Evidence-Based Counterterrorism Policy

Springer Series on Evidence-Based Crime Policy Series Editors: Lawrence W. Sherman, University of Pennsylvania Heather ...

Author: Cynthia Lum | Leslie W. Kennedy

130 downloads 1349 Views 6MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!

Report copyright / DMCA form

DOWNLOAD PDF

Springer Series on Evidence-Based Crime Policy Series Editors: Lawrence W. Sherman, University of Pennsylvania Heather Strang, Australian National University

Crime prevention and criminal justice policies are domains of great and growing importance around the world. Despite the rigorous research done in this field, policy decisions are often based more on ideology or speculation than on science. One reason for this may be a lack of comprehensive presentations of the key research affecting policy deliberations. While scientific studies of crime prevention and criminal policy have become more numerous in recent years, they remain widely scattered across a wide range of journals and countries The Springer Series on Evidence-Based Crime Policy aims to pull this evidence together while presenting new research results. This combination in each book should provide, between two covers (or in electronic searches), the best evidence on each topic of crime policy. The series will publish primary research on crime policies and criminal justice practices, raising critical questions or providing guidance to policy change. The series will try to make it easier for research findings to become key components in decisions about crime and justice policy. The editors welcome proposals for both monographs and edited volumes. There will be a special emphasis on studies using rigorous methods (especially field experiments) to assess crime prevention interventions in areas such as policing, corrections, juvenile justice, and crime prevention. Published in Cooperation with the Campbell Crime and Justice Group

For further volumes: http://www.springer.com/series/8396

Cynthia Lum • Leslie W. Kennedy Editors

Evidence-Based Counterterrorism Policy

Editors: Cynthia Lum Center for Evidence-Based Crime Policy Department of Criminology, Law & Society George Mason University Fairfax, VA, USA [email protected]

Leslie W. Kennedy Rutgers School of Criminal Justice Rutgers University Newark, NJ, USA [email protected]

ISBN 978-1-4614-0952-6 e-ISBN 978-1-4614-0953-3 DOI 10.1007/978-1-4614-0953-3 Springer New York Dordrecht Heidelberg London Library of Congress Control Number: 2011936394 © Springer Science+Business Media, LLC 2012 All rights reserved. This work may not be translated or copied in whole or in part without the written permission of the publisher (Springer Science+Business Media, LLC, 233 Spring Street, New York, NY 10013, USA), except for brief excerpts in connection with reviews or scholarly analysis. Use in connection with any form of information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed is forbidden. The use in this publication of trade names, trademarks, service marks, and similar terms, even if they are not identified as such, is not to be taken as an expression of opinion as to whether or not they are subject to proprietary rights. Printed on acid-free paper Springer is part of Springer Science+Business Media (www.springer.com)

Contents

Part I 1

Evidence-Based Counterterrorism Policy ........................................... Cynthia Lum and Leslie W. Kennedy

Part II 2

3

4

Introduction

Data Sources for Evaluating Terrorism and Counterterrorism

Assessing and Comparing Data Sources for Terrorism Research ......................................................................... Ivan Sascha Sheehan

13

Generating Terrorism Event Databases: Results from the Global Terrorism Database, 1970 to 2008 .............. Gary LaFree

41

Evidence-Based Intelligence Practices: Examining the Role of Fusion Centers as a Critical Source of Information .............. Jeremy Carter and Steven Chermak

65

Part III

Methodological Innovations for Counterterrorism Policy

5

Innovative Methods for Terrorism and Counterterrorism Data ....... Michael D. Porter, Gentry White, and Lorraine Mazerolle

6

Introducing Group-Based Trajectory Analysis and Series Hazard Modeling: Two Innovative Methods to Systematically Examine Terrorism Over Time............................... Laura Dugan and Sue-Ming Yang

7

3

A Complexity Method for Assessing Counterterrorism Policies ..................................................................... Claudio Cioffi-Revilla

91

113

151

v

vi

8

9

10

Contents

Analyzing Terrorism Using Spatial AnalysisTechniques: A Case Study of Turkish Cities ............................................................. Danielle M. Rusnak, Leslie W. Kennedy, Ibrahim S. Eldivan, and Joel M. Caplan The Importance of Instrument Validity in Evaluating Security Screening Programs ........................................ Tracy E. Costigan

187

Translational Criminology: Using Existing Evidence for Assessing TSA’s Comprehensive Security Strategy at Airports ................................................................ Cynthia Lum, Charlotte Gill, Breanne Cave, Julie Hibdon, and David Weisburd

209

Part IV 11

12

13

14

15

Perspectives in Evaluating Counterterrorism Policy

Terrorist Finance, Informal Markets, Trade and Regulation: Challenges of Evidence Regarding International Efforts .................. Nikos Passas

255

Evaluating the Legal Challenges and Effects of Counterterrorism Policy ................................................................... Linda M. Merola

281

Public Opinion Research and Evidence-Based Counterinsurgency ................................................................................. Clay Ramsay

301

Counterinsurgency and Criminology: Applying Routine Activities Theory to Military Approaches to Counterterrorism ......................................................... Breanne Cave Toughness vs. Fairness: Police Policies and Practices for Managing the Risk of Terrorism .................................................... Tom R. Tyler

Part V 16

167

325

353

Epilogue

The Next Steps: A Need for a Research Infrastructure for Evaluating Counterterrorism ................................ Cynthia Lum and Leslie W. Kennedy

367

Biographies .....................................................................................................

377

Index ................................................................................................................

381

Contributors

Joel M. Caplan, Ph.D School of Criminal Justice, Rutgers University, Newark, NJ, USA Rutgers Center on Public Security, Rutgers University, Newark, NJ, USA Jeremy Carter, Ph.D Department of Criminology and Criminal Justice, University of North Florida, Jacksonville, FL, USA Breanne Cave, M.A Center for Evidence-Based Crime Policy, Department of Criminology, Law & Society, George Mason University, Fairfax, VA, USA Steven Chermak, Ph.D School of Criminal Justice, Michigan State University, East Lansing, MI, USA Claudio Cioffi-Revilla, D.Sc.Pol., Ph.D Center for Social Complexity and Department of Computational Social Science, Krasnow Institute for Advanced Study, George Mason University, Fairfax, VA, USA Tracy E. Costigan, Ph.D American Institutes for Research (AIR), Washington, DC, USA Laura Dugan, Ph.D Department of Criminology and Criminal Justice, University of Maryland, College Park, MD, USA Ibrahim S. Eldivan School of Criminal Justice, Rutgers University, Newark, NJ, USA Rutgers Center on Public Security, Rutgers University, Newark, NJ, USA Charlotte Gill, Ph.D Center for Evidence-Based Crime Policy, Department of Criminology, Law & Society, George Mason University, Fairfax, VA, USA Julie Hibdon, Ph.D Center for Evidence-Based Crime Policy, Department of Criminology, Law & Society, George Mason University, Fairfax, VA, USA

vii

viii

Contributors

Leslie W. Kennedy, Ph.D School of Criminal Justice and the Center on Public Security, Rutgers University, Newark, NJ, USA Gary LaFree, Ph.D National Center for the Study of Terrorism and Responses to Terrorism (START), and the Department of Criminology and Criminal Justice, University of Maryland, College Park, MD, USA Cynthia Lum, Ph.D Center for Evidence-Based Crime Policy, Department of Criminology, Law & Society, George Mason University, Fairfax, VA, USA Lorraine Mazerolle, Ph.D ARC-Centre of Excellence in Policing and Security, Institute for Social Science Research, The University of Queensland, Brisbane, Queensland, Australia Linda M. Merola, J.D., Ph.D Department of Criminology, Law & Society, George Mason University, Fairfax, VA, USA Nikos Passas, Ph.D School of Criminology and Criminal Justice, Northeastern University, Boston, MA, USA Michael D. Porter, Ph.D GeoEye Analytics, McLean, VA, USA Clay Ramsay, Ph.D Program on International Policy Attitudes (PIPA) and the Center for International and Security Studies (CISSM), University of Maryland, College Park, MD, USA Danielle M. Rusnak, M.A School of Criminal Justice, Rutgers University, Newark, NJ, USA Ivan Sascha Sheehan, Ph.D Negotiation and Conflict Management, School of Public and International Affairs, University of Baltimore, Baltimore, MD, USA Tom R. Tyler, Ph.D Department of Psychology, New York University, New York, NY, USA David Weisburd, Ph.D Center for Evidence-Based Crime Policy, Department of Criminology, Law & Society, George Mason University, Fairfax, VA, USA Gentry White, Ph.D ARC-Centre of Excellence in Policing and Security, Institute for Social Science Research, The University of Queensland, Brisbane, Queensland, Australia Sue-Ming Yang, Ph.D Institute of Criminology, National Chung Cheng University, Chia-Yi, Taiwan

Part I Introduction

Chapter 1

Evidence-Based Counterterrorism Policy Cynthia Lum and Leslie W. Kennedy

As we approach the tenth anniversary of the terror attacks on September 11th, one fact is clear among the many unknowns about terrorism: there has been an exponential increase in spending on counterterrorism measures. For the United States, this increase in spending is not just reflected in federal homeland security measures or military efforts in Iraq, Afghanistan, and Pakistan, but also among private individuals, corporations, and public entities at the municipal, county, regional, and state levels. Most recently, the killing of Osama Bin Laden by the United States has led to a renewed increase in resource allocation at home by local law enforcement agencies (van Natta, 2011). In the scientific fields, there has also been more funding for a variety of academic and technological research and development related to terrorism and counterterrorism. An addendum to this fact: This massive increase in capacity building has not been matched by evaluation and assessment regarding the cost-effectiveness of those expenditures. The many efforts to detect, prevent, deter, and reduce the risk of terrorism-related violence at home and aboard have rarely been scientifically evaluated using rigorous methods. Lum, Kennedy, and Sherley (2006) discovered this state of the research when conducting a Campbell Collaboration systematic review of counterterrorism research. The Campbell review sought to determine “what we know” about the types of counterterrorism strategies and tactics that can reduce the likelihood of, or damage from, terrorism events. However, after reviewing thousands of research articles on terrorism and counterterrorism, the comprehensive search only produced seven evaluation studies of at least moderate scientific rigor, primarily advanced by a few scholars (see, e.g., Enders & Sandler, 1993, 2000,

C. Lum (*) Center for Evidence-Based Crime Policy, Department of Criminology, Law & Society, George Mason University, Fairfax, VA, USA e-mail: [email protected] L.W. Kennedy School of Criminal Justice, Rutgers University, Newark, NJ, USA C. Lum and L.W. Kennedy (eds.), Evidence-Based Counterterrorism Policy, Springer Series on Evidence-Based Crime Policy 3, DOI 10.1007/978-1-4614-0953-3_1, © Springer Science+Business Media, LLC 2012

3

4

C. Lum and L.W. Kennedy

2006; Enders, Sandler, & Cauley, 1990). Most of the analysis focused on the use of metal detectors at airports to prevent hijacking. Some research also suggested that interventions (e.g., military ones) could increase terrorism in the short run, while in the long run have little significant effect on levels of terroristic violence. Lum and colleagues also emphasized that the only way to determine the costeffectiveness of counterterrorism interventions, as with crime prevention more generally, was through scientifically rigorous evaluations. Yet evaluation research on counterterrorism interventions in the 10 years since September 11th (and 5 years after the Lum et al. review) remains dismal compared to crime prevention research or even other areas of terrorism studies (i.e., etiological or case studies on terrorism or groups). Lum recently reported to the National Academies that evaluations of police interventions outnumber those on security and counterterrorism more than tenfold (National Research Council, 2010). Only recently has the Transportation Security Agency (TSA) requested more scientific and comprehensive assessments of interventions to ensure airport security (see Lum et al., this volume). Setting aside evaluations of outcome effectiveness, there are also very few studies of the implementation of security measures to understand how they work and how their functioning could be enhanced. Modern democracies are supposed to be accountable for their use of public resources, and the lack of a balance between implementation and assessment reflects negatively on them. How nations choose to prevent terroristic violence both at home and abroad, and how they evaluate those interventions, matters not only to outcomes but also to how citizens view the legitimacy of government actions. Thus, the decision to commit large amounts of funds to antiterrorism strategies must be coupled with the responsibility of assessing whether programs are effective, ineffective, or harmful – that is, interventions should be evidence-based. The term “evidencebased” – as a descriptor for policy and decision-making – means that choices to implement interventions, like those that attempt to counter terrorism, are based on scientific and analytic knowledge that rigorously examines their impact on outcomes (Cullen & Gendreau, 2000; Davies, Nutley, & Smith, 2000; MacKenzie, 2000; Nutley & Davies, 1999; Sherman, 1998; Sherman, Farrington, Welsh, & MacKenzie, 2002; Weisburd, Lum, & Petrosino, 2001). This approach to decisionmaking not only provides both scientific and fiscal justification for programs but also appeals ethically (see Chalmers, 2003), and can temper rash responses to crises and moral panics (Lum, 2009). In theory, the principles underpinning evidence-based policy can also support counterterrorism (Lum & Koper, 2011). These principles include pursuing methodologically rigorous evaluations of interventions; systematically reviewing research; disseminating, translating, and using research to inform practice; engaging in partnerships that foster evaluation (i.e., between practitioners and researchers); and expanding the collection of high-quality data. In practice, especially with concerns such as terrorism and random violence, there are many challenges to fostering such an approach (see Petrosino, Boruch, Soydan, Duggan, & Sanchez-Meca, 2001; Van Brunschot & Kennedy, 2008). While this collection is governed by the presumption that such assessments are not only scientifically valid, useful, and

1 Evidence-Based Counterterrorism Policy

5

necessary, but also possible, the reality of such an assumption is questionable. Even the most liberal governments are resistant to self-evaluation, especially regarding subject matters (terrorism) and actions (counterterrorism), that bureaucrats may believe are beyond the reach of assessment because of national security concerns (whether real or imagined). This tug-of-war between high-quality governance and national security lies at the heart of the current lack of evaluations of counterterrorism policy, and it is likely one major reason why research in etiology of terrorism has thrived, while evaluation research in counterterrorism has not. The chapters within this volume, individually and collectively, underscore the principles of an evidence-based approach while exploring these challenges to scientific evaluation of counterterrorism interventions. Specifically, we assembled a set of papers that together would move toward two goals of evidence-based crime policy. First, the papers would speak to the basic issues of creating an evaluation and scientific base for counterterrorism interventions. Second, each chapter would contribute to research innovation in this area by expanding and creating depth in our thinking about evaluating counterterrorism measures.

Innovation in Research Three strategies can create innovation in research for subjects (like counterterrorism) that need a stronger evidence base. First, the range, depth, and quality of available data to measure outcomes and to evaluate interventions can be scrutinized and improved. Observations and data, whether qualitative or quantitative, are the foundations of analysis and provide important access into both theoretical understanding and assessment. Second, innovation in research also is provoked by challenging and developing the methods used in evaluation. The use of the myriad of available rigorous scientific methods, from modeling to experimentation, helps us see complex relationships, including causal ones that are not immediately obvious. And finally, research innovation occurs when new perspectives are offered. Idea creation can help inspire others, including new generations of researchers, to consider alternative ways of looking at problems. Such innovation can expand the evidence base of topics that seem difficult for science to touch. The Data, Methods, and Perspectives sections of this volume reflect this research innovation and provide reasonable, scientific, and neutral suggestions in improving the evidence base of counterterrorism. When discussing terrorism data, LaFree and Sheehan detail and compare the utility of major terrorism data sources, including the Global Terrorism Database (GTD). While the types of data available are interesting, there are two important takeaway points from both chapters regarding how we interpret terrorism trends across time: How the data is collected and also the motivations behind that data collection matter when measuring the effects of a counterterrorism intervention. Further, unlike in crime prevention studies, where data are relatively plentiful and provided with fewer obstacles, terrorism research relies on a more tenuous supply of data, whose sources are often shrouded in secrecy,

6

C. Lum and L.W. Kennedy

sometimes unnecessarily. As Sheehan notes, validation between datasets can also be difficult, as the definitions, political motivations, and quality of efforts behind different data collection efforts can vary significantly. While the National Consortium for the Study of Terrorism and Responses to Terrorism (START Center)1 at the University of Maryland has made major strides in the collection and open publication of terrorism incident data, as LaFree notes, other types of information remain more mysterious. Sheehan provides a comparison across terrorism data sources in his chapter. Related to the discussion of developing data sources is data sharing for prevention purposes, which has become a major concern post September 11th. Carter and Chermak discuss both the advantages and challenges in intelligence flow, sharing and use of preventative data, and analysis with fusion centers. Their survey indicates that while there is much interest in cooperation among law enforcement entities, the quality, sharing, and use of fusion center data remain less well established, and the effectiveness of fusion centers still remains unevaluated. All three chapters emphasize one major point: that the scrutiny of terrorism-related data and their use is a central exercise in increasing our knowledge of the impact of counterterrorism efforts. Even with better availability and quality of data, determining how to analyze rare events and evaluate interventions that attempt to prevent rare events are major challenges to evidence-based counterterrorism, as the section on methodology indicates. Terrorism event trend analysis can be used in evaluating counterterrorism (see Lum et al., 2006), and four articles in this section focus specifically on exploring new methods to evaluate terrorism trends over time. Porter’s change-point analysis shows how combining both frequency and impact of attacks presents important nuances that can result in different conclusions about the nature of terrorism and counterterrorism over time than previously thought. Dugan and Yang’s group-based trajectory approach also provides an alternative to previously used interrupted time series and vector autoregressive models. They argue that trajectory modeling can reveal important variations in group activity that may not be seen using other approaches. Cioffi-Revilla, using a social complexity framework, suggests that the nonnormal distribution of terrorism over time requires rethinking models of terroristic attacks and how to counter them. Finally, Rusnak and her colleagues shift the focus to spatial trends, looking for the factors in the environment (in this case major features in Turkey) that would, when combined, increase the risk of terrorism attack. This approach provides a basis of examining the environmental backdrop of terrorism, making forecasts of events less dependent on previous incidents and more informed by the important correlates of the event. The fact that terrorism is considered a rare event often leads to a focus on statistical modeling of the effects of counterterrorism interventions as opposed to other evaluation methods, such as, experimental designs. However, as Lum et al. and Costigan discuss in their chapters, experimental evaluation and assessment, as well as the concerns about testing and assessing the instrument validity of specific security measures and interventions, are also important for those building the evidence-base

1

See http://www.start.umd.edu/start/

1 Evidence-Based Counterterrorism Policy

7

of counterterrorism interventions. Costigan’s detailed discussion of instrument validity and reliability for screening systems to prevent air travel violence is an important step in evaluating the effectiveness of screening technologies, especially since these are widely used. Lum and colleagues, also focusing on air transportation security, suggest a “translational criminological” approach to preliminary evaluation. That is, if evaluations of existing practices are not readily available to judge programs, or if evaluations are not immediately possible, perhaps a way to preliminarily assess, and then to build an agenda for evaluation, is to apply existing criminological knowledge to counterterrorism measures in a systematic way. All of the discussions in the methods section of this volume present interesting and unique approaches and considerations if more opportunities for evaluation arise for counterterrorism interventions. The contributions in the data and methods section of this volume all point to the need for more creative thinking to encourage the building of an evidence base for counterterrorism evaluation. However, new and creative ideas still must fall within the confines of good science and structured thinking about developing facts and knowledge. To bring this point home, we offer a series of articles that discuss fresh, alternative, and varied perspectives on scientific approaches to knowledge about counterterrorism interventions. These chapters provide a unique addition to the counterterrorism literature and are intended to broaden thinking about types of counterterrorism interventions that can be evaluated, as well as from what perspective they might be assessed. For example, Passas’ chapter shows how a traditional emphasis on examining financial institutions has led to a general neglect of scientific evaluation of other measures, such as counterterrorism finance. Examining informal markets, as well as pursuing trade transparency and regulation, may be just as, if not more, important as a counterterrorism measure than law enforcement activities. Ramsay’s and Cave’s articles both focus on what might be considered a very specific type of counterterrorism intervention – counterinsurgency and military interventions – to quell political violence and stabilize states. Ramsay and Cave apply common social science methods (surveys) and theories (opportunity, routine activities) respectively, and by doing so, they emphasize the contradictions to theory building, evaluation, and research given the unique subject matters at hand (insurgency and counterinsurgency). Indeed, terrorism and political violence often contradict traditional moral perspectives of the etiology and (lack of) support for violent crimes, which suggests that countering these types of social violence requires rethinking the scope of traditional social science theories. Whether survey or polling research can assist with such evaluations is, as Ramsay discusses, itself an open question. The legitimacy of the state and the law is a tenuous assumption in times of terrorism and insurgency. Merola advises the legal and law and society communities that much can be gained from adding social science and evaluation research approaches to understanding the impact of legal changes brought on by counterterrorism. As Tyler suggests, being more sensitive to the complexities of social relationships and also society’s relations with the state and legal processes when engaging in prevention approaches can make the difference between evaluating

8

C. Lum and L.W. Kennedy

success or failure, especially in future support for terroristic violence as a means to an end. These ideas regarding data, methods, and perspectives are not specific to terrorism, but are generally important in criminal justice, policing, and security policies. And, similar to policing and security, moving forward in counterterrorism policy requires a stronger focus on the measurement and evaluation of the effectiveness of interventions; the rationale behind evaluation; lessons learned from other areas of research; the obstacles to good evaluation; and the resources needed to affect these changes. Given the range of issues covered in this volume, where does this lead us in developing a future agenda around evaluation? In our epilogue, we outline ideas for developing a research infrastructure that might support more generation of research evidence in this area. Research infrastructures are necessary across many research agendas and provide support for conducting research and promoting the use of science in practice. This is especially the case with counterterrorism policy, as we still have many opportunities to test and assess interventions as they are in their early stages of development.

References Chalmers, I. (2003). Trying to do more good than harm in policy and practice: The role of rigorous, transparent, up-to-date evaluations. Annals of the American Academy of Political and Social Sciences, 589, 22–40. Cullen, F. T., & Gendreau, P. (2000). Assessing correctional rehabilitation: Policy, practice, and prospects. In J. Horney (Ed.), Policies, processes, and decisions of the criminal justice system: Criminal justice 3 (pp. 109–175). Washington: National Institute of Justice. Davies, H. O., Nutley, S., & Smith, P. C. (2000). What works: Evidence-based policy and practice in public services. London: Policy Press. Enders, W., & Sandler, T. (1993). The effectiveness of antiterrorism policies: A vector-autoregression-intervention analysis. The American Political Science Review, 87(4), 829–844. Enders, W., & Sandler, T. (2000). Is transnational terrorism becoming more threatening? Journal of Conflict Resolution, 44, 307–332. Enders, W., & Sandler, T. (2006). The political economy of terrorism. New York: Cambridge University Press. Enders, W., Sandler, T., & Cauley, J. (1990). UN conventions, terrorism, and retaliation in the fight against terrorism: An econometric evaluation. Terrorism and Political Violence, 2(1), 83–105. Lum, C. (2009). Translating police research into practice. Ideas in American Policing. Washington: Police Foundation. Lum, C., Kennedy, L., & Sherley, A. (2006). Are counter-terrorism strategies effective? The results of the Campbell Systematic Review on counter-terrorism evaluation research. Journal of Experimental Criminology, 2(4), 489–516. Lum, C., & Koper, C. (2011). Is crime prevention relevant to counter-terrorism? In B. Forst, J. Greene, & J. Lynch (Eds.), Criminologists on terrorism and homeland security. New York: Cambridge University Press. MacKenzie, D. L. (2000). Evidence-based corrections: Identifying what works. Crime and Delinquency, 48(4), 457–471.

1 Evidence-Based Counterterrorism Policy

9

National Research Council, The National Academies. (2010). Field evaluation in the intelligence and counterintelligence context: Workshop summary. Washington: The National Academies Press. Nutley, S., & Davies, H. T. O. (1999). The Fall and Rise of Evidence in Criminal Justice. Public Money and Management, 19, 47–54. Petrosino, A., Boruch, R., Soydan, H., Duggan, L., & Sanchez-Meca, J. (2001). Meeting the challenges of evidence-based crime policy: The Campbell Collaboration. Annals of the American Academy of Political and Social Sciences, 578, 14–34. Sherman, L. W. (1998). Evidence-based policing. Second Invitational Lecture on Ideas in Policing. Washington: Police Foundation. Sherman, L. W., Farrington, D. P., Welsh, B. C., & MacKenzie, D. L. (Eds.). (2002). Evidence based crime prevention. London: Routledge. van Brunschot, E., & Kennedy, L. W. (2008). Risk balance and security. Thousand Oaks: Sage. van Natta, D. (2011). Cities nationwide heighten vigilance on terror. The New York Times. Retrieved May 11, 2011, from http://www.nytimes.com/2011/05/14/us/14threat.html?_r=2&hpw. Weisburd, D., Lum, C., & Petrosino, A. (2001). Does research design affect study outcomes in criminal justice? Annals of the American Academy of Political and Social Sciences, 578, 50–70.

Part II

Data Sources for Evaluating Terrorism and Counterterrorism

Chapter 2

Assessing and Comparing Data Sources for Terrorism Research Ivan Sascha Sheehan

Introduction A journey of a thousand miles begins with a single step. Lao Tzu

Much of the early work on terrorism was based on “small-n” qualitative case studies. Little by little narrative chronologies were added. Today, large-n quantitative databases of terrorist events containing thousands and even tens of thousands of events and a wide range of variables are only a click way on the Internet. These databases have provided enormous opportunities for terrorism researchers to identify cases and test hypotheses that are relevant to the field. But how good is the quality of the data? And how should terrorism researchers go about choosing between competing datasets? One of the assumptions behind this chapter is that insights gained from the study of small-n data may be relevant to the development of standards to assess and compare large-n terrorism datasets. Ever since Geddes’ paper on “how the cases you choose affect the answers you get” (Geddes, 1990), small-n qualitative researchers have quite self-consciously tried to improve the quality of small-n data. One result has been the growth of a large body of scholarship around the concept of “best practices” norms and standards to maximize transparency, reliability, and validity in this kind of data (Brady & Collier, 2004). At the same time recognition has been growing that large-n datasets, including data on political events and processes, are often riddled with the same problems that plague small-n data. Collier, Brady, and Seawright (2004), for example, have drawn attention to the problems large and small-n researchers both face in making contextually sensitive judgments in terms of coding. Others have gone a step further suggesting that large-n datasets should be able to convey the kind of “detailed knowledge I.S. Sheehan (*) School of Public and International Affairs, University of Baltimore, Baltimore, MD, USA e-mail: [email protected] C. Lum and L.W. Kennedy (eds.), Evidence-Based Counterterrorism Policy, Springer Series on Evidence-Based Crime Policy 3, DOI 10.1007/978-1-4614-0953-3_2, © Springer Science+Business Media, LLC 2012

13

14

I.S. Sheehan

and sensitivity to context” that are the hallmarks and strength of case-oriented studies (Munck & Verkuilen, 2002). Still others have argued that improving data quality in databases such as Polity IV requires heightened critical attention to questions normally raised by qualitative researchers on subjects such as data construction and what Herrera and Devesh call “data supply chains”: Who produced the data? Why? What were the producer’s incentives and capabilities? Did they work as an independent agency or were they influenced by external actors? (2007, p.366)

These observations have led to increased calls for shared standards to evaluate small-n and large-n research. In this context, John Gerring (2001, 2010) introduced the concept of a “criterial framework” that bridges small-n/large-n, qualitative/quantitative chasms. Drawing on this concept, Evan Lieberman (2010) has shown that normative criteria commonly associated with the evaluation of small-n studies (e.g., citation transparency and handling of issues of uncertainty) can potentially improve political datasets such as Polity IV, the Annual Survey of Freedom (Freedom House index), the Minorities at Risk dataset (MAR), and the Uppsala Conflict Data Program dataset (UCDP). In this chapter I will build on the concept of a “criterial framework” by extending it to large-n terrorism datasets and by proposing a best practices framework to help users evaluate the validity and reliability of a range of terrorism datasets. I will begin by discussing why we need norms or best practice standards to compare and evaluate quantitative terrorism data sources. I will then make a case for extending the concept of a “criterial framework,” such as the one described by Gerring (2001, 2010), to terrorism data. I will highlight challenges and problems that occur in applying such a framework to five of the most well-known and respected terrorism events datasets in the field. All five of the databases selected are translations of narrative records (usually news reports) into numerical data in the form of counts, indexes, or dummy variables indicating the presence or absence of a phenomenon related to a coded terrorism event. In selecting the databases I have chosen ones that are publically available on the Internet and could be considered elite or the best in the field. I will conclude with proposals for implementing best practices in terrorism databases.

Developing Standards for Terrorism Data Why Do We Need Them? There are practical as well as methodological reasons for developing standards to assess and compare large-n terrorism datasets.

Practical Reasons Terrorism datasets are unique in that most of them were first developed, maintained, and used outside universities in the intelligence and defense communities. First

2

Assessing and Comparing Data Sources for Terrorism Research

15

published in 1975, the RAND Terrorism Chronologies, which later became the basis for the well-known RAND-MIPT terrorism database, now subsumed into the Rand Database of Worldwide Terrorism Incidents (RDTWI), were originally developed by policy analysts such as Brian Michael Jenkins under a defense department grant for intelligence purposes (Fowler, 1981). The RAND chronologies were subsequently used by CIA analyst, Edward Mickolus, to produce the International Terrorism: Attributes of Terrorist Events dataset (ITERATE), a dataset that later formed the basis for the CIA’s annual terrorism report, Patterns of Global Terrorism (Schmid, 1983, p. 257). Much of the Global Terrorism Database (GTD) originated with reports made by Pinkerton Global Intelligence Service (PGIS), a private global security firm (La Free, 2010), and the World Incident Tracking System (WITS) is a replacement for the previous annual Patterns of Terrorism report put out by the U.S. National Terrorism and Counter Terrorism Center (NCTC) (Wigle, 2010). Because of these origins, outside an academic environment, terrorism data were not always subjected to the kinds of rigorous norms in terms of collection or coding that are usually expected in academia (Schmid, 2004). This situation led to considerable embarrassment when two Princeton scholars reviewed the data tables at the end of the State Department’s annual Patterns of Global Terrorism report for 2003 and found that the numbers in the tables did not add up and that the conclusion of the report, namely that global terrorism had decreased that year, was in error and that terrorism had actually increased. When they subsequently published that information in an op-ed piece in the Washington Post (Krueger & Laitin, 2004a) and in an article in Foreign Affairs (Krueger & Laitin, 2004b), the State Department admitted that the report was wrong and retracted it. Today large numbers of the users of terrorism data, however, are academic scholars. Many of these academics were nurtured in programs that emphasized the importance of best practices in collecting data and while they may agree with Gerring that the objects of social science often “refuse to lie still in the manner of rocks, animals, cells and atoms” (Gerring, 1999, p. 393) there is a much greater expectation that terrorism data should be based on solid norms and that it should be subjected to evaluation and questioning. Terrorism datasets differ from other political and social science data in another important way. Since much of the data is derived from media sources in real time, and since its developers have frequently used different definitions and coding rules, no one dataset is completely comprehensive or exhaustive and there is a great deal of variability across datasets. For example, in a previous comparison of transnational terrorism events data from two terrorism databases, ITERATE and RANDMIPT, this author found several large discrepancies in quarterly events counts for the time period 1993–2004. As shown in Fig. 2.1, there were distinct differences in counts at the outset of the series in 1993 and again between 2001 and 2002 and in 2004. A likely explanation for these differences was that the databases operationalized what constituted a “transnational” event very differently. In general, transnational terrorist events are viewed as ones that involve perpetrators and victims from different countries.1 However, what constitutes a country may be differently defined

16

I.S. Sheehan

250 ITERATE RAND

200

150

100

50

0 1993 1994

1995

1996 1997

1998

1999 2000

2001 2002

2003 2004

Fig. 2.1 ITERATE vs. RAND-MIPT: quarterly number of transnational terrorist incidents (1993–2004). ITERATE International Terrorism: Attributes of Terrorist Events; RAND-MIPT RAND-Memorial Institute for Prevention of Terrorism Database

when a territory is still disputed. A likely explanation for the discrepancy at the outset of the series in 1993 is that ITERATE counted more events that started in one of the newly formed Soviet states and ended in another or targeted people as transnational whereas RAND excluded those events under the assumption that they were not yet transnational. By contrast, closer inspection of the data for the period 2001– 2002 and for 2004 suggests that RAND counted more incidents associated with

1

When a terrorist incident in one country involves victims, targets, institutions, or citizens of another country, it is considered transnational or international and is included. The 9/11 hijackings, for example, are included as transnational terrorist incidents for at least three reasons. First, the perpetrators came from outside of the United States. Second, the victims were from over 80 countries. And third, the incidents had worldwide economic and security ramifications. The bombings of the US embassies in Kenya and Tanzania on August 7, 1998, as well as the suicide car bombings aimed at British and Jewish targets in Istanbul, Turkey on November 20, 2003, are similarly included as transnational terrorist incidents since they involve perpetrators and victims from different countries. On the other hand, the bombing of the Murrah Federal Building in Oklahoma City by Timothy McVeigh is not included since it is considered to be a purely domestic event. Similarly, bombings by the IRA in Northern Ireland are not included as transnational terrorist acts. However, IRA attacks in England would be included.

2

Assessing and Comparing Data Sources for Terrorism Research

180

GTD

160

RAND

17

WITS

140 120 100 80 60 40 20 0 2004

2005

2006

2007

2008

2009

2010

Fig. 2.2 GTD, RAND, and WITS (2004–2008): quarterly number of suicide attacks. GTD Global Terrorism Database; RAND Rand Database of Worldwide Terrorism Incidents (RDTWI); WITS Worldwide Incidents Tracking System

Palestinian uprisings against Israel (Intifada) as transnational whereas ITERATE, making the assumption that such events were domestic, did not include them (Sheehan, 2007, 2009). Similar disparities are evident in more recent data. Figure 2.2, for example, shows the quarterly number of suicide attacks for more recent data on international and domestic terrorism events from three current terrorism datasets (the GTD, RAND, and WITS) for the time period 2004–2008. In this case the RAND and WITS datasets both show much higher frequencies of suicide attacks at almost every time point compared to the GTD. Disparities like these can be a source of consternation for terrorism researchers. Just looking at the plots, we do not know if the disparities are a function of differences between the two databases in the ways they define terrorism, in the sources they use, in coding rules, or something else. And without more information, it is not possible to tell if one or the other dataset capturing more “true” events or including more “false” ones. Part of the problem is that terrorism databases often tap into different data sources and information. The plot below (Fig. 2.3) of international terrorist incidents from ITERATE and RAND-MIPT for the years 1993–2004 illustrates this point. The plot above shows that unique incidents, ones covered in only one of the two databases, outnumbered overlapping ones at almost every quarterly period shown. Overall for the time period, overlapping incidents constituted only about one third

18

I.S. Sheehan

250 Overlapping 200

Unique

150

100

50

0 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004

Fig. 2.3 ITERATE vs. RAND-MIPT (1993–2004): overlapping and unique international terrorist incidents. ITERATE International Terrorism: Attributes of Terrorist Events; RAND-MIPT RANDMemorial Institute for Prevention of Terrorism Database

of all incidents. Overlapping terrorist incidents are the ones the databases agree on. In general they are more likely to be reported and more likely to be deadly (Sheehan, 2007). However, they constitute a relatively small proportion of all incidents. This may be because the databases have selection biases in terms of sources, because they use different operational definitions of terrorism, because they have different inclusion or exclusion rules, or simply because the enormity of capturing so much data in real time is so great that each database is only able to cover a segment of terrorism events. Unique events, however, constitute larger portions of each dataset and they are responsibile for the variability in terrorism datasets. The point is that from a practical viewpoint to assess the quality and comparability of information across terrorism databases, researchers need some benchmarks or normative criteria.

Theoretical Reasons There are also good theoretical reasons for developing best practices criteria for terrorism data. As Alex Schmid observes, scholars seek out terrorism data for a number of reasons. They may want to use the data to uncover underlying patterns or

2

Assessing and Comparing Data Sources for Terrorism Research

19

trends in terrorism, to compare terrorist campaigns cross-nationally and over time, to predict future events, to examine the causes, concomitants, or consequences of terrorism in relation to other phenomena, or to evaluate the success of counterterrorism (Schmid, 2004). To meet these different requirements, terrorism databases should ideally fulfill a number of criteria. In the best of worlds, it should be relevant and transparent. The definition of terrorism should be consistent over time and across regions. In addition, the data should be replicable and reliable and the validity and integrity of observations should be such that they can be checked (Drakos, 2009). Unfortunately, terrorism data like much of what Colin Robson (2002) calls “real world” data rarely lives up to these expectations. However, to the extent that a terrorism database considers and at least tries to approach some norms and standards, we can have greater confidence in its credibility. The existence of norms and standards, moreover, can be a guide to particular researchers in their search for the best database for a particular project.

Developing Standards for Terrorism Data The Case for a “Criterial Framework” Users of large-n datasets usually look for tests of validity and reliability to evaluate a given dataset. Unfortunately, traditional tests of validity and reliability are problematic for terrorism data. Validity tests typically rely on a gold standard, but there is no gold standard for terrorism data. There is not even a universally accepted definition of terrorism. By one count there are as many as 109 definitions of terrorism (Schmid & Jongman, 1988). Reliability tests typically depend on consistency, but terrorism databases are often viewed as “living databases” (Wigle, 2010) that can be changed, even retrospectively, as new information emerges, as perspectives change or as the operational definition of terrorism is revised (Paull, 1982; Schmid, 1983; Wilkinson, 1986; Reid, 1997).2 This means that inter-coder reliability tests, even if conducted at the time of collection of a discrete item of data (e.g., a terrorist event from a report in the media), are rendered meaningless and the conditions for test– retest may not exist. To complicate matters, terrorism databases have relied almost

2

In the 1970s and 1980s, it was not at all uncommon for terrorism databases to redefine and reclassify terrorist incidents for political reasons (Paull, 1982, p. 46; Schmid, 1983, p. 260). This practice led to wide disparities in the annual Patterns of International Terrorism report across the Ford, Carter, and Reagan administrations. For example, although only eight types of incidents were classified as terrorism under Ford, as many as 17 were classified as terrorism under Reagan. Reclassification, moreover, was applied retrospectively. As a result the 1980 Patterns of Global Terrorism report estimated that the number of worldwide terrorism incidents for the period 1968–1980 was 6,714 although only a year earlier it estimated that terrorism incidents for approximately the same time period 1968–1979 were half that number or 3,336 (Wilkinson, 1986, p.44, cited in Reid, 1997).

20

I.S. Sheehan

exclusively on reports in the news media (Fowler, 1981) and all too often such reports have been accepted unquestioningly despite their known biases and unreliability (Wardlaw, 1982; Wilkinson, 1986; Herman & Chomsky, 1988). Alternative criteria, developed for small-n research may be well suited to evaluating terrorism data with its naturalistic roots in “real world” inquiry and its emergent qualities (openness to adaptation and change as new information emerges). The mindset that informs these criteria is also well suited to terrorism data. Although large-n research is typically guided by a positivist belief that a real world of objects apart from people exists out there and that researchers can accurately describe it and compare their descriptions with this objective reality, this is not the case with small-n research. Small-n researchers, coming from post-positivist, constructionist, and interpretivist traditions, are more likely to take the position that researchers can only know the world from their own perspective of it.3 One implication of this view is that researchers’ and informants’ own preconceptions, values, and biases are relevant to all phases of research and are critical to its credibility. Another implication of this view is that knowledge (the truth) can only be approximated, never fully known, and uncertainty needs to be acknowledged. This mindset may be particularly helpful in evaluating terrorism data where, because of the clandestine nature of the phenomenon, objective information is not always available. At the same time, small-n researchers, mindful of the importance of validity, have developed criteria such as credibility, transferability, generalizability, and dependability that parallel the concepts of validity and reliability (see Guba, 1981). This situation has led to increasing recognition that shared norms for designing and evaluating research can be built across the two approaches (Brady & Collier, 2004). The concept of a criterial framework was introduced and developed by Gerring (2001, 2010) as a means of helping social science researchers bridge small-n/large-n and other divides (qualitative/quantitative, positivist/interpretivist) to find common ground in designing research. It has since been extended to evaluating large-n political data (Lieberman, 2010). One of the advantages of applying such a framework large-n data is that it has the potential to bring rich detailed descriptions that can serve as an alternative to traditional validity and reliability testing when such testing is not possible.

Proposed Criteria In this section, I propose six criteria derived from small-n research that could be used to evaluate and compare terrorism datasets. To make them easier to remember, I have chosen words that all start with the letter C. They include Conceptual clarity, Context and immediacy of observation, Citation transparency, Coding

3

It should be noted that these approaches are not uniform. Post-positivists are more likely to accept the “objective” nature of reality than interpretivists and constructionists.

2

Assessing and Comparing Data Sources for Terrorism Research

21

Table 2.1 Criterial framework for evaluating terrorism databases Criterion Rationale Assessment Conceptual The relevance of a database Is the definition of terrorism used in clarity depends on the definition(s) construction of the database of terrorism used specified? Context and Data collected from primary and Do the authors report the context and immediacy of secondary sources are often immediacy of the observations? observations more valued than those collected from tertiary sources Citation The replicability of a dataset Are the actual sources of the data transparency depends on citation described? Are clear references to transparency original data provided? Coding The reliability of a dataset Do the authors provide a codebook? consistency depends on coding Do they discuss how they resolve consistency over time and coding conflicts and make across raters decisions in ambiguous cases? Certainty The validity of observations Are contradictory facts reported? depends on certainty. Contradictions and ambiguity in data should be reported Conflict of The integrity of observations can Are funding sources and other interest issues be compromised in the potential conflicts of interest presence of competing reported? interests Convenience/ Differences in data sources can How accessible is the database? functionality be uncovered more easily Can it be downloaded? Can it be when those sources are easily disaggregated for fine analyses? accessible and the data can be disaggregated

consistency, Certainty, and Conflict of interest. I have added an additional criterion, Convenience and functionality, since many users of terrorism data want data that they can find easily and that will fulfill different functions. My hope is that the criteria proposed here will help researchers be better able to evaluate the quality of terrorism data and be in a better position to choose between terrorism data sources (Table 2.1).

Conceptual Clarity How well do the authors of the database communicate the underlying concept of terrorism they use to choose individual cases? How well do they define it? As Lazarsfeld and Barfeld (1951, p. 155) once wrote, Before we can investigate the presence or absence of some attribute… or before we can rank objects or measure them in terms of some variables we must form the concept of that variable.

22

I.S. Sheehan

Concepts, as Goertz (2005, p 5) observes, are ontological: “they are theories about the fundamental constitutive elements of a phenomenon.” Traditionally, concepts implied necessary and sufficient conditions. In the last few decades, however, what Collier and Mahon (1993) call “family resemblance” approaches have often been used as a substitute for necessary and sufficient conditions. In these approaches, one condition substitutes for another. Such substitutions can lead to “conceptual stretching” or “traveling” and more permissive inclusion of cases (Collier & Mahon, 1993, p. 845). While conceptual stretching and traveling may have benefits (ambiguous cases will not as easily be lost), they pose problems for researchers who want to generalize from a set of data. This was an issue when the U.S. State Department released its 2003 Patterns of Global Terrorism Report. This report specified terrorism as “premeditated, politically motivated violence perpetrated against noncombatant targets by subnational groups or clandestine agents, usually intended to influence an audience” (U.S. State Department, Patterns of Global Terrorism 2003, p. vii). It further specified that an international terrorist attack was an act committed by substate actors from one nation against citizens or property of another and that an incident was “judged significant if it results in loss of life or serious injury to persons, major property damage, and/or is an act or attempted act that could reasonably be expected to create the conditions noted” (U.S. State Department, Patterns of Global Terrorism 2003, Appendix A). The key problem here was the use of the word “significant.” As Krueger and Laitin (2004a, 2004b) quickly pointed out, almost no information was provided about how the government authors distinguished significant from nonsignificant events. In the end, a reanalysis of the data with better specification of what the government authors meant by the word “significant” produced a very different set of data and one which embarrassingly contradicted previous findings that terrorism events had decreased that year. In fact, the new evidence indicated that terrorist events had increased. Conceptual clarity has implications for inclusion rules. Different concepts of what constitutes an event, for example, can lead to different inclusion rules and widely varying estimates. This was the case when in its original database RAND counted 40 bombings by one group in one city as one event when ITERATE counted 40 separate events (Jenkins & Johnson, 1975). Similarly, different concepts of what is international or transnational can lead to other discrepancies. As discussed earlier in this chapter, RAND but not ITERATE appears to have treated incidents involving Palestinians and Israelis as transnational with the result that its estimates of total transnational terrorist incidents were much higher than ITERATE’s at some intervals in the last 12 years.

Context and Immediacy of Observation What is the context of the data? How was it collected? How close were the authors to the source? Terrorism data may be generated from primary sources, from secondary sources, tertiary ones, and from experts. Academic scholars often put a premium on

2

Assessing and Comparing Data Sources for Terrorism Research

23

data collected through direct observation because of its immediacy. Interview data, such as that collected by Merari (2010) in his interviews with suicide terrorists, is especially valued because of its immediacy. The value of such data, however, has to be weighed against the potential for selection bias. After direct observation, scholars usually value primary sources such as newspaper accounts at the time of an event over secondary sources (e.g., journal articles) and tertiary ones (e.g., textbooks, other datasets).

Citation Transparency Do the authors provide the sources for the data they report? Terrorism data is derived from a wide variety of sources. The sources may include United States and foreign news and wire services. They may also include information from interviews, books, memoirs, and interviews with principals (Mickolus, 2002). Knowing the sources of a particular event matters since coders often have to make decisions. They may have to choose, for example, between conflicting estimates from different sources on the number of fatalities associated with an event. Confidence in data is heightened when the actual source of the data is provided. Citation transparency is also critical to correcting and replicating datasets.

Coding and Consistency Do the authors use consistent rules to code data? Do they use a codebook? And have they institutionalized systems to ensure coding consistency across raters (inter-rater reliability) and over time (test–retest reliability)? If so, how do these systems work? In cases where there are conflicting reports for example, about claims of responsibility for an event, or the number of casualties, how are final decisions made? By fiat? By consensus, majority rule or some other way? Finally, is there an “audit trail” that users can follow if the operational definition of terrorism or inclusion rules is altered?

Certainty of Record Do the authors report the presence of uncertainty in the coding of a particular variable or attribute of a terrorist event? Conflicting accounts, as discussed above, may lead to uncertainty about whether an event was actually a terrorist event or something else. Or, an attribute of an event may not quite fit within a given coding scheme. The scheme, for example, may allow coding of one or two targets, but the terrorist attack had multiple targets. How do the authors resolve such quandaries and do they highlight them?

24

I.S. Sheehan

Conflict of Interest Do the creators and maintainers of the database have conflicts of interest and do they report them? This is tricky. Since terrorism is a national security concern and since the creation and maintenance of terrorism databases is costly, most of the databases we cover here have received significant funding from a government body and one, WITS, is a direct output of a government agency, the NCTC. The question is really the extent to which sponsors, funders, or others influence the content and the extent to which operators of the database are upfront about such real or potential influences. In election years, in particular, political pressure may be exerted to show that terrorism rates have diminished. How do the authors handle such potential conflicts of interest? For example, do they disclose funding? Do they discuss the extent to which the sponsors have control over the data?

Convenience/Accessibility/Functions How convenient is the database to use? How accessible is it? Is it available online? Is it searchable with key words? How functional is it? Can users perform online searches with key words? Can they use the data to create graphs and tables online? Is the dataset fully downloadable so that researchers can conduct their own analyses on it? Can the data be disaggregated for fine analyses? And finally, is there a fee to use it?

Application of the Framework Assessment of Existing Terrorism Databases Below I review and assess a range of existing terrorism databases and try to apply the criteria developed above as a basis for comparison. The databases I cover include five terrorism events databases. Terrorism events databases are systematic numeric records, usually derived from newspapers, wire, and other media, of the occurrence of individual terrorist events and the events’ characteristics (e.g., date, location, name or type of perpetrator group when it can be identified, type of attack, and number of casualties). Terrorism events data can be linked in turn with other data to study the causes and consequences of terrorism. Events databases allow researchers to examine trends and patterns in terrorism over long periods of time and geographical space. They can be used in conjunction with other data (e.g., political or economic indicators) in analyses of the causes and consequences of terrorism and can contribute information to analyses of how, when and why and terrorism events and campaigns decline or end (Schmid, 2004).

2

Assessing and Comparing Data Sources for Terrorism Research

25

Over the years, terrorism events databases have been used in time series analyses to assess impact of terrorist incidents on tourism (Enders, Sandler, & Parise, 1992; Fleischer & Buccola, 2002), on foreign investment (Enders & Sandler, 1996), and on gross domestic income and trade (Abadie & Gardeazabal, 2003; Nitsch & Schumacher, 2004). They have also been used to assess the impacts of such initiatives as the installation of metal detectors in airports on skyjackings (Enders & Sandler, 1993) and get tough police responses on violence in Northern Ireland (La Free, 2010). Five of the best-known terrorism events databases are as follows: 1. 2. 3. 4. 5.

International Terrorism: Attributes of Terrorist Events (ITERATE) Rand Database of Worldwide Terrorism Incidents (RDWTI) Global Terrorism Database (GTD) World Incident Tracking System (WITS) Terrorism in Western Europe: Events Data (TWEED)

Their scope, the time spans they include, and the numbers of incidents they cover are shown in Table 2.2. This table also provides an evaluation of how well each of the five databases meets each of the seven best standards criteria I set forth above.

General Comparisons All of the databases with the exception of TWEED were created and are maintained in the United States. Although most were created outside academia (TWEED developed by Jan Engene at the University of Bergen is an exception), all are currently operated, directed, or at least partially managed by academics who provide consulting and oversight. ITERATE was started and continues to be maintained by former CIA analyst Edward Mickolus, but is now updated by Todd Sandler and his colleagues at the University of Texas at Dallas. Similarly, the RAND database, which originated with policy analysts such as Brian Michael Jenkins, is at least partially under the direction of Bruce Hoffmann at Georgetown University. The basis for the GTD was data collected by a private security service, PGIS, but the data is now under the direction of Gary La Free at the University of Maryland. WITS is a product of the US government National Terrorism and Counterterrorism Center (NCTC) but consulting oversight is provided by John Wigle at Johns Hopkins University. As shown in Table 2.2, the five databases vary considerably in terms of the time spans they cover. With incidents dating back to 1950, TWEED spans almost 60 years. ITERATE, RAND, and GTD, with start dates in 1968, 1968, and 1970 respectively, each cover approximately 40 years. WITS, on the other hand, is relatively recent. With a start date in 2004, it only covers incidents for the past 7 years. The five databases also differ in scope. ITERATE is uniquely restricted to transnational terrorist events, defined as incidents that start in one country and end in another or involve victims, targets, institutions, or citizens of more than one nationality (Sandler & Enders, 2004). In ITERATE, domestic incidents of terrorism, i.e., incidents which begin and end in the same country and which only have ramifications

1970–2008 ~88,000 ~75 in 15 categories

Time span # of events # of variables

Context/source type

Provides definition and criteria Reports using primary sources: news articles, wire services, interviews with principals and secondary and tertiary documents: journals, books, and chronologies

Provides definition and criteria Reports using primary documents including newspapers, journals, radio broadcasts, and foreign press

Provides definition and criteria Reports using primary sources (news and wire services) in real time, secondary and tertiary sources (books, journals, existing datasets)

Events International + domestic

Events International + domestic

Assessment Conceptual clarity

https://wits.nctc.gov

http://www.start.umd.edu/ gtd

Provides definition and criteria Reports using open sources including subscription news services, local news websites in English, and foreign languages

2004–2010 ~69,000 ~15

NCTC

START University Maryland (LaFree)

1968–2009 ~36,000 15 + narrative description

Worldwide Incidents Tracking System

Global Terrorism Database

1968–2008 ~13,000 42

WITS

GTD

Table 2.2 Evaluation of five databases on terrorist events ITERATE RDWTI Overview International Terrorism – Rand Database of Attributes of Terrorist Worldwide Events Terrorism Incidents DB Operator(PI) University Michigan and RAND (Original PIs: Vinyard Software Jenkins, Hoffman) (Mickolus) Website www.icpsr.umich.edu http://www.rand.org/ nsrd/projects/ terrorism-incidents/ Unit of analysis Events Events Scope of events International International + domestic

Provides definition and criteria Reports using only one source, Keesing’s Record of World Events. Most of the data (events from ~1950–1998) constructed retrospectively

Events Events in Western Europe 1950–2004 11,245 52

http://folk.uib.no/ sspje/tweed.html

Terrorism in Western Europe: Events Data University of Bergen (Engene)

TWEED

26 I.S. Sheehan

Provides detailed codebook Reports using identical criteria and maintaining continuity among coders through the use of overlapping coders and monitors

Coding transparency and consistency

Certainty of record

Reports generally using AP, UPI, Reuters, Foreign Broadcast Information, and major US newspapers

ITERATE

Citation transparency

RDWTI

No discussion of uncertainty at website

Provides basic information for coding of variables on website Procedures for achieving coding consistency provided in separate papers

Reports using two or more sources for most events and that all source documentation is kept in paper form for each event

GTD

Provides criteria for incident inclusion and coding scheme in a codebook. Cautions that data were collected in real time for GTD I (1970–1997), retrospectively for GTD 2 (1998–2007) and in real time again after 2007 Includes a “doubt terrorism proper” field to record any reservation in the eyes of GTD analysts that the incident in question is truly terrorism for incidents after 1997

Reports using >3,500 news articles and 25,000 news sources for 1998–2007 alone. Clear citations to sources provided for recent events data

WITS

Addresses potential for uncertainty. Includes a “Confidence” field to designate whether attribution to a particular group is unknown, likely, plausible or inferred

Does not provide actual codebook Does address methodological issues in coding related to some variables on website and in papers

Reports using commercial subscription news services, the US government’s sources, local news websites, and use of news websites in foreign languages

TWEED

(continued)

Reports using only one source, Keesing’s Record of World Events, a world news archive that has recorded world events since 1931 and is updated monthly Provides detailed codebook Does not address who did the coding or how coding consistency was achieved on website

2 Assessing and Comparing Data Sources for Terrorism Research 27

ITERATE

Data was originally developed by a CIA analyst, E. Mickolus. ICPSR website does not report government funding

Free to students/faculty at subscribing universities; charges otherwise apply

RDWTI

Full Yes Yes Yes Yes Yes (through a code number). User can lease or purchase textual database No cost to user

RDWTI website reports US government contract to develop original database and continuous advising to US government on terrorism

GTD

No cost to user

Full Yes Yes Yes Yes Short narrative included with data

GTD website reports current funding from DOJ, DHS. Clearly states that the GTD does not purport to represent inclusion decisions of DHS, DOJ, or other funding agencies

No cost to user

Full Yes Yes Yes Yes

WITS

TWEED

No cost to user

Full Yes No No Yes

No funding sources reported on website

The RDWTI is composed of two earlier databases, the RAND Terrorism Chronology (1968–1997) which contained ~10,000 incidents and was limited to international incidents and the RAND-MIPT Terrorism Incident Database (1998–2008) which contained about 26,000 incidents and included both national and international incidents (check numbers)

Pricing

Convenience/functionality Online availability Partial Browsing features No Key word searching No Graphing features No Downloadable No Linkages Numeric dataset is linked to narrative (text) database

Conflict of interest (website report)

Table 2.2 (continued)

28 I.S. Sheehan

2

Assessing and Comparing Data Sources for Terrorism Research

29

for that country, are specifically excluded. For example, the Oklahoma City bombing on April 19, 1995 is excluded, as are terrorist attacks by ethno-national groups within their own countries. This exclusion explains the relatively low number of incidents in ITERATE compared to the other databases. It has been estimated that domestic terrorist events outnumber international ones by eight to one (Sandler & Enders, 2008). By contrast, TWEED is limited to acts of internal terrorism in Western Europe. That is, it only includes events initiated by agents originating in the18 countries of Western Europe it covers. TWEED expressly excludes terrorist acts “imported” from outside the West European countries (Engene, 2007). The other three databases include both domestic and international incidents but do so for different time spans. The RAND database provides international incidents dating back to 1968 but contains domestic incidents only for the past 13 years, since 1997. The GTD provides domestic and international incidents dating back to 1970 while WITS provides domestic and international incidents only from 2004 to the present. As shown in Table 2.2, the number of variables ranges from about 15 in the RAND database to 75 in the GTD. These differences, however, are partially a function of how variables are counted. The date of an incident may be shown as one variable or as three distinct ones that include day, month, and year. Similarly, the location of an incident can be presented as one variable or three (city, state or province, and country). Although ITERATE and the GTD do provide many more variables than the other databases, not all are what might be called “usable.” ITERATE, for example, includes separate variables for the number of terrorists and the number of nationalities of terrorists in an attack force. Because of the clandestine nature of terrorism and the fact that a large majority of them are unclaimed, information like this is often unknown and it is not uncommon to have large numbers of missing values for such variables (Table 2.3).

How Well Do the Five Events Databases Clarify the Concept of Terrorism That Informs Their Selection of Terrorist Acts or Incidents? All five of the databases we cover here give fairly clear definitions of the concept of terrorism they employ in selecting cases (see Appendix). ITERATE and RAND have consistently used the same or approximately the same definition of a terrorist incident since they began collecting incident data in the early 1970s. TWEED and WITS have also employed one definition since their inceptions. As discussed in a separate article in this monograph (La Free, 2011), the GTD is based on three datasets. GTD1 (1970–1997) was collected in real time by the PGIS, a private security firm. GTD2 (1998–2007) was collected by START at the U Maryland retrospectively, and GTD3 (2007-present) has been collected in real time by START. The GTD clearly provides the definition used by PGIS for GTD1 data and it also provides the definition used by START for GTD2 and 3 data. As discussed below, START uniquely allows the researcher to apply his or her own definition to select cases from the full GTD.

30

I.S. Sheehan

Table 2.3 Comparison of variables included in five terrorism databases Variable ITERATE RAND WITS Date Incident start date (year/month/day) √ √ √ Incident end date (year/month/day) √ √ Location City Country Province or state Region Site of attack for example, home, office, car, airplane Attack type Type of attack State sponsorship if known Suicide attack (yes/no) Part of multiple incident/ coordinated attack Logistical success of attack (yes/no) International incident (yes/no) Perpetrator Perpetrator group name(s) Perpetrator group type/or ideological profile Number of perpetrators in attack force Number of female perpetrators Nationalities of perpetrators in attack force Number of perpetrators captured Claim(s) of responsibility

√

√

√ √ ?

√ √

√ √ √ √

TWEED

√

√

√ √ √

√ √ √ √

√ ? √

√ ? √ √

√

√ √

√ √

√

√

√

√

√

Text √

√

√ √

√

√ √ √ √

√

Weapon Means or weapon type used

√

√

√

Victims Number of victims Nationalities of victims Number killed Number wounded/injured Number US citizens killed Number US citizens wounded Number of perpetrators killed

√ √ √ √ √ √ √

√ √ √

√ √ √ √

Type of target/victim Type of victim/target US victim (yes/no) Type of US victim Nationality of target

GTD

√ √ √

√ √

√

√

√

√ √ √ √ √

√ √

√

√

√ (continued)

2

Assessing and Comparing Data Sources for Terrorism Research

Table 2.3 (continued) Variable

ITERATE

Consequences Property damage (yes/no) Extent of property damage Value of property damage Logistic success

√ √ √ √

Hostage/kidnapping Hostages (yes/no) Number of hostages Number of US hostages

√ √ √

RAND

31

WITS

GTD

√

√ √ √ √

√

TWEED

√ √

Doubt/uncertainty

√

Multiple incident

√

Source citation(s)

√

Government reaction Armed response √ Arrest √ Conviction √ Demonstration control √ Number killed by state √ institution Number wounded by √ state institution Examples for type of attack = assassination, bombing, hostage taking Examples for type of target = business, government, airports, journalists and media, private citizens, and property Examples of weapon type = biological, chemical, nuclear, firearms, explosives Some databases (e.g., ITERATE and GTD) allow coding of group information for up to three “terrorist” groups Some databases (e.g., GTD) allow coding of target characteristics for >1 type of target Citations to sources available in GTD for incidents from 1997 onward

As shown, in Table 2.4 the databases do differ in terms of operational inclusion rules. For example, although ITERATE, RAND, WITS, and GTD restrict inclusion to terrorist acts committed by substate actors (substate terrorism), TWEED allows inclusion of terrorist acts by states (state terrorism). TWEED also differs from the others in its omission from its definition of the word violence and its use instead of the concept of “personal injury” or “attacks against material targets (property).” While this difference may appear semantic, it suggests a slightly lower level of tolerance for inclusion (an injury is not necessarily the result of violence). As another example, some but not all of the databases specify that the act must be premeditated or intentional. Some but not all include threats of violence and some but not all specify that the act must be committed against a civilian or noncombatant target (see Appendix). The extent to which each of the databases stretches the concept of terrorism is difficult to evaluate. Schmid and Jongman (2005, p. 146), however, contend that even in ITERATE, long considered the most authoritative database on terrorist

32

I.S. Sheehan

Table 2.4 Similarities and differences in factors included in database definitions of terrorism ITERATE RAND WITS GTD TWEED Premeditated, intentional, or deliberate √ √ √ Use of violence √ √ √ √ Threat of violence √ √ √ Act must be politically motivated √ √ √ Intended to influence or coerce an audience √ √ √ √ Civilians or noncombatants targeted √ √ Perpetrators are substate groups or √ √ √ √ √ clandestine agents Perpetrators are states or state agents √ Calculated to cause anxiety, fear, or terror √ √ Can be motivated by political, religious, √ economic, or social goal Carried out to achieve publicity √ Act is outside the context of legitimate √ √ √ √ √ warfare or a coup d’état Note: GTD includes incidents that meet elements #1, 2, 7 and at least 2 of #5, 10, and 12

events, there is sometimes a “strained relationship” between the cases it includes and its “operational definition” of terrorism. While some types of incidents – such as kidnappings, letter bombings, assassinations and murders and aerial hijackings – fit the general notion of terrorism and are compatible with the working definition, other types of incidents – such as sabotage, arms smugglings shootouts with police, occupations, thefts or break-ins, conspiracies and snipings – are not.

As noted earlier in this chapter, stretching of the definition was common in the early years of Patterns of Global Terrorism (the predecessor of WITS) when the definition was redefined several times to make it more inclusive.

What Is the Context of the Data? How Was It Collected? Where Does It Come from? How Immediate Were the Sources? Most of the data for ITERATE, RAND, GTD, and WITS was collected by analysts in real time shortly after an incident occurred and was reported in the media. This is not the case with TWEED. For this database, created as part of the author’s doctoral dissertation (1998), all of the data was collected retrospectively from one source. Also, as discussed above, although the bulk of the GTD dataset was collected in real time, a portion of it was collected retrospectively. These differences can affect the numbers of events included since not all sources that are available in real time are still available or accessible years later. The extent to which the databases rely on primary vs. secondary and tertiary sources is difficult to assess. Sandler, Arce, and Enders (2008) report that the ITERATE dataset is derived from sources such as the AP, UPI, Reuters tickers, the Foreign Broadcast Information Service (FBIS), Daily Reports, and major US newspapers. Such sources are usually classified as primary, but Edward Mickolus has indicated that earlier data is also based at least in part on direct observation (interviews with

2

Assessing and Comparing Data Sources for Terrorism Research

33

principles), secondary sources (journals), and tertiary sources (books). The RDWTI website reports that most events included in its dataset are based on at least two sources and that all source documentation is kept in paper for each event record. The website for the GTD indicates that over 3,500,000 news articles and 25,000 news sources were reviewed to collect incident data for its database from 1998 to 2007 alone. According to John Wigle, the NCTC gathers data for WITS from open sources manually using commercial subscription news services, the U.S. Government’s Open Source Center (OSC), local news websites reported in English, and, as permitted by the linguistic capabilities of the team, local news websites reported in foreign languages. TWEED data is uniquely based on one source, Keesings Record of World Events, an archive of articles on political, social, and economic events around the world that has existed since 1931 is updated monthly (Engene, 2007).

Do the Authors Cite Their Sources? Most of the databases provide facts about their sources but do not go beyond referring to a general range of sources. At this time, only the GTD has made an attempt to cite sources for individual incidents and then only for its most recent incidents (see GTD website, home page announcements).

