ASSESSING THE VULNERABILITY IN TARGETS OF LETHAL DOMESTIC EXTREMISM By Sarah St. George A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of Criminal Justice- Doctor of Philosophy 2017 ABSTRACT ASSESSING THE VULNERABILITY IN TARGETS OF LETHAL DOMESTIC EXTREMISM By Sarah St. George Domestic terrorism is a significant issue of concern and in recent years there has been a notable rise in deadly attacks committed by extremists. Extensive government resources have been allocated to prevent domestic terrorist attacks and to harden vulnerable targets. Scholars have conducted numerous studies on domestic terrorism and target selection. However, very little is known about target vulnerability and more specifically about the relationship between target vulnerability, victims and lethality. This dissertation fills this gap and examines the victims of lethal domestic extremist attacks and the situational context that surrounds these incidents. Eight separate measures of vulnerability are examined that are derived from Clarke and Newman (2006)’s EVIL DONE framework. This dissertation expands this framework by applying the framework to human targets and considering the victim and the situational context the victim was in at the time of the attack. The ideological motivation for each attack is examined to determine if there are differences in vulnerability based on motive (ideologically motivated homicides vs. non-ideologically motivated homicides) as well as the ideology of the suspect (right-wing vs. jihadist). Several factors relating to the victim and suspect are also examined. This project utilizes the Extremist Crimes Database (ECDB) and examines lethal incidents of domestic extremism that occurred between 1990-2014. This research makes several important contributions by filling a gap in terrorism literature and helps policymakers with target prioritization. Copyright by SARAH ST. GEORGE 2017 This dissertation is dedicated to my husband, Matthew. He has patiently been by my side throughout this long and tedious process. He has made me a better person and I am lucky to have him in my life. I look forward to the journey we have ahead and I will forever be a proud member of Team Scuba. iv ACKNOWLEDGEMENTS I would first like to acknowledge the National Consortium for the Study of Terrorism and Responses to Terrorism (START) for the funding and resources provided by them through the START Pre-Doctoral Fellowship program. This support has greatly helped with the completion of the dissertation. There are many individuals who have guided me in the completion of this dissertation. I would like to thank Dr. Thomas Holt, Dr. Edmund McGarrell and Dr. Janka Thomas for their helpful feedback as I worked from early stages of developing my topic up until its conclusion. I would also like to thank the chair of my dissertation committee, Dr. Steve Chermak who has given me the opportunity to work on the Extremist Crime Database and mentored me throughout graduate school. I would also like to thank Dr. Joshua Freilich and Dr. Jeff Gruenewald for providing feedback on EVIL DONE and measuring target vulnerability. Additionally, I would like to thank Brent Klein for helping with intercoder reliability. I would also like to thank my family and friends for being supportive throughout my graduate career and encouraging me through difficult times. Finally, I would like to thank my late cat, Cookie, for being by my side and on my lap day after day as I worked on this dissertation. v TABLE OF CONTENTS LIST OF TABLES ....................................................................................................................... viii Chapter 1: Introduction ....................................................................................................................1 Relevance of Topic ..........................................................................................................................3 Threat of Domestic Terrorism ..................................................................................................4 Theoretical Contributions .........................................................................................................5 Methodological Contributions ..................................................................................................9 Terrorism Prevention and Policy Contributions .....................................................................12 Research Questions ................................................................................................................13 Proposed Project .........................................................................................................................14 Outline of Chapters ....................................................................................................................15 Chapter 2: Theoretical Framework and Relevant Literature .........................................................17 Introduction ................................................................................................................................17 Theoretical Foundations .............................................................................................................17 Terrorism vs. Traditional Crimes ...........................................................................................17 Rational Choice Theory..........................................................................................................20 Routine Activities Theory ......................................................................................................25 Situational Crime Prevention .................................................................................................29 Vulnerability and EVIL DONE .....................................................................................................31 Vulnerability ...........................................................................................................................31 EVIL DONE and Prior Research ...........................................................................................31 EVIL DONE Framework .......................................................................................................36 Lethality and Target Choice .......................................................................................................40 Ideological Factors .................................................................................................................41 Suspect Factors .......................................................................................................................42 Victim Factors ........................................................................................................................43 Hypotheses .....................................................................................................................................46 Conclusion ..................................................................................................................................49 Chapter 3: Methodology ................................................................................................................50 Data ............................................................................................................................................50 Inclusion Criteria ........................................................................................................................52 Coding ........................................................................................................................................53 Dependent Variable ....................................................................................................................54 EVIL DONE Operationalization ................................................................................................55 Coding and Reliability............................................................................................................61 Victim, Suspect and Ideological Variables ................................................................................62 Data Analysis .............................................................................................................................69 Conclusion ..................................................................................................................................70 vi Chapter 4: EVIL DONE Variables and Other Vulnerability Factors ............................................71 Descriptive Findings for Ideology, Suspect and Victim Characteristics ...................................72 EVIL DONE Descriptive Findings ............................................................................................77 EVIL DONE and Ideological Motivation ..................................................................................83 EVIL DONE and Ideological Affiliation ...................................................................................94 EVIL DONE and Suspect Characteristics ................................................................................100 EVIL DONE and Victim Characteristics .................................................................................113 Chapter 5: EVIL DONE Vulnerability and Lethality: Multivariate Analysis .............................126 Lethality Bivariate Results .......................................................................................................126 Multivariate Results: All Incidents ..........................................................................................135 Multivariate Results: Ideological v. Non-Ideological Incidents ..............................................146 Multivariate Results: Right-Wing v. Jihadist Incidents ...........................................................150 Chapter 6: Discussion ..................................................................................................................154 Review of Major Findings........................................................................................................154 Policy Recommendations and Research Implications .............................................................159 Research Limitations ................................................................................................................163 Future Research ........................................................................................................................164 REFERENCES ............................................................................................................................167 vii LIST OF TABLES Table 1.1 EVIL DONE Components ...............................................................................................2 Table 2.1 Hypotheses .....................................................................................................................48 Table 3.1 Deaths in Extremist Incidents 1990-2014 ......................................................................54 Table 3.2 EVIL DONE Codebook (n=434) ...................................................................................57 Table 3.3 Reliability Analysis of EVIL DONE .............................................................................61 Table 3.4 Victim, Suspect and Ideology Codebooks .....................................................................64 Table 3.5 Incident Motivation Frequencies ...................................................................................68 Table 4.1 Descriptive Findings for Independent Variables (n=434) .............................................74 Table 4.2 EVIL DONE Descriptive Results (n=434) ....................................................................79 Table 4.3 EVIL DONE and Ideological Motivation (n=434) ........................................................86 Table 4.4 EVIL DONE and Suspect Ideology (n=434) .................................................................96 Table 4.5 EVIL DONE and Suspect Characteristics (n=434) .....................................................102 Table 4.6 EVIL DONE and Suspect Characteristics 2 (n=434) ..................................................105 Table 4.7 EVIL DONE and Suspect Characteristics 3 (n=434) ..................................................108 Table 4.8 EVIL DONE and Victim Characteristics (n=434) .......................................................115 Table 4.9 EVIL DONE and Victim Characteristics 2 (n=434) ....................................................118 Table 4.10 EVIL DONE and Victim Characteristics 3 (n=434) ..................................................121 Table 5.1 EVIL DONE Variables and Lethality (n=434) ............................................................129 Table 5.2 Independent Variables and Lethality (n=434) .............................................................132 viii Table 5.3 Multivariate Results: EVIL DONE and Lethality (n=434) .........................................136 Table 5.4 Multivariate Results: Lethality and Ideology (n=434) ................................................138 Table 5.5 Multivariate Results: Lethality and Suspect Characteristics (n=434) .........................139 Table 5.6 Multivariate Results: Lethality and Victim Characteristics (n=434) ...........................140 Table 5.7 Multivariate Results: Lethality and Victim Characteristics with Occupation (n=246) ......................................................................................................................................................141 Table 5.8 Multivariate Results: Equation with Previously Significant Variables (n=434) .........143 Table 5.9 Multivariate Results: Equation with Previously Significant Variables with Victim Occupation(n=246) ......................................................................................................................145 Table 5.10 Multivariate Results: EVIL DONE Variables and Lethality for Ideological Incidents (n=256).........................................................................................................................................148 Table 5.11 Multivariate Results: EVIL DONE Variables and Lethality for Non-Ideological Incidents (n=199) .........................................................................................................................149 Table 5.12 Multivariate Results: EVIL DONE Variables and Lethality for Right Wing Incidents (n=374).........................................................................................................................................151 Table 5.13 Multivariate Results: EVIL DONE and Lethality for Jihadist Incidents (n=60) .......152 ix Chapter 1: Introduction The purpose of this study is to better understand the relationship between vulnerability and lethality in domestic extremist incidents. Why is lethality high in some incidents of domestic extremist homicide and not in others? Is lethality higher in incidents that are ideologically motivated or those that are non-ideologically motivated? Is lethality higher in attacks committed by right-wing extremists or attacks committed by jihadists? What factors related to the victims, suspects and location influence the likelihood that more people are killed in an attack? These questions are answered through the examination of vulnerability utilizing variables within the EVIL DONE framework as well as several factors related to ideology, victims, and suspects. Specifically, this study examines variations in victim/target characteristics of lethal incidents, and how these characteristics are related to the number of deaths in an extremist attack. This project assesses gaps in the literature and provides a better understanding of vulnerability and lethality in target selection by domestic extremists. This study seeks to examine vulnerability by using an innovative framework that was developed to identify vulnerable targets. Clarke and Newman (2006) created the EVIL DONE framework as a method to be able to assess and rank potential targets of terrorist attacks. This framework is a situational crime prevention (SCP) technique that can be used by policymakers to identify and rank targets and determine which targets are at the most risk. Clarke and Newman (2006) assume that terrorists are rational decision makers who weigh the costs and benefits of choosing particular targets. Vulnerable targets are more attractive because the benefits (destruction, death, publicity) of attacking them outweigh the risks (such as being caught or the attack failing). What makes one target more vulnerable to attack than other targets? Clarke and Newman's (2006) framework provides eight primary factors that terrorists may consider when 1 choosing a target to attack. These factors include how exposed it is (E), how vital it is (V), how iconic it is (I), how legitimate it is (L), how destructible it is (D), how occupied it is (O), how near it is to the perpetrator (N) and how easy it is (E) to carry out an attack on. Table 1.1 provides a summary of the eight components of EVIL DONE. Table 1.1 EVIL DONE Components Exposed How visible is the target; how much attention does the target attract? Vital How critical is target to day to day functioning of society? Iconic Does the target hold symbolic value? How will the public perceive this target being attacked; will the attack garner Legitimate support? Destructible How easily can the target be destroyed? Occupied Is the target occupied with people? Near How close is the target to the would-be terrorist? Easy How easy is it to successfully accomplish the mission? First, this study expands on the EVIL DONE framework by applying it to human targets. Clarke and Newman (2006) originally introduced this framework and applied it to physical targets. This study will expand on this and will determine what features make the targets and victims of extremist homicides vulnerable. Specifically, this study examines differences in EVIL DONE factors compared by victim characteristics (gender, race, age, occupation, victim-suspect relationship), suspect characteristics (gender, race, age, occupation, lone wolf offender, number of suspects, weapon choice) and ideology (incident ideology, suspect ideology). Second, this study extends the prior literature and analyses if EVIL DONE can also be used to determine whether the vulnerability of a target impacts the degree of lethality of terrorist incidents in the United States. Until now EVIL DONE has only been used for risk assessment purposes, to identify which targets are more likely to be attacked. 2 This dissertation accomplishes this by examining lethal domestic extremist incidents that have occurred in the United States between 1990 and 2014 and specifically looking at the individuals killed and the surrounding situational characteristics. The primary targets of most terrorist attacks are people (Le Vine, 1997) therefore it is important to consider the victim when examining target vulnerability. Since victims are the targets of these incidents, the term target is used interchangeably with the term victim. The variables this study examines are related to the victims, the suspect and the ideological motivation of the attack. Several new variables are created through this dissertation. These variables enhance scholarly value by allowing for the answering of several research questions related to terrorist targets and target vulnerability. This study’s results determine who is most vulnerable to being attacked and what characteristics of targets are most vulnerable. The identification of vulnerable populations allows for more effective policy and practical decisions to prioritize targets based on vulnerability (through mechanisms such as situational crime prevention (SCP) and target hardening). This first chapter is presented in several sections. First, the relevance of this study and its four main contributions are discussed. Second, the two primary research questions are stated. The research project is then proposed. Finally, the chapter concludes with an outline of the dissertation and remaining chapters. Relevance of Topic This study makes four important contributions to scholarly study of terrorism: (1) it addresses an important topic, domestic extremist homicide; (2) it makes several theoretical contributions; (3) it makes several methodological contributions; and (4) it makes significant terrorism prevention and policy contributions. 3 Threat of Domestic Terrorism First, this study addresses extremist violence which poses a significant social problem and is of concern to practitioners, academics and the general society. There are many noteworthy lethal incidents of extremist violence in recent years. For example, a mass shooting took place at the Inland Regional Center in San Bernardino, California on December 2nd, 2015 when Rizwan Farook and Tashfeen Malik killed 14 individuals using semi-automatic rifles and semi-automatic pistols. The suspects spent more than a year planning the attack and are believed to have become self-radicalized while living in the United States and to have committed the attack to further ISIS’ goals (Schuppe, 2015). Another notable example is the attack committed on June 17th, 2015, by Dylann Roof, who shot and killed nine people at the African American Emanuel AME church in Charleston, South Carolina. Roof had become a radicalized white supremacist after watching coverage of the Trayvon Martin case (Robles & McPhate, 2016). Six years prior to this, on November 11th, 2009, Nidal Malik Hasan, a U.S. army psychiatrist, shot and killed thirteen people and wounded more than thirty at Fort Hood in Texas. Hasan is believed to have become disillusioned with the U.S. military and at his trial he declared that his motive was to defend the lives of Taliban leadership in Afghanistan and to support the goals of the Al Qaeda in the Arabian Peninsula organization (Ferran, 2013). These three incidents showcase how different the victims and situational contexts can be surrounding extremist homicides in the United States and further demonstrates the potential for extreme lethality. There is additional evidence that also raises serious concerns. For example, there are far more domestic terrorist attacks than international attacks against the United States, with seven domestic attacks occurring for every one international attack (LaFree, Yang & Crenshaw, 2009). Authorities recognize the risk posed by extremist violence and the FBI lists domestic terrorism as 4 a top priority (FBI, 2015). Recently, there has been a notable spike in domestic attacks committed by lone wolf suspects and militia extremists who hold violent anti-government beliefs (DHS, 2014). Not only is the number of attacks concerning but it is also worth noting the number of fatalities caused by these attacks. Between 1970 and 2012 there were more than 2,600 terrorist attacks in the United States which resulted in more than 3,500 deaths (Miller, 2014).1 With this threat and concern about the lethality of terrorist attacks, a related question arises of whom or what is most vulnerable to being attacked? Internationally, 43% of the attacks against the United States are against military installations, whereas domestically 42.9% of attacks are against businesses, followed by 24.2% against private citizens or property (Muhlhausen & McNeill, 2011); however, there is evidence that the target choices of terrorists are shifting (Brandt & Sandler, 2010). The problem is that there has been very little scholarly attention put forth to understand factors predicting the lethality of incidents, and relatedly, there has been virtually no research on victims as targets of terrorism (Parkin, Freilich & Chermak, 2014; Parkin & Freilich, 2015). Theoretical Contributions Since September 11th 2001, there has been a substantial increase in funding for terrorism research (Kingshott, 2003; Sedgwick, 2004, Bellia, 2005) and an increase in studies conducted that examine international terrorism (Silke, 2001; Silke, 2008). However, there has been significantly less research undertaken on domestic terrorism. Current terrorism literature fails to fully examine the vulnerability of targets, and how target selection impacts incident lethality. Further, current literature fails to examine the victim and situational context for extremist 1 Of these fatalities, 86% were a result of the attacks of September 11th, 2001. 5 homicides that are ideologically motivated versus those that are committed for other reasons. This study fills these important gaps. Terrorism literature has begun to address the connection between terrorism and rational choice theory (Boba, 2009; Clarke & Newman, 2006; Dugan & Chenoweth, 2012; Dugan, LaFree & Piquero, 2005; Ekici, Ozkan, Celik & Maxfield, 2008; Freilich & Chermak, 2009; Gruenewald, Gruenewald & Klein, 2015; Parkin & Freilich, 2015; Pridemore & Freilich, 2007) and terrorism and situational crime prevention (SCP) (Clarke & Newman, 2006; Freilich & Newman, 2009). However, there are still significant gaps in this area. Research on criminal behavior generally, and terrorism specifically, has primarily been concerned with identifying the motivations of offenders, whereas Clarke and Newman (2006) argue that motivations are likely to vary and that they are not critical for a deep understanding of terrorist decision-making. Instead, they argue that it is necessary to look at the situational opportunities in which crime arises. Since motivations for crimes may vary significantly, the goal of SCP is simply to prevent opportunities thus neutralizing motivations. Clarke and Newman (2006) argue that there is little difference between terrorism and ordinary crimes in that the planning occurs in relation to opportunities. In the past, research has focused on the motivation of offenders, rather than the situational environment (Agnew, 1992; DeLisi, 2005; Gottfredson & Hirschi, 1990). Law enforcement has similarly focused on the individual (suspect) and put primary concern to catching terrorist(s) as opposed to reducing opportunities for terrorism to occur (Clarke & Newman, 2006). The first step in SCP is to identify the opportunities for crime (in this case a terrorist attack). To have a better understanding of opportunities, it is necessary to determine who is most 6 vulnerable to being attacked. Terrorists do not find all potential victims or targets to be equally vulnerable or attractive and they often choose the most vulnerable target to attack. This study is unique in specifically examining vulnerability of victims and their situational environments as it relates to lethality through a situational crime perspective. Vulnerable victims and targets are defined as those that are easiest to attack. Targets that are easy to attack are more attractive choices to terrorists since terrorists seek the maximum benefit for the least amount of risk or cost. What EVIL DONE factors are related to victim attributes and lethality? What situational factors increased the level of lethality in the incident? Why did more people die in some incidents than in others? This study is the first to start answering these questions. Another theoretical gap that this study will address is the current failure to fully utilize Clarke and Newman’s (2006) terrorist target vulnerability framework known as EVIL DONE. This framework is a SCP technique that was introduced a decade ago and has been neglected in the literature with the exception of some mostly descriptive applications (Boba, 2009; Ekici et al., 2008; Gruenewald et al., 2015) with only one exception (Paton, 2013). Boba (2009) has discussed EVIL DONE as a framework for assessing target vulnerability and provided an easy system for analyzing targets. Research has utilized this operationalization to assess potential targets in Turkey (Ekici et al., 2008). Paton (2013) examined EVIL DONE using multivariate analysis but examined incidents internationally. There has been only one application of this framework to domestic targets. Gruenewald, Gruenewald and Klein (2015) operationalized EVIL DONE and directly applied it to eco-terror incidents in the United States. They found support for terrorists generally picking targets that met the vulnerability criteria of EVIL DONE. This study will take EVIL DONE, a framework from situational crime prevention, and expand it in four ways. First, it will expand EVIL DONE by applying it to human targets, the 7 victims of prior extremist attacks. Clarke and Newman (2006) created this framework as a method for law enforcement to use to assess the vulnerability of physical targets. However, while buildings and locations can be vulnerable targets, they fail to consider human targets. Second, this study applies EVIL DONE to lethality to better understand vulnerabilities of victims in lethal domestic extremist incidents. This will help to see if EVIL DONE vulnerability characteristics can be used as predictors of lethality. In other words, this research seeks to see if certain vulnerability characteristics are more likely to be present in attacks with multiple deaths as compared to attacks with only one death. Third, this study will expand on Gruenewald and colleague’s prior study. Their study examined EVIL DONE and vulnerability in eco-terror attacks. This study will compare attacks committed by right-wing extremists and jihadist extremists. Finally, this study will see what differences exist in ideologically motivated attacks and non-ideologically motivated attacks. This study examines variations in victim/target characteristics of lethal incidents and how these characteristics are related to the number of deaths in an extremist attack. Prior research has examined lethality but not its relationship with victims and vulnerability. Research has examined characteristics of terrorist groups and found several that influence the lethality of an attack. First, groups that are organized in a hierarchical manner are more likely to be lethal than those that are more loosely organized (Heger, Jung & Wong, 2012). Second, groups that possess a religious or supernatural ideology are more likely to be lethal than secular groups (Asal & Rethemeyer, 2008). Third, groups who have more affiliations with other organizations are deadlier (Asal & Rethemeyer, 2008; Horowitz & Potter, 2013). Fourth, groups who are more centrally located in a network are deadlier (Caspi, Freilich & Chermak, 2012). Lethality is also much higher in incidents of suicide bombings (Nilsson, 2015; Pape, 2003) compared to deaths from more 8 conventional attacks. Further, lethality has been examined in terms of weapon choice and incidents that involve the use of guns or bombs are deadlier than those involving other weapons (Bogen & Jones, 2006). Research has, however, neglected to examine what characteristics of the victim and the situational circumstances are related to lethality. In other words, why are some incidents significantly more lethal than others-what is different about the targets in these incidents? This study will help answer this question. Methodological Contributions There are also key methodological contributions that this study makes. There are several methodological issues with terrorism research which include lack of data and lack of empirical and statistical analyses (Sageman, 2014; Silke, 2008; Silke, 2001). The first methodological contribution this study makes is using a unique database of open source information. Data collection for crime typically comes from official reports, victimization information and selfreport data. LaFree and Dugan (2004) describe the unique methodological and definitional issues that terrorism researchers face that those researching more common crimes do not encounter. Terrorism data provided from the government has several potential methodological red flags. Government data may be influenced by politics and may not be easily available or released for analysis due to the nature of terrorism and the fact that many suspected terrorists are not charged with terrorism directly but rather related criminal offenses (LaFree & Dugan, 2004). Despite difficulties gaining access to terrorists, some researchers have been able to interview past or current terrorists. Berko and Erez (2007) as well as Bloom (2005) were able to interview wouldbe suicide bombers; Post, Sprinzak and Denny (2003) interviewed incarcerated Middle Eastern terrorists; Alonzo (2006) interviewed IRA members and Horgan (2009) was able to interview individuals who had left extremist organizations. While interviews have been conducted, it is 9 generally difficult to gain access to known terrorists to conduct an interview (Silke, 2001) and many of the interviews are with a select sample such as incarcerated terrorists (Post, Sprinzak & Denny, 2003) or those who had already left the extremist lifestyle (Bubolz & Simi, 2015; Harris, Simi & Ligon, 2016; Horgan, 2009; Horgan, 2012). Importantly, most interviews are often only conducted once and long after the terrorist was active, thus raising both retrospective construction and “bounding” issues (Freilich & LaFree, 2016). There are also issues with victimization data and it is not readily available to researchers. The victimization data that is available is not very useful since being a victim of a terrorist act is very rare and victims may be chosen at random as compared to being a victim of a more common crime (LaFree & Dugan, 2004, see Freilich & Parkin, 2015; Parkin, 2012). Another limitation with self-report data is that terrorism is a relatively rare event and there is already a limited pool of potential interviewees. Documents may serve as an important source of selfreport data on terrorism. As an example of how self-report data can be used, McCauley (2003) examined the notebooks and letters that were left behind by the suicide bombers from the September 11th attacks. Due to the difficulties with self-report data, victimization data and government data, a significant portion of terrorism research involves the use of secondary data. Silke (2001; 2008) found that 80% of studies on terrorism involve the use of secondary data and archival data. These data can be inaccurate since secondary data often relies on media reports that may contain factual errors and many reports have some form of bias (Silke, 2001). Research has also shown that the frames used to portray terrorism threats in the media influence the perceptions citizens feel of the threat of terrorism (Haider-Markel, Joslyn & Al-Baghal, 2006). There have been several terrorist event databases that have been developed that utilize secondary data. Studies 10 using secondary terrorism data include those using data from the International Terrorism Attributes of Terrorist Events (ITERATE) database (Sandler & Enders, 2004), the Global Terrorism Database (GTD) (Borooah, 2009; Dugan, LaFree & Piquero, 2005; Drakos, 2010;; LaFree & Dugan, 2007; LaFree, 2010) and the Extremist Crime Database (ECDB) (Chermak, Freilich, Parkin & Lynch, 2012; Freilich et al., 2014; Gruenewald & Pridemore, 2012; Sullivan, Chermak, Wilson & Freilich, 2014; Sullivan, Freilich & Chermak, 2016).These databases have been used to answer a variety of research questions pertaining to international and domestic terrorism. This dissertation contributes to the domestic terrorism literature by utilizing a unique national database called the Extremist Crime Database (ECDB). This is a database funded by the Department of Homeland Security that contains detailed information of violent incidents committed by domestic extremists from 1990 to present (Freilich, Chermak, Belli, Gruenewald & Parkin, 2014). The ECDB is unique from other databases and contains extensive information (hundreds of variables) about these violent incidents. EVIL DONE factors have not previously been examined in lethal domestic extremist incidents and this database contains the important information that allows for this analysis. This project also innovatively devised several new variables related to target vulnerability to the ECDB and this will greatly enhance scholarly value and allow for the answering of several research questions related to terrorist targets and target vulnerability. Additionally, this research makes a significant contribution by using quantitative analyses to test target vulnerability at the multivariate level. Few studies have used empirical tests in analyzing terrorism and until recently most terrorism research has not been empirical (Silke, 2001; Silke, 2009). Additionally, few studies have examined target choice, and most of this 11 research has been descriptive (Bloom, 2005; Boba, 2009; Ekici et al., 2008; Gruenewald et al., 2015; Roislien & Roislien, 2010; Santifort, Sandler & Brandt, 2013). One reason for this may be that most scholars do not generate their own data (Silke, 2001). This research makes a significant contribution by operationalizing and testing the various characteristics of vulnerability in the EVIL DONE framework using multivariate techniques. This study takes EVIL DONE, a framework created for physical targets, and expands it to human targets. Scholars have failed to utilize EVIL DONE with a few key exceptions (Boba, 2009; Ekici et al., 2008; Gruenewald et al., 2015; Paton, 2013) but these exceptions have been limited in scope and only one study has examined EVIL DONE through multivariate analysis (Paton, 2013). Paton (2013) utilized Boba’s (2009) coding of EVIL DONE to apply it to incidents in the Global Terrorism Database (GTD). Paton’s study, however, was broader in scope and focused more on WMDs and did not examine domestic terrorism. This study is the first to examine domestic extremist incidents through examining the EVIL DONE framework at a multivariate level. Terrorism Prevention and Policy Contributions There have been many problems with counterterrorism resource allocation (Willis, 2007) and this study produces results that can be used to help in best allocating resources. These measures involve increasing the costs of terrorism (difficulty and risks) and reducing the rewards by allocating resources to the most vulnerable targets and populations. This study provides results that examine whether EVIL DONE factors vary across victim characteristics and if this is related to lethality. Characteristics of victims and targets are also examined to see their 12 relationship with lethality. By determining what situational environments had the greatest lethality, then more effective policies and practical decisions can be made. These decisions may include the use of SCP strategies that can help identify and harden targets that are the most vulnerable to being attacked. For example, the 25 techniques of SCP have been applied to terrorism (Clarke & Newman, 2006). Efforts such as controlling weapons technology, increasing security training, increasing formal surveillance, and helping to conceal or remove targets are all examples of techniques that may be implemented to protect vulnerable targets. Targets that are identified as being high risk can be prioritized and assessed to make changes to reduce their vulnerability and prevent acts of terrorism.2 Research Questions This dissertation examines two important research questions. The first question focuses on providing a better understanding of EVIL DONE vulnerability in prior lethal domestic extremist attacks. Differences in EVIL DONE factors are compared by ideology (suspect ideology, incident ideological motivation), victim characteristics (gender, race, age, occupation, victim-suspect relationship) and suspect characteristics (gender, race, age, occupation, lone wolf offender, number of suspects, weapon choice). The second research question examines the relationship between vulnerability and lethality. The relationship between all EVIL DONE vulnerability factors, victim factors, suspect factors and ideological factors and lethality is examined to see which factors are the strongest predictors of lethality. 