University of Southern California Dissertations and Theses

Using decision and risk analysis to assist in policy making about terrorism

(USC Thesis Other)

Using decision and risk analysis to assist in policy making about terrorism

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content USING DECISION AND RISK ANALYSIS TO ASSIST IN POLICY MAKING ABOUT TERRORISM by Heather Beth Rosoff A Dissertation Presented to the FACULTY OF THE GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree DOCTOR OF PHILOSOPHY (POLICY, PLANNING AND DEVELOPMENT) August 2009 Copyright 2009 Heather Beth Rosoff ii TABLE OF CONTENTS Acknowledgements v List of Tables vi List of Figures viii Abstract x Chapter 1: Introduction 1 1. The Terrorism Threat 1 2. Why Use Risk Analysis to Study Terrorism 6 2.1. Practical Analytic Tool 8 2.2. Quality of Decision Making 8 2.3. Government Accountability 9 3. Applications of Terrorism Risk Analysis 10 3.1. Chapter 2: Terrorism Leader Motivations 11 3.2. Chapter 3: Los Angeles-Based Dirty Bomb Attack 12 Project Risk Analysis 3.3. Chapter 4: Terrorism Risk Perception 14 3.4. General Applicability of Risk Analyses 15 Chapter 2: Decision Analysis by Proxy for the Rational Terrorist 17 1. Introduction 17 1.1. Terrorism Context 19 1.2. Proxy Multiple Objectives Values Modeling 22 2. Methodology 23 2.1. Decision Maker and Context 23 2.2. Objectives Hierarchy 24 2.3. Attack Alternatives 26 2.4. Decision Tree Probability Estimates 27 2.5. Attribute Definition and Measurement 30 2.6. Risk Attitude Across Attributes 35 2.7. Value Tradeoffs Across Attributes 38 3. Results 40 3.1. Attack Alternative Ranking 40 3.2. Sensitivity Analysis 46 3.2.1. Attribute Interaction Effect 46 3.2.2. Attribute Dependencies 52 4. Conclusions 53 4.1. Model Challenges 54 4.2. Model Applications 55 iii Chapter 3: A Risk and Economic Analysis of Dirty Bomb Attacks 58 on the Ports of Los Angeles and Long Beach 1. Introduction 58 1.1. The Dirty Bomb Threat 58 2. Sources of Radioactive Material 60 2.1. Nuclear Reactor and Waste Facilities 61 2.2. Medical, Research, and Industrial Facilities 62 2.3. Foreign Sources of Radioactive Material 63 3. Scenarios and Probabilities 63 3.1. Type of Bomb Constructed 66 3.2. Delivery Modes 67 3.3. Detonation Site 67 3.4. Pruning Scenarios and Assessing Relative Likelihoods 68 3.5. Probability of Success 69 4. Consequences 74 4.1. Blast Effects and Acute Radiation 76 4.2. Health Effects Due to Airborne Releases 76 4.3. Economic Consequences 80 5. Countermeasures 82 6. Conclusions 84 Chapter 4: The Perceived Risk of Terrorism 89 1. Introduction 89 2. Overview 93 2.1. Hierarchical Linear Modeling (HLM) 93 3. Study One 96 3.1. Method 96 3.2. Results 99 3.3 Summary and Discussion of Study One 103 4. Study Two 104 4.1. Method 104 4.2. Results 106 4.3. Summary and Discussion of Study Two 113 5. Study Three 113 5.1. Method 113 5.2. Within Subjects (Level 1) Research Questions 116 5.3. Within Subjects (Level 1) Results 116 5.4. Summary and Discussion Level 1 Analysis 119 5.5. Between Subjects (Level 2) Research Questions 120 5.6. Between Subjects (Level 2) Results 120 5.6.1. Gender Subgroup 121 5.6.2. Location Subgroup 121 5.6.3. Age Subgroup 123 iv 5.6.4. Education Level Subgroup 124 5.6.5. Personal Annual Income Subgroup 125 5.7. Summary and Discussion of Level 2 Analysis 127 6. Conclusions 127 6.1. Policy Implications 133 Chapter 5: Conclusion 140 1. Identifying Policy Problems 141 2. Eliminating Assessment Bias and Uncertainty 143 3. Quantifying Policy Problems 146 4. Closing Remarks 146 Alphabetized Bibliography 149 Appendix A: Risk Perception Study One 157 Appendix B: Risk Perception Study Two 163 Appendix C: Risk Perception Study Three 170 v DEDICATION Though my name is the only one to appear on the cover of this dissertation, a great many people have contributed to its development. I owe my gratitude to all those who have made this dissertation possible and because of whom my graduate experience has been one that I will cherish forever. My deepest gratitude goes to my doctoral committee, consisting of Drs. Detlof von Winterfeldt, Elisabeth Graddy and Richard John. I would like to extend a special thanks to Richard for the knowledge he shared and the countless hours he spent working with me. I was consistently amazed by how he was able to encourage my research curiosity, while simultaneously offering me the guidance I needed to stay on course. Richard provided me with the comfort to ask whatever questions I felt necessary (no matter how ridiculous they seemed), helped me sort out the technical details of my projects, and turned me into the quasi-mathematical psychologist that I am today. Lastly, I am thankful to my family and friends, particularly my mom and Ryan, who spent countless hours listening to my rants about the complexities of the dissertation development process. I am pretty sure I wouldn’t have made it without all their loving support and encouragement. vi LIST OF TABLES Table 2.1: Attribute Units and Scales Defined for One Proxy 30 Table 2.2: Attribute Units and Scales Across Four Proxies 31 Table 2.3: Impact of “No Attack” on Attributes by Alternative’s 34 Score Matrix Table 2.4: Proxy Swing Weight Assignments 38 Table 2.5: Normalized Swing Weight Assessments 40 Table 2.6: Attack Alternatives Expected Utilities by Proxy 43 Table 2.7: Probability of Attack 45 Table 3.1: Sources of Radioactive Material 61 Table 3.2: Transportation and Location Scenarios 69 Table 3.3: Vulnerable Tasks of the Medium Scenario 73 Table 3.4: Ranges of Consequence Estimates 75 Table 4.1: Survey Questions and Rating Scales in Sequential Order 97 Table 4.2: Correlation Coefficients for Mean Likelihood Estimates 100 by Attack and Risk Type Table 4.3: Correlation Coefficients for Mean Fatality Estimates 102 by Attack and Risk Type Table 4.4: HLM Level 1 Risk Relationships 103 Table 4.5: Survey Questions and Rating Scales in Sequential Order 105 Table 4.6: Summary of Risk Relationships with Cognitive and 112 Emotional Indicators Table 4.7: Study Three’s Sample Demographic Data 114 Table 4.8: Survey Questions and Rating Scales in Sequential Order 115 vii Table 4.9: HLM Level 1 Cognitive Variables Risk Relationships 117 Table 4.10: HLM Level 1 Emotional Variables Risk Relationships 118 Table 4.11: Location Analysis Contrast Groups 122 Table 4.12: Age Analysis Contrast Groups 123 Table 4.13: Education Level Analysis Contrast Groups 124 Table 4.14: Annual Income Analysis Contrast Groups 126 viii LIST OF FIGURES Figure 2.1: Terrorist Leader Fundamental Objectives Hierarchy 25 Figure 2.2: Attack Alternatives 26 Figure 2.3: Attack Event Tree 27 Figure 2.4: Proxy Beta Distributions Describing Uncertainty in the 29 Probability of a Terrorist Avoiding Interdiction for Two Attack Strategies Figure 2.5: Attribute “Americans Killed” Uncertainty 33 Figure 2.6: Long Term Economic Impact Certainty Equivalents for 36 Two Proxies Figure 2.7: Long Term Economic Impact Uncertainty Distributions 37 For Two Proxies from Figure 2.6 Figure 2.8: Swing Weight Uncertainty Distributions 39 Figure 2.9: Utility Distributions (Risk Profile) for 42 “No Attack” and “Explosions on Mass Transportation” Attack Alternatives Figure 2.10: CDF Utility Distributions 44 Figure 2.11: Multi-attribute Risk Attitude - Complementing 49 Figure 2.12: Multi-attribute Risk Attitude - Substituting 51 Figure 2.13: Effect of Positive Correlations Upon Attribute 52 Utility Outcomes Figure 3.1: Microsoft Project Tasks – Building the Dirty Bomb 71 Figure 3.2: Microsoft Project Tasks – Transporting the Dirty Bomb 71 Figure 3.3: Schematic View of the Complete Project 72 Figure 3.4: Distribution over the Probability of a Successful Attack 74 (Medium Radioactivity Scenario) ix Figure 3.5: Hypothetical Plume from a Medium Radioactivity Scenario 78 Figure 4.1: Mean Likelihood Estimates vs. Societal Risk Ratings 99 Figure 4.2: Mean Fatalities Estimates (Log10) vs. Societal Risk Ratings 101 Figure 4.3: Mean Likelihood and Fatality (Log10) Estimates vs. Risk Rating 107 Figure 4.4: Mean Familiarity and Capability Estimates vs. Risk Rating 108 Figure 4.5: Mean Disaster Potential and Dread Estimates vs. Risk Rating 109 Figure (a): Mean Disaster Potential vs. Personal Risk Rating Figure 4.5: Mean Disaster Potential and Dread Estimates vs. Risk Rating 110 Figure (b): Mean Dread Estimates vs. Personal Risk Rating Figure 4.6: Mean Disaster Potential vs. Dread Estimates 111 x ABSTRACT Risk has been characterized as a function of a potential threat, vulnerability to the threat, and the consequences were the threat to be carried out. In the context of terrorism, threats are the individuals who might wage an attack against a specific target; vulnerabilities are the people and targets whose safety is contingent upon the effectiveness of security policies; and consequences are the possible negative outcomes from an attack. The intent of this paper is to use different risk and decision analysis techniques to assist in the policy making relative to assessing these three different components that define terrorism risk. First, a methodology is described for representing terrorist leader preferences for alternative attack strategies against the U.S. A multi-attribute utility model embedded within a @Risk simulation model was developed to characterize terrorist motivations and values. Ultimately, relative likelihood of a terrorist attack is determined as a function of the terrorists' attack utility. While the model’s outputs are mostly illustrative of the methodology used, the policy implications of the approach are considered. Next, the threat of attacks on the ports of Los Angeles and Long Beach is analyzed. Terrorists are assumed to be using a radiological dispersal device (RDD, also known as a “dirty bomb”) to shut down port operations and cause substantial economic and psychological impacts. The analysis is an exploratory investigation of a combination of several risk analysis tools, including scenario generation and pruning, project risk analysis, direct consequence modeling, and indirect economic impact assessment. The xi implications for countering a dirty bomb, including the protection of the radiological sources and intercepting an ongoing dirty bomb attack, are discussed. Lastly, a compilation of three studies were conducted to assess how individuals perceive the risks of terrorism. The psychometric paradigm is employed to evaluate the influence of various predictor variables, both cognitive and emotional, on this calculation. Results describing the findings’ policy implications on preparedness and response efforts, such as what efforts are needed to keep people educated about terrorism and how that information should be directed, are included. 1 Chapter 1: Introduction On September 11 th , 2001 (9/11) the risk associated with the threat of terrorism became a major concern to the United States (U.S.). While international terrorism had been plaguing the globe for decades, this was the first major attack to successfully infiltrate U.S. borders. 1 The 9/11 attacks reinforced the need for government and public attention to the terrorism threat. The Department of Homeland Security (DHS) was created to protect the nation, while, locally, steps have been taken to protect vulnerable targets and restore a sense of security to the people. However, these defensive measures, coupled with the fortunate lack of another attack, have resulted in ongoing uncertainty and limited understanding of terrorism. 1. THE TERRORISM THREAT Prior to 9/11, the majority of the U.S.’s emergency resources were directed toward preparing for, and responding to, natural and technology-based events. California was known for its earthquakes, Florida had a history of hurricanes, and New England prepared annually for at least one Nor’easter, to name a few. These were threats with which the country was familiar and had experienced for decades. Compared to natural and technological disasters, the origin and uncertainty about the consequences and management of terrorism is foreign to Americans. This is illustrated by the varied level of human involvement associated with the causation of 1 There was a prior successful attack against the World Trade Center in New York on February 26, 1993. The World Trade Center was damaged when a car bomb planted by Islamic terrorists exploded in an underground garage. 2 each disaster type. Natural disasters are defined as naturally occurring processes that impact people’s daily living and routines. However, human activities have little involvement in the orientation and occurrence of such events. 2 Examples of natural disasters include landslides, earthquakes, blizzards, and naturally occurring diseases. In the case of a natural disaster, hazards such as volcanic ash from an erupting volcano, falling debris during a landslide, or edifice structural damage following an earthquake pose a serious threat to human safety. Technological disasters are the risks associated with operational and engineering systems. These technology-based systems are developed, operated, and repaired by humans. 1 Unfortunately, systems such as those developed for air travel, nuclear power plants, and space shuttles lend themselves to various hazards. The exposure to radioisotopes used to run power plants, an explosion during a space shuttle launch, and airplane damage from poor weather conditions are some of the accidents that might occur. Yet, without the means to travel across the country, explore space, or produce energy, we would be lost. Terror disasters result from the use of human aggression and conflict to express political, social, and/or economic beliefs. They are the antithesis of natural disasters because they are directed completely by human activity, thought, and decision. While technological disasters are an unfortunate by product of human efforts to solve problems and avoid accidents, terrorism is driven by humans motivated to create disastrous 2 This assumption does not account for theories about how human degradation of the natural environment is triggering natural disasters. 3 outcomes. Terrorists act through, and rely upon, the manipulation of various known and potentially unknown hazardous sources (explosive devices, radioactive isotopes, biological warfare) for disaster execution. Terrorists are also sensitive to available resources and the political environment within which they operate, resulting in very strategic and adaptable attack plans. Ultimately, a terror attack is contingent upon human creation and execution of the event. 2 A second difference among the three disaster types is their level of inherent uncertainty. With any disaster, there are a myriad of unanswerable questions about the consequences of a given event, the level of risk associated with this event, and public acceptability of limited control over an event. 3 The extent of this uncertainty depends upon the likelihood of the disaster occurring, and awareness of the damage or potential damage caused by this disaster. Natural disasters tend to be easier to study because they have occurred, continue to occur, and as such, provide plenty of data for comparative analysis. The challenge lies in predicting that which is controlled by the forces of nature – the uncertainty surrounding the location, severity, and frequency of the event. 4 However, since scientists have recorded the occurrence of natural disasters for decades, much knowledge has been acquired. For example, the physical manifestation of future earthquakes is estimated by examining geological records, historical events, and conducting experiments that assess how the ground responds to seismic waves. The uncertainty of structural damage is assessed by considering the design, distribution, and degree of damage following past 4 earthquakes. The added sense of security induced by the availability of information lessens the extent of perceived disaster consequences. Technological uncertainties are similar to those of natural disasters in that they are more manageable because of knowledge derived from prior experience; however, in both instances, questions remain as to when and where such events may occur, and the extent of damage they cause. For example, there are uncertain outcomes associated with various forms of transportation – bikes, cars, space shuttles, and so forth. These are estimated by the accident rates for each type of travel, as well as information on a variety of failure modes, such as the probability of bicycle tire blowout per mile traveled, or probability of brake failure in cars. There tends to be a higher degree of control over the probability of such consequences since humans are the designers of the systems. 5 While human involvement does not ameliorate chances of an uncertain outcome, it allows for greater accountability relative to the extent of a disaster’s consequences. Terrorism disaster uncertainties are unique in that they involve human beliefs and values foreign to what is familiar to Americans. 6 Researchers’ understanding of human beings and their problems are deficient compared to their experience with natural and technological disasters. Attacks of the most recent type and size are almost impossible to anticipate because the understanding of terrorist motivations, capabilities, and intent is limited. Furthermore, uncertainties related to the planning, preparing, and execution of a terror attack include unknown variables because of the lack of available information or the inability of the observer to interpret the known information. For example, researchers find it difficult to assess the probability of a terrorist entering the U.S., as multiple 5 approaches have been attempted – some successful, others not – with many more left to be explored. Similarly, the probability of a terrorist successfully acquiring radioactive material is uncertain given the various strategies that might be employed and the effectiveness of human-run security efforts in mitigating the threat. The vast potential and unpredictability of terrorism is considerably more uncertain compared to that of natural and technological disasters. Therefore, while some terror attacks have occurred and valuable information has been collected, uncertainty still remains. The uncertainties surrounding the risk of terrorism disasters have made efforts to prepare for and respond to the threat a priority. Though a priority, the study of terrorism does not become any less challenging. In a RAND report entitled “Estimating Terrorism Risk” (2005), researchers stressed how “the data are not available to answer questions about the effectiveness of available risk reduction alternatives or to determine reasonable minimum standards for community preparedness.” 7 The number of possible attacks and degrees of severity is indeterminable. Furthermore, since few attacks have occurred on U.S. soil, ensuring that the appropriate response capabilities and services are available is in the nascent stages of development. While certain techniques and lessons learned from natural and technological incidents can be applied, emergency response is compounded by unique complications. First, it is difficult to prepare for an innumerable amount of possible attacks varying in type and severity. Second, terrorists have the ability to modify their decisions relative to what they perceive U.S. defenses to include; thus, making it difficult to track a moving and adaptable adversary. 6 The vastness of the terrorism threat has led to more questions than available answers. What constitutes terrorism? How can one analyze and measure the risk of terrorism? What steps might be taken to mitigate and manage the terrorism threat? There is a clear need for study into the risks of terrorism, and the demand will continue as long as the threat persists. Through the use of risk analysis, structure is provided to address the uncertainty/problem of terrorism. The identification and analysis of terrorism threats and vulnerabilities allows researchers to ascertain the attack consequences. These findings are used to indicate ways in which the impact of the disaster might be reduced or eliminated. 2. WHY USE RISK ANALYSIS TO STUDY TERRORISM? The literature on the subject of risk continues to grow and explore the concept’s many different meanings and applications. Below, “risk” is defined mathematically as: Risk = [Threat (T) x Vulnerability (V)] x Consequence (C) Where: (1) Threat measures the probability of what can happen. A natural disaster threat might be a tsunami; a technological disaster might be a nuclear plant explosion. (2) Vulnerability measures whether the safeguards to protect people and the environment from a threat will fail in the case of an event. (3) Consequence measures the negative impact of an event, including lives lost, economic implications and social impacts. In its simplest form, risk analysis is the process of defining and analyzing uncertainty surrounding the threats, vulnerabilities, and consequences posed by potential 7 natural and human-caused adverse events. In quantitative risk analysis, an attempt is made to numerically determine the probability of an adverse event and the extent of the consequences if the particular disaster were to take place. Initial developments in the area of risk analysis were made in the aerospace and nuclear sectors. Following a series of space shuttle accidents (1960s), a demand emerged for quantitative risk analyses to support safety goals during the design and operation phases of manned space travel. Similarly, following Eisenhower’s “Atoms for Peace” program (1950s), nuclear facilities started quantifying the effects of how design improvements might reduce the risk of accidents. Risk analysis continues to be applied successfully to engineered systems, as well as financial, environmental, health security/emergency preparedness, and others industries. The ultimate goal of risk analysis is to allow for informed decision making about an uncertainty’s threats, vulnerabilities, and consequences. Risk analysis brings structure and order to an otherwise uncertain and amorphous risky situation. The technique helps wed information and values, as well as balances (a) quantitative and qualitative information, (b) diverse priorities and perspectives, and (c) uncertainty. 3 This is an important step in formulating a true understanding of the risk, as it allows decision makers to avoid rash decisions about risk estimates. Furthermore, it provides the foundation upon which informed decisions might be based. Risk analysis can be used for pure measurement of a risk, as the input into a decision, or as the decision rule itself. Whatever the purpose, risk analysis as a technique helps to provide guidance toward the development of public policy. 8 Decision makers 8 find applications of risk analysis appealing because the technique (1) presents a practical analytic tool, (2) improves the quality of decision making, and (3) is insulated from, and held accountable by, government agencies. 2.1. Practical Analytic Tool According to Adler and Posner (2000), a primary benefit of formal tools, such as risk analysis, is that they are simply pragmatic. 9 For starters, risk analysis is problem- oriented, rather than method-driven. Method-driven techniques often result in an analyst becoming more focused on proving a position as opposed to seeking an accurate understanding of an issue. Method-driven techniques also are often the only presented tool, thus providing the analyst with an incentive to make that tool work. Conversely, risk analysis presents an alternative approach that is designed to assist decision makers in analyzing policy decisions. The tool affords clear problem definition, presents decision alternatives, and accounts for uncertainty to ensure the risk under evaluation is assessed accurately and comprehensively. Thus, analysts are able to avoid the method-driven complications. In addition, while some formal analysis techniques have fallen under criticism as not being representative of all the values critical to a problem definition (e.g. cost benefit analysis only focuses on economic efficiency), risk analysis can be defended as more of a normative approach to policy analysis. The technique allows for problems to be addressed in terms entirety; as such, account for their economic feasibility and the contributions of subjective values as needed. 2.2. Quality of Decision Making 9 Risk analysis improves decision making by eliminating some of the complexities associated with problem definition through the use of techniques better than existing beliefs and understandings. 3 Risk analysis breaks down the problem, so when presented in a decision making context it is easier to comprehend. 10 Furthermore, the explicit layout of policy problems helps uncover research anomalies not identified through traditional means. The application of risk analysis brings a consistent and objective approach to risk assessments. Coupling quantitative risk analysis with qualitative tools also improves decision making. 3 Researchers have extended the boundaries beyond once commonly used case studies, storytelling, and other forms of qualitative analyses. They have worked to bridge the gap between qualitative models and more technocratic and quantitative tools. In doing so, they have learned the value of empirically, evidence-based perspectives, and how they contribute to strengthening and validating the analysis of policy issues. 2.3. Government Accountability The usefulness of risk analysis to government problem solving is gauged by how insulated a decision process is from political influence. Transparency, stakeholder identification, and communication are all components of the risk analysis process that hinder chances of a political monopoly over decision making. Risk analysis calls for transparency of underlying assumptions and uncertainties to reinforce long term confidence in the risk assessment process. Involving stakeholders in the risk analysis process avoids public skepticism over the objectivity of the decision outcomes. Lastly, the success of the risk analysis process is contingent upon committing to an ongoing 10 dialogue among stakeholders about the processes’ assumptions, uncertainties, and other critical decision issues. The promulgation and use of risk analysis has resulted in government acceptance of the process for the evaluation of policy decisions. The technique serves as a foundation from which policy positions are developed, policy challenges are based, and the integrity of policy decisions are preserved. As such, government oversight and regulatory agencies consider concerns based on definitive facts pulled from a quantitative and qualitative risk analyses more substantiated than equity and fairness claims. 3. APPLICATIONS OF TERRORISM RISK ANALYSIS Following the attacks of 9/11, the U.S.’s immediate response was to direct resources toward protecting and preparing for ALL terrorism-related events and outcomes. Through the application of risk analysis, approaches for reducing or eliminating the terrorism threat will be developed. These findings will guide policy decisions about improving public safety, preparing efficiently for an attack, and meeting the aforementioned objectives through cost effective approaches. Traditional approaches to risk analysis are not effective in capturing terrorism risk. This is because they fail to account for its unique nature in being driven by human resourcefulness and adaptability, of which only limited knowledge is available at this time. Terrorism risk analysis certainly draws from approaches used and lessons learned from natural and technological risk assessment. 11 However, significant modifications to the models must be made to capture the uncertainty of terrorism risk in its entirety. A 11 major challenge for risk analysts is formulating an understanding of the terrorism risk itself and then identifying the optimal approach to use for its analysis. According to its mathematical definition (page 6), a risk is made up of threats, vulnerabilities, and consequences. Terrorism threats are represented by the terrorist organizations, leaders, and individuals that might wage an attack against a specific target. Terrorism vulnerabilities are the people, buildings and other targets whose safety is contingent upon the effectiveness of security (mitigation and preparedness efforts) policies. Vulnerabilities also can be measured in terms of public perception and reaction to a terror attack scenario, and how in turn such a reaction might affect social structure and continuity. Lastly, terrorism consequences are the possible negative outcomes from an attack in terms of the economic, psychological, and health implications. The subsequent three chapters introduce risk-based models designed to address one or more of the aforementioned components of the terrorism risk definition. Within each chapter, a risk-based decision model is described and assessed in terms of its policy implications. The three chapters are intended to answer the following research questions: 1. How can risk analysis be used to assess the threat of terrorism? 2. Given the results of terrorism risk analyses, what are potential policy recommendations for more effective mitigation, preparedness, and/or response programs? The remainder of this chapter provides an overview of the risk analysis techniques applied in chapters two through four. 3.1. Chapter Two: Terrorist Leader Motivations 12 This chapter describes a methodology for representing terrorist leader preferences for alternative modes of attack against the U.S. A multi-attribute utility model embedded within a @Risk simulation model was developed to characterize terrorist motivations and values. The risk model is directed from the terrorist perspective and considers how the motivations driving a terrorist shape terror risk. To analyze terrorist leaders’ attack preferences, extensive interviews were conducted to elicit information about terrorist beliefs, value trade-offs, attitude towards risk, and uncertainty about the success of particular attack strategies. Interviewed subjects are individuals knowledgeable about terrorist motivations and beliefs, and capable of providing assessments of relevant terrorist leader uncertainty, as well as uncertainty in their own knowledge about the terrorist beliefs and values (we refer to these individuals as “proxy” terrorists). Each proxy assessed 9 varied attack strategies across 11 attack attributes. The results of these interviews are four separate multi- attribute models of alternative attack modes for al Qaeda. Each multi-attribute utility model is embedded within a simulation model that generates risk profiles for each attack strategy, as well as estimated probabilities that a particular terrorist leader will select each attack strategy over a fixed time horizon. While the proxy utility model’s outputs are mostly illustrative of the methodology used, the policy implications of the approach to contributing to DHS analyses are considered. 3.2. Chapter Three: Los Angeles-Based Dirty Bomb Attack Project Risk Analysis This chapter analyzes possible terrorist attacks on the ports of Los Angeles and Long Beach using a radiological dispersal device (RDD, also known as a “dirty bomb”) 13 to shut down port operations and cause substantial economic and psychological impacts. The analysis is an exploratory investigation of a combination of several risk analysis tools, including scenario generation and pruning, project risk analysis, direct consequence modeling, and indirect economic impact assessment. The development and execution of the attack is considered from the terrorist’s perspective so as to fully address the terror risk in its entirety – the dirty bomb threat, the port vulnerabilities, and the economic, psychological, and health consequences. To analyze the dirty bomb threat, the danger of varying sources and quantities of radioactive material (measured in curies – Ci) was explored, as well as the differences in such attacks when the material originates from domestic versus international locations. Thirty six attack scenarios were examined and reduced to two plausible or likely scenarios using qualitative judgments. The attack scenarios were laid out in more detail using Microsoft Project software. This model was developed to formulate an understanding of the tasks terrorists need to perform to carry out the attacks and to determine the likelihood of the project’s success. To determine how the probability of a task’s interdiction affects an RDD’s overall attack success, each project task was assessed in terms of its complexity, the number of people involved and the time required to perform it. These figures were used as inputs for the probabilistic simulation model (@Risk by Palisades, Inc.) created to simulate the uncertainty around the overall success probability of each task. The consequences of a successful RDD attack are described in terms of a radiological plume model and resulting human health and economic impacts. In addition, 14 the policy implications of countering a dirty bomb attack, including the protection of the original radiological sources and intercepting an ongoing dirty bomb attack, are discussed. 3.3. Chapter Four: Terrorism Risk Perception This chapter explores how risk defined by vulnerabilities depends not just upon the susceptibility of targets, but also upon an individual’s perceived risk of a terror threat. While the public’s perception of terrorism by no means diminishes a disaster’s existence, their interpretation certainly affects their reaction and response in the face of an attack. A compilation of three studies assesses how the public perceives the risks associated with terrorist events. The psychometric paradigm is employed to evaluate the influence of various predictor variables on this calculation. The studies were designed to answer the question of how subjects perceive the risk of terrorism to themselves and their family, and to society at large. All three studies start with basic questions asking subjects to assess disasters based on the perceived risk, probability of disaster occurrence in the next year, and the expected number of fatalities from a given event. Each subsequent study was expanded to include additional affective indicators to test the influence of emotive values upon risk predictions. Survey respondents evaluate the risk of various natural, manmade, and terror-related disasters. The data were analyzed using hierarchical linear modeling (HLM) to assess relationships among the indicator variables. When the sample population was large enough, additional analyses were conducted to test how subgroup variables (such as gender) moderate the relationships among primary indicators. 15 Results indicate whether additional efforts are needed to keep people educated about terrorism. A complete discussion of the findings’ policy implications on preparedness and response efforts are included. For example, general findings about event likelihood and dread characteristics provide insight into how to manage the public’s risk perceptions. We also found that presenting information through directed messaging to different subgroups will further contribute to policy development. 3.4. General Applicability of the Risk Analyses The overall intent of these projects is to introduce new and different ways to assess terror risk and make contributions to counter-terrorism efforts already underway. Chapter Two’s model on the risks posed by terrorist leader preferences and Chapter Three’s model on the risks of a dirty bomb are illustrations of methodologies using data estimates acquired from open source locations. The objective is to present these tools to persons with access to the information needed to successfully apply the models toward the assessment of terrorism risk. Chapter Four’s work on the perceptions of risk produces results with more direct utility because, prior to 9/11, limited research had been conducted on how Americans perceive the risk of terrorism. Given this, the studies’ benefits are two-fold: (1) they identify the cognitive and predictor variables that define terrorism risk, and (2) produce findings about what people think and this understanding can be used to help frame communications about terrorism. 16 Chapter 1 Endnotes 1 Evan, William and Mark Manion. Minding the Machines: Preventing Technological Disasters. New Jersey: Prentice Hall PTR, 2002. 2 Bandura, Albert. “Mechanisms of moral disengagement,” Origins of terrorism: Psychologies, ideologies, theologies, states of mind. Ed. IW. Reich. Cambridge: Cambridge University Press, 1990, 161-191. 3 Kammen, Daniel and David Hassenzahl. Should We Risk It? Exploring Environmental, Health, and Technological Problem Solving. Princeton, New Jersey: Princeton University Press, 1999. 4 Ammann, Walter J., Stefanie Dannenmann and Laurent Vulliet eds. Proceedings of the RISK21 Workshop, Monte Verità, Ascona, Switzerland, 28 November - 3 December 2004. London: Taylor & Francis, 2006. 5 Ursano, Robert J., Carol S. Fullerton and Ann E. Norwood eds. Terrorism and Disaster: Individual and Community Mental Health Interventions. Cambridge, UK: Cambridge University Press, 2003. 6 Smelser, Neil J. and Faith Mitchell, eds. Terrorism: Perspectives from the Behavioral and Social Sciences. Panel on Behavioral, Social, and Institutional Issues, Committee on Science and Technology for Countering Terrorism, Division of Behavioral and Social Sciences and Education. Washington, DC: National Academies Press, 2002. 7 Willis, Henry, et al. Estimating Terrorism Risk. Santa Monica: RAND Corporation, 2005, 2. 8 Haimes, Yacov Y. Risk Modeling, Assessment, and Management. New York: John Wiley & Sons, Inc., 1998. 9 Adler, Matthew D. and Eric A. Posner. “Rethinking Cost-Benefit Analysis.” Yale Law Journal. 109 (1999). 10 Shapiro, Sidney A and Christopher H Schroeder. “Beyond Cost-Benefit Analysis: A Pragmatic Reorientation.” Harvard Environmental Law Review 32.2 (2008):434-502. 11 Demuth, Julie L. A Summary to National Disasters Roundtable: Countering Terrorism: Lessons Learned from Natural and Technological Disasters. Washington D.C.: National Academies Press, February 28 – March 1, 2002. 17 Chapter 2: Decision Analysis by Proxy for the Rational Terrorist 1. INTRODUCTION Understanding the objectives and motivations that drive terrorist group behavior is critical. Insight into this aspect of terrorism assists (1) the United States (U.S.) government in making decisions about how to respond to terrorist violence, (2) the U.S. public in formulating a perspective on the terrorism threat and (3) the terrorist community in distinguishing themselves from what has categorically been labeled as ‘extremist practices’. Current methods for terrorism risk assessment focus on target vulnerability, terrorist capability and resources, and attack consequence. What many researchers have yet to consider is the influence of terrorist group values and beliefs on deciphering the root cause of their militant behavior. This understanding has the potential to contribute to probabilistic estimates of terrorism threats. In this paper, we introduce an approach to representing the influence of various terrorist motivations and capabilities on the selection of potential attack strategies in the U.S. This approach is analogous to representing the various subtasks in a terror attack through a project management framework in order to identify the time required for a particular attack and the probability of its success (See Chapter 2). Using a value- focused decision framework that we refer to as “proxy utility modeling,” we assess how the values and beliefs of terrorist leaders might influence the selection of an attack strategy. We then use a random utility modeling approach to compare the risk profiles 18 of alternative attack strategies and estimate the relative likelihood of a terrorist leader selecting a particular attack strategy. Since we cannot collect information directly from terrorists, published writings served as a sufficient source to infer information about the beliefs and motivations of terrorist leaders. However, to fully capture terrorist leader objectives, people who have studied the general topic of terrorism as well as Islamic terrorist groups were asked to act as Al Qaeda terrorist leader value experts (proxies). These proxies included people holding positions as former intelligence specialists, policy analysts and researchers familiar with the study of terrorism and associated events, and former residents of Middle Eastern Islamic countries familiar with the perspective and motivations of Islamic terrorist organizations. We expected some proxies to be more informed, others to disagree, and of course none are perfect. The diversity across proxies was critical for the assessment of various perspectives on how terrorist leader values and beliefs motivate the attack strategy decision process. Through the proxy utility modeling approach, we strive to answer the following research questions: 1. Does the proxy utility modeling approach capture proxies’ estimates and uncertainties about these terrorist values and beliefs? 