Benjamin Acosta a & Steven J. Childs b a Department of Politics & Policy, Claremont Graduate University,

Transcription

1 This article was downloaded by: [American University of Beirut] On: 28 June 2014, At: 12:59 Publisher: Routledge Informa Ltd Registered in England and Wales Registered Number: Registered office: Mortimer House, Mortimer Street, London W1T 3JH, UK Studies in Conflict & Terrorism Publication details, including instructions for authors and subscription information: Illuminating the Global Suicide-Attack Network Benjamin Acosta a & Steven J. Childs b a Department of Politics & Policy, Claremont Graduate University, Claremont, CA, USA b School of International Relations, University of Southern California, Los Angeles, CA, USA Published online: 14 Dec To cite this article: Benjamin Acosta & Steven J. Childs (2013) Illuminating the Global Suicide-Attack Network, Studies in Conflict & Terrorism, 36:1, 49-76, DOI: / X To link to this article: PLEASE SCROLL DOWN FOR ARTICLE Taylor & Francis makes every effort to ensure the accuracy of all the information (the Content ) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content. This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at

2 Studies in Conflict & Terrorism, 36:49 76,2013 Copyright Taylor & Francis Group, LLC ISSN: X print / online DOI: / X Illuminating the Global Suicide-Attack Network BENJAMIN ACOSTA Department of Politics & Policy Claremont Graduate University Claremont, CA, USA STEVEN J. CHILDS School of International Relations University of Southern California Los Angeles, CA, USA This article examines the contemporary phenomenon of suicide attacks by fusing network analysis and time-series econometrics. We find that a global network of militant organizations drives the reproduction of the suicide-attack phenomenon, and brokers within the network mark the primary perpetrators and diffusers of the tactic. The introduction of a fourth level of analysis of political violence demonstrates that network connections between organizations form a system that perpetuates suicide attacks. An organization-level analysis reveals that ideological congruence facilitates the establishment of network connections. As exemplified by the wide range of employers and targets, and moreover by the generation of an autogamous function, contemporary suicide attacks represent a unique sociopolitical phenomenon. Accordingly, organizations that use the tactic warrant a distinct classification. Over the last three decades, suicide attacks have come to represent a unique sociopolitical phenomenon. 1 Occurring in contexts of insurgency, guerrilla warfare, terrorism, interstate and civil war, as well as individual political and legal protest, suicide attacks have gradually become a mainstay of contemporary political violence. From 1980 to 2012, 3,658 suicide attacks killed 33,556 people. 2 More than 90 percent of these attacks have taken place in the last ten years, and 54 percent in the last five years. The use of suicide attacks proliferates among militant organizations that exchange justifications, strategies, and logistics, and perhaps more significantly concepts for mobilizing individuals to undertake such attacks (Moghadam 2008). Consequently, the contemporary suicide-attack phenomenon evolves according to aglobalnetwork of militant organizations that employs the tactic. 3 Within its 32-year Received 11 February 2012; accepted 25 July The idea for this article emerged in January 2010 during a seminar on social network analysis at Claremont Graduate University. The authors would like to thank Mark Abdollahian, Rita Chiang, the participants of the seminar, Melissa Rogers, Piotr Zagorowski, and the Claremont Linking Identities Organizations Networks (LION) Initiative. Address correspondence to Benjamin Acosta, Department of Politics & Policy, Claremont Graduate University, P.O. Box 1863, Claremont, CA 91711, USA. benjamin.acosta@ cgu.edu 49

3 50 B. Acosta and S. J. Childs development, this suicide-attack network has formed a system that perpetuates the use of suicide-attacks. The annual number of suicide attacks largely relies on the level of connectivity within the network-system. As such, network brokers, or the chief organizational operators of the tactic, lead a transnational endeavor to proliferate the suicide-attack phenomenon. By increasing the rate and quality of information exchange, brokers within the suicide-attack network have made suicide attacks commonplace in contemporary asymmetric warfare. Notably, the expansion of the network regularly runs along ideological lines. A shared worldview and values-based objectives (i.e., ideological congruence ) aids considerably in overcoming traditional problems of collective action between multiple actors. Further, just as an individual seeks to increase his or her legacy by carrying out a martyrdom operation /suicide attack (Berko and Erez 2005), so does an organization by diffusing the modus operandi to organizations with congruent ideological platforms. Thus, the establishment of network connections, or ties, frequently transpires in accordance with ideological similitude. Importantly, the dynamic and evolutionary character of the suicide-attack phenomenon has fostered an autogamous function. The use of suicide attacks consistently attracts new organizations to adopt the tactic, adapting the modus operandi to fit and abide by their own ideological goals and targeting constraints, and reestablishing a fresh set of allures that in turn attract an even newer cadre of employers. As a result, the general use of suicide attacks engenders subsequent employment of the modus operandi. Like a trendy fashion, the use of the tactic itself repeatedly spawns new adherents. The generation of this self-reproducing function, as attested to by the wide range of disparate employers and targets of suicide attacks, demonstrates that the suicide-attack phenomenon represents more than a mere subset of terrorism. 4 Accordingly, organizations that employ suicide attacks warrant a distinct classification implying a particular set of guidelines to counter the phenomenon. Approaches to the Study of Suicide Attacks One can divide the literature on suicide attacks into three paradigms. Each conceptual prism addresses the two central issues of the contemporary suicide-attack phenomenon: (1) the motivations that support the use of the tactic and (2) the ways in which the tactic diffuses. The first two paradigms work from the assumptions of the rational actor model, treating suicide attacks as either a reactionary effort to coerce foreign enemies or the product of intra-group domestic competition. The third paradigm synthesizes multiple levels of analysis and allots a causal role for ideology. The first two paradigms concentrate on the tangible goals of organizations that employ suicide attacks, whereas the latter additionally outlines intangible aspirations. Simply, the three paradigms have different audience foci, viewing suicide attacks as respective games: coercion of an external audience, competition with an internal audience, and ideological articulation for a universal audience. The Coercion Paradigm and Occupation Thesis The first paradigm emphasizes the coercive element in launching suicide attacks. From this perspective, suicide attacks generally mark a move in a signaling game, wherein organizations use violence in efforts to coerce an enemy to change certain policy (Sprinzak 2000; Hoffman and McCormick 2004). More specifically, the paradigm presumes that suicide attacks represent a response to foreign occupation by armed forces, and that the