2 A similar model entitled CRAVED was created to determine what products are likely to be stolen (Newman & Clarke, 2006) and it has been tested and found to be successful in designing products that are more difficult to steal. 13 Research Question 1 What factors of vulnerability are present in lethal domestic extremist incidents? a. How do these factors vary across ideologically motivated and non-ideologically motivated incidents? b. How do these factors vary across the ideological affiliation of the suspect? c. How do these factors vary based on suspect and victim characteristics? Research Question 2 What factors of vulnerability predict the lethality of an incident? a. Are these factors important when controlling for victim and suspect characteristics? b. What factors predict the lethality of ideologically motivated and non-ideologically motivated incidents? c. What factors predict the lethality of jihadist homicides and right-wing homicides? Proposed Project This project examines the relationship between target vulnerability and lethality. This study relies on rational choice theory, routine activities theory and situational crime prevention. A motivated offender is assumed and the focus is on understanding the situational context of each homicide. The data used for this project comes from the Extremist Crime Database which contains violent incidents committed by extremists from 1990present. Incidents that occurred between 1990-2014 and involve the death of at least one individual are examined. 14 This study expands on Clarke and Newman’s (2006) EVIL DONE vulnerability framework by applying it to human targets. EVIL DONE is operationalized and applied to lethal domestic extremist attacks. This study is unique because it examines the victims of lethal attacks as well as the situational context of the attack. Gruenewald and colleagues (2015) examined the distribution of EVIL DONE vulnerability factors for eco-terror attacks. This study expands on this research by examining these factors for right-wing extremist homicides and jihadist homicides. The descriptive results of EVIL DONE vulnerability variables are presented and bivariate analyses are conducted between EVIL DONE variables, ideology variables, suspect variables and victim variables. EVIL DONE vulnerability factors are then examined through multivariate analysis to see which of them are predictors of the lethality of an incident and if these predictors remain significant when controlling for victim, suspect and ideological characteristics. Outline of Chapters The second chapter of this dissertation provides a more detailed discussion of the theoretical framework and highlights existing literature relevant to the topic. It starts with a discussion of differences between terrorism and traditional crimes. This study then discusses rational choice theory, routine activities theory and situational crime prevention and their application to terrorism. EVIL DONE vulnerability factors are then presented. Then terrorism target research is examined as well as prior research on situational, suspect, suspect and ideology factors and how they relate to lethality. The third chapter addresses the methodology used and begins by discussing the procedures for gathering and coding the data, and the ECDB. The variables of EVIL DONE are 15 examined relative to how they have been operationalized by Clarke and Newman (2006), as well as by other scholars. The operationalization of EVIL DONE used in this study is presented and intercoder reliability results are presented. The victim, suspect and ideological variables are discussed and operationalized. The chapter concludes with a discussion of the data analysis techniques used. The fourth and fifth chapters present the study’s results. Chapter four presents descriptive statistics for the EVIL DONE variables and answers the first research question. Descriptive results are presented for all incidents and then separately based on ideology as well as victim and suspect characteristics. Chapter five presents multivariate results for the study and answers the second research question. Specifically, the fifth chapter presents bivariate logistic regression analyses that examine the relationship between vulnerability characteristics and lethality. The sixth chapter reviews the major research findings, discusses policy and research implications from the study, examines research limitations and discusses future research. 16 Chapter 2: Theoretical Framework and Relevant Literature Introduction This chapter is presented in several sections. First, this chapter compares terrorism to more traditional crimes. This is necessary to highlight similarities and differences between terrorism and traditional crimes and determine if criminological theories (originally applied to more conventional crimes) can be applied to terrorism. Next, the theoretical framework for this study is presented. This includes a discussion of the three environmental theories that this study relies on and how terrorism research has begun to utilize these theories. The three theories discussed are rational choice theory (RCT), routine activities theory (RAT), and situational crime prevention (SCP). The third section of this chapter discusses the introduction of EVIL DONE as a SCP technique. The eight elements of EVIL DONE are outlined as well how research has begun to apply this framework, but has failed to consider application to human targets. The fourth and final section of this chapter, discusses several factors related to victims, suspects and ideology and their relationship with lethality. Theoretical Foundations Terrorism vs. Traditional Crimes To examine the application of theory to terrorism, it is necessary to define terrorism and outline how terrorism relates to traditional crimes. There has been great debate surrounding the definition of terrorism (Crenshaw, 2000; Freilich, Chermak & Simone, 2009; Ruby, 2002; Schbley, 2003; Schmid, 2004; Schwartz, Dunkel & Waterman, 2009; Turk, 1982; Weinberg, Pedahzur & Hirsch-Hoefler, 2004). Terrorism has been defined in various contexts such as religion, war and politics (Schmid, 2004); and can generally be defined as the deliberate violent targeting of civilian targets to further an ideological goal (Freilich et al., 2009; Schwartz et al., 17 2009). Domestic terrorism is defined as violent acts that specifically impact a nation’s citizens, policy and property and seek to accomplish a political goal (Sandler & Lapan, 1988). There are several ways in which terrorism and traditional crimes are very similar (Clarke & Newman, 2006). First, both are interdisciplinary social constructions, and are acts disproportionately committed by young males (LaFree & Dugan, 2004).3 Second, the motivations for both are similar in that offenders may often be motivated by a sense of peer pressure or commit an act and try to find a sense of belonging or excitement. Not all terrorists, especially those who are lower in the organization, are working for a higher cause such as an ideological goal (Clarke & Newman, 2006; McGarrell, Freilich & Chermak, 2007). Third, many terrorists commit conventional crimes to fund their activities (Clarke & Newman, 2006; Hamm, 2007). They may sell drugs or other illegal goods or commit burglaries or robberies to garner funds to help the movement. Fourth, the planning and opportunity involved in terrorist acts is much like the planning involved in more traditional crimes (Clarke & Newman, 2006; Hamm, 2007). For example, many terrorist attacks require a great deal of planning but so do more traditional crimes such as bank robbery, fraud, kidnapping, or murder. Fifth, there is a misconception that terrorism usually involves large scale attacks. There are many nonviolent crimes committed by extremists, such as counterfeiting (Sullivan et al., 2014) and other financial crimes (Sullivan, Freilich & Chermak, 2016). Similarly, there are mass casualty crimes committed by traditional criminals. These crimes are often committed for motives such as personal disputes (spouse, family, friends, etc.) or are acts of workplace violence. Another similarity between criminal acts and terrorist acts is that they are often organized group 3 Some research examining domestic extremists has found that terrorist incidents are more commonly committed by older men (Smith & Morgan, 1994), especially among lone wolf offenders (Gruenewald, Chermak & Freilich, 2013). 18 activities. Since terrorism is similar to more traditional crimes, we can begin to apply criminological theories to the study of terrorism. Many criminology scholars have stressed positivist characteristics such as criminal dispositions or specific individual attributes that make one more prone to commit criminal acts. Early criminological theories argued that crime was caused by such factors as genetics, upbringing, personality and sociological influences (Clarke, 1992). For example, crime may be a result of low self-control (Gottfredson & Hirschi, 1990), the breakdown of social bonds (Hirschi & Gottfredson, 2003) or social learning (Akers, 1998). The problem with etiological theories is that they over predict and offer an “embarrassment of riches” (Matzah, 1964). It is also often difficult to determine what could be done to change the relationships between broad background, sociological problems such as economic disadvantage, discrimination, education and their relationship with crime. Similarly, law enforcement has put a great deal of focus into identifying terrorist threats and capturing or eliminating the offender, which are tactics that lend themselves to more positivist offender-based theories. However, terrorism research has found that there is no “terrorist personality” or particular illness that can differentiate the terrorist population from the general population (Taylor, 1988). Criminologists have begun to move past distal theories and have started to examine crime as a result of environmental factors and opportunity. These opportunity theories argue that crime is a result of deliberate choices individuals make and that these choices are dependent on available opportunities (Clarke, 1992). Environmental theories move the focus away from the offender and look at how situational characteristics impact the likelihood of a crime. 19 In recent years, many environmental theories have begun to be applied to terrorism research, such as rational choice theory (Carson, 2014; Carson, LaFree & Dugan, 2012; Dugan & Chenoweth, 2012; Dugan et al., 2005; Perry & Hasisi, 2015; Pridemore & Freilich, 2007), routine activities theory (Canetti-Nisim, Mesch & Pedahzur, 2006; Hamm, 2007; Parkin & Freilich, 2015) and situational crime prevention (Freilich & Newman, 2009; Clarke & Newman, 2006). Rational Choice Theory Rational choice theory (RCT) has its roots in classical criminology and was first introduced in the late 18th century (Beccaria, 1764). This theory argues that individuals have agency and make cogent decisions (Clarke, 1992; Cornish & Clarke, 1986) and that crime occurs as an interaction between a rational offender and the situational environment (Guerette, Stenius & McGloin, 2005). One of the seminal works of RCT is The Reasoning Criminal by Cornish and Clarke (1986) which reintroduced rational choice theory to the field and argued that offenders seek benefits from their criminal behavior and weigh the costs and benefits of these actions. Those who commit criminal acts are expected to follow the same decision-making process as those who choose not to participate in criminal acts. People are expected to follow the utility model, which states that individuals make rational decisions based on what they anticipate the gains or profits to be and that they attempt to minimize the losses or costs (Becker, 1968; Sampson, Piquero & Paternoster, 2002). Offenders make the decision to become involved in crime as well as the decision to which type of criminal act they will commit. In other words, if an offender sees the potential gains of committing a criminal act (profit, excitement, etc.) as 20 more enticing than the potential costs (failed attempt, getting caught, etc.) then they will choose to commit the act. The decisions that individuals make in terms of criminal involvement are influenced by their own experiences and learning (Cornish & Clarke, 1986). People make decisions based on the limited information they have and make these decisions under time constraints. The decisions they make are therefore the optimal choice to them at the time. Two individuals put into the same circumstances may make different choices; which is represented by the idea of bounded rationality. Offenders make decisions that are rational to them considering their own knowledge and experiences at the time. Further, these decisions are heavily based on opportunities that an offender has, and often are based upon incomplete information, which is why the framework is referred as “bounded” rationality. This moves the focus from the offender to the setting. The outcome of these crimes, the benefits or costs of the crime, all influence the chances of the offender choosing to reoffend in the future (Tillyer & Eck, 2011). Several studies have examined RCT and the role it plays in various crimes as well as how it relates to offenders choosing a target/victim. Offenders have been found to weigh the costs and benefits of selecting certain targets and the distribution of crime in an area has been found to be related to the number of viable targets, such as bars, parking lots and certain businesses (Engstad, 1975). Target characteristics have been found to be related to the likelihood of victimization in a variety of crimes including vandalism (Ley & Cybriwsky, 1974; Wilcox, Madensen, & Tillyer, 2007), shoplifting (Walsh, 1978), burglary (Brantingham & Brantingham, 1975; Maguire, 1982; Repetto, 1974; Scarr, 1973; Waller & Okihiro, 1978), childhood sexual abuse (Terry & Ackerman, 2008), and cyberstalking (Reyns, Henson & Fisher, 2012). With the knowledge that certain targets/victims are more vulnerable than others, crime prevention tactics can be tailored 21 to make those targets less attractive. Crime can then be designed out through such tactics as increasing the perceived effort, increasing perceived risks, inducing guilt or shame on potential offenders and reducing the anticipated rewards (Cornish & Clarke, 2003). One of the key elements of RCT is the assumption that offenders are rational individuals. Extremists and terrorists are much like any other criminal in that they are rational and weigh the costs and benefits before committing a crime, or in this case a violent extremist act (Clarke & Newman, 2006). To prevent terrorism, it is necessary to “think terrorist” which involves understanding the logic and reasoning that goes into the various decisions (target/victim selection, weapon selection, etc.) that terrorists make (Clarke & Newman, 2006). Terrorists identify what opportunities are present to commit an act of terrorism and then determine which of these opportunities is the most attractive. The idea of “thinking terrorist” comes from “thinking thief” which is a theoretical approach to help understand the choices thieves make (Poyner, 2005). In examining burglaries in one city, Poyner and Webb (1991) found that there were two distinct types of burglaries, one committed in the inner-city that included theft of items such as cash and jewelry and another type that was committed in the suburbs that primarily included electronics. The opportunities in each of these environments determined the type of theft. For example, in the inner city the houses were designed in a way that gave less privacy so offenders needed to be able to escape quickly and did not have time to take multiple trips carrying heavy electronic goods. In the suburbs, people were primarily away from home during the day and homes were further apart, allowing more time and privacy to burglarize. 22 Research has examined RCT as it pertains to the target selection of terrorists (Asal, Rethemeyer, Anderson, Rizzo, Rozea & Stein, 2009; Crenshaw, 2000; Hoffman, 1998). Terrorists seek to cause maximum damage through destruction and loss of life as well as heightening public fear (Asal et al., 2009; Clarke & Newman, 2006; Crenshaw, 2000). Many terrorists seek to produce a shock value to garner attention to their cause (Le Vine, 1997). Additionally, terrorists often choose their targets based on perceived public support they will receive from attacking that target as well as the likelihood they believe they will succeed in their mission (Nemeth, 2010; Sandler & Lapan, 1988; Sandler & Siqueira, 2006). Terrorists seek a target that if attacked will garner them significant media attention in order for them to relay their grievances, ideology and goals to a larger audience (Hoffman, 1998; LaFree & Dugan, 2004). It is clear that terrorists have a variety of goals that they consider in their selection of a target or victim. Abrahms (2012; 2014) argues that terrorists are motivated by three types of goals which include process goals, outcome goals and personal goals. Process goals are goals such as media attention, financial support, garnering public support and increasing membership. Outcome goals are the stated political goals of a terrorist. These goals are different from process goals because for these goals to be met the government would need to collapse or comply with the demands. Additionally, terrorists are often motivated to provoke retaliation from the government as well as coerce the government into concessions (Kydd & Walter, 2006; McCauley, 2006). For example, an Al Qaeda leader was quoted as saying the main goal of the attacks of September 11th was to provoke the United States government to retaliate (Kydd & Walter, 2006). Personal goals that motivate terrorists include solidarity and acceptance in groups, a feeling of purpose, respect, allowing them to travel and even giving them something to do to alleviate boredom. 23 Several studies have examined the rationality of terrorists and found terrorists generally to be much like any other criminal in that they are rational decision makers. Carson, LaFree and Dugan (2012) interviewed animal rights and environmental rights extremists and found that they weigh the costs and benefits before getting involved in illegal protests. Another study examined bombings of London railways and compared them to bombings of Tokyo and Budapest railways and found that the targets for these attacks were not chosen randomly, showcasing another instance of rationality in terrorist choices (Jordan, 2008). Motivations of suicide bomber terrorists were examined through RCT and findings show that even suicide terrorists act in a rational manner (despite the fact that many people consider suicide terrorism to be an irrational behavior); these individuals consider the benefits of committing the act and are no different than other criminals (Perry & Hasisi, 2015). Suicide terrorists often justify their actions as rational and believe that they will be rewarded religiously, personally, or socially. The religious rewards consist of the promise of spending eternity in paradise after death, prior sins being forgiven and a path being set for loved ones to also receive eternity in paradise. Personal rewards include being depicted as a martyr or hero, revenge or even committing the act to alleviate depression or hopelessness. Social rewards include improving their family’s status after the attack or a monetary gain for the family by completing the act. For example, the families of suicide bombers in Palestine were offered money through what is known as the Martyr’s Fund which consists of funds gathered by private Saudi funders (Singer, 2003). As demonstrated, RCT has begun to be applied to terrorism but there are still gaps in this literature. This research has mostly been descriptive in nature (Boba, 2009; Bloom, 2005; Ekici et al., 2008, Gruenewald et al., 2015; Roislien & Roislien, 2010; Santifort et al., 2013). RCT 24 literature has provided some insights into the rationalizations of decisions terrorists make, but has not examined how terrorists choose targets based on the vulnerability of the victims. This study will help fill this gap in the literature by examining what specific factors of vulnerability are present in lethal extremist incidents and their impact on the outcome of such incidents. It will further expand this literature by seeing how these vulnerability factors vary across the ideological motivation as well as victim and suspect characteristics. Routine Activities Theory Routine activities theory (RAT), like rational choice theory, focuses on situational factors that impact crime (Brantingham & Brantingham, 1981). RAT was developed by Cohen and Felson (1979) and argues that crime results when there is a convergence of a suitable target, a motivated offender and the lack of a capable guardian. The probability of this convergence is influenced by routine activities. These routine activities may include family, leisure, work and consumption activities (Cohen & Felson, 1979) and they include activities of potential offenders in the area as well as those of potential victims (Brantingham & Brantingham, 1993). A guiding principle of the theory is that there has to be an opportunity for a crime to occur. Felson (1994) argues that RCT can be used to explain the content of the decisions that offenders make whereas RAT is focused on ecological and situational contexts that provide the context where offenders are given the options that they can choose from. In other words, RAT explains what opportunities may exist that are desirable for offenders and RCT explains what factors into the choice that offenders make in whether or not to commit a specific crime. Many factors influence an offender choosing to commit a crime and what opportunities are appealing. Tillyer and Eck (2011) suggest that there are controllers that can influence the 25 three components of crime as outlined in RAT, the setting, the offender and the target. Place managers control specific places; guardians control specific targets and handlers have an impact on the decision-making process of offenders. Research has found that crime is not evenly distributed spatially (Sherman, Gartin & Buerger, 1989; Weisburd, Bushway, Lum & Yang, 2004). Crime is expected to occur within an activity space, which then makes it beneficial to examine individuals’ daily routines. Places that harbor a large number of people at the same time are more likely to involve the convergence of a victim and an offender (Brantingham & Brantingham, 1993). It is when there is a convergence of an offender, a suitable target/victim and a place that lacks a capable guardian that a crime is more likely to occur. When a motivated offender is present, the focus then turns to the location. If under certain conditions crime is more likely to occur, then these conditions can be changed. This study’s research questions are focused on understanding what factors make an opportunity attractive and what can be done to make these opportunities less attractive. This study examines situational characteristics of homicide events and pinpoints what factors make a target most vulnerable and what factors are related to lethality when controlling for victim and suspect characteristics. This research further examines what factors are important based on the ideological motivation, victim characteristics and suspect characteristics. RAT has been applied to terrorism studies (Canetti-Nisim et al., 2006; Hamm, 2007; Parkin & Freilich, 2015; Parkin et al., 2014). In Mark Hamm’s (2007) book Terrorism as Crime: From Oklahoma City to Al-Qaeda and Beyond he applies RAT to various case studies including the first bombing of the World Trade Center (1993) and the U.S. embassy bombings in Kenya and Tanzania. There were several situational characteristics, such as inadequate on-site security, 26 that influenced the decisions that offenders made in these attacks. Hamm’s approach focuses on how offenders have created opportunities for their attacks. For example, Hamm describes the preparation that would have been needed to complete the Oklahoma City bombing including an explosives expert and “experience in the logistics of urban terrorism.” In order to complete an attack as large as the bombing of the Alfred Murrah Federal Building it was necessary for the suspects to conduct surveillance and carefully study the routine activities that occurred in the building (opening hours, security) as well as that of the area (traffic at the time, how to plant the bomb and exit without being noticed). The suspects utilized this knowledge to create an opportunity to successfully achieve the attack which killed 168 citizens. Cohen and Felson (1979) argue that an individual’s lifestyle influences the likelihood of being selected as a target. For example, young women are at greater risk of being personally victimized (Rodgers & Roberts, 1995). The activities that individuals partake in are directly related to their likelihood of victimization with some routine activities being riskier than others (Hindelang, Gottfredson & Garofalo, 1978). It is also important to consider individual victim level factors as they may have an impact on the likelihood of being a victim of a terrorist attack. For example, income may be related to vulnerability because people of lower income are more likely to use public transportation and therefore are at greater risk of being a victim of a suicide bombing (since suicide bombings often take place on public transportation) (Canetti-Nisim et al., 2006). Canneti-Nism, Mesch, and Pedahzur (2006) examined RAT by comparing characteristics of victims of suicide bombings to victims of other types of terrorism. They found that there are key differences in victims of suicide bombings. They argue that this refutes the hypothesis that being a victim of terrorism is random and that it is indeed impacted by lifestyle choices. 27 Individual characteristics such as profession, age and gender were found to be related to victimization. Younger people, women and students were found to be more likely to be victims of suicide bombers because of their lifestyle choices, such as using public transportation more often and being in highly populated locations more often. Parkin and Freilich (2015) applied RAT to examine differences in victims of far right extremist lethal ideologically motivated homicides versus non-ideologically motivated homicides committed by far-right extremists. Victims of non-ideological attacks were more likely to be white males and were also more likely to be known to the offender. Further, they found that ideological victims were more likely to be murdered while doing routine activities that take place outside of their homes and non-ideologically motivated homicides were more likely to occur inside the home (domestic violence, disputes within extremist groups and among friends). Research is beginning to use RAT to examine the situational environment of terrorist attacks as well as the factors that influence victimization. This study will expand upon the study by Parkin and Freilich (2015) by including homicides committed by jihadists. It will further expand on this by looking at victim, suspect and ideological characteristics. Studies have utilized the victim as the unit of analysis (Parkin & Freilich, 2015; Parkin et al., 2014) but have not considered the level of lethality. Law enforcement has heavily focused on offender apprehension, profiling and the characteristics of the terrorist. Some descriptive research has been done examining terrorist targets, however, research has failed to examine the victim and the situational environment. This study will examine lethal domestic extremist incidents and look at characteristics of the victim, suspect and the situational environment (EVIL DONE) the victim was in at the time of the attack. This makes an important contribution in that it will allow for 28 there to be comparisons of vulnerability characteristics and how they influence the level of lethality in an incident. Situational Crime Prevention Situational crime prevention (SCP) is an action based research paradigm and is grounded in RCT and RAT (Clarke, 1992). Like RCT, which it is closely associated with, it stresses the immediate crime situation as opposed to both background factors (poverty, schools, etc.) and the formal criminal justice system and punishment, that are thought to be too far removed to influence offender decisions (Freilich, 2015). SCP measures are therefore directed at very specific types of crime, and involve the design and manipulation of the immediate environment to make crime more difficult, or rather less rewarding to would be offenders (Clarke, 1992). One of the main tenants of SCP is that crime is an interaction between an individual with a criminal disposition and an opportunity for crime (Clarke, 1992). It is expected that if the opportunities for crime are reduced and the perceived risk is increased (through mechanisms such as target hardening) then crime will be less likely to occur. Several studies have shown positive results for SCP’s effectiveness. These studies involve examining how targets or opportunities were changed in an attempt to reduce crime. These studies have examined a variety of SCP measures that were applied to help reduce crime, including improving street lighting (Painter & Farrington, 1997), adding in street barricades (Atlas & LeBlanc, 1994), redesigning stores’ layouts (Eck, 2006) and adding security guards to car parks (Welsh & Farrington, 2009). Such strategies have been examined as they relate to reducing shoplifting (Farrington & Burrows, 1993), convenience store robbery (Bellamy, 1996), child sexual abuse (Terry & Ackerman, 2008), information security theft (Willison & Siponen, 29 2009) and cyberstalking (Reyns, 2010). For example, Brown (1979) found that reductions in suicide in Great Britain were related to the reduction in toxic gas used in homes for heating and cooking. If individuals do not have easy access to toxic gas, then they are less likely to use it commit suicide. In another study, street lighting was improved in several areas while other areas were used as control groups and findings indicate that areas with improved lighting had a significant drop in crime rates (23%) as compared to the control group areas (3% drop in crime rate) (Painter & Farrington, 1997). The increase in lighting in areas increased the risk for offenders and the likelihood that they may be caught and it therefore deterred them. In a recent study, Hart and Miethe (2014) found that individuals greatly reduced their risk of burglary victimization after taking SCP preventative actions and displacement did not appear to occur. This study shows that actions can be taken to reduce risk of victimization and at as a result an offender may choose not to victimize as opposed to simply finding a new target. The techniques used in these three examples vary in nature and are dependent upon the area, specifics of the situation and the type of crime that is trying to be prevented. The criminology literature is just beginning to apply environmental theories to the study of terrorism and there are several gaps in this research that this study seeks to fill. This study will apply SCP and EVIL DONE to understand vulnerabilities of victims in lethal domestic extremist incidents which will further this theoretical perspective. The next section introduces the EVIL DONE framework, which is a framework developed to determine target vulnerability. The elements of this framework are used and expanded upon to examine vulnerability and its relationship with lethality. Finally, the third section examines ideological, victim and suspect related factors that may be related to vulnerability and lethality. 30 Vulnerability and EVIL DONE Vulnerability To reduce opportunities for violent attacks there needs to be an understanding of who and what is vulnerable to being attacked. Vulnerability represents how likely a victim or target is to be attacked. For this study, all incidents that are examined resulted in the death of at least one victim. Some incidents had two, three or even dozens of deaths. Why were there more deaths in some incidents? The EVIL DONE framework includes eight factors that will help provide insight into understanding target vulnerability. EVIL DONE and Prior Research Clarke and Newman (2006) published a book titled Outsmarting the Terrorists. This book outlines how SCP can be applied to terrorism. Law enforcement, and prior research has focused on why terrorism occurs rather than “how”. They argue that the best way to reduce terrorism is by identifying and removing opportunities for terrorism. In order to identify these opportunities, the process by which terrorism occurs must be understood. They argue that there are four pillars to terrorism: targets, weapons, tools and facilitating conditions. Tools include vehicles, credit cards, cell phones, cash, fake documents and similar products that are used to plan or conduct an attack. Weapons are what are used in the attack and may include small fire arms, bombs, knives, blunt objects or biological/chemical weapons. Facilitating conditions are the physical and social arrangements that exist in society that make it possible for an attack to occur. These conditions may be the way media reports attacks, the way police operate in a certain area and general societal beliefs about terrorists and attacks that terrorists may use to their advantage. Facilitating conditions and societal arrangements are much more difficult to identify 31 and change. All three of the aforementioned pillars are important in the opportunity structure for terrorism. The fourth pillar that Clarke and Newman discuss is targets, and this is the focus of this study. Targets represent who and what terrorists seek to attack. Not all potential targets have the same risk to being attacked. Clarke and Newman argue that law enforcement needs to evaluate the risk of potential targets being attacked. They argue that cities should develop a catalog of targets within the city and then rank these targets based on various vulnerability characteristics in order to determine which potential targets are at the greatest risk. Practitioners have limited resources to prevent terrorism and the EVIL DONE framework can be used to identify which targets are at most risk to being attacked. The EVIL DONE framework consists of eight components: exposure, vital, iconic, legitimate, destructible, occupied, near and easy. Targets who rank high on these components are arguably more vulnerable and attractive choices to terrorists. When Clarke and Newman (2006) first introduced this framework, they applied it to targets in Washington D.C. in a demonstrative fashion. They did a “rough rating” of targets to demonstrate how EVIL DONE could be used to rank target vulnerability but noted that all of the structures they analyzed would be relatively high in vulnerability as compared to lesser-known buildings. They ranked targets from one (low) to five (high) in each of the eight categories of EVIL DONE. The rankings for all categories were then combined to determine a score between eight and 40 for each target, with higher scores indicating more attractive targets. Despite ranking targets, Clarke and Newman (2006) fail to discuss a breakdown of these rankings or further operationalize them. The numbers were provided with little justification. For example, there is no discussion of the difference in a score of two on the vital variable versus a score of three. They note that the scale they make is simply 32 done as a demonstration of how EVIL DONE could be operationalized and that objective rating scales should be developed for each of the EVIL DONE criteria. Additionally, Clarke and Newman (2006) argue that to policymakers some variables may hold greater weight, such as the ‘occupied’ variable as opposed to how iconic a target is. When applying this framework to targets in Washington D.C. they found that the National Zoo (receiving a total score of 10) is not a very attractive target and it ranked relatively low on most of the components of EVIL DONE. Specifically, the National Zoo ranked low on the vital variable, the iconic variable and the legitimate variable. Whereas the U.S. Capitol (receiving a total score of 29) was found to be the most attractive target of those assessed followed by the White House and the Pentagon. EVIL DONE has been adapted and re-operationalized using prior research in order to apply it to lethal domestic extremist attacks (Boba, 2009; Gruenewald et al., 2015; Paton, 2013). This study does not compare attacked targets with targets that have not been attacked but extends the literature by seeing if the EVIL DONE framework can explain lethality in prior attacks. Further, this study expands the literature by comparing EVIL DONE characteristics across ideology, suspect and victim characteristics to better understand vulnerability in prior domestic extremist attacks. The first researchers to utilize EVIL DONE were Ekici and colleagues (2008) and they applied the framework to international targets. They had analysts from the Turkish National Police Anti-Terrorism Department, who were chosen based on their expertise of certain terrorist groups, apply the coding scheme to targets. The three primary terrorist groups in Turkey were examined, all three of which had differing ideologies and goals. They found that what was attractive across targets did not vary much depending on ideology. Three raters that were experts on three different terrorist groups ranked six targets and their rankings across all targets were 33 nearly identical. The target that received the highest rating was the underground metro station and it received this rating because all raters perceived it to be high on the variables destructible, occupied, vital and exposed. This study will expand on this by examining different ideologies on a domestic level, and specifically seeing what differences in vulnerability exist between victims in jihadist attacks versus victims in attacks committed by right-wing extremists. Another study examined terrorism in Turkey using the EVIL DONE framework. Ozer and Akbas (2011) extended the framework by applying it to attacked targets. They examined buildings that were attacked by the Kurdistan Worker’s Party (PKK) in Istanbul, Turkey between 1998 and 2008. In terms of the EVIL DONE variable near, targets were found to be located near where PKK members lived. They also found targets chosen were easily destructible and generally had little to no security. However, they only examined sixteen attacks and they did not operationalize the EVIL DONE variables but rather discussed the eight EVIL DONE variables in relation to these targets. Boba (2009) introduced a new operationalization for the EVIL DONE framework to be used by other researchers. The focus of her operationalization primarily involved international terrorism and targets that were attacked using a WMD. Paton (2013) applied Boba’s operationalization of EVIL DONE to targets that have been attacked and used incidents from the Global Terrorism Database (GTD). Her results found that EVIL DONE factors such as destructible, vital, legitimate and near were strong predictors of vulnerability in these attacks. She found the destructible variable to be the strongest predictor of incident lethality. Boba (2009) and Paton (2013) provide a useful operationalization for international attacks but due to its heavy focus on WMDs it has limited application to domestic attacks. 