Targets and Tactics: Testing for a Duality within Al Qaeda’s Network

ABSTRACT

One prevailing view treats Al Qaeda as a monolithic entity with a global network of affiliates. Yet, certain affiliates appear more committed to local political and territorial goals – parochial, not global, terrorists. We construct a classification scheme to differentiate affiliates conceptually and empirically, and then track their types over time. We sort Al Qaeda’s network of affiliates using a principal components analysis of terrorist attacks from 1988 to 2012. We show that this aids in identifying latent affiliate types, and interpreting shifts in behavior. We find that despite Al Qaeda’s anti-western rhetoric, there exists a global-parochial divide in which most affiliates are parochial – with anti-Western groups pursuing local political goals even when Western targets remain. By providing an empirical strategy to identify which affiliates are more or less aligned with global terrorism, this research holds implications for the literatures on terrorism and civil war, terrorism and democracy, and the effects of counterterrorism.

Acknowledgments

We thank Jeffrey Arnold, Brianna Bardos, Janina Beiser, Dana D’Amelio, Jacob Eisenstein, Dan Reiter, Joshua Rovner and anonymous reviewers and colleagues for their comments and questions. We also thank Daniel Arnon, Rachel Harmon, Tina Huang, Peyton Sammons, Ryan Tans, and Kalyn Zaman for their exceptional research assistance. The authors are responsible for any errors.

Notes

¹ Al Shabaab, a Somali militant group affliated with Al Qaeda, executed more than 150 civilians after holding them hostage at a Kenyan university in 2014. Although Al Shabaab is subject to internal divisions, most of its fighters are predominantly interested in the nationalistic battle, and do not support a global jihad. Boko Haram { a group that for years claimed to be an Al Qaeda affliate, but was not formally acknowledged by Al Qaeda, and more recently allied with ISIS} has targeted civilians in an insurgency in Nigeria since 2009 with the goal of creating an Islamic state.

² In 2002, Jemaah Islamiyaah ordered a strategy of hitting soft targets like Bali nightclubs and bars, due to the belief these were visited by Americans (Byman 2015).

³ Al Qaeda lists five reasons to kidnap: to coerce concessions; to shame a government that cannot protect its citizens; to gain valuable information; as propaganda for a specific cause; and to obtain ransom money ‘like what our brothers did in Philippines, Chechnya, and Algeria.’ Abdelaziz al-Muqrin, ‘Al Qaeda Guide to Kidnapping,’ New York Times, 29th July 2014.

⁴ We include analyses using all tactics in the Appendix.

⁵ We inclusively examine all groups where a relationship has been documented in tracing all of Al Qaeda’s partnerships, formal and informal affliations since Al Qaeda’s inception.

⁶ Mapping Militant Organizations Project (2016), MIPT Terrorism Knowledge Base (TKB): Profiles archived through START’s Big, Allied and Dangerous (BAAD) (2016) and National Consortium for the Study of Terrorism and Responses to Terrorism (START) (2016).

⁷ We exclude the following affliates due to the lack of data: Al Qaeda Kurdish Battalions, Lashkar-e-Zil, Moroccan Islamic Combatant Group, Mujahideen Army, Second Soran Unit.

⁸ Data for this come from GTD’s natlty1, which codes the target’s nationality.

⁹ We used two conditions: first, the GTD’s targtype1 must indicate that the entity targeted was the government, police, military, or government-diplomatic categories. Second, the attack had to take place in the state in which the affiliate is operationally based, as coded in TKB and listed in Table 1.

¹⁰ We do not use factor analysis because that requires specification of the number of factors in advance. Since we do not know the correct number of factors, we let the PCA extract the optimal number of dimensions.

¹¹ In addition, if the differences in variances were a problem, then the PCA might find State-Suicide to form Component 1, and Killed-West to form Component 2. This does not occur.

¹² The more complex model shows that affliates that target the West are more likely to use suicide attacks, and attacks using bombs and explosives. Affiliates that target the State are more likely to use kidnappings, assault, assassinations, and infrastructure attacks. Beyond this, certain affiliates that target the West also use kidnappings, and a few affliates that target the State use suicide attacks. Overall, these fit with our findings here: Global affiliates target the West and use suicide attacks; parochial affiliates target the State and use kidnappings; all affiliates are using kidnappings more recently.

¹³ PCA is not an inferential statistical method; thus, it does not return standard errors for its estimated components. The limits indicate the ranges within which the affiliate means are located 95% of the time.