Version 1.0 April Appendix A: Suicide Terrorism Group Linkages

Size: px

Start display at page:

Download "Version 1.0 April Appendix A: Suicide Terrorism Group Linkages"

Homer Oliver
5 years ago
Views:

1 Appendices And Extensions For: Michael C. Horowitz, Nonstate Actors and the Diffusion of Innovations: The Case of Suicide, International Organization 64: (Winter 2), pp Version. April 2 Appendix A: Suicide Group Linkages Figure 3 in the article raises the question of connections between non-suicide adopters. To answer that question, I gathered data from MIPT on the connections between all groups. Combined with the data on adoption of suicide attacks, I created three variables for each group, a number of connections variable, a number of connections to suicide attack adopters variable, and a ratio of the two. The raw connections data shows that, on average, adopters have significantly more ties to other groups than non-adopters (verified through a t-test). More importantly, the ratio data shows the percentage of connections a given group has to groups that have adopted suicide attacks. A score of.5 means 5% of a groups connections are to groups that have adopted suicide attacks, while a score of. means only % of a groups connections are to groups that have adopted. The mean score for suicide adopters is.43, compared to.9 for non-adopters. The difference is significant according to a t-test. Appendix A, Table Mean Percentage Of Connections Suicide With Suicide Terror Adopters Total Just looking at the number of connections to other groups reveals the same thing. Those groups that adopt suicide attacks have more linkages than non-adopters Appendix A, Table 2 Suicide Mean Number Of Connections Total This shows there are significant differences in the connection patterns of suicide groups versus non-suicide groups. While this analysis does not cover the indirect connections shown in figure 3, it suggests, based on the available evidence, that figure 3 is not simply an artifact of the data on suicide attack adopters. However, there is an endogeneity problem in the data since it does not show the timing of connections between groups. It is therefore not possible to tell whether groups connected before or after they adopted suicide attacks. Thus, while this data is potentially useful, care has to be taken in

2 interpretation. Appendix A, Table 3 below shows that the results from Models 3 and 4 in Table are consistent when including the number of connections variable. Appendix A, Table 3: Including A Network Connections Variable Into Models 3 and 4 of Table Connections Model Paper Table, Model 3) Connections Model 2 Paper Table, Model 4) Organizational Age -.97*** -.55 (.284) (.34) Religious.788*** 3.68*** (.638) (.922) Religious * Organizational Age -.6** (.52) Number of Connections to Other.399***.4*** Groups (.4) (.7) Communist/Socialist (.945) (.986) Leftist (.36) (.447) Nationalist/Separatist (.586) (.583) Other.4.55 (.29) (.77) Lebanon 2.8** 2.4** (.57) (.884) Iraq War (.693) (.747) Israel (.844) (.943) Link to Al Qaeda (.823) (.853) Constant -2.99*** *** (.742) (.839) Observations Degrees of Freedom. 2. Wald Chi log pseudolikelihood Pseudo R Robust standard errors in parentheses. *** p<., ** p<.5, * p<. 2

3 Appendix A, Table 4: Description of Links Between Groups in Figure 3 Name Location Initial Link Abu Hafs Al-Masri Brigade Nature of Link Europe Al Qaeda Al-Aarifeen Al Qaeda Indirect Al-Aqsa Martyrs Brigade Al-Bara Bin Malek Brigades Al-Fatah Al-Islambouli Brigades of Al-Qaeda Israel al-fatah Iraq Al Qaeda Israel Russia; Hamas; Palestinian Islamic Jihad Al Qaeda; Riyad us-saliheyn Martyrs' Brigade Indirect Al-Mansoorain Kashmir Al Qaeda Indirect Al-Qaeda World Hezbollah Al-Qaeda in the Arabian Peninsula (AP) Al-Qaeda Organization in the Land of the Two Rivers Al-Qanoon Al-Quds Brigades Saudi Arabia Al Qaeda Middle East Al Qaeda Israel Al Qaeda; Lashkar-e- Jhangvi (LeJ) Palestinian Islamic Jihad Citation (Bloom 25; Pape 25; Pedahzur 25). (National Commission on Terrorist Attacks upon the United States 24). Further Grouped (Ricolfi 25). Grouped With Europe al-fatah Iraq ME Iraq Palestinian Islamic Jihad Amal Lebanon Hezbollah (Kramer 99). Ansar Al-Islam Iraq Al Qaeda Iraq Ansar Allah Lebanon Hezbollah Ansar al-sunnah Iraq Al Qaeda Iraq Also note that the suicide attack adopter classifications in Figure 3 are based on a maximalist interpretation of suicide attack adoption. 3