How Consistent Is the Coding? All five of the databases provide a codebook or basic information on coding of variables associated with incidents. Most, however, give only limited information about who does the coding, how coding consistency over time is achieved, whether multiple coders are used, and if so whether tests of inter-coder reliability are performed. According to Sandler et al. (2008), coding consistency over time is sought in ITERATE by applying identical criteria and maintaining continuity among coders through the use of overlapping coders and monitors. According to Wigle (2010), the WITS team strives to maintain consistency in collection by having a central intelligence officer maintain knowledge of the search strings and Internet web sites commonly used by the analysts. This “knowledge capital,” writes Wigle, provides consistency during turnover on the team. To reduce interpretation bias further (or increase inter-rater reliability), NCTC analysts maintain account notes of commonly used terms and phrases found in the press, recurring political and ethnic issues, terrain notes, weather-related trends, and other factors that influence a mastery of context surrounding acts of violence in countries assigned to their area of responsibility.

What About Handling of Uncertainty? Almost all of the authors address this issue. For most of the databases covered here, uncertainty is addressed by omitting information that cannot be verified. The GTD

34

I.S. Sheehan

is unique in building issues of uncertainty into its coding. For post 1997 data, it includes a field labeled “Doubt Terrorism Proper.” This field allows the coder to indicate uncertainty about whether an incident meets all of the criteria for inclusion in the database. Users can filter the data and exclude cases in which there is a doubt. As indicated above, users can also filter results based on whether they meet all or only some of the criteria for its definition of terrorism. Do the Authors Disclose Funding and Potential Conflict of Interest? Conflict of interest occurs when an individual or organization is involved in multiple interests, one of which could corrupt the motivation for the act of another. Many of the events databases covered here were originally developed with government grants or under government sponsorship for specific political purposes. The RAND database, begun in the early 1970s, was first developed under a US government contract. RAND data was subsequently used as a basis for ITERATE, which later became a basis for the CIA’s annual terrorism report in 1977 (Schmid, 1983, p. 257– 260). The WITS database is a much more recent product of a US government agency, the NCTC, and the GTD has received funding from the DOJ and DHS. Most of the databases we cover here do disclose their funding sources. Some are more explicit than others about the extent to which they have independent control over the data they produce. Ultimately, the extent to which government sponsorship or funding has in fact produced biases in content is not known. How Convenient and Accessible Are the Different Databases? Is there a cost? Are the databases available online? Are they searchable, interactive, and can datasets be downloaded? ITERATE: The ITERATE database is not fully accessible online. ITERATE datasets and documentation for incidents from 1968 to 1977 can be downloaded at no cost from the Inter-University Consortium for Political and Social Research (ICPSR) at www.icpsr.umich.edu/icpsrweb/ICPSR/studies/7947 at the University of Michigan. More recent data (1978 to the present), however, are available at no cost only to students and faculty of subscribing universities or for a fee from Vinyardsoftware@ hotmail.com or via postal services at Vinyard Software, Inc. 2305 Sandburg Street, Dunn Loring, VA 22027–1124. WITS: The WITS database is available online at no cost at http://www.nctc.gov/ wits/witsnextgen.html. Researchers can search this database. This database can be searched using a variety of parameters (date, location, attack type, weapon type) to generate subsets of data and reports. The data can also be exported in subsets or in its entirety in a tab-delimited data file from the RAND website. RDWTI: The full RDWTI database is available online at no cost at http://www. rand.org/nsrd/projects/terrorism-incidents/. This database can be searched using a

2

Assessing and Comparing Data Sources for Terrorism Research

35

variety of parameters (date, location, attack type, weapon type) to generate subsets of data and reports. The data can also be exported in subsets or in its entirety in a tab-delimited data file from the RAND website. GTD: The GTD is also fully accessible online at no charge. It can be accessed at http://www.start.umd.edu/gtd/. Researchers can browse the GTD and use search terms and filters to generate subsets of data and produce graphs and tables. They can also download the dataset in different formats. TWEED: The TWEED database is not interactive online. However, the full TWEED dataset in SPSS format and the accompanying codebook can be downloaded for educational and research purposes at http://folk.uib.no/sspje/tweed.htm.

Conclusion and Challenges for the Future This chapter has reviewed the need for norms and standards to assess the quality of terrorism data and to help researchers choose between competing datasets. Using insights from scholarship on small-n qualitative data and the concept of a criterial framework (Gerring, 2001), the chapter proposes a set of criteria for evaluating large-n terrorism data and applies these criteria to five existing terrorism datasets. The findings suggest that despite remarkable strides in the construction and availability of terrorism data, there is still room for considerable improvement. There is a need in some cases for better specification of the definition of terrorism. There is a need in almost all cases for greater transparency in source citation and almost all of the databases could benefit from more explicit descriptions of coding rules and acknowledgement, where relevant, of doubt and uncertainty. Going forward, the author has several recommendations. 1. In the long run, terrorism databases are only as good as the concepts they are built on. There is still a need to fine-tune our conceptualizations of terrorism. In the meantime, greater recognition of conceptual differences in the definitions of terrorism used in existing datasets will also benefit researchers struggling with the problem of interpreting differing results from different datasets. 2. Greater transparency in terms of the context and citation for sources of terrorism data would generate increased confidence in data and give researchers an opportunity to check original citations. It could also facilitate more mixed method analyses in which researchers, for example, perform additional tests of the truth of quantitative findings using qualitative case study analyses. The GTD’s explicit citation of actual sources for its most recent events data is an important step in this direction. 3. More explicit acknowledgement of coding issues (who does the coding, how many coders are used, what is the process, how are coding conflicts resolved?) will help researchers better evaluate discrepancies in results and have greater confidence in the data.

36

I.S. Sheehan

4. More widespread recognition and acknowledgement that absolute certainty in coding terrorist events is not always possible will enrich interpretation of data. The inclusions by the GTD of a “doubt terrorism proper” field and by WITS of a “confidence” field are both valuable steps in this direction. 5. More transparency in acknowledgement of funding sources and who controls terrorism data will help researchers better evaluate potential conflicts of interest. Finally, the near canonical reputation of datasets such as ITERATE needs to be reevaluated in light of the valuable contributions of newcomers to the field. Over the years ITERATE data has been used so often in academic publications that it has come to be seen by some as the only authoritative database on terrorism. But ITERATE is confined to international or transnational events and it is becoming much more obvious that distinctions between international and domestic terrorist events are not as clear-cut as previously thought. Moreover, ITERATE data is only available to subscribing universities and is not otherwise accessible on the web.

Appendix Definitions of Terrorism ITERATE: For the purpose of the dataset, ITERATE defines a terrorist event as . . . the use, or threat of use, of anxiety-inducing, extra-normal violence for political purposes by any individual or group, whether acting for or in opposition to established governmental authority, when such action is intended to influence the attitudes and behavior of a target group wider than the immediate victims. (Mickolus, 2003, p. 2)

RAND: For the RAND database terrorism is defined as …violence calculated to create an atmosphere of fear and alarm to coerce others into actions they would not otherwise undertake, or refrain from actions they desired to take. Acts of terrorism are generally directed against civilian targets. The motives of all terrorists are political, and terrorist actions are generally carried out in a way that will achieve maximum publicity. (RAND)4

4

The RAND website clarifies its definition further with this excerpt from Defining Terrorism by Bruce Hoffman….We may therefore now attempt to define terrorist as the deliberate creation and exploitation of fear through violence or the threat of violence in the pursuit of political change. All terrorist acts involve violence or the threat of violence. Terrorism is specifically designed to have far-reaching psychological effects beyond the immediate victim(s) or object of the terrorist attack. It is meant to instill fear within, and thereby intimidate, a wider “target audience” that might include rival ethnic or religious group, an entire country, a national government or political party, or public opinion in general. Terrorism is designed to create power where there is none or to consolidate power where there is little. Through the publicity generated by their violence, terrorists seek to obtain the leverage, influence and power they otherwise lack to effect political change on either local or international scale.

2

Assessing and Comparing Data Sources for Terrorism Research

37

RAND specifies that terrorism is defined by the nature of the act, not by the identity of the perpetrators or the nature of the cause. For RAND the key elements are as follows: • • • • • •

Violence or the threat of violence Calculated to create fear and alarm Intended to coerce certain actions Motive must include a political objective Generally directed against civilian targets Can be a group or an individual

GTD: GTD data for 1970–1997, collected by the Pinkerton Global Intelligence Service (PGIS) used the following definition of terrorism: the threatened or actual use of illegal force and violence by a non state actor to attain a political, economic, religious, or social goal through fear, coercion, or intimidation. (START, 2011)

GTD Data for 1998–2007 coded incidents in a way to allow users to identify cases that met their own definition of terrorism. Using the original definition, each incident had to be an intentional act of violence or threat of violence by a nonstate actor. In addition two of the following three criteria had to be met to be included. • The act was aimed at attaining a political, economic, religious, or social goal. • There included evidence of an intention to coerce, intimidate, or convey some other message to a larger audience (or audiences) than the immediate victims. • The act was outside the context of legitimate warfare activities. These criteria continued to be included for data collected in real time after 2007 (START, 2011). WITS: For its database WITS uses the definition of terrorism prescribed in the congressional reporting statute 22 U.S.C. § 2656f (d)(2). This statute reads …the term ‘terrorism’ means premeditated, politically motivated violence perpetrated against noncombatant targets by subnational groups or clandestine agents.

TWEED: For its database, TWEED uses the following explanation An act of terrorism is counted an act that has inflicted personal injury or attacks against material targets (property) if the act is of a nature that could have led to personal injury or in another way would have a noticeable impact on an audience, while at the same time the act was committed to direct demands of or raise attention from others than those immediately inflicted with personal or material injury. (Engene, 2006)

For TWEED the following events are counted as “violent actions of a terrorist nature: bombings, explosions, arson, fires, rocket attacks, killings, attempted killings, abductions, kidnaps, shootings, sieges, violent actions, violent attacks, attacks and similar violent actions.” Further, the event must be brought about “by an agent that has deliberately initiated the action.” While TWEED excludes events in which

38

I.S. Sheehan

the purpose might be a coup d’état, it includes events in which government authorities engage in actions against terrorist or put the public in a state of fear: …events in which state authorities, police, secret services, military institutions, etc. are involved in actions directed against terrorists and terrorist groups are to be included. Also violent acts of state institutions directed against civilians are to be included, for instance in conjunction with demonstrations, strikes, and the like, when the state institution acts in a way that might put the public or sections of it in a state of fear. (Engene, 2006)

References Abadie, A., & Gardeazabal, J. (2003). The economic costs of conflict: A Case Study for the Basque Country. American Economic Review, 93(1), 113–132. Brady, H. E., & Collier, D. (2004). Rethinking Social Inquiry: Diverse tools, shared standards. Berkeley: Rowman & Littlefield and Berkeley Public Policy Press. Collier, D., Brady, H. E., & Seawright, J. (2004). Sources of leverage in causal inference: Toward an alternative view of methodology. In H. Brady & D. Collier (Eds.), Rethinking social inquiry: Diverse tools, shared standards (pp. 229–266). Berkeley: Rowman & Littlefield and Berkeley Public Policy Press. Collier, D., & Mahon, J. E., Jr. (1993). Conceptual ‘stretching’ revisited: Adapting categories in comparative analysis. American Political Science Review, 87, 845–855. Drakos, K. (2009). Security economics: A guide for data availability and needs. Economics of Security Working Paper 6, Berlin: Economics of Security. Retrieved July 29, 2011 from https:// www.diw.de/documents/publikationen/73/diw_01.c.94892.de/diw_econsec0006.pdf. Enders, W., & Sandler, T. (1993). The effectiveness of anti-terrorism policies: Vector-autoregressionintervention analysis. American Political Science Review, 87, 829–844. Enders, W., & Sandler, T. (1996). Terrorism and foreign direct investment in Spain and Greece. Kyklos, 49(3), 331–352. Enders, W., Sandler, T., & Parise, G. F. (1992). An econometric analysis of the impact of terrorism on tourism. Kyklos, 45, 531–554. Engene, J. O. (2006). TWEED code book. Bergen: University of Bergen: Department of Comparative Politics. Engene, J. O. (2007). Five decades of terrorism in Europe: The TWEED data set. Journal of Peace Research, 44(1), 109–121. Fleischer, A., & Buccola, S. (2002). War, terror, and the tourism market in Israel. Applied Economics, 34(11), 1335–1343. Fowler, W. W. (1981). Terrorism data bases: A comparison of missions methods, and systems. Retrieved July 29, 2011 from http://www.rand.org/content/dam/rand/pubs/notes/2005/ N1503.pdf. Geddes, B. (1990). How the cases you choose affect the answers you get: Selection bias in comparative politics. Political Analysis, 2, 131–150. Gerring, J. (1999). What makes a concept good? A criterial framework for understanding concept formation in the social sciences. Polity, 31(3), 357–393. Gerring, J. (2001). Social Science Methodology: A criterial framework. Cambridge: Cambridge University Press. Gerring, J. (2010). Social science methodology: Tasks, strategies, criteria. Cambridge: Cambridge University Press. Goertz, G. (2005). Social Science Concepts: A user’s guide. Princeton: Princeton University Press. Guba, E. (1981). Criteria for assessing the trustworthiness of naturalistic inquiries. Educational Communication and Technology Journal, 29(2), 75–91.

2

Assessing and Comparing Data Sources for Terrorism Research

39

Herman, E. S., & Chomsky, N. (1988). Manufacturing Consent. New York, NY: Pantheon. Herrera, Y., & Devesh, K. (2007). Improving data quality: Actors, incentives and capabilities. Political Analysis, 15, 365–386. Jenkins, B., & Johnson, J. (1975). International Terrorism: A chronology, 1968–1974. Santa Monica: Rand Corporation. Krueger, A. B., & Laitin, D. D. (2004, May 17). Faulty terror report card. Washington Post, p. A21. Krueger, A. B., & Laitin, D. D. (2004). Misunderestimating terrorism. Foreign Affairs. Retrieved July 29, 2011 from http://www.krueger.princeton.edu/terrorism1.html. La Free, G. L. (2010). The global terrorism database: Accomplishments and challenges. Perspectives on Terrorism, IV(1), 24–46. La Free, G. L. (2011). Generating terrorism event data bases: Results from the global terrorism database, 1970–2008, College Park, MD: University of Maryland. Lazarsfeld, P., & Barfeld, A. H. (1951). Qualitative measurement in the social sciences. Classification, typologies and indices. In D. Lerner & H. Lasswell (Eds.), The Policy Sciences (pp. 155–192). Stanford: Stanford University Press. Lieberman, E. S. (2010). Bridging the qualitative-quantitative divide: Best practices in the development of historically oriented replication databases. Annual Review of Political Science, 13, 37–59. Merari, A. (2010). Driven to death: Psychological and social aspects of suicide terrorism. Oxford: Oxford University Press. Mickolus, E. F. (2002). How do we know we’re winning the war against terrorists? Issues in measurement. Studies in Conflict & Terrorism, 25, 151–160. Mickolus, E. (2003). International terrorism attributes of terrorist events (ITERATE) data codebook. Dunn Loring: Vinyard Software. Munck, G. L., & Verkuilen, J. (2002). Measuring democracy: Evaluating alternative indices. Comparative Political Studies, 35, 5–34. Nitsch, V., & Schumacher, D. (2004). Terrorism and international trade: An empirical investigation. European Journal of Political Economy, 20(2), 423–433. Paull, P. (1982). International terrorism: the propaganda war. Master’s thesis, University of San Francisco. Reid, E. O. (1997). Evolution of a body of knowledge: An analysis of terrorism research. Information Processing & Management, 33(1), 91–106. Robson, C. (2002). Real world research: A resource for social scientists and practitioner-researchers. Malden: Blackwell Publishers. Sandler, T., Arce, D. G., & Enders, W. E. (2008). Copenhagen consensus 2008 challenge paper: Terrorism. Retrieved July 29, 2011 from http://www.copenhagenconsensus.com/Files/Filer/ CC08/Papers/0%20Challenge%20Papers/CP_Terrrorism_-_Sandler.pdf. Sandler, T., & Enders, W. (2004). An economic perspective on transnational terrorism. European Journal of Political Economy, 20(2), 301–316. Sandler, T., & Enders, W. (2008). Economic consequences of terrorism in developed and developing countries: An overview. In P. Keefer & N. Loayza (Eds.), Terrorism and economic development (pp. 17–47). Cambridge: Cambridge University Press. Schmid, A. (1983). Political terrorism: A research guide to concepts, theories, databases, and literature. New Brunswick: Transaction Books. Schmid, A. P., & Jongman, A. J. (1988). Political Terrorism: A New Guide to Actors, Authors, Concepts, Data Bases, Theories, and Literature. New Brunswick, New Jersey: Transaction Books, pp. 5–6. Schmid, A. (2004). Statistics on terrorism: The challenge of measuring trends, global terrorism. Forum on Crime and Society, 4(1–2), 49–69. Schmid, A., & Jongman, A. J. (2005). Political Terrorism: A new guide to actors, authors, concepts, data, theories, and literature. New Brunswick: Transaction Publishers. Sheehan, I. S. (2007). When Terrorism and Counterterrorism Clash: The war on terrorism and the transformation of terrorist activity. Amherst: Cambria Press. Sheehan, I. S. (2009). Has the war on terrorism changed the terrorist threat? Studies in Conflict & Terrorism, 32(8), 743–761.

40

I.S. Sheehan

START National Consortium for the Study of Terrorism and Responses to Terrorism. (2011). Codebook GTD Variables and Inclusion Criteria. Retrieved July 29, 2011 from http://www. start.umd.edu/gtd/downloads/Codebook.pdf. Wardlaw, G. (1982). Political Terrorism. Melbourne: Press Syndicate of the University of Cambridge. Wigle, J. (2010). Introducing the worldwide incidents tracking system (WITS). Perspectives on Terrorism, 4(1), 3–23. Wilkinson, P. (1986). Trends in international terrorism and the American response. In L. Freedman & C. Hill (Eds.), Terrorism & International Order (pp. 37–55). London: Routledge & Kegan Paul.

Chapter 3

Generating Terrorism Event Databases: Results from the Global Terrorism Database, 1970 to 2008 Gary LaFree

Developing valid data on terrorism raises unique challenges. To begin with, terrorism represents a behavior that is difficult to define and measure. As PLO Chairman Arafat famously noted in a 1974 speech before the United Nations, “One man’s terrorist is another man’s freedom fighter.” For example, while the United States regards Hamas as a Foreign Terrorist Organization, many Palestinians regard it as a legitimate political party that recently won a major democratically held election. By contrast, while many individuals in China regard the ethnic Uyghurs who have been detained by the United States at Guantanamo Bay as terrorists, much of the rest of the world appears to disagree. Even Osama bin Laden’s execution in May 2011 did not result in universal support. Indeed, many of the most prominent terrorist groups in the world (including Shining Path, ETA, the Farabundo Marti National Liberation Front [FMLN], the IRA, Revolutionary Armed Forces of Colombia [FARC], the ELN, and the PKK) often define themselves as freedom fighters and have a loyal constituency who may denounce terrorism but are, indeed, relying on these groups to advance their own political agenda. This fundamental characteristic of terrorism no doubt explains in large part why international organizations such as the United Nations have not succeeded in adopting a universally accepted definition of terrorism (European Commission, 2008). Defining terrorism is no less complex for scientists. Schmid and Jongman’s (1988) influential review found 109 different research definitions of terrorism. Indeed, the first chapter of many books on terrorism is devoted to exploring and defending competing definitions (cf., Hoffman, 2007; Smelser, 2007). And beyond the challenge of arriving at a defensible definition of terrorism is the considerable difficulty of collecting valid data on terrorism (for a review, see LaFree & Dugan, 2009). Data on illegal violence has come traditionally from

G. LaFree (*) Department of Criminology and Criminal Justice, University of Maryland, College Park, MD, USA e-mail: [email protected] C. Lum and L.W. Kennedy (eds.), Evidence-Based Counterterrorism Policy, Springer Series on Evidence-Based Crime Policy 3, DOI 10.1007/978-1-4614-0953-3_3, © Springer Science+Business Media, LLC 2012

41

42

G. LaFree

three sources, corresponding to the major social roles connected to criminal events: “official” data collected by legal agents, especially the police; “victimization” data collected from the general population of victims and nonvictims; and “self-report” data collected from offenders. Victimization surveys have been of little use in the study of terrorism. Despite the attention it gets in the global media, terrorism is much rarer than more ordinary types of violent crimes. This means that even with extremely large sample sizes, only few individuals in most countries will have been victimized by terrorists. Moreover, because victims of terrorism are often chosen at random, they are unlikely to know the perpetrators, making it impossible to produce details about offenders. And finally, in many cases, victims of terrorism are killed by their attackers. Self-report data on terrorists have been more important than victimization data, but they also face serious limitations. Most active terrorists are unwilling to participate in interviews. And even if willing to participate, getting access to known terrorists for research purposes raises obvious challenges. As Merari (1991, p. 88) has put it, “The clandestine nature of terrorist organizations and the ways and means by which intelligence can be obtained will rarely enable data collection which meets commonly accepted academic standards.” Although governments in some countries do collect official data on terrorism (e.g., the U.S. National Counter Terrorism Center), data collected by governments are suspicious either because they are influenced by political considerations or because many fear that they might be so influenced. Moreover, while vast amounts of detailed official data on common crimes are routinely produced by the various branches of the criminal justice system in most countries, this is rarely the case for terrorism. For example, the majority of offenders suspected of terrorism against the United States are not legally processed for terrorism, but rather for other related offenses, such as weapons violations and money laundering (Smith, Damphouse, Jackson, & Sellers, 2002). In addition, much primary data are collected by intelligence agents (e.g., informers, communications intercepts) and are not available to researchers working in an unclassified environment. In response to these data and definitional challenges, in the last few decades researchers began to collect and analyze open source unclassified data on terrorist attacks. These terrorism event databases generally use news reports from electronic and print media to collect detailed information on the characteristics of terrorist attacks. Several of these open source databases started in the late 1960s – at the same time that portable cameras coupled with satellite technology first allowed reporters to send pictures and stories almost instantaneously from anywhere on the planet. By the 1970s, there was a minor cottage industry comprised of individuals and companies in several countries collecting data on terrorist attacks from unclassified media sources (see Sheehan, 2011). Many of those collecting data on terrorism in the early days had armed forces backgrounds and many had worked for military intelligence before starting new careers as terrorism data collectors in the private sector. While all of these terrorism event databases had unique individual characteristics, they all shared a reliance on open media accounts of terrorism – collected originally from some combination of newswire services, unclassified government reports such as those issued by the U.S. State Department, and leading

3

Generating Terrorism Event Databases…

43

international newspapers. These early data collection efforts provided a value-added service by systematizing many of the disparate sources of data on terrorism in a single database that could then be analyzed by researchers. This chapter reports new results for both international and domestic terrorism attacks from the most recent version of the Global Terrorism Database (GTD) maintained since 2005 by the National Consortium for the Study of Terrorism and Responses to Terrorism (START; LaFree & Dugan, 2009). The GTD currently includes data on the characteristics of nearly 88,000 terrorist attacks that occurred worldwide between 1970 and 2008. We began the GTD in 2002 by computerizing data originally collected by the Pinkerton Global Intelligence Service (PGIS), a private company that recorded terrorism incidents from 1970 to 1997 from wire services (including Reuters and the Foreign Broadcast Information Service [FBIS]), U.S. State Department reports, other U.S. and foreign government reporting, U.S. and foreign newspapers (including the New York Times, British Financial Times, Christian Science Monitor, Washington Post, Washington Times, and Wall Street Journal), and information provided by PGIS offices around the world. We completed computerizing the PGIS data in December 20051 and in April 2006, the START Consortium contracted with a team led by Gary Ackerman and Charles Blair at the Center for Terrorism and Intelligence Studies (CETIS) to extend data collection for the GTD beyond 1997. CETIS collected raw data for GTD for 1998 through March 2008. Starting in April 2008, raw data for GTD have been collected by a team led by Richard Ward and Dan Mabrey at the Institute for the Study of Violent Groups at New Haven University. The most recent data available when this chapter was being prepared were for 2008. The operational definition of terrorism I use in this chapter is: the threatened or actual use of illegal force, directed against civilian targets, by non state actors, in order to attain a political, economic, religious or social goal, through fear, coercion, or intimidation.2 It is important to note that terrorism depends as much on threats as the actual use of violence: for example, individuals who seize an aircraft and threaten to low it up unless their demands are met. Note also that by specifying the threatened or actual use of force we exclude hoaxes. The requirement that these events be limited to the actions of “nonstate actors” means that we exclude the considerable violence and terrorism that is directly attributable to states or their militaries.

1

Most of the 1993 data in the GTD were lost by the original data collectors and we have never been able to recover them fully (LaFree & Dugan, 2007). 2 This was the original PGIS definition of terrorism applied from 1970 to 1997. When data collection was taken over by the START Consortium in 2005, we required that two of the following three criteria also had to be met for inclusion in the database: (1) the violent act was aimed at attaining a political, economic, religious, or social goal; (2) the violent act included evidence of an intention to coerce, intimidate, or convey some other message to a larger audience (or audiences) other than the immediate victims; and (3) the violent act was outside the precepts of International Humanitarian Law. These criteria were constructed to allow analysts and scholars flexibility in applying various definitions of terrorism to meet different operational needs. The data presented in this chapter include all cases that meet any two of these three criteria.

44

G. LaFree

And the requirement that the act have a direct political goal means that we exclude ordinary criminal violence. Thus, the GTD excludes state terrorism and genocide, topics that are important and complex enough to warrant their own separate analysis. The GTD currently provides the most comprehensive unclassified data source for measuring terrorist attacks, including structured data on more than 120 variables for over 87,000 terrorist attacks committed by more than 2,000 terrorist organizations around the world since 1970. During the past 6 years, the GTD has become a public resource, playing an important role for those who need access to objective, unbiased information about the dynamics of terrorism. Dozens of policy professionals and researchers have downloaded the database or requested hard copies, and the GTD website (http://www.start.umd.edu/gtd) averages 1.5 million unique page hits per month. In summer 2011, we plan to release GTD data through 2010. After this release, our goal is to provide annual GTD data for each year in the spring of the following year. In this chapter, I provide baseline information on the worldwide distribution of terrorist attacks and fatalities, the regional distribution of terrorism, terrorist organizations responsible for the largest number of attacks, terrorist targets, tactics and weapons used by terrorists, and the regional distribution of terrorist attacks since 1970.

The Extent of Terrorism The tragic events of September 11, 2001 had an immediate and dramatic impact on levels of public concern about terrorism in the United States and well beyond. Similarly, more recent attacks such as those in Madrid on March 11, 2004, in London on July 7, 2005, and in Mumbai on November 26–29, 2008 also raised concerns about terrorism among citizens, not only in the countries attacked but also from observers around the world. Accordingly, many might assume that terrorist attacks and fatalities were up sharply in the years leading up to the twenty-first century. But an examination of the GTD indicates that the patterns of terrorist attacks since 1970 are complex. Figure 3.1 shows trends in total attacks and fatal attacks from 1970 to 2008. According to Fig. 3.1, terrorist attacks reached their twentieth century zenith in 1992 (with over 5,100 attacks worldwide), declined substantially in the years leading up to the 9/11 attacks, and then began to increase sharply following 2003. So, our conclusions about terrorism trends depend greatly on our reference point. On the one hand, total attacks the year before the 9/11 attacks (1,379) were at about the same level as total attacks in 1977 (1,320). On the other hand, total attacks in 2008 (4,668) had once again reached levels that had not been seen since the early 1990s. Looking more broadly at overall trends, Fig. 3.1 shows that worldwide terrorist attacks through the mid-1970s were relatively infrequent, with fewer than 1,000 incidents each year. But from 1976 to 1979 the frequency of events nearly tripled. The number of terrorist attacks continued to increase until the 1992 peak, with smaller

3

Generating Terrorism Event Databases…

45

Fig. 3.1 Global terrorism attacks, 1970–2008 (n = 87,708)

peaks in 1984, at almost 3,500 incidents, and 1989, with over 4,300 attacks. After the first major peak in 1992, the number of terrorist attacks declined until the end of the twentieth century, before rising steeply to a 10-year high of nearly 4,700 in 2008 – 5 years after the start of the Iraq war. Still, total attacks in 2008 were 8.4% lower than total attacks for the 1992 peak. Fatal attacks also declined in the years prior to the 9/11 attacks. In fact, fatal attacks in 2000 (584) were considerably lower than they had been more than 2 decades earlier, in 1980 (1,113). In general, the number of fatal attacks clearly followed the pattern of total attacks (r = 0.93), but at a substantially lower magnitude (averaging 948 fatal attacks per year compared to 2,359 total attacks per year worldwide). Fatal attacks rose above 1,000 per year for the first time in 1980. After hovering close to 1,000 attacks annually for most of the 1980s, they more than doubled between 1988 and 1992. Like total attacks, fatal attacks declined somewhat after 1992, bottoming out in 1998 with 426 attacks and then rising again to a global peak of more than 2,100 fatal attacks in 2008. The 2008 peak was similar to the peak in 1992 (2,178). It is also important to be critical when interpreting these patterns. First, recall that most of the 1993 data in the GTD were lost by the original data collectors and we have never been able to recover them fully (see Footnote 1; and LaFree & Dugan, 2009). For Fig. 3.1 and the other trend analysis we estimate 1993 rates by taking the average value for 1992 and 1994. It turns out that between 1992 (the peak year) and 1994, there were steep declines in total terrorist attacks in the GTD (they dropped from 5,120 to 3,462). So, it could be that if we had the 1993 data, the shape of the drop in the early 1990s would look a bit different. According to the original PGIS

46

G. LaFree

report (Pinkerton Risk Assessment Services, 1995, p. 34), total attacks in 1993 were 4,954, suggesting only a slight decrease since 1992. Unfortunately, we cannot compare the 1993 figure directly to the results in Fig. 3.1 because when we computerized the original PGIS data we eliminated some cases that either did not fit the PGIS definition of terrorism or had been eliminated by additional information on incidents that was only available in more recent years. Moreover, we also added cases that were missed by the original data collectors. Second, by the time we received funding to extend the GTD data collection beyond 1997, it was already 2006. This meant that we were collecting data on events that were in some cases as much as 8 years old. We have since learned that the extensiveness of open source data on terrorism declines over time. Some smaller, local newspapers are not archived; some electronic sources are not indefinitely maintained. In short, our data likely undercount total attacks after 1997 and this undercount is likely most serious for 1998, and becomes less serious as our data collection process got closer to real time. Based on these considerations, I would offer the following conclusions about worldwide terrorist attacks over the past 4 decades based on the GTD. First, both total and fatal attacks increased steadily until reaching a peak in the early 1990s and then declined substantially after the collapse of the Soviet Union in 1991. Second, even though we likely undercount total attacks in 1998, the trajectory of attacks was already steeply down before that year and our ability to identify attacks likely increases as we get closer to the new data collection point in 2006. Third, despite the challenges of collecting open source event data, it is safe to conclude that during the 4 years prior to 9/11, worldwide terrorist attacks and fatal attacks were at the lowest level they had been at for 20 years. And finally, both total and fatal attacks have increased considerably since 9/11 so that in 2008 both total and fatal attacks were back to the historically high levels that they had been at in the mid-1990s.

The Global and Regional Distribution of Terrorism The ubiquity of modern communication systems means that individuals are now routinely bombarded by images of terrorist attacks from around the globe. This blanket coverage leaves the impression that no location on the planet is safe from terrorism. But in fact, our analysis of the GTD indicates that terrorist attacks are highly concentrated in geographic space. This concentration can be demonstrated at the national level by examining the proportion of all terrorist attacks that take place in those countries with the most terrorist activity. In Table 3.1, I present the top 20 countries in terms of terrorist attacks and compare the cumulative percentage of total attacks within these countries. According to Table 3.1, the top 20 countries and territories in terms of terrorist attacks account for nearly 71% of all terrorist activities in the world. The top ten countries account for more than 48% of all terrorist attacks and the top five countries account for more than 30% of the world’s terrorist attacks.

3

Generating Terrorism Event Databases…

47

Table 3.1 The 20 top ranking countries in terms of total terrorist attacks Rank Country Frequency Cumulative percentage 1 Colombia 6,911 11.06 2 Peru 6,041 20.73 3 El Salvador 5,330 29.26 4 India 4,799 36.94 5 Iraq 4,168 43.61 6 Northern Ireland 3,806 49.70 7 Spain 3,182 54.79 8 Pakistan 3,064 59.69 9 Philippines 2,733 64.06 10 Turkey 2,714 68.40 11 Sri Lanka 2,493 72.39 12 Chile 2,292 76.06 13 United States 2,264 79.68 14 Guatemala 2,032 82.93 15 Lebanon 1,997 86.12 16 Nicaragua 1,987 89.30 17 South Africa 1,921 92.37 18 Algeria 1,677 95.05 19 Afghanistan 1,542 97.51 20 Italy 1,494 99.97

Considerable concentration is also apparent if we examine terrorist attacks within smaller geospatial units. For example, in a recent analysis of terrorist attacks by ETA in Spain, LaFree, Dugan, Xie, and Singh (2011) found that the vast majority of attacks were concentrated in three Basque provinces in the North, Madrid, and Barcelona. Similarly, in an analysis of known terrorist attacks in India from the GTD, Cutter (2005) found that most attacks were concentrated in just two regions: the Punjab and Kashmir region, on the border with Pakistan, and the area around Bangladesh. The vast majority of Indian territory suffered few terrorist attacks from 1970 until quite recently – which is one of the reasons why the 2008 attacks on Mumbai were so shocking. To develop geospatial comparisons for terrorist attacks, we next divided the countries of the world into nine major regions (for a list of countries in each region, see Appendix).3 In Fig. 3.2, we compare the total distribution of attacks and fatalities by region. Figure 3.2 shows that by far the largest proportion of terrorist attacks

3

For this classification we treat the country or territory as the target. Thus, an attack on the U.S. embassy in Switzerland is treated here as a Swiss attack. Similarly, an attack on a Swiss ambassador living in the U.S. is counted here as a U.S. attack. Although the vast majority of cases in the GTD involve attacks where the location of the target and the nationality of the target are the same, there are some interesting variations across attacks depending on the geographical country attacked, the nationality of the perpetrators, and the nationality of the target. We are exploring these issues in much greater detail in ongoing research.

48

G. LaFree

35%

30%

Attacks Fatalities 25%

20%

15%

10%

5%

0% SubLatin Middle East Western South Asia Southeast North Saharan America & North Europe Asia & America Africa Africa Oceania

Eastern Europe

East & Central Asia

Fig. 3.2 Percent distribution of attacks and fatalities by region, 1970–2008

in the GTD (32%) occurred in Latin America. The Middle East–North Africa (17%), Western Europe (16%), and South Asia (15%) all had roughly similar rates and together, account for another 48% of all terrorist attacks. The countries of SubSaharan Africa and Southeast Asia-Oceania have the next highest levels of total attacks, accounting respectively for nearly 6 and 5% of the total. None of the remaining regions account for more than 5% of total attacks. North America accounts for nearly 3% of the total, Eastern Europe for 2.7%, and East and Central Asia for 1.2% of total attacks. Figure 3.2 also shows substantial variation across regions in terms of the relationship of attacks to fatalities. In particular, Sub-Saharan Africa, South Asia, and the Middle East–North Africa stand out for having a larger percent of total fatalities than attacks while Western Europe (in particular) stands out for having a larger percent of attacks than fatalities. When we looked more closely at the fatality rates for Sub-Saharan Africa, we found that three countries averaged ten or more fatalities per attack: Mozambique (28.2%), Chad (27.2%), and Burundi (24.1%). Much of the terrorist violence in Burundi was driven by the conflict between the Tutsis and the Hutus. The Lord’s Resistance Army in Uganda also staged a series of terrorist attacks with large numbers of fatalities during the period spanned by the data. By contrast, groups like the IRA and ETA operating in Western Europe staged large numbers of attacks with relatively few fatalities. We next examine trends in regional terrorism rates over time. Figure 3.3 shows the trends for the four most highly active regions: Latin America, the Middle

3

Generating Terrorism Event Databases…

49

Fig. 3.3 Terrorist attacks by region, 1970–2008 (high-frequency regions; n = 70,697)

East–North Africa, Western Europe, and South Asia. Latin America includes all of the countries in Central and South America and the island nations of the Caribbean (see Appendix). The Middle East–North Africa region includes 24 countries and territories and Western Europe includes 26 countries. South Asia includes 15 countries and territories including several with large numbers of terrorist attacks. Perhaps the most obvious conclusion from Fig. 3.3 is that attacks in the highly active regions vary greatly over time. For the years 1970–1979, terrorism was dominated by attacks taking place in Western Europe. Well-known attacks include terrorist campaigns by Republicans and Loyalists in Northern Ireland, the Red Brigades in Italy, and ETA in Spain. However, after peaking in 1979 with 1,015 attacks, terrorism in Western Europe drops to an average of 350 attacks per year by 1981. By contrast, worldwide terrorism in the 1980s was in large part a Latin American phenomenon. Annual attacks in Latin America rose precipitously in the late 1970s, reaching a peak in 1984 with over 2,150 attacks – driven especially by Sendero Luminoso in Peru, the FMLN in El Salvador, and the FARC and the Manuel Rodriguez Patriotic Front (FPMR) in Chile. After 1984, Latin America averages about 1,700 attacks per year through 1991 but with large fluctuations. It is not until the mid-1990s that total Latin American attacks fell below total attacks in the Middle East–North Africa (594) and South Asia (1,084).

50

G. LaFree

Fig. 3.4 Terrorist attacks by region, 1970–2008 (low-frequency regions; n = 18,044)

From 2002 to the end of the series, the Middle East–North Africa and South Asia overtake Western Europe and Latin America to have the highest number of terrorist attacks per year. In 2008, the GTD records a total of 1,691 attacks for the countries of South Asia and 1,448 attacks for the countries of the Middle East– North Africa. Combined, the countries of South Asia and the Middle East/Persian Gulf had nearly 10 times more terrorist attacks in 2008 than the combined totals for the countries of Latin America and Western Europe. To summarize, global terrorist hot spots moved from Western Europe in the 1970s, to Latin American in the 1980s and finally to the Middle East–North Africa and South Asia in the beginning of the twenty-first century. Figure 3.4 shows the remaining five regions of the world; those with on average fewer terrorist attacks than the four high frequency regions just reviewed. Among this group of regions with relatively low attack frequencies, North America dominated in the 1970s. Leading groups in the database from the United States and Canada in the 1970s included Fuerzas Armadas de Liberacion Nacional (FALN; n = 86) and the New World Liberation Front (NWLF; n = 87). By contrast, the 1980s were dominated by attacks from the countries of SubSaharan Africa and South East Asia–Oceania. Especially active organizations from Sub-Saharan Africa during this period include the African National Congress

3

Generating Terrorism Event Databases…

51

(n = 546) and the National Union for the Total Independence of Angola (UNITA; n = 248). Active organizations from Southeast Asia–Oceania during the period include the New People’s Army (NPA; n = 768) and the Moro National Liberation Front (MILF; n = 95). Finally, among this group of generally less active regions, the regions with the greatest rise in terrorist attacks since 2004 are Southeast Asia–Oceania, Sub-Saharan Africa, and Eastern Europe. Since 2004, the most active organizations in Southeast Asia–Oceania include the New People’s Army (n = 135) and the Moro Islamic Liberation Front (n = 67); in Sub-Saharan Africa, the Lord’s Resistance Army (n = 40) and Al-Shabaab (n = 28); and in Eastern Europe, the Chechen rebels (n = 49).

The Most Active Terrorist Organizations In Table 3.2 we present rankings of the 20 most active terrorist organizations found in the GTD and the 20 that claimed the most fatalities. While the specific rank orderings of the most active and deadliest vary, 14 of the top 20 most active are also in the top 20 for deadliest. The six organizations that are in the top 20 for most active but not in the top 20 for deadliest are ETA, the Manuel Rodriguez Patriotic Front (Chile), the African National Congress (South Africa), the Corsican National Liberation Front (France), the Tupac Amaru Revolutionary Movement (Peru), and the Movement of the Revolutionary Left (Chile). The six organizations that are among the 20 deadliest but not among the top 20 most active include Al-Qaida and Al-Qaida in Iraq (counted here as separate organizations), the Mozambique National Resistance Movement, the Democratic Revolutionary Alliance (Nicaragua), the Lord’s Resistance Army (Uganda), and the Armed Islamic Group (Algeria). The range of attacks among the top 20 most active terrorist organizations is large: number one Shining Path has over 4,500 known attacks while number 20 Hizballah has less than 300 known attacks. Similarly, the GTD attributes more than 11,000 fatalities to Shining Path compared to fewer than 900 fatalities to Hizballah. Altogether, nine terrorist groups in the GTD have over 1,000 attacks and 18 groups claimed more than 1,000 lives during the 39 years spanned by the data. One of the most interesting observations about these two lists of terrorist organizations is how many of the groups are organized around disputes having to do with political control over territory. Although there are major differences in terms of their orientation, this explains in large part virtually all of the top 20 groups, including Shining Path, ETA, the IRA, FARC, Hamas, and the LTTE. South Africa is an especially striking example of this phenomenon. Both pro- and antiapartheid terrorist attacks in South Africa resulted in an enormous amount of political violence during much of the 1980s and early 1990s. However, in an analysis of terrorist attacks in South Africa using GTD data, Van Brakle and LaFree (2007) found that violence ended rapidly following the first democratic elections at the end of apartheid. In fact, a major reason why it is so difficult to develop a universal definition of terrorism is that it is often a method used to advance political goals on which nations and individuals disagree.

52 Table 3.2 The 20 most active 1970–2008 Most frequent perpetrators Organization Shining Path (SL) Farabundo Marti National Liberation Front (FMLN) Irish Republican Army (IRA) Basque Fatherland and Freedom (ETA) Revolutionary Armed Forces of Colombia (FARC) National Liberation Army of Colombia (ELN) Liberation Tigers of Tamil Eelam (LTTE) Kurdistan Workers’ Party (PKK) New People’s Army (NPA) Taliban

Nicaraguan Democratic Force (FDN)

G. LaFree terrorist organizations in terms of attack frequency and fatalities,

Frequency 4,513 3,357

2,671

Most fatalities Organization Shining Path (SL) Liberation Tigers of Tamil Eelam (LTTE)

Fatality count 11,6477 9,534

Farabundo Marti National Liberation Front (FMLN) Nicaraguan Democratic Force (FDN) Revolutionary Armed Forces of Colombia (FARC)

8,508

1,258

Al-Qa`ida

4,299

1,253

Kurdistan Workers’ Party (PKK) New People’s Army (NPA)

3,558

1,991 1,668

1,173 1,168 977

900

Taliban National Union for the Total Independence of Angola (UNITA) Mozambique National Resistance Movement (MNR) Irish Republican Army (IRA)

Manuel Rodriguez Patriotic 830 Front (FPMR) African National Congress 606 Democratic Revolutionary (South Africa) Alliance (ARDE) Corsican National 569 Al-Qa`ida in Iraq Liberation Front (FLNC) Tupac Amaru 557 Lord’s Resistance Army Revolutionary (LRA) Movement (MRTA) M-19 (Movement of 554 National Liberation Army of April 19) Colombia (ELN) People’s Liberation Front 434 M-19 (Movement of April 19) (JVP) National Union for the 421 Armed Islamic Group (GIA) Total Independence of Angola (UNITA) People’s Liberation Front Movement of the 421 (JVP) Revolutionary Left (MIR) (Chile) Hizballah 293 Hizballah Note: highlighted groups appear in both classifications

7,268 4,835

3,330 2,867 2,562

2,443

1,829 1,803 1,607

1,596

1,449 1,323 1,092

891

837

3

Generating Terrorism Event Databases…

53

Terrorist Fatalities Figure 3.5 shows the distribution of fatalities per attack. Some may be surprised by the fact that in nearly 60% of the attacks there were no fatalities. In many cases terrorist groups target property and do not intend to cause casualties. In other cases, they plan to cause casualties but fail. Moreover, some well-known terrorist groups such as the IRA and ETA frequently provide warnings before attacks to minimize casualties. Thirty-five years ago these considerations led Jenkins (1975, p. 12) to a suggestion that “terrorists want a lot of people watching, not a lot of people dead.” Of course, it is still the case that more than 40% of the attacks in the GTD (or more than 36,000 attacks) involved at least one fatality. Attacks that are especially worrisome are the 112 attacks that produced more than 100 fatalities. And in fact, Jenkins (2007) has recently revisited his earlier statement and after reviewing the stated plans of terrorist groups operating in the early twenty-first century concluded that indeed “many of today’s terrorists want a lot of people watching and a lot of people dead.” Nevertheless, the majority of terrorist attacks in the GTD since 1970 produced no fatalities.

Terrorist Targets Figure 3.6 presents the distribution of terrorist targets worldwide. We can see that there is considerable variation in terrorist targeting with the most common target (private citizens) representing just over 20% of the total. Together, private citizens

55000

51,506 (58.7%)

50000

Number of Attacks

45000 40000 35000

28,299 (32.2%)

30000 25000 20000 15000 10000 5000

3,279 (3.7%)

3,427 (3.9%)

6 to 9

10 to 25

799 (0.9%)

288 (0.3%)

112 (0.1%)

26 to 50

51 to 100

over 100

0 0

1 to 5

Number of Fatalities

Fig. 3.5 Fatalities per attack, 1970–2008 (n = 87,708)

54

G. LaFree

Fig. 3.6 Distribution of terrorist targets globally, 1970–2008 (n = 87,708)

and businesses account for more than 38% of all terrorist attacks. The next most common targets are the military, the government, and the police.4 Utilities and transportation were each targeted about 5% of the time. The remaining targets are attacked even less frequently, and include diplomats, religious figures or institutions, journalists and other media, and educational institutions. Important target types in the “other” category include other terrorists and criminals. In Fig. 3.7 we present the percentages of terrorist targets for each region. We limit the comparisons here to the four most common targets. According to Fig. 3.7, private citizens are the modal target for Sub-Saharan Africa, the Middle East–North Africa, South Asia, and Southeast Asia–Oceania. Many attacks on private citizens took place in public spaces, such as market places and sidewalk cafés. The most common target in East and Central Asia and Eastern Europe is government. Many of these attacks were directed toward government offices, or the homes or vehicles of government officials. Businesses are most commonly targeted in Latin America,

4

The GTD generally includes attacks against the police only if they were not acting in concert with and under the direct authority of a military unit. We also include attacks on the military, if they occur either in situations where the military personnel were not performing military duties (such as the 2005 attacks in Bali) or where they were not in an active military theater (such as the 2000 attack on the USS Cole).

3

Generating Terrorism Event Databases…

55

Fig. 3.7 Targets of terrorism, 1970–2008 (n = 63,483)

North America, and Western Europe, followed closely by the military in Latin America, private citizens in Western Europe, and government in North America. Transportation infrastructure was among the top four targets in East and Central Asia and utilities were among the top four targets in Latin America. Note that compared to other regions, terrorists operating in the Middle East and North Africa are more likely to target private citizens and police and less likely to target government and businesses.

Terrorist Tactics We turn now to a discussion of the tactical choices made by terrorists. Figure 3.8 shows the total distribution of terrorist tactics. Perhaps, the most striking result here is that more than two-fifth of all terrorist attacks are bombings and nearly a quarter is armed assault. Taken together, bombings and armed assaults account for nearly 70% of the total. Because armed assaults require weapons rather than explosives, terrorists in these cases are more likely to have direct contact with targets. The next most common tactic is assassination, followed by facility attacks and kidnappings. Assassinations are about 3 times more common than facility attacks and kidnappings. Barricade/hostage situations, unarmed assaults, and hijackings each account for less than 1% of the total. In Fig. 3.9 we present distributions of the five most common terrorist tactics by region. Perhaps, the most striking comparison is between bombings and armed assaults. While in the aggregate bombings are nearly twice as common as armed

56

G. LaFree Hostage/Barricade 0.80%

kidnapping 4.66%

Unarmed Assault 0.69%

Unknown 2.92%

Hijacking 0.44% Facility Attack 6.88% Armed Assault 24.68%

Assassination 15.24% Bombing 43.69%

Fig. 3.8 Distribution of terrorism tactics, 1970–2008 (n = 87,708)

Fig. 3.9 Distribution of the four most common tactics by region, 1970–2008 (n = 77,410)

3

Generating Terrorism Event Databases…

57

Fig. 3.10 Global frequency of terrorist tactics, 1970–2008 (n = 81,071)

assaults (see Fig. 3.8) Fig. 3.9 shows that there is substantial regional variation in bombings. Bombings were especially common in Eastern Europe, the Middle East– North Africa, North America, and Western Europe, but somewhat less common in Southeast Asia–Oceania, Latin America, East and Central Asia, South Asia, and Sub-Saharan Africa. In fact, for Sub-Saharan Africa and Southeast Asia–Oceania, armed assaults were about as common as bombings. In Western Europe, assassinations were more common than armed assaults. Kidnappings were most common in Southeast Asia–Oceania and Sub-Saharan Africa, least common in Western Europe. Facility attacks were considerably more common in North America and East and Central Asia than in other regions. In Fig. 3.10 we show trends in the five major terrorist tactics over time. Bombing is the most common tactic throughout the entire period. Armed assaults were the second most common tactic throughout the series with the exceptions of a short period between 1972 and 1973, when they were overtaken by assassinations. The gap between bombings and armed assaults widened throughout the 1980s and then closed considerably after the early 1990s. However, bombings end the series in 2008 with a commanding lead over all other tactics. Assassinations rose steadily throughout much of the 1980s, reaching a series peak in 1989 with a total of 980 attacks. Assassinations then declined into the present century and in 2008 they were the least common of the five tactics. Facility attacks are the fourth most common of

58

G. LaFree

the tactics surveyed here. They reached a peak in 1992 with 493 attacks. And finally, kidnappings are the least common of the five tactics included. Kidnappings experienced their series high point in 2008 with 357 attacks. The two groups with the largest number of kidnappings in that year were the Taliban of Afghanistan (47), the Communist Party of India–Maoist (19), and the Tehrik-i-Taliban of Pakistan (12).

Terrorist Weapons Figure 3.11 shows the weapons used in the terrorist attacks included in the GTD. Not surprisingly, explosives and firearms are the dominant weapons used by terrorists, jointly accounting for over 80% of all attacks. The most common explosives used are dynamite, car bombs, grenades, and mortars. The most common firearms used are shot guns, pistols, and automatic weapons. Incendiaries (including fire and firebombs) contribute nearly 8% of total weapons. Melee attacks, where the perpetrator comes into direct contact with the target, account for just over 2% of the total. These attacks usually depend on low technology weapons such as knives, or even stones or fists. In short, most of the weapons used in these cases were conventional, low technology, and readily available. Fortunately, sophisticated weaponry, especially chemical, biological, and nuclear weapons are quite rare, accounting for less than three-tenth of 1% of all attacks. In nearly 9% of the attacks there was not enough detail to classify weapon type.

Fig. 3.11 Weapons used in terrorist attacks, 1970–2008 (n = 87,708)

3

Generating Terrorism Event Databases…

59

Fig. 3.12 Weapon use across regions, 1970–2008 (n = 79,720)

In Fig. 3.12 we examine the distribution of the five most common weapons over the nine regions of the world. According to Fig. 3.12, explosives are the modal weapon of choice for terrorists operating in Eastern Europe, Western Europe, the Middle East–North Africa, East and Central Asia, and North America whereas firearms are the most common weapon used in South East Asia–Oceania, Latin America, South Asia, and Sub-Saharan Africa. Explosives and firearms account for the highest proportion of total attacks in Southeast Asia–Oceania (80%) and the lowest proportion of total attacks in East and Central Asia (54%). Incendiaries are disproportionately common in East and Central Asia and in North America. Melee style weapons are the least common of these four weapon types. They appear most frequently in attacks in East and Central Asia where they constitute 6% of total attacks and in the Middle East and North Africa where they constitute 4% of total attacks. In Fig. 3.13 we examine trends in weapon use over time. Here, explosives and firearms are closely related (r = 0.85). Explosives were more common than firearms for much of the 1970s and the first decade of the twenty-first century; firearms were more common than explosives throughout much of the 1990s. In part, these patterns reflect regional differences in terrorist attacks. Explosives are the most common weapon for terrorists in Western Europe, which dominated terrorism rates in the 1970s, and the Middle East–North Africa, which has been the location for a high proportion of terrorist attacks since 2000. By contrast, Sub-Saharan Africa and Latin America had a good deal of terrorist activity in the 1990s, a period where firearms were a very common weapon for terrorist attacks. The large increase in explosives and firearms use in the last few years of the series is in part a consequence of attacks from Iraq following the 2004 US-led invasion. The use of incendiary

60

G. LaFree

Fig. 3.13 Terrorist weapon use, 1970–2008 (n = 83,578)

devices increased slowly from the 1970s, reaching a series peak in 1992 and then falling off somewhat. Melee attacks are concentrated in the late 1980s and early 1990s. This reflects in large part the importance of terrorist attacks in the Middle East–North Africa and in East and Central Asia during this period.

Discussion and Conclusions The scope of open source databases on terrorist attacks has greatly expanded since the early 1970s. As I have tried to demonstrate in this short review, the GTD contributes to this development by creating a comprehensive open source database on terrorist attacks that now spans 4 decades. Strengths of this database are that it was collected by a nongovernmental organization using a consistent coding framework throughout the process and that it includes both domestic and transnational attacks. However, open source databases like the GTD that are generated from the print and electronic media also have important limitations. In particular, the media may report inaccuracies and falsehoods. In some cases there may be conflicting information or false and multiple or no claims of responsibility. The database originated with predominantly English-speaking researchers located in western democracies and

3

Generating Terrorism Event Databases…

61

despite our efforts, the coverage no doubt still relies more on western than nonwestern sources. Government censorship and disinformation may also affect results. It is especially challenging to disentangle terrorism from acts of war, insurrection, or massive civil unrest. Of all the complexities of separating terrorist violence from war-related violence, perhaps the most vexing is distinguishing between armed targets with civilian collateral damage and civilian attack with noncivilian casualties. This is an especially big problem in interpreting terrorism in Iraq, following the US-led invasion in 2003. Even though the media now seemingly peer into every corner of the world, media coverage still varies across time and geographic space. A promising method for assessing the quality of open source databases on terrorism would be to do more systematic comparisons between event data on terrorism and terrorism data drawn from other sources of the type represented by Sheehan in this volume. Thus far, there have been very few studies of this type owing to many of the methodological problems we discussed above in the section on defining and measuring terrorism: no universally accepted definition of terrorism, the absence of international data from official sources, and difficulties in conducting victimization or self report surveys. An exception is a recent study by Chermak, Freilich, Parkin, and Lynch (2011) in which the authors compare estimates of terrorism and violent extremist crime for the United States from ten open source event databases (including the GTD). Based on the inclusion criteria for each database, the authors examine whether sources correctly included or incorrectly excluded specific attacks. The authors find considerable variation in the number of events captured by the different sources and further, that some sources include events that appear to be contrary to their inclusion criteria and other sources exclude events that appear to meet their criteria. Importantly, though, the authors conclude that the general attributes of victim, suspect, and incident are surprisingly similar across diverse data sources. More comparisons between open source databases like the GTD and other official and unofficial event databases would be helpful. It would also be possible to examine media sources used in the GTD to look for differential patterns of bias and incorrect or incomplete reporting. In addition, much primary data are collected by intelligence agents, including data from communications intercepts, surveillance, informers, defectors, interrogation of prisoners, and captured internal documents (e.g., memos, training manuals). While most of these sources are not readily available to researchers working in an open, unclassified environment, there are important opportunities provided by official data on terrorism that have not been adequately exploited. In particular, researchers could do more to examine court records (as suggested by Damphousse, this volume) and transcripts, government reports and hearings, and unclassified intelligence reports. While the GTD has obvious limitations, it also offers a wide variety of analysis opportunities. GTD data are likely to be particularly useful for assessing the impact of specific policies or events on the future risk of terrorist activity of a particular type. Thus, we can use the database to examine the impact of specific counterterrorism policies on specific terrorist groups in specific countries over time. The data have particular promise for geospatial analysis (Behlendorf, LaFree, & Legault, 2011; LaFree et al., 2011). The data can also be merged with other databases to allow analysis of

62

G. LaFree

global or regional determinants of terrorist attacks or to examine the effect of global or regional terrorist attacks on other variables (Fahey, LaFree, Dugan, & Piquero, 2011). We close by offering two general areas where future research could build directly on terrorism event databases like the GTD. First, while considerable progress has been made in collecting more objective data on terrorism and its aftermath, there is quite often little theoretical or conceptual grounding for resulting analyses. The variation in event databases for variables like the decision of terrorists to target or kill, by which tactic, and using what type of weapon are stark. To better understand these variations, much more research is needed that integrates data from event databases with theoretical and conceptual insights from the social and behavioral sciences, including sociocultural studies of the influence of ideology on strategy and tactics, theoretical lessons gained from formal decision analysis, research on social movements, and collective violence and insights from organizational and social psychology. Second, as this volume clearly demonstrates, there is a dearth of social science research on how specific antiterrorism and counterterrorism efforts impact the behavior and activities of terrorists or potential terrorists as chronicled by event databases like the GTD. We need far more information on the impact of counterand antiterrorism interventions and their expected impact on trajectories of terrorist activities. Such research could help us better understand what separates successful and unsuccessful counter measures, what factors predict the countermeasures implemented by governments, and whether the same counter measures may have different impacts when implemented in response to different groups. Acknowledgments Prepared for Evidence-Based Counterterrorism, edited by Cynthia Lum and Les Kennedy. Support for this research was provided by the Department of Homeland Security through the National Consortium for the Study of Terrorism and Responses to Terrorism (START), grant number N00140510629. Any opinions, findings, and conclusions or recommendations in this document are those of the authors and do not necessarily reflect views of the Department of Homeland Security. I want to thank Erin Miller and Sumit Kumar for database support and Cynthia Lum, Leslie Kennedy, and several anonymous reviewers for helpful editorial suggestions.

Appendix Countries Listed Within Each Region Region East and Central Asia

Countries/territories China, Hong Kong, Japan, Kazakhstan, Kyrgyzstan, Macao, North Korea, South Korea, Taiwan, Tajikistan, Turkmenistan, and Uzbekistan

Eastern Europe

Albania, Armenia, Azerbaijan, Bosnia-Herzegovina, Bulgaria, Belarus, Croatia, Czechoslovakia, Czech Republic, Estonia, Georgia, Hungary, Latvia, Lithuania, Kashmir, Kosovo, Macedonia, Moldova, Montenegro, Poland, Romania, Russia, San Marino, Serbia, SerbiaMontenegro, Slovak Republic, Slovenia, Soviet Union, Ukraine, and Yugoslavia (continued)

3

Generating Terrorism Event Databases…

63

Appendix (continued) Region Countries/territories Latin America

Antigua and Barbuda, Argentina, Bahamas, Barbados, Belize, Bermuda, Bolivia, Brazil, Cayman Islands, Chile, Colombia, Costa Rica, Cuba, Dominica, Dominican Republic, Ecuador, El Salvador, Falkland Islands, French Guiana, Grenada, Guadeloupe, Guatemala, Guyana, Haiti, Honduras, Jamaica, Martinique, Nicaragua, Panama, Paraguay, Peru, Puerto Rico, St. Kitts and Nevis, Suriname, Trinidad and Tobago, Uruguay, Venezuela, and the Virgin Islands (US)

Middle East and North Africa

Algeria, Bahrain, Cyprus, Egypt, Iran, Iraq, Israel, Jordan, Kuwait, Lebanon, Libya, Morocco, North Yemen, Oman, Qatar, Saudi Arabia, South Yemen, Syria, Tunisia, Turkey, United Arab Emirates, West Bank and Gaza Strip, Western Sahara, and Yemen

North America

Canada, Mexico, and the United States

South Asia

Afghanistan, Bangladesh, Bhutan, India, Maldives, Mauritius, Nepal, New Caledonia, Pakistan, Seychelles, Sri Lanka, Tonga, Vanuatu, Wallis and Futuna, and Western Samoa

Southeast Asia and Oceana

Australia, Brunei, Cambodia, Fiji, French Polynesia, Guam, Indonesia, Laos, Malaysia, Myanmar, New Caledonia, New Hebrides, New Zealand, Papua New Guinea, Philippines, Samoa (Western Samoa), Solomon Islands, Singapore, South Vietnam, Thailand, Timor-Leste, Vanuatu, Vietnam, and Wallis and Futuna

Sub-Saharan Africa

Angola, Benin, Botswana, Burkina Faso, Burundi, Cameroon, Central African Republic, Chad, Comoros, Congo (Brazzaville), Congo (Kinshasa), Djibouti, Equatorial Guinea, Eritrea, Ethiopia, Gabon, Gambia, Ghana, Guinea, Guinea-Bissau, Ivory Coast, Kenya, Lesotho, Liberia, Madagascar, Malawi, Mali, Mauritania, Mauritius, Mozambique, Namibia, Niger, Nigeria, Rwanda, Senegal, Sierra Leone, Somalia, South Africa, Sudan, Swaziland, Tanzania, Togo, Uganda, Zaire, Zambia, and Zimbabwe

Western Europe

Andorra, Austria, Belgium, Corsica, Denmark, East Germany, Finland, France, Germany, Gibraltar, Great Britain, Greece, Iceland, Ireland, Italy, Luxembourg, Malta, Isle of Man, Netherlands, Northern Ireland, Norway, Portugal, Spain, Sweden, Switzerland, and West Germany

References Behlendorf, B., LaFree, G., & Legault, R. (2011). Predicting microcyles of violence: Evidence from terrorist attacks by the FMLN and ETA. University of Maryland: Unpublished manuscript. Chermak, S. M., Freilich, J. D., Parkin, W. S., & Lynch, J. P. (2011). Comparing data sources of American terrorism and extremist crime: Investigating selectivity bias. Michigan State University: Unpublished manuscript. Cutter, S. (2005). Geo-coded terrorism attacks in India. University of South Carolina: Unpublished manuscript. European Commission. (2008). Defining terrorism. Transnational terrorism, security and the rule of law. Retrieved October 10, 2008, from http://www.transnationalterrorism.eu. Fahey, S., LaFree, G., Dugan, L., & Piquero, A. (2011). Situational determinants of terrorist and nonterrorist aerial hijackings. Justice Quarterly. DOI: 10.1080/07418825.2011.583265.

64

G. LaFree

Hoffman, B. (2007). The new age of terrorism. RAND: National Security Research Division. Retrieved from http://www.rand.org/pubs/reprints/2006/RAND_RP1215.pdf. Jenkins, B. M. (1975). International terrorism: A new model of conflict. In D. Carlton & C. Schaerf (Eds.), International terrorism and world security. London: Croom Helm. LaFree, G., & Dugan, L. (2009). Tracking global terrorism, 1970–2004. In D. Weisburd, T. Feucht, I. Hakimi, L. Mock, & S. Perry (Eds.), To protect and to serve: Police and policing in an age of terrorism (pp. 43–80). New York: Springer. LaFree, G., Dugan, L., Xie, M., & Singh P. (2011). Geospatial and temporal patterns of terrorist attacks by ETA, 1970 to 2007. Journal of Quantitative Criminology. DOI: 10.1007/s10940-011-9133-y. Merari, A. (1991). Academic research and government policy on terrorism. Terrorism and Political Violence, 3, 88–102. Pinkerton Global Intelligence Service. (1995). Annual report for 1993. Washington: Unpublished. Schmid, A., & Jongman, A. J. (1988). Political terrorism: A new guide to actors, authors, concepts, databases, theories and literature. Amsterdam: North-Holland. Sheehan, I. S. (2011). Assessing and comparing terrorism sources. In C. Lum & L. Kennedy (Eds.), Evidence-based counter terrorism. New York: Springer. Smelser, N. (2007). The faces of terrorism: Social and psychological dimensions. Princeton: Princeton University Press. Smith, B. L., Damphouse, K. R., Jackson, F., & Sellers, A. (2002). The prosecution and punishment of international terrorists in federal courts: 1980–1998. Criminology and Public Policy, 1, 311–338. Van Brakle, M., & LaFree, G. (2007). Rational choice and terrorist target selection: Lessons from the South African experience. Paper presented at the annual meetings of the American Society of Criminology. Atlanta, Georgia.