4 Illuminating the Global Suicide-Attack Network 51 causal mechanics of casualties reduces occupier resolve (Pape 2003, 2005; Gambetta 2006; Wade and Reiter 2007; Pape and Feldman 2010). Within the coercion paradigm, scholars largely agree that foreign military occupation causes suicide attacks. That is, organizations rationally match their limited means in the pursuit of ending the occupation game. Accordingly, this paradigm views the diffusion of suicide attacks resulting from a strategic logic, wherein terrorist groups [learn] from each other s coercive successes (Pape 2005, 73). The Competition Paradigm and Outbidding Thesis As an outgrowth of the ethnic outbidding literature, the second paradigm treats suicide attacks as the product of domestic competition. The use of the tactic then derives from internal competition between rival organizations for influence over a shared constituency (Bloom 2004, 2005; Gupta and Mundra 2005; Hoffman 2006; Kydd and Walter 2006). Others find that organizations employ suicide attacks in order to compete domestically by disrupting peace negotiations with enemy forces (Kydd and Walter 2002). Similar to the coercion perspective, the competition paradigm views the diffusion of suicide attacks as arational,orstrategic,learningprocess. Thiskindofdiffusiontakesvariousforms:a spillover of violence to neighboring areas within the same country or to other countries.... In some instances, insurgent factions have been physically trained by other organizations.... On other occasions, factions observe the successful operations of groups from afar (Bloom 2005, 122). The Ideology Paradigm and Contextualized Theses The third paradigm allocates a role for ideology and culture (Hafez 2006a; Moghadam 2008/2009; Acosta 2010a) within a multilevel analytical framework that makes a distinction between individual, organizational, and environmentally induced motivations for suicide attacks (Moghadam 2007). Generally, most accounts [now] focus on the interaction of individuals, organizations, and societies (Crenshaw 2007, 141). A variety of studies work from such a paradigm, although authors frequently emphasize one of the three levels of analysis to spearhead causality whether individual primacy (Berko and Erez 2005), organizational (Merari et al. 2010), or environmental (Hafez 2007a; Hafez 2007b). Others maintain the interdependency of each level (Pedahzur 2005; Hafez 2006a), or that context can dictate the centrality one motivational level might take over the others at a given time and/or place (Fine 2008). The third paradigm emphasizes the matching of sociocultural means to ideological ends, refuting many rational-actor-based studies from the first two paradigms (Hafez 2006b; Moghadam 2006; Hafez 2007b). Unlike the first two paradigms that attempt to address the spread of the tactic as the dispersal of an impassive strategy, the third paradigm finds the vast diffusion of suicide attacks resulting from a parallel spread of an alluring type of Islamic martyrdom (Moghadam 2008; Moghadam 2008/2009; Acosta 2010b). Despite the prevalence of the coercion and competition paradigms, key puzzles abound pertaining to their explanations of the suicide-attack phenomenon. 5 Alimitationofthecoercion paradigm rests in its fixation on externally induced motivation as the causal factor for suicide attacks, given that such motivations would also thrive in numerous instances where suicide attacks do not occur. Similarly, if the internal competition of domestic organizations fuels the phenomenon, then one questions why competing domestic organizations do not harness the tactic in more locales. In short, an abundance of foreign intervention and

