You are on page 1of 28

450573

an and ChenowethAmerican Sociological Review


2012
ASRXXX10.1177/0003122412450573Dug

American Sociological Review

Moving Beyond 77(4) 597624


American Sociological
Association 2012
Deterrence: The Effectiveness DOI: 10.1177/0003122412450573
http://asr.sagepub.com

of Raising the Expected Utility


of Abstaining from Terrorism in
Israel

Laura Dugana and Erica Chenowethb

Abstract
Rational choice approaches to reducing terrorist violence would suggest raising the costs of
terrorism through punishment, thereby reducing the overall expected utility of terrorism. In
this article, we argue that states should also consider raising the expected utility of abstaining
from terrorism through rewards. We test effects of repressive (or punishing) and conciliatory
(or rewarding) actions on terrorist behavior using the newly developed GATE-Israel dataset,
which identifies events by Israeli state actors toward Palestinian targets on a full range of
counterterrorism tactics and policies from 1987 to 2004. Results show that repressive actions
are either unrelated to terror or related to subsequent increases in terror, and conciliatory
actions are generally related to decreases in terror, depending on the tactical period. Findings
also reveal the importance of understanding the role of terrorists constituencies for reducing
violence.

Keywords
conciliation, counterterrorism, deterrence, expected utility, Israel, Palestine, rational choice
theory, terrorism

Rational choice approaches have long sug- well as a broad range of offending behaviors
gested that reducing unwanted behavior (Matsueda, Kreager, and Huizinga 2006;
requires raising the costs (or perceived costs) Nagin 1998; Paternoster 1987), including ter-
of a behavior through the threat of punish- rorist violence (LaFree, Dugan, and Korte
menta process commonly referred to as 2009). The clear appeal of deterrence theory
deterrence. In the eighteenth century, Becca- is its parsimony, as well as the fact that pun-
ria ([1764] 1983) argued that the state should ishment can be imposed with relative ease.
punish law-breakers just enough so that the
burdens of punishment outweigh any pleasure a
University of Maryland
derived from perpetrating the crimean idea b
University of Denver
that directly informs U.S. criminal law, mili-
Corresponding Author:
tary strategy, and a host of other social policy
Laura Dugan, University of Maryland,
domains. Scholars have since applied the Department of Criminology and Criminal Justice,
notion of deterrence to a range of areas, 2220 LeFrak Hall, College Park, MD 20742
including nuclear doctrine (Brodie 1959), as E-mail: ldugan@umd.edu
598 American Sociological Review 77(4)

All the theory requires is that punishment be yield lower levels of subsequent Palestinian
calibrated in such a way that specific behav- violence, even when the primary targets are
iors are no longer appealing. religiously inspired terrorist groups, empha-
Despite the popularity of deterrence, two sizing an importantyet understudied
important problems remain. First, empirically, dimension of rational choice approaches. In
unwanted behaviors continueand indeed summary, Israeli policymakers should con-
often increasedespite the threat of punish- sider conciliatory tactics as potentially viable
ment. Second, theoretically, implications of in reducing terrorism and disempowering
rational choice theory are broader than simply extremists within the Palestinian Territories.
deterring unwanted behavior through punish- Scholars should study more carefully the con-
ment. As Becker (1968) demonstrates, actors ditions under which conciliatory policies may
choose whether to break the law by comparing be effective in reducing terrorism.
the expected utility of committing a crime with
the expected utility of making a different
choice. Deterrence-based policies are naturally RATIONal choice
directed toward lowering the anticipated gains Predictions of Terrorist
of illegal behavior by raising its costs, but rela- Behavior
tively little attention has been given to raising
the anticipated value of legal behaviors rela- Any application of rational choice theory
tive to illegal ones. assumes that actors make decisions designed
In this article, we argue that raising the to optimize their own well-being while mini-
expected utility of abstaining from an unwanted mizing costs (Bentham [1781] 1996). We
behavior may be an effective policy choice in argue that terrorist actors, despite the grue-
certain circumstances. To support this argu- some nature of their crimes, can be consid-
ment, we use a newly developed dataset that ered rational actors (Crenshaw 2001; LaFree
documents specific terrorism-relevant actions and Ackerman 2009). Kruglanski and col-
by the Israeli government directed toward sub- leagues (2009) characterize the reasons for
state actors in the Palestinian Territories to violent participation as a quest for personal
assess effects of repressive actions (which significance; other scholars highlight the
raise the costs of terrorism) and conciliatory common terrorist goal of recognition and
actions (which raise the benefits of abstaining fame (Hamm 2004). These views suggest that
from terrorism) on terrorist activity during terrorists are generally less concerned about
three decision regimes: the First Intifada, the being punished and more concerned about
Oslo Lull, and the Second Intifada (Brym and their role in ensuring the well-being of their
Andersen 2011; Kuperman 2007; Rasler 2000). movement and its constituency, so strategies
We find that during the First Intifada and that successfully deter common criminals
the Oslo Lull, small numbers of conciliatory may be ineffective for terrorists (for an exam-
tactics led to increases in terror; however, as ple of this difference among airline hijackers,
Israel initiated more conciliatory actions, the see Dugan, LaFree, and Piquero 2005).
number of terrorist attacks declined. During Therefore, we deliberately adjust the costs
the Second Intifada, conciliatory tactics had a and benefits of perpetrating terrorist attacks
much stronger and linear effect on reducing so that they also relate to the larger goals of
Palestinian terrorist attacks. Furthermore, the terrorist movement as well as to any per-
despite the conventional confidence in deter- sonal fame (LaFree and Dugan 2004).
rence approaches, repressive actions have We illustrate this point by examining the
never led to decreases in terrorism and have equation for the expected utility of perpetrat-
sometimes led to increases in terrorism. Our ing a terror attack (Equation 1). Here, the
study is the first to show empirically that expected utility for person i, [E(uterror)i], is a
whereas solely repressive tactics tend to function of the perceived costs of and benefits
backfire, conciliation toward Palestinians can from perpetrating the act:
Dugan and Chenoweth 599

E(uterror)i = pi U( yi Fi) + (1 pi) U( yi), (1) finds that fewer people report intentions to
engage in illicit behaviors when they perceive
where p is the perceived probability of being a high risk of detection (Nagin 1998), illumi-
punished, y is the anticipated benefits of per- nating the importance of the perceived cer-
petrating the act, and F is the perceived pen- tainty of being punished (or pi from Equation
alty for the act.1 Thus, the decision to 1). Evidence supporting the importance of
perpetrate an act of terror for person i depends severity (or Fi from Equation 1) is mixed
on whether E(uterror)i > E(unonterror)i. The value (Nagin 1998; Paternoster 1987). Nagin (1998)
of this expectation varies across individuals cautions, however, that empirical support for
depending on their unique preferences, pro- deterrence based only on individual-level
pensities toward violence, and other individ- perception studies is insufficient to conclude
ual differences (see Tibbetts and Gibson that policies can deter crimeother than per-
2002). For terrorists, y typically advances the haps policies designed to alter perceptions.
movements progress toward its larger goals Instead, because policies are designed to
and improves their personal status, so any change the behavior of aggregate groups of
penalty directed toward individual i has less people, informative research should also be
of an impact on the overall expected utility. conducted at the aggregate.
This disproportion between y and F is well Aggregate studies of deterrence often take
illustrated when we consider suicide attacks. the form of interrupted time-series analyses
Even when F is death, the expected utility is that estimate the impact of a specific interven-
clearly higher for some individuals than are tion on crime, or ecological studies that assess
any alternative actions (see Hafez 2006). It the natural variation between sanction levels
naturally follows that to deter terrorism, F and crime rates (Nagin 1998). The appeal of
must affect more than just the individual. interrupted time-series analysis is that it allows
For this reason, and because policy is researchers to evaluate the effectiveness of
designed to influence the population as an specific types of interventions or policies,
aggregatenot just specific individualswe such as stricter laws or police crackdowns, on
generalize Equation 1 by averaging across all the particular behaviors these policies are try-
individuals to produce Equation 2: ing to affect (Goldkamp and Vilcica 2008;
Ross 1982; Sherman 1990). Efforts to reduce
E(uterror) = p U( y F) + (1 p) U( y) (2) terrorist behavior include terrorist apprehen-
sion and extended prison sentencing (Landes
This naturally leads to a policy strategy 1978), passage of anti-terrorism laws (Enders
designed to reduce the overall expected utility and Sandler 1993), assassination (Byman
of terror by increasing the certainty (p) and 2006; Hafez and Hatfield 2006; Jaeger and
severity (F ) of punishment. Rational choice Paserman 2008; Maoz 2007; Plaw 2008;
theory predicts that when states are able to Zussman and Zussman 2006), curfews and
raise the cost of perpetrating terror high containment strategies (LaFree et al. 2009;
enough so that the overall expected utility of Maoz 2007), deportation (Maoz 2007), home
terror is lower than that for abstaining from demolitions (Benmelech, Berrebi, and Klor
terror, rates of terrorism will drop. 2010), violent repression and military retalia-
tion (Brophy-Baermann and Conybeare 1994;
Brym and Araj 2006; Enders and Sandler
Raising the Costs of Perpetrating
1993; LaFree et al. 2009; Maoz 2007; Testas
Terror
2004), and indiscriminate repression (Lyall
Many studies have evaluated the effective- 2009). Some scholars have also examined
ness of raising the costs of illicit behavior, effects of containment policies, such as instal-
such as drunk driving and sexual assault lation of metal detectors (Cauley and Im 1988;
(Nagin 1998; Nagin and Paternoster 1993; Dugan et al. 2005; Enders and Sandler 1993;
Paternoster 1987). Research consistently Landes 1978).
600 American Sociological Review 77(4)

The second category of aggregate research Republican Army (IRA) terrorists as crimi-
includes ecological studies that examine the nals rather than political prisoners (LaFree et
natural variation between intervention levels al. 2009). In contrast, Britain behaved indis-
and illegal behavior to more clearly identify a criminately when it imposed a 36-hour
causal link (Nagin 1998). This methodology military curfew in Northern Ireland and
measures interventions continuously rather searched all homes for evidence of IRA mem-
than by a single shift in a discrete value before bership and weaponry (LaFree et al. 2009).
and after its implementation. Continuous We expect discriminate and indiscriminate
measures are far superior because they portray repression to have different effects on the
a more accurate measure of authorities level expected utility of terrorism (E[uterror] in
of effort, such as increases in the prison popu- Equation 2) because indiscriminate repres-
lation or the number of police on the street sion affects the larger constituency, on whose
(Levitt 1996, 1997). Ecological studies must interests terrorists generally base their goals.
be methodologically sophisticated in order to In other words, indiscriminate repression
identify any causal relationship between sanc- likely raises the costs of terrorism much
tions and crime and avoid issues of reverse higher than any penalty directed toward a
causality (e.g., more police are hired in specific offender. For instance, Lyall (2009)
response to increases in crime [Nagin 1998]). finds that indiscriminate violent repression
Although studies have examined effects of reduced the number of insurgent attacks in
repression on protest (Della Porta 1995; Kha- Chechnya over a three-month period by about
waja 1994; Koopmans 1993; Lichbach 1987; 24 percent.
Moore 1998) and effects of annual human
rights violations on subsequent terrorist attacks Possible backlash. Despite the popular-
(Piazza and Walsh 2009), no studies to date ity of deterrence theories, research shows that
have examined both repressive and concilia- punishment sometimes fails to deter and can
tory tactics and their effects on terrorism.2 even lead to more crime. For example, label
Until now, the only information available on theorists argue that when states impose pun-
state actions describes discrete interventions. ishment, offenders will begin to identify more
These data are usually gleaned from case stud- thoroughly with their role as law-breakers and
ies or well-publicized media reports (e.g., the will then fortify their criminal (or terrorist)
killing of Osama Bin Laden), yielding data that lifestyles (Becker 1963; Farrington 1977;
are ideal for models of interrupted time series Schwartz and Skolnick 1962). Other scholars
but not for ecological studies. claim that when punishment compromises the
Ours is thus the first published ecological perceived legitimacy of the punisher, it could
study, of which we are aware, that estimates elicit acts of defiance (Sherman 1993; Tyler
the effects of aggregate measures of state 2006). Indiscriminate repression, often
repression and conciliation on terrorism. viewed as illegitimate, may lead to defiance
Before we continue, we consider two impor- and increased violence. For instance, although
tant components of deterrence: the nature of Benmelech and colleagues (2010) find that
the targets and the possibility of backlash. home demolitions that targeted suicide terror-
ists families reduced subsequent Palestinian
Targets of deterrence (or repression). suicide attacks during the Second Intifada,
Deterrence is indiscriminate when it targets incidental or preventive home demolitions
individuals who have not yet broken the law resulted in a sharp increase in suicide terrorist
(general deterrence) and discriminate when it attacks.
targets known offenders (specific deterrence) Other research also finds that state repres-
(Andenaes 1971; Gibbs 1975). For example, sion exacerbates terrorism, or at least mobili-
the British government behaved discrimi- zation in general. Peroff and Hewitts (1980)
nately when it treated jailed, suspected Irish analysis of Northern Ireland indicates that
Dugan and Chenoweth 601