Chapter 4

Evidence-Based Intelligence Practices: Examining the Role of Fusion Centers as a Critical Source of Information Jeremy Carter and Steven Chermak

The September 11 attacks impacted society generally, and law enforcement specifically, in dramatic ways. One of its major impacts has been changing expectations regarding criminal intelligence practices among state, local, and tribal (SLT) law enforcement agencies in the United States, and the need to coordinate new intelligence efforts and share information at all levels of government. In fact, enhancing intelligence efforts has emerged as a critical issue for the prevention of terrorist acts. There have been fundamental changes in the national, state, and local information sharing infrastructure, including efforts to expand and enhance intelligence gathering. One major innovation to better coordinate efforts across levels of government is the creation of state-run fusion centers. At present, there are 72 official fusion centers funded primarily by the Department of Homeland Security (Saari, 2010). Despite dramatic changes to the information sharing infrastructure and the subsequent growth of fusion centers, and the acknowledgement that local intelligence is critical to the prevention and deterrence of terrorist acts, very little research exists in this area. In addition, there has been little systemic examination of evidencebased practices in the area of information sharing. This chapter draws from fusion center personnel survey data to consider three important issues related to enhancing evidence-based practices within fusion centers. First, we explore the current state of information sharing and communication among agencies and fusion centers. This is important in that such an exploration begins to establish a need to identify best practices for enhancing the flow of relevant and timely intelligence and considering obstacles by documenting the current experiences of SLT agencies in building an intelligence capacity. Second, we explore the type of information, data, and analysis

J. Carter (*) Department of Criminology and Criminal Justice, University of North Florida, Jacksonville, FL, USA e-mail: [email protected] S. Chermak School of Criminal Justice, Michigan State University, East Lansing, MI, USA C. Lum and L.W. Kennedy (eds.), Evidence-Based Counterterrorism Policy, Springer Series on Evidence-Based Crime Policy 3, DOI 10.1007/978-1-4614-0953-3_4, © Springer Science+Business Media, LLC 2012

65

66

J. Carter and S. Chermak

currently used within fusion centers. We are particularly interested in understanding the types of projects completed by fusion centers and how this research impacts policies and procedures. Third, we discuss opportunities for research-practitioner partnerships and recommend ways that scholarly attention to intelligence issues and terrorism generally could enhance understanding of intelligence practices and move us toward a more scientific approach to the study of law enforcement intelligence.

The Current State of Information Sharing for Counterterrorism in the United States Since the terrorist attacks of September 11, there has been a considerable investment of resources across federal, state, and local sectors to better prepare, respond, and recover from terrorist acts. One critical commitment has been in improving the law enforcement intelligence capacity at all levels of government. The changes in the intelligence practices for SLT agencies have been particularly pronounced. Many law enforcement agencies had eliminated their intelligence units, starting in the late 1960s, in reaction to a proliferation of section 1983 civil rights lawsuits, in which agencies were being found liable and forced to pay monetary damages (Carter, 2009). While there had been some restructuring of local law enforcement information sharing capabilities beginning in the 1980s to increase “tips and leads,” largely as a result of the global drug trade, this restructuring was largely limited to major cities. Consequently, after the September 11 attacks, most law enforcement agencies had to develop an intelligence capacity (i.e., capability to collect, analyze, and share information) from the ground up at a time when the threats they faced and the discipline of intelligence had changed dramatically. In fact, every law enforcement agency, regardless of size, has had to build an intelligence capacity to “understand the implications of information collection, analysis, and intelligence sharing,” and “must have an organized mechanism to receive and manage intelligence as well as a mechanism to report and share critical information with other law enforcement agencies” (Carter, 2009, pp. 1). The development of this capacity has resulted in a significant expansion of the intelligence function in law enforcement agencies, the institutionalization of intelligence units, and a significant need for providing intelligence training to all levels of law enforcement. These changes have also resulted in the widespread creation of intelligence fusion centers and expectations that such centers will play a critical role in preventing terrorism. The growth and evolution of intelligence practices (i.e., connectivity, formal policies and procedures, and privacy protection) has coincided with an increasing acknowledgment within various levels of government of the importance of SLT law enforcement for enhancing the value of intelligence related to terrorism, and the coordination of information sharing across levels of government. Congress made it generally clear in the Homeland Security Act of 2002 that state and local information was critical for preventing and preparing for terrorist events, and that federal,

4

Evidence-Based Intelligence Practices…

67

state, and local entities should work to embrace strategies that would dramatically increase the sharing of information (see General Accounting Office, 2003, 2007a; President’s National Strategy on Information Sharing, 2007). This conclusion highlights the appreciation that each level of government best understands its specific environment and is in a position to harness its various assets in the most effective way to accomplish the broad goals of responding effectively to terrorism. There needs to be, however, some mechanism for the coordination of these enhanced efforts. Fusion centers were developed to play this role – to “fuse” information together from disparate sources. The importance of SLT’s contribution to the intelligence process and considering how this information can be used by other levels of government can be highlighted in several ways. First, the National Strategy for Information Sharing (2007), established by the White House under President George W. Bush, highlights the importance of sharing threat information with many sectors of society. Specifically, it highlights the need for SLT law enforcement agencies to foster a culture of fusing information on crime and terrorist-related incidents, support efforts to detect and prevent attacks, and develop training and awareness programs on terrorism. Second, although the Federal Bureau of Investigation (FBI) is the lead agency for investigating terrorism, the types of information provided by various sources (i.e., fusion centers, state and local law enforcement, private sector, etc.) and the sheer number of cases and leads requiring follow-up is simply too burdensome for the FBI alone. Such a strain on personnel highlights the importance of involving SLT law enforcement in terrorist investigations (Davis et al., 2004). An emphasis to focus on terrorism as a local event has been well documented (Boba, 2009; Clarke & Newman, 2006; Newman & Clarke, 2008), SLT law enforcement agencies are in a unique position to contribute important intelligence because of their knowledge about individuals, groups, and organizations operating in local communities as part of their day-to-day operational work. As Bayley and Weisburd (2009) noted, “low policing” that focuses on street-level interactions among law enforcement and community members is perhaps one of the most significant benefits police have in their counterterrorism efforts. Specifically, the authors note that “uniformed police have more opportunities to observe activities that may be associated with terrorism than specialists, especially specialists not deployed routinely in local areas (Bayley & Weisburd, 2009, pp. 91).” A report for the Congressional Research Service states, “The 800,000 plus law enforcement officers across the country know their communities most intimately and, therefore, are best placed to function as the ‘eyes and ears’ of an extended national security community. They have the experience to recognize what constitutes anomalous behavior in their areas of responsibility and can either stop it at the point of discovery (a more traditional law enforcement approach) or follow the anomaly or criminal behavior, either unilaterally or jointly with the FBI, to extract the maximum intelligence value from the activity (a more intelligence-based approach)” (Masse, O’Neil, & Rollins, 2007, pp. 7). While partnerships for utilizing the community for information gathering have found success in the last decade – such as the National Institute of Justice-funded Community Mapping, Planning, and Analysis for Safety Strategies (COMPASS)

68

J. Carter and S. Chermak

and the Strategic Approaches to Community Safety Initiative (SACSI) projects, which took place in the early 2000s – there is currently a lack of community engagement with respect to counterterrorism. A recent study by Lum, Haberfield, Fachner, and Lieberman (2009) reviewed current empirical counterterrorism studies, focusing largely on three empirical examinations,1 as well as presented original data from a random sample of local US law enforcement. The three studies shared a commonality that most state and local law enforcement agencies are shifting resources toward counterterrorism prevention. However, this shift is concentrated heavily on larger agencies that focus on strategic planning for large cities and critical infrastructures. Such an approach is a step in the right direction. However, it was noted in the LEMAS study that agencies are more likely to focus their community efforts on an increased presence at critical areas (i.e., infrastructure) rather than developing partnership campaigns (Lum et al., pp. 108) – which serve as information gathering mechanisms. In their own empirical exploration, Lum et al. found that, on average, less than 25% of agencies sampled had processes in place for building any form of relationship with local community members. When the questions related to community partnership building are parceled out to be counterterrorism specific, the percentage drops to less than 10% of agencies having community relationships for counterterrorism. Such a disconnect presents a challenge for enhancing the information available to local law enforcement agencies and therefore fusion centers. In addition, the local nature of terrorism clearly highlights that SLT law enforcement agencies must have access to timely and actionable intelligence for the prevention and response to terrorist acts. For example, Kevin Strom and colleagues from the Institute of Homeland Security Solutions identified and analyzed 68 foiled plots that have occurred since 1999, and found over half of them were initiated by local law enforcement or citizens (Strom et al., 2010). Critical infrastructures and highvalue targets are dispersed widely in the United States, and many of these potential targets are located in rural and less-populated areas. Local law enforcement agencies in these communities are in the best position to recognize when suspicious situations occur near these critical targets (Boba, 2009). Finally, survey research indicates that the terrorism experiences and expectations of the general public regarding intelligence responsibilities of state and local agencies increased after September 11 (Davis et al., 2004). One important issue is how SLT agencies can build and expand on their intelligence capabilities. Not only has the federal government mandated that agencies do this, but there has also been an effort to provide training and technical assistance so that the efforts by agencies are more likely to be effective in response to concerns

1

The three studies included in Lum et al. (2009) were Davis et al. (2004) When Terrorism Hits Home: How Prepared are State and Local Law Enforcement? by the RAND corporation; The Impact of Terrorism on State Law Enforcement: Adjusting to New Roles and Changing Conditions (2006) by the Council of State Governments and Eastern Kentucky University (CSG/EKU) funded by the National Institute of Justice; and the Law Enforcement Management and Administrative Statistics (LEMAS) (2003) Sample Survey of Law Enforcement Agencies by the Bureau of Justice (U.S. Department of Justice, 2006).

4

Evidence-Based Intelligence Practices…

69

about privacy and civil rights. A fundamental obstacle for the sharing and processing of intelligence is the fragmented structure of law enforcement in the United States. The sheer number of public agencies with law enforcement responsibilities is significant as this function is delivered at all levels of government. There is an appreciation of the need to coordinate and share information within and across agencies, and there are many specific initiatives, such as multiagency task forces, that are designed to overcome the “stovepiping of information.” “The intent was to integrate – that is, fuse – information from diverse sources to better understand and prevent multijurisdictional crime problems. Hence, the foundation was laid for intelligence fusion centers.” The idea for analyzing information from various resources at some centralized location was not developed solely as a crisis response to the September 11 attacks. In the 1980s, cities, states, and the federal government created regional intelligence centers in response to specific crimes, such as drug trafficking or gun crime, in a specific geographic region (Carter & Carter, 2009). For example, in response to concerns about drug trafficking, the federal government institutionalized High Intensity Drug Trafficking Area (HIDTA) intelligence centers. Similarly, in the 1990s, the Bureau of Alcohol, Tobacco, and Firearms developed a number of programs related to reducing gun violence, including enhancing their intelligence capacity through the creation of Crime Gun Centers (Carter & Carter, 2009). According to Carter and Carter (2009, pp. 1324), “the HIDTA and the bureau intelligence centers had a great deal of interaction with SLT law enforcement agencies.” The September 11 attacks caused the government to respond in many different ways, but one major concern was significant gaps in intelligence and information sharing. Although the 9/11 attacks did not initiate the call for better information sharing among government agencies, the attacks did enhance the urgency to take action. There are many examples that can be taken from congressional hearings, reports, and legislative initiatives that demonstrate the widespread conclusion that the sharing of information must be improved. For example, the USA PATRIOT Act specifically cites the need to improve information sharing and that the “wall” between the intelligence and law enforcement communities must be torn down (U.S. Congress, 2001). The 9/11 Commission report (2004) highlights multiple information sharing failures and missed opportunities to prevent the attacks, and importantly concludes: “The culture of agencies feeling they own the information they gathered at taxpayer expense must be replaced by a culture in which the agencies instead feel they have a duty to the information – to repay the taxpayers’ investment by making that information available” (National Commission on Terrorist Attacks Upon the United States, 2004, pp. 417). Other examples of government commissions and reports that highlight the problems with information flow and the need to improve information sharing include the Senate Select Committee on Intelligence, The National Strategy for Homeland Security (2007) and The National Security for the Physical Protection of Critical Infrastructures and Key Assets (The White House, 2003), the General Accounting Office’s, 2003 and 2007a reports on strengthening information sharing, and The Weapons of Mass Destruction Commission report (2005). In fact, the Commission

70

J. Carter and S. Chermak

on the Intelligence Capabilities Regarding Weapons of Mass Destruction concluded that the intelligence community has instigated over 100 initiatives to respond to concerns about information sharing (2005, pp. 430). For example, the creation of the Global Intelligence Working Group and the National Criminal Intelligence Sharing Plan (NCISP) has been critically important for more effective intelligence sharing among federal, SLT law enforcement agencies (GIWG, 2003). In addition, the number of Joint Terrorism Task Forces has increased dramatically since 9/11, and the Intelligence Reform and Terrorism Prevent Act of 2004 mandates that the President establish an Information Sharing Environment (ISE) Implementation Plan. The implementation plan for this ISE, which was released in November 2006, states “This environment will create a powerful national capability to share, search, and analyze terrorism information across jurisdictional boundaries and provide a distributed, secure, and trusted environment for transforming data into actionable information. The resulting environment will also recognize and leverage the vital roles played by State and major urban area information fusion centers, which represent crucial investments toward improving the nation’s counterterrorism capacity” (Program Manager for the Information Sharing Environment, 2006, pp. xiv).

The Role of Fusion Centers While there are multiple definitions of fusion centers, they commonly agree that fusion centers can be explained as a collaborative effort of multiple law enforcement and non-law enforcement entities that combine resources and information with the intent to “fuse” disparate pieces of information in an attempt to prevent or mitigate threats (Carter, 2009; Carter & Carter, 2009; Randol, 2009). In response to concerns about information and intelligence gaps, the hope was that the establishment of such centers would support the prevention of terrorism by improving information sharing between government levels, across government agencies, and with private citizens. The rationale behind fusion centers is sound: critical information is captured daily by local law enforcement officials and other sources, such as private citizens, that may be relevant to understanding and prioritizing extant threats. The amount of information, the structure of law enforcement in the United States, and the fragmentation of sources make it impossible to have a single national warehouse that could process intelligence information directly. Instead, it is important that intelligence centers are organized at least at the state level so that the gathering, processing, and sharing of information become more manageable. These independent centers would be connected to various federal agencies so that the boundaries between agencies are crossed and interagency sharing and cooperation is enhanced. Empirical research on fusion centers is virtually nonexistent, and our survey, discussed shortly, revealed three important issues which remain open to examination. The first is how information sharing occurs and the efforts of fusion centers to fuse intelligence from various sources. The second is the capacity of these fusion

4

Evidence-Based Intelligence Practices…

71

centers to conduct data analysis relevant to understanding terrorist threats and evaluate the effectiveness of responses to terrorism. Finally, given the lack of empirical evaluation of the impact of intelligence gathering on counterterrorism, are fusion centers systematically evaluating the impact of their efforts, and what are the opportunities for assessment? We now discuss these three issues below, and then show the results of our survey of fusion centers regarding these three concerns.

Information Sharing One of the key elements to the successful use of intelligence for terrorism prevention is widespread information sharing. According to the ISE, “Strengthening our nation’s ability to share terrorism information constitutes a cornerstone of our national strategy to protect the American people and our institutions and to defeat terrorists and their support networks at home and abroad” (McNamara, 2006, pp. xiii). Similarly, President Bush’s National Strategy for Information Sharing (2007, pp. 1) states, “Our success in preventing future terrorist attacks depends upon our ability to gather, analyze, and share information and intelligence regarding those who want to attack us, the tactics they use, and the targets that they intend to attack.” As a consequence, organizations and agencies must know how to identify relevant information to be analyzed, collect it without violating civil liberties, know who the information should be shared with, and must be willing to share it. Although there are many agencies that will provide information that will have to be shared to be successful in preventing terrorist attacks, it was demonstrated earlier that SLT law enforcement agencies play a particularly important role. There is reason to suspect that, despite substantial effort, information sharing in the area of intelligence is significantly limited (Carter, 2011; Rojek, Kaminski, Smith, & Cooney, 2010). Although there has not been much empirical research that attempts to examine issues related to information sharing among law enforcement agencies, two studies provide a general understanding of relatively recent concerns. First, the General Accounting Office (2003) reviewed critical documents related to law enforcement information sharing, interviewed officials from various agencies, and surveyed 29 federal law enforcement agencies, all 50 home security offices, all cities with a population of 100,000 or greater (N = 485), and a random sample of smaller cities (N = 242). The surveys were sent to the mayor who either completed the survey or delegated the completion to the chief of police, an assistant, or other emergency management personnel. Of the many important findings highlighted in this report, several concern information sharing limitations. Among these findings were that: (1). Officials from federal, state, and local governments do not think the process of sharing information is “effective” or “very effective”; (2). They do not routinely receive the information they need to protect the homeland; (3). The information received is not timely; (4). Opportunities are routinely missed to obtain and provide information to the federal government; and (5). Law enforcement agencies are not receiving the types of information they need to effectively prevent terrorist attacks.

72

J. Carter and S. Chermak

The RAND Corporation conducted two national surveys related to domestic preparedness and intelligence (Riley & Hoffman, 1995; Riley, Gregory, Treverton, Wilson, & Davis, 2005). The 1995 survey focused on preparedness issues for state and local law enforcement. The important conclusion of the study was that there was very little intelligence and strategic assessment capability and poor information sharing between federal and state law enforcement officials – a conclusion similarly reached by the 9/11 Commission Report. Prior to the establishment of the Department of Homeland Security in 2002, RAND did a second survey and several case studies to examine issues related to local and state intelligence efforts (Davis et al., 2004). The study concludes that SLT law enforcement agencies have played an increasingly important role in responding to and preventing terrorism. Law enforcement agencies wanted better intelligence sharing, needed improvements in communication interoperability, and thought that training improvements were necessary. In addition, even small agencies, if assessing their threat risk as high, were very proactive in focusing their preparedness efforts (Riley et al., 2005). Although it is generally understood that intelligence must be shared widely, and it might seem on its face that fusion centers provide an ideal source of data for important evidence-based analysis, there has been very little empirical research that identifies key obstacles to information sharing. The studies discussed above provide valuable background information and highlights some of the key obstacles in effectively using state and local intelligence in the war of terrorism. However, the GAO study does not specifically focus on law enforcement efforts and the RAND study was conducted in 2002 prior to the establishment of the Department of Homeland Security. The field of intelligence has changed incredibly since 2002, and it is important to examine current issues specific to law enforcement efforts in the area of intelligence. In addition, neither the study focuses on the efforts to improve intelligence flow nor there has been a systematic attempt to examine how fusion centers strategically fuse intelligence and what promising strategies exist to enhance information sharing.

What Is Good Intelligence? As it has been discussed previously in this chapter, there have been monumental changes within SLT law enforcement agencies regarding intelligence, building an intelligence capacity, and demands for information sharing. The intelligence environment is changing at a rapid pace, and thus it has been difficult to assess the impacts of these changes: • How can we measure intelligence reliably to know if it has prevented terrorist acts? • Do we know if SLT law enforcement agencies have actually developed an operational “intelligence capacity?” • How do we assess the quality of intelligence?

4

Evidence-Based Intelligence Practices…

73

• How can the quality be improved? • How do we know how well SLT agencies coordinate with federal intelligence access points and state level access points? • Do SLT law enforcement agencies understand the types of information that should be collected and shared? There is clearly a need for the development of standards related to intelligence gathering, processing, and information sharing. A key maxim of organizational behavior is that what gets measured gets done (Osborne & Gaebler, 1992). The absence of performance measures and metrics for law enforcement intelligence makes it impossible for policymakers to assess progress toward enhancing the ISE, and leaves fusion centers and individual agencies vulnerable to intelligence gaps. In general, the accurate measurement of performance related to intelligence has several potential benefits (see General Accounting Office, 2006; Johnston, 2005). First, these measures are valuable for the improvement of programs and strategies. Second, change is most likely to occur in iterative steps, and thus accurate performance measures can help an organization monitor improvement steps, highlight problem areas, and suggest approaches for accomplishing goals. Third, such measures would be valuable in making promotion decisions and allocating resources within the organization. Fourth, such measures can hold organizations and individuals accountable for accomplishing goals. Finally, according to a GAO report (2006, pp. 11), “in a risk management process, agencies can use performance measurement to assess progress towards meeting homeland security goals. The intended effect of assessing such progress, when coupled with other aspects of the risk management process, is the reduction of risk.” There is very little evidence that intelligence practices and the products of these practices are being measured in any meaningful way. Intelligence leaders and analysts, however, have provided anecdotal support for the conclusion that there are significant limits to both the amount and quality of information shared and have voiced frustrations about the inability to accurately assess performance. The problem may not be a lack of data and information, but just the opposite: state fusion centers and analysts have been overwhelmed with data but are only receiving little actionable information. This problem is significant because such information still has to be processed, thus leaving little time to focus on producing helpful analytic products and distributing reports. In many instances, it also appears that intelligence reports disseminated by fusion centers may simply be a “repackaging” of intelligence products, not new information. According to Treverton, Jones, Boraz, and Lipscy (2006, pp. 14), “the United States has been obsessed with data, and that has come at the expense of judgment. Rather than maintaining the ideal of speaking truth to power, intelligence has focused on gathering information. In many ways, this is a function of wealth – a big budget can buy lots of gadgets. The problem is that with all these so-called added capabilities, technologists assert we can collect everything.” It is important to note that there is not widespread agreement about whether performance measurement in the intelligence arena is even possible or desirable

74

J. Carter and S. Chermak

(Treverton et al., 2006). This would appear to be an important empirical question that is worth pursuing in future research. It would be difficult for this research to proceed, however, without a foundational understanding of how law enforcement agencies currently collect information, what types of data are processed, and what capabilities fusion centers have for analysis.

The Need for Assessment A critical component of evidence-based practices is incremental change in reaction to scientific study. A classic way to think about this feedback loop is through the consideration of problem-solving in policing – problems are identified, understood, responded to, and then evaluated. The evaluation should provide directions for additional or different types of responses to enhance the impact. Within the policing literature, there is little knowledge about how the utility of information sharing and intelligence contribute to a reduction in crime and/or terrorism. The point to emphasize from the review of the present state of counterterrorism information sharing research is that information sharing and fusion centers are perhaps the most salient areas for research in the context of counterterrorism in the United States, but there is little understanding of, and a scarce literature on, information sharing in SLT law enforcement agencies. Furthermore, there is not only little knowledge about intelligence sharing and what constitutes good intelligence, as well as the nature of intelligence-sharing entities such as fusion centers, but also little knowledge about whether such efforts can work to reduce risk or increase the capacity to share information. What is needed is an assessment of the current state of law enforcement fusion centers information sharing and data analysis with respect to counterterrorism efforts. The review presented in this chapter is not one of complex scientific rigor, but an exploration of current practices which begins to establish an empirical foundation on which more rigorous methodologies can be based to provide more acute insights on what sources of information and types of data analysis are most effective in the prevention of terrorism. The current scholarly knowledge base of law enforcement information sharing for counterterrorism lacks this foundation. Lum, Kennedy, and Sherley (2006) employed a Campbell systematic review and meta-analysis of US law enforcement counterterrorism efforts to determine the extent to which counterterrorism assessments utilized any form of rigorous methodology. Of more than 20,000 published works that mentioned terrorism, only 112 articles from peer-reviewed sources identified law enforcement counterterrorism efforts. Of these 112 articles, none had a methodological rigor that met the standard of the Campbell systematic review. The need for further examination is solidified further as the policing and counterterrorism literature currently lacks empirical examination of intelligence/information sharing practices for the purposes of counterterrorism (Carter, 2011). A further and interrelated issue to the research gaps discussed above concerns our exploration of how SLT law enforcement agencies communicate information about intelligence issues and how fusion centers might connect or partner with other sources that would enhance the effectiveness of their

4

Evidence-Based Intelligence Practices…

75

efforts. The assessment – or exploration – presented in this chapter seeks to shed light on the current information sources, data analysis, and potential partnerships of state of law enforcement fusion centers.

A Survey of Fusion Centers The data presented in this chapter was gleaned from the “Understanding the Intelligence Practices of State, Local, and Tribal Law Enforcement Agencies” grant funded by the National Institute of Justice.2 In that project we sought to explore fusion centers throughout the United States and their intelligence experience especially as it pertains to adopting intelligence-led policing, information sharing and strategies that could promote better information sharing, how fusion centers intelligence practices are assessed and what metrics are being used to measure performance, and what formal and informal communication networks exist for fusion centers to share information. Toward this end, we used a self-administered questionnaire to be completed through a web-designed survey provider. The survey method targeted key informants through a purposive sample of all persons who had attended the 2007 and 2008 National Fusion Center Conferences (NFCC). The NFCC is sponsored by the leading law enforcement intelligence organizations3 and is considered to be the prominent gathering of key personnel from every fusion center in the United States. Attendees of the NFCC included fusion center directors, operational personnel, and intelligence analysts. This compilation of personnel that attended the NFCCs provides the most comprehensive population of persons with specific knowledge regarding the information sharing practices of state and regional fusion centers in the United States. Given the infancy and ever-evolving nature of fusion centers in the United States, such specialized knowledge of key concepts is critical to the issues discussed in this chapter. This sample includes 96 responses from regional and state fusion center personnel. On average, respondents had been employed by their fusion center for two and a half years at the time they completed the survey. The section to follow will provide a variety of information with respect to fusion center data sources and analytic activities. While all the information provided is not specifically discussed, it is provided with the intent to give context to the fusion center environment. The section to follow will provide a variety of information with respect to fusion center data sources and analytic activities. While all the information provided

2

Grant number 2008-IJ-CX-0007 awarded to Michigan State University, School of Criminal Justice in 2009. Principal Investigators include Dr. David Carter, Dr. Edmund McGarrell, and Dr. Steve Chermak – all from Michigan State University, School of Criminal Justice. 3 The National Fusion Center Conference is sponsored by the following agencies: Bureau of Justice Assistance, Office of Justice Programs, U.S. Department of Justice (DOJ), U.S. Department of Homeland Security, DOJ’s Global Justice Information Sharing Initiative, Federal Bureau of Investigation, Office of the Director of National Intelligence, Office of the Program Manager, Information Sharing Environment, Bureau of Alcohol, Tobacco, Firearms and Explosives, Office of Community Oriented Policing Services.

76

J. Carter and S. Chermak

is not specifically discussed, it is provided with the intent to give context to the fusion center environment. In particular, themes related to the protection of civil rights, the current state of information sharing and analysis, fusion center data analysis, and the potential for fusion center partnerships with researchers and academics are explored. These exploratory findings represent fusion center personnel surveyed and their perceptions of the operations and practices of their fusion center.

Protection of Civil Rights First and foremost, counterterrorism efforts – including intelligence practices within fusion centers – must observe citizens’ civil rights and civil liberties (Carter, 2009). The sensitivity to this issue has become so magnified that the Department of Homeland Security created the “Office for Civil Rights and Civil Liberties4” to safeguard all citizens who may be affected by the activities of the Department of Homeland Security – which is the primary funding source for the majority of fusion centers. In short, law enforcement organizations and personnel must establish a criminal predicate5 (or reasonable suspicion) prior to collecting and retaining information on individuals within a permanent intelligence management system. Safeguards for protecting individuals’ civil rights include having legal counsel review fusion center policies adhering to federal privacy regulations.6 Of the fusion centers surveyed, 75% indicated having policies in place consistent with federal guidelines while 59% of these centers indicated that legal counsel had reviewed their guidelines.

Fusion Centers’ Current State of Information Sharing Information sharing practices among regional and state fusion centers are centered on two key factors. First, there must be a willingness to work collaboratively with other law enforcement agencies to engage in information sharing. Second, there must be a willingness to cooperate; essentially, are the products being created and shared those that are most effective in preventing threats and crimes? Interrelated to this second characteristic is the necessity to get effective products to persons who need to know, when they need to know it. 4

For more information on this new department, visit: http://www.dhs.gov/xabout/structure/crcl. shtm. 5 “Reasonable Suspicion” or “Criminal Predicate” is established when information exists that establishes sufficient facts to give a trained law enforcement or criminal investigative agency officer, investigator, or employee a basis to believe there is a reasonable possibility that an individual or organization is involved in a definable criminal activity or enterprise (U.S. Department of Justice, 2006). 6 Federal regulations guiding law enforcement intelligence pertain to 28 Code of Federal Regulation Part 23.

4

Evidence-Based Intelligence Practices…

77

Fig. 4.1 Percentage of fusion centers with a good working relationship with other law enforcement organizations (N = 96)

Law enforcement’s willingness to engage in information sharing with regional and state fusion centers can be determined through the perceived working relationships between law enforcement agencies; the likelihood of fusion centers to reach out to other law enforcement organizations; and the frequency by which fusion centers disseminate actionable intelligence products to other organizations. These perceived working relationships are illustrated in Fig. 4.1, while the likelihood of fusion centers reaching out to other law enforcement organizations is illustrated in Fig. 4.2 and frequency of product dissemination in Fig. 4.3. Overall, fusion center personnel in this survey indicated having positive relationships with other organizations with the exception of tribal law enforcement. This lack of a working relationship with tribal agencies can perhaps be explained by the minimal number of fusion centers that have a tribal presence in their region as well as a lack of connectivity resources for many tribal departments. Moreover, the results from this study illustrated in Fig. 4.2 indicate that fusion center personnel, on average, are likely to consult other organizations. Lastly, Fig. 4.3 illustrates that, on average, fusion center personnel frequently disseminate their intelligence products to other organizations. While the majority of information sharing appears to occur between law enforcement organizations, and less between law enforcement and nonlaw enforcement

78

J. Carter and S. Chermak

Fig. 4.2 Percentage of fusion centers that are likely to consult representatives from other organizations (N = 96)

(i.e., public health, private sector), this is likely due to fusion centers being law enforcement oriented and rely most heavily on other law enforcement organizations. Given these three indicators, it can be assumed that there is at least a willingness to share information exists among fusion centers and other organizations. Given this finding, the next section explores the type of analytic products created by fusion centers.

Fusion Center Data Analysis Fusion centers are designed to serve as a lynchpin among federal and SLT law enforcement to improve a fragmented information sharing structure. While fusion centers have closed the information sharing gap, significant shortcomings still exist. The majority of local law enforcement agencies remain unaware of the fusion center mission and the resources fusion centers can provide. Local law enforcement expressed a strong willingness to share information with fusion centers; however, they simply lack the connectivity and knowledge of how to do so. This unawareness among local law enforcement creates an inhibitor to successful information sharing, primarily based on two associated factors. First, local law enforcement serves as the primary information collectors as fusion centers rely on local law enforcement for the input of raw street-level information to be incorporated into the analytic process. Second, local law enforcement do not currently utilize analytic intelligence products provided by fusion centers to develop collection requirements to guide the collection of the most relevant raw information.

4

Evidence-Based Intelligence Practices…

79

Fig. 4.3 Percentage of fusion centers that perceive to frequently provide actionable intelligence products to the organizations (N = 96)

Developing collection requirements to guide the collection of raw information is critical. Rather than not having enough information, fusion centers are hampered by information overload. Collection requirements are developed based on an assessment of threats (e.g., crime, terrorism, or extremism) within geographic jurisdictions. For example, if a law enforcement jurisdiction identifies a right-wing extremist threat within the community, collection requirements can be developed to focus on raw information pertaining to this specific threat rather than collecting random raw information to be included in the analytic process. State and local law enforcements are not alone with respect to this shortcoming of lacking collection requirements for identifying “good information.” Fusion centers currently lack this component as well. In general, 98% of fusion centers across the country indicate being aware of, and prepared for, threats in their region. Despite almost all of the fusion centers that responded to the survey indicating being aware of threats, only 48% indicate having developed collection requirements based upon these risk assessments.

80

J. Carter and S. Chermak

Fig. 4.4 Sources of information for fusion center analysts (N = 96)

Although the sharing of information between state and local law enforcement and fusion centers is improving, fusion centers must rely on other sources of information to enhance analytic products. Figure 4.4 displays the percentage of fusion centers that currently have access to a variety of information sources. Access to these information sources provides raw data input for creating, or enhancing, intelligence products. To further enhance such products, comprehensive analytic processes are desired to produce the type of products most utilizable by fusion centers. Figure 4.5 illustrates the frequency of information sources utilized by fusion centers across the United States. The difference between information and intelligence is an analytic process. Fusion centers are tasked with the responsibility of collecting/receiving raw information to be included in an analytic process. To further enhance these efforts, fusion centers can receive information that is not raw, but that has gone through some form of analytic – or critical thinking – process. The two most frequent sources of information commonly received are news reports and open source information. These sources require little, if any, analysis or critical thinking before they are disseminated. Fusion center reports, crime reports, and suspicious activity reports comprise

4

Evidence-Based Intelligence Practices…

81

Fig. 4.5 Percentage of fusion centers frequently receiving information products from outside agencies

the second most frequent types of information. Once again, these information sources lack the application of a comprehensive analytic component.7 Fusion center reports are typically generated through critical thinking and some baseline analytic procedures (e.g., social networking analysis), crime reports mirror crime mapping and CompStat reports (e.g., number of and types of incidents in geographic areas), and suspicious activity reports are documented suspicious behavior (e.g., tip from the field). The most comprehensive source of information, from an analytic perspective, is threat assessments, which are received less than annually in most instances. While these sources are all valuable – and critical – to the analytic process of fusion centers, they lack a higher level of predictive analysis that could increase the quality and effectiveness of intelligence products. Similar to these survey findings, a recent study by Rojek et al. (2010) examining the South Carolina Intelligence and Information Center (SCIIC) also indicated a lack of high-level predictive analysis. They found that 74% of the products created by the SCIIC were routine intelligence alerts and warnings as compared to only 31% that were intelligence threat assessments. Their study also found that even though threat assessments were not conducted very often, they ranked as one of the highest forms of analytic products to be reviewed by members of the South Carolina law enforcement community. 7

Crime reports and fusion center reports do require a level of analysis; however, the analytic process applied to these reports is consistent with identifying trends based on incidents that have already occurred.

82

J. Carter and S. Chermak

Fig. 4.6 Percentage of fusion centers that frequently use different analytic methods (N = 96)

To provide context for the type of analytic work currently being conducted by fusion centers, Fig. 4.6 presents the varying types of analyses fusion centers incorporate. While this figure presents a great deal of information, it should be noted that the focus, in general, of fusion centers is on the analysis of suspicious activity reports and the creation of alerts and notifications as well as link analysis charts. These types of analytic procedures are necessary to law enforcement’s mission to identify criminal threats and clear existing criminal cases. However, there is a lack of predictive analytic methods currently utilized by fusion centers. The method that is most predictive in nature is geographic profiling8 – and as Fig. 4.6 illustrates, less than half of the fusion centers surveyed indicated doing this type of analysis. Graphia-Joyal (2010) found similar results with respect to the types of analyses currently carried out within a select group of fusion centers, noting that “that a significant portion of their activities continue to revolve around tactical and operational

8

Geographic profiling is a technique designed to identify areas where habitual offenders or commonly occurring crimes are likely to be located. This technique, also referred to as geospatial analysis, requires the manipulation of large amounts of crime, offender, and geographic data.

4

Evidence-Based Intelligence Practices…

83

activities, such as background checks, digital line ups, and driver’s license look ups, as well as investigative case-support, and preparing products for prosecutorial use” (Graphia-Joyal, 2010, pp. 73). An emerging initiative among law enforcement is the concept of predictive policing which posits the use of high-level predictive statistical analysis to forecast criminality (Osborne, 2006). Predictive policing, much like intelligence-led policing, encompasses a variety of analytical processes and information sources which serve as different pieces of the larger threat picture. These analytical processes can include hot spot analysis, crime mapping, and even social network analysis. While these analytic processes are not in themselves the most rigorous, they can provide critical pieces of information to be included in the predictive analytical processes on which the predictive policing philosophy relies. A recent report from the National Institute of Justice on enhancing predictive policing programs focused on law enforcement agencies engaged in strategies and tactics that improve the situational awareness of law enforcement concerning individuals or locations before criminal activity occurs (Uchida, 2010). A separate solicitation from the National Institute of Justice also emphasized the role of fusion centers that were using predictive tools to provide leads or guide specific investigations which include the use of models of criminal activity or innovative methods to integrate a variety of information sources (U.S. Department of Justice, 2006).

Potential for Fusion Center and Academic Partnerships Fusion center executives have emphasized the desire to have more comprehensive analytics, but acknowledged such analytic methods were beyond the scope of most analysts’ skills and knowledge base (Carter, 2011). This lack of a comprehensive analytic capability is due to a lack of funding to employ a statistically trained researcher (Carter, 2011). In a recent study of four fusion centers, Graphia-Joyal (2010) found that fusion center executives indicated the center’s analytical capabilities have yet to be fully developed internally or utilized externally. More specifically, she found that “a robust and reliable analytical component with estimative or predictive capabilities has yet to be fostered and institutionalized” (Graphia-Joyal, 2010: 64). Among the fusion centers included in the study, she explained this lack of robust analytic methods. Fusion center executives interviewed for the study said that the analysts’ collective skill sets have yet to mature at the state and local levels and that “developing a robust analytical capability is partially influenced by the resources a fusion center is able to secure, namely a sufficient number of analysts with the experience and skill set necessary to engage in more sophisticated analyses” (Graphia-Joyal, 2010: 64). This lack of personnel training in the area of sophisticated analytic methods is not necessarily a shortcoming of the intelligence analyst profession – intelligence analysts simply are not required to employ such methods on a regular basis and,

84

J. Carter and S. Chermak

therefore, training on these methods is not a priority. A potential solution for fusion centers to fill this predictive analytic gap is by engaging in active partnerships with academic researchers who have been trained to utilize such methods. However, there is currently a lack of active engagement between law enforcement and academic scholars. Lum and Fachner (Lum et al., 2009) found police partnerships with academic researchers to be rather infrequent (11%). This finding reflects a belief among law enforcement that a gap exists between the outputs of research and the needs of practitioners (Lum et al.: 123). The authors note that such a partnership would aid law enforcement in more scientifically rigorous approaches to analytic products. The application of predictive modeling is commonplace in the academic arena and may provide common ground for fostering relationships between researchers and the intelligence community. Contingent upon a variety of security and clearance factors, collaboration with academic personnel provides a potential to fill two significant gaps within fusion centers. First, academics can provide a more comprehensive level of predictive analysis once they are vetted and given access to fusion center data. The caveat is that this is certainly not as straightforward as it is explained here. A multitude of policies and procedures would have to be established prior to such a partnership. Potential obstacles include agreements of memorandums of understanding pertaining to the disclosure of information, limitations to predictive analyses, and protections of civil liberties – a discussion of which is beyond the undertaking of the current chapter. Second, partnerships with academics provide opportunities for evaluating the effectiveness of fusion centers. The importance of active partnerships between law enforcement and academics is tangible as these partnerships provide the means by which operational practices are explored compared to established constructs (Boba, 2010). Within the years of 2004–2006, the Department of Homeland Security provided almost $131 million to fund regional and state fusion centers (General Accounting Office, 2007b). In a time of financial strain, such investments by the federal government have drawn scrutiny and criticism (Russell & Taylor, 2011) for a lack of return on investment. Such assessments are unfounded as there has yet to be an empirical evaluation of the effectiveness of fusion center capabilities.9 It is believed the investment in fusion centers – not only by the federal government but also by the state and municipal governments – has yielded a positive return on investment. Collaboration among academics and fusion centers would allow for the evaluation of practices and initiatives to begin to empirically justifying such investments. Moreover, beyond financial justification, evaluations of fusion center practices can identify and translate “best practices” as centers across the country are in different stages of development.

9

Data used for this chapter is from the “Understanding the Intelligence Practices of State, Local, and Tribal Law Enforcement Agencies,” grant funded by the National Institute of Justice which will be the first empirical exploration of regional and state fusion centers at the national level.

4

Evidence-Based Intelligence Practices…

85

Discussion and Conclusion The ability of fusion centers to collect, analyze, and share information with other organizations is dependent on working relationships with external organizations. Previously discussed findings have indicated a general willingness among fusion centers to engage with other organizations. More specifically, the organizations that have more counterterrorism responsibility – FBI and other federal agencies – appear to be engaged with fusion centers. While these agencies may be tasked more toward counterterrorism than local agencies, the role local law enforcement plays in the counterterrorism landscape should not be underestimated. Aside from their role as first responders to terrorism incidents, local law enforcement serve as force multipliers for both enhancing information collection for fusion centers and carrying out counterterrorism initiatives.10 These working relationships are also indicative of fusion centers’ efforts to gather a variety of information from a variety of sources. An “all threats, all crimes, all hazards” approach to safeguarding America’s communities relies on an array of information sources that range from traditional law enforcement to private security and even public health. Information received by fusion centers from external law enforcement agencies is currently in need of improvement. On average, fusion centers indicated a moderate frequency of information products being pushed to them from other agencies. This information is perhaps most pertinent to the counterterrorism mission as the raw information that is a derivative of police–citizen interactions is often the “missing piece” of the puzzle and is critically important to be incorporated into the analytic process. This information can come to the fusion center in the form of a suspicious activity report or open source product. While the percentages indicate “less” as opposed to “more” agencies are receiving these types of information, the numbers themselves are promising given the current status of information sharing. As mentioned previously, key concerns for fusion center relationships with local agencies hinge on awareness among local agencies as to the mission and capabilities of fusion centers. Fusion centers are not dependent on information pushed to them from outside entities; they have access to varied information systems which they can query. At present, these data sources are more along the traditional lines and include motor vehicle records, driver’s license information, and other databases managed by other criminal justice organizations – such as the National Crime Information Center. Indications are positive with respect to sources of information for counterterrorism efforts. Approximately, three quarters of fusion centers indicated have access to the most terrorism-specific information systems. These systems include FBI’s Law Enforcement Online, the Regional Information Sharing System, and the Homeland Security Intelligence Network. Connectivity to these information sharing systems allows for fusion centers to create products that are best suited for developing strategic and operational planning.

10

Such as the Nationwide Suspicious Activity Reporting Initiative – focused on local law enforcement’s collection of documented suspicious behaviors – this initiative is also known as Department of Homeland Security Secretary Janet Napolitano’s “See Something, Say Something” campaign.

86

J. Carter and S. Chermak

As it was illustrated in Fig. 4.6, fusion centers have the ability to produce a wide variety of analytic products useful for the prevention of threats and crimes. Most critical to the current discussion is the ability of fusion centers to identify threats and evaluate the effectiveness of their efforts in preventing these threats. It is worth repeating that 98% of the fusion center personnel surveyed indicated they had identified threats in their region. Awareness of a threat and predicting likelihoods of threats are two different issues. Of the products created by fusion centers, the most powerful “predictive” analytic method employed (that is reported) is geographic profiling. However in the most simplistic definition, geographic profiling identifies physical areas where crimes are likely to occur based on previous crimes with similar geographic characteristics. This analytic process does not yield any form of predictive odds or likelihoods with respect to the threat actually occurring.11 This chapter has begun to explore the current state of information sources and analysis utilized by law enforcement fusion centers in efforts to prevent terrorism. Such an exploration is the necessary foundation for an empirical assessment of these efforts. Empirical evaluations to determine what sources of information, and which analytic processes, are most effective in law enforcement’s counterterrorism efforts are the next step to better understand the current state of fusion center information sharing. Establishing outcomes for counterterrorism intelligence practices would be best achieved through partnerships among fusion center personnel with detailed counterterrorism knowledge and academic researchers possessing program evaluation skills. Beyond program evaluation, academic researchers can also fill the predictive analytic gap faced by most fusion centers. These partnerships would also allow for best practices and lessons learned to be identified to help facilitate the development of fusion center practices across the United States by informing the counterterrorism discipline. Emerging from these partnerships could be advancements in law enforcement counterterrorism information sharing that would be beneficial for both academics and practitioners. Such advancements include, but certainly are not limited to, (1) identifiable metrics for effective information sharing (i.e., timely/organic information flow from the streets to fusion centers and back to the streets), (2) increased analytic capabilities for more acute prevention/forecasting efforts, (3) increased satisfaction among personnel toward the counterterrorism mission (i.e., feedback loops on “good intelligence”), and (4) identifiable return on investment (i.e., impact of effective information sharing on monetary and personnel expenditures).

Conclusion Regional and state fusion centers are facilitating the counterterrorism mission in the United States. Given their infancy, significant strides have been made to increase information sharing effectiveness as a result of fusion centers’ productivity and 11

It should be noted that not even the most advanced and comprehensive statistical data techniques can be used to identify specific crime and threat locations as human behavior is not an exact science.

4

Evidence-Based Intelligence Practices…

87

lynchpin design. However, room for improvement still exists. While their analytic strategies are diverse, they are not yet on par with expectations of fusion centers. As a fusion center executive explained, fusion centers somewhat suffer from the “CSI effect” in that people have unrealistic expectations of their abilities (Carter, 2011). While no analytic strategy will serve as the crystal ball and predict the future, enhancements can be made by partnering with academic researchers trained in statistical data analysis. Such partnerships are likely to prove fruitful for both law enforcement and academia.

References Bayley, D. H., & Weisburd, D. (2009). Cops and spooks: The role of the police in counterterrorism. In D. Weisburd, T. Feucht, I. Hakimi, L. Mock, & S. Perry (Eds.), To protect and to serve: Police and policing in an age of terrorism – and beyond (pp. 81–99). New York: Springer. Boba, R. (2010). A practice-based evidence approach in Florida. Police Practice and Research, 11(2), 122–128. Boba, R. (2009). Evil done. In G. Newman & R. Clarke (Eds.), Reducing terrorism through situational crime prevention. Crime prevention studies series (pp. 71–92). Monsey: Criminal Justice Press. Carter, D. (2009). Law enforcement intelligence: A guide for state, local, and tribal law enforcement agencies (2nd ed.). Washington: US Department of Justice. Carter, D. L., & Carter, J. G. (2009). The intelligence fusion process for state, local and tribal law enforcement. Criminal Justice and Behavior, 36(12), 1323–1339. Carter, J. (2011). Police innovation: Exploring the adoption of intelligence-led policing. Dissertation, School of Criminal Justice, Michigan State University, East Lansing. Clarke, R. V., & Newman, G. (2006). Outsmarting the terrorists. Portsmouth: Greenwood Publishing Group. Davis, L. M., Riley, K. J., Ridgeway, G., Pace, J., Cotton, S. K., Steinberg, P. S., et al. (2004). When terrorism hits home: How prepared are state and local law enforcement? Santa Monica: RAND Corporation. General Accounting Office. (2003). Efforts to improve information sharing need to be strengthened. Washington: General Accounting Office. General Accounting Office. (2006). Homeland Security guidance and standards are needed for measuring the effective of agencies’ facility protection efforts. Washington: General Accounting Office. General Accounting Office. (2007a). Numerous federal networks used to support homeland security need to be better coordinated with key state and local information sharing initiatives. Washington: General Accounting Office. General Accounting Office. (2007b). Federal efforts are helping to alleviate some challenges encountered by state and local information fusion centers. Washington: General Accounting Office. Global Intelligence Working Group. (2003). National criminal intelligence sharing plan. U.S. Departent of Justice. Retrieved February 12, 2011 from http://www.it.ojp.gov/documents/ National_Criminal_Intelligence_Sharing_Plan.pdf. Graphia-Joyal, R. (2010). Are fusion centers achieving their intended purposes? Findings from a qualitative study on the internal efficacy of state fusion centers. International Association of Law Enforcement Intelligence Analysts Journal, 19(1), 54–76. Johnston, R. (2005). Analytic culture in the US intelligence community: An ethnographic study. Washington: Central Intelligence Agency. Lum, C., Haberfield, M., Fachner, G., & Lieberman, C. (2009). Police activities to counter terrorism: What we know and what we need to know. In D. Weisburd, T. Feucht, I. Hakimi, L. Mock, & S. Perry (Eds.), To protect and to serve: Police and policing in an age of terrorism – and beyond (pp. 101–142). New York: Springer.

88

J. Carter and S. Chermak

Lum, C., Kennedy, L. W., & Sherley, A. (2006). Are counter-terrorism strategies effective? The results of the Campbell systematic review on counter-terrorism evaluation research. Journal of Experimental Criminology, 2(4), 489–516. Masse, T., O’Neil, S., & Rollins, J. (2007). Fusion centers: Issues and options for Congress. Washington: Congressional Research Service. McNamara, T. F. (2006). Information sharing environment implementation plan. Washington: Office of the Director of National Intelligence. National Commission on Terrorist Attacks Upon the United States. (2004). The 9/11 Commission Report. Retrieved February 23, 2011 from: http://www.9-11commission.gov/. Newman, G., & Clarke, R. (2008). Policing terrorism: An executives guide. Washington: Office of Community Oriented Policing Services. Osborne, D. (2006). Out of bounds: Innovation and change in law enforcement intelligence analysis. Thesis, Joint Military Intelligence College, Washington. Osborne, D., & Gaebler, T. (1992). Reinventing government. Reading: Addison Wesley. President’s National Strategy on Information Sharing. (2007). Retrieved February 25, 2011 from: http://www.whitehouse.gov/nsc/infosharing/index.html. President’s National Strategy on Homeland Security. (2007). Retrieved June 2, 2011 from http:// www.dhs.gov/xlibrary/assets/nat_strat_homelandsecurity_2007.pdf. Program Manager for the Information Sharing Environment. (2006). Information sharing environment implementation plan. Washington: Office of the Director of National Intelligence. Randol, M. A. (2009). Terrorism information sharing and the nationwide suspicious activity report initiative: Background and issues for Congress. Washington: Congressional Research Service. Riley, K. J., & Hoffman, B. (1995). Domestic terrorism: A national assessment of state and local law enforcement preparedness. Santa Monica: RAND Corporation. Riley, K., Gregory, J., Treverton, F., Wilson, J. M., & Davis, L. M. (2005). State and local intelligence in the war on terrorism. Santa Monica: RAND Corporation. Rojek, J. R., Kaminski, J., Smith, H., & Cooney, M. (2010). 2010 South Carolina law enforcement census: Local law enforcement use and evaluation of the South Carolina Intelligence and Information Center. Department of Criminal Justice and Criminology. Columbia: University of South Carolina. Russell, A., & Taylor, R. (2011). The failure of police “fusion centers” and the concept of a national intelligence sharing plan. Presentation. Panel: Emerging trends in intelligence-led policing and homeland security. Academy of Criminal Justice Sciences. Toronto, Ontario. Saari, S. C. (2010). Fusion centers: Securing America’s heartland from threats. Master’s Thesis, Naval Post Graduate School, Monterey. Strom, K., Hollywood, J., Pope, M., Weintraub, G., Daye, C., & Gemeinhardt, D. (2010). Building on clues: Examining successes and failures in detecting U.S. terrorist plots, 1999–2009. Institute for Homeland Security Solutions. Retrieved March 2, 2011 from https://www.ihssnc. org/portals/0/Building_on_Clues_Strom.pdf. The Commission on the Intelligence Capabilities of the United States Regarding Weapons of Mass Destruction. (2005). Report of The commission on the intelligence capabilities of the United States regarding weapons of mass destruction. Retrieved June 2, 2011 from http://www.fas.org/ irp/offdocs/wmd_report.pdf. The White House. (2003). The National Strategy for the Physical Protection of Critical Infrastructures and Key Assets. Washington. Retrieved March 6, 2011 from: http://www.dhs. gov/xlibrary/assets/Physical_Strategy.pdf. Treverton, G. F., Jones, S. G., Boraz, S., & Lipscy, P. (2006). Toward a theory of intelligence: Workshop report. Santa Monica: RAND Corporation. Uchida, C. (2010). A national discussion on predictive policing: Defining our terms and mapping successful implementation strategies. Washington: National Institute of Justice. U.S. Congress. (2001). Uniting and Strengthening America by Providing Appropriate Tools Required to Intercept and Obstruct Terrorism Act (USA PATRIOT Act). Passed on October 26. U.S. Department of Justice. (2006). Criminal intelligence systems. Washington: Office of Justice Programs. Retrieved March 4, 2011 from: http://www.ojp.usdoj.gov/BJA/txt/chap13.txt.

Part III

Methodological Innovations for Counterterrorism Policy

Chapter 5

Innovative Methods for Terrorism and Counterterrorism Data Michael D. Porter, Gentry White, and Lorraine Mazerolle

Introduction Better understanding of the dynamics of terrorism allows for a more complete picture of the complexities involved in measuring success or failure and can assist the 110th Congress as it coordinates, funds, and oversees anti-terrorism policy and programs. Perl (2007)

Terrorism is a complex dynamic process involving multiple factors. Each attack is the result of a terrorist’s decision-making process regarding the number, timing, and characteristics of the ensuing event. As the terrorists’ objectives, capabilities, and preferences change, so do their attack profiles. Attempting to model the dynamics of such a process is a daunting task, especially in the context of how the process responds to counterterrorism interventions. However, scientists in catastrophe modeling have been successful in describing complex events in terms of temporal and spatial distribution, severity, and frequency or risk; the insurance industry can empirically evaluate the probable impacts of many different types of disasters on a particular target; and researchers in economics and psychology are adept at describing and predicting human behavior through quantitative models. These advances demonstrate that sophisticated conceptual models can be developed to account for the dynamic and complex nature of terrorism, and that quantitative data can be incorporated into these models to give robust empirical information on the nature of terrorism and the success or failure of counterterrorism interventions.

M.D. Porter (*) GeoEye Analytics, McLean, VA, USA e-mail: [email protected] G. White • L. Mazerolle ARC-Centre of Excellence in Policing and Security, Institute for Social Science Research, The University of Queensland, Brisbane, Queensland e-mail: [email protected]; [email protected]

C. Lum and L.W. Kennedy (eds.), Evidence-Based Counterterrorism Policy, Springer Series on Evidence-Based Crime Policy 3, DOI 10.1007/978-1-4614-0953-3_5, © Springer Science+Business Media, LLC 2012

91

92

M.D. Porter et al.

Developing a useful model for terrorism requires two things. The first is a reliable and valid measurement for terrorist activity. Measurements can vary from simply counting the number of attacks to more complex measures of severity, economic loss, tourism decline, fear, etc. (see Frey, Luechinger, & Stutzer, 2007, for a good discussion of severity measures). The choice of measure will not only impact what models are used but it will also impact how counterterrorism efforts are assessed. Second, a parsimonious conceptual model of terrorism is required in order to evaluate these measures. The specification of an accurate model including all relevant factors is a prerequisite for useful and robust model results. The particular challenge of this is that it can be difficult to select the relevant factors for inclusion in a conceptual model, and it can be even more difficult to define the form of the relationships between all of the factors mathematically. Most efforts in this direction have relied on models from the time series or survival analysis literature. Examples include Enders and Sandler (1993, 2000, 2002, 2006) and Barros (2003) using intervention analysis and vector autoregressive (VAR) time series models and Dugan, LaFree, and Piquero (2005) and Lafree, Dugan, and Korte (2009) using Cox proportional hazards models to assess the effectiveness of counterterrorism activities. Our objective is to develop a useful mathematical framework for accommodating a wide variety of terrorism measures, use these measures to construct models describing how terrorism varies over time, and develop appropriate statistical tests to assess the effectiveness of counterterrorism interventions. In the section “Measuring Terrorism,” we describe how a marked point process framework can establish a comprehensive measure of terrorism based not only on the frequency of terrorist attacks, but also on their combined impacts. Section “Measuring Counterterrorism” describes how the marked point process framework can be useful in evaluating counterterrorism efforts. We present an example of assessing the influence that the formation of Detachment-88 had on the terrorism process in Indonesia, exposing some dangers of using poor fitting models and pointing out some limitations of intervention analysis. We contrast intervention analysis with change point analysis in the section “Change Point Analysis.” Change point analysis is an approach to detect when a process experienced a significant change. By testing over all possible change points, it can reveal the most likely times when changes occurred. When this matches the intervention time, it provides added confidence that the change in the process was actually due to the intervention.

Measuring Terrorism In order to quantify our understanding of terrorism the first task is to construct a meaningful measure of the loss associated with a terrorist attack. This is a difficult endeavor as a terrorist attack can result in deaths, injuries, damage, economic

5

Innovative Methods for Terrorism and Counterterrorism Data

93

losses, emotional and psychological damage and pain, political consequences, and any number of a myriad of possible impacts. Because it may strike an observer as cold to seek to reduce the impact of a terrorist attack to mathematical computations, the sensitivity and importance of this task is not to be understated. Unfortunately, the tools we have to assess the costs and impact of a terrorist event are limited to those same mathematical computations based on actuarial tables and economic data, computing the cost of human life and injury with the same indifference as they calculate the cost of damage to buildings or businesses. The limitations of these tools should serve as a reminder that the methods developed here do not intend to replace the decision-making skill and responsibilities of people, but merely to serve as an aid for interpreting the data in order to make the best decisions possible. There are, however, various measures of terrorist activity that are easily implemented. The simplest is the raw count of terrorist events. More sophisticated measures consider the severity of events through counts of injuries and fatalities (Clauset, Young, & Gleditsch, 2007). Economic measures such as the impact of terrorism on tourism (Enders & Sandler, 1991), investment and personal savings (Abadie & Gardeazabal, 2003; Fielding, 2003), and stock markets and foreign investments (Chen & Siems, 2004; Collier, 1999), are all available and provide some additional information on the impact of terrorism. Additional measures for concepts such as personal safety and well-being have also been explored (Frey & Luechinger, 2005). Such measurements may be a function of directly observable quantities such as attack type, group attribution, and target. In addition, the outcome of the attack (e.g., infrastructure damage, fatalities, publicity) may also be relevant in constructing appropriate measures. The specific objectives of the analysis will drive what measures are used and how they are combined. An insurance company, for example, will be interested in different measurements than a government agency or political scientist. How to construct a proper measurement for terrorism is beyond the scope of this paper. Rather, we focus on developing a mathematical framework that can incorporate a wide variety of terrorism measures. Our ensuing models can be used to analyze how terrorism varies over time and in response to counterterrorism activities. In general, we assume that any measure of terrorism will be a function of both the number of attacks as well as their combined impact. It is also important to consider both frequency and impact when evaluating counterterrorism efforts. Interventions such as arrests and imprisonment of terrorists may reduce the frequency of attacks, but do little to reduce the impact that future attacks have. However, policy changes targeted at reducing a terrorist group’s finances, communication, or access to material support may not diminish their attack frequency, but will reduce the impact of those attacks. This leads us to propose a general measurement framework for terrorism, based on a marked point process, that takes into account both the frequency of attacks, their individual impacts, and their cumulative effects. This framework consists of two components. The first is a temporal point process to model the frequency

94

M.D. Porter et al.

of attacks, and the second is an impact score to quantify the severity or ensuring consequences of a terrorist attack. Together, these two components provide a quantitative measure of terrorism.

Point Process A temporal point process is a type of model that governs the random occurrence of events at points in time. The distinction should be made clearly that in a point process the time that events occur is random. This is in contrast to a time-series model in which observations are taken at known regular times and only the outcome is considered random. In the marked point process framework presented subsequently, we assume that both the time and the outcome of an event are random variables. This allows us to make inference and perform statistical tests on both the frequency and impact of terrorist events. The benefit of this approach will become clearer in the section “Change Point Analysis” when we discuss assessing the effectiveness of counterterrorism interventions and examine how interventions may affect these two aspects differently. The observed data from a terrorist point process are the event times (or point pattern) t1 £ t2 £ , …, £ tn, where ti is the time of the ith attack and n is the total number of observed attacks.1 Given this sequence of random event times, they can be conceptualized in a variety of ways. For example, we could measure the time between events (called inter-arrival times) {x1, x2, …, xn}, where xi = ti - ti -1 . Small inter-arrival times indicate periods of high attack frequency. This perspective and the resulting inter-arrival times is employed in survival analysis (Cox, 1972). Dugan, LaFree, and Piquero (2005) and Lafree, Dugan, and Korte (2009) used such an approach for analyzing the effectiveness of counterterrorism interventions by modeling the impact of the interventions on the subsequent inter-arrival times. We could also consider the counting process, N (s, t ) = number of attacks between time s and t Intervals with large counts indicate periods of high attack frequency. Time series models for intervention analysis, such as those used by Barros (2003) and Enders and Sandler (1993, 2000), use counts in adjacent, nonoverlapping time intervals. The assumptions on the time series models necessitate that the intervals be sufficiently wide to contain a large number of events (e.g., yearly or quarterly). However, this restriction to large regular time intervals is not necessary in the point process

1

We are assuming that the terrorist process is observed over discrete time units (e.g., days) and hence multiple events could occur at the same time. While this differs from most continuous time point process models that often require no events to occur at the same time, this causes no problem for the models used here.

5

Innovative Methods for Terrorism and Counterterrorism Data

95

counting process N(0, t)

15

10

5

0 0

100

200

300

Time (t)

inter−arrival time (xi)

70 60 50 40 30 20 10 0 2

4

6

8

10

12

14

Event index (i) Fig. 5.1 Two views of a point process. Top: Counting process N(0, t) for a point pattern. The rug indicates the event times. Bottom: The inter-arrival times for the same point pattern

formulation. By considering the event times directly, the point process approach allows analysis at the highest resolution (e.g., daily). Figure 5.1 shows the counting process and inter-arrival times for an example point process. We want to use these point process models to test specific hypotheses, such as the effects of a counterterrorism intervention or the presence of a change point in the terrorism process. To carry out such statistical tests, we need models that capture the inherent randomness in the event data. The sampling distributions of the inter-arrival times and the counts can be defined using a variety of models (Kabelfleisch & Prentice, 2002). For example, a Poisson point process2 has a 2

A Poisson point process has the properties that the number of events in any time interval [s, t] is Poisson random variable with mean E[ N (s, t )] and the number of events in any set of disjoint (nonoverlapping) intervals are independent.

96

M.D. Porter et al.

probability distribution that is fully specified by an intensity function l(t) which is the expected number of events at time t. Thus, the number of events in a given time window [s, t] will have a Poisson distribution with mean t

E[ N (s, t )] = å λ (u), u=s

where the expected number of events in the interval is simply the sum of the intensity function over that interval.

Marked Point Process As previously mentioned, in addition to evaluating the timing and number of terrorist attacks, it is also important to consider the impact or severity of the terrorist attacks. This additional information about the terrorist events can be captured by a marked point process. The marked point process extends a standard point process by including with each event time ti an associated mark, or impact score Yi. The impact score indicates an attack’s magnitude or severity. The data for a marked point process are pairs of event times and impact scores: {(ti, Yi), i = 1, 2, …, n} (see Daley & Vere-Jones, 2003, for a thorough coverage of point processes). For the interval [s, t], the compound score is the sum of the impact scores for the events occurring in that interval Z (s, t ) = cumulative impact of attacks between time s and t =

å

Yi

(5.1)

{i : s £ ti £ t}

where ti is the time and Yi is the impact score for the ith terrorist event. Each terrorist attack has an associated impact score Yi and the compound score (over an interval) is the sum of impact scores for the attacks occurring in the interval. The compound score can be increased in two ways: by having more attacks or attacks of larger magnitude. Thus, the compound score provides a quantitative measure of terrorism by combining the frequency and severity of the attacks into one measure. If all impact scores are one (i.e., Y1 = Y2 =,¼, = Yn = 1 ), then the marked point process reduces to the standard point process and the compound score is equal to the count (i.e., Z(s, t) = N(s, t)). Figure 5.2 illustrates an example of a compound process along with its equivalent marked point process view. Notice that the heights of the jumps in Z(0, t) correspond to the heights of the impact scores. Compare this with the counting process in the top of Fig. 5.1 where all jumps are the same size.

5

Innovative Methods for Terrorism and Counterterrorism Data

97

compound score Z(0,t)

50 40 30 20 10 0 0

100

200

300

Time (t)

impact score Yi

8 6 4 2 0 0

100

200

300

Time (t) Fig. 5.2 Two views of a marked point process. Top: Compound process Z(0, t). The rug indicates the event times. Bottom: The marked point process view showing the impact scores associated with each event time

Impact Score There are multiple perspectives from which to examine the development of the impact score Yi. In the process of constructing this measure, it is important to remember that this construct is just a function to take data concerning the characteristics and outcome of an attack and calculate a single number score to quantify its severity or impact. It could be the result of a mathematical formula, elicited expert opinion, or even a combination of the two. In practice, the methods used to generate the impact scores will differ depending on the objectives and interests of the analysis

98

M.D. Porter et al.

(e.g., economists may limit their interest to specific areas such as tourism, while political leaders may have different measures of interest such as public feelings of safety). Regardless of the impact score employed, we assume that larger values correspond to more severe attacks.3 Furthermore, we assume that each impact score is a random variable and has a density function f(y | ti) denoting the probability that an attack at time ti will result in an impact score of y. Modeling efforts will focus on the proper specification of the impact density function f(y | t). If the impact scores are independent random variables then the expected compound score (at any single point in time) is the intensity multiplied by the expected impact score (if an event were to occur at that time): E[ Z (t )] = λ (t ) ⋅ E[Y (t )], where Z(t) ≡ Z(t, t) is the compound score for the single day t and E[Y (t )] = ò yf ( y | t )dy . This representation makes explicit the relationship that both the attack rate and impact score has on the compound score. Another useful measure is the average impact score Y ( s, t ) =

Z ( s, t ) N ( s, t )

(5.2)

which normalizes the compound score by the number of attacks occurring in the interval [s, t]. This measure estimates the expected impact score for the interval. Writing (5.2) another way, Z (s, t ) = N (s, t ) ⋅ Y (s, t ) shows how the compound score is the product of the count and average impact score.

