– Executive Summary –

This thesis explores the development and application of probabilistic risk assessment tools that predict a domestic terrorist attack’s possible human consequence (i.e., civilians killed or wounded) if conducted against a mass gathering or special event. Special events make up a significant part of the fabric of any healthy community yet are commonly targeted by violent extremists worldwide. The shared vulnerability of mass gatherings across the United States warrants a better suite of risk assessment tools for public safety planners and practitioners than what is currently available. This study contends that a more realistic understanding of these consequences can help inform the deployment of limited resources and personnel to mitigate an attack’s lethal outcomes.

To model terrorism consequences, the research must first determine the relationship between common attack variables that influence their outcomes. This research examines the relationship between common attack types, the overall crowd size at the time of the attack, and the ultimate killed and wounded rates resulting from the attack. The five common attack types include attacks using improvised explosive devices (IEDs), firearms, vehicles as a weapon (ramming vehicles), edged weapons, and incendiary devices. The data used to analyze common attack types and their resulting casualties come from actual terror attacks documented in the Global Terrorism Database (GTD).[1] The crowd size for each incident selected for analysis represents documented, publicly available, open-source information.

The literature review identifies how terrorism databases evolved and explains the selection of the GTD as the most comprehensive and appropriate dataset available for this study. The literature review also examines how open-source information is now a standard component of terrorism research, which lends credibility to the research design. The available literature identifies a gap in risk assessment tools for public safety planners, which highlights the need for continued and expanded quantitative research into terrorism risk and consequences.

After applying a series of exclusions and removing outliers, only two attack types (IEDs and firearms) yielded a large enough sample size for regression analysis. The statistical significance provided by regression analysis determines the confidence level to answer this study’s secondary question: How can the results be used to develop a consequence model that predicts future lethality rates of terror attacks against mass gatherings? The results can be evaluated for their applicability as a consequence model only if a relationship exists.

The findings reveal a weak correlation between an IED attack and the overall crowd size’s influence over the resulting casualties. Attacks involving firearms show a slightly stronger (moderate) correlation. Although the findings of this study indicate a lack of correlation between crowd size and casualty rates within the IED and active shooter attack vectors, the results still have merit. The descriptive analysis provides helpful markers that can be used to set parameters for Monte Carlo simulations. Monte Carlo simulations are computer-generated random samplings within a predefined set of boundaries. The historical data and analysis completed by this research can be used to establish those boundaries for the Monte Carlo simulations. Those simulations can then generate a large enough sample size for all five attack types within realistic, data-driven limits. Although this method may not be as accurate as creating a consequence model solely based on historical events, the use of Monte Carlo simulation is a good compromise and a common alternative when sample sizes are small. The appendices provide a prototype model using Monte Carlo simulation, as well as the detailed results of all statistical analyses completed in this study.

The implications of this thesis relate to the potential of the continued study and advancement of this topic that is only possible with more rigorous application of academic research and statistical analysis. Although this project stopped short of a fully developed consequence model for all attack types considered, the findings reveal some powerful benefits of thoroughly researched, tested, and vetted predictive analysis tools. Four key implications were identified during this research. They serve as recommended topics for further study:

(1)               The overall crowd size matters less than hypothesized. Instead, the future pursuit of a probabilistic consequence model would benefit from research into other crowd dynamics, such as the maximum effective range of the attack type, crowd density, and ease of egress. An assessment into each of these components’ impact on the attack type’s outcomes can be incorporated into a more holistic model that factors in more than just the crowd size.

(2)               Pockets of risk within an event are just as important as an overall risk score. A model that can assess individual pockets of risk (to include variables like crowd size, crowd density, and ease of egress at those pockets) within a single event is worth pursuing since most attacks observed in this study only affected a fraction of the overall crowd. Providing these professionals with an easily accessible, predictive analysis tool that identifies the most probable outcomes of an attack within those individual pockets of risk supports better preparedness through data-driven prevention and mitigation strategies.

(3)               Outlier events must be accounted for and merit additional research. A clear implication of this study is the influence of outlier events. This research defined an outlier as a terrorist incident that exceeded 2.5 standard deviations (2.5σ) from the norm. The findings indicate the value of appropriately accounting for outliers, as evidenced in the differences between initial and adjusted results. Furthermore, analyzing the variables and tactics that create casualties at such a high rate can be incorporated into a consequence model as a maximum of maximums. The result would be a simulation that would set realistic ranges of the most likely casualty rates while still noting potential worst-case scenarios based on historical data.

(4)               Prioritize the use of artificial intelligence in public sector research. The role and value of artificial intelligence (AI) in the future of public safety research cannot be understated. A properly written algorithm implemented as an AI function could scour the same open-source information to pair crowd sizes with GTD casualty rates in a fraction of the time it took this research team. In this way, AI can supplement this type of qualitative research, which reduces the overall burden on staff members. Such an AI program can also be modified to search for multiple characteristics of crowds or terrorist attacks in a way that is constantly learning and improving. Well-resourced organizations that host and maintain extensive databases, such as the GTD, are ideal laboratories to experiment and refine AI’s potential use in public safety research.

The overarching recommendation is to advance this research so that it may lead to a widely accepted and easily accessible casualty prediction tool for public safety practitioners. Just as the U.S. military developed and maintains a set of risk assessment calculations and products, the same set of resources should be created and implemented to help mitigate domestic threats against special events and mass gatherings. A pathway to implementation includes drawing on preexisting resources and networks because most of the pieces already exist within the Department of Homeland Security (DHS). The DHS Science and Technology Directorate oversees a Centers of Excellence network that can function as the academic, research arm of developing and vetting risk assessment tools for mass gatherings. Once ready, the DHS Special Event Working Group is an ideal host for providing broad access and training of these risk assessment tools through the DHS Homeland Security Information Network.

The threats against soft targets are unlikely to lessen. However, their consequences can be more effectively mitigated with properly researched, tested, and implemented risk analysis products. Regardless of whether the approach used in this thesis is followed or another methodology is chosen, the continued focus on developing easily accessible and realistic consequence models is greatly needed in the homeland security community. A clear understanding of the potential consequences is required for risk to be appropriately managed, mitigated, transferred, or accepted. Focusing the efforts of researchers, statisticians, and policymakers to develop and maintain a domestic terrorism consequence model is sorely needed but well within reach.

[1] “Global Terrorism Database,” National Consortium for the Study of Terrorism and Responses to Terrorism, accessed November 27, 2019, https://www.start.umd.edu/gtd/.

Leave a Comment

Your email address will not be published. Required fields are marked *

Scroll to Top