2. Can the proxy utility modeling approach differentiate among competing terrorist attack strategies? The next section of this paper describes the use of risk assessment in the evaluation of the terrorism threat. Section three summarizes an analysis of terrorist attack 19 alternatives and describes the assessment and elicitation procedures employed for proxy utility model. Section four presents results from the proxy utility model and the implications of the findings upon the overall probability of attack. 1.1. Terrorism Context Since the attacks of September 11, 2001, researchers have been exploring the science of risk assessment as an approach to eliminating or limiting the short- and long- term risks posed by terrorism threats. 1 Over the years some of the various models that have been developed include economic analyses, applied case studies, vulnerability and capability assessments, and game-theoretic models. Several applied case studies exist assessing the risk of disruption to the electrical grid (Simonoff et al.), the effects on the U.S. economy of a seven-day shutdown of the commercial aviation system following an attack (Gordon et al.), and a dirty bomb attack upon the ports of Los Angeles and Long Beach (See Chapter 3), to name a few. These reports offer tremendous insight into the technical and resource (manpower) capabilities of the terrorist organization, the relative feasibility of carrying out and defending against an attack, and the economic, health, and psychological consequences that might ensue. To fully assess the threat of terrorism, studying potential terrorist attack targets alone may not be the most effective counter- terrorism strategy. This paper considers incorporating the probability of an attack being selected for execution into the analysis. A terrorist organization’s commitment to an attack’s execution is part of a complex decision process. Much like the Department of Homeland Security (DHS) makes decisions on national counter-terrorism policies, 2 terrorists must decide upon the 20 best attack strategy given their perceived security needs. However, the terrorist organization’s decision is unique in that it is from the perspective of the adversary. The decision to carry out an attack comes as a directive from a terrorist leader. It is the responsibility of this leader to decide what attack, if any, is representative of the organization’s best interests and should be pursued. If certain beliefs or motivations weigh heavily on a leader’s decision making process, then certain attack types may have an increased likelihood of occurring. This chapter describes an approach to modeling the decision problem of a terrorist leader. By modeling the terrorist leader mindset, additional information is acquired about the decision making relative to attack selection. Traditionally, decision analysis techniques have been used to assess decisions associated with a certain level of uncertainty. 3 This uncertainty may be attributed to the complexity of the technical conditions, political environment, or economic variables. In addition, certain variables might be more difficult to characterize because they are more qualitative and non-numeric in nature. By design, decision analyses systemically lay out the decision problem and evaluate different alternatives in order to determine the most feasible outcome. According to Keeney, the attractiveness of alternatives depends on the likelihood of each alternative’s outcome, and the preferences of the decision maker for that outcome. 2 In the study of terrorism, estimating the likelihood of alternative outcomes is complicated by the nature of the attack type being unpredictable in terms of the time and location of the event. Other various disaster situations, whether they are technologically, manmade or naturally occurring have faced similar predictive challenges. Yet, 21 researchers still attempt to characterize the probability of these events. They conduct geological studies to evaluate earthquakes, oceanographic studies to understand hurricanes and risk studies to assess the threat of industrial accidents. The study of terrorism is further complicated by the fact that it is difficult to identify terrorist leader attack preferences. Collecting data on terrorist leaders’ values and beliefs is a formidable task, given the sensitive nature of the information and limited number of public resources. Also, terrorist attacks are created and caused by human agents and thus, extremely dynamic in nature. Knowledge about the functionality of terrorist leaders, their organizations, and their capabilities, is perpetually evolving and difficult to acquire. For example, research might show that a suicide truck bombing might be the most feasible attack alternative, but it remains unclear whether the attack’s outcome meets the objectives of the terrorist leader’s values and beliefs. Alternatively, a dirty bomb attack might be desirable to a terrorist, but this is conditional upon the adversary’s success in acquiring the radioactive material. Any decision model build around terrorism will have to account for the uncertainty that the terrorist has about the alternatives, as well as the uncertainty of the analyst’s assumption relative to the terrorist’s preferences. There is a need for a systemic approach to assessing the uncertainty associated with the decision making threat posed by terrorist leaders. A greater understanding of the opponent’s objective function may give some direction as to the probabilities associated with different attack types. This article describes the construction of a value model used to decompose the decision of a terrorist leader. To accomplish this, we seek to utilize and 22 expand upon the multiple objective decision analysis approach (Keeney and Raiffa 1976). 4,5 1.2. Proxy Multiple Objectives Values Modeling The primary objective of the proxy utility model is to use proxies’ interpretation of Al Qaeda leaders’ beliefs and motivations to determine what might be their preferred attack type. The construction of the model involves six primary steps: (1) Select the fundamental objectives for values analysis, (2) Identify and define attributes for the fundamental objectives, (3) Define the single-attribute utility functions (illustrative of proxy risk attitude), (4) Define the value tradeoffs that prioritize the different objectives and attributes, (5) Specify the uncertainties of the attributes and all model parameters, (6) Use Monte Carlo simulation to obtain risk profiles (CDFs) for each alternative, and probabilities that the utility of each alternative attack strategy is its maximum. The basic frame of the proxy utility model was expanded upon to account for the significant uncertainty associated with terrorist leaders and their attack preferences. As previously noted, since we cannot directly collect information from terrorists, proxies were used for elicitation. As a result, probability distributions were assessed over attribute scale scores, utility function parameters, and trade-off (weight) parameters to address any uncertainty in terrorist prediction of future outcomes and about the proxy’s uncertainty about terrorist beliefs and uncertainty about the event tree probabilities. In addition to the uncertainty about model parameters, there are several unknowns associated with the feasibility of attack execution. Analysts incorporated an event tree 23 into the proxy utility model to account for variations in attack execution, and to decipher how, if at all, variability in success probability might impact the proxy’s preference for an attack alternative. Overall, this value model design characterizes what we think Al Qaeda leaders believe, and ultimately will provide us with insight into making better decisions about defending against the terrorism threat. 2. METHODOLOGY This section describes the development of the proxy Al Qaeda experts’ value model being used to represent the values and beliefs behind terrorist leader attack strategy selections. An introduction to the proxy decision maker is initially provided, then a set of objectives and attributes are defined, and lastly, the assessment of a utility function over these attributes is developed. 2.1. Decision Maker and Context To identify objectives critical to a decision analysis, it is important to assess information about the decision makers’ value and beliefs; in the context of this study, the decision maker is a hypothetical leader of Al Qaeda. While the actions carried out by terrorist organizations are interpreted by many to indicate that terrorist leaders are irrational published writings suggest otherwise. 6,7 Dating back to the early 19 th century, the violence tied with terrorism was rationally justified as a means to an end. 8 Terrorists pursued goals recognizing that the consequences might be grim, yet they had a practical determination. Under such pretenses, terrorists were assumed to operate as a collective unit that required a high level of organization and careful planning to succeed. 9 24 The idea of a terrorist group being rationale also translates into how terrorist leaders are often well educated 10 and use the lessons from the classroom to make logical and strategic decisions. In this framework, terrorism is perceived as an instrumental activity designed to achieve a set of goals. 11 Like any such decision, the terrorist leader evaluates a decision by looking ahead and evaluating consequences, which in this case refers to the decision to commit an act of terror and the nature of the attack selected. Much like any other major business or social development decision, a terrorist leader attempts to maximize expected returns while minimizing the expected costs in terms of lives and dollars spent. 2.2. Objectives Hierarchy The basic way to derive objectives is to start by asking individual decision makers about the meaning and reasoning behind what drives a terrorist organization to commit acts of terror. Then for each fundamental objective, discussions about mechanisms for obtaining them are pursued. To elicit this information, the proxy terrorists were first interviewed individually, and then from the individual assessments a union of the provided objectives was developed. The findings were organized into the objectives hierarchy illustrated in Figure 2.1. 25 Figure 2.1: Terrorist Leader Fundamental Objectives Hierarchy It was determined that an Al Qaeda leader’s perceived primary objectives fall into three categories: (1) Maintaining Al Qaeda’s organizational strength, (2) managing Al Qaeda’s operational expenditures, and (3) ensuring Al Qaeda has an impact upon the U.S. Further investigation into the primary objectives resulted in a compilation of attributes, or sub- objectives, that are used to evaluate and measure the aforementioned primary objectives. For example, a terrorist leader’s effort to continue the formation of training camps is an ongoing process. This is a means to maximizing recruitment, which will in turn contribute to maintaining internal organizational strength in the event of an attack. This is also a means to contributing to the fear of terrorism in the U.S., as the suggested existence of training camps implies the Al Qaeda threat is in fact a reality. Overall, the attributes identified were of a health, economic, or psychological nature. Impact on the United States Minimize U.S. support worldwide Preferred Terrorist Attack Maximize recruitment Al Qaeda Organizational Strength Maximize pop. support (sympathizers) Minimize “backlash” To Al Qaeda Maximize funding Operational Expenditures Minimize cost Minimize resources Human causalities Economic impact Instill fear ST immediate damage LT ripple effects 26 2.3. Attack Alternatives Proxies were asked to suggest attack modes (strategies) that Al Qaeda leaders would contemplate to achieve ideal attack feasibility. These suggestions were collected during open dialogue about the terrorist organization’s general objectives and operations. Analysts opted for this style of interview format to ensure that each proxy formulated his suggestions from whatever considerations, such as attack feasibility, attractiveness, and consequences, he so chose. Figure 2.2 is the compiled list of strategies. Figure 2.2: Attack Alternatives The strategy list resulted in a range of attack modes including explosive, nuclear, and biological alternatives. The “no attack” alternative was included as an obvious status quo alternative for comparison. Some proxies expressed doubt that Al Qaeda leaders were actively preparing to execute attacks at this time. They posited that the organization might be directing their resources to internal development or toward the execution of attacks outside of the U.S. The attack strategies deliberately were broadly defined to Attack Alternatives No attack (baseline) IED* in the engine room of naval vessel Explosion resulting in dam failure MANPADS** attack on an airplane Portable nuclear bomb in a major city Explosion n mass transport(s) Release of anthrax (movie or sports arena) Detonation of a dirty bomb Smallpox release in a major city *Improvised explosive device **Man Portable Air Defense Systems 27 allow each proxy to flexibly interpret the attacks in terms of size, frequency, and location. The intent of using attack strategies rather than specific attack scenarios was to fully capture the range of possible attack alternatives Al Qaeda leadership might be considering. 2.4. Decision Tree Probability Estimates A critical component to modeling an attack strategy’s desirability is an Al Qaeda leader’s uncertainty about successfully executing each of the attack types. Research has shown that a terrorist attack operates much like any other complex business project, starting with an attack planning phase, followed by the actual preparations for the attack, and culminating with attack execution. 12 Figure 2.3 is a simple event tree illustrating these three critical phases; event nodes to the right of an arc are conditional on the preceding events (to the left). Figure 2.3: Attack Event Tree Success Failure Success Failure Success Failure PREPARING PLANNING EXECUTION P M 1-P M P I 1-P I P S 1-P S Successful Attack Interdiction? Trigger Event Material Acquisition? Failed Attack Failed Attack Failed Attack 28 In the planning phase, the acquisition of material is instrumental to ensuring that the attack strategy is viable. Avoiding interdiction by anti-terror forces is an ongoing concern for the terrorist leader during the preparing phase, including such intermediary tasks as bomb building and casing of the target. Lastly, the execution phase refers to the critical final steps involved in whether or not the attack will be carried out successfully. Here emphasis is on whether the triggering device is effective or if the executioner carries out the attack. If at any point within the planning, preparing, or execution phases a task is not successfully completed, the attack is assumed terminated. Each of the planning, preparing, and execution phases is associated with a probability of detection and disruption of the attack. To determine how these probability estimates affect the overall attack success probability, the proxy terrorists estimated the probability of success of each phase for each attack type. Each probability estimate varied depending upon the difficulty of the task as perceived by the proxy. The elicitation process was two-fold. First, proxies provided a preliminary estimate of the probability of success for a given attack strategy of obtaining material, successfully avoiding interdiction, and execution of attack. Next, they considered the uncertainty in their estimates. General beta distributions were specified for each probability estimate. Proxies specified an interquartile range (25 th and 75 th percentiles on the cumulative distribution) for each probability estimate. Beta parameters were estimated using the @Risk software (Best Fit module) for each distribution using these 2 points on the cumulative beta distribution (F(x) = .25 and F(x) = .75. The beta distributions captured both the degree of overall uncertainty in the proxy estimates, as well as any skew in the 29 direction of the uncertainty. The median of the resulting probability distribution was then compared to the original probability assessed and proxies were allowed to adjust their estimates to resolve inconsistencies. For example, one expert sited that the probability of a terrorist successfully acquiring radioactive material might vary depending on the number of persons involved and the type of material acquired. Another suggested that the probability of successful execution might vary depending on whether the attack was carried out using a triggering device or a suicide bomber. Figure 2.4 depicts sample distributions of the probability of the terrorist organization avoiding interdiction while carrying out the portable nuclear bomb and improvised explosive device (IED) strategies. Figure 2.4: Proxy Beta Distributions Describing Uncertainty in the Probability of a Terrorist Avoiding Interdiction for Two Attack Strategies The IED distribution is skewed right and has a relatively low perceived probability of interdiction; on a scale of 10% - 99%, the highest probability of interdiction falls between 22% - 36%. This is quite different from the portable nuclear device distribution on the right, which has a larger range (on a scale of 40% - 95%, highest probability of interdiction falls between 60% - 80%), and is skewed in the opposite direction (left Portable Nuclear Bomb BetaGeneral(25% = .6, 75% = .804, min = .4, max = .95) X <= 0.481 3 5.0% X <= 0.9006 95.0% 0 0.5 1 1.5 2 2.5 3 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 IED BetaGeneral(25% = .218, 75% = .366, min = .1, max = .99) X <= 0.1 52 5.0% X <= 0.500 95.0% 0 0.5 1 1.5 2 2.5 3 3.5 4 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 30 skewed); suggesting a perceived higher chance of attack interdiction. While this is a sample of the 96 distributions elicited (4 proxies X 8 attack strategies X 3 events), it illustrates the range of variations possible in the interpretation of each attack strategy’s feasibility. 2.5. Attribute Definition and Measurement An important part of proxy utility modeling approach is specifying attributes that allow for comparison of the alternative attack strategies relative to the fundamental objectives defined above. Proxies provided both measures, units by which the attributes would be defined, as well as scales indicating the range over which the measures are defined. Table 2.1 shows how the attribute units and scales were defined by one proxy. When thinking like a terrorist, he found it easiest to base his units and scales off his understanding of outcomes following the attacks of September 11 th (9/11). Table 2.1: Attribute Units and Scales Defined for One Proxy ATTRIBUTE MEASURE (UNIT) SCALE Short term economic % of 9/11 0- 400% Long term economic % of 9/11 0- 500% Recruitment % of 9/11 militaristic recruitment 0- 400% Popular support (for Al Qaeda) % of 9/11 popular support 0- 170% Damage to Al Qaeda % of 9/11 organization damaged 0- 60% Instill fear in U.S. % life change (relative to 9/11) 0- 80% Worldwide U.S. support (post attack) % of 9/11 worldwide support 0- 100% Kill Americans # killed 0- 100,000 Funding % of 9/11 annual funding 0- 400% Cost Dollars (relative to 9/11) 0- $450,000 Resources # of people required (relative to 9/11) 0- 25 31 Proxies were encouraged to define the attribute units and scales as they desired (as opposed to forcing consensus among proxies). This enabled proxies to be comfortable with their choices and justification of input. Table 2.2 demonstrates the variation in measure units and scales across all four proxies for the short term economic cost and cost attributes. Table 2.2: Attribute Units and Scales Across Four Proxies To account for proxies uncertainty about terrorist leader preferences and the analysts’ interpretation of the proxy’s perspective, proxies provided uncertainty distributions as seen in Figure 2.5 over the impact of the attributes relative to each attack type. Uncertainty scales were defined by each proxy using general beta distributions on the entire matrix of 9 attack strategies by 11 attributes. For each of the 99 cells of the matrix, each proxy provided an estimate of the score obtained for the particular attack strategy on the particular scale. ATTRIBUTE MEASURE (UNIT) SCALE Proxy 1 Short term economic % of $400 billion 0-100% Proxy 2 Short term economic % of 9/11 0-400% Proxy 3 Short term economic % of 9/11 0-190% Proxy 4 Short term economic % of $1 trillion 0-100% ATTRIBUTE MEASURE (UNIT) SCALE Proxy 1 Cost dollars 0- $200,000 Proxy 2 Cost dollars (relative to 9/11) 0- $450,000 Proxy 3 Cost dollars (relative to 9/11) 0- $135,000 Proxy 4 Cost dollars (over 10 years) 0- 50 million 32 The initial assessment was conducted assuming a successful attack, but modified assessments were also obtained for the three unsuccessful end nodes of the event tree described previously in Figure 2.3. As with the attack event tree in Section 2.4, the uncertainty distribution elicitation process was two-fold. First, the proxies’ estimates were assessed as a median, such that the proxy indicated a 50-50 chance that the true score was above or below the estimate. Then, a beta distribution was obtained by assessing both a range (minimum and maximum possible scores) and an interquartile range (25 th and 75 th percentiles on the cumulative distribution). The Best fit module of @Risk was again used to calculate parameters of the beta distribution consistent with these four estimates. The median of the beta distribution fit to these estimates was then compared to the median obtained in the original estimate, and the proxy was allowed to make adjustments to work out discrepancies. The proxies were provided with several fractiles of the resulting distributions, including more extreme percentiles (5 th and 95 th ), in addition to the values corresponding to the matching points (25 th , 50 th , and 75 th ). The resulting beta distribution is intended to quantify uncertainty about how well a given measure meets what is perceived. In Figure 2.5, the distributions reflect proxy uncertainty in the percent change in “the the number of Americans killed” caused by a mass transportation explosion and a dirty bomb. 33 Figure 2.5: Attribute “Americans Killed” Uncertainty Both the ranges and distributions for the two attack strategies are quite different. The dirty bomb is perceived to be associated with a greater number of expected casualties because the maximum estimate of number killed is 15,000, compared to a mass transportation explosion impact estimated at 1,200 fatalities. Furthermore, the distribution for the mass transportation explosion is left skewed, suggesting that the number of casualties is perceived to fall on the higher end of that 1,200 fatality range. Conversely, the dirty bomb attack distribution is right skewed. While there is greater uncertainty surrounding the number of potential deaths associated with a dirty bomb, the distribution suggests that the majority of deaths will fall on the lower end of 15,000 fatality range. These finding are somewhat expected, given the familiarity with the impact and consequences of a mass transportation explosion, compared to the less predictable and known nature of a dirty bomb. By having proxies characterize the attributes using different scales, ranges and uncertainty distributions, the uniqueness of the model as a tool for capturing perceived Al Qaeda leader motivations and beliefs is best preserved. Also, analysts were unable to BetaGeneral(25% = 32, 75% = 85, min = 6, max = 150) X <= 1 2.5 5.0% X <= 1 21 .3 95.0% 0 0.2 0.4 0.6 0.8 1 1.2 0 20 40 60 80 100 120 140 160 Values x 10^-2 BetaGeneral(25% = 7.56, 75% = 10.83, min = 1.5, max = 12) X <= 4.75 5.0% X <= 1 1 .77 95.0% 0 0.05 0.1 0.15 0.2 0.25 0 2 4 6 8 10 12 14 Dirty Bomb Mass Transit Explosion 34 compare attribute values across proxies. However, given the nature of utility models, analysts still were able to compare relative expected utilities and probability estimates produced as the model final output. As noted above, the initial scores provided for each proxy’s attribute matrix were constructed conditional upon a successful attack. When accounting for attack feasibility, there are instances where the attack fails – whether it is during material acquisition, interdiction throughout attack development, or unsuccessful execution. For unsuccessful attack outcomes, the scores assigned to attack attributes change. If, for instance, the terrorist is unable to acquire the radioactive material for a dirty bomb, certain costs and resources still are expended in the process. However, only a fraction of the originally estimated costs and resources for the full attack is tapped. There also would be “no impact” on many attributes, resulting in a score from one of the scale endpoints. Table 2.3 displays how the utility values of these select attributes were reconfigured to account for a failed attack. Table 2.3: Impact of “No Attack” on Attributes by Alternative’s Score Matrix Failed Material Acquisition Attack Interdiction During Preparing Failed Attack Execution Minimize blowback to Al Qaeda No impact (utility = 1) % of execution utility No impact (utility = 1) Minimize U.S. worldwide support No impact (utility = 1) % of execution utility No impact (utility = 1) Attack cost % of execution utility full execution utility full execution utility Attack resources % of execution utility full execution utility full execution utility Execution utility refers to the utility value assigned to the attribute assuming a successful attack. 35 2.6. Risk Attitudes Across Attributes The first instance of variation in proxy perceptions of Al Qaeda leaders’ motivations and beliefs was exemplified through attribute definition and uncertainty about attack outcomes on the scores for attack strategies on each attribute. A second variation across proxies is found in the single attribute utility functions used to capture the terrorist leader’s attitude toward risk. While the proxies may share similar attack attributes, their perspectives toward their value are not always in agreement. Each proxy can have different attitudes toward risk and in turn, may be willing to accept different levels of risk relative to satisfying an attribute’s objective. Utility functions are the measurement tool traditionally used to capture an individual’s attitude toward risk. The direction of the utility function indicates whether an individual is risk averse, neutral, or seeking. Through the acquisition of certainty equivalents from the proxies, analysts were able to estimate the nature of their risk attitudes toward each attribute. To accomplish this, an exponential utility function was estimated for each proxy for each of the 11 attributes by assessing a certainty equivalent on each attribute for a 50-50 gamble between the worst and best outcomes for each attribute. The proxy was asked to estimate a sure outcome that would make a terrorist leader indifferent between playing the gamble and taking the sure thing. Figure 2.6 displays two proxy’s decision trees capturing the certainty equivalent for the long term economic impact attribute. 36 Figure 2.6: Long Term Economic Impact Certainty Equivalents for Two Proxies Proxy 1 evaluated long term economic impact in terms of damage relative to 9/11. Proxy 1 perceives 55% of the economic impact following 9/11 to be of equal value to the gamble between his best (195%) and worst (0%) estimates relative to 9/11. Proxy 2 evaluated long term economic impact in terms of a percentage of $2 trillion. Proxy 2 perceived a terrorist leader would value 30% of 2 trillion as much as the best-worst gamble. Interestingly, while the proxies used different units for attribute definition, their certainty equivalents represented a similar percentage of the total measure, roughly 30%. As with the definition of attribute uncertainties across proxies (see page 32), generalized beta distributions were assessed for the proxies’ uncertainty about an Al Qaeda leader’s certainty equivalent for the gamble. The same 4-point elicitation (b) Proxy 2 (a) Proxy 1 LT economic impact .50 sure thing .50 gamble Proxy Best: 195% relative to 9/11 Proxy Worst: 0% relative to 9/11 …. 55% relative to 9/11 .50 sure thing .50 gamble Proxy Best: 100% of $2 trillion Proxy Worst: 0% of $2 trillion …. 32% of $2 trillion LT economic impact 37 procedure of obtaining a minimum, maximum, and an interquartile range was used as for the score matrix. An iterative process for comparing the median of the obtained distribution to the original estimate was also used to resolve inconsistencies.) Figure 2.7 compares the risk attitudes of the two proxies from Figure 2.6 toward the long term economic impact of an attack. Figure 2.7: Long Term Economic Impact Uncertainty Distributions For Two Proxies from Figure 2.6 Proxy 1’s long term economic impact uncertainty distribution ranges from 5% - 90% relative to 9/11; suggesting the proxy was very uncertain as to at what point a terrorist leader would estimate a sure outcome. In addition, the left skewed curve indicates that the proxy believes the terrorist leader is willing to take greater risk to ensure his gains from the attribute are significant (risk seeking). Conversely, proxy 2’s distribution range, 19% - 55% of 2 trillion, falls on the lower half of the overall scale for the attribute (0% - 100% of 2 trillion). Furthermore, the distribution is right skewed, suggesting that the terrorist leader is risk averse. The distribution’s range and skew BetaGeneral(25% = .4, 75% = 0.7, min = 0.05, max = 0.9) X <= 0.208 5.0% X <= 0.834 95.0% 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 BetaGeneral(25% = 0.258, 75% = 0.373, min = 0.1925, max = 0.55) X <= 0.21 23 5.0% X <= 0.4577 95.0% 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 Proxy 2 Unit: % of 2 trillion Proxy 1 Unit: % relative to 9/11 38 imply that while uncertain about his desired sure outcome, the proxy believes the terrorist leader is willing to accept a low to moderate amount of long term economic impact. 2.7. Value Tradeoffs Across Attributes Another essential component of the proxy utility model is the proxies’ perception of how terrorist leaders assess the tradeoffs among the attack attributes. Each proxy rank ordered the attack attributes using swing weights. 13,14 They assigned one attribute for which the change (swing) from worst to best represented the largest impact for the terrorist leader in terms of the overall objective – committing a terror attack. All remaining attributes were assigned a percentage between 0 and 100% to reflect relative desirability of changing (swinging) a score from worst to best on that attribute. Table 2.4 displays a sample of a proxy’s swing weight assignments. Of greatest perceived value was the long term economic implication of an attack, while attack funding required was the least. Table 2.4: Proxy Swing Weight Assignments Generalized beta distributions were assessed over each assigned swing weight to reflect uncertainty about terrorist leaders’ relative value of the swing from worst to best. The same 4-point elicitation procedure of obtaining a minimum, maximum, and an LT economic impact 100% Kill Americans 95% ST economic impact 90% Instill fear 85% Worldwide U.S. support 70% Popular support 60% Al Qaeda compromised 55% Cost of the attack 50% Recruitment 45% Resources (people) 40% Funding 30% 39 interquartile range was used as for the tradeoff uncertainty matrix. Figure 2.8 illustrates a proxy’s uncertainty distributions for Al Qaeda popular support and attack cost. Figure 2.8: Swing Weight Uncertainty Distributions For Al Qaeda popular support, the proxy assigned a normal distribution over a moderate to high range (55% - 85%). The curve shape suggests that the proxy is relatively uncertain whether the swing in popular support would fall above or below the median. The distribution assumes that popular support has a significant impact on the overall model objective. For attack cost, the distribution range is relatively low (20% - 35%), suggesting that the attribute has a small perceived influence on the overall model objective. Furthermore, since the swing in attack costs is left skewed, there is only small likelihood that the swing from worst to best for attack cost can fall somewhere on the lower end of the range. Lastly, the swing weights were normalized to sum to 1.0. Table 2.5 shows the mean normalized swing weight assessments for all four proxies. While two of the proxies prioritized logistical feasibility, such as cost of the attack, a third emphasized BetaGeneral(25% = 73.2, 50% = 75.42, min = 65, max = 80) X <= 69.99 5.0% X <= 79.02 95.0% 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 64 66 68 70 72 74 76 78 80 82 Attack Cost Al Qaeda Popular Support BetaGeneral(25% = 87.6, 50% = 88.4, min = 80, max = 92) X <= 86.385 5.0% X <= 90.051 95.0% 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 85 86 87 88 89 90 91 92 93 40 economic impacts, and the fourth stressed an attack’s psychological impacts – the ability to instill fear. Table 2.5: Normalized Swing Weight Assessments 3. RESULTS 3.1. Attack Alternative Ranking The model includes full implementation of the event tree with 4 possible outcomes (3 failures and 1 success), as well as the additive multi-attribute utility function, utilizing exponential single attribute utility functions and trade-offs defined by swing weights. Uncertainty is captured in all proxy assessments, including event tree probabilities, the attributes by alternatives score matrix, certainty equivalents, and swing weights. A risk profile for each attack strategy is generated using Monte Carlo simulation. Expected utilities are calculated from the means of these distributions, obtained from 10,000 iterations using the @Risk software (Palisades, Inc.). ST Economic Impact LT Economic Impact Recruitment Popular Support Retaliation Instill Fear Proxy 1 .15 .16 .08 .10 .04 .12 Proxy 2 .10 .14 .07 .10 .05 .10 Proxy 3 .05 .05 .07 .08 .08 .07 Proxy 4 .15 .14 .04 .07 .08 .16 U.S. Support Kill Americans Funding Cost Resources Proxy 1 .08 .12 .08 .05 .05 Proxy 2 .09 .04 .00 .24 .07 Proxy 3 .09 .08 .07 .29 .06 Proxy 4 .11 .15 .02 .03 .05 41 The two proxies’ utility distributions in Figure 2.9 represent risk profiles for the no attack and mass transportation explosion alternatives. As illustrated by the risk profile ranges and curve shapes, there was some uncertainty around the desirability of the two alternatives to the Al Qaeda leader. The no attack utility ranges from .54 - .61, indicating the alternative is moderately desirable (on a 0 - 1 scale). Plus, the areas around the mean (.57) are equally distributed, suggesting there is limited uncertainty and variability as to the alternative’s desirability within this range. Comparatively, an Al Qaeda leader’s desirability for an attack involving a transportation explosion was considerably less. The utility ranges from .21 - .41, and the distribution weight falls to the left (right skewed), meaning it is likely that the terrorist’s desirability for this alternative falls below the mean (.34). 42 Figure 2.9: Utility Distributions (Risk Profile) for “No Attack” and “Explosions on Mass Transportation” Attack Alternatives Results in Table 2.6 rank order the attack alternatives in terms of expected utility for each proxy. Findings indicate that when accounting for the possibility of attack failures, the attack with the highest mean expected utility was no attack for three of the four proxies (all except Proxy 2). Proxy 2’s utility preference was for a smallpox attack, and this utility was only .01 greater than that for no attack. Interestingly, the utility outputs for Proxy 2 are all very similar. This suggests he might not feel that Al Qaeda leaders’ preferences for attack type vary, or alternatively he has considerable uncertainty about terrorist leaders’ preferences. Proxy 2 Proxy 1 Distribution for "Explosions on Mass Transport" M ean = 0.345861 9 X <=0.3 5% X <=0.41 95% 0 2 4 6 8 10 12 14 0.26 0.31 0.36 0.41 0.46 Distribution for 'No attack" M ean = 0.57961 57 X <=0.56 5% X <=0.59 95% 0 5 10 15 20 25 30 35 40 45 0.54 0.5575 0.575 0.5925 0.61 43 Table 2.6: Attack Alternatives Expected Utilities by Proxy Cumulative utility distributions are presented for two proxies for all nine attack strategies in Figure 2.10. Each curve represents a different attack strategy. Utility increases from left to right, so curves on the right generally reflect the more desired attack strategies than curves on the left. Attack Alternative Proxy 1 Proxy 2 Proxy 3 Proxy 4 No attack 0.18 0.49 0.58 0.41 IED 0.15 0.45 0.34 0.21 Dam Explosion 0.14 0.47 0.29 0.15 MANPADS 0.14 0.46 0.38 0.33 Portable Nuclear Device 0.06 0.44 0.10 0.14 Transportation System 0.14 0.48 0.35 0.36 Anthrax 0.16 0.45 0.32 0.37 Dirty Bomb 0.14 0.42 0.22 0.27 Smallpox 0.09 0.51 0.17 0.14 44 Figure 2.10: CDF Utility Distributions The diagrams depict the considerable variability across attack utility and the uncertainty over that utility. For both proxies, the distributions intersect, meaning there is no stochastic dominance (one attack utility distribution preferred over the others). Also, the variations in curve shape suggest there is more uncertainty associated with some attacks compared to others. For example, the greater ‘S’ formation in proxy 1’s no attack (the last curve on the far right) distribution compared to that of the smallpox attack (the first 0 0.2 0.4 0.6 0.8 1 0 0.225 0.45 0.675 0.9 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 Proxy 1 Proxy 2 45 curve on the far left) implies there is greater uncertainty over the utility of the smallpox attack. From the model output, analysts were able to determine estimated attack probabilities. These estimates were derived by sampling from the expected utility distributions. We used the risk profiles to calculate the probability that the utility for each alternative attack strategy is the maximum. Table 2.7 displays each proxy’s estimated subjective probabilities and specifies the most attractive attack strategy for each. Table 2.7: Probability of Attack Given the possibility of not acquiring necessary material, getting caught, and failing to successfully execute an attack, the most probable attack for two of the proxies was “No attack.” Of the remaining two proxies, proxy 2 was partial to a smallpox attack, and proxy 4 to a MANPADS attack. The probability estimates for Proxies 2 and 4 are considerable lower than those for Proxies 1 and 3. This suggests that Proxies 2 and 4 Attack Alternative Proxy 1 Proxy 2 Proxy 3 Proxy 4 No attack 0.69 0.13 0.94 0.10 IED 0.01 0.01 0.05 0.00 Dam Explosion 0.00 0.03 0.00 0.01 MANPADS 0.00 0.09 0.00 0.24 Portable Nuclear Device 0.03 0.12 0.01 0.09 Transportation System 0.00 0.16 0.00 0.12 Anthrax 0.08 0.05 0.00 0.22 Dirty Bomb 0.14 0.07 0.00 0.17 Smallpox 0.05 0.34 0.00 0.05 46 believe the attack strategies in this study were not strongly favored by Al Qaeda leaders. However, since an attack strategy was preferred to the “no attack” alternative, it is possible the proxies felt other attack strategies (not included in the list) might better capture a terrorist leader’s preferences. Lastly, the probability estimates across some of the attack strategies are extremely low for all four proxies, such as an IED attack and dam failure. This consensus brings to question whether these threats might now be excluded from consideration. 3.2. Sensitivity Analysis 3.2.1. Attribute Interaction Effect – Multiplicative Analysis If a comprehensive set of fundamental objectives are elicited for a model, the additive utility function is typically appropriate. 