or the threat of violence against humans is central
are unchanged. In general, while some classifications in our scale are to the attack. In other words, this allows us to combine kidnappings and
certainly open to debate, it seems that the overall order and grouping are the more violent categories into a single classification to avoid the sub-
accurately capturing a single latent variable, something that is supported jective decisions outlined here. The results using this consolidated variable
by the statistical significance of the cut points in each model below. are comparable to those reported in the article.

and armed assaults (like al-Shabbab’s attack on the mall in Table 1. Frequency Distribution of Maximum Target
Nairobi) are far more likely to garner media coverage and and Attack Severity Levels
attract recruits than traditional kidnappings and hostage
Target Attack
takings.
Severity Level Observations Observations Severity Level
Ideally, we would use information on each attack and
target type to see if organizations are turning to more ex-
No Attacks 1361 1359 No Attacks
treme forms of violence in the presence of increased com-
(44.73) (44.66)
petition. But one of the persistent problems with terrorism Infrastructure 61 29 Infrastructure/
data is the lack of attribution for most attacks (Hoffman (2.00) (0.95) Unarmed
1997). Indeed, in the GTD database, more than 41% of at- Assault
tacks are not attributed to any particular terrorist organi- Combatants and 359 554 Kidnapping
zation. government (11.80) (18.21)
Because of this data limitation, we instead aggregate in- Civilians 1262 1101 Armed
(41.47) (36.18) Assualt
formation about attack and target types to the state level. In
each case, we code the highest level of severity that a state
Note—Unit of analysis is country year. Percentages in parentheses.
experienced in a given year, which produces our two de-
pendent variables: Attack Severity and Target Severity. If a
ative to other organizations. Chenoweth (2010) uses a mea-
country experiences no attacks in a given year, the variables
sure of the competitiveness of political institutions in a
are coded as “0.” We include the “no attacks” categories be-
country as a proxy for competition itself. In all cases, how-
cause we believe that this represents both a quantitative and
ever, these measures provide a “snapshot” of the competi-
qualitative difference in terror. For example, just as terror-
tive environment at a single point in time. While we include
ist organizations have the ability to choose higher levels of
the active Number of Terrorist Groups as a key independent
severity by attacking civilians, they likewise have the ability
variable, we also assess the impact of dynamic changes in
to reduce their severity levels. We argue that moving from
the number of active groups by including a measure, Δ
attacks against infrastructure to refraining entirely from
Number of Terrorist Attacks, which measures the annual
terrorism (a movement from “1” to “0” on the Target Sever-
change in the number of groups (the number of active
ity scale) represents a meaningful difference in severity lev-
groups in time t minus the number of groups in time t-1).
els. Further, as discussed below, we include several models
This may be a better way of capturing the logic of out-
where the independent variable is a count of the number of
bidding theories, since rapid increases in the number of
rebel groups in a country. Many of these organizations do
organizations is likely to represent a more competitive en-
not engage in terrorism, so it is relevant to testing our hy-
vironment than one in which multiple organizations have
potheses to examine if some of these groups, when faced with
existed for some time. In both cases, data on the number of
increasing political competition, choose to use terrorism at
active organizations is taken from information compiled
all, let alone use more severe types of attacks. Nonetheless,
by Young and Dugan (2014). For robustness purposes, and
for robustness purposes, we also analyze models where the
to explore the implications of our theory for competition
“0s” in our dependent variables were dropped entirely, and
among violent groups in general, we follow Findley and
we discuss those results below. Table 1 displays descriptive
Young (2012) and also include the number of active rebel
statistics about the two dependent variables. For the tem-
groups in each country year. Data on the number of active
poral period of this study (1981–2004), a majority of coun-
rebel groups is taken from the Uppsala Conflict Data Pro-
try years involve at least one terrorist attack. Around 55%
gram (UCDP).9 This information provides two additional
of observations involve some type of attack, with a majority
independent variables, Number of Rebel Groups and Δ Num-
in each case featuring attacks against civilians or the most
ber of Rebel Groups, constructed in the same fashion as the
violent tactics (armed assaults, bombings, assassinations).
previous variables. To assess evidence for our second hy-
Turning now to our independent variables, previous em-
pothesis about the effect of the quantity of violence on the
pirical assessments of terrorist competition have recom-
severity level of attacks, we include a moving average of the
mended multiple measures, with no consensus on which is
best. Findley and Young (2012) use a count of active ter-
rorist organizations in a given year, while Nemeth (2013) 9. To be considered active in a given year, the rebel groups must be
includes a measure of the organization’s market share rel- associated with at least 25 battle deaths during that time.

