All Terrorism is Local? A Quantitative Analysis of Al Qaeda Affiliation and Rebel Group Behavior


Scholarship on Al Qaeda affiliates (AQAs) has focused largely on these groups’ lethality or relationship to Al Qaeda’s transnational structure. This study addresses the queries: Are AQAs “hijacked” by Al Qaeda’s transnational jihadist ideology, becoming mere franchises in “Al Qaeda, Inc?” Or are they “hijacking” the Al Qaeda brand, using the benefits it provides to fuel their local struggles? This paper quantitatively examines the effects of a group’s affiliation on a group’s behavior. As over 96 percent of AQA attacks from 1976-2012 occurred in civil war contexts, AQAs are best characterized as rebel groups in civil wars that use terrorism as a strategy to achieve largely parochial political goals. Chi-squared and difference in proportions tests are conducted to determine the relationships between Al Qaeda affiliates, rebel groups, and civil war. An original map overlaying AQA attacks on civil war zones integrates these concepts using ArcGIS software. There is strong evidence that civil war context is a strong indicator of AQA behavior, while actual Al Qaeda affiliation is not. This supports the characterization of AQAs as rebel groups.

Table of Contents: 


    As of June 14th, 2014, a group calling itself the Islamic State of Iraq and Syria (ISIS) held position 60 miles north of the Iraqi capital, Baghdad. ISIS moved almost 200 miles south in the span of three days, capturing territory stretching 200 miles north to Mosul and 200 miles west to the deserts of Anbar province, including Iraq’s second largest city (Nordland & Rubin, 2014). ISIS’s foray and accompanying brutality have also brought it to the forefront of global media coverage, permeating across television, social networks, and academic discourse worldwide. News outlets rushed to provide profiles of the group’s origins, actions, and development into a major regional player. They have also attempted to clarify ISIS’s connections to a group that is much more recognizable to the west: Al Qaeda.

    ISIS originated as “Al Qaeda in Iraq,” just one of dozens of Al Qaeda-affiliated groups worldwide. In the post-9/11 rush to understand more about the shadowy transnational organization Al Qaeda, policymakers, journalists, and academics delved into the available information on Al Qaeda’s militant jihadist ideology and network structure. A great deal of work focused on Al Qaeda’s ideology and development from a group of former anti-Soviet fighters in Afghanistan drawn from all over the Arab world to a globally recognizable organization capable of threatening the strongest world powers at home (Gerges, 2009; Cragin & Daly, 2004; Hafez, 2003). Other researchers from a variety of fields subsequently examined the Al Qaeda “brand’s” diffusion through its network, and the advantages and disadvantages of that particular network (Duffield, 2002; Eilstrup-Sangiovanni & Jones, 2008; Helfstein & Wright, 2011; Kenney, 2007; Kilberg, 2012; Sageman, 2004; Siqueira & Sandler, 2010).

    Less scholarship focuses on these Al Qaeda-affiliated groups themselves, or AQAs. At the core of an examination of AQAs lie the broad queries: Are AQAs “hijacked” by Al Qaeda’s transnational jihadist ideology, becoming mere franchises in “Al Qaeda, Inc?” Or are these groups “hijacking” the Al Qaeda brand, using the benefits it provides to fuel their parochial struggles? Scholars and analysts have argued about the pros and cons of groups affiliating with Al Qaeda, using almost entirely a qualitative approach. I quantitatively examine the actual effects of affiliation on a group’s violent behavior. It is intuitive that a group would become more deadly after affiliating with an experienced, well-equipped organization such as Al Qaeda. But do affiliates change their targeting behaviors after affiliation? How well does their target choice reflect their local aims versus AQ Central’s “far enemy” focus? My results show that AQAs do not proportionally increase attacks on “far enemy” (US/allies) targets after affiliation. I also situate these questions within the well-established scholarship on civil conflict, as most AQAs originated and continue to operate in civil war contexts. AQAs in civil war contexts are less likely to attack “far enemy” targets. Additionally, an analysis of transnational AQA attacks shows that the vast majority of AQA attacks are perpetrated in the group’s home state or contiguous states, which aligns with rebel group behavior in civil wars. This study shows that civil war may be a stronger predictor of AQA behavior than mere affiliation.

    Background: Al Qaeda Central – Then and Now

    The roots of the Al Qaeda organization lie in 1980s Afghanistan, where thousands of radical Muslims from around the world converged to fight the Soviet occupation in a holy jihad. One of those fighters was Osama Bin Laden, who found a group of fellow mujahideen who shared his radical, fundamentalist view of Sunni Islam. Forged and tested in the ultimately successful struggle against the Soviets, the group coalesced and began to advocate for violent struggle against those who did not adhere to its radical outlook, mostly “apostate” regimes in Muslim-dominated states. During the first Gulf War, Osama Bin Laden developed his enmity towards the United States further when the his native Saudi Arabia’s leadership chose U.S. and allied support against Saddam Hussein over Bin Laden’s offer of assistance from the mujahideen. Bin Laden believed that the presence of U.S. troops in Saudi Arabia was profaning the most holy place for Muslims, making their presence a major grievance (Nelson & Sanderson, 2011).

    Bin Laden and the core group of what had become Al Qaeda (AQ Central) were banished to Sudan, where they were offered safe haven from 1992-1996. During this period, the core group of Al Qaeda leadership tightened further. In 1996, the Al Qaeda leadership moved to Taliban-controlled Afghanistan, where it was offered safe haven. It was during this period, in the late 1990s, that Bin Laden shifted the organization’s focus to the “far enemy” – America and its allies – from the “near enemy” (apostate Muslim regimes). Bin Laden issued fatwas, religious edicts, in 1996 and 1998 calling on all Muslims to kill Americans, Israelis, and their allies whenever possible (Nelson & Sanderson, 2011).