15 There are instances, however, where the additive independence assumption is not appropriate due to the presence of interactions across attributes. Under such circumstances, the multiplicative model is applied. This model expands on the additive multi-attribute objective function to account for variability in the scaling constants (as seen in equation x.1). Where an additive utility model’s scaling constants (k) sum to one, in a multiplicative model the scaling constants (k) can be positive or negative depending on the interaction type. 47 To illustrate, assume the utility u function for levels of the attributes X and Y is: u (x, y) = k x u x (x) +k y u y (y) + ku x (x)u y (y) (x.1) where u x and u y are the utility functions for attributes X and Y that are scaled from 0 to 1, k x and k y are scaling constants that represent the importance of the attributes, and k is a constant, so that k x +k y + k = 1 if an additive model or >/< 1 if a multiplicative model. (x.2) If the sum of k is greater than one, this is indicative of a complementary relationship between the two attributes. To illustrate, consider a situation where a terrorist leader might be evaluating the value of two attributes, economic impact and recruitment, relative to an attack strategy. A terrorist leader would likely prefer an attack that negatively impacts the U.S. economy, as well as enhances the appeal of joining Al Qaeda forces. However, if recruitment seems like it would be unaffected or the economic impact is perceived to be uncertain, it is likely the terrorist leader will opt for no attack. The nature of this relationship implies that the attributes are compliments. Alternatively, if the sum of k is less than one, this is indicative that the two attributes are substitutes for one another. This time consider the following scenarios evaluating the value of the number of casualties and instill fear attributes. Ideally, a terrorist leader would prefer an attack where both casualties and fear are maximized. However, it is possible that due to complications in pulling together the resources for an attack, the terrorist leader might decide that neither attribute is feasible. In a second scenario of the same attributes, it is possible the targeted population will elect to be resilient and show no fear. Alternatively, the terrorist organization might select an attack 48 that is associated with fewer fatalities and greater psychological implications, such as a dirty bomb. For illustrative purposes, assume the terrorist leader wants to carry out the attack immediately. When choosing among the scenarios, he likely would prefer the second or third because he is guaranteed to reach the desired level of one of the two attributes. In this sense, the two attributes are substitutes for one another because many casualties or a significant fear factor have the same implications upon the overall objective – a successful terrorist attack. To test whether the assumption of additive independence is violated in the proxy value model, analysts conducted sensitivity analyses on the multiplicative parameter k. A range of both positive and negative k values is explored across the eleven attributes to determine whether the model was sensitive to complementary or substitution effects. Figure 2.11 depicts the risk profiles resulting from tests a mild, moderate, and high positive k values – complementing. The graphs show that the larger the complementary relationship across attributes, the greater the value and uncertainty of the utility curves. Despite the moderate shift in the utility curves, the order of attack strategy preference (from the left with the triangle-lined curve to the far right with the thicker-lined curve) is consistent with the findings from the additive model for all four proxy Al Qaeda leaders. 49 Figure 2.11: Multi-attribute Risk Attitude - Complementing 0 0.2 0.4 0.6 0.8 1 0 0.25 0.5 0.75 1 High Complementing, K=-0.99 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 Mild Complementing, K=-0.5 0 0.2 0.4 0.6 0.8 1 0 0.225 0.45 0.675 0.9 Moderate Complementing K=-.9 50 Comparatively, Figure 2.12 shows the risk profiles for a test of mild, moderate, and high negative k values – substituting. The graphs show that as the substitution effect across attack modes increases, the value of the utility curve increases and the uncertainty around that value decreases. As with the risk profiles in Figure 2.10, despite the mild shift in the utility curves, the order of attack strategy preference remains constant with findings from the additive model for all four proxy Al Qaeda leaders. 51 Figure 2.12: Multi-attribute Risk Attitude - Substituting 0 0.2 0.4 0.6 0.8 1 0 0.15 0.3 0.45 0.6 Mild Substitution, K=2 0 0.2 0.4 0.6 0.8 1 0 0.1125 0.225 0.3375 0.45 Moderate Substitution, K=10 0 0.2 0.4 0.6 0.8 1 0 0.0625 0.125 0.1875 0.25 High Substitution, K=100 52 3.2.2. Attribute Dependencies – Correlation Analysis While an assumption of Monte Carlo simulation is that probability distributions assigned to different attributes for the same attack strategy are independent, it is possible for correlations to emerge. This might occur because the measures used to define a subset of the attributes share a common cause, and in turn produce positive correlations. Correlations are often caused by external factors that are more naturally occurring and difficult to control for in the context of the proxy utility model design. To test for dependence across attributes, correlation matrices were incorporated into each proxy’s multi-attribute utility model. Analysts reevaluated the attack strategy risk profiles by assuming low (.5), medium (.7) and high (.87) correlations across attributes. These values were selected to span the range of possible variance, as evidenced by the square of the correlations (.25, .49 and about .75, respectively). The result is the generation of new risk profiles for each attack strategy as seen in Figure 2.13. Figure 2.13: Effect of Positive Correlations Upon Attribute Utility Outcomes Findings suggest that variations in the dependencies across model attributes, no matter what the strength of the correlation, did not impact the proxy’s attack strategy utility 0 0.2 0.4 0.6 0.8 1 0 0.225 0.45 0.675 0.9 Avg. Correlation = 0.50 0 0.2 0.4 0.6 0.8 1 0 0.225 0.45 0.675 0.9 Avg. Correlation = 0.87 53 preferences. Attack strategy preferences remained constant as illustrated by the order and range of the utility curves from left to right in Figure 2.13. Variations in curve shape are consistent as well. This pattern of results was found across all four proxies. 4. CONCLUSIONS The proxy utility modeling approach introduced in this paper is comprised of several tasks. For starters, analysts worked closely with proxies to formulate an understanding of how an Al Qaeda leader’s values and beliefs influence their attack strategy preferences. Data was collected from proxies on attack strategy attributes, their risk preferences for the attributes, their value tradeoffs across attributes, and their estimates of attack strategy feasibility. Probability distributions also were placed over all model parameters to account for the terrorist’s uncertainty about the attack strategy alternatives, as well as the uncertainty of the proxy’s assumptions relative to the terrorist’s preferences. Analysts then used the information collected from proxies to estimate the threat of the assessed attack strategies, in terms of expected utility and probability estimates. All pieced together, the model evaluates how perceived values and preferences intersect with perceived alternative feasibility to produce an overall probability of attack strategy selection. Results indicate that after taking into account the possibility of not acquiring the necessary material for an attack, getting caught, or not successfully executing an attack, the alternative with the highest mean expected utility was “No attack” for three of the four proxies. The remaining proxy’s utility preference was for a smallpox attack. When the probability estimates were calculated, two of the three proxies were consistent in 54 favoring the “No attack” alternative. The proxy that favored the smallpox attack preferences also did not waiver. However, the fourth proxy’s output showed a preference switch from “No attack” to a MANPADS attack. 4.1. Model Challenges By design, the proxy value model is unable to account for changes in the objective and attribute inputs over time. Some of the variability in proxy terrorist leader values will be captured by the uncertainty distributions within the model. However, significant political, economic, and social changes will likely occur, resulting in the need to restructure and redefine some of the core inputs of the fundamental objectives hierarchy. For instance, a terror organization’s leadership may change. Whether it is because their leadership structure has evolved, or a leader is killed or captured, new leaders rise in the ranks. The new leaders’ beliefs and motivations might diverge from that of their predecessor. Alternatively, as DHS counterterrorism efforts are introduced, the terrorist leader’s attack preferences and strategies might be altered to adjust to the new environment. Whatever the reason, both terrorist organizations and DHS values and preferences are bound to change, and the associated utility model should be adjusted accordingly. This might be accomplished by reevaluating the utility model on a regular basis. However, ongoing assessments would require resources and commitment on the part of the proxies and trained analysts. To avoid some of the complications with frequent data elicitation, the development of a dynamic utility feedback model of a terrorist 55 organization that accounts for the fluctuations in terrorist leader beliefs and motivations might have greater long term returns. 4.2. Model Applications While the proxy utility model’s outputs are mostly illustrative of the methodology used, they do show how the model may help the DHS make better decisions when responding to the terrorism threat. In any decision context, it is difficult to determine how to allocate security resources when one does not understand the nature of the threat, vulnerability, and consequences. The proxy value model presents a formalized approach for understanding the influence of various terrorist motivations and capabilities on the selection of potential attack strategies. The model also provides an assessment of the terrorism threat from the perspective of the adversary, as opposed to assuming that what the U.S. government values is reflective of the terrorist’s preferences. By focusing on what terrorists’ value, policy makers can gain a new perspective and additional insight into the terrorist threat. The comparative nature of the decision model output might further help DHS address some of the challenges and complexity associated with the allocation of the department’s annual budget. Taking time to consider the uncertainties of terrorist leader objectives and the resulting impact this might have on alternative selection (whether the alternative is an attack type or target, or other decision problem) can help to organize the scope of the department’s knowledge base. By initially investing in value modeling, the remaining DHS funds can be better directed toward making our country and citizens safer from the risk of terrorism. 56 Lastly, DHS also might explore the contributions of the proxy value model to counter intelligence analyses. The model described in this paper focuses on using the evaluation of proxy terrorist leaders’ values and beliefs. Intelligence analysts present completely different perspectives and sources of data input for the model. They are trained to counter the terrorism threat (as opposed to assess it from the perspective of the terrorist), and have access to more proprietary and recent information about terrorist activities and communications. Adjustment to the model to account for different data sources can provide valuable insight and/or an interesting comparative perspective on the terrorism threat. 57 Chapter 2 Endnotes 1 Anderson, Elizabeth L. “Editorial Commentary: Assessing the Risks of Terrorism: A Special Collection of Perspectives Articles by Former Presidents of the Society for Risk Analysis.” Risk Analysis. 33.3 (2002): 401-402. 2 Keeney, Ralph . “Modeling Values for Anti-Terrorism Analysis.” Risk Analysis. 27.3 (2007): 585-596. 3 Keeney, Ralph . “Decision Analysis: An Overview.” Operations Research. 30.5 (1982): 803-838. 4 Keeney, Ralph. “Developing Objectives and Attributes.” Advances in Decision Analysis: From Foundations to Applications. Eds. Ward Edwards, Ralph F. Miles and Detlof von Winterfeldt. New York: Cambridge University Press, 2007. 104-128. 5 Keeney, Ralph, and Howard Raiffa. Decisions with Multiple Objectives: Preferences and Value Trade- Offs. Cambridge: Cambridge University Press, 1993. 6 Sprinak, Ehud. “Rational Fanatics.” Foreign Policy. 120 (2000): 66-73. 7 Victoroff, J. “The Mind of the Terrorist: A Review and Critique of Psychological Approaches.” Journal of Conflict Resolution, 49.1 (2005): 3-42. 8 Wittke, Carl. “Against the Current: The Life of Karl Heinzen (1809-80).” The American Historical Review, 50:815 (1945): 342. 9 Rosoff, Heather. “A Risk and Economic Analysis of Dirty Bomb Attacks on the Ports of Los Angeles and Long Beach.” Chapter 3. 10 Fouda, Yosri. “The laughing 9/11 bombers.” Times Online. 1 October 2006. World Wide Web, April 2009 < http://www.timesonline.co.uk/tol/news/world/article656683.ece>. 11 Enders, Walter and Todd Sandler. The Political Economy of Terrorism. Cambridge: Cambridge University Press, 2006. 12 Rosoff, Chapter 3. 13 Keeney, Ralph, Detlof von Winterfeldt, and Thomas Eppel. “Eliciting Public Values for Complex Policy Decisions.” Management Science, 36.9 (1990): 1011-1030. 14 Borcherding, Katrin., Thomas Eppel and Detlof von Winterfeldt. “Comparison of Weighting Judgments in Multiattribute Utility Measurement.” Management Science. 37.12 (1991):1603-1619. 15 Keeney, Ralph and Detlof von Winterfeldt. “Practical Value Models. Advances in Decision Analysis: From Foundations to Applications. Eds. Ward Edwards, Ralph F. Miles and Detlof von Winterfeldt. New York: Cambridge University Press, 2007. 232-252. 58 Chapter 3: A Risk and Economic Analysis of Dirty Bomb Attacks on the Ports of Los Angeles and Long Beach 1. INTRODUCTION 1.1. The Dirty Bomb Threat Since the events on September 11, 2001, the prospect of a terrorist attack using a radiological dispersal device (dirty bomb) is cited as among one the most serious terrorist threats. 1 Several recently reported incidents confirm the concerns of security officials. In June 2002, the United States (U.S.) arrested Jose Padilla for his involvement with Al Qaeda in planning a dirty bomb attack on the U.S., 2 and in January 2003, British officials found documents in the Afghan city of Herat indicating Al Qaeda successfully built a small dirty bomb as well as possessed training manuals on using the explosive device. 3 There are several reasons, why terrorists may consider dirty bombs to be an attractive weapon. Radioactive materials are relatively easy to obtain and building a dirty bomb is a fairly simple process, requiring little more than the skills required for assembling a conventional bomb. 4 Furthermore, dirty bombs can create large radioactive plumes, cause health and psychological effects, and have major economic impacts due to the need for decontaminating large areas. The primary challenge faced by terrorists is procuring the radioactive material. The International Atomic Energy Agency (IAEA) states that nearly every country has devices containing radioactive material useful for the creation of dirty bombs and questions whether security in many of these locations is adequate. 5 Significant quantities of radioactive material have been lost, stolen, or abandoned – referred to as “orphan 59 sources” – from U.S. and international facilities. According to an August 2003 General Accounting Office report, since 1998 more than 1,300 radioactive sources have become orphaned in the U.S. 6 A primary concern of U.S. and international security experts is the number of orphan sources scattered throughout the former states of the Soviet Union and the security of nuclear facilities in Pakistan, India and other developing countries. A dirty bomb consists of radioactive material packaged in conventional explosives. When detonated, the radioactive material scatters into the environment, some forming a radioactive plume, and the remaining quantity falling in clumps or large particulate matter near the location of the explosion. No nuclear-fission and/or fusion reaction takes places as in a nuclear weapon. However, a dirty bomb can result in both death and injuries from the initial blast of the conventional explosives as well as radiation sickness and cancer from exposure to the radioactive material. Furthermore, the dirty bomb is widely recognized as having psychological and long-term economic impacts that could outweigh its health consequences. More specifically, depending on the amount of radioactive material released and dispersed, the contaminated area could require complete evacuation, followed by decontamination efforts that could take months or even years. Locally, evacuations and decontamination efforts impact the economy and instill public fear about returning to the contaminated area. Nationally, this could result in dirty bomb scares, both real and hoaxes, and instigate residual repercussions throughout the economy. 60 This article presents a risk and economic analysis of a dirty bomb attack on the ports of Los Angeles and Long Beach. We attempt to answer the following three questions: 1. What are the threats and vulnerabilities of a dirty bomb attack upon the ports? 2. If a dirty bomb attack was successfully carried out at the ports, what might be the health and economic impacts? 3. Given our risk and economic analysis, what are potential policy recommendations for more effective countermeasures? The next section of this article describes the sources of radioactive material in the U.S. and abroad that could be used to construct a dirty bomb. Section 3 summarizes an analysis of 36 attack scenarios and describes a methodology and some preliminary findings for estimating the relative likelihood of a successful attack. Section 4 presents an analysis of the consequences of the most likely attack scenarios in terms of the health effects and economic impact of a port shutdown. Section 5 examines possible countermeasures and their cost effectiveness. 2. SOURCES OF RADIOACTIVE MATERIAL Millions of radioactive sources are distributed worldwide, with hundreds of thousands in varying quantities and sizes currently being used, stored and produced. In the U.S. alone, approximately 2 million licensed sealed sources are in use. 6 Among the 15 member states of the European Union, the European Commission reported that about 500,000 sealed sources have been located. 6 As seen in Table 3.1, there are multiple sources of radioactive material that pose different levels of security risk given the amount 61 of curies (unit of measurement of radioactivity) they could generate. Spent fuel rods from nuclear reactors and waste facilities, industrial and blood irradiators, and radiography equipment are among some of the primary sources that contain radioactive material. For a terrorist to build a dirty bomb, any of the radioactive material necessary for these applications could be employed. Most reports of trafficking incidents or unauthorized movement of radioactive material involve sealed sources, with a few incidents involving unsealed sources such as contaminated scrap metal. Table 3.1: Sources of Radioactive Material 2.1. Nuclear Reactor and Waste Facilities In the U.S., nuclear power and waste facilities contain millions of curies of radioactive material that is the mostly deadly in nature, but also extremely difficult to Source Radioisotope Radioactivity Levels (Curies) Spent fuel assembly Multiple sources 300,000 – 2,000,000 Industrial irradiator (sterilization and food preservation) Cobalt 0 (Co 60) Cesium 137 (Cs 137) Up to 4,000,000 Up to 3,000,000 Blood irradiator Co 60 Cs 137 2,400 – 25,000 50 – 50,000 Radiotherapy (single and multibeam) Co 60 Cs 137 4,000 – 27,000 500 – 13,500 Medical radiography Co 60 Iridium 192 (Ir192) 1,000 1 - 200 Industrial radiography Co 60 Ir 192 3 – 250 3 – 250 Calibration Co 60 Cs 137 Americium 241 20 60 10 Sources: Modified (1) Center for Nonproliferation Studies (CNS), The Four Faces of Nuclear Terrorism, 2005; (2) CNS, Commercial Radioactive Sources: Surveying the Security Risks, 2003; (3) IAEA, Categorization of Radioactive Sources, 2003; (4) personal communication with nuclear safety expert, Pacific Northwest National Laboratory, August 2004. 62 obtain and handle. Special licenses are issued by the U.S. Nuclear Regulatory Commission (NRC) to ensure the facilities are designed, constructed and operated in accordance with safety standards. In addition, security surrounding nuclear power and waste sites is extremely high. While the large inventories of radioactive material may be appealing to terrorists, such precautions present a formidable challenge to acquiring the material. However, material from the nuclear fuel cycle are less protected in some countries and may be available from “rogue” countries that are developing a nuclear power capability (see below). 2.2. Medical, Research, and Industrial Facilities The NRC also issues licenses for medical, research and industrial applications requiring radioactive material. Medical and research institutions use radioactive material in medical diagnosis, sterilization of medical equipment, radiotherapy (both internal and external), and for research in nuclear medicine. Radiotherapy, the treatment of disease with radiation, employs radioisotopes that are susceptible to security risk. 7 In contrast the material used for sterilizing equipment and medical diagnosis present a smaller security concern since they require relatively low amounts of radioactive material with short half- lives. Industrial facilities use radioactive material to operate machinery such as food irradiators, gauging devices, well-logging devices and industrial radiography systems. Irradiators pose the greatest security risk because they typically contain thousands to millions of curies. 7 Industrial radiography contains lower quantities of radioactive material, but they are placed in portable devices that present a security risk. 7 Gauging and 63 well logging devices typically contain inconsequential amounts of radioactive material. 7 While the NRC is responsible for issuing licenses and monitoring such facilities, security requirements are less stringent than those found at nuclear reactor and waste facilities. 2.3. Foreign Sources of Radioactive Material Internationally, experts are concerned about the security risk associated with spent fuel assemblies and reprocessed material abandoned, lost or poorly guarded in the former states of the Soviet Union. There are also approximately 1,000 radioisotope thermoelectric generators (RTGs) that have exhausted their design life and are in need of dismantlement. The amount of radioactivity generated by these sources can be in the millions of curies. Surplus radioactive material coupled with a large number of sites with inadequate protection present opportunities for illegal stealing, selling and trafficking. Compared to the U.S., acquiring material of this quantity in some foreign countries may be less challenging mostly because of less stringent accountability and security standards. The former Soviet Union also houses weapons-grade plutonium and highly- enriched uranium produced in excess during the Cold War. If a terrorist were to acquire plutonium or highly-enriched uranium, he or she would most likely save these materials for the construction of a nuclear weapon. However, experts have noted that of all known cases of attempted trafficking of weapons-grade nuclear materials, the total acquired material is not enough to build a single nuclear bomb. 8 3. SCENARIOS AND PROBABILITIES Ports are attractive terrorist targets because of the potential for a successful attack to result in lives lost and economic damage to local businesses, harbor operations and the 64 flow of trade worldwide. Overall, ports are major trade nodes, have complex business infrastructures and are difficult to secure due to their extensive size and accessibility by water and land. Most ports are located near major metropolitan regions that rely heavily on the resources and jobs provided by the businesses within the harbors. Also, ports are connected through several transportation modes (e.g. road, ship and rail), and often industries, businesses, and tourist attractions are close by, presenting terrorists with several options for deception and attack scenarios. In this analysis, we examined possible dirty bomb attacks on the Los Angeles and Long Beach harbors as an example. Combined, they are the fifth busiest port in the world, which handles 15.7 million 20-foot unit equivalent containers annually with a value of about $384 billion. 9,10 Dispersed across the harbors are oil refineries, business offices, storage facilities for hazardous materials and cargo, container terminals and more. Cargo is transported to the ports via land, ship, or rail, increasing the challenge of securing the region. And whether coming to the ports for work or to make a delivery, many people enter the Los Angeles and Long Beach harbors daily. Immediately surrounding the ports are parks and various roads leading to fishing wharfs and tourist attractions such as the Queen Mary and cruise line terminals. Also, in the proximity are downtown Long Beach and San Pedro. Major highways, roads, and bridges pass through or alongside the ports. The activity in the nearby metropolis and recreational areas makes a terrorist attack on the ports of significant consequence both to the local livelihood as well as to the regional and national economy. 65 To analyze the dirty bomb threat to the ports of Los Angeles and Long Beach, we explored the danger of varying sources and quantities of radioactive material (measured in curies – Ci), as well as the differences in such attacks when the material originates from domestic versus international locations. We considered three scenarios, each depicting either a small, medium or large-scale attack: 1. Low radioactivity scenario: Theft of radioactive material from a radiotherapy device in a U.S. hospital. 2. Medium radioactivity scenario: Theft of radioactive material from an industrial irradiator in a U.S. facility. 3. High radioactivity scenario: Purchase of a spent fuel assembly from a former Soviet Union nuclear power or reprocessing plant. In collaboration with a counterintelligence expert, we examined these three scenarios in more detail. In particular, we studied the motivations and capabilities of terrorists to engage in any of the three scenarios to attack the ports of Los Angeles and Long Beach and conducted a qualitative “red teaming” exercise for each. In a red teaming exercise, the scenario is played out from the perspective of the terrorists to better understand the opponents’ thinking and plans. For each source scenario, we considered four transportation modes (truck, ship, train, and plane or helicopter) and three locations (bridge, harbor-elevated, harbor-ground). We examined a total of 36 possible terrorist attack scenarios. For illustrative purposes, this article will focus primarily on the medium radioactivity scenario and its transportation and location subscenarios. We assumed that 66 a moderate quantity of radioactive material (100,000 curies) is stolen from a U.S. blood or industrial irradiator. Once the radioactive material is acquired, we assumed it would be transported to a warehouse near the port for dirty bomb construction. A separate terrorist cell, equipped with technical expertise, would be accountable for building the dirty bomb during this phase of preparation for the attack. Finally, upon construction completion, a third cell would drive the dirty bomb to the selected site and remote detonate the device at a safe distance from the explosion. 3.1. Type of Bomb Constructed The type of dirty bomb constructed can vary in sophistication depending on the quantity and type of radioactive material used and the amount of time provided to assemble the device. Furthermore, the level of the terrorist’s expertise in balancing the use of explosives with the nature and quantity of radioactive material determine the severity of the blast effect and plume formation. A successfully built dirty bomb might result in very minor consequences (dispersing a few clumps of radioactive material over a fairly small area) or significant consequences (dispersing a large fraction of radioactive material as aerosols or fine particulates into the air). Also, the time allocated for bomb construction is sensitive to the possibility of detection following material theft or black market purchase. If detected, only limited time may be provided for building the bomb. Under time constraints, the terrorists might simply use the vehicle carrying the radioactive material as the detonation device. 67 3.2. Delivery Modes Terrorists are likely to select a delivery mode that has a low probability of detection by port security, yet maximizes the potential for damage to the ports. As such, the vehicle of choice is based upon what is the ideal means of dirty bomb transport to the detonation site. The ports of Los Angeles and Long Beach are accessible by land, air and sea. A truck, car, or train might be the best mode of transport if entering the port by one of the surrounding access roads or as a package on a cargo train. With respect to arriving through the ports’ waterways, a cargo ship, or recreational boat most likely provide the most flexibility. Nearby helicopter landing pads and airports make planes and helicopters alternative modes of transport, although less likely because of additional security barriers associated with gaining access to their launch sites. In addition, the vehicle selected depends on the size and weight of the dirty bomb. A bomb’s dimensions vary based on the amount of conventional explosive and radioactive material used in construction. Typically, radioactive material tends to be easily packaged because it comes in either a powder or pellet form. However, the shielding material can be bulky and heavy. The bomb’s surface area is altered most significantly when explosives are packaged around the radioactive material. Ultimately, the bomb can be designed to fit into something as small as a suitcase or as large as a van. 3.3. Detonation Site To increase the effects of the dirty bomb, the detonation site is carefully selected based upon factors such as the ease with which it can be accessed and its compatibility with the weather conditions surrounding the ports. Detonation site access is evaluated 68 based on variables such as population density, location within or outside of the ports, and the selected mode of transport for executing the attack. Finally, weather conditions as well as wind direction and velocity are considered as they affect the size and directional flow of the radioactive plume. Overall, to a terrorist, the optimal detonation site causes damage resulting in lives lost and economic consequences. A location that is less visible and susceptible to suspicious behavior is critical to enhancing the probability of attack success. However, too few people in the surrounding vicinity, winds blowing out to sea, and a detonation site located miles from the harbors might deem the attack insignificant. 3.4. Pruning Scenarios and Assessing Relative Likelihoods An important step was to determine which combinations of radioactive source × transportation mode × delivery location were implausible, not likely, or likely, given the intent to attack the ports of Los Angeles and Long Beach. We worked with a counterintelligence expert to help make these qualitative judgments. Plausibility was judged simply by considering the logical combination of the scenario factor (amount and type of bomb, delivery mode, location of detonation). For example, we considered it implausible that a terrorist group would use a small amount of radioactive material to attack the ports. If terrorists were to obtain a small quantity of radioactive materials, they probably would plan for its release within an enclosed facility or building where the dispersal effects would have a greater impact. Another example is the combination of a truck as a delivery mode and an elevated harbor location. This would require getting the truck into the harbor, waiting for the container to be placed on a ship, and exploding it in 69 mid air. This all seems very complicated compared to simply detonating the truck either within or close to the perimeter of the harbor. Table 3.2 shows the four transportation scenarios and the three detonations site scenarios considered for the medium radioactivity source scenario. Using qualitative judgments, we were able to narrow the 12 scenarios down to two likely ones. These two transportation/location scenarios were not significantly different in judged probability or consequences, so only one was analyzed for this scenario (a similar process was conducted for the high radioactivity scenario, though different cells were judged to be likely). Because of the sensitivity of information, the plausibility and qualitative likelihood judgments of this portion of the project are not included in Table 3.2. Table 3.2: Transportation and Location Scenarios 3.5. Probabilities of Success We used Microsoft Project© to lay out the details for the selected scenarios. This software originally was created to provide businesses with a computer tool that tracks a project’s progress by task, timeline, and resources. A terrorist attack operates much like any other complex business project, starting with an attack planning phase, followed by the actual preparations for the attack, and culminating with the attack execution. TRUCK SHIP TRAIN PLANE/HELI Bridge Harbor - Ground Harbor - Elevated TRANSPORTATION LOCATION 70 Microsoft Project was used to outline planning, preparing, and execution tasks, and defined each in terms of task duration and number of resources (people) required. For example, in the medium radioactivity scenario, the project starts with tasks such as planning how and where the attack will take place, determining who will be involved in the attack scenario, and establishing a means of communication among the operatives. Next, preparations begin, which include tasks such as traveling into the U.S. and purchasing explosives for the dirty bomb. Ultimately, the planning and preparation tasks come together with the execution of the dirty bomb attack on the ports of Los Angeles and Long Beach. Each task was entered into Microsoft Project© through a table format known as a Gantt chart. Tasks were inserted chronologically and described by relevant details, such as predecessor information, task duration, and resources needed. Once the Gantt chart was completed, the tasks were grouped together to form what is termed a network diagram, also known as a PERT chart. The network diagram is a graphic layout of the entire attack scenario from start to finish. Figures 2.1 and 2.2 are snapshots taken from the medium radioactivity scenario network diagram. They illustrate the steps involved for two separate tasks, building the dirty bomb and transporting the dirty bomb into the harbors. For example, Figure 3.1 shows how building the dirty bomb involves obtaining the explosive and radioactive material prior to assembling the device. Figure 3.2 shows how detonating the dirty bomb involves transporting and dropping off the bomb at the selected site and then using a remote detonation device to generate the explosion. Figure 3.3 depicts how all the individual tasks come together to form the network diagram. For 71 security reasons, we do not provide the details of each of the boxes, but only show the overall schematic. The upper-left parallelogram represents the start of the initial planning for the dirty bomb attack. The box on the far right signifies project completion with dirty bomb detonation. Figure 3.1: Microsoft Project Tasks – Building the Dirty Bomb Figure 3.2: Microsoft Project Tasks – Transporting the Dirty Bomb Figure 3.3: Microsoft Project Network Diagram The Dirty Bomb Attack Start: 11/21/05 ID: 46 Finish: 11/21/05 Dur: 0.18 wks Comp: 0% Building the Dirty Bomb Start: 10/3/05 ID: 38 Finish: 11/18/05 Dur: 7.2 wks Comp: 0% Obtaining the explosives Start: 10/3/05 ID: 39 Finish: 10/4/05 Dur: 0.4 wks Comp: 0% Obtaining the RAD material Start: 10/10/05 ID: 41 Finish: 11/4/05 Dur: 4 wks Comp: 0% Assembling the dirty bomb Start: 11/7/05 ID: 44 Finish: 11/18/05 Dur: 2 wks Comp: 0% Drop off the dirty bomb at detonation site Start: 11/21/05 ID: 46 Finish: 11/21/05 Dur: 3 hrs Res: DB Terrorist Remote detonate from an off-site location Start: 11/21/05 ID: 47 Finish: 11/21/05 Dur: 1 hr Res: DB Terrorist Pick up dirty bomb Start: 11/21/05 ID: 45 Finish: 11/21/05 Dur: 3 hrs Res: DB Terrorist Transport the dirty bomb into the port Start: 11/21/05 ID: 44 Finish: 11/21/05 Dur: 0.15 wks Comp: 0% 72 Figure 3.3: Schematic View of the Complete Project Each of the planning, preparing, and execution tasks was associated with a probability of detection and disruption of the project. To determine how the probability of detection affects overall attack success, we collaborated with a counterintelligence analyst with whom we identified the most vulnerable tasks and assigned a probability of success to each. Table 3.3 lists some of these tasks for the medium radioactivity scenario. For example, the theft of radioactive material is clearly a very vulnerable task from the perspective of the terrorists. What ki nd of at t ack - - a D B at t ack agai nst t he U ni t ed S t at es St ar t : 10/6/03 ID : 2 Fi ni sh: 10/10/03 D ur : 1 w k Res: M ast er m i nd Where - - som ew here at t he port St ar t : 10/11/03 ID : 3 Fi ni sh: 10/1/05 D ur : 102. 4 w ks Res: M ast er m i nd, At t ack Funder , Advi sor , D B Ter r or i st When - - dat e and t i m e St ar t : 10/11/03 ID : 4 Fi ni sh: 10/1/05 D ur : 102. 4 w ks Res: Ever yone (10) H ow - - rem ot e det onat i on at t he P ort of Los A ngel es ( need som e sort of l ag f or 2 and 3 t o occur bef ore) St ar t : 10/11/03 ID : 5 Fi ni sh: 10/1/05 D ur : 102. 4 w ks Res: M ast er m i nd, Advi sor , At t ack Funder , BE 1, BE 2, D B Ter r or i st Who - - num ber of peopl e necessary t o carry out t he at t ack ( 2- 5 b4) St ar t : 10/11/03 ID : 6 Fi ni sh: 3/28/05 D ur : 75. 4 w ks Res: Ever yone (10) O bt ai n and t ransport radi oact i ve m at eri al St ar t : 10/11/04 ID : 9 Fi ni sh: 10/9/05 D ur : 51. 1 w ks Com p: 0% S cope out dri vi ng rout es ( t hi s i ncl udes obt ai ni ng a car i f needed) St ar t : 5/23/05 ID : 12 Fi ni sh: 7/6/05 D ur : 6. 5 w ks Res: Ter r or i st 1, Ter r or i st 2 O bt ai n and t ransport expl osi ve m at eri al St ar t : 4/28/04 ID : 14 Fi ni sh: 10/2/05 D ur : 73. 8 w ks Com p: 0% C ase and i dent i f y sources f or expl osi ve purchase St ar t : 4/28/04 ID : 15 Fi ni sh: 10/11/04 D ur : 24 w ks Res: Expl osi ves Assi st ant P urchase expl osi ve St ar t : 10/12/04 ID : 16 Fi ni sh: 4/6/05 D ur : 24 w ks Res: Expl osi ves Assi st ant R eci eve t rai ni ng St ar t : 10/11/04 ID : 19 Fi ni sh: 10/27/04 D ur : 2. 5 w ks Res: BE 1, BE 2 C ase and i dent i f y l ocat i on ( hi de- out ) w here t he di rt y bom b w i l l be const ruct ed St ar t : 10/27/04 ID : 20 Fi ni sh: 2/25/05 D ur : 16 w ks Res: BE 1, BE 2 S cope out dri vi ng rout e f or t ransport i ng expl osi ves and R A D St ar t : 2/25/05 ID : 21 Fi ni sh: 4/12/05 D ur : 6. 5 w ks Res: BE 1, BE 2 O bt ai n m at eri al s f or bui l di ng a di rt y bom b - - hazm at sui t s, w i res, et c. St ar t : 2/25/05 ID : 22 Fi ni sh: 4/12/05 D ur : 6. 5 w ks Res: BE 1, BE 2 O bt ai n vehi cl e ( s) f or pi cki ng up di rt y bom b m at eri al s St ar t : 10/5/05 ID : 23 Fi ni sh: 10/7/05 D ur : 3 days Res: BE 2, BE 1 B ui l d t he di rt y bom b St ar t : 10/11/04 ID : 18 Fi ni sh: 10/7/05 D ur : 51 w ks Com p: 0% O bt ai n a vehi cl e f or t ransport i ng t he di rt y bom b i nt o t he port St ar t : 10/10/05 ID : 31 Fi ni sh: 10/11/05 D ur : 2 days Res: D B Ter r or i st Travel - - how t he operat i ves w i l l get i nt o t he U ni t ed S t at es ( w e need t o det erm i ne w ho needs t o be brought i nt o t he U S ) St ar t : 10/11/03 ID : 7 Fi ni sh: 3/28/05 D ur : 75. 4 w ks Res: M ast er m i nd, Advi sor , At t ack Funder Trai ni ng St ar t : 10/11/04 ID : 25 Fi ni sh: 10/27/04 D ur : 2. 5 w ks Res: D B Ter r or i st P l an on how t o get i nt o t he port St ar t : 10/27/04 ID : 26 Fi ni sh: 5/20/05 D ur : 28 w ks Com p: 0% S cope out ent rance securi t y St ar t : 10/28/04 ID : 27 Fi ni sh: 5/20/05 D ur : 28 w ks Res: D B Ter r or i st O pt i m al det onat i on l ocat i on M i l est one D at e: Wed 10/27/04 ID : 28 D ri vi ng rout es M i l est one D at e: Wed 10/27/04 ID : 29 Wi nd pat t erns M i l est one D at e: Wed 10/27/04 ID : 30 O bt ai ni ng t he expl osi ves St ar t : 10/9/05 ID : 33 Fi ni sh: 10/10/05 D ur : 0. 4 w ks Com p: 0% Transport expl osi ves - pi cki ng up and t ransport i ng t o D B const ruct i on si t e St ar t : 10/9/05 ID : 34 Fi ni sh: 10/10/05 D ur : 2 days Res: Expl osi ves Assi st ant , BE 2 O bt ai ni ng t he R A D m at eri al St ar t : 10/10/05 ID : 35 Fi ni sh: 10/11/05 D ur : 0. 3 w ks Com p: 0% S t eal R A D St ar t : 10/10/05 ID : 36 Fi ni sh: 10/10/05 D ur : 1 day Res: Ter r or i st 1, Ter r or i st 2 Transport R A D St ar t : 10/10/05 ID : 37 Fi ni sh: 10/11/05 D ur : 1 day Res: Ter r or i st 1, Ter r or i st 2, BE 2 The B om b St ar t : 10/10/05 ID : 38 Fi ni sh: 10/10/05 D ur : 0. 2 w ks Com p: 0% B ui l di ng t he bom b St ar t : 10/10/05 ID : 39 Fi ni sh: 10/10/05 D ur : 1 day Res: BE 1, BE 2 B om b det onat i on ( t i m es t o be speci f i ed l at er) St ar t : 10/11/05 ID : 40 Fi ni sh: 10/11/05 D ur : 0. 1 w ks Com p: 0% P i ck up di rt y bom b St ar t : 10/11/05 ID : 41 Fi ni sh: 10/11/05 D ur : 1 hr Res: D B Ter r or i st C ase and i dent i f y l ocat i ons w here radi oact i ve m at eri al i s st ored St ar t : 10/28/04 ID : 10 Fi ni sh: 5/20/05 D ur : 28 w ks Res: Ter r or i st 1, Ter r or i st 2 Transport and det onat e t he di rt y bom b St ar t : 10/11/04 ID : 24 Fi ni sh: 10/11/05 D ur : 51. 5 w ks Com p: 0% O bt ai n vehi cl e f or t ransport i ng R A D f rom si t e w here st ol en and f or t aki ng t o drop of f si t e St ar t : 10/7/05 ID : 13 Fi ni sh: 10/9/05 D ur : 2 days Res: Ter r or i st 1, Ter r or i st 2 Trai ni ng on how t o handl e radi oact i ve m at eri al St ar t : 10/11/04 ID : 11 Fi ni sh: 10/27/04 D ur : 2. 