number of terrorist attacks that the state has experienced Table 2. Competition and Target Severity Level
since 1970.
We include a number of control variables that are likely Dependent Variable: Maximum Target Severity by Country Year
to influence not only the level of political competition within
a country, but potentially the severity of that competition. Model Model Model Model
1 2 3 4
We include a measure of democraticness of a state’s insti-
tutions, drawn from the Polity IV project (Marshall et al.
Number of terrorist 0.73* - - -
2002). The measure ranges from 210 to 10 with higher val-
groups (0.16) - - -
ues indicating states that are more democratic. We include
DNumber of terrorist - 0.16* - -
two measures which capture the demographic makeup of a groups - (0.07) - -
country, the level of Ethnic Fractionalization and Religious Number of rebel groups - - 0.78* -
Fractionalization. Compiled by Alesina et al. (2003), each - - (0.24) -
measure is the probability that two people chosen at ran- DNumber of rebel - - - 20.07
dom in a country will be from a different ethnic or reli- groups - - - (0.09)
gious group. Higher values therefore represent countries Annual terrorist attacks 0.01* 0.04* 0.04* 0.04*
(moving avg.) (0.01) (0.02) (0.02) (0.02)
with many distinct ethnic or religious groups. State treat-
Democracy 0.07* 0.08* 0.08* 0.08*
ment of their citizens has long been thought to affect vio-
(0.01) (0.01) (0.01) (0.01)
lence, so we also include a measure of Physical Integrity Ethnic fractionalization 20.10 20.08 20.13 –0.08
Rights from Cingranelli and Richards (2010). The variable is (0.36) (0.39) (0.38) (0.39)
an ordinal scale, ranging from “0” to “8,” with higher values Religious 0.31 –0.28 –0.22 –0.29
assigned to states that demonstrate greater respect for the fractionalization (0.32) (0.37) (0.37) (0.37)
physical integrity rights of their citizens. We also include a Physical integrity rights –0.22* –0.30* –0.28* –0.30*
measure which equals “1” if the state is experiencing a civil (0.04) (0.04) (0.05) (0.04)
Civil war 0.48* 0.91* 0.18 0.94*
war in a given year, and “0” otherwise, using the UCDP
(0.19) (0.21) (0.23) (0.21)
dataset. We expect both the quantity and quality of violence,
Ln(GDP) 0.13* 0.20* 0.21* 0.20*
as well as the level of political competition, to be particularly (0.06) (0.06) (0.06) (0.06)
acute during such conflicts. To account for potential eco- Regime durability 20.01 20.01 20.01 20.01
nomic determinants of competition and violence, we include (0.01) (0.01) (0.01) (0.01)
the natural log of the state’s GDP in a given year (United Cut 1 2.55 3.27 3.61 3.21
Nations Statistics Division 2009). Finally, we include a mea- (1.40) (1.51) (1.48) (1.49)
sure of Regime Durability to account for the possibility that Cut 2 2.67 3.38 3.72 3.32
(1.40) (1.51) (1.49) (1.50)
some organizational behavior may simply be a response to
Cut 3 3.43 4.04 4.39 3.98
new and/or weak regimes. This variable is a count of the
(1.40) (1.51) (1.48) (1.49)
number of years since the last major institutional change in Observations 3043 3041 3043 3043
the state’s political structure (measured as 3-point change
on the Polity scale) (Marshall et al. 2002). Note—Ordered logistic regression. Robust standard errors clustered on
We analyze all models using ordinal logistic regression, the country in parentheses.
and we calculate robust standard errors, clustered on the * p ! 0.10 (two-tailed).
country. In the next section, we present the results from our

primary set of statistical analyses.
variables, but the measure of the key independent vari-
RESULTS AND DISCUSSION able (competition) differs in each case. In Model 1, the ab-
To examine the evidence for our hypotheses, we first look solute number of terrorist groups in a given year is associ-
at the relationship between political competition and the ated with higher target severity levels. In other words, states
types of targets chosen by terrorist organizations. We then with greater numbers of active terrorist organizations are
analyze the link between competition and the chosen meth- more likely to experience more severe attacks (by target
ods of attack. Table 2 lists the results of a series of models type) in a given year. Similarly, in Model 3, states with
in which the dependent variable is our ordinal scale of target greater numbers of active rebel groups experience more se-
severity. Models 1 through 4 include the same set of control vere types of attacks compared to states with fewer groups.

Additionally, the dynamic measure of competition has a Table 3 lists the substantive effects of each of the key
positive and significant relationship with target severity. In independent variables from Table 2. The substantive effects
Model 2, a greater annual change in the number of terrorist of the significant control variables are also included, as cal-
groups is associated with higher severity levels. Only the culated using the results from Model 1. The odds ratio in-
measure of changes in the number of rebel groups (Model 4) dicates how an increase in one unit of the independent
has an insignificant relationship with the dependent vari- variable affects the odds of a state seeing a higher category
able. Overall, however, the models provide strong evidence of attack severity in a given year. For instance, the largest
in favor of our first hypothesis: states with more competi- substantive effect is generated by an increase in the number
tive political environments are likely to experience more of terrorist and rebel groups. An increase of one group of
severe types of attacks than those with less competitive either type increases the odds that the state will experience a
environments. higher category of severity by 108 and 119%, respectively.
Turning to our second hypothesis, which has a related The absolute number of active organizations seems to have
expectation about the effect of the quantity of violence on an even more dramatic effect than the level of democracy
the type of violence that a state is likely to experience, we or the occurrence of civil war. Additionally, each terrorist
again find strong support. The average number of attacks group that forms between the previous and current years
that a state experiences since 1970 influences the type of (ΔNumber of Terrorist Groups) increases the odds of higher
violence that is used and does so in each of our model spec- severity levels by 18%. And an increase of just one attack in a
ifications. States that experience more terrorism, therefore, state’s annual average increases the odds of higher severity
are also likely to experience more severe types of attacks. levels by 1%. The information gleaned from Tables 2 and 3
This lends support to our contention that competitive en- therefore demonstrates a strong relationship between po-
vironments where terrorist attacks are commonplace pro- litical competition and the quality of violence.
vide greater incentives for organizations to differentiate their The relationship is also evident if we examine the quality
activities. of violence in terms of the methods of attack. Table 4 lists
The control variables influence severity levels consis- the results of a series of models that are identical to those in
tently across the four models, with the level of democracy, Table 2, but using our second ordinal dependent variable.
GDP, and the occurrence of civil war having strong positive This variable captures the severity level of the methods used
effects on the severity of violence. States that offer greater in terrorist attacks, rather than the target types. Once again,
protection of their citizens’ physical integrity rights, on the in the first three models, we see that the absolute number
other hand, are less likely to witness more severe forms of of active terrorist and rebel groups, as well as the annual
violence. Neither of the fractionalization variables are sig- change in the number of terrorist groups, has a positive
nificant in any of the models presented here. Finally, each effect on the severity level of attack methods. Also similar
of the three cut points in the models are significantly dif- to our first set of models, the annual change in the number
ferent from each other, indicating that the categories we of rebel groups is not significantly associated with changes
have designed are effectively capturing distinct levels of tar- in the severity level. The results from Models 5 through 7
get severity. suggest, however, that states with higher levels of political
competition are likely to experience more severe types of
attacks, such as armed assaults, rather than attacks on in-
Table 3. Odds Ratios from Table 2 frastructure, for instance. Additionally, the Annual Terror-
ist Attacks variable is consistently significant and positive
Variable Odds Ratio across all four models. Hypotheses 1 and 2 are therefore
supported using evidence of the chosen methods of terrorist
Number of terrorist groups 2.08 attacks.10 Substantive effects for all significant variables are
DNumber of terrorist groups 1.18 listed in Table 5, and again we see that the absolute num-
Number of rebel groups 2.19 bers of terrorist groups and rebel groups have the strongest
Average annual terrorist attacks (Model 1) 1.01
Democracy (Model 1) 1.08
Physical integrity rights (Model 1) 0.74
Civil war (Model 1) 2.49 10. In Table 4, the cut points are all statistically significant different
Ln(GDP) (Model 1) 1.22 from each, once again indicating that the each of the categories in the
dependent variable is tapping into a distinct level of violence.

effects on the odds of higher severity levels. An increase of

Table 5. Odds Ratios from Table 4
one terrorist group, for instance, increases the odds that a
country will experience a more severe method of attack by Variable Odds Ratio
73%. All control variables exhibit similar relationships with
the dependent variable in this set of analyses. Among the Number of terrorist groups 1.73
controls, the occurrence of a civil war continues to have a DNumber of terrorist groups 1.15
strong substantive effect, increasing the odds of more severe Number of rebel groups 1.49
attacks by 33%. Average annual terrorist attacks (Model 5) 1.01
In a set of analyses not reported here, we also consider Democracy (Model 5) 1.06
the possibility that the severity of violence is actually be- Physical integrity rights (Model 5) 0.83
Civil war (Model 5) 1.33
Ln(GDP) 1.18
Table 4. Competition and Attack Severity Level
Dependent Variable: Maximum Attack Severity by Country Year ing driven by the “acceptability of violence” (Atran 2003;
Bloom 2005; Nemeth 2013).11 That is, violent organizations
Model Model Model Model are likely to behave differently within societies where vio-
5 6 7 8 lence is more readily accepted by the general population as
a justified form of political protest. We conducted robust-
Number of terrorist 0.55* - - - ness checks, including a measure of the Acceptability of
groups (0.11) - - - Violence in a country created by Nemeth (2013). The re-
DNumber of terrorist - 0.14* - - sulting analysis dramatically reduced our sample size (by
groups - (0.06) - - roughly 80%), but nonetheless, the conclusions reached
Number of rebel groups - - 0.40* -
are comparable to those we have already outlined. The ab-
- - (0.17) -
DNumber of rebel - - - 20.05
solute number of terrorist groups, as well as the year-to-
groups - - - (0.07) year change in the number of terrorist groups, continues to
Annual terrorist attacks 0.01* 0.03* 0.03* 0.03* positively affect the severity of both the target and attack
(moving avg.) (0.01) (0.02) (0.01) (0.01) types.
Democracy 0.06* 0.07* 0.07* 0.07* We also conducted a series of analyses in which we re-
(0.01) (0.01) (0.01) (0.01) moved the “0” or “no attack” categories from the dependent
Ethnic fractionalization 20.32 20.31 20.31 20.31 variables, as it can be argued that the difference between
(0.35) (0.38) (0.38) (0.38)
the “0” categories and the other categories does not rep-
Religious 0.30 20.24 20.21 20.25
fractionalization (0.32) (0.36) (0.36) (0.36)
resent an increase in severity, per se. As mentioned previ-
Physical integrity rights 20.19* 20.26* 20.25* 20.27* ously, however, particularly for rebel groups, we believe
(0.04) (0.04) (0.04) (0.04) that groups who do not use terrorism and subsequently
Civil war 0.28* 0.68* 0.26 0.70* change their tactics and start using some form of terrorist
(0.17) (0.18) (0.22) (0.18) attacks represents an interesting qualitative difference in
Ln(GDP) 0.16* 0.23* 0.23* 0.23* their tactical choices. In robustness checks when leaving out
(0.05) (0.06) (0.05) (0.06) the “0s” entirely, the results are generally comparable to
Regime durability 20.01 20.01 20.01 20.01
those reported in this article. The coefficients for the key
(0.01) (0.01) (0.01) (0.01)
Cut 1 3.28 4.00 4.20 3.93
independent variables in the first three models in Table 2
(1.30) (1.39) (1.38) (1.38) remain statistically significant in the expected direction. The
Cut 2 3.33 4.05 4.25 3.98 Annual Terrorist Attacks variable remains significant across
(1.30) (1.39) (1.38) (1.38) three of the four models, as well. In the models from Ta-
Cut 3 4.45 5.05 5.26 4.99 ble 4, the Annual Terrorist Attacks variable is significant
(1.30) (1.40) (1.38) (1.39) in all models, and the Number of Terrorist Groups also
Observations 3043 3041 3043 3043
Note—Ordered logistic regression. Robust standard errors clustered on 11. Nemeth creates an index based on the state’s level of political
the country in parentheses. terror, use of capital punishment, and involvement in interstate and in-
* p ! 0.10 (two-tailed). trastate conflicts.

significantly and positively affects the severity level. Because the ones we used in the previous analysis. Our variable
of the loss of observations, the substantive effects in most ranges from “0” to “3,” with a “0” indicating that the orga-
cases are reduced, but these results at least give us confidence nization did not commit violence against any of the fol-
that our primary conclusions are not an artifact of including lowing in a given year: infrastructure, security personnel,
the “0” categories.12 or noncombatants. If a state scores a “1” on the scale, the
Whether we measure the severity of terrorist attacks by highest level of violence it engaged in was attacks against
the type of target or the method of the attack, the evidence infrastructure. A score of “2” indicates that the organiza-
clearly points to a relationship with political competition tion engaged in violence against security personnel, while
among combatant groups. In all cases, the number of ter- a score of “3” indicates violence against noncombatants
rorist groups and rebel groups operating within a country (civilians or noncombatant government personnel).
has a significant, positive, and substantial effect on the type To control for other potential determinants of organi-
of terrorism the country is likely to experience. States with zational violence, we include a series of binary and cate-
greater levels of competition among such nonstate groups
are far more likely to experience severe or “shocking” types Table 6. Domestic Violence by Middle Eastern Political
of attacks compared to states with lower levels of domestic Organizations, 1980–2004
competition. The variable capturing the annual change in
the number of rebel groups is the only measurement of the Dependent Variable: Maximum Target Severity by Organization
independent variable which does not conform to our theo-
retical expectations. This may be related to the smaller range Model Model Model Model
of values for this variable relative to the terrorist group 9 10 11 12
version but nonetheless is an interesting result. Generally
speaking, though, we have found evidence supporting our Number of terrorist 0.08* - - -
hypotheses across a range of various measurements of both groups (0.02) - - -
the independent and dependent variables. Further, the ef- DNumber of terrorist - 0.02 - -
groups - (0.02) - -
fect of competition on the severity of violence is indepen-
Number of rebel groups - - 0.23* -
dent of influences such as civil conflict and human rights
- - (0.09) -
abuses. DNumber of rebel - - - 0.05
groups - - - (0.08)
Organization Level Analysis Annual terrorist attacks 0.01 0.03* 0.02 0.03*
As mentioned previously, one reason why we have con- (Moving Avg.) (0.01) (0.01) (0.01) (0.01)
ducted a state-level analysis thus far is because the primary Foreign state support 0.97* 1.12* 1.14* 1.12*
source of the data, GTD, includes a large amount of terrorist (0.34) (0.34) (0.34) (0.34)
Diaspora support 1.92* 1.80* 1.88* 1.79*
attacks that are not attributed to a particular organization.
(0.71) (0.80) (0.82) (0.80)
Another challenge is that even among the attacks for which
Separatist ideology 0.16 20.06 20.13 20.05
there is attribution, the “group” may only be categorized by (0.38) (0.39) (0.38) (0.39)
nationality (e.g., “Palestinians”) or by an even vaguer clas- Religious ideology 0.64* 0.60 0.57 0.60
sification (e.g., “students”). It is difficult, therefore, to iden- (0.36) (0.38) (0.40) (0.39)
tify group-level trends using the existing GTD data. Participates in elections 20.29 20.39* 20.37* 20.39*
As a limited test of our hypotheses at the organization (0.21) (0.20) (0.20) (0.20)
level, we turn to the Minorities at Risk-Organizational Be- State targets organiza- 2.39* 2.28* 2.12* 2.28*
tion with violence (0.35) (0.35) (0.32) (0.34)
havior Dataset-Middle East (MAROB-ME), which contains
Cut 1 2.67 2.33 2.41 2.32
annual data on all ethnopolitical organizations in the Mid-
(0.34) (0.35) (0.35) (0.35)
dle East and North Africa from 1980 to 2004 (Asal, Pate, Cut 2 2.69 2.35 2.43 2.35
and Wilkenfeld 2008). We use the dataset’s DOMORG (0.34) (0.35) (0.35) (0.35)
VIOLENCE variable to create an ordinal variable similar to Cut 3 4.81 4.42 4.52 4.41
(0.47) (0.48) (0.49) (0.48)
Observations 1484 1484 1484 1484
12. In the models from Table 2, for instance, the addition of one Note—Ordered logistic regression. Robust standard errors clustered on
terrorist group in a state increases the odds of a more severe category of the organization in parentheses.
targets by 52%. * p ! 0.10 (two-tailed).