5 52 B. Acosta and S. J. Childs domestic organizational rivalries persist across the globe without a concomitant manifestation of suicide attacks. Moreover, as Brym and Araj note, [on] close inspection, monocausal explanations of suicide [attacks] are bound to fail (2008, 500). For example, the increasing use of suicide attacks against domestic rather than foreign targets (Fine 2008) often carried out by organizations predominately made up of foreign fighters as with the case of al-tawhid wal-jihad and later Al Qaeda in Iraq (Rosen 2006; Hafez 2007a; Moghadam 2008/2009) renders both the coercion and outbidding schools of thought obsolete. Adding a Fourth Level of Analysis and Predictive Dimension Advancing research on suicide attacks requires a novel and scientific approach that takes into account the nuanced and multilevel analytical approach of the third paradigm. Importantly, scholars have recently reconsidered the role of social network analysis in explaining suicide attacks and political violence (Jordan, Horsburgh, and Manas 2008; Horowitz 2010; Pedahzur and Perliger 2011). 6 Previously, studies have shown that familial and clan networks can supersede the motivations of vertically structured organizations (Pedahzur and Perliger 2006). The finding that networks maintain brokers rather than leaders (Sageman 2004) stands in contrast to hierarchical and strategic choice explanations insofar as the prime operative considerations rest in contextualized network dynamics (e.g., tribal prestige and power) rather than grand strategic assessments (Pedahzur and Perliger 2006). Studies have also described the aggregated pathways of diffusion for suicide attacks (Horowitz 2010). Nevertheless, scholarship has yet to use network analysis to holistically address the larger system that perpetuates the suicide-attack phenomenon. 7 By mapping the dynamic shifts in the connections between organizations, a network approach can add a heretofore-overlooked system-level analysis of the suicide-attack phenomenon. Viewing the suicide-attack phenomenon dynamically and across time can reveal the underlying relations and connections that systemically perpetuate the modus operandi. Accordingly, this study presents network analysis as a tool for shedding light on the system that reproduces the phenomenon adding a complementary fourth dimension to Moghadam s multilevel analytical framework of the individual, organization, and sociocultural environment (2007a). 8 We also seek to fuse a predictive element to the network analysis. Pedahzur and Perliger note the considerable descriptive and explanatory potential of social network analysis in the study of political [violence] (2011, 45). However, if coupled with timeseries econometrics, network analysis can also enable a rich predictive capability. Where a network approach provides for adding a fourth level of analysis, time-series econometrics fosters predictive analysis. A system-level network analysis can translate the qualitative narrative of the suicide-attack phenomenon s development and evolution into quantitative data, which then sets up the testing of hypotheses. Together, the approaches make for a unique methodology that adds a robust scientific element to the nuanced scholarship of the third paradigm. Network Analysis and the Narrative of the Suicide-Attack Phenomenon Data 9 Documenting known suicide attacks from 1980 to 2012, we have compiled and utilize anewandmoreconcisedatasetthanthosereliedoninpreviousstudies. 10 Essential to advancing research on suicide attacks, network metrics allow for scholarly treatment of the

6 Illuminating the Global Suicide-Attack Network 53 network of militant organizations that contemporaneously employ suicide attacks or the suicide-attack network as an emergent whole. That is, the data abets in the analysis of the suicide-attack phenomenon as a system (i.e., the network of militant organizations that employ the tactic) as well as its constituent parts (i.e., the individual organizations that recruit and sponsor the martyrs that carry out such attacks). Network Measures and Quantitative Variables Integral to the veracity of the discussion lays the definition of terms, measures, and variables. The term node stands for a single entity within a social network. In this study, each node signifies a militant organization or other type of entity that conducts suicide attacks. Ties represent a known physical relationship between agents from different but connected organizations. 11 Brokerage determines the number of times that a node serves as the shortest path between two other nodes (Freeman 1978/1979). In effect, brokerage captures the extent to which a node acts as a bridge linking other nodes. In addition to the network measures of ties and brokerage, this study analyzes a traditional quantitative variable. Attacks, or frequency of suicide attacks, stands for the annual number of suicide attacks, either overall which refers to the aggregated output of the entire suicide-attack network, or for individual organizations. Visual Attributes Amainadvantageofnetworkanalysisrestsinitsabilitytoqualitativelyconsiderthehistorical evolution of the network, and in turn exhibit a time series of the network visualizations and set of common attributes among nodes. 12 As noted above, the suicide-attack network refers to the continuously changing set of relations between organizations, individuals, and other entities that conduct suicide attacks. To determine the effect that ties between organizations might have on the collective output of the suicide-attack network first entails reviewing a qualitative and visual narrative. As such, we have identified physical ties between organizations, constructing the network in annual intervals covering 1980 to In the network visualizations below, lines connecting one node to another represent ties. 14 If a node does not participate in any suicide attack in a specific year, it is considered inactive and removed from that year s network visual. Within the annual visualizations, the size of the node reflects the amount of an organization s attacks in the given year relative to all other organizations. The shapes of the nodes denote organizational ideology: a circle indicates an ideology based on political Islam, a square signifies an ethno-nationalist or separatist ideology, a diamond represents a hybrid between an ideology based on political Islam and nationalism, a down-triangle denotes a national-socialist ideology, an up-triangle stands for a Marxist-Leninist or Maoist ideology, and a plus symbol refers to an unidentified or other ideology (see Figure 1). White-colored nodes mark organizations that use suicide attacks only against combatants or political targets (i.e., insurgent or guerrilla organizations). Black-colored nodes stand for organizations that only target civilians (i.e., terrorist organizations), and gray-colored nodes mark organizations that target both civilians and non-civilians. The visual attributes of size, shape, and shade help expose which types of ideological organizations carry out suicide attacks, which more commonly connect to one another, how often an organization uses suicide attacks relative to other organizations, and whether organizations solely target civilians. Per our primary theoretical assertions, the network visuals show (1) connections, (2) ideological type of organizations, and (3) targeting preferences.