5 w ks Res: Ter r or i st 2, Ter r or i st 1 R em ot e det onat e f rom an of f - si t e l ocat i on St ar t : 10/11/05 ID : 44 Fi ni sh: 10/11/05 D ur : 1 hr Res: D B Ter r or i st D rop of f t he di rt y bom b St ar t : 10/11/05 ID : 43 Fi ni sh: 10/11/05 D ur : 1 hr Res: D B Ter r or i st Transport t he di rt y bom b i nt o t he port St ar t : 10/11/05 ID : 42 Fi ni sh: 10/11/05 D ur : 1 hr Res: D B Ter r or i st The A t t ack St ar t : 10/9/05 ID : 32 Fi ni sh: 10/11/05 D ur : 0. 6 w ks Com p: 0% P repari ng St ar t : 4/28/04 ID : 8 Fi ni sh: 10/11/05 D ur : 75. 4 w ks Com p: 0% Wai t f or i nst ruct i ons on how t o t ransport expl osi ve m at eri al St ar t : 10/2/05 ID : 17 Fi ni sh: 10/2/05 D ur : 1 day Res: Expl osi ves Assi st ant P l anni ng St ar t : 10/6/03 ID : 1 Fi ni sh: 10/1/05 D ur : 103. 4 w ks Com p: 0% 73 Table 3.3: Vulnerable Tasks of the Medium Scenario The probability of success for each of these tasks depends upon the complexity of the task, the number of people involved, and the time required to perform the task. Preliminary assessments of success probabilities were made for a given estimate of the number of people involved and task duration. A logit model was used to estimate variations in these probabilities as a function of changes in the number of people and time to task completion. We then developed probability distributions over the number of people and time for each task and used a probabilistic simulation model (@Risk by Palisades, Inc.) to simulate the uncertainty around the overall success probability of each task. The research team, with the help of a counterintelligence analyst, used only publicly available, open-source data to make all assessments. The data represent very preliminary estimates and are largely illustrative of the methodology used. Refinements Tasks Travel in the U.S. – the coordinator Obtain a job at the selected facility (for steal the radioactive material) Steal radioactive material from a research hospital Transport radioactive material from the research hospital Casing of the Los Angeles Port Travel into the U.S. – attack executioners Assemble the dirty bomb Transport the dirty bomb in the Los Angeles Port Dirty Bomb detonation – first explosion Second explosion 74 of these probability estimates would require access to classified data as well as the use of established procedures for formal elicitation of probabilities from personnel currently working in counterintelligence and counterterrorism operations. An example of the results from the medium radioactivity scenario probabilistic simulation is shown in Figure 3.4. Interestingly, the probabilities of success are relatively small (less than 20-40%). This is because for the overall project to be successful, all individual tasks must be successful. As the uncertainty and risk affecting the success of the vulnerable tasks listed in Table 3.3 varies, this in turn affects the overall probability of project success. Of course, terrorists may engage in multiple, independent projects, thus increasing the probability that at least one of them succeeds. Figure 3.4: Distribution over the Probability of a Successful Attack (Medium Radioactivity Scenario) 4. Consequences The consequences of a dirty bomb attack fall into three categories: (1) immediate fatalities and injuries due to blast effects and acute radiation exposure, (2) medium- and M ean = 0.3090534 X <=0.43 95% X <=0.1 6 5% 0 1 2 3 4 5 6 0 0.15 0.3 0.45 0.6 75 long-term health effects caused by airborne dispersal of radioactive material, and (3) economic impacts resulting from shutting down port operations – including evacuations, business losses, property losses, and decontamination costs. In the medium radioactivity scenario, we assumed that 5-30% of the material contained in the bomb was released into the air as aerosols or fine particulates. This results in a plume carrying roughly 500- 30,000 Ci. The ranges of various damage estimates are shown in Table 3.4. Explanations for these ranges are provided below. We tried to be conservative on the upper end of the ranges, using information, model assumptions, and existing estimates that are at the high end. The low end of these ranges is usually self-explanatory, resulting from a failure of a successful dispersal of radioactive materials into the air. Table 3.4: Ranges of Consequence Estimates 1 1 The lower end of the health effect ranges includes cases where one or more of the following might occur: (1) unsuccessful airborne releases due to faulty construction of the dirty bomb; (2) wind direction towards populated areas; and (3) low radioactive doses (10 mrem or less) that produce no health effects. The high end of the health effects were based on (1) 20% release of the source term; (2) wind directed at populated areas; and (3) NARAC estimates of doses (up to 100 mrem) and a linear dose-response function. The ranges in Table 3.4 should be considered a preliminary illustration of the analysis’ capabilities. Consequences Medium Scenario High Scenario Measures Blast and acute radiation effects 0 – 10 0 – 50 Fatalities Latent caners 0 – 20 0 – 1,000 Fatalities Port shutdown and related business losses 0 – 200 million 30 – 100 billion Dollars Evacuation cost (plume) Negligible 10 – 100 billion Dollars Business loss (plume) Negligible 1 – 3 billion Dollars Property values (plume) Negligible 100 – 200 million Dollars Decontamination costs (plume) 10 – 100 million 10 – 100 billion Dollars 76 4.1. Blast Effects and Acute Radiation The immediate fatalities and injuries following the explosion of the dirty bomb depend on the amount of explosives used and the population density in the area near the detonation site. To explode a dirty bomb, only a limited amount of explosive material is needed and, therefore, the blast effects are limited to an area within 100 feet of the detonation point. 11 Unless the bomb is set off in a very densely populated area, the effects are likely to cause only a few fatalities and several injuries. Acute radiation sickness might occur if bystanders or emergency workers who rush to assist blast victims suffer from prolonged exposure to highly radioactive material. For example, during a 2004 dirty bomb exercise held in Long Beach, emergency workers rushed to the blast site, unaware of the radioactive material and without protective clothing. Had this been a real attack, they probably would have suffered from some level of radiation exposure, though most likely not in a range that produced acute radiation effects. Overall, the severity of radiation sickness depends on the dose and duration of exposure. For example, total body exposure of about 100 rem can result in radiation sickness, where 400 rem causes radiation sickness and death in half of the exposed individual. 22 4.2. Health Effects Due to Airborne Releases The incidence of health effects following the detonation of a dirty bomb depend largely on the source and amount of radioactive material used, and the sophistication of the detonation device. If successfully detonated, a respirable fraction of the material will be released into the air that varies from about 1% to 80% of the original source. 13 77 The remaining material will fall in clumps or larger particles within hundreds of feet of the detonation site. In addition, weather conditions, wind direction, and wind velocity exacerbate the situation as they predicate the formation of the radioactive plume. Figure 3.5 shows the medium radioactivity scenario Gaussian plume. This plume is hypothetical and not based on specific models. However, we have obtained similar plumes from the National Atmospheric Release Advisory Center (NARAC) to verify that these examples are realistic. While the following calculations were conducted with the NARAC plumes (not included), the results would be very similar when applied to the plumes shown (see reference for a summary of the plume and dose modeling capabilities of NARAC). The plume in Figure 3.5 defines an inner ellipse with more than 1 mrem exposure per hour and an outer ellipse covering an area exposed to more than 0.1 mrem/hour. NARAC model calculations for a similar plume suggest that the total four-day effective dose equivalent exceeds 1,000 mrem or 1 rem in the inner ellipse and 100 mrem in the outer ellipse. To put these numbers into perspective:  Public background radiation exposure is about 300 mrem/year  A single CAT scan (for medical diagnostic purposes) creates an exposure of 1.3 rem  Worker radiation standards are set at 5 rem/year  Radiation effects occur around 400 rem or higher. 78 Figure 3.5: Hypothetical Plume from a Medium Radioactivity Scenario While these numbers may not be comforting to those exposed to 100 mrem or more, it is clear that the health impacts will be relatively small. Initial exposure to radioactivity occurs through inhalation of contaminated material as the plume passes over an area. Typical calculations assess the amount of exposure during the first four days following the event. To get a rough first-order approximation of the four-day exposure, the analysis assumed median exposure values (500 mrem) in the outer ellipse of the plume and higher exposures in the inner ellipse (2 rem) to calculate and integrate population doses. All persons located in the area covered by the radioactive plume are susceptible to radiation exposure and contamination (both internal and external). Using this crude approximation and a linear dose-response 79 function for the population estimates provided by NARAC, results indicated there would be no more than 10 latent cancers for the medium scenario and no more than 500 latent cancers for the high scenario. All assumptions made in these calculations were conservative, so the actual latent cancers are likely to be much lower. While Figure 3.5 identifies the area in which short- and medium-term exposure to radioactive material could occur, there also might be a significant level of ground deposition resulting in long-term exposure consequences. Radiation from deposition is usually referred to as “ground shine.” The process by which deposed material is resuspended, inhaled, or gets into the food chain is complicated. Only a fraction of this radioactive material eventually is absorbed by people, thus creating the same effect as the inhalation of material transported through the plume. This process will occur continually until decontamination procedures are effective. According to the NARAC models, the ground shine contours are similar to those shown in Figure 3.5 with the outer ellipse defining areas above 100 mrem/year and the inner ellipse defining areas exceeding 1 rem/year. To get a first-order approximation of the health effects, we assumed all the ground shine would be absorbed by people living in the plume area during the first year following the attack. This assumption is clearly conservative, since only a fraction of the ground contamination would be resuspended or get into the food chain during this time. Furthermore, we assumed that workers and the public would return to the contaminated areas and not take any particular precautions. Next, we assumed decontamination would be successful within a year following the attack and that no additional ground shine occurs thereafter. Together, these assumptions 80 imply that the health effects due to ground shine are approximately the same as those due to the first four days of plume exposure. Both estimates are included in the health effect ranges shown in Table 3.4. 4.3. Economic Consequences One of the major concerns about the dirty bomb threat to the ports of Los Angeles and Long Beach is the potential for an extended shutdown of the region’s operations. While it is very hard to predict how long the ports would be inoperable following the medium radioactivity scenario, it is understood that large areas of the ports would be subjected to short-, medium- or even long-term closures because of:  Concerns of dock workers about returning to work  Concerns of shippers about delivering goods to the harbors  Extensive procedures related to decontamination activities. Several shutdown scenarios were analyzed, ranging from short (15 days) to medium (120 days) to long (one year). To assess the economic impacts, the Southern California Planning Model (SCPM) was used. This is a highly disaggregated regional input-output model of the Southern California economy that was previously used to estimate the impacts of earthquakes and other disasters in Southern California. 14 The results are shown in Table 3.4. The 15 day shutdown has a small impact (about $300 million) because most ships would simply wait out the port closures and businesses would be supplied through other ports. The 120-day and one-year shutdowns, in contrast, have significant impacts ($63 and $252 billion, respectively) because they account for the economic impacts of a delay of delivering goods as well as all ripple 81 effects throughout the nation’s economy that such long-term delays involve. This includes costs ranging from the loss of local dock worker jobs to the reduced income and possible forced closure of nationwide businesses not receiving necessary parts or retail products. Additional analysis focused on the costs associated with the evacuation of the plume area, reductions of property values, and business losses resulting from stigmatization of businesses in the contaminated region. We assumed that all residents and businesses would evacuate for one week from a plume with higher than 100 mrem activity (see Figure 3.5). In addition, property values in the plume area were estimated to drop by 25% during the first year following the attack and then recover to previous levels. 15 Finally, we assumed business activity would be reduced by 10% for the first year following the attack and then return to former levels. 15 The results in Table 3.4 show that the economic impacts of the evacuation are small. This occurs because the evacuees would likely continue their business as usual, albeit from shelters, homes of family or friends, or hotels. The cost of the (temporary) reduction in property values can be in the hundreds of millions, but not nearly of the same magnitude as the cost of shutting down the ports. The costs of business disruptions could be fairly large, certainly in the billions of dollars, but only if one assumes the majority of businesses relocate outside of the region or cease to exist. In addition to the social costs inflicted upon the contaminated region, there are extensive costs associated with decontaminating surfaces with depositions of radioactive material. More specifically, the cost of decontamination depends on the required clean- 82 up level and the cost of disposing low-level radioactive material. One study estimated extremely large costs (in the trillion dollars) even for the high radioactivity scenario plume. 16 This was based on the assumption that the clean-up standards would be those promulgated by the Environmental Protection Agency (15 mrem/year) and the cost of disposal would be similar to that imposed by the current low-level radioactive waste sites at Barnwell in North Carolina or at Envirocare in Utah. Using less stringent clean-up standards (e.g. 100/year) and disposal costs closer to those of a landfill, these cost estimates can be reduced by a factor of 1,000. Nevertheless, the clean-up costs are still in the billions (see Table 3.4). 5. COUNTERMEASURES Current efforts to counter the threat of a dirty bomb attack involve plans to check all cargo for radiological materials – both dirty bombs and actual nuclear devices. 17 For example, on June 4, 2005, Secretary Chertoff announced that the Los Angeles and Long Beach ports will be equipped with sensitive radiological detection devices in the form of portals to screen all international cargo entering the harbor. 18 This is certainly a step in the right direction, as radiation portals are very effective and relatively unobtrusive measures to detect even very low levels of radiation. 19 However, the following discussion shows that significant threats remain, even within the specific set of scenarios analyzed in this article. In addition to radiation portals at the entry and exit points of the harbor, it would also be useful to install radiation detection devices in the outer perimeter of the harbor, especially in areas where an RDD device could do damage. Furthermore, one of the 83 complicated aspects of countering terrorism is that terrorists shift their attack modes in response to our defensive actions. In the case of radiological detection devices, it seems likely that terrorists would attempt to develop attack scenarios that avoid any newly installed radiation detection devices. Thus, trucks or cars having to go through screening checkpoints would be a less likely method of attack. Instead, terrorists might opt for delivery vehicles that completely bypass detection measures. Another problem with radiological detection devices is the anticipated rate of false alarms. These devices can detect very low radioactivity levels. They have the potential to pick up radiation from many sources other than weapons-grade material or radioactive material used in dirty bombs. For example, some naturally occurring material, such as granite, gives off low levels of radioactivity that might be detected. People who recently received medical procedures involving radiography also are likely to set off alarms. It is very important to define the optimal sensitivity level of the detection devices (balancing the costs of missing a threatening device against the cost of too many false alarms). Significant research exists in this area, known as “signal detection theory” that can guide the operators of these systems to set the optimal level of sensitivity (see Reference 20 for a general introduction and Reference 21 for a specific example). When optimizing the sensitivity of the detection devices, the costs and benefits of false alarms, hits, misses, and correct rejections (using the signal detection terminology) have to be considered carefully together with the probability that a piece of cargo might contain a radiological device. The initial inspection at the radiation portal is a relatively efficient process. However, if the alarm is set off, the truck or container must go into a 84 special inspection cue. Such secondary inspections create shipment delays, require significant amounts of manpower and incur large operational costs. 22 In addition to highlighting ways of modifying current countermeasure efforts at the ports of Los Angeles and Long Beach, our research demonstrated how a terrorist attack can be interrupted at many stages. The project risk analysis identifies the attack tasks most susceptible (from the terrorists’ point of view) to disruption and thus defines the terrorists’ vulnerabilities (see Table 3.2 for an example). In the dirty bomb scenarios discussed in this article, the findings suggest that the most cost-effective solution is to prevent or interdict the purchase or theft of radiological material. Radioactive material in the U.S. is highly regulated by the NRC and thefts are difficult to carry out successfully. In our attack scenario involving theft from a research or industrial facility, we hypothesized that an employee would assist in attempting to bypass NRC barriers. As such, one implication of focusing on this phase of the attack would be the benefit associated with improving security of the facility, particularly management of employees with access to radioactive sources. Similarly, in the scenario involving theft or purchase of significant material in the former Soviet Union and other foreign countries, we recognize the importance of improving safeguards and security at these facilities. 6. CONCLUSIONS A terrorist attack using a dirty bomb in the U.S. is possible, perhaps even moderately likely, but would not kill many people. Instead, such an attack primarily would result in economic and psychological consequences. Moreover, it would not be easy to carry out a dirty bomb attack. Considering the difficulties associated with 85 obtaining and transporting radioactive material, building the dirty bomb and detonating the device successfully, our preliminary analyses suggest that the chances of a successful attempt are no better than 15-40% for the medium radioactivity scenario, less likely for the high radioactivity scenario. Of course, multiple independent attempts would increase these chances. While our probability estimates are mostly illustrative, the chances of terrorists succeeding with an attack that involves relatively low-level radioactive material from a U.S. facility are larger than their chances of succeeding with the import of a large quantity of foreign sources. This is because transporting foreign source material through a number of international ports increases susceptibility to detection. If a dirty bomb attack is successful, the consequences depend primarily on the amount of radioactive material in the detonated source term, the amount released into the air, weather conditions and the population density in the impacted region. The medium radioactivity scenario analyzed in detail suggests there would be some, but fairly limited health effects and possibly significant economic impacts. The most costly economic impact would result from a lengthy shutdown of the ports and decontamination efforts. The length of the harbor shut down would in part depend on the decision to declare access to the ports as safe. In a national emergency, standards of safety different from those promulgated by the EPA may be appropriate. For example, worker safety standards may be more appropriate than public safety standards. The same also holds true for clean-up standards. Because we don’t know how policy makers and harbor workers will react in such an emergency, we have 86 parameterized the length of the harbor shutdown, from 15 days to one year, corresponding to roughly $130 million to $100 billion in costs. The economic consequences of evacuations, property value impacts, and business losses due to stigmatization in the plume area are in the billions, but not in the tens or hundreds of billions. People and the economy are likely to respond in a resilient way. Many people would relocate for some time out of the areas with relatively high levels of radioactivity (100 mrem or more), but they would not stop working. Also, businesses may relocate and later return to their original location. Similarly, effects on property values may be severe in the short-term, but like in many other disasters, return back to normal in a year or so. Regarding countermeasures, our analysis clearly supports ongoing programs to install radiation detection technology around the harbor. In addition, the analysis raises concerns regarding the security risks associated with cargo material as it is offloaded from ships, but not yet transported through the portals, incoming containers from the U.S. mainland (by truck, small boat or air), and harbor perimeter control. Finally, the analysis suggests preventing terrorism by interdicting vulnerable activities during the planning and preparing stages of an attack scenario. Such action might include being more proactive in controlling and protecting the original sources of radioactive material. 87 Chapter 3 Endnotes 1 Blair, Bruce. “What if Terrorists Go Nuclear?” Center for Defense Information. 1 October 2001. <http://www.cdi.org/terrorism/nuclear-pr.cfm>. 2 Johnson, K., & T. Locy. “Threat of dirty bomb softened.” USA Today. 12 June 12 2002. <http://www.usatoday.com/news/nation/2002/06/11/bushothers.htm>. 3 Gardner, F. “Al Qaeda was making a dirty bomb.” BBC News Online. 31 January 31 2003. <http://news.bbc.co.uk/2/hi/uk news/2711645.stm>. 4 King, Gilbert. Dirty Bomb: Weapons of Mass Disruption. New York: Penguin Group, 2004, 25. 5 International Atomic Energy Agency. (2002). “Inadequate Control of World’s Radioactive Sources.” 25 June 2002. < http://hps.org/documents/iaeapressrelease.pdf>. 6 General Accounting Office. ”Nuclear Security: Federal and State Action Needed to Improve Security of Sealed Radioactive Sources.” 6 August 2003. < http://www.gao.gov/new.items/d03804.pdf>. 7 Ferguson, Charles. D., Tahseen Kazi, and Judith Perera. “Commercial Radioactive Sources: Surveying the Security Risks.” Monterey Institute of International Studies, Center for Nonproliferation Studies. 11 (2003). 8 Council on Foreign Relations. “Terrorism Q&A: Making a Bomb.” (2004). <http://cfrterrorism.org/weapons/making print.html>. 9 The Port of Long Beach. “About the Port—Port Statistics Index.” Accessed July 2009. <http://www.polb.com/about/facts.asp>. 10 The Port of Los Angeles. “About the Port—FAQs.” Accessed July 2009. <http://www.portoflosangeles.org/newsroom/press_kit/facts.asp>. 11 National Council on Radiation Protection and Measurements (NCRP). “Management of Terrorist Events Involving Radiological Material.” NCRP 138 (2001). 12 Medline Plus. “Medical Encyclopedia Radiation Sickness.” Accessed February 19, 2005. <http://www.nlm.nih.gov/medlineplus/ency/article/000026.htm>. 13 Harper, F. “Working Together: R&D Partnerships in Homeland Security.” Department of Homeland Security Conference on “Working Together: R&D Partnerships in Homeland Security.” Boston, Massachusetts: April 27–28, 2005. 14 Gordon, Peter, et al. “The economic impact of a terrorist attack on the twin ports of Los Angeles-Long Beach.” The Economic Impacts of Terrorist Attacks. Eds. Harry Richardson, Peter Gordon and James Moore. Northampton, Massachusetts: Edward Elgar Publishing Limited, 2005. 262-286. 15 Gordon, Peter, et al. “Los Angeles–Long Beach seaports radioactive plume economic impact.” Draft Report, Center for Risk and Economic Analysis of Terrorist Events. Los Angeles, California: University of Southern California, 2005. 88 16 Reichmuth, B. “Economic consequences of a rad/nuc attack: Cleanup standards significantly affect cost.” Department of Homeland Security Conference on “Working Together: R&D Partnerships in Homeland Security.” Boston, Massachusetts: April 27–28, 2005. 17 Martonosi, Susan, David Ortiz and Henry Willis. Evaluating the Viability of 100 Per cent Container Inspection at America’s Ports. Santa Monica: RAND Corporation, 2005. 18 Krikorian, Greg. “Port complex to get radiation detection devices to forestall terrorism.” Los Angeles Times. June 4, 2005. < http://www.latimes.com/news/local/la- mechertoff4jun04,1,7330940.story?ctrack=1&cset=tru>. 19 Magee, Richard S. The Disposition Dilemma: Controlling the Release of Solid Materials from Nuclear Regulatory Commission-Licensed Facilities. Washington, DC: National Academy Press, 2002. 20 Green, David, and John Swets. Signal Detection Theory and Psychophysics. New York: John Wiley and Sons Inc., 1966. 21 Heasler, P., and T. Woods. “Game theoretic modeling of detection and deterrence.” Department of Homeland Security Conference on “Working Together: R&D Partnerships in Homeland Security.” Boston, Massachusetts: April 27–28, 2005. 22 U.S. Government Accountability Office. (2006). “Combat Nuclear Smuggling: DHS Has Made Progress Deploying Radiation Detection Equipment at U.S. Ports-of-Entry, But Concerns Remain.” GAO-06-389 (2006). 23 Bradley, Michael. (2005). “NARAC: An Emergency Response Resource for Predicting the Atmospheric Dispersion and Assessing the Consequences of Airborne Radionuclides.” Journal of Environmental Radioactivity. 96.1-3 (2007): 116-121. 89 Chapter 4: The Perceived Risks of Terrorism 1. INTRODUCTION The events of September 11, 2001 (9/11) confronted the American public with a stark reality that a change was needed in how the country chose to face the risks of terrorism. Public attention and government resources have been directed primarily towards conventional disasters, such as floods, hurricanes, earthquakes, or other natural or man-made disasters because they occur with greater frequency. The concern posed by terror disasters has been defined as a new species of hazard. 1 Terrorist attacks are driven by human beliefs and values foreign to what is common within the American mindset. Furthermore, terrorism involves an unusual level of human involvement, as the events are created and motivated by human desire to cause extensive social disruption. These unique aspects of terrorist attacks have created a demand for different approaches to address the uncertainties associated with terror-related risk. Since 9/11, Americans have been forced to accept that terrorism is a fact of contemporary life and that it is important to take intelligent and reasonable steps to protect themselves and surroundings. Such precautionary measures are critical to maintaining public morale and sustaining economic activity. If the public were to feel that their safety and security were comprised, there lies the potential for widespread social, political, and economic disruption. Thus, part of countering the terrorism threat is formulating an understanding of what the risks entail in order to effectively communicate about them. 90 To communicate effectively with the American public, government officials and policymakers are working to formulate a better understanding of how the threat of terrorism is perceived by the public. Since there have been a limited number of terror- related attacks on United States (U.S.) soil, a,2 it is still relatively unclear how Americans perceive the likelihood and consequences of different attack types and how they might react and respond if they were to occur. The most common approach for studying risk perception is employed through the use of the psychometric paradigm. 3,4 The psychometric paradigm uses scaling and multivariate analytic techniques to produce quantitative representations of a perceived risk. The paradigm calls for people to make quantitative judgments about a perceived risk (to themselves and others). These judgments are then compared to other factors that influence the perception of that same risk. Numerous studies have demonstrated that the psychometric approach is an effective predictive tool for quantifying risk perceptions, as well as characterizing a disaster’s qualities in terms of its perceived risk. 5 Initial studies using the psychometric approach processed risk through actuarial predictors, such as a risk’s likelihood and the consequences of its occurrence. In Study One the psychometric paradigm was applied to test whether predicting risk perception through cognitive variables would hold in the evaluation of terror and non-terror related disasters. Fischhoff et al. (2005) and La Porte (2005) show that risk perception is contingent upon the actual type of disaster (natural, manmade or terror- a It is estimated that since the early 1900’s, there have been a total of 5 terrorist attacks on U.S. soil (including domestic attacks, such as Oklahoma City Bombing in 1995) and 18 attacks against Americans carried out internationally (not including the attacks in response to U.S. actions in Iraq). 91 related) under consideration. 6,7 In many of the studies conducted, researchers distinguish among disasters by characteristics of the event, such as the motives behind the event and the mechanism by which the attack was executed, rather than specific disaster types. Motives refer to terrorist’s personal motivations for carrying out an attack in the Sjöberg (2005) study, and whether the disaster motivation was intentional or accidental in Burns and Slovic (2006). When considering attack execution mechanism, Fischhoff (2005) focuses primarily on the plane crashes of 9/11, while Burns and Slovic (2006) compare explosions, infectious disease releases, and radiation releases. Study One formulates an understanding of risk by focusing on differences in perception across specified terror and non-terror related disasters. A list of natural, manmade and terror-related events was systematically developed to serve as the basis from which respondents assessed their perceived risk. Early psychometric studies were criticized for only explaining a fraction of the variance of perceived risk. Empirical results suggest that there are in fact two fundamentally different ways in which humans understand risk. 8,9 First, there is the analytical/rational mode which is associated with evaluating risk through actuarial variables. 5 Second, there is a literature that considers the interplay of emotions and risk perceptions (emotional or affective mode). 10 This mode of thinking is considered more intuitive and automatic, and is experienced more as a feeling state (without much conscious awareness). As such, “affect” refers to a person’s positive or negative feeling toward a risk. Slovic has posited that “reliance on affect and emotion is quicker, easier, 92 and a more efficient way to navigate in a complex, uncertain, and sometimes dangerous world.” 8 Study Two’s design includes both groups of variables, actuarial and emotional, in its design. Researchers have found that a hybrid model of risk perception is optimal for the assessment of risky situations with uncertain outcomes. 11,12 In a potentially dangerous situation, emotional processing serves as a signal that some action is needed to reduce a perceived risk. Cognitive processing then brings logic and rational deliberation to bear on the risk under assessment. This study continued to examine how people perceive the magnitude and likelihood of disastrous events, but also considered human reaction in the form of fear, dread, disaster potential, and personal familiarity with the disaster type. Another criticism of the psychometric model is its inability to account for individual risk perceptions. In its nascent form, the psychometric paradigm calls for researchers to base their analyses on ordinary least squares (OLS) regressions of mean risk ratings, rather than the raw/individual data. 13,14 Means data are less subject to error and more likely to produce models that fit. However, results calculated through averaging have the potential to provide an inflated sense of explanatory power. This produces findings that hide individual risk perception differences and generate patterns that may not be representative of any individual or of a majority of individuals in a sample. As such, Study Three’s focus was on understanding how individuals perceive the risk of terrorism. The study was designed to use a heterogeneous sample to systematically investigate general risk relationships and to assess whether perceptions of 93 risk vary by subgroups – e.g. gender, age, education level, personal annual income, and geographic location. The survey tool applies the same predictor variables (actuarial and emotional) and assesses the perception of risk in terms of terror and non-terror related events as in Studies One and Two. In order to conduct the subgroup analysis, this survey’s distribution was expanded to a national sample of roughly 1,000 people. Section two of this paper provides an overview of the research objectives and the statistical technique applied for the data analysis. Sections three through five describe the methodology and results of the three risk perception studies. Lastly section six describes conclusions generated by the findings and examines the policy implications of the research. 2. OVERVIEW This paper encompasses three separate studies designed and implemented to evaluate risk perception as it pertains to terrorism. Study One applies the traditional psychometric model design to garner fundamental insight into people’s risk perspectives. With each additional study, the analysis was expanded to explore more developed experimental designs. Study Two tests additional risk factors, and Study Three tests how perceptions vary among population subgroups across a national sample. 2.1. Hierarchical Linear Modeling (HLM) All three studies employed hierarchical linear modeling (HLM) as a statistical technique for assessing the predictors of risk perception. For Studies One and Two, 94 HLM was used in addition to risk analysis of mean responses, and in Study Three as the primary analytic approach. HLM is a statistical procedure used to analyze data collected within groups. 15 The term hierarchical refers to the existence of relationships between variables with a hierarchical or nested structure. HLM allows for the regression of at least one lower level variable on higher level variables. For example, in the study of education, HLM could be used to analyze data about subjects’ math scores as determined by their class attendance. The subjects are nested within schools, so the model allows for the investigation of the relationship between math scores and class attendance by school-level factors (such as gender and socio-economic status). In this context, “within subjects” unit of analysis Level 1) describes the variables that might predict a subject’s math score. The Level 1 variables are treated as nested within the groups of school-level factors referred to as “between subjects” (Level 2) units of analysis. For the first two studies, Level 1 HLM analyses consisted of regression models estimating each survey respondent’s perception of event risk (as seen in Equation 2.1). The subjects’ responses for each of the terror and non-terror events were regressed on cognitive and/or emotional factors. The resulting equation captures the individuals’ responses. 95 To illustrate, assume the y ij is perceived event risk represented by the relationship with the random beta (ß) coefficients: Level 1 y i = ß 0i + ß 1i (emotional) + ß 2i (cognitive) + …. r i (2.1) where y is the overall perceived event risk for the i individuals appearing in each as defined by ß 0i , the coefficient representing the mean of y, ß 1i and ß 2i , the emotional and cognitive coefficients representing the relationship between y and ß’s, and r i representing the error associated with each observation. For Study Three, Level 1 and Level 2 regression models were run to estimate whether subgroups affect the relationship between Level 1’s predictor variables and the perception of event risk (as seen in Equation 2.2 and 2.3). As in Studies One and Two, the resulting Level 1 equation captures the individuals’ responses: Level 1 y ij = ß 0j + ß 1j (emotional) + ß 2j (cognitive) + …. r ij (2.2) where y ij is the overall perceived event risk for group j (across the i individuals appearing in each group) as defined by ß 0j, the coefficient representing the mean of y, ß 1j and B 2j , the emotional and cognitive coefficients representing the relationship between y and ß’s, and r ij representing the error associated with each observation. The intercept and slopes from the within subjects (Level 1) regression are used as the dependent variables for the between subjects (Level 2) analyses. Level 2 ß 0j (Intercept) = γ 00 + γ 01 (Gender) + u 0j (2.3) ß 1j ( Cognitive) = γ 10 + γ 11 (Gender) + u 1j where γ k0 represents the mean of the corresponding Level 1 ß coefficient (as expressed by the equations’ dependent variable), and γ k1 captures the group effects (gender in this 96 case) on ß’s in Level 1. The effect of these subgroups is calculated using maximum likelihood estimates. 3. STUDY ONE The first study was designed to answer the following questions: 1. How are risk perceptions of terror and non-terror events related to perceived likelihood and consequence judgments about risk? 2. Are terror and non-terror events perceived differently? 3. Are personal and societal risk judgments evaluated differently? 