    It is this far enemy ideology that persists to this day, advocated by the Al Qaeda leadership operating in Afghanistan and Pakistan as well as many of the organization’s adherents worldwide. Al Qaeda’s most successful operation to date remains the deadly September 11, 2001, attacks on the World Trade Center in New York, which perfectly aligned with the far enemy ideology. Since the killing of Bin Laden in May 2011 by American forces, Al Qaeda veteran Ayman al-Zawahiri has arisen as the head of the Al Qaeda organization. Though Zawahiri’s dedication to the far enemy ideology is not as strong as Bin Laden’s, (Gerges, 2009), the official ideology of Al Qaeda remains focused on violence toward America, Israel, and their allies. 

    During the late 1980s and 1990s, Bin Laden worked to broaden the reach of his group. He built alliances with militant groups throughout North Africa and the Middle East, including outfits in Egypt, Oman, Jordan, and Iraq. He also worked to forge ties with groups beyond the Arab world, aiding entities such as Jemaah Islamiyah in the Philippines and groups in Burma, Chad, Malaysia, Uganda, and others. Bin Laden’s Al Qaeda trained thousands of militants in its camps in Afghanistan, and provided varying levels of strategic, operational, and financial support to groups and plots around the world (Nelson & Sanderson, 2011). What would a group of Saudis and Egyptians based in Afghanistan have to gain from jihadist violence in Burma, Mali, or Tajikistan? The answer lies in the radical fundamentalist Al Qaeda ideology, which envisions a caliphate of pure Muslim rule in the Middle East and part of Africa and Central Asia.

    This goal is the central motivator for thousands of militants who pass through Al Qaeda’s training camps, for adherents worldwide, and for those who pledge fealty to this strain of radical Sunni ideology. Its sheer unattainability, however, is a major detriment to the Al Qaeda organization. It is like a major corporation with a strategic plan that is next to impossible to execute; how, then, will employees and financial backers stay committed to the organization? The answer lies in the affiliate/alliance network that Bin Laden has been building since the 1980s. Al Qaeda can draw on the constant militancy of these scattered outfits to show progress to its adherents. This constant action is vital to show that the organization is working toward its goal and progress is being made. In a sense, Al Qaeda gains whenever a group that (even only nominally) shares its ideology sows chaos in its name. It is when the machine pauses that chinks in the ideological armor show through.

    Literature Review

    The literature relevant to the aforementioned research questions can be divided into four categories, each illuminating an important facet of this interdisciplinary discussion of terrorism, networks, and civil conflict: contemporary think tank reports, academic literature on terrorism, academic literature on network organizational structures, and academic literature on civil wars and terrorism. Each category yields a strong contribution to an empirical study of AQAs, and also leaves gaps that can be filled by other areas of literature.

    Think Tank Reports

    The most specific, current information regarding AQAs’ origins, ideologies, and current practices can be found in the many well-researched, comprehensive reports issued by policy-oriented think tanks such as RAND, the Center for Strategic and International Studies, the American Enterprise Institute, the Congressional Research Service, and the Combating Terrorism Center at West Point. These reports are much more policy-oriented than they are academic, and are not theory-driven. They often reference primary source documents on Al Qaeda (AQ) ideology and practices, case-by-case discussions of AQAs that draw heavily from news sources, and generally include counterterrorism recommendations for relevant policymakers.

    Several authors in this discipline engage the question of what constitutes an affiliate, and utilize different methods of categorization that are relevant to this discussion. Thomas Joscelyn of the Long War Journal classifies AQAs as groups that have sworn the official oath of bayat, or loyalty, to AQ core, while others simply emphasize the diversity within the set of formal and informal AQA-AQ alliances (Ibrahim, 2014). In Congressional testimony (United States, 2013) Seth Jones from RAND divides AQ’s organizational structure into four tiers: AQ Central, affiliated groups that have become formal branches of AQ, allied groups that have established direct relationships with AQ but are not formal members, and inspired networks that have no direct contact with AQ. Nelson and Sanderson (2011) utilize a three-tiered framework in their report for the CSIS, effectively combining RAND’s separate affiliated and allied tiers into one cluster they call “Al Qaeda affiliates and like-minded groups.”

    Regarding the question of the AQAs’ global versus local allegiances, most of the analysts  argue that though affiliating with Al Qaeda can yield new sources of funding, recruits, and other benefits, most AQAs continue to pursue largely local agendas (Nelson & Sanderson, 2011; Rollins, 2010;  Loidolt, 2011; Chivvis & Liepman, 2013; Mudd, 2012; United States, 2013). These reports relate case by case analyses of AQAs’ locally-based origins, motivations, and continuing violent actions, despite affiliation with AQ. Their arguments are often supported by analyses of internal AQ documents that have been seized and made available to the public (Lahoud et al., 2012). A weakness of these arguments is that they are almost entirely on a qualitative case-by-case basis, despite their claims about trends in the AQA network as a whole. Additionally, they focus on a few well-known cases or limit their analyses to particular regions of interest to policymakers. Countering or qualifying these assessments, some argue that it is dangerous to assume that AQAs are simply local groups nominally supporting the AQ “brand.” In their view, AQAs continue to serve AQ Central’s goals and maintain strong ties to AQ Central through leadership and shared experiences in Afghanistan in the 1980s and Iraq (United States, 2013).

    These reports integrate a great deal of regional, political, and historical context with authority; the think tanks they represent have been analyzing terrorist threats and the Al Qaeda network for decades, even before much of the current academic scholarship (Pape, 2009). However, they lack theoretical framing and do not attempt to build on prior scholarship. They also lack quantitative evidence to support their claims in all but a few cases, and those are very basic representations of the Al Qaeda and AQA violence (Jones, 2013); Cragin & Daly, 2004).