between 1968 and 1973, an increased British E[unonterror]). Therefore, we next consider how
troop presence led to more rioting, a finding raising the benefits of abstaining from terror
that White (1989) corroborates. Similarly, might reduce subsequent incentives to engage
LaFree and colleagues (2009) find that three in terrorism.
of six British interventions in Northern Ire-
land led to increases in the risk of Republican
Raising the Benefits of Abstaining
terrorist attacks. Khawaja (1993) finds that
from Terror
repressive acts by Israel increased the rate of
collective action by Palestinians in the West Equation 3 presents the expected utility of
Bank. In her study on protest events during nonterror, which parallels that for the expected
the Iranian Revolution, Rasler (1996) finds utility of terror in Equation 2:
that repression decreased protests in the short
run but increased them in the long run. Testas E(unonterror) = q U(x + G) + (1 q) U(x) (3)
(2004) concludes that political repression
may be negatively associated with levels of Here, q represents the probability of receiv-
terrorism over the short term, but that contin- ing rewards for abstaining from terrorism, x
ued use of repressive policies will eventually represents the value of the current situation
increase terrorist activity. Piazza and Walsh (i.e., the status quo), and G represents the
(2009) corroborate these findings on a global anticipated rewards of abstaining from ter-
scale, finding that countries that violate rorism. Aside from the different meanings
human rights are more likely to suffer terror- attributed to each component of the equation,
ism than are countries that adhere to human the primary difference between Equations 3
rightsparticularly rights that affect physical and 2 is that G (anticipated rewards) adds to
safety. Furthermore, Brym and Araj (2006) the value of x in Equation 3, whereas F
and Araj (2008) argue that terrorism emerges (anticipated punishment) detracts from the
as a response to perceived injustices, such as value of y in Equation 2. This small contrast
government repression, that inspire groups to establishes that this is a carrot, rather than a
mobilize in retaliation. stick, approach to countering terrorism (Frey
Another less evident form of backlash 2004).
results when offenders substitute one prohib- We can better appreciate the role of U(x +
ited activity for other illegal acts. For exam- G) when we take a closer look at x. By repre-
ple, despite finding that metal detectors led to senting the status quo, x directly relates to the
a reduction in airline hijacking, Enders and grievances that motivate people to commit
Sandler (1993) note a subsequent increase in acts of terror. For example, if we consider the
hostage taking events. Displacement thus case of ETA (Euskadi Ta Askatasuna), the
occurs when the expected utility of commit- nationalist terror organization in the Basque
ting one type of offense (e.g., hijacking) region of Spain, the status quo (x) is that the
drops below the expected utility of a different Basque region remains under the sovereignty
offense (e.g., hostage taking) (see Cornish of Spain, despite ETAs intent to establish an
and Clarke 1987). independent reunified Basque state (Clark
One could conclude that policies must be 1990; Mees 2003). If the Spanish government
designed to lower the expected utility of all offered rewards (G) to ETA or its Basque
reasonable offenses. But this would be pro- constituency (e.g., allowing the Basque peo-
hibitively costly for any state. Instead, we ple greater freedoms to practice their cultural
argue that by focusing efforts only on reduc- traditions), then the utility to ETA and the
ing the expected utility of perpetrating an act Basque people, if Spain were to follow
(i.e., E[uterror]), we ignore a potentially impor- through on these rewards, would be U(x + G)
tant component of the rational choice deci- (note that q is the probability that Spain actu-
sion: the utility of abstaining from crime (i.e., ally does follow through on these rewards). In
602 American Sociological Review 77(4)

other words, for ETA and the Basque people, the innocent (general) and known offenders
the utility would be a function of the advan- (specific), conciliatory actions can also target
tages of the rewards in addition to the status indiscriminately and discriminately.
quo (i.e., more freedoms while still under
Spains sovereignty). Although these rewards Targets of rewards (or conciliation).
might not offset their grievances, they would Neumann (2007) and others suggest that gov-
indeed improve upon the status quo. If the ernments can win the legitimacy battle in part by
incentive to behave according to the law is approaching terrorists constituents with a more
only the status quo (despite the absence of conciliatory tone and set of actions (i.e., indis-
punishment that deterrence promises), it may criminate conciliation). Through improved
not prevent rational actors from executing legitimacy, governments offer fair treatment (G
their illegal prerogatives. Quite simply, by in Equation 3) that could be lost after a terrorist
raising the overall E(unonterror) to be greater attack. Yet patience and consistency are required
than the E(uterror), a country may be more for conciliatory actions to influence terrorist
likely to experience a drop in terrorism.3 behavior, because these actions rely on trust that
With a few exceptions (Bueno de Mes- can only develop over time.
quita 2005; Kydd and Walter 2002; Lapan Many scholars note the importance of
and Sandler1988; Neumann 2007), little appealing to terrorists constituencies. The
research explicitly assesses effects of concili- Armenian terrorist group ASALA reduced
ation on crime or terrorism. Our interest is not attacks quickly after losing the support of its
limited to bargaining with terrorists, but also primary constituency, the Armenian diaspora
includes rewarding nonterrorist behavior. For (Dugan et al. 2008). Terrorist organizations
example, many programs (e.g., after-school rely on their constituencies for financial sup-
programs) provide legal alternatives to crime, port and as an ongoing recruitment pool.
which inherently raise the expected utility of Indeed, Crenshaw (2001) explains that the key
leading crime-free lives (Gottfredson et al. component for group survival is recruiting and
2004; Newman et al. 2000). Governments maintaining a strong membership. To preserve
have also attempted to provide alternatives to the loyalty of their constituencies, some terror-
terrorist violence. Small movements in this ist organizations provide social services to
direction include the establishment of the accommodate their needs. Poor constituencies
Basque Autonomous Community in the post- are especially dependent on terrorist organiza-
Franco Spanish constitution, which improved tions when they are the only service provider
the status quo for the Basque people (Clark (Flanigan 2010). This strategy has paid off for
1990). The Turkish army attempted to Hamas, which developed a network of charita-
improve the status quo for Turkish Kurds in ble organizations and services for the Palestin-
the mid-1990s by opening educational and ian people and then won a majority of the
health facilities to the Kurdish population in Palestinian parliamentary seats in January
the southeast (Cornell 2001). Such actions 2006 (Malka 2007). Hezbollah, which is better
raise the expected utility of not engaging in equipped to provide services to Southern Leb-
terrorism, and in these contexts they can be anon and the southern suburbs of Beirut than is
considered conciliatory rather than repres- the Lebanese government (Flanigan and
sive. Although these actions fall short of Abdel-Samad 2009), also won electoral victo-
compensating for the original grievances, ries. Constituencies dependence on terrorist
they do reward nonviolent behavior and may organizations demonstrates the strategic
encourage people to refrain from terrorism. importance of governments providing compet-
Notice that these efforts target individuals ing social support through overt conciliatory
who have engaged in illegal behaviors as well actions. We expect that when governments are
as those who were only at risk of such behav- able to improve the status quo (G) for indis-
ior. Just as repressive actions can affect both criminate yet relevant populations, terrorist
Dugan and Chenoweth 603

organizations will have difficulty maintaining created in 1964 consisting of four main fac-
strong membership and will subsequently lose tions: Fatah, the Democratic Front for the
the capacity to inflict harm. Liberation of Palestine (DFLP), the Popular
Conciliatory actions can also be discrimi- Front for the Liberation of Palestine (PFLP),
nate when they target known offenders. and the Palestinian Communist Party (PCP).
Deradicalization programs engage convicted A core group of exiled Palestinian Fatah
terrorists in religious dialogue to dismantle members, including Yassir Arafat and Khalil
the ideological beliefs that justify terrorism. Al-Wazir (Abu Jihad), dominated the PLO.
Concurrently, these programs work closely The groups leadership looked to other anti-
with detainees families to prepare them to colonial movements, such as the Algerian
lead normal, nonviolent lives by providing Liberation Front, for inspiration and adopted
financial support to educate the children, armed struggle as the sole method of con-
training wives, and helping to reintegrate fronting the Israeli occupation in the late
detainees into the community (Kruglanski, 1960s (see Article 9 of the Palestinian
Gelfand, and Gunaratna 2010). Similar con- National Charter, drafted in 1968) (Kadi
ciliatory efforts took place in Europe in the 1969). Early examples of armed actions
1980s when Spain pardoned imprisoned ETA include multiple airline hijackings, high-pro-
members after they publicly renounced the file assassinations, and the famed Munich
organization and its use of violence. This Massacre, a high-profile kidnapping and mur-
reinsertion policy allowed ETA members to der of 11 Israeli athletes at the 1972 Munich
live normal lives, free from ETA (Mees 2003). Olympic Games by the Palestinian group
The Italian government offered leniency to Black September. However, due to popular
members of the Red Brigades when they pro- backlash against the latter incident, Palestinian
vided information that led to the apprehen- groups refocused their efforts on armed strug-
sion of other members (Crenshaw 2001; gle against Israel within Israel itself.
Cronin 2006). In all of these examples, gov- We now turn our focus to the three time
ernments strategically raised the expected periods under inquiry. Scholars typically
utility of choosing a violence-free life for describe the First Intifada, the Oslo Lull, and
known terrorists. Although some of these the Second Intifada as distinct periods in
efforts show promise (Kruglanski et al. 2010), which the Israeli government adopted fairly
we expect conciliatory efforts that target the uniform approaches to managing the terror
broader constituency will be more effective in environment (Brym and Andersen 2011;
reducing terrorism in the long term, because Kuperman 2007; Rasler 2000). Previous stud-
they will eventually shift popular opinion ies have sought to explain the causes of these
away from terrorism, depleting terrorist regimes, whereas we are more interested in
groups of public support. the effects of Israeli policies during these
periods. We note that, importantly, different
terrorist groups predominated during each of
Tactical Regimes and the distinct regimes. Moreover, despite adop-
the IsraeliPalestinian tion of particular regimes, we observe a mix
Conflict Environment of conciliatory and repressive tactics during
each of these periods, demonstrating that the
Much can be said about Israels relationship tactical regimes were not as uniform as is
with its contentious neighbors,4 but our pri- often suggested.
mary goal is to explore Israels objective of
reducing Palestinian terrorist attacks. The
The First Intifada (1987 to 1993)
modern terrorist environment can be traced to
the Palestine Liberation Organization (PLO), Although the PLO endorsed and pursued
a secular, nationalist, umbrella organization violent methods of resistance against Israeli
604 American Sociological Review 77(4)