Further Considerations We have outlined three useful measures of terrorism: the count N(s, t), the compound score Z(s, t), and the average impact score Y (s, t ) = Z (s, t ) / N (s, t ) . Each of these summarize certain aspects about the nature of the terrorist activity in a time interval. However, they are based on the assumption that the impact of the events are instantaneous and persistent. This is not reasonable in some cases as the impact of events may diminish over time. Feelings like fear may abate as the physical reminders of the attack are reduced or eliminated. This can still be accommodated in a marked point process framework. By slightly modifying (5.1), a shot-noise process (see Rice, 1977, for a succinct description)

3

While our examples will employ an impact score that is always greater than or equal to 0, there is no reason not to allow for negative values. For example, failed or foiled attacks, might actually improve measures like public perceptions of safety and security.

5

Innovative Methods for Terrorism and Counterterrorism Data

99

S (t ) = å Yi g(t - ti ), ti < t

where g is a weight function to account for the decay in influence an event has over time, may be beneficial to incorporate. By allowing the cumulative impact of the events to decay over time, this can provide a measure of the perceived terrorism impact at a particular time t. The flexibility in the choice of g allows for a wide range of possibilities in modeling how the effects of a terrorist event will be felt over time. As a summary measure, the average of S over a period of time would give an idea of the overall impact of events for that period.

Measuring Counterterrorism We have attempted to show that the marked point process provides a mathematical framework suitable for modeling and analyzing the total impact of terrorist activity. The capability for statistical inference in this framework also allows for the quantitative analysis of counterterrorism. As counterterrorism efforts can influence one or both components of the compound process (i.e., the event rate or impact score distribution), measures of effectiveness can be created for each of these individually or combined.

Example: Detachment-88 As an illustration of how the marked point process can be used in evaluating the effects of counterterrorism efforts, we consider the history of terrorist activity in Indonesia. In October 2002, Indonesia suffered a deadly terrorist attack on the resort island of Bali. Three bombs exploded killing 202 people and injuring approximately 300. In response to this and other deadly terrorist attacks, the government of Indonesia, with support and assistance from the USA and Australia, formed a specialized counterterrorism unit called Detachment-88 (D-88). We will evaluate how the operational start of D-88 on June 30, 2003 affected terrorism in Indonesia. We used the terrorist event data from the Global Terrorism Database (GTD) (LaFree & Dugan, 2007). These data for Indonesia includes 454 incidents between 1977 and 2007. Along with the event times, the data includes several event characteristics such as the number injured, number killed, success indication, group responsible, etc. We considered a terrorist attack to be the combination of all coordinated incidents (i.e., multiple incidents may make up one attack) occurring on the same day as specified by the multiple indicator in the GTD data. This resulted in 362 terrorist attacks recorded between 1977 and 2007. Figure 5.3 shows the number of attacks, killed, and wounded per year.

100

M.D. Porter et al.

Number of Attacks

60 50 40 30 20 10 0

Number Killed

400 300 200 100

Number Wounded

0 400 300 200 100 0 1977

1979

1981

1983

1985

1987

1989

1991

1993

1995

1997

1999

2001

2003

2005

2007

Fig. 5.3 Terrorist attacks in Indonesia, by year. Top: Number of attacks. Middle: Number killed. Bottom: Number wounded. The rugs shows the attack days and the vertical line the D-88 intervention on June 30, 2003

For purposes of illustration, we construct the following impact score Y = (10 ´ success) + (1 ´ # killed) + (0.1 ´ # wounded)

(5.3)

4

where success indicates if the attack was successful. This states that every successful attack starts with an impact score of 10, every person killed in an attack increases the impact score by 1, and every ten people injured in an attack increases the score by 1. While these numbers are somewhat arbitrary, they do reflect the belief that deaths have more of an impact than injuries on most terrorism measures. This simple impact score is constructed to illustrate the use of the marked point process model in measuring terrorism and counterterrorism efforts; more complex scores (possibly including economic, social, and political factors) would likely be used in practice. Figure 5.4 shows the resulting compound score for nonoverlapping yearly intervals.

4

See http://www.start.umd.edu/gtd/downloads/Codebook.pdf for the definition of a successful attack.

5

Innovative Methods for Terrorism and Counterterrorism Data

101

Compound Score

800 600 400 200 0 1977

1979

1981

1983

1985

1987

1989

1991

1993

1995

1997

1999

2001

2003

2005

2007

Fig. 5.4 Compound score for Indonesia terrorist attacks, by year, using the impact score in (5.3)

Assessing Effectiveness There are a number of possible approaches to take in assessing the effectiveness of D-88 in combating terrorism. We will make a comparison between two of these. The first is intervention analysis, (Box & Tiao, 1975), designed to test the influence of a particular set of interventions on a process. The other technique, change point analysis, (Basseville & Nikiforov, 1993), searches for the times when a process undergoes significant changes. If the formation of D-88 had its desired effect on terrorism, then we would expect either the frequency, the severity, or both to decrease at, or a little after, the intervention time. Intervention analysis (Box & Tiao, 1975) is a common approach for testing the effectiveness of such an intervention. Commonly used with time series data, intervention analysis assumes that the process under consideration may change at the prespecified intervention time(s). Often, the form of the change (e.g., step change, pulse, or ramp) is specified in advance for each considered intervention, but the magnitude is not and estimated as part of the analysis. In this way, intervention analysis seeks to find which interventions lead to significant changes in the process along with estimating the corresponding magnitude of the changes. However, two major problems arise with such an approach. First, the model only allows for changes at the known intervention times. This will be problematic if the process actually changes due to other factors not included in the model or intervention series. Second, the form of the intervention effect could be incorrectly specified. This is a concern if there is a possibility that the true response to the intervention does not match the assumed form. For example, if a step change in the mean rate is assumed after an intervention, but the true response is gradual, then the model might not attribute the intervention to be significant when it really is or alternatively attribute a false significance when the intervention had no true effect. We note that it is not a property of intervention analysis to limit the form of intervention response. Indeed, intervention analysis will allow any form of change and it is recommend to test for multiple forms before settling on the final model (Box & Tiao, 1975; Dugan, 2010; McDowall, McCleary, Meidinger, & Hay, 1980). Nevertheless, restricting the form of the intervention effect (e.g., to a step function) appears to be a common practice and one that can lead to false inferences. We will illustrate these problems by using intervention analysis to model the D-88 intervention on the terrorism point process. Before we formulate the model,

102

M.D. Porter et al.

we need to decide what time periods we want to include in the analysis. We want to examine the times around the intervention (I=June 30, 2003), but if we include too large of a period then we will need more complicated models to explain the data. However, if we use too little data, our tests may lack a sufficient sample size to make good inference. We compromised on using the time period between 7 ⁄ 1 ⁄ 1999 and 12 ⁄ 31 ⁄ 2007. This uses 4 years before and about 4.5 years after the intervention time. This avoids including the major reduction in recorded terrorist events around 1998, but still includes several years to build models and estimate parameters. As an exploratory tool, we constructed kernel estimates of the daily compound score Z(t) = Z(t, t), rate l(t), and expected impact score Y (t ) = Y (t , t ) during the analysis period (see Fig. 5.5). The estimated scores are given by (see Wand & Jones, 1995) n

Zˆ (t ) = å K (t - ti )Yi , i =1

λˆ (t )

n

= å K (t - ti ), i =1 n

Yˆ (t ) =

å å

i =1 n

K (t - ti )Yi

i =1

K (t - t i )

Zˆ (t ) = , λˆ (t )

(5.4)

where K(u) is a mean zero Gaussian kernel with standard deviation of 120 days and Yi is the impact score for the ith attack.

Intervention Analysis We begin by assuming the event times come from a Poisson point process with intensity5 l(t). Furthermore, we assume that the intensity is constant with a possible step change at the intervention time. This can be written log λ (t ) = β + θ DI (t ),

(5.5)

where DI(t) = 1 for t ³ I and 0 for t < I models the step function response, I is the intervention time, and q is the magnitude of the change. The maximum likelihood parameter estimates6 are β = -1.86 and θ = -1.59 (p-value < 0.001). This suggests that the intervention results in a significant reduction in the number of terrorist attacks.

5

Note that this is equivalently to the hazard function used in survival analysis. See Brown, Barbieri, Eden, and Frank (2003) for a discussion of the general correspondence between the intensity and hazard functions for more complex models. 6 Using Poisson regression via a generalized linear model (GLM) (McCullagh & Nelder, 1989).

5

Innovative Methods for Terrorism and Counterterrorism Data

103

Compound Score

2.5 2.0 1.5 1.0 0.5 0.0

Attack Rate

0.15

0.10

0.05

Impact Score

0.00

20 15 10 5 2000

2001

2002

2003

2004

2005

2006

2007

Fig. 5.5 Kernel estimate of scores using Gaussian kernels with a standard deviation of 120 days. Top: Compound score Zˆ (t ) . Middle: Attack rate λˆ (t ) . Bottom: Average impact score Yˆ (t ) . The rugs shows the attack days and the vertical line the D-88 intervention on June 30, 2003

Figure 5.6 shows the intensity estimate from the fitted intervention model with step change (5.5). Also plotted is the kernel intensity estimate (5.4), revealing more subtle patterns in the event rate and two potential problems with the intervention analysis. First, it appears that the piecewise constant rate imposed by our model was a poor choice. The kernel estimate shows the initial rate to be at a high level with a peak during 2001 and then steadily decreases down to the intervention time. This raises concern over the implicit assumption of a constant mean for the intensity. Similarly, the post-intervention rate appears to decrease steadily, not step down as was postulated in the model. Violation of these assumptions brings into question our original finding of significance. We note that this is not a problem inherent to intervention analysis, but rather due to the incorrect restriction on the form of the model and post-intervention response. A second concern, and one directly linked to the method of intervention analysis, is related to the timing of the change. We have used the formation of D-88 as our intervention time, but what if the change occurred at a different time? If there was a period of training before D-88 actually began operations, then the actual change

104

M.D. Porter et al.

Intervention Estimate Kernel Estimate

^ λ(t)

0.15

0.10

0.05

0.00 2000

2001

2002

2003

2004

2005

2006

2007

Fig. 5.6 Estimated intensities from the intervention analysis with step response and kernel model (Gaussian kernel with standard deviation of 120 days)

could be some time after the originally proposed intervention time. Alternatively, there could be other factors (in addition to the formation of D-88) driving the change in intensity. By only testing for one change point, we may be missing the real interventions or changes that affected the incident rate. Without considering the possibility of alternative change points, how can we be confident that the intervention was responsible for the shift?

Change Point Analysis An alternative to intervention analysis, and one that overcomes some of its limitations, is change point analysis. Change point analysis can be described in terms of a hypothesis test. The null hypothesis assumes that no changes have taken place in the process, while the alternative hypothesis assumes that (at least) one change has occurred. However, there is a further specification of the alternative hypothesis that indicates the timing of the change(s). Thus, a change point analysis not only tests for the presence of changes in the process, but also tests for when the changes occurred. This can help identify spurious claims of intervention significance due to actual changes (not related to the intervention) that happened around the same time. Intervention analysis, on the other hand, uses an alternative hypothesis that the change only occurs at the intervention time. By limiting the alternative hypothesis to only consider a change at the intervention time, intervention analysis is implicitly assuming that the process can not change at other times. This is a crucial distinction. Intervention analysis only tests to determine if the model with the change point fits the data better than the model without the change point. It does not consider that there could be a better explanation for the change.

5

Innovative Methods for Terrorism and Counterterrorism Data

105

If the true change occurs at another time point (especially before the intervention), analysis results could lead to the false conclusion that the intervention was significant. For intervention analysis to give an unbiased answer it must be determined that there are no other possible change points. Alternatively, change point analysis, by considering the possibility that the change could be due to other causes, can provide more assurance that if a change is found in the process it can be correctly attributed to the intervention. To illustrate the differences with intervention analysis, consider the assumption that only one change can take place in the process. While both approaches specify the same null model of no change, they differ on how they specify the alternative. For change point analysis, there are actually a number of sub-hypotheses, each one corresponding to the change occurring at a particular time. For example, if we assume that the actual change point could be any day between June 30, 2000 and June 30, 2006 (a 6-year window around the intervention time), then we would specify 2,1907 alternative sub-hypotheses. Consider the original problem of testing the significance of the D-88 intervention on the point process intensity. Keeping for the moment the assumption of a piecewise constant intensity, we can explicitly state the hypotheses: H0 :

log λ (t ) = β all t,

(5.6) ì β pre t
1. Estimate b for the model under H0. 2. Estimate the model parameters (bpre, bpost), for every change point k. 3. Calculate R(k), the log-likelihood ratio of Ha(k) over H0 to identify possible change point(s). 4. Test the significance of each R(k). The following example will illustrate the use of change point analysis for testing the D-88 intervention.

Testing Change in Attack Rate In order to begin the change point analysis we plot the change profile (top of Fig. 5.7) to see how much evidence there is for a change at all possible change points. The change profile R(k) is a measure of the evidence that change occurred at time k. If there is no prior information concerning the timing of the change point, then this is 7

365 days per year times 6 years.

log−likelihood ratio R(k)

106

M.D. Porter et al.

50 40 30 20 10 0 2000

2001

2002

2003

2004

2005

2007

Max ChgPt Estimate Kernel Estimate

0.15

^ λ(t)

2006

0.10

0.05

0.00 2000

2001

2002

2003

2004

2005

2006

2007

Fig. 5.7 Top: Change profile plot of the log-likelihood ratio for testing (5.6). The vertical line is the D-88 intervention and the dotted horizontal line shows the 0. 01 reject level (R0. 01 = 6. 12). Bottom: The estimated piecewise constant intensity using k = 1 / 14 / 2002 , the most likely change point

proportional to the probability that time k is the change point and we can normalize to convert this into a probability density function. This shows that R(k) takes its maximum value at k = January 14, 2002 and scores about 15% higher than R(I), the score at the intervention time. Overall, this change profile suggests that if the change is found to be significant, there is strong evidence to suggest that it actually occurred sometime prior to the intervention time – as far back as mid to late-2001. The mid 2001 spike in the change profile corresponds to a cluster of attacks that occurred in August. Based on the form of the model, there is little evidence of change before this cluster of attacks. The bottom of Fig. 5.7 shows the estimated values of bpre and bpost corresponding to a change occurring at the most likely change point. This is the time that provides the most significant separation between the two intensity estimates. But is this a statistically significant change? Testing for significance in a change point model is more complex than in an intervention analysis because we are testing over multiple hypotheses (one test for each possible change point). If we do not adjust for this, we will have an overinflated false positive rate. One way to address the multiple testing problem is to use a bootstrap resampling method. We use the parametric bootstrap by first generating a large number of “bootstrap” samples by simulating data from the fitted null model. The test statistic is then calculated for every possible change point in

Innovative Methods for Terrorism and Counterterrorism Data

log−likelihood ratio R(k)

5

107

10 8 6 4 2 0 2000

2001

2002

2003

2004

2005

2006

2007

Max ChgPt Estimate Kernel Estimate

0.20

^λ(t)

0.15 0.10 0.05 0.00 2000

2001

2002

2003

2004

2005

2006

2007

Fig. 5.8 Top: Change profile plot of the log-likelihood ratio for testing (5.7). The vertical line is the D-88 intervention and the dotted horizontal line shows the 0. 01 reject level (R0. 01 = 7. 45). Bottom: The estimated intensity from the trend model using k = 7 / 22 / 2001 , the most likely change point

the bootstrap sample giving a set of test statistics Rboot(k). The maximum value of the test statistic, over all possible change points k ∈ K, is recorded for each bootstrap sample. The estimated significance of the original observation is by the proportion of bootstrap values that exceed the original test scores (see Efron & Tibrshirani, 1993, for details). Running 1,000 bootstrap samples on our data gives a threshold of R0. 01 = 6. 12 for rejection at the 0. 01 level. Since most values of R(k) exceed this level (see top of Fig. 5.7), this model does suggest that there is a significant change in the process, however the timing of the change appears to be well before the intervention time. We still need to be cautious about reaching premature conclusions as the constant intensity model does not seem to provide a sufficient fit to the data. Looking at the kernel intensity estimate of the attack rate (Fig. 5.5), it appears the rate is decreasing (starting around late-2001), violating the assumption of a piecewise constant attack rate. A better model can be obtained by adding a linear trend term. The new hypotheses can be stated as: H0 :

log λ (t ) = β0 + β1t all t,

Ha (k ) :

ì β0pre + β1pre t log λ (t ) = í post pre îβ0 + β1 t

Figure 5.8 shows the change profile plot for this model.

t
(5.7)

108

M.D. Porter et al.

The model results are quite revealing, the linear trend improves the fit of the null model, but more importantly puts the change point in mid-2001, about 2 years before Detachment-88 becomes operational and a year prior to the Bali bombings. Additional bootstrap testing reveals that this is a significant change point (R0. 01 = 7. 45) for the attack rate. We can also examine both the impact and compound scores for the presence of significant change points.

Testing Change in Impact Distribution In order to test for a change in the impact densities the hypotheses can be stated as: H0 : Ha (k ) :

f ( y | t ) = f null ( y | t ) all t , ì f pre ( y | t ) f ( y | t ) = í post î f (y | t)

t
(5.8)

where f(y | t) is the density of the impact score at time t. Since the impact scores do not fit a standard distribution,8 we estimated the impact densities using histograms.9 Figure 5.9 shows the change profile plot using the log-likelihood ratio test statistic. Once again, the most likely change point is in mid-2001, although the bootstrap testing reveals that it is not significant (R0. 01 = 10. 55). Despite this, the Bali bombing in 2002 and Marriot Hotel bombing in 2003 (occurring after the estimate change point) were large attacks that made international headlines. Using a different impact measure including additional factors, such as economic impact, may provide a significant change point.

Testing Change in Compound Score Testing the compound score is straightforward due to the structure of the marked point process. The log-likelihood ratio for the compound process is simply the addition of the log-likelihood ratios from the intensity and impact processes. This implies that the change profiles from the trend model for the intensity and histogram model for the impact distribution can be added together to produce the results as shown in Fig. 5.10. Once again the potential change point is identified in mid-2001 (May 27 to Aug 12, 2001 all exceed the 0. 01 reject level of R0. 01 = 13. 44) and bootstrapping reveals 8

The impact score (5.3) is actually a discrete distribution with a mass at 0 corresponding to events that are not successful, a mass at 10 corresponding to events that have no injuries or fatalities, and also a long tail due to the extreme values of some large, deadly attacks. 9 Eighteen equal width bins were constructed over the range [0, 360] and the estimated likelihood of observing an observation in a bin is the number of impact scores that fell into the bin divided by the total number of impact scores considered.

5

Innovative Methods for Terrorism and Counterterrorism Data

109

log−likelihood ratio R(k)

7 6 5 4 3 2

2000

2001

2002

2003

2004

2005

2006

2007

log−likelihood ratio R(k)

Fig. 5.9 Change profile plot of the log-likelihood ratio for testing the change in impact score distribution (5.8). The vertical line is the D-88 intervention. With a p-value of 0.25, the change point is not found to be significant

15

10

5

2000

2001

2002

2003

2004

2005

2006

2007

4

Max ChgPt Estimate Kernel Estimate

^ Z(t)

3

2

1

0 2000

2001

2002

2003

2004

2005

2006

2007

Fig. 5.10 Top: Change profile plot of the log-likelihood ratio for the compound score. The vertical line is the D-88 intervention and the dotted horizontal line shows the 0. 01 reject level (R0. 01 = 13. 44). Bottom: The estimated mean compound score using k = 7 / 22 / 2001 , the most likely change point. This is the product of the trend intensity and mean impact scores

that this is a significant result. This result can lead us to state with some confidence that there was indeed a change in the nature of Indonesian terrorism in 2001. The downward trend in the frequency of attacks begins in late-2001 and continues through 2007. According to our choice of impact score, there is mild evidence for a change around 2001. However, when combined, the compound score provides strong evidence for a change in the terrorist process during this time period. The data do

110

M.D. Porter et al.

not, however, suggest what caused this change in 2001; only that significant evidence supports a much earlier change point than the formation of D-88.

Discussion/Conclusion We have presented a marked point process approach to measuring terrorism that accounts for the both the frequency and impact of the terrorist attacks. In addition, we have presented the use of change point analysis to provide more insight into the timing and significance of an intervention. It was shown how intervention analysis, using a poor model, resulted in claims that the formation of D-88 was successful at significantly reducing the frequency of terrorist attacks. However, using a more appropriate trend model in a change point analysis revealed that there was little evidence to suggest that the intervention was significant. Instead, the change point analysis suggested that a significant change in the terrorist process occurred almost 2 years prior. The analysis and data provide evidence to support the assertion that mid-2001 marks the beginning of a significant shift in terrorist behavior with a decrease in the frequency of terrorist activities and change in attack severity. It is interesting to note that this change point occurs in very close proximity to the attacks of September 11, 2001. Is it possible also that this shift in strategy is not unrelated? As the dominant terrorist group in Indonesia at that time, Jemaah Islamiyah had prior known ties to Al-Qaeda. Is it possible that this hypothetical shift of strategy was part of a global shift in strategy for the whole Al-Qaeda network, and that the attacks of September 11, 2001 were the first of this new campaign focusing on fewer, more sophisticated, and higher impact attacks? These questions and others are important and need to be addressed in detail. While our analysis does not provide the answers to all of these questions, it does illustrate the utility of mathematical-based analysis techniques in helping to illuminate areas for future investigation as well as a framework for measuring and evaluating terrorism and counterterrorism methods. In addition, we highlighted the need to review the choices of tests and models to ensure important assumptions are not violated, and correct inferences are made. More complex models examining seasonality, larger numbers of covariates, or processes like self-excitation may require the development of resources, such as partnerships with experts from the fields of statistics or mathematics. We feel these challenges will be rewarded by a deeper understanding of terrorism. In addition to more complex models, future research should incorporate methodologies like Bayesian statistics which facilitate the incorporation of expert knowledge into the mathematical analysis. It is our ultimate hope that the examples and analysis we have presented will serve as the basis for future cooperative exploration and research between social scientists, policy makers, and those from the mathematical sciences.

5

Innovative Methods for Terrorism and Counterterrorism Data

111

An Afterthought It is important to note that while the results from the change point analysis may cast doubt that the establishment of D-88 resulted in a significant change to the terrorism process in Indonesia, it does not suggest that the operational activities of D-88 were ineffective. The continuation of the downward trend in attack frequency and lack of massive attacks after D-88 was formed supports the conjecture that D-88 is having a positive influence on the reduction of terrorism – even if it did not lead to an abrupt change in the current trend. It would be beneficial to evaluate the individual operations of D-88 at a smaller time scale in order to assess their effectiveness. Acknowledgements We would like to thank CEPS/ISSR Research Assistant Jacqui Davis for her help in editing and reviewing various drafts of this chapter.

References Abadie, A., & Gardeazabal, J. (2003). The economic costs of conflict: A case study of the basque country. American Economic Review, 93, 113–132. Barros, C. P. (2003). An intervention analysis of terrorism: The Spanish ETA case. Defence and Peace Economics, 6(14), 401–412. Basseville, M., & Nikiforov, I. (1993). Detection of abrupt changes: Theory and application. Englewood Cliffs: Prentice Hall. Box, G. E. P., & Tiao, G. C. (1975). Intervention analysis with application to economic and environmental problems. Journal of the American Statistical Association, 70(349), 70–79. Brown, E., Barbieri, R., Eden, U., & Frank, L. (2003). Likelihood methods for neural spike train data analysis. In J. Feng (Ed.), Computational neuroscience: A comprehensive approach, chap. 9. London/Boca Raton: Chapman & Hall/CRC. Chen, A. H., & Siems, T. F. (2004). The effects of terrorism on global capital markets. European Journal of Political Economy, 20, 349–366. Clauset, A., Young, M., & Gleditsch, K. S. (2007). On the frequency of severe terrorist events. Journal of Conflict Resolution, 51(1), 58–87. Collier, P. (1999). On the economic consequences of civil war. Oxford Economic Papers, 51, 168–183. Cox, D. R. (1972). Regression models and life tables (with discussion). Journal of the Roayl Statistical Society, Series B, 34, 187–220. Daley, D. J., & Vere-Jones, D. (2003). An introduction to the theory of point processes (2nd ed., Vol. I). New York: Springer. Dugan, L. (2010). Estimating effects over time for single and multiple units. In A. R. Piquero & D. Weisburd (Eds.), Handbook of quantitative criminology (pp. 741–763). New York: Springer. Dugan, L., LaFree, G., & Piquero, A. (2005). Testing a rational choice model of airline hijiackings. Criminology, 43, 1031–1066. Efron, B., & Tibrshirani, R. (1993). An introduction to the bootstrap. London/Boca Raton: Chapman & Hall/CRC. Enders, W., & Sandler, T. (1991). Causality between transnational terrorism and tourism: The case of Spain. Terrorism, 14, 49–58. Enders, W., & Sandler, T. (1993). The effectiveness of antiterrorism policies: A vector-autoregression intervention analysis. The American Political Science Review, 4(87), 829–844.

112

M.D. Porter et al.

Enders, W., & Sandler, T. (2000). Is transnational terrorism becoming more threatening? Journal of Conflict Resolution, 44, 307–332. Enders, W., & Sandler, T. (2002). Patterns of transnational terrorism, 1970-1999: Alternative timeseries estimates. International Studies Quarterly, 2(46), 145–165. Enders, W., & Sandler, T. (2006). The political economy of terrorism. New York: Cambridge University Press. Fielding, D. (2003). Counting the cost of the Intifada: Consumption, saving and political instability in Israel. Public Choice, 116, 297–312. Frey, B. S., & Luechinger, S. (2005). Measuring terrorism. In A. Marciano & J.-M. Josselin (Eds.), Law and the state: A political economy approach (pp. 142–181). Cheltenham, UK/Northampton, MA: Edward Elgar. Frey, B. S., Luechinger, S., & Stutzer, A. (2007). Calculating tragedy: Assessing the costs of terrorism. Journal of Economic Surveys, 21(1), 1–24. Kabelfleisch, J. D., & Prentice, R. L. (2002). The statistical analysis of faliure time data. New York: Wiley. LaFree, G., & Dugan, L. (2007). Introducing the global terrorism database. Terrorism and Political Violence, 19, 181–204. Lafree, G., Dugan, L., & Korte, R. (2009). The impact of British counter terrorist strategies on political violence in Northern Ireland: Comparing deterrence and backlash models. Criminology, 47, 17–45. McCullagh, P., & Nelder, J. (1989). Generalized linear models. London/Boca Raton: Chapman & Hall/CRC. McDowall, D., McCleary, R., Meidinger, E. E., & Hay, R. A. (1980). Interrupted time series analysis (Vol. 21). Thousand Oaks: Sage. Perl, R. (2007). Combating terrorism: The challenge of measuring effectiveness (Technical Report, Congressional Research Services Report RL33160). Rice, J. (1977). On generalized shot noise. Advances in Applied Probability, 9(3), 553–565. Wand, M. P., & Jones, M. C. (1995). Kernel smoothing. London/Boca Raton: Chapman & Hall/CRC.

Chapter 6

Introducing Group-Based Trajectory Analysis and Series Hazard Modeling: Two Innovative Methods to Systematically Examine Terrorism Over Time Laura Dugan and Sue-Ming Yang

Introduction Terrorism research has historically been characterized by a high number of case studies (Bennett & Elman, 2007; Odell, 2004); though important, the ability of making causal inference is considered the biggest weakness of qualitative research. Additionally, qualitative studies provide little support for generalization that could inform policy makers on the best strategies to reduce the terrorist threat in their jurisdiction. In these studies, scholars have translated documentation on and communiqués by clandestine terrorist organizations into insightful discussions on topics such as terrorist organizational structure (Crenshaw, 2001), terrorist tactics (Jackson et al., 2005), effectiveness of counter terrorist strategies (LaFree, Dugan, & Korte, 2009; Soule, 1989), terrorist mindsets (Cordes, 2001), and how terrorist groups might end (Cronin, 2006; Jones & Libicki, 2008; United States Institute for Peace, 1999). While useful individually, as a group, case studies provide a distorted picture of terrorism, as qualitative research tends to focus on more high profile cases. In other words, even if we were to compile a list of all terrorism case studies and combined them in an extensive narrative review to generate broader conclusions, they would be biased toward the more extraordinary cases. Further, most – if not all – of these micro investigations were conducted in the absence of the broader, more global view of all terrorist organizations, their campaigns, and states’ efforts to stop them. In other words, in order to capture a more objective characterization of terrorism, scholars must draw upon the entire universe of terrorist events.

L. Dugan (*) Department of Criminology and Criminal Justice, University of Maryland, College Park, MD, USA e-mail: [email protected] S.-M. Yang Institute of Criminology, National Chung Cheng University, Chia-Yi, Taiwan e-mail: [email protected] C. Lum and L.W. Kennedy (eds.), Evidence-Based Counterterrorism Policy, Springer Series on Evidence-Based Crime Policy 3, DOI 10.1007/978-1-4614-0953-3_6, © Springer Science+Business Media, LLC 2012

113

114

L. Dugan and S.-M. Yang

The reason for the absence of systematic analysis of the universe of terrorism is twofold. First, many of the home disciplines of the scholars who study terrorism favor qualitative approaches to analysis, such as history, political science, and social psychology (Crenshaw, 2001; McCauley, 2001; Rapoport, 2004). Disciplinary norms and preferences determine the ways in which research is typically conducted; and the training offered by each discipline largely restricts the methods used to conduct research (see Lum & Yang, 2005, for discussion). Second, until very recently, a comprehensive dataset that chronicles both transnational and domestic terrorist attacks over an extended period of time simply did not exist. In the late 1960s and early 1970s two efforts began to systematically chronicle transnational terrorist attacks. International Terrorism Attributes of Terrorist Events (ITERATE) is an ongoing collection of transnational terrorist attacks managed by Edward Mickolus – a former CIA analyst. In 1972 the RAND Corporation, a nonprofit policy research institution, also began collecting information on international terrorist attacks (starting the series in 1968). The initiation of this research arose amidst the United States government’s growing concern with the escalating occurrence of international terrorist attacks and the potential threat posed to Americans abroad. Brian Michael Jenkins led this effort and undertook a multidisciplinary approach, using the expertise of staff from various fields such as psychology, political science, intelligence, weapons technology and computer science to address a broad array of terrorism-related topics. While both of these datasets have been used in numerous important studies, we now know that domestic acts of terror outnumber transnational attacks by as much as 7 to 1 (LaFree, Dugan, Fogg, & Scott, 2006). Thus, even efforts to generalize from these comprehensive datasets bias our understanding toward only those organizations or campaigns with enough resources to be able to operate across national borders. In April 2001, things changed, as the RAND Corporation, under the sponsorship of the National Memorial Institute for the Prevention of Terrorism (MIPT) began collecting both domestic and international terrorist attacks starting in January 1998. RAND recognized that the distinction between international and domestic terrorism was becoming much less obvious. Terrorist groups were increasingly difficult to classify along purely ethnic and nationalist lines, and as information on domestic acts of terrorism was more readily available, it made sense to include these attacks to more accurately understand emerging trends (Dugan, LaFree, Cragin, & Kasupski, 2008). The RAND-MIPT Terrorist Chronology was part of a larger Terrorism Knowledge Base found on MIPT’s website, and available for public perusal. Despite the changing availability of comprehensive terrorist chronology, only information on international terrorist attacks was available prior to 1998. Excluding domestic terrorism is problematic for many reasons. First, by examining only international attacks, the activity of any organization is truncated to show only their efforts against foreign targets and ignores completely those terrorist organizations that only attack domestic targets (LaFree, Yang, & Crenshaw, 2010). While some terrorist groups (e.g., al Qaeda, Mujahedin-E-Khalq) have global operations that cut across domestic and international lines, others (e.g., Abu Nidal, Aum Shinrikyo, Kurdistan Workers’ Party, and Popular Front for the Liberation of Palestine) have operations in multiple countries and hence, may simultaneously be engaged in acts

6 Introducing Group-Based Trajectory Analysis and Series Hazard Modeling…

115

of both domestic and international terrorism. Still others (e.g., Revolutionary Armed Forces of Colombia [FARC], IRA, Sendero Luminoso [SL]) operate almost entirely in the domestic realm. By recording only international terrorist attacks, we capture only part of the picture. In fact, LaFree, Yang, and Crenshaw (2009) conducted a study examining attacks of some “anti-American” terrorist organizations over 35 years and found that 95% of the attacks of those high profile groups were against their home countries. Thus, excluding domestic terrorist cases would lead to a huge loss of data and result in incomplete conclusions of activity patterns for those terrorist groups. Furthermore, when analyzing root causes of terrorism, studies often rely on variables that would be theoretically more relevant to intrastate terrorism rather than interstate terrorism. In his highly cited article Li (2005) finds a relationship between democratic participation and transnational terrorism relying on a theory claiming that through democratic participation, different groups have the means to bring about social change, precluding the need to resort to violence. Yet, the benefits of democratic participation are strictly domestic, making his argument illogical for explaining attacks by foreign organizations. In fact, Young and Dugan (2011) later reevaluate Li’s argument using measures of domestic terrorism in the Global Terrorism Database (GTD) and find that when states have greater political constraints, terrorism is more likely and more frequent. The GTD is the only database to chronicle both international and domestic terrorist attacks over more than 35 years (data span will reach 40 years by the end of 2011) across the entire globe.

Global Terrorism Database The GTD began as 58 boxes containing nearly 70,000 index cards, each recording the details of a single terrorist attack or coordinated attack (often including the perpetrator) for the entire world from 1970 to 1997, collected by the Pinkerton Corporation’s Global Intelligence Service (PGIS). The collectors of the PGIS data aimed to record every known terrorist event across nations and over time. They originally collected this information from multilingual news sources for the purpose of performing risk analysis for United States business interests (LaFree & Dugan, 2007). These data, combined with more recent data collected by the National Consortium to Study Terrorism and the Response to Terrorism (START), form the GTD. The key features of the GTD that distinguish it from RAND-MIPT and ITERATE is that it includes both domestic and international attacks over 4 decades and that the criteria for inclusion is broader, allowing more flexibility with analysis.1 1

In order to consider an incident for inclusion in the GTD, all three of the following attributes must be present: (1) The incident must be intentional – the result of a conscious calculation on the part of a perpetrator. (2) The incident must entail some level of violence or threat of violence – including property violence, as well as violence against people. (3) The perpetrators of the incidents must be sub-national actors. This database does not include acts of state terrorism. In addition, at least

116

L. Dugan and S.-M. Yang

Methods to Systematically Analyze All Terror Lum, Kennedy, and Sherley (2006) criticize the lack of rigorous quantitative evaluation research of counter terrorism programs. Specifically, out of 20,000 studies, only 3–4% relied on empirical data or methods to support their conclusions. However, since the publication of this study, more rigorous empirical research evaluating terrorist outcomes have gradually been conducted (see Dugan, LaFree, & Miller, 2011; Dugan, LaFree, & Piquero, 2005; Enders & Sandler, 2006; LaFree, Dugan, et al., 2009; LaFree, Morris, & Dugan, 2009; LaFree, Morris, Dugan, & Fahey, 2006). To better understand terrorism and the effects of counter terrorism tactics and strategies, innovative methodological advances have been directed toward terrorist studies. Most importantly, longitudinal analyses have been conducted to study the effects of democratic attributes (Li, 2005; Young & Dugan, 2011), economic well-being (Kruger & Maleckova, 2003; Li & Schaub, 2004; Piazza, 2006), and others (Callaway & Harrelson-Stephens, 2006) on terrorism. For example, zero-inflated count models accommodate models that include countries that never experienced terrorism (LaFree, Morris, et al., 2006; Young & Dugan, 2011). Time series has been used to examine trends over time, and the effects of events on changes in those trends (Enders & Sandler, 2006). More sophisticated analysis using multiple time series has been used to examine how interventions influence changes in different types of terrorist activity simultaneously (Enders & Sandler, 1993). In this chapter, we introduce two innovative methodologies that were used in the field of criminology and have then been applied to systematically analyze terrorism over time. The first innovation, group-based trajectory analysis (GBTA), presents the big picture of terrorism by grouping countries or terrorist organizations with similar attack patterns. The second innovation, series hazard modeling, drills down to specific cases (e.g., countries, organizations, or movements) to estimate changes in the hazard of another attack based on changes in independent variables, such as government interventions or other potential turning points. These two approaches help us understand longitudinal terrorism patterns in different way. Each method has its unique strengths and provides answers to different questions regarding terrorism trends. Specifically, we think the GBTA (or latent class growth analysis) is useful to identify distinct developmental patterns across units (terrorist organizations or countries) and to

two of the following three criteria must be present for an incident to be included in the GTD: Criterion 1: The act must be aimed at attaining a political, economic, religious, or social goal. In terms of economic goals, the exclusive pursuit of profit does not satisfy this criterion. It must involve the pursuit of more profound, systemic economic change. Criterion 2: There must be evidence of an intention to coerce, intimidate, or convey some other message to a larger audience (or audiences) than the immediate victims. It is the act taken as a totality that is considered, irrespective if every individual involved in carrying out the act was aware of this intention. As long as any of the planners or decision-makers behind the attack intended to coerce, intimidate or publicize, the intentionality criterion is met. Criterion 3: The action must be outside the context of legitimate warfare activities. That is, the act must be outside the parameters permitted by international humanitarian law (particularly the prohibition against deliberately targeting civilians or noncombatants).

6 Introducing Group-Based Trajectory Analysis and Series Hazard Modeling…

117

separate terrorism “hot-spots” from the low-risk cases. The series hazard model can be used to estimate the role of specific factors in shifting the attack risk for specific units. Together these methods can provide “group-level” and “individual-level” information within the context of the larger picture.

Group-Based Trajectory Analysis In this section, we introduce GBTA and review some empirical examples that have applied GBTA to analyze patterns of terrorism. Because conducting GBTA is a nuanced process, we present that process through an example using a subset of terrorism data drawn from GTD. The analysis results will be presented and the strengths and limitations of GBTA will also be discussed. Because terrorism is an evolving phenomenon, it is critical to examine terrorism longitudinally, accounting for historical and social changes. Thus, in order to accurately model the trends of terrorism and its related impacts, researchers must adopt methodological strategies that can account for both change and continuity over a long period of time. Many familiar quantitative methods are well equipped to estimate these dynamic processes, such as time series analysis, survival analysis, hierarchical modeling, latent growth curve analysis, and trajectory analysis, just to name a few. Among which, GBTA has been increasingly applied to the study of terrorism (Dugan et al., 2011; LaFree, Morris, et al., 2006; LaFree, Morris, et al., 2009; LaFree, Yang, et al., 2009; Miller, 2009; Yang & LaFree, 2011), and is considered a promising tool for scholars who are interested in exploring population heterogeneity as well as behavioral continuity among terrorist organizations or geographic units (like countries, provinces, or regions). GBTA was first introduced by Nagin and Land (1993) to model developmental patterns of individual criminality (e.g., Blokland, Nagin, & Nieuwbeerta, 2005; Bushway, Sweeten, & Nieuwbeerta, 2009; Nagin, 1999; Nagin, Farrington, & Moffit, 1995; Nagin & Tremblay, 1999, 2001). Essentially, GBTA approximates the population with a set number of groups that follow distinctive developmental pathways. As such, GBTA helps identify the latent trajectories and estimate their developmental patterns. It is a very useful tool for scholars who want to explore differences in trends across terrorist organizations or countries over time. Though this method was initially developed to model individual criminal trajectories, it has been applied to study phenomena at other units of analysis. GBTA has also been used in studies of crime distribution at places (Griffiths & Chavez, 2004; Weisburd, Bushway, Lum, & Yang, 2004; Weisburd, Groff, & Yang, 2010; Weisburd, Morris, & Groff, 2009; Yang, 2010), terrorism patterns over time at the national (LaFree, Morris, et al., 2009), provincial (Yang & LaFree, 2011) and organizational levels (Dugan et al., 2011; LaFree, Yang, et al., 2009; Miller, 2009). For example, LaFree, Morris, et al. (2006) examined terrorist activities across 201 countries over 28 years. Using GBTA, they found that the activity patterns of these 201 countries followed 4 developmental trajectories. Additionally, they

118

L. Dugan and S.-M. Yang

pointed out that terrorist incidents, just like crime, are concentrated in a few countries, or “hot spots,”2 around the world. The extent of concentration is pretty substantial. In a follow-up study using GBTA, LaFree, Morris, et al. (2009) showed that out of 206 countries and territories, 10 countries accounted for 38% of total attacks over 37 years. They further concluded that the concentrated phenomenon remained stable over a long time period. Based on the concentration patterns and the stability of those “hot spots,” the findings provide a clear policy directive to focus scarce resource toward those places that are most fraught with terrorism, and could achieve the most prevention benefit. In addition to identifying the “hot spots” where activities tend to concentrate, GBTA has also been used to examine the variability between cases. For example, Yang and LaFree (2011) use it to examine province-level terrorist activity across Indonesia, the Philippines, and Thailand. In this study, GBTA helps to identify the variability of terrorist attacks across provinces within and between countries. The findings highlight the great heterogeneity of geographic distribution of terrorist attacks within the same country. As such, they concluded that it is important not to apply one-size-fits-all counter measures on all areas without first assessing their associated risk. GBTA has also been used to study terrorism-related subjects that are more volatile than geographic units. LaFree, Yang, et al. (2009), for instance, applied GBTA on scrutinizing attack patterns of 53 terrorist organizations that were identified by the U.S. State Department and other authorities as highly dangerous to the U.S. They identified four different activity trajectories of those “anti-U.S.” terrorist organizations based on 35 years of data. They found that within these high profile terrorist organizations, there is substantial case-by-case variability of their activity patterns. That is, not all those terrorist organizations flagged by the State Department were equally active in attacking American targets. Moreover, while a few terrorist organizations were extremely active and accounted for the majority of the attacks, some of these labeled as “dangerous” terrorist organizations had never attacked any U.S. targets and about half of the “high risk” organizations struck only 2–3 times over the 35 years of the study period. In this study, GBTA helped classify terrorist organizations into different activity trajectories based on their attack frequencies and trends.

Analyzing Attack Patterns of “Anti-U.S.” Foreign Groups In this section of the chapter, we use the same data as the LaFree, Yang, et al. (2009) study but do not make the distinction about whether the attacks were against U.S. or non-U.S. targets. This analysis includes the same 53 foreign nonstate actors that 2

The term “hot spot” tends to be used to describe a high concentration of social problems like crime in a very small number of places (see Sherman & Weisburd, 1995). In this case, LaFree, Morris, et al. (2009) follow the same fashion and refer the small number countries with a large share of terrorist attacks as terrorist hot spots.

6 Introducing Group-Based Trajectory Analysis and Series Hazard Modeling…

119

have posed a serious threat to U.S. citizens since 1970 based on classifications done by the U.S. State Department, Office of the Historian, and literature searches. More detail about the designation of these “high-risk” groups can be found in LaFree, Yang, et al. (2009). The list of the 53 terrorist organizations3 included in the study can be found in Table 6.1. While it may seem limited to focus on a subset of the total number of terrorist organizations, these 53 terrorist organizations were responsible for 16,916 terrorist incidents from 1970 to 2004. In this chapter, we analyze the activity patterns of these terrorist organizations over 35 years as an attempt to identify groups that follow different trajectories. We present the results along with discussions of the strengths and weaknesses of GBTA. To understand the general “contribution” of these 53 terrorist organizations collectively by terrorist attacks, we present Fig. 6.1, which shows a simple trend of the average number of attacks across the 53 organizations from 1970 to 2004. We see that their activity began to pick up in the mid-70s, reached the peak in the 80s, and then began a rapid decline in the early 90s. The trend remained stable over the last 10 years of the series. From this figure, we can see the general trend of these organizations follow the traditional boom and bust cycle. Nonetheless, this figure fails to tell us whether all the organizations contributed to this overall trend similarly, following the same behavioral pattern; or if there is substantial amount of variability among these organizations. Looking at this trend, it might be interesting to compare it with David Rapoport’s wave argument. Specifically, Rapoport (1992: 1064) has stated that the historical trend of terrorism can be divided into four waves based on different historical backgrounds: the Anarchist Wave, the Anticolonial Wave (1920s–1960s), the New Left Wave (1960s–2000s), and the Religious Wave (1979-present). Though the time span of our data only covers the last two waves; we still could see if terrorist organizations that were active in different time period also had different orientations as predicted by Rapoport. In other words, are the leftist organizations more active earlier in the series and the religious organizations more active near the end of the series? To disaggregate the overall trend and explore the variability of trajectory of individual terrorist organization, we apply GBTA to the database composed of 16,916 terrorist incidents committed by the 53 target terrorist organizations. GBTA has several strengths that are appealing to terrorism scholars who want to explore longitudinal trends. In the section below, we will demonstrate how to use trajectory analysis using the example mentioned above and point out its strengths and weaknesses.4

3

The term terrorist “organization” and terrorist “group” tend to be used interchangeably. In this chapter, however, we want to avoid confusion between trajectory group and terrorist group and use the term “organization” when we talk about individual terrorist groups. 4 In this study, we use macro Proc Traj, a SAS procedure, to execute the analysis (see Jones et al., 2001). Other statistical programs can also be used for GBTA including M Plus©. More information about Proc Traj, programming examples, and documentation can be found at the following website http://www.andrew.cmu.edu/user/bjones/index.html.

120 Table 6.1 Terrorist organizations included in the analysis Name of group Country of origin Abu Nidal Organization (ANO) Iraq, Israel/Palestine Abu Sayyaf Group (ASG) Philippines al-Gama’at al-Islamiyya (IG) Egypt, Afghanistan Al Faran/Harkat-ul Mujahidin (HuM) Pakistan Al Qaeda International Al Qaeda in the Arabian International Penninsula (AQAP) Al Qaeda in Mesopotamia International Ansar al-Islam Iraq Black September Jordan, Lebanon, Israel/Palestine Central American Revolutionary Workers El Salvador Party (PRTC) Dev Sol Turkey Ejercito Revolucionaria del Pueblo (ERP) Argentina Eritrean Liberation Front Eritrea, Ethiopia Farabundo Marti National Liberation El Salvador Front (FMLN) Islamic Movement of Uzbekistan (IMU) Uzbekistan Jaish-e-Mohammad Pakistan Japanese Red Army (JRA) Japan Jemaah Islamiya (JI) Indonesia Lashkar-e Taiba Pakistan Lashkar I Jhangvi Pakistan Lebanese Armed Revolutionary Faction Lebanon (LARF) Lebanese Socialist Revolutionary Lebanon Organization M-19 (Movement of April 19) Colombia Manuel Rodriguez Patriotic Front Chile (FPMR) Marxist-Leninist Armed Propaganda Unit Turkey Montoneros Argentina Moro Islamic Liberation Front (MILF) Philippines Moro National Liberation Front (MNLF) Philippines Mujahideen-I-Khalq (MK) Iran National Liberation Army of Colombia Colombia (ELN) Nestor Paz Zamora Commission (CNPZ) Bolivia New People’s Army (NPA) Philippines November 17 Revolutionary Organization Greece (N17RO) Palestine Liberation Front (PLF) Israel/Palestine Patriotic Morazanista Front (FPM) Honduras People’s Liberation Forces (FPL) El Salvador

L. Dugan and S.-M. Yang

Years of activities (in GTD) 1976–1998 1993–2004 1991–1997 1995–2004 1998–2004 2004 2004 2002–2004 1971–1988 1979 1979–1996 1970–2004 1970–1992 1978–1994 2000–2004 2000–2003 1972–1988 1993–2004 1999–2004 1998–2004 1981–1985 1973–1974 1976–1997 1984–1997 1977–1980 1970–1991 1986–2004 1975–2001 1972–2001 1972–2001 1990 1970–2004 1976–2001 1979–2004 1988–1995 1977–1979 (continued)

6 Introducing Group-Based Trajectory Analysis and Series Hazard Modeling… Table 6.1 (continued) Name of group

Country of origin

121

Years of activities (in GTD)

Popular Front for the Liberation of Israel/Palestine, 1970–2004 Palestine (PFLP) Syria Popular Front for the Liberation of Israel/Palestine 1970–2003 Palestine, Gen Cmd (PFLP-GC) Popular Liberation Army (EPL) Colombia 1976–1999 Popular Revolutionary Vanguard (VPR) Brazil 1970–1976 Rebel Armed Forces of Guatemala (FAR) Guatemala 1970–1989 Red Army Faction (RAF) West Germany 1977–1993 Red Army for the Liberation of Catalonia Spain 1987 Red Brigades Italy 1974–2003 Red Brigades Fighting Communist Party Italy 1983–1987 (BR-PCC) and Fighting Communist Union (BR-UCC) Revolutionary Armed Forces of Colombia 1975–2004 Colombia (FARC) Revolutionary People’s Struggle (ELA) Greece 1976–1995 Shining Path (SL) Peru 1978–2004 Taliban Afghanistan 2001–2004 Tupac Amaru Revolutionary Movement Peru 1984–1997 (MRTA) Tupamaros Uruguay 1970–2001 Turkish People’s Liberation Army Turkey 1970–2000 Note: it is worth recalling that our analysis ends in 2004 – before al Qaeda in Mesopotamia and its successors in Iraq staged most of their attacks

Average Number of Attacks

35 30 25 20 15 10 5

1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004

0

Fig. 6.1 Average number of terrorist attacks of the 53 “anti-U.S.” Terrorist organizations from 1970 to 2004

122

L. Dugan and S.-M. Yang

Applying GBTA to the Activity of the 53 Terrorist Groups The first and the most important strength of GBTA is its ability to identify cases that follow qualitatively distinct paths. This information contributes to both policy and theory because by understanding why certain terrorist groups follow one pathway instead of another, we can gain important insight into their differing etiologies. Similar benefits apply to scholars who study terrorism at the country level; identifying countries in the same region that experienced different levels of terrorist threats or exhibited different developmental patterns can also assist in customizing appropriate counter terrorism measures. The underlying strategy used to model the trajectories is to characterize each trajectory group by a specific set of parameters that can vary freely across groups. In order to determine the optimal estimates of these parameters, a set of decisions first needs to be made regarding (1) the type of distribution, (2) the number of groups, and (3) the shape of the model. Each decision-making point will be discussed below using the 53 terrorist organizations example.

Type of Distribution In GBTA there are only three ways to model the dependent variable, logistic for binary outcomes, censored normal for continuous data, and Poisson for count data. Since the GTD dependent variable is aggregated to a count of attacks, the Poisson model was the best choice among all three possibilities. Additionally, the inspection of the data showed that its distribution is negatively skewed as not all active terrorist groups strike every year. The large number of zero observations in the database leads to over-dispersion problem; that is, when the variance of the response variable exceeds the mean of the variable. To accommodate for these issues, the zero-inflated Poisson model was used to correct for the bias and an intermittency5 parameter was added into the model to account for high probability of zero activity observed in the population (Nagin, 2005; Roeder, Lynch, & Nagin, 1999). It is also worth noting that when the range of values is too big, the high variability might make the convergence of the model estimation very difficult. Thus, it is common practice to truncate the extreme values in order to enable model convergence (see LaFree, Morris, et al., 2006; LaFree, Yang, et al., 2009, Weisburd et al., 2004, 2009). These data are no exception, as the lowest number of attacks is zero and the highest is 510 for 1989. Thus, we truncated values, by replacing all that exceed 100 with 100. Reasonable truncation values vary case by case, and we try to only truncate cases with the highest 5–7% of values under the condition that it still

5

Intermittency “refers to episodes of inactivity in a criminal career” (Nagin, 2005, page 34). The observation of zero activity can be a result of two possibilities: the true lack of activity or the lack of opportunity for such activity for the given time period. The intermittency parameter is added in the model to account for both situations.

6 Introducing Group-Based Trajectory Analysis and Series Hazard Modeling…

123

preserves the shape of the overall distribution. In this case, the results will be unaffected by the truncation process.6 Nonetheless, researchers should carefully examine their data before making a decision to truncate the data.

Number of Trajectory Groups Perhaps the most important analytical decision in the GBTA is to determine the number of trajectory groups. Similar to principle component analysis, this decision determines the optimal number of trajectory groups in order to minimize the within group differences while maximizing the heterogeneity between groups.7 The determination of the number of trajectory groups is based on both diagnostic criteria and the substantive meaning of the results. This will be demonstrated in the example below.

Shape of Trajectory Ideally, the shape of each trajectory group should be flexible enough to capture its movement, yet simple enough to maintain a parsimonious interpretation. The estimation process of GBTA allows each trajectory group to be estimated with a different polynomial order; and thus separate sets of parameters are estimated for each trajectory. Bushway, Thornberry, and Krohn (2003) pointed out that quadratic form was uniformly a better fit than a linear model when estimating individual trajectory, but cubic models perform better in the case of a small number of groups. This has been confirmed by using place-based data as well as terrorism data (LaFree, Morris, et al., 2009; LaFree, Yang, et al., 2009; Weisburd et al., 2004; Yang, 2010). In this analysis, we also found that the cubic form generally fit the data better.8 Overall Model Selection To determine the most appropriate number of groups and the shapes of each trajectory group, we follow the exhaustive approach detailed by Nagin (2005). That is, we tested for all possible combinations of number of groups and polynomial order for each trajectory group. Specifically, we began our modeling exercise by fitting the

6

In our study, the only trajectory group affected by the truncation practice is the 1980s Boom. When plotting the projected trajectories and the actual values, it is clear that the projections and actual values almost totally overlap (except for 1980s Boom and a very minor deviation of the 1990s Boom) over the 35 years. 7 Using the same data, Gregory Ihrie (2010) ran longitudinal Principle Analysis and reached very similar results of the LaFree, Yang, et al. (2009) findings using GBTA. 8 The exact polynomial orders of each trajectory in the final solution are as following Sporadic (quadratic), twenty-first Century (cubic), 1970s Onset (quartic), 1990s Onset (cubic), and 1980s Onset (cubic).

L. Dugan and S.-M. Yang

124

data to a two-trajectory group intercept-only model. We then fit the data to three trajectories and compared its fit with the two-group solution using criteria described below. When the three-group model proved better than the two-group model, we then estimated the four-group model and compared it to the three-group solution. We continued adding groups, one at a time, each time finding an improved statistical fit until we arrived at six groups. The same process was repeated for quadratic models and cubic models. When we went beyond six groups, the model became unstable and parameters lost statistical significance; thus, we decided that extracting more groups would not provide any additional useful information. During this process, we rely on a specific set of criteria to determine the best fitting trajectory model and to evaluate its adequacy. Analysts have a range of diagnostic tools to help select the optimal model including the Bayesian Information Criterion (BIC), the posterior probability, and the odds of correct classification (OCC). First, trajectory results can be evaluated using the BIC to determine the optimal number of groups in an analysis. BIC = log(L ) - 0.5* log(n)* (k ).

(6.1)

In (6.1), “L” represents the model’s maximized likelihood estimates, “n” is the sample size, and “k” is the number of parameters estimated in the given model. Because more complex models will generally improve the fit of a given analysis, the BIC encourages a parsimonious solution by penalizing models that increase the number of groups or the model’s complexity unless they substantially improve fit. However, focusing only on the absolute BIC values could be shortsighted because it raises the risk that the model selection process will become too mechanical at the expense of thoughtful theoretical consideration and lead to proper attention not being paid toward the substantive patterns in the data. The relative difference of BIC values between two models is only one key to determine whether the new model is better than the previous specification. The actual decision-making process regarding the final number of groups should include an additional set of considerations, such as the relevant theories that underlie expected trajectory groups, the posterior probabilities of group assignment, the OCC, the estimated group probabilities, and whether meaningful groups are revealed. The equation of the computation of OCC represents the accuracy of trajectory group assignment relative to random group assignment. Its function is shown below in (6.2). The notation “AvePPj” denotes for average posterior probability of trajectory “j” and “ pˆ j” represents the group proportion of each trajectory relative to population. In (6.2), the numerator represents the OCC into group j based on the maximum probability classification rule while the denominator represents the OCC into group j based on group proportion (Nagin, 2005; Sweeten, 2006). AvePP j 1- AvePP j OCC j = . pˆ j 1-pˆ j

(6.2)

6 Introducing Group-Based Trajectory Analysis and Series Hazard Modeling… Table 6.2 Diagnostic statistics Number of Traj. group organizations Sporadic 18 1970s onset 13 1980s boom 10 1990s boom 5 Twenty-first 7 century boom

% of total organizations 34.1 24.4 18.9 9.6 13.0

Avg. posterior Prob. 0.998 0.995 1.000 1.000 0.982

125

Odds correct classification 964.34 616.57 +¥ +¥ 365.10

After reviewing the BIC and considering the patterns observed in each solution for the 53 terrorist organizations, we determined that a five-group model was the optimal model to best represent the terrorist attacks for the 53 terrorist organizations over 35 years (BIC = −4999.00). The diagnostics statistics of the final model are presented in Table 6.2. From the table, we can see the minimum average within-group posterior probability in the model is 0.982, and the lowest value of the odds of correction classification (OCC) is 365.10. Nagin (2005) suggests that when average posterior probability is higher than 0.7 and OCC values are higher than 5, the group assignment represents a high level of accuracy. Judging by these standards, the five-group model performs satisfactorily in classifying the 53 terrorist organizations into separate trajectories.9 The second and very appealing strength of GBTA is it visually presents the different developmental patterns of terrorist organizations in a very straightforward and illustrative manner. The average number of attacks for the organizations falling into each trajectory group is shown in Fig. 6.2, demonstrating the value of GBTA’s visualization tools.10 From the figure, we see five different trajectories, each following somewhat different pathways – one low-rate group in the bottom while the other four groups represent separate and sequential waves patterns. We labeled the trajectories based on their activity patterns: Sporadic, 1970s Onset, 1980s Boom, 1990s Boom, and twenty-first Century Boom. The detailed list of terrorist organizations classified in each trajectory can be found in Table 6.3.

9

The process of executing GBTA is computationally intensive and could take several iterations before reaching the final convergence. Thus, start values can be added to the program to help Proc Traj identify the solution more efficiently. Additionally, specifying start values can help avoid finding solution with local maximum likelihood rather than global maximum likelihood value. Due to the simplicity of this example, we did not specify start values in our program. For readers who wish to understand more about start values and examples in programming, please refer to http://www. andrew.cmu.edu/user/bjones/example.html. 10 The Proc Traj program also produces a set of predicted number of attacks based on the estimated parameters for each trajectory group. Others have found it useful to produce both the projected trajectories and the average number of events to represent each trajectory group.