4 Army Armed Islamic Group (GIA) Army for the Liberation of Kurdistan Black Widows DHKP/C (Revolutionary People's Liberation Party/Front) Algeria None Lebanon None Chechnya/R ussia Riyad us- Saliheyn Martyrs' Brigade; Al Qaeda Turkey PKK Indirect Hamas Israel Hezbollah Harakat ul-mujahidin (HuM) Hezbollah Hizbul Mujahideen (HM) Kashmir Al Qaeda Lebanon Kashmir Iranian Government (Lashkar-e- Toiba) (Pedahzur 25; (Aboul-Enein 25, 9; Dolnik and Bhattacharjee 22, 9; Levitt 24; Ricolfi 25, 9-92; Schweitzer 22, 3). (Kramer 99). Iraqi Liberation Army Lebanon None Islami Chhatra Shibir Al Qaeda-related Bangladesh (ICS) groups Islamic Army in Iraq Islamic Glory Brigades in the Land of the Nile Islamic Jihad Group (Uzbekistan) Jamatul Mujahedin Bangladesh Iraq Al Qaeda-related groups Egypt Al Qaeda Indirect Uzbekistan Bangladesh Islamic Movement of Uzbekistan (IMU); Al Qaeda Al Qaedainspired; Islami Chhatra Shibir (ICS) Indirect Jemaah Islamiya (JI) Indonesia Al Qaeda Jenin Martyrs' Brigade Israel Hamas (Pedahzur 25; (Hoffman and McCormick 24). Chehnyan PIJ Bangladesh Iraq ME Uzbekistan Bangladesh Hamas 4

5 Jihad Pegah Iraq Al Qaeda Indirect Jund Al-Sham (Army of the Levant) Kurdistan Workers' Party (PKK) Lashkar-e-Jhangvi (LeJ) Lebanese National Resistance Front Liberation Tigers of Tamil Eelam (LTTE) Mujahideen Shura Council Palestinian Islamic Jihad (PIJ) People's Liberation Army of Kurdistan Popular Front for the Liberation of Palestine Riyad us-saliheyn Martyrs' Brigade Secret Organization of al-qaeda in Europe Soldiers of the Prophet's Companions Syrian-Inspired (Assorted) Lebanon; Qatar Al Qaeda Turkey (Pedahzur 25). Al Qaeda; Tailban Lebanon Hezbollah Indirect Sri Lanka Hezbollah Iraq Al Qaeda Israel Hezbollah Turkey PKK Israel Chechnya/R ussia Hamas; Palestinian Islamic Jihad Al Qaeda Indirect Europe Al Qaeda Iraq Al Qaeda Indirect (Pedahzur 25; (Hoffman and McCormick 24; Pape 25). (Aboul-Enein 25, 9; Dolnik and Bhattacharjee 22, 9; Levitt 24; Ricolfi 25, 9-92; Schweitzer 22, 3). (Bloom 25; Pedahzur 25). (Pedahzur 25; Lebanon Hezbollah Indirect (Kramer 99). Taliban Afghanistan Al Qaeda Tanzim Israel al-fatah (Hoffman and McCormick 24; Iraq ME Iraq Hamas Chechnya Europe Iraq al-fatah 5