34 Gruenewald, Gruenewald and Klein (2015) refined EVIL DONE even further by applying it (and operationalizing it) to eco-terror incidents in the United States. Findings indicate that eco-terrorists generally chose targets that had limited security(easy) and targets that were open to the general public but rarely frequented and that terrorists often attacked at night(exposed). They also found that eco-terrorists primarily attacked targets that were considered to be legitimate, meaning targets that the eco-terrorists thought were responsible for harming the environment. They take note that they are applying the framework to these targets but that it is not a comparison but simply done to look at the distributions of the variables. This framework is the most helpful since it is applied to targets that have been attacked and it also applies to domestic incidents. The number codes assigned to each variable, however, aren’t clearly defined. Gruenewald and colleague’s study provides descriptive results of what components of EVIL DONE are more evident in incidents of domestic eco-terror. The EVIL DONE variables have varying measurements in their framework and this approach is taken because the scores are not combined into one large EVIL DONE score. Rather, the goal of their study was to operationalize each variable and apply them to prior environmental and animal rights attacks to see if they are applicable. Their study provides a helpful operationalization of EVIL DONE but is more specific to environmental extremists and does not go beyond examining the EVIL DONE variables at the univariate level. The current study is valuable because it expands on this work by comparing attacks committed by right-wing extremists and jihadists, the results of which can be found in Chapter four and Chapter five. This dissertation is also unique in that it takes analysis beyond the univariate level and looks at the relationship between EVIL DONE vulnerability factors and lethality at a multivariate level while controlling for victim and suspect characteristics. 35 In sum, the prior research examining EVIL DONE is limited in nature. Ekici and colleagues (2008) as well as Ozer and Akbas (2011) expanded the framework by using Clarke and Newman’s operationalization and applying it on an international level by looking at targets in Turkey. Boba (2009) expanded EVIL DONE by operationalizing it in a way that could be useful for practitioners and made the framework more applicable to international targets as well as WMDs. Paton (2013) then took Boba's (2009) framework and applied it to prior incidents using the GTD and found support for several of the variables in EVIL DONE. Grunewald and colleagues (2015)’s study has been the only study to apply it to domestic terrorism but the framework created was specific to only eco-terrorism. To code and analyze the vulnerability of victims using EVIL DONE it is necessary “think terrorist” (Clarke & Newman, 2006). Through rationality and opportunity theories, offenders are viewed as being rational individuals who weigh the costs and benefits of their actions. Incidents are coded to capture the characteristics of the victim and their surroundings at the time of the attack. This is very important because target locations can change drastically over even short periods of time (e.g., a potential victim may be an ideal target when walking alone down the street at night but becomes less ideal when accompanied by three of four others). The next section will discuss the EVIL DONE framework. EVIL DONE Framework The first component of EVIL DONE is the exposure of a target. A target is considered exposed if it is visible and attracts attention; those targets that are more visible are argued to be more attractive to terrorists (Clarke & Newman, 2006; Gruenewald et al., 2015). Examples of exposed targets include many outside areas (streets, parks), malls, businesses and areas where 36 large groups of people gather. Locations that are not very exposed are those that have limited access like many government buildings or private residences. It is expected that victims who are at a location that is more exposed will be more likely to have been attacked in an incident that caused a higher number of casualties as compared to victims in less exposed areas. Clarke and Newman (2006) propose that a target is vital if it is necessary to the daily functioning of society. For example, major power plants and transportation hubs if destroyed would greatly impact the day to day functioning of society. The destruction of the only grocery store or gas station in a small town may also have a large impact on the day to day functioning in that community. Boba (2009) goes on to argue that the purpose of attacking vital targets is not necessarily to cause harm but may be to demoralize or paralyze the community. This variable is measured based on the vitalness of the location that the victim was in when attacked. Locations that are not considered to be vital may include many outside areas, private residences and businesses. Iconic targets are those that are symbolic or representative of something deemed important to society (Clarke & Newman, 2006). Symbolism and the iconic nature of a target may be important in target selection because terrorists often consider the audience of the target beyond the actual or potential victims (Fussey, 2011). Examples of iconic targets include the Pentagon, the White House and the Statute of Liberty. The federal building in Oklahoma City that Timothy McVeigh bombed was iconic because it was representative of the government, which McVeigh despised.4 4 It is also very important to consider the type of terrorist group or ideology when considering the iconic dimension. Different types of terrorists may view the iconic nature of targets differently. 37 A target is considered legitimate to a terrorist if the terrorist views the public reaction to that target being attacked as acceptable (Clarke & Newman, 2006). Domestic terrorists target individuals for a variety of reasons: their ethnicity, their sexuality or even their occupation. Terrorists typically consider how the public will react when they attack a certain target. Often terrorists want to gain a positive reaction by their sympathizers and gain new sympathizers (Fussey, 2011). Government and military institutions are often viewed as legitimate targets (Clarke & Newman, 2006). Children are universally viewed as innocent and illegitimate targets and often after a terrorist incident has killed children the offenders will issue regrets or apologies for harming children (Clarke & Newman, 2006). The more legitimate a target is, the more attractive it is to the terrorist. Law enforcement, military and security personnel are often considered to be legitimate targets of domestic extremists. Alternatively, law enforcement, military and security personnel are significantly more likely to be in locations with security/have weapons on them and may be considered “hard” targets. Therefore, it may be more difficult for these targets to be successfully attacked (Seifert & McCauley, 2014) which may decrease their attractiveness and vulnerability to would be attackers. Additionally, research has shown that target choice has begun to move from harder targets (government/military) to softer targets (civilians) (Santifort et al., 2013). These competing viewpoints are tested with the prediction that despite EVIL DONE’s expectation of security personnel being more attractive, that instead civilian targets will be more common and the attacks on civilians will be more lethal. Destructible targets are those that a terrorist can destroy (Clarke & Newman, 2006). If the target has a number of security measures or precautions in place it may range from difficult to impossible for a terrorist to cause significant damage to the target. Alternatively, if the target is large or expansive it may be more difficult to destroy. If there are too many security measures in 38 place or specific measures that will prevent a certain method a terrorist may decide to pick a different target that is easier to destroy. A target is considered occupied if it has people inside of it (Clarke & Newman, 2006). Terrorists try to destroy targets (secured or moving) that have individuals inside (Boba, 2009). Airplanes, busses, malls, and stadiums are ideal because damage to the structure also harms those inside those structures. The more often people are brought together at a location the more opportunities a terrorist has to complete a successful attack on that target (Boba, 2009). The World Trade Center was an ideal target for terrorists on 9/11 because of the thousands of people working in and visiting the buildings as well as the hundreds that were on the planes that were flown into the buildings. Criminological studies have found that generally offenders commit crimes near their homes because they know the opportunities in those areas (Eck, 1993). Brent Smith (2008) examined behaviors of terrorists and found that in 44% of terrorist attacks the terrorists lived within 30 miles of the target. Clarke and Newman (2006) argue that the single most crucial element of the attractiveness of a target is how close it is located to the terrorist's home. Offenders typically choose targets that are close to minimize their effort. These targets are more convenient and allow the suspect the ability to blend back in without having to travel far after completing their attack. It is expected that the closer the target is to the suspect’s residence, that the more attractive and vulnerable it is. A target is considered to be easy if it has limited security or obstructions in the way of the offender (Clarke & Newman, 2006). Buildings that have no security or houses that do not have locks can be viewed as easy targets. Several questions can be asked to determine level of security 39 such as: does the target have armed security? Does the target have security cameras or CCTV? Does the target have any system of surveillance? Does the target have an alarm system? Is the alarm system of the target continually monitored? Who has access to the building? Targets that are easy are more attractive to terrorists than those targets that appear to be high risk (Clarke & Newman, 2006). This study will fill a unique gap in literature. Clarke and Newman (2006) called for researchers to test this framework. This study therefore begins to fill a gap in the literature by operationalizing and applying the EVIL DONE framework to domestic incidents and further determining which elements of EVIL DONE vulnerability are more prominent or important predictors than others. Similarly, this study brings a significant methodological contribution in operationalizing and testing the framework by applying it to prior incidents of domestic extremism at a multivariate level. Scholars have examined EVIL DONE but only one study has examined the framework in a multivariate framework and applied it to international targets (Paton, 2013). This dissertation helps to further the vulnerability literature by specifically utilizing this framework and expanding it to consider victims and situational context to explain factors that impact the vulnerability and lethality of victims in prior incidents. Lethality and Target Choice Research has examined the factors that impact target choice and lethality but this research has been limited and is generally descriptive in nature. There are no studies that have explored if and how the EVIL DONE factors predict lethality, controlling for other potential influences in domestic terrorist attacks. This final section highlights prior research on lethality. 40 Ideological Factors It is very important to consider what the motivation or the intent was for a specific crime. SCP argues for examining very specific crimes separately (Clarke, 1992), and in applying this to terrorism there is the need to examine the ideological motivation of the individual or group separately (Freilich & Chermak, 2009). It should be expected that individuals or groups with different ideological beliefs will find different targets attractive (Drake, 1998). Some ideology types may explicitly promote deviance; groups that promote deviance through their ideology are more likely to legitimize deviance (Freilich, Almanzar & Rivera, 1999). For example, a group’s ideology is specifically related to its likelihood to participate in kidnapping (Forest, 2012). More broadly a group’s ideology influences its likelihood to radicalize (Post, Ruby & Shaw, 2002), and its level of lethality in committing attacks (Asal & Rethemeyer, 2008). Religion has been found to be directly related to a group’s level of violence (Asal et al., 2009; Cook & Lounsbury, 2011). Ideology can legitimize targets and can be used to justify attacks (Drake, 1998). A terrorist’s target choice is dependent on their ideology (Brandt & Sandler, 2010). Since the 1990s there has been an increase in Islamic fundamentalist terrorists (Brandt & Sandler, 2010). While there is growing concern over Islamist terrorism, there has been a significant portion of right-wing extremist terrorist acts. Attacks by these ideological types may have stark differences. Therefore, it is important to examine target attractiveness and how it varies based on the ideology of the terrorist. Due to the increase in attacks and fatalities committed by right-wing extremists, it is expected that they will be more likely to be involved in attacks with greater casualties. 41 Suspect Factors Several suspect demographics are examined in this study. Far right extremists have been found to be more likely to be males (Bloom, Gill & Horgan, 2012; Gruenewald et al., 2009; Gruenewald, 2011; LaFree & Dugan, 2004; Smith & Morgan, 1994). Some research has found that terrorism is disproportionately committed by young men (LaFree & Dugan, 2004) but other research has found that terrorism is more likely to be committed by older men (Smith & Morgan, 1994; van Zuijdewijn & Bakker, 2016) and lone wolves are likely to be older (Gruenewald et al., 2013). For example, Smith and Morgan (1994) found that offenders who were prosecuted for federal related terrorism crimes were likely to be white males in their late 30s. The gender, age and race of suspects are examined in this study. These demographic characteristics are examined and it is expected that older white males will be involved in more lethal attacks. Whether or not an offender was a lone wolf offender is examined in this study, the number of suspects that were involved in the attack and the weapon the suspect used to commit the attack. Suspects that are intensely affiliated within their movement may be more likely to offend alone (Gruenewald, 2011) as opposed to those who have looser movement connections. Suspects who work with others may be able to pool together new skills and resources and better conduct surveillance or other activities necessary to complete a successful attack (Klein, Gruenewald & Smith, 2016). There may also be other key differences between attacks that involve loners versus multiple individuals. For example, lone offenders have been found to generally live further from target locations and lone offenders may be able to more easily evade arrest (Smith, Gruenewald, Roberts & Damphousse, 2015). Additionally, loners have been found to be more likely to commit bombings or armed attacks with guns than those operating in conjunction with others (Phillips, 2015) and are further more likely to target civilians (Spaaij, 42 2010). Klein and colleagues (2016), found that terrorist acts committed by far-right extremists in the United States were more likely to be successful if they were committed by a lone offender as compared to by a group. It is expected that acts committed by lone wolf offenders will be more lethal than those committed by multiple offenders. Victim Factors Research has found demographic differences and lifestyle choices of victims to be related to the likelihood of victimization (Cohen & Felson, 1979; Meier & Meithe, 1993). There are several factors of victims that make some victims more vulnerable to being victimized or killed than others. Environmental theories have been applied to homicides and have begun to be applied to victims of extremist attacks (Canetti-Nisim et al., 2006; Feniger & Yuchtman-Yaar, 2011; Parkin & Freilich, 2015). These factors related to the victim include sex, race, age, occupation, and the victim’s relationship with the offender. Research has found that different sexes have different routine activities and therefore different victimization likelihoods (Caywood, 1998; Fox & Zawitz, 2007; Kposowa & Breaull 1998; Messner & Tardiff, 1985; Silverman & Kennedy, 1987). Males are disproportionately more likely to be victims of homicide than females (Fox & Zawitz, 2007; Kposowa & Breaull, 1998; Rogers & Roberts, 1995) and even more likely to be victimized if they live in the inner city (Lauritsen, 2001). Not only are women disproportionally less likely to be a victim of homicide as compared to men they are also disproportionately less likely to offend (Silverman & Kennedy, 1987). Race has found to be significantly related to risk of homicide (Ezell & Tanner-Smith, 2009) with blacks and Asians having a higher risk of being a victim of homicide than whites 43 (Breaull & Kposowa, 1997). According to the FBI Supplementary Homicide Reports, 50% of homicide victims are white (Gruenewald & Pridemore, 2012). Interestingly, Parkin and Freilich (2015) found that white males were more likely to be victims of non-ideologically motivated homicides committed by far-right suspects. Additionally, increases in age have generally been found to be related to a decrease in homicide risk (Breaull & Kposowa, 1997; Messner & Tardiff, 1985). Messner and Tardiff (1985) found that those considered very young or very old are more likely to be killed while at home. However, younger individuals are more likely to be victims of a terrorist attack committed by a suicide bomber (Canetti-Nisim et al., 2006). Parkin and colleagues (2014) found that victims who were killed in an anti-racial minority attack were more likely to be under the age of 25 whereas those killed in anti-government attacks were more likely to be 35 years of age or older. It is expected that younger individuals are more likely to be victims of terrorist attacks because younger individuals are believed to participate in lifestyles and activities that are riskier than those in other age groups. An individual’s occupation has been found to be related to their risk of being a victim of homicide. It is expected that those who are unemployed have a higher risk of being a victim of homicide. The risk of homicide has been found to be higher for those working in the service industry as compared to individuals who worked in professional occupations (Breaull & Kposowa, 1997). Employed individuals are likely to be killed further away from their residences than unemployed individuals (Caywood, 1998; Messner & Tardiff, 1985). Students are also more likely to be victims of terrorist attacks than other occupations (Canetti-Nisim et al., 2006). Dugan and Apel (2003) found that women who were employed but had low income were more at risk for violence victimization. Pridemore and Freilich (2005) also found a relationship between 44 women and community status and found that women were more likely to be a victim of homicide in areas where women’s employment levels were similar to that of males in the area. Homeless individuals are another population that may be at risk. Law enforcement, military and security personnel are often considered to be legitimate targets of domestic extremists. Alternatively, law enforcement, military and security personnel are significantly more likely to be harmed and may be considered “hard” targets. It may therefore be more difficult for these targets to be successfully attacked (Seifert & McCauley, 2014) which may decrease their attractiveness and vulnerability to would be attackers. Additionally, research has shown that target choice has begun to move from harder targets (government/military) to softer targets (civilians) (Santifort et al., 2013). It is important to examine the relationship between the victim and suspect (Messner & Tardiff, 1985; Nelson & Huff-Cazine, 1998; Silverman & Kennedy, 1997). Parkin and Freilich (2015) examined ideological and non-ideological homicides committed by far-right extremists in the United Sates and found that victims of non-ideological homicides were more likely to be known to the suspect. This makes sense since it can be expected that non-ideological homicides are more likely to be disputes between friends as well as occurrences of domestic violence. It is expected that many non-ideologically motivated attacks would involve not only the suspect knowing the victim but also fewer deaths. Extremists often seek to cause mass casualties and destruction in ideologically motivated attacks and it is expected that these attacks will be against strangers. While arguably an incident characteristic, the weapon the suspect chooses to use in the attack is also examined. It is included as a suspect variable since it does not easily fit into the other three categories of EVIL DONE factors, ideological factors and victim factors. 45 Additionally, suspects must choose what weapon they can obtain, use and transport. Weapons that are readily available will generally be more attractive to terrorists whereas those that are difficult to produce or acquire are less attractive choices. By examining weapon choice, it can be determined how successful attacks are with certain types of weapons as well as what characteristics of target attractiveness and vulnerability are associated with various weapon choices. Clarke and Newman (2006) also argue that the choice of weapon is an important factor to terrorists when they are making a decision on which target to attack. The weapon choice and availability will have an impact on what targets or victims would make the best choices to attack (Clarke & Newman, 2006) Weapons can be categorized into three broad categories including guns/firearms, explosives and unconventional weapons. Unconventional weapons include such weapons as chemical or biological weapons, nuclear weapons or weapons that are not as easily available as firearms or explosives. For this study, weapons will be coded into a binary variable of firearms vs. non-firearms weapon. The variable is coded this way because of very few cases within each other non-firearms category (especially once the data is split into ideological and non-ideological incidents). However, this coding can still provide useful information as firearms especially as firearms legislation has become a hotly debated political topic. Therefore, this can shed light on the differences in lethality in domestic extremist incidents for incidents involving firearms versus cases involving the usage of any other weapon. It is expected that attacks involving firearms will be deadlier than attacks that are committed using other weapons. Hypotheses It is expected that the EVIL DONE vulnerability characteristics will be related to the lethality of an incident, with incidents ranking higher on these characteristics having more deaths. Prior research has not explored this relationship, but argues that several victim and 46 suspect characteristics are related to victimization. Although it is anticipated that victim and suspect characteristics should be important predictors of lethality at the bivariate level, it is expected that EVIL DONE vulnerability factors will be the strongest predictors of lethality in a multivariate model. Table 2.1 presents the hypotheses of this study. 47 Table 2.1 Hypotheses EVIL DONE Hypotheses Lethality will be higher in incidents that occurred in… Exposed locations that are more accessible to the public Vital locations that are considered to be vital to the day to dayfunctioning of society Iconic an iconic location Legitimate locations that house only general citizenry Destructible locations that can easily be destroyed Occupied crowded locations Near the same city as the suspect's residence Easy a location with no security Ideology Hypotheses Lethality will be higher in incidents… Incident ideology that are ideologically motivated Suspect ideology that are committed by right-wing extremists Victim Hypotheses Lethality will be higher in incidents where the victims… Gender are male Race are non-white Age are between the ages of 15-24 Occupation are law enforcement/security officers Victim-suspect Relationship did not know the suspect Suspect Hypotheses Lethality will be higher in incidents where the suspects… Gender are male Race are white Age are older Occupation are homeless/unemployed/blue collar employees Lone wolf are lone wolf suspects Number of suspects acted alone Weapon choice used a firearm as compared to a different type of weapon 48 Conclusion This chapter has covered the theoretical foundations for this study. This dissertation is based on the foundations of RCT, RAT and SCP. Extremists in this study are considered to be rational suspects who weigh the costs and benefits when choosing who and what to attack. This chapter also discussed the theoretical and methodological gaps in terrorist literature and how this study will fill these gaps by examining a vulnerability framework and considering victim, suspect and ideology characteristics and how they are related to incident lethality. The next chapter outlines the methodology of this study and how these elements are tested and how the dependent and independent variables for this study are operationalized. 49 Chapter 3: Methodology This chapter begins by discussing the data that is used and the benefits and potential pitfalls of open source information. Second, the specific inclusion criteria or this study is examined. Third, the dependent variable is presented. The fourth section discusses the operationalization of EVIL DONE, and then the operationalization of the other independent variables is discussed. Finally, the chapter concludes with a discussion of the analytical (univariate, bivariate and multivariate) plan and how this will answer the research questions. Data There has been an increase in the use of secondary data to study terrorism and specifically the usage of terrorist event databases (LaFree & Dugan, 2004). The data for this study is drawn from the U.S. Extremist Crime Database (ECDB) (Freilich, Chermak, Belli, Gruenewald & Parkin, 2014). Research using data from the ECDB has also been utilized in several studies that have been published in prominent criminal justice journals, including such topics as lone wolf extremism (Gruenewald, Chermak & Freilich, 2013), right-wing homicides (Gruenewald & Pridemore, 2012), the life trajectories of extremist organizations (Freilich, Chermak & Caspi, 2009), financial crimes (Sullivan, 2015; Sullivan et al., 2014), and countylevel variation in attacks and suspect residency (Chermak & Gruenewald, 2015; Freilich, Adamczyk, Chermak, Boyd & Parkin, 2015). The unit of analysis for this study is the incident. All lethal incidents that were committed in the United States from 1990 to 2014 by at least one far right or jihadist extremist are included. The criteria for labeling extremists was created based on existing extremist typologies, feedback from terrorism scholars and extensive literature reviews (Freilich et al., 2014). This research 50 expands on Gruenewald and colleagues (2015), who previously examined vulnerability in environmental extremist attacks and applies it to right-wing and jihadist attacks. This database contains lethal ideological and non-ideological incidents committed by extremists in the United States. This allows for an important comparison that helps us better understand vulnerability and whether the motivation for the attack impacts lethality. The data contained in the ECDB was gathered in a three-step process that includes the identification of incidents, collecting open source information about incidents, and then the coding of incidents. Researchers have described this process in detail and it is summarized here (see Freilich et al., 2014). Homicides were identified from existing sources including terrorist databases, official reports, scholarly accounts, reports from private watchdog groups and media reports. These crimes were then compiled and each incident was searched through 30 web search engines. Some of these search engines include Lexis Nexis, Proquest, Yahoo, News Library, Westlaw and the Homeland Security Digital Library. The homicide events included in the ECDB have been examined for selection bias utilizing reliability methods comparing the estimates of homicide events from ten different sources. Findings indicate that the characteristics of the suspect, victim and incident are similar across the sources, which offers support for the use and accuracy of the ECDB data (Chermak, Freilich, Parkin & Lynch, 2012). It is worth noting that there are potential validity and reliability concerns with the usage of open source information. One primary concern is misinformation, which means that the information provided may be biased or inaccurate (LaFree et al., 2006). To help prevent this, all of the information that was gathered for the ECDB was assessed for its quality, giving heavier weight to the accuracy of information from more reliable sources. Another potential concern with using open source data is that there may be limited information. Some cases occurred years 51 ago and have little information available on the internet or have received limited media coverage. Additionally, there may be limited information on the victims or suspects’ personal lives in some of the cases that aren’t as widely reported. Since several search engines were used and multiple coders examined each case this allowed for a greater possibility to find specific information about cases. Also, coders conducted targeted searches to try to fill in specific variables. A final concern of open source data is with the consistency of data collection. Several searchers worked on gathering the open source information for the ECDB. It is important that all searchers gathered information and assessed it in the same manner. In order to increase reliability in data collection all of the coders were trained and on a probationary period where their work was checked by a supervisor. Furthermore, specific protocols and steps for searching cases were created in order to keep the process systematic. For example, an examination of homicide data from the ECDB previously found that coder reliability with situational variables related to homicide incidents had an agreement of 90% and higher (Freilich & LaFree, 2016). Inclusion Criteria For an incident to be included in the ECDB it must have occurred in the United States, involved at least one extremist suspect and occurred between 1990-2014. This study examines only incidents where at least one individual was killed in order to examine how various variables are related to the lethality of an incident.5 There are 434 incidents included in this study (rightwing=374, jihadist=60).6 These incidents are coded to determine vulnerability factors and characteristics of victims, situations and suspects. 5 There must be at least one victim killed in the incident. Incidents where the only death(s) were of suspects are not included. 6 Forty murders that occurred in prison are excluded from analyses. Attacks occurring in prisons are inherently different than those that occur outside of prison in regards to the victim and situational environment. 52 The approach for this study relies upon situational crime prevention, rational choice theory and routine activities theory. Routine activities theory argues that it is the convergence of a suitable target, a motivated suspect and the lack of a capable guardian at a specific time and place that allows a crime to occur. All three aspects of the crime triangle are examined in this study. Characteristics of victims have been found to be related to the likelihood of being victimized and specifically of being a victim of a right-wing extremist attack (Parkin & Freilich, 2015) or of a suicide bombing (Canetti-Nisim et al., 2006). Several variables are used to examine the characteristics of the victims of these homicides including age, gender, sex, occupation, and the victim’s relationship to the suspect. Additionally, variables to examine the suspect have been added and include the suspect’s race, age, occupation, the number of suspects involved in the attack and whether the attack was committed by a lone wolf. Two ideological variables are examined that include the ideological affiliation of the suspect and the ideological motivation for the incident. Finally, the eight elements of EVIL DONE are examined and they consider the situational environment of the attack. Coding The ECDB is relational and each incident has a specific incident, suspect, and victim codebook linked to it. All of this information is compiled and variables of interest are extracted. For an act to be considered a unique incident in the database it must have occurred at a specific location at a specific time. For example, an individual shooting people at one location over a short period of time is coded as one incident. However, incidents that are part of a spree are coded separately. A spree could consist of a perpetrator committing a murder at his house and then hours later driving to another location to commit another murder. 53 Dependent Variable The dependent variable in this study is lethality. All of the incidents included are lethal in that at least one victim was killed, but some attacks had multiple deaths. The dependent variable was thus coded into a dichotomous variable with one victim homicides coded as zero (0), and all homicides with more than one victim coded as one (1). The vast majority of attacks involve the death of only one victim (80%). The mean number of deaths is 8.32, however, when excluding outliers (the Oklahoma City Bombing and the 9/11 attacks) the mean is 1.52. Table 3.1 displays frequencies of deaths and the variation of frequencies based on ideological motivation of the attack. Nearly 80% of ideologically motivated attacks, as well as non-ideologically motivated attacks involve the death of only one victim. However, there is a lot more variation in the number of deaths among ideologically motivated attacks, with these attacks appearing to be more lethal than non-ideologically motivated attacks. Table 3.1 Deaths in Extremist Incidents 1990-2014 All Incidents Ideological Non-Ideological (n=434) Incidents (n=235) Incidents (n=199) Deaths N % N % N % 1 344 79.1% 184 78.3% 160 80.4% 2 52 12.0% 28 11.9% 24 12.1% 3 13 3.0% 7 3.0% 6 3.0% 4 5 1.2% 1 0.4% 4 2.0% 5 5 1.2% 3 1.3% 2 1.0% 6 5 1.2% 3 1.3% 2 1.0% 7 2 0.5% 1 0.4% 1 0.5% 13 3 0.7% 3 1.3% 40 1 0.2% 1 0.4% 168 1 0.2% 1 0.4% 184 1 0.2% 1 0.4% 1303 2 0.5% 2 0.9% - 54 EVIL DONE Operationalization Victims of lethal domestic extremist attacks were targeted for a variety of reasons. A victim may have been specifically targeted because of their status (police officer, race, gender, etc.) or they may have been a victim targeted by chance because they were at a vulnerable location. Alternatively, it may be a combination of these factors that ultimately led to them being a victim of a lethal extremist attack. Clarke and Newman (2006) have proposed eight characteristics specific to targets/locations that should be critical to vulnerability. Prior research has failed to examine the relationship between EVIL DONE variables and lethality while controlling for victim and suspect characteristics. The following section discusses the operationalization and coding of all eight of the EVIL DONE characteristics. Several characteristics are used to represent EVIL DONE variables and many are adapted from prior research and examine features of the specific location that the victim was in when they were attacked (Boba, 2009; Gruenewald et al., 2015). These characteristics represent qualities of the target at the time the target was attacked. All 434 incidents are coded for these eight characteristics. Categorical variables are recoded into dummy variables, which are variables that are binary coded (0, 1). The coding for EVIL DONE variables can be located in Table 3.2. The first EVIL DONE variable is exposure. Here, the coding captures the accessibility of the target, and its categories are adapted from Gruenewald, Gruenewald and Klein (2015). This variable is coded into three dummy variables. The three variables are coded one if a location is inaccessible without permission, if a location is accessible but rarely frequented by the public day or night and if the location is accessible and routinely frequented by the public day or night. A location that is routinely frequented by the public day or night would be a location such as a business that is open and busy during the day. These are locations that are frequented only during 55 the day or only during the night. A location that is frequented only during the night may include some bars or nightclubs that are not open during daytime hours. A location that is frequented only during the day would include many businesses. Locations that are accessible but rarely frequented by the public day or night include some government buildings as well as remote or secluded outside areas. The final category represents locations that are inaccessible to the public day or night and it includes private residences and some corporate headquarters. These three variables are being compared to the reference category, which is locations that are accessible and routinely frequented by the public day and night. The reference category includes outside locations and businesses that are open 24 hours a day, such as some department store locations like Wal-Mart, hotels and many convenience stores. A location is considered to be vital when if it were totally destroyed it would impact the day to day functioning of that community (Clarke & Newman, 2006). Vital targets often include power grids, transportation hubs and water supplies. Vital is operationalized into a binary variable. Victims who are in locations that if destroyed would have a great impact on the day to day functioning of that community receive a score of one. Many government buildings, court houses, and utility companies would be considered vital. The only gas station or grocery store in a small town would also be coded as a one and considered vital because destruction of this location would greatly impact the day to day functioning of that community. The reference category represents victims in locations that if destroyed would have no impact on the day to day functioning of their community are coded as a zero. For example, private residences, many outside areas and many businesses would receive a score of zero and would not be considered to be vital. 