L. Dugan and S.-M. Yang

126 140 Sporadic (34.1%) 21st Century Boom (13%)

Average Number of Attacks

120

1970s onset (24.4%)

100 80 60 40 20

1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004

0

Average Number of Attacks

140 120 100 80 60 40 20

Sporadic (34.1%) 1990s Boom (9.6%)

21st Century Boom (13%) 1980s Boom (18.9%)

2004

2002

2000

1998

1996

1994

1992

1990

1988

1986

1984

1982

1980

1978

1976

1974

1972

1970

0

1970s onset (24.4%) Projection of Sporadic

Fig. 6.2 Attack trajectories of the 53 terrorist organizations

The Sporadic trajectory includes 18 terrorist organizations (34.1% of total sample) that tend to have sporadic and infrequent attacks during the study period. Terrorist organizations classified in this trajectory generally have fewer than ten incidents throughout 35 years, except for Eritrean Liberation Front (19 attacks), Jaish-e-Mohammad

6 Introducing Group-Based Trajectory Analysis and Series Hazard Modeling…

127

Table 6.3 Terrorist organizations by trajectory assignment Trajectory group Terrorist organization Sporadic (n = 18) Al Faran/Harkat-ul Mujahidin (HuM) Al Qaeda in the Arabian Penninsula (AQAP) Al Qaeda in Mesopotamia Ansar al-Islam Central American Revolutionary Workers Party (PRTC) Eritrean Liberation Front Islamic Movement of Uzbekistan (IMU) Jaish-e-Mohammad Japanese Red Army (JRA) Jemaah Islamiya (JI) Lashkar I Jhangvi Lebanese Armed Revolutionary Faction (LARF) Lebanese Socialist Revolutionary Organization Marxist-Leninist Armed Propaganda Unit Palestine Liberation Front (PLF) Popular Front for the Liberation of Palestine, Gen Cmd (PFLP-GC) Popular Revolutionary Vanguard (VPR) Red Army for the Liberation of Catalonia Twenty-first century boom (n = 7)

1970s onset (n = 13)

Twenty-first century boom (n = 9)

Abu Sayyaf Group (ASG) Al Qaeda Lashkar-e Taiba Moro Islamic Liberation Front (MILF) Nestor Paz Zamora Commission (CNPZ) Patriotic Morazanista Front (FPM) Taliban Abu Nidal Organization (ANO) Black September Organization Ejercito Revolucionaria del Pueblo (ERP) (Argentina) Montoneros (Argentina) Moro National Liberation Front (MNLF) Mujahideen-I-Khalq (MK) November 17 Revolutionary Organization (N17RO) Popular Front for the Liberation of Palestine (PFLP) Rebel Armed Forces of Guatemala (FAR) Red Army Faction (RAF) Red Brigades Fighting Communist Party (BR-PCC) and Fighting Communist Union (BR-UCC) Revolutionary People’s Struggle (ELA) Turkish People’s Liberation Army al-Gama’at al-Islamiyya (IG) Dev Sol Hizballah Popular Liberation Army (EPL) Tupamaros (Uruguay) al-Gama’at al-Islamiyya (IG) (continued)

L. Dugan and S.-M. Yang

128 Table 6.3 (continued) Trajectory group 1980s boom (n = 10)

Terrorist organization Farabundo Marti National Liberation Front (FMLN) M-19 (Movement of April 19) Manuel Rodriguez Patriotic Front (FPMR) National Liberation Army of Colombia (ELN) New People’s Army (NPA) People’s Liberation Forces (FPL) Red Brigades Revolutionary Armed Forces of Colombia (FARC) Sendero Luminoso (SL) Tupac Amaru Revolutionary Movement (MRTA)

Table 6.4 Characteristics of attacks of the 53 terrorist organizations Number of Number of fatal Traj. group attacks (%) attacks (%) Sporadic 121 (0.7%) 67 (1.0%) 1970s onset 1,123 (6.6%) 362 (5.6%) 1980s boom 13,985 (82.7%) 5,209 (80.8%) 1990s boom 1,084 (6.4%) 473 (7.3%) Twenty-first century boom 603 (3.6%) 336 (5.2%) Total 16,916 (100%) 6,447(100%)

Number of fatalities (%) 761 (1.8%) 1,922 (4.6%) 31,847 (75.7%) 1,756 (4.2%) 5,770 (13.7%) 42,056 (100%)

(14 attacks), and Lashkar I Jhangvi (14 attacks). The 1970s Onset trajectory represents 13 terrorist organizations (24.4% of the total sample). This trajectory reached peak in the early 70s, gradually declined since the 80s and then remained low in frequency afterwards. The third trajectory group, 1980s Boom, includes only ten terrorist organizations (18.9%), but accounts for the great majority of terrorist attacks (13,985 attacks, i.e., 82.7% of total attacks). The 1980s Boom trajectory was active from the late 1970s and remained active until the early 1990s. Because of its overwhelmingly high frequency of attacks (on average almost 1,400 attacks per organization), it overshadowed the peak of the 1970s Onset trajectory in 1976. Similarly, the 1980s Boom dwarfed the peaks for the 1990s Boom and the twentyfirst Century Boom. The statistics presented in Table 6.4 further demonstrate the “danger” of the organizations classified in the 1980s Boom trajectory. Together, they account for 13,985 attacks (82.7% of total attacks), resulting in 5,209 fatal attacks (80.8% of total fatal attacks), and 31,847 deaths (75.7% of total fatalities). Table 6.4 clearly shows that the terrorist organizations identified in the 1980s Boom trajectory are the “hot spots” of the 53 organizations. Thus, it would be interesting to closely examine the cases belong to this trajectory to see if Rapoport’s predictions hold true. Furthermore, we can also seek to understand the factors that lead to increases and decreases in attacks perpetrated by the organizations in this group. The 1980s Boom trajectory includes organizations such as Farabundo Marti National Liberation Front (FMLN), M-19 (Movement of April 19), Manuel Rodriguez Patriotic Front (FPMR), National Liberation Army of Colombia (ELN),

6 Introducing Group-Based Trajectory Analysis and Series Hazard Modeling…

129

New People’s Army (NPA), People’s Liberation Forces (FPL), Red Brigades, FARC, SL, and Tupac Amaru Revolutionary Movement (MRTA). Most of the organizations identified in this trajectory are either Latin American revolutionary organizations, leftist (NPA and Red Brigades), or both. The characteristics of the 1980s Boom seem to be consistent with Rapoport’s New Left Wave. With the limited information we have for each group and the available time span of the data, we cannot do a full test of Rapoport’s theory. Nonetheless, the findings from trajectory analysis provide us with the first stepping stone to a more holistic understanding of terrorism trends over centuries. Out of all the organizations classified in the 1980s Boom, we turn our attention in the second part of the chapter toward one of the most active of these organizations, SL, to demonstrate how series hazard modeling can reveal insights into the rises and falls of its trajectory. The flexibility of GBTA extends well beyond the example here. First, it can account for both time-varying and time-stable covariates and use the covariates to help predict the trajectory assignment. In another study, Yang and LaFree (2011) examine terrorism trends in 189 provinces across Thailand, Indonesia, and the Philippines. Because political stability was believed to be an important risk factor for terrorism, they included a “fail-state” variable designating whether the providence is located in a country that is in failure. This is a time-varying covariate in the trajectory model. The results confirm other findings that political instability is associated with statistically significant increases in terrorist attacks for all trajectory groups. Moreover, they also found that the effect of instability is greatest for provinces with a chronic terrorist threat. Additionally, GBTA can also use time-stable covariates to help estimate trajectory groups. Some static features such as population characteristics, geographic features, or ideologies of terrorist groups could impact the number of terrorist attacks. The examination and interpretation of the effects of these risk factors is the same as of a conventional regression, except the effects are now examined separately for each trajectory group. Including covariates can be especially important when there are high dependencies between covariates and the dependent variable. As Nagin (2005) states, trajectory models without covariates estimate “…the prototypical development path of trajectory group members, averaged over all contingencies…” (page 124). After adding important covariates into the specification, the shape of trajectories can vary and the interpretation of the parameters also changes depending upon the effects of covariates on the dependent variable. By omitting important risk factors, the model is vulnerable to omitted variable bias, just as in conventional regression. Thus, it is important to incorporate key risk factors into the model whenever possible to perfect the model specification. Furthermore, GBTA can track two or more characteristics simultaneously to help us understand the linkage between related trends (see Nagin & Tremblay, 2001). Joint Trajectory Analysis (JTA) is an extension of trajectory analysis designed to account for the comorbidity of two distinct but theoretically connected developmental courses. For example, Nagin and Tremblay examined the link between conduct disorder and hyperactivity and see if higher rates of one increase the occurrence of another (Nagin & Tremblay, 2001:18). Recently, Yang (2010) used JTA to explore

130

L. Dugan and S.-M. Yang

the link between longitudinal trends in disorder and violence and to determine whether the hypothesized connection proposed by the broken windows thesis (Wilson & Kelling, 1982) holds true in the data. Dugan et al. (2011) use JTA to jointly model the lethality and frequency of attacks for all terrorist organizations that were active for at least 365 days and began attacking prior to 1999. JTA allows terrorism scholars to explore two different time trends and examine their corresponding strength without making any assumptions about the underlying mechanism linking the two dependent variables. It is worthwhile to note that unlike the co-integration modeling used in economics, JTA does not require the 2 time trends to have overlapping durations. Finally, GBTA is particularly useful for scholars who simply want to explore the data without having any strong theoretical basis (Muthén, 2001; Nagin, 1999, 2005; Raudenbush, 2001). Terrorism is a complex phenomenon; thus, GBTA provides a tool to describe both continuity and change of longitudinal patterns without oversimplifying the trends (Jung & Wickrama, 2008). In other words, trajectory analysis can be used to extract basic patterns existing in the population (Skardhamar, 2010), perhaps to guide theory development and to motivate research hypotheses that can be tested using more conventional methods. In sum, GBTA can be used to identify cases that follow distinct developmental patterns. As such, because the different trajectories have different parameters, their shapes are driven by the data rather than imposed upon by the researchers. Furthermore, this method can be applied to a number of different units of analysis. For example, it can be used to estimate long-term activity patterns of terrorist organization or trends of terrorist threats in countries (or other geographic units). Other available static or dynamic covariates can also be included in the model to help improve model accuracy. Of course, as demonstrated in this chapter, the results of GBTA are valuable even without covariates. This feature is important because it is often extremely difficult to collect reliable longitudinal data on the characteristics of organizations (LaFree et al., 2011). Thus, GBTA can be a powerful tool for scholars who want to disentangle the trends as the first step of research. Its extension, JTA, also provides a unique tool to model two distinct but related trends simultaneously. Despite the benefits of GBTA, like other methods, GBTA has some weaknesses that researchers should consider before using. First and the most widely cited issue of GBTA is whether the groups extracted represent “reality” or simply an approximation of an underlying continuous distribution? While GBTA is useful in reducing complex phenomenon into a more parsimonious solution, some have argued that the trajectory groups obtained in the analysis are misleading if used to support evidence for crime taxonomy (Laub, 2006, 2009; Moffitt, 2006; Skardhamar, 2010). The key debate is centered on the issue of whether the trajectory groups are “real” groups existing in the population. While the group-based trajectory model allows us to place terrorist organizations into groups with similar patterns, analysts must remember that these categories are just ideal types rather than the actual trends of any specific terrorist organizations. Conducting trajectory analysis can be a complicated process and caution should be taken at each step. In addition to the BIC, trajectory analysis requires researchers

6 Introducing Group-Based Trajectory Analysis and Series Hazard Modeling…

131

to also consider posterior probabilities of group assignment, OCC, estimated group probabilities, and whether meaningful groups are revealed before deciding upon a final model (Nagin, 2005). Group assignments in trajectory analysis can potentially be made with error (see Nagin, 2005; Roeder et al., 1999), and as such, they should be used as an approximation of a complex underlying reality or to reveal population heterogeneity – not as a theory testing tool (Nagin, 2005:173). The most reasonable use of GBTA in terrorism research is to reveal population heterogeneity rather than theory testing. Another caution that was raised by Eggleston, Laub, and Sampson (2004) was that the trajectory classification could change with more data being collected over time. That is, when time span extends or when more cases are included in the analysis, the number of trajectory groups or shape of distribution might be altered (also see Skardhamar, 2010). While a valid concern, Nagin (2005) points out that this very same concern would apply to all other statistical approaches. When more valid information is included in any analysis, the results will almost always improve. The final disadvantage of GBTA is the restriction placed on the within group parameters. GBTA can be viewed as a more restrictive version of Generalized Growth Mixture Modeling (GGMM) (Bushway et al., 2003; Jung & Wickrama, 2008; Morris & Slocum, in progress; Muthén, 2001). Both GGMM and GBTA model changes within individual cases and classify them into several latent classes (or trajectory groups). However, GGMM allows the individuals within the groups to have different intercepts and slopes, while GBTA restricts the individuals within each group to have the same parameter estimates. Although GGMM is more flexible, it is also much more computationally intense than GBTA, and its findings are more complex. Morris and Slocum (in progress) compare results from both GBTA and GGMM and concluded that each has its unique strengths and weaknesses when applied to the study of terrorism. Nonetheless, GBTA is more robust in identifying terrorism hot spots than GGMM (Morris & Slocum, in progress). Readers should consider the utility of each method and make choices based on their research interest. Recall that we identified five trajectory groups with different activity patterns. The most interesting group is the 1980s Boom trajectory due to both its large magnitudes and persistent impacts. After identifying groups that belong to this chronic trajectory, it might be of interest to scholars to further examine the nuanced changes in activity of a particular group in response to an intervention or other government effort to control the organization. The following section introduces another method designed to serve that purpose. We use it to examine, as an example, the changing longitudinal trend of SL.

Series Hazard Modeling The series hazard model is used with event-based data, like the GTD, that records the exact date of each terrorist attack or other relevant event. Thus, in contrast to GBTA, which tallies the number of attacks in a given year (or other time intervals),

132

L. Dugan and S.-M. Yang

series hazard modeling uses information from every attack to model risk of continued attacks. More specifically, is an extension of the Cox proportional hazard model that estimates the changes in risk for subsequent events conditioned upon characteristics of past events and other event-specific or date-specific covariates – such as the implementation date for a policy or intervention (Dugan, 2011a). While an intuitive modeling strategy, very few terrorism researchers have adopted this approach, perhaps because unfamiliarity within the field until its first use in 2005. In that article, Dugan et al. (2005) used the series hazard model (then unnamed) to evaluate the success of three policy efforts to end aerial hijackings. They found, among other things, that once metal detectors were installed the hazard of hijackings dropped, especially those flights that originated in the United States. They also found that after Cuba passed a law criminalizing aerial hijacking, the risk of hijacked flights to Cuba fell. These findings are intriguing because they suggest that simply tightening screening with additional security guards had no effect on reducing the risk of continued hijacking. A second research project that relied upon series hazard modeling estimated the effects of six major initiatives by the British government to stop republican violence (LaFree, Dugan, et al., 2009). Results showed that only Operation Motorman seemed to have the intended effect of reducing the hazard of continued attacks by republican terrorist organizations. Three other efforts (internment, criminalization, and targeted assassinations in Gibraltar) instead appeared to increase terrorist activity, while the other two showed no real change. While both of these studies were mostly focused on assessing government efforts to reduce violence, a third study by Dugan, Huang, LaFree, and McCauley (2008) examined the effects of a terrorist organization’s own misstep. More specifically, they estimated the impact of one excessively damaging terrorist attack by the Armenian Secret Army for the Liberation of Armenia (ASALA) on the continued violence by two Armenian terrorist organizations, ASALA and the Justice Commandos of the Armenian Genocide (JCAG). The findings did, indeed, show that attacks by both groups subsided after the deadly attack by ASALA at Paris’ Orly Airport, suggesting that by “putting off” the Armenian constituency, both organizations within the movement suffered. All of these projects estimate changes in activity for a single unit after one or more interventions or other noteworthy events. Currently, the most common analytical strategy for this type of problem is to use some form of time series (e.g., Brandt & Williams, 2007; D’Alessio & Stolzenberg, 1995; Lewis-Beck, 1986; McDowall, Lizotte, & Wiersema, 1991). While appropriate in many contexts, Dugan (2011a) demonstrates that when the data are event-based, the series hazard model can be used as a more flexible alternative to time series. For this section of the chapter, we demonstrate the utility of the series hazard model by more thoroughly examining the behavior of SL, the Peruvian terrorist organization that was identified as a member of the 1980s Boom trajectory group in the GBTA analysis above. While founded in 1970 by a professor of philosophy, SL’s campaign of violence did not start until 1978, when it stated as its goal the destruction of all Peruvian political institutions and creating a “new state of workers and

6 Introducing Group-Based Trajectory Analysis and Series Hazard Modeling…

133

peasants” (McCormick, 2001). Its charismatic leader, college professor Abimael Guzman, led a Maoist revolt across Peru by recruiting faculty, students, and poor Indian populations from the highlands. The organization perpetrated thousands of attacks over the years, averaging two and half deaths per attack (Dugan, 2011b).

Strengths and Weaknesses of the Series Hazard Model Strengths The series hazard model offers several advantages over time series when estimating effects of potential turning points, such as the date of implementation for programs, policies, or interventions. First, because the model relies on the variation between events (rather than across tabulations), more information is used to estimate the effects of interventions or other independent variables. To demonstrate the importance of this variation, Fig. 6.3 presents the frequency of attacks by SL aggregated to different periods of time. In Fig. 6.3a, we see the annual counts of attacks from 1978 through 2007. For each of these years, the attack pattern is summarized by a single number. For example, we summarize the peak year, 1989, by reporting that SL attacked 509 times. While informative, this summary value ignores all variation of the attack patterns within 1989. In other words, a yearly time series analysis would be relying on only that value to describe 1989, implicitly assuming uniformity of attacks within that year. Figure 6.3b presents the monthly patterns of attack within 1989. Here we see that until August, attacks ranged between 30 and 60 a month. In September, attacks dropped to only 11, but then rose to their peak in October with 98 attacks. After that month of violence, attacks by SL dropped once again to just under 30 a month. The unusual drop and rise between September and October could indicate critical patterns that would be missed with annual aggregation. Taking this one step further, Fig. 6.3c presents the frequency of weekly attacks by SL in October 1989. This demonstrates that by summarizing that month with the total count of 98, we are missing another part of the story. The weekly graph shows a huge drop in attacks from week 2 at 31 attacks to 6 in week 3. Attacks pick up again in week 4 with 26. Again, this week to week variation gives the researcher more information that would allow for more precise estimation. Because the series hazard model preserves the date of each event, it is estimated using even more information than that shown in Fig. 6.3c. The clustering of attacks in early October 1989 will appear in the data as having relatively short times until the next attack, whereas those attacks in the middle of October will on average show longer times until the next attack. Furthermore, Fig. 6.3a suggests that those attacks after 1997 will have even larger periods until the next attack. Thus, because the series hazard model avoids aggregating events, it uses all available variation in the data to estimate its parameters.

L. Dugan and S.-M. Yang

134

a

Yearly Attacks by Sendero Luminoso 600

Attack Frequency

500 400 300 200 100

07

05

20

03

20

01

20

99

20

97

19

95

19

92

19

90

19

88

19

86

19

84

19

82

19

80

19

19

19

78

0

Year

b

Monthly Attacks by Sendero Luminoso in 1989 120

Attack Frequency

100

80

60

40

20

0 Jan

Feb

Mar

Apr

May

Jun

Jul

Aug

Sep

Oct

Nov

Dec

Month in 1989

c

Weekly Attacks by Sendero Luminoso in October, 1989 35

Attack Frequency

30 25 20 15 10 5 0 1

2

3

4

5

Week in October 1989

Fig. 6.3 Attack frequencies by Sendero Luminoso for years (a), months (b), and weeks (c)

6 Introducing Group-Based Trajectory Analysis and Series Hazard Modeling…

135

A second advantage of avoiding arbitrary aggregation is that different levels of aggregation may produce different findings (Freeman, 1989). Important research by Shellman (2004a) demonstrates that conclusions can dramatically change when event data are aggregated to broader units when using vector autoregression. More generally, Alt, King, and Signorino (2000) show the importance of understanding and avoiding pitfalls of aggregation bias (see also Shellman, 2004b). Related to these strengths found in modeling with disaggregated units, a third strength is that the temporal ordering of the data used in the series hazard model can be exact to the day. As Dugan (2011a) shows, because the exact day of an intervention of interest can fall anywhere within an aggregated unit, time series estimates might actually include the reverse effect of the events on the adoption of the intervention, leading to bias. While many avoid this bias by lagging the intervention by 1 time unit, this raises the risk of missing any short term-effects from the intervention biasing the estimates toward zero. To demonstrate this issue, let us consider the day that the Peruvian government captured SL’s leader, Guzman, September 12, 1992. Had we used yearly data to estimate the effects of Guzman’s capture on the activity of Sendero Lumino, the estimates would include any effects of terrorist attacks leading to the capture of Guzman. In fact, more than half of the attacks (179 out of 286) perpetrated by the group in 1992 were before September, which surely indicates a high risk of simultaneity bias for nonlagged estimations. Yet, lagging the capture of Guzman by 1 year would essentially exclude from the analysis any change in activity between September 12 and December 31st of 1992, which could be the months with the most direct impact. Though this is less of a problem with smaller temporal units, as Dugan (2011a) points out, researcher must always be aware of the precise location of the intervention within the time interval in order to properly evaluate the risks of simultaneity bias and diluted effects. Because the series hazard model incorporates the exact dates of each event as well as any intervention, the temporal ordering is persevered, and both problems are avoided. The fourth strength is that dependency between events can be directly modeled.11 Oftentimes, theoretical reasoning would suggest that the specific characteristics of one event would increase or decrease the hazard of another event. This contagion has been described by Holden (1986) as a reason why successful aerial hijackings would lead to more similar events. Holden, and later, Dugan et al. (2005), tested this contagion hypothesis directly and indeed found that successful hijackings seem to lead to additional attempted hijacking events. Holden (1986) tested the contagion hypothesis directly by using a linear self-exciting model to estimate daily counts of hijacking attempts between 1968 and 1972. The self-exciting model estimates the cumulative effects of characteristics in all previous hijacking events on the intensity of future hijacking.12 Instead of using frequency of daily hijackings, Dugan et al. 11

In fact, in order to avoid bias in the standard errors, dependency between events should be modeled to establish conditional independence (Dugan, 2011a). 12 The self-excitement model is designed to weigh more recent events heavier than earlier events using an exponential function that eventually converges to zero for the earliest events, incorporating the assumption that “every incident tends to be forgotten eventually” (Holden, 1986: 888).

136

L. Dugan and S.-M. Yang

(2005) used the series hazard model to estimate the effects of contagion on the hazard of additional hijackings. They operationalized contagion with a measure of “success density” that incorporated the temporal distance between the current and two previous hijackings and their proportion of success. While this measure ignores all events prior to the two preceding events, the series hazard model can easily incorporate information on as many previous events as theoretically relevant. In fact, in the model presented in this chapter, we operationalize success density by incorporating the seven most recent events. Because the first N events will be deleted from the analysis, seven was chosen as a balance between including additional information, while preserving much of the earliest events. The first seven attacks by SL are missing values for success density and are therefore excluded from the model. Dependency can be operationalized in several ways, such as the cumulative number of events to date or the amount of time that has passed since the first. The key is that the researcher be guided by theory in order to construct the most reasonable measure. Other indirect measures of dependency can be incorporated if theory suggests that specific characteristics of one event will increase or decrease the hazard of an additional event. For example, findings by Dugan, Huang, et al. (2008), suggested that severely violent attacks could sabotage the future of a terrorist organization. If true, it might be especially important to include a measure of lethality in the series hazard model. In other cases, highly fatal attacks could increase the momentum of continued attacks, raising the hazard of the next event. Regardless of the different predictions, these examples demonstrate the added value of including event-specific measures. A fifth, and possibly the most appealing, strength is that the series hazard model is simple to estimate because the researcher can simply use the Cox proportional hazard model command from any software package on events rather than units, by setting the time until the next event as the dependent variable. This leads directly to the final strength. Statistical power is directly related to the number of events rather than the number of units (as in traditional hazard modeling) or time periods (as in time series). For example, instead of tying the degrees of freedom to the length and precision of the series (e.g., 7 years, 28 quarters, or 84 months), the degrees of freedom would represent the additional information gained from including each event as its own unit in the model (e.g., 2,000 terrorist attacks). While this works nicely for events that occur relatively frequently, the series hazard model will not function well for relatively rare events. This leads to the first weakness. Weaknesses The events must occur with relative frequency. At the most basic level, because the event is the unit of analysis, rare events will unlikely produce enough statistical power to efficiently estimate parameters. For example, if the SL analysis were restricted to only those years beyond 1997, we would only have 19 observations from which to accrue statistical power. Furthermore, low statistical power is only one limitation of applying the series hazard model to rare events. Because changes in temporal covariates are only measured on the date of the event, rare events could

6 Introducing Group-Based Trajectory Analysis and Series Hazard Modeling…

137

reduce statistical variation making it difficult to detect any effect. In an extreme example, if the intervention successfully reduced events to zero, the intervention dummy variable would simply be a vector of zeros because no events occurred after the intervention. Having said this, if the events become less frequent after the intervention, time-varying covariates that capture the change in the intervention condition could be incorporated into the model, as done with conventional hazard modeling. While this can increase statistical power and document successful interventions, it will also produce missing values for the incident-specific variables since no incident is associated with the additional observations. Depending on the severity of the problem and the importance of the event-specific covariates, one might be able to cleverly interpolate missing values in order to optimize the efficiency of the model while reducing bias. To date, nobody has attempted this adjustment. High frequency events could also produce a problem with the series hazard model if multiple events regularly occur on the same day. For example we would run into this problem if we wanted to model the risk of alcohol related deaths in the United States. Because there are dozens of these types of deaths on a given day, and because the dependent variable is typically the number of days until the next event, the dependent variable for data like these would appear dichotomous, where the 1 value would simply mark the last event ordered for a specific day. Given that daily event ordering is often unsystematic, this value is likely to be meaningless. One strategy to overcome this limitation would be to rescale the dependent variable so it measures the number of days for every 10 or every 100 events, in order to create meaningful variation in the dependent variable. While nobody has done this to date, it seems like a reasonable solution to this issue. For example, in the alcohol related deaths example, we could use as the dependent variable, the number of days until the 100th death. The next two limitations are really just issues that need to be addressed. The first is an issue with any hazard model, and the second is unique to the series hazard model. When two or more observations have the same value for the dependent variable, they are referred to as tied data (Allison, 1995). For example, a simple frequency of the dependent variable in the SL data shows that 18% of the attacks, had a second attack the following day (nextinc = 1). Because each of these duplicate observations (or tied data) will have its distinct set of covariate values, these differences must be resolved prior to estimation. Most statistical packages can easily resolve tied data in the traditional hazard model by either accounting for all possible orderings (exact marginal) or by adopting a multinomial strategy to estimate the likelihood function (exact partial) (Allison, 1995). Regardless, these strategies also work well with the series hazard model. A specific tie that is handled differently is when two or more events occur on the exact same day. For these cases, n−1 of them would be coded as 0 and the last event listed would be coded as the number of days until the next event. For example, if SL had seven attacks on the same day in different locations, the first six events would have a zero as the dependent variable, and the seventh would have an integer that measures the number of days until the next event. If multiple events typically occur daily, then the researchers could simply rescale the dependent variable by modeling the number of days until the 10th or 100th event, as mentioned earlier. However, if

138

L. Dugan and S.-M. Yang

multiple events are atypical, meaning that there is sufficient variation in the number of between events, then the data will need to be recoded in order to avoid problems. Most statistical packages will delete case when the dependent variable to the hazard model is set to zero. Thus, in our SL example, the first six events of the day would be excluded from the estimation, undoubtedly biasing the results. Furthermore, removing cases reduces important information and statistical power that could be used to better estimate the model. Information and power need not be lost. In order to thoughtfully recode the data, researcher should consider the context of the multiple events. For example, because SL is a single organization, it is likely that multiple attacks on a single day are part of a coordinated effort. Another scenario is that separate cells operating under purview of SL acted independently on the same day. In either of these contexts, the best measure of time until the next attack would be the days beyond the current day until the next event. Thus, for our SL example all attacks where the dependent variable equals zero were recoded to have the same value as the last event listed on their days. Prior to this recoding, more than 60% of SL’s attacks had at least one other attack perpetrated on the same day. Not surprisingly, the years with the most multiple event days correspond to the years with the most attacks (1983, 1984, and 1989).

Using the Series Hazard Model to Study Sendero Luminoso The goal for the SL analysis is to estimate the effects of one major intervention, the capture of leader Guzman on September 12, 1992. Data on SL activities are drawn from the GTD (http://www.start.umd.edu/gtd/). While this is the best source of terrorist event data available, it does present one very relevant problem. Cases from 1993 were lost. However, it is not difficult to incorporate missing data into the series hazard model because just as with the Cox proportional hazard model, the series hazard model can easily accommodate censoring. In other word, because we do not know the timing of the earliest events in 1993, the last event in 1992 is censored. In this case, the last three attacks in 1992 were perpetrated on December 29th and were simply marked as censored. If the arrest of Guzman did, indeed, lead to a reduction in attacks by SL, then we would expect that the distribution of the number of days until the next attack to differ before and after the arrest. Figure 6.4 presents these distributions, with the white bars representing the distribution of days for attacks prior to Guzman’s arrest and the black bars representing the distribution after the arrest. Most apparent is that the modal number of days for both the pre and postarrest period is one. However, prior to Guzman’s attack, more often than not (52.0%), attacks were perpetrated on sequential days. This pattern drops by more than half (25.7%) after the arrest. Similarly, we see that prior to the arrest the distribution drops quickly toward 0% after several days. While there are a few long period of time between attacks prior to the arrest, 95% of all attacks occurred within 6 days of the last attack. After Guzman’s arrest, only 53% of the attacks occurred within 6 days of the last attack. Finally, we see in Fig. 6.4 that the black bars extend over the entire distribution. In

6 Introducing Group-Based Trajectory Analysis and Series Hazard Modeling…

139

60%

50%

Pre-Arrest Post-Arrest 40%

30%

20%

10%

0% 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57

Number of Days until the Next Attack

Fig. 6.4 Distribution of the number of days until the next attack before and after Guzman was arrested

fact, we only reach the 95th percentile after 96 days. In other words, 95% of all attacks occurred within 96 days of the last attack (compared to only 6 days prior to Guzman’s arrest). Together, these distributions strongly suggest that something changed after Guzman’s arrest. We can test this hypothesis more formally with the series hazard model. As mentioned above, one of the more appealing properties of the series hazard model is that it can be run using the same commands as the Cox proportional hazard model. The only difference is that instead of each observation measuring a different unit, it measures a different event. To be explicit, the hazard function for the Cox model takes the following form, li (t | X i ) = l0 (t )exp(X i b),

(6.3)

where Xi is a vector of covariate values for unit i, b is a vector of unknown parameters for X and l0 (t ) is an unspecified baseline hazard function for all units. In other words, it is the hazard for an individual whose value of X = 0. Equation (6.4) shows the hazard function of the series hazard model, le (t | X e ) = l0 (t )exp(X eb).

(6.4)

Note that the only difference between (6.3) and (6.4) is that the subscript now refers to the number of events (e), instead of units (i). Dugan (2011a) expands this function

L. Dugan and S.-M. Yang

140

to explicitly include covariates that model the dependency across events in order to avoid bias in the standard errors by establishing conditional independence. le (t | X e ) = l0 (t )exp(X eb + Z eg ).

(6.5)

In other words the specification of all series hazard models should include some measure of dependency across observations, which is shown in (6.5) as Ze. In practice, we first prepare the data so that (1) the dependent variable measures the number of days until the next event, (2) events on the same day are recoded to match the value for the last event on that day, (3) the date of the intervention is made into an independent variable, and (4) other key independent variables are created. All of these processes will be shown as do file code from Stata. The coding strategy can easily be translated into other packages. The first two of these steps are shown in the code below:

First the data are sorted by the date variable, which is called idate in the GTD. Because the Var[_n + 1] feature points to the value of the variable on the following line, nextinc is generated by taking the difference between the date of the next incident (idate[_n + 1]) and the current incident (idate). Thus, nextinc is the dependent variable and it measures the number of days until the next event. The next step is to look for multiple attacks on the same day. If the frequency distribution of nextinc includes any zeros, the next set of code should be run until there are no more zeros in the frequency distribution. That code simply takes those cases where nextinc = 0 and replaces them with the value on the following line. In order to operationalize the key independent variable we created dummy variables for each event by turning the value on at the date of the event and extending it to the end of the series.

The code shows how Stata reads dates as it creates the dummy variable guzarst. In this case it is entirely appropriate to code the values at 1 through to the end of the series because Guzman remained in jail throughout that period. However, some interventions might be more short-term and should be coded as such. For example,

6 Introducing Group-Based Trajectory Analysis and Series Hazard Modeling…

141

in LaFree, Dugan, et al. (2009), several of the interventions were 1-day initiatives meant to deter terrorist attacks. Because their effect was not expected to be longlasting, the authors set the variables back to zero again after a year. They later ran sensitivity tests to determine if the substantive findings were dependent on the length of the intervention but did not find that to be the case. Because it might be naïve to assume a long-term constant intervention effect throughout the end of the series, interactions can be created with some sort of temporal variable. The following code creates a variable that counts the number of months since the beginning of 1978 for each event that marks how deep into the series the event is located (num_month). It then creates an interaction variable (garst_m) with the monthly count and the intervention.13

The first series of code creates a program that calculated the count of the number of months into the series starting with January 1978 (1977 + i). Because this value is only calculated on the dates of attacks, some months are not included and others can be included more than once. For example, the first five attacks by SL occurred in months 8, 13, 31, 31, and 32. Once the number of months since the beginning of 1978 was calculated, then that variable is interacted with the intervention dummy variable in order to test for a temporal trend in the intervention effect. Finally, we created two variables that capture the dependency across observations. The first, lastinc measures the number of days since the last attack. As with nextinc, we replaced this value for all attacks that occurred on the same day with the number of days excluding the current day. See the code below.

13

While the month was used in this example, the program can be easily modified to use a more refined unit, such as day.

L. Dugan and S.-M. Yang

142

The measure of success density (succ7) was created using the following code.

In order to calculate the final formula, P(success for the current and six previous attempts) (event date current - event date 6th previous ) / 365

(6.6)

several intermediary variables needed to be created. First, asucc7 measures the proportion of the current and six previous attacks that were successful.14 Note that this is the numerator for (6.6). The notation days7 is the number of days between the current event and the sixth previous event. This variable is divided by 365 to get the denominator for (6.6), which is the number of years between the current and the sixth previous attack (years7). Thus, when the final variable (succ7) is at its largest, the numerator is likely close to 1 and the denominator is small, suggesting that the seven most recent attacks were relatively recent and successful. When the success density is at its smallest, very few of the seven most recent attacks were successful and/or they occurred over a relatively long period of time. The values for success density range from 0.187 to 365, where 365 represents the case where the seven most recent attacks were all successful and they all occurred within the same day. The last coding step is to create the censored variable, which we traditionally name dead. This is a dummy variable, and is coded as 1 if the next attack occurred before the observation period ended, and zero if it is censored. Had 1993 not been missing from the GTD, there would only be one censored value at the end of the series. Because 1993 is missing, we also mark the attack(s) on the last day in 1992 as censored. By looking at the frequency distribution of dates, we see that three attacks were perpetrated on December 29, 1992. We use the following code to create the censoring variable.

14

The original coders of the GTD define success by the details of the terrorist event. If a bomb exploded, then it was considered successful, even if the larger intent of the organization was not achieved (Dugan et al., 2005).

6 Introducing Group-Based Trajectory Analysis and Series Hazard Modeling…

143

As with the traditional Cox proportional hazard model, we need to tell Stata that the data are survival-time data. We use the following command.

The command tells Stata that the dependent variable is nextinc and the censoring variable is dead. The output from this command tells us that there are 4,513 observations with 4,509 failures, or known times until the next attack. It also tells us that the relevant interval in the analysis is (0, nextinc). As mentioned above, once the data are constructed, the model is run with the same command as the Cox proportional hazard model.

Note that we simply list the independent variables because the dependent and censoring variable were reported in the stset command. Also, to be clear, all independent variables are measured at the current event, in order to estimate the hazard using the number of days until the next event as the dependent variable. Thus, the appropriate temporal ordering is imposed. We use the subcommand exactm to tell Stata to use the exact marginal strategy to resolve the tied data, which accounts for all possible orderings. The default in Stata is to use the breslow method, which approximates a multinomial strategy to resolve the ties under the assumption that an ordering does exist (Allison, 1995). Because these data have a large number of ties, it is unrealistic to assume that there is a precise ordering amongst them. The nohr subcommand is used so that the coefficient estimates are presented instead of the hazard ratios. Because we include an interaction between the monthly count and the intervention (Guzman’s arrest), the hazard ratio should be calculated from a combination of the main and interaction effects. Table 6.5 presents the results of this model. We see from this table that the only statistically significant finding related to the intervention is the interaction between Guzman’s arrest and the number of months into the series. The main effect is statistically insignificant, which has little meaning

L. Dugan and S.-M. Yang

144

Table 6.5 Coefficient estimates for the series hazard model estimating the hazard of the next attack Variables Code name Coefficient SE Intervention Guzman’s arrest guzarst 0.343 0.717 G. arrest × number of months garst_m −0.009* 0.004 Event specific Number of fatalities

nkill

Dependency measures Number of months num_month Time since last attack lastinc Success density succ7 * = p < 0.05, ** = p < 0.01, all tests are two-tailed

0.009

0.007

0.003** −0.002 0.002**

0.001 0.001 0.000

within the context of an interaction model. When we run the model with only the main effect, it produces a strong negative and statistically significant coefficient estimate. However, a likelihood ratio test concludes that the more flexible interaction model is the better choice as a final model (p = 0.0075). Because these estimates should be interpreted in conjunction with the coefficient estimate for the number of months, we will interpret them below using a graph showing trend of the marginal hazard for the number of months into the series. The event specific measure of fatalities is positive, but null. We see from the value of the number of month variable, attacks that are deeper into the series have a higher hazard rate than those earlier in the series. There appears to be no direct relationship between the time since the last attack and the time until the next attack. However, if there was a higher concentration of recent successful attacks, the hazard of another attack will increase. Figure 6.5 shows the marginal change in the hazard ratio throughout the series before and after Guzman’s arrest. Up until the arrest, the hazard ratio was increasing gradually over time. Once SL’s leader was captured, the hazard ratio dropped from 1.77 to 0.36, and continued to drop throughout the remainder of the series. This decline reflects the negative coefficient of the interaction between Guzman’s arrest and the number of months into the series.15 In summary, the series hazard model shows that the capture of SL’s leader Guzman seemed to have dramatically reduced the risk of continued attacks by that organization. This particular finding is unsurprising, giving other case studies of the organization came to the same conclusion, leading us back to where we started at the beginning of the chapter (Cronin, 2006). However, we got to this conclusion using an entirely different approach, by first examining the trends of all of the original 53 terrorist organizations that were identified as threatening to the U.S. and then honing in on one organization that fell into the most threatening trajectory group.

15

The post-arrest downward trend in this figure makes the reason that the main effect was statistically insignificant more obvious. If we extend the after-arrest line toward zero months, the hazard ratio would be very close to one when it crosses the zero month (not shown in graph).

6 Introducing Group-Based Trajectory Analysis and Series Hazard Modeling…

145

2 1.8

Estimated Hazard Ratio

1.6 1.4 1.2 1 0.8 0.6 0.4 0.2

8 18 28 38 48 58 68 78 88 98 108 118 128 138 148 158 168 178 188 198 208 218 228 238 248 258 268 278 288 298 308 318 328 338 348 358

0

Month

Fig. 6.5 The marginal monthly hazard ratio before and after Guzman was arrested

This chapter was meant to demonstrate how both methods can add value to the growing body of literature that investigates terrorist behavior, regardless of the unit of analysis. The value of GBTA is that it allows the researcher to identify patterns across the larger context, and the value of series hazard modeling is that it uses all of the information found in the timing between events to identify factors related to increases and decreases in the risk of continued attacks. Together, these methods provide a more systematic way to understand terrorism and its impact that can be easily replicated by other researchers who follow the same procedures. They also help set the context for empirically testing hypotheses that might have been generated by the case studies.

References Allison, P. D. (1995). Survival analysis using the SAS system: A practical guide. Cary: SAS Institute Inc. Alt, J. E., King, G., & Signorino, C. S. (2000). Aggregation among binary, count, and duration models: Estimating the same quantities from different levels of data. Political Analysis, 9(1), 21–44. Bennett, A., & Elman, C. (2007). Case study methods in the international relations subfield. Comparative Political Studies, 40(2), 170–195. Blokland, A. A. J., Nagin, D. S., & Nieuwbeerta, P. (2005). Life span offending trajectories of a Dutch conviction cohort. Criminology, 43, 919–954.

146

L. Dugan and S.-M. Yang

Brandt, P. T., & Williams, J. T. (2007). Multiple time series models. Thousand Oaks: Sage Publications, Inc. Bushway, S. D., Sweeten, G., & Nieuwbeerta, P. (2009). Measuring long term individual trajectories of offending using multiple methods. Journal of Quantitative Criminology, 25(3), 259–286. Bushway, S. D., Thornberry, T. P., & Krohn, M. D. (2003). Desistance as a developmental process: A comparison of static and dynamic approaches. Journal of Quantitative Criminology, 19(2), 129–153. Callaway, R., & Harrelson-Stephens, J. (2006). Toward a theory of terrorism: Human security as a determinant of terrorism. Studies in Conflict and Terrorism, 29(7), 679–702. Cordes, B. (2001). When terrorist do the talking: Reflections on terrorist literature. In D. C. Rapoport (Ed.), Inside terrorism organizations (pp. 150–171). London: Frank Cass Publishers. Crenshaw, M. (2001). Theories of terrorism: Instrumental and organizational approaches. In D. C. Rapoport (Ed.), Inside terrorism organizations (pp. 13–29). London: Frank Cass Publishers. Cronin, A. K. (2006). How al-Qaida ends: The decline and demise of terrorist groups. International Security, 31(1), 7–48. D’Alessio, S. J., & Stolzenberg, L. (1995). The impact of sentencing guidelines on jail incarceration in Minnesota. Criminology, 33(2), 283–302. Dugan, L. (2011a). The series hazard model: An alternative to time series for event data. Journal of Quantitative Criminology. Dugan, L. (2011b). The making of the Global Terrorism Database and its applicability to studying the lifecycles of terrorist organizations. In D. Gadd, S. Karstedt, & S. Messner (Eds.), The sage handbook of criminological research methods. Dugan, L., Huang, J., LaFree, G., & McCauley, C. (2008). Sudden desistance from terrorism: The American Secret Army for the Liberation of Armenia and the Justice Commandos of the Armenian Genocide. Dynamics of Asymmetric Conflict, 1(3), 231–249. Dugan, L., Lafree, G. L., Cragin, K., & Kasupski, A. (2008). Building and analyzing an open source comprehensive database on global terrorist events (DOJ Publication No. 223287). Washington: U.S. Government Printing Office. Dugan, L., LaFree, G., & Miller, E. (2011). Using rational choice to better identify the most threatening terrorist organizations. Working paper. The University of Maryland. Dugan, L., LaFree, G., & Piquero, A. R. (2005). Testing a rational choice model of airline hijackings. Criminology, 43(4), 1031–1066. Eggleston, E. P., Laub, J. H., & Sampson, R. J. (2004). Methodological sensitivities to latent class analysis of long-term criminal trajectories. Journal of Quantitative Criminology, 20(1), 1–26. Enders, W., & Sandler, T. (1993). The effectiveness of antiterrorism policies: A vector-autoregressionintervention analysis. American Political Science Review, 87(4), 829–844. Enders, W., & Sandler, T. (2006). The political economy of terrorism. New York: Cambridge University Press. Freeman, J. R. (1989). Systematic sampling: Temporal aggregation, and the study of political relationships. Political Analysis, 1(1), 61–98. Griffiths, E., & Chavez, J. M. (2004). Communities, street guns, and homicide trajectories in Chicago, 1980–1995: Merging methods for examining homicide trends across space and time. Criminology, 42(4), 941–978. Holden, R. T. (1986). The contagiousness of aircraft hijacking. American Journal of Sociology, 91(4), 874–904. Ihrie, G. (2010). Functional data analysis. Unpublished manuscript. Jackson, B., Baker, J. C., Cragin, K., Parachini, J., Trujillo, H. R., & Chalk, P. (2005). Aptitude for destruction volume 2: Case studies of organizational learning in five terrorist groups. Santa Monica: RAND Corporation. Jones, B. L., Nagin, D. S., and Roeder, K. (2001). A SAS procedure based on mixture models for estimating developmental trajectories. Sociol. Meth. Res. 29, 374–393. Jones, S. G., & Libicki, M. C. (2008). How terrorist groups end: Lessens for countering al Qa’ida. Santa Monica: RAND Corporation. Jung, T., & Wickrama, K. A. S. (2008). An introduction to latent class growth analysis and growth mixture modeling. Social and Personality Psychology Compass, 2(1), 302–317.

6 Introducing Group-Based Trajectory Analysis and Series Hazard Modeling…

147

Kruger, A. B., & Maleckova, J. (2003). Education, poverty, and terrorism: Is there a causal connection? Journal of Economic Perspectives, 17(4), 119–144. LaFree, G., & Dugan, L. (2007). Introducing the Global Terrorism Database. Terrorism and Political Violence, 19(2), 181–204. LaFree, G., Dugan, L., Fogg, H., & Scott, J. (2006). Building a Global Terrorism Database (DOJ Publication No. 214260). Washington: U.S. Givernment Printing Office. LaFree, G., Dugan, L., & Korte, R. (2009). The impact of British counterterrorist strategies on political violence in Northern Ireland: Comparing deterrence and backlash models. Criminology, 47(1), 17–45. LaFree, G., Mazerolle, L., Denning, R., Miller, E., White, G., & Yang, S. -M. (2011). Modeling the Effectiveness of Counter-Terrorism Strategies in Indonesia, the Philippines and Thailand. Final Report. National Consortium for the Study of Terrorism and Responses to Terrorism (U.S.) and Centre of Excellence in Policing and Security (Australia). LaFree, G., Morris, N. A., & Dugan, L. (2009). Cross-national patterns of terrorism: Comparing trajectories for total attributed and fatal attacks, 1970–2006. British Journal of Criminology, 50(4), 622–649. Lafree, G., Morris, N. A., Dugan, L., & Fahey, S. (2006). Identifying cross-national global terrorist hot spots. In J. Vicktoroff (Ed.), Tangled roots: Social and psychological factors in the genesis of terrorism (pp. 98–114). Amsterdam: Ios Press. LaFree, G., Yang, S.-M., & Crenshaw, M. (2009). Trajectories of terrorism: Attack patterns of foreign groups that have targeted the United States, 1970–2004. Criminology and Public Policy, 8(3), 445–473. LaFree, G., Yang, S.-M., & Crenshaw, M. (2010). International cooperation, not unilateral policies may be the best counterterrorist strategy. In N. A. Frost, J. D. Freilich, & T. R. Clear (Eds.), Contemporary issues in criminal justice policy: Policy proposals from the American Society of Criminology Conference (pp. 121–128). Belmont: Cengage/Wadsworth. Laub, J. H. (2006). Edwin H. Sutherland and the Michael-Adler report: Searching for the soul of criminology seventy years later. Criminology, 44(2), 235–257. Lewis-Beck, M. S. (1986). Interrupted time series. In W. D. Berry & M. S. Lewis-Beck (Eds.), New tools for social scientists (pp. 209–240). Beverly Hills: Sage Publications. Li, Q. (2005). Does democracy promote or reduce transnational terrorist incidents? Journal of Conflict Resolution, 49(2), 278–297. Li, Q., & Schaub, D. (2004). Economic globalization and transnational terrorism: A pooled time-series analysis. Journal of Conflict Resolution, 48(2), 23–58. Lum, C. (2009). Theoretical and methodological innovations in terrorism research: A response to LaFree, Yang and Crenshaw. In N. A. Frost, J. D. Freilich, & T. R. Clear (Eds.), Contemporary issues in criminal justice policy: Policy proposals from the American Society of Criminology Conference (pp. 133–138). Belmont: Cengage/Wadsworth. Lum, C., Kennedy, L., & Sherley, A. (2006). Are counter-terrorism strategies effective? The results of the Campbell systematic review on counter-terrorism evaluation research. Journal of Experimental Criminology, 2(4), 489–516. Lum, C., & Yang, S.-M. (2005). Why do evaluation researchers in crime and justice choose nonexperimental methods? Journal of Experimental Criminology, 1(2), 191–213. McCauley, C. (2001). What do terrorists want? Psychological Science Agenda, 14(6), 5. McCormick, G. H. (2001). The shining path and Peruvian terrorism. In D. C. Rapoport (Ed.), Inside terrorist organizations (pp. 109–128). London: Frank Cass Publishers. McDowall, D., Lizotte, A. J., & Wiersema, B. (1991). General deterrence through civilian gun ownership: An evaluation of the quasi-experimental evidence. Criminology, 29(4), 541–560. Miller, E. (2009). Patterns of decline among terrorist movements. Presented at the American Society of Criminology Conference, Philadelphia. Moffitt, T. E. (2006). A review of research on the taxonomy of life-course persistent and adolescencelimited offender. In F. T. Cullen, J. P. Wright, & M. Coleman (Eds.), Taking stock: The status of criminological theory (pp. 277–311). New Brunswick: Transaction Publishers. Morris, N., & Slocum, L. (In progress). Estimating county level terrorism trends using group-based trajectory analysis: A comparison of latent class growth analysis and general mixture modeling.

148

L. Dugan and S.-M. Yang

Muthén, B. (2001). Second general structural equation modeling with a combination of categorical and continuous latent variables: New opportunities for latent class-latent growth modeling. In L. M. Collins & A. G. Sayers (Eds.), New methods for the analysis of change (pp. 289–322). Washington: American Psychological Association. Nagin, D. S. (1999). Analyzing developmental trajectories: A semi-parametric, group-based approach. Psychological Methods, 4(2), 139–157. Nagin, D. S. (2005). Group-based modeling of development. Cambridge: Harvard University Press. Nagin, D. S., Farrington, D. P., & Moffit, T. E. (1995). Life-course trajectories of different types of offenders. Criminology, 33(1), 111–139. Nagin, D. S., & Land, K. C. (1993). Age, criminal careers, and population heterogeneity: Specification and estimation of a nonparametric, mixed Poisson model. Criminology, 31(3), 327–362. Nagin, D. S., & Tremblay, R. E. (1999). Trajectories of boys’ physical aggression, opposition, and hyperactivity on the path to physically violent and nonviolent juvenile delinquency. Child Development, 70(5), 1181–1196. Nagin, D. S., & Tremblay, R. E. (2001). Parental and early childhood predictors of persistent physical aggression in boys from kindergarten to high school. Archives of General Psychiatry, 58(4), 389–394. Odell, J. S. (2004). Case study methods in international political economy. In D. F. Sprinz & Y. Wolinsky-Nahmias (Eds.), Models, numbers and cases: Methods for studying international relations (pp. 56–80). Ann Arbor: University of Michigan Press. Piazza, J. A. (2006). Rooted in poverty? Terrorism, poor economic development, and social cleavages. Terrorism and Political Violence, 18(1), 159–177. Rapoport, D. C. (1992). Terrorism. In M. Hawkesworth & M. Kogan (Eds.), Routledge encyclopedia of government and politics (Vol. 2). London: Routledge. Rapoport, D. C. (2004). The four waves of modern terrorism. In A. K. Cronin & J. Ludes (Eds.), The campaign against international terrorism (pp. 46–73). Washington: Georgetown University Press. Raudenbush, S. (2001). Toward a coherent framework for comparing trajectories of individual change. In L. M. Collins & A. G. Sayers (Eds.), New methods for the analysis of change (pp. 35–63). Washington: American Psychological Association. Roeder, K., Lynch, K. G., & Nagin, D. S. (1999). Modeling uncertainty in latent class membership: A case study in criminology. Journal of the American Statistical Association, 94(447), 766–776. Shellman, S. M. (2004a). Time series intervals and statistical inference: The effects of temporal aggregation on event data analysis. Political Analysis, 12(1), 97–104. Shellman, S. M. (2004b). Measuring the intensity of intranational political events data: Two interval-like scales. International Interactions, 30(2), 109–141. Sherman, L. W., & Weisburd, D. (1995). General deterrent effects of police patrol in crime “Hot Spots”: A randomized, controlled trial. Justice Quarterly, 12(4), 625–648. Skardhamar, T. (2010). Distinguishing facts and artifacts in group-based modeling. Criminology, 48(1), 295–320. Soule, J. W. (1989). Problems in applying counter terrorism to prevent terrorism: Two decades of violence in Northern Ireland reconsidered. Terrorism, 12(1), 31–46. Sweeten, G. (2006). Causal inference with group-based trajectories and propensity score matching: Is high school dropout a turning point? Unpublished doctoral dissertation, University of Maryland, College Park. United States Institute of Peace. (1999). How terrorism ends. Special Report, No. 48. Washington: United States Institute for Peace. Weisburd, D. L., Bushway, S., Lum, C., & Yang, S.-M. (2004). Trajectories of crime at places: A longitudinal study of street segments in the city of Seattle. Criminology, 42(2), 283–321. Weisburd, D. L., Groff, E., & Yang, S. -M. (2010). Understanding Developmental Crime Trajectories at Places: Social Disorganization and Opportunity Perspectives at Micro Units of Geography. Final Report. National Institute of Justice Grant. Number 2005-IJ-CX-0006. Weisburd, D. L., Morris, N., & Groff, E. R. (2009). Hot spots of juvenile crime: A longitudinal study of street segments in Seattle, Washington. Journal of Quantitative Criminology, 25(4), 443–467.

6 Introducing Group-Based Trajectory Analysis and Series Hazard Modeling…

149

Wilson, J. Q., & Kelling, G. L. (1982). Broken windows: The police and neighborhood safety. The Atlantic Monthly, 249(3), 29–38. Yang, S.-M. (2010). Assessing the spatial-temporal relationship between disorder and violence. Journal of Quantitative Criminology, 26(1), 139–163. Yang, S. -M., & LaFree, G. (2011). Trajectories and geospatial patterns (Chapter 4). In G. LaFree, L. Mazerolle, R. Denning, E. Miller, G. White, & S.-M. Yang (Eds.), Modeling the effectiveness of counter-terrorism strategies in Indonesia, the Phillipines and Thailand. A final report for the National Consortium for the Study of Terrorism and Responses to Terrorism (U.S.) and Centre of Excellence in Policing and Security (Australia). Young, J., & Dugan, L. (2011). Veto players and terror. Journal of Peace Research, 48(1), 19–33.

Chapter 7

A Complexity Method for Assessing Counterterrorism Policies * Claudio Cioffi-Revilla

Assessing the performance of counterterrorism (CT) policies and interventions poses some special challenges beyond those normally encountered in assessing the impact of other policies. This is primarily because the incidence of terrorism events is marked bysignificant uncertainty along several dimensions, such as time of onset, location, intensity, and other incident-related attributes. The analysis presented here applies the theory of political uncertainty and complexity theory to assessment of counterterrorism (CT) policies. Results from this approach can provide new and potentially actionable insights on the effect of CT policies byexamining changes in the time between terrorism incidents T and the severity of such events S (fatalities). Empirical hazard force analysis of pre- and post-CT interventions can also provide insights on event severity as well as dynamical change. Selected policy implications are discussed, including the usefulness of real-time and anticipatory analytical strategies. This paper proceeds as follows. The first section provides motivation for the methods presented and a brief discussion of earlier relevant literature. The second sections presents an integrated methodology for data analysis and model testing, based on the theory of political uncertainty and social complexity theory. The essence of these methods is to use terrorist incident data as signals for understanding patterns of occurrence, such as onset and severity, and more importantly, the latent, underlying dynamics that are causally responsible for observed occurrences. Although technically these methods are statistical, mathematical, and computational, they are essentially information extraction methods for understanding terrorist

* Chapter prepared for Cynthia Lum and Leslie Kennedy, eds. 2011. Evidence-Based Counterterrorism Policy.Springer-Verlag. C. Cioffi-Revilla (*) Center for Social Complexity and Department of Computational Social Science, Krasnow Institute for Advanced Study, George Mason University, Fairfax, VA, USA e-mail: [email protected] C. Lum and L.W. Kennedy (eds.), Evidence-Based Counterterrorism Policy, Springer Series on Evidence-Based Crime Policy 3, DOI 10.1007/978-1-4614-0953-3_7, © Springer Science+Business Media, LLC 2012

151

152

C. Cioffi-Revilla

incidence patterns and deriving new metrics for evaluation. The last section presents a discussion of main results available through these methods and some general conclusions, including discussion of evidence-based CT policy evaluation. The discussion of policy implications is innovative for the integrated multidisciplinary methods used in this analysis, which combine political uncertainty theory and complexity modeling or complex systems theory.

Introduction Motivation Assessing the effectiveness of counterterrorism (CT) policies poses numerous challenges, not least of which is the fundamental uncertainty associated with the incidence of terrorism events. Uncertainty is a universal characteristic of terrorism events, not an artifact of measuring their incidence. Such uncertainty presents a multifaceted challenge for both basic science and applied policy analysis. First, the remote and proximate causes of terrorist attacks are multiple and the interaction of such causes remains poorly understood. Second, some spatio-temporal features of terrorism events (e.g., associated statistics and distributions) may be measurable, but understanding and forecasting require a deeper theoretical approach. Third, interactive linkages between different types of events – such as, for instance, between bombings and highjacking, or between attacks across time and space – are not well understood. Most terrorism incidence patterns (not necessarily CT interventions) lack the “normal” (bell-shaped) or Gaussian distribution that is characteristic of equilibrium systems. Instead, terrorism event distributions are often skewed, showing heavy tails, symptomatic of nonequilibrium dynamics, in some cases approximating a power law with critical or near-critical exponent value of 2. (Power laws arise in many domains, not just complex systems – where they play a special role.) Heavy tails in the incidence of terrorism are a significant nonequilibrium property, because they imply that extreme events occur with much greater than normal frequency (i.e., much higher probability density). From an applied policy perspective, governmental and nongovernmental agencies involved in CT can benefit from scientific insights concerning the likely onset, severity, location, and other features of terrorist attacks. Because engineered systems are normally designed, built, and operated on the basis of reliable information concerning the operating environment and patterns of “calls on the system,” when it comes to agencies in the CT domain there seems to be less reliance on comparable information. For example, scientifically derived patterns of onset, severity, and location of attacks should be valuable for improving design, implementation, and operation of CT agencies and organizations.

7 A Complexity Method for Assessing Counterterrorism Policies

153

Earlier Literature The quantitative literature on violence is vast, but studies on the statistical dynamics of terrorism event processes are scant, even after decades of measurement, data development, and research in this area. Although the literature on counterterrorism is also large, the quantitative contributions that are relevant to the methods described in this chapter are few. Most earlier work in this area has been either on quantitative measurement and hypothesis testing (how to generate data and test propositions on terrorist events) or based on statistical regression models. In the former category, the data sets of the START Center at the University of Maryland, especially the Global Terrorism Dataset (GTD), are among the most valuable (START, 2010). Additional quantitative research on insurgency is also instructive (e.g., Townsley, Johnson, & Ratcliffe, 2008), but less focused on specific features of terrorism. In the statistical modeling area, most studies have applied regression analysis to the ITERATE data (Mickolus, Sandler, Murdock, & Flemming, 2004) and more recently the START GTD (e.g., LaFree, Dugan, & Korte, 2004), or similar data sets. While both bodies of literature are valuable, not much has been used for purposes of advancing an empirically grounded, process-oriented theory of terrorism to better understand the phenomenon, or for improving design and implementation of CT interventions and policy evaluation. The claim in this chapter is that some potentially valuable progress is feasible through the application of concepts and principles from the theory of political uncertainty (Cioffi, 1998) combined with complexity theory (Cioffi & Romero, 2009).

Method This section presents an analytical procedure for assessing the impact of CT policies, given a threat environment where terrorist attacks are occurring. The emphasis is on the actual computations and data treatments required in the analysis, without actual data. Events drawn from ITERATE, the GTD, or other country- or region-specific data set can serve as practice data for the reader interested in applying this method. Let X denote a terrorism incident-related random variable, such as the timeinterval between attacks T (measured in days), fatalities F produced by each incident (e.g., measured in deaths or other statistic of interest), distance D from the previous incident (or from a base reference location, such as a command center or logistics hub), or other variable(s) associated with terrorist events. Formally, an event process of this kind P (X áτ ñ ) can be modeled as an n-tuple of random variables with realizations ordered in historical time t (so-called “epochal time”; Feller, 1968), where each random variable is defined by its probability functions, p(x) and F(x), or probability density function (p.d.f.) and cumulative density function (c.d.f.), respectively.

154

C. Cioffi-Revilla

Fig. 7.1 Overall computational data analysis procedure for obtaining process features on the incidence of attacks. Hazard force analysis and power law analysis are parallel computational data analysis processes that can use identical data for extracting a range of inferences. Source: Prepared by the author, adapted from Cioffi and Romero (2009)

Figure 7.1 illustrates the overall methodological procedure, as detailed in the following subsections. Empirically, this approach is applicable to terrorism incident data, such as the GTD archive or other comparable data set. Following Cioffi (1998, chaps. 2–4), the procedure begins with data collection and preprocessing (top part of Fig. 7.1) continuing with the main computational event data analysis (bottom part of Fig. 7.1). Synchronic analysis is based on the entire population of data, whereas diachronic analyses are based on epochs as subsets (temporal samples) of the overall population (e.g., pre- and post-attack intervention epochs). Other comparative analyses of interesting subpopulations are also feasible (e.g., by geographic criteria, according to different CT policies, different resources, or other criteria).

7 A Complexity Method for Assessing Counterterrorism Policies

155

The core procedures involving computational events data analysis consist of two distinct but interrelated types of quantitative methods (bottom parallel processes in Fig. 7.1): (1) hazard force analysis, founded on the theory of political uncertainty (Cioffi, 1998), and (2) power law analysis from complexity theory (Boccara, 2004). Although traditionally autonomous from each other, in this approach the synergy of the two methods is exploited to obtain new inferences that advance our understanding of terrorist incidents.

Data As mentioned earlier, a specific study will use a data set such as ITERATE, the GTD, or comparable. For illustrative purposes, suppose that two coded variables used in this study were Date of Attack, used for computing time lapsed between attacks (T) and event severity (S [number of people affected]). Other variables included in the data set might include, for instance: country or region affected, type of event, cumulative death up to a given year, dollar amount invested in CT operations, and a brief description of the type of CT interventions implemented by agencies. Some uses of these and other data are mentioned in the section “Discussion.”

Analyses As shown in Fig. 7.1, the two analytical methods – hazard force analysis and power law analysis – are applied to the same data for time-between-events T and for severity S. In turn, each analysis is conducted synchronically and diachronically, as explained below.

Temporal Analysis For each variable (say, time T and severity S) and type of analysis (hazard force and power law) an overall synchronic analysis is first conducted, based on all data for an entire period, followed by a more historically detailed diachronic analysis. The latter is based on several epochs marked by CT interventions: • Epoch 1: dd/mm/yyyy to dd/mm/yyyy. Pre-intervention phase. • Epoch 2: dd/mm/yyyy to dd/mm/yyyy. Intervention phase. • Epoch 3: dd/mm/yyyy to dd/mm/yyyy. Post-intervention phase. The significance of these three “phases” (epochs, in quantitative analysis terminology) is defined in terms of CT interventions. The main theoretical motivation for

156

C. Cioffi-Revilla

these epochs – and additional reason why epochs matter – is that terrorist dynamics, in terms of forces of onset FT and forces of severity FS, which drive the onset and severity of attacks, undergo fundamental changes across epochs due to the presence or absence of CT interventions. Hazard force analysis and power law analysis aim at detecting fundamental change in such latent dynamics, as described in the next subsections. If CT interventions have an effect, then different epochs should reveal significant transitions in terms of forces affecting the event process.

Hazard Force Analysis Hazard force analysis is a quantitative method for discovering or testing distinct qualitative patterns of unobservable but nonetheless real, causal forces that operate in a process consisting of behavioral events, such as attacks and related violence. This is a method particularly suitable for understanding uncertainty – more precisely, the pattern of uncertainty and its possible cause – and risk (Singpurwalla, 2006) as described by a time series of events data. Definition 1 (Intensity Function) The intensity function H(x) of a c.r.v. X is defined by the ratio of the value of the p.d.f. to the value of the complementary c.d.f. of X. Formally, H(x) is defined by the equation H ( x) =

p( x ) , [1 - F( x )]

(7.1)

where p(x) and F(x) are the p.d.f. and c.d.f. of X, respectively. Note that, although the intensity or hazard force H(x) is a latent (i.e., unobservable or not directly observable) variable, (7.1) renders H(x) measurable, because both p(x) and F(x) can be computed from a sufficiently large set of observed realizations x i Î X . The original interpretation of (7.1) as an intensity or force is probably due to Cox (1962), based on Bartholomew (1973, p. 138). Unfortunately, most of the standard social statistical and econometric literature (e.g., Greene, 2011) treats the estimation of Hˆ (x) as just another case of regression, ignoring the much deeper dynamical implications used in this study. Accordingly, by (7.1), the specific form of H(x) (i.e., constant, increasing, decreasing, non-monotonic) depends directly on the form of the associated probability functions (c.d.f. or p.d.f.). Specifically, four qualitative cases are fundamentally important. To illustrate, let X = T, the time interval between attacks, measured – for instance – in days. Case 1. Constant Intensity: H(t) = k. In this special case, the propensity for the next attack to occur – i.e., the hazard rate or event intensity – does not change between realizations, consistent with the notion that escalating and mitigating forces of terrorism are in balance. This case also corresponds to the Poisson distribution with corresponding simple negative exponential density, p(t ) = k e -kt and t = 1 / kˆ = σ 2 (t ) .

7 A Complexity Method for Assessing Counterterrorism Policies

157

This case has arguably the strongest empirical support for many types of socio-political events, both domestic and international (Cioffi, 1998, pp. 113–116). In terms of epochs, we expected to detect a constant or perhaps increasing intensity during Period 1 (pre-intervention), consistent with earlier findings (e.g., Cioffi & Romero, 2009), because the attack process is running a “more-or-less-natural” course unmitigated by CT policy. Case 2. Increasing Intensity: dH ⁄ dt > 0. In this case, the hazard force or event intensity increases between attacks, symptomatic of a fundamentally unstable situation where attacks occur under rising pressure or increasing propensity. This situation is akin to a driven threshold system (Rundle, Klein, Tiampo, & Gross, 2000). Following this interpretation, latent onset forces build up and occasionally trigger attack events (Cioffi & Romero, 2009). In terms of epochs, we expected to observe increasing force intensity prior to interventions. Case 3. Decreasing Intensity: dH ⁄ dt < 0. In this case, the hazard force or event intensity decreases between attacks, symptomatic of a stable situation where attacks occur under diminishing pressure or decreasing propensity. This situation is akin to a leaky threshold system that dissipates forces as they build up (Rundle, Tiampo, Klein, & Sa Martins, 2002). For example, an effective CT policy may be responsible for dissipation and decreasing propensity for attacks. In terms of epochs, we expected to see this force pattern only in Epochs 2 and 3 (section “Temporal Analysis”). The above three dynamic cases can be modeled by the two-parameter Weibull equation: H ( x ) = kt β -1 ,

(7.2)

where k and b are the scale and shape parameters, respectively. The Weibull model is appropriate for modeling terrorism data for several reasons. First, numerous empirical studies of terrorism data (beginning with ITERATE) have shown that the Poisson process with constant hazard is frequently (albeit not universally) observed, as in Case 1. Second, the Weibull function for (7.1) (i.e., (7.2)) captures a variety of common threat environments, as discussed in the next paragraph ((7.3)–(7.5) below). Third, the Weibull function for βˆ » 3.5 approximates a normal Gaussian process (equilibrium dynamics) and the Rayleigh process for βˆ » 2 (linear driven system). Finally, the model is formally parsimonious relative to the variety of implications it is capable of generating (Cioffi, 1998, pp. 117–124). Using the Weibull model, the estimate βˆ computed directly from data supports the follow inferences concerning terrorist dynamics: βˆ < 1 : decreasing terrorism force Þ stable situation

(7.3)

βˆ W 1 : constant terrorism force Þ borderline situation

(7.4)

βˆ > 1 : increasing terrorism force Þ unstable situation

(7.5)

158

C. Cioffi-Revilla

Clearly, these three scenarios are qualitatively distinct, and from a policy perspective they correspond to desirable, indifferent, and undesirable conditions (threat environments), respectively. Interestingly, the mean or first moment of T is given by t = kG(1 + 1 / β ),

(7.6)

where G is the gamma function. Therefore, commonly used estimates based of mean values or other simple statistics are not as intuitive as they might seem and must be computed exactly because t is notoriously sensitive to βˆ (by (7.6)). In other words, the relationship between t and b (7.6) implies that βˆ -estimates will be highly unstable. Finally, a fourth policy-relevant case in the qualitative form of terrorism force is also interesting: Case 4. Non-monotonic Intensity. After an attack occurs, terrorism force may rise as in Case 2, but then subside, as in a lognormal function. Alternatively, the terrorism force may subside following an attack and then begin to rise again sometime after, as in a so-called “bathtub” function (Singpurwalla, 2006). These non-monotonic situations are also considered in this analysis, given their plausibility and policy relevance for planning purposes. In terms of epochs, their logic is nonlinear, ruling out monotonic forces. Empirical examples related to these cases are reported elsewhere (Cioffi, 1998, 2006; Cioffi & Romero, 2009). Summarizing the hazard force analysis procedure, terrorism events data on the time interval between attacks (T) and severity (fatalities) caused by each attack (S) are used to compute the corresponding empirical hazard functions Hˆ (t) and Hˆ (s). These empirical functions are then closely examined to determine their qualitative shape and draw inferences concerning the threat environment. This procedure is repeated for the entire population of data, as well as for each pertinent epoch. The initial expectation is that these estimates would yield mostly Case 1 (constant force), consistent with most earlier studies, but with rising value of k as the epochs progress and CT interventions are implemented. Power Law Analysis Power law analysis is a complexity-theoretic method that can be used for drawing additional unique inferences when applied to a set of terrorism data. This method is particularly useful for drawing inferences about the criticality of a process based on observed time series of events over time, combined with some size or intensity, such as attacks. Here we use the so-called type IV power law (Cioffi-Revilla, 2003, chap. 2). Other types of power laws include the rank-size law or Zipfian, various algebraic forms, and others (Kleiber & Kotz, 2003). In this approach we apply the type IV power law (Definition 2 below) because in the case of terrorist attacks it provides the most powerful complexity-theoretic inferences compared to rank-size (Zipfian) and other types of power laws.