7 54 B. Acosta and S. J. Childs Figure 1. Key. Development of the Suicide-Attack Phenomenon and Global Network 15 The act of detonating oneself to kill others has passed through many organizational hands and evolved in various ways over the last three decades. In 1980, near the beginning of the Iran Iraq war, amid a skirmish, a young Iranian named Mohammed Hussein Fahmideh gathered hand-grenades and rushed an Iraqi tank blowing himself up and disabling the tank. Fahmideh s story spread across Iran as an example of true Shi a martyrdom and heroism. Shortly thereafter, Iran s Islamic Revolutionary Guard (IRG) Corps recruited thousands of young men into the Bassidj organization under the promise that they too would attain martyrdom (Betty and Toolis 2006). The IRG deployed the Bassidj to clear Iraqi minefields throughout the conduct of the war. By walking into certain death, these mine jumpers ultimately helped to turn back the mechanized advancements of the Iraqi military (Reuter 2004). Recognizing the utility of the Bassidj mine jumpers, the IRG sought to export the selfannihilating Shi a martyr to the Shi a effort in Lebanon to combat U.S., French, Israeli, Sunni, and Christian-Lebanese forces. By 1981, the Shi a organization Hezb al-da awa

8 Illuminating the Global Suicide-Attack Network 55 Figure 2. Suicide-attack network: 1980 to al-islamiyya launched a suicide attack in Lebanon, followed by Hezbollah in A year after its creation, Hezbollah gained international notoriety with its suicide attacks on Western targets, including the attack on 18 April 1983 when a suicide bomber drove a van into the U.S. embassy in Beirut, killing 63 people and wounding 120. Hezbollah followed up with the 23 October 1983 attack on the U.S. Marine and French paratrooper barracks in Beirut, which resulted in 304 deaths. 17 Consequently, the United States withdrew its Marines from Lebanon on 26 February 1984, lending a foundation for the perception that suicide attacks hold a strategic capability. In 1984, another Lebanese-Shi a organization, Amal, joined the network (see Figure 2). In 1985, the burgeoning suicide-attack network began to show the fruits of its labor. From 1984 to 1985, ties between organizations in the suicide-attack network jumped from two to 28, and overall suicide attacks increased from four to 31. With its high-casualty suicide attacks, Hezbollah inspired pro-syrian Lebanese organizations to recruit and launch their own suicide bombers, expanding the network beyond the Shi a paradigm (Shay 2004). By 1985, the tactic had spread to the Syrian Social-Nationalist Party (SSNP), the Syrian Ba ath Organization (SBO), and their umbrella organization Jammoul, 18 al-gama a al- Arabiyya Misr, and al-hezb ash-shu aye al-lubnani, and the Nasserist al-mourabitoun, all of which operated within Lebanon (see Figure 2). Between the years of 1986 and 1992, the suicide-attack network dissipated. The tactic remained mostly at the margins of asymmetrical warfare, with an average of only 3.3 attacks per year. The number of ties in the network went from 12 in 1986 to zero in 1987, 1988, 1989, 1991, and This decline derived in large part from Hezbollah s recognition of the negative social potentialities of mass istishhad (deliberate Islamic martyrdom). In

9 56 B. Acosta and S. J. Childs Figure 3. Suicide-attack network: 1986 to , Hezbollah s spiritual advisor Mohammed Hussein Fadlallah issued a fatwa (Islamic degree) authorizing the use of suicide attacks only on special occasions, in order to prevent exaggerated use [by] over-zealous youth (Israeli 2002, 30). The network, however, did gain some new members. In 1987, the Liberation Tigers of Tamil Eelam (LTTE) became the first suicide-attack organization to operate outside of the Islamic world in its guerrilla war against the Sri Lankan government. In the same year, Jabhat al-tahrir al-lubnaniyye also launched its first attack, as did the Popular Front for the Liberation of Palestine (PFLP) in 1990 (see Figure 3). In 1993, Harakat al-muqawama al-islamiyya (Hamas) and al-jihad al-islami fi Filastin or Palestinian Islamic Jihad (PIJ) evolved the suicide-attack phenomenon by altering targeting preference, as well as changing the collective character of the suicide-attack network. Prior to the Palestinian adoption of suicide attacks, organizations had used the modus operandi primarily against military forces, keeping the tactic mostly within the domain of guerrilla warfare and insurgency. 19 Palestinian organizations transitioned suicide attacks for use against civilian targets. In late 1992, as a response to the First Intifada, Israel s Rabin government deported the uprising s leaders and other Palestinian activists, including both Hamas and PIJ members, to Southern Lebanon (Atran 2003). 20 Under the auspices of Hezbollah, these Hamas and PIJ members learned the functional elements of suicide attacks (Hoffman 2006), but more importantly matured their conceptual understanding of the istishhadi,ordeliberateislamicmartyr,forpurposesofmobilizingindividualstocarry out suicide attacks. Aligning with the return of many of the 418 deportees by the end of the year, Hamas and PIJ conducted their first successful suicide attacks against Israel in 1993 (Stern 2003; Acosta 2010a). In the process, the Palestinian organizations reinvigorated the