    Terrorism Literature

    The academic study of terrorism can be roughly categorized into two waves (Pape, 2009). From the 1970’s through the 1990s, research focused mainly on the causes of terrorism at the individual and societal level, and focused on single causal factors, essentially in isolation. This scholarship focused on the mostly leftist, revolutionary terrorists of the time, and emphasized the role of irrational factors in individual and group terrorism motivation and action (Rapoport, 1971; Hoffman, 1998; Crenshaw, 1981). This work formed the vital basis for later terrorism scholarship, which is situated in the post-9/11 period. In this period, researchers have built on and in some instances countered prior work, using quantitative social science methods and more complex analytical tools (Byman, 2006; Hafez, 2003; Krueger, 2007; Kydd & Walter, 2006; Sageman, 2004). A major line of theory has developed around more rational, strategic explanations for terrorist behavior and decision-making (Kydd & Walter, 2006).

    In order to analyze the actions and motivations of AQAs, it is necessary to address the debate in the literature on terrorist decision-making. There are three general bodies of theory on this topic: strategic, organizational, and psychological (McCormick, 2003). The strategic model posits that terrorist violence is a form of costly signaling. Terrorists are too weak to impose their will directly through armed force, and normal communication is insufficient to achieve a group’s aims in bargaining. Terrorist groups are treated as players in a game of strategy, making decisions based upon their perception of their opponents, political constituency, and other actors in order to maximize political returns and minimize costs. This body of theory acknowledges that individual terrorists have a variety of motives (rewards in the afterlife, financial payoffs, ideologically-motivated revenge), but treats terrorist groups as having directly political goals (Kydd & Walter, 2006; Schelling, 1960). Overall, I find the strategic theory of terrorist decision-making to be the most compelling, and the most applicable to this study of Al Qaeda affiliates. In the strategic model, AQAs are individual, politically-motivated entities; a characterization that dovetails with the contextual think tank assessments of AQAs as more locally-focused groups.

    The second body of theory is the psychological frame, which focuses on common psychological traits of terrorists. Scholars in this field have sought to explain terrorist behavior through individual personality traits and background (Ross, 1996), mental pathologies (Silke, 1998), and other psychological models. Though these models have varying levels of support, they are all rooted in an argument that McCormick accurately characterizes as impossible to disprove: the roots of terrorism are in the mind. As McCormick argues, the psychological framework ignores significant variation in terrorists’ social and cultural environments, though it is outside the scope of this study to analyze psychological factors concerning terrorism (McCormick, 2003). Additionally, this paper addresses terrorism at a group level, not an individual one, and is primarily concerned with differences in context that the psychological frame ignores. The third broad theory of terrorist decision-making is the organizational frame. In this paradigm, terrorist behavior can be interpreted by examining the structure and makeup of the group itself. This literature on this dynamic will be discussed in the next section, which discusses network organizational structures.

    Literature on Network Organizational Structures

    Political scientists have adopted the concept of networks, social and economic systems where actors are linked through ongoing formal and informal relations from other disciplines. It seems straightforward to apply this model to networks such as Al Qaeda, an organization made up of many links to individuals and groups.

    A great deal of recent scholarship has focused on the threats posed by illicit networks (Raab & Milward, 2003; Duffield, 2002; Sageman, 2004). According to this literature, the main confrontation in world politics is actually between states and networks, organizational structures that offer significant advantages. These scholars argue that networks’ fluid structures enables adaptability, resilience, capacity for rapid innovation and learning, and wide scale recruitment that make them formidable opponents for hierarchically organized states.

    However, Eilstrup-Sangiovanni and Jones (2008) convincingly argue that these analyses of networks neglect important structural disadvantages in illicit networks. They find that a network structure (decentralized with decision making and action dispersed among largely autonomous local actors that share personal contacts) makes it easier for illicit organizations to survive, but harder for them to engage in concerted action – like planning a major attack. Essentially, illicit networks are forced to trade efficiency of communication, information-sharing, collective action, and learning for increased security. The researchers apply this analysis to Al Qaeda, finding that Al Qaeda’s most successful and spectacular operations, including the 9/11 attacks, occurred when Al Qaeda was hierarchically structured. They argue that as the organization has decentralized to include dozens of affiliates and require the maintenance of contacts worldwide, it has lost unity, cohesion, and collective action capacity. A corollary to this assessment is that though the loss of a more hierarchical structure may inhibit the kind of large-scale attacks exemplified by 9/11, the growth of the decentralized core-affiliate network may be more conducive to smaller-scale, less costly (and risky) operations such as suicide bombings in less-developed countries. Gerges expresses this evolution most eloquently in The Far Enemy, writing that the Al Qaeda core is now “an ideological label, a state of mind, and a mobilizational outreach program to incite attacks worldwide” (2009, p. 49).

    Along these lines, Helfstein and Wright (2011) empirically test the argument that terrorist networks will tend to organize in “all-channel” (diffuse structure) and “hub-spoke” (a few individuals are well-connected) structures (Kilberg, 2012), and find that the groups tested do not demonstrate either structure. The empirical study was limited to only the networks involved in six well-documented attacks, which is a major limitation of the project. However, it illustrates the necessity of empirically testing the network claim. This study will challenge the network claim with regard to Al Qaeda, which has frequently been classified as a dangerous transnational network with local “franchises”, implying a level of coordination and discipline that may not exist.

    Literature on Civil Wars, Rebel Groups, and Terrorism

    This study aims to fill a notable gap between terrorism research and the extensive literature on rebel groups and violence in civil wars. Research on Al Qaeda and its affiliates has overwhelmingly fallen into the former category, while little scholarship exists on terrorism in civil war. This study’s focus on Al Qaeda affiliate groups, which tend to originate from and operate in civil war contexts, requires a bridging of the gap between civil war and terrorism scholarship. Findley and Young make a convincing case that, within the rationalist approach discussed earlier, actors referred to as dissidents, rebels, insurgents, terrorists, or revolutionaries in civil war contexts use different strains of violence but a similar strategic approach (2012; Byman, 2006; Kydd & Walter, 2006). This is compelling in the context of the study of Al Qaeda affiliate groups, which take on many of those categorizations. Countering this theory, Sambanis (2004) argues that terrorism and civil war are like water and ice – connected states but taking on different forms. This study aligns more with the Findley and Young approach, arguing that, regardless of the nomenclature used to describe the groups, they are using terrorism as a strategy to achieve political or territorial aims – just as rebel, dissident, or insurgent groups do. These groups should not be characterized based upon one type of tactic used; their choice among tactics is based on the circumstances and context of their situation.