occupation, the first Palestinian Intifada that that turned the focus to defending against threats
began in December 1987 was initially a non- from regional rivals, such as Iran.
violent popular uprising that erupted in a Gaza Commentators often point out that neither
refugee camp and spread throughout the Gaza side has lived up to its Oslo obligations (Mar-
Strip and the West Bank (King 2007). The shall 2009). Israeli settlements continued to
uprising succumbed to factional divisions and expand throughout the 1990s, and the PA
violence by mid-1990 (Pearlman 2008/2009). failed to maintain security in the Palestinian
Although secular Palestinian nationalists dom- Territories, mismanaged economic affairs,
inated the campaign and the consequent Oslo and engaged in widespread corruption. Pales-
Accords, two offshoots of the Muslim tinian extremist groupsparticularly Hamas
Brotherhood, Islamic Jihad and the Islamic and the Palestinian Islamic Jihad (PIJ)con-
Resistance Movement (Hamas), emerged as tinued to launch sporadic violent attacks
important players in the latter part of the First against Israeli civilians after the Oslo Accords
Intifada. These two religious groups aspired to were signed, and Jewish extremists carried
create an Islamic Palestine governed by sharia out violence against Palestinian civilians as
law and refused to recognize Israel as a legiti- well. Yet Israel responded to these incidents
mate state (Brym and Andersen 2011; Kydd with considerable restraint until the Second
and Walter 2002). The coexistence of multiple Intifada began (Brym and Andersen 2011).
Palestinian groups undermined the unity of
Palestinian resistance during the First Intifada,
The Second Intifada (2000 to 2004)5
and by spring 1990, Palestinians were killing
more fellow Palestinians than Israeli soldiers In September 2000, tensions erupted when
were (Pearlman 2008/2009; Rigby 1991). then-Defense Minister Ariel Sharon visited
Nevertheless, popular opinion in Israel the Temple Mount. Viewed as an affront to
began to turn against Israeli occupation of the Islamic faith and a dishonor to Palestinian
Palestinian territories, and in October 1991, traditions, Palestinian extremist groups initi-
Israeli and Palestinian negotiators met for ated a sustained violent campaign that became
bilateral talks at the Madrid Conference. After known as the Second Intifada, also known as
more than eight months of talks, key issues the al-Aqsa Intifada (Beitler 2004). In con-
such as the status of the Jewish settlements trast to the First Intifada, the Second Intifada
inside the Palestinian Territoriesremained was primarily violent.
unresolved. But in 1993, Israeli and PLO Religious groups (Hamas and PIJ) carried
officials began to meet in secret in Oslo, Nor- out the majority of terrorist attacks during this
way, paving the way for a series of agree- period, which became distinctive for the sui-
ments known as the Oslo Accords. cide bombing campaigns perpetrated against
Israeli civilians and occupation forces. From
2000 to 2005, suicide bombing was the
The Oslo Lull (1993 to 2000)
favored tactic of Islamist Palestinian groups
As part of the Oslo Accords, negotiators estab- such as Hamas and PIJ. Secular groups such
lished the Palestinian Authority (PA), a semi- as the Popular Front for the Liberation of
independent governing body that assumed Palestine (PFLP), the Democratic Front for
limited control over parts of Gaza and the West the Liberation of Palestine (DFLP), and the
Bank. In return, Arafat and the Palestinian lead- Al-Aqsa Martyrs Brigades, a collection of
ership agreed to recognize Israels territory Fatah-affiliated cells, deployed this tactic as
within the 1967 borders. Brym and Andersen well (Bloom 2004; Hafez 2006), but not to
(2011) write that a new decision regime began the same extent as religious groups.
to dominate Israel at this point: a regime that Israel responded by intensifying occupation
privileged political solutions over military ones of the West Bank and Gaza Strip, initiating pro-
with regard to the Palestinian Territories, and grams such as bulldozing suicide bombers
Dugan and Chenoweth 605

homes, increasing curfews, assassinating mili- of abstaining from terrorism is akin to initiating
tant leaders, and constructing a highly contro- conciliatory actions by the Israeli government
versial concrete and barbed-wire barrier that toward Palestinians. Israeli actions may target
weaves between Israeli and Palestinian occu- known Palestinian terrorists (discriminate) or
pied land. Although Israels military occupation Palestinian civilians in general (indiscriminate).
of the West Bank ended in 2005, Hamas won
the 2006 elections to become the legitimate
Hypotheses
government of the PA. Hamas has refused to
negotiate with Israel or recognize Israels right As a guide to the hypotheses we present a
to exist, and indeed has continued to support the final equation that combines Equations 2 and
use of violence against Israelis. At the same 3 into one inequality. When
time, Israeli elected leaders have also become
increasingly hawkish, adopting provocative E(uterror) < E(unonterror), (4)
policies such as further settlement expansion. In
2009, Israel launched Operation Cast Lead, a we expect less terrorism. We now present three
war in the Gaza Strip that resulted in thousands hypotheses that follow directly this inequality
of deaths. With no resolution in sight, Israel condition, and two hypotheses that relate to
continues to face attacks from within the Pales- specific components of the expected utility of
tinian Territories, and the Israeli government terror and nonterror.
continues to rely overwhelmingly on force to
attempt to deter yet more violence. Hypothesis 1: Any Israeli action leads to fewer
Israel continues to experience violent attacks terrorist attacks by Palestinians.
by Hamas, PIJ, and Fatah-affiliated organiza-
tionsall of whom are sensitive to and depend- We present this hypothesis because
ent on sympathy and support from the Palestinian whether Israel lowers the expected utility of
population, the vast majority of whom would not terrorism through repressive actions or raises
engage in terrorist activities as their default the expected utility of nonterrorism through
choice (Bloom 2004; Hafez 2006). As men- conciliatory actions, the result should pro-
tioned earlier, most terrorist groups rely on will- duce a higher utility for nonterrorism, thus
ing recruits to survive (Crenshaw 2001). leading a rational actor away from terrorist
Moreover, these groups are much more likely to behavior. The next two hypotheses are spe-
endure when the surrounding population is either cific to each type of expected utility.
complicit in their activities, sympathetic toward
them, or convinced that informing Israelis will Hypothesis 2: Conciliatory actions lead to few-
result in swift retaliation from the terrorist groups er terrorist attacks by Palestinians.
themselves (Kalyvas 2006; Kocher and Kalyvas Hypothesis 3: Repressive actions lead to fewer
2007; Lyall and Wilson 2009). As many analysts terrorist attacks by Palestinians.
have noted, the civilian population is the central
fulcrum of any protracted civil conflict (Kalyvas These hypotheses allow effects of the spe-
2006; Lyall and Wilson 2009), and the Israeli cific type of action to behave independently
Palestinian conflict is no exception. from effects of the other type. Because both
punishment and reward are expected to tilt
inequalities in favor of nonterrorism, we pre-
Applying Rational dict that both types of actions will lead to
Choice to the IsraelI fewer attacks.
Palestinian Conflict The two secondary hypotheses are generated
from other components of rational choice.
Within the context of Israel, raising the costs of
terror is akin to initiating repressive actions Hypothesis 4: Indiscriminate repressive actions
against the Palestinians; and raising the benefits lead to more terrorist attacks.
606 American Sociological Review 77(4)

Here, we expect repressive actions that Dugan 2011). Third, we use newly collected
affect the Palestinian people in general will event data on the specific actions taken by the
cause a backlash of violence. Prior literature state of Israel directed toward substate actors
has found evidence of backlash (Benmelech relevant to the Israeli conflicts with Palestinian
et al. 2010), and an important source of back- populations. This allows us to aggregate the
lash is the Israeli governments compromised data to any temporal unit. For this research,
legitimacy. Without legitimacy, Palestinians we chose to aggregate to the month because
have little reason to trust that the Israeli gov- the IsraeliPalestinian conflict is highly
ernment will behave fairly in times of peace dynamic, and information is outdated after a
(Braithwaite 2005).6 few months. Finally, we adopted a method
This reaction by the larger constituency that allowed us to estimate effects of various
can favor Israel if it provides peaceful alter- tactics on terrorist attacks while accounting
natives to violence. This leads to the next for reciprocal effects that terrorist attacks may
secondary hypothesis: have on counterterrorism actions.

Hypothesis 5: Indiscriminate conciliatory ac-


Data
tions lead to a larger decrease in terrorist
violence than do other actions. Data for these analyses come from two
sources. The dependent variable, Palestinian
Quite simply, if Israel builds legitimacy terrorist attacks, comes from the Global
and trust through conciliatory actions, the Terrorism Database (GTD), which was col-
Palestinian people will be more reluctant to lected by scholars at the Center for the Study
sabotage the possibility of peace by support- of Terrorism and Response to Terrorism
ing terrorism. Without their support, terrorists (START) (LaFree and Dugan 2007). The
will be unable to maintain an ongoing cam- independent variables, Israeli state actions,
paign of violence. come from a new database collected by the
authors called Government Actions in a
Terrorist Environment-Israel (GATE-Israel)
Research Strategy (Chenoweth and Dugan 2011). The current
The study of counterterrorism has long been analysis is constrained to the years covered
constrained by a lack of high-quality data by this dataset, June 1987 through December
with which to evaluate these approaches in a 2004. This collection is part of a larger effort
robust way (Chenoweth and Dugan 2011). To to document tactics used by states to reduce
test our hypotheses, we provide several terrorist threats (Chenoweth and Dugan
empirical innovations. First, we focus on rela- 2011).7
tional data between state and Palestinian
actors in Israel. Second, we include a wide Palestinian terrorist attacks. The GTD
range of state actions under the category of is an event-based database that documents all
counterterrorism. Many studies focus terrorist attacks across the globe from 1970
exclusively on repression or concessions, but through 2010.8 The collection was originally
we collected data on thousands of types of compiled by the Pinkerton Global Intelligence
state actionsfrom raids and arrests to allow- Services (PGIS) from 1970 through 1997, and
ing telephone lines to be built in refugee then cleaned and updated by START staff and
campsso that we can explore the relative contractors (Dugan 2012; LaFree and Dugan
effects of different types of interactions. This 2007). Regardless of the collecting agent, all
allows us to move away from misleading cases came from open sources available
characterizations of conflict as a series of through media and other reporting agencies.
dichotomous choices and to consider a wider The reliance on open sources produces
range of conflict actions (Chenoweth and some strengths and some weaknesses (see
Dugan and Chenoweth 607