10 Illuminating the Global Suicide-Attack Network 57 suicide-attack network. From the West Bank and Gaza, the tactic transmitted further into the Sunni realm after the Egyptian Islamic Jihad (EIJ) carried out two suicide attacks in Despite the Sunni spark and Hezbollah s use of suicide attacks in Argentina under the name Ansar Allah, the mid-1990s saw another devolution of the suicide-attack network (see Figure 3). Although the suicide-attack network grew mildly during the second half of the 1990s, many organizations that adopted the tactic largely failed to connect to the wider network. In 1995, a variety of organizations began using suicide attacks, including the Egyptian al-gama a Islamiyya, the Groupe Islamique Armé (GIA),andtheSikh-nationalistBabbar Khalsa International. Hezbollah sponsored a Gulf-based group s suicide attack in Saudi Arabia (Shay 2005). Nine organizations launched suicide attacks but only six ties existed between the organizations a development that set the tone for the remainder of the decade. In 1996, Partiya Karkerên Kurdistan (PKK) and the Pakistani Harakat ul-ansar began using suicide attacks, and network ties increased to eight. From 1997 to 1998, ties increased from two to 12 and overall attacks went from 18 to 32. In August of 1998, Al Qaeda entered the network with its attacks on U.S. embassies in Africa. In 1999, ties dropped to four and overall suicide attacks decreased to 26. Pakistani Lashkar-e Toiba (LeT) and the Turkish Worker-Peasant Liberation Army (TIKKO) began employing suicide attacks (see Figure 4). In 2000, the suicide-attack network expanded as two cliques began to form one centered on Hamas and the other on Al Qaeda, both Sunni organizations with Muslim Brotherhood origins. The year 2000 witnessed Palestinian Fatah, the Pakistani Jaish-e Mohammad, and the Chechen Special Purpose Islamic Regiment (SPIR) carry out successful suicide attacks. Nevertheless, not until Sheik Yousef al-qaradawi issued a fatwa in 2001 in support of Palestinian suicide bombers did the Sunni world widely embrace suicide terrorism (Acosta 2010b). Qaradawi s affiliation with the Muslim Brotherhood placed an ideological link of legitimacy within the context of Sunni Islam. From 2000 to 2001, ties within the network more than doubled from 12 to 28, and the overall number of suicide attacks went from 50 to 80. In 2001, Indonesian Jema ah Islamiyyah (JI), the Islamic Movement of Uzbekistan (IMU), the Taliban in Afghanistan, the Turkish Revolutionary People s Liberation Party-Front (DHKP-C), and the Bangladeshi Islami Chhatra Shibir undertook their first suicide attacks (see Figure 5). In 2002, ties slid to 22 and overall suicide attacks increased to 90. Amid the height of the Second Intifada, Palestinian organizations stepped up suicide attacks against Israel. The Democratic Front for the Liberation of Palestine (DFLP) joined fellow Palestinian organizations Hamas, PIJ, Fatah, and the PFLP in launching suicide attacks. The year also brought the further Islamicization of the Chechen effort against the Russian Federation, as demonstrated by the suicide attacks of the newly established Riyadh as-saliheyn Martyrs Brigade (RSMB). The period saw the suicide-attack phenomenon begin its vast diffusion across the Sunni world. The individuals that carried out the attacks were accepted as shuhada (martyrs), bringing an allure of heroism and fame to those who became suicide bombers (Moghadam 2008). Capitalizing on Palestinian advancements within a freshly evolved Sunni martyrdom, Al Qaeda and others within the international Salafi-jihadi movement began to mobilize thousands of Sunni martyrs/suicide bombers for their efforts against Western and secular-arab nation-states, and Shi a and other Muslim minorities (Acosta 2010a). In 2003, ties jumped to 50 and the overall amount of suicide attacks increased to 126. With the refocusing of jihadi efforts to Iraq, the suicide-attack network began shifting from apalestinian-drivennetworktoonedominatedbyalqaeda.still,themajorityofties

11 58 B. Acosta and S. J. Childs Figure 4. Suicide-attack network: 1996 to between organizations continued to emerge along ideological lines, specifically rooted in political Islam. In Iraq, Abu Musab al-zarqawi s al-tawhid wal-jihad, Ansar al-sunnah, Ansar al-islam, Ansar Allah, former-ba athists, in Pakistan Lashkar-e Jhangvi (LeJ), and in North Africa as-sirat al-moustaquim began conducting suicide attacks. By this point, the evidence had become increasingly clear that the widespread employment of suicide attacks did not spring from strategic capability but rather a rapidly spreading culture of martyrdom (Moghadam 2006; Acosta 2010b). In 2004, the number of ties increased drastically to 70, as did the overall amount of suicide attacks to 227. Zarqawi pledged a bayah (loyalty oath) to Osama bin Laden and shortly thereafter Al Qaeda gave Zarqawi permission to operate his al-tawhid wal-jihad