56 Table 3.2 EVIL DONE Codebook (n=434) Exposed Victim(s) in a location that is accessible and routinely frequented by public during both day and night Victim(s) in a location that is inaccessible to public without permission both during the day and night Victim(s) in a location that is accessible but rarely frequented by the public either during the day or night Victim(s) in a location that is accessible and routinely frequented by public during the day or night 0 1 2 3 Vital Victim(s) in a location that if eliminated would have no impact on day to day functioning of the community Victim(s) in a location that if eliminated would have a great impact on the day to day functioning of the community 0 1 Iconic Victim(s) in a location that is not iconic Victim(s) in a location that is iconic 0 1 Legitimate Victim(s) in a location that houses only general citizenry Victim(s) in a location that houses general citizenry and those working for target organization Victim(s) in a location that houses only those working for target organization 0 1 2 Destructible Victim(s) in a location that is easily destructible Victim(s) in a location that is difficult to destroy Victim(s) in a location that is moderately destructible 0 1 2 Occupied There are no other individuals around at time of incident Victim(s) in a location with 1-5 other people near them. Victim(s) in a location with 6-25 other people near them. Victim(s) in a location with 26-100 people near them. Victim(s) in a location with 101 or more people near them 0 1 2 3 4 Near Offender(s) lived in same city as target Offender(s) lived 101 or more miles from target Offender(s) lived outside city within 100 miles 0 1 2 Easy Victim(s) in a location with no security Victim(s) in a location with high security Victim(s) in a location with some security 0 1 2 The iconicity of a location is also binary coded into a dummy variable. Iconicity represents how iconic or representative a target is. Victims that were in a location that is 57 considered to be iconic receive a score of one and the reference category, victims who are in a location that is not iconic, receive a score of zero. Iconic locations include local buildings (religious building, police department, town hall), major commercial symbols, federal state and government buildings as well as major national and political symbols. Locations that are not iconic include many outside locations, private residences and commercial buildings. The operationalization for the legitimacy of a target is borrowed from Gruenewald and colleagues (2015). Legitimacy is coded into two dummy variables. Targets in the first variable receive a score of one if the victims are at a location that housed general citizenry and individuals working for the target organization and a zero is given to any other location types. An example of this type of location would be an abortion clinic. Anti-abortion extremists may attack a clinic and employees of the clinic represent the target organization but there may also be private citizen clients there who are harmed in the attack. With the second variable, locations receive a score of one if they house only individuals working for the target organization. Examples of these locations include military bases are highly attractive to anti-government extremists for the legitimacy variable since they almost exclusively house those who work for the target organization. These two types of legitimacy are compared to the reference category, which represents locations that housed only general citizenry. Locations that house only general citizenry include malls, department stores, office buildings and private residences. This variable examines the effect of the victims being at a location that houses only members of the target organization, or a location that houses target organization members and general citizens compared to locations housing only civilians, on the lethality of an incident. The destructible variable considers how easy it is to destroy the location of where the attack occurred. This variable is coded into two dummy variables. The first variable represents 58 incidents that occur at a location that is difficult to destroy. These locations include concrete buildings, multi-story buildings and large structures that would require large weapons or equipment in order for them to be destroyed. A code of one is given to all incidents that meet these criteria and all other incidents receive a code of zero. The second variable represents incidents that occurred at a location that is moderately difficult to destroy. Locations that are moderately difficult to destroy include many locations that would require small IEDs to destroy such as private residences and small businesses/buildings. Similarly, incidents that meet these criteria receive a score of one and all other incidents receive a score of zero. The reference category for the destructible variable is locations that are easily destructible. This includes locations that can be easily destroyed by weapons such as guns and homemade pipe bombs. Many of these locations are outdoor areas, small buildings and structures such as sheds. Lethality is examined to see differences between locations that are moderately difficult or very difficult to destroy as compared to the reference category, locations that are can be easily destroyed. Occupied represents the number of individuals, other than the victims and suspects, who are at the location at the time of the attack. Prior research has measured the occupied variable by whether or not any individuals are at the location (Gruenewald et al., 2015), but for this study since all incidents involve at least one death there are individuals at all locations. This variable is coded into four dummy variables. Each appropriate occupied variable is coded one if there were 1-5 others around, 6-25 others around, 26-100 others around or 101 more others around. Locations with 1-5 others around often include small businesses and private residences whereas locations with 6-25 others around include larger businesses, parks and some outside areas. Similarly, locations with 26-100 others around include large department stores and businesses. Finally, locations with 101 or more others around include concerts, movie theaters, large multi59 story office buildings, and special occasions or events that involve a large gathering of people. With each of these variables the remaining category is coded as zero. These variables are compared to the reference category which are incidents that occurred in locations with no other individuals around. Many of the locations where no other individuals are near the victim(s) are private residences and remote areas. The operationalization of the near variable is borrowed from Gruenewald and colleagues (2015). Near is coded into two dummy variables. For the first variable, incidents where the attacker lives within 100 miles of the target (but not in the same city) are coded as a one whereas all other incidents receive a score of zero. For the second variable, incidents where the attacker lives 101 or more miles from the target receive a score of one and all other incidents receive a score of zero. Both of these variables are compared to homicides that are committed by attackers that lived in the same city as the target. This coding is done by using Google Maps and inputting the address of the nearest suspect’s home with the address of the target. If a suspect’s home address is not available, then distance is computed between the attack location and the city the closest suspect lived in. Finally, easy is coded into two dummy variables. Homicides that occur at a location that have one form of security receive a score of one and homicides that occurred at a location with no security or that have multiple security measures are coded as zero. For a target to be coded as having some security then one security measure needs to be identified. Examples of security include metal detectors, CCTV, access/key cards and guards. This variable will examine the effect of a location having security, compared to the effect of a location having no security, on the lethality of an incident. Examples of locations with some security include many small businesses and offices. The second dummy coded variable includes targets that have high 60 security. For a target to be considered as having high security it must have two or more security measures in please. Examples of locations with high security include many government buildings, large office buildings, corporate headquarters and court houses. Coding and Reliability All EVIL DONE variables were coded by the primary researcher for this project. To ensure reliability, an intercoder reliability assessment was done. Intercoder reliability looks at how accurate independent coders are at evaluating the same data and reaching similar conclusions (Tinsley & Weiss, 2000). Another graduate student researcher, who has previously conducted work with the EVIL DONE framework, was trained using this codebook. He was given a set of randomly selected incidents to code (N=50). Table 3.3 Reliability Analysis of EVIL DONE Chronbach's Alpha Exposed 0.955 Vital 0.814 Iconic 0.824 Legitimate 0.817 Destructible 0.831 Occupied 0.916 Near 0.992 Easy 0.926 Cronbach’s alpha is a measure of reliability to determine internal consistency to see how consistent two ratings for the same measure are. Table 3.3 presents the results of the intercoder reliability analysis and Cronbach’s alpha for each of the eight EVIL DONE variables. In most social science research, a Cronbach’s alpha measure of .70 or greater is considered to have high 61 internal consistency and be an acceptable score. As you can see in Table 3.3, measures for all eight of the EVIL DONE variables fall well within this range. Victim, Suspect and Ideological Variables This study examines several characteristics related to the victim, suspect and ideological motivation for the attack. This section will discuss the coding of these variables and how it draws on extremist and homicide literature. The coding for all of these variables can be found in Table 3.4. For multivariate analysis, categorical variables are dummy coded. Dummy coded variables are variables that are binary coded (0 or 1). Six characteristics of victims are examined. If an incident had multiple victims the most common/consistent characteristic for each variable is coded. For example, if seven individuals died in an incident and five of them were females then that ‘victim’ is coded as female. For the age variable, the average is taken of all known victim ages. It is important to note that the coding for many victim variables measures the presence or absence of an attribute. The victim coding process is conducted in a similar manner to Parkin and Freilich (2015)’s study examining homicides committed by right-wing extremists. They argue that news articles have no reason to report a negative response. For example, if a victim is a police officer it is expected that media reports will indicate this. On the other hand, it is not expected that reports will identify that an individual is ‘not a police officer.’ In other words, the default response would be a negative response. The frequency results of affirmative responses should then be considered the minimum number for the presence of that characteristic. Several characteristics of victims are examined including sex, race, age, occupation, and victim/suspect relationship. The sex of the victim is coded as male or female, with men receiving 62 a score of one and women a score of zero. Race is also coded as a binary variable with whites receiving a score of one and non-whites receiving a score of zero. Age is coded into three dummy variables. Victims who are 17 and under, victims who are 25-49 and victims who are 50 or older are compared to the reference group of victims aged 18-24 years old. The young are argued to be more at risk of homicide victimization (Breaull & Kposowa, 1997; Caywood, 1998; Messner & Tardiff, 1985). Individuals are often more active during the ages of 18-24 and potentially involved in more dangerous situations. Similarly, research has found that younger individuals are more at risk to being a victim of terrorism due to their lifestyle choices such as staying out late and utilizing public transportation (Canetti-Nisim et. al, 2006). 63 Table 3.4 Victim, Suspect and Ideology Codebooks Victim Codebook Suspect Codebook Gender Gender Female= 0 Female= 0 Male= 1 Male= 1 Ideology Codebook Suspect Ideology Jihadist= 0 Right-Wing= 1 Race White= 0 Non-white= 1 Race White= 0 Non-white= 1 Motivation Non-Ideological= 0 Ideological= 1 Age 17 and under= 1 *18-24= 1 25-49= 2 50+= 3 Age *14-24= 0 25-49= 1 50+= 3 Occupation Unemployed= 0 Blue collar= 1 White collar= 2 *Police/government= 3 Occupation *Unemployed= 0 Blue collar= 1 White collar= 2 Victim-Suspect Relationship Strangers= 0 Victim knew suspect= 1 Number of Suspects *One suspect= 0 Two suspects= 1 Three suspects= 2 Lone Wolf Not a lone wolf= 0 Lone wolf= 1 Weapon Other= 0 Firearms= 1 *some categories collapsed to conduct multivariate analyses 64 Occupation is coded into three dummy variables. Victims who are blue collar workers, white collar workers, and those who are unemployed are compared to the reference category of victims who are police/government employees. Police, military and government officials are often viewed as legitimate targets. However, police and government personnel are also often considered “hard” targets, meaning they are difficult to successfully attack (Seifert & McCauley, 2014). If the occupation of the victim is unknown then this is coded as missing. The occupation variable is excluded from multivariate analysis because 43% of all cases are missing a code for the victim’s occupation. A separate analysis is conducted that just examines all incidents that there is a code for the victim’s occupation and that model is discussed. Finally, the relationship between the suspect and victim is coded in a binary manner to compare lethality between victims who knew the attacker (1) and those who did not (0). Victims are coded as knowing the suspect if they had any form of relationship with the suspect including romantic relationship, familial relationship, friendship, co-worker relationship or acquaintances. There are six suspect level variables that are examined. The gender of the suspect is also coded into a dummy variable. Males are coded as a one and females as a zero, which will allow for the effect of the suspect being female to be compared to the effect of the suspect being male. Age is coded into two dummy variables. Suspects 25-49 years old and suspects 50 and older are compared to the reference category of suspects who are between the age of 14-24. Some research argues that young men are more likely to commit extremist acts (LaFree & Dugan, 2004). This coding will allow for there to be a comparison in the lethality of homicides committed by suspects who were between 25-49 years old and suspects 50+ years old and suspects who are 1424 years old. The race of a suspect is also coded into a dummy variable. Suspects who are white receive a score of one and suspects who are non-white receive a score of zero. This coding will 65 help determine the effect of the suspect being white, compared to the suspect being non-white on the lethality of an incident. The occupation of suspects is coded into two dummy variables. Blue collar suspects and white-collar suspects are compared to the reference category of unemployed suspects7. Similar to the victim’s occupation, the occupation of the suspect is missing for many of the cases (56%). This variable is excluded from multivariate analysis. A separate model is conducted that examines cases with information on the suspect’s occupation and no significant relationships are found for the occupation variable. There are two variables related to the number of suspects involved in the attack. First, the number of suspects involved in the incident is dummy coded into two variables. Incidents that involved two suspects and incidents that involved three or more suspects are compared to a reference category of incidents with only one suspect. The second variable examining the number of suspects looks at whether the incident was committed by a lone wolf suspect. The operationalization of this variable is based on prior research (Gruenewald et al., 2013, Pantucci, 2011; Spaaij, 2010). A suspect is considered to be a lone wolf suspect if they meet the following three criteria: operated alone at all stages of the attack, the suspect was not a member of an extremist or hate group and the actions of the extremist were conceived of and completed without the direction of any external sources (Gruenewald et al., 2013). Additionally, only suspects that are committing an ideologically motivated attack are considered to be lone wolves. Whether or not there is group affiliation is not relevant if the attack was committed for nongroup reasons. This variable is dummy coded with lone wolf suspects being coded as one and all other suspects receiving a score of zero. 7 The police/government occupation category is collapsed into the blue collar since only five suspects were coded as police/government workers. 66 The final suspect level variable is the weapon the suspect chose to use in the attack. This is coded in a dichotomous manner with one indicating that the attack was committed using a firearm and a zero indicating that the attack was committed using some other type of weapon. The other type of weapon can vary drastically and include weapons such as knives, blunt objects and bombs. It is worth noting that ideally it would be best to look at this variable to compare differences based on each different type of weapon choice but due to the makeup of the variable there are two few cases in some of the categories in order to conduct that type of analysis. It should still be fruitful to examine firearms vs. other types of weapons and see what differences in lethality exist. There are two variables that examine the ideological motivation of the incident. The first dummy variable that is coded represents the ideological affiliation of the suspect involved in the incident. Suspects who are far right extremists receive a score of one and jihadists receive a score of zero. Individuals who are labeled as adhering to a far-right belief system generally have the following ideals: fiercely nationalistic, anti-global, a desire to fight for individual liberty, suspicion of centralized federal authority, a belief in conspiracy theories, a need to be prepared for an attack and a belief that his or her way of life is being threatened. For an individual to be labeled as a jihadist he or she typically adheres the following ideals: only acceptance of Islamic faith, belief in jihad, rejection of traditional Muslim respect for ‘People of the Book’ (Christians and Jews), belief in the Islamic law- Sharia, belief that the Islamic faith is under attack and a general anti-West or anti-United States belief system. Additionally, many jihadists believe that people from the West (U.S. primarily) are to be held responsible for the actions of their governments and culture and a believe that it is their religious obligation to promote violent Islamic revolution to combat corruption, oppression and assault on Islam by the West. 67 The motivation for the incident is also examined. This is coded into a dummy variable. Incidents that were perpetrated for ideological reasons receive a score of one and incidents that were perpetrated for non-ideological reasons receive a score of zero. A descriptive presentation of motives for attacks can be found in Table 3.5. Attacks that are committed for an ideological purpose seek to promote a specific agenda or ideology whereas attacks committed for nonideological reasons are more routine homicides committed by domestic extremists for other purposes. There are several types of attacks committed for ideological purposes which include, but are not limited to: anti-race, anti-gay, anti-sex offender, anti-immigration, anti-global, antifederal government, anti-homeless, abortion related and promoting a global or local jihad. Table 3.5 Incident Motivation Frequencies All Cases (n=434) Ideological Incidents (n=256) N % N % N % Domestic dispute 34 7.8% - - 34 17.2% Evading arrest 6 1.4% - - 6 3.0% Robbery 28 6.5% - - 28 14.1% Gang related/group member 54 12.4% 2 0.9% 52 26.1% Work dispute 4 0.9% - - 4 2.0% Drug related 7 1.6% - - 7 3.5% Non-ideological other 57 13.1% - - 57 28.6% Honor killing 15 3.5% 4 1.7% 11 5.5% Anti-government/law enforcement 40 9.2% 40 17.0% - - Promote jihad 31 7.1% 31 13.2% - - Anti-race 97 22.4% 97 41.1% - - Anti-homeless 16 3.7% 16 6.8% - - Anti-abortion 6 1.4% 6 2.6% - - Anti-gay 23 5.3% 23 9.8% - - Anti-female 3 0.7% 3 1.3% - - Anti-sex offender 3 0.7% 3 1.3% - - Ideological other 10 2.3% 10 4.3% - - 68 Non-Ideological Incidents (n=222) Data Analysis The data used for this study was imported from Microsoft Access into a Microsoft Excel workbook where it was cleaned and coded. These data were then imported into SPSS for analysis. The first research question was examined through descriptive statistics to look at each individual EVIL DONE variable. This examination will determine if these factors of vulnerability are evident within incidents of lethal domestic extremist violence. Gruenewald and colleagues (2015) applied the EVIL DONE framework to domestic attacks committed by ecoterrorists and this study expands on that by applying it to attacks committed by jihadists and right-wing extremists. Additionally, EVIL DONE characteristics are examined to determine differences based on ideology, suspect characteristics and victim characteristics. Significance bivariate results are presented in Chapter four. Bivariate analysis was used to test relationships between EVIL DONE variables and victim and suspect related characteristics with the dependent variable lethality and the results of this are presented in Chapter five. Binary logistic regression is utilized to examine the EVIL DONE vulnerability characteristics to see which of them are the strongest predictors of lethality. A binary logistic regression analysis is conducted because the dependent variable is not normally distributed. Binary logistic regression is the appropriate analytical technique to use when the dependent variable is dichotomous (Long, 1997). It is used to examine the conditional expectations of the dependent variable (lethality) given all of the independent variables being held constant (Bachman & Paternoster, 2003). The dependent variable could be examined through ordinal logistic regression and the variable could be split into three categories consisting of one death, two to five deaths and six or more deaths in order to keep some of the variation. Several ordinal logistic regression models were conducted with the data but due to the 69 distribution (too few cases in the latter categories) the data was failing key tests of ordinal logistic regression, such as the test of parallel lines. Therefore, the most appropriate analysis with the distribution of this dependent variable was to code the dependent variable in a dichotomous manner and conduct binary logistic regressions. A binary logistic regression is conducted with the EVIL DONE factors with controlling for suspect and victim characteristics to see which EVIL DONE factors remain strong predictors of lethality. These results are presented in Chapter five. Conclusion The following chapters present the results from the univariate, bivariate and multivariate analyses. Chapter four presents descriptive statistics for EVIL DONE and answers the first research question. Relationships between EVIL DONE characteristics and victim and suspect characteristics are also presented. The fifth chapter addresses the second research question and looks at the relationship between vulnerability and lethality. Specifically, this chapter presents multivariate results that look at which EVIL DONE variables are the strongest predictors of lethality when controlling for victim and suspect characteristics. 70 Chapter 4: EVIL DONE Variables and Other Vulnerability Factors This chapter addresses the first research question by examining what factors of vulnerability are present in prior lethal domestic extremist attacks and how these factors vary across ideology, suspect characteristics and victim characteristics. This chapter begins with presenting descriptive findings for all independent variables which include ideological factors, suspect factors and victim factors. Many of these factors have previously been found to be related to vulnerability. This chapter presents a picture of the general characteristics and nature of domestic extremist attacks that occurred in the United States between 1990-2014. Second, this chapter presents descriptive statistics for the eight EVIL DONE vulnerability factors. Prior research argues that these eight factors are related to the vulnerability of a terrorist target (Clarke & Newman, 2006). This study takes the EVIL DONE framework and applies it to human targets. This framework has been applied at the international level (Ekici et al., 2008; Ozer & Akbas, 2011; Paton, 2013) and at the domestic level for eco-terror incidents (Gruenewald et al., 2015). This study fills an important gap by examining these factors of vulnerability at the domestic level and including victim and suspect characteristics, which prior research has failed to do. Third, all EVIL DONE variables are examined by whether the incident was ideologically motivated and the across suspect’s ideological affiliation8. This will determine if there are differences in what is vulnerable about victims who are killed in ideological incidents versus those who are killed for a non-ideological reason. It also expands on Gruenewald and colleagues’ (2015) research which applied EVIL DONE to eco-terror incidents, by now applying the 8 All variables were subjected to collinearity diagnostics and all VIF scores are in an acceptable range of 3.2 or below. 71 framework to far right and jihadist incidents. Finally, this chapter compares EVIL DONE vulnerability factors across suspect and victim characteristics such as gender, race, age, occupation, victim/suspect relationship, number of suspects, weapon used, and whether it was a lone wolf attack. Descriptive Findings for Ideology, Suspect and Victim Characteristics All descriptive statistics for victim characteristics, suspect characteristics and ideology are located in Table 4.1. There are several interesting findings with the victim characteristics. The majority of victims in this study are white males between the ages of 25 and 49. Nearly three-fourths of all victims are male (75.3%). Caucasian victims represent 65.2% of all victims. These findings are similar to prior victimization research which has found men to be more likely victims of homicide (Caywood, 1998; Fox & Zawit, 2007; Kposowa & Breaull, 1998; Messner & Tardiff, 1985; Silverman & Kennedy, 1987). This study finds Caucasians to be more likely victims of extremist homicide. However, prior research has found minorities to be more at risk (Ezell & Tanner-Smith, 2009; Breaull & Kposowa, 1997). This may vary based on the ideological motivation of the homicide and if the homicide is committed by an extremist or not. In a study that examined attacks committed by far-right suspects, Parkin and Freilich (2015) found that white males were more likely to be victims of non-ideological homicide. Interestingly, this study shows that middle aged persons are the most likely victims of extremist homicide. Prior research on victimization has found that younger individuals are at a greater risk for victimization (Breaull & Kposowa, 1997; Messner & Tardiff, 1985). It is difficult to determine the occupation of most victims because this information is infrequently discussed in open source materials. There is no clear occupation of the victim in 72 43.3% of the incidents. Of the incidents where the occupation is identifiable (n=246), 32.9% of all victims are unemployed or homeless. One-fourth of all victims are police or government employees, while 20% work a blue-collar job and 23% work in a white-collar position. Prior research has found that those working in the service industry (blue collar jobs) are more likely to be victims of homicide (Breaull & Kposowa, 1997) and students (coded as unemployed) have also been found to be more likely victims of terrorist attacks (Canetti-Nisim et al., 2006). Over one-half (56.5%) of victims did not know their attacker. Freilich and Parkin (2015) found that victims of right-wing extremist attacks were more likely to know their suspect. This is an important variable to examine when looking at differences between ideologically motivated and non-ideologically motivated attacks. 73 Table 4.1 Descriptive Findings for Independent Variables (n= 434) N % Victim Characteristics Gender Female 107 24.7% Male 327 75.3% Race Caucasian African American Other 283 78 73 65.2% 18.0% 16.8% 24 and under 25-49 50+ 107 241 85 24.7% 55.5% 19.8% Unemployed/homeless Blue collar White collar Police/government 81 48 57 60 32.9% 19.5% 23.2% 24.4% Strangers Victim knew suspect 245 189 56.5% 43.5% Age Occupation Victim-Suspect Relationship *occupation variable excludes missing cases 74 Table 4.1 (cont’d) N % Female Male 4 430 0.9% 99.1% Caucasian African American Other 362 29 43 83.4% 6.7% 9.9% 14-24 25-49 50+ 152 251 32 35.0% 57.6% 7.4% Unemployed/homeless Blue collar White collar 98 64 30 51.0% 33.4% 15.6% One suspect Two suspects Three or more suspects 234 101 99 53.9% 23.3% 22.8% Not a lone wolf Lone wolf 325 109 74.9% 25.1% Other Firearms 154 280 35.5% 64.5% Suspect Characteristics Gender Race Age Occupation Number of suspects Lone wolf Weapon choice *occupation variable excludes missing cases 75 Table 4.1 (cont’d) N % Ideology Suspect ideology Jihadist Right Wing 60 374 13.8% 86.2% Non-ideological Ideological 199 235 45.9% 54.1% Incident ideology *occupation variable excludes missing cases The majority of suspects are white males between the ages of 25 and 49. The finding that suspects are more likely to be men is consistent with prior homicide research (Bloom et al., 2012; Gruenewald et al., 2009; Gruenewald, 2011; LaFree & Dugan, 2004; Silverman & Kennedy, 1987; Smith & Morgan, 1994). In regards to age, the finding that most suspects are middle-aged is very interesting. Prior research in regards to the average age of terrorists is conflicting. Some research has found that terrorist suspects are more likely to be young (LaFree & Dugan, 2004) and some has found suspects are more likely to be older (Smith & Morgan, 1994; van Zuijdewijn & Bakker, 2016; Gruenewald et al., 2013). The occupation of the suspect is not known for 56% of all incidents. Of the incidents where the suspect’s occupation is identifiable, 51% of the suspects are unemployed or homeless. One-third of all suspects work a blue-collar job and 15% have a white-collar job. There has been a recent growing concern on lone wolf attacks. Lone wolves are individuals who operated alone at all stages of attack, are not a member of an extremist or hate group and who commit the attack without the direction of external resources. One-fourth of all homicides in this study are committed by a lone wolf suspect. Of the attacks that are ideologically motivated, 54% of them are committed by lone wolf 76 suspects. In nearly two-thirds of all attacks examined, the suspect uses a firearm as opposed to another type of weapon. This finding is not surprising given how easily accessible firearms are in the United States. In regards to ideology, 86.2% of all attacks are committed by right-wing extremists, compared to 13.8% of attacks committed by jihadists. There is nearly an even split in terms of the motivation for the incident with 54.1% of incidents being ideologically motivated and 45.9% being motivated by other reasons. Ideological motivations include anti-law enforcement/government, anti-race, anti-homeless, anti-abortion, anti-gay, anti-female and antisex suspect attacks. Incidents that are not ideologically motivated have very different motivations including domestic disputes, evading arrest, robbery, drug-related attacks and gang member attacks. EVIL DONE Descriptive Findings All EVIL DONE variables are examined to assess what vulnerability exists in prior incidents of domestic extremism (n=434). These descriptive statistics can be found in Table 4.2. The exposure variable includes elements of accessibility and whether the site of the attack is frequented by people (Clarke & Newman, 2006; Gruenewald et al., 2015). Findings indicate that nearly 34% of victims are at a location that is inaccessible to the public without permission day and night, and 12% of victims are at a location that is accessible to the public but rarely frequented during the day or the night. Many of the locations that homicides occur at are private residences that are not accessible to the public day or night. Locations that are accessible but rarely frequented are often wooded areas or deserted public areas that people could frequent but rarely do. Incidents that occur in locations where the public frequented day or night represent 77 14.1% of the all incidents. These include areas that are frequented only during certain times of the day. For example, this may include stadiums, court houses and many businesses that are only frequented during the day time. Finally, 40.1% of victims are at a location that is accessible to the public and frequented often day and night. These locations include busy streets, gas stations, parks and train stations. These findings offer support of a relationship between exposure and vulnerability; exposure of a target increases its vulnerability. When applying, EVIL DONE to international incidents, Paton (2013) found that the majority of locations are not highly exposed. However, on a domestic level these findings are similar to Gruenewald and colleagues (2015) who found that environmental extremists preferred targets that were accessible. 