    The civil war element to AQA analysis has been alluded to in an anecdotal or qualitative manner. On a group-by-group basis, even AQAs that purport to be multinational or regional players operate almost exclusively at the state level. For example, Al Qaeda in the Islamic Maghreb (AQIM), claims to represent several North African countries when in fact it originated as an Algerian rebel group, remains largely focused on Algerian issues, and features homogenously Algerian leadership (Filiu, 2009; Chivvis & Liepman, 2013). This type of parochial focus will be quantitatively tested in this study, so it is imperative to incorporate theory from civil war literature. The nearest overlap between terrorism and civil war literature concerns the use of violence in civil wars. Kalyvas (2006) addresses this when he argues that if violence is used for the intentional creation of fear then it is a means, not an end. In civil war, insurgents and incumbents compete for monopoly on the use of legitimate force, or secession, with a third player: civilians (Kalyvas, 2006). This assessment dovetails perfectly with the use of terrorism by certain groups who also seek to bargain with their local governments.

    Theory and Hypotheses

    In order to develop a theory on terrorist groups’ behavior and decision-making, it is necessary to clearly define “terrorism.” As Charles Tilly writes, the terms “terror”, “terrorism”, and “terrorist” have been used by political scientists to “sprawl across a wide range of human cruelties” in a confusing and unscientific manner. He recommends a more fundamental definition: terror is the use of an imprecisely bounded political strategy. It is characterized by the “asymmetrical deployment of threats and violence against enemies using means that fall outside of the forms of political struggle routinely operating within some current regime” (2004). Kydd and Walter use a similar definition: “the use of violence against civilians by nonstate actors to attain political goals” (2006). Terrorism is, then, a strategy, not a designation for an actor or group. “Terrorist groups” are simply groups that have elected to use the strategy of terrorism to achieve their political or territorial goals.

    Though I will build on Kydd and Walter’s theories of rational, strategic decision-making by terrorist groups, Tilly’s definition of the strategy of terror is more applicable to this study of Al Qaeda affiliates (AQAs). AQAs are largely perceived as terrorist groups, and are identified with the transnational organization of Al Qaeda. In addition, there is a tendency to conflate terrorist acts with terrorist networks, and to assume that Al Qaeda has built a multinational corporation with franchises that adhere to the same ideology and pose the same transnational threat as AQ Central. However, I will argue that AQAs are complex entities based in civil war contexts that utilize terrorism as a strategy to achieve largely parochial political goals, despite their affiliation with the most infamous global terrorist organization in the modern era.

    In this study, I will operate under the strategic model of terrorist decision-making put forward by Kydd and Walter, assuming that terrorist groups are rational actors resorting to costly signals to influence their adversaries.  They identify five goals of terrorist organizations (regime change, policy change, social control, and status quo maintenance). Kydd and Walter’s categorizations illustrate the diversity of AQAs. Some, like the Abu Sayyaf Group in the Philippines and Lashkar-e-Taiba in India, are secessionist groups seeking mainly territorial change. Many are revolutionary groups seeking political change, using terrorism as a strategy to establish an Islamic state in their home countries. This is evidence that AQAs do not simply become franchises of AQ Central upon affiliation; the groups may profess to adhere to the lofty (and impossible to achieve) “far enemy” ideology espoused by AQ Central, but their goals remain parochial.

    The decision to ally with Al Qaeda is highly strategic. Alliance allows the groups to aggregate capabilities, access technology and weaponry, and train to develop tactical skills. Affiliation with AQ Central brings funding, training, weapons, and recruits to a group that takes on the famous Al Qaeda brand name. AQ Central can offer a very high-quality alliance as well; Horowitz and Potter (2014) show that AQ has the deepest alliance network (the most ties at the center of the universe of groups), and that the most lethal groups have the deepest alliance networks. AQAs have been shown to be more lethal than other types of groups, including other radical Islamist groups (Piazza, 2009). Allying with AQ Central does make a group more lethal (Horowitz & Potter, 2014), but no analysis has yet determined whether an AQA becomes more lethal to the “far enemy.” Not all groups believe that these benefits outweigh the costs of affiliation, which can alienate the local population and make the group a bigger target to the opposition or counterterrorism forces. Signing on to a radical Sunni global jihad ideology also limits the universe of groups that are eligible for affiliation. However, for those that fit the criteria and pursue affiliation, the alliance can pay off: AQ Central contributes resources, a deep network, a famous brand name, and an AQA signs on to the global jihad ideology and represents the spread of the “AQ network” across the globe while becoming more deadly.

    Findley and Young (2012) found that most incidents of terrorism take place in geographic regions where civil war is occurring. I expand on this by showing through GIS mapping that AQAs are overwhelmingly operating in civil war contexts. I also argue that most AQAs originated as rebel groups in civil wars, then made a series of strategic decisions that brought them into the universe of groups known as “Al Qaeda Affiliates.” They have pursued a deep alliance or affiliation with Al Qaeda, motivated by the costs and benefits detailed above and a requisite similarity in professed ideology. They have also chosen to utilize the strategy of terrorism in pursuit of their political goals. Rebel groups, including AQAs, use terrorism against civilian targets to coerce the government into granting concessions, to highlight the government’s lack of monopoly on the use of violence, and to intimidate the population into collaborating with the group. AQAs thus use the resources and skills gained from affiliation to more effectively attack civilians in what is actually a dialogue with their local government.

    Groups reveal their preferences through resource allocation to particular targets and types of violence, enabling us to navigate the dichotomy between professed ideology and unknowable, actual intentions. It is more rational for a group to take the money, weapons, training, and brand recognition that Al Qaeda provides and funnel them into the struggle that most impacts their cost-benefit analyses: the local political fight. The choice of target and tactic used is constrained by resources and personnel, and groups reveal information to their constituencies, rivals, and opponents about both their preferences and capability through targeting decisions.