7 A Complexity Method for Assessing Counterterrorism Policies

159

Definition 2 (Power Law). A power law of a terrorism event process P (X áτ ñ ) is a parametric distribution model where increasing values xi Î X of a given terrorism variable X occur with decreasing frequency, or f(x) ¥ x − b, with b > 0. Formally, f(x) is a p.d.f. given by p( x ) =

a(b - 1) , xb

(7.7)

where a and b are scale and shape parameters, respectively. From this two-parameter hyperbolic equation for the p.d.f., it is easily shown that the complementary cumulative density function (c.c.d.f.), defined as 1 − F(x) ≡ Pr(X > x) (a.k.a. survival function when X = T, denoted by S(x)), has the following form in log–log space: log[1 - F( x )] = a¢ - (b - 1) log x.

(7.8)

Equation (7.8) yields F( x ) = 1 -

a x

( b -1)

= 1 - ax1- b .

(7.9)

Equation (7.8) is commonly used for empirical analysis, because it can be obtained directly from the set of observed values x i . The empirical estimate bˆ is of major analytical interest, because the first moment of a power law is given by ¥ ¥ E( x) = ò xp( x )dx = a(b - 1)ò x1- b dx (7.10) min{ x} min{x} a(b - 1) 2 - b ¥ = x ½ , min{x} 2-b which goes to infinity as b → 2. In other words, there is no theoretical mean size (no expected value E(x) exists) for the terrorism variable X (such as onset times T or fatalities F) when X is governed by a power law with exponent b approaching the critical value of 2, or (b − 1) < 1 (below unit elasticity). This is an insightful theoretical result for terrorism patterns such as organizational sizes, fatalities (Cioffi et al., 2004; Richardson, 1945), and terrorist attacks (Cioffi, 2003, chap. 16). The critical thresholdbcritical = 2 marks the dynamical boundary between terror regimes or threat environments that have a finite average and computable size (b > 2) and a highly volatile regime that lacks an expected value or mean size (b £ 2). This is a theoretical insight derived directly from the empirically estimated value of the power law exponent b – by no means an observable property for any ongoing terrorism process. In practice, any data set will yield a complete set of moments, so it is easy to miss the significance of special values (such as 2.0) in the power law exponent. While it is true that only a finite number of people can die from a terrorist attack, finding that bˆ has been drifting toward the critical value of 2 is a clear signal that the threat environment is deteriorating.

160

C. Cioffi-Revilla

Based on previous studies (Cioffi, 1998, p. 52, table 2.1), time-between-attacks T is expected to obey the simple (one parameter) negative exponential p.d.f. of a Poisson process, p(t ) = λ e - λ t ,

(7.11)

where λˆ = 1 / ( t ) ; and attack severity S is expected to obey a power law. Moreover, with respect to the diachronic epochs of CT interventions discussed earlier, we would expect t to increase across periods (epochal time) and bˆ to approach criticality. Summarizing the power law analysis procedure, terrorism events data on time interval between attacks (T) and fatalities produced by each attack (S) are used to ˆ (t )] and compute the corresponding empirical power law functions log[1 - F ˆ ( s)] , for onsets and fatalities, respectively. These empirical functions are log[1 - F then closely examined to determine their qualitative shape and draw inferences concerning security conditions. One can also examine the p.d.f.s directly using kernel estimation. This procedure is repeated for the entire population of data, as well as for each epoch. The initial expectation is that these estimates will yield mostly a poor fit of the power law for the overall synchronic analysis. As epochs pass and CT policies become more effective, any power law pattern should move away from criticality (decreasing variance).

Discussion Results from the procedures described above can suggest new insights and implications for CT research and policy. The following discussion focuses on the main findings and selected policy implications in reference to issues raised in the Introduction.

Empirical Findings Results from the onset of attacks T (time between events) in the analysis of overall synchronic patterns often show a non-normal distribution with a heavy right tail, as pointed out earlier (Cioffi, 1998, p. 52, Table 2.1, 2006; Cioffi & Romero, 2009). The formal normality tests (Shapiro-Wilk) can reject the null hypothesis for the presence of a log-normal distribution in the data. The empirical c.d.f and p.d.f. both allow us to visualize this non-normal pattern in the distribution of T. These statistical properties suggest a high degree of political uncertainty, meaning conditions far from the equilibrium conditions of normality with marked central tendency and unlikely to highly improbably extreme events. Normality does not follow from equilibrium; but nonequilibrium conditions are implied by a non-normal process generating heavy upper tail because extreme events (terrorist attacks with fatalities in the thousands,

7 A Complexity Method for Assessing Counterterrorism Policies

161

i.e., several orders of magnitude above the mean) have much higher probability density than in a normal, Gaussian process. The Kaplan–Meier estimate of the survival function Sˆ (t) can demonstrate that T has a higher probability of realizing very short time spans between attacks, with rapidly increasing cumulative probability (much faster than Poisson). In addition, the empirical hazard force function can show that the intensity of the force for attacks to take place can decrease, after which such intensity can fluctuate below some value. The average hazard rate for the complete period can also vary across epochs, indicative of the effectiveness of CT policies. These and other findings are not available through plain observation or even field visits to areas of CT interventions. The two methods based on complexity theory and uncertainty theory should be viewed as separate scientific instruments for assessing CT interventions, each providing different readings on CT performance. While more traditional methods provide significant information of a different nature, these analytical results provide insights concerning terrorist dynamics that are latent albeit measurable. Such insights can shed new light on terrorist activity and underlying processes. As such, these insights can help inform policy-makers on the effectiveness of CT policies implemented or under consideration.

Policy Implications The following discussion of policy implications moves from some basic aspects of theoretical science in applied CT domains to more practical institutional issues. Throughout, the science–policy nexus dominates the discussion, but several important themes are only summarized due to space limitations. To begin, the scientific principle according to which “there is nothing more practical than a good theory” (Lewin, 1952, p. 169) is or should be as valid for CT policy analysis as it has been for social psychology – a science that evolved from humanistic origins dating back to Aristotle. In fact, as Vansteenkiste and Sheldon (2006) have noted, Lewin intended to convey a two-way relationship between scientists and practitioners, such that the two would gain from each others’ insights and specialized familiarity with information, issues, and methods – as well as toolkits. Whereas computational terrorism scientists could and should develop research that yields more actionable results, practitioners could and should make greater use of available scientific advances, including viable areas of social science and CT research. The difficulties for each are many but the potential payoff is significant. Kline’s thesis is as true for terrorism analysts as it is for physicists – some of whom, such as L. F. Richardson (founder of scientific conflict analysis) have made contributions to the science of conflict. Another way to appreciate the power of scientific approaches to CT analysis is by recalling a thesis formulated by the late mathematician Morris Kline (1985), that scientists do not learn mathematics for its own sake, but because mathematics provides a unique and powerful method for discovering fundamental features of the real empirical world that are not accessible through other methods – including direct observation, measurement, or experience.

162

C. Cioffi-Revilla

Gravity, pressure, and radiation are among the many natural phenomena that we understand through the exclusive medium of mathematics, even when we can observe their effects. Much the same is true of the terrorism features revealed by the medium of theories such as those applied in this study. Hazard rates (the latent intensity for attacks), half-life (the greater-than-even-odds tipping point for attacks to occur), and criticality (the phase transition to an extreme threat environment) are specific features of terrorist attacks that we know exclusively through the medium of mathematics, not through direct experience or observation. Within a CT context, the situational awareness dashboard of analysts and policymakers could be significantly enriched by adding new panels for viewing computational indicators, such as those applied in this analysis or others with comparable theoretical foundation. For example, application of these methods soon after Phase I may reveal the gathering momentum for upcoming attacks, perhaps in time to avoid the entrenchment and maturation of effective radical networks by reformulating an appropriate policy. While this study hypothesized a post hoc situation, by necessity, real-time or near-real-time analysis of uncertainty and complexity indicators is becoming increasingly feasible. This is also significant within a CT context. Already increased interest in open-source data and analysis on the part of the intelligence community is stimulating a new generation of information processing tools that will one day provide real-time capabilities in events analysis and related methodologies (Cioffi & O’Brien, 2007). In addition, the merging of real-time facilities with advanced data visualization and cartographic tools (e.g., social GIS, spatial social science models) – combined with Moore’s Law on exponentially increasing computing power – will soon render feasible information awareness environments that would have been close to unthinkable just a few years ago. Real-time events data analysis will provide significant support not just for CT analysts but also for planners, decision-makers, and others who can benefit from input and feedback – even if Moore’s Law may begin to encounter some limitations. Besides these improvements, sequential event process modeling of attacks – such as for suicide bombings or IED attacks – could prove helpful for practitioners, as well as challenging from a scientific perspective. For instance, a detailed empirically based event process model (sometimes known as a “business model” in organizational theory or “workflow analysis” in management science) of IED attacks could shed significant light on the attackers’ vulnerabilities, by revealing actionable information that a defender could exploit to prevent attacks or mitigate their effects. Models like this already exist for weapons of mass destruction (Allison, 2004, chaps. 1–5); they should be developed for a broad variety of terrorist attacks. More specifically, event process models should focus on phases in the overall life cycle of an attack: 1. Decision-Making: Attackers deciding to act, including cognitive processes and alternative choice mechanisms 2. Planning: Attackers organizing a schedule for implementing an attack, including operational security 3. Preparation: Attackers coordinating tasks deemed necessary to execute the attack

7 A Complexity Method for Assessing Counterterrorism Policies

163

4. Execution: Attackers carrying out the attack that causes undesirable effects for the defender 5. Effects: Consequences affecting the defender 6. Recovery: Defenders partially or fully restoring their preattack condition, including socio-psychological aspects 7. Investigation: Defenders engaging in a fact-finding campaign to apprehend attackers and their confederates 8. Prosecution: Defenders apprehending and processing attackers through the criminal justice system, and last but not least 9. Learning: Defenders harvesting lessons learned and using such knowledge to improve preparedness The simple fact that the operational causal structure of an attack processes is serialized – not parallelized – holds fundamental and inescapable policy and practical implications: all serialized behavior is vulnerable to disruption by elimination of one or more necessary conjunction. Effective defenders must therefore learn how to exploit the inescapable serialization of an attacker’s process – by making the difficult life of terrorists impossible or as difficult as possible. Of course, when it comes to the complex conflict dynamics of terrorism and asymmetric warfare, another important consideration within a CT context is that not all the necessary science is known – not even for selected regions of the world or for subsets of actors – and much will remain unknown for a long time, even as better data and better theories are developed and become available to the policy community. But this situation in the CT domain is not different from what occurs in medicine, engineering, or economics; and yet, public policy in these areas does attempt to draw on the best existing scientific understanding. Understanding what we do not know is as important as mastering what we do know. It is important to increase the availability and desirability of scientific knowledge on terrorism. The main results from development and application of methods such as these – summarized in the previous section – are to offer new actionable insights that are worth considering in the domain of policy analysis and planning. This science-based strategy – and others like it that apply computational social science approaches to the analysis of real-world conflict events (Cioffi, 1990; King & Lowe, 2003; O’Brien, 2002; Schrodt, 1989; Tsvetovat & Carley, 2005) – should become increasingly available to policy analysts and practitioners. Much remains to be demonstrated, but some evidence of increasing relevance is already available. Finally, the main emphasis in this chapter has been on assessment of CT policies with a rather narrow scope on terrorism attacks. This was intended as a way to provide focus and specificity. A more generalized perspective should cover the broader lifecycle of terrorism in its various stages, including prevention (e.g., counter-radicalization policies), preparedness, mitigation, recovery, prosecution, and learning. The methods described in this chapter are susceptible to applications in each stage, assuming a proper identification of relevant event processes. Acknowledgements Thanks to two anonymous reviewers who offered comments and suggestions, and to Pedro Romero for initial testing of these ideas in the context of counterinsurgency analysis.

164

C. Cioffi-Revilla

Funding for this study was provided by the Center for Social Complexity of George Mason University and by the Office of Naval Research (ONR) under grant no. N000140810378 (Mason Baseera Project). Opinions, findings, conclusions, and recommendations expressed in this work are those of the author and do not necessarily reflect the views of the funding agencies.

References Allison, G. (2004). Nuclear terrorism: The ultimate preventable catastrophe. New York: Henry Holt and Company. Bartholomew, D. J. (1973, 1982). Stochastic Models for Social Processes. New York: John Wiley and Sons. Boccara, N. (2004). Modeling complex systems. New York: Springer. Cioffi-Revilla, C. A. (1990). The scientific measurement of international conflict: Handbook of datasets on crises and wars, 1495–1988. Boulder, CO: Lynne Rienner. Cioffi-Revilla, C. (1998). Politics and uncertainty: Theory, models and applications. Cambridge and New York: Cambridge University Press. Cioffi-Revilla, C. (2003). Power Laws of Conflict: Scaling in Warfare and Terrorism. In Claudio Cioffi-Revilla (Ed.), Power Laws and Non-Equilibrium Distributions in the Social Sciences, Mason Center for Social Complexity, Fairfax, VA 22030. Book manuscript. Cioffi-Revilla, C., Sean, P., Sean, L., James, L.O., & Jason, T. (2004). Mnemonic Structure and Sociality: A Computational Agent-Based Simulation Model. In David Sallach and Charles Macal (Eds.), Proceedings of the Agent 2004 Conference on Social Dynamics: Interaction, Reflexivity and Emergence, Chicago, IL: Argonne National Laboratory and University of Chicago. Cioffi-Revilla, C. (2006). Power laws of conflict: Scaling in warfare and terrorism. In C. CioffiRevilla (Ed.), Power laws and non-equilibrium dynamics in the social sciences. Unpublished edited volume. Cioffi-Revilla, C., & O’Brien, S. P. (2007). Computational analysis in US foreign and defense policy. In D. Nau & J. Wilkenfeld (Eds.), Proceedings of the First International Conference on Computational Cultural Dynamics, University of Maryland, College Park, MD, 27–28 August, 2007. Available online. Cioffi-Revilla, C., & Romero, P. P. (2009). Modeling uncertainty in adversary behavior: Attacks in Diyala Province, Iraq, 2002–2006. Studies in Conflict & Terrorism,32(3), 253–276. Cox, D. R. (1962). Renewal theory. London: Methuen. Feller, W. (1968). An Introduction to Probability Theory and its Applications. 3rd ed. New York: John Wiley and Sons. Greene, W. H. (2011). Econometric analysis (7th ed.). New York: Prentice Hall. King, G., & Lowe, W. (2003). An automated information extraction tool for international conflict data with performance as good as human coders: A rare events evaluation design. International Organization, 57, 617–642. Kleiber, C., & Kotz, S. (2003). Statistical size distributions in economics and actuarial sciences. New York: Wiley Inter-Science. Kline, M. (1985). Mathematics and the search for knowledge. Oxford: Oxford University Press. LaFree, G., Dugan, L., & Korte, R. (2009). The impact of British counterterrorist strategies on political violence in Northern Ireland: Comparing deterrence and backlash models. Criminology, 47(1), 17–45. Lewin, K. (1952). Field theory in social science: Selected theoretical papers. Chicago and London: University of Chicago Press. Mickolus, E. F., Sandler, T., Murdock, J. M., & Flemming, P. (2004). International terrorism: Attributes of terrorist events, 19682003 (ITERATE 5). Dunn Loring, VA: Vineyard Software. O’Brien, S. P. (2002). Anticipating the good, the bad, and the ugly: An early warning approach to conflict and instability analysis. Journal of Conflict Resolution, 46(6), 808–828.

7 A Complexity Method for Assessing Counterterrorism Policies

165

Richardson, L.F. (1945). The distribution of wars in time. Journal of the Royal Statistical Society (Series A), 107(3–4), 242–250. Rundle, J. B., Klein, W., Tiampo, K. F., & Gross, S. (2000). Linear pattern dynamics in nonlinear threshold systems. Physical Review E, 61(3), 2418–2431. Rundle, J. B., Tiampo, K. F., Klein, W., & Sa Martins, J. S. (2002). Self-organization in leaky threshold systems: The influence of near-mean field dynamics and its implications for earthquakes, neurobiology, and forecasting. Proceedings of the National Academy of Sciences of the United States of America, 99(Supplement 1), 2514–2521. Schrodt, P. A. (1989). Short term prediction of international events using a Holland classifier. Mathematical and Computer Modelling, 12, 589–600. Singpurwalla, N. D. (2006). Reliability and risk: A Bayesian perspective. New York: Wiley. START [National Consortium for the Study of Terrorism and Responses to Terrorism]. (2010). Global terrorism database: GTD variables & inclusion criteria. College Park, MD: START Center, University of Maryland. May 2010. Available online. Townsley, M., Johnson, S. D., & Ratcliffe, J. H. (2008). Space time dynamics of insurgency activity in Iraq. Security Journal, 21, 139–146. Tsvetovat, M., & Carley, K. (2005). Structural knowledge and success of anti-terrorist activity: The downside of structural equivalence. Journal of Social Structure, 6(2), 23–28. Vansteenkiste, M., & Sheldon, K. M. (2006). There’s nothing more practical than a good theory: Integrating motivational interviewing and self-determination theory. British Journal of Clinical Psychology, 45(1), 63–82.

Chapter 8

Analyzing Terrorism Using Spatial Analysis Techniques: A Case Study of Turkish Cities Danielle M. Rusnak, Leslie W. Kennedy, Ibrahim S. Eldivan, and Joel M. Caplan

Introduction There is a general belief that with proper intelligence, achieved through precise data collection (e.g., improved micro-level data on group characteristics and infrastructure), counterterrorism agencies can more effectively manage the threat of terrorism. However, perfect data are regularly not obtainable as they are often unstandardized and unreliable, a problem discussed in other chapters in this book. As data collection improves, the assessment of terrorism risk will improve. In anticipation of the increased availability of better data, we will demonstrate the use of location quotients (LQs) to control for social and physical contextual risk in Turkey, a nation that has experienced high levels of terrorism. Using this analytical strategy, we discuss how counterterrorism resources combined with proper data can be directed to geographic areas where attacks are most likely to occur.

Terrorism and Risk Assessment Risk Layers There is an extensive literature on risk and risk management dealing with a range of topics from health security (Glass & Schoch-Spana, 2002) to crime control and crime prevention (Bradley & Morss, 2002; Simon, 1988, respectively). Since 9/11, D.M. Rusnak (*) School of Criminal Justice, Rutgers University, Newark, NJ, USA e-mail: [email protected] L.W. Kennedy • I.S. Eldivan • J.M. Caplan School of Criminal Justice, Rutgers University, Newark, NJ, USA Rutgers Center on Public Security, Rutgers University, Newark, NJ, USA C. Lum and L.W. Kennedy (eds.), Evidence-Based Counterterrorism Policy, Springer Series on Evidence-Based Crime Policy 3, DOI 10.1007/978-1-4614-0953-3_8, © Springer Science+Business Media, LLC 2012

167

168

D.M. Rusnak et al.

researchers have begun to apply risk models to the study of terrorism (Cummins & Lewis, 2003; Van Brunschot & Kennedy, 2008; Viscusi & Zeckhauser, 2003). Special attention has been directed to the uneven distribution of terrorism across nations. Researchers have found that 10% of the countries in the world (20 countries/ territories) account for almost 72% of all incidents. To look at this in another way, “Two percent of the world’s countries account for more than 27% of the world’s terrorist attacks. Five percent of the world’s countries account for half of the world’s terrorist attacks” (LaFree, 2010, p. 28). Incidents are also highly concentrated within affected countries. The concentration or clustering of multiple factors related to terrorism creates an “environmental backcloth” (Brantingham & Brantingham, 1981, p. 19). This backcloth has been the focus of risk terrain modeling (RTM) in the study of crime (Caplan & Kennedy, 2009, 2010). Risk terrains represent layers of important correlates that, when combined, identify locations in which the outcome variable, in this case terrorism, has a high probability of occurrence. The conceptual approach offered by RTM is useful in framing our study of terrorism although the full explication of the model relies on address-level georeferenced data that are not currently available for Turkey. It is possible, however, to approximate the continuous surfaces needed for the risk index calculations that can be used to create an index of relative risk of terrorism by cities across the country. The current analysis categorizes the correlates of terrorism into three general layers of risk: attractor, infrastructure, and crime terrains. The attractor terrain includes data on structures or individuals that might entice terrorists to attack a particular location or target, either by a religious or politically motivated terrorist group. The infrastructure terrain specifically addresses issues that may aggravate any political, national, social, or economic situation and may cause that city to be a breeding ground of terrorism. The crime terrain is included to control for any incidents that may have been attributed to criminal rather than terrorist activity. Both the infrastructure and crime terrains include correlates that aggravate terror risk for a particular geographic location and shed light on disorder in cities. Correlates will be discussed in detail shortly.

Location Quotients In order to address the terror risk for a particular geographic location, LQs were used to assess relative risk. LQs are generally used in economics and planning disciplines. They deal with the relative structure and activity of a phenomenon (in this case, terrorism risk). LQs were originally developed to determine activity in one area compared to its surroundings to account for the larger populations of some cities relative to others. The primary purpose of an analysis using LQs (from a planning perspective) is to address future activity, in this instance a terrorist attack, using forecasts based on the relative importance of one city in relation to the whole country (Andresen, 2007; Brantingham & Brantingham, 1998). “What happens in one city is seen to depend not only on what happens in other cities but also on what happens or what exists in surrounding resources” (Brantingham & Brantingham, 1998, p. 268).

8

Analyzing Terrorism Using Spatial Analysis Techniques…

169

Fig. 8.1 Location quotient for terrorism

A terrorist attack in one city within Turkey can be influenced by what happens across the rest of the country and LQs provide a way to control differential contextual risk. The equation used in the calculation of the LQs for terrorism (LQ) is as follows (adapted from LQC in Brantingham & Brantingham, 1998) Fig. 8.1.

Assessing Terrorism Risk: A Turkish Case Study The Study Area: Turkey Turkey is a good case study for two reasons. First, the country is in a “crossroad of conflict,” geographically connected to the Balkans, Mediterranean, Middle East, Caucasus, and, even beyond, into Asia. The distinctiveness of Turkey’s geography generates “…one of the most unstable and uncertain areas of the world” (Cakar, 1996, p. 12). Turkey has been striving for democracy and a free market economy since the end of the Cold War, and “has achieved an enviable record of economic growth and democratization and has shown its ability to play an ever-increasing role in geostrategic and geopolitical matters” (p. 12). Turkey’s unique geography promotes religious as well as political conflicts. Although Islamic, Turkey is more secular than other Muslim countries. Turkey has experienced attacks stemming from three major terrorist ideologies. The first ideological trend is religiously motivated (e.g., al-Qaeda in Turkey or Hizbollah); the second is derived from the remnants of the PKK’s (Partiya Karkarei Kurdistan – Kurdistan Workers’ Party) Kurdish liberation movement and their separatist/rebellion ideology; and the third is in support of far-left Marxist ideologies, without an ethnic focus.

170 Table 8.1 Correlates of Turkish terrorism risk used in layers Attractor terrain Infrastructure terrain Number of assembly Socioeconomic development members Net trade (exports – imports) Number of mosques City development Population

D.M. Rusnak et al.

Crime terrain Number of murder convictions

In the last 40 years, Turkey has experienced many attacks by these terrorist groups that have caused thousands of casualties. These attacks have occurred both in the metropolitan cities and in the border and mountain locations of the Eastern part of the country, placing the country as a whole into an unstable situation. For the purpose of current analysis, correlates of terrorism, or precursors to a terrorist incident, include the factors discussed as risk layers in Table 8.1. The correlate data were abstracted from the Turkish Prime Ministry State Planning Organization’s study from 2000 to 2005 that contains information regarding socioeconomic, education, and health data. The data obtained from this study include socioeconomic development, city development, population and number of assembly members, and number of mosques. Additionally, open source data were obtained and used from turkstat.gov and included net trade. The number of murder convictions per city was provided by the Turkish National Police (TNP).

Attractor Terrain Independent Variables Number of Assembly Members. For the purpose of counterterrorism efforts, there is a broader political context that needs to be considered when not only combating terrorism (Lum, Kennedy, & Sherley, 2006) but also assessing risk. “Because of the seeming irrationality of high profile Al Qaeda attacks in recent years” and the constant link between terrorism and radicalized religious groups, “it is easy to lose sight of the fact that a large number of terrorist attacks involve political disputes over territory” (LaFree, 2010). In the case of Turkey, following the capture of Ocalan, the PKK terrorist leader, the PKK began to commit more attacks toward civilians in Turkey which led to social unrest within the country. Residents in areas in which there were attacks migrated to more stable and unproblematic locations. This led to a change in the number of the assembly members over time since the number of assembly members is calculated according to the population of a city. Therefore, this warrants the inclusion of the number of assembly members into the attractor terrain as possible “attractors” or targets for political terrorism. Number of Mosques. The number of mosques per city1 in 2005 was included into the attractor terrain as an aggravating factor since they may attract terrorist attacks from a

1

For reference purposes, in other articles or studies, cities in Turkey have also been called provinces but the data was obtained from our Turkish sources who advised that these were deemed cities.

8

Analyzing Terrorism Using Spatial Analysis Techniques…

171

religious terrorism ideology within Turkey. Inclusion of both politically and religiously linked targets for terrorist activity aids in a comprehensive analysis of terrorism overall, rather than merely one theoretical/ideological categorization of terrorism.

Infrastructure Terrain Independent Variables Socioeconomic Development. The research literature demonstrates that special socioeconomic conditions such as poverty, inequality, and characteristics of a population may be related to the occurrence and/or number of terrorist incidents (Koseli, 2006). Socioeconomic problems are “a good starting place for terrorists to influence and make their propaganda” and terrorism has been justified “…as a response to inequality…” (p. 173). Economic growth is negatively related to the production of domestic terrorism2 and evidence shows economic deprivation is associated with the onset of terrorism in Europe (Gries, Krieger, & Meierrieks, 2009). The data on cities’ socioeconomic development for the current analysis were created from more than 50 selected social and economic indicators/variables.3 Economic variables included manufacturing, construction, agricultural, and financial indicators. All indicators were converted using “principal component analysis” which can be defined by summarizing many different indicators and grouping them with the help of common characteristics resulting in the socioeconomic development index score utilized in analysis. Net Trade. Eckstein and Tsiddon (2004) demonstrated that “terror has significant and detrimental short-term effects on major macroeconomic variables such as consumption, investment and net exports” (Araz-Takay, Arin, & Omay, 2009, p. 2). Using open source data from turkstat.gov, the number of imports was subtracted from the number of exports per city to determine the net amount of goods or trade within each city for 2005 and the relevance to terrorism risk. City Development. Research shows that the number of terorist incidents is negatively associated with the level of development (Araz-Takay et al., 2009, p. 4). However, more urbanized areas might experience greater frequency of attacks based on accessibility and number of targets centralized in such areas. To assess this notion, a city development index score was included in analysis. The city development index includes the demographic characteristics of the city, labor force demands, success of education and health service as well as production and income levels. Population. Population size is important in judging terrorism risk (Krieger & Meierrieks, 2009). However, the exact population for 2005 could not be obtained. Therefore, the population growth rate by year was used to estimate the expected population of 2005 Turkish cities using the population growth rate from turkstat. gov. The 2000 census population by city was used as a baseline for population and

2

Domestic terrorism indicates terrorist activity by a group or groups within the country of attack. This analysis was provided to the authors by the Turkish National Police, as were the data for city development, population, and numbers for assembly members, mosques, and murder convictions.

3

172

D.M. Rusnak et al.

the 2005 population was projected from the annual growth rate available on turkstat. gov. Results were verified against 2007 Turkish city population data and appeared to be equivalent to expected counts which reaffirmed inclusion in analysis.

Crime Terrain Independent Variable Number of Murder Convictions. “A decline in state-sponsored terrorism has caused many terrorist organizations to resort to criminal activity as an alternative means of support” (Hamm, 2005, p. iv).4 The crime terrain addresses instances of crime that may confound the terror problem. Terrorist attack victims may be attributed as murder victims with poor record keeping and skew the actual vs. the recorded threat of terrorism risk. Therefore, inclusion of the number of murder convictions per city controls for any influence murder counts may have on terrorism incident risk. Evidence suggests that most fatal terrorist attacks were recorded as murder which indicates that a ‘murder effect’ may be evident and due to incorrect data classification (Gould & Stecklov, 2009). To control for this, we use as our measure the number of murder convictions for 2005.

Turkish Terrorist Incident Data (Dependent Variable) Terrorism incident data were obtained from the TNP for 2006 (n = 87). According to the data, 23 of 81 cities experienced terrorist incidents in 2006. Of the terrorist incidents in 2006 with known cities of attack (n = 82), about 77% were bombings, 21% were an armed attack/assault, and the rest (2%) were arson and kidnapping. Around 28% of the attacks were against the government/politicians, 21% against the police/security/military, 20% against transportation/utilities, 10% against businesses, and 10% on civilians/private property. A little less than 4% were attacks on religious structures and 2% were against the media/journalists. Six percent of the attacks were against other/unknown structures/people. Since analysis was not centered around target type, this data are included for illustrative purposes Table 8.2.

Analysis A contextual risk index including all correlates was constructed for each city in Turkey (see Appendix). For the purpose of this analysis, since not all X, Y coordinates or address level data were obtainable for Turkish terrorist incidents, some estimates were made. Locations of terrorist incidents were abstracted from the data 4

Hamm (2005) examined crimes ranging from motor vehicle violations, immigration fraud, and manufacturing illegal firearms to counterfeiting, armed bank robbery, smuggling weapons of mass destruction, and transnational organized crime.

8

Analyzing Terrorism Using Spatial Analysis Techniques…

Table 8.2 2006 terrorist incidents Citya Number of attacks Adana 3

Agri

3

Antalya Batman

1 2

Bingöl

4

Diyarbakir

5

Elazig

2

Erzincan

4

Erzurum Gaziantep

1 2

Hakkari Hatay Igdir Istanbul

1 1 1 23

Target type Business Civilian Other Police Utilities Utilities Government Government Police Business Police Transportation Transportation Business Government Transportation Unknown Utilities Transportation Transportation Civilian Government Other Transportation Transportation Government Police Police Transportation Police Business Business Business Business Civilian Government Government Government Government Government Government Government Government Journalist/media

173

Attack type Bombing Bombing Bombing Armed attack Bombing Bombing Bombing Armed attack Armed attack Kidnapping Armed attack Bombing Bombing Bombing Bombing Bombing Bombing Bombing Bombing Bombing Bombing Bombing Bombing Bombing Bombing Bombing Bombing Armed attack Armed attack Bombing Bombing Bombing Bombing Bombing Bombing Armed attack Bombing Bombing Bombing Bombing Bombing Bombing Bombing Bombing (continued)

174 Table 8.2 (continued) Citya Number of attacks

D.M. Rusnak et al.

Target type

Attack type

Journalist/media Bombing Police Bombing Private property Bombing Security Bombing Transportation Arson Transportation Bombing Transportation Bombing Transportation Bombing Transportation Bombing Izmir 3 Government Bombing Government Bombing Police Bombing Kirikkale 1 Police Armed attack Malatya 3 Government Bombing Government Bombing Police Bombing Mardin 1 Civilian Bombing Mersin/Icel 5 Civilian Bombing Government Bombing Politician Bombing Politician Bombing Private property Bombing Ordu 1 Religious Bombing Sirnak 4 Government Armed attack Other Bombing Police Armed attack Police Bombing Trabzon 3 Government Bombing Religious Armed attack Religious Armed attack Van 8 Business Bombing Civilian Bombing Government Armed attack Military Armed attack Police Armed attack Police Armed attack Police Armed attack Unknown Bombing a Only cities with terrorist incidents are shown. Six cities are missing along with 5 incidents which had unknown cities

file as was other demographic information on each incident and attributed to the city of origin. Risk layers were calculated using the raw numbers of each correlate presented in Table 8.1. The risk terrain index was created using a modified version of RTM to fit available data for the assessment of terrorism risk. Once all relevant correlate data were obtained and operationalized within ArcGIS, each correlate had to be converted into

8

Analyzing Terrorism Using Spatial Analysis Techniques…

175

raster format in order to utilize map algebra (using the raster calculator in Spatial Analyst). To convert the data into a format compatible for raster calculation, each correlate (or feature within the joined attribute table) was converted via feature-toraster conversion for each of the seven variables: socioeconomic development, city development, net trade, population, number of assembly members, number of mosques, and number of murder convictions. Data were reclassified into quartiles of risk (“1” being the lowest/no possible risk and “4” being the highest possible risk). When reclassifying, the new values should reflect the accompanying research. Socioeconomic development was reclassified as “4” for cities with lower levels of development and “1” for areas of higher socioeconomic development because economic deprivation is noted to be associated with the onset of terrorism. Net trade was reclassified as “4” for cities with higher net trade and “1” for cities with lower net trade because countries with higher GDP were stated as more likely targets of terrorism and at greater risk. City development was reclassified as “1” for areas of less development and “4” for areas of more development; for the current study, more developed areas are indicative of greater risk. The crime terrain layer was reclassified as “1” for the least risky areas (or lowest amount of murder convictions), and up to “4” (highest 25%) for greatest number of murder convictions and presumed “riskiest” area. Number of assembly members, number of mosques, and population were reclassified in a similar approach to number of murder convictions. Reclassified correlates, now identified as risk layers, ideally could range in value from 7 to 28. The resulting risk index illustrates the aggregate contextual risk of terrorism for each city (see Appendix for risk index values).5 Risk layers of each correlate were divided into quartiles (top 25% = high risk, bottom 25% = low risk, and the middle 50% = medium risk6) and a value of 1–4 placed on each quartile, “1” being the least risk and “4” being the highest.7 Istanbul has the highest risk value (Risk Value = 21) of future terrorist attack and is the only city in the “high risk” category. The actual risk value range is 17–21 whereas the possible values for risk ranged from 7 to 28. Those cities with a medium risk level are located primarily in the eastern part of Turkey, supporting both the literature and practitioner knowledge expectations. Thirteen out of 46 predicted cities experienced an attack in 2006. Around 33% with medium level risk for terrorist attack actually experienced at least one subsequent terrorist attack (see Fig. 8.2 and Appendix). Logistic regression analysis measured the extent to which the Period 1 (2005) risk terrain index explained the patterns of terrorist incidents during Period 2 (Independent Variable = “Risk Value” [7–28] and Dependent Variable = “Presence of Any Terrorist Incident” [Yes (1) or No (0)]). As shown in Table 8.3, the odds ratio suggests that for every increased unit of risk, the likelihood of a future incident 5

Please keep in mind that these steps are for use of RTM without address level or XY coordinate data; if you have that coordinate information, use the steps in Caplan & Kennedy RTM Manual from (http://www.riskterrainmodeling.com). 6 “No risk” was not a categorization as no city lacked data that could plausibly aggravate terrorism risk according to the available empirical literature. 7 Osmaniye, Duzce, Kilis, Karabuk, and Yalova did not have a risk value calculated (data were missing for each of these cities in one or more risk terrain).

Fig. 8.2 Risk index and 2006 terrorist incidents

176 D.M. Rusnak et al.

8

Analyzing Terrorism Using Spatial Analysis Techniques…

177

Table 8.3 Logistic regressions for risk value of terrorist incidents 95% C.I. for Exp(B) B S.E. Wald df Sig. Exp(B) Lower Upper Risk value 0.391 0.186 4.401 1 0.036 1.479 1.026 2.131 a −2 Log likelihood = 84.021; Nagelkerke R2 = 0.091; n = 81 cities; 87 incidents; risk value 7–28 a Period 1 risk terrain

significantly increased by almost 48% (p < 0.05). The model has predictive validity (c2 = 4.922, df = 1, p < 0.05) and correctly identified a city at risk of future terror attack(s) almost half the time. However, it is important to note that only about 9% of the variance is accounted for in the model (Nag. R2 = 0.091) indicating that other factors might be affecting results (see Table 8.3). However, such results are in-line with findings when RTM is used for crime analysis. Unlike RTM use in crime analysis, disaggregation of risk values for terrorist incidents displays a more useful event picture (see Fig. 8.2). Ankara, predicted to be the second highest city for attack (Risk Value = 16), had no terrorist incidents present in 2006. By contrast, there were six cities (Ezrincan et al.) that had medium level risk but had four or more attacks in 2006. To illustrate, Van had a risk value of 14 but had 8 terrorist incidents occur in 2006. Additionally, within the “low risk” category, Mersin (also called Icel) had a risk value of 10 but had 5 incidents in 2006. These findings suggest that environmental factors within Turkey may have different effects in different parts of the country. So, the question is raised concerning the predictive validity of this approach and how we might improve the error variance and, in turn, the model by controlling for contextual risk using LQs.

Location Quotients for Terrorism LQs were calculated to represent relative risk for a terrorist attack as well as control for contextual risk. If a city has an LQ value equal to “1” then incidents within the city are proportionate to the risk of terrorism when compared to Turkey as a nation. If the value is below “1” then there is a less than normal proportion in that city than with Turkey. If a city has a value of “0” or is white/unshaded, then relatively, a terrorist incident and future risk of terrorism is in normal proportion when compared to the whole study area of Turkey. The higher the value gets above 1, the greater the relative proportionate risk in that particular city compared to Turkey as a whole. Relative risk overall for Turkish cities with incidents in 2005 and/or 2006 is actually quite low based on the fact that most LQ values are close to 0 (see Appendix for LQ values). LQ values do, however, improve the accuracy of assessing terrorism risk within Turkey over solely using RTM methods. LQs were calculated with the risk index held constant in an attempt to control for contextual risk and increase the visual reliability of the terrorism risk assessment model. Incidents for the Period 1 risk terrain (2005) were included in an analysis to further account for contextual risk and the notion that where previous attacks occur,

Fig. 8.3 LQ values and 2006 terrorist incidents

178 D.M. Rusnak et al.

8

Analyzing Terrorism Using Spatial Analysis Techniques…

179

so will future attacks. The LQ value indicates where efforts can be targeted to address other factors (see Fig. 8.3). In the relative higher value LQ places something “super contextual” (i.e., political, economic, religious, etc.) might be playing a role and would need to be researched and addressed to either reject or accept this notion. Nineteen of the 81 cities (approximately 24%) had LQ values greater than “1” indicating that these cities have the highest relative risk of a future terrorist attack. Eight cities have values between “0” and “1” and therefore have less than proportionate relative risk of a future terrorist attack (see Fig. 8.3 and Appendix). The rest of the cities were categorized as having normal proportionate risk of a terrorist attack when compared to Turkey as a whole. Agri, Bitlis, Trabzon, Hatay, Konya, Mugla, Bursa, and Sakarya all had values between 0 and 1, and for response purposes, these areas (with values between 0 and 1) may be of secondary concern to those cities with an LQ value above 1. Of the cities that were expected to have greater proportionate risk, 10 of 19 (around 53%) experienced a future attack. Therefore, suggesting that even with limited micro-level data, this method can accurately identify geographic locations where terrorist attacks occur more than half of the time.

Considerations When Spatially Analyzing Terrorism Data Study Conclusions There are rarely “ideal” situations in real life let alone when conducting research. Nonetheless, by applying an innovative approach to assess place-based terrorism risk, this chapter demonstrates that it is reasonable to apply existing data, even if they are limited, to the study of terrorism risk if the risk assessment framework is rigorous and conceptually sound. The current analysis demonstrates that terrorist incidents within Turkey are not randomly distributed throughout the landscape but rather are concentrated in a statistically significant way among certain high risk cities. Based on this analysis, a risk analysis framework can be constructed that assists in defining where problems are likely to occur and where counterterrorism efforts and resources may be most effective. The current study primarily illustrates the need for more precise (i.e., streetlevel/address) data for geographic analysis of terrorism to improve the reliability and validity of data used to forecast future locations at risk of terrorist attacks. As data collection improves, so will the strength of the terrorism risk model, ensuring that future research can be more robust and accurate. Second, the use of LQs replaces our reliance on rates, which only strengthens the statistical model. LQs present a more accurate portrayal of terrorism risk through the control of contextual risk. Third, the current research adds a proactive assessment strategy of risk combined with statistical validation of the error in risk assessment that has been lacking within terrorism research, especially with regard

180

D.M. Rusnak et al.

to mapping. With proper replication and refinement, current counterterrorism techniques and/or programs may be able to become proactive rather than reactive; nations may be able to preempt a terror attack by ascertaining geographic levels of risk.

Implications for Evidence-Based Counterterrorism A nation’s susceptibility to risk is affected by various natural and/or socially constructed factors (such as, the limited access to information, resources, and structures). Better approaches for obtaining access to information and resources, in particular collection of data on terrorism and its correlates, are now needed, especially now that spatial risk assessment techniques have improved. Rossmo and Harries’ (2011) analytical approach for terrorist incidents demonstrates the use of localized (address level) data for analysis but does not include aggravating factors such as those included in the current study. A study including aggravating factors as well as more localized data would allow for more precise analysis and subsequent allocation of resources for counterterrorism response and prevention. To fully realize the potential of more localized analysis in aiding in counterterrorism, a shift in theoretical as well as practical thought, like that which occurred within criminal policing during the 1990s, is needed. Just as the NYPD in 1990 used problem-solving in policing to respond to particular crimes, counterterrorism agencies today must use these same techniques and customize proactive responses for terrorism; and “[T]to attack terrorism proactively, police need special training” (Kelling & Bratton, 2006). With special training analyzing prospective risk, agencies would be better prepared to deal with the problem of terrorism, as a whole, rather than addressing each incident independent of context. Instead of reacting to individual terrorist incidents, analysts can proactively use risk modeling to address general sociopolitical, sociocultural, infrastructure, and socioeconomic problems. In sum, there are areas of spatial analysis dealing with terrorism which need improvement before a proactive counterterrorism strategy can be achieved: 1. Data Accessibility • In countries that experience high levels of terrorism, such as Turkey, there is limited access to free publicly available spatial data. The current data that are available are not organized well nor are they easily navigable (especially international data since it is oftentimes in need of translation). The implementation of open source geo-databases may make spatial analysis and consequently, risk modeling, less tedious and increase the ease of accessibility. 2. Data Reliability and Validity • Since data are from various sources, reliability of collection comes into play. Data congruence also plays a role in a model’s strength and accuracy. 3. Methods • The basis for correlate selection is more theoretical than empirically grounded in literature on terrorism. Therefore, more rigorous methods are needed to

8

Analyzing Terrorism Using Spatial Analysis Techniques…

181

ensure reliability of predictive validity and, until an exhaustive correlate list can be implemented based on empirical results, no method will be completely reliable. 4. Theory • Determination of whether or not there are distinct terrorist groups with distinct correlates and/or targets is necessary to create a solid framework to explain terrorism and terrorist group behavior. There currently is a discord between current theories on terrorism and correlates, only emphasizing the need for more empirical research on which factors are associated with terrorism prospective risk; especially since firsthand knowledge of the phenomenon is not always available. With more comprehensive research and a better understanding of the structural sources of terrorism, a common and rigorous methodology and assessment procedure might emerge. In the meantime, some recommendations to keep in mind for future research follow: 1. Future research should consider using various units of analysis. Smaller units of analysis, such as neighborhood-level, may provide more precise results for resource allocation. Although it is difficult to obtain neighborhood-level data, it may be theoretically the best level of analysis since there may be a large amount of variation in social and economic variables within a given city. 2. Analyzing whether risk is different for each terrorism ideological type when assessing risk is important to do prior to analysis. This disaggregation of terrorist ideology may aid in counterterrorism policing because it could shed light on whether efforts should be focused on terrorism as a whole or as separate ideological categories. If governments can deal with terrorism as a whole, then environmental, criminological, and problem-oriented policing strategies may be of more use in prevention of a terror threat. If disaggregated terrorist ideological types have higher individual predictive validity, then more focused and context-specific counterterrorist strategies will be required for adequate risk assessment. 3. Longitudinal research is necessary to observe how changes in socioeconomic conditions affect the number of terrorist incidents over time as well as their targets. Economic and social policies have been drastically changing within Turkey in the past 10 years, and an examination of the changes would indicate if or how these factors affect the number of terrorist incidents per city in subsequent years. 4. Inclusion of data on origin/location of terrorists may shed more light onto socioeconomic factors correlated to the production of terrorists, rather than the terrorist incidents. Current research indicates that being able to analyze the number of terrorists captured according to their province of origin would produce better results (Koseli, 2006). 5. The findings of Krieger and Meierrieks (2009) additionally stress the importance of good institutions (e.g., sound welfare systems) in decreasing terrorism risk. The analysis of other institutions and their effect on risk only furthers counterterrorism efforts and risk reduction strategies.

182

D.M. Rusnak et al.

Future work on this topic should look at the ways in which different types of terrorism spatially correlate to certain risk factors and the importance of counterterrorism efforts as well as crime prevention strategies for improving the terrorism threat within Turkey. In addition, the ways in which certain correlates aggravate or mitigate a location’s risk and threat level need to be more clearly understood. The current analysis provides an initial step in an attempt to articulate the dynamics underlying terrorism risk and provides an innovative way to assess terrorism risk. In the future, interactions that may intensify the influence of correlates should be analyzed by proximity in both space and time for terrorist attacks. For example, is terrorism risk influenced differently when socioeconomic status is combined with city development vs. trade productivity? These types of creative attempts at applying spatial analysis innovations to the problem of terrorism will only further the understanding of the relationship between the features of a landscape and the terror incidents that occur within, and consequently further the reliability of counterterrorism efforts.

Appendix Cities’ Risk Index, LQ Value, and Number of 2006 Terrorist Incidents City Istanbul Van Ankara Mardin Adana Sirnak Izmir Batman Hakkari Kocaeli Mersin/Icel Tunceli Diyarbakir Gaziantep Elazig Bingöl Gümüshane Mus Siirt Mugla Hatay

Risk 21 14 16 14 11 13 12 13 13 10 10 12 14 11 12 13 13 13 13 10 11

LQ 12.544 5.702 4.989 3.991 3.628 3.070 2.661 2.456 2.456 2.395 2.395 1.996 1.711 1.451 1.330 1.228 1.228 1.228 1.228 0.798 0.726

Incidents 23 8 0 1 3 4 3 2 1 0 5 0 5 2 2 4 0 0 0 0 1 (continued)

8

Analyzing Terrorism Using Spatial Analysis Techniques…

Appendix (continued) City

Risk

LQ

Sakarya Bursa Agri Bitlis Konya Trabzon Adiyaman Afyon Aksaray Amasya Antalya Ardahan Artvin Aydin Balikesir Bartin Bayburt Bilecik Bolu Burdur Çanakkale Çankiri Çorum Denizli Edirne Erzincan Erzurum Eskisehir Giresun Igdir Isparta K.maras Karaman Kars Kastamonu Kayseri Kirikkale Kirklareli Kirsehir Kütahya Malatya Manisa Nevsehir Nigde Ordu

11 12 13 13 13 13 12 12 12 11 11 13 12 10 11 12 13 9 10 10 9 12 12 10 9 12 13 9 13 13 10 13 11 13 15 10 11 9 11 12 12 11 11 12 14

0.726 0.665 0.614 0.614 0.614 0.614 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000

183

Incidents 0 0 3 0 0 3 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 4 1 0 0 1 0 0 0 0 0 0 1 0 0 0 3 0 0 0 1 (continued)

184 Appendix (continued) City

D.M. Rusnak et al.

Risk

LQ

Incidents

Rize 12 0.000 0 Samsun 14 0.000 0 Sanliurfa 14 0.000 0 Sinop 13 0.000 0 Sivas 13 0.000 0 Tekirdag 9 0.000 0 Tokat 13 0.000 0 Usak 10 0.000 0 Yozgat 13 0.000 0 Zonguldak 10 0.000 0 Italicized and bold cities could indicate type I error; Italicized, bold, and underlined could indicate type II error

References Andresen, M. A. (2007). Location quotients, ambient populations, and the spatial analysis of crime in Vancouver, Canada. Environment and Planning, 39, 2423–2444. Araz-Takay, B., Arin, K. P., & Omay, T. (2009). The endogenous and non-linear relationship between terrorism and economic performance: Turkish evidence. Defense and Peace Economics, 20(1), 1–10. Retrieved March 2, 2011, from https://netfiles.uiuc.edu/esfahani/ www/MEEA/Conferences/ArazTakay_Arin_Omay.pdf. Bradley, B. S., & Morss, J. R. (2002). Social construction in a world at risk: Toward a psychology of experience. Theory & Psychology, 12, 509–532. Brantingham, P. J., & Brantingham, P. L. (1981). Environmental criminology. Beverly Hills: Sage. Brantingham, P. L., & Brantingham, P. J. (1998). Mapping crime for analytic purposes: Location quotients, counts and rates. In D. Weisburd & T. McEwen (Eds.), Crime mapping and crime prevention (pp. 263–288). Monsey: Criminal Justice Press. Cakar, N. (1996). Turkey’s security challenges. Perceptions, 2, 12–21. Retrieved March 10, 2011, from http://www.sam.gov.tr/perceptions/Volume1/June-August1996/TURKEYSSECURITY CHALLENGES.pdf. Caplan, J. M., & Kennedy, L. W. (2009). Risk terrains as spatial intelligence: Threat suppression by tactical response. Rutgers Center on Public Security Brief. Retrieved March 2011, from http://www.rutgerscps.org. Caplan, J. M., & Kennedy, L. W. (2010). Risk terrain modeling manual. Newark: Rutgers Center on Public Security. Cummins, J. D., & Lewis, C. M. (2003). Catastrophic events, parameter uncertainty and the breakdown of implicit long-term contracting: The case of terrorism insurance. The Journal of Risk and Uncertainty, 26, 153–178. Eckstein, Z., & Tsiddon, D. (2004). Macroeconomic consequences of terror: Theory and the case of Israel. Journal of Monetary Economics, 51(5), 971–1002. Glass, T. A., & Schoch-Spana, M. (2002). Bioterrorism and the people: How to vaccinate a city against panic. Clinical Infectious Diseases, 34, 217–223. Gould, E. D., & Stecklov, G. (2009). Terror and the costs of crime. Journal of Public Economics, 93(11–12), 1175–1188. Retrieved February 22, 2011, from http://repec.iza.org/RePEc/ Discussionpaper/dp4347.pdf. Gries, T., Krieger, T., & Meierrieks, D. (2009). Causal linkages between domestic terrorism and economic growth. University of Paderborn, CIE Working Papers, 20. Retrieved March 4, 2009, from http://ideas.repec.org/f/pgr301.html.

8

Analyzing Terrorism Using Spatial Analysis Techniques…

185

Hamm, M. S. (2005). Crimes committed by terrorist groups: Theory, research, and prevention. Washington: U.S. Department of Justice, Office of Justice Programs. Retrieved March 4, 2011, from http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.76.3702&rep=rep1&type=pdf. Kelling, G. L., & Bratton, W. J. (2006). Policing terrorism. New York: Manhattan Institute. Retrieved February 10, 2011, from http://www.manhattan-institute.org/html/cb_43.html. Koseli, M. (2006). Poverty, inequality & terrorism relationship in Turkey (p. 217). Virginia Dissertation, Wilder School of Government and Public Affairs, Virginia Commonwealth University, Richmond. Retrieved January 13, 2011, from http://digarchive.library.vcu.edu/bitstream/10156/1633/1/koselim.phd.pdf. Krieger, T., & Meierrieks, D. (2009). Terrorism in the Worlds of Welfare Capitalism. CIE Working Paper, 22. Retrieved March 1, 2011, from http://groups.uni-paderborn.de/fiwi/RePEc/ Working%20Paper%20neutral/WP22%20-%202009-04.pdf. LaFree, G. (2010). The Global Terrorism Database (GTD): Accomplishments and challenges. Perspective in Terrorism, 4(1), 24–46. Lum, C., Kennedy, L. W., & Sherley, A. (2006). Are counter-terrorism strategies effective? The results of the Campbell systematic review on counter-terrorism evaluation research. Journal of Experimental Criminology, 2(4), 489–516. Rossmo, D. K., & Harries, K. (2011). The geospatial structure of terrorist cells. Justice Quarterly, 28(2), 221–248. Simon, J. (1988). The ideological effects of actuarial practices. Law and Society Review, 22, 772–800. Van Brunschot, E., & Kennedy, L. W. (2008). Risk balance and security. Thousand Oaks: Sage Publications. Viscusi, W. K., & Zeckhauser, R. J. (2003). Sacrificing civil liberties to reduce terrorism risks. Journal of Risk and Uncertainty, 26, 99–120.

Chapter 9

The Importance of Instrument Validity in Evaluating Security Screening Programs Tracy E. Costigan

Introduction Any measurement, selection, or decision making instrument must be accurate in order to draw reasonable and valid inferences from it. In the fields of education and psychology this refers to examining the psychometrics of an instrument to assure reliability and validity. Instrument validity refers to demonstrable evidence “to support the intended interpretation of test scores and their relevance to the proposed use” (American Educational Research Association, American Psychological Association, & National Council on Measurement in Education, 1999). Instrument reliability refers to whether an instrument provides consistent results across items, raters, and time at the unit of analysis (i.e., item-, instrument-, and decision threshold level). While these concepts of instrument validity are based in the fields of education and psychology, such concepts should not be restricted to measurement in only these domains. Rather, it is important to consider instrument validity in broader contexts, including criminal justice and counterterrorism, in which there exist many screening instruments aimed to identify serious threats to national security. Given the highstakes decisions and outcomes associated with such screening instruments and methods, consideration of instrument validity is of utmost importance. Despite the wealth of counterterrorism programs in general and security programs in particular, to date, there has been little to no analysis of instrument validity for screening methods or counterterrorism strategies. A review of unclassified, publically available literature reveals nearly no publications about instrument validity in this domain. This lack of validation assessment as an element in developing a scientific evidence base for counterterrorism programming is surprising, given our nation’s strong national security agenda. Furthermore, in a recent systematic review, Lum, Kennedy, and Sherley (2006) came to the conclusion that in the field T.E. Costigan (*) American Institutes for Research (AIR), Washington, DC, USA e-mail: [email protected] C. Lum and L.W. Kennedy (eds.), Evidence-Based Counterterrorism Policy, Springer Series on Evidence-Based Crime Policy 3, DOI 10.1007/978-1-4614-0953-3_9, © Springer Science+Business Media, LLC 2012

187

188

T.E. Costigan

of counterterrorism, there was a lack of scientific evidence on counterterrorism strategies overall. From a review of over 20,000 studies on terrorism, only seven contained moderately rigorous evaluations. These seven evaluations that showed scientifically sound methods covered a range of counterterrorism strategies, not just security screening. Thus, it appears that program evaluation, including examinations of instrument validity, has not been a focus in the implementation of counterterrorism strategies. Yet, there is increased interest in formalized program evaluation and validation assessments from the U.S. Congress and its General Accountability Office (GAO). Given the sheer volume of counterterrorism programs and strategies, their impact on individuals’ activities, rights, and privacy, and the proportion of the U.S. budget allocated to national security, such programs should be subjected to rigorous evaluation. Consider, for example, the screening method of Advanced Imaging Technology (AIT), also known as the whole body image scanner. In 2010, the Transportation Security Administration (TSA) began large scale deployment of AIT scanners. This imaging technology is designed to detect a wide range of threats to transportation security, particularly looking for certain types of objects hidden on a person’s body. Such threats include metallic and nonmetallic weapons, explosives, and other prohibited items concealed under layers of clothing (TSA, 2011). There was much controversy when these were initially put into operation, because many passengers believed this technology was overly intrusive. However, TSA and the Department of Homeland Security (DHS) have developed systems to protect passenger privacy and maintain that scanners do not collect nor retain personal information (DHS, 2011). At the time of writing this chapter, this technology was still a relatively new security screening method and given the controversy surrounding it, the AIT serves as a useful example to consider as a candidate for validity examination. At present, the AIT screening method is not deployed in every airport location, nor is every traveler subjected to it. Rather, individuals are selected at random for this screening, depending on location capabilities, staffing levels, and other factors affecting operational implementation. For anyone who has been screened by this method, it is clear that errors are made, including false positive (FP) errors (i.e., selecting an individual for possession of a prohibited item and upon further screening the identified object was a nonthreat) and false negative (FN) errors (i.e., missing an individual in possession of a prohibited item that is a system threat). While no screening instrument is expected to be 100% accurate, it is important to consider validity of this screening method from a psychometrics, or instrument validity, standpoint. That is, the AIT should be subjected to an examination to determine whether this screening method is valid in terms of its ability to achieve its intended goal of identifying transportation system threats. How can we evaluate the validity of the AIT as a security screening instrument? How can we determine if the screening measurement is sufficiently achieving its goal of identifying transportation security threats so that we can indeed evaluate its effectiveness? This chapter aims to set up a framework to answer these types of questions by presenting methodologies for validation assessment. These methods

9 Examining Instrument Validity

189

originate from educational, psychological, and medical sciences, but are nonetheless applicable to criminal justice or counterterrorism. Validation assessments of counterterrorism and national security programs serve several purposes. First, given that little work has been done in this area, there is a need for a methodological framework to conduct such studies using best practices from a variety of fields. Second, an evidence base for what does and does not work would benefit policymakers and the public in terms of evaluating which programs provide the most value in terms of national security, and validation assessments are one source of scientific evidence. Finally, examination of instrument validity is an important first step in broader program evaluation. When it comes to examining the impact of a national security screening program, a determination of the extent to which its instrumentation and methods accurately identify targets is necessary. If a screening program’s primary instrument lacks validity; that is if the instrument lacks evidence to support its intended uses or, stated differently, it does not collect the necessary information to counter terrorist activity, then a broader program evaluation of its impact and effectiveness is not necessary. The chapter begins with an introduction and overview to methodologies for validation assessment. This is followed by considerations for designing a rigorous validation assessment. The chapter concludes with recommendations about approaches to validation assessment.

Instrument Validity As a first step to program evaluation, the examination of instrument validity addresses the extent to which a given measurement is appropriate for its intended purpose and that the measurement leads to appropriate inferences about persons who are and are not selected for further screening. That is, before conducting studies to determine whether the screening program achieves its intended goal (e.g., reduce terrorist activity), a researcher should first ask the more specific question of whether the instrument and method used is valid and reliable. When considering instrument development, there are standard types of reliability and validity that are typically examined to establish instrument psychometrics (see, for example, Crocker & Algina, 1986; Nunnally & Bernstein, 1994; Pedhazur & Schmelkin, 1991). These are summarized in Table 9.1. These types of reliability and validity typically apply to educational, psychological, and medical measurement instruments like surveys or rating scales, selection instruments, educational and cognitive assessments, and diagnostic tests. As an extension, such validation assessments should also apply to the areas of criminal justice and national security screening instruments. Regardless of domain, the relative importance of each type of reliability and validity depends on the goal of the instrument under investigation. In the next sections, these concepts are described in further detail, with examples of how each may be relevant to national security screening.

190

T.E. Costigan

Table 9.1 Sources of evidence for instrument validity Types of validity Criterion-related validity Extent to which the instrument performs against various criteria. Extent to which scores on the instrument predict the expected outcomes (criterion) Construct-related validity Extent to which the instrument adequately measures the underlying construct that it purports to measure Content-related validity Extent to which the content of the instrument covers all or a representative sample of the domains under examination Types of reliability Inter-rater reliability Test-retest reliability Internal consistency

Consistency of item- and instrument-level scores, across raters Consistency of scores over time, at the item and instrument level Consistency of scores on item-level indicators

Criterion-Related Validity When it comes to screening paradigms in general, and national security screening programs in particular, criterion-related validity is the most critical piece of validity evidence. Criterion-related validity refers to the extent to which scores on the instrument predict the expected outcomes. In terms of security screening, this means that the most important element of a screening program is the ability for the measurement to accurately identify the individuals of interest to the system. Strong classification accuracy for a given system – i.e., discriminating targets from nontargets – is evidence that a given method (measurement system) has criterion-related validity. If other types of validity are met but criterion-related validity is not, then program justification is difficult to establish. In order to examine criterion-related validity of a security screening instrument, one must calculate classification accuracy, or the rate of correct decisions made by the screening instrument. Classification accuracy is typically illustrated in a classification matrix, in which there are four possible scenarios, two are correct decisions and two are incorrect decisions, or errors (Fig. 9.1). In this matrix, positive refers to individuals who are selected by the screening; for example, the AIT result shows some foreign object that appears to be a weapon or explosive. Such a finding might result in a true positive (TP), in which further screening determines that the person is a target (positive for the criterion) or a FP, in which further screening determines that the individual is not a target (not positive for the criterion). In the case of the AIT, a TP would be an individual found with a weapon, explosive or other prohibited item that is a threat to the system, and a FP would be an individual found with an item that is not a threat (e.g., a business card in their pocket). Generally, the relative counts of TP and FP can be easily calculated for any screening program that retains data on those selected by the measurement. That is, of the total individuals screened, one could examine the total that were correctly identified as targets and the total that were incorrectly identified as targets. From these numbers, several classification metrics (described below) can be calculated.

9 Examining Instrument Validity

191 Meets Criterion? YES

NO

YES

(A) True Positive (TP) count

(B) False Positive (FP) count

NO

(C) False Negative (FN) count

(D) True Negative (TN) count

Selected by screening method?

Fig. 9.1 Classification matrix

An examination of TP and FP counts, however, does not provide a complete analysis of classification accuracy. An examination of the negative cases is equally important to the study of criterion-related validity. Negative cases are those individuals who are subjected to screening and they do not meet instrument criteria for selection as a target. In the case of AIT, the screening result would be that no indications of explosives, weapons, or other questionable concealed items were present. A true negative (TN) refers to those who were not positive on the screening and had that individual been subjected to further assessment, the results would have indicated that he was not a target. In contrast, a FN refers to any individual who was not positive on the screening, but had they been further screened, it would have been determined that the individual was a target. In the case of AIT, this latter category refers to cases in which the scan did not find any weapon or explosive, but had the individual been subject to further screening, such a prohibited item would have been found. Because most screening programs do not track those individuals who were not positive on a screening, the rate of correct (TN) and incorrect (FN) decisions around negative screening results is unknown.1 Rather these individuals are allowed to proceed as usual; as a result TN and FN rates are unknown through available operational data. However, this information is necessary for a full examination of criterion-related validity. Thus, although it is difficult to calculate these values and estimation is required, such calculations are necessary. Once the classification matrix is constructed, with counts in each cell, metrics can be calculated to determine classification accuracy. There are several types of metrics, each of which take into consideration various types of correct and incorrect decisions in screening classification. In the literature examining metrics to quantitatively evaluate screening programs, various researchers recommend using different metrics depending on scientific discipline, goal of analysis, instrument functions, and population characteristics (including low base rates). The following table (Table 9.2) and discussion present an introduction and overview to these metrics.

1

There is a second level of negative results that should also be considered; that is, the individuals who were not selected for the screening at all (i.e., not sent through the AIT). For these cases, it is unknown whether they would have correctly or incorrectly been identified as a positive or negative on the AIT.