3.1. Method Student volunteers from undergraduate classes at the University of Southern California (USC) were asked to participate in a terrorism risk perception study. Subjects received class credit toward a requirement for participation. A total of 151 students completed the survey. Of the 151 students, 94 were females and 57 males. Each respondent was presented with a range of terror and non-terror events (20 disasters in total). Terror attacks include three related to radiological risk (dirty bomb, nuclear weapon, portable nuclear device), three risks from chemical/biological attacks (smallpox, anthrax, chemical release), four risks involving transportation nodes and/or conventional explosives attacks (truck bomb, 9/11 copycats, MANPADS, transportation node explosion), and two risks from food and water contamination. Non-terror attacks include five naturally created risks (tsunami, hurricane, earthquake, wild fires, Avian flu 97 outbreak), and three manmade accidents (dam failure, toxic gas release, nuclear power plant meltdown). These stimuli were selected to span a broad range of events that students would find familiar. They were ordered randomly so as not to suggest that one event was of greater importance than another. Subjects were told they would be responding to questions asking them to make risk judgments about events. The ordering of survey questions was held constant since researchers wanted subjects to report their raw risk estimates prior to thinking of event risk in terms of likelihood and fatalities. Table 4.1 lists the survey questions asked in sequential order, as well as the graphical rating scales used for each. Table 4.1: Survey Questions and Rating Scales in Sequential Order Primary Survey Question Scale Risk to you and your family 0 (no risk) to 10 (high risk) Risk to society 0 (no risk) to 10 (high risk) Event likelihood (in the next year) 0 (no likelihood) to 100 (most likely) Estimated fatalities (from worst case disastrous event) Number killed estimate Preliminary Survey Question Scale Event likelihood (next year) Probability estimate Estimated fatalities (next year) Number killed estimate Most likely type of terrorist attack in the U.S. Attack type Reason for no additional attack on the U.S. since 9/11 Multiple choice To begin, participants were presented with five preliminary questions asking them to estimate the probability of a terrorist attack in the next year, the number of U.S. fatalities from terror attacks in the next year, the most likely attack type to occur in the U.S. and 98 why they believe there has not been another attack on the U.S. since 9/11. These questions were intended to collect general information about subjects’ risk perceptions prior to providing them with more technical questions. After finishing the opening section, subjects were shown the list of 20 events and asked to make judgments about each. For each event, the subjects made a judgment of the perceived personal and societal risk, and then gave estimates for two cognitive predictors: likelihood of event occurrence in the next year and the number of expected fatalities from “the worst case” event scenario. Perceived personal and societal risk estimates were marked on a graphical rating scale anchored at 0 (low risk) and 10 (high risk). The risk ratings were meant to reflect the subject’s overall perception of the event’s risk to themselves and their families, and then separately to society at large. To estimate event likelihood, subjects were asked to rate the likelihood of each event occurring in the next year using a percentage scale from 0 (no likelihood) to 100 (most likely). Lastly, to estimate fatalities, subjects were asked to think about the worst-case event that was “reasonably possible,” yet catastrophic. They were then instructed to provide an estimate of the number of persons that would be killed if such an event were to occur. Lastly, the survey included two questions that drew answers not from the list of 20 events. The first question asked subjects to estimate the probability of a list of “attack types” occurring. The second question asked subjects to estimate the probability of a list of “attack targets” being selected for a terrorist attack. For both questions the scale was anchored at 0% and 100%. These questions were included but not analyzed for the 99 purposes of this evaluation. The entire task took students on average 45 minutes. A copy of the survey instrument is available in Appendix A. 3.2. Results Study One data was assessed using both mean judgments and HLM analysis. The mean data was used to understand how risk perceptions are related to beliefs about event probability and fatality estimates. Emphasis was placed on whether patterns in the data demonstrated the presence of relationships across variables. Subsequently, researchers tested for significant relationships in individual’s risk perceptions using HLM Level 1 analyses. Figure 4.1 illustrates scatterplots comparing the subjects’ average perceived societal risk and probability of attack estimates for the 20 events. Figure 4.1: Mean Likelihood Estimates vs. Societal Risk Ratings 100 There was a positive linear relationship between the two variables (r 2 = .436). While not illustrated, there was a similar but weaker relationship found between perceived personal risk and probability of attack (r 2 = .127). These findings suggest that as the perception of risk increases so do estimates for the probability of a given event occurring in the next year. However, while most risk judgments, both personal and societal, fall within a relatively fixed range (2 to 6), probability estimates cover a much larger range (5% - 40%), indicating that the variability in the relationship is more sensitive to likelihood estimates. These findings are consistent with data trends found across non-terror and terror events. The relationships for societal risk are considerably stronger than those for personal risk as seen in Table 4.2. The three terror attacks perceived as riskiest were those using a truck bomb, detonating a portable nuclear device, and attacking a transportation node. The three non-terror events perceived as riskiest were a hurricane, earthquake, and wild fires. Table 4.2: Correlation Coefficients for Mean Likelihood Estimates by Attack and Risk Type Personal Risk Societal Risk Non-terror events .145 .401 Terror events .078 .697 Figure 4.2 shows scatterplots comparing the subjects’ average perceived societal risk and fatality estimates for the 20 events. Log transformations were conducted over fatality estimates for illustrative purposes. 101 Figure 4.2: Mean Fatalities Estimates (Log10) vs. Societal Risk Ratings There is a very slight positive linear relationship between the two variables, (r 2 = .05). While not illustrated, there was a similar relationship found between perceived personal risk and fatality estimates (r 2 = .08). This finding suggests that there is no relationship between perceived risk and estimated fatalities. Table 4.3 shows that significant relationships were found when the events were assessed as two separate groups, terror and non-terror, except for the relationship between perceived societal risk and fatality estimates of terror events. 102 Table 4.3: Correlation Coefficients for Mean Fatality Estimates by Attack and Risk Type Personal Risk Societal Risk Non-terror events .303 .128 Terror events .559 .014 The three terror attacks perceived as riskiest were those involving the detonation of a portable nuclear device, release of a nuclear missile, and attack upon a transportation node. The three non-terror events perceived as riskiest were a tsunami, earthquake, and avian flu outbreak. HLM was used to run two Level 1 regression models that evaluated the influence of attack likelihood and fatality judgments on perceived personal and societal risk. The dependent variables were personal and societal risk. The independent variables were attack likelihood and estimated fatalities. Prior to analyzing the findings, data were transformed to satisfy regression assumptions. The log odds of an event were calculated for risk and likelihood rankings and log transformations were performed on event fatality estimates. 16 Table 4.4 reveals the risk relationships between the predictor variables. 103 Table 4.4: HLM Level 1 Risk Relationships Estimated fatalities Attack likelihood Societal risk Personal risk β = .17** β = .32** β = .27** β = .39** TERROR EVENTS ** p < .01 Estimated fatalities Attack likelihood Societal risk Personal risk β = .19** β = .34** β = .26** β = .37** NON-TERROR EVENTS Results indicate that for terror related events, as subjects’ estimates of attack likelihood increased, so did perceptions of both personal and societal risk. A similar pattern was seen relative to the impact of likelihood estimates of non-terror related events upon personal and societal risk. The means scatterplots convey similar risk relationships. When considering fatality estimates of terror-related events, as respondents’ estimates increased, so did their perception of both personal and societal risk. For non- terror events, personal and societal risk estimates increased with fatality estimates. In the means scatterplots, these relationships were found only to be significant when analyzed independently as terror and non-terror events. 3.3. Summary and Discussion of Study One Means data scatterplots and HLM analyses showed that terror and non-terror risk ratings are more strongly related to likelihood estimates than to consequence (fatality) estimates. In addition, perceived terror attack risks rated similarly to non-terror event risks for all relationships. These patterns were consistent for perceived personal and 104 societal risk judgments. Lastly, HLM analysis was used to demonstrate that the assessment of risk is likely to be better captured through analysis of individual risk perceptions as opposed to through the study of aggregated data. 4. STUDY TWO The motivation for Study Two is to explore non-actuarial characteristics thought to be predictive of risk judgments. This study applies the psychometric paradigm to test how subjects make judgments about perceived personal risk of terror events based on cognitive and emotional predictor variables. Researchers assessed whether the perceptions of terrorism risk are determined more by emotional or cognitive attack characteristics. 4.1. Method A second sample of USC students were asked to participate in a terrorism risk perception study through the university’s psychology subject pool. A total of 331 students completed the survey. Study Two presented each subject with a list of 12 terror attack stimuli. Respondents were asked to make overall judgments about perceived personal risk. Subject’s risk judgments were regressed on both cognitive and emotional predictors. In addition, the perceived capability of a terrorist to execute an attack was incorporated into the analysis as a cognitive predictor in order to consider the subject’s perceived risk based on their understanding of the adversary. The emotional predictors included perceived dread and disaster potential of a terror attack, and the subject’s familiarity with the attack type. 105 Similar to Study One, the ordering of survey questions was held constant since researchers wanted subjects to report their raw risk estimates prior to thinking of event risk in terms of likelihood and fatalities. However, the ordering of the terror events was varied to ensure that subjects were evaluating each terror attack stimuli independently; 12 versions were created and 30 students on average completed each. Table 4.5 lists the survey questions asked in sequential order, as well as the graphical rating scales used for each. Table 4.5: Survey Questions and Rating Scales in Sequential Order Primary Survey Question Scale Familiarity and knowledge of terrorist event 0 (not familiar) to 100 (very familiar) Terrorist capability of attack execution 0 (not capable) to 100 (very capable) Terrorist attack dread 0 (no dread) to 100 (very dreaded) Disaster potential of terrorist event 0 (not disastrous) to 100 (very disastrous) Risk of terrorist event to you and your family 0 (no risk) to 10 (high risk) Terrorist event likelihood (in the next year) 0 (no likelihood) to 100 (most likely) Terrorist event estimated fatalities (from worst case disastrous event) Number killed estimate Rating scales for shared predictor variables with Study One remained the same. To estimate perceived terrorist capability, subjects were asked to think about the terrorist’s sophistication and feasibility of executing attacks in terms of dollars, manpower, materials, and other resources needed. They were then instructed to rate a terrorist’s capability using a percentage scale from 0 (Not capable) to 100 (Very capable). The perceived dread associated with an attack type was marked on a graphical rating scale anchored at 0 (no dread) and 100 (very dreaded). The term ‘dread’ is an all- 106 encompassing term that captured a subject’s fear, scare, and desire to avoid an attack. To estimate an attack’s disaster potential, subjects were asked to rate how disastrous each terrorist event would be in terms of the political, economic, psychological, and casualty outcomes using a scale from 0 (Not disastrous) to 100 (Very disastrous). To estimate attack familiarity, subjects considered their knowledge of an attack type and rated it on a scale from 0 (Not familiar) to 100 (Very familiar). The entire task took students on average 45 minutes. A copy of the survey instrument is available in Appendix B. 4.2. Results As in Study One, Study Two data were analyzed using both mean judgments and HLM. The mean data were used to assess the relationships between risk perceptions and cognitive and emotional predictors. Subsequently, researchers tested for the presence of the same relationships using HLM Level 1 analyses. Figure 4.3’s scatterplots compare the subjects’ average perceived personal risk with likelihood and fatality (Log10) b estimates across the 12 terror events. There are very slight, almost negligible, significant relationships between perceived personal risk and likelihood (r = .019, p<.01) and fatality estimates (r = .023, p< .01). Perceived risk and likelihood are not positively correlated because for 6 of the 12 events, subjects ranked them roughly the same, suggesting that subjects found it difficult to estimate the probability of an event. A possible explanation for the low correlation between fatality estimates and perceived risk is that the number of deaths associated with a nuclear attack is significantly greater compared to other terror events. When the nuclear missile attack b Log transformations were conducted over fatality estimates for illustrative purposes. 107 was removed from the analysis, the relationship between risk and fatalities proved to be stronger (r = .525, p<.01) Figure 4.3: Mean Likelihood and Fatality (Log10) Estimates vs. Risk Rating Figure (a): Mean Likelihood Estimates vs Personal Risk Rating Figure (b): Mean Fatalities Estimates vs Personal Risk Rating 108 Figure 4.4 compares the relationships between judgments of attack familiarity and terrorist capability with risk ratings. Figure 4.4: Mean Familiarity and Capability Estimates vs. Risk Rating Figure (a): Mean Capability Estimates vs Personal Risk Rating Figure (b): Mean Familiarity Estimates vs Personal Risk Rating 109 There are strong significant relationships between perceived personal risk and familiarity (r = -.694, p <.05) and capability (r = -.738, p<.05). The results indicate that as subjects’ familiarity and capability ratings increased, the perceived risk of an attack decreased. It appears subjects believe terrorists are more capable of executing the attacks with which they (the subjects) are more familiar. Figure 4.5 illustrates the relationships between judgments of attack dread and disaster potential with risk ratings. Figure 4.5: Mean Disaster Potential and Dread Estimates vs. Risk Rating Figure (a): Mean Disaster Potential vs Personal Risk Rating 110 Figure 4.5: Mean Disaster Potential and Dread Estimates vs. Risk Rating (cont’d) Figure (b): Mean Dread Estimates vs Personal Risk Rating Disaster potential (Figure 4.5a) has a strong positive linear relationship with perceived risk (r =.725, p<.05). Conversely, attack dread (Figure 4.5b) is not significantly correlated with perceived risk (r = .076, p<.01). These findings indicate that as subjects’ disaster potential increased, the perceived risk of an attack also increased. Interestingly, a significant negative relationship was found between disaster potential and attack dread (r = -.595, p<.05). Figure 4.6 shows that as the perceived disaster potential of a terrorist attack increase, the perceived dread of the attack decreases. 111 Figure 4.6: Mean Disaster Potential vs. Dread Estimates To examine variations in how emotional and cognitive judgments characterize risk perception, HLM was used to run two regression models. The first addressed how judgments of personal risk are related to cognitive characteristics of terrorist attacks, and the second regressed the same dependent variable on emotional predictors. Prior to analysis, data transformation procedures described in Study One (see page 102) were applied. Table 4.6 presents a summary of the model’s results. 112 Table 4.6: Summary of Risk Relationships with Cognitive and Emotional Indicators COGNITIVE PREDICTORS Estimated fatalities Attack likelihood Personal risk β = .05 β = (.07)* β = .61* * p < .05 Terrorist capability EMOTIONAL PREDICTORS Attack dread Disaster potential Personal risk β = .35* β = (.06)* β = .29* Attack familiarity Although these results were drawn from a different pool of subjects than Study One, a similar relationship was observed between the two studies with regard to perceived personal risk and estimated fatalities. Findings indicate that as fatality estimates increase, so do subjects’ perceived risk to themselves and their families. Perceived attack likelihood did not have a significant relationship with personal risk as in Study One. Lastly, it was found that terrorist capability and attack familiarity were not significantly related to perceived risk. Emotional variable analysis suggests that as estimates of attack consequence increased, namely attack dread and disaster potential, so did personal risk estimates. Additionally, there was a negligible significant relationship between attack familiarity and personal risk. 113 4.3. Summary and Discussion of Study Two With a larger and different sample from Study One, Study Two’s HLM analyses also showed that individual risk ratings are positively related to fatality estimates. Likelihood estimates, however, were found not to have a significant impact upon perceived risk as in Study One. Study Two also found that perceived personal risk of terrorism is most closely related to possible consequences of the attack (disaster potential and estimated fatalities). Furthermore, these results support findings from Study One that risk judgments are characterized by cognitive predictors, and in addition demonstrate that emotional indicators proved to be important for capturing components of perceived risk as well. 5. STUDY THREE The motivation for Study Three was to assess two aspects of risk perception through the use of a heterogeneous sample. First, HLM Level 1 models were estimated to replicate and draw comparisons to Studies One and Two. Next, HLM Level 2 analyses were used to test for differences in risk perceptions across groups subdivided based on gender, age, education level, income level and location (within the U.S). 5.1. Method A nationwide sample of 1,991 Qualtrics panel members received a survey about the perception of disaster risk. The study sample was reduced to 1,026 people who met the screening criteria and completed the survey questions satisfactorily. The screening criteria were developed to establish equal group sizes across the gender, age, and personal annual income demographic variables, as well as to assure that geographic 114 location was distributed proportionally to national census statistics. Table 4.7 provides a complete breakdown of the sample. Equal groups defined by annual income proved difficult to acquire given the sample population interested in online surveys. To manage this discrepancy, income was divided into three groups for analysis, low (below $20,000 - $39,999), moderate ($40,000 - $69,999) and high ($70,000 - $90,000 or more) levels of income. Manipulation checks were also built into the survey to evaluate whether subjects read the survey or provided a string of random responses. Questions for which the answers are common knowledge (e.g., one question read: “From the options provided, please select the current month.”) were inserted randomly at the beginning, middle and end of the survey. If a subject answered any of these questions incorrectly, they were removed from the final sample. Table 4.7: Study Three’s Sample Demographic Data Gender Response Percent Female 521 51% Male 508 49% TOTAL 1029 100% Age Range Response Percent 18 to 24 years 173 17% 25 to 34 years 169 16% 35 to 44 years 172 17% 45 to 54 years 172 17% 55 to 64 years 168 16% 65 years & over 175 17% TOTAL 1029 100% Income Range Response Percent Below $20,000 183 18% $20,000 - $29,999 158 15% $30,000 - $39,999 139 14% $40,000 - $49,999 103 10% $50,000 - $59,999 110 11% $60,000 - $69,999 56 5% $70,000 - $79,999 70 7% $80,000 - $89,999 35 3% $90,000 or more 115 11% No Response 60 6% TOTAL 1029 100% 115 All subjects, depending on the extent of survey completion, received compensation from Qualtrics for their participation. Additional information on Qualtrics recruitment service is available at the company’s web site. 17 In Study Three, subjects were asked to make overall judgments of perceived personal and societal risk. Risk was evaluated in terms of the cognitive and emotional predictors used in Study Two. In addition, risk was assessed for each of the terror and non-terror events (20 events) from Study One. As seen in Table 4.8, the survey design coupled previously defined screening variables with the same preliminary and primary survey questions from Study Two. Table 4.8: Survey Questions and Rating Scales in Sequential Order Primary Survey Question Scale Risk of terrorist event to you and your family 0 (no risk) to 10 (high risk) Risk to society 0 (no risk) to 10 (high risk) Terrorist event likelihood (in the next year) 0 (no likelihood) to 100 (most likely) Terrorist event estimated fatalities (from worst case disastrous event) Number killed estimate Familiarity and knowledge of terrorist event 0 (not familiar) to 100 (very familiar) Terrorist capability of attack execution 0 (not capable) to 100 (very capable) Terrorist attack dread 0 (no dread) to 100 (very dreaded) Disaster potential of terrorist event 0 (not disastrous) to 100 (very disastrous) Preliminary Survey Question Scale Event likelihood (next year) Probability estimate Estimated fatalities (next year) Number killed Most likely type of terrorist attack in the U.S. Attack type Reason for no additional attack on the U.S. since 9/11 Multiple choice Demographic Data 1. Gender 5. Demographic location 2. Age 6. Highest education level 3. Religion 7. Personal annual income 4. Race 8. Marital status 116 Question and event order remained constant throughout the survey. The survey closed with a series of questions asking subjects how much they have changed their life activities since 9/11 because of the possibility of future terrorism in the U.S. These findings are not including in Study Three’s analysis. A copy of the survey instrument is available in Appendix C. 5.2. Within Subjects (Level 1) Research Questions The Level 1 analysis was designed to answer the following research questions: 1. Are there differences between personal and societal risk perceptions of terror and non-terror events? 2. How are risk perceptions of terror and non-terror events related to cognitive and emotional judgments about risk? 3. Are terror and non-terror events perceived differently? 5.3. Within Subjects (Level 1) Results HLM was used to run eight Level 1 regression models. The dependent variables were personal and societal risk. The groups of independent variables were cognitive and emotional predictors, and two separate event groups, non-terror and terror, were assessed. Each dependent variable was regressed on the two sets of predictor variables for each event group type. As in Studies One and Two, prior to analyzing the findings, the data were transformed to satisfy regression assumptions. 117 Table 4.9 shows the relationships from the Level 1 regression models between risk and the cognitive variables across both terror and non-terror events Table 4.9: HLM Level 1 Cognitive Variables Risk Relationships * p < .05 TERROR EVENTS β = .34* β = .18* Terrorist capability Estimated fatalities Attack likelihood Societal risk Personal risk NON-TERROR EVENTS Attack likelihood Estimated Fatalities Personal risk Societal risk β = .35* β = .16* As in Studies One and Two, perceived personal and societal risks increased with attack likelihood and fatality estimates for both event types. Likelihood emerged as the best predictor of both risk types. Perceived personal and societal risks also were found to increase with perceived terrorist capability of terror events. The relationship between personal risk and terrorist attack capability, however, shifted from a negative relationship in Study Two to a positive relationship in Study Three. 118 Table 4.10 depicts the relationships from the Level 1 regression models between risk and the emotional variables across both terror and non-terror events. Table 4.10: HLM Level 1 Emotional Variables Risk Relationships * p < .05 TERROR EVENTS β = .20* β = .08* Attack familiarity Disaster potential Attack dread Societal risk Personal risk NON-TERROR EVENTS β = .29* β = .20* Attack familiarity Disaster potential Attack dread Societal risk Personal risk As in Study Two, perceived personal risk increased with judgments about attack dread and disaster potential for terror attacks. However, in this study dread proved to be the best predictor of risk, while in Study Two it was disaster potential. In addition, the relationship between personal risk and attack familiarity shifted from a negative relationship in Study Two to a positive relationship in Study Three. While not assessed until Study Three, similar relationship patterns between personal risk and emotional predictors were found for non-terror events. Also, the relationship between societal risk and emotional indicators was evaluated. Results in Table 4.10 show a similar pattern across societal risk relationships as those found for personal risk for both terror and non- terror events. 119 5.4. Summary and Discussion of Level 1 Analysis The patterns of perceived risk relationships derived from the national sample were similar to the USC student samples in Studies One and Two. However, the relationships between perceived risk (personal and societal) and judgments about estimated fatalities, attack dread, and attack disaster potential overall were found to be weaker across terror and non-terror events in the national sample. This was expected given the heterogeneity of the Study Three sample. Study Three’s Level 1 analysis produced new information about the following: 1. The risk relationships with likelihood and fatality estimates were the same for both risk types (personal and societal) for all events, with likelihood being the best predictor of perceived risk. This finding is consistent with results from Study One. 2. Among the emotional variables, the strongest predictor of perceived personal and societal risk was attack dread for all events. The results for terror events are different than those from Study Two for which the strongest predictor was attack disaster potential. 3. Judgments about terrorist capability and attack familiarity were found to be positively correlated with perceived risk (personal and societal) in Study Three for all events. These findings suggest that subjects feel the more likely attack types (in terms of greater familiarity and capability of execution) are riskier. This result is the opposite of findings from Study Two that were specific to terror attacks only. Study Two’s findings suggest subjects feel the less likely terror events (in terms of less familiarity and capability of execution) are riskier. This dichotomy in the perception of terror 120 events may be attributed to differences in interpretation of the predictors by the national versus USC student samples. Alternatively, these results may indicate that the perceived risk relationships are weakly correlated and possibly worthy of dismissal. 5.5. Between Subjects (Level 2) Research Questions The Level 2 subgroup analysis of perceived risk relationships sought to answer the following research questions: 1. Are there differences across gender, age, income, education, and geographic location (subgroups) in how cognitive and emotional components relate to risk perceptions? 2. Are there differences across subgroups in how predictors relate to personal versus societal risk? 3. Are there differences across subgroups in how terror and non-terror events are perceived? 5.6. Between Subjects (Level 2) Results HLM Level 2 analyses assessed how the impact of cognitive and emotional components of risk perception varies across different groups categorized by demographic data. More specifically, researchers evaluated whether gender, age, location, education, or income affected the relationship between perceived risk (both personal and societal) and the cognitive predictors (perceived attack likelihood, estimated fatalities and terrorist capability) for both terror and non-terror events. The same analysis was repeated with the emotional predictors (perceived attack dread, disaster potential and attack familiarity). Furthermore, to fully evaluate the impact of the demographic groups, various contrasts, 121 or categorizations of the demographic data were explored to ensure that the group analysis was exhaustive. 5.6.1. Gender Subgroup For the gender analysis, there was one contrast divided into two groups; males coded as 0 and females coded as 1. Consequence related predictors had the greatest impact on risk perceptions of terror events. Females had higher fatality estimates than males when estimating the perceived personal risk to themselves and their families (γ = 0.06, p < .05). Females also had larger judgments of attack disaster potential relative to perceived societal risk (γ = 0.04, p < .01). A similar relationship between perceived risk perceptions and consequence predictors was found for non-terror events. Females had higher fatality estimates when estimating personal risk (γ = 0.07, p < .05). Females also felt the dread of a non-terror event was greater when estimating their personal risk (γ = 0.05, p < .05). The only other significant coefficient for non-terror events was attack likelihood. Females estimated the likelihood of a non-terror event to be a greater perceived societal risk compared to males (γ = 0.06, p < .05). 5.6.2. Location Subgroup The location analysis tested for risk perception differences by comparing different segmented regions of the U.S. to the Northeast, which was the 9/11 target area. Table 4.11 shows how location was divided and coded into four contrast groups. 122 Table 4.11: Location Analysis Contrast Groups Contrast 1 Contrast 2 Contrast 3 Contrast 4 Northwest -1 -1 0 0 Southwest -1 -1 0 0 Midwest -1 0 0 +1 Mid-Atlantic +1 0 +1 +1 Northeast +1 +2 -2 -2 Southeast +1 0 +1 0 Contrast 1 compares the western part of the country to the east. Contrast 2 focuses on possible differences between the West Coast and the Northeast. Contrast 3 is centered on the East and draws a comparison between the Northeast and the Southeast/mid-Atlantic regions. Lastly, contrast 4 compared Middle America (Midwest and Mid-Atlantic) and the Northeast. Contrasts 1 and 2 produced significant results indicating that disparities in perceived risk by location are more evident when comparing western and eastern regions of the U.S. Contrast 1 showed that for terror attacks, subjects in the North- and South- west had higher likelihood estimates of attacks when evaluating their perceived personal risk (γ = -0.03, p < .05). Contrast 2 indicated that subjects in the West also estimated the disaster potential of terror attacks to be greater in regions outside of their own (γ = -0.04, p < .01). For non-terror events, contrast 2 found that the coefficients were significant for attack likelihood across both risk types (personal risk: γ = -0.03, p < .05 for both risk types). This suggests that subjects in the north- and south-west regions of the country feel 123 there is a greater probability of an attack occurring compared to subjects from the Midwest, Mideast, and East. 5.6.3. Age Subgroup For the age subgroup analysis, the three contrasts defined in Table 4.12 were explored. Contrast 1’s design considers possible difference in risk perception within the older population, specifically examining differences between 45-64 year olds and those 65 years and above. Conversely, contrast 2 considers differences among the younger population, specifically between 18-24 year olds and 25-44 year olds. And contrast 3 compares the younger population (18-44 years) to the older population (45 and up). Table 4.12: Age Analysis Contrast Groups Contrast 1 Contrast 2 Contrast 3 18-24 years 0 -2 0 25-34 years 0 +1 0 35-44 years 0 +1 0 45-54 years +1 0 +1 55-64 years +1 0 +1 65+ years -2 0 +1 Contrast 4 produced significant results suggesting that differences in risk perception arise when comparing younger and older populations. The perceived risks of terror attacks leads younger subjects (18-44 years) to place greater weight on the predictive power of cognitive variables, while older subjects are influenced more by emotional predictors. More specifically, younger subjects’ fatality estimates were higher than their elders when estimating the risk to themselves and society (γ = -0.08, p < .01 124 and γ = -0.03, p < .05, respectively). Conversely, older subjects dread the risk of terror attacks more than their younger counterparts when estimating their perceived personal and societal risk (γ = 0.02, p < .05 and γ = 0.04, p < .01, respectively). For non-terror events, older subjects’ perception of the risk to themselves and society was most strongly associated with event likelihood (γ = 0.06, p < .05 and γ = 0.12, p < .01, respectively). There were no significant relationships between non-terror event risk and emotional predictors. 5.6.4. Education Level Subgroup The education analysis was broken into two contrast groups (see Table 4.13). Contrast 1divided the sample into three groups for comparison, those with low, moderate, and high levels of education. Contrast 2 assessed differences between those persons with and without college or advanced degrees. Table 4.13: Education Level Analysis Contrast Groups Contrast 1 Contrast 2 High School (HS)/ HS Graduate 0 -2 Some college/ 2-yr college degree -1 +1 4 yr college degree/ Masters degree/ Doctorate/ Professional degree +1 +1 Contrasts 1and 2 produced significant findings for terror events. Contrast 1 results indicate that less-educated subjects’ (people with either a high school or two year college degree) terror risk perceptions were influenced more by consequence predictors. They had higher fatality estimates when assessing their personal and societal risk 125 (γ = -0.03, p < .05 for both risk types). They also perceived an event’s disaster potential to be greater relative to perceived societal risks (γ = -0.03, p < .01). To the contrary, educated subjects perceived the risk to themselves and society as greater relative to the terrorist’s attack capability (γ = 0.02, p < .05 for both risk types). Contrast 2 findings suggest that subjects with only high school education have higher judgments of attack familiarity with respect to perceived personal and societal risk (γ = 0.03, p < .05 for both risk types). Also, higher educated subjects dreaded attacks to society more than less educated subjects (γ = 0.05, p < .05). For non-terror events, only one significant finding emerged from contrasts 1 and 2. Educated subjects (college level or above education experience) were more likely to associate disaster potential with perceived societal risk (γ = -0.06, p < .05). 5.6.5. Personal Annual Income Subgroup For annual income subgroup analysis, the three contrasts in Figure 4.14 were evaluated. Contrast 1 divides the sample into two groups by comparing the very low income subjects to the rest of the sample. Contrast 2 again divides the sample into two groups by comparing the wealthy subjects to the rest of the sample. Lastly, Contrast 3 assesses differences in risk perceptions across three groups: those with low, moderate, and high levels of income. 126 Table 4.14: Annual Income Analysis Contrast Groups Contrast 1 Contrast 2 Contrast 3 < 20K $20 - $29 K $30 - $39 K -2 +1 0 $40 - $49 K $40 - $59 K $60 - $69 K +1 +1 -1 $70 - $79 K $80 - $89 K $90 + +1 -2 +1 Contrasts 1 and 3 produced significant findings, indicating income influences the subjects’ perceptions of terror events. The results suggest that low income subjects ($39K and below) placed greater weight on fatality estimates as a predictor of personal risk of terror events (γ = 0.10, p < .01). Conversely, findings from contrast 1 indicate that higher income subjects ($40K and above) perceived societal risk to be slightly influenced by terrorist capability (γ = 0.01, p < .05). For the emotional predictors, results from contrast 3 suggest that the subjects with the highest income ($70K and above) relied more on perceptions of attack when evaluating societal risk (γ = 0.03, p < .05). When considering their perceived personal risk, the higher income subjects ($40K and above) were influenced more by the predictive power of attack familiarity (γ = -0.03, p < .05) and the moderately wealthy ($40 - $60K) by perceived disaster potential (γ = -0.03, p < .05). There were no significant relationships for non-terror events. 127 5.7. Summary and Discussion of Level 2 Analyses Given the limited subgroup analysis of terror attacks, Study Three’s Level 2 analyses were intended to be exploratory. Results indicate the following: 1. Overall, the subgroups had a greater impact on the relationship between cognitive and emotional predictors and perceived risk of terror events compared to non-terror events. 2. When assessing how subgroups affect the relationship between predictors and risk perception, no consistent patterns emerged except for the education subgroup. The relationship between emotional predictors and perceived societal risk was sensitive to changes in education level. 3. Annual income, education level, and age were more sensitive to terror attacks, while geographic location had a stronger association with non-terror events. Gender was the only demographic variable equally sensitive to both. 4. Gender, location, age, and annual income findings showed no significant impact upon personal or societal risk perceptions. Changes in education level, however, proved to have a greater impact on perceived societal risk. 6. Conclusion The motivation for the three risk perception studies was to use the psychometric paradigm to test various hypotheses about the perception of terror and non-terror risk. What follows is an overview of the paper’s risk perception findings characterized by cognitive predictors, emotional factors, and lastly by comparisons of risk perceptions across subgroup populations. Each predictor is summarized in terms of relationship to 128 perceived risk in terms of personal and societal risk and terror and non-terror events, when applicable. This review is followed by a brief discussion of how the study findings provide policy-relevant knowledge. In its most nascent form, perceived risk was estimated through the assessment of probabilities and consequences of an uncertain adverse event. 18 Research has shown that fatality estimates increase an individual’s perceived risk when the event is associated with uncertainty and limited scientific understanding. 19,20,21 In all three of the aforementioned studies, subjects were asked to make judgments about the number of fatalities resulting from the worst case scenario. Such scenarios often are associated with little to no exposure and consequently, extreme uncertainty. Therefore, to no surprise, for all three studies attack fatality estimates were positively correlated with attack risk, both personal and societal, for both terror and non-terror events. Probability or likelihood estimates also have been associated with increased risk perception. 22,22 For Studies One and Three, likelihood judgments were significantly related to risk for both terror and non-terror events. In addition, data patterns indicate these relationships were more strongly associated with perceived societal - rather than personal - risk. This finding is consistent with the tendency toward optimism relative to one’s personal chances of harm compared to that of society. 23 In Study Two, however, likelihood judgments were not positively correlated with risk perceptions. This disparity could be explained by recognized difficulties in understanding probabilistic processes that have resulted in misjudged risks (sometimes overestimated and other times underestimated) and judgments of fact wrongfully held in 129 confidence. 20,24 Study Two’s subjects consisted of college students who might have less experience with probabilistic assessment and, as a result, are more inclined to be misguided in their judgments. In addition, Study Two was focused only on terror events for which difficulty in understanding would have further complicated subjects’ abilities to make judgments about likelihood. However, people’s perceptions have been shown to be sensitive to other factors besides those judged by assessments of probability and consequence. Dread, disaster potential, and familiarity (old vs. new risk) are some of the characteristics of risk commonly used in the assessment of adverse events. 9,11,25 Research shows that both perceptions of event dread and disaster potential are consistently related to perceived risk. 9 An earlier risk perception study by Fischhoff et al. (1978) found dread to be one of the main explanatory factors of risk. In Study 3, judgments about event dread were of greater significance than disaster potential across both attack and risk types. However, in a more recent study conducted by Sjöberg (2005), subjects’ judgments about event dread exhibited a decreased correlation with risk compared to disaster potential. 26 This pattern is consistent with dread and disaster findings (about perceived personal risk of terror events) in Study 2. This shift in predictor dominance might be explained through Study 3 Level 2 age subgroup findings that suggest older subjects dread the risk of terror attacks more than their younger counterparts (see page 123). In Studies Two and Three, the evaluation of new versus old risk was characterized in terms of event familiarity. Research has found that subjects perceive the events with which they are more familiar as less risky. 