    I argue that targeting is the key to assessing AQAs’ actual adherence to AQ Central’s far-enemy ideology versus the pursuit of largely local aims. Lethality or mere quantity of attacks is confounded by the demonstrated material benefits of affiliating with Al Qaeda. Attack targets should reveal more about the group’s real motivations. Significant qualitative evidence has shown that, true to its “far-enemy” ideology, AQ Central strongly advocates for attacks on foreign interests, particularly American/Western interests. As stated previously, AQAs have different, more parochial aims and targets. One way of examining this claim is to compare AQAs’ targeting behavior before and after affiliation with AQ Central. I expect that, after affiliation, there is no significant increase in proportion of attacks perpetrated against US/allied targets. In fact, I expect that there is no relationship at all between a group affiliating with AQ Central and its targeting of US/allied targets proportional to total attacks:

    H1: There is no significant difference in targeting of US/allies relative to total attacks before and after a group’s affiliation with Al Qaeda Central.

    The civil war element is key here, however. The AQA-civil war connection has never been quantitatively examined. Using ArcGIS software, the Global Terrorism Database data on AQA attacks was overlaid on a map of civil war zones, with intensity represented by UCDP/PRIO data on battle-related deaths. The battle-related deaths data was only available from 1981-2013 and the GTD data from 1976-2012, so the map displays aggregated data from 1981-2012. The map shows that almost all AQA attacks in this time period occurred in countries experiencing civil war.

    If AQAs in civil war contexts are really rebel groups facing off against the government, we would expect them to have similar targeting behaviors to other, non-Al Qaeda affiliated rebel groups in similar environments:

    H2: In civil war contexts, there is no relationship between presence of Al Qaeda affiliation and targeting of US/allies.

    Another question of interest concerns those AQAs that are not operating in civil war contexts. An example is Al Qaeda in the Arabian Peninsula, which operated in Saudi Arabia, a very stable state, until the government successfully mounted a campaign to eradicate it from 2003-2007. The group then moved to a state teetering on the brink of civil war, Yemen (Rollins, 2010). These groups may also be seeking a dialogue with their local governments through terrorism, but it is important to disaggregate them from those operating in civil war environments where the government’s monopoly on the legitimate use of violence is already highly in question. We expect that groups operating in civil war contexts will concentrate their resources on their violent dialogue with the government, but groups operating outside of civil war have different resource limitations and ideological focuses.

    H3: An AQA is more likely to target US/allies when not operating in a civil war context.

    The alleged transnational nature of threats posed by AQAs is highlighted every time such a group perpetrates a high-profile attack across borders. Projecting force beyond the national borders of a group’s home country would seem to indicate a higher level of organization, potential coordination with transnational or foreign groups, and perhaps an ideological adherence to transnational aims. I argue that most instances of such cross-border attacks by AQAs still fit the civil war pattern by occurring in bordering countries, often across porous borders. A wealth of prior civil conflict research has shown that civil wars tend to cluster in space and spread to neighboring states, even when individual country characteristics are controlled for (Buhaug & Gleditsch, 2008). Thus, AQA attacks in contiguous states fit into the AQA-rebel group characterization.

    H4: Transnational AQA attacks are more likely to occur in states contiguous to the AQA’s home state than in noncontiguous states.

    Research Design

    This section will lay out the method of operationalization and testing of the hypotheses presented above. The unit of analysis used in this paper is incidents of terrorism, or events. The set of groups categorized as “Al Qaeda Affiliates” is drawn from Martha Crenshaw’s list of 28 “Global Al Qaeda” groups from the Stanford Mapping Militants Project. This list contains the groups that have a publicly declared alliance with Al Qaeda or are widely believed to be operationally associated with Al Qaeda. This list includes all of the groups categorized as AQAs from the think tank reports and academic literature, as well as a set of other groups with strong demonstrated ties with AQ Central. Thus, it is taken as a comprehensive list of groups with a high degree of alliance or affiliation with AQ Central. The temporal domain is 1976-2012, the operational lifespan of all of the groups of interest. The spatial domain is global, encompassing the activities of all of the AQA groups and the non-AQA affiliated rebel groups used to test H4. The primary dataset used is the Global Terrorism Database (GTD), an open-source database covering over 100,000 incidents of domestic and international terrorism from 1970 to 2012.

    The GTD defines an incident of terrorism as: “the threatened or actual use of force or violence by a non-state actor to attain a political, economic, religious, or social goal through fear, coercion, or intimidation.” This definition fits neatly with this paper’s use of a rational-strategic definition of terrorism; the incidents included in the GTD have been perpetrated by a wide variety of groups, all of which used terrorism as a strategy to achieve certain goals. The GTD definition does not narrow the dataset by some assumed characterization of certain groups as “terrorist groups.”

    Dependent Variables

    The dependent variable for H1-H3 is groups’ targeting of Americans and American allies in violent attacks. Targeting behavior is tied to this paper’s assessment of AQ Central’s main goals: hurting the “far enemy” – the United States and its Western allies. Targeting behavior is operationalized using the GTD variable natlty1, which gives the primary nationality of the attack target. The dependent variable for H4 is the number of attacks perpetrated in a country.

    A new variable, usalliestarget, must be created to capture a group’s targeting of AQ Central’s wider set of professed enemies, America’s allies and Israel. This is achieved by creating a dummy variable coded with a value of 1 if the target/victim’s nationality aligns with one of the 28 member states of the North Atlantic Treaty Organization (NATO) or Israel, 0 if the target/victim’s nationality is not ­on the NATO list or Israel, and left blank if the target/victim’s nationality is unknown. Only 57 of the 5859 total attacks perpetrated by AQAs had a target/victim nationality coded as unknown by the GTD, so this operationalization is justified. The set of NATO member states and Israel were chosen as a proxy for AQ Central’s rather nebulous categorization of America’s western allies and Israel as the “far enemy” (Gerges, 2009).

    Independent Variables

    The independent variable capturing the effect of Al Qaeda affiliation on a rebel group is a dummy variable indicating if an event in question occurred before or after affiliation, aqaattack. An event is coded 0 if it occurred prior to affiliation and 1 if it occurred after affiliation. The dates of affiliation were coded using the Stanford Mapping Militants Project’s qualitative group profiles and the Terrorism Knowledge Base®’s Terrorist Organization Profiles (TOPs). The TKB® was created in 2004 and maintained through 2008 by the Department of Justice, the Department of Homeland Security, and the Memorial Institute for the Prevention of Terrorism. It is now made available through the University of Maryland’s Study of Terrorism and Responses to Terrorism (START) website. Where the date of affiliation was unavailable through either of those sources, it was obtained through an additional reliable source such as the Long War Journal and the Combating Terrorism Center at West Point; all sources are documented in the appendix. The universe of cases tested for affiliation is the set of all terrorist attacks in the GTD perpetrated by the 28 AQAs drawn from the Stanford Mapping Militants project. This allows for within-group comparisons of targeting behavior before and after Al Qaeda affiliation, which controls for group-specific confounds, such as location, size, age, wealth, territorial control, and host state characteristics.

    The independent variables capturing the comparison of targeting behavior between AQAs and non-AQAs is a binary variable, affiliate, indicating whether a group in question is an AQA or not. A group is coded 0 if it is not an AQA according to the Stanford Mapping Militants list, and 1 if it is an AQA on the list.

    The independent variable concerning civil war and terrorist attacks is a binary variable indicating whether or not an event is perpetrated in a civil war context, civil. The country-years of civil war are obtained from the UCDP/PRIO Armed Conflict dataset. The event data is coded with a 1 if the country-year of an attack matches a country-year of civil war and a 0 if it does not (as indicated by the UCDP/PRIO data).

    The final independent variable concerns transnational AQA attacks and state contiguity. It is operationalizes using two original binary variables. The variable aqahomeattack is coded 1 if the country location of the attack matches the “home base” country of the AQA, indicating that the attack was perpetrated within the home country, and 0 if the attack location country does not match the “home base” country, indicating a transnational attack. “Home base” AQA countries were coded using the Stanford Mapping Militants Project’s qualitative AQA profiles and the START TKB TOPs. The variable contigattack is coded 0 if an AQA attack was perpetrated in a country noncontiguous to the AQA’s “home base” state or was not transnational, and 1 if an AQA attack was perpetrated in a contiguous state.

    The variables used in this analysis are summarized in Table 1.


    The hypotheses presented in this paper are structured to test the presence of relationships between binary independent and dependent variables in various subsets of the terror attack-civil war data. Thus, chi-squared tests are used on a binary independent and binary dependent variable to test for the presence non-random associations between independent variables and the dependent variable (attack perpetrated against US/allies). The chi-squared test is appropriate because it measures if an observed or alleged association between two binary variables is due to chance, by comparing the observed results with those that would be expected if no real relationship existed. Additionally, difference in proportions tests are conducted to detect significant differences between the proportion of total attacks that were perpetrated against the US/allies across the data subsets. Each hypothesis required a different data subset and the use of different variables, as detailed below.

    Hypothesis 1 argues that there is no significant difference in targeting of US/allies relative to total attacks before and after a group’s affiliation with Al Qaeda Central. For H1, the dataset is trimmed to only attacks perpetrated by AQAs. A chi-squared test is performed on the binary variable reflecting an attack occurring pre- or post-affiliation (aqaattack) and the binary dependent variable, usalliestarget. A difference in proportions test is conducted on the proportion of total attacks perpetrated against the US/allies pre-affiliation vs. post-affiliation.

    Hypothesis 2 argues that in civil war contexts, there is no relationship between the presence of Al Qaeda affiliation and targeting of US/allies. For H2, the dataset is trimmed to only those attacks perpetrated in civil war country-years (including both AQA and non-AQA attacks). A chi-squared test is performed on the binary variable reflecting a group being an AQA (affiliate), and the binary dependent variable, usalliestarget. A difference in proportions test is conducted on the proportion of total attacks perpetrated against the US/allies by AQAs vs. non-AQAs.

     Hypothesis 3 argues that an AQA is more likely to target US/allies when not operating in a civil war context. For H3, the dataset is trimmed to only attacks perpetrated by AQAs. A chi-squared test is performed on the binary variable reflecting an attack being perpetrated in a civil war country-year (civil), and the binary dependent variable, usalliestarget. A difference in proportions test is conducted on the proportion of total attacks perpetrated against the US/allies by AQAs in civil war contexts vs. non-civil war contexts. This analysis is designed to test the null hypothesis that there is no significant difference between targeting of US/allies relative to total attacks inside versus outside of civil war contexts.

    Hypothesis 4 argues that transnational AQA attacks are more likely to occur in states contiguous to the AQA’s home state, due to the theory that AQA operations outside of the group’s home state are still tied to the civil war dynamic. For H4, the dataset is trimmed to only attacks perpetrated by AQAs. A chi-squared test was performed on the binary variable reflecting an attack being perpetrated by an AQA in its own “home base” state (aqahomeattack) and the binary variable contigattack, which shows if the attack was perpetrated in a contiguous state.

    Results, Analysis, and Discussion

    Results and Analysis

    The findings of the analysis supported H1 (no significant difference in targeting of US/allies relative to total attacks before and after a group’s affiliation with Al Qaeda Central). The p-value of the chi-squared test was 0.829, indicating that there is an 83% chance that any association between an attack occurring before or after an AQA’s affiliation with AQ Central and targeting of US/allies is due to chance. See Table 2. The p-value of the difference in proportions test is the same, showing that there is insufficient evidence to conclude that the difference in proportions (US/allies targeting pre- vs. post-affiliation) is statistically significant. See Table 3. This result is especially strong when considering the cases of AQA activity in Iraq and Afghanistan; given the invasion of those countries by US/allied forces in years proximate to the local AQAs’ dates of affiliation, we would expect that those groups’ targeting of US/allies relative to total attacks may have gone up significantly after affiliation/invasion. Since it did not, H1 is even more robust as these cases provide the strictest test of this hypothesis. This supports my expectation that though affiliation with AQ Central may make an AQA more lethal, it does not make an AQA more likely to attack “far enemy” (US/allies) targets relative to total attacks – the groups remain focused on their local fight.