192

T.E. Costigan

Table 9.2 Classification metrics Metric Abbr. Calculation a Positive predictive PPV A/[A + B] value Negative predictive NPV D/[D + C] value False positive index Odds ratios

FPI

B/A

OR

[AD]/[BC]

Relative risk

RR

[A/(A + B)]/[C/(C + D)]

Percent correct

PC

[A + D]/[A + B + C + D]

True positive rate

TPR

A/[A + C]

True negative rate

TNR

D/[D + B]

False positive rate

FPR

B/[D + B]

False negative rate

FNR

C/[A + C]

Area under the curve a See Fig. 9.1

A¢

1 − 1/4 × ([FPR/TPR] + [(1 − TPR)/(1 − FPR)])

Definition Of those selected by instrument, how many are targets? Of those not selected by the instrument, how many are not targets? For every correct selection (TP), how many over selection (FP) errors? The ratio of odds of target among those selected compared to those not-selected The ratio of the probability of a target among those selected compared to those not selected Total targets and nontargets correctly classified by method Of the targets in the population, how many did the instrument correctly select? Of the nontargets in the population, how many did the instrument correctly not select? Of the nontargets in the population, how many did the instrument incorrectly select? Of the targets in the population, how many did the instrument incorrectly not select? Instrument sensitivity that takes into account TP and FP rates

Positive and negative predictive values. When screening a large population to identify a rare target, positive predictive value (PPV) is a frequently used metric. PPV is defined as the proportion of actual targets among those selected by the instrument. This metric is particularly relevant in the use of the polygraph, or lie detection instrument, when used for pre-employment screening to determine whether an individual may have plans to engage in illegal or inappropriate behavior as an employee of a given company or agency (National Research Council [NRC], 2003). In the case of the AIT, PPV would represent the number of individuals with serious prohibited items from the total number of individuals with positive AIT scans. As a complement to PPV, negative predictive value (NPV) describes the proportion of nontargets among those not selected by the instrument or screening method. For AIT, this is the ratio of individuals who were not in possession of weapons or explosives (or other items that represented a threat) relative to the total number of individuals with negative AIT scan results. This is another measure of correct screening decisions and is an important metric in providing the full picture of

9 Examining Instrument Validity

193

classification accuracy. However, because screening programs often do not retain data on cases not selected by the instrument, this metric is often not considered in screening program validation assessments. Nonetheless, if information about the rate of targets in the general population is known or can be reliably estimated, this is a useful metric for examination of criterion-related validity. False positive index. Another metric recommended by polygraph researchers is the false positive index (FPI), which is the proportion of FP counts to TP counts (NRC, 2003). This is a practical metric in that it represents the ratio of the number of positive screening results that are erroneously examined to every correct positive screening result. It may be useful to decision makers when considering the cost of screening errors in terms of added resource expenditure. For example, if a positive screening result leads to a 10 minute secondary screening to verify that result, then an FPI of 50 would indicate that the particular system would spend an additional 500 minutes of unnecessary follow-up screening for every 10 minutes of correct follow-up screening. While no screening system will be absolutely perfect in classification decisions, the FPI provides a meaningful estimate of resource expenditures due to screening errors. Like PPV, FPI is concerned with accuracy and errors in the screening instrument among those that the instrument selects; thus, this metric can be evaluated based on existing data. Moreover, PPV and FPI are preferred metrics for screening of low base rate conditions or outcomes (NRC, 2003; Sackett & Decker, 1979). Odds ratio and relative risk. Two useful metrics that come from epidemiology and medicine are the odds ratio (OR) and relative risk (RR) statistics. The OR is defined as the ratio of the odds of a positive outcome among those selected by the instrument to the odds of a positive outcome among those not selected by the instrument (Edwards, 1963). The OR can also be thought of as the ratio of the odds of an event occurring to the odds of that event not occurring. An OR greater than 1.0 represents an increased likelihood of the event and an OR less than 1.0 represents a decreased likelihood. For example, if the OR for identifying a threat using AIT was 4 and the baseline odds of an individual carrying a serious prohibited item was 1 in 10,000 in the general population (i.e., odds = 1 to 9,999 or 0.0001%), then use of the AIT would increase the odds of finding such individuals fourfold or to 4 in 10,000 (i.e., odds = 4 to 9,996 or 0.0004%). The RR, sometimes called the risk ratio, gets at the same concept as OR, proportion of events occurring to not occurring, but is defined slightly differently based on probabilities rather than odds. The RR represents the ratio of the probability of a positive outcome among those selected by the instrument to the probability of a positive outcome among those not selected (Cornfield, 1951). For example, if the RR for AIT was 4, and the baseline probability of an individual carrying a serious prohibited item was 1 in 10,000 in the general population (i.e., probability = 1 in 10,000 or 0.0001%), then use of the AIT would increase the probability of finding such individuals fourfold or to 4 in 10,000 (i.e., probability = 4 in 10,000 or 0.0004%). Again, a value greater than 1.0 represents increased risk and a value of less than 1.0 represents decreased risk.

194

T.E. Costigan

These two metrics are very similar in concept; however, they differ because of differences in the definition and meaning of odds and probability. Depending on the data in a 2 × 2 classification matrix, interpretations of these metrics can vary substantially. While RR and OR statistics can lead to different values and interpretations in commonly occurring events, when the outcome is a rare event, as in the example of AIT, the two metrics are essentially equivalent. In those cases, the OR is the more appropriate metric to report. Ultimately, decisions about which metric to select often depend on the researcher’s background and preference. For example, many researchers report the RR because it is more interpretable to nonstatisticians. In contrast, however, the OR is often preferred because of its use in regression analyses, case-controlled studies, and meta-analyses. Many researchers have presented discussions about the advantages and limitations to each metric (see for example, Davies, Crombie, & Tavakoli, 1998; Sistrom & Garvan, 2004). True positive rate and true negative rate. Signal detection theory examines how well a signal detects targets and nontargets from the population (Green & Swets, 1966; Swets, 1996). Whereas PPV, NPV, and FPI metrics represent the conditional probabilities of identifying targets or nontargets given screening decisions, signal detection theory describes the inverse; that is, the conditional probability of a positive (or negative) screening decision, given the population of positive (or negative) conditions (i.e., target [or nontarget]). These metrics are represented by sensitivity or true positive rate (TPR), specificity or true negative rate (TNR), false positive rate (FPR), and false negative rate (FNR). TPR is defined as the number of correct positive screening decisions from the total population of targets. TNR is defined as the number of correct negative screening decisions from the total population of nontargets. The FPR and FNR metrics represent screening errors, as described in Table 9.2. For AIT, TPR would represent the number of individuals identified with weapons or other prohibited items from the total population of individuals with weapons or other contraband, and TNR represents the number of individuals identified by the AIT without weapons or prohibited items from the total population of individuals without weapons or prohibited items. These values require known population base rates of individuals traveling with items that are considered threats to the system, and thus involve some estimation. While these metrics are not always recommended in the screening of a low base rate phenomenon (Sackett & Decker, 1979), these metrics are often used in describing classification accuracy. Nonetheless, TPR/TNR and their complements FPR/FNR are value metrics for considering system utility and these represent one more perspective on criterion-related validity. An examination of criterion-related validity should examine an instrument’s ability to both accurately identifying targets (i.e., TPR and PPV) and rule-out nontargets (i.e., TPR and NPV). This is important because there are costs associated with errors due to incorrect decisions about targets and nontargets. It is often assumed that there is a higher cost associated with missing a target (e.g., in the case of an AIT scan result allowing an individual to proceed, who is carrying an item that is a threat to the system) than with missing a nontarget (e.g., the AIT scan selects an individual for secondary screening, who does not have a prohibited item). Of course,

9 Examining Instrument Validity

195

the cost of missing a single threat can be extremely high in that such an individual could be involved in an imminent terrorist attack. However, high rates of overselection of nontargets results in costs that include increased system burden, higher resource expenditures, screening delays (e.g., passengers have to wait longer to get through security), and generally negative reactions to the system. These overselection costs can add up to the point that a system is not efficient. Thus, it is important to consider both types of correct and incorrect decisions in screening systems. Within the context of signal detection theory and these metrics, one way to consider the overall accuracy is to examine the instrument’s receiver operating characteristic (ROC) curve (Green & Swets, 1966; Swets, 1996). The ROC is a plot of the instrument’s TPR vs. FPR, for various settings of the instrument (here, screening method). An instrument that can achieve a high TPR while maintaining a low FPR will have a curve that gets close to the upper left corner of the 1.0 × 1.0 bounds of the plot. The area under the ROC curve (AUC) is a way of measuring how close the instrument comes to these bounds. There are various ways of calculating AUC; one of which is A¢ as defined by Norman (1964; see also Craig, 1979), which is preferred because it does not require a normal distribution and equal variances of TP and FP (Caldeira, 1980). This summary metric is used by many in the field of selection research to examine instrument accuracy, and it is often used to compare across studies, instruments, and paradigms. Confidence intervals. Each of the metrics above provides a point estimate of classification accuracy. Sampling error is always associated with point estimates because these values are calculated based on data from a sample of the population. To understand the precision of these metrics, then, inclusion of a confidence interval (CI) is recommended. The CI is a range, based on the point estimate and sampling error, and describes the interval in which the exact value for the population falls. CIs can be calculated with a range of confidence, or certainty, levels. For example, a 95% CI indicates that one can be 95% certain that the true population value for a given metric falls within the range of the CI. Narrow intervals are preferred because they represent more stable and precise estimates. If the PPV for AIT was 0.22 with a 95% CI from 0.20 to 0.24, then one could be 95% certain that the true population metric for this PPV fell within that range. CIs can be calculated for all of the metrics described in this section, with the exception of the FPI. In comparing screening programs, CIs can also allow for an examination of whether accuracy metrics differ significantly between methods. Because the CI represents the interval in which the true population parameter falls, nonoverlapping CIs represent statistically significant differences between the methods for a given metric. Sensitivity analyses. Above, it is noted that when calculating classification accuracy for a screening program, certain metrics involve estimated data. Specifically, of the individuals not selected for screening, one must estimate the rate of correct and incorrect decisions. To do some, the validation assessment must estimate of the population base rate (or parameter) of travelers with weapons, explosives, or other materials that are threats to the system. Because this parameter is an estimate, it is important to consider potential variation in this value. This is particularly a concern

196

T.E. Costigan

in the case of low base rate phenomena, like in counterterrorism screening programs, in which such population parameters may be overestimated. A sensitivity analysis, in which varying base rate values are considered, is a useful way to examine the impact of potential error in this estimate. One can examine metrics that involve estimated data – TPR, TNR, FPR, FNR, A¢, NPV, OR, and RR – by considering the impact of varying base rates on these values. Specifically, these metrics can be calculated for varying scenarios using ranges of hypothetical base rates above and below the estimated parameter used in the study. The range of estimates selected for a sensitivity analysis will depend on the estimated base rate for the given study. For example, if it were estimated that the population parameter for the outcome of interest was 0.01%, then the sensitivity analysis range may span 0.005% (i.e., half the estimate) to 0.02% (double the estimate). In contrast, if the population parameter for the outcome of interest was 6%, then the sensitivity analysis may range from 1 to 11% (±5% points). The range selected for a sensitivity analysis depends on a variety of factors, including but not limited to, the magnitude of the estimated population parameter, the precision of this estimate, and the relative seriousness, or the cost of a screening error.

Construct-Related Validity Once criterion-related validity is established for a screening program, it is useful to examine content- and construct-related validity as supplemental evidence to support instrument validity. Construct-related validity is defined as the extent to which the instrument adequately measures the underlying construct that it purports to measure (Crocker & Algina, 1986; Farrington, 2003; Nunnally & Bernstein, 1994; Shadish, Cook, & Campbell, 2002). A construct refers to an abstract concept that an instrument aims to measure; therefore, the measurements selected to represent a given construct need to be examined to determine the extent to which they truly do represent that construct. With respect to examining instrument validity, constructrelated validity is primarily relevant to behavioral, attitudinal, and judgment-based instruments. For example, this is relevant to screening programs requiring officers to score individuals based on a list of behaviorally oriented items, such as the case for behavioral recognition programs used by TSA, the Federal Bureau of Investigation (FBI), law enforcement agencies, and others. In these types of screening programs, the focus is on identifying suspicious individual behaviors, rather than suspicious or prohibited items (Frank, Maccario, & Govindaraju, 2009). In standardized behavioral recognition programs, officers usually work from a list of items that represent anomalous or suspicious behavior; these items are therefore meant to represent persons who are threats to the system, whether due to their intentions to engage in a hostile act or due to what they may be carrying. It would thus be important to examine whether the listed behaviors indeed represent the overarching construct(s) that each is trying to measure (e.g., suspicious behavior or devious intent to commit a crime). That is, a validation assessment would benefit

9 Examining Instrument Validity

197

from examining the extent to which the instrument items really measure what they were designed to measure. In contrast, when the goal of a screening instrument is to identify a clearly measureable characteristic of an individual (e.g., gender) or a concealed object that the individual may be carrying, construct-related validity is not relevant to examining screening program validity. For example, construct-related validity is less relevant to the AIT in which the scanners and operators aim to find materials that are considered threats to the system. Rather, in this case there is a set of test results that are clearly defined as positives. While there may be some operator judgment required in the use of AITs, variability in these judgments is a matter of reliability (discussed below) rather than validity. Thus, this type of validity does not necessarily apply to all types of security screening programs. Evaluation of this type of validity requires examining correlations among various measures of the construct. Some methods used to examine construct-related validity, when relevant to a validation assessment, include correlation analysis and factor analyses (exploratory and confirmatory). The goal of such analyses is to examine relationships among measurable variables (instrument items) and latent variables (constructs) to determine whether the instrument items are measuring what they are designed to measure. Specific approaches to the analysis of construct-related validity are fairly involved and depend on the instrument. Further reading can be found in texts devoted to this subject (see for example, Nunnally & Bernstein, 1994; Pedhazur & Schmelkin, 1991).

Content-Related Validity Content-related validity is defined as the extent to which the content of the instrument covers all or a representative sample of the domains to be measured. In other words, content-related validity is concerned with whether the instrument items represent all facets of a given construct or outcome. For programs where content-related validity is relevant, it should be appropriately examined in a validation assessment, as it has an impact on instrument generalization. As with construct-related validity, this source of validity evidence applies more to behavioral and attitudinal scales and less to measurements of items that individuals may be in possession of. Thus, this type of validity evidence is also less critical to security screening programs and may only be relevant to certain types of screening instruments. Content-related validity is more relevant, for example, to a final examination in an academic course, in which it is important that such a test cover a representative sample of course content. In the case of national security screening programs, content-related validity may apply to behavioral recognition programs, as described above. Content-related validity should be considered in this case because it is important that the items used for such behavior-based screening instruments are representative of the large range of behaviors that are indicative of suspicious persons.

198

T.E. Costigan

Examination of content-related validity primarily involves subject matter expert (SME) review. The goal of such a review is to determine whether items on the instrument are representative of, and essential for, measurement of the construct that the instrument aims to measure. There exist some methods to quantify SME ratings of content-related validity, including the content validity ratio (CVR; Lawshe, 1975) and Cohen’s Kappa, or k (Cohen, 1960), both of which represent measures of agreement.

Inter-Rater Reliability Reliability, which refers to consistency in the implementation of a measurement tool, is extremely important to any screening program. Reliability from a psychometrics perspective relates particularly to consistency in human judgments and decision-making (Nunnally & Bernstein, 1994; Pedhazur & Schmelkin, 1991). While the mechanics of AIT and other screening technologies should be examined for reliability during laboratory testing prior to operational use, the types of reliability assessments of interest to this discussion of instrument validity focus on how the operators handled AIT scanner results. That is, a technology could be perfectly reliable and consistently provide the same output when presented repeatedly with a given stimulus, but the human decisions that are made based on the technology’s output may vary, which would be an issue. When it comes to human judgments, whether in the interpretation of automated technology output (e.g., AIT or X-ray) or nontechnology involved methods (e.g., a pat down, behavior recognition, or interview), reliability is important. Three primary types of measurement reliability are listed in Table 9.1; the first of which is inter-rater reliability. Inter-rater reliability refers to the extent to which there is consistency in decision-making in a screening method across raters. That is, regardless of who is making judgments or scoring an individual during screening, a reliable instrument should produce the same results for that individual’s screening. Any pair of operators should come to the same decision, regardless of the individual or object scanned and regardless of the setting where the screening occurred. To examine inter-rater reliability, a study can be designed in which screening personnel are required to score a standard set of items. For example, with AIT, screeners may be asked to examine a set of AIT scanned images and make judgments about what is or is not in the scan. Judgments can be made at the decision level (i.e., Should individual be further screened or allowed to proceed?) or item- level (i.e., What exactly is the positive finding?). Judgments can be correlated or scored for agreement across raters (i.e., screeners) as well as between raters and SME judgments. In the case of methods like AIT, in which the outcome is a known, physical piece of evidence, screeners’ judgments can also be scored in terms of items correct. High levels of inter-rater reliability suggest clear-cut definitions of the instrument items, good training of personnel using the method, and stability in judgments across settings and situations.

9 Examining Instrument Validity

199

Test-Retest Reliability Also concerning human judgments, test-retest reliability refers to consistency of scores over time. That is, regardless of when the screening occurred, judgments about the same scenario should not change. Holding all else constant – rater, individual being screened, object on the person, etc. – the results of a screening should be the same across time. This type of reliability must be assessed across a time period that is long enough that the raters do not recall their determination for a given sample, but not so long that conditions have changed. A 2–4-week period is fairly standard delay for testretest reliability. To examine test-retest reliability, the same set of screening scenarios is used for the two test periods, with the same group of personnel rating the material. Decisions or scores are correlated for each rater to examine intra-individual consistency. Low levels of test-retest reliability suggest lack of system consistency, including issues with operational definitions, training of personnel, or coding drift (i.e., changes in scoring/decisions over time).

Internal Consistency Reliability Internal consistency reliability refers to consistency of responses across instrument items. This type of reliability assessment is noted here for completeness; however, it is most relevant to multi-item scales that represent one or more constructs. For example, internal consistency may be examined for an intelligence test or a customer satisfaction survey in which one would expect patterns of responses to be consistent on the subsets of items that aim to measure the same constructs. This type of reliability assessment is not particularly relevant to screening instruments because, in most cases, there is a breadth of signals to detect an outcome and only one, or a small number of, signal is needed for a positive result.

Design Considerations The previous section provided an overview of the primary elements of instrument validity and reliability, with examples of how each piece of validity evidence applies to national security screening programs and methods for analysis. In this section, important considerations in designing a validation assessment are discussed. As described above, a validation assessment consists of examining instrument validity and reliability as a critical first step in program evaluation. Validation assessment focuses specifically on the question of, does the instrument measure what it purports to measure, perform as expected, and lead to valid inferences, whereas program evaluation involves a broader set of questions examining goals, design, implementation, utility, cost-effectiveness, etc.

200

T.E. Costigan

There exist several design challenges related to validation assessments that are important to consider, particularly when examining programs already in operation. These include maintaining experimental validity in a field study; addressing issues of unknown and low base rates of outcomes; and addressing issues of varied and low base rates of signals. Ultimately, to conduct a validation assessment for a national security screening program that is methodologically rigorous it must be conducted in the setting where the program is in operation. These issues must be considered in this context.

Experimental Validity The experimental methodology of a research study undoubtedly has an impact on its results. A researcher must consider the study design, including data collection methods, sample size, adequacy of sampling procedures, reliability of measurements, and appropriate statistical analyses. In doing so, it is important to consider how to best balance experimental controls, which assures scientific rigor and minimizes experimental error, with a design that will produce results that are generalizable to real-world, operational settings. Whereas the earlier discussion of instrument validity referred to evidence to support the intended purpose and use of measurements in evaluation, issues of study design refer to experimental validity, the impact of study design on research results. Experimental validity will have an effect on validation assessment (i.e., examining instrument validity through an analysis of reliability and validity), as well as an effect on program evaluation. While the latter is extremely important in terms of outcome analysis, this discussion of issues associated with experimental design focuses on the former, instrument validity. Advantages of basic research in the laboratory setting include the researcher’s ability to control experimental design, measurements, variables to investigate, and confounding variables. Experimental controls are important so that reasonable conclusions can be drawn about relationships between independent variables (predictors) and dependent variables (outcomes) from the study results. Internal validity refers to the experimental controls necessary to draw such conclusions, and in designing a study to maximize internal validity, the researcher seeks to minimize bias and errors that would affect the reliability of the results. There are many potential threats to internal validity that must be considered in research design. These threats to internal validity are issues that may impact study results, and the concern is that study outcomes could be attributed to variables associated with these issues rather than to the variables of interest in the experiment. While many of these threats to validity are most relevant to treatment studies, which are longitudinal and have pre- and post-tests, these types of issues should still be considered in a validation assessment to the extent that each is relevant. Some of the most common threats to internal validity are listed in Table 9.3. More detailed discussions of threats to validity can be found in research design textbooks, including the seminal work of Campbell and Cook (e.g., Cook & Campbell, 1979; Shadish et al., 2002).

9 Examining Instrument Validity

201

Table 9.3 Threats to internal validity Threat Definition History Events that occur during the course of the experiment Maturation Changes in individuals over time Testing Improved scores on a second test from exposure to the first test Instrumentation Changes in instrument, observers, or scoring systems over the course of the experiment Attrition Loss of study participants in repeated measurement studies Selection Biased selection or assignment of participants to experimental conditions Selection–maturation interaction Interaction between selection bias and maturation bias that can produce confounding variables Regression to the mean Selection of participants who score on the extreme ends of variables of interest Table 9.4 Threats to external validity Threat Definition Nonrepresentative sampling Varying results for different participant characteristics, which may or may not be due to participant variation Nonrepresentative setting Varying results in different settings, which may or may not be due to setting variation Nonrepresentative timing Varying results across time periods, which may or may not be due to time variation; rather, results may be due to unique situations or circumstances during a particular time period Reactiveness Participants’ reactions about being included in an experiment, which may be different than typical reaction to the same stimulus outside of the experiment

When examining instrument validity for a national security program, particularly one that is already in operation, the degree of experimental control and internal validity must be balanced with the ability to generalize results to real world situations, which is referred to as external validity. That is, how a counterterrorism screening instrument performs in a controlled laboratory setting will have little to no bearing on how it performs in an operational setting. Too many other factors are necessary to consider, like the total volume of individuals passing through a checkpoint, variability in individuals (including their behavior, demeanor, appearance), other setting characteristics, security personnel administering the screening method, and most importantly the motivations of the high-risk individuals and the stakes associated with attempting to defeat the system. If a security program was examined in the laboratory or controlled setting, it would be extremely difficult, if not impossible, to account for all of the variability that exists in real world screening settings. Thus, while internal validity is important to research design, it is equally important to assure external validity in this domain of applied research. Table 9.4 lists some threats to external validity that must also be considered in designing a study. Again, these are described in full detail elsewhere (Cook & Campbell, 1979; Pedhazur & Schmelkin, 1991; Shadish et al., 2002).

202

T.E. Costigan

The challenge in designing a scientifically rigorous validation assessment is in balancing internal and external validity. One recommendation in designing a validity assessment for a counterterrorism screening method is to conduct a field study that collects data in an operational setting, where the instrument is already in use. The operational setting is less controllable by the researcher; thus, it requires some innovative design to assure that the needed data are collected. Nonetheless, it is possible to conduct a rigorous field study. In addition, such a field study requires buy-in from program managers, staff, and stakeholders involved in the security screening method. While such a study, if designed well, will produce strong external validity, it may be relatively lacking in internal validity. However, internal validity can, and should, be addressed in applied research by developing rigorous methods for sampling, data collection, and statistical analyses. Although all threats to internal and external validity cannot be controlled, a well-designed study can produce reliable and stable results.

Unknown Base Rate of Outcomes Another major consideration in examining the validity of a national security screening program is the base rate of the target of interest. Ultimately, the primary goal of this type of screening program is to identify individuals who are engaged in terrorist activity. Although the population prevalence of this outcome is not precisely known, it can be reasonably assumed that the base rate for individuals engaging in terrorism is extremely low. This low and unknown base rate affects approaches to validity assessments. As noted above, classification accuracy metrics are not complete without an assessment of the correct and incorrect decisions about those screened as negative by the instrument. That is, in order to fully examine criterion-related validity, one must be able to describe the system’s ability to discriminate between targets and nontargets. Yet, screening instruments typically only retain information about those identified as positive cases (i.e., those selected by the screening method) and the only way to complete the analysis is to estimate the negative cases (i.e., those not selected by the screening method). There are several ways to do this, but the most rigorous method is to use the population prevalence of the outcome of interest in constructing the classification matrix and associated accuracy metrics. Grant (1974) introduces this issue for early disease detection screening methods and describes four methods to estimate the population prevalence of the outcome. The most precise method, of course, is to obtain the actual population prevalence of the outcome. While most precise, this is usually not feasible unless other studies of identical outcomes have been conducted in matching populations and settings (and even then there may be considerable error in those studies). The second method is to conduct a sampling study to obtain the population prevalence. In a sampling study, individuals are randomly selected and examined for the outcome of interest.

9 Examining Instrument Validity

203

In doing so, it is critical that appropriate sampling procedures be incorporated into the study design to assure adequate representativeness of the population of interest. If these first two options are not possible, then the next two options include using prevalence information from other studies of screening programs used for similar populations and outcomes or coming up with a “best guess” based on the researcher’s experience in the area. It should be clear how each method, other than knowing the exact population parameter, has some error in its estimate, and the degree of error increases as the method becomes less precise. Thus, of these four methods, Grant recommends the second method, a sampling study, to most accurately estimate the population prevalence of the target outcome if it is unknown. This is the recommended method for validation assessments for national security screening programs. Such a sampling study provides information about the negative cases. That is, using the estimated population parameter for the outcome, a classification matrix can be constructed with actual data for screened cases (top row of Fig. 9.1) and estimated data for nonscreened cases (bottom row of Fig. 9.1). Then all of the criterion-related validity metrics may be calculated. Given that the negative cases will most likely be estimated as in a sampling study, the precision of this estimate is of utmost importance. Any classification metrics derived from estimated data in the bottom row are only as accurate as the information used to derive these estimated cell values. This caveat applies to all metrics and corresponding CIs, with the exception of PPV and FPI, which do not require estimated data. As noted earlier, one can conduct a sensitivity analysis to consider how these estimates affect outcome metrics, thus addressing the error of the estimate. In addition, the CI around each metric, as described above, provides information about the precision of the metrics.

Low Base Rate of Outcomes Although the population base rate of terrorist activity or presence of a security risk is unknown precisely, it is certainly an extremely infrequent or rare event. The low base rate of the outcome of interest is another issue when evaluating a screening instrument. In the case of national security screening programs, a single catch or a single miss of an individual who is planning to engage in a terrorist attack has an enormous impact on society. Because the prevalence of individuals intending to engage in a terrorist attack is so rare compared to the total population, there are not enough instances in any setting to allow for a precise calculation of classification accuracy. That is, to get a reliable estimate of classification accuracy, with reasonable CIs, one must have sufficient samples in each cell in the classification matrix (Fig. 9.1). The exact number of cases per cell can be determined by a power analysis and even small sample sizes (i.e., under 20 cases in certain cells) may be acceptable. However, if the outcome of a validation assessment was defined as individuals engaging in terrorist activity, it would take an inordinate number of total cases to get

204

T.E. Costigan

even a small number of cases in the TP cell (Fig. 9.1; Cell A) of the classification matrix. Such a study would require data collection for an exceedingly long period – years or decade – to achieve enough positive cases for a reasonable assessment of the screening method. Thus, in conducting a validation assessment, a researcher must consider measures broader than terrorist activity. There are two ways to think about these broader measures. On the one hand, a researcher could select measures that serve as proxies for terrorism. This may include criminal activity; noncriminal suspicious activity; or possession of items that are prohibited in the sterile area. One must make the argument why any such proxy measures are reasonable, given the security method under examination. On the other hand, a national security screening program may actually be interested in identifying targets that are threats to a given system and these targets may be broader than terrorists. For example, in using the AIT, while our nation’s security programs may ultimately aim to identify terrorists, if an individual is found with weapons, explosives, or other items viewed as security threats, TSA is not likely concerned with the reason for possession of such materials. Regardless of whether the individual found with contraband is engaged in some level of terrorist activity or engaged in criminal activity, it is in TSA’s interest to remove that weapon from the transportation system. Moreover, whether carrying of a weapon was inadvertent (i.e., individual has a license for a gun and forgot to place it in checked baggage) or purposeful, again, it is in TSA’s interest to remove that weapon from the system. Thus, in the case of AIT, and similar screening methods, while the nation’s broader interests may be to prevent terrorist activity, the near term or proximate outcome is removing threats from the system. Such an outcome, while still relatively low in frequency, will result in a higher base rate, as compared to the outcome of “terrorist,” making a validation assessment more feasible. Furthermore, using this broader outcome is not only higher in frequency, but, in this case it also allows for a precise operational definition. Working with the system users, the researcher can specifically define what types of weapons, explosives, or related materials count as positive for an outcome (e.g., gun) and what does not (e.g., pocket knife) depending on the screening program goals. In contrast, because there is less agreement upon the definition of “terrorism” as LaFree (2011) and Sheehan (2011) point out, it would be difficult to make reliable and systematic judgments about which cases counted as positives if terrorism were used as the outcome. Using an outcome that is operationally defined well will result in less measurement error and therefore more internal validity. This discussion should highlight the challenges associated with low base rate outcomes and the importance of precise definitions. Though it may seem that choosing the target outcome in a validation assessment would be straightforward, from a measurement perspective it can be quite complex. Yet, the precision in this measurement is of utmost importance to the validation assessment. Given these issues, it is recommended that the researcher conducting a validation assessment work closely with the operational staff to determine the screening method’s target of interest and develop clear operational definitions of reliably measureable outcomes.

9 Examining Instrument Validity

205

Varied and Low Base Rate of Signals In addition to the infrequency of outcomes in national security screening programs, the system inputs – the signals examined by the system – are also often rare and highly varied. For example, the AIT screening is looking for a range of items that are considered weapons, explosives, or other system threats. The signals to detect these outcomes (i.e., scanner results) are low frequency in their occurrence and highly varied, given the range of materials that could be used as weapons. Moreover, given expected shifts in terrorist modus operandi, materials may change over time, increasing the variety of potential signals that the AIT must detect. The low frequency and high variety of signals present a challenge for validity and reliability assessments. This is an issue for more than just AIT and other screening methods that aim to identify weapons, explosives, or other threats. For example, behavioral recognition programs usually include a wide range of indicators that are meant to represent suspicious behavior by individuals attempting to circumvent the system. Given the vast number of indicators related to deception and the great variability in individuals’ leakage of deception cues (DePaulo et al., 2003; Frank et al., 2009), one can expect that the behavioral patterns of targets in these screening methods are highly varied and the frequency of the presence of each indicator is low. Regarding validation assessment, classification accuracy of the method as a whole is minimally affected by varied and rare signals. At the instrument level, the method’s accuracy would be calculated for individuals identified for any reason, examined as an aggregate; therefore, the specific signal is not an issue. However, if there is interest in understanding item-level classification accuracy; that is, the accuracy of each different type of signal in the prediction of outcomes, then it would take a great deal of data collection before a sufficient sample size is achieved for reliable calculations of item-level classification accuracy. Reliability assessments are also more challenging with rare and varied signals. As described above, reliability assessments are concerned with human judgments of whether the selection method items, or signals, are present or absent. This applies to both inter-rater and test-retest reliability. Reliability assessment is best done with real, operational data, rather than simulated data. However, to obtain enough examples of each item, a great deal of data would have to be collected to conduct a reliability assessment. This issue is important to consider in designing the validation assessment. However, the specific approach to examining system inputs depends on the selection method and the goal of the validation assessment. For example, if the researcher is primarily interested in criterion-related validity at the level of general decisionmaking, then this issue is not a primary concern, as noted above. However, if there is interest in the function of individual signals, then one must consider how to collect sufficient instances of each signal in order to conduct reliability and validity examinations.

206

T.E. Costigan

Conclusions In summary, instrument validity is an important first step in program evaluation. While these methods originated in education, psychology, and medicine, these are completely relevant to the domains of criminal justice, national security, and counterterrorism. This is because it is important to understand the extent to which any instrument or method is collecting the necessary data to examine outcomes in a systematic, reliable, and valid way. Once instrument validity is established, a broader program evaluation can be conducted that asks questions about the implementation and impact of the program and whether it achieves its goals. Conducting validation assessments for counterterrorism strategies, especially screening programs, serves several purposes. First, such analyses will help to develop a scientific evidence base in the area. Knowing what works from an instrument validity standpoint is extremely important to help policymakers with decisions about future directions in counterterrorism strategies, as well as help the public understand the benefit of such screening requirements. In addition, because this type of analysis is new to the field, completed assessments will help to develop a framework for best practices in the area. Methods of validation assessment include systematic examinations of criterion-, construct-, and content-related validity, as well as inter-rater, test-retest, and internal consistency reliability. A researcher must determine which of these elements of instrument validity are appropriate to their analysis, and this depends on the design and goals of the screening instrument under investigation. In developing the validation assessment, the researcher should consider several important points. These were introduced throughout this chapter and are summarized here. First, in all cases, an examination of criterion-related validity is required to determine the instrument’s classification accuracy. That is, the first examination in a validation assessment is to determine the extent to which the instrument can correctly select targets and nontargets. If criterion-related validity is established, then the screening program has shown evidence for utility and the researcher can test other elements of reliability and validity to further examine instrument validity. Second, experimental validity is important in developing the methodology for a validation assessment. Given the applied nature of instrument validity research, it is most important to have strong external validity; however, maintaining internal validity is also crucial to the analyses. To establish external validity, such assessments must be conducted in the field; that is, where the program is in operation. This requires buy-in from program stakeholders and staff who are involved in daily operations. Security personnel will likely collect the data and/or make judgments about test results. In order to maintain internal validity in such a study, the researcher should make efforts to develop a standardized protocol for data collection that has good experimental controls while also taking into account real-world security issues. For example, the researcher would need to build into their protocol a plan if security personnel need to stray from the data collection due to a serious system threat or security issue. Working closely with operational staff to develop and implement the field research will bring about the best success in such a study.

9 Examining Instrument Validity

207

Another way to increase internal validity is to develop well-defined outcome metrics. This holds for validation assessment as well as for program evaluation. For national security screening programs, the target outcome likely won’t be “terrorist” or “terrorist activity.” Rather, a given screening instrument may be interested in an outcome that represents a wider range of high-risk individuals, such as aviation security threats or individuals in possession of prohibited items. This broader outcome may be the actual class of targets that the instrument aims to identify or it may serve as a proxy to a more specific, but lower frequency outcome. Regardless of the outcome, a precise operational definition of the outcome is needed to design a rigorous study. While the researcher is concerned with developing a clear and measurable operational definition for the outcome, they should also work with program stakeholders to assure that the selected outcome and its operational definition is relevant to the instrument under evaluation. Finally, in order to establish criterion-related validity, classification accuracy metrics for the negative cases (i.e., those not selected by the screening instrument) should be calculated as accurately as possible. To do so, the population base rate for the outcome of interest must be determined. If this value is not known, then a sampling study should be conducted to estimate it. Implementation of a sampling study is quite involved, requiring the screening of a large number of cases, regardless of whether these individuals would have met criteria on the screening instrument. Thus, support from program stakeholders will be extremely important to conduct this study. In addition, because there will be error in this important estimate, a researcher should conduct an examination of the interval associated with this point estimate, using CIs around metrics and conducting a post hoc sensitivity analysis. No screening program is expected to be 100% accurate in discriminating targets from nontargets. After all, screeners are only designed to select cases for further assessment. Nonetheless, there needs to be solid scientific evidence of a screening program’s instrument validity in order to justify its use.

References American Educational Research Association, American Psychological Association, & National Council on Measurement in Education. (1999). Standards for educational and psychological testing. Washington: American Educational Research Association, American Psychological Association, & National Council on Measurement in Education. Caldeira, J. D. (1980). Parametric assumptions of some “nonparametric” measures of sensory efficiency. Human Factors, 22(1), 119–120. Cohen, J. (1960). A coefficient of agreement for nominal scales. Educational and Psychological Measurement, 20(1), 37–46. Cook, T. D., & Campbell, D. T. (1979). Quasi experimentation: Design and analytical issues for field settings. Chicago: Rand McNally. Cornfield, J. (1951). A method of estimating comparable rates from clinical data. Journal of the National Cancer Institute, 11, 1269–1275. Craig, A. (1979). Nonparametric measures of sensory efficiency for sustained monitoring tasks. Human Factors, 21(1), 69–78.

208

T.E. Costigan

Crocker, L., & Algina, J. (1986). Introduction to classical and modern test theory. New York: Holt, Rinehart, and Winston. Davies, H. T. O., Crombie, I. K., & Tavakoli, M. (1998). When can odds ratios mislead? British Medical Journal, 316, 989–991. DePaulo, B. M., Lindsay, J. J., Malone, B. E., Muhlenbruck, L., Charlton, K., & Cooper, H. (2003). Cues to deception. Psychological Bulletin, 129, 74–118. Edwards, A. W. F. (1963). The measure of association in a 2 × 2 table. Journal of the Royal Statistical Society, Series A, 126(1), 109–114. Farrington, D. P. (2003). Methodological quality standards for evaluation research. The Annals of the American Academy of Political and Social Science, 587(1), 49–68. Frank, M. G., Maccario, C. J., & Govindaraju, V. (2009). Behavior and security. In P. Seidenstat & F. X. Splane (Eds.), Protecting airline passengers in the age of terrorism. Santa Barbara: Praeger. Grant, J. A. (1974). Quantitative evaluation of a screening program. American Journal of Public Health, 64(1), 66–71. Green, D. M., & Swets, J. A. (1966). Signal detection theory and psychophysics. New York: Wiley. (Reprinted by Peninsula Publishing, Los Altos, 1988). LaFree, G. (2011). Generating terrorism event data bases: Results from the global terrorism database, 1970 to 2008. In C. Lum & L. W. Kennedy (Eds.), Evidence-based counterterrorism policy. New York: Springer. Lawshe, C. H. (1975). A quantitative approach to content validity. Personnel Psychology, 28, 563–575. Lum, C., Kennedy, L. W., & Sherley, A. J. (2006). Are counter-terrorism strategies effective?: The results of the Campbell systematic reviews on counter-terrorism evaluation research. Journal of Experimental Criminology, 2(4), 489–516. National Research Council. (2003). The polygraph and lie detection. Washington: National Academies Press. Norman, D. A. (1964). A comparison of data obtained under different false-alarm rates. Psychological Science, 1, 125–126. Nunnally, J. C., & Bernstein, I. H. (1994). Psychometric theory (3rd ed.). New York: McGraw Hill. Pedhazur, E. J., & Schmelkin, L. P. (1991). Measurement, design, and analysis: An integrated approach. Hillsdale: Lawrence Erlbaum Associates. Sackett, P. R., & Decker, P. J. (1979). Detection of deception in the employment context: A review and critical analysis. Personnel Psychology, 32, 487–506. Shadish, W. R., Cook, T. D., & Campbell, D. T. (2002). Experimental and quasi-experimental designs for generalized causal inference. Boston: Houghton Mifflin. Sheehan, I. S. (2011). Assessing and comparing data sources for terrorism research. In C. Lum & L. W. Kennedy (Eds.), Evidence-based counterterrorism policy. New York: Springer. Sistrom, C. L., & Garvan, C. W. (2004). Proportions, odds, and risk. Radiology, 230(1), 12–19. Swets, J. A. (1996). Signal detection theory and ROC analysis in psychology and diagnostics: Collected papers. Mahwah: Lawrence Erlbaum Associates. U.S. Department of Homeland Security. (2011). Privacy impact assessment update for TSA advanced imaging technology. Washington: U.S. Government Printing Office. U.S. Transportation Security Administration. (2011). Retrieved April 13, 2011, from http://www. tsa.gov/approach/tech/ait/how_it_works.shtm.

Chapter 10

Translational Criminology: Using Existing Evidence for Assessing TSA’s Comprehensive Security Strategy at Airports Cynthia Lum, Charlotte Gill, Breanne Cave, Julie Hibdon, and David Weisburd

In September, 2011, before this book was about to go to press, the Transportation Security Agency (TSA) ordered the authors of this study and the editors of this volume to redact specific information from this chapter. The TSA’s Office of Secured Sensitive Information (SSI), and the TSA’s Playbook Program Office determined some material, and almost all of the findings from this study to be “Sensitive, Secured Information”, subsequent to 49 Code of Federal Regulations Part 1520. The determination of what is “secured sensitive information” is at the discretion of the TSA’s SSI’s Office and the program manager of TSA’s Comprehensive Security Strategy at Airports program. The redaction is indicated by the blackening of text. Although specific redactions cannot be discussed (as they are now deemed “SSI”), two types of redactions appeared most frequent. These were: (1) any findings, numbers, percentages, words that described numbers, or empirical information, no matter how broad; and (2) phrases that named security measures, including those that the public are well aware. It should also be noted that the publication of this chapter and subsequently this book, could not continue without the authors conceding to TSA’s redactions. Further questions about this process can be directed to the Transportation Safety Administration’s Security Sensitive Information Office, TSA-31 OSC, 601 South 12th Street, Arlington, VA 20598.

Introduction Transportation security at the nation’s airports has become a major priority in United States homeland security since the events of September 11, 2001. The establishment of the Transportation Security Administration (TSA), the advancement of new C. Lum (*) • C. Gill • B. Cave • J. Hibdon • D. Weisburd Center for Evidence-Based Crime Policy, Department of Criminology, Law and Society, George Mason University, Fairfax, VA, USA e-mail: [email protected] C. Lum and L.W. Kennedy (eds.), Evidence-Based Counterterrorism Policy, Springer Series on Evidence-Based Crime Policy 3, DOI 10.1007/978-1-4614-0953-3_10, © Springer Science+Business Media, LLC 2012

209

210

C. Lum et al.

scanning and detection technologies, the increased use and sharing of information, and greater coordination between various law enforcement, security, and civilian agencies, all emphasize the importance of airport security. Even prior to 9/11, airports have been the focus of more security efforts than any other transportation system. Their size, complexity, use, and multiple functions have presented opportunities for a range of criminal and terrorist activities and consequentially, crime-prevention efforts. One of the recent developments in airport security has been the call for a more coordinated security strategy. In 2009, the TSA revised and reimplemented its Comprehensive Strategy to Security at Airports – also known as “the Playbook”1 – to supplement and coordinate existing security at airports. The Playbook is part of TSA’s “layers of security” concept,2 which seeks to provide a holistic security apparatus for air transportation. It consists of a myriad of situational tactics and strategies which span various domains, sectors, and environs of the airport and are designed to prevent, detect, deter, and protect against crime. The purpose of the Playbook, as described by TSA, is “to create a transportation security system that increases unpredictability, thereby frustrating terrorist plans and potentially deterring attacks” (U.S. Transportation Security Administration, 2010). To date, there has been no independent assessment of either the implementation or effectiveness of the Playbook, or of airport security as a whole. Yet, given the importance of and increased attention toward airport security, it is surprising that airports remain relatively understudied in social science and evaluation research. Research can play a crucial role in providing objective assessments of the nature and effectiveness of airport security, which encompasses prevention, control, and deterrence of many types of crime, from the most “ordinary” to the most severe. Both the US Government Accountability Office (see, e.g., U.S. Government Accountability Office, 2007, 2009, 2010) and the Transportation Research Board of the National Academies (2003) have formally called for more evaluation, assessment, and research cooperation in airport security. Thus, to advance research in this area, the Department of Homeland Security (DHS) Science and Technology Directorate, at the request of the TSA, tasked the Center for Evidence-Based Crime Policy (CEBCP) at George Mason University (GMU) with carrying out a comprehensive four-phase assessment of the Playbook. Here, we provide the results of the first phase of this project, which uses “translational criminology”3 to preliminarily assess the evidence base of the prevention and deterrence mechanisms of the Playbook using existing knowledge from crime prevention. While recognizing the importance of empirical assessments of the Playbook

1

This paper takes a general approach to discussing the Playbook. Per the CEBCP’s confidentiality agreements with DHS and TSA, we will not discuss specific Playbook content. 2 See http://www.tsa.gov/what_we_do/layers/index.shtm. 3 This term arises from a lecture given by the director of the National Institute of Justice, John Laub, at the University of Pennsylvania in April 2009, to describe using research to shape policy and practice. We use it here in the spirit of Director Laub’s lecture, but more specifically focus on applying existing research of one type of policy to similar policies when evaluation research is unavailable.

10

Translational Criminology: Assessing Airport Security

211

and other aspects of airport security, evaluations of airport security need not start from scratch. The translational criminological approach posits that existing evaluations and theory enable a preliminary assessment of whether prevention mechanisms within the Playbook are consistent with what is known about crime prevention in a more general manner (see Lum & Koper, 2011). This assessment will become the research toehold by which experimental evaluations and hypothesizing about possible effects of interventions will be explored. It is uncertain whether the nature of offending at airports differs so significantly from other types of offending that criminological research is not applicable. A comparison of the Playbook against similar research may prove useful in hypothesizing about the fruitfulness of such research.

Airport Security and TSA’s Playbook Strategy Aviation assets are both visible and symbolic, making them an attractive target for terrorism. The images and aftermath of 9/11 and subsequent incidents and attempts of violence have solidified air transportation attacks as “one of the most deadly and spectacular tactics employed by terrorists” (Asal, Rethemeyer, Bellandi, Legault, & Tynes, 2010, p. 2). Such attacks are not confined to hijackings and in-air violence. In 2002, a limousine driver opened fire in the unsecure, publicly accessible ticketing area of El Al Airlines at Los Angeles International Airport. More recently, on January 24, 2011, a bomber targeted the unsecured reception area outside of customs at Domodevo Airport in Moscow, Russia (Englund & Lally, 2011). These security breaches, as well as crime more generally at airports, have led airport authorities and DHS to prioritize security in airports and air transportation (U.S. Department of Homeland Security, 2008). Most recently, the establishment of the TSA by the Aviation and Transportation Security Act on November 19, 2001, emphasized this priority. The new agency took over security and other regulatory responsibilities previously vested in the Federal Aviation Administration (FAA) of the Department of Transportation (DOT), and was also given new duties. Its original mandate was to provide for the “security for all modes of transportation; recruit, assess, hire, train, and deploy Security Officers for 450 commercial airports from Guam to Alaska in 12 months; and provide 100% screening of all checked luggage for explosives by December 31, 2002” (U.S. Transportation Security Administration, 2011). In November 2003, governance of TSA moved from the DOT to DHS, as part of a national effort to centralize homeland security functions. Towards these security goals, the TSA piloted the Playbook in 2008 and then developed, revised, and implemented it more fully in 2009. The Playbook reflects a comprehensive approach to the physical security of airports, consisting of a myriad of situational tactics and strategies (or “plays”) that span various domains of airport security. They can include increasing surveillance and screening of passengers, employees, and airport personnel at different locations of the airport; monitoring luggage, planes, and other equipment; restricting or blocking access to secured

212

C. Lum et al.

locations; or redirecting passenger or vehicular traffic. Types of activities are generally discussed in Fig. 10.10 further in this chapter.4 Many of the plays within the Playbook are not new to airport security, as the purpose of the Playbook was to add to the overall TSA goal of strengthening “layers of security” by capitalizing on existing security tactics. Plays mimic concepts similar to those found in situational crime-prevention (Clarke, 1983, 1995, 1997; Cornish & Clarke, 1987, 2003; Eck, 2002) and deterrence research (for reviews, see Durlauf & Nagin, 2011; Nagin, 1998). Through physical and environmental prevention and deterrence mechanisms, plays are intended to stop and discourage offenders, provide increased guardianship, increase the perception of a risk of apprehension, and also strengthen the vigilance of passengers and other targets. They include a wide array of prevention mechanisms such as hardening targets, blocking opportunities, reducing vulnerabilities, controlling access, or increasing surveillance. At the time of this writing, the plays are organized into three playbooks, totaling approximately 164 deployment security activities, and a handful of combined plays. The three sub-books are the Federal Security Director’s Playbook (hereafter “FSD Playbook”), the Nonrandomized (NR) Playbook (hereafter “NR Playbook”), and the Headquarters (HQ) Playbook (hereafter “HQ Playbook”). To increase unpredictability of security measures, TSA officials use a “randomizer,” or computer program that randomly selects a group of plays from some of the plays for implementation at any given time. Local TSA personnel can then supplement sets of randomly selected plays with other existing plays or new plays a locale may develop. The combination of random selection and discretion in choosing plays is designed to both thwart attempted security breaches by reducing the ability of offenders to anticipate TSA’s actions, and allow officials to take account of local security conditions in deciding how to deploy their assets. The TSA also believes that the Playbook promotes interagency cooperation, and so increases the ability of units to carry out specialized tasks. TSA personnel emphasize coordination as the key to successfully establishing airport security generally and to deploying the Playbook strategy specifically. While most plays are conducted solely by TSA staff, some require coordination and cooperation between the TSA and other airport groups such as the airport authority, the public, employees inside of the airport and within airlines, and local businesses. Although it is not mandatory for all airports to subscribe to the Playbook, many TSA units across the nation’s airports know of and use the Playbook in various ways.5 On its face, the Playbook appears to be rational and useful. However, whether this is true is important to examine. Given that we know that many interventions and programs in crime prevention once thought to be effective are not, or can even 4

The Playbook was provided by the TSA under a cooperative agreement, protected by a nondisclosure agreement for the purposes of this study. Thus, no specific play can be discussed in detail because the Playbook is classified as “Sensitive Security Information” (see http://www.tsa.gov/ assets/pdf/stakeholder_brochure.pdf). 5 A detailed analysis of Playbook implementation at selected airports is the focus of later phases of this study.

10

Translational Criminology: Assessing Airport Security

213

increase harm (McCord, 2003; Sherman, Farrington, Welsh, & MacKenzie, 2002), it is important for both fiscal and prevention purposes to examine Playbook effectiveness. Four key questions of evidence-based crime policy apply: (1) Is there a theoretical, scientific, and research evidence base for the Playbook? (2) Is the deployment of the Playbook effective in maintaining security at airports? (3) Following the second question, what is the reality of implementation of the Playbook across airports? (4) Can airport security be scientifically evaluated, and are there measurable outcomes we can use for evaluation? This article focuses on the first of these issues, with the goal of preliminarily assessing the Playbook for the continued research by the CEBCP team.

Relevant Research Perspectives for Assessing the Playbook Currently, only one counterterrorism measure has been shown to be effective through systematic evaluation: metal detectors in airports. In a Campbell Collaboration6 systematic review of counterterrorism interventions, Lum, Kennedy, and Sherley (2006) summarized 16 outcomes of metal detector evaluations across five research studies that used interrupted time series methods (Cauley & Im, 1988; Enders & Sandler, 1993, 2000; Enders, Sandler, & Cauley, 1990; Landes, 1978). Overall, these studies showed that passenger screening using metal detectors deters and reduces hijackings over time, although two studies (Cauley & Im, 1988; Enders & Sandler, 1993) found possible substitution effects of terrorist activity to other types of terrorism. Dugan, LaFree, and Piquero (2005), using the Global Terrorism Database from the START Center (discussed by LaFree, 2011; Sheehan, 2011),7 also discovered a deterrent effect of metal detectors when the certainty of apprehension was increased. However, there are many other ways to secure airports that extend beyond metal detectors (and many more ways they could be evaluated). These include no-fly lists and prescreening, random searching and screening of both passengers and employees, general visibility of law enforcement personnel, canine units trained to detect explosives or contraband, or other security measures used to block or restrict access into secure areas. Many of these exist as plays within the Playbook. Research on these interventions, unlike other crime-prevention measures, is almost nonexistent. Lum recently reported to the National Research Council of the National Academies that evaluative studies in policing outnumber those on counterterrorism (at least those that are publicly available) more than tenfold, and she found no rigorous experimental or non-time series quasi-experimental evaluations of airport security 6

The Campbell Collaboration (see http://www.campbellcollaboration.org/) is an internationally recognized research network that produces systematic reviews and meta-analyses on the effects of interventions in crime and justice, education, social welfare, and international development. It also examines the methods used for evaluation. 7 http://www.start.umd.edu/start/.

214

C. Lum et al.

strategies compared to approximately 25 randomized controlled experimental trials and 10 rigorous quasi-experiments of policing strategies at the time of writing (National Research Council, 2010; see also Lum et al., 2006; Weisburd, Feucht, Hakimi, Mock, & Perry, 2009). Yet, the consequences of interventions with regard to cost, safety, passenger and employee satisfaction, and civil liberties make evaluations imperative. Is there previous theoretical and evaluation research that can help evaluate tactics within the Playbook? Terrorism is sometimes viewed as a unique problem, given its ideological nature, political or religious motivations, and the involvement of social institutions beyond law enforcement and criminal justice agencies (e.g., Deflem, 2004; LaFree & Dugan, 2004; Mythen & Walklate, 2006; Rosenfeld, 2004). However, even with the dearth of empirical research specifically addressing airport security, there are other sources of research evidence from which we can preliminarily assess the Playbook. Acts of, and preparations for, terrorism involve wellunderstood illegal activities that at least in theory can be interpreted through a criminal justice lens and potentially be deterred or blocked (Lum & Koper, 2011). Lessons, concepts, and ideas from theories and evaluation of crime and crime prevention may be useful, specifically rational choice, opportunities, and routine activities theory, as well as research on situational crime prevention, deterrence and unpredictability in deployment, and interagency cooperation. Each of these concepts is now discussed.

Rational Choice, Opportunity, and Routine Activities as a Theoretical Framework for Evaluating the Playbook A rational choice perspective of offending can be relevant to terrorism and counterterrorism and might help us hypothesize about the potential effects the Playbook can have on motivated offenders. Rational choice theories propose that offenders, given certain constraints, are decision makers who “respond selectively to… the opportunities, costs and benefits” associated with specific crimes (Cornish & Clarke, 1987, p. 934). It is reasonable to hypothesize that although acts of terrorism may seem irrational and rare, terrorism can reflect elastic deterrence structures (see Durlauf & Nagin, 2011) that are responsive to interventions. Indeed, a number of scholars have argued that rational choice theory is relevant to explaining terrorist offending (Berrebi, 2009; Clarke & Newman, 2006; Crenshaw, 1990; Dugan et al., 2005; Jackson, 2009). The ability to deter offenders at airports by increasing the cost of offending (risk of apprehension) has some empirical support. Dugan et al. (2005) found that airplane hijackers respond rationally to variation in opportunity. New hijacking attempts were less likely when the use of metal detectors and law enforcement tactics at screening checkpoints increased the certainty of apprehension. This deterrent effect was not found for criminalization policies on terrorist-related hijackings (although it was found on non-terrorist-related hijackings). Their work suggests that

10

Translational Criminology: Assessing Airport Security

215

prevention mechanisms in airports should focus on identifying and eliminating opportunities created by security and infrastructure weaknesses, thereby increasing detection, prevention, and deterrence while minimizing the possible displacement or substitution effects suggested by Cauley and Im (1988). Although a rational choice perspective provides an umbrella explanation for an individual’s decision to offend, we have to consider why rational decisions to commit crime are often concentrated in certain situations, and at specific places and times. Additional criminological perspectives, such as routine activities (Cohen & Felson, 1979; Felson, 1994) and opportunity theories (see Clarke, 1980, 1983, 1992, 1995, 1997) are helpful. Clarke argued that an opportunity to commit crime must be present prior to the occurrence of a rational decision. Opportunities arise, as Cohen and Felson (1979) argue, when motivated offenders, suitable targets, and a lack of capable guardianship converge in space and time. Felson (1994) described these and other situational attributes as the “chemistry for crime,” which when present in the right environment could beget criminal events. Such convergences are not random, and result from daily routines, or “rhythms,” “tempos,” and “timings” of daily activities (See Cohen & Felson, 1979, p. 590). Further, research on the “criminology of place” (Sherman, Gartin, & Buerger, 1989; Weisburd, 2002; Weisburd, Groff, & Yang, 2009) views places as important attractors of these opportunities (i.e., high-risk locations) and daily routines. A strong body of research suggests that crime opportunities are not spread randomly across place but are highly concentrated. For example, research has found that half of all crime in a city occurs in just 4–6% of places (Sherman et al., 1989; Weisburd, Bushway, Lum, & Yang, 2004), and in some cases, about 1% of chronic hot spot street segments have been found to consistently produce more than 20% of crime (Weisburd et al., 2004). This research emphasizes the importance of identifying “vulnerable” locations within larger administrative or geographic units as targets of intervention. It also identifies the importance of not only crime opportunities in creating vulnerability, but also social and structural characteristics of places (e.g., the social fabric of locations created by the types of people who live or work there) that make them more resistant to criminal activity (Weisburd et al., 2009). In total, rational choice, opportunity, routine activities, and criminology of place theories all provide theoretical context for hypothesizing about the place-focused crime-prevention and deterrence strategies found in the Playbook.

Situational and Place-Based Prevention and Deterrence The interplay of rationality, routine activities, and opportunity structures explains the nonrandom concentration of crime at specific places, times, and situations, and has direct implications for crime prevention and deterrence (Clarke, 1980; Felson, 1994; Weisburd, 2002, 2008). Within these frameworks, offending can be predictable and

216

C. Lum et al.

therefore preventable by altering the potential for the convergence of a motivated offender, lack of guardian, and suitable target. A wealth of prevention research has developed around these ideas, most notably situational crime prevention (see Clarke, 1997; Cornish & Clarke, 1987; Eck, 1997) and place-based, hot spots policing (Braga & Weisburd, 2010; Sherman & Weisburd, 1995; Weisburd, 2008). Clarke and Newman (2006) have extended their work in situational crime prevention to counterterrorism. Situational crime prevention and deterrence measures are actions, environmental changes, or other tactics that prevent crime by blocking opportunities or access to crime, hardening targets, strengthening potential victims against offenders, and increasing the risk of detection and apprehension by increasing guardianship. Although overlapping and related, prevention and deterrence measures are distinguished often by perspective: prevention refers to blocking opportunities so crime cannot occur, while deterrence measures often increase the risk of certainty, severity, or celerity of punishment and apprehension. Both can cause an alteration in the decision-making process of an individual to commit crime, and both prevention and deterrence mechanisms can be found in the same interventions (metal detectors are a good example). By changing the availability of opportunities offered by the environment, such interventions are intended to persuade the potential offender that an area is not conducive to committing a crime – the risks are too great and the rewards too modest to make the effort worthwhile. Examples of situational crime prevention measures include putting locks on doors, blocking off streets, reducing the maximum number of drinks at a bar, increasing street lighting, marking property, or using video surveillance. Cornish and Clarke (2003) detail 25 techniques (reproduced in Fig. 10.1) that could encompass a wide variety of measures. Many of the plays in the Playbook mimic these techniques through screening, searching, surveillance, access blocking, and vulnerability-reducing measures. Extensive theoretical development and testing of situational crime prevention approaches have already been undertaken within the rational choice, routine activity, deterrence, and opportunity frameworks (see Clarke, 1983, 1992, 1995; Clarke & Newman, 2006; Cornish & Clarke, 1986; Eck, 2002; Weisburd, 1997). A large body of literature indicates that target hardening and access control can be promising ways to block motivated offenders and reduce opportunities for crime (Clarke, 1992, 1995; Eck, 2002; Newman, 1972). Further, in a systematic review of the collateral effects of situational crime prevention interventions, Guerette and Bowers (2009) found that crime displacement – the shift of criminal activity to other locations, crime types, offenders, or strategies as a result of an intervention – occurs less frequently than believed, and is seldom total if it does (see also Braga, 2005, 2007; Clarke & Weisburd, 1994; Weisburd, 2002; Weisburd et al., 2006). Whether this is the case at airports remains to be seen. Researchers have also found that the most effective situational prevention strategies are those tailored to the targeted offense and the context (Clarke, 1997). For example, although Lum et al. (2006) found that metal detectors at airports were effective in reducing attacks at airports, more general situational measures meant to

10

Translational Criminology: Assessing Airport Security

Increase the Effort 1. Harden Targets

217

Increase the Risks 6. Extend guardianship

Reduce the Rewards 11. Conceal targets

Reduce Provocations 16. Reduce frustration and stress

Cocooning; neighbourhood watch

gender-neutral phone directories; off-street parking

efficient queuing; soothing lighting

2. Control access to facilities

7. Assist natural surveillance

12. Remove targets

17. Avoid disputes

alley-gating; entry phones

improved street lighting; neighbourhood watch hotlines

3. Screen exits

8. Reduce anonymity

13. Identify property

taxi driver IDs; ‘how’s my driving?’ signs

property marking; controls on violent vehicle licensing porn; prohibit paedophiles working with children

roadside speed display signs; ‘shoplifting is stealing’

4. Deflect offenders

9. Utilise place managers

14. Disrupt markets

19. Neutralise peer pressure

24. Assist compliance

street closures in red light district; separate toilets for women

train employees to checks on pawn prevent crime; brokers; licensed support whistle street vendors blowers

‘idiots drink and drive;’ ‘it’s ok to say no’

litter bins; public lavatories

5. Control tools/ weapons

10. Strengthen formal surveillance

20. Discourage imitation

25. Control drugs/ alcohol

rapid vandalism repair; V-chips in TVs

breathalysers in pubs; alcohol-free events

immobilisers in cars; anti-robbery screens

tickets needed; electronic tags for libraries

toughened beer glasses; photos on speed cameras; CCTV in town credit cards centres

Remove the Excuses 21. Set rules rental agreements; hotel registration

22. Post instructions

fixed cab fares; removable car reduce crowding in ‘No parking;’ radios; pre-paid ‘Private property’ pubs public phone cards

15. Deny benefits ink merchandise tags; graffiti cleaning

18. Reduce emotional arousal

23. Alert conscience

Fig. 10.1 Twenty-five techniques of situational crime prevention (adapted from Cornish & Clarke, 2003)

fortify embassies and protect diplomats against a range of potential attacks were not similarly effective. This might be due to different motivations of these two types of crimes, which can also be tied into environmental settings (see Clarke, 1997). The effectiveness of greater specificity has been found in policing interventions as well (see Lum, Koper, & Telep, 2011; Weisburd & Eck, 2004). While situational crime prevention approaches can be implemented at very specific places, there is also strong research showing that law enforcement deterrence and guardianship efforts directed at hot spots of crime and risky places can have significant crime-reduction effects (Weisburd, 1997, 2002, 2008). Beginning with the Minneapolis hot spots experiment (Sherman & Weisburd, 1995), numerous studies have shown repeatedly that directed patrol can deter would-be offenders (for

218

C. Lum et al.

reviews, see Braga, 2005, 2007; Lum et al., 2011; National Research Council, 2004; Sherman, 1997; Sherman & Eck, 2002; Weisburd & Eck, 2004). Further, Durlauf and Nagin (2011) argue that certainty-based deterrence approaches like hot spot policing seem more effective than “severity-enhancing” deterrence approaches in which threats of severe punishment are used to create a deterrent effect. Both situational crime prevention and place-based deterrence approaches are used throughout the Playbook, which is highly oriented to certain places at the airport. The question is whether the locations chosen are indeed high risk, and whether the implementation of the plays at those locations leads to a measurable effect.

Deterrence and Unpredictability One key feature that TSA highlights within the Playbook is the element of unpredictability of the implementation of specific interventions. The Playbook attempts to achieve this through a computer-based application that allows personnel to randomly select sets of plays and supplement them with other nonrandomly selected tactics to be implemented during a period of time. Ideally, potential offenders would be deterred because it is difficult for them to anticipate the location, timing, and nature of security measures. Under a rational choice perspective, this uncertainty of the opportunity structure could prevent them from making an informed decision about risk and rewards, increasing the risks (and costs) of their offending, and consequently discourage them from crime. The idea that unpredictability might create a perceived “omnipresence” that can then deter crime has parallels in both the traditional preventive police patrol research and also the hot spots studies mentioned above. Before the hot spots studies, it was argued that spreading police randomly and widely across the city would deter potential offenders (e.g., Repetto, 1976). However, a large field study in the 1970s conducted by the Police Foundation in Kansas City (Kelling, Pate, Dieckman, & Brown, 1974) appeared to counter this argument, showing that increases or decreases in preventive patrol in large areas had no measurable impact on crime or on citizens’ feelings of security and safety. Sherman and Weisburd (1995), in the Minneapolis Hot Spots experiment, argued that random preventive patrol across large areas was not likely to be successful because of the relatively small dosage that most city police were able to deploy and the fact that most crime in a city is concentrated in very specific places (“hot spots”). They contend that patrol should not be spread randomly but rather be focused on high-risk places. Further the research indicates that unpredictability in security and law enforcement efforts can be effective when there is an initial targeting of high-risk places (or even people). The Koper Curve principle (Koper, 1995), which indicates that directed patrol can have diminished residual deterrent effects after a peak time of direct patrol, supports this notion. Moving patrol from hot spot to hot spot in an unpredictable manner for shorter periods of time may be more effective in reducing crime rates in a city (or place) than keeping officers at a single hot spot all day. Plays

10

Translational Criminology: Assessing Airport Security

219

within the Playbook, and the Playbook more generally, have the potential for both targeted, unpredictable approaches and nontargeted, unpredictable approaches. Which is more efficacious in an airport setting is a question for evaluation.