27 Study Two findings are consistent with this 130 assumption in the assessment of perceived personal risk of terror events. However, Study Three’s results suggest that risk perceptions increase as subjects’ judgments about familiarity increase. These findings were constant across perceptions of personal and societal risk, as well as for terror and non-terror events. Research on the study of hurricane insurance has shown that people prefer to insure against high probability - low loss events. 27 Naturally, one becomes more familiar with those events that occur with greater frequency. While economic theory suggests that persons should protect themselves against rare incidents, human intuition tends to run counter to this assumption. In addition to assessing the relationships between perceived risk and subjects’ judgments of cognitive and emotional predictors, subgroup analyses offered a different perspective on perceived risk. In Study Three, the demographic subgroups evaluated (gender, age, geographic location, education level, and annual income) provided insight into how perceptions may vary across sub-sectors of the population. Gender risk perception research findings have shown that risk tends to be judged lower by men than by women. 28 Women are believed to be more vulnerable than males and thus more sensitive to risk. 29 Others posit that such gender differences are apparent from an early age, as girls have higher perceived risk of injury from play and are more able to identify risks. 30 This pattern is consistent with the gender findings from Study Three for both non-terror and terror events. In addition, in Study Three females placed greater weight on consequence variables that males. While literature tends to show that there are no gender differences across fatality estimates, 31 it is not surprising that females 131 are preoccupied with the impact of an event. Females tend to have a strong desire to take preventative and preparedness measures. 32 Such actions are often driven by one’s perception of the potential event consequences. Proximity to attack (location) traditionally has been associated with increased perceived risk by those who are in the immediate vicinity or are strongly linked through a personal story. 33 Given this, individuals closer to the event location often have higher probability and consequence estimates of risk. 33 The non-terror findings in Study Three are consistent with this assumption. Our results show that subjects in the West placed greater weight on attack likelihood relative to perceived risk. For terror attacks, an inverse relationship emerged. Those farthest from the attacks of 9/11 (persons located in the West of the U.S.) placed greater weight on attack likelihood and consequence (disaster potential). A possible explanation for these findings is that subjects in the West believe that while natural disasters are likely to occur in similar locations, terrorists are more likely to strike a new target. Similar patterns of perceived event risk were found among the education and income level subgroups. People with more advanced degrees tend to place greater weight on the same cognitive and emotional indicators as those with higher incomes and vice versa. Educated and wealthier subjects had higher familiarity and capability estimates than those of lower levels of education and income. These subjects tended to be more contemplative and intellectualize the nature of the threat when evaluating risk. Conversely, less educated and poorer persons placed more weight on estimated fatalities 132 and disaster potential compared to their counterparts. These subjects seemed preoccupied with attack consequences and less concerned about the seriousness of the threat. A recent study by Ho et al. highlights how education was found to play a role in assisting a person’s “control” disaster risk perceptions. Such analyses suggest that a more educated person is capable of understanding new information more easily, and therefore feels a higher degree of “control” over a disaster (in terms of understanding and making decisions relative to it). Many of the Study Three indicators favored by those with higher education and income levels aligned with what Ho et al. term “controllability.” Control is gained by formulating an understanding of the emotional predictors, attack differences, and terrorist attack execution capabilities. This information is then used to make more informed decisions about the perceived dread associated with an attack and the appropriate mitigation and/or response measures needed. The survey findings about age and risk perception are mixed. Some studies have found that older people tend to have higher risk estimates, 9 while others have demonstrated the opposite tendency. 33 Still others have found that there are no discernable differences by age group. 34 The findings in Study Three show that both age groups, younger and older, are associated with significant terror and non-terror event risk perceptions. These relationships, however, are based on different cognitive and emotional indicators. Older persons tend to place greater weight on likelihood and dread estimates, and younger people are influenced more by estimated fatalities. 133 6.1. Policy Implications A major challenge for policy analysts is managing how risks are perceived and responded to by individuals. A conceptual framework known as social amplification of risk 35 suggests there are risks perceived by experts as minor that often elicit strong public concerns. These concerns are quickly amplified by the way in which risk information is conveyed, whether it be through scientists, the media, or social networks. Ultimately, the string of reactions translates into substantial social and economic impacts. The problems cited with disaster emergency planning and preparedness efforts correspond with the principles of social amplification of risk. Research shows that people tend to protect against what they perceive as high-probability, low loss events. 21,36 This assumption reflects a different public perception of event types, their potential consequences, and the appeal of their associated preparedness options compared to scientists. 37 The diversity of risk perspectives coupled with the influx of multiple and varied sources of information quickly amplify into a negative situation. People ultimately show limited interest in taking preparedness measures because they don’t feel that preparations will make a difference in the case of an extreme event. 21 When people are not willing to prevent and prepare for events, their only option is to respond and react (social disruption). Many social, political, economic, and psychological factors contribute to policy analysts’ understandings of risky events. Any added understanding of how individuals think about risk could play a significant role in informing policy. Through risk perception studies, researchers assess individuals’ risk knowledge and use their findings 134 to contribute to policy makers’ efforts to develop preparedness and response programs for terror and/or non-terror events. This chapter’s studies indicate that cognitive and emotional predictors shape individuals’ perception of risk. These results are consistent with other researchers’ findings (Plapp, 2006; Sjöberg, 2000; Slovic, 1992; Slovic et al., 1982) suggesting that perceived risk is most strongly related to judgments about event dread and likelihood. 38 As such, general education about an event’s likelihood and dread characteristics could help redirect the public’s inflated and/or deflated perceptions of event risk. For example, informing the public that different levels of radiological threats exist is a critical distinction policy makers need to make relative to correcting public risk perceptions and ensuring public safety. Individuals perceive the risk of nuclear events to be incredibly high because the risks of radioactive material are unknown, dreaded, uncontrollable, and catastrophic. 39 Consequently, any event that involves radioactive material is perceived as being of equally high risk. A dirty bomb, however, is characterized more as a weapon of mass “disruption” than “destruction”, a label often associated with nuclear events. As such, there is added value in conveying the nature of a dirty bomb, and consequently, how a dirty bomb’s perceived likelihood and dread is different than that of a nuclear weapon. Overall, if well informed and prepared, the public and economy are expected to respond in a resilient way to a dirty bomb. Yet, if the public’s responses are emotional and misguided, more harm and damage will ensue from public panic and lack of compliance with safety procedures. 135 The challenge for policy analysts is not only conveying the basics about the differences across event risks, but also presenting the information using messaging the public will respond to. Morgan et al. (1992) advocate for presenting people with information they need in a form that fits their intuitive way of thinking. 40 The subgroup differences identified in Study Three contribute to policy analysts’ efforts to create more directed, preparedness-based outreach programs. For instance, subgroup analysis found that less educated, low income, and younger subjects placed greater emphasis on consequence predictors. Referring back to the dirty bomb example, the less educated subgroup might perceive a dirty bomb as less risky if educated about its forecasted limited fatalities and health consequences, and directed on how to prepare (such as by sheltering in place) in the event of an attack. Conversely, more educated, wealthier, and older people tend to intellectualize the threat and rely more on capability and familiarity estimates. As such, the more educated group is likely aware of appropriate preparedness and response measures in place. Thus, policy efforts should be directed toward keeping this population amply apprised of and involved with ongoing risk management efforts to the greatest extent possible. This is because having an “enhanced” awareness of their safety options provides this population with a greater sense of control, which has been associated with lower perceived risk. 21 The policy implications of this effort are twofold. First, the allocation of resources to increase public preparedness efforts will be more streamlined and effective because they are directed by subgroup differences in risk perception. Second, a shift in the lower educated groups’ focus toward preparedness efforts might produce a greater 136 understanding of the event threat, which in turn might translate into better public reactions to an attack in the event it might occur. In addition to being sensitive to how information is communicated to the public, it is equally important to share the studies with researchers investigating the economic impacts of terrorism events. An ongoing challenge for researchers has been to forecast the costs of outcome and response services needed in the event of a disaster for which we fortunately have had little experience. Consequently, when producing economic analyses of an attack’s consequences (e.g. evacuation costs, business shutdown, reductions in property value, and clean-up costs), existing reports have produced varied estimates. For example, Reichmuth (2005) estimated the cost of a dirty bomb attack to be in the trillions, 41 where Gordon et al.’s (2005) cost estimates were reduced by a factor of 1,000. 42 A panicked versus calm public response can lead to the demand for very different response services resulting in a wide range of cost estimates. In the case of a dirty bomb attack, a panicked public will likely require more immediate efforts directed at mitigating over-exposure to radiation and managing evacuation processes, where a calmer public is likely to require less support. Ultimately, the idea is for economic analysts to take the added knowledge about how the public might respond to an attack and use it to develop more efficient and cost effective response consequences and mitigation measures. 137 Chapter 4 Endnotes 1 Slovic, P. “Terrorism As Hazard: A New Species of Trouble.” Risk Analysis. 22.3 (2002): 425-426. 2 “Terrorist Attacks (within the United States or against Americans abroad).”Infoplease Page. June 2009. <http://www.infoplease.com/ipa/A0001454.html>. 3 Fischhoff, Baruch, et al. “How safe is safe enough? A psychometric study of attitudes towards technological risks and benefits.” Policy Sciences, 9 (1978):127-152. 4 Slovic, Paul, Baruch Fischhoff and Sarah Lichtenstein. “Behavioral decision theory perspectives on risk and safety.” Acta Psychologica. 56 (1984): 183-203. 5 Slovic, Paul and Elke Weber.“Perception of Risk Posed By Extreme Events.” Risk Management Strategies in an Uncertain World. Palisades, New York: April 12-13, 2002. 6 La Porte, Todd. “Governance and the Specter of Infrastructure Collapse.” The 8th National Public Management Research Conference. Los Angeles: University of Southern California, 2005. 7 Fischhoff, Baruch, et al. “Evolving Judgment of Terror Risks: Foresight, Hindsight & Emotion.” Journal of Experimental Psychology. 11.2 (2005): 124-139. 8 Slovic, Paul, et al. “Risk as Analysis and Risk as Feelings: Some Thoughts about Affect, Reason, Risk, and Rationality.” Risk Analysis. 24.2 (2002): 1-12. 9 Sjöberg, L. “The Perceived Risk of Terrorism.” Risk Management. 7.1(2005): 43-61. 10 Slovic, Paul, et al. “The affect heuristic.” European Journal of Operational Research. 177.3 (2007): 1333-1352. 11 Holtgrave, D. and E.U. Weber. “Dimensions of risk perception for financial and health risks. Insurance: Mathematics and Economics. 17.3 (1996): 249-250. 12 Loewenstein, G. F., et al. “Risk as feelings.” Psychological Bulletin. 127 (2001): 267-286. 13 Gardner, G.T. and L.C. Gould. “Public Perceptions of the Risks and Benefits of Technology.” Risk Analysis. 9.2 (1989): 225-242. 14 Sjoberb, L. “Factors in Risk Perception.” Risk Analysis. 20.1 (2000): 1-11. 15 Bryk, A.S. and S.W. Raudenbush. Hierarchical linear models: Applications and data analysis methods. Thousand Oaks, CA: Sage Publications, Inc., 1992. 16 Ashton, W.D. The logit transformation with special reference to its uses in bioassay. London: Griffin and Co., 1972. 17 For more details on Qualtrics subject recruitment services, see http://www.qualtrics.com/consumer- panels.html. 18 Slovic, Paul, et al. “Rating the Risks.” Environment. 21.3 (1979): 14-20. 138 19 Slovic, Paul, et al.. “Modeling the societal impact of fatal accidents.” Management Science. 30.4 (1984): 464-474. 20 Slovic, Paul, et al. “Why Study Risk Perception?” Risk Analysis. 2.2 (1982): 83-93. 21 Von Winterfeldt, Detlof, et al. “Cognitive Components of Risk Ratings.” Risk Analysis, 1.4 (1981): 277- 287. 22 Rogers, G. “The Dynamics of Risk Perception: How Does Perceived Risk Respond to Risk Events?” Risk Analysis. 17:6 (1997): 745-757. 23 Svenson, O., et al. “Perceived driving safety and seatbelt usage.” Accident Analysis and Prevention. 17.2 (1985): 119-133. 24 Slovic, Paul, et al. “Facts and Fears: Understanding perceived risk.” Eds. R. Schwing and W.A. Albers Jr. Societal Risk Assessment: How Safe is Safe Enough?. New York: Plenum, 1980. 25 Plapp, T. and U. Werner. “Understanding risk perception from natural hazards: examples from Germany.” Eds. Amman, Dannenmann and Vulliet. RISK 21 – Coping with Risks due to Natural Hazards in the 21st Century. London: Taylor & Francis Group, 2006. 26 Sjöberg, L. “Explaining Individual Risk Perception: The Case of Nuclear Waste.” Risk Management. 6.1 (2004): 51-64. 27 Kunreuther, H. “The role of insurance in managing extreme events: Implications.” Business Economics. 37.2 (2002): 6-16. 28 Flynn, James , et al. “Gender, Race, and Perception of Environmental Health Risks.” Risk Analysis. 14.6 (1994): 1101-1108. 29 Slovic, Paul. “Trust, emotion, sex, politics and science: Surveying the risk-assessment battlefield.” Eds. M. Bazerman, D. Messick, A. Tenbrunsel, & K. Wade-Benzoni. Environment, Ethics and Behavior. San Francisco, California: New Lexington Press, 1997. 277–313. 30 Morrongiello, B. and H. Rennie. “Why do boys engage in more risk taking than girls? The role of attributions, beliefs, and risk appraisals.” Journal of Pediatric Psychology. 23.1 (1998): 33– 43. 31 Hakes, J. and W. Viscusi. “Dead Reckoning: Demographic Determinants of the Accuracy of Mortality Risk Perceptions.” Risk Analysis. 24.3 (2004): 651-664. 32 Ho, M. and D. Shaw, S. Lin and Y. Chiu. “How Do Disaster Characteristics Influence Risk Perception?” Risk Analysis. 28.3 (2008): 635-643. 33 Fischhoff, Baruch. “Judged Terror Risk and Proximity to the World Trade Center.” The Journal of Risk and Uncertainty. 26.2-3 (2003): 137-151. 34 Hellesoy, O., et al. “Profiling the high hazards perceivers: An exploratory study.” Risk Analysis. 18.3 (1998): 253-272. 139 35 Kasperson, R. et al. “The Social Amplification of Risk: A Conceptual Framework.” Risk Analysis. 8.2 (1998): 177-187. 36 McDaniels, T.L., et al. “Risk perception and the value of safety.” Risk Analysis. 12.4( 1992): 495-503. 37 Larsson, G. and A. Enander. “Preparing for Disaster: Public Attitudes and Actions.” Disaster Prevention and Management. 6.1 (1997): 11–21. 38 Slovic, Paul. “Perception of risk: reflections on the psychometric paradigm.” Eds. S.Krimsky and D. Golding. Social theories of risk. Westport, CT: Praeger, 1992. 117-152. 39 Slovic, Paul, et al. “Images of disaster: Perception and acceptance of risks from nuclear power.” Eds. G. Goodman and W. Rowe. Energy Risk Management,. London: Academic Press, 1979. 40 Granger, Morgan, et al. “Communicating Risk to the Public: First Learn What People Know and Believe.” Environmental Science & Technology. 26.11 (1992): 2048-2056. 41 Reichmuth, B. “Economic consequences of a rad/nuc attack: Cleanup standards significantly affect cost.” Department of Homeland Security Conference on “Working Together: R&D Partnerships in Homeland Security.” Boston, Massachusetts: April 27–28, 2005. 42 Gordon, Peter, et al. “Los Angeles–Long Beach seaports radioactive plume economic impact.” Draft Report, Center for Risk and Economic Analysis of Terrorist Events. Los Angeles, California: University of Southern California, 2005. 140 Chapter 5: Conclusion The risk analysis techniques applied in Chapter 2, 3, and 4 present novel approaches to the assessment of terrorism risk. In Chapter 2, multi-attribute utility modeling and probabilistic assessment were used to assess how the values and beliefs of terrorist leaders might influence the selection of an attack strategy. Chapter 3 was an exploratory investigation of a combination of several risk analysis tools, including scenario generation and pruning, project risk analysis, direct consequence modeling, and indirect economic impact assessment, all drawn upon to evaluate the threat of a dirty bomb attack on the ports of Los Angeles and Long Beach. And Chapter 4 applied the psychometric paradigm to assess how individuals think about the risk of terrorism. Considered in their entirety, the project designs cut across multiple disciplines (policy, psychology, and engineering) and apply a divergent set of risk analytic techniques. In addition, each project’s outcomes make unique contributions to policy efforts to mitigate, prepare for, and respond to the terrorism threat. There are, however, policy makers that criticize quantitative-based analytic tools like risk analysis for not reflecting the processes that shape policy decisions. They are concerned with how analytic tools are used in: (1) identifying policy problems, (2) eliminating assessment bias (in terms of the decision maker and policy analyst), and (3) defining policy variables quantitatively. While critics are not arguing that quantitative tools be banished from policy analysis all together, they do posit that the tools be improved so as to maximize upon their contributions to policy decision making. 141 These closing pages argue that the novel approaches introduced in the previous chapters are careful to address the issues critics have with quantitative approaches and present viable contributions to the processes that shape policy making. Each of the concerns are defined and then refuted relative to the risk analytic approaches applied in Chapters 2, 3 and 4. 1. IDENTIFYING POLICY PROBLEMS Traditional quantitative tools of analysis have been cited as limited because they characterize policy problems only in terms of their economic and technical feasibility. Technical analyses may be entirely appropriate for purely engineering decisions and economic assessments for projects only related to cost efficiency. However, social problems in all their complexity cannot be captured strictly in terms of technical and/or economic feasibility. These approaches fail to account for the pervasiveness of stakeholder values and beliefs in shaping the decisions relevant to a societal policy issue. The omission of this key component in problem definition leads to attempts to apply solutions that ultimately solve the wrong policy problems, and even create new dilemmas. The proxy terrorist leader modeling approach applied in Chapter 2 addresses many of the complexities associated with problem definition. The premise was to model attack strategy selection by using the proxy terrorist leader’s values and objectives as the launching point for the decision problem, as opposed to the alternatives she or he faces. One of the more critical steps in reaching this end is the process of problem definition. For this decision approach to be effective, it is critical that analysts take the time to wean 142 through the difficulties in defining the objectives and alternatives associated with a problem definition. Rather than jumping into reformulating a problem so it fits into a technical of economic model, value focused thinking emphasizes keeping an open mind about problem formulations and taking the time to become familiar with the problem environment. In the development of the proxy utility model extensive time was spent speaking with the proxies individually and in repeated follow-up conversations. Initial meetings were directed at getting proxies into the terrorist mindset through general conversation about what they believed to be Al Qaeda’s objectives. These were followed by a series of discussions about how to accomplish the organization’s objectives through the selection of an attack type. The identification and nature of the decision problem quickly becomes complex, yet a comprehensive understanding is formulated. A similar argument about improved problem definition can be made for the probabilistic risk assessment approach used in Chapter 3. Several steps were taken to characterize the dirty bomb attack decision problem through a project management framework. Researchers collaborated with a counterintelligence expert to study the motivations and capability of a terrorist to pursue the attack option. Several rounds of meetings were held and extensive research using only open source data was conducted to outline the planning, preparing, and execution tasks of a dirty bomb attack. By formulating an understanding of the challenges in carrying out a dirty bomb attack, the true capability and feasibility of the problem is assessed in its entirety as defined by its parts. 143 In Chapters 2 and 3 the analyses are sensitive to accounting for values and beliefs in the elicitation of problem definition from the perspective of the terrorist. In Chapter 4, through the application of the psychometric paradigm, the risk perception studies focus more on what the public perceives and believes about terrorism and using this information to clarify problem definition. Clarify in this sense refers to ways in which the problems created by the terrorist are perceived, and in turn how they should be responded to. The risk perception study designs include both cognitive and emotional variables to fully capture the problem environments in terms of the critical risk predictors and stakeholder beliefs. 2. ELIMINATING ASSESSMENT BIAS AND UNCERTAINTY When evaluating a social problem, part of the policy maker’s job is to assist decision makers in framing an issue. The quantitative tools used for problem definition are also associated with elicitation techniques that assist in the evaluation of measures, probabilities, and utilities for a policy problem. Some of the more commonly used techniques include interviews and surveys. For values more difficult to elicit, approaches are used that frame problems in terms of bets, lotteries, importance ranking, utility functions, and tradeoff values. As policy decisions become more complex, it is critical to ensure that the quality of their decision does not suffer. While researchers understand that evaluation techniques are intended to transform the assessment of policies through more critical analysis, they are concerned that the nature of the relationship between the policy analyst and decision maker, and the personal objectives of each relevant party, could produce biased results. This creates the potential for policy decisions based on 144 misrepresentations of the issue and an analysis guided by the personal, political, economic, or academic motivations of analysts and decision makers. The ultimate goal of risk analyses is informed decision making about uncertain events. The previously described elicitation techniques are critical to defining and analyzing uncertain problems, but also can be used to account for concerns relative to assessment bias and uncertainty. These concerns can be managed through (1) the establishment of a good relationship between the analysts and proxies, and (2) the processes employed to elicit the information. 1 For the proxy utility model developed in Chapter 2, the accuracy of model definition was contingent upon the effectiveness of methods used by researchers to collect information from the proxy terrorist leaders. Elicitation techniques were used to identify and define the attributes for each attack strategy, specify relative risk preferences across the attack attributes, and prioritize through value tradeoffs among the different attack attributes. Bias and uncertainty were controlled for in how analysts conducted several rounds of interviews with proxies so as to provide researchers with a firm understanding of the proxies’ motives, and terrorist proxies with a comfortable, established repertoire. In addition, assessment techniques were facilitated by the use of careful anchors for scales, and relative judgments about the values (as opposed to absolute ones) to assist proxies in the quantification of values. However, without access to actual terrorist leaders, there is still uncertainty as to how accurate the proxies’ assessments might be. As a result, probability distributions were placed over elicited 145 values to address any uncertainty in terrorist predictions of future outcomes and in the proxy’s uncertainty about terrorist beliefs. A similar approach to avoiding assessment bias and uncertainty was employed in the dirty bomb model developed in Chapter 3. Each of the tasks involved in carrying out a dirty bomb attack were associated with a probability of detection and disruption of the event. As was the challenge with identifying terrorist leader values and beliefs in Chapter 2, without actual access to terrorists knowledgeable about plans for an attack, all estimates relative to a dirty bomb’s execution were formulated through educated means, although associated with a certain level of uncertainty. To avoid unintentionally optimistic or pessimistic assessments, researchers worked with a former counterintelligence expert to place probability distributions over the number of people and time for each task. In the study of risk perception, bias is less of a concern than uncertainty relative to the subjects’ approach to completing the survey. Researchers took precautions to account for whether subjects consistently over- or under-estimated their risk judgments, and to ensure that subjects carefully read and followed instructions. This was accomplished by being sensitive to sample size, inserting manipulations checks and altering question and attack list order in the survey designs, and lastly including various measures of cognitive and emotional predictors of risk so as to capture each subject’s unique perception of risk. 146 3. QUANTIFYING POLICY PROBLEMS To define the components of an uncertain policy problem, terrorist proxies are used to place a numeric value on variables not typically quantified, probabilities are used to estimate the likelihood of a variable’s occurrence, and utilities are used to exemplify how much value is placed on a certain variable. Critics are concerned that the proxies, probabilities, and utilities applied in risk analytic models produce an inaccurate representation of the policy problem. 2 Consequently, over reliance on quantitative evidence, coupled with the lack of appropriate data sources, weakens decision maker’s ability to argue for certain policy positions. The intent of Chapters 2, 3, and 4 were to assess the terrorism threat through the application of risk analysis techniques. As such, all three chapters employ some level of quantitative analysis and use terrorist proxies, probabilities, and utilities to enhance the analysis. Aside from the motivations for this dissertation, terrorism is a challenging field to study given the tremendous uncertainty surrounding its threats, vulnerabilities, and consequences. Fortunately, aside from the attacks of 9/11, there have been no new attacks upon the U.S. However, this limits the amount of available data for the study of terrorism and contributes to the demand for risk analysis techniques to address its unknowns. 4. CLOSING REMARKS Alvin Weinberg (1972) proposed that policy decisions and the application of the quantitative analysis tools such as risk analysis techniques fall into a special category called trans-science. 3 Trans-science refers to the questions that can be asked of science, 147 yet cannot be answered by science. The idea being that people need to learn how to make decisions for problems whose outcomes are associated with uncertainty and risk, as opposed to relying on definitive solutions. If one assumes the “trans-scientific” nature of risk analysis to hold true, there always will be some level of debate about the optimal method, model, or outcome. While there are recognized challenges with risk analysis techniques, they are becoming more common practice in policy decision making. All researchers may not be in agreement as to the benefits of these technique and other quantitative tools, but few would dispute the notion that some understanding of quantitative analysis is invaluable to policy decision making, particularly in relation to topics associated with extreme uncertainty like terrorism. This is not to suggest that policy analysts’ decisions should be driven by analytic models alone. Rather, coupling quantitative analyses with qualitative tools will likely provide the most solid and comprehensive foundation for policy decision making. 148 Chapter 5 Endnotes 1 Von Winterfeldt, Detlof and Ward Edwards. Decision Analysis and Behavioral Research. New York: Cambridge University Press, 1986. 2 Edwards, Ward and Robert Newman. “Multiattribute Evaluation.” Eds. Terry Connolly, Hal R. Arkes, and Kenneth R. Hammond. Judgment and Decision Making: An interdisciplinary Reader. New York: Cambridge University Press, 2000. 17-34. 3 Weinberg, A. “Science and trans-science.” Minerva. 10.2(1972): 209-21. 149 BIBLIOGRAPHY Adler, Matthew D. and Eric A. Posner. “Rethinking Cost-Benefit Analysis.” Yale Law Journal. (1999): 109. Ammann, Walter J., Stefanie Dannenmann and Laurent Vulliet eds. Proceedings of the RISK21 Workshop, Monte Verità, Ascona, Switzerland, 28 November - 3 December 2004. London: Taylor & Francis, 2006. Anderson, Elizabeth L. “Editorial Commentary: Assessing the Risks of Terrorism: A Special Collection of Perspectives Articles by Former Presidents of the Society for Risk Analysis.” Risk Analysis. 33.3 (2002): 401-402. Ashton, W.D. The logit transformation with special reference to its uses in bioassay. London: Griffin and Co., 1972. Bandura, Albert. “Mechanisms of moral disengagement,” Origins of terrorism: Psychologies, ideologies, theologies, states of mind. Ed. IW. Reich. Cambridge: Cambridge University Press, 1990, 161-191. Borcherding, Katrin., Thomas Eppel and Detlof von Winterfeldt. “Comparison of Weighting Judgments in Multiattribute Utility Measurement.” Management Science. 37.12 (1991):1603-1619. Blair, Bruce. “What if Terrorists Go Nuclear?” Center for Defense Information. 1 October 2001. <http://www.cdi.org/terrorism/nuclear-pr.cfm>. Bradley, Michael. “NARAC: An Emergency Response Resource for Predicting the Atmospheric Dispersion and Assessing the Consequences of Airborne Radionuclides.” Journal of Environmental Radioactivity. 96.1-3 (2007): 116-121. Bryk, A.S. and S.W. Raudenbush. Hierarchical linear models: Applications and data analysis methods. Thousand Oaks, CA: Sage Publications, Inc., 1992. Burns, W. & Slovic, P. “Predicting Public Response to a Terrorist Strike.” (Unpublished paper) 2006. Council on Foreign Relations. “Terrorism Q&A: Making a Bomb.” 2004. <http://cfrterrorism.org/weapons/making print.html>. Demuth, Julie L. A Summary to National Disasters Roundtable: Countering Terrorism: Lessons Learned from Natural and Technological Disasters. Washington D.C.: National Academies Press, February 28 – March 1, 2002. 150 Edwards, Ward and Robert Newman. “Multiattribute Evaluation.” Eds. Terry Connolly, Hal R. Arkes, and Kenneth R. Hammond. Judgment and Decision Making: An interdisciplinary Reader. New York: Cambridge University Press, 2000. 17-34. Enders, Walter and Todd Sandler. The Political Economy of Terrorism. Cambridge: Cambridge University Press, 2006. Erickson, K. “Toxic Reckoning: Business Faces a New Kind of Fear.” Harvard Business Review, (1990): 118-126. Erickson, K. A New Species of Trouble. New York: W.W. North & Company, Inc, 2006. Evan, William and Mark Manion. Minding the Machines: Preventing Technological Disasters. New Jersey: Prentice Hall PTR, 2002. Ferguson, Charles. D., Tahseen Kazi, and Judith Perera. “Commercial Radioactive Sources: Surveying the Security Risks.” Monterey Institute of International Studies, Center for Nonproliferation Studies. 11 (2003). Fischhoff, Baruch, et al. “How safe is safe enough? A psychometric study of attitudes towards technological risks and benefits.” Policy Sciences, 9 (1978): 127-152. Fischhoff, Baruch. “Judged Terror Risk and Proximity to the World Trade Center.” The Journal of Risk and Uncertainty. 26.2-3 (2003): 137-151. Fischoff, B. “Risk Perception and Risk Communication.” Ed. D. Kamien. The McGraw- Hill Handbook of Terrorism. New York: McGraw-Hill, 2004. Fischoff, B., Gonzales, R., Lerner, J. & Small, D. “Evolving Judgment of Terror Risks: Foresight, Hindsight & Emotion.” Journal of Experimental Psychology. 11 (2005): 124-139. Flynn, James, et al. “Gender, Race, and Perception of Environmental Health Risks.” Risk Analysis. 14.6 (1994): 1101-1108. Fouda, Yosri. “The laughing 9/11 bombers.” Times Online. 1 October 2006. World Wide Web, April 2009 < http://www.timesonline.co.uk/tol/news/world/article656683.ece>. Gardner, G.T. and L.C. Gould. “Public Perceptions of the Risks and Benefits of Technology.” Risk Analysis. 9.2 (1989): 225-242. Gardner, F. “Al Qaeda was making a dirty bomb.” BBC News Online. 31 January 2003. <http://news.bbc.co.uk/2/hi/uk news/2711645.stm>. General Accounting Office. ”Nuclear Security: Federal and State Action Needed to Improve Security 151 of Sealed Radioactive Sources.” 6 August 2003. < http://www.gao.gov/new.items/d03804.pdf>. Gordon, Peter, et al. “The economic impact of a terrorist attack on the twin ports of Los Angeles-Long Beach.” The Economic Impacts of Terrorist Attacks. Eds. Harry Richardson, Peter Gordon and James Moore. Northampton, Massachusetts: Edward Elgar Publishing Limited, 2005. 262-286. Gordon, Peter, et al. “Los Angeles–Long Beach seaports radioactive plume economic impact.” Draft Report, Center for Risk and Economic Analysis of Terrorist Events. Los Angeles, California: University of Southern California, 2005. Granger, Morgan, et al. “Communicating Risk to the Public: First Learn What People Know and Believe.” Environmental Science & Technology. 26.11 (1992): 2048-2056. Green, David, and John Swets. Signal Detection Theory and Psychophysics. New York: John Wiley and Sons Inc., 1966. Haimes, Yacov Y. Risk Modeling, Assessment, and Management. New York: John Wiley & Sons, Inc., 1998. Hakes, J. and W. Viscusi. “Dead Reckoning: Demographic Determinants of the Accuracy of Mortality Risk Perceptions.” Risk Analysis. 24.3 (2004): 651-664. Hammond, J.S., and Keeney, R.L, and Raiffa, H. Smart Choices: A practical guide to making better decisions. Boston, MA: Harvard Business School Press, 1999. Harper, F. “Working Together: R&D Partnerships in Homeland Security.” Department of Homeland Security Conference on “Working Together: R&D Partnerships in Homeland Security.” Boston, Massachusetts: April 27–28, 2005. Heasler, P., and T. Woods. “Game theoretic modeling of detection and deterrence.” Department of Homeland Security Conference on “Working Together: R&D Partnerships in Homeland Security.” Boston, Massachusetts: April 27–28, 2005. Hellesoy, O., et al. “Profiling the high hazards perceivers: An exploratory study.” Risk Analysis. 18.3 (1998): 253-272. Ho, M. and D. Shaw, S. Lin and Y. Chiu. “How Do Disaster Characteristics Influence Risk Perception?” Risk Analysis. 28.3 (2008): 635-643. Holtgrave, D. & Weber, E.U. “Dimensions of risk perception for financial and health risks.” Risk Analysis, 13 (1993): 553-558. International Atomic Energy Agency. (2002). “Inadequate Control of World’s Radioactive Sources.” 25 June 2002. < http://hps.org/documents/iaeapressrelease.pdf>. 152 Johnson, K., & T. Locy. “Threat of dirty bomb softened.” USA Today. 12 June 2002. <http://www.usatoday.com/news/nation/2002/06/11/bushothers.htm>. Kammen, Daniel and David Hassenzahl. Should We Risk It? Exploring Environmental, Health, and Technological Problem Solving. Princeton, New Jersey: Princeton University Press, 1999. Kasperson, R. et al. “The Social Amplification of Risk: A Conceptual Framework.” Risk Analysis. 8.2 (1998): 177-187. Keeney, Ralph. “Decision Analysis: An Overview.” Operations Research. 30.5 (1982): 803-838. Keeney, Ralph, Detlof von Winterfeldt, and Thomas Eppel. “Eliciting Public Values for Complex Policy Decisions.” Management Science, 36.9 (1990): 1011-1030. Keeney, Ralph. Value Focused Thinking. Cambridge, CA: Harvard University Press, 1992. Keeney, Ralph, and Howard Raiffa. Decisions with Multiple Objectives: Preferences and Value Trade-Offs. Cambridge: Cambridge University Press, 1993. Keeney, Ralph. “Modeling Values for Anti-Terrorism Analysis.” Risk Analysis. 27.3 (2007): 585-596. Keeney, Ralph. “Developing Objectives and Attributes.” Advances in Decision Analysis: From Foundations to Applications. Eds. Ward Edwards, Ralph F. Miles and Detlof von Winterfeldt. New York: Cambridge University Press, 2007. 104-128. Keeney, Ralph and Detlof von Winterfeldt. “Practical Value Models. Advances in Decision Analysis: From Foundations to Applications. Eds. Ward Edwards, Ralph F. Miles and Detlof von Winterfeldt. New York: Cambridge University Press, 2007. 232- 252. King, Gilbert. Dirty Bomb: Weapons of Mass Disruption. New York: Penguin Group, 2004. 25. Krikorian, Greg. “Port complex to get radiation detection devices to forestall terrorism.” Los Angeles Times. 4 June 2005. < http://www.latimes.com/news/local/la- mechertoff4jun04,1,7330940.story?ctrack=1&cset=tru>. Kunreuther, H. “The role of insurance in managing extreme events: Implications.” Business Economics. 37.2 (2002): 6-16. La Porte, Todd. “Governance and the Specter of Infrastructure Collapse.” The 8 th National Public Management Research Conference. Los Angeles: University of Southern California, 2005. 153 Larsson, G. and A. Enander. “Preparing for Disaster: Public Attitudes and Actions.” Disaster Prevention and Management. 6.1 (1997): 11–21. Loewenstein, G. F., et al. “Risk as feelings.” Psychological Bulletin. 127 (2001): 267- 286. Magee, Richard S. The Disposition Dilemma: Controlling the Release of Solid Materials from Nuclear Regulatory Commission-Licensed Facilities. Washington, DC: National Academy Press, 2002. Martonosi, Susan, David Ortiz and Henry Willis. Evaluating the Viability of 100 Per cent Container Inspection at America’s Ports. Santa Monica: RAND Corporation, 2005. McDaniels, T.L., et al. “Risk perception and the value of safety.” Risk Analysis. 12.4 (1992): 495-503. Medline Plus. “Medical Encyclopedia Radiation Sickness.” 19 February 2005. <http://www.nlm.nih.gov/medlineplus/ency/article/000026.htm>. Morrongiello, B. and H. Rennie. “Why do boys engage in more risk taking than girls? The role of attributions, beliefs, and risk appraisals.” Journal of Pediatric Psychology. 