    The analysis did not support H2 (no relationship between presence of Al Qaeda affiliation and targeting of US/allies in civil war contexts). The p-values of the chi-squared test and the difference in proportions test were 0, indicating that there is a strong relationship between a group being an AQA and targeting of US/allies, in civil war contexts. See Table 4 and Table 5. This finding is weakened, however, by the fact that the tests compared 5,666 AQA attacks in civil war country-years with 75,072 attacks by non-AQAs in civil war country-years, with no controls on the type of group, type of war, or country context during the attack. It is important to revisit this hypothesis with regression analysis, which will incorporate these controls for a more accurate comparison across groups.

    The analysis supported H3 (an AQA is more likely to target US/allies when not operating in a civil war context). The p-value for the chi-squared test is 0, indicating that the relationship between an AQA being in vs. out of a civil war context and its targeting of US/allies is not due to chance. See Table 6. The p-value for the difference in proportions test is also 0, which means we can reject the null hypothesis and conclude that there is a significant difference between AQA targeting of US/allies (relative to total attacks) in civil war and out of civil war. See Table 7. This supports our expectation that the presence of a civil war context matters in targeting of US/allies. Given these results, an illustration of this finding would be the case of ISIS – ISIS in Saudi Arabia is likely to be far more threatening toward US/allies than ISIS in Iraq/Syria due to the divergent contexts of these states.

    Finally, the analysis supported H4 (transnational AQA attacks are more likely to occur in states contiguous to the AQA’s home state). The p-value of the chi-squared test was 0, indicating that the relationship between an attack being transnational and an attack being in a contiguous state is not due to chance. See Table 6.  The descriptive statistics are particularly convincing in this analysis: out of 5,867 total AQA attacks from 1960-2012, 93% were in the perpetrator’s “home base” state. 5.9% were perpetrated across borders but in contiguous states, and only 0.39% were perpetrated across borders in noncontiguous states. This lends strong support to the theory that AQAs are operating as rebel groups in civil wars; as even transnational AQA attacks occur overwhelmingly in contiguous states, a practice consistent with rebel groups in civil wars that often bleed across borders.


    The analysis shows strong support for this paper’s arguments and theoretical framework. H1, H3, and H4 are supported, showing that AQAs are not more threatening towards the US/allies “far enemy” after affiliation with AQ Central, that an AQA is more threatening towards the US/allies when operating outside of civil war, and that transnational AQA attacks are rare and perpetrated mostly in contiguous states. These findings, along with the map of AQA attacks and civil war by battle death intensity shown in Figure 1, strongly support the characterization of AQAs as rebel groups in civil wars. They refute the commonly held notion that AQAs are mere franchises in “Al Qaeda, Inc.”

    The H2 results appear to refute this hypothesis (no relationship between presence of Al Qaeda affiliation and targeting of US/allies in civil war contexts), yet is it highly likely that this is more evidence of an incomplete model than of an unsupported theory. In future research, I aim to address these issues to more accurately test this hypothesis.


    By now, it is a cliché to argue that the nature of war is changing in our time. The most significant national and global threats are posed by groups or individuals, not states, and the effort to understand this dynamic is ongoing. Academically, this study offers an important addition to both terrorism and civil war scholarship. These disciplines have long been distinct, to the detriment of developing a comprehensive understanding and meaningful analyses of complex groups and contexts. In bridging the academic divide, this paper seeks a more accurate characterization of a fascinating, threatening, and opaque set of groups: Al Qaeda affiliates. These groups should be characterized not as franchises of an international terrorism network, but as rebel groups in civil war who have formed a particular strategic alliance to achieve their own particular set of objectives. Though this affiliation makes AQAs more deadly, it does not make them more deadly to the professed primary “far enemy” – the US and its allies. Whether a group operates within a civil war context is shown to be a stronger determinant of “far enemy” targeting by AQAs than mere affiliation.

    This paper has important policy implications. Counterterrorism efforts against the entire AQ network and AQAs need to incorporate the AQA-civil war connection, and operate under the understanding that these AQAs are not mere outposts for a broader enemy. The threats posed by AQAs towards the “far enemy” and transnational targets in general are actually quite weak. These groups pose the greatest threat to their own home states by fomenting instability and through the sheer levels of violence that they impose on their own countrymen. As ISIS marches through Iraq and Syria, we need to recognize that the threats AQAs pose may not concern buildings in New York or L.A., but instead are destabilizing entire regions locked in civil war.