gorical variables that capture characteristics of the organi- bidding are often missing important evidence of the pro-
zation and its political environment. We include whether cess.
the organization received financial, humanitarian, political, This research also suggests, more broadly, that focusing
or military support from a foreign state or members of its solely on raw counts of terrorist attacks, although the most
ethnic diaspora. Both are expected to increase the likeli- popular dependent variable used in terrorism research, may
hood of the organization engaging in violence overall. We be inappropriate for answering many research questions.
also include whether the organization espouses a primarily While detailed information on terrorist organizations and
separatist or religious ideology. A religious ideology is ex- attacks is often lacking, as there are incentives for both
pected to increase the likelihood of violence overall, while governments and terrorist organizations to limit the distri-
a separatist ideology has been found to decrease the like- bution of such information, scholars should invest in more
lihood of transnational violence specifically (Asal, Conrad, nuanced ways of capturing the impact of terrorist attacks.
and White 2014). Participates in Elections is a categorical Weighting attacks based on information such as attack type,
variable that equals “1” if the organization is formally en- target type, fatality levels, and the amount of subsequent
gaged in the electoral process and “2” if it is currently an media coverage may provide better leverage on standard
election year (and “0” if it is not engaged in the electoral questions of terrorism. The United States has experienced
process). We expect incorporation into the political process, relatively few terrorist attacks compared to countries like
particularly during an election year, to reduce the organi- India and Colombia, but several of these attacks had greater
zation’s likelihood of using violence. Finally, State Repres- disproportionate impacts. In order to effectively combat
sion is a binary variable which equals “1” if the organization terrorism, we must distinguish the effects of more severe
is subject to targeted “lethal” violence by the state. attacks from attacks that occur more frequently but garner
Table 6 displays the results of the organization-level less attention.
analysis. While the generalizability of the results is limited Past research (e.g., Shapiro and Siegel 2007) finds that en-
due to the data covering only a single region, they never- couraging schisms within terrorist organizations decreases
theless provide additional evidence for our hypotheses. In their effectiveness. However, if these schisms lead to the cre-
this case, the absolute counts of rebel and terrorist groups ation of direct competition for the original organization, the
significantly increase the probability that an organization results here suggest that violence (at least temporarily) may
will engage in more severe attacks. The annual changes in become more severe. Based on our findings, the addition of
the counts, however, no longer have a significant influence. just one terrorist organization more than doubles the odds
The average number of terrorist attacks has a significant that the state will experience a higher category of severity.
and positive effect on the severity of violence in two of the While encouraging divisions and fractionalization among or-
models (and the effects are significant in one-tailed tests in ganizations might be an effective means of undermining an
the remaining models). The controls suggest that outside original terrorist organization, careful attention must be paid
support (from foreign states or diasporas) and state violence to the possibility of unintended changes in the nature of vio-
also significantly increase the severity of violence. Again, the lence that is used.
results in Table 6 should be interpreted with caution, given Finally, the results here emphasize that the effectiveness
the limited spatial domain, but even these results suggest of terrorist campaigns often relies on how “effectiveness” is
that domestic competition influences the quality of violence defined. As Abrahms (2012) has pointed out, if effective-
chosen by political organizations. ness is measured as bargaining outcomes, the success rate
for most terrorist campaigns is mixed, at best. This study
CONCLUSION has provided evidence that terrorist organizations may at
This study has identified evidence of outbidding among vi- least expect short-term returns from extreme violence, even
olent groups that has largely been ignored by scholars. Un- if the strategy is risky in the long run. To improve our un-
like previous studies, we categorize terrorist violence by its derstanding of terrorist decision making, future research
“shock value” or severity. The new measures indicate a sig- might focus on just how successful terrorists are in achiev-
nificant relationship between domestic competition and the ing these more limited goals.
tactical choices of terrorists. Our analysis shows that the
severity of terrorist attacks based on both the type of target REFERENCES
Abrahms, M. 2006. “Why Terrorism Does Not Work.” International Se-
and the type of method employed increases dramatically
curity 31 (2): 42–78.
as organizations face greater levels of competition. In light Abrahms, M. 2008. “What Terrorists Really Want: Terrorist Motives and
of this research, we argue that empirical treatments of out- Counterterrorism Strategy.” International Security 32 (4): 78–105.