12 Illuminating the Global Suicide-Attack Network 59 Figure 5. Suicide-attack network: 2001 to 2004.

13 60 B. Acosta and S. J. Childs organization under the name Tanzim Qaedat al-jihad fi Bilad al-rafidayn (the Base of the Jihad Squad in the Land of Two Rivers) or Al Qaeda in Iraq (AQI). The rebranding increased the prestige of Zarqawi s organization among jihadis by displaying official Al Qaeda sponsorship. New employers of suicide attacks included Al Qaeda in Arabia, the Pakistani Harakat-ul Mujahedin, the Islamic Army in Iraq, Kitaeb Thourat al- Ashrin, and the Central Asian Islamic Jihad Group (IJG). In 2005, suicide attacks jumped from the prior year to 451 and ties slightly rose to 74. In February, the previously unknown al-nosra wal-jihad fi Bilad al-sham used a suicide bomber to assassinate former-lebanese Prime Minister Rafik Hariri in Beirut. In Iraq, Jund al-sahaba carried out it first suicide attack, as did the Levantine Jund al-sham and Abdallah Azzam Shahid Brigades (AASB), the Palestinian Popular Resistance Committees (PRC), the Islamic Glory Brigades in the Land of the Nile, the Pakistani Sipah-e Sahaba, and the Bangladeshi Jamaat-ul Mujahedin (see Figure 6). In 2006, ties remained at 74, while overall suicide attacks decreased marginally to 434. Baitullah Mehsud s organization in Pakistan began launching suicide attacks, as did Islamic Jihad of Yemen (also known as Al Qaeda in Yemen), Ittihad al-mahakim al-islamiyya in Somalia, a group in Iraq named the Supporters of the Sunni Community, and the Pakistani Hizb-ul Mujahedin. In January, Al Qaeda in Iraq merged with smaller groups and changed its name again to the Mujahedin Shura Council, which in October merged with more groups including Jund al-sahaba to form the Islamic State of Iraq. In 2007, ties starkly decreased to 34, as overall suicide attacks reached the highest annual level of 613. A collection of North African Islamist groups united under the name Al Qaeda in the Islamic Maghreb (AQIM) and began carrying out suicide attacks, as did Hizb-e Islami in Afghanistan, Harakat al-shabaab al-mujahedin (HSM) in Somalia, and Tehrik-e Nafaz-e Shariat-e Mohammadi (TNSM), and the Amjad Farooqi Group in Pakistan. Mehsud s organization began conducting suicide attacks under the name Tehrik-e Taliban or Pakistani Taliban. In 2008, ties more than doubled to 78 and overall suicide attacks decreased to 447. During this period, the Taliban and Tehrik-e Taliban became key brokers within the network. Tekrik-e Taliban established a number of schools for potential shuhada (Waraich 2009). Organizations launching their first suicide attack included the Levantine Fatah al-islam, a Shi a-houthi organization in Yemen al-shabaab al-muminim, the Turkestan Islamic Party, the Baluchi organization Jundallah, the Haqqani Network in Afghanistan, and the Pakistani groups Lashkar-e Abdullah and Karwan-e Naimatullah. The Iranian affiliate of the PKK, called Free Existence of Kurdistan Party (PJAK), began launching suicide attacks. By 2008, many Pakistani organizations had formed alliances, and began conducting attacks under the umbrella organizations Harakat ul-jihad al-islami (HJI) and Fedayin-e Islam. In 2009, ties decreased to 56, as did overall attacks to 332. During the year, the RSMB helped form the Caucasus Emirate, and Islamic Jihad of Yemen and Al Qaeda in Arabia merged into Al Qaeda in the Arabian Peninsula (AQAP). Pakistani organizations Lashkar-e Islam and Lashkar-e Jhangvi al-alami (LJA) carried out their first suicide attacks, as did Al Qaeda s special operations Shadow Army, Lashkar al-zil (see Figure 7). In 2010, ties increased sharply to 94 yet overall attacks stayed nearly the same at 330. The Pakistani Ansar ul-islam began utilizing suicide attacks against members of its rival Lashkar-e Islam. A splinter group of the PKK started conducting suicide attacks under the name the Kurdistan Freedom Falcons (TAK). The Tajiki Jama at Ansar Allah began using suicide attacks as well. In 2011, ties dropped greatly to 42 and overall attacks decreased to 245. A variety of organizations and militants began using suicide attacks, including the Palestinian Jaish

14 Illuminating the Global Suicide-Attack Network 61 Figure 6. Suicide-attack network: 2005 to 2008.

15 62 B. Acosta and S. J. Childs Figure 7. Suicide-attack network: 2009 to al-islam, the Indonesian Jema ah Ashorut Tauhid, the Kazakhi Jund al-khalifah, the Balochi Liberation Army, as well as anti-qadaffi rebels and Misurata Sons of Lions Brigade in Libya. Additionally, AQAP began sponsoring attacks under the name Ansar al-shari a.

16 Illuminating the Global Suicide-Attack Network 63 Modeling the Suicide-Attack Network To more thoroughly analyze the findings of the network visuals, we apply an econometric analysis of the related quantitative and network measures. We apply our method at the system as well as organizational levels of analysis, in order to decipher the relation between suicide attacks, organizational network ties, and time. 21 We seek to reverse engineer the suicide-attack phenomenon by first unraveling the output mechanisms of the system, and then determining causal elements among the system s constituent parts. At the system level, we theorize that attacks and ties beget attacks. Under this theory, the frequency of suicide attacks in a given year alone affects the perpetuation of the tactic in subsequent years. Network ties between employing organizations serve to increase or diminish the frequency of attacks. Additionally, as the network matures and main nodes, or brokers, expand influence to like-minded organizations, these new diffusion channels increase the frequency of attacks. Moreover, we test the proposition that as a network s connectivity increases, knowledge related to the product (i.e., suicide-attacks) diffuses and as a result the connecting ties facilitate a higher quality of information flow and subsequent rise in product output. Connectivity at t 0 implies a period of preparation and knowledge-transfer; therefore, one expects it to result in an increase in the frequency of attacks at subsequent periods following this increased systemic connectivity. These increases result from organizations that shift from internalizing the tactic to employing it. Moving beyond the system level, we examine organizational level effects, using ties and brokerage measures to uncover the bridging role of organizations on an individual basis. We enumerate testable propositions below, first addressing the aggregated network (system level) and then the disaggregated actors (organization level). Hypotheses We test the following hypotheses at the network-system level of analysis: H 1 (Connectivity): An increase in the number of ties in the network produces a subsequent increase in the overall number of attacks. Moretotalconnectionswithinthesystem spawns greater diffusion and aggregated use of the tactic. Put differently, as the network becomes more interconnected, knowledge of the tactic and influence to use the tactic spread, thus generating more attacks. Notably, more ties indicate more connections among the ideologically congruent. H 2 (Brokerage): An increase of brokerage in the network relates to a subsequent increase in the overall number of attacks.asbrokerorganizationsgarnermoreinfluenceinthe network, it precipitates an increase in the output of attacks. Brokers have the capability to apply pressure on like-minded organizations, and establish expectations, such as the adoption and promotion of concepts like self-annihilating martyrdom. H 3 (Impulsion): An increase of attacks foreshadows a subsequent increase in the volume of attacks. Simply,attackstriggeranimpulsionforfutureattacks.Effectively,previous systemic attacks advertise and build momentum for new attacks. When established organizations within the network become more active in producing attacks, it reinforces the impetus among affiliates to launch attacks. In one sense, this represents a snowball effect, in which the number of attacks progressively grows across time when all other factors are held constant. In another sense, it exposes the fashionable feature within the suicide-attack phenomenon. Imitating a trend of fashion, the martyrdom