    • Buhaug, H. & Gleditsch, K.S. (2008). Contagion or confusion? Why conflicts cluster in space. International Studies Quarterly, 52(2), 215-233. Retrieved from
    • Byman, D.L. (2006). Friends like these: Counterinsurgency and the war on terrorism." International Security, 31(2), 79-115. Retrieved from Project Muse.
    • Chivvis, C. & Liepman, A. (2013). North Africa's menace: AQIM's evolution and the U.S. policy response. RAND Corporation. Retrieved from
    • Cragin, K. & Daly, S.A. (2004). The dynamic terrorist threat: An assessment of group motivations and capabilities in a changing world. Santa Monica: RAND Corporation.
    • Crenshaw, M. (1981). The causes of terrorism. Comparative Politics, 13, 397–99. Retrieved from
    • Duffield, M. (2002). War as a network enterprise: The new security terrain and its implications. Cultural Values, 6(1/2), 153-165. doi: 110.1080/1362517022019793.
    • Eilstrup-Sangiovanni, M. & Jones, C. (2008). Assessing the dangers of illicit networks: Why Al-Qaida may be less threatening than many think. International Security, 33(2), 7-44. Retrieved from
    • Filiu, J. (2009). The local and global jihad of al-Qa'ida in the Islamic maghrib. Middle East Journal, 63(2), 213-226. Retrieved from
    • Findley, M. & Young, J. K.. (2012). Terrorism and civil war: A spatial and temporal approach to a conceptual problem. Perspectives on Politics, 10(2): 285-305. doi:
    • Gerges, F. (2009). The far enemy, 2 ed. Boston: Cambridge University Press.
    • Hafez, M.M. (2003). Why muslims rebel. Boulder: Lynne Rienner Publishers.
    • Helfstein, S. & Wright, D. (2011). Covert or convenient? Evolution of terror attack networks. Journal of Conflict Resolution, 55,785-813. doi: 10.1177/0022002710393919.
    • Hoffman, B. (1998). Inside terrorism. New York: Columbia University Press.
    • Horowitz, M.C., & Potter, P.B.K. (2014). "Allying to kill terrorist intergroup cooperation and the consequences for lethality." Journal of Conflict Resolution, 58(2), 199-225. doi: 10.1177/0022002712468726.
    • Ibrahim, A. (2014). The resurgence of Al-Qaeda in Syria and Iraq. Strategic Studies Institute and U.S. Army War College Press.
    • Kalyvas, S.N. (2006). The Logic of Violence in Civil War. Boston: Cambridge University Press.
    • Kenney, M. (2007). From Pablo to Osama: Trafficking and terrorist networks, government bureaucracies, and competitive adaptation. University Park: Pennsylvania State University Press.
    • Kilberg, J. (2012). A basic model explaining terrorist group organizational structure. Studies In Conflict and Terrorism, 35(11), 810-830. doi:10.1080/1057610X.2012.720240.
    • Krueger, A. (2007). What Makes a Terrorist: Economics and the Root Causes of Terrorism. Princeton: Princeton University Press.
    • Kydd, A. and B. Walter. (2006). The strategies of terrorism. International Security, 31(1), 49-80. Retrieved from Project Muse.
    • Kydd, A. & Walter, B.F. (2002). Sabotaging the peace: The politics of extremist violence. International Organization, 56(2): 263-96. Retrieved from
    • Lahoud, N., Caudill, S., Collins, L., Koehler-Derrick, G., Rassler, D. & al-'Ubaydi, M. (2012). Letters from Abbottabad: Bin Laden Sidelined? Combating Terrorism Center. Retrieved from
    • Loidolt, B. (2011). Managing the global and local: The dual agendas of Al Qaeda in the Arabian Peninsula. Studies in Conflict & Terrorism, 34(2), 102-123. Retrieved from Academic Search Complete.
    • McCormick, G. (2003). Terrorist decision making. Annual Review of Political Science, 6, 473-507. doi:10.1146/annurev.polisci.6.121901.085601.
    • Mudd, P. (2012). Are jihadist groups shifting their focus from the far enemy? Sentinel, 5(5). Retrieved from
    • Nelson, R. & Sanderson, T. (2011). A threat transformed Al Qaeda and associated movements in 2011. Washington, DC: Center for Strategic and International Studies.
    • Nordland, R. & Rubin, A. (2014, June 15). Iraq rebels stall north of Baghdad as residents brace for a siege. The New York Times. Accessed June 16, 2014 from
    • Pape, R. (2009). Introduction: What is new about research on terrorism. Security Studies, 18(4): 643. doi: 10.1080/09636410903369100
    • Piazza, J.A. (2009). Is Islamist terrorism more dangerous?: An empirical study of group ideology, organization, and goal structure. Terrorism and Political Violence, 21(1): 62-88. doi: 10.1080/09546550802544698
    • Raab, J., & Milward, H.B.. (2003). Dark networks as problems. Journal of Public Administration Research and Theory, 13(4): 413-439. Retrieved from
    • Rapoport, D.C. (1971). Assassination and terrorism. Toronto: CBC Merchandising.
    • Rollins, J. (2010). Al Qaeda and affiliates historical perspective, global presence, and implications for U.S. policy. Washington, D.C.: Congressional Research Service.
    • Ross, J.I. (1996). A model of the psychological causes of oppositional terrorism. Peace and Conflict: Journal of Peace Psychology, 2(2): 129-141.
    • Sageman, M. (2004). Understanding terror networks. Philadelphia: University of Pennsylvania.
    • Sambanis, N. (2004). Using case studies to expand economic models of civil war. Perspectives on Politics, 2(2): 259–279. Retrieved from
    • Schelling, T.C. (1960). The strategy of conflict. Cambridge: Harvard University Press.
    • Silke, A. (1998). Cheshire-cat logic: The recurring theme of terrorist abnormality in psychology research.” Psychology, Crime & Law, 4(1): 51-69.
    • Siqueira, K. & Sandler, T. (2010). Terrorist networks, support, and delegation. Public Choice, 142(1/2): 237-253. doi:
    • Tilly, C. (2004). Terror, terrorism, terrorists. Sociological Theory, 22(1), 5-13. Retrieved from
    • United States. (2013). Global Al-Qaeda: Affiliates, Objectives, and Future Challenges: Hearing Before the Subcommittee on Terrorism, Nonproliferation, and Trade of the Committee on Foreign Affairs, House of Representatives, One Hundred Thirteenth Congress, First Session, July 18, 2013. 2013.

    Figure 1: Al Qaeda Affiliate Attacks and Civil War by Battle Intensity, 1981-2012

    Table 1: Summary of Measurement of Variables Used

    Table 2: Chi-squared test of H1

    Table 3: Two-sample test of proportions for H1

    Table 4: Chi-squared test of H2

    Table 5: Two-sample test of proportions for H2

    Table 6: Chi-squared test of H3

    Table 7: Two-sample test of proportions for H3

    Table 8: Chi-squared test of H4

    Appendix: Coding of AQA Affiliation Dates