Abrahms, M. 2012. “The Political Effectiveness of Terrorism Revisited.” Dickson, P. R. 1997. Marketing Management. 2nd ed. Orlando, FL: Dryden
Comparative Political Studies 45 (3): 366–93. Press.
Alesina, A., A. Devleeschauwer, W. Easterly, S. Kurlat, and R. Wacziarg. Distler, M., O. Hodwitz, M. Jensen, G. LaFree, E. Miller, and A. Safer-
2003. “Fractionalization.” Journal of Economic Growth 8 (2): 155–94. Lichtenstein. 2014. “The Challenges of Collecting Terrorism Data.”
Asal, V., J. Conrad, and P. White. 2014. “Going Abroad: Transnational August 6. http://www.washingtonpost.com/blogs/monkey-cage/wp/2014
Solicitation and Contention by Ethnopolitical Organizations.” Inter- /08/06/the-challenges-of-collecting-terrorism-data/ (accessed Septem-
national Organization 68 (4): 945–78. ber 9, 2014).
Asal, V., A. Pate, and J. Wilkenfeld. 2008. “Minorities at Risk Organiza- Downes, A. B. 2006. “Desperate Times, Desperate Measures: The Causes
tional Behavior Data and Codebook.” Version 9. http://www.cidcm of Civilian Victimization in War. International Security 30 (4): 152–
.umd.edu/mar/data.asp (accessed September 9, 2014). 95.
Atran, S. 2003. “Genesis of Suicide Terrorism.” Science 299 (5612): 1534–39. Enders, W., and T. Sandler. 2006. The Political Economy of Terrorism.
Baker, M. J. 1996. Marketing: An Introductory Text. 6th ed. Bath, UK: Bath Cambridge: Cambridge University Press.
Press. Findley, M., and J. Young. 2012. “More Combatant Groups, More Terror?
Blank, R. L., and P.G. Noone. 2013. International Law and Armed Conflict: Empirical Tests of Outbidding Logic.” Terrorism and Political Violence
Fundamental Principles and Contemporary Challenges in the Law of 24 (5): 706–21.
War. New York: Wolters Kluwer Law & Business. Fromkin, D. 1975. “The Strategy of Terrorism.” Foreign Affairs 53 (4):
Bleich, A., M. Gelkopf, and Z. Solomon. 2003. “Exposure to Terrorism, 683–98.
Stress-Related Mental Health Symptoms, and Coping Behaviors among Guiltinan, J. P., and G. W. Paul. 1991. Marketing Management. 4th ed. New
a Nationally Representative Sample in Israel.” Journal of the American York: McGraw-Hill.
Medical Association 290 (5): 612–20. Hoffman, A. M., C. Shelton, and E. Cleven. 2013. “Press Freedom, Pub-
Blomberg, S. B., K. Gaibulloev, and T. Sandler. 2011. “Terrorist Group licity, and the Cross-National Incidence of Transnational Terrorism.”
Survival: Ideology, Tactics, and Base of Operations.” Public Choice 149 Political Research Quarterly 66 (4): 896–909.
(3–4): 441–63. Hoffman, B. 1997. “Why Terrorists "Don’t Claim Credit.” Terrorism and
Bloom, M. 2005. Dying to Kill: The Allure of Suicide Bombing. New York: Political Violence 9 (1): 1–6.
Columbia University Press. Hoffman, B. 2006. Inside Terrorism. New York: Columbia University Press.
Briggs, R. 2001. The Business of Kidnapping. London: Central Books. Hoffman, B., and G. McCormick. 2004. “Terrorism, Signaling, and Suicide
Canes-Wrone, B., and K. Shotts. 2007. “When Do Elections Encourage Ideo- Attack.” Studies in Conflict and Terrorism 27 (4): 243–81.
logical Rigidity?” American Political Science Review 101 (2): 273–88. Holman, E. A., D. R. Garfin, and R. C. Silver. 2014. “Media’s Role in
Caplan, B. 2006. “Terrorism: The Relevance of the Rational Choice Broadcasting Acute Stress Following the Boston Marathon Bomb-
Model.” Public Choice 128 (1–2): 91–107. ings.” Proceedings of the National Academy of Sciences 111 (1): 93–
Chai, S. K. 1993. “An Organizational Economics Theory of Antigovern- 98.
ment Violence.” Comparative Politics 26 (1): 99–110. Kalyvas, S. N. 2006. The Logic of Violence in Civil War. Cambridge:
Chamberlin, E. H. 1933. The Theory of Monopolistic Competition. 1st ed. Cambridge University Press.
Cambridge, MA: Harvard University Press. Keinan, G., A. Sadeh, and S. Rosen. 2003. “Attitudes and Reactions to
Chenoweth, E. 2010. “Democratic Competition and Terrorist Activity.” Media Coverage of Terrorist Acts.” Journal of Community Psychology
Journal of Politics 72 (1): 16–30. 31 (2): 149–65.
Chenoweth, E., and M. J. Stephan. 2011. Why Civil Resistance Works: The Kotler, P., S. H. Ang, S. M. Leong, and C. T. Tan. 1996. Marketing
Strategic Logic of Nonviolent Conflict. New York: Columbia University Management—An Asian Perspective. Singapore: Prentice Hall.
Press. Kydd, A. H., and B. F. Walter. 2006. “The Strategies of Terrorism.” In-
Chermak, S. M., and J. Gruenewald. 2006. “The Media’s Coverage of ternational Security 31 (1): 49–80.
Domestic Terrorism.” Justice Quarterly 23 (4): 428–61. Lake, D. A. 2002. “Rational Extremism: Understanding Terrorism in the
Cingranelli, D. L., and D. L. Richards. 2010. “The Cingranelli and Richards Twenty-First Century.” Dialogue IO 1 (1): 15–29.
(CIRI) Human Rights Data Project.” Human Rights Quarterly 32 (2): Lee, C. Y. 2013. “Democracy, Civil Liberties, and Hostage-Taking Ter-
401–24. rorism.” Journal of Peace Research 50 (2): 235–48.
Crenshaw, M. 1987. “Theories of Terrorism: Instrumental and Organi- Marshall, M. G., T. R. Gurr, C. Davenport, and K. Jaggers. 2002. “Polity
zational Approaches.” Journal of Strategic Studies 10 (4): 13–31. IV, 1800–1999. Comments on Munck and Verkuilen.” Comparative
Crenshaw, M. 1991. “How Terrorism Declines.” Terrorism and Political Politics Studies 35 (1): 40–45.
Violence 3 (1): 69–87. Mercer, D. 1992. Marketing. Oxford, UK: Blackwell Publishers.
Crenshaw, M. 2002. “The Logic of Terrorism: Terrorist Behavior as a Nacos, B. L. 2003. “Terrorism as Breaking News: Attack on America.”
Product of Strategic Choice.” In Origins of Terrorism: Psychologies, Political Science Quarterly 118 (1): 23–52.
Ideologies, Theologies, States of Mind, ed. Walter Reich. Washington, National Consortium for the Study of Terrorism and Responses to Ter-
DC: Woodrow Wilson Center Press, 54–66. rorism (START). 2011. Global Terrorism Database [Data file]. Re-
Crenshaw, Martha. 1985. “An Organizational Approach to the Analysis of trieved from http://www.start.umd.edu/gtd (accessed September 9,
Political Terrorism.” Orbis 29 (3): 465–89. 2014).
Cronin, A. K. 2006. “How al-Qaida Ends: The Decline and Demise of Nemeth, S. 2013. “The Effect of Competition on Terrorist Group Oper-
Terrorist Groups. International Security 31 (1): 7–48. ations.” Journal of Conflict Resolution 58 (2) 336–62.
Cronin, A. K. 2009. How Terrorism Ends: Understanding the Decline and Oots, K. L. 1989. “Organizational Perspectives on the Formation and
Demise of Terrorist Campaigns. Princeton, NJ: Princeton University Disintegration of Terrorist Groups.” Studies in Conflict & Terrorism 12
Press. (3): 139–52.
Delli Carpini, M. X., and B. A. Williams. 1987. “Television and Terrorism: Paletz, D. L., P. A. Fozzard, and J. Z. Ayanian. 1982. “The IRA, the Red
Patterns of Presentation and Occurrence, 1969 to 1980.” Western Brigades, and the FALN in the New York Times.” Journal of Com-
Political Quarterly 40 (1): 45–64. munication 32 (2): 162–71.