6 Tawhid and Jihad Iraq Al Qaeda Iraq APPENDIX A SOURCES Aboul-Enein, Youssef. 25. Hamas: A Further Exploration of Jihadist Tactics Strategic Insights IV(9): 2. Bloom, Mia. 25. Dying to Kill: The Allure of Suicide Terror. New York: Columbia University Press. Dolnik, Adam, and Anjali Bhattacharjee. 22. Hamas: Suicide Bombings, Rockets, or Wmd? and Political Violence 4(3): Hoffman, Bruce, and Gordon H. McCormick. 24., Signaling, and Suicide Attack. Studies in Conflict and 27(4): Kramer, Martin S. 99. The Moral Logic of Hizballah. In Origins of : Psychologies, Ideologies, Theologies, States of Mind, edited by Walter Reich, Cambridge, UK: Cambridge University Press. Levitt, Matthew. 24. Hamas from Cradle to Grave. Middle East Quarterly (). National Commission on Terrorist Attacks upon the United States. 24. The 9/ Commission Report: Final Report of the National Commission on Terrorist Attacks Upon the United States. Authorized ed. New York: Norton. Pape, Robert A. 25. Dying to Win: The Strategic Logic of Suicide. st ed. New York: Random House. Pedahzur, Ami. 25. Suicide. Cambridge ; Malden, MA: Polity. Ricolfi, Luca. 25. Palestinians, In Making Sense of Suicide Missions, edited by Diego Gambetta, New York: Oxford University Press. Schweitzer, Yoram. 22. Suicide : Development and Characteristics. Lecture presented at the International Conference on Countering Suicide, Table : ory: Groups. Available from (Accessed July-August, 6

7 Appendix B, Table : Summary Statistics Summary Statistics for Universe of Cases Used In Models -4 Minimum Maximum P5 Mean Standard Deviation Use of Suicide Organizational Age Religion Religion * Organizational Age Communist/Socialist Leftist Nationalist/Separatist Other Lebanon Iraq War Israel Al Qaeda Link

8 Appendix C: Limitations, ions for Future Research, and Other Extensions There are important limitations to this analysis that could serve as a starting point for future research. In particular, there are some potential biases within the MIPT-RAND data. Like the Correlates of War and Militarized Interstate Dispute datasets, the MIPT- RAND data is imperfect. However, use of this data is consistent with recent peer reviewed work (Asal and Rethemeyer 28; Berman and Laitin 28). One important limitation is the switch from only coding international terrorist groups from to coding both domestic and international groups from Purely internal groups that ceased to operate prior to 998 were excluded by MIPT, potentially biasing the results. A factor mitigating against this concern is that the coders appear to have been generous in their interpretation of international, including groups if their attacks affected foreign nationals. Those excluded, if purely domestic, are also probably disproportionately likely to be coded as nationalist/separatist. They are also all nonadopters, since the dataset on the universe of suicide attacks is independent of the MIPT- RAND data. This suggests inclusion would only strengthen the results. While it is possible those domestic groups that existed prior to 998 were all younger non-adopters, they could have also been involved in long-term struggles that simply ended. Therefore, it is not possible to draw inferences about the likely age of excluded groups; the bias was not systematic in a way that helped the findings of this paper. However, given that some bias in the data may exist, even if the exact direction is unknown, I tried to hedge against bias in two additional ways. First, I dropped all groups that did not conduct attacks in the post-997 period and re-ran the results. They are consistent with the findings in the paper. Second, cross-checking the MIPT-RAND universe with Sambanis civil war data revealed only 9 definitively excluded rebel groups (Sambanis 24). This low number suggests the extent of bias is likely tolerable. Future research could create and identify more robust measures of alliances between groups as well as clarify some of the hard choices made about which groups conducted certain attacks. Unfortunately, the MIPT-RAND data is no longer being updated (it has now migrated to the START database hosted by the University of Maryland), so research efforts will have to shift to other databases. Additionally, selection models focusing on the link between adoption and casualties or count models looking at the number of times groups used suicide attacks might also help shed light on these issues. This discussion of suicide bombing may also be helpful for understanding the diffusion of practices. New research by Bloom suggests suicide attacks are now a more regular practice for terrorist groups with the capacity to adopt (Bloom, 28). However, this process seems to be just beginning for suicide bombing so it is outside the scope of this paper. Though this idea is very tentative, the level of organizational capital required to adopt innovations may also shape what practices groups and societies find attractive, in a modified form, by constraining what is possible. Below are additional extension models designed to show the robustness of the results in the article. The first, Appendix C, Table, restricts the universe of cases to those groups that definitively engaged in attacks after 98. The purpose of this model is to make sure that the results are not biased by potentially including groups that went out 8