LaFree and Dugan 2007). The most relevant by Palestinians against Israelis from June
concern for this research is that open sources 1987 through December 2004, the months
are biased toward the most noteworthy events used in the current analysis.
and are likely to underrepresent attacks on
more remote parts of the globe. However, Israeli state actions. The GATE-Israel
because the IsraeliPalestinian conflict has Database includes all Israeli actions toward
been central to the international arena, we are substate actors from June 1987 through
not very concerned about missing events. Fur- December 2004. Our focus is on actions
thermore, because it is inevitable that some directed toward Palestinian terrorists or civil-
events might have missed public scrutiny, we ians. We collected data using Textual Analysis
have no reason to believe this issue would be by Augmented Replacement Instructions
systematic. (TABARI), which searches news articles and
Two nuances in GTD data are relevant to identifies observations that match the criteria
the current analysis. First, all events from the of an extensive set of dictionaries designed to
year 1993 are missing. Boxes that held the capture international and domestic activity
original data for that year were lost while still (Schrodt 2001, 2006).10 TABARI is an auto-
under the control of PGIS. All analyses thus mated text-coding program that codes news
exclude the months of that year. We note that articles based on noun and verb pattern recog-
1993 spans the tactical regimes of the First nition. This method is surprisingly accurate
Intifada and the Oslo Lull. Second, although and considerably more efficient than human
the original data were collected prospectively coding of entire stories (Schrodt 2001,
by PGIS, data since 1998 were collected ret- 2006).11 For other recent applications of
rospectively, inevitably undercounting attacks TABARI, see Clauset and colleagues (2010);
that were documented only in sources that are Shellman (2008); and Shellman, Hatfield, and
no longer available. To adjust for this, all Mills (2010).
models include an indicator variable that dis- In our study, we used TABARI to code
tinguishes the retrospective period from the 243,448 Reuters articles downloaded from
prospective period (GTD2 = 1 if year > 1997 Factiva using the word Israel* as the search
and 0 otherwise).9 We expect the coefficient criterion for the period January 1, 1987 to
for this estimate to be negative, absorbing December 31, 2004. Reuters archives begin
the undercounting for the retrospective data in June 1987, delineating the beginning point
collection. of this research. We determined the end date
Dependent variables for all analyses come at the time we wrote the original grant pro-
from the GTD and are counts of Palestinian posal. We chose Reuters over other wire ser-
terrorist attacks for each month. We used only vices because of its consistent editorial
GTD cases that involved at least one Israeli control and its tendency to use a simpler
target in Israel or the Palestinian territories. sentence structure and vocabulary than alter-
Furthermore, because the GTD provides native news sources such as the Washington
some information about the perpetrator in Post and the New York Times (Schrodt, Davis,
nearly 70 percent of these attacks, and because and Weddle 1994; Schrodt and Gerner 1994).
the majority of attacks are Palestinian related, After TABARI identified relevant news
we included in the current analysis all attacks articles, we filtered the output to keep only
by unknown perpetrators (30.2 percent) and actions that the Israeli government imple-
excluded attacks by non-Palestinians (e.g., mented toward substate targets. We chose not
Lebanese or Israeli terrorists). Although it is to select on the types of actions (or verbs) to
possible that some of the unknown attacks make sure that all unexpected actions would
were by non-Palestinians, we doubt that any be captured. This method ensures that we
error is systematic. After filtering attacks by captured a wide range of actions that may not
these criteria, we found 1,208 terrorist attacks immediately seem like counterterrorism but
608 American Sociological Review 77(4)

Table 1. Seven-Point Guide for the ConciliatoryRepression Scale


1 = Accommodation/Full Concessions
Appeasing or surrendering to adversary
Making full concessions to opponents demands
Action required
2 = Conciliatory Action
Making material concessions
Taking action that signals intention to cooperate or negotiate with opponent
3 = Conciliatory Statement or Intentions
Expressing intention to cooperate or showing support
Verbal expression short of physical action
4 = Neutral or Ambiguous
No clear moves toward or away from resolution of conflict
Includes all attempts to ask for help from a third party to resolve the conflict
Requires more context to determine whether it is conciliatory or repressive
Includes all infighting over Palestinians within the Israeli government
5 = Verbal Conflict
Express intent to engage in conflict or threaten
Decline to cease ongoing conflict; maintain the status quo during conflict
Short of physical action
6 = Physical Conflict
Physical or violent action aimed at coercing opponent
No apparent intention to kill
7 = Extremely Deadly Repression
Physical action exhibiting intent to kill
Torture or severe violence (such as severe beatings), which could easily kill someone

are relevant to the overall conflict, such as people (i.e., individuals who are not suspected
allowing developers to build better water of involvement in terrorist activity).
wells in the Palestinian territories. Following the autocoding stage, research
Furthermore, we autocoded each action assistants hand-checked each observation to
according to several additional criteria. Rele- ensure that TABARI coded each story correctly
vant to this study, we established a Concilia- and to mark for removal any irrelevant cases.
toryRepression scale for each action, During this cleaning process, we also attributed
illustrated in Table 1. We based all codes on the each government action to politicians, the mili-
Palestinian perspective. The scale features dis- tary, the judiciary, or the police. This process
tinctions in the intensity of the action as well as revealed a relatively high degree of error (about
its relative placement of the action on a concil- 30 percent). Research assistants corrected these
iationrepression spectrum, similar to the errors, and both authors checked their coding to
Goldstein (1992) scale. Table 2 lists specific ensure intercoder reliability. The resulting file
actions found in our data that commonly fell contains the lead sentence to the article, the
into each category. We also autocoded each actor, action, target, the new codes mentioned
observation for whether the actions target was earlier for 6,070 Israeli government actions, and
discriminate or indiscriminate. Discriminate other variables not relevant to the current study.
actions attempt to single out guilty or sus- This dataset gives an action-by-action view of
pected parties from uninvolved parties. Indis- Israeli attempts to resolve conflicts with various
criminate actions directly affect uninvolved non-state actors, including Palestinian, Israeli,
Dugan and Chenoweth 609

Table 2. Examples of Common Actions for Each Scale Item


Accommodation/Full Concessions Verbal Conflict
Withdrew from town Made pessimistic comment
Signed peace accord Dismissed
Handed town to Palestinians Blamed for attack
Denied Responsibility
Conciliatory Action Threatened military force
Met to discuss
Released Physical Conflict
Lifted curfew Demolished
Pulled out Barred
Investigated abuse Sealed off
Imposed curfew
Conciliatory Statement or Intensions Arrested
Expressed optimism
Agreed to hold talks Extreme Deadly Repression
Praised Palestinians Shot dead
Expressed desire to cooperate Fired missiles
Admitted mistake Clashed with
Raided
Neutral or Ambiguous Helicopter attack
Infighting over
Failed to reach agreement
Hosted a visit
Appealed for third-party assistance
Investigated

and Lebanese militants. Over 90 percent of months during the First Intifada (December
actions were directed toward Palestinians or 1987 through August 1993), noting that the
Palestinian terrorists (Chenoweth and Dugan eight months in 1993 are missing. The third
2011). For the current analysis, we retained dataset includes only the months during the
only actions relevant to the IsraeliPalestinian Oslo Lull (September 1993 through August
conflict. 2000), with the four months of 1993 and the
first four months of 1994 missing. The final
dataset includes only the months during the
Analysis
Second Intifada (September 2000 through
We combined the GTD and GATE-Israel data December 2004). We conducted all analyses
into four monthly time-series datasets. The using the three smaller datasets to assess the
first includes most of the 211 months from robustness of the findings to the different
June 1987 through December 2004. We omit- tactical regimes.
ted all months from 1993 (211 12 = 199) We used a two-step methodological
because the GTD data are missing; we approachone parametric and the other non-
dropped the first four months of the set due to parametricto assess the robustness of the
the lagged dependent variable, which we findings and to provide a visual representa-
describe below (199 4 = 195). Furthermore, tion of the relationship between government
because 1993 is missing, we also dropped the actions and terrorist attacks. First, we para-
first four months of 1994 from the models metrically tested these relationships by mod-
due to the lagged dependent variable (195 4 eling them using a negative binomial
= 191). The second dataset includes only the regression (NBR). Because the relationships
610 American Sociological Review 77(4)

can be nonlinear (i.e., the effect of govern- Table 3. Primary Independent Variables
ment actions on terrorism depends on the Measuring Israeli Government Actions
number of actions), we tested for both linear Model 1. All Actions (Hypothesis 1)
and nonlinear relations using squared terms.
Model 2. Conciliatory and Repressive Actions
We used NBR because the dependent variable (Hypotheses 2 and 3)
is a count of the number of attacks in the cur-
Model 3. Conciliatory-Discriminate, Concil-
rent month, which is a relatively rare event. iatory-Indiscriminate, Repressive-
Furthermore, because there is a chance that Discriminate, and Repressive-
the variance is over-dispersed, we chose the Indiscriminate (Hypotheses 4
more flexible negative binomial over the and 5)
more restrictive Poisson model (Greene
2008).
The nonparametric approach, Generalized
Additive Models (GAM), allows us to visu- retrospective data collection (1998 to 2004),
ally examine relationships between govern- Attacks represents the number of Palestinian
ment actions and the number of terrorist attacks directed toward Israelis for the current
attacks during the next month, while control- month (t) and the four previous months (t 1
ling for all the same variables that are in the through t 4).13 By including measures of
NBR. Because the dependent variable is a lagged attacks in the models, we are better able
count, we used a log link function for a Pois- to isolate the effect of actions in the previous
son distribution. This methodology uses a months on attacks in the current month. Without
smoothing function to isolate the relationship controlling for lagged attacks, the estimated
between actions and attacks without imposing relationship between actions and attacks would
assumptions about linearity (Hastie and Tib- likely be distorted because it would include any
shirani 1990). The method produces graphs effects that earlier attacks had on both govern-
that show partial predictions of our independ- ment actions and current attacks.
ent variables with confidence intervals, allow- For each dataset, we ran three models,
ing us to visually examine the nature of the measuring government actions according to
relationships for consistency with our hypoth- the several dimensions listed in Table 3. Most
eses (Xiang 2001). Using both methods to apparent is that models go from least granular
examine these relationships allowed us, first, (All Actions) to most granular, where repres-
to test the robustness of the findings with and sive and conciliatory actions are partitioned
without parametric assumptions and, second, by whether they were discriminate or indis-
to visually examine the nature of the relation- criminate. We ran each set of independent
ship in the absence of statistical significance. variables using NBR and GAM for all four
The general format of all NBR and GAM datasets, totaling 12 models.
models follows the form shown in Equation 5,

Attackst = f(Actionst1, Regimes, GTD2,


Results
(5)
Attackst1, Attackst2, Attackst3, Attackst4), Because this research presents a new dataset,
we begin our analysis by presenting descrip-
where the vector Regimes is included only in the tive statistics for the primary dependent and
model for all months and includes an indicator independent variables for all months speci-
variable for each Intifada (First and Second). fied to the tactical regime (see Table 4). For
These are important controls because levels of each regime, we present the means, standard
terrorist activity and the Israeli decision regime deviations, and proportion of months that
differed substantially depending on the tactical have a value of zero. We included only the
regime (Brym and Anderson 2011).12 GTD2 months that are in the analyses, so we lagged
is an indicator variable depicting the years of all actions by one month, excluded the first
Dugan and Chenoweth 611

Table 4. Means, Standard Deviation, and Proportion of Zeros


All Months (n = 191) First Intifada (n = 61)

Variable Mean SD P(0) Mean SD P(0)


Attacks 5.84 5.84 .13 7.72 5.09 .07
All Actions 27.90 16.33 .00 19.57 9.30 .00
Conciliatory 7.55 5.91 .07 4.03 3.16 .16
Repressive 17.95 12.17 .00 14.38 7.25 .00
Conciliatory-Discriminate 1.47 1.51 .32 1.38 1.29 .31
Conciliatory-Indiscriminate 6.04 5.31 .09 2.61 2.49 .21
Repressive-Discriminate 4.13 4.17 .08 3.23 2.65 .10
Repressive-Indiscriminate 13.82 9.53 .01 11.13 5.58 .00

Oslo Lull (n = 76) Second Intifada (n = 52)