Pape, R. A. 2003. “The Strategic Logic of Suicide Terrorism. American Thomas, J. 2014. “Rewarding Bad Behavior: How Governments Respond
Political Science Review 97 (3): 343–61. to Terrorism in Civil War.” American Journal of Political Science 58
Pape, R A., K. Ruby, and V. Bauer. 2014. “Government Data Exaggerates (4): 804–18.
the Increase in Terrorist Attacks.” July 21. http://www.washingtonpost United Nations Statistics Division. 2009. “National Accounts Main Ag-
.com/blogs/monkey-cage/wp/2014/07/21/government-data-exaggerate gregates Database GDP.” http://unstats.un.org/unsd/snaama/Introduction
-the-increase-in-terrorist-attacks/ (accessed September 9, 2014). .asp (accessed September 9, 2014).
Piazza, J. A. 2008. “A Supply-Side View of Suicide Terrorism: A Cross- Weimann, G., and C. Winn. 1994. The Theater of Terror: Mass Media and
National Study.” Journal of Politics 70 (1): 28–39. International Terrorism. New York: Longman.
Pinson, C., and D. J. Brosdahl. 2014. “The Church of Mac: Exploratory Wilson, J. Q. 1995. Political Organizations. Princeton, NJ: Princeton Uni-
Examination on the Loyalty of Apple Customers.” Journal of Man- versity Press.
agement & Marketing Research 14 (1): 1–15. Wood, R. M. 2010. “Rebel Capability and Strategic Violence against Ci-
Powers, T. L. 1991. Modern Business Marketing. Saint Paul, MN: West vilians.” Journal of Peace Research 47 (5): 601–14.
Publishing. Wood, R. M. 2014. “From Loss to Looting? Battlefield Costs and Rebel
Scott, J. L. 2001. “Media Congestion Limits Media Terrorism.” Defence Incentives for Violence.” International Organization 68 (3): 633–61.
and Peace Economics 12 (3): 215–27. Young, J. K. 2013. “The Strength of al-Shabaab.” Political Violence at a
Shapiro, J. N., and D. A. Siegel. 2007. “Underfunding in Terrorist Orga- Glance. September 26. http://politicalviolenceataglance.org/2013/09/26
nizations.” International Studies Quarterly 51 (2): 405–29. /the-strength-of-al-shabab/ (accessed September 9, 2014).
Sharp, B., and J. Dawes. 2001. “What is Differentiation and How Does it Young, J. K., and L. Dugan. 2014. “Survival of the Fittest: Why Terrorist
Work?” Journal of Marketing Management 17 (7–8): 739–59. Groups Endure.” Perspectives on Terrorism 8 (2): 2–23.