78 Table 4.2 EVIL DONE Descriptive Results (n=434) N % 174 40.1% 147 33.8% 52 12.0% 61 14.1% 288 66.4% 146 33.6% Victim(s) in a location that is not iconic 384 88.5% Victim(s) in a location that is iconic 50 11.5% Victim(s) in a location that houses only general citizenry 336 77.4% Victim(s) in a location that houses general citizenry and those working for target organization 83 19.1% Victim(s) in a location that houses only those working for target organization 15 3.5% Victim(s) in a location that is easily destructible 195 44.9% Victim(s) in a location that is difficult to destroy 25 5.8% Victim(s) in a location that is moderately destructible 214 49.3% There are no other individuals around at time of incident 237 54.6% Victim(s) in a location with 1-5 other people near them. 100 23.0% Victim(s) in a location with 6-25 other people near them. 49 11.3% Victim(s) in a location with 26-100 people near them. 23 5.3% Victim(s) in a location with 101 or more people near them 25 5.8% Exposed Victim(s) in a location that is accessible and routinely frequented by public during both day and night Victim(s) in a location that is inaccessible to public without permission both during the day and night Victim(s) in a location that is accessible but rarely frequented by the public either during the day or night Victim(s) in a location that is accessible and routinely frequented by public during the day or night Vital Victim(s) in a location that if eliminated would have no impact on day to day functioning of the community Victim(s) in a location that if eliminated would have a great impact on the day to day functioning of the community Iconic Legitimate Destructible Occupied 79 Table 4.2 (cont’d) N % Offender(s) lived in same city as target 217 50.0% Offender(s) lived 101 or more miles from target 52 12.0% Offender(s) lived outside city within 100 miles 165 38.0% Victim(s) in a location with no security 328 75.6% Victim(s) in a location with high security 14 3.2% Victim(s) in a location with some security 92 21.2% Near Easy The second variable is how vital the location is in terms of the impact it has on the community (Clarke & Newman, 2006). Boba (2009) describes the vitality of targets as being important in that terrorists may not always try to target specific people but may be trying to paralyze or demoralize a community. Two-thirds of the incidents (66.4%) occur at locations that if destroyed would have no impact on the local community (or beyond, i.e. regionally or nationally). Locations that are not vital include private residences and many small businesses. There are several reasons why only one-third of attacks occur at vital locations. Arguably, locations that are vital to the functioning of society are likely to have higher security than nonvital locations. High security may act as a deterrent to attackers. Additionally, ideology may play an important role in target choice. Attackers committing an ideologically motivated homicide may be more likely to target a vital location than attackers committing a homicide for other reasons. Iconicity is the third measure of EVIL DONE vulnerability and examines whether a victim is at a location that holds symbolic value. Terrorists often consider the audience of a 80 target beyond the victims (Fussey, 2011) so will choose targets that are representative of their cause and that may gain them favor or support. However, the majority (88.5%) of victims in this study are not attacked at a location that is considered to be iconic. Many iconic locations have higher security so may be more difficult to attack. The next variable of vulnerability is the legitimacy of a target. This variable is measured by examining the types of individuals at the location that is targeted. Terrorists are expected to be more likely to target places or people whom they deem to be legitimate targets and want to pick a target that will gain an acceptable public reaction (Clarke & Newman, 2006). The government, as well as military institutions, are often viewed as being legitimate targets (Clarke & Newman, 2006; Santifort et al., 2013). Most attacks occur at locations that house only general citizenry (77.4%). Nearly one-fifth of attacks (19.1%) occur at a location that included both citizens and individuals working for the target organization. Only 3.5% of attacks targeted locations that housed only individuals working for that target location. This is an interesting finding because prior research has argued that police/military/government are viewed as legitimate targets. However, these types of individuals may be more difficult to attack than average citizens and this may deter suspects. Additionally, terrorists may be interested in targeting a civilian population to instill fear and promote their ideology. Destructibility is related to how easily the location of the attack could be destroyed. Only 5.8% of victims are at a location that is difficult to destroy. Locations that are difficult to destroy include large office buildings, stadiums and many schools. Most victims (49.3%) are at a location that is moderately destructible. These locations include small stores, most homes, and small buildings. Incidents where victims are at a location that is easily destructible represent 44.9% of all incidents coded. Many of these incidents are located in outdoor areas. These 81 findings make sense since locations that are easily destructible are expected to be more attractive choices. Very few attacks occur at a location that is considered to be difficult to destroy (large office buildings, stadiums, military bases, etc.). Suspects are going to commit attacks they can successfully complete using weapons they can easily obtain. The next measure of vulnerability is whether or not a target is occupied. Most victims (54.6%) are at a location where no other individuals are around them when they are attacked. Just over one-fifth, 23%, of individuals are at a location with 1-5 other individuals near them. Only 5.3% of incidents occur at a location with 26-100 others around and 5.8% of incidents involved victims in locations where there are 101 or more other people around. It is expected that suspects would be more likely to target places that have large crowds, however, only 11.1% of attacks occur at a location with 26 or more others around. Suspects may fear they will be more likely to be detected in crowded locations. Near is the seventh variable of vulnerability and examines where the suspect lived during the time of the attack. Criminological studies generally find that suspects commit crimes near their homes (Eck, 1993) because they are familiar with that area (Bernasco, 2010). Smith (2008) found that terrorists generally commit attacks within 30 miles of where they reside. These findings are consistent with this prior research. Findings show that one-half of the extremists live in the same city as the target and 38% live within 100 miles of the target. Only 12% of suspects live 101 or miles away from the location of the incident. The final variable examined is how easy the target is to attack, which is a representation of the amount of security at the location where the victim is at the time of the attack. There are many security obstructions that might hinder the suspect, including surveillance systems, locks, 82 armed guards, CCTV, alarm systems and key card entry to buildings. Suspects often choose targets with little to no security. The majority (75.6%) of victims are at a location that has no known security. Many of these locations include outside areas and private residences. Nearly one-fifth (21.2%) of incidents involved victims at a location with some security, such as local businesses, and only 3.2% of incidents are located in places with high security, such as government buildings and corporate headquarters. Not surprisingly, most locations have no security. Suspects are often avoiding locations with security because they want to remain undetected. EVIL DONE and Ideological Motivation This section examines the eight EVIL DONE characteristics based on the ideological motivation for the incident. Ideological motivation is examined for all incidents and then it is examined separately to compare right-wing ideologically motivated incidents to right-wing nonideologically motivated incidents and jihadist ideologically motivated incidents to jihadist nonideologically motivated incidents. All of the EVIL DONE vulnerability variables are significantly related to the ideological motivation of the incident when examining all incidents. This means that there are significant differences in the types of targets extremists are attacking based on ideology. The relationships that are found indicate that the proportional distributions between the variables are significant but it does not show what these differences are. This means we do not know the direction of the relationship but just that there is a significant difference in proportional distributions in vulnerability between ideologically motivated attacks and nonideologically motivated attacks. These results can be found in Table 4.3. 83 The chi square statistic for exposure and the ideological motivation of the incident is significant for all incidents. Incidents that are not ideologically motivated rank low on exposure. Nearly one-half (45.2%) of non-ideologically motivated incidents occur at a location that is inaccessible to the public without permission day and night. Many of these incidents occur in private residences. However, most of the incidents that are ideologically motivated occur at a location that is routinely frequented by the public day and night (45.5%). Most of these incidents occur in an outside area or a business. This may be related to the notion that ideologically motivated crimes are more likely to be attacks against strangers and it is easier for suspects to find these potential victims (strangers) in public locations. Whereas, many non-ideologically motivated attacks are against individuals the suspect knows (i.e. family and friends) so suspects may have easier access to their victims and can attack them in locations not easily accessible to the general public. Similarly, there is a significant relationship between exposure and ideological type for right-wing and jihadist incidents. The proportional differences are similar in comparison to all ideologically motivated incidents and all non-ideologically motivated incidents. This indicates that regardless of type of motivation (right wring or jihadist) that in ideologically motivated incidents suspects are more likely to attack victims at locations that are highly exposed as compared to suspects in non-ideologically motivated incidents. Vitality of the location is also compared with ideological motivation across all incidents, right-wing incidents, and jihadist incidents. Extremists are more likely to target locations that hold a symbolic value or relevance (Asal et al., 2009; Clarke & Newman, 2006; Crenshaw, 1998). Nearly one-half (44.3%) of ideologically motivated attacks occur at a location that is vital, a location that if eliminated would have an impact on the community, but only 21.1% of 84 non-ideologically motivated attacks occur at a location that is considered vital. There are similar proportional distributions across right-wing only and jihadist only incidents. Interestingly, it appears that proportionally speaking, ideologically motivated jihadist attacks are more likely to be against a vital target (68.2% of attacks) compared to only 38.7% of ideologically motivated right-wing attacks. Victims that are killed in ideologically motivated incidents are more likely to be at a location that is considered vital as compared to victims killed in incidents that are in nonvital locations. Both Paton (2013), who examined international attacks, and Gruenewald and colleague’s (2015) who examined environmental extremist incidents found that vitality did not play a key role. When examining all incidents in this study, vitality did not play an important role since only one-third of all incidents are attacks against vital targets. However, vitality plays an important role in attacks that are ideologically motivated. 85 Table 4.3 EVIL DONE and Ideological Motivation (n=434) Ideological Motivation Ideologically Motivated (n= 235) NonIdeologically Motivated (n=199) Right Wing Only Ideologically Motivated (n=191) Jihadist Only NonIdeologically Motivated (n=183) Ideologically Motivated (n=44) NonIdeologically Motivated (n= 16) N % N % N % N % N % N % 107*** 45.5% 67 33.7% 83* 43.5% 64 35.0% 24 54.5% 3*** 18.8% 57 24.3% 90 45.2% 52 27.2% 78 42.6% 5 11.4% 12 75.0% 28 11.9% 24 12.1% 26 13.6% 24 13.1% 2 4.5% 0 0.0% 43 18.3% 18 9.0% 30 15.7% 17 9.3% 13 29.5% 1 6.3% 131*** 55.7% 157 78.9% 117*** 61.3% 145 79.2% 14** 31.8% 12 75.0% 104 44.3% 42 21.1% 74 38.7% 38 20.8% 30 68.2% 4 25.0% 193*** 82.1% 191 96.0% 165*** 86.4% 176 96.2% 28* 63.6% 15 93.8% 42 17.9% 8 4.0% 26 13.6% 7 3.8% 16 36.4% 1 6.3% Exposed Victim(s) in a location that was accessible and routinely frequented by public during both day and night Victim(s) in a location that was inaccessible to public without permission both during the day and night Victim(s) in a location that was accessible but rarely frequented by the public either during the day or night Victim(s) in a location that was accessible and routinely frequented by public during the day or night Vital Victim(s) in a location that if eliminated would have no impact on day to day functioning of the community Victim(s) in a location that if eliminated would have a great impact on the day to day functioning of the community Iconic Victim(s) in a location that is not iconic Victim(s) in a location that is iconic ***p<.001, **p<.01,*p<.05 86 Table 4.3 (cont’d) Ideological Motivation Ideologically Motivated (n= 235) NonIdeologically Motivated (n=199) Right Wing Only Ideologically Motivated (n=191) Jihadist Only NonIdeologically Motivated (n=183) Ideologically Motivated (n=44) NonIdeologically Motivated (n= 16) N % N % N % N % N % N % 163*** 23.7% 173 17.4% 143** 19.8% 159 17.2% 20* 11.1% 14 17.5% 59 8.6% 24 2.4% 42 5.8% 23 2.5% 17 9.4% 1 1.3% 13 1.9% 2 0.2% 6 0.8% 1 0.1% 7 3.9% 1 1.3% Victim(s) in a location that is easily destructible 115*** 48.9% 80 40.2% 93 48.7% 77 42.1% 22** 50.0% 3 18.8% Victim(s) in a location that is difficult to destroy 20 8.5% 5 2.5% 11 5.8% 4 2.2% 9 20.5% 1 6.3% Victim(s) in a location that is moderately destructible 100 42.6% 114 57.3% 87 45.5% 102 55.7% 13 29.5% 12 75.0% Legitimate Victim(s) in a location that houses only general citizenry Victim(s) in a location that houses general citizenry and those working for target organization Victim(s) in a location that houses only those working for target organization Destructible ***p<.001, **p<.01,*p<.05 87 Table 4.3 (cont’d) Ideological Motivation Ideologically Motivated (n= 235) NonIdeologically Motivated (n=199) Right Wing Only Ideologically Motivated (n=191) Jihadist Only NonIdeologically Motivated (n=183) Ideologically Motivated (n=44) NonIdeologically Motivated (n= 16) N % N % N % N % N % N % 111*** 47.2% 126 63.3% 97*** 50.8% 116 63.4% 14 31.8% 10 62.5% 50 21.3% 50 25.1% 38 19.9% 46 25.1% 12 27.3% 4 25.0% 34 14.5% 15 7.5% 32 16.8% 14 7.7% 2 4.5% 1 6.3% 19 8.1% 4 2.0% 13 6.8% 4 2.2% 6 13.6% 0 0.0% 21 8.9% 4 2.0% 11 5.8% 3 1.6% 10 22.7% 1 6.3% Occupied There are no other individuals around at time of incident Victim(s) in a location with 1-5 other people near them Victim(s) in a location with 6-25 other people near them Victim(s) in a location with 26-100 people near them Victim(s) in a location with 101 or more people near them Near Offender(s) lived in same city as target 106** 45.1% 111 55.8% 92 48.2% 104 54.2% 14 31.8% 7 43.8% Offender(s) lived 101 or more miles from target 39 16.6% 13 6.5% 18 9.4% 9 4.7% 21 47.7% 4 25.0% Offender(s) lived outside city within 100 miles 90 38.3% 75 37.7% 81 42.4% 79 41.1% 9 20.5% 5 31.3% 155*** 66.0% 173 86.9% 137*** 71.7% 159 86.9% 18** 40.9% 14 87.5% Victim(s) in a location with high security 12 5.1% 2 1.0% 7 3.7% 1 0.5% 21 47.7% 1 6.3% Victim(s) in a location with some security 68 28.9% 24 12.1% 47 24.6% 23 12.6% 5 11.4% 1 6.3% Easy Victim(s) in a location with no security ***p<.001, **p<.01,*p<.05 88 The chi square statistic for iconicity is significant, indicating that there is a relationship between the iconicity of the location the victim was in at the time of the attack and whether or not the attack is ideologically motivated. Significance is found when examining all incidents together, right-wing only incidents and jihadist only incidents. For all incidents, in incidents that are not ideologically motivated, victims are at a location that is iconic only 4% of the time, compared to 17.9% in incidents that are ideologically motivated. Jihadists appear to be more likely to attack iconic targets. More than one-third (36.4%) of ideologically motivated attacks committed by jihadists are at iconic locations whereas only 13.6% of ideologically motivated attacks committed by right-wing extremists are at similar locations. These findings illustrate that ideological incidents are more likely to rank higher on the iconicity measure of EVIL DONE. In examining environmental extremists’ targets, Gruenewald and colleagues (2015) found iconicity to be somewhat important, attacks committed are often targeted at local agricultural operations or local commercial targets, targets important to the attackers. When examining international incidents, however, Paton (2013) found iconicity of targets to be one of the least predictive factors in terms of level of lethality of an incident. Therefore, iconicity may not be one of the most important vulnerability predictors for EVIL DONE or target vulnerability but it appears that it is more relevant in ideologically motivated attacks than in non-ideologically motivated attacks. A significant relationship was found between the legitimate variable and ideological motivation for all incidents, for right-wing only incidents, and for jihadist only incidents. The majority of attacks that are not ideologically motivated are attacks that occur at locations that house only general citizenry (86.9%), compared to ideologically motivated incidents that occur at locations that housed only general citizenry 69.4% of the time. This finding is not surprising, it 89 was expected that attacks would be more likely to occur at locations housing only general citizenry because these targets typically have less security and are easier to successfully attack. Terrorists may target soft targets (civilians) because they are value maximizing by trying to cause harm and to communicate fear (Asal et al., 2009) and targeting civilians for the shock value (Le Vine, 1997). Prior research on the legitimate variable is mixed. Paton (2013) found legitimacy to be one of the least predictive factors of EVIL DONE with most targets being civilians, however, Gruenewald and colleagues (2015) found it to be an important predictor in eco-terror incidents. A significant relationship is found, when examining all incidents, between destructibility of the location a victim was at when attacked and whether or not the attack was ideologically motivated. For incidents that are not ideologically motivated, the majority of incidents occur at a location that could be easily destroyed (57.3%) and 42.7% occur at locations that are moderately destructible. The latter category includes many residential locations and small businesses whereas the former includes many outdoor locations. Of the incidents that are ideologically motivated 49% of them occur at an area that can be easily destroyed, 42.5% at an area that is moderately destructible and 8.5% at a location that is difficult to destroy. These findings are consistent with what would be expected, in that suspects are more likely to attack locations that are easily destructible. Hoffman (1998) argues there is a new era of terrorism, and terrorists are seeking to cause mass destruction, which makes areas that are easily destroyable highly attractive options. Interestingly, for incidents that are not ideologically motivated suspects appear to be more likely to attack locations that are moderately destructible (compared to easily destructible). This may be the case because many of these locations are small businesses and homes that the suspect may target because of their relationship with the victim. Additionally, they may be 90 targeting a business to commit another crime, such as a robbery, and in the process committed the homicide. Interestingly, when examining right-wing only incidents by ideological motivation no significant relationship is found between the destructibility of the target location and the motivation of the attack. However, a significant relationship is found within jihadist incidents between target destructibility and ideological motivation. One-half of ideologically motivated jihadist incidents are attacks against locations that can be easily destroyed, whereas under 20% of non-ideologically motivated attacks occur at similar locations. The proportional distributions for the destructibility of locations of right-wing attacks are very similar across whether or not the incident is ideologically motivated. This indicates that right-wing extremists may be less likely to consider the destructibility of a target when committing an attack, or rather they may not find it to be as important of a factor as other vulnerability factors. For all incidents, those that are not ideologically motivated had no one other than the victim at the location during the time of the incident 63.3% of the time and 25.1% of nonideologically motivated incidents occur at a location with one to five others around. Only 7.5% of incidents that are not ideologically motivated occur at a location with six or more other individuals around. These findings make intuitive sense in that if a suspect is committing a nonideologically motivated crime, they may be more concerned with being caught and therefore want fewer witnesses around. In examining far right extremists, Caspi, Freilich and Chermak (2012) found they are more likely to kill potential witnesses at the scene of an attack if they are a lone wolf suspect, working with no accomplices. A significant chi square statistic is found between the occupied variable and whether or not the incident is ideologically motivated. In examining cases that are ideologically motivated, the majority of attacks occur at a location with 91 no others around (47.2%) but there is more variation among these incidents than there is between non-ideologically motivated incidents. Within ideological incidents, victims are at a location with one to five others around 21.3% of the time, six to twenty-five others around 14.5% of the time, 26-100 others around 8.1% of the time and 8.9% of the time the victim is at a location with 101 or more others around. These results show that there are clear differences in ideologically motivated versus non-ideologically motivated incidents, with ideologically motivated incidents scoring higher on this EVIL DONE factor and occurring in locations with more individuals around as compared to non-ideologically motivated attacks. A significant relationship is found when examining right-wing only incidents between the ideological motivation for the incident and the occupied variable. Nearly one-half of all ideologically motivated incidents by right-wing extremists occur at a location with no others around the victim(s) at the time of the attack as compared to 63.4% of non-ideologically motivated right-wing attacks. Nearly 13% of ideologically motivated right-wing extremist attacks occur at locations with at least 26 other individuals nearby whereas only 3.8% of nonideologically motivated right-wing extremist attacks occur at similar locations. It appears that proportionately, right-wing extremists are more likely to attack victims who are at locations that have a greater number of individuals around when they are attacking for ideological purposes. There is no significant relationship found when examining jihadist only incidents between the occupied variable and the ideological motivation of the incident. The chi square statistic for the near variable is statistically significant with whether the incident was ideologically motivated or not. This variable examines how close the suspect lives to the target. In cases that are not ideologically motivated, 55.8% of the time suspects live in the same city as where the incident occurs and in 37.7% of the cases they live within 100 miles of 92 the incident location. Similar findings are found in ideologically motivated cases where 45.1% of the time suspects live in the same city as the location of the incident and 38.3% of the time they live within 100 miles of the attack site. In 16.6% of ideologically motivated incidents the suspect travels more than 100 miles to the incident location, whereas in non-ideological incidents the suspect travels this far only 6.5% of the time. For both ideologically motivated and nonideologically motivated incidents these findings support the notion that suspects will be more likely to commit their attack near where they live. However, it appears suspects may be more likely to travel when they are committing an ideologically motivated attack as compared to a non-ideologically motivated attack. This supports prior research which has found that terrorists are likely to live and engage in preplanning activities within 30 miles of the target (Smith, Cothren, Roberts & Damphousse, 2008). The near variable is the only one of the eight EVIL DONE variables that is not significant when examining ideological motivation between far-right attacks and between jihadist attacks separately. This is interesting since significance is found when examining all cases together but disappears when looking at the two types of attackers separately. The proportional distributions between far-right attackers and where they lived at the time of the attack are very similar between ideologically motivated incidents and non-ideologically motivated incidents with nearly 50% living in the same city as the target and nearly 40% living within 100 miles of the target location. The proportional distributions for jihadist attacks are also very similar across ideological motivation, however, jihadists appear to be much more likely to travel to commit an attack. Nearly 50% of ideologically motivated attacks by jihadists are against a target that was 100 or more miles away from the suspect’s residence and 25% of non-ideologically motivated attacks are a similar distance. 93 The chi square statistic is significant for the easy variable and whether or not the incident is ideologically motivated for all incidents, for right-wing only incidents and for jihadist only incidents. This indicates that there is a relationship between ideological motivation and the level of security at the homicide site. In incidents that are not ideologically motivated, 87% of the time the incident occurs at a location with no security, 12% occur in locations with some security and only 1% occur in locations with high security. Similar results are found in incidents that are ideologically motivated with 66% of them occurring at a location with no security, 29% at a location with some security and 5% at a location with high security. Gruenewald and colleagues (2015) had similar findings in that eco-terrorists chose targets that are unprotected by security. These findings for both ideologically and non-ideologically motivated incidents support the expectation that suspects are more likely to commit an act of extremist violence at a location with no or limited security. It is interesting to see that suspects appear to be more likely to commit an attack at a location with high security when the motivation is ideological as compared to when they are committing the attack for other purposes. EVIL DONE and Ideological Affiliation While ideological motivation plays a significant role in target selection, different aspects of targets will be more attractive depending on the ideological motivation of the suspect. This dissertation expands on Gruenewald and colleagues (2015) examination of EVIL DONE by applying these vulnerability factors to jihadist and right-wing extremist incidents. This section discusses EVIL DONE factors and the suspect’s ideological affiliation, highlighting significant results. 94 Significant relationships are found between all of the EVIL DONE variables and the suspect’s ideological affiliation (right-wing v. jihadist). Table 4.4 presents these results. Significant relationships are also found between right-wing and jihadist incidents for ideologically motivated attacks only for all eight EVIL DONE variables, but only one significant result is found between right-wing attacks and jihadist attacks for non-ideologically motivated incidents and it was the near variable. Similar proportional distributions are found between all incidents and then between ideologically motivated incidents for right-wing attackers and for jihadist attackers, therefore this section will discuss the distributions for all of the incidents combined. These findings show clear differences between what vulnerability features existed in targets that jihadists choose versus those that right-wing extremists choose. Jihadists are more likely than right-wing extremists to attack victims in locations that rank high on the vital, iconic and occupied measures. Right-wing extremists, on the other hand, are more likely to choose victims who are located in highly exposed areas near where they live and areas that have no security. 95 Table 4.4 EVIL DONE and Suspect Ideology (n=434) Ideological Affiliation Right Wing (n= 374) Jihadist (n= 60) Ideological Only Right Wing (n=191) Non-Ideological Only Jihadist (n=44) Right Wing (n=183) Jihadist (n=16) N % N % N % N % N % N % 147* 39.3% 27 45.0% 83* 43.5% 24 54.5% 64 35.0% 3 18.8% 130 34.8% 17 28.3% 52 27.2% 5 11.4% 78 42.6% 12 75.0% 50 13.4% 2 3.3% 26 13.6% 2 4.5% 24 13.1% 0 0.0% 47 12.6% 14 23.3% 30 15.7% 13 29.5% 17 9.3% 1 6.3% 262*** 70.1% 26 43.3% 117*** 61.3% 14 31.8% 145 79.2% 12 75.0% 112 29.9% 34 56.7% 74 38.7% 30 68.2% 38 20.8% 4 25.0% 341*** 91.2% 43 71.7% 165*** 86.4% 28 63.6% 176 96.2% 15 93.8% 33 8.8% 17 28.3% 26 13.6% 16 36.4% 7 3.8% 1 6.3% Exposed Victim(s) in a location that was accessible and routinely frequented by public during both day and night Victim(s) in a location that was inaccessible to public without permission both during the day and night Victim(s) in a location that was accessible but rarely frequented by the public either during the day or night Victim(s) in a location that was accessible and routinely frequented by public during the day or night Vital Victim(s) in a location that if eliminated would have no impact on day to day functioning of the community Victim(s) in a location that if eliminated would have a great impact on the day to day functioning of the community Iconic Victim(s) in a location that is not iconic Victim(s) in a location that is iconic ***p<.001, **p<.01,*p<.05 96 Table 4.4 (cont’d) Ideological Affiliation Right Wing (n= 374) Jihadist (n= 60) Ideological Only Right Wing (n=191) Non-Ideological Only Jihadist (n=44) Right Wing (n=183) Jihadist (n=16) N % N % N % N % N % N % 302*** 30.4% 34 11.3% 143*** 26.7% 20 9.1% 159 19.6% 14 17.5% 65 6.5% 18 6.0% 42 7.8% 17 7.7% 23 2.8% 1 1.3% 7 0.7% 8 2.7% 6 1.1% 7 3.2% 1 0.1% 1 1.3% Victim(s) in a location that is easily destructible 170*** 45.5% 25 41.7% 93** 48.7% 22 50.0% 77 42.1% 3 18.8% Victim(s) in a location that is difficult to destroy 15 4.0% 10 16.7% 11 5.8% 9 20.5% 4 2.2% 1 6.3% Victim(s) in a location that is moderately destructible 189 50.5% 25 41.7% 87 45.5% 13 29.5% 102 55.7% 12 75.0% 213*** 57.0% 24 40.0% 97*** 50.8% 14 31.8% 116 63.4% 10 62.5% 84 22.5% 16 26.7% 38 19.9% 12 27.3% 46 25.1% 4 25.0% 46 12.3% 3 5.0% 32 16.8% 2 4.5% 14 7.7% 1 6.3% 17 4.5% 6 10.0% 13 6.8% 6 13.6% 4 2.2% 0 0.0% 14 3.7% 11 18.3% 11 5.8% 10 22.7% 3 1.6% 1 6.3% Legitimate Victim(s) in a location that houses only general citizenry Victim(s) in a location that houses general citizenry and those working for target organization Victim(s) in a location that houses only those working for target organization Destructible Occupied There are no other individuals around at time of incident Victim(s) in a location with 1-5 other people near them Victim(s) in a location with 6-25 other people near them Victim(s) in a location with 26-100 people near them Victim(s) in a location with 101 or more people near them ***p<.001, **p<.01,*p<.05 97 Table 4.4 (cont’d) Ideological Affiliation Right Wing (n= 374) Jihadist (n= 60) Ideological Only Right Wing (n=191) Non-Ideological Only Jihadist (n=44) Right Wing (n=183) Jihadist (n=16) N % N % N % N % N % N % 196*** 52.4% 21 35.0% 92*** 48.2% 14 31.8% 104** 56.8% 7 43.8% Offender(s) lived 101 or more miles from target 27 7.2% 25 41.7% 18 9.4% 21 47.7% 9 4.9% 4 25.0% Offender(s) lived outside city within 100 miles 151 40.4% 14 23.3% 81 42.4% 9 20.5% 70 38.3% 5 31.3% 296*** 79.1% 32 53.3% 137*** 71.7% 18 40.9% 159 86.9% 14 87.5% Victim(s) in a location with high security 8 2.14% 6 10.00% 7 3.7% 5 11.4% 1 0.5% 1 6.3% Victim(s) in a location with some security 70 18.7% 22 36.7% 47 24.6% 21 47.7% 23 12.6% 1 6.3% Near Offender(s) lived in same city as target Easy Victim(s) in a location with no security ***p<.001, **p<.01,*p<.05 98 Attacks that are committed by right-wing extremists occur at a location that are inaccessible to the public without permission day or night 34.8% of the time and 28.3% of the time jihadists committed their attacks at inaccessible locations. Right-wing extremists appear to be more willing to attack a victim who is located in an area that is not easily accessible as compared to jihadists. The near variable represents how close the suspect lives to the location of the incident. For right-wing extremists, 52.4% of the time the suspect lives in the same city the incident occurs in and 35% of the time jihadist attackers live in the same city as the location of the incident. This is the only variable that is significant for non-ideologically motivated attacks. Jihadists appear to be proportionally more likely to travel to commit a non-ideologically motivated attack compared to right-wing extremists. The easy variable examines the security at the location the victim was in at the time of the attack. Right-wing extremists commit attacks at a location with no security 79% of the time whereas jihadists commit attacks at locations with no security only 53.3% of the time. Right-wing extremists appear to be more likely to attack victims who are in highly exposed, non-secure locations and who are in areas near where they live. Jihadists appear to plan more and consider other vulnerability factors more attractive when choosing victims so may be more willing to attack a victim who is located at an area that is not highly exposed or at a location that is far from where they reside. For all incidents, right-wing extremists commit attacks against a victim who was at a location considered to be vital to the community 29.9% of the time and jihadists commit attacks against victims at vital locations 56.7% of the time. These general findings are interesting in showing that jihadists may view the vitality of a target as being more important than right-wing extremists view it. In 8.8% of attacks right-wing extremists commit the victim is located at a place that is considered to be iconic and in 28.3% of the incidents jihadists commit the victim is 99 at an iconic location. This again, shows that for many of these variables jihadists are choosing victims and locations that are considered more vulnerable or attractive than right-wing extremists are. Only 4% of attacks right-wing extremists commit occur at a location that is difficult to destroy compared to 16.6% of jihadist attacks occurring in these types of areas. Significance is found with this variable and while most attacks right-wing extremists and jihadists commit are at locations that are at least somewhat destructible (moderately destructible), jihadists appear to be more likely to attack victims at locations that are difficult to destroy. The variable occupied measures how many others are located near the victim at the time of the attack. For right-wing extremists, 57% of all attacks occur at a location where no others are present at the time of the incident and 40% of jihadist incidents occur at a setting with no others present besides the perpetrator(s) and victim(s). Jihadists appear to be more likely to attack victims who are vulnerable based on the EVIL DONE factors vital, iconic, destructible and occupied as compared to right-wing extremists. These findings show that jihadists may value different vulnerability factors more than right-wing extremists do. EVIL DONE and Suspect Characteristics This section examines the relationship between EVIL DONE vulnerability characteristics and suspect variables of gender9, age, race, occupation10, number of suspects, whether or not the suspect is a lone wolf and the weapon used in the attack. Results can be located in Table 4.5, Table 4.6, and Table 4.7. 9 No significant relationship is found between suspect gender and EVIL DONE characteristics which is not surprising given only four incidents had female suspects. This variable will be excluded from multivariate analysis. 10 No significant relationships are found between suspect occupation and EVIL DONE characteristics. A large portion of cases are missing occupation data (nearly 50%) therefore this variable is excluded from multivariate analysis. 100 Significant relationships are found between all of the EVIL DONE vulnerability characteristics and the race of the suspect. Whites are proportionately more likely to attack victims at locations that are easily destructible, near where the suspect lives and at locations that have no security. Whereas non-whites rank higher on the other five EVIL DONE characteristics. Non-whites are more likely to attack victims at locations that are more exposed, vital, iconic, legitimate and occupied. Non-whites are proportionately more likely to attack victims who are located in areas that are accessible and routinely frequented by the public day or night as compared to whites. One-fourth of locations non-whites attacked are considered to be vital and only 8.8% of locations attacked by whites are considered to be vital. Whites appear to be more concerned with attacking locations that can be easily destroyed, lack security and are nearby where they live. 101 Table 4.5 EVIL DONE and Suspect Characteristics (n=434) Gender Male (n=430) Race Female (n=4) White (n=362) Weapon Non-White (n=72) Firearm (n=280) Other (n=154) N % N % N % N % N % N % 171 39.8% 3 75.0% 142** 39.2% 32 44.4% 118* 42.1% 56 36.4% 146 34.0% 1 25.0% 124 34.3% 23 31.9% 89 31.8% 58 37.7% 52 12.1% 0 0.0% 51 14.1% 1 1.4% 27 9.6% 25 16.2% 61 14.2% 0 0.0% 45 12.4% 16 22.2% 46 16.4% 15 9.7% 287 66.7% 1 25.0% 252** 69.6% 36 50.0% 175* 62.5% 113 73.4% 143 33.3% 3 75.0% 110 30.4% 36 50.0% 105 37.5% 41 26.6% Victim(s) in a location that is not iconic 380 88.4% 4 100.0% 330*** 91.2% 54 75.0% 244 87.1% 36 72.0% Victim(s) in a location that is iconic 50 11.6% 0 0.0% 32 8.8% 18 25.0% 36 12.9% 14 28.0% Exposed Victim(s) in a location that was accessible and routinely frequented by public during both day and night Victim(s) in a location that was inaccessible to public without permission both during the day and night Victim(s) in a location that was accessible but rarely frequented by the public either during the day or night Victim(s) in a location that was accessible and routinely frequented by public during the day or night Vital Victim(s) in a location that if eliminated would have no impact on day to day functioning of the community Victim(s) in a location that if eliminated would have a great impact on the day to day functioning of the community Iconic ***p<.001, **p<.01,*p<.05 102 Table 4.5 (cont’d) Gender Male (n=430) Race Female (n=4) White (n=362) Weapon Non-White (n=72) Firearm (n=280) Other (n=154) N % N % N % N % N % N % 332 15.4% 4 20.0% 292*** 30.3% 44 12.2% 203** 19.0% 133 17.3% 83 3.9% 0 0.0% 62 6.4% 21 5.8% 67 6.3% 16 2.1% 15 0.7% 0 0.0% 8 0.8% 7 1.9% 10 0.9% 5 0.6% Victim(s) in a location that is easily destructible 213 49.5% 1 25.0% 167*** 46.1% 28 38.9% 118 42.1% 77 50.0% Victim(s) in a location that is difficult to destroy 25 5.8% 0 0.0% 12 3.3% 13 18.1% 18 6.4% 7 4.5% Victim(s) in a location that is moderately destructible 192 44.7% 3 75.0% 183 50.6% 31 43.1% 144 51.4% 70 45.5% 236 54.9% 1 25.0% 211*** 58.3% 26 36.1% 134** 47.9% 103 66.9% 98 22.8% 2 50.0% 79 21.8% 21 29.2% 72 25.7% 28 18.2% 48 11.2% 1 25.0% 42 11.6% 7 9.7% 40 14.3% 9 5.8% 23 5.3% 0 0.0% 15 4.1% 8 11.1% 18 6.4% 5 3.2% 25 5.8% 0 0.0% 15 4.1% 10 13.9% 16 5.7% 9 5.8% Legitimate Victim(s) in a location that houses only general citizenry Victim(s) in a location that houses general citizenry and those working for target organization Victim(s) in a location that houses only those working for target organization Destructible Occupied There are no other individuals around at time of incident Victim(s) in a location with 1-5 other people near them Victim(s) in a location with 6-25 other people near them Victim(s) in a location with 26-100 people near them Victim(s) in a location with 101 or more people near them ***p<.001, **p<.01,*p<.05 103 Table 4.5 (cont’d) Gender Male (n=430) Race Female (n=4) White (n=362) Weapon Non-White (n=72) Firearm (n=280) Other (n=154) N % N % N % N % N % N % Offender(s) lived in same city as target 216 50.2% 1 25.0% 184*** 50.8% 33 45.8% 139 49.6% 78 50.6% Offender(s) lived 101 or more miles from target 52 12.1% 0 0.0% 28 7.7% 24 33.3% 37 13.2% 15 9.7% Offender(s) lived outside city within 100 miles 162 37.7% 3 75.0% 150 41.4% 15 20.8% 104 37.1% 61 39.6% Victim(s) in a location with no security 324 75.3% 4 100.0% 284*** 78.5% 44 61.1% 198** 70.7% 130 84.4% Victim(s) in a location with high security 14 3.3% 0 0.0% 7 1.9% 7 9.7% 10 3.6% 4 2.6% Victim(s) in a location with some security 92 21.4% 0 0.0% 71 19.6% 21 29.2% 72 25.7% 20 13.0% Near Easy ***p<.001, **p<.01,*p<.05 104 Table 4.6 EVIL DONE and Suspect Characteristics 2 (n=434) Occupation Unemployed (n=98) Blue collar (n=64) Number of Suspects White collar (n=30) One suspect (n=234) Two suspects (n=101) Three + suspects (n=99) N % N % N % N % N % N % 35 35.7% 29 45.3% 6 20.0% 90* 38.5% 48 47.5% 36 36.4% 35 35.7% 22 34.4% 14 46.7% 86 36.8% 28 27.7% 33 33.3% 9 9.2% 5 7.8% 4 13.3% 20 8.5% 12 11.9% 20 20.2% 19 19.4% 8 12.5% 6 20.0% 38 16.2% 13 12.9% 10 10.1% 62 63.3% 40 62.5% 20 66.7% 153 65.4% 67 66.3% 68 68.7% 36 36.7% 24 37.5% 10 33.3% 81 34.6% 34 33.7% 31 31.3% Victim(s) in a location that is not iconic 80 81.6% 52 81.3% 24 80.0% 199* 85.0% 95 94.1% 90 90.9% Victim(s) in a location that is iconic 18 18.4% 12 18.8% 6 20.0% 35 15.0% 6 5.9% 9 9.1% Exposed Victim(s) in a location that was accessible and routinely frequented by public during both day and night Victim(s) in a location that was inaccessible to public without permission both during the day and night Victim(s) in a location that was accessible but rarely frequented by the public either during the day or night Victim(s) in a location that was accessible and routinely frequented by public during the day or night Vital Victim(s) in a location that if eliminated would have no impact on day to day functioning of the community Victim(s) in a location that if eliminated would have a great impact on the day to day functioning of the community Iconic ***p<.001, **p<.01,*p<.05 105 Table 4.6 (cont’d) Occupation Unemployed (n=98) Blue collar (n=64) Number of Suspects White collar (n=30) One suspect (n=234) Two suspects (n=101) Three + suspects (n=99) N % N % N % N % N % N % 73 76.0% 44 68.8% 22 73.3% 170* 72.6% 79 78.2% 87 87.9% 21 21.9% 15 23.4% 7 23.3% 53 22.6% 21 20.8% 9 9.1% 2 2.1% 5 7.8% 1 3.3% 11 4.7% 1 1.0% 3 3.0% Victim(s) in a location that is easily destructible 37 37.8% 26 40.6% 7 23.3% 139** 57.2% 46 45.5% 6 6.1% Victim(s) in a location that is difficult to destroy 10 10.2% 7 10.9% 4 13.3% 17 7.0% 2 2.0% 38 38.4% Victim(s) in a location that is moderately destructible 51 52.0% 31 48.4% 19 63.3% 87 35.8% 53 52.5% 55 55.6% 52 53.1% 27 42.2% 18 60.0% 111* 47.4% 61 60.4% 65 65.7% 15 15.3% 12 18.8% 5 16.7% 56 23.9% 25 24.8% 19 19.2% 16 16.3% 10 15.6% 4 13.3% 33 14.1% 10 9.9% 6 6.1% 8 8.2% 6 9.4% 1 3.3% 19 8.1% 1 1.0% 3 3.0% 7 7.1% 9 14.1% 2 6.7% 15 6.4% 4 4.0% 6 6.1% Legitimate Victim(s) in a location that houses only general citizenry Victim(s) in a location that houses general citizenry and those working for target organization Victim(s) in a location that houses only those working for target organization Destructible Occupied There are no other individuals around at time of incident Victim(s) in a location with 1-5 other people near them Victim(s) in a location with 6-25 other people near them Victim(s) in a location with 26-100 people near them Victim(s) in a location with 101 or more people near them ***p<.001, **p<.01,*p<.05 106 Table 4.6 (cont’d) Occupation Unemployed (n=98) Blue collar (n=64) Number of Suspects White collar (n=30) One suspect (n=234) Two suspects (n=101) Three + suspects (n=99) N % N % N % N % N % N % Offender(s) lived in same city as target 50 51.0% 33 51.6% 16 53.3% 121* 51.7% 43 42.6% 53 53.5% Offender(s) lived 101 or more miles from target 10 10.2% 7 10.9% 6 20.0% 21 9.0% 21 20.8% 10 10.1% Offender(s) lived outside city within 100 miles 38 38.8% 24 37.5% 8 26.7% 92 39.3% 37 36.6% 36 36.4% Victim(s) in a location with no security 72 73.5% 42 65.6% 22 73.3% 172 73.5% 73 72.3% 83 83.8% Victim(s) in a location with high security 3 3.1% 4 6.3% 4 13.3% 10 4.3% 2 2.0% 2 2.0% Victim(s) in a location with some security 23 23.5% 18 28.1% 4 13.3% 52 22.2% 26 25.7% 14 14.1% Near Easy ***p<.001, **p<.01,*p<.05 107 Table 4.7 EVIL DONE and Suspect Characteristics 3 (n=434) Lone Wolf Lone wolf (n=109) Age Not a lone wolf (n=325) 14-24 (n=152) 25-49 (n=250) 50+ (n=32) N % N % N % N % N % 45*** 41.3% 129 39.7% 66 43.4% 97 38.8% 11 34.4% 26 23.9% 121 37.2% 41 27.0% 97 38.8% 9 28.1% 8 7.3% 44 13.5% 24 15.8% 24 9.6% 4 12.5% 30 27.5% 31 9.5% 21 13.8% 32 12.8% 8 25.0% 56*** 51.4% 232 71.4% 107 70.4% 162 64.8% 19 59.4% 53 48.6% 93 28.6% 45 29.6% 88 35.2% 13 40.6% 80*** 73.4% 304 93.5% 138** 90.8% 223 89.2% 23 71.9% 29 26.6% 21 6.5% 14 9.2% 27 10.8% 9 28.1% Exposed Victim(s) in a location that was accessible and routinely frequented by public during both day and night Victim(s) in a location that was inaccessible to public without permission both during the day and night Victim(s) in a location that was accessible but rarely frequented by the public either during the day or night Victim(s) in a location that was accessible and routinely frequented by public during the day or night Vital Victim(s) in a location that if eliminated would have no impact on day to day functioning of the community Victim(s) in a location that if eliminated would have a great impact on the day to day functioning of the community Iconic Victim(s) in a location that is not iconic Victim(s) in a location that is iconic ***p<.001, **p<.01,*p<.05 108 Table 4.7 (cont’d) Lone Wolf Lone wolf (n=109) Age Not a lone wolf (n=325) 14-24 (n=152) 25-49 (n=250) 50+ (n=32) N % N % N % N % N % 60*** 55.0% 276 84.9% 120 21.6% 196 15.7% 20 12.5% 40 36.7% 43 13.2% 30 5.4% 43 3.4% 10 6.3% 9 8.3% 6 1.8% 2 0.4% 11 0.9% 2 1.3% Victim(s) in a location that is easily destructible 38*** 34.9% 157 48.3% 80 52.6% 105 42.0% 10 31.3% Victim(s) in a location that is difficult to destroy 14 12.8% 11 3.4% 7 4.6% 14 5.6% 4 12.5% Victim(s) in a location that is moderately destructible 57 52.3% 157 48.3% 65 42.8% 131 52.4% 18 56.3% 46*** 42.2% 191 58.8% 84* 55.3% 142 56.8% 11 34.4% 17 15.6% 83 25.5% 39 25.7% 56 22.4% 5 15.6% 21 19.3% 28 8.6% 15 9.9% 25 10.0% 9 28.1% 13 11.9% 10 3.1% 7 4.6% 12 4.8% 4 12.5% 12 11.0% 13 4.0% 7 4.6% 15 6.0% 3 9.4% Legitimate Victim(s) in a location that houses only general citizenry Victim(s) in a location that houses general citizenry and those working for target organization Victim(s) in a location that houses only those working for target organization Destructible Occupied There are no other individuals around at time of incident Victim(s) in a location with 1-5 other people near them Victim(s) in a location with 6-25 other people near them Victim(s) in a location with 26-100 people near them Victim(s) in a location with 101 or more people near them ***p<.001, **p<.01,*p<.05 109 Table 4.7 (cont’d) Lone Wolf Lone wolf (n=109) Age Not a lone wolf (n=325) 14-24 (n=152) 25-49 (n=250) 50+ (n=32) N % N % N % N % N % Offender(s) lived in same city as target 55 50.5% 162 49.8% 80 52.6% 124 49.6% 13 40.6% Offender(s) lived 101 or more miles from target 14 12.8% 38 11.7% 14 9.2% 33 13.2% 5 15.6% Offender(s) lived outside city within 100 miles 40 36.7% 125 38.5% 58 38.2% 93 37.2% 14 43.8% 58*** 53.2% 270 83.1% 120*** 78.9% 190 76.0% 18 56.3% Victim(s) in a location with high security 10 9.2% 4 1.2% 1 0.7% 8 3.2% 5 15.6% Victim(s) in a location with some security 41 37.6% 51 15.7% 31 20.4% 52 20.8% 9 28.1% Near Easy Victim(s) in a location with no security ***p<.001, **p<.01,*p<.05 110 Three EVIL DONE vulnerability factors are related to the age of the suspect. Older suspects rank higher on the iconic and occupied variables and appear to be more likely to attack crowded locations and iconic locations. Roughly 10% of attacks are committed by suspects between 14-24, 10.8% by suspects 25-49, occur at locations considered to be iconic. Interestingly, nearly 30% of attacks that are committed by suspects over the age of 50 occur at an iconic location. As suspects age, it appears that the iconicity of a target becomes a more important factor. Another interesting finding is the relationship between the suspect’s age and the easy variable. It appears that older suspects are more likely to attack victims who are at locations that have some security or high security. As suspects age and become more aware of various techniques to evade security they may develop more elaborate plans of attack and be more willing to attack more difficult targets than younger suspects. Six EVIL DONE vulnerability characteristics are significantly related to the number of suspects involved in an attack. Attacks that involve only one suspect are proportionately more likely to occur at locations that are considered to be iconic (15%) as compared to attacks committed by two suspects (5.9%) or attacks committed by three or more suspects (9.1%). A significant relationship is also found with the legitimate variable. Attacks that are committed by one suspect are proportionately more likely to occur at locations that house only individuals working for the target organization as compared to attacks with two or more suspects. Attacks involving one suspect also appear to be more likely to be attacks that are against targets that are easily destructible, whereas when there are multiple suspects involved it appears the suspects are more willing to attack a location that is more difficult to destroy. When multiple suspects are involved it may be easier to obtain weaponry and plan larger scale attacks that can destroy larger targets. Interestingly, attacks that are committed by one suspect are proportionately more likely 111 to occur at locations with others around as compared to attacks with more suspects. Nearly one half (47.4%) of attacks committed by one suspect occur at a location with no others around, and 60.4% of attacks with two suspects occur at a similar location. Individuals who commit an attack alone appear to be more willing to target a large group of individuals. All EVIL DONE vulnerability characteristics, except for near, are significantly related to whether or not an attack is committed by a lone wolf. Lone wolves are individuals who commit an attack by themselves, who have no specific group affiliation and commit the attack without help of any other individuals. Lone wolves rank higher on exposed, vital, iconic, legitimate and occupied than non-lone wolf attackers but rank lower on destructible and easy. For example, lone wolves attack vital locations 48.5% of the time compared to 28.5% for other attackers. Lone wolves attack victims at iconic locations in 26.6% of their attacks compared to only 6.5% of attacks committed by non-lone wolf suspects occurring at an iconic location. In 8.3% of attacks committed by lone wolf suspects the attack occurs at a location that houses only those working for the target organization as compared to only 1.8% of attacks committed by non-lone wolf suspects. Interestingly, lone wolf suspects rank lower on destructible and easy. Lone wolf suspects are more likely to attack locations that are difficult to destroy as compared to other types of suspects. Additionally, lone wolf suspects are more likely to attack locations that have high security as compared to non-lone wolf suspects. The majority of attacks committed by nonlone wolves (83.1%) are at locations that have no security whereas 53.2% of attacks committed by lone wolf suspects are at locations with no security. Lone wolf suspects appear to be more willing to take the extra effort to attack a location that is difficult to destroy and has security. Finally, five of the EVIL DONE vulnerability characteristics are significantly related to weapon choice. Proportionately speaking, attacks involving firearms appear to be in more open 112 areas as compared to attacks using other weapons. Attacks involving firearms are also more likely to be against targets that are considered iconic (37.5%) as compared to attacks using other weapons (26.6%). Attacks targeting government, military and police are more likely to involve the use of a firearm as compared to a different type of weapon. This makes sense in that these types of attacks may be at highly secure locations or against individuals who are also protecting themselves with firearms so may require a suspect to have a firearm. The occupied variable is also significantly related to weapon choice. When there is a firearm used in an attack then it is more likely there are other individuals near the suspect. Finally, in nearly 30% of attacks involving firearms the suspect did not live in the same city as their target, compared to 15% of attacks that didn’t involve firearms. This may be because firearms are easier to transport compared to larger weapons (such as bombs). EVIL DONE and Victim Characteristics EVIL DONE vulnerability factors and victim characteristics are examined in this section. The gender, age, race and occupation of the victim is examined as well as the relationship between the victim and the suspect. These variables are compared to EVIL DONE vulnerability characteristics. Significant relationships that are found between factors are discussed. Results are located in Table 4.8, Table 4.9, and Table 4.10. Only two EVIL DONE vulnerability characteristics are significantly related to the gender of the victim. These two variables are exposed and destructible. Men are proportionately more likely to be attacked in locations that are frequented day and night (44.3% of the time) as compared to women who are attacked in similar locations only 27.1% of the time. Men appear to be more likely to be attacked in public settings and women more likely to be attacked in a private 113 setting. Women are often the primary victims of domestic violence and often this type of violence occurs behind closed doors. Two EVIL DONE variables are significantly related to the race of the victim and those include exposed and vital. Non-whites are proportionately more likely to be in locations that are frequented day and night (52.3% of the time) than whites who are in these locations only 33.6% of the time. Similar to the findings for gender, this is interesting that there are differences in the exposure of a victim. One reason for this finding may be that attacks committed by white supremacists often target minorities and often occur in public settings whereas attacks against whites by white supremacists may not be ideologically motivated and may be more likely to be in a private setting. Interestingly, non-whites are more likely to be killed in vital locations (41.7% of the time) compared to whites who are killed in vital locations 29.3% of the time. This finding may be related to the above finding in that nonwhites may be more likely to be attacked for ideologically motivated reasons and may be more likely to be in a public area. Because they are attacked for ideologically motivated reasons they may also be in areas that rank high on other EVIL DONE factors, such as vitality. Additionally, two EVIL DONE variables are significantly related to the age of the victim. Younger victims (those who are 17 and under) are more likely to live in the same city as the suspect. Older victims are proportionately more likely to be at locations with some security or high security as compared to younger victims. This makes intuitive sense in that older victims may be more likely to be at vital or iconic locations because of their status or occupation and many of these locations have some security or high security. 114 Table 4.8 EVIL DONE and Victim Characteristics (n=434) Gender Male (n=327) Race Female (n=107) White (n=283) Non-White (n=151) N % N % N % N % 145*** 44.3% 29 27.1% 95*** 33.6% 79 52.3% 94 28.7% 53 49.5% 107 37.8% 40 26.5% 42 12.8% 10 9.3% 40 14.1% 12 7.9% 46 14.1% 15 14.0% 41 14.5% 20 13.2% 209 63.9% 79 73.8% 200* 70.7% 88 58.3% 118 36.1% 28 26.2% 83 29.3% 63 41.7% Victim(s) in a location that is not iconic 288 88.1% 96 89.7% 249 88.0% 135 89.4% Victim(s) in a location that is iconic 39 11.9% 11 10.3% 34 12.0% 16 10.6% Exposed Victim(s) in a location that was accessible and routinely frequented by public during both day and night Victim(s) in a location that was inaccessible to public without permission both during the day and night Victim(s) in a location that was accessible but rarely frequented by the public either during the day or night Victim(s) in a location that was accessible and routinely frequented by public during the day or night Vital Victim(s) in a location that if eliminated would have no impact on day to day functioning of the community Victim(s) in a location that if eliminated would have a great impact on the day to day functioning of the community Iconic ***p<.001, **p<.01,*p<.05 115 Table 4.8 (cont’d) Gender Male (n=327) Race Female (n=107) White (n=283) Non-White (n=151) N % N % N % N % 248 75.8% 88 82.2% 224 79.2% 112 74.2% 67 20.5% 16 15.0% 49 17.3% 34 22.5% 12 3.7% 3 2.8% 10 3.5% 5 3.3% Victim(s) in a location that is easily destructible 159* 48.6% 36 33.6% 123 43.5% 72 47.7% Victim(s) in a location that is difficult to destroy 20 6.1% 5 4.7% 14 4.9% 11 7.3% Victim(s) in a location that is moderately destructible 148 45.3% 66 61.7% 146 51.6% 68 45.0% 173 52.9% 64 59.8% 160 56.5% 77 51.0% 77 23.5% 23 21.5% 60 21.2% 40 26.5% 38 11.6% 11 10.3% 33 11.7% 16 10.6% 20 6.1% 3 2.8% 14 4.9% 9 6.0% 19 5.8% 6 5.6% 16 5.7% 9 6.0% Legitimate Victim(s) in a location that houses only general citizenry Victim(s) in a location that houses general citizenry and those working for target organization Victim(s) in a location that houses only those working for target organization Destructible Occupied There are no other individuals around at time of incident Victim(s) in a location with 1-5 other people near them Victim(s) in a location with 6-25 other people near them Victim(s) in a location with 26-100 people near them Victim(s) in a location with 101 or more people near them ***p<.001, **p<.01,*p<.05 116 Table 4.8 (cont’d) Gender Male (n=327) Race Female (n=107) White (n=283) Non-White (n=151) N % N % N % N % Offender(s) lived in same city as target 160 48.9% 57 53.3% 140 49.5% 77 51.0% Offender(s) lived 101 or more miles from target 41 12.5% 11 10.3% 32 11.3% 20 13.2% Offender(s) lived outside city within 100 miles 126 38.5% 39 36.4% 111 39.2% 54 35.8% Victim(s) in a location with no security 243 74.3% 85 79.4% 215 76.0% 113 74.8% Victim(s) in a location with high security 13 4.0% 1 0.9% 8 2.8% 6 4.0% Victim(s) in a location with some security 71 21.7% 21 19.6% 60 21.2% 32 21.2% Near Easy ***p<.001, **p<.01,*p<.05 117 Table 4.9 EVIL DONE and Victim Characteristics 2 (n=434) Age 17 and under (n=27) N % 18-24 (n=81) N % 25-49 (n=241) N % 50+ (n=85) N % Exposed Victim(s) in a location that was accessible and routinely frequented by public during both day and night Victim(s) in a location that was inaccessible to public without permission both during the day and night Victim(s) in a location that was accessible but rarely frequented by the public either during the day or night Victim(s) in a location that was accessible and routinely frequented by public during the day or night 9 33.3% 24 29.6% 110 45.6% 31 36.5% 14 51.9% 32 39.5% 70 29.0% 31 36.5% 2 7.4% 12 14.8% 29 12.0% 9 10.6% 2 7.4% 13 16.0% 32 13.3% 14 16.5% Vital Victim(s) in a location that if eliminated would have no impact on day to day functioning of the community Victim(s) in a location that if eliminated would have a great impact on the day to day functioning of the community 17 63.0% 62 76.5% 157 65.1% 52 61.2% 10 37.0% 19 23.5% 84 34.9% 33 38.8% 24 3 88.9% 11.1% 75 6 92.6% 7.4% 211 30 87.6% 12.4% 74 11 87.1% 12.9% Iconic Victim(s) in a location that is not iconic Victim(s) in a location that is iconic ***p<.001, **p<.01,*p<.05 118 Table 4.9 (cont’d) Age 17 and under (n=27) N % Legitimate Victim(s) in a location that houses only general citizenry Victim(s) in a location that houses general citizenry and those working for target organization Victim(s) in a location that houses only those working for target organization Destructible Victim(s) in a location that is easily destructible Victim(s) in a location that is difficult to destroy Victim(s) in a location that is moderately destructible Occupied There are no other individuals around at time of incident Victim(s) in a location with 1-5 other people near them Victim(s) in a location with 6-25 other people near them Victim(s) in a location with 26-100 people near them Victim(s) in a location with 101 or more people near them 18-24 (n=81) N % 25-49 (n=241) N % 50+ (n=85) N % 25 92.6% 68 84.0% 184 76.3% 59 69.4% 2 7.4% 12 14.8% 48 19.9% 21 24.7% 0 0.0% 1 1.2% 9 3.7% 5 5.9% 13 1 48.1% 3.7% 37 3 45.7% 3.7% 113 17 46.9% 7.1% 32 4 37.6% 4.7% 13 48.1% 41 50.6% 111 46.1% 49 57.6% 14 51.9% 41 50.6% 130 53.9% 52 61.2% 7 25.9% 20 24.7% 57 23.7% 16 18.8% 2 7.4% 13 16.0% 29 12.0% 5 5.9% 2 7.4% 3 3.7% 11 4.6% 7 8.2% 2 7.4% 4 4.9% 14 5.8% 5 5.9% ***p<.001, **p<.01,*p<.05 119 Table 4.9 (cont’d) Age 17 and under (n=27) N % 18-24 (n=81) N % Near Offender(s) lived in same city as target Offender(s) lived 101 or more miles from target Offender(s) lived outside city within 100 miles 21* 1 5 77.8% 3.7% 18.5% 45 6 30 55.6% 7.4% 37.0% 109 33 99 45.2% 13.7% 41.1% 42 12 31 49.4% 14.1% 36.5% Easy Victim(s) in a location with no security Victim(s) in a location with high security Victim(s) in a location with some security 22* 0 5 81.5% 0.0% 18.5% 72 0 9 88.9% 0.0% 11.1% 175 11 55 72.6% 4.6% 22.8% 59 3 23 69.4% 3.5% 27.1% ***p<.001, **p<.01,*p<.05 120 25-49 (n=241) N % 50+ (n=85) N % Table 4.10 EVIL DONE and Victim Characteristics 3 (n=434) Victim and Suspect Relationship Occupation Unemployed (n=69) N % Exposed Victim(s) in a location that was accessible and routinely frequented by public during both day and night Victim(s) in a location that was inaccessible to public without permission both during the day and night Victim(s) in a location that was accessible but rarely frequented by the public either during the day or night Victim(s) in a location that was accessible and routinely frequented by public during the day or night Vital Victim(s) in a location that if eliminated would have no impact on day to day functioning of the community Victim(s) in a location that if eliminated would have a great impact on the day to day functioning of the community Iconic Victim(s) in a location that is not iconic Victim(s) in a location that is iconic Blue collar (n=48) N % White collar (n=57) N % Police/ government (n=60) N % Victim knew suspect (n=189) N % Strangers (n=245) N % 29 42.0% 26 54.2% 20 35.1% 28 46.7% 43*** 22.8% 131 53.5% 16 23.2% 13 27.1% 12 21.1% 20 33.3% 101 53.4% 46 18.8% 15 21.7% 2 4.2% 5 8.8% 6 10.0% 27 14.3% 25 10.2% 9 13.0% 7 14.6% 20 35.1% 6 10.0% 18 9.5% 43 17.6% 41 59.4% 36 75.0% 25 43.9% 34 56.7% 151*** 79.9% 137 55.9% 28 40.6% 12 25.0% 32 56.1% 26 43.3% 38 20.1% 108 44.1% 59* 10 85.5% 14.5% 44 4 91.7% 8.3% 40 17 70.2% 29.8% 47 13 78.3% 21.7% 175* 14 92.6% 7.4% 209 36 85.3% 14.7% ***p<.001, **p<.01,*p<.05 121 Table 4.10 (cont’d) Victim and Suspect Relationship Occupation Unemployed (n=69) N % Legitimate Victim(s) in a location that houses only general citizenry Victim(s) in a location that houses general citizenry and those working for target organization Victim(s) in a location that houses only those working for target organization Destructible Victim(s) in a location that is easily destructible Victim(s) in a location that is difficult to destroy Victim(s) in a location that is moderately destructible Occupied There are no other individuals around at time of incident Victim(s) in a location with 1-5 other people near them Victim(s) in a location with 6-25 other people near them Victim(s) in a location with 26-100 people near them Victim(s) in a location with 101 or more people near them Blue collar (n=48) N % White collar (n=57) N % Police/ government (n=60) N % Victim knew suspect (n=189) N % Strangers (n=245) N % 57*** 82.6% 33 68.8% 25 43.9% 45 75.0% 166*** 87.8% 170 69.4% 12 17.4% 13 27.1% 26 45.6% 10 16.7% 17 9.0% 66 26.9% 0 0.0% 2 4.2% 6 10.5% 5 8.3% 6 3.2% 9 3.7% 42* 5 60.9% 7.2% 24 4 50.0% 8.3% 13 10 22.8% 17.5% 28 5 46.7% 8.3% 65*** 10 34.4% 5.3% 130 15 53.1% 6.1% 22 31.9% 20 41.7% 34 59.6% 27 45.0% 114 60.3% 100 40.8% 39* 56.5% 28 58.3% 25 43.9% 19 31.7% 131*** 69.3% 106 43.3% 8 11.6% 10 20.8% 14 24.6% 15 25.0% 33 17.5% 67 27.3% 9 13.0% 2 4.2% 7 12.3% 17 28.3% 9 4.8% 40 16.3% 5 7.2% 5 10.4% 3 5.3% 5 8.3% 7 3.7% 16 6.5% 8 11.6% 3 6.3% 8 14.0% 4 6.7% 9 4.8% 16 6.5% ***p<.001, **p<.01,*p<.05 122 Table 4.10 (cont’d) Victim and Suspect Relationship Occupation Unemployed (n=69) N % Near Offender(s) lived in same city as target Offender(s) lived 101 or more miles from target Offender(s) lived outside city within 100 miles Easy Victim(s) in a location with no security Victim(s) in a location with high security Victim(s) in a location with some security Blue collar (n=48) N % White collar (n=57) N % Police/ government (n=60) N % Victim knew suspect (n=189) N % Strangers (n=245) N % 37 3 29 53.6% 4.3% 42.0% 24 7 17 50.0% 14.6% 35.4% 17 12 28 29.8% 21.1% 49.1% 33 11 16 55.0% 18.3% 26.7% 111*** 12 66 58.7% 6.3% 34.9% 106 40 99 43.3% 16.3% 40.4% 58*** 0 11 84.1% 0.0% 15.9% 32 2 14 66.7% 4.2% 29.2% 25 4 28 43.9% 7.0% 49.1% 37 7 16 61.7% 11.7% 26.7% 167*** 2 20 88.4% 1.1% 10.6% 161 12 72 65.7% 4.9% 29.4% ***p<.001, **p<.01,*p<.05 123 Five of the EVIL DONE vulnerability characteristics are significantly related to the occupation of the victim. White collar employees and police/government employees are proportionately more likely to be in iconic, legitimate and occupied locations compared to victims who are blue collar employees or victims who are unemployed. However, victims with white collar employment and government workers are proportionately more likely to be in locations that are difficult to destroy and locations that have high security. Many of the locations that white collar employees work in are large office buildings that are difficult to destroy, have some security/high security and often house a large number of individuals. The relationship between the victim and the suspect is significant for all EVIL DONE variables. Attacks that are against strangers rank high on several EVIL DONE variables. Strangers are more likely to be at an easily accessible location (exposed), a vital location, an iconic location, a legitimate location, an easily destructible location, and an occupied location. If a suspect is attacking a victim for ideological purposes the victim is likely a stranger and the suspect may be looking for victims in locations that rank high on several EVIL DONE factors. Two EVIL DONE variables have higher vulnerability ratings for victims who know the suspect and those are the near and easy variables. Suspects who know their victim live in the same city as the location of the attack 59.7%, as compared to 43.3% of the time for suspects who did not know the victim. This makes sense in that if a suspect is attacking someone they know they likely interact with them somewhat regularly and therefore live in the same general vicinity. Finally, suspects who know their victim are more likely to target the victim at a location with no security (88.4% of the time) as compared to when they attacked strangers it is at a location with no security 65.7% of the time. If a suspect knows the victim, they may be able to more easily 124 gain access to the victim’s home or to a setting where the victim is that lacks security as opposed to trying to attack a stranger. This chapter presented a clear picture of vulnerability in lethal domestic attacks in the United States. The next chapter seeks to answer the second research question and examines which of these factors are the greatest predictors of lethality in an incident. 125 Chapter 5: EVIL DONE Vulnerability and Lethality: Multivariate Analysis This chapter is presented in three sections. The first section examines the relationship between each EVIL DONE variable and lethality and discusses significant relationships. This first section also examines the relationship between victim, suspect, and ideological variables and lethality at a bivariate level. The second section presents multivariate results for each set of variables. Specifically, all EVIL DONE variables are examined in a binary logistic regression model to see which variables are significant when controlling for the others. A model is also conducted for all victim variables, a model for all suspect variables, and a model for the ideological variables. Finally, a model is presented that includes all significant variables from the previous models. The third section of this chapter examines the relationship between EVIL DONE variables and lethality based on ideological factors. Four models are presented which compare across ideological motivation and ideological affiliation. Specifically, incidents that are not ideologically motivated are examined in one model and in a separate model, incidents that are ideologically motivated are examined. Jihadist incidents are examined in their own model and then right-wing extremist attacks in a separate model. These results show differences in lethality based on ideology. Lethality Bivariate Results Bivariate relationships are examined between each EVIL DONE variable and the dependent variable lethality. These results are located in Table 5.1. For all incidents, 344 (79.3%) of these incidents involve the death of only one victim whereas 90 (20.7%) incidents involve the death of more than one victim. Six of the EVIL DONE variables are significantly 126 related to lethality: exposed, iconic, legitimate, destructible, occupied and easy, but not all of these findings are consistent with what was expected. The iconic and occupied variables are significantly related to the lethality of an incident. Incidents that rank higher on these variables (more vulnerable) are more likely to have multiple deaths as compared to incidents with lower rankings on these two variables. Iconic locations are significantly more likely to involve the deaths of multiple victims as compared to locations that are not iconic. Nearly one-fourth of incidents that involve multiple deaths occur at an iconic location, whereas less than 10% of incidents that involve the death of one individual occur at an iconic location. The variable occupied examines how many other individuals are near the victim when they are attacked. Incidents that involve the death of only one individual occur at a location with no others (other than victim and suspect(s)) around 57.1% of the time. When multiple individuals are killed, 45.1% of the time the incident occurs at a location with no others around. Nearly 17% of incidents that involve multiple deaths occur at a location with 101 or more others around as compared to 3.1% of incidents that involve the death of only one individual. This last finding is very intuitive in that locations that have more individuals around have more deaths. Four other EVIL DONE variables are significant: exposed, legitimate, destructible and easy, but these findings are different than expected. Attacks that involve one death occur at locations that are accessible and routinely frequented by the public day and night 43% of the time compared to attacks that involve multiple deaths that occur at similar settings 30% of the time. The majority of attacks that involve multiple deaths occur at locations that are inaccessible to the public without permission day and night. The attacks on 9/11 of the Twin Towers and the 127 Pentagon, and the attack on the Alfred P. Murrah Federal Building are examples of attacks that caused mass casualties and occurred at a location that is not considered easily accessible. Another example, in 2008, Charles Lee Thornton killed five people at the City Hall in Kirkwood Missouri. This attack occurred in a location that is not easily accessible to the public day or night and involved multiple deaths. When multiple people are killed, it is more likely that the attack is at a location that houses individuals working for a target organization as opposed to general citizenry. The majority of attacks that involve the death of only one individual occurred at a location that houses only general citizenry (80.5%) compared to 65.6% of attacks that involved multiple deaths. An example of an attack involving multiple deaths that targeted non-civilians is the 2009 attack on Fort Hood by Nidal Malik Hasan that killed thirteen individuals. 128 Table 5.1 EVIL DONE Variables and Lethality (n=434) One Death (n=344) Exposed Victim(s) in a location that was accessible and routinely frequented by public during both day and night Victim(s) in a location that was inaccessible to public without permission both during the day and night Victim(s) in a location that was accessible but rarely frequented by the public either during the day or night Victim(s) in a location that was accessible and routinely frequented by public during the day or night Vital Victim(s) in a location that if eliminated would have no impact on day to day functioning of the community Victim(s) in a location that if eliminated would have a great impact on the day to day functioning of the community Iconic Victim(s) in a location that is not iconic Victim(s) in a location that is iconic Legitimate Victim(s) in a location that houses only general citizenry Victim(s) in a location that houses general citizenry and those working for target organization Victim(s) in a location that houses only those working for target organization ***p<.001, **p<.01,*p<.05 129 Multiple Deaths (n=90) N % N % 147* 42.7% 27 30.0% 106 30.8% 41 45.6% 46 13.4% 6 6.7% 45 13.1% 16 17.8% 231 67.2% 57 63.3% 113 32.8% 33 36.7% 315*** 91.6% 29 8.4% 69 21 76.7% 23.3% 277*** 80.5% 59 65.6% 60 17.4% 23 25.6% 7 2.0% 8 8.9% Table 5.1 (cont’d) One Death (n=344) N Destructible Victim(s) in a location that is easily destructible Victim(s) in a location that is difficult to destroy Victim(s) in a location that is moderately destructible % 173*** 50.3% 8 2.3% 163 47.4% Occupied There are no other individuals around at time of incident 196*** 57.0% Victim(s) in a location with 1-5 other people near them 81 23.5% Victim(s) in a location with 6-25 other people near them 41 11.9% Victim(s) in a location with 26-100 people near them 16 4.7% Victim(s) in a location with 101 or more people near them 10 2.9% Multiple Deaths (n=90) N % 22 17 24.4% 18.9% 51 56.7% 41 45.6% 19 21.1% 8 7 8.9% 7.8% 15 16.7% Near Offender(s) lived in same city as target Offender(s) lived 101 or more miles from target Offender(s) lived outside city within 100 miles 170 40 134 49.4% 11.6% 39.0% 47 12 31 52.2% 13.3% 34.4% Easy Victim(s) in a location with no security Victim(s) in a location with high security Victim(s) in a location with some security 269* 8 67 78.2% 2.3% 19.5% 59 6 25 65.6% 6.7% 27.8% ***p<.001, **p<.01,*p<.05 The destructible findings are very interesting. Attacks that occur at locations that are difficult to destroy appear to be more lethal. Nearly 50% of incidents that involve one death are at easily destructible locations compared to only 24.2% of attacks involving multiple deaths at such locations. Attacks against victims who are located in areas that are difficult to destroy involve multiple victims 18.9% of the time and only 2.3% of the time when there is only one 130 victim. This makes sense that if a large structure is able to be successfully attacked that it would result in multiple deaths. Finally, the easy variable is also significant. Lethality is higher in attacks that occur at a location with security. Seventy-eight percent of attacks that involve the death of only one victim occurred at a location with no security compared to 64.8% of incidents with multiple deaths. It appears that suspects in incidents where multiple victims are killed are more willing to attack victims in a location with at least some security (35% of the time compared to 22% of the time in incidents with one death). Correlations between suspect characteristics, victim characteristics and ideological variables and lethality are presented in Table 5.2. The ideology of the suspect is significantly related to lethality. When jihadists commit attacks, they are more likely to involve multiple deaths as compared to attacks that are committed by right-wing extremists. Nearly 90% of attacks committed by right-wing extremists involve the death of only one individual, but onefourth of attacks committed by jihadists involve the death of multiple individuals. Gender and occupation are also significantly related to lethality. Men are more likely to be victims of extremist homicide in attacks involving one death (79.4% of the time) and in attacks where multiple individuals die (60% of the time). However, it is worth noting how the proportional distributions change based on the number of deaths. Women are the victims in 20% of the attacks involving one death but that number increases to 40% when considering multiple deaths. This may be because these attacks are often ideological and occur in settings where there are more likely to be a mixed group of individuals. The occupation variable is also significantly related to the lethality of the incident. This variable only contains 60% of all incidents due to missing cases. Of the incidents examined, police/government workers appear to be proportionately more likely to be killed in an incident involving multiple deaths and blue-collar 131 workers are proportionately less likely to be killed in an incident involving multiple deaths as compared to all other occupations. Table 5.2 Independent Variables and Lethality (n=434) One Death Multiple (n=344) Deaths(n=90) N % N % Victim Characteristics Gender Female 71*** 20.6% 36 40.0% Male 273 79.4% 54 60.0% Race White Non-white 225 119 65.4% 34.6% 58 32 64.4% 35.6% 17 and under 18-24 25-49 50+ 19 65 186 74 5.5% 18.9% 54.1% 21.5% 8 16 55 11 8.9% 17.8% 61.1% 12.2% Unemployed/homeless Blue collar White collar Police/government 66** 41 43 38 35.1% 21.8% 22.9% 20.2% 15 7 14 22 25.9% 12.1% 24.1% 37.9% Strangers Victim knew suspect 201 143 58.4% 41.6% 44 46 48.9% 51.1% Age Occupation Victim-suspect relationship ***p<.001, **p<.01,*p<.05 132 Table 5.2 (cont’d) One Death (n=344) N % Multiple Deaths(n=90) N % Suspect Characteristics Gender Female Male 3 341 0.9% 99.1% 1 89 1.1% 98.9% White 301*** 87.5% Non-White 43 12.5% 61 29 67.8% 32.2% Race Age 14-24 25-49 50+ 125 198 21 36.3% 57.6% 6.1% 27 52 11 30.0% 57.8% 12.2% Unemployed/homeless Blue collar White collar 73* 50 15 52.9% 36.2% 10.9% 25 14 15 46.3% 25.9% 27.8% One suspect Two suspects Three or more suspects 179 87 78 52.0% 25.3% 22.7% 55 14 21 61.1% 15.6% 23.3% Not a lone wolf Lone wolf 259 85 54.3% 17.8% 66 24 36.7% 13.3% 211** 133 61% 39% 69 21 77% 23% Occupation Number of suspects Lone wolf Weapon Firearm Other ***p<.001, **p<.01,*p<.05 133 Table 5.2 (cont’d) One Death (n=344) N % Multiple Deaths(n=90) N % Ideology Suspect ideology Jihadist Right Wing 39** 305 11.3% 88.7% 21 69 23.3% 76.7% Non-ideological Ideological 160 184 46.5% 53.5% 39 51 43.3% 56.7% Incident ideology ***p<.001, **p<.01,*p<.05 There are three suspect variables that are related to the lethality of an attack. First, there appear to be differences in lethality based on the occupation of the suspect Attacks that are committed by suspects who hold a blue collar or white-collar job are more likely to cause multiple casualties as compared to attacks committed by suspects who are unemployed. Second, the race of suspects is found to be significantly related to the lethality of an incident. Nearly 90% of the attacks that involve the death of only one victim are committed by a white suspect and 67.8% of attacks that involve the death of multiple victims are committed by white suspects. While most attacks are committed by white suspects, non-white suspects are proportionately more likely to commit attacks that involve the death of multiple victims as compared to white suspects. Finally, the weapon used in an attack is significantly related to the lethality of an incident. The majority of attacks that caused multiple deaths (77%) are committed with a firearm, whereas attacks that caused only one death are committed with a firearm 60% of the time. This shows that firearms are often the choice weapon for suspects in all attacks but more specifically when they are committing an attack, if they use a firearm they are more likely to cause multiple deaths. 134 Multivariate Results: All Incidents This section examines how each grouping of variables impacts lethality (EVIL DONE, ideology, suspect, and victim variables). The final equation includes all the variables that are significant from the previous four equations in one model. The first binary logistic model includes all eight EVIL DONE variables and these results are located in Table 5.3.11 Only one EVIL DONE variable is significantly related to lethality when controlling for the other EVIL DONE variables. The variable difficult to destroy is significantly related to lethality in comparison to the reference category of locations that are easy to destroy. Locations that are difficult to destroy are more than eight times as likely to involve multiple deaths as compared to locations that are easy to destroy. This relationship is the opposite of the direction that was predicted. A reason for this finding may be that locations that are difficult to destroy are often large buildings and structures. On a surface level, it may be more attractive for a suspect to attack a victim in an easily destructible location but if the suspect wants to commit a large-scale attack they often seek large structures. Therefore, when the suspect can successfully attack a difficult to destroy location they are more likely to cause multiple deaths. 11 Binary logistic regression models were also conducted that excluded the 9/11 attacks and the Oklahoma City Bombing to see if significant vulnerability characteristics remain significant when outlier attacks are not included. No differences in significance are found when these incidents are excluded from analyses. 135 Table 5.3 Multivariate Results: EVIL DONE and Lethality (n=434) B Exposed Victim(s) in a location that was accessible and routinely frequented by public during both day and night (ref. category) Victim(s) in a location that was inaccessible to public without permission both during the day and night .614 Victim(s) in a location that was accessible but rarely frequented by the public either during the day or night -.214 Victim(s) in a location that was accessible and routinely frequented by public during the day or night -.849 Vital Victim(s) in a location that if eliminated would have no impact on day to day functioning of the community Victim(s) in a location that if eliminated would have a great impact on the day to day functioning of the community Iconic Victim(s) in a location that is not iconic Victim(s) in a location that is iconic Legitimate Victim(s) in a location that houses only general citizenry Victim(s) in a location that houses general citizenry and those working for target organization Victim(s) in a location that houses only those working for target organization Destructible Victim(s) in a location that is easily destructible Victim(s) in a location that is difficult to destroy Victim(s) in a location that is moderately destructible ***p<.001, **p<.01,*p<.05 136 S.E. Exp(B) 0.599 1.849 0.551 0.807 0.573 0.428 -.322 0.37 0.724 (ref. category) .780 0.597 2.182 .415 0.722 1.514 .377 0.939 1.458 0.941 8.139 0.577 1.465 (ref. category) (ref. category) (ref. category) 2.097* .382 Table 5.3 (cont’d) B Occupied There are no other individuals around at time of incident (ref. category) Victim(s) in a location with 1-5 other people near them .122 Victim(s) in a location with 6-25 other people near them -.056 Victim(s) in a location with 26-100 people near them -.056 Victim(s) in a location with 101 or more people near them 1.011 S.E. Exp(B) 0.327 1.13 0.463 0.584 0.945 1.65 0.66 2.748 Near Offender(s) lived in same city as target Offender(s) lived 101 or more miles from target Offender(s) lived outside city within 100 miles (ref. category) .049 .072 0.431 0.288 1.05 1.075 Easy Victim(s) in a location with no security Victim(s) in a location with high security Victim(s) in a location with some security (ref. category) -1.362 .179 0.986 0.497 0.256 1.195 0.381 0.126 Constant Nagelkerke R2 Cox and Snell R2 -2 Log Likelihood -2.072*** .169 .108 393.485 ***p<.001, **p<.01,*p<.05 The next regression model examines ideological variables and includes the ideological motivation of incidents and the suspect’s ideological affiliation. The results of this analysis are located in Table 5.4. Suspects who possess a jihadist ideology appear to be more likely to be involved in attacks that involve multiple deaths as compared to suspects who adhere to a rightwing ideological belief system. The ideological motivation is not significantly related to lethality. 137 Table 5.4 Multivariate Results: Lethality and Ideology (n=434) Suspect ideology B S.E. Exp(B) Jihadist (ref. category) Right Wing -0.862** 0.31 0.422 Incident ideology Non-ideological (ref. category) Ideological 0.022 Constant Nagelkerke R2 Cox and Snell R2 -2 Log Likelihood -.635 0.25 1.022 0.33 0.53 .028 0.018 435.33 ***p<.001, **p<.01,*p<.05 Suspect characteristics are presented in Table 5.512 Two variables are significant in this model. The suspect’s race is significantly related to the lethality of an incident. Non-whites are more than three times as likely to be involved in multiple death attacks than whites. This variable is similar to the suspect ideology variable since many non-whites are jihadists. The weapon the suspect uses is also related to the lethality of the attack. Incidents that involve firearms are twice as likely to have multiple deaths as compared to attacks committed with another type of weapon. 12 A regression model was conducted that includes the occupation variable and no significance was found for this variable. This variable is excluded from the analysis due to the number of missing cases. 138 Table 5.5 Multivariate Results: Lethality and Suspect Characteristics (n=434) B S.E. Exp(B) Race White (ref. category) Non-white 1.225*** 0.29 3.405 Age 14-24 (ref. category) 25-49 50+ .124 .781 0.277 0.477 1.132 2.183 Number of suspects One suspect (ref. category) Two suspects Three or more suspects -.693 .195 0.359 0.335 0.5 1.215 Not a lone wolf (ref. category) Lone wolf -.349 0.329 0.706 Other (ref. category) Firearm .669* 0.293 1.952 -2.2027 0.347 0.132 Lone wolf Weapon Constant Nagelkerke R2 Cox and Snell R2 -2 Log Likelihood .110 .070 411.436 ***p<.001, **p<.01,*p<.05 Regression models are conducted to examine victim characteristics and lethality when controlling for other victim characteristics and results are located in Table 5.6 and Table 5.7. The first analysis excludes the occupation variable because of the number of missing cases. In this model, only one victim variable is significant. Women are nearly three times as likely to be 139 killed in an attack that causes multiple deaths as compared to men. Victims who know the suspect are more likely to be killed in an attack involving multiple deaths as compared to victims who did not know the suspect. Several domestic violence homicide attacks involve multiple victims. The suspect may kill their significant other but also children, other family members, and co-workers. Table 5.6 Multivariate Results: Lethality and Victim Characteristics (n=434) B S.E. Exp(B) Gender Male (ref. category) Female .924*** 0.266 2.519 Race White (ref. category) Non-white .015 0.261 1.015 Age 17 and under 18-24 (ref. category) 25-49 50+ .443 0.518 1.558 .397 -.332 0.334 0.441 1.487 0.718 Victim-suspect relationship Strangers (ref. category) Victim knew suspect .208 0.261 1.232 -1.924*** 0.365 0.146 Constant Nagelkerke R2 Cox and Snell R2 -2 Log Likelihood 0.068 0.044 423.71 ***p<.001, **p<.01,*p<.05 140 Table 5.7 Multivariate Results: Lethality and Victim Characteristics with Occupation (n=246) B S.E. Exp(B) Gender Male (ref. category) Female 1.6*** 0.414 4.989 Race White (ref. category) Non-white .094 0.362 1.098 0.706 0.667 0.569 0.646 0.865 0.509 0.396 2.216 0.549 0.547 0.471 0.157 0.194 0.352 0.619 0.565 Age 17 and under -.405 18-24 (ref. category) 25-49 -.145 50+ -.675 Victim-suspect relationship Strangers (ref. category) Victim knew suspect .796* Occupation Police/government (ref. category) Unemployed/homeless -1.850*** Blue collar -1.640** White collar -1.045* Constant Nagelkerke R2 Cox and Snell R2 -2 Log Likelihood -.571 .214 .142 231.063 ***p<.001, **p<.01,*p<.05 The second analysis involving victim variables includes the occupation variable. Several variables are significant in this analysis. The gender of the victim remains significant. The variable victim and suspect relationship variable becomes significant. Victims who know the suspect are twice as likely to be killed in an attack that involves multiple deaths as compared to 141 strangers. The reason for this may be that many domestic abuse situations that turn into homicides involve the deaths of multiple family members. Additionally, many hate crimes are targeted at one individual. Victims who are police officers or government employees are significantly more likely to be killed in an attack involving multiple deaths as compared to victims who have a blue-collar job, white-collar job or who are unemployed/homeless. Finally, all significant variables from the first four equations are included in the equation13. Two final equations are conducted. The first equation includes all incidents (n=434) and therefore excludes the victim occupation variable. The second equation includes the victim occupation variable as well as the victim-suspect relationship variable and looks at that subset of the dataset (n=246). These results are located in Table 5.8 and Table 5.9. In the first analysis, including all incidents, all variables remain significant except for suspect ideology. Women are more likely to be killed in multiple death attacks than men are, specifically 2.4 times more likely. Suspect race is also significantly related to lethality. Non-whites are 3.6 times more likely to commit an attack that kills multiple individuals as compared to whites. Attacks that occur at locations that are moderately difficult to destroy are more than two times as likely to involve the death of multiple victims as compared to attacks that occur at a location that is easy to destroy. When an attack occurs at a location that is difficult to destroy it is 14.1 times more likely to involve multiple deaths as compared to locations that are easy to destroy. 13 There is concern that the suspect race variable the suspect ideology variable is measuring the same concept. A regression is conducted that includes both variables. A regression is also conducted that excludes the race variable and another regression is conducted that excludes the suspect ideology variable. The race variable remains significant whether or not ideology is included and ideology on its own is not significant at the multivariate level. 142 Table 5.8 Multivariate Results: Equation with Previously Significant Variables (n=434) B S.E. Exp(B) Ideology Variables Suspect ideology Jihadist (ref. category) Right Wing .455 0.595 2.56 Victim Variables Gender Male (ref. category) Female .983*** 0.282 2.672 White (ref. category) Non-white 1.060 0.553 2.885 Other (ref. category) Firearm .757* 0.302 2.131 0.289 0.51 2.208 14.702 -3.409*** 0.656 0.033 Suspect Variables Race Weapon EVIL DONE Variables Destructible Victim(s) in a location that is easily destructible (ref. category) Victim(s) in a location that is moderately difficult to destroy .792** Victim(s) in a location that is difficult to destroy 2.688*** Constant Nagelkerke R2 Cox and Snell R2 -2 Log Likelihood 0.222 0.142 376.728 ***p<.001, **p<.01,*p<.05 The next equation examines all previously significant variables including the victim’s occupation and the relationship between the victim and the suspect. The gender of the victim remains significant and women are nearly six times as likely to be killed in an attack that involves the death of multiple victims. All three victim occupation variables remain significant. 143 Victims who work as a police officer or government employee are six times as likely to be killed in an attack that involves multiple deaths as compared to victims who have other occupations. Additionally, the weapon the suspect used remains significant. Suspects who commit an attack using a firearm are twice as likely to cause multiple fatalities as suspects who commit an attack using a different weapon. The destructible variable also remains significant and attacks that occur at a location that is difficult to destroy are nearly 18 times as likely to involve multiple deaths as compared to attacks that occur at a location that is easy to destroy. 144 Table 5.9 Multivariate Results: Equation with Previously Significant Variables with Victim Occupation (n=246) B S.E. Exp(B) Ideology Variables Suspect Ideology Jihadist (ref. category) Right Wing .803 0.898 2.233 Victim Variables Gender Male (ref. category) Female 1.778*** 0.475 5.916 (ref. category) -1.907*** -1.824** -1.691** 0.561 0.607 0.541 0.149 0.161 0.184 (ref. category) .494 0.438 1.638 (ref. category) 1.142 0.829 3.134 (ref. category) -.119 0.441 0.888 0.409 0.629 1.88 17.604 0.923 0.132 Occupation Police/government Unemployed/Homeless Blue Collar White collar Victim-Suspect Relationship Strangers Victim knew suspect Suspect Variables Race White Non-white Weapon Other Firearm EVIL DONE Variables Destructible Victim(s) in a location that is easily destructible (ref. category) Victim(s) in a location that is moderately difficult to destroy .631 Victim(s) in a location that is difficult to destroy 2.868*** Constant Nagelkerke R2 Cox and Snell R2 -2 Log Likelihood -2.028 .351 0.234 203.266 ***p<.001, **p<.01,*p<.05 145 Multivariate Results: Ideological v. Non-Ideological Incidents Table 5.10 and Table 5.11 display multivariate results for EVIL DONE variables and lethality for ideological incidents and non-ideological incidents respectively. Only one EVIL DONE variable, destructible, is significant throughout all four models. It is important to acknowledge potential concerns of multicollinearity that are worth addressing. Multicollinearity diagnostics are conducted prior to any regression analysis. All EVIL DONE variables had VIF scores less than 2.7 and no two EVIL DONE variables has a correlation above .6. There does not appear to be any collinearity issues when examining EVIL DONE variables for all incidents. Incidents are examined separately based on if the ideological motivation of the attack. When examining just non-ideologically motivated attacks, all VIF scores are 2.8 or below and all VIF scores are 3.1 or below for ideologically motivated attacks and all correlations for both remain .7 or below. EVIL DONE collinearity is also examined for jihadist incidents and right-wing extremist incidents with all VIF scores being 2.6 or below and no correlations being above .6. Collinearity diagnostics do not point to any significant concerns between the variables, however, on a theoretical level there is concern that the variables may be overlapping or measuring a similar concept. When subsections of the data are examined (right-wing v. jihadist) and (non-ideological v. ideological) some of the EVIL DONE variables have limited variability (i.e. very few cases in certain predictors). This can cause the standard errors to rise. Four EVIL DONE variables are chosen to be examined for the next four regression analyses. When examining subsets of the data in this manner, there is not enough statistical power to include all eight EVIL DONE variables. The four variables that are included are chosen 146 because significant relationships at the bivariate level between EVIL DONE variables and ideological factors and the potential theoretical contribution. The variables that are examined are: vital, iconic, destructible and near. These variables are chosen because they are all measuring distinct concepts and have high variability within the variables. Vital and iconic are chosen because they are two EVIL DONE variables that are significant across bivariate ideological relationships in Chapter four. Vital and iconic are significantly related to the ideological motivation of an incident as well as the ideological affiliation of the suspect. Chapter four looked at the relationship between EVIL DONE variables and ideology and found there to be significant differences between right wing and jihadist attacks (for all incidents, for ideological incidents only and for non-ideological incidents). The next variable included is destructible. This variable is included because it appears to be an important predictor of lethality in prior regression models. There appears to be a relationship between how easily destructible a target is and the number of deaths that occur in an attack. The final variable chosen is the near variable. This variable is chosen because Clarke and Newman (2006) argue that near will be the most important predictor variable. They argue that suspects are more likely to commit attacks near where they live because they know the opportunities in that area. Only one of the four EVIL DONE variables is significant for ideological incidents, the destructible variable. In attacks that are ideologically motivated, if they occur at a location that is difficult to destroy they are more than 14 times as likely to involve multiple deaths as compared to attacks that occur at a location that is easy to destroy. 147 Table 5.10 Multivariate Results: EVIL DONE Variables and Lethality for Ideological Incidents (n=256) B S.E. Vital Victim(s) in a location that if eliminated would have no impact on day to day functioning of the community (ref. category) Victim(s) in a location that if eliminated would have a great impact on the day to day functioning of the community -.114 0.413 Iconic Victim(s) in a location that is not iconic Victim(s) in a location that is iconic Destructible Victim(s) in a location that is easily destructible Victim(s) in a location that is difficult to destroy Victim(s) in a location that is moderately destructible Near Offender(s) lived in same city as target Offender(s) lived 101 or more miles from target Offender(s) lived outside city within 100 miles 0.57 1.455 (ref. category) 2.647*** 0.713 14.11 .659 0.384 1.933 (ref. category) -.638 -.141 0.571 0.381 0.529 0.868 0.37 0.168 1.787*** .185 .120 215.806 ***p<.001, **p<.01,*p<.05 148 0.892 (ref. category) .375 Constant Nagelkerke R2 Cox and Snell R2 -2 Log Likelihood Exp(B) Table 5.11 Multivariate Results: EVIL DONE Variables and Lethality for Non-Ideological Incidents (n=199) B S.E. Iconic Victim(s) in a location that is not iconic (ref. category) Victim(s) in a location that is iconic -.238 1.318 Vital Victim(s) in a location that if eliminated would have no impact on day to day functioning of the community Victim(s) in a location that if eliminated would have a great impact on the day to day functioning of the community Constant Nagelkerke R2 Cox and Snell R2 -2 Log Likelihood 0.788 (ref. category) -.149 Destructible Victim(s) in a location that is easily destructible (ref. category) Victim(s) in a location that is difficult to destroy 2.966* Victim(s) in a location that is moderately destructible 1.035* Near Offender(s) lived in same city as target Offender(s) lived 101 or more miles from target Offender(s) lived outside city within 100 miles Exp(B) (ref. category) 1.038 .045 -2.249*** 0.609 0.862 1.391 19.417 0.461 2.816 0.641 0.42 2.823 1.046 0.469 0.106 0.107 0.067 183.033 ***p<.001, **p<.01,*p<.05 The analysis for non-ideologically motivated incidents is presented in Table 5.11. Similar to the ideological incident results, attacks that are committed for non-ideological reasons are likely to involve multiple deaths if they occur at a location that is difficult to destroy as compared to attacks that occur at locations that are easy to destroy. Attacks that occur at 149 locations that are difficult to destroy are more than 19 times as likely to involve multiple deaths as compared to attacks that occur at a location that is easy to destroy. For non-ideological attacks, moderately destructible locations are also significant, with attacks that occur at a moderately destructible location being nearly three times as likely to involve multiple deaths as compared to attacks that occur at locations that can easily be destroyed. Z tests are conducted to see if there are differences between these two models. Z tests are helpful to examine differences between two populations when the variance is known. No significant differences are found for vital, iconic or near. The destructible variable is significant between the two models. Differences in EVIL DONE variables were found when examining EVIL DONE and ideological factors in Chapter four. However, it is interesting to see here that when looking at EVIL DONE variables there only appear to be differences in lethality based on the destructible variable. Multivariate Results: Right-Wing v. Jihadist Incidents Incidents are also examined to see differences in EVIL DONE vulnerability by the suspect’s ideological affiliation, whether they are a right-wing extremist or a jihadist extremist. These results are located in Table 5.12 and Table 5.13. 150 Table 5.12 Multivariate Results: EVIL DONE Variables and Lethality for Right-Wing Incidents (n=374) B Vital Victim(s) in a location that if eliminated would have no impact on day to day functioning of the community (ref. category) Victim(s) in a location that if eliminated would have a great impact on the day to day functioning of the community -.080 S.E. Exp(B) 0.371 0.923 Iconic Victim(s) in a location that is not iconic Victim(s) in a location that is iconic (ref. category) .241 0.607 1.273 Destructible Victim(s) in a location that is easily destructible Victim(s) in a location that is difficult to destroy Victim(s) in a location that is moderately destructible (ref. category) 2.641*** .813* 0.76 0.318 14.03 2.256 Near Offender(s) lived in same city as target Offender(s) lived 101 or more miles from target Offender(s) lived outside city within 100 miles (ref. category) .056 -.157 0.561 0.301 1.057 0.855 0.313 0.129 Constant Nagelkerke R2 Cox and Snell R2 -2 Log Likelihood -2.049*** 0.111 0.068 331.17 ***p<.001, **p<.01,*p<.05 151 Table 5.13 Multivariate Results: EVIL DONE Variables and Lethality for Jihadist Incidents (n=60) B S.E. Vital Victim(s) in a location that if eliminated would have no impact on day to day functioning of the community (ref. category) Victim(s) in a location that if eliminated would have a great impact on the day to day functioning of the community -1.160 0.785 Iconic Victim(s) in a location that is not iconic Victim(s) in a location that is iconic Destructible Victim(s) in a location that is easily destructible Victim(s) in a location that is difficult to destroy Victim(s) in a location that is moderately destructible Near Offender(s) lived in same city as target Offender(s) lived 101 or more miles from target Offender(s) lived outside city within 100 miles Constant Nagelkerke R2 Cox and Snell R2 -2 Log Likelihood Exp(B) 0.313 (ref. category) -.381 0.975 0.683 (ref. category) 3.407** 1.233 30.176 .967 0.751 2.63 (ref. category) -.208 1.249 0.743 0.847 0.812 3.486 -1.214 0.826 0.297 .315 .228 62.139 ***p<.001, **p<.01,*p<.05 The same variable, destructible, that is significant in ideological and non-ideological models is also significant in the jihadist model and the right-wing model. Attacks committed by right-wing extremists are 14 times as likely to involve multiple deaths if at a location that is difficult to destroy as compared to attacks that are at locations that are easy to destroy. The difference is even greater for attacks committed by jihadists, when jihadists commit an attack at a 152 location that is difficult to destroy it is more than thirty times as likely that the attack will involve multiple deaths as compared to when jihadists commit an attack at a location that can easily be destroyed. These findings are interesting yet not surprising given the destructible variable is the only EVIL DONE variable that is significant in previous models. It appears clear that regardless of ideological motivation or affiliation that if an attack is successfully committed at a location that is difficult to destroy it is likely that multiple people will be killed. Z tests are also conducted between these two models in order to see if there are differences between coefficients in the two separate regressions. Coefficients are significantly different for two variables between jihadists and right-wing extremists. The coefficients for the difficult to destroy variable are significantly different between the two models. Attacks committed by jihadists that are committed against a target that is difficult to destroy are 30 times as likely to result in multiple deaths as compared to when they attack targets that are easily destructible. Right-wing extremists are also more likely to commit attacks that result in multiple deaths when they are attacking targets that are difficult to destroy, however, only one-half as likely as jihadists are. Significance was also found with one of the near variables. Attacks committed by an extremist living within 100 miles appear to be different between the two models. However, significance is not found for this variable in either of the models separately. 153 Chapter 6: Discussion The final chapter seeks to accomplish several goals. First, there is a review and discussion of major research findings. After this, policy recommendations and research implications are addressed. Third, this chapter describes the limitations of this research project. Finally, the chapter concludes with a discussion of areas for future research. Review of Major Findings This study sought to examine the relationship between vulnerability and lethality in domestic extremist incidents. This study sought to apply a target identification technique from SCP called EVIL DONE. This framework was originally proposed as a technique to be used by practitioners to identify vulnerable physical targets. This study sought to expand this framework and apply it to human targets and further to see if it could explain lethality in prior domestic extremist attacks. Suspect, victim and ideological characteristics are also examined to help explain the situational context of prior lethal attacks and see what factors may predict lethality. This study applied a SCP framework that was intended for physical targets to human targets. There appears to be value in examining the relationship between the situational environment and lethality. Despite not finding support for the EVIL DONE framework in predicting lethality, there are several important findings from this study about other important predictors of lethality. These findings will be addressed by closely examining each research question. The first research question is addressed in Chapter four and examines the presence of EVIL DONE vulnerability characteristics in prior lethal domestic extremist attacks. This question further looks to see how these characteristics vary based on ideology, suspect factors and victim factors. These findings help paint a better picture of vulnerability in lethal attacks. 154 Findings indicate there are clear differences in vulnerability based on the ideological motivation of the attack and these differences mirror those of attacks based on suspect ideology. In other words, attacks committed for ideological purposes rank high in the same vulnerability characteristics as attacks committed by jihadists. Further, attacks committed for non-ideological purposes rank higher in the same vulnerability characteristics as attacks committed by right-wing extremists. Attacks that are committed for ideological purposes rank higher on exposed, vital, iconic, legitimate and occupied as compared to non-ideologically motivated attacks. This indicates that when a suspect commits an attack for ideological purposes they find targets that are easily accessible, vital, iconic, legitimate and occupied to be more attractive than if a suspect is committing an attack for a non-ideological reason. Ideologically motivated incidents, however, rank lower on destructible, near and easy. Suspects committing ideologically motivated attacks appear to be more willing to attack targets that are difficult to destroy, are further from their residences and have higher security as compared to suspects who are committing nonideologically motivated homicides. The majority, 86.2%, of incidents in this study are attacks committed by right-wing extremists. Jihadists are more likely to value the vulnerability characteristics of exposed, vital, iconic, legitimate and occupied, while right-wing extremists rank higher on the destructible, near and easy variables. Interestingly, these findings coincide with the incident motivation findings with the same five EVIL DONE factors being proportionately higher for jihadists as compared to right-wing extremists and then for ideological attacks as compared to non-ideological attacks. In other words, for jihadists or those committing an ideological attack certain features of targets appear to be more valuable than features of targets of non-ideological or of right-wing attacks. 155 This study also paints a picture of what the average lethal domestic extremist attack looks like. Most suspects in this study are white males (83.4%), between the ages 25-49 (57.6%), and most suspects commit their attacks using a firearm (60%). Differences in EVIL DONE vulnerability are found between several suspect characteristics. White suspects seem to value the vulnerability characteristics of destructible, near and easy while non-white suspects value the characteristics of exposed, vital, iconic, legitimate and occupied. Suspects who are 50 and older also value these same characteristics while younger suspects (14-24) are more likely to attack victims at locations that rank high on the characteristics of exposed, destructible, near and easy. Blue collar suspects are more likely to attack highly occupied areas. White collar suspects are more likely to attack victims at locations that are difficult or moderately difficult to destroy as well as locations with high security. Lone wolves rank higher on exposed, vital, iconic, legitimate and occupied as compared to attacks that are committed by non-lone wolf suspects. Lone wolves appear to value several EVIL DONE vulnerability characteristics more than those who are not lone wolves. The average victim in this study is a white (65%) male (75%) between the ages of 25 and 49 (56%). Males are more likely to be in exposed and destructible areas while non-whites are more likely to be in exposed and vital areas. Victims who hold white collar jobs rank higher on several EVIL DONE vulnerability characteristics including vital, iconic, legitimate and occupied. All categories of victim occupation rank higher on the near variable than white-collar victims. Victims who are strangers to the suspect rank higher on six EVIL DONE factors including exposed, vital, iconic, legitimate, destructible and occupied. Victims who know the suspect rank high on near and easy as compared to victims who are strangers. 156 The second research question is addressed in Chapter five and examines the impact of vulnerability factors on lethality. The relationship between EVIL DONE factors, suspect factors, victim factors and ideological factors and lethality are examined. The findings for two EVIL DONE variables are as expected. Multiple deaths appear to be more likely at locations that are iconic and in locations that house a great deal of people (occupied). Four EVIL DONE variables have findings that are the opposite of what was expected. Attacks appear to be more likely to cause multiple deaths if they are at locations that are not exposed, locations with security, locations that house those working for the target organization and at locations that are difficult to destroy. The sixth variable of EVIL DONE, destructible, has findings that refute the sixth hypothesis. Specifically, incidents that occur at locations that are difficult to destroy are more likely to have multiple deaths as compared to victims at a location that can easily be destroyed. Gruenewald and colleagues (2015), when applying EVIL DONE to attacks committed by environmental extremists had similar findings. They found that most attacks committed by environmental extremists are committed at a location that required either a large or small IED to be destroyed. In examining EVIL DONE in relation to international incidents, Paton (2013) found that no combination of EVIL DONE factors could significantly predict lethality except for destructible by itself which indicates that this may be one of the most important factors related to vulnerability and lethality. She found that the destructibility of a location was negatively related to lethality. This conflicts with prior literature that argues that difficult to destroy locations are going to be less attractive because the perceived effort to attack them will be too high. This finding makes sense in that this study applied EVIL DONE to locations that have been successfully attacked. Locations that are difficult to destroy are often large buildings that house 157 many individuals including office buildings, schools, malls and government buildings. These locations may have more security or other factors that make them less attractive but if they are able to successfully attack a location that is difficult to destroy then they are likely to cause more than one death. Therefore, if a suspect is able to successfully commit a crime at that location there is likely to be mass casualties. For example, the attacks on the Twin Towers or the Oklahoma City Bombing. These were large structures and these attacks produced a large number of casualties. Finally, incidents committed by jihadists or other extremists are significantly more likely to involve more than one death as compared to incidents committed by far-right extremists. Ideology has been found to play an important role in target selection (Cook & Lounsbury, 2011). These findings indicate that attacks committed by right-wing extremists are not as deadly as those committed by jihadists. Recent trends in attacks may further display this discrepancy such as the 2015 shooting in San Bernardino, California that killed 14 individuals or the 2016 shooting at an Orlando night club which killed 49 persons, both of which were committed by jihadists. All incidents are examined based on ideological motivation (ideologically motivated vs. non-ideologically motivated) and suspect ideology (right-wing vs. jihadist). Destructible is the only EVIL DONE variable that remains significant throughout all of these analyses. This finding is similar to Paton (2013), finding that destructible is consistently the only EVIL DONE variable that predicts lethality. Regardless of ideological motivation or affiliation, when suspects attack locations that are difficult to destroy more deaths are likely to occur than if suspects attack locations that are easy to destroy. 158 Policy Recommendations and Research Implications There are several research implications from this study for law enforcement, practitioners, as well as academia. The first implication of this study is its contribution to criminological literature. This study provides some answers to EVIL DONE and its applicability beyond a vulnerability assessment tool. This study shows that EVIL DONE can be extended beyond physical targets and applied to human targets. Prior research has examined EVIL DONE but only one study has done so at a multivariate level and that study was examining international attacks, which makes this the first study to examine EVIL DONE at the domestic level using multivariate techniques. This study sought to apply EVIL DONE, a SCP technique to prior lethal domestic extremist attacks. When Clarke and Newman (2006) introduced this framework, it may not have been their intent to have it be applied to targets that have already been attacked and more specifically to be used to try to understand lethality. This study sought to examine the relationship between vulnerability and lethality, using EVIL DONE, victim characteristics, suspect characteristics and ideological characteristics. However, it appears that the EVIL DONE framework may not appear to be an appropriate tool to understand the lethality of attacks. Significance was found at the multivariate level with only one of the eight variables and the relationship found was not found in the expected direction. Second, this study showcases implications for law enforcement and the need for target reduction strategies. Through the SCP framework, crime is a result of the convergence of a motivated offender with the opportunity for a crime (Clarke, 1992). Therefore, it is expected if opportunities are reduced then crime will also decrease. However, implementing these strategies 159 to help protect all potential targets is not feasible. The risk for crime is not evenly distributed and therefore it is necessary to prioritize targets and victims (Gruenewald et al., 2015; Clarke & Eck, 2005). This study sought to identify characteristics that make targets vulnerable. Key relationships between vulnerability and lethality with six EVIL DONE variables can be seen at the bivariate level. There are also key relationships between EVIL DONE variables and other characteristics such as suspect ideology and incident ideological motivation. Law enforcement should be focusing on the exposure of targets, how iconic they are, how occupied, how legitimate, how easy they are to attack and how destructible they are. These are the six variables that are found to be related to the lethality of an attack at the bivariate level. These findings can help practitioners and policy makers to better understand what different types of extremists find attractive or vulnerable depending on the motivation for the attack (ideological or nonideological). Clarke and Newman (2006) have applied 25 techniques that can be used to reduce the opportunity for terrorism which include removing excuses, reducing provocations, reducing rewards, increasing the effort and increasing the risks. These techniques may include such strategies as controlling building access (increasing effort), adding in alarms or security guards (increasing risks) and posting signs with rules clearly stated (removing excuses). Increasing the effort involves identifying vulnerable targets and prioritizing protection of targets. Another way to increase effort is to restrict weapon sales and access to weapons. Stricter laws on gun sales and ways to identify suspects purchasing bomb related materials and reporting them to authorities is essential. Utilizing techniques of crime prevention along with publicizing these efforts may help to deter offenders. 160 Third, this study provides an important contribution by identifying the importance of target destructibility. More specifically, the relationship between destructibility and lethality. If locations that are difficult to destroy are attacked then more deaths are likely to occur as compared to attacks on locations that are easily destructible. This finding is the opposite of what was anticipated. Prior research has argued that extremists will seek locations that house a large number of people, are easily accessible, and can be easily destroyed through conventional weapons (Clarke & Newman, 2006). One example of this is the Boston Marathon bombings. People were clustered together in the race and at the finish line and were in an open setting. One would expect an attack like such to cause mass casualties but only three deaths occurred during this attack. This may be a result of a quick response by law enforcement. These locations and events, on the surface, appear to be highly attractive target choices. However, locations, such as large buildings, have many obstacles for suspects to overcome to successfully attack them. Examples of locations that are difficult to destroy include high schools, college campuses, government buildings, airports, large churches, large office buildings/businesses and military bases. What this research shows though is that when these obstacles are overcome and suspects attack a location that is not easily destructible then there are likely to be more casualties. These findings are novel and research needs to further examine the importance of destructibility in not only lethality but in success of terrorist attacks and target selection. Fourth, law enforcement should focus on developing a catalog of targets within their local community. Targets need to be analyzed and assessed at a local level to determine what locations are the most attractive or vulnerable. Additionally, law enforcement needs to be cognizant that locations that already have high security (such as large office buildings) are still at great risk if they are successfully attacked. Law enforcement needs to be focused on strategies 161 that can identify suspects who are in the planning stages of attacks before they are able to implement them. Law enforcement agencies may benefit from working more closely with joint terrorist task forces (JTTFs) to gain information from analysts in Washington (Jenkins, Liepman & Willis, 2014) and help determine vulnerable targets within their jurisdiction. Police officers may benefit from additional training to help identify potential targets in their area. Finally, these research findings are also important in that they contribute to a discussion about policy and extremist violence. These findings showcase differences between far-right attacks and jihadist attacks as well as differences based on ideological motivation. This showcases that there are certain vulnerability characteristics like exposed, vital, iconic, legitimate and occupied that are more present in ideological attacks and in attacks committed by jihadists. Characteristics such as destructible, near and easy are more likely present in non-ideologically motivated attacks and attacks committed by far-right extremists. This research helps to fill a gap in understanding what influences target choices based on ideology. By examining these differences, lawmakers can become informed policymakers and better allocate government resources. Relatedly, this study adds to the growing conversation about firearms and firearms restrictions in the United States. It is a hotly debated topic and this research showcases that attacks that are committed with firearms are more likely to have multiple casualties as compared to attacks committed with other weapons. There is the argument that a suspect is going to commit a crime regardless of the availability of firearms. However, when a suspect commits a crime with another type of weapon it is not as lethal as when a firearm is used. 162 Research Limitations There are several research limitations to this study that are worth noting. First, one research limitation of this study is the use of open source materials. All incidents and information used in analysis comes from open source data. For some incidents, there is limited information and there is even less information for incidents that occur in the early 1990s, before internet news became popular. Relatedly, a second limitation of this study is the selection of the time frame. The years 1990-2014 are chosen because there is not easily accessible open source information for cases prior to 1990. Third, one of the largest limitations of this study is that EVIL DONE is examined in application to targets that have already been attacked. By examining targets that have already been attacked, these targets are clearly vulnerable in some manner since they were chosen and were successfully attacked. This study is very limited in that there is not a comparison group. Targets that have been attacked should be matched with similar targets that have not been attacked so we can better understand what differences exist and try to understand why one target was chosen and not another. A fourth limitation of this study is the measurement of variables. There is some concern that the EVIL DONE variables are measuring similar concepts. There is also an argument that some of these variables should be weighed more and are more important than other variables. Future research should examine differences in these variables and how they should be weighted. Relatedly, statistical significance is found between several EVIL DONE variables and lethality, however, only one variable is significant at the multivariate level. There are many reasons that this may be the case. One reason that significance disappears at the multivariate 163 level may be because of the way the dependent variable, lethality, is measured. Future research should consider new ways of measuring lethality and also look at not only deaths but injured victims. Due to limited variability in the lethality measure the measure is coded in a binary fashion. Because of this coding important information is lost on why some incidents had multiple deaths. However, due to the nature of the data there are limited options in this regard. Lethality is measured as a dichotomous variable with a one indicating that multiple individuals die and a zero indicating that the attack resulted in only one death. A lot of information is lost by coding the variable in this manner. As discussed in the methods section, the data is skewed toward one death incidents. The vast majority of attacks (80%) involve the death of only one victim. The lethality variable is tested in several ways. It does not meet the criteria necessary for OLS regression, which is what would be needed if the variable is kept as a continuous variable. The variable is split into three categories: one death, two to five deaths and six or more deaths in order to keep some of the variation within the variable. Ordinal logistic regression models are conducted with the data but because of the distribution (there being too few cases in the latter categories) the data was failed key tests that are necessary for ordinal logistic regression. Therefore, binary logistic regression is chosen as the method of analysis. Future Research This study sought to examine the relationship between vulnerability and lethality and further to see if an innovative SCP framework for identifying vulnerable physical targets could be applied to human targets in prior attacks and more specifically to understand lethality. This is one of the first studies to examine EVIL DONE and is the first to examine it in application to domestic terrorism through multivariate analysis. Further, this study is one of the first to examine the relationship between vulnerability and lethality in domestic extremist incidents. This study 164 did not find support for the EVIL DONE framework at being able to predict lethality, meaning that many factors of EVIL DONE may not be associated with targets that are actually chosen by terrorists. Despite this, this study has several important findings and helps contribute to a new and growing body of research on terrorist target vulnerability. There are still several areas that future research should examine. First, no significance is found between several vulnerability factors and lethality. Future research should consider new ways to measure vulnerability as well as lethality. This study is restricted because of how skewed the lethality variable was toward one death incidents. Other measurements of lethality could include the number of injuries, media attention, damage/destruction caused, local community impact and even terrorist group recruitment or response from attack by the public. Second, if EVIL DONE or a similar framework is examined, there needs to be careful consideration for the measurement of each vulnerability factor. Certain factors of vulnerability, such as the destructible variable, may be more important than others and should therefore hold more weight on a vulnerability scale. There may also be important factors that are impacting vulnerability that are not included in the EVIL DONE framework. If future research seeks to utilize this framework, it would be important to identify which EVIL DONE factors may be most important and consider how these factors should be weighted and how that might vary based on ideological motivation. Additionally, this research may be flawed by taking a framework and trying to apply it beyond what it was intended to be used as. This framework may be best suited to be used by law enforcement at a local level and not to be used to explain lethality or to explain prior attacks. 165 Finally, future research should utilize a control group of targets that have not been attacked to assess the relationship between vulnerability and lethality. By doing this, comparisons can be made between attacked targets and those that have not been attacked to determine what differences in vulnerability exist. A specific city, state or regional area can be chosen and targets that have been attacked in the past can be matched to a similar target (in a similar area or the same area) that has not been attacked. This is very important in order to determine why suspects are finding some targets attractive choices and not others. Even if future research doesn’t utilize EVIL DONE as a whole, many important vulnerability characteristics can help guide future research to look at the intersection of situational characteristics and how they may be related to target choice and lethality. 166 REFERENCES 167 REFERENCES Abrahms, M. (2012). The political effectiveness of terrorism revisited. Comparative Political Studies, 45(3): 366-393. Abrahms, M. (2014). Deterring terrorism: A new strategy. Perspectives on Terrorism, 8(3): 2-15. Agnew, R. (1992). Foundation for a general strain theory of crime and delinquency. Criminology, 30(1): 47-88. Akers, R.L. (1998). Social learning and social structure: A general theory of crime and deviance. Boston: Northeastern University Press. Asal, V. & Rethemeyer, K. (2008). The nature of the beast: organizational structures and the lethality of terrorist attacks. Journal of Politics, 70(2): 437-449. Asal, V., Rethemeyer, K., Anderson, I., Stein, A., Rizzo, J. & Rozea, M. (2009). The softest of targets: A study on terrorist target selection. Journal of Applied Security Research, 4(3): 258-278. Atlas, R., & LeBlanc, W. (1994). Environmental barriers to crime. Ergonomics in Design, 9-16. Bachman, R., & Paternoster, R. (2003). Statistics for criminology and criminal justice. New York: McGraw-Hill Humanities/Social Sciences/Languages. Beccaria, C. (1986). On crimes and punishments. Hackett Publishing Company. (Original work published 1764). Becker, G. (1968). Crime and punishment: An economic approach. Journal of Political Economy, 76(2): 169-217. Bellamy, L. (1996). Situational crime prevention and convenience store robbery. Security Journal, 7(1):41–52. Bellia, P. (2005). The "lone wolf" amendment and the future of foreign intelligence surveillance law. Villanova Law Review, 50: 425-478. Berko, A. & Erez, E. (2007). Gender, Palestinian women, and terrorism: Women’s liberation or oppression? Studies in Conflict & Terrorism, 30(6): 493-519. Bernasco, W. (2010). Modeling micro-level crime location choice: Application of the discrete choice framework to crime at places. Journal of Quantitative Criminology, 26(1):113138. 168 Bloom, M. (2005). Dying to kill: The allure of suicide terror. New York: Columbia University Press. Bloom, M., Gill, P. & Horgan, J. (2012). Ticofaidh ar Mna: Women in the provisional Irish republican army. Behavioral Sciences of Terrorism and Political Aggression, 4(1): 6076. Boba, R. (2009). EVIL DONE in reducing terrorism through situational crime prevention. Crime Prevention Studies, 25: 71-91. Bogen, K. & Jones, E. (2006). Risks of mortality and morbidity from worldwide terrorism: 19682004. Risk Analysis, 26(1): 45-59. Borooah, V. (2009). Terrorist incidents in India, 1998-2004: A quantitative analysis of fatality rates. Terrorism & Political Violence, 21: 476-498. Brandt, P. & Sandler, T. (2010). What do international terrorists target? Has it changed? Are we safer? Journal of Conflict Resolution, 54(2): 214-236. Brantingham, P. & Brantingham, P. (1975). The spatial patterning of burglary. The Howard Journal of Crime and Justice, 14(2): 11-23. Brantingham, P. & Brantingham, P. (1981). Environmental criminology. Beverly Hills: Sage. Brantingham, P. & Brantingham, P. (1993). Nodes, paths and edges: considerations on the complexity of crime and the physical environment. Journal of Environmental Psychology, 13(1): 3-28. Breault, K. & Kposowa, J. (1997). The effects of marital status on adult female homicides in the United States. Journal of Quantitative Criminology, 13(2): 217-230. Brown, J. (1979). Suicide in Britain. More attempts, fewer deaths, lessons for public policy. Archives of General Psychiatry, 36(10): 1119-1124. Bubolz, B. & Simi, P. (2015). Leaving the world of hate: Life-course transitions and self-change. American Behavioral Scientist, 59(12): 1-21. Canetti-Nisim, D., Mesch, G. & Pedahzur. (2006). Victimization from terrorist attacks: Randomness or routine activities? Terrorism and Political Violence, 18(4): 485-501. Carson, J. (2014). Counterterrorism and radical eco-groups: A context for exploring the series hazard model. Journal of Quantitative Criminology, 30: 485-504. Carson, J., LaFree, G. & Dugan, L. (2012). Terrorist and non-terrorist criminal attacks by radical environmental and animal rights groups in the United States, 1970-2007. Terrorism and Political Violence, 24(2): 295-319. 169 Caspi, D., Freilich, J. & Chermak, S. (2012). Worst of the bad: Violent white supremacist groups and lethality. Dynamics of Asymmetric Conflict, 5(1): 1-17. Caywood, T. (1998). Routine activities and urban homicides. Homicide Studies, 2(1): 64-82. Chermak, S. & Gruenewald, J. (2015). Laying a foundation for the criminological examination of right-wing, left-wing and Al Qaeda-inspired extremism in the United States. Terrorism and Political Violence, 27(1): 133-159. Chermak, S., Freilich, J., Parkin, W. & Lynch, J. (2012). American terrorism and extremist crime data sources and selectivity bias: An investigation focusing on homicide events committed by far-right extremists. Journal of Quantitative Criminology, 28(1): 191-218. Clarke, R. (1992). Situational crime prevention: Successful case studies. New York: Harrow and Heston Publishers. Clarke, R. & Eck, J. (2005). Crime analysis for problem solvers. United States Department of Justice, pp. 1-150. Clarke, R. & Newman, G. (2006). Outsmarting the terrorists. Westport. Praeger Security International. Cohen, L. & Felson, M. (1979). Social change and crime rate trends: A routine activity approach. American Sociological Review, 44(4): 588-608. Cook, A. & Lounsbury, O. (2011). Assessing the risk posed by terrorist groups: Identifying motivating factors and threats. Terrorism and Political Violence, 23(5): 711-729. Cornish, D., & Clarke, R. (1986). The reasoning criminal: Rational choice perspectives on offending. New York: Springer-Verlag. Cornish, D. & Clarke, R. (2003). Opportunities, precipitators and criminal decisions: A reply to Wortley’s critique of situational crime prevention. Crime Prevention Studies, 16: 41-96. Crenshaw, M. (2000). The psychology of terrorism: An agenda for the 21st century. Political Psychology, 21(2), 405-420. DeLisi, M. (2005). Career criminals in society. Thousand Oaks: SAGE Publications. Department of Homeland Security. (2014). The 2014 quadrennial homeland security review. Drake, C. (1998). The role of ideology in terrorists’ target selection. Terrorism and Political Violence, 10(2): 53-85. 170 Drakos, K. (2010). Terrorism activity, investor sentiment and stock returns. Review of Financial Economics, 19: 128-135. Dugan, L. & Apel, R. (2003). An exploratory study of the violent victimization of women: Race/ethnicity and situational context. Criminology, 41(3): 959-980. Dugan, L. & Chenoweth, E. (2012). Moving beyond deterrence: The effectiveness of raising the expected utility of abstaining from terrorism in Israel. American Sociological Review, 77(4): 597-624. Dugan, L., LaFree, G. & Piquero, A. (2005). Testing a rational choice model of airline hijackings. Criminology, 43(4): 1031-1064. Eck, J. (1993). The threat of crime displacement. Problem Solving Quarterly, 6(3): 1-2. Eck, J. (2006). Science, values, and problem oriented policing. In Prospects and problems in an era of police innovation. New York: Cambridge University Press. Ekici, N., Ozkan, M., Celik, A., & Maxfield, M. (2008). Outsmarting terrorists in Turkey. Crime Prevention & Community Safety, 10: 126-139. Engstad, P. (1975). Environmental opportunities and the ecology of crime. In Crime in Canadian society. Toronto: Butterworth. Ezell, M. & Tanner-Smith, E. (2009). Examining the role of lifestyle and criminal history variables on the risk of homicide victimization. Homicide Studies, 13: 144-173. Farrington, D. & Burrows, J. (1993). Did shoplifting really decrease? The British Journal of Criminology, 33(1): 57-69. FBI. (2015). What we investigate. Retrieved January 24, 2015, from http://www.fbi.gov/aboutus/investigate/what_we_investigate Felson, M. (1994). Crime and everyday life: Insights and implications for society. Thousand Oaks: Pine Forge Press. Feniger, Y. & Yuchtman-Yaar, E. (2011). Risk groups in exposure to terror: The case of Israel’s citizens. Social Forces, 88: 1451-1462. Ferran, L. (2013, June 5). Nidal Hasan admitted jihadist motive, Ft. Hood victims’ attorneys say. ABC News. Retrieved from http://www.abcnews.go.com Forest, J. (2012). Kidnapping by terrorist groups, 1970-2010: Is ideological orientation relevant?. Crime & Delinquency, 58(5): 769-797. 171 Fox, J. & Zawitz, M. (2007). Homicide trends in the United States. Washington D.C. Bureau of Justice Statistics, U.S. Department of Corrections. Freilich, J. & Chermak, S. (2009). Preventing deadly encounters between law enforcement and American far-rightists. Crime Prevention Studies, 25: 141- 172. Freilich, J. & Newman, G. (2009). Reducing terrorism through situational crime prevention. New York: Criminal Justice Press. Freilich, J., Adamczyk, A., Chermak, S., Boyd, K. & Parkin, W. (2015). Investigating the applicability of macro-level criminology theory to terrorism: A county-level analysis. Journal of Quantitative Criminology, 31(3): 383-411. Freilich, J., Chermak, S., & Caspi, D. (2009). Critical events in the life trajectories of domestic extremist white supremacist groups. A case study analysis of four violent organizations. Criminology & Public Policy, 8(5): 497-530. Freilich, J., Chermak, S., & Simone, J. (2009). Surveying American state police agencies about terrorism threats, terrorism sources, and terrorism definitions. Terrorism and Political Violence, 21(3): 450-475. Freilich, J., Pichardo-Almanzar, N. & Rivera, C. (1999). How social movement organizations explicitly and implicitly promote deviant behavior: The case of the militia movement. Justice Quarterly, 16(3): 655-683. Freilich, J., Chermak, S., Belli, R., Gruenewald, J. & Parkin, W. (2014). Introducing the United States Extremist Crime Database (ECDB). Terrorism and Political Violence 26(2): 372384. Fussey, P. (2011). An economy of choice? Terrorist decision making and criminological rational choice theories reconsidered. Security Journal, 24(1): 85-99. Gottfredson, M. & Hirschi, T. (1990). A general theory of crime. Palo Alto: Stanford University Press. Gruenewald, J. (2009). A comparative examination of far-right extremist homicide events. Homicide Studies, 15: 188-203. Gruenewald, J., Chermak, S., & Freilich, J. Far-right lone wolf terrorism in the United States. Studies in Conflict & Terrorism, 36(12): 1005-1024. Gruenewald, J., Gruenewald, K. & Klein, B. (2015). Assessing the attractiveness and vulnerability of eco-terrorism targets: A situational crime prevention approach. Studies in Conflict and terrorism, 38(6): 433-455. 172 Gruenewald, J., & Pridemore, W. (2012). A comparison of ideologically-motivated homicides from the new Extremist Crime Database and homicides from the Supplementary Homicides Reports using multiple imputation by chained equations to handle missing values. Journal of Quantitative Criminology, 28: 141-162. Guerette, R., Stenius, V. & McGloin, J. (2005). Understanding offense specialization and versatility: A reapplication of the rational choice perspective. Journal of Criminal Justice, 33: 77-87. Haider-Markel, D., Joslyn, M. & Al-Baghal, M. (2006). Can we frame the terrorist threat? Issue frames, the perception of threat, and opinions on counterterrorism policies. Terrorism and Political Violence, 18(4): 545-559. Hamm, M. (2007). Terrorism as crime: From Oklahoma City to Al-Qaeda and beyond, New York: New York University Press. Harris, D., Simi, P. & Ligon, G. (2016). Reporting practices of journal articles that include interviews with extremists. Studies in Conflict & Terrorism, 39: 7-8. Hart, T. & Miethe, T. (2014). Street robbery and public bus stops: A case study of activity nodes and situational risk. Security Journal, 27(2): 180-193. Heger, L., Jung, D. & Wong, W. (2012). Organizing for resistance: How group structure impacts the character of violence. Terrorism and Political Violence, 24(5): 743-768. Hindelang, M., Gottfredson, M. & Garofalo, J. (1978). Victims of personal crime: An empirical foundation for a theory of personal victimization. Cambridge: Ballinger Publishing Co. Hirschi, T. & Gottfredson, M. (2003). Commentary: Testing the general theory of crime. Journal of Research in Crime and Delinquency, 30: 47-54. Hoffman, B. (1998). Inside terrorism. Columbia University Press. Horgan, J. (2009). Walking away from terrorism: Accounts of disengagement from radical and extremist movements. Routledge. Horgan, J. (2012). Interviewing the terrorists: Reflections on fieldwork and implications for psychological research. Behavioral Sciences of Terrorism and Political Aggression, 4(3): 195-211. Horowitz, M., & Potter, P. (2013). Allying to kill: Terrorist intergroup cooperation and the consequences of lethality. Journal of Conflict Resolution, 10: 1-27. 173 Jenkins, B., Liepman, A. & Willis, H. (2014). Identifying enemies among us: Evolving terrorist threats and continuing challenges of intelligence collection and information sharing. Rand Corporation, Volume 317. Jordan, F. (2008). Predicting target selection by terrorists: A network analysis of the 2005 London underground attacks. International Journal of Critical Infrastructures, 4(1): 206214. Kingshott, B. (2003). Terrorism: The “new” religious war. Criminal Justice Studies, 16(1): 15-27. Klein, B., Gruenewald, J. & Smith, B. (2016). Opportunity, group structure, temporal patterns, and successful outcomes of far-right terrorism incidents in the United States. Crime & Delinquency, June: 1-26. Kposowa, A. & Breault, K. (1998). Seasonal variations and homicide in the United States. Journal of Interpersonal Violence, 13: 590-604. Kydd, A. & Walter, B. (2006). The strategies of terrorism. International Security, 31(1): 49-80. LaFree, G. (2010). The Global Terrorism Database: Accomplishments and challenges. Perspectives on Terrorism, 4: 24-46. LaFree, G. & Dugan, L. (2004). How does studying terrorism compare to studying crime? In Terrorism and counter-terrorism: Criminological perspectives (pp. 53-74). New York: Elsevier. LaFree, G. & Dugan, L. (2007). Introducing the global terrorism database. Terrorism and Political Violence, 19: 181-204. LaFree, G., Yang, S. & Crenshaw, M. (2009). Trajectories of terrorism: Attack patterns of foreign groups that have targeted the United States, 1970 to 2004. Criminology and Public Policy, 8: 445-473. LaFree, G. & Dugan, L. (2012). Trends in global terrorism, 1970-2008. In Peace and Conflict. (pp. 41-64). New York: Springer. Lauritsen, J. (2001). The social ecology of violent victimization: Individual and contextual effects in the NCVS. Journal of Quantitative Criminology, 17: 3-32. Le Vine, V. (1997). On the victims of terrorism and their innocence. Terrorism and Political violence, 9(3): 55-62. Ley, D. & Cybriwsky, R. (1974). Urban graffiti as territorial markers. Annals of the Association of American Geographers, 64(4): 491-505. 174 Long, J. (1997). Regression models for categorical and limited dependent variables. Advanced quantitative techniques in the social sciences. Thousand Oaks: Sage Publications. Maguire, E. (1982). Burglary in a dwelling. London: Heinemann Educational Books. Matzah, D. (1967). Delinquency and drift. New Brunswick: Transaction Publishers. McCauley, C. (2006). Jujitsu politics: Terrorism and response to terrorism. In Collateral Damage: The Psychological Consequences of America’s War on Terrorism. West Port: Praeger. McCauley, C. (2007). Psychological issues in understanding terrorism and the response to terrorism. In The Psychology of Terrorism. London: Routledge. Meier, R. & Miethe, T. (1993). Understanding theories of criminal victimization. Research and Justice, 17: 459-499. Messner, S. & Tardiff, K. (1985). The social ecology of urban homicide: An application of the “routine activities” approach. Criminology, 23(2): 241-267. Miller, E. (2014). Terrorist attacks in the U.S. between 1970 and 2012: Data from the global terrorism database (GTD). National Consortium for the Study of Terrorism and Responses to Terrorism. Muhlhausen, D. & McNeill, J. (2011). Terror trends: 40 years’ data on international and domestic terrorism. Heritage Special Report, 1-11. Nelson, C. & Huff-Corzine, L. (1998). Strangers in the night: An application of the lifestyle routine activities approach to elderly homicide-victimization. Homicide Studies, 2: 130-159. Nemeth, S. (2010). A rational explanation of terrorist targeting. PhD (Doctor of Philosophy) thesis, University of Iowa. Nilsson, M. (2015). Foreign fighters and the radicalization of local jihad: Interview evidence from Swedish jihadists. Studies in Conflict & Terrorism, 28(5): 343-358. Ozer, M. & Akbas, H. (2011). The application of situational crime prevention to terrorism. Turkish Journal of Police Studies, 13(2): 179-194. Painter, K. & Farrington, D. (1997). Street lighting and crime: Diffusion of benefits in the stoke on-trent project. Crime Prevention Studies, 10: 71-122. Pantucci, R. (2011). A typology of lone wolves: Preliminary analysis of lone Islamist terrorists. In The International Centre for the Study of Radicalization and Political Violence. 175 Pape, R. (2003). The strategic logic of suicide terrorism. American Political Science Review, 97(3): 343-361. Parkin, W. & Freilich, J. (2015). Routine activities and right-wing extremists: An empirical comparison of the victims of ideologically- and non-ideologically- motivated homicides committed by American far-rightists. Terrorism and Political Violence, 27(1): 182-203. Parkin, W., Freilich, J. & Chermak, S. (2014). Ideological victimization: Homicides perpetrated by far-right extremists. Homicide Studies, 19(3): 211-236. Paton, S. (2013). EVIL DONE vulnerability assessment: Examining terrorist targets through situational crime prevention. (unpublished master’s thesis). Florida Atlantic University, Boca Raton, Florida. Perry, S. & Hasisi, B. Rational choice rewards and the jihadist suicide bomber. Terrorism and Political Violence, 27(1): 53-80. Post, J., Ruby, K. & Shaw, E. (2002). The radical group in context: 1. An integrated framework for the analysis of group risk for terrorism. Studies in Conflict & Terrorism, 25(2): 73-100. Phillips, B. (2015). Deadlier in the U.S.? On lone wolves, terrorist groups, and attack lethality. Terrorism and Political Violence, 29(3): 533-549. Post, J., Sprinzak, E. & Denny, L. (2003). The terrorists in their own words: Interviews with 35 incarcerated Middle Eastern terrorists. Terrorism and Political Violence, 15(1): 171-184. Poyner, B. (2005). Crime-free housing in the 21st century. London: Jill Dando Institute of Crime Science. Poyner, B. & Webb, B. (1991). Crime-free housing. Oxford: Butterworth-Architecture. Pridemore, W. & Freilich, J. (2005). A reassessment of state-level covariates of militia groups. Behavioral Sciences and the Law, 23(4): 527-546. Pridemore, W. & Freilich, J. (2007). The impact of state laws protecting abortion clinics and reproductive rights on crimes against abortion providers: Deterrence, backlash or neither? Law and Human Behavior, 31(6): 611-627. Repetto, T. (1974). Residential Crime. Cambridge: Ballinger Press. Reyns, B., Henson, B. & Fisher, B. (2012). Stalking in the twilight zone: Extent of cyberstalking victimization and offending among college students. Deviant Behavior, 33(1): 1-25. Robles, F. & McPhate, M. (2016, May 12). George Zimmerman tries to auction gun used to kill Trayvon Martin. The New York Times, Retrieved from http://www.nytimes.com 176 Rodgers, K. & Roberts, G. (1995). Women’s non-spousal multiple victimization: A test of the routine activities theory. Canadian Journal of Criminology, 37(3): 363-381. Roislien, H. & Roislien, J. (2010). The logic of Palestinian terrorist target choice? Examining the Israel defense forces’ official statistics on Palestinian terrorist attacks 2000-2004. Studies in Conflict & Terrorism, 33(2): 134-148. Ruby, C. (2002). The definition of terrorism. Analysis of Social Issues and Public Policy, 2(1): 9-14. Sageman, M. (2014). The stagnation in terrorism research. Terrorism and Political Violence, 26(4): 565-580. Sampson, S., Piquero, N., & Paternoster, R. (2002). Rationality and corporate offending decisions. In Rational Choice and Criminal Behavior: Recent Research and Future Challenges (pp. 25-39). New York: Routledge. Sandler, T. & Enders, W. (2004). An economic perspective on transnational terrorism. European Journal of Political Economy, 20(2): 301-316. Sandler, T. & Lapan, H. (1988). The calculus of dissent: An analysis of terrorists’ choice of targets. Synthese, 76(2): 245-261. Sandler, T. & Siqueira, K. (2006). Global terrorism: Deterrence versus preemption. Canadian Journal of Economics, 50(4): 1370-1387. Santifort, C., Sandler, T. & Brandt, P. (2013). Terrorist attack and target diversity: Changepoints and their drivers. Journal of Peace Research, 50(1): 75-90. Scarr, H. (1973). Patterns of burglary. (National Institute of Law Enforcement and Criminal Justice). Washington, DC: U.S. Government Printing Office. Schbley, A. (2003). Defining religious terrorism: A cause and anthological profile. Studies in Conflict & Terrorism, 26(2): 105-134. Schmid, A. (2004). Frameworks for conceptualizing terrorism. Terrorism and Political Violence, 16(2): 197-221. Schuppe, J. (2015). From housewife to killer: The mystery of San Bernardino shooter Tashfeen Malik. NBC News, retrieved from http://www.nbcnews.com Schwartz, S., Dunkel, C. & Waterman, A. (2009). Terrorism: An identity theory perspective. Studies in Conflict & Terrorism, 32: 537-559. 177 Sedgwick, M. (2004). Al-Qaeda and the nature of religious terrorism. Terrorism and Political Violence, 16(4): 795-814. Seifert, K. & McCauley, C. (2014). Suicide bombers in Iraq, 2003-2010: Disaggregating targets can reveal insurgent motives and priorities. Terrorism and Political Violence, 26(5): 803-820. Sherman, L., Gartin, P. & Buerger, M. (1989). Hot spots of predatory crime: Routine activates and the criminology of place. Criminology, 27(1): 27-56. Silke, A. (2001). The devil you know: Continuing problems with research on terrorism. Terrorism and Political Violence, 13(4): 1-14. Silke, A. (2008). Holy Warriors: Exploring the psychological processes of jihadi radicalization. European Journal of Criminology, 5(1): 99-123. Silverman, R. & Kennedy, L. (1987). Relational distance and homicide: The role of the stranger. The Journal of Crime Law and Criminology, 78: 272-308. Smith, B. (2008). A look at terrorist behavior: How they prepare, where they strike. NIJ Journal, 260: 1-6. Smith, B. & Morgan, K. (1994). Terrorists right and left: Empirical issues in profiling American terrorists. Conflict & Terrorism, 17(4): 39-57. Smith, B., Cothren, J., Roberts, P. & Damphousse, K. (2008). Geospatial analysis of terrorist activities: The identification of spatial and temporal patterns of preparatory behavior of international and environmental terrorists. United States Department of Justice, 1-87. Smith, B., Gruenewald, J., Roberts, P. & Damphousse, K. (2015). The emergence of lone wolf terrorism: Patterns of behavior and implications for intervention. Terrorism and Counterterrorism Today, 89-110. Spaaij, R. (2010). The enigma of lone wolf terrorism: An assessment. Studies in Conflict & Terrorism, 33(9): 854-870. Sullivan, B., Chermak, S., Wilson, J. & Freilich, J. (2014). The nexus between terrorism and product counterfeiting in the United States. Global Crime: Special Issue: Transnational Organized Crime and Terrorism: Different Peas, Same Pod? 15(3-4): 357-378. Sullivan, B. (2015). Corporate-financial crime scandals: A comparative analysis of the collapses of Insull and Enron. In Barak, G. (Ed.), International Handbook of the Crimes of the Powerful. (pp.172-185). New York: Routledge. 178 Sullivan, B., Freilich, J. & Chermak, S. (2016). Financial terror: Financial crime schemes involving extremists linked to the American far-right and al Qaida and affiliated movements. In on the Criminology of Terrorism (pp. 420-432). Hoboken: Wiley Press. Taylor, M. (1988). The terrorist. London: Brassey’s. Terry, K. & Ackerman, A. (2008). Child sexual abuse in the Catholic church. Criminal Justice and Behavior, 35(5): 643-657. Tillyer, M., & Eck, J. (2011). Getting a handle on crime: A further extension of routine activities theory. Security Journal, 24: 179-193. Turk, A. (1982). Social dynamics of terrorism. The Annals of the American Academy of Political and Social Science, 463: 119-128. van Zuijdewijn, R. & Bakker, E. (2016). Analyzing personal characteristics of lone-actor terrorists: Research findings and recommendations. Perspectives on Terrorism, 10(2): 1-8. Waller, I. & Okihiro, N. (1978). Burglary: The victim and the public. Centre of Criminology, University of Toronto by University of Toronto Press. Walsh, D. (1978). Shoplifting: Controlling a major crime. Palgrave. Weinberg, L., Pedahzur, A. & Hirsch-Hoefler, S. (2004). The challenges of conceptualizing terrorism. Terrorism and Political Science, 16(4): 777-794. Weisburd, D., Bushway, S., Lum, C. & Yang, S. (2004). Trajectories of crime at places: A longitudinal study of street segments in the city of Seattle. Criminology, 42(2): 283-322. Welsh, B. & Farrington, D. (2009). Public area CCTV and crime prevention: A systematic review and meta-analysis. Justice Quarterly, 26(4): 716-745. Wilcox, P., Madensen, T & Tillyer, M. (2007). Guardianship in context: Implications for burglary victimization risk and prevention. Criminology, 45(4): 771-803. Willis, H. (2007). Guiding resource allocations based on terrorism risk. Risk Analysis, 27(3): 597-606. Willison, R. & Siponen, M. (2009). Overcoming the insider: Reducing employee computer crime through situational crime prevention. Communications of the ACM, 52(9): 133-137. Wright, A. (2013). Terrorism, ideology and target selection. (Unpublished) Princeton University. 179