23.1 (1998): 33-43. National Council on Radiation Protection and Measurements (NCRP). “Management of Terrorist Events Involving Radiological Material.” NCRP 138 (2001). Plapp, T. and U. Werner. “Understanding risk perception from natural hazards: examples from Germany.” Eds. Amman, Dannenmann and Vulliet. RISK 21 – Coping with Risks due to Natural Hazards in the 21st Century. London: Taylor & Francis Group, 2006. Pratt, J. W., Raiffa, H., and Schlaifer, R.0. “The Foundations of Decision Under Uncertainty: An Elementary Exposition.” J. Am. Statist. Assoc. 59 (1964): 353-375. Pratt, J. W., Raiffa, H., and Schlaifer,R.0. Introduction to Statistical Decision Theory. McGraw-Hill, New York, 1965. Raiffa, H. Decision Analysis. Addison-Wesley, Reading, Mass, 1968. Reichmuth, B. “Economic consequences of a rad/nuc attack: Cleanup standards significantly affect cost.” Department of Homeland Security Conference on “Working Together: R&D Partnerships in Homeland Security.” Boston, Massachusetts: April 27– 28, 2005. 154 Rogers, G. “The Dynamics of Risk Perception: How Does Perceived Risk Respond to Risk Events?” Risk Analysis. 17:6 (1997): 745-757. Shapiro, Sidney A and Christopher H Schroeder. “Beyond Cost-Benefit Analysis: A Pragmatic Reorientation.” Harvard Environmental Law Review 32.2 (2008): 434-502. Sjöberg, L. “Factors in Risk Perception.” Risk Analysis. 20.1 (2000): 1-11. Sjöberg, L. “Limits of Knowledge and the Limited Importance of Trust.” Risk Analysis. 21 (2001): 189-198. Sjöberg, L. “Are Received Risk Perception Models Alive and Well?” Risk Analysis. 22.4 (2002): 665-70. Sjöberg, L. “Explaining Individual Risk Perception: The Case of Nuclear Waste.” Risk Management. 6.1 (2004): 51-64. Sjöberg, L. “The Perceived Risk of Terrorism.” Risk Management. 7.1 (2005): 43-61. Slovic, Paul, et al. “Rating the Risks.” Environment. 21.3 (1979): 14-20. Slovic, Paul, et al. “Images of disaster: Perception and acceptance of risks from nuclear power.” Eds. G. Goodman and W. Rowe. Energy Risk Management,. London: Academic Press, 1979. Slovic, Paul, et al. “Facts and Fears: Understanding perceived risk.” Eds. R. Schwing and W.A. Albers Jr. Societal Risk Assessment: How Safe is Safe Enough?. New York: Plenum, 1980. Slovic, Paul, et al. “Why Study Risk Perception?” Risk Analysis. 2.2 (1982): 83-93. Slovic, P., Fischhoff, B., & Lichtenstein, S. “Behavioral decision theory perspectives on risk and safety.” Acta Psychologica, 56 (1984): 183-203. Slovic, Paul, et al. “Modeling the societal impact of fatal accidents.” Management Science. 30.4 (1984): 464-474. Slovic, Paul. “Perception of risk: reflections on the psychometric paradigm.” Eds. S.Krimsky and D. Golding. Social theories of risk. Westport, CT: Praeger, 1992. 117- 152. Slovic, Paul. “Trust, emotion, sex, politics and science: Surveying the risk-assessment battlefield.” Eds. M. Bazerman, D. Messick, A. Tenbrunsel, & K. Wade-Benzoni. Environment, Ethics and Behavior. San Francisco, California: New Lexington Press, 1997. 277–313. 155 Slovic, P. & Peters, E. “The Importance of Worldviews in Risk Perception.” Risk Decision and Policy. 3 (1998): 165-170. Slovic, Paul and Elke Weber.“Perception of Risk Posed By Extreme Events.” Risk Management Strategies in an Uncertain World. Palisades, New York: April 12-13, 2002. Slovic, P. “Terrorism as Hazard: A New Species of Trouble.” Risk Analysis. 22.3 (2002): 425-426. Slovic, Paul, et al. “Risk as Analysis and Risk as Feelings: Some Thoughts about Affect, Reason, Risk, and Rationality.” Risk Analysis. 24.2 (2002): 1-12. Slovic, Paul, et al. “The affect heuristic.” European Journal of Operational Research. 177.3 (2007): 1333-1352. Smelser, Neil J. and Faith Mitchell, eds. Terrorism: Perspectives from the Behavioral and Social Sciences. Panel on Behavioral, Social, and Institutional Issues, Committee on Science and Technology for Countering Terrorism, Division of Behavioral and Social Sciences and Education. Washington, DC: National Academies Press, 2002. Sprinak, Ehud. “Rational Fanatics.” Foreign Policy. 120 (2000): 66-73. Svenson, O., et al. “Perceived driving safety and seatbelt usage.” Accident Analysis and Prevention. 17.2 (1985): 119-133. “Terrorist Attacks (within the United States or against Americans abroad).”Infoplease Page. June 2009. <http://www.infoplease.com/ipa/A0001454.html>. The Port of Long Beach. “About the Port—Port Statistics Index.” July 2009. <http://www.polb.com/about/facts.asp>. The Port of Los Angeles. “About the Port—FAQs.” July 2009. <http://www.portoflosangeles.org/newsroom/press_kit/facts.asp>. Ursano, Robert J., Carol S. Fullerton and Ann E. Norwood eds. Terrorism and Disaster: Individual and Community Mental Health Interventions. Cambridge, UK: Cambridge University Press, 2003. U.S. Government Accountability Office. “Combat Nuclear Smuggling: DHS Has Made Progress Deploying Radiation Detection Equipment at U.S. Ports-of-Entry, But Concerns Remain.” GAO-06-389, 2006. Victoroff, J. “The Mind of the Terrorist: A Review and Critique of Psychological Approaches.” Journal of Conflict Resolution, 49.1 (2005): 3-42. 156 Von Neumann, J., and Morgenstern, 0. Theory of Games and Economic Behavior, Ed. 2. Princeton University Press, Princeton, N.J., 1947. Von Winterfeldt, D. “Structuring Decision Problems for Decision Analysis.” Acta Psychol. 45 (1980): 71-93. Von Winterfeldt, Detlof, et al. “Cognitive Components of Risk Ratings.” Risk Analysis, 1.4 (1981): 277-287. Von Winterfeldt, Detlof and Ward Edwards. Decision Analysis and Behavioral Research. New York: Cambridge University Press, 1986. Weinberg, A. “Science and trans-science.” Minerva. 10.2(1972): 209-21. Willis, Henry, et al. Estimating Terrorism Risk. Santa Monica: RAND Corporation, 2005. Wittke, Carl. “Against the Current: The Life of Karl Heinzen (1809-80).” The American Historical Review, 50:815 (1945): 342. Woods, J. et al. “Terrorism Risk Perceptions and Proximity to Primary Terrorist Targets: How Close is Too Close?” Research in Human Ecology. 15.1 (2008): 63-70. 157 Appendix A: Risk Perception Study One Question 1: What is the probability of a terrorist attack in the United States in the next year? Question 2: How many people in the U.S. will be killed from terrorist attack(s) in the next year (select a number resulting in a 50/50 chance over or under your estimate)? Question 3: What is the most likely type of terrorist attack to occur in the U.S.? Question 4: The United States has spent a lot of money on homeland security since the attacks of 9/11. There also have been no additional attacks since this time. Do you attribute this to: (a) Terrorists having decided not to attack the U.S. because of fears of greater reprisal (b) Effective intelligence that has thwarted terrorist attacks (c) The effectiveness of the war on terror in Afghanistan and Iraq leaving terrorists running scared (d) Terrorist have been planning another significant attack and just taking time to prepare (e) Better measures toward homeland security having discouraged terrorists from attempting an attack 158 Question 5: Rate the overall risk of each scenario to you and your family on a scale from 0 (no risk) to 10 (high risk). Disaster Scenario NOTE: Presume all effects of the disaster scenarios take place in the United States. 1. Dam failure resulting in substantial flooding 2. Terrorists using a truck bomb attack a major tourist attraction 3. Shoulder-fired missile attack on commercial passenger airliner 4. Terrorists using a portable nuclear bomb attack a major metropolitan region 5. Disastrous category 5 hurricane on landfall 6. Terrorists carry out simultaneous massive explosions on subway cars and city buses 7. 9/11 copy-cats strike high-rise buildings with commercial passenger airliners 8. Terrorists release a poisonous chemical during an event at a large sports arena 9. Nuclear missile launched and exploded over major U.S. city 10. Avian (bird) flu outbreak in a major metropolitan region 11. Monstrous tsunami rolls in off the coast 12. Terrorists release of a lethal chemical into the regional drinking water supply 13. Terrorists cause a contagious smallpox epidemic in a heavily populated region 14. Massive wild fires spread into heavily populated residential communities 15. Nuclear power plant meltdown resulting in release of radioactive material 16. Terrorists release airborne anthrax into a heavily populated metropolitan region 17. Deadly toxic gas release following an industrial plant explosion 18. Major earthquake on the San Andreas Fault 19. Terrorists detonate an explosive device surrounded by radioactive material (also known as a 'dirty bomb') 20. Terrorists contaminate the regional food supply with a deadly chemical (No risk) (High risk) (Moderate risk) 0 1 2 3 4 5 6 7 8 9 10 159 Question 6: Rate the overall risk of each scenario to society at large (across the United States) on a scale from 0 (no risk) to 10 (high risk). Disaster Scenario NOTE: Presume all effects of the disaster scenarios take place in the United States. 1. Dam failure resulting in substantial flooding 2. Terrorists using a truck bomb attack a major tourist attraction 3. Shoulder-fired missile attack on commercial passenger airliner 4. Terrorists using a portable nuclear bomb attack a major metropolitan region 5. Disastrous category 5 hurricane on landfall 6. Terrorists carry out simultaneous massive explosions on subway cars and city buses 7. 9/11 copy-cats strike high-rise buildings with commercial passenger airliners 8. Terrorists release a poisonous chemical during an event at a large sports arena 9. Nuclear missile launched and exploded over major U.S. city 10. Avian (bird) flu outbreak in a major metropolitan region 11. Monstrous tsunami rolls in off the coast 12. Terrorists release of a lethal chemical into the regional drinking water supply 13. Terrorists cause a contagious smallpox epidemic in a heavily populated region 14. Massive wild fires spread into heavily populated residential communities 15. Nuclear power plant meltdown resulting in release of radioactive material 16. Terrorists release airborne anthrax into a heavily populated metropolitan region 17. Deadly toxic gas release following an industrial plant explosion 18. Major earthquake on the San Andreas Fault 19. Terrorists detonate an explosive device surrounded by radioactive material (also known as a 'dirty bomb') 20. Terrorists contaminate the regional food supply with a deadly chemical (No risk) (High risk) (Moderate risk) 0 1 2 3 4 5 6 7 8 9 10 160 Question 7: Rate the likelihood of each disastrous event occurring in the next year using a percentage scale from 0 (no likelihood) to 100 (most likely). Disaster Scenario NOTE: Presume all effects of the disaster scenarios take place in the United States. 1. Dam failure resulting in substantial flooding 2. Terrorists using a truck bomb attack a major tourist attraction 3. Shoulder-fired missile attack on commercial passenger airliner 4. Terrorists using a portable nuclear bomb attack a major metropolitan region 5. Disastrous category 5 hurricane on landfall 6. Terrorists carry out simultaneous massive explosions on subway cars and city buses 7. 9/11 copy-cats strike high-rise buildings with commercial passenger airliners 8. Terrorists release a poisonous chemical during an event at a large sports arena 9. Nuclear missile launched and exploded over major U.S. city 10. Avian (bird) flu outbreak in a major metropolitan region 11. Monstrous tsunami rolls in off the coast 12. Terrorists release of a lethal chemical into the regional drinking water supply 13. Terrorists cause a contagious smallpox epidemic in a heavily populated region 14. Massive wild fires spread into heavily populated residential communities 15. Nuclear power plant meltdown resulting in release of radioactive material 16. Terrorists release airborne anthrax into a heavily populated metropolitan region 17. Deadly toxic gas release following an industrial plant explosion 18. Major earthquake on the San Andreas Fault 19. Terrorists detonate an explosive device surrounded by radioactive material (also known as a 'dirty bomb') 20. Terrorists contaminate the regional food supply with a deadly chemical (No likelihood) (Most likely to occur) (Moderate likelihood) 0 1 2 3 4 5 6 7 10 8 9 161 Question 8: What is your estimate of the number of fatalities resulting from the worst case disastrous event? Disaster Scenario NOTE: Presume all effects of the disaster scenarios take place within the United States. 1. Dam failure resulting in substantial flooding 2. Terrorists using a truck bomb attack a major tourist attraction 3. Shoulder-fired missile attack on commercial passenger airliner 4. Terrorists using a portable nuclear bomb attack a major metropolitan region 5. Disastrous category 5 hurricane on landfall 6. Terrorists carry out simultaneous massive explosions on subway cars and city buses 7. 9/11 copy-cats strike high-rise buildings with commercial passenger airliners 8. Terrorists release a poisonous chemical during an event at a large sports arena 9. Nuclear missile launched and exploded over major U.S. city 10. Avian (bird) flu outbreak in a major metropolitan region 11. Monstrous tsunami rolls in off the coast 12. Terrorists release of a lethal chemical into the regional drinking water supply 13. Terrorists cause a contagious smallpox epidemic in a heavily populated region 14. Massive wild fires spread into heavily populated residential communities 15. Nuclear power plant meltdown resulting in release of radioactive material 16. Terrorists release airborne anthrax into a heavily populated metropolitan region 17. Deadly toxic gas release following an industrial plant explosion 18. Major earthquake on the San Andreas Fault 19. Terrorists detonate an explosive device surrounded by radioactive material (also known as a 'dirty bomb') 20. Terrorists contaminate the regional food supply with a deadly chemical 162 Q9: What is the probability (from 0% to 100%) of each ‘attack type’ occurring? Attack Type Probability 1. Suicide bomb attack ____________ 2. Nuclear attack ____________ 3. Chemical attack ____________ 4. Product (food and water) contamination ____________ 5. Bioterror attack ____________ 6. Conventional explosive attack ____________ 7. Radiological attack ____________ 8. Hand grenade ____________ 9. Hijacking ____________ 10. Kidnapping ____________ 11. Missile attack ____________ 12. Shooting ____________ 13. Arson ____________ Q10: What is the probability (from 0% to 100%) of each ‘target type’ being targeted for a terrorist attack? Target Type Probability 1. Major U.S. landmark ____________ 2. Transportation node (e.g. train or port) ____________ 3. Tourist attraction ____________ 4. Plant/Factory ____________ 5. Marketplace/restaurant/shopping mall ____________ 6. Hotel ____________ 7. Government building ____________ 8. Financial building ____________ 9. Place of worship ____________ 10. School/University ____________ 11. Ship ____________ 12. Pipeline/Powerline ____________ 163 Appendix B: Risk Perception Study Two A Study of the Perceived Risk Associated with Terrorist Events Question 1: Rate your familiarity and knowledge of each terrorist event using a scale from 0 (Not familiar) to 100 (Very familiar). Terrorist Event NOTE: Presume all effects of the terrorist events take place in the United States. 1. Terrorists using a truck bomb attack a major tourist attraction 2. Shoulder-fired missile attack on commercial passenger airliner 3. Terrorists using a portable nuclear bomb attack a major metropolitan region 4. Terrorists carry out simultaneous massive explosions on subway cars and city buses 5. 9/11 copy-cats strike high-rise buildings with commercial passenger airliners 6. Terrorists release a poisonous chemical during an event at a large sports arena 7. Nuclear missile launched and exploded over major U.S. city 8. Terrorists release of a lethal chemical into the regional drinking water supply 9. Terrorists cause a contagious smallpox epidemic in a heavily populated region 10. Terrorists release airborne anthrax into a heavily populated metropolitan region 11. Terrorists detonate an explosive device surrounded by radioactive material (also known as a 'dirty bomb') 12. Terrorists contaminate the regional food supply with a deadly chemical (Not familiar ) (Very familiar) (Somewhat familiar) 0 10 20 30 40 50 60 70 80 90 100 164 Question 2: Rate terrorists’ capability to successfully carry out each terrorist event using a percentage scale from 0 (Not capable) to 100 (Very capable). Terrorist Event NOTE: Presume all effects of the terrorist events take place in the United States. 1. Terrorists using a truck bomb attack a major tourist attraction 2. Shoulder-fired missile attack on commercial passenger airliner 3. Terrorists using a portable nuclear bomb attack a major metropolitan region 4. Terrorists carry out simultaneous massive explosions on subway cars and city buses 5. 9/11 copy-cats strike high-rise buildings with commercial passenger airliners 6. Terrorists release a poisonous chemical during an event at a large sports arena 7. Nuclear missile launched and exploded over major U.S. city 8. Terrorists release of a lethal chemical into the regional drinking water supply 9. Terrorists cause a contagious smallpox epidemic in a heavily populated region 10. Terrorists release airborne anthrax into a heavily populated metropolitan region 11. Terrorists detonate an explosive device surrounded by radioactive material (also known as a 'dirty bomb') 12. Terrorists contaminate the regional food supply with a deadly chemical (Not capable ) (Very familiar) (Somewhat capable) 0 10 20 30 40 50 60 70 80 90 100 165 Question 3: Rate how much you dread each terrorist event using a scale from 0 (No dread) to 100 (Very dreaded). Terrorist Event NOTE: Presume all effects of the terrorist events take place in the United States. 1. Terrorists using a truck bomb attack a major tourist attraction 2. Shoulder-fired missile attack on commercial passenger airliner 3. Terrorists using a portable nuclear bomb attack a major metropolitan region 4. Terrorists carry out simultaneous massive explosions on subway cars and city buses 5. 9/11 copy-cats strike high-rise buildings with commercial passenger airliners 6. Terrorists release a poisonous chemical during an event at a large sports arena 7. Nuclear missile launched and exploded over major U.S. city 8. Terrorists release of a lethal chemical into the regional drinking water supply 9. Terrorists cause a contagious smallpox epidemic in a heavily populated region 10. Terrorists release airborne anthrax into a heavily populated metropolitan region 11. Terrorists detonate an explosive device surrounded by radioactive material (also known as a 'dirty bomb') 12. Terrorists contaminate the regional food supply with a deadly chemical (No dread) (Very dreaded) (Some dread) 0 10 20 30 40 50 60 70 80 90 100 166 Question 4: Rate how disastrous each terrorist event would be in terms of political, economic, psychological, and casualty outcomes using a scale from 0 (Not disastrous) to 100 (Very disastrous). Terrorist Event NOTE: Presume all effects of the terrorist events take place in the United States. 1. Terrorists using a truck bomb attack a major tourist attraction 2. Shoulder-fired missile attack on commercial passenger airliner 3. Terrorists using a portable nuclear bomb attack a major metropolitan region 4. Terrorists carry out simultaneous massive explosions on subway cars and city buses 5. 9/11 copy-cats strike high-rise buildings with commercial passenger airliners 6. Terrorists release a poisonous chemical during an event at a large sports arena 7. Nuclear missile launched and exploded over major U.S. city 8. Terrorists release of a lethal chemical into the regional drinking water supply 9. Terrorists cause a contagious smallpox epidemic in a heavily populated region 10. Terrorists release airborne anthrax into a heavily populated metropolitan region 11. Terrorists detonate an explosive device surrounded by radioactive material (also known as a 'dirty bomb') 12. Terrorists contaminate the regional food supply with a deadly chemical (Not disastrous) (Very disastrous) (Somewhat disastrous) 0 10 20 30 40 50 60 70 80 90 100 167 Question 5: Rate the overall risk of each terrorist event to you and your family on a scale from 0 (No risk) to 100 (High risk). Terrorist Event NOTE: Presume all effects of the terrorist events take place in the United States. 1. Terrorists using a truck bomb attack a major tourist attraction 2. Shoulder-fired missile attack on commercial passenger airliner 3. Terrorists using a portable nuclear bomb attack a major metropolitan region 4. Terrorists carry out simultaneous massive explosions on subway cars and city buses 5. 9/11 copy-cats strike high-rise buildings with commercial passenger airliners 6. Terrorists release a poisonous chemical during an event at a large sports arena 7. Nuclear missile launched and exploded over major U.S. city 8. Terrorists release of a lethal chemical into the regional drinking water supply 9. Terrorists cause a contagious smallpox epidemic in a heavily populated region 10. Terrorists release airborne anthrax into a heavily populated metropolitan region 11. Terrorists detonate an explosive device surrounded by radioactive material (also known as a 'dirty bomb') 12. Terrorists contaminate the regional food supply with a deadly chemical (No risk) (High risk) (Moderate risk) 0 10 20 30 40 50 60 70 80 90 100 168 Question 6: Rate the likelihood of each terrorist event occurring in the next year using a percentage scale from 0 (No likelihood) to 100 (Most likely). Terrorist Event NOTE: Presume all effects of the terrorist events take place in the United States. 1. Terrorists using a truck bomb attack a major tourist attraction 2. Shoulder-fired missile attack on commercial passenger airliner 3. Terrorists using a portable nuclear bomb attack a major metropolitan region 4. Terrorists carry out simultaneous massive explosions on subway cars and city buses 5. 9/11 copy-cats strike high-rise buildings with commercial passenger airliners 6. Terrorists release a poisonous chemical during an event at a large sports arena 7. Nuclear missile launched and exploded over major U.S. city 8. Terrorists release of a lethal chemical into the regional drinking water supply 9. Terrorists cause a contagious smallpox epidemic in a heavily populated region 10. Terrorists release airborne anthrax into a heavily populated metropolitan region 11. Terrorists detonate an explosive device surrounded by radioactive material (also known as a 'dirty bomb') 12. Terrorists contaminate the regional food supply with a deadly chemical (No likelihood) (Most likely to occur) (Moderate likelihood) 0 10 20 30 40 50 60 70 80 90 100 169 Question 7: What is your estimate of the number of fatalities (deaths) resulting from a “worst case” terrorist event? Terrorist Event NOTE: Presume all effects of the terrorist events take place within the United States. 1. Terrorists using a truck bomb attack a major tourist attraction 2. Shoulder-fired missile attack on commercial passenger airliner 3. Terrorists using a portable nuclear bomb attack a major metropolitan region 4. Terrorists carry out simultaneous massive explosions on subway cars and city buses 5. 9/11 copy-cats strike high-rise buildings with commercial passenger airliners 6. Terrorists release a poisonous chemical during an event at a large sports arena 7. Nuclear missile launched and exploded over major U.S. city 8. Terrorists release of a lethal chemical into the regional drinking water supply 9. Terrorists cause a contagious smallpox epidemic in a heavily populated region 10. Terrorists release airborne anthrax into a heavily populated metropolitan region 11. Terrorists detonate an explosive device surrounded by radioactive material (also known as a 'dirty bomb') 12. Terrorists contaminate the regional food supply with a deadly chemical 170 Appendix C: Risk Perception Study Three Question 1: What is the probability of a terrorist attack in the United States in the next year? Question 2: How many people in the U.S. will be killed from terrorist attack(s) in the next year (select a number resulting in a 50/50 chance over or under your estimate)? Question 3: What is the most likely type of terrorist attack to occur in the U.S.? Question 4: The United States has spent a lot of money on homeland security since the attacks of 9/11. There also have been no additional attacks since this time. Do you attribute this to: (a) Terrorists having decided not to attack the U.S. because of fears of greater reprisal (b) Effective intelligence that has thwarted terrorist attacks (c) The effectiveness of the war on terror in Afghanistan and Iraq leaving terrorists running scared (d) Terrorist have been planning another significant attack and just taking time to prepare (e) Better measures toward homeland security having discouraged terrorists from attempting an attack 171 Question 5: Rate the overall risk of each scenario to you and your family on a scale from 0 (no risk) to 10 (high risk). Disaster Scenario NOTE: Presume all effects of the disaster scenarios take place in the United States. 1. Dam failure resulting in substantial flooding 2. Terrorists using a truck bomb attack a major tourist attraction 3. Shoulder-fired missile attack on commercial passenger airliner 4. Terrorists using a portable nuclear bomb attack a major metropolitan region 5. Disastrous category 5 hurricane on landfall 6. Terrorists carry out simultaneous massive explosions on subway cars and city buses 7. 9/11 copy-cats strike high-rise buildings with commercial passenger airliners 8. Terrorists release a poisonous chemical during an event at a large sports arena 9. Nuclear missile launched and exploded over major U.S. city 10. Avian (bird) flu outbreak in a major metropolitan region 11. Monstrous tsunami rolls in off the coast 12. Terrorists release of a lethal chemical into the regional drinking water supply 13. Terrorists cause a contagious smallpox epidemic in a heavily populated region 14. Massive wild fires spread into heavily populated residential communities 15. Nuclear power plant meltdown resulting in release of radioactive material 16. Terrorists release airborne anthrax into a heavily populated metropolitan region 17. Deadly toxic gas release following an industrial plant explosion 18. Major earthquake on the San Andreas Fault 19. Terrorists detonate an explosive device surrounded by radioactive material (also known as a 'dirty bomb') 20. Terrorists contaminate the regional food supply with a deadly chemical (No risk) (High risk) (Moderate risk) 0 1 2 3 4 5 6 7 8 9 10 172 Question 6: Rate the overall risk of each scenario to society at large (across the United States) on a scale from 0 (no risk) to 10 (high risk). Disaster Scenario NOTE: Presume all effects of the disaster scenarios take place in the United States. 1. Dam failure resulting in substantial flooding 2. Terrorists using a truck bomb attack a major tourist attraction 3. Shoulder-fired missile attack on commercial passenger airliner 4. using a portable nuclear bomb attack a major metropolitan region 5. Disastrous category 5 hurricane on landfall 6. Terrorists carry out simultaneous massive explosions on subway cars and city buses 7. 9/11 copy-cats strike high-rise buildings with commercial passenger airliners 8. Terrorists release a poisonous chemical during an event at a large sports arena 9. Nuclear missile launched and exploded over major U.S. city 10. Avian (bird) flu outbreak in a major metropolitan region 11. Monstrous tsunami rolls in off the coast 12. Terrorists release of a lethal chemical into the regional drinking water supply 13. Terrorists cause a contagious smallpox epidemic in a heavily populated region 14. Massive wild fires spread into heavily populated residential communities 15. Nuclear power plant meltdown resulting in release of radioactive material 16. Terrorists release airborne anthrax into a heavily populated metropolitan region 17. Deadly toxic gas release following an industrial plant explosion 18. Major earthquake on the San Andreas Fault 19. Terrorists detonate an explosive device surrounded by radioactive material (also known as a 'dirty bomb') 20. Terrorists contaminate the regional food supply with a deadly chemical (No risk) (High risk) (Moderate risk) 0 1 2 3 4 5 6 7 8 9 10 173 Question 7: Rate the likelihood of each disastrous event occurring in the next year using a percentage scale from 0 (no likelihood) to 100 (most likely). Disaster Scenario NOTE: Presume all effects of the disaster scenarios take place in the United States. 1. Dam failure resulting in substantial flooding 2. Terrorists using a truck bomb attack a major tourist attraction 3. Shoulder-fired missile attack on commercial passenger airliner 4. Terrorists using a portable nuclear bomb attack a major metropolitan region 5. Disastrous category 5 hurricane on landfall 6. Terrorists carry out simultaneous massive explosions on subway cars and city buses 7. 9/11 copy-cats strike high-rise buildings with commercial passenger airliners 8. Terrorists release a poisonous chemical during an event at a large sports arena 9. Nuclear missile launched and exploded over major U.S. city 10. Avian (bird) flu outbreak in a major metropolitan region 11. Monstrous tsunami rolls in off the coast 12. Terrorists release of a lethal chemical into the regional drinking water supply 13. Terrorists cause a contagious smallpox epidemic in a heavily populated region 14. Massive wild fires spread into heavily populated residential communities 15. Nuclear power plant meltdown resulting in release of radioactive material 16. Terrorists release airborne anthrax into a heavily populated metropolitan region 17. Deadly toxic gas release following an industrial plant explosion 18. Major earthquake on the San Andreas Fault 19. Terrorists detonate an explosive device surrounded by radioactive material (also known as a 'dirty bomb') 20. Terrorists contaminate the regional food supply with a deadly chemical (No likelihood) (Most likely to occur) (Moderate likelihood) 0 10 20 30 40 50 60 70 80 90 100 174 Question 8: What is your estimate of the number of fatalities resulting from the worst case disastrous event? Disaster Scenario NOTE: Presume all effects of the disaster scenarios take place within the United States. 1. Dam failure resulting in substantial flooding 2. Terrorists using a truck bomb attack a major tourist attraction 3. Shoulder-fired missile attack on commercial passenger airliner 4. Terrorists using a portable nuclear bomb attack a major metropolitan region 5. Disastrous category 5 hurricane on landfall 6. Terrorists carry out simultaneous massive explosions on subway cars and city buses 7. 9/11 copy-cats strike high-rise buildings with commercial passenger airliners 8. Terrorists release a poisonous chemical during an event at a large sports arena 9. Nuclear missile launched and exploded over major U.S. city 10. Avian (bird) flu outbreak in a major metropolitan region 11. Monstrous tsunami rolls in off the coast 12. Terrorists release of a lethal chemical into the regional drinking water supply 13. Terrorists cause a contagious smallpox epidemic in a heavily populated region 14. Massive wild fires spread out into heavily populated residential communities 15. Nuclear power plant meltdown resulting in release of radioactive material 16. Terrorists release airborne anthrax into a heavily populated metropolitan region 17. Deadly toxic gas release following an industrial plant explosion 18. Major earthquake on the San Andreas Fault 19. Terrorists detonate an explosive device surrounded by radioactive material (also known as a 'dirty bomb') 20. Terrorists contaminate the regional food supply with a deadly chemical 175 Question 9: Rate your familiarity with each disastrous event using a scale from 0 (Not familiar) to 100 (Very familiar). Disaster Scenario NOTE: Presume all effects of the disaster scenarios take place in the United States. 1. Dam failure resulting in substantial flooding 2. Terrorists using a truck bomb attack a major tourist attraction 3. Shoulder-fired missile attack on commercial passenger airliner 4. Terrorists using a portable nuclear bomb attack a major metropolitan region 5. Disastrous category 5 hurricane on landfall 6. Terrorists carry out simultaneous massive explosions on subway cars and city buses 7. 9/11 copy-cats strike high-rise buildings with commercial passenger airliners 8. Terrorists release a poisonous chemical during an event at a large sports arena 9. Nuclear missile launched and exploded over major U.S. city 10. Avian (bird) flu outbreak in a major metropolitan region 11. Monstrous tsunami rolls in off the coast 12. Terrorists release of a lethal chemical into the regional drinking water supply 13. Terrorists cause a contagious smallpox epidemic in a heavily populated region 14. Massive wild fires spread into heavily populated residential communities 15. Nuclear power plant meltdown resulting in release of radioactive material 16. Terrorists release airborne anthrax into a heavily populated metropolitan region 17. Deadly toxic gas release following an industrial plant explosion 18. Major earthquake on the San Andreas Fault 19. Terrorists detonate an explosive device surrounded by radioactive material (also known as a 'dirty bomb') 20. Terrorists contaminate the regional food supply with a deadly chemical (Not familiar) (Very familiar ) (Somewhat familiar) 0 10 20 30 40 50 60 70 80 90 100 176 Question 10: Rate the terrorist’s capability to carry out each disastrous event using a percentage scale from 0 (Not capable) to 100 (Very capable). Disaster Scenario NOTE: Presume all effects of the disaster scenarios take place in the United States. 1. Dam failure resulting in substantial flooding 2. Terrorists using a truck bomb attack a major tourist attraction 3. Shoulder-fired missile attack on commercial passenger airliner 4. Terrorists using a portable nuclear bomb attack a major metropolitan region 5. Disastrous category 5 hurricane on landfall 6. Terrorists carry out simultaneous massive explosions on subway cars and city buses 7. 9/11 copy-cats strike high-rise buildings with commercial passenger airliners 8. Terrorists release a poisonous chemical during an event at a large sports arena 9. Nuclear missile launched and exploded over major U.S. city 10. Avian (bird) flu outbreak in a major metropolitan region 11. Monstrous tsunami rolls in off the coast 12. Terrorists release of a lethal chemical into the regional drinking water supply 13. Terrorists cause a contagious smallpox epidemic in a heavily populated region 14. Massive wild fires spread into heavily populated residential communities 15. Nuclear power plant meltdown resulting in release of radioactive material 16. Terrorists release airborne anthrax into a heavily populated metropolitan region 17. Deadly toxic gas release following an industrial plant explosion 18. Major earthquake on the San Andreas Fault 19. Terrorists detonate an explosive device surrounded by radioactive material (also known as a 'dirty bomb') 20. Terrorists contaminate the regional food supply with a deadly chemical (Not capable) (Very capable) (Somewhat capable) 0 10 20 30 40 50 60 70 80 90 100 177 Question 11: Rate how much each disastrous event scares you using a scale from 0 (Not scared) to 100 (Very scared). Disaster Scenario NOTE: Presume all effects of the disaster scenarios take place in the United States. 1. Dam failure resulting in substantial flooding 2. Terrorists using a truck bomb attack a major tourist attraction 3. Shoulder-fired missile attack on commercial passenger airliner 4. Terrorists using a portable nuclear bomb attack a major metropolitan region 5. Disastrous category 5 hurricane on landfall 6. Terrorists carry out simultaneous massive explosions on subway cars and city buses 7. 9/11 copy-cats strike high-rise buildings with commercial passenger airliners 8. Terrorists release a poisonous chemical during an event at a large sports arena 9. Nuclear missile launched and exploded over major U.S. city 10. Avian (bird) flu outbreak in a major metropolitan region 11. Monstrous tsunami rolls in off the coast 12. Terrorists release of a lethal chemical into the regional drinking water supply 13. Terrorists cause a contagious smallpox epidemic in a heavily populated region 14. Massive wild fires spread into heavily populated residential communities 15. Nuclear power plant meltdown resulting in release of radioactive material 16. Terrorists release airborne anthrax into a heavily populated metropolitan region 17. Deadly toxic gas release following an industrial plant explosion 18. Major earthquake on the San Andreas Fault 19. Terrorists detonate an explosive device surrounded by radioactive material (also known as a 'dirty bomb') 20. Terrorists contaminate the regional food supply with a deadly chemical (Not scared) (Very scared) (Somewhat scared) 0 10 20 30 40 50 60 70 80 90 100 178 Question 12: Rate how disastrous each event might be using a scale from 0 (Not disastrous) to 100 (Very disastrous). Disaster Scenario NOTE: Presume all effects of the disaster scenarios take place in the United States. 1. Dam failure resulting in substantial flooding 2. Terrorists using a truck bomb attack a major tourist attraction 3. Shoulder-fired missile attack on commercial passenger airliner 4. Terrorists using a portable nuclear bomb attack a major metropolitan region 5. Disastrous category 5 hurricane on landfall 6. Terrorists carry out simultaneous massive explosions on subway cars and city buses 7. 9/11 copy-cats strike high-rise buildings with commercial passenger airliners 8. Terrorists release a poisonous chemical during an event at a large sports arena 9. Nuclear missile launched and exploded over major U.S. city 10. Avian (bird) flu outbreak in a major metropolitan region 11. Monstrous tsunami rolls in off the coast 12. Terrorists release of a lethal chemical into the regional drinking water supply 13. Terrorists cause a contagious smallpox epidemic in a heavily populated region 14. Massive wild fires spread into heavily populated residential communities 15. Nuclear power plant meltdown resulting in release of radioactive material 16. Terrorists release airborne anthrax into a heavily populated metropolitan region 17. Deadly toxic gas release following an industrial plant explosion 18. Major earthquake on the San Andreas Fault 19. Terrorists detonate an explosive device surrounded by radioactive material (also known as a 'dirty bomb') 20. Terrorists contaminate the regional food supply with a deadly chemical (Not disastrous) (Very disastrous) (Somewhat disastrous) 0 10 20 30 40 50 60 70 80 90 100 179 Questions 13-17. The next few questions ask about how much you have changed the way you do various life activities since 9/11/2001 because of the possibility of future terrorism in the U.S. As a result of the potential for future terrorism, how have you changed in terms of . . . Please Mark ONE Box on Each Line. Questions 18-21. The next few questions ask about how much you have changed the way you do various life activities since 9/11/2001 because of the possibility of future terrorism in the U.S. As a result of the potential for future terrorism, how have you changed in terms of . . . Please Mark ONE Box on Each Line. I do not do it anymore I do it much less I do it somewhat less I do it a little less I do it about the same 13. flying on commercial airplanes? 14. using public transportation, such as subways, buses, or commuter trains? 15. going to public places, such as malls, restaurants, or sports stadiums? 16. voting in national or local elections? 17. interacting with others who are of Middle Eastern or Arab descent? Not at All A Little Somewhat Very Much An Extreme Amount 18. vacationing, such as in selecting some places over others? 19. working or going to school, such as in staying away from large cities or skyscrapers? 20. deciding on places to live, such as avoiding cities or high-rise apartment buildings? 21. whether you watch or read less on the TV, newspaper, or internet that is about terrorism? 180 The following are EIGHT demographic questions. 21. What is your gender? ____ Male ____Female 22. What is your age? ___ 18 to 24 years ___25 to 34 years ___35 to 44 years ___45 to 54 years ___55 to 64 years ___65 years and over 23. What is your race? ___ White/Caucasian ___African American ___Hispanic ___Asian ___Native American ___Pacific Islander ___Other: ___No Response 24. What is your religious affiliation? ___Protestant Christian ___Roman Catholic ___Evangelical Christian ___Jewish ___Muslim ___Hindu ___Buddhist ___Other: ___No Response 25. What is your marital status? ___Single never married __Widowed ___Divorced/separated ___Married 26. What is the highest level of education you have completed? ___ Less than high school ___High School/GED ___Some College ___2 yr College Degree ___4 yr College Degree ___Master’s Degree ___Doctoral Degree ___Professional Degree (JD/MD) 27. What is your annual income range? ___ Below $20,000 ___$20K-$29,999 ___$30K-$39,999 ___$40K-$49,999 ___ $50K-$59,999 ___$60K-$69,999 ___$70K-$79,999 ___$80K-$89,999 ___$90K or more ___No Response 28. In which state do you currently reside? ___________________________________