Attacks 3.07 3.92 .26 7.79 7.98 .00
All Actions 25.42 11.44 .00 42.00 19.76 .00
Conciliatory 10.16 6.57 .03 8.12 5.28 .00
Repressive 12.92 7.62 .00 29.89 14.21 .00
Conciliatory-Discriminate 1.63 1.73 .33 1.38 1.42 .31
Conciliatory-Indiscriminate 8.46 5.79 .03 6.73 4.88 .02
Repressive-Discriminate 3.01 2.39 .12 6.85 6.11 .00
Repressive-Indiscriminate 9.91 6.13 .01 23.04 11.17 .00

Note: We generated all statistics from the data used to estimate the models. This means the first four
months of each series (and in 1994) were excluded due to the lagged dependent variable, and all
months from 1993 were excluded.

four months of each series (and in 1994), and and indiscriminate; averaging nearly 14 indis-
omitted all 1993 months. Turning first to the criminate acts each month. In fact, according
dependent variable, we see that over the to the data, in all but one month there was at
entire period, there were, on average, almost least one repressive-indiscriminate action
six attacks per month with only 13 percent of (shown by P(0) = .01). The table also shows
months free of terrorist attacks. When we turn there were more than twice as many repres-
to the three tactical regimes, we see, as sive acts (scale items 5, 6, and 7) as concilia-
expected, that terror attacks were higher in tory acts (scale items 1, 2, and 3) each month,
the First and Second Intifada compared to the and both actions were more often indiscrimi-
Oslo Lull (7.72 and 7.79 versus 3.07, respec- nate than discriminate. Conciliatory-discrimi-
tively). Similarly, more than a quarter of the nate acts were rare, averaging 1.47 each
months during the Oslo Lull saw no terror month. More revealing is that in 32 percent of
attacks, whereas every month during the months, Israel offered no discriminate concil-
Second Intifada had at least one attack and iatory actions.
almost every month in the First Intifada had Although these patterns generally hold over
an attack. all three tactical regimes, there are important
Turning to the statistics describing Israels differences worth noting. First, Israels average
actions, we see that over the entire period, number of total actions increased over time. It
Israel initiated an average of nearly 28 actions was smallest during the First Intifada (19.57); it
a month toward Palestinians. By scanning the rose during the Oslo Lull (25.42); and it reached
averages for actions partitioned by type, we a peak during the Second Intifada (42.00). Israel
see Israels actions were most often repressive most frequently relied on conciliatory actions
612 American Sociological Review 77(4)

Figure 1. Quarterly Repressive and Conciliatory Actions by Israel and Palestinian Terrorist
Attacks

during the Oslo Lull (10.16) and most fre- sive actions are marked with a light color and
quently relied on repressive actions during the conciliatory actions are shown with a solid
Second Intifada (29.89). Finally, almost every black bar. The three tactical regimes are sepa-
month saw a wide range of types of actions by rated by vertical dashed lines. Figure 1 shows
Israelexcluding conciliatory-discriminate, of that both repressive and conciliatory actions
course. appear to track terrorist attacks rather closely
Because we are interested in the temporal (r = .49 and r = .24, respectively). The fre-
relationship between government actions and quency of terrorist attacks and the number of
terrorist attacks, we now map the quarterly conciliatory actions rose during the First Inti-
counts of repressive and conciliatory actions fada, and repressive government actions show
onto the quarterly count of Palestinian terror- no distinct pattern. Having said that, all three
ist attacks directed toward Israelis over time.14 series peaked around the time of the Oslo
This comparison is especially important Agreement and then declined until the begin-
because both measures come from different ning of the Second Intifada. During the Sec-
sources that were collected independently of ond Intifada, all three trends rose dramatically
one another. Because we expect both govern- and then declined at different rates.16
ment actions and terrorist attacks to vary with We now turn to results for the NBR and
the intensity of the IsraeliPalestinian con- GAM models that used data for all months
flict, we would also expect these measures to from June 1987 through December 2004.
track one another. Figure 1 presents a bar Table 5 presents coefficients and standard
chart of Israeli actions with a line depicting errors from the NBR models. Although we
Palestinian attacks mapped over it.15 Note tested all nonlinear relationships, this table
that actions are scaled by the left axis and includes only squared terms if the tests con-
attacks are scaled by the right axis. Repres- cluded nonlinearity. Table 5 also includes
Dugan and Chenoweth 613

Table 5. Negative Binomial Coefficients and (SE), June 1987 through December 2004, n = 191

Model 1 Model 2 Model 3


Government Actions
All Actions .005
(.004)
Conciliatory .048
(.031)
Conciliatory2 .003*
(.001)
Repressive .009
(.006)
Conciliatory-Discriminate .006
(.041)
Conciliatory-Indiscriminate .051
(.036)
(Conciliatory-Indiscriminate)2 .003*
(.002)
Repressive-Discriminate .016
(.015)
Repressive-Indiscriminate .005
(.008)

Tactical Regime
First Intifada .732** .626** .660**
(.156) (.185) (.191)
Second Intifada 1.263** 1.120** 1.153**
(.264) (.277) (.278)

Controls
GTD2 1.023** 1.032** 1.028**
(.255) (.252) (.253)
First Lagged Attacks .023* .025* .025*
(.011) (.011) (.011)
Second Lagged Attacks .001 .004 .002
(.010) (.010) (.010)
Third Lagged Attacks .035** .035** .034**
(.011) (.011) (.011)
Fourth Lagged Attacks .018 .020 .020
(.011) (.011) (.011)
*p .05; **p .01 (two-tailed tests).

results for the tactical regimes and control more, terror attacks during the retrospective
variables. These findings confirm what we data collection period (GTD2) were lower
expected: attacks were highest during the than when data were collected prospectively.
Second Intifadaas evidenced by the large Finally, coefficients for the lagged attacks
and significant coefficient estimates in all show that the first and third lagged attacks
three models corresponding to the Second were most important.17
Intifadaand second highest during the First Turning now to the hypotheses, Model 1 in
Intifada (compared to the Oslo Lull). Further- Table 5 shows the parametric estimate of the
614 American Sociological Review 77(4)

1.61374

.360915
1 121
Lagged all actions

Figure 2. Partial Predictions of All Actions from the Past Month on Terrorist Attacks in the
Current Month
Note: Smoothing component of lagged actions has three degrees of freedom.

relationship between all government actions p = .12);19 however, when Israel initiated
and Palestinian terrorist attacks. Accordingly, more than eight conciliatory actions, terror
the value of the coefficient is positive but statis- attacks appeared to drop (the quadratic term
tically null ( p = .26). Figure 2 presents partial is negative, p = .024). Conversely, the NBR
predictions (with 95 percent confidence bands) suggests a positive relationship between
of all government attacks for the past month on repressive actions and terror attacks the fol-
terrorist attacks during the current month. We lowing month, although the significance is
include a horizontal line to mark zero (or no less than marginal ( p = .15).
relationship). Also, because the x-axis is the Figures 3a and 3b present partial predic-
number of actions in the previous month, any tions of conciliatory and repressive actions
nonlinear relationship suggests that the effect of from the past month on terrorist attacks in the
actions depends on the quantity of actions by the current month, respectively. Both figures
Israeli government in the previous month. This mimic the NBR findings. Figure 3a shows
graph suggests that the number of Israeli actions that with a low number of conciliatory actions,
in the past month is unrelated to the number of attacks appeared to increase; however, as
Palestinian terrorist attacks targeting Israelis. Israel initiated more conciliatory actions, the
The increase at the end of this graph is negligi- expected number of attacks in the next month
ble because it is driven entirely by one month droppedsupporting the prediction of
that had 121 actions; all other months had fewer Hypothesis 2. In fact, Israel initiated more
than 90 actions. We conclude, therefore, that than eight conciliatory actions in 38 percent
Hypothesis 1 is unsupported.18 of months, suggesting that the drop in Figure
Turning now to findings related to our 3a is not driven by outliers. As we examine
second and third hypotheses, the NBR coef- the repressive actions in Figure 3b, a small
ficient estimates presented in Table 5 under number of actions seem to have produced no
Model 2 suggest that when Israel initiated effect on attacks. However, as Israel initiated
eight or fewer conciliatory actions there was more repressive actions, the expected number
more terrorism the following month (the main of attacks rose. We interpret this finding with
effect is positive yet less than marginal, caution because in most months (87 percent),
Dugan and Chenoweth 615

a. Conciliatory Actions b. Repressive Actions


.142908 1.57031

1.04907 .171465
0 27 1 80
Lagged Conciliatory Acts Lagged Repressive Acts

Figure 3. Partial Predictions of Conciliatory and Repressive Actions from the Past Month on
Terrorist Attacks in the Current Month
Note: Smoothing component of lagged actions has three degrees of freedom.

Israel initiated 30 or fewer repressive actions, and present the results in Tables 6a, 6b, and
suggesting that the increase at the end of the 6c, respectively. These tables include the
graph is driven by only 13 percent of the original findings from all months, for com-
months. At best, Hypothesis 3 is unsupported, parison. We included controls in the estima-
and at worst it is opposed, suggesting some tion, but we omit them here for parsimony.
support for Hypothesis 4 (backlash). Due to space constraints, partial predictions
We now turn to results for Model 3 in Table from the GAM models are excluded, but they
5, which evaluates effects of discriminate and are available in the online supplement.
indiscriminate actions. We expect a positive and Table 6a presents NBR coefficients for the
significant coefficient for repressive-indiscrimi- effect of all actions on Palestinian terror
nate actions (Hypothesis 4) and a negative and attacks. We see that the relationship is null,
significant coefficient for conciliatory-indis- regardless of the period. Yet coefficients are
criminate actions (Hypothesis 5). Results show positive during the First and Second Intifada
some support for Hypothesis 5, but no support and negative during the Oslo Lull, suggesting
for Hypothesis 4. Like results for conciliatory that different dynamics were at play during
actions in general, conciliatory-indiscriminate different tactical regimes. Regardless, it is
actions seem to affect terrorism only after safe to conclude that these data do not support
months in which Israel initiated relatively large Hypothesis 1, but a closer look at each of the
numbers (more than 8.5) of such actions.20 The tactical regimes is in order.
GAM partial predictors for all four variables Table 6b presents coefficient estimates for
presented in Figure 4 conform to the findings Model 2, testing the relationship between
from Table 5. The drop below zero in Figure 4b conciliatory and repressive actions separately
is driven by more than a quarter of the months, on terror attacks across each of the three
suggesting it is sound.21 Figure 4c seems to sug- regimes. Recall that Hypotheses 2 and 3 pre-
gest some backlash from repressive-discrimi- dicted negative relationships between all
nate acts; however, that increase appears to be coefficients. Table 6b shows that only the
driven by relatively few months, as suggested Second Intifada provides clear support for
by the wide confidence bound. Hypothesis 2. Here the relationship between
To investigate whether relationships differ conciliatory actions and terror attacks is
across the three tactical regimes, we re-esti- unambiguously negative. Interestingly, this
mated Models 1, 2 and 3 for the First Intifada, finding is contrary to much conventional wis-
the Oslo Lull, and the Second Intifada months dom, in that the period when suicide missions
616 American Sociological Review 77(4)

a. Conciliatory-Discriminate b. Conciliatory-Indiscriminate
.376635 .165583

.691987 1.48244
0 8 0 26
Lagged Conciliatory Discriminate Lagged Conciliatory Indiscriminate

c. Repressive-Discriminate d. Repressive-Indiscriminate
1.21061 .6765

.356292 .212605
0 32 0 48
Lagged Repressive Discriminate Lagged Repressive Indiscriminate

Figure 4. Partial Predictions of the Past Months ActionsConciliatory or Repressive and


Discriminate or Indiscriminateon Terrorist Attacks in the Current Month
Note: Smoothing component of lagged actions has three degrees of freedom.