9 of existence prior to the beginning of the suicide terrorism era. As you can see below, the results are consistent both with and without the organizational age * religion interaction term. Appendix C, Table : Ensuring the results are not an artifact of pre-98 groups Post-98 Model (Extension of Paper Table, Model 3) Post-98 Model 2 (Extension of Paper Table, Model 4) Organizational Age -.45** (.225) (.27) Religious.865*** 3.242*** (.67) (.92) Religious * Organizational Age -.983** (.436) Communist/Socialist (.877) (.85) Leftist.68.5 (.375) (.5) Nationalist/Separatist (.536) (.546) Other (.78) (.93) Lebanon.63**.643** (.796) (.742) Iraq War (.67) (.72) Israel.94**.972* (.954) (.33) Link to Al Qaeda.85**.766** (.752) (.794) Constant *** *** (.697) (.842) Observations Degrees of Freedom.. Wald Chi log pseudolikelihood Pseudo R Robust standard errors in parentheses. *** p<., ** p<.5, * p<. Additionally, as noted above, MIPT switched in 998 from only coding international terrorism incidents to coding both domestic and international terrorism incidents. In Appendix C, Table 2 below I show the results restricting to only post-997 cases. I also add an organizational size variable drawn from Asal and Rethemeyer (28) to show the results are still consistent even including their key variable of interest. The organizational size variable is significant and in the predicted direction as well. 9

10 Appendix C, Table 2: Ensuring the results are consistent with post-997 groups and organizational size variable Organizational Size Model Paper Table, Model 3) Organizational Size Model 2 Paper Table, Model 4) Organizational Age -.447** -.34 (.22) (.24) Religious.329** 2.95*** (.62) (.937) Religious * Organizational Age -.9*** (.458) Organizational Size.682***.753*** (.247) (.243) Communist/Socialist (.989) (.6) Leftist 3.548** 3.39* (.483) (.77) Nationalist/Separatist (.7) (.73) Other () () Lebanon ** (.449) (.97) Iraq War (.787) (.854) Israel 2.57** 2.59** (.884) (.6) Link to Al Qaeda.889**.75** (.848) (.87) Constant -3.7*** *** (.99) (.964) Observations Degrees of Freedom.. Wald Chi log pseudolikelihood Pseudo R Robust standard errors in parentheses. *** p<., ** p<.5, * p<. Finally, to ensure that the results, especially with the interaction term, are not just a function of the logit model, Appendix C, Table 3 below shows the results are consistent with a regression specification as well. Replication of Figure with a regression specification shows the substantive results are also very similar.

11 Appendix C, Table 3: Ensuring the results are consistent with a regression specification Regression Model Paper Table, Model 3) Regression Model 2 Paper Table, Model 4) Organizational Age -.396** (.66) (.48) Religious.223***.447*** (.684) (.2) Religious * Organizational Age -.36*** (.432) Communist/Socialist.6.77 (.662) (.657) Leftist (.832) (.883) Nationalist/Separatist (.596) (.58) Other (.768) (.755) Lebanon.52.5 (.6) (.6) Iraq War (.3) (.34) Israel.3*.287* (.55) (.59) Link to Al Qaeda.366***.38** (.34) (.37) Constant (.656) (.67) Observations R-squared F 5.74 (, 222) 5.29 (, 22) Prob > F Root MSE Robust standard errors in parentheses. *** p<., ** p<.5, * p<. APPENDIX C SOURCES Asal, Victor and R. Karl Rethemeyer. 28. The Nature of the Beast: Organizational Structures and the Lethality of Terrorist Attacks. Journal of Politics 7(2):

Main State Actor/ Adversary. Afghan & Coalition forces. Afghan & Coalition forces. Afghan & Coalition forces. Afghan & Coalition forces

Main State Actor/ Adversary. Afghan & Coalition forces. Afghan & Coalition forces. Afghan & Coalition forces. Afghan & Coalition forces 1 Who Is a Rebel? Typology and Rebel Groups in the Contemporary Middle East Hyeran Jo Rotem Dvir Yvette Isidori Table 1. Rebel groups operating in the Middle East between 1946 and 2015 Taliban 1994-Present