17 64 B. Acosta and S. J. Childs operation /suicide attack consistently attracts new adherents, driving the reproduction and perpetual growth of the phenomenon. H 4 (Autogamy): An initial increase in network ties initiates a cycle, in which the aggregate number of ties diminishes over subsequent years.asaself-reproducingnetwork,brokers remain embedded and mostly immune to productive decay, yet other organizations leave the network for various reasons, including defunction, newfound disapproval, or ineptitude. With a high turnover of membership in the suicide-attack network, particular sets of ties generally have a short lifespan. Additionally, we disaggregate the data and test following hypotheses at the organization level of analysis: H 5 (Conduction): Increases in an organization s conductive avenues brings about an increase in that organization s frequency of attacks.asanorganizationexpandsitsdirect ties and brokerage, it conducts more attacks. By doing so, such dominant organizations, fulfill a need to set an example to adherents recently tied in to the network. H 6 (Political Islam): Following an ideology based on political Islam comparatively heightens an organization s output of suicide attacks. PoliticalIslamdirectlyaffectsand establishes an organization s ability and expectation to carry out suicide attacks and diffuse the tactic to other organizations. H 7 (Range): Using suicide attacks against civilians comparatively lowers an organization s output of suicide attacks.overtime,organizationshaveusedthetacticmorefrequently outside of the context of terrorism, preferring to strike combatant or political targets. Models and Results 22 Before exploring the temporal causality, we provide a snapshot of the aggregate associations with the network measures of ties and brokerage and the dependent variable of attacks through Ordinary Least Squares (OLS). Table 1 displays the OLS results of four log-log models. The findings validate the network approach, confirming that the network matters in the perpetuation of suicide attacks. At this aggregated level, the number of ties in the system alone explains 87 percent of the variance; each 1 percent increase in the number of ties in the network is associated with a 128 percent increase in the number of attacks. Model 3evaluatesbrokerage,orthelevelofinfluenceanodemaintainsinconnectingothernodes. The brokerage measure captures 73 percent of the variance in explaining attacks; each 1 percent increase in brokerage is associated with a 279 percent increase in the frequency of attacks. To allay concerns of endogeneity, a one-year lag is examined. The results of the one-year lag lose 15 percent of explained variation, but the coefficients remain largely unchanged (see models 2 and 4). All in all, the findings of the OLS show that relationships are important in spurring an increase in attacks. In order to further decipher the dynamics, we turn to isolating temporal effects. Using an annual time series from 1980 to 2012, we highlight the network s evolution by addressing the relation between of attacks and network ties providing a quantitative representation of the suicide-attack phenomenon s qualitative narrative. We analyze the relationship using Vector Auto-Regression (VAR), as the system-level hypotheses require accounting for the temporal domain (i.e., time) in tests for significance. 23 In order to conform to the stationary unit-root assumption necessary for time-series analysis, the data is differenced. Therefore, the VAR findings entail an understanding of the immediate growth of the variables. The equation considers lags out to the second-order per goodness of fit

18 Illuminating the Global Suicide-Attack Network 65 Table 1 Ordinary least squares results DV = Attacks (logged) (1) (2) (3) (4) Ties logged (0.090) Ties logged (t-1) (0.133) Brokerage logged Brokerage logged (t-1) (0.308) (0.375) Constant (0.196) (0.284) (0.218) (0.266) R-Squared Observations F-statistic Root Mean Squared Error p <.01. measures, 24 with the following specification: Attacks = β 10 + β 11 Attacks t 1 + β 12 Attacks t 2 + β 13 Ties t 1 + β 14 Ties t 2 + ε 1 Ties = β 20 + β 21 Attacks t 1 + β 22 Attacks t 2 + β 23 Ties t 1 + β 24 Ties t 2 + ε 2 Table 2 reports the VAR results. 25 Model 5 provides evidence of a snowball effect with respect to the frequency of attacks. Demonstrating sustained momentum, every new attack from one year to the next results in an increase of one-third of an attack in the following two years. Regarding the network measure of ties, each additional tie from one year to the next results in an increase in six attacks at two years out, and three attacks at one year. Thus, as more connections are made in the network, the output of attacks rises. The Granger-causality test confirms the finding. The VAR approach provides the added advantage of exploring feedback loops between attacks and ties. Accordingly, model 6 treats ties as the dependent variable, finding no evidence that a change in attacks results in a change in ties. There is, however, a negative association of ties on ties. On average, each additional tie from one year to the next brings adecreaseofhalfatie.thedatashowsthatafeedbackmechanismisatwork:increased network ties lead to a growth in the number of attacks, yet increases in the number of ties