Table 6a. Negative Binomial Coefficients and (SE) for Government Actions in Model 1 for
Each Tactical Regime
All Months First Intifada Oslo Lull Second Intifada
(n = 191) (n = 61) (n = 76) (n = 52)
All Actions .005 .015 .020 .007
(.004) (.011) (.012) (.005)

Note: Control variables were included in the estimation but excluded from this table.
*p .05; **p .01 (two-tailed tests).

dominated the terrorist environment was also Hypothesis 2 is directly supported only dur-
the period when conciliatory tactics had the ing the Second Intifada and receives only
most powerful effect on reducing Palestinian qualified support during the other two
terrorism. regimes. Conciliatory actions seem to have
In contrast, during the First Intifada and reduced terrorist attacks only when Israel
Oslo Lull, small numbers of Israeli concilia- initiated larger numbers of conciliatory
tory actions led to increases in Palestinian actions during the previous month.
terror attacks, but the negative coefficient of Hypothesis 3 also finds no support in
the squared term shows that larger numbers of Table 6b. Repression did not decrease Pales-
conciliatory actions reduced terrorist attacks. tinian terrorism during any of the regimes. In
The GAM partial predictions, found in the fact, the positive and significant estimate dur-
online supplement, show this effect was more ing the Second Intifada suggests that during
dramatic during the Oslo Lull. In summary, that period, Israel experienced backlash when
Dugan and Chenoweth 617

Table 6b. Negative Binomial Coefficients and (SE) for Government Actions in Model 2 for
Each Tactical Regime
All Months First Intifada Oslo Lull Second Intifada
(n = 191) (n = 61) (n = 76) (n = 52)
Conciliatory .048 .280** .173**
.054*
(.031) (.100) (.064) (.023)
Conciliatory2 .003* .019* .007**
(.001) (.008) (.002)
Repressive .009 .002 .011 .021**
(.006) (.014) (.015) (.008)

Note: Control variables were included in the estimation but excluded from this table.
*p .05; **p .01 (two-tailed tests).

Table 6c. Negative Binomial Coefficients and (SE) for Government Actions in Model 3 for
Each Tactical Regime
All Months First Intifada Oslo Lull Second Inti-
(n = 191) (n = 61) (n = 76) fada (n = 52)
Conciliatory-Discriminate .006 .476* .063 .041
(.041) (.217) (.045) (.077)
(Conciliatory-Discriminate)2 .107*
(.051)

Conciliatory-Indiscriminate .051 .030 .172** .055*


(.036) (.042) (.059) (.024)
(Conciliatory-Indiscriminate)2 .003* .009**
(.002) (.002)
Repressive-Discriminate .016 .033 .490** .021
(.015) (.037) (.154) (.017)
(Repressive-Discriminate)2 .065**
(.017)
Repressive-Indiscriminate .005 .011 .009 .020*
(.008) (.019) (.013) (.010)

Note: Control variables were included in the estimation but excluded from this table.
*p .05; **p .01 (two-tailed tests).

they initiated repression, partially supporting Lull provide insight into results from Table 6b.
Hypothesis 4. During the First Intifada, conciliatory-dis-
To look for further evidence for Hypothe- criminate actions, where Israel gave conces-
ses 4 and 5, we turn to Table 6c, which lists sions to terrorists rather than to the Palestinian
estimates for conciliatory and repressive population, seem to have led to more terror.
actions partitioned by whether they were dis- Conciliatory-indiscriminate actions did not.
criminate or indiscriminate. Again, these This means that during that regime, there was
hypotheses are directly supported only in find- often more terror following the months when
ings from the Second Intifada. During that Israel offered concessions to terrorists. How-
regime, conciliatory-indiscriminate actions ever, during the Oslo Lull regime, terror was
were associated with fewer terror attacks, and more frequent after Israel offered a few con-
repressive-indiscriminate actions were associ- cessions to Palestinians in general, but the
ated with more. Findings related to concilia- effect reversed as the number of conciliatory
tory acts during the First Intifada and the Oslo actions accumulated (see GAM results in the
618 American Sociological Review 77(4)

online supplement for a graphical depiction of Conclusions


this relationship). During that regime, we also
find evidence that attacks were more frequent In this study, we found that governments can
after months when Israel targeted repressive often influence the number of terrorist attacks
actions toward terrorists. perpetrated against their people. Our argu-
These differences in findings across regimes ment expands beyond traditional strategies of
demonstrate some support for the notion that deterrence by incorporating actions that raise
effects of counterterrorism tactics may shift dur- the expected utility of refraining from terror-
ing different time periods, but not in the direc- ist behavior. This strategy paid off, as we
tions commonly expected. The Second Intifada, found that this often-overlooked dimension
often thought to be among the most extreme seemed to significantly reduce terrorist
periods of Palestinian violence, is the period attacks. Had we only measured effects of tra-
when conciliatory actions had the most direct ditional deterrence, the analysis would have
effect on reducing terrorist attacks. These find- appeared inconclusive at best, and the policy
ings suggest that instead of demonstrating irra- implications would have been misleading: we
tional behaviors, groups like Hamas and PIJ conclude that repressive actions by the Israeli
may be far more sensitive to opinions and government are unlikely to deter Palestinian
demands of the Palestinian constituency than is terrorism and may lead to a backlash of ter-
often suggestedperhaps even more so than rorist violence. This backlash effect is espe-
secular Palestinian groups. This implication cially prominent if repression is directed
echoes previous findings that emphasize the toward Palestinians in general.
hypersensitivity of religious organizations to Considering that the utility of terrorist vio-
public opinion (Iannaccone and Berman 2006). lence is much greater than the self-interest that
Moreover, our findings directly contest the typically motivates common criminals, this
claim that repression reduces terrorist attacks makes sense. Without additional conciliation,
by raising the cost of terrorist activitya the only value offered to terrorists and their
claim that receives no support in any of the constituencies for disengaging from terrorism is
tactical regimes examined. Instead, we find an absence of punishmentwhich is really just
that during the First Intifada and the Oslo the status quo. Had the status quo been suffi-
Lull, reductions in terrorist attacks followed cient to avert terrorism, no terrorism would
months with relatively high numbers of con- occur in the first place. Furthermore, terrorist
ciliatory actions, although small numbers of organizations are invested in ensuring that the
conciliatory actions were associated with a status quo remains unsatisfying to their constitu-
modest increase in terrorist activity. Perhaps encies (Malka 2007; Yaalon 2007). Thus, one
during these periods, adopting only a handful important contribution is demonstrating the
of conciliatory actions made Palestinians value of offering concessions to Palestinian
question Israels sincerity, thereby making the people that reward alternatives to violence.
Palestinian public less responsive to potential Constituent populations need evidence that
rewards of nonterror and emboldening terror- opposes terrorist propaganda rather than evi-
ist groups to push for more conciliation. dence that reinforces it.
Importantly, however, in months where con- Results also suggest that conciliatory
ciliatory actions were more forthcoming and actions must be sustained if they are to effec-
frequent, the credible benefits of nonterror- tively reduce terrorist violence. A few concil-
ism were clearer. During the Second Intifada, iatory efforts are unlikely to show an effect;
effects of conciliatory actions on reducing and in fact, they could lead to increased vio-
terrorism were more immediate, indicating, lence, as we saw in the First Intifada and the
perhaps, that during this particularly violent Oslo Lull. However, an ongoing and consist-
period, rewards for nonterrorism were ent campaign of conciliation can lead to a
accepted more swiftly. drop in terrorism as early as the following
Dugan and Chenoweth 619

month. This is good news, but Israeli repres- (Olds et al. 1986; Rich and Jacoby 1999;
sion can easily dismantle such progress, as Schweinhart 2005). Similarly, nurturing the
the backlash can be swiftespecially if the relationship between a government and a ter-
repression affects innocent Palestinians. rorist organizations constituency also requires
Moreover, according to the results, the consistency.
scope of concessions matters. Indiscriminate Having said this, we do not recommend
conciliatory actionssuch as making ges- that governments adopt purely conciliatory
tures toward peace talks, announcing plans to policies. It is likely still important to punish
withdraw troops, and criticizing abuses individuals who break the law, even if it pro-
against Palestiniansmay decrease subse- duces no obvious deterrence benefits. In fact,
quent terror attacks because they reward our analysis cannot speak to what would hap-
nonterrorist behavior. On the other hand, pen if Israel practiced only conciliatory
actions that single out particular terrorist behavior, because every month had at least
actors for conciliation (e.g., releasing prison- one repressive action by Israel. Instead, our
ers) benefit terrorism rather than nonterror- hope is that this research provides alternatives
ismand therefore, unsurprisingly, may not to solely focusing policy efforts on reducing
reduce terrorism. Indeed, whereas indiscrimi- the expected utility of bad behavior by also
nate conciliatory actions reduced terrorism considering the value of raising the expected
during two of the three periods (as concilia- utility of good behavior.
tory actions accumulated and appeared more
credible), discriminate conciliatory actions Authors Note
reduced terrorism only during the First Inti- Equal authorship is implied.
fada. This indicates that discriminate concil-
iatory actions typically yield fewer benefits in
counterterrorism than do more indiscriminate Funding
actions that reward nonterrorist behavior This material is based on work supported by the Science
and Technology directorate of the U.S. Department of
among the constituent population.
Homeland Security under Grant Award Number 2008-ST-
Combined, the full set of findings rein- 061-ST0004, made to the National Consortium for the
forces Braithwaites (2005) speculation that Study of Terrorism and Responses to Terrorism (START,
terrorist organizations may even benefit from http://www.start.umd.edu). The views and conclusions
repressive actions, and that they likely strate- contained in this document are those of the authors and
should not be interpreted as necessarily representing the
gically elicit repressive responses that will
official policies, either expressed or implied, of the U.S.
sabotage any goodwill that might be develop- Department of Homeland Security or START.
ing between Israel and the Palestinians. If the
Palestinian people begin to trust that Israel is
sincere about a mutually sustainable resolu- Acknowledgments
tion to the conflict, terrorist organizations will We gratefully acknowledge Phil Schrodt, Baris Kesgin, and
lose their base of support. Paradoxically, Pal- Matthias Heilke for helping us learn and use TABARI; and
estinian terrorists probably rely on Israels Richard Berk for his methodological advice. We also thank
our research assistants, Jeremy Berkowitz, Miranda Berry,
hawkish policies to preserve their longevity. Michael Distler, Robert Gambo, Max Livingston, Elizabeth
This article is the first to empirically dem- McClellan, Nicholas Miller, Valentina Postolache, Nicholas
onstrate the important role that conciliatory Quah, Joanna Seirup, Max Slater, Elsie Smith, Rachel
actions can play in reducing terrorist violence. Tecott, Elizabeth Weisman, Gregory Wong, and Nicholas
The importance of providing lawful alterna- Yulinsky, who assisted with data collection, coding, and lit-
erature review.
tives to violence is mirrored in the crimino-
logical literature. Indeed, many programs that
research shows positively affect the lives Notes
of those most at risk require consistent and 1. This notation was partially borrowed from Piliavin
long-term exposure to healthier alternatives and colleagues (1986).
620 American Sociological Review 77(4)

