This project involves the analysis of terror network members in terms of their personal, social, and educational background. Investigators used data from START's Global Terrorism Database to conduct a two-mode network analysis to identify clusters of terrorist groups connected by their engagement in similar terrorist actions, and conversely, clusters of terrorist actions (e.g., hostage-taking, bombing, suicide bombing, etc.) connected by the terrorist groups. With the defined clusters, investigators collected additional data specifically about these clusters to be able to predict future groups and events.
The project identified a number of characteristics that make terrorist organizations more likely to be lethal and more likely to be prolifically lethal. We found that certain ideologies make it more likely (or less likely) that organizations will kill in the first place. Specifically, organizations that are religious or combine religious ideology with ethno-nationalist ideology are more likely to kill and to kill a lot while organizations that have an environmental, animal rights, leftist or anarchist ideology are less likely to kill. Organizations that are large are more likely to kill and kill a lot while state sponsorship makes organizations more likely to kill but not more likely to kill a lot. Organizations that are small and young as well as those that have not carried out a lot of attacks are less likely to kill. Organizations that are more networked are much more likely to kill a great number of people. We also identified factors that make it more or less likely that an organization will target or attack US interests. Islamist ideology is important for targeting (that is saying that you would like to attack US interests) but not for actually attacking US interests. US troop presence encourages threats but when it comes to attacks US troop presence only has an impact in non-democratic countries. Finally, organizations that are highly networked are much more likely to attack US interests.
The project married several different methodologies. We coded data from the MIPT semantic data base into numerical data of two types. One type of data was coded in regular spreadsheet format with the unit of analysis being the organization. The other kind of data was network data where we coded relations in a matrix format that then produced network variables that were integrated into the organizational level data set. The resulting data set (Big, Allied, and Dangerous, or BAAD I) captured organizational and relational characteristics of 400 different terrorist organizations for the years 1998-2005. This data was then used in a variety of statistical analyses including logit and negative binomial regression depending on the nature of the dependent variable we were analyzing.