19 66 B. Acosta and S. J. Childs Table 2 Vector auto-regression results Model 5 Model fitness Y = Attacks Lag Coef. Std. Err. Sample: AIC = Attacks t (0.171) LL = HQIC = Attacks t (0.170) Observations = 29 SBIC = Ties t (1.591) Equation F-Stat R-sq RMSE Ties t (1.733) Attacks Constant t (12.089) Ties Model 6 Granger causality Y = Ties Lag Coef. Std. Err. Attacks t (0.023) Y = Attacks F-Stat Attacks t (0.023) Ties Ties t (0.212) Ties t (0.231) Y = Ties F-Stat Constant t (1.613) Attacks p <.01, p <.05, p <.1. Degrees of freedom and standard errors adjusted for small samples. briefly brings about a decrease in network ties. In summation of the VAR results, a growth in ties and attacks causes a growth in attacks. In addition to incorporating the dynamic element by testing lag orders, we disaggregate the data by approaching the phenomenon from an organizational perspective. With the unit of analysis as the organization by each year, Table 3 displays random-effects panel regression 26 results for contemporaneous models with the dependent variable of attacks. 27 Using this disaggregated organization-level analysis, models 7 and 8 explain the elasticity in attacks. Unlike the aggregated system findings, the panel regressions allow for estimating the effects of each individual organization in the system. These effects are based on the within variation of each organization (every individual organization over time) as well as the variation between organizations. We test the effect of the number of connections for each organization along with two control variables for ideology and targeting. Model 7 demonstrates that every 1 percent increase in the amount of connections an organization makes results in a 48 percent increase in that organization s number of attacks. Model 8 shows that broker organizations increase their number of attacks by 50 percent for each 1 percent increase in their brokerage score. Models 7 and 8 also show that organizations that subscribe to an ideology incorporating political Islam have a positive association with attacks in comparison to all other employers of suicide attacks; Islamist organizations produce 30 percent more attacks compared to all other organizations. If organizations predominately target civilians, they have a negative association with attacks in comparison to organizations with a non-civilian targeting preference. Organizations that solely target civilians alone produce 47 to 51 percent fewer attacks than all other organizations. The econometric findings of the combined network and time-series approach confirm the connectivity, brokerage, impulsion, autogamy, conduction, political Islam, and range hypotheses. At the network-system level, models 1, 2, and 5 confirm hypothesis 1 that

20 Illuminating the Global Suicide-Attack Network 67 Table 3 Random effects panel regression results DV = Attacks (logged) (7) (8) Ties (logged) Brokerage (logged) (0.153) (0.183) Islamist ideology (0.179) (0.164) Civilian targeting (0.103) (0.096) Constant (0.195) (0.148) R-Squared (within) R-Squared (between) Years (minimum) 1 1 Years (mean) Years (maximum) Organizations Observations (organization-year) Chi-squared Root Mean Squared Error p <.01, p <.1. Standard errors clustered by organization. network ties produce attacks. Models 3 and 4 confirm hypothesis 2 that network brokerage increases attacks. Model 5 confirms hypothesis 3 that attacks beget attacks. Model 6 confirms hypothesis 4 that network ties have low longevity but are regularly replaced. At the organization level, models 7 and 8 confirm hypothesis 5 that an organization s brokerage and ties increase its attack output. Models 7 and 8 also confirm hypothesis 6 that political Islam contributes to an organization s heightened output of attacks and hypothesis 7 that attacking civilian targets factors in an organization s lowered output of attacks. These confirmations indicate that the network itself, and especially broker organizations, perpetuates the phenomenon. As connectivity within the network increases, it produces more attacks. An increase in attacks, itself, precipitates a greater increase in attacks over a two-year period. Specific tie-sets tend not to last very long, though new ties emerge rapidly to take the place of severed ties. Broker organizations mark the chief organizational employers and diffusers of the tactic, and they commonly subscribe to an ideology based on political Islam and prefer using suicide attacks to strike combatants and other non-civilian targets (see Figure 8). Network Connections, Ideology, and the Autogamy of a Distinct Phenomenon The hypotheses this study puts forth and confirms bolster three key theoretical assertions. First, connections between organizations matter in the perpetuation and diffusion of suicide attacks. Second, ideology largely facilitates the establishment of network connections.

21 68 B. Acosta and S. J. Childs Figure 8. Broker organizations. (Color figure available online). Third, suicide attacks represent a distinct phenomenon of political violence, and not a mere sub-category of terrorism. The Suicide-Attack Network The findings demonstrate that the network matters. The network of militant organizations that contemporaneously carries out suicide attacks forms an autogamous system continuously promoting new brokers and attracting new members as others go defunct or abandon the tactic. Simply, the actions of one organization affect connected organizations, and this system of interactions reproduces itself over time, often with new actors. 28 Studying these interdependent effects makes for a complementary fourth, networksystem, level of analysis. Ideology and Natural Alliances Time has shown, while the actors within the network change frequently, the connectivity of the system remains rooted in ideological similitude among actors. In essence, ideological congruence contributes to establishing the physical alliances among individual organizations that comprise the system. Ideological ties resolve innate collective action problems between different actors. Naturally, organizations with similar goals cooperate and exchange information, technology, and innovations more readily than organizations with dissimilar or opposing ideological objectives. Ideology can also establish expectations for affiliates of powerful organizations, or between organizations with a shared political constituency. Structurally, ideological fashions, and their signifying social products such as martyrdom operations, come and go. 29 Yet, along the way, numerous actors jump on the bandwagon of modi operandi made popular by powerful organizations that actively assert influence in efforts to keep their ideological outlook dominant.