16-conrad

Uploaded by

Copyright:

Available Formats

16-conrad

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

16-conrad

Uploaded by

Copyright:

Available Formats

Competition, Differentiation, and the Severity of Terrorist Attacks

Author(s): Justin Conrad and Kevin Greene

This content downloaded from 132.203.227.61 on Thu, 1 Oct 2015 03:30:02 AM

W hen explaining the behavior of terrorist orga-

This content downloaded from 132.203.227.61 on Thu, 1 Oct 2015 03:30:02 AM

This content downloaded from 132.203.227.61 on Thu, 1 Oct 2015 03:30:02 AM

This content downloaded from 132.203.227.61 on Thu, 1 Oct 2015 03:30:02 AM

seek to control a larger market share and to overcome their

This content downloaded from 132.203.227.61 on Thu, 1 Oct 2015 03:30:02 AM

This content downloaded from 132.203.227.61 on Thu, 1 Oct 2015 03:30:02 AM

This content downloaded from 132.203.227.61 on Thu, 1 Oct 2015 03:30:02 AM

This content downloaded from 132.203.227.61 on Thu, 1 Oct 2015 03:30:02 AM

8. As with the target severity variable, we developed alternative coding

This content downloaded from 132.203.227.61 on Thu, 1 Oct 2015 03:30:02 AM

This content downloaded from 132.203.227.61 on Thu, 1 Oct 2015 03:30:02 AM

country. In the next section, we present the results from our

This content downloaded from 132.203.227.61 on Thu, 1 Oct 2015 03:30:02 AM

This content downloaded from 132.203.227.61 on Thu, 1 Oct 2015 03:30:02 AM

effects on the odds of higher severity levels. An increase of

This content downloaded from 132.203.227.61 on Thu, 1 Oct 2015 03:30:02 AM

This content downloaded from 132.203.227.61 on Thu, 1 Oct 2015 03:30:02 AM

This content downloaded from 132.203.227.61 on Thu, 1 Oct 2015 03:30:02 AM

This content downloaded from 132.203.227.61 on Thu, 1 Oct 2015 03:30:02 AM

This content downloaded from 132.203.227.61 on Thu, 1 Oct 2015 03:30:02 AM

You might also like