2. An important piece of scholarship by Sharvit and col- 16. We have considered the possibility that the increase in
leagues (2012) estimates effects of Israels coercive Israeli actions over time might be driven by an
and conciliatory policies on Palestinian terrorist vio- increase in reporting rather than an increase in actual
lence from 2000 through 2006. events. Although we cannot entirely resolve this
3. Like backlash, it is possible that conciliatory acts problem here, our strategy was to select the most
could lead to increased terrorism if they lead terrorists impartial and extensive coverage available (Reuters)
to perceive the state as weak, thereby emboldening and to carefully analyze each observation for inter-
terrorists to use still more violence (Iyengar and coder reliability. Furthermore, we examined the
Monten 2008). Some scholars also argue that concil- actions very closely to make sure that each is distinct
iatory tactics may create incentives for extremist from the others, ensuring that each action is counted
elements to sabotage efforts toward peace, depending only once in the GATE Database.
on the perceived strength of the negotiating partner 17. Adding additional lags provided no additional bene-
(Kydd and Walter 2002). fit; all statistical tests were null.
4. Some useful sources on the IsraeliPalestinian con- 18. Granger tests confirm the findings in the tables. When
flict include Bloom (2004), Gelvin (2005), Hafez government actions are statistically significant,
(2006), Jamal (2005), Pearlman (2008/2009), and Granger tests favor the more flexible model. When
Tessler (1995). government actions are statistically null, Granger
5. Although some scholars argue that the Second Inti- tests favor the model that excludes government
fada persisted until 2007 (Brym and Andersen 2011), actions.
we limit our study to 2004 because of data 19. We estimated eight as the maximum number of
availability. attacks using this equation: .048/(2 .003).
6. According to some arguments, conciliation may 20. This finding is unsurprising because the relatively
make the opposing government look weaker, there- few discriminate-conciliatory acts (e.g., release of
fore emboldening terrorists (Iyengar and Monten terrorist prisoners) suggest that results for Model 2
2008; Kydd and Walter 2002). were driven by indiscriminate acts.
7. Access to the data, statistical commands, and supple- 21. Twenty-eight percent of months had more than eight
mental analysis is available in an online supplement conciliatory-indiscriminate actions. We chose eight
(http://asr.sage pub.com/supplemental). actions because the maximum number of attacks fol-
8. These analyses used GTD data that were downloaded lowed 8.5 conciliatory-indiscriminate actions [.51/
on August 25, 2010. Because the GTD is continu- (2 = .003)].
ously being updated, newer versions of data could
produce slightly different results. Having said that,
we know of no systematic changes to Israeli or Pales- References
tinian related terrorist attacks for the years in the Andenaes, Johannes. 1971. The Moral or Educative
current analysis. Influence of Criminal Law. Journal of Social Issues
9. Before the earlier (1970 to 1997) GTD data were syn- 27:1731.
thesized with the later data, the GTD data since 1997 Araj, Bader. 2008. Harsh State Repression as a Cause of
were called GTD2. Since both collections have now Suicide Bombing: The Case of the Palestinian-Israeli
been synthesized, the data are called GTD as a whole Conflict. Studies in Conflict and Terrorism 31:284303.
(Dugan 2012). Beccaria, Cesare. [1764] 1983. On Crimes and Punish-
10. In our case, we used the CAMEO coding scheme. ment. Boston: Branden Books.
11. Schrodt (2006) estimates that TABARI codes 33 mil- Becker, Gary S. 1968. Crime and Punishment: An Eco-
lion times faster than the average human coder. nomic Approach. Journal of Political Economy
12. Note that the reference category is Oslo Lull and the 76:169217.
five months prior to the First Intifada. Furthermore, Becker, Howard S. 1963. Outsiders: Studies in the Soci-
because we lagged the dependent variable by four ology of Deviance. New York: The Free Press.
months, only one month prior to the First Intifada is Beitler, Ruth. 2004. The Path to Mass Rebellion: A Tale
included in the estimation. That one month would of Two Intifadas. New York: Lexington Books.
make an inadequate reference period, and there are no Benmelech, Efraim, Claude Berrebi, and Esteban Klor.
compelling theoretical reasons to distinguish it from 2010. Counter-Suicide Terrorism: Evidence from Home
the Oslo Lull. Demolitions. Working Paper No. 16493, National
13. We use the population in thousands for Israel in a Bureau of Economic Research, Cambridge, MA.
given year as the exposure measure for both the GAM Bentham, Jeremy. [1781] 1996. An Introduction to the
and negative binomial models. Principles of Morals and Legislation. Oxford: Oxford
14. A graphical representation of monthly counts is too University Press.
cluttered for the reader to easily interpret. Bloom, Mia. 2004. Outbidding, Market Share, and Pal-
15. Terrorist attacks are missing in the months during estinian Suicide Bombing. Political Science Quar-
1993, the period for which the GTD is missing data. terly 119:6188.
Dugan and Chenoweth 621

Braithwaite, John. 2005. Pre-empting Terrorism. Cur- Dugan, Laura, Julie Huang, Gary LaFree, and Clark
rent Issues in Criminal Justice 17:96114. McCauley. 2008. The Armenian Secret Army for the
Brodie, Bernard. 1959. Strategy in the Missile Age. Liberation of Armenia and the Justice Commandos of
Princeton, NJ: Princeton University Press. the Armenian Genocide. Dynamics of Asymmetric
Brophy-Baermann, Bryan and John A. C. Conybeare. Conflict 1:23149.
1994. Retaliating against Terrorism: Rational Dugan, Laura, Gary LaFree, and Alex R. Piquero. 2005.
Expectations and the Optimality of Rules versus Testing a Rational Choice Model of Airline Hijack-
Discretion. American Journal of Political Science ings. Criminology 43:10311066.
38:196210. Enders, Walter and Todd Sandler. 1993. The Effective-
Brym, Robert J. and Robert Andersen. 2011. Rational ness of Antiterrorism Policies: A Vector-Autore-
Choice and the Political Bases of Changing Israeli gression-Intervention Analysis. American Political
Counterinsurgency Strategy. British Journal of Soci- Science Review 87:82944.
ology 62:482503. Farrington, David P. 1977. The Effects of Public Label-
Brym, Robert J. and Bader Araj. 2006. Suicide Bombing ing. British Journal of Criminology 17:11225.
as Strategy and Interaction: The Case of the Second Flanigan, Shawn Teresa. 2010. For the Love of God:
Intifada. Social Forces 84:196986. NGOs and Religious Identity in a Violent World. Ster-
Bueno de Mesquita, Ethan. 2005. Conciliation, Conces- ling, VA: Kumarian Press.
sions, and Patterns of Terrorist Violence. Interna- Flanigan, Shawn Teresa and Mounah Abdel-Samad.
tional Organization 59:14576. 2009. Hezbollahs Social Jihad: Nonprofits as Resis-
Byman, Daniel. 2006. Do Targeted Killings Work? tance Organizations. Middle East Policy 16:12237.
Foreign Affairs (March/April):95107. Frey, Bruno. 2004. Dealing with Terrorism: Stick or Car-
Cauley, Jon and Eric Iksoon Im. 1988. Intervention rot? Northampton, MA: Edward Elgar.
Policy Analysis of Skyjackings and Other Terrorist Gelvin, James. 2005. The Israel-Palestine Conflict: One
Incidents. American Economic Review 78:2731. Hundred Years of War. New York: Cambridge Uni-
Chenoweth, Erica and Laura Dugan. 2011. Explor- versity Press.
ing Counterterrorism in the Middle East and North Gibbs, Jack P. 1975. Crime, Punishment, and Deterrence.
Africa: A New Data Set. Presented at the annual New York: Elsevier.
meeting of the American Political Science Associa- Goldkamp John S. and E. Rely Vilcica. 2008. Targeted
tion, September 2, Seattle, WA. Enforcement and Adverse System Side Effects: The
Clark, Robert P. 1990. Negotiating with ETA: Obstacles Generation of Fugitives in Philadelphia. Criminol-
to Peace in the Basque Country, 19751988. New ogy 46:371409.
York: Routledge. Goldstein, Joshua S. 1992. A Conflict-Cooperation
Clauset, Aaron, Lindsay Heger, Maxwell Young, and Scale for WEIS Event Data. Journal of Conflict
Kristian Skrede Gleditsch. 2010. The Strategic Cal- Resolution 36:36985.
culus of Terrorism: Substitution and Competition in Gottfredson, Denise C., Stephanie A. Gerstenbligh,
the Israel-Palestine Conflict. Cooperation and Con- David A. Soule, Shannon C. Womer, and Shaoli Lu.
flict 45:633. 2004. Do After School Programs Reduce Delin-
Cornell, Svante E. 2001. The Land of Many Crossroads: quency? Prevention Science 5:25366.
The Kurdish Question in Turkish Politics. Orbis Greene, William H. 2008. Econometric Analysis, 6th ed.
45:3146. Upper Saddle River, NJ: Pearson Prentice Hall.
Cornish, Derek B. and Ronald V. Clarke. 1987. Under- Hafez, Mohammed. 2006. Rationality, Culture, and
standing Crime Displacement: An Application of Structure in the Making of Suicide Bombers: A Pre-
Rational Choice Theory. Criminology 45:93347. liminary Theoretical Synthesis and Illustrative Case
Crenshaw, Martha. 2001. Theories of Terrorism: Instru- Study. Studies in Conflict and Terrorism 29:16585.
mental and Organizational Approaches. Pp. 1329 Hafez, Mohammed and Joseph Hatfield. 2006. Do Tar-
in Inside Terrorism Organizations, edited by D. C. geted Assassinations Work? A Multivariate Analysis
Rapoport. London: Frank Cass. of Israels Controversial Tactic during the Al-Aqsa
Cronin, Audrey Kurth. 2006. How al-Qaida Ends. Uprising. Studies in Conflict and Terrorism 29:359
International Security 31:748. 82.
Della Porta, Donatella. 1995. Left-Wing Terrorism in Hamm, Mark S. 2004. Apocalyptic Violence: The
Italy. Pp. 105159 in Terrorism in Context, edited by Seduction of Terrorist Subcultures. Theoretical
M. Crenshaw. State College, PA: Pennsylvania State Criminology 8:32339.
University Press. Hastie, T. J. and R. J. Tibshirani. 1990. Generalized Addi-
Dugan, Laura. 2012. The Making of the Global Terror- tive Models. New York: Chapman and Hall.
ism Database and Its Applicability to Studying the Iannaccone, Laurence R. and Eli Berman. 2006. Reli-
Lifecycles of Terrorist Organizations. Pp. 17598 gious Extremism: The Good, the Bad, and the Ugly.
in The Sage Handbook of Criminological Research Public Choice 128:109129.
Methods, edited by D. Gadd, S. Karstedt, and S. Iyengar, Radha and Jonathan Monten. 2008. Is there an
Messner. Los Angeles, CA: Sage. Emboldenment Effect? Evidence from the Insurgency
622 American Sociological Review 77(4)

in Iraq. Working Paper No. 13839, National Bureau on Political Violence in Northern Ireland: Compar-
of Economic Research, Cambridge, MA. ing Deterrence and Backlash Models. Criminology
Jaeger, David A. and M. Danielle Paserman. 2008. The 47:501530.
Cycle of Violence? An Empirical Analysis of Fatali- Landes, William. 1978. An Economic Study of U.S.
ties in the Palestinian-Israeli Conflict. American Aircraft Hijacking, 19611976. Journal of Law and
Economic Review 98:15911604. Economics 21:131.
Jamal, Amal. 2005. The Palestinian National Movement: Lapan, Harvey E. and Todd Sandler. 1988. To Bargain
Politics of Contention, 19672005. Bloomington: or Not To Bargain: That Is the Question. American
Indiana University Press Economic Review 78:1621.
Kadi, Leila S., ed. 1969. Basic Political Documents of Levitt, Steven. 1996. The Effect of Prison Population
the Armed Palestinian Resistance Movement. Beirut: Size on Crime Rates: Evidence from Prison Over-
Palestine Research Centre. crowding Litigation. Quarterly Journal of Econom-
Kalyvas, Stathis. 2006. The Logic of Violence in Civil ics 111:31952.
War. New York: Cambridge University Press. Levitt, Steven. 1997. Using Electoral Cycles in Police
Khawaja, Marwan. 1993. Repression and Popular Col- Hiring to Estimate the Effect of Police on Crime.
lective Action: Evidence from the West Bank. Socio- American Economic Review 87:27090.
logical Forum 8:4771. Lichbach, Mark Irving. 1987. Deterrence or Escalation?
Khawaja, Marwan. 1994. Resource Mobilization, Hard- The Puzzle of Aggregate Studies of Repression and
ship, and Popular Collective Action in the West Dissent. Journal of Conflict Resolution 31:26697.
Bank. Social Forces 73:191220. Lyall, Jason. 2009. Does Indiscriminate Violence Incite
King, Mary E. 2007. A Quiet Revolution: The First Pales- Insurgent Attacks? Evidence from Chechnya. Jour-
tinian Intifada and Nonviolent Resistance. New York: nal of Conflict Resolution 53:33162.
Nation Books. Lyall, Jason and Isaiah Wilson. 2009. Rage against the
Kocher, Matthew Adam and Stathis Kalyvas. 2007. How Machines: Explaining Outcomes in Counterinsur-
Free is Free Riding in Civil Wars? Violence, Insur- gency Wars. International Organization 63:67106.
gency, and the Collective Action Problem. World Malka, Haim. 2007. Hamas: Resistance and Transfor-
Politics 59:177216. mation of Palestinian Society. Pp. 98126 in Under-
Koopmans, Ruud. 1993. The Dynamics of Protest standing Islamic Charities, edited by J. B. Alterman
Waves: West Germany, 1965 to 1989. American and K. Von Hippel. Washington, DC: Center for Stra-
Sociological Review 58:63758. tegic and International Studies.
Kruglanski, Arie W., Xiaoyan Chen, Mark Dechesne, Maoz, Zeev. 2007. Evaluating Israels Strategy of Low-
Shira Fishman, and Edward Orehek. 2009. Fully Intensity Warfare, 19492006. Security Studies
Committed: Suicide Bombers Motivation and the 16:31949.
Quest for Personal Significance. Political Psychol- Marshall, Rachelle. 2009. Middle East Leaders Hold
ogy 30:33157. Open the Door to PeaceWill the U.S. and Israel
Kruglanski, Arie W., Michele Gelfand, and Rohan Respond? Washington Report on Middle East
Gunaratna. 2010. Detainee Deradicalization: A Affairs, Special Report, (January-February):79.
Challenge for Psychological Science. Observer Matsueda, Ross L., Derek A. Kreager, and David Huiz-
(http://www.psychologicalscience.org/observer/ inga. 2006. Deterring Delinquents: A Rational
getArticle.cfm?id=2607). Choice Model of Theft and Violence. American
Kuperman, Ranan D. 2007. Cycles of Violence: The Evo- Sociological Review 71:95122.
lution of the Israeli Decision Regime Governing the Mees, Ludger. 2003. Nationalism, Violence, and Democ-
Use of Limited Military Force. Lanham, MD: Lex- racy: The Basque Clash of Identities. New York: Pal-
ington. grave Macmillian Ltd.
Kydd, Andrew and Barbara F. Walter. 2002. Sabotaging Moore, William H. 1998. Repression and Dissent: Sub-
the Peace: The Politics of Extremist Violence. Inter- stitution, Context and Timing. American Journal of
national Organization 56:26396. Political Science 42:85173.
LaFree, Gary and Gary Ackerman. 2009. The Empiri- Nagin, Daniel S. 1998. Criminal Deterrence Research
cal Study of Terrorism: Social and Legal Research. at the Outset of the Twenty-first Century. Crime and
Annual Review of Law and Social Science 5:34774. Justice 23:142.
LaFree, Gary and Laura Dugan. 2004. How Does Study- Nagin, Daniel S. and Raymond Paternoster. 1993.
ing Terrorism Compare to Studying Crime? Pp. 5374 Enduring Individual Differences and Rational
in Terrorism and Counter-Terrorism: Criminological Choice Theories of Crime. Criminology 27:46796.
Perspectives, edited by M. DeFlem. New York: Elsevier. Neumann, Peter. 2007. Negotiating with Terrorists.
LaFree, Gary and Laura Dugan. 2007. Introducing the Foreign Affairs 86:12838.
Global Terrorism Database. Terrorism and Political Newman, Sanford A., James Allen Fox, Edward A.
Violence 19:181204. Flynn, and William Christeson. 2000. Americas After
LaFree, Gary, Laura Dugan, and Raven Korte. 2009. School Choice: The Prime Time for Juvenile Crime,
The Impact of British Counterterrorist Strategies or Youth Enrichment and Achievement, a research
Dugan and Chenoweth 623

report. Washington, DC: Fight Crime Invest in Kids. Schweinhart, Lawrence J. 2005. The High/Scope Perry
Olds, David L., Charles R. Henderson Jr., Robert Tatel- Preschool Study through Age 40. Ypsilanti, MI: High/
baum, and Robert Chamberlin. 1986. Improving the Scope Press.
Delivery of Prenatal Care and Outcomes of Preg- Sharvit, Keren, Arie W. Kruglanski, Mo Wang, Xiaoyan
nancy: A Randomized Trial of Nurse Home Visita- Chen, Lauren Minacapelli, Boaz Ganor, and Eitan
tion. Pediatrics 77:1628. Azani. 2012. The Effects of Israeli Coercive and
Paternoster, Raymond. 1987. The Deterrent Effect of Conciliatory Policies on Palestinian Terrorist Activ-
the Perceived Certainty and Severity of Punishment: ity: 20002006. Department of Psychology, Univer-
A Review of the Evidence and Issues. Justice Quar- sity of Haifa, Israel. Unpublished manuscript.
terly 4:173217. Shellman, Stephen M. 2008. Coding Disaggregated
Pearlman, Wendy. 2008/2009. Spoiling from the Inside Intrastate Conflict: Machine Processing the Behavior
Out: Internal Political Contestation and the Middle of Substate Actors over Time and Space. Political
East Peace Process. International Security 33:79 Analysis 16:46477.
109. Shellman, Stephen M., Claire Hatfield, and Maggie J.
Peroff, Kathleen and Christopher Hewitt. 1980. Rioting Mills. 2010. Disaggregating Actors in Intranational
in Northern Ireland: The Effects of Different Poli- Conflict. Journal of Peace Research 47:8390.
cies. Journal of Conflict Resolution 24:593612. Sherman, Lawrence W. 1990. Police Crackdowns: Ini-
Piazza, James A. and James Igoe Walsh. 2009. Trans- tial and Residual Deterrence. Crime and Justice
national Terrorism and Human Rights. International 12:148.
Studies Quarterly 53:12548. Sherman, Lawrence W. 1993. Defiance, Deterrence,
Piliavin, Irving, Rosemary Gartner, Craig Thorton, and and Irrelevance: A Theory of the Criminal Sanc-
Ross L. Matsueda. 1986. Crime, Deterrence, and tion. Journal of Research in Crime and Delinquency
Rational Choice. American Sociological Review 30:44573.
51:101119. Tessler, Mark. 1995. A History of the Israeli-Palestinian
Plaw, Avery. 2008. Targeting Terrorists: A License to Conflict. Bloomington: Indiana University Press.
Kill? London: Ashgate. Testas, Abdelaziz. 2004. Determinants of Terrorism in
Rasler, Karen. 1996. Concessions, Repression, and the Middle East. Terrorism and Political Violence
Political Protest in the Iranian Revolution. American 16:25373.
Sociological Review 61:13252. Tibbetts, Stephen G. and Chris L. Gibson. 2002. Indi-
Rasler, Karen. 2000. Shock, Expectancy Revision, and vidual Propensities and Rational Decision-Making:
De-escalation of Protracted Conflicts. Journal of Recent Findings and Promising Approaches. Pp.
Peace Research 37:699720. 314 in Rational Choice and Criminal Behavior
Rich, Thomas F. and Kristen Jacoby. 1999. National Recent Research and Future Challenges, edited by A.
Evaluation of Weed and Seed Hartford Case Study, R. Piquero and S. G. Tibbetts. New York: Routledge.
research report. Washington, DC: Office of Justice Tyler, Tom R. 2006. Psychological Perspectives on
Programs. Legitimacy and Legitimation. Annual Review of
Rigby, Andrew. 1991. Living the Intifada. London: Zed. Psychology 57:375400.
Ross, Laurence H. 1982. Deterring the Drunk Driver: White, Robert. 1989. From Peaceful Protest to Guer-
Legal Policy and Social Control. Lexington, MA: rilla War: Micromobilization of the Provisional Irish
D.C. Health. Republican Army. American Journal of Sociology
Schrodt, Philip A. 2001. Automated Coding of Interna- 94:12771302.
tional Event Data Using Sparse Parsing Techniques. Xiang, Dong. 2001. Fitting Generalized Additive Mod-
Department of Political Science, University of Kan- els with the GAM Procedure. SUGI Proceedings.
sas, Lawrence, KS. Unpublished manuscript. Cary, NC: SAS Institute, Inc.
Schrodt, Philip A. 2006. Twenty Years of the Kansas Yaalon, Moshe. 2007. Lessons from the Palestinian
Event Data System Project. Department of Politi- War Against Israel. Washington, DC: The Wash-
cal Science, University of Kansas, Lawrence, KS. ington Institute for Near East Policy.
Unpublished manuscript. Zussman, Asaf and Noam Zussman. 2006. Assassi-
Schrodt, Philip A., Shannon G. Davis, and Judith L. Wed- nations: Evaluating the Effectiveness of an Israeli
dle. 1994. Political Science: KEDS: A Program for Counterterrorism Policy Using Stock Market Data.
Machine Coding Events Data. Social Science Com- Journal of Economic Perspectives 20:193206.
puter Review 12:56188.
Schrodt, Philip A. and Deborah J. Gerner. 1994. Validity Laura Dugan is an Associate Professor in the Depart-
Assessment of a Machine-Coded Event Data Set for ment of Criminology and Criminal Justice at the
the Middle East, 19821992. American Journal of University of Maryland. She is a lead investigator at the
Political Science 38:82554. National Center for the Study of Terrorism and the
Schwartz, Richard D. and Jerome H. Skolnick. 1962. Response to Terrorism (START). Her research examines
Two Studies of Legal Stigma. Social Problems the consequences of violence and the efficacy of violence
10:13342. prevention/intervention policy and practice. She also
624 American Sociological Review 77(4)

designs methodological strategies to overcome data limi- START and an Associate Senior Researcher at the
tations inherent in the social sciences. She has published Peace Research Institute of Oslo (PRIO). A specialist in
scholarly articles in Criminology, Journal of Quantitative political violence and its alternatives, she is the co-
Criminology, Crime and Justice, Law and Society, Ter- author of Why Civil Resistance Works: The Strategic
rorism and Political Violence, and others. Logic of Nonviolent Conflict and co-editor of Rethink-
ing Violence: States and Non-State Actors in Conflict.
Erica Chenoweth is an Assistant Professor at the Josef She has published scholarly articles in the Journal of
Korbel School of International Studies at the University Politics, International Security, Political Research
of Denver, where she directs the Program on Terrorism Quarterly, and the Review of Policy Research, among
and Insurgency Research. She is a lead investigator at other venues.