Abstract (if available)

Abstract Risk has been characterized as a function of a potential threat, vulnerability to the threat, and the consequences were the threat to be carried out. In the context of terrorism, threats are the individuals who might wage an attack against a specific target

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Making terrorism risk assessments more useful for decision-making

PDF

A robust model for disaster risk and decision analysis

PDF

A multiattribute decision model for the selection of radioisotope and nuclear detection devices

PDF

The influence of risk on decision-making during walking

PDF

Threat assessment of physical attacks of electrical infrastructure: how do terrorist groups select their targets?

PDF

Data use in middle schools: a multiple case study of three middle schools’ experiences with data-driven decision making

PDF

Beyond the numbers chase: how urban high school teachers make sense of data use

PDF

Environmental ABC's -- acting on behalf of children: rethinking representation, risk and responsibility in US environmental policy implementation

PDF

Information sharing, deliberation, and collective decision-making: A computational model of collaborative governance

PDF

How urban school superintendents effectively use data-driven decision making to improve student achievement

PDF

Macroeconomic conditions, systematic risk factors, and the time series dynamics of commercial mortgage credit risk

PDF

A two-level analysis of foreign policy decision making: an empirical investigation of the case of China-Taiwan

PDF

Resilience by anticipating change: simple and robust decision making for coastal adaptation planners, communities and elected officials

PDF

Decisions to ratify the Kyoto Protocol: a Latin American perspective on poliheuristic theory

PDF

A risk analysis methodology to address human and organizational factors in offshore drilling safety: with an emphasis on negative pressure test

PDF

Emerging technology and its impact on security policy

PDF

Challenging migrant worker policies in Korea: settlement and local citizenship

PDF

Beyond regulation: special improvement districts, design review, and place-making in New Jersey

PDF

Homegrown violent extremism: designing a community-based model to reduce the risk of recruitment and radicalization

PDF

Collaborating to provide microfinance to caregivers of orphans and vulnerable children in Ethiopia

Asset Metadata

Creator Rosoff, Heather Beth (author)

Core Title Using decision and risk analysis to assist in policy making about terrorism

School School of Policy, Planning, and Development

Degree Doctor of Public Administration

Degree Program Public Policy

Publication Date 08/07/2009

Defense Date 05/20/2009

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag decision analysis,Decision making,OAI-PMH Harvest,policy analysis,risk analysis,terrorism

Place Name California (states), Long Beach (city or populated place), Los Angeles (city or populated place)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor von Winterfeldt, Detlof (committee chair), Graddy, Elizabeth A. (committee member), John, Richard S. (committee member)

Creator Email hrosoff5@yahoo.com,rosoff@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-m2540

Unique identifier UC1136608

Identifier etd-Rosoff-2890 (filename),usctheses-m40 (legacy collection record id),usctheses-c127-177451 (legacy record id),usctheses-m2540 (legacy record id)

Legacy Identifier etd-Rosoff-2890.pdf

Dmrecord 177451

Document Type Dissertation

Rights Rosoff, Heather Beth

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Repository Name Libraries, University of Southern California

Repository Location Los Angeles, California

Repository Email cisadmin@lib.usc.edu