Interagency Cooperation Finally, the Playbook emphasizes the importance of interagency cooperation for preventing terrorism and crime in many of its plays. Interagency cooperation is also a theme in existing crime-prevention and counterterrorism literature. In law enforcement, multiagency cooperation grew primarily from policing models such as community-oriented and problem-oriented policing, as well as “all-hazards” approaches to disaster response. These models of security and policing center around general notions of inclusiveness, democratic decision-making, tailored and multifaceted problem-solving, and “thinking outside of the box,” all of which invoke principles of cooperation across stakeholders toward common goals. There is growing evidence in the research literature that interagency cooperation may be a promising part of crime reduction and deterrence, especially in law enforcement. Both Weisburd and Eck (2004) and Lum et al. (2011) discovered that tailored policing approaches, which often employ third-party partnerships, are effective in reducing crime (see also Mazerolle & Ransley, 2005; Weisburd, Telep, Hinkle, & Eck, 2010). Further, Riley and Hoffman (1995) and Lum, Haberfeld, Fachner, and Lieberman (2009) found a strong post-9/11 trend in law enforcement to pursue and support information sharing across agencies and interagency collaboration for purposes of establishing homeland security systems (see also Carter & Chermak, 2011). In our discussions with TSA managers, they emphasize that coordination and collaboration must occur for airport security to be effective. This includes cooperation between TSA officers and managers, local police departments both in and outside of the airport, the airport authority, those doing business in and around the airport, and passengers themselves. Whether the Playbook achieves collaboration that proves to be effective and efficient in providing security is therefore an important question in assessing the Playbook. In conclusion, the overall conceptual framework of the Playbook generally reflects criminological theory as well as research in various areas of crime prevention and deterrence. Thus, to use a translational criminological approach to preliminarily assess airport security measures is not farfetched. In the next section, we more specifically analyze all of the 164 plays within the Playbook against the evaluation research within these research areas. To do this, we dig into the prevention and deterrence mechanisms of the plays in the Playbook, and compare them to the evidence on similar mechanisms of prevention in existing crime prevention research. Although the constraints of the nondisclosure agreements of this project prohibit us from discussing specific plays, we present the general tendencies of the plays for purposes of thinking about advancing airport security.

220

C. Lum et al.

Translational Criminology: A Preliminary Assessment of Whether the Playbook Is Evidence-Based by Adapting the Evidence-Based Crime-Prevention Matrix Classification Approach

Z:

R

EA C

TI

LE VE

L

VE

O

F

PR

PR

O

O

AC

TI VE

AC TI VI

TY

H

IG H

LY

PR

O

.

FOCUSED GENERAL

Y: SPECIFICITY OF PREVENTION MECHANISM

Given the similarities of the theoretical contexts of airport security and criminology, the purpose of this analysis is to assess whether the Playbook fits the current evidence base of crime prevention. This exercise is useful in the absence of experimental evaluations on airport security, and can provide a general sense of whether the Playbook reflects what we know works in security in a more general manner. To accomplish this, the first step is to categorize plays in such a way that allows us to make generalizations from them that apply to the broader crime-prevention literature. A categorization approach for evidence-based policy has been developed in the area of policing by Lum, Koper, and Telep called the Evidence-Based Policing Matrix.© In 2007, Lum and Koper developed a visual organization scheme to systematically display evaluation evidence called the Crime Prevention Matrix (see Lum & Koper, 2011). In 2008, Lum, Koper, and Telep extended this concept to the policing literature, creating the Evidence-Based Policing Matrix (see Lum, 2009; Lum, Koper & Telep, 2009, 2011) as shown in Fig. 10.2. The purpose of this was to create a tool by which existing research knowledge could be easily summa-

X: TYPE OR SCOPE OF TARGET

Significant Backfire

Non-Significant Finding

Mixed Results

Significant /Effective

Fig. 10.2 The evidence-based policing matrix (Lum, Koper & Telep, 2009, 2011)

10

Translational Criminology: Assessing Airport Security

221

rized according to common attributes of interventions (target of the intervention, level of specificity and proactivity of the intervention). By mapping all moderate to highly rigorous police evaluation research according to common dimensions (shown by each axis) of tested interventions, clustering patterns of effective tactics emerge. These clusters can be described in the areas of the matrix in which specific nodes on the X, Y, and Z axes intersect, creating generalizations of promising interventions. The purpose of creating generalizations from such mapping was to create a tool that could eventually be used to assess tactics that have yet to be evaluated. For example, if a police agency wanted to know if a neighborhood-level, proactive, and general intervention could be effective, the totality of the research in that area of the Matrix indicates that such an approach might be promising. We think the Matrix concept offers an opportunity to categorize the Playbook and assess its evidence base in a similar fashion. However, it is impossible to create a matrix of airport security measures given the lack of evaluations in this area. Further, a matrix of situational crime-prevention and deterrence measures has yet to be created and is well beyond the scope of this project. To approximate this approach, we carried out a “reverse matrix” exercise. Lum (2009) initially advised police agencies to map their own deployment tactics into the Matrix, and then compare their deployment clusters with what dimensions the research evidence patterned. In similar fashion, we developed a hypothetical airport security matrix, according to common characteristics developed from the situational crime-prevention, deterrence, and interagency coordination literature. We then mapped the plays into the airport security matrix, which allowed us to discern common tendencies across plays. Then, because there is a large amount of evaluation research on situational crime-prevention, we compared clustering of play tendencies along common dimensions to what we know about those dimensions from that literature. Although this approach does not substitute evaluation, it provides some understanding of how well the evidence supports Playbook activities.

Developing the Airport Security Matrix To develop a Matrix for airport security, we examined all 164 plays within the Playbook to identify similar characteristics with regard to prevention, deterrence, and interagency cooperation. The main Playbook volume, the Federal Security Director’s (FSD) Playbook, contains just over half (52%) of the total plays available to TSA and are organized by the location within the airport in which they are intended to be implemented. The FSD Playbook also comprises the set of plays that are subject to randomization. The NR Playbook contains plays supplemental to the FSD plays that can be selected at the discretion of TSA staff. These NR plays make up 12% of all plays and are organized according to purpose (covert operations, strategies designed to observe suspicious behavior, and responses to insider threats). Finally, the HQ Playbook contains plays intended for direct selection by TSA’s centralized operations (not security coordinators in specific airports) in response to a specific threat. This HQ

222

C. Lum et al.

Playbook also contains a variety of strategic plays. By “strategic”, we mean plays that are intended to be more long-term in nature, and that usually has a planning or proactive nature to them (for example, training exercises to practice responses to a potential situation). Thirty-six percent of the total plays are HQ plays. Like the Evidence-Based Policing Matrix methodology (see Lum et al., 2011), we began building a classification scheme by first examining a sample of plays in the context of common crime-prevention principles informed by the literature. The plays were then coded using several key features of each play, including the nature of the activity involved in the play, the classification of the play according to Cornish and Clarke’s 25-item situational crime-prevention typology (Fig. 10.1), and criminological theories reflected in the play, timing, target population, agencies involved, and the mechanism of prevention. All members of the research team were involved in this coding exercise, which fostered consensus about the nature of airport security. What emerged are three dimensions that we believe are critical to understanding the prevention activities central to the Playbook, as shown in our “Airport Security Matrix” (Fig. 10.3). The X-axis classifies the location at which a particular play is implemented, the Y-axis indicates the mechanism of the preventative strategy, as informed by situational crime-prevention scholars (see Cornish & Clarke, 2003), and the Z-axis indicates the level of interagency cooperation.

X-Axis: The Location or Place of the Playbook Intervention A common way to describe a crime-prevention tactic is the target of the prevention, usually a person, group, location, or place. Although airport counterterrorism measures do target specific persons and groups (i.e., no-fly lists, employees with criminal histories), the Playbook tactics primary are place-based. Indeed, TSA’s “layers of security” approach emphasizes implementing different tactics at different airport locations to create a secure environment. For our matrix, we delineated target locations, and thus the X-axis, moving from the outermost to innermost layers, closest to the airplane itself: External public airport areas: Places within the airport perimeter but outside the physical airport building and entry terminals, such as taxi areas, walkways, curb dropoff and pick-up locations, rental car parking lots, etc. Also places outside the airport perimeter, such as mass transit systems or streets that connect to airport entrances. Internal public airport areas: Publicly accessible areas of the interior of airport buildings, prior to security screening; for example, ticket counters and baggage claim areas. Screening areas: Perimeter check points for either passengers or employees (which may be the same or separate areas). These screening areas divide publicly accessible areas from secure areas. Secure public areas: Areas that can only be accessed after passing through a formal screening process. These areas are available to ticketed passengers and authorized and screened employees.

Translational Criminology: Assessing Airport Security

223

Fig. 10.3 The airport security matrix

10

224

C. Lum et al.

Secure employee areas: Areas that can only be accessed by employees after passing through a formal screening process (including vetting prior to employment and/or entry screening with an identification check). These areas may be inside buildings and outdoors in areas enclosed in a security perimeter, such as baggage handling and fueling areas. Gate: Places in the immediate vicinity of the gate area for a specific departing flight. Aircraft: The interior or immediate exterior of aircraft bodies.

Y-Axis: Mechanism of Prevention The Y-axis reflects the prevention mechanism used by a specific tactic. We derived the Y-axis categories using three common constructs of situational crime-prevention and opportunity theories: (1) to deter offenders by increasing their effort, (2) to increase general guardianship, and (3) to reduce the vulnerability of passengers or other targets. In many cases, plays could be classified into multiple categories on this dimension, given their theoretical overlap. For example, tactics intended to harden targets for deterrence could also be interpreted as reducing the vulnerability of a potential victim. Where there was overlap, we selected the primary mechanism of prevention based on the intended target of the tactics: Deter offenders/increase their effort: These involve plays that primarily focus on blocking offenders by increasing the effort they would have to use in order to succeed in a specific activity. Examples include keypad locks on secure doors, screening at entrances to secured areas, and hand swabs to test for explosive traces. Increase guardianship: These plays generally attempt to increase the level of general watchfulness and oversight to detect criminal activity. Broadly, this classification is used for plays that intend to increase the risk of being apprehended through increasing surveillance (e.g., directed security patrols, identification checks of employees, watching the airfield). Reduce vulnerability of passengers and other targets: These plays are designed to decrease the vulnerability of targets (both people and places) or to make criminal activity less worthwhile for the offenders by making the passengers and employees more alert or less available for potential victimization. Such plays are designed to protect people and locations even in the presence of a motivated offender. Examples include internal and external inspections of aircraft or audio warnings. Z-Axis: Interagency Cooperation and Reliance on Other Agencies The Z-axis of the matrix describes the extent to which TSA must rely on other agencies to implement plays. The Playbook and the TSA emphasize increased cooperation with agencies at the airport not under TSA jurisdiction. These might include local law enforcement responsible for security in nonscreening areas; the airport

10

Translational Criminology: Assessing Airport Security

225

authority and business managers of the airport; vendors; and employees of the airport, airline, and outside agencies. Thus, the Z-axis is divided accordingly: Independent or TSA-Primary: These are plays that are primarily conducted by TSA officers/employees. Cooperation of other agencies (such as law enforcement) may be sought or needed for arrest but is not necessary to initiate or carry out the play. Cooperative: These plays require cooperation between TSA and another agency, such as law enforcement, in order for them to be initiated. The Playbook specifically references the agency whose cooperation is required, and without the support of the other agency the play could not occur. Because the Playbook arises from TSA, there are no plays in which the TSA does not take a lead or cooperative role.

Method for Mapping Plays into the Airport Security Matrix After the Airport Security Matrix was adapted and the axes were defined, each play was initially coded by two members of the research team. To maximize inter-rater reliability, each was given half of the plays to code, and then each examined the other’s coding to check for inconsistencies. Any disputes were resolved by the project manager and principal investigators who oversaw the play coding and frequently examined individual plays. To further strengthen coding, the project manager checked a random sample of plays, and the principal investigators served as a final point of dispute resolution. During this process, the definition of the categories for each axis continued to be refined, which often led to further examination of plays and recoding. This process increased the amount of individuals examining and discussing the coding of each play. Where more than one mechanism of prevention applied to a single play, the research team discussed the apparent intent of the play and selected the primary intended mechanism on this basis. For example, a play involving checks of airplane crawl spaces might be interpreted on the Y-axis as intended to increase guardianship or decrease the vulnerability of the target. However, following discussions, the research team agreed that increased guardianship referred primarily to “general watchfulness,” looking for suspicious activity but not necessarily a specific threat. Crawl space searching was more specific and therefore was categorized as “reducing vulnerability of targets.” Similarly, checks of bags and hand swabs could be coded as either deterring offenders or reducing vulnerability of aircraft. Again, the team agreed that if decreasing vulnerability was primarily concerned with minimizing the impact of a threat that had not otherwise been blocked, bag and person searches should be coded as deterring offenders by providing increased effort because they are designed to block or deter a specific threat (i.e., contraband, weapons, or explosives penetrating a layer of security). It is important to note that this process of coding the plays into the airport security matrix was a dynamic process, for the purpose of illustration and categorization only. We did not treat this process in the same way one might validate a measurement or coding instrument through reliability statistics for classification consistency.

226

C. Lum et al.

The goal of this exercise was to develop consensus across the team about describing plays according to common dimensions. The final definitions provided in each of the dimensions above reflect this consensus building, and the details within each description provide others with guidance on further mapping new plays (if any) into the matrix. Yet, readers should be cautious; even within the dimensions, there is definitional overlap, as described above. Further, unlike the Evidence-Based Policing Matrix, there is not a populated airport security matrix to compare plays against. More important to the overall project is determining how to evaluate the Playbook’s effectiveness. The purpose of this exercise is to gain a better sense of the nature and potential of the Playbook and to find ways to compare activities against existing research in a qualitative manner.

A Note on Strategic Plays Twenty-two A minority of the 164 individual plays were more strategic than tactical and were mostly found in the HQ Playbook. While tactical plays are those carried out immediately, for a specific situation or preventative effect and usually at a specific location, strategic plays are intended to be more long-term in nature, and have a planning or proactive aspect to them. Strategic plays, for example, may emphasize building the capacity for security for security through running drills. Given that our prevention matrix focuses primarily on situational and deterrent actions, these plays are discussed separately in our results section, as they could not be easily categorized along any of the axes. Thus, 142 immediate/tactical plays were mapped into the Airport Security Matrix, and the 22 strategic plays are discussed separately in the results. Finally, it should be noted that during this first phase of our multiphase project, we were in constant contact with DHS and TSA personnel about questions and clarifications of the Playbook. We also conducted two site visits to one international and one domestic airport that used the Playbook regularly, to discuss how the Playbook and the randomization software were implemented and to obtain a better sense of the realities of the Playbook. In-depth examination of airport use of the Playbook occurs in later phases of this project. However, for the purpose of establishing an evidence base for the Playbook here, these visits and interactions provided an important realistic context in which the Playbook operates.

Results For ease of visualization and discussion, we divide our results into two two-dimensional “cuts” of the Airport Security Matrix along its Z-axis. Figure 10.4 shows just the Matrix codings for independent plays, and Fig. 10.5 shows just the mapping of the cooperative plays. The independent (or TSA-Primary) plays reflect the great majority (89%) of all of the tactical plays, and only a small minority of the tactical plays are cooperative in nature. Even when including strategic plays, which by nature rely more

10

Translational Criminology: Assessing Airport Security

227

Y: MECHANISM OF PREVENTION

Deter offenders/ increase their effort

The Transportation Security Administration required the authors to remove any showing of results in this figure.

Increase Guardianship

Reduce vulnerability External to airport

Public airport areas

Screening area

Employee secure areas

Passenger secure areas

Gate

Aircraft

No place specified

X: LOCATION

Fig. 10.4 Independent plays [ORIGINAL FIGURE REDACTED BY ORDER OF THE TRANSPORTATION SECURITY AGENCY AND REPLACED WITH THIS FIGURE]

Y: MECHANISM OF PREVENTION

Deter offenders/ increase their effort

The Transportation Security Administration required the authors to remove any showing of results in this figure.

Increase Guardianship

Reduce vulnerability External to airport

Public airport areas

Screening area

Employee secure areas

Passenger secure areas

Gate

Aircraft

No place specified

X: LOCATION

Fig. 10.5 Cooperative plays [ORIGINAL FIGURE REDACTED BY ORDER OF THE TRANSPORTATION SECURITY AGENCY AND REPLACED WITH THIS FIGURE]

228

C. Lum et al.

on the cooperation of other airport entities and law enforcement, the independent TSA-Primary plays still make up the great majority of all plays. While cooperation is highly emphasized as one of the main goals of the Playbook, it is clear that the vast majority of plays are conducted primarily by TSA resources and personnel.

Independent, TSA-Only Plays Figure 10.4 shows the Matrix mapping of all the plays in which TSA acted independently (N = 126) shaded by the specific Playbook (FSD, HQ, and Non Random) in which the play appears. The three Playbooks are included simultaneously to reflect the fact that TSA personnel draw on all three when creating sets of plays for deployment. A number of independent plays were tactical but could not be pinpointed to a specific location (these appear at the right of the Matrix, in the “No place specified” column). A number of interesting findings are immediately apparent. The first is the heavy clustering of Playbook tactics that focus on preventing and deterring offenders by increasing the required effort to undertake a crime or the possibility of detection if carrying a weapon. These represent a majority 54% of all plays. Note, we clarify that this does not mean that TSA officers are using these plays 54% of the time, just that 54% of the existing plays available are focused on deterring offenders in these ways. A minority of the plays approximately 29% of the plays focused on increasing guardianship and even smaller minority 16% % of the plays were about decreasing vulnerability of passengers and other targets. During the discussion of our results, we generally describe these types of plays in Fig. 10.10. These offender blocking, target hardening, and deterrence-focused plays are most often conducted in the screening area, as expected, and are fairly equally divided among passengers and employees. Unexpectedly, we found that there are many deterrence-based plays for the secure employee areas but few for the secure passenger areas except at the gates. This might reflect an avoidance of deterrence-based approaches on passengers after security screening, perhaps to reduce customer dissatisfaction or delays. Secure employee area plays include checking identification and conducting searches and tests on materials and personnel once they are past employee or public screening checkpoints. From this mapping, it appears one emphasis of the Playbook is increasing security measures toward airport staff, vendors, contractors, airline representatives, and TSA employees. This tendency is further supported when considering the three Playbooks separately (FSD, NR, and HQ). The nonrandom plays are much less likely to occur in screening areas or secured employee areas than mandated HQ plays or FSD plays. Only one offender-based deterrence play occurs in public areas outside or directly inside of the main terminal entrance (e.g., passenger loading, rental car locations, taxi stands, ticketing, check-in, baggage claim). Instead, more deterrence plays were found at the gate areas near aircraft (33% of all deterrence plays took place at the gate). These efforts include final screening, searches, and target hardening at gates as passengers board planes. TSA might consider deterrence at much earlier stages (external to airport) given recent events like those in Los Angeles or Moscow.

10

Translational Criminology: Assessing Airport Security

229

The plays that are more likely to occur in public areas outside and directly inside of the main airport terminal are those that attempt to increase general guardianship and watchfulness. Indeed, a great majority 81% of plays in these areas were guardianship-oriented. TSA may rely more on general guardianship plays because these strategies are less intrusive and easier to implement in public spaces. Plays that increase guardianship primarily include general patrol and alertness, general oversight at checkpoints, audio alerts to passengers to report suspicious activity, and periodic checks with vendors. There are no guardianship plays that occur within the aircraft itself in the Playbook, as it focuses more on airport, not in-flight, security. Of course, other tactics outside of the purview of the Playbook, including TSA’s Federal Air Marshals program, as well as airline employee activities, still occur. While the guardianship plays were more evenly spread across multiple location settings, plays intended to decrease the vulnerability of passengers or other targets (or reduce rewards for offending), like deterrence plays, were more clustered. These plays took place in and around the aircraft itself, (61.9% of vulnerability-reducing plays occur in and around the aircraft) and primarily involved inspections of aircraft bodies and hidden spaces. Additionally, this mechanism of prevention tended to employ K-9 teams.8

Cooperative Plays Far fewer plays fell into the “cooperative” area of the Matrix (Fig. 10.5). Of those plays that are tactical and immediate in nature, only a small minority of 11%, or 16 of the 142 tactical tactical plays relied on the cooperation of other non-TSA entities in a substantial way.9 Cooperative plays were also more likely to be strategic and longterm, rather than tactical the strategic plays. Three cooperative tactical plays could not be categorized in terms of a specific location and appear at the right of Fig. 10.5. These were plays that have the potential for deployment across multiple areas of the airport or that were on-the-spot audits of badges and access at various locations. The majority of the cooperative plays took place in areas external to the airport, either outside the building (curbside, taxi stands, etc.) or in adjacent spaces connected to the airports (parking lots, rental car locations, taxi waiting areas). Tactics included setting up barriers, checking vehicles, general patrol, or adjusting traffic patterns. However, within airport secure areas, either in public or employee access areas or at the gate or on the plane, the vast majority of plays are primarily carried out by TSA. 8

It is unclear whether the use of K-9 is a cooperative or TSA-independent venture. In visits to two airports, the research team learned that there may be variation across airports of who owns this particular resource and how often it can be used. 9 We recognize that many plays require law enforcement to follow through with arrest. However, to be “cooperative” in our classification requires more than this use of law enforcement after the play is implemented.

230

C. Lum et al.

Differences Across the Three Playbooks We found significant differences across the three Playbooks in tendencies within the Matrix, as indicated in Fig. 10.6. For instance, the emphasized mechanism of prevention differed significantly between Playbooks,10 which was expected given the different functions of each. The FSD playbook and the HQ playbook had plays that were more target hardening/increased effort/deterrence-oriented, while the Non-Randomized (NR) playbook had more plays focused on increasing guardianship. Recall that these NR plays can be chosen at the discretion of TSA and added onto the randomly selected FSD plays and mandated HQ plays. Similarly, significant differences were found between Playbooks with regards to the location where plays are deployed.11 As Fig. 10.7 shows, FSD-randomized plays focused mostly on screening and secure employee areas, while HQ plays centered on public areas both external and immediately inside the terminal. NR plays tended to center on secure passenger areas, including gate locations. It appears that one important emphasis of the Playbook strategy is focusing prevention efforts on nonpassengers with official business in the airport. However, the existence of a variety of NR discretionary plays in passenger secure areas shows that there is an attempt to focus on passengers as well. Most plays were categorized as independent and required no interagency coordination for implementation. This in itself is a useful finding, especially since a goal of the Playbook is to facilitate interagency cooperation. There were significant differences in the amount of collaboration required to implement plays across the three Playbooks. As Fig. 10.8 illustrates, HQ plays were far more likely to require collaboration between TSA and an external agency than either FSD or NR plays.12 None of the plays in the NR Playbook required cooperation from other entities. As one reviewer pointed out, this should not be surprising given the nature of NR plays from an organizational perspective. NR plays may be dependent on staffing levels in TSA for their own purposes. Involving other entities may reduce the control TSA has over personnel and security, or may tax resources within and outside of TSA enough to discourage cooperative approaches. HQ plays, on the other hand, may be focused on developing broad-range, long term, airport-wide strategies that focus on interagency cooperation. Finally, a contingency table between the location and mechanism of prevention of plays revealed significant findings.13 As Fig. 10.9 depicts, we found that the majority of plays that intended to increase effort or deter offenders were located in screening (28.6%) or secure employee areas (37.7%). Plays intended to increase guardianship was more likely to target public (39.5%), secure public 10

Although significance testing is not entirely meaningful for this type of analysis, we report significance values in footnotes for those interested. Here, Chi-Squared = 9.78, df = 4, p < 0.05. 11 Chi-Squared = 33.6, df = 10, p < 0.001. 12 Chi-Squared = 24.2, df = 2, p < 0.001. 13 Chi-Squared = 105.0, df = 10, p < 0.001.

10

Translational Criminology: Assessing Airport Security

231

FSD (N=82)

HQ (N=41)

NR (N=19)

Increase effort/ deter offenders (N=77)

62.2

48.8

31.6

Increase guardianship (N=43)

26.8

26.8

52.6

Reduce vulnerability (N=22)

11.0

24.4

15.8

100.0

100.0

100.0

Total (N=142)

Fig. 10.6 Differences in mechanisms of prevention across the three playbooks (% of plays)

FSD (N=82)

HQ (N=41)

NR (N=19)

Public airport areas (outside and inside) (N=29)

14.6

36.6

10.5

Screening areas (all) (N=26)

25.6

12.2

0.0

Secure employee areas (N=34)

30.5

12.2

21.1

Secure passenger areas including gates (N=24)

13.4

9.8

47.4

Aircraft only (N=16)

8.5

14.6

15.8

No place specified (N=13)

7.3

14.6

5.3

100.0

100.0

100.0

Total (N=142)

Fig. 10.7 Differences in location of implementation across the three playbooks (% of plays)

Independent (N=126) Cooperative (N=16) Total

FSD (N=82)

HQ (N=41)

NR (N=19)

96.3 3.7 100.0

68.3 31.7 100.0

100.0 0.0 100.0

Fig. 10.8 Differences in interagency cooperation and reliance on other agencies (% of plays)

(23.3%), or nonspecific areas (23.3%). The majority of plays intended to reduce target vulnerability were located at the aircraft itself (59.1%).

Strategic Plays Twenty-two of the 164 plays (13.4%) across the Playbook were classified as strategic – that is, long term, operational activities rather than immediate, tactical deployments. As we noted above, strategic plays were excluded from placement in the Matrix because they aim to address broader goals across locations and mechanisms

232

C. Lum et al. Increase effort/ deter offenders (N=77) 11.7

Increase guardianship (N=43) 39.5

Reduce target vulnerability (N=22) 13.6

Screening areas (N=26)

28.6

4.7

9.1

Secure employee areas (N=35)

37.7

9.3

4.5

Secure public areas (including gates) (N=24) Airplane (N=16)

16.9

23.3

4.5

3.9

0.0

59.1

1.3

23.3

9.1

100.0

100.0

100.0

Public areas (N=29)

No place specified (N=13) Total (N=142)

Fig. 10.9 Differences in mechanism of prevention across locations within the airport (% of plays)

of prevention and, as such, could not easily be classified along the three axes. Nineteen of the 22 strategic plays (86.4%) appear in the HQ Playbook, reflecting the nature of these plays as long-term management activities, rather than ground-level specific operations. The 22 strategic plays may be broadly classified into three themes: security planning, security awareness, and security infrastructure management. Securityplanning activities, which broadly encompass the development of protocols for incident management and cooperation with stakeholders, comprise the majority of strategic plays (N = 12; 54.6%). Seven plays (31.8%) focused on ensuring familiarity among airport staff and the traveling public with threat levels. Three plays (13.6%) were concerned with the management of security and communications infrastructure within airports. As might be expected from the long term, operational nature of strategic plays, these plays tended to rely on the cooperation of other non-TSA agencies more than on the immediate plays, but not always. Only half of the 22 strategic plays required cooperation. Further, the majority of the strategic plays (72.7%), unlike the tactical plays, focus not on deterring offenders but on increasing the capacity for general guardianship and watchfulness for suspicious activity through planning and awareness activities. Finally, and again unlike the tactical plays, most of the strategic plays (N = 18, 81.8%) did not specify a location for deployment. They focused more on addressing the capacity of the security apparatus itself over the long term through management and planning.

Discussion: Is the Playbook Evidence-Based? The results above show the classification of the Playbook across common crimeprevention mechanisms and where they cluster in the airport security matrix. However, plotting unevaluated plays into this matrix does not provide knowledge of the effectiveness of these specific interventions, only a categorization of common

10

Translational Criminology: Assessing Airport Security

233

prevention and deterrence tendencies of similar plays. The question for evidencebased security policy is whether these tendencies are supported by research evidence on evaluated interventions that use similar prevention approaches. Unfortunately, the lack of evaluation research in airport security means that we do not have an evidence base to develop a matrix by which to compare our mapping of the Playbook plays (as was done in policing by Lum et al., 2011). However, a translational criminological approach allows us to examine existing research in similar subject areas to see whether the evidence supports the mechanisms of prevention of the Playbook plays. This provides a preliminary assessment of effectiveness for better informed decisions now, and builds the case for more evaluation. Here, we use the aforementioned literature on situational crime-prevention, deterrence, and interagency cooperation to make such comparisons. In Fig. 10.10 we first describe the groupings of plays that fall within each cell of the airport security matrices in Figs. 10.4 and 10.5.14 These descriptions provide an overview of the types of plays that fall within a cell (e.g., “external public areas” which “deter offenders by hardening targets”). Using these general descriptions, we then looked for existing evaluations of similar situational crime-prevention, security, and policing mechanisms. We relied heavily on Eck’s (2002) review of situational crime-prevention measures, and sometimes on policing studies collected in Weisburd & Eck (2004) and Lum, Koper & Telep (2009, 2011). We choose research evidence that uses at least a “moderately rigorous” evaluation design of which to compare. “Moderately rigorous” is labeled by Sherman et al. (2002) in their Scientific Methods Score (SMS) as “Level 3” (of five levels). A Level 3 evaluation design “measures crime before and after the program in [nonequivalent] experimental and comparable control conditions” and is considered the minimally acceptable research design in determining “what works” (Sherman et al., 2002, p. 17; see also Cook & Campbell, 1979). Level “4” and “5” studies are those that use more rigorous quasi-experimental and experimental designs, respectively. Figure 10.11 summarizes this evidence using the same cells as Fig. 10.10, with the SMS score provided in parentheses after each citation.

Prevention, Deterrence, and Unpredictability The translation from the research evidence in Fig. 10.11 back to the descriptions of mechanisms for prevention in the Playbook plays in Fig. 10.10 is challenging. Not only do the subject matters differ, but the methodological rigor of many studies in this area of criminology is also modest. Thus, only cautious translation is encouraged here. 14

Again, in September 2011, prior to this book going to print, the Transportation Security Administration required us to redact numerous sections of this chapter, or the authors and the editor could not move forward with publishing this volume. We recognize this results in missing information in this chapter.

234

C. Lum et al. Deterring offenders and increasing their efforts

Increasing guardianship

Checking ID and other ways of controlling access to airport; cooperation from other entities.

Increased surveillance of transit and other portals leading up to and into airports. Increased surveillance of indoor public areas.

NO PLAYS Screening areas

Secure areas (Employee)

Secure areas (Passengers) (Except boarding gates) Gate

Aircraft

No area specified

Reducing vulnerability of passengers and other targets Keep vehicles further away from airport entrance or slowing their approach. NO PLAYS

Screening activities the TSA is involved in, including searching passengers, employees, contractors and vendors, and baggage. Restricting access of people and vehicles; variety of screening options for employees.

Increased surveillance NO PLAYS for suspicious behavior at checkpoints.

Increased surveillance for baggage handling; employee compliance regarding security.

Ensure that baggage handling procedures are being properly followed.

NO PLAYS

Increased surveillance of people inside secured public areas.

NO PLAYS

Random checks and screening of passengers, baggage, and crew. Screening of personnel and restricting flight privileges or not allowing access to flights.

Increased surveillance of passengers and crews in gate areas. NO PLAYS

Monitoring materials brought on flights.

Monitor and control access to various secured areas throughout the entire airport via screening and pre-screening procedures.

Generally increased surveillance, encouraging awareness random checks of vendor items, testing security.

Checking and inspecting airplanes, examining compartments for hiding of people and things. Developing strategies for pre-screening, escorts, raising awareness of threat level.

Fig. 10.10 General description of plays of intersection of location and prevention mechanism

For example, one major element of the Playbook is the focus on tactics and strategies intended to prevent and deter offenders by increasing the amount of effort they would have to undertake to commit an offense (e.g., by target hardening). This is embodied both within the individual plays and more generally in the concept of randomly selecting and deploying sets of plays to increase unpredictability of security actions. The most relevant studies in this area are those directly related to metal detectors at airports, the effects of which have been highlighted above. However, other studies in Eck’s (2002) review of situational crime-prevention measures that are intended to deflect and deter offenders may also apply, although again, caution should be exercised, given that many of these evaluations have only been conducted using moderately rigorous methods (Weisburd, 1997). Methods of access control,

10

Translational Criminology: Assessing Airport Security

235

Deter offenders

Guardianship

External public

Challinger, 1996 (3), identification requirements. Poyner, 1994 (3), redesign of residence. Ekblom,1988 (3), bullet-proof screen and barriers. Atlas & LeBlanc, 1994 (3) and Lasley, 1998 (3) street closures and barricades. Enders & Sandler, 1993, 2000 (3) fortifying embassies showed NO effect.

Mixed findings regarding guardianship in outdoor places that have not had preliminary screening. Hesseling, 1995 (3), guards have no impact, but Barclay, Buchley, Brantingham, Brantingham, & WhinnYates, 1996 (3), Laycock & Austin, 1992 (3), and Poyner, 1994 (3), found an effect. CCTV in open public spaces showed some positive effects, only in UK, not in US (Welsh & Farrington, 2008).

Evidence on street closures, rerouting and barricades. Mixed but mostly effective. (Eck = “promising’) Atlas & LeBlanc, 1994 (3); Lasley, 1996, 1998(3); Newman 1996 (3). Restricting pedestrian movement is effectivePoyner 1994 (3).

Internal public

Although no plays are here, strategies that may be comparable in Eck (2002) are public-space interventions that increase lighting (or in this case, “visibility”) (see Eck, 2002, p. 270– 271).Hot spot patrol targeting high risk locations may be relevant, i.e., Sherman & Weisburd, 1995.

Mixed findings regarding guardianship in indoor places that have not had preliminary screening. Kenney, 1986 (3), Guardian Angels have no impact in subways. Webb & Laycock, 1992 (3), CCTV in subways.

Strategies here may be similar to those two cells to the left, especially in increasing lighting.

Screening areas

Metal detectors are effective - Cauley & Im, 1988 (3); Enders & Sandler, 1993 (3); 2000 (3); Enders, Sandler, & Cauley, 1990 (3); Landes, 1978 (3). Centers for Disease Control and Prevention, 1993 (4), metal detectors in schools.

Due to secure sensitive information policies of DHS and TSA, some information here cannot be disclosed. However, confidential studies exist in this area.

No comparable evaluation could be located.

Secure areas (E)

Interventions that stop employees from becoming facilitators may be applicable, including timelock cash boxes and safes-Clarke & McGrath, 1990 (3).

May be similar to bar and tavern evaluations, such as codes of practice and training for employees. Felson, Berends, Richardson, & Veno, 1997 (3), Putnam, Rockett, & Campbell., 1993 (3).

Could be comparable to changing codes of practice through training of employees in alcohol outlets or convenience stores-Felson et al., 1997 (3); Putnam et al., 1993 (3); Saltz, 1987 (3).

Fig. 10.11 Comparable evaluations of situational crime-prevention and deterrence interventions

236

C. Lum et al.

Secure areas (P)

Strategies that may be comparable in Eck (2002) are the CCTV studies already mentioned (3) (for purposes of deterrence, as opposed to guardianship,see below) as well as hot spots studies by Sherman & Weisburd, 1995 (5) in specific high risk areas.

First two cells in this column above may apply. However, this is an area of the airport security matrix that focuses on public past the initial screening and security stage. There are no studies comparable in the crime prevention literature on this type of prevention strategy.

No comparable studies found.

Gate

Passenger screening mentioned in above cells are relevant, if repeated at gates. Targeting highrisk flights and gates may be more effective than a random selection strategy, given what we know from hot spots studies (see Braga’s reviews 2005, 2007).

Could be compared to security interventions at gate areas of pass transit systems, but evaluations of this are weak (SMS=2); See e.g., DesChamps, Brantingham, & Brantingham., 1992 (2); van Andel, 1989 (2).

Bank studies on reducing vulnerability found that screens protecting tellers resulted in burglary reductions—Ekblom, 1987 (3); Barclay et al., 1996, Hesseling, 1995, Laycock & Austin, 1992 (3); Poyner, 1994 (3).

Aircraft

Some studies have examined protective screens for drivers, but were not scientifically rigorous (see Poyner & Warner, 1986). Target hardening of public coin machines may have similar mechanisms of prevention (Decker, 1972 (4))

No studies found, but evaluations in this area could include those that would examine business checks by police or placespecific targets.

No comparable studies found.

No area specified

Some parallels might be found regarding cooperation element— Weisburd & Eck (2004) and Lum et al. (2011) argue the more tailored and multi-agency strategies (sometimes needing outside agency) are more effective.

van Andel, 1989 (2) is a very weak study that indicates getting public to monitor may reduce crime. Neighborhood watch might apply here (see Bennett et al., 2006 (3)), which showed some positive effects, but many are weak studies.

Reducing vulnerability in general: tagging and property marking has mixed results, some backfires (Gabor 1981 (3)). Laycock 1985, 1991 (3)—highly publicized property marking is effective; Farrington et al 1993 (3)—effective. Repeat victimization studies have been effective (see Farrell’s 2005 review, most studies level 3, a few 4).

Fig. 10.11

(continued)

10

Translational Criminology: Assessing Airport Security

237

entry/exit screening, and target hardening, such as improving the security of doors and windows (Tilley & Webb, 1994), closing walkways connecting buildings (Poyner, 1994), or blocking off or restricting access to streets and alleyways (Bowers, Johnson, & Hirschfield, 2004), have all been shown to help reduce robberies and burglaries. Clarke (1997) asserts that entry/exit screening also significantly reduces crimes such as theft, shoplifting, and fare evasion. In a systematic review of publicarea surveillance and crime prevention, Welsh, Mudge, and Farrington (2010) provide further evidence that techniques such as street closures and barricades are effective crime-prevention methods. Again, although these findings are not directly related to TSA operations, they show strong support for similar crime-prevention mechanisms in many plays that reduce access, improve guardianship, and decrease the vulnerability of targets at airports. Although the vast majority of the relevant literature is not policing related, some policing evaluation studies may be useful, especially those that relate to gun carrying and drug enforcement. For example, McGarrell, Chermak, Weiss, and Wilson (2001) and Sherman, Shaw, and Rogan (1995) showed mixed but promising results in targeting places that have high incidents of gun crime (and thus gun carrying) using directed traffic and pedestrian patrols and stops in high gun-crime neighborhoods. This strategy could be loosely interpreted as a “screening” mechanism as well as random searching. However – and this is a key distinction – these studies also depend on one very important caveat not present in Playbook deployment: a specific location that is identified as high risk for contraband carrying. Most Playbook deterrence plays take place at screening checkpoints and employee-secure areas rather than in public areas prior to checkpoints or passenger-secured areas. Some deterrence plays do occur at gates for identified high-risk flights, which may present an opportunity for better place-based targeting and gate selection for play deployment. However, it does not appear that this type of patrol is present in public areas at airports (at check-in, curbside or baggage claim), where previous attacks have occurred. Existing research in police patrol suggests that combining these types of deterrence-oriented plays with more specific place-based targeting may increase the deterrent effect without significant spatial displacement (Braga, 2007). As already emphasized, this approach is contrary to a nondirected patrol approach, in which little to no analysis is done to try to anticipate high-risk places, people, times, or situations. There is little to no evidence that ad hoc deployment of tactics is an effective strategy against reducing crime in a jurisdiction (Lum et al., 2011; National Research Council, 2004; Weisburd & Eck, 2004). However, it is important to note the differences between policing street crime versus screening at airports when applying police-related research to airport security. These differences may affect whether or not translation between crime prevention evaluations and counterterrosim or security is possible. In policing, high-risk locations are easily identified and fairly stable (Weisburd et al., 2004), given the regularity and “everyday” aspects of crime and disorder. Not only is targeting high-risk locations easier given the high predictability of these places, but offending is usually more visible, especially in areas with high crime rates. Further, the motivations of the

238

C. Lum et al.

average offender may (or may not) be different or more tenuous than someone intending to do harm in an airport. While Dugan et al. (2005) indicated that hijackers generally succumb to rational choice forces (see also Berrebi, 2009; Crenshaw, 1990; Jackson, 2009), Nagin (1998) and Durlauf and Nagin (2011) note that the deterrent effect of strategies may have varying ranges of elasticity with regards to the responsiveness of offenders to interventions. We might hypothesize that an airport offender has a less elastic deterrence response to prevention interventions than an offender in a more “everyday” crime situation. While we use policing literature here to emphasize that place-based targeting is important in general patrol and increased guardianship, airport security efforts may be more translatable from situational crime-prevention research. Most importantly, translation statements rely on the assumption that plays have been properly implemented. For example, with regard to the random selection of plays from the Playbook, our initial assessments of the implementation of the Playbook in two airports found the randomization scheme of the Playbook is in reality less random than its intention. TSA personnel have the prerogative to deselect plays from the randomization software prior to running it. Plays may be deselected for a variety of reasons, including irrelevance to a specific airport (e.g., airports with minimal cargo operations can deselect cargo plays), to lack of resources to employ the play, to just individual discretion. Further, once the randomizer selects plays, TSA personnel can supplement the selection with specifically selected plays. Once the suite of plays are selected across the period of time in which that particular randomized selection is used (i.e., 1 or 2 weeks, a month), TSA personnel can select which plays they implement on any given day. Finally, the Playbook only represents one part of many other security measures provided by TSA, local law enforcement, passengers, airport workers, and other employees that are not in the Playbook.

Increasing Guardianship Situational crime-prevention studies, as well as police patrol studies, on increasing guardianship have also been conducted using moderately rigorous to highly rigorous evaluation methods. These strategies primarily focus on increasing formal and informal surveillance through law enforcement, although some evaluate other types of guardians. The most translatable studies for airport security are evaluations of interventions that increase general watchfulness of areas, such as CCTV and Neighborhood Watch. Campbell Collaboration systematic reviews of these interventions indicate generally positive but highly variable results depending on the location of the intervention, the nature of the guardianship, and the strength of the evaluation method used. For example, Bennett, Holloway, and Farrington (2006) found that Neighborhood Watch programs that attempted to increase civilian detection of suspicious activity through increased guardianship were associated with crime reductions of between

10

Translational Criminology: Assessing Airport Security

239

16 and 26%. However, many of the evaluations they reviewed involved nonequivalent comparison groups and weak designs, which Weisburd, Lum, and Petrosino (2001) found are biased towards showing positive results. Further, it was not always possible to distinguish whether the crime reduction effect stemmed from the guardianship element of the program or from other security measures employed alongside it (such as property marking). Additionally, Welsh and Farrington (2008a) reviewed studies of formal surveillance using closed-circuit television cameras. They also found a modest crimereduction effect overall, but further analyses suggested that this effect was driven by successes in very specific situations and contexts, such as reducing vehicle crime in parking lots. The cameras were less successful in reducing other crime types and in city centers or residential areas (see also Farrington, Gill, Waples, & Argomaniz, 2007; Sherman & Eck, 2002). The variety of findings is also evident in a recent quasi-experimental evaluation of the deterrent effect of CCTV by Caplan, Kennedy, and Petrossian (2011). Interestingly, the deterrent effect of CCTV did not change depending on whether cameras were placed in strategic or random locations, possibly challenging ideas related to directed vs. random deployment for this particular situational crime-prevention intervention. Some of the CCTV research examines surveillance cameras in conjunction with other crime-reduction interventions like street lighting, created difficulties in identifying a direct causal effect. Improved street lighting in itself has been shown in a Campbell systematic review to significantly decrease personal and property crime in public spaces (Welsh & Farrington, 2008b), but it is unclear whether increased guardianship facilitated by better visibility is the primary causal mechanism, or whether improved lighting creates an image of community investment and social cohesion that operates to reduce crime through informal social control. However, the comparability of better visibility to airport security is questionable. Welsh et al. (2010) conducted a more general systematic review of a variety of public-area surveillance strategies using at minimum studies which employed at least a nonequivalent comparison group design. They found some support for the effectiveness of both security guards and citizen patrols (such as “Guardian Angels”) in reducing crime, although again there are limitations in the findings. For instance, the few evaluations of security guards were all conducted in parking lots, so they are only shown to be effective in that limited context. Additionally, the Guardian Angels studies indicated reductions in property crime but not violent crime. Welsh et al. also examined two promising studies of place managers (a concierge in a housing block and a taxi firm operating in a parking lot), but again it was not possible to distinguish whether increased guardianship was the primary causal mechanism, or whether other factors were more important in reducing crime (e.g., access control performed by the concierge). Although the results of the research on increased guardianship are mixed, the use of surveillance and general watchfulness strategies do represent a promising approach to airport security. The evidence base indicates that increased guardianship may play some role in reducing crime in specific contexts and when used in conjunction with other situational mechanisms. It remains unclear what types of guardianship work

240

C. Lum et al.

best in the context of airport security, as well as the effects such guardianship might have on passenger and employee perceptions of (and thus legitimacy they afford to) airport security personnel. However, given the extent to which guardianship is a key aspect of securing large spaces, more research is clearly warranted.

Decreasing Vulnerabilities of Passengers and other Targets Passengers and employees can also become effective guardians themselves, not only in improving guardianship more generally, but also in decreasing their own risk of victimization. Further, airport security systems which encourage passenger and employee empowerment may have additional benefits of garnering support, legitimacy, and willingness to cooperate from people. In the language of the Playbook, decreasing vulnerability of passengers and other targets primarily translates into checking aircraft and areas around aircraft for explosives and hiding spaces, conducting final search and screening of passenger bags and persons at gates, as well as erecting barriers, redirecting traffic, and also ensuring baggage procedures are followed. General security announcements and reminders also play a role in decreasing vulnerability, although these strategies apply to airport security more generally, not just to the Playbook. While there is no research in the situational crime-prevention literature that examines these specific strategies, there have been evaluations of related techniques to make targets less attractive, thus reducing their vulnerability, and increasing awareness. Property marking or tagging is one such intervention, albeit questionable with regard to similarities with airport security. Only a handful of moderately rigorous studies have examined property marking, and while a few found positive results (Farrington et al., 1993; Laycock, 1985, 1991), Gabor (1981) found an increase in burglaries following property marking schemes. The repeat victimization literature also provides some information on the effects of reducing victim vulnerability (summarized in Farrell, 2005). This body of research focuses specifically on interventions that reduce the tendency for some people, places, or targets to experience victimization more frequently than others. The evaluations of at least moderate rigor in this field have focused on repeat residential burglary victimization and encompass a range of tactics, including citizen awareness and advice schemes, property marking, security assessments, and target hardening. Taken together, these studies suggest that using multiple tactics that are tailored to the specific context in which the victimization occurs may be the most successful approach. However, the most successful combinations of tactics tended to be those focused on target hardening or target removal, whereas studies involving mainly victim advice and awareness schemes appear to have less of an impact on crime. TSA’s Playbook provides for a number of vulnerability-reducing plays at the innermost layers of security, such as at the gate area and around the aircraft itself. As noted above, these are important points of intervention for ensuring the safety of

10

Translational Criminology: Assessing Airport Security

241

the aircraft and passengers. Some plays, like traffic redirections and barriers, also aim to reduce vulnerabilities at the airport perimeter. Less vulnerability-reducing plays at the intermediate levels may reflect the fact that most activity in these areas is more directly concerned with screening for and blocking specific threats. However, given that the literature on repeat victimization suggests the effectiveness of tailored, multitactic approaches to reduce vulnerability, TSA may also consider focusing a similar variety of efforts within the public areas and secure passenger areas.

Interagency Cooperation Our site visits and discussions with TSA personnel revealed that Playbook operations take place in a complex environment. Security activities potentially influence (and sometimes directly engage) a wide range of stakeholders, including employees, vendors, the public, local law enforcement agencies, and other agencies that perform security and law enforcement functions near airport areas. Each of these individuals and organizations has a different set of interests related to airport areas – not just in terms of security but also business and commercial interests that may come into conflict with safety concerns. Research on interagency cooperation in criminal justice has examined how cooperation between agencies occurs in general, as well as how cooperation relates to specific crime-prevention activities, such as community-based and situational crime-prevention (e.g., see Crawford & Jones, 1995; Knutsson & Clarke, 2006). This body of research highlights the difficulties that justice agencies can face when attempting to establish cooperative processes, such as managing the exchange of information, organizational resistance to change, and sometimes conflicting organizational goals (Gil-Garcia, Schneider, Pardo, & Cresswell, 2005). The plays that we classified as strategic and cooperative required TSA personnel to accomplish a wide variety of tasks, such as interacting with experts in threat mitigation, ensuring that established security procedures were being followed, and checking that all vendors operating at the airport are authorized to do so. Such activities involve challenges that are similar to those experienced by law enforcement agencies that are trying to establish cooperative processes with third parties. Some research exists that indicates that multiagency approaches to problems can achieve significant crime-reduction gains. Weisburd and Eck (2004), as well as Lum et al. (2011), highlight a number of “tailored” evaluations of policing drugs and violence that involve third parties and use of civil remedies. Many of the highest quality studies include experimental evaluations of the use of nuisance abatement, finding that cooperation between law enforcement and property managers, prosecutors, and other city agencies can reduce drug problems at hot spots (see Braga et al., 1999; Eck & Wartell, 1998; Mazerolle, Price, & Roehl, 2000). However, in these cases, the third parties provided a specific service needed to shut down a crime facilitator, rather than to ensure security. With regard to homeland security, while Lum, Haberfeld, Fachner, and Lieberman (2009) found that many law

242

C. Lum et al.

enforcement agencies see interagency cooperation and information sharing as important to counterterrorism, there remains no evaluation of this assertion with regard to either crime prevention or homeland security. One criticism of American intelligence efforts before 9/11 was the lack of interagency cooperation, which was discussed extensively in the 9/11 Commission Report (National Commission on Terrorist Attacks upon the United States, 2004). This criticism led to the creation of “fusion centers,” which sought to create greater cooperation between police and security agencies (see Carter & Chermak, 2011). A study of counterterrorism operations in Israel reinforces the idea that strong cooperation among agencies will aid effective prevention efforts (Weisburd, Jonathan, & Perry, 2009). However, there is also some evidence that interagency collaboration may not always be desirable or add to specific and focused crime-prevention efforts. For example, Boba, Weisburd, and Meeker (2009) report that efforts to develop regional data sharing in Redlands Valley, California were hindered by the simple reality that most ordinary prevention efforts occurred within a specific enough context that data sharing between agencies was unnecessary. This may also be the case for Playbook operations. Indeed, most plays did not require interagency cooperation. While this may be because most plays take place in areas over which TSA has primary jurisdiction, it is uncertain whether assistance and cooperation with other agencies could improve outcomes. Our preliminary site visits indicated that, in practice, the Playbook activities presented both a challenge and an opportunity to TSA and other airport personnel. Challenges include educating and soliciting the cooperation of other airport entities, including law enforcement, airport authorities, airport employees, contractors and vendors, and passengers about the Playbook. However, the Playbook also provides tangible tasks and activities that can serve as a forum by which interaction and relationship-building can occur. It seems reasonable to hypothesize (for future analysis) that improved cooperation between entities responsible for security can provide for a stronger security apparatus against both passenger and insider threats.

Possible Collateral Effects of Playbook Implementation Any implementation of prevention and deterrence measures can have collateral effects. These might include displacement, net widening of tactics to low or no-risk individuals, perceptions of bias and differential treatment in implementation, or even violations of privacy or Fourth Amendment rights. While passengers have expressed their willingness to submit to searches in airports for the sake of security (USA Today/Gallup Poll, Nov. 19–21, 2010), airport authorities, the TSA, and law enforcement agencies are still concerned with how broader security measures can affect their legitimacy with the public, or potentially weaken security. For example, one collateral effect often mentioned in situational crime-prevention strategies is displacement. Displacement refers to the change in either the method or location of crime in response to a crime-prevention strategy. Some studies

10

Translational Criminology: Assessing Airport Security

243

point to a lack of major spatial displacement for general crime trends from focused, place-based interventions, countered with diffusion of crime control benefits (see Clarke & Weisburd, 1994; Guerette & Bowers, 2009). However, these studies concern spatial displacement of directed patrol in high crime places. Research on potential displacement effects of airport security measures has not yet been conducted, with the exception of the substitution effects from hijacking to different types of incidents noted by Cauley and Im (1988). Displacement of method or location is important to consider and evaluate when examining counterterrorism and security programs. A second potential collateral effect of implementing the Playbook is the possibility of reduced legitimacy of security authorities, especially those involving screening and searches of passengers and employees (Elias, 2010). Not all security activities are equally legitimate in the public’s eyes, and even in an environment with a relatively supportive public there may be wariness about more intrusive screening tactics among passengers. General deterrence strategies have the drawback of creating “false positives” when broadly applied to a population. As noted above, this creates a challenge to the values of democratic societies, such as privacy or human dignity, which in turn can lead to reduced legitimacy. Tyler (1990, 2003, 2004, 2011) has argued that when practices and procedures of governing institutions are perceived as fair and equitable by the individuals subject to them (procedural justice) – even if outcomes are not favorable – the legitimacy of the institution will not be eroded, and individuals will be more likely to cooperate with authorities. This cooperation is crucial in airports because many security measures rely on citizens and employees to act as additional “eyes and ears” to maintain security. Related to this, collateral effects of airport security may be distributed differently along racial, ethnic, and religious lines, which can further erode legitimacy and notions of fairness. In an on-site survey of 500 passengers who had just passed through airport security at a major international East Coast hub, Lum et al. (2007) discovered that despite findings of overall customer satisfaction with TSA airport screenings, there were significant differences between White and non-White passengers with regard to treatment at screening. Non-White passengers reported being more likely to be subjected to additional TSA screening, more likely to receive a higher number of additional screening actions, and were less likely than their White counterparts to receive a verbal explanation of why they were pulled aside for additional screening. Non-White passengers were also more likely to feel inconvenienced and embarrassed by additional screening than Whites. Similar findings are reported in Israel by Hasisi and Weisburd (forthcoming), who found that citizen assessments of the legitimacy of airport security are influenced strongly by whether they have been subject to specific interventions, such as having their baggage identified for selective screening. These are only perceptions, and must be further studied by independent observations of officer behavior. However, they do reinforce the more general police literature that indicate that ethnic minorities are more likely to perceive their treatment by law enforcement as unfair (Langan, Greenfeld, Smith, Durose, & Levin, 2001; National Research Council, 2004). These are important considerations in the implementation of any crime/terrorism prevention intervention.

244

C. Lum et al.

Conclusion This chapter describes the first systematic, evidence-based review and assessment of TSA’s Playbook strategy to prevent and deter crime and terrorist activity at our nation’s airports using a translational criminological approach. As we have seen, there are very few evaluations of counterterrorism measures or airport security compared to other law enforcement sectors. Given the massive amount of money spent on such measures since 9/11, evaluation of the efficiency and outcome effectiveness of such measures is imperative. However, many of the crime-prevention measures at airports mirror a broader criminological literature on situational crime-prevention, deterrence, and interagency cooperation. Here, we have used these parallels in our preliminary assessment and evaluation of the TSA Playbook. In classifying the Playbook using an “Airport Security Matrix,” we found that most plays are immediate and tactical in nature, and few are strategic. Further, the vast majority of plays do not require cooperative deployment. Thus, much of our analysis focuses on immediate and tactical plays that are primarily carried out by TSA personnel. For these plays, we discovered four general tendencies. The first is that these plays more often involve mechanisms of prevention that aim to harden targets, deter and prevent offenders by increasing their perceived effort, rather than increase guardianship, or reduce vulnerabilities of passengers or other targets. Second, most of these plays focus on the public and employee screening areas; there is definitely a focus in the Playbook on employees rather than passengers. Third, plays occurring in public areas outside or directly inside of the airport entrance tend to be guardianshiporiented rather than specifically focused on deterring offenders. Finally, the Playbook tends to focus on reducing passenger and target vulnerability largely at the final “layer of security” located at gates and airplanes. When we examined the immediate/tactical plays within each of the sub-books, we found additional concentrations of plays in both mechanism type and location of play. For instance, FSD plays primarily occur in screening and secure areas (both passenger and employee) and mainly involve approaches designed to increase offender efforts. HQ plays are also designed to deter offenders, but unlike the FSD plays, they are typically designed for public areas. The HQ Playbook also contains a significant majority of the plays that require cooperation between TSA and other non-TSA agencies. NR plays typically occur at secure passenger areas and gate locations and tend to use increased guardianship as their main mechanism of prevention. A small minority of the plays was strategic in nature, and most focus on long-term management activities that incorporate the use of general watchfulness and increased guardianship. It is expectedly in the strategic plays where requirements for cooperation are found. When comparing more general descriptions of plays at intersecting Matrix dimensions, we found that the Playbook generally and loosely incorporates many evidence-based practices for prevention and deterrence, although this evidence base varies across studies by design rigor as well as applicability to airport security and counterterrorism. Of course, how and which plays are implemented at any given time ultimately tempers the Playbook’s effectiveness. The majority of plays within

10

Translational Criminology: Assessing Airport Security

245

the Playbook use situational crime-prevention mechanisms (e.g., blocking offender access and target hardening), which have been supported in other crime-prevention evaluations. Additionally, studies confirm and support the use of tailored, placespecific interventions for crime prevention and deterrence. The Playbook illustrates some compliance with this evidence-based mechanism through the location focus of many of the plays. However, how places are chosen for play implementation is not clear. More importantly, exactly how such studies translate to the context of terroristic violence within a confined location (airports) is still unknown. With regard to the notion of randomization as a deterrence mechanism, the research indicates that randomly allocating patrol at selected high-risk places can increase crime-prevention effects. However, whether the locations in which the plays are implemented are indeed the highest-risk locations in the airport is unknown. Further, although the Playbook has a built-in randomization component with regard to selection of the set of plays used at any particular time, this element of the Playbook may be manipulated in such a way that reduces randomization. However, whether this is a negative or positive change with regards to increasing security is also unknown in the absence of evaluation. Reducing random deployment of plays may not be problematic depending on whether such randomization increases or decreases deterrence. This is not clearly understood in criminological research and is not researched at all in counterterrorism studies. Further, although there is research supporting some of the prevention mechanisms that are found in both situational crime-prevention measures and airport security (which itself needs to be more closely scrutinized for comparison), there are some types of airport security measures for which we could not easily identify parallel evidence in the crime-prevention literature. Ultimately, the determination of effectiveness must be supported by evaluations, through experimentation and simulation, of the actual interventions within airports. Finally, we think the Playbook, which uses plays that involve interagency cooperation, can actually serve as a means of facilitating and fostering working relationships between the TSA and other agencies that operate in and around the airport. It might be worthwhile to explore how these interagency relationships and efforts could benefit from involvement in additional plays beyond public airport areas and areas external to the airport. The Playbook attempts a broad range of prevention and deterrence tactics across multiple contexts. Understanding the prospects and challenges of implementing such a strategy and identifying ways in which measures of success might be derived are imperative in accurately judging this method of airport security.

References Asal, V., Rethemeyer, K., Bellandi, R., Legault, R., & Tynes, R. (2010). Making the wrong connection: The determinants of terrorist targeting of airplanes and airports. Presented at the International Studies Association Annual Convention, New Orleans. Atlas, R., & Leblanc, W. (1994). The impact on crime of street closures and barricades: A Florida case study. Security Journal, 5, 140–145.

246

C. Lum et al.

Barclay, P., Buchley, J., Brantingham, P., Brantingham, P., & Whinn-Yates, T. (1996). Preventing auto theft in suburban Vancouver commuter lots: Effects of a bike patrol. In R. Clarke (Ed.), Preventing mass transit crime (Vol. 6, pp. 133–162). Monsey: Criminal Justice Press. Bennett, T., Holloway, K., & Farrington, D. (2006). Does neighborhood watch reduce crime? A systematic review and meta-analysis. Journal of Experimental Criminology, 2(4), 437–458. Berrebi, C. (2009). The economics of terrorism and counterterrorism: What matters and is rational-choice theory helpful? In P. Davis and K. Cragin (eds.), Social science for counterterrorism: Putting the pieces together. Santa Monica, CA: RAND. Boba, R., Weisburd, D., & Meeker, J. (2009). The limits of regional data sharing and regional problem solving: Observations from the East Valley, CA COMPASS initiative. Police Quarterly, 12(1), 22–41. Bowers, K., Johnson, S., & Hirschfield, A. (2004). Closing off opportunities for crime: An evaluation of alley-gating. European Journal on Crime Policy & Research, 10(4), 285–308. Braga, A. (2005). Hot spots policing and crime prevention: A systematic review of randomized controlled trials. Journal of Experimental Criminology, 1(3), 317–342. Braga, A. (2007). Effect of hot spots policing on crime. Campbell Systematic Reviews 2007:1. DOI: 10.4073/csr.2007.1. Braga, A., & Weisburd, D. (2010). Policing problem places. New York: Oxford University Press. Braga, A., Weisburd, D., Waring, E., Mazerolle, L., Spelman, W., & Gajewski, F. (1999). Problemoriented policing in violent crime places: A randomized controlled experiment. Criminology, 37(3), 541–580. Caplan, J., Kennedy, L., & Petrossian, G. (2011). Police-monitored CCTV cameras in Newark, NJ: A quasi-experimental test of crime deterrence. Journal of Experimental Criminology. doi: 10.1007/s11292-011-9125-9. Carter, J., & Chermak, S. (2011). Evidence-based intelligence practices: Examining the role of fusion centers as a critical source of information. In C. Lum & L. W. Kennedy (Eds.), Evidencebased counterterrorism policy. New York: Springer. Cauley, J., & Im, E. (1988). Intervention policy analysis of skyjackings and other terrorist incidents. The American Economic Review, 78(2), 27–31. Centers for Disease Control and Prevention. (1993). Violence-related attitudes and behaviors of high school students: New York City, 1992. Morbidity and Mortality Weekly Report, 42, 773–777. Challinger, D. (1996). Refund fraud in retail stores. Security Journal, 7, 27–35. Clarke, R. (1980). Situational crime prevention: Theory and practice. British Journal of Criminology, 20(2), 136–147. Clarke, R. (1983). Situational crime prevention: Its theoretical basis and practical scope. In M. Tonry & N. Morris (Eds.), Crime and justice: An annual review of research (Vol. 14). Chicago: The University of Chicago Press. Clarke, R. (1992). Situational crime prevention: Successful case studies. New York: Harrow and Heston. Clarke, R. (1995). Situational crime prevention. In M. Tonry & D. Farrington (Eds.), Building a safer society: Strategic approaches to crime prevention. Chicago: The University of Chicago Press. Clarke, R. (1997). Introduction. In R. V. Clarke (Ed.), Situational crime prevention: Successful case studies (2nd ed., pp. 1–43). Albany: Harrow and Heston. Clarke, R., & McGrath, G. (1990). Cash reduction and robbery prevention in Australian betting shops. Security Journal, 1, 160–163. Clarke, R., & Newman, G. (2006). Outsmarting the terrorists. Westport: Praeger Security International. Clarke, R., & Weisburd, D. (1994). Diffusion of crime control benefits: Observations on the reverse of displacement. In R. Clarke (Ed.), Crime prevention studies (Vol. D). Monsey: Criminal Justice Press. Cohen, L., & Felson, M. (1979). Social change and crime rate trends: A routine activity approach. American Sociological Review, 44(4), 588–608.

10

Translational Criminology: Assessing Airport Security

247

Cook, T., & Campbell, D. (1979). Quasi-experimentation: Design and analysis issues for field settings. Chicago: Rand McNally. Cornish, D., & Clarke, R. (Eds.). (1986). The reasoning criminal. New York: Springer. Cornish, D., & Clarke, R. (1987). Understanding crime displacement: An application of rational choice theory. In S. Henry & W. Einstader (Eds.), The criminology theory reader (pp. 45–56). New York: New York University Press. Cornish, D., & Clarke, R. (2003). Opportunities, precipitators, and criminal decisions: A reply to Wortley’s critique of situational crime prevention. In M. Smith & D. Cornish (Eds.), Theory for situational crime prevention. Crime prevention studies (Vol. 16). Monsey: Criminal Justice Press. Crawford, A., & Jones, M. (1995). Interagency cooperation and community-based crime prevention: Some reflections on the work of Pearson and Colleagues. British Journal of Criminology, 35(1), 17–33. Crenshaw, M. (1990). The logic of terrorism: Terrorist behavior as a product of strategic choice. In W. Reich (ed.), Origins of Terrorism: Psychologies, Ideologies, Theologies, States of Mind. Washington, DC: Woodrow Wilson International Center for Scholars. Decker, J. (1972). Curbside deterrence: An analysis of the effect of a slug rejection device, coin view window and warning labels on slug usage in New York City parking meters. Criminology, 10, 127–142. Deflem, M. (2004). Terrorism and counter-terrorism: Criminological perspectives. San Diego: Elsevier Ltd. DesChamps, S., Brantingham, P., & Brantingham, P. (1992). The British Columbia transit fare evasion audit. In R. Clarke (Ed.), Situational crime prevention: Successful case studies (1st ed., pp. 139–150). Albany: Harrow and Heston. Dugan, L., LaFree, G., & Piquero, A. (2005). Testing a rational choice model of airline hijackings. Criminology, 43(4), 1031–1066. Durlauf, S., & Nagin, D. (2011). Imprisonment and crime: Can both be reduced? Criminology and Public Policy, 10(1), 13–54. Eck, J. (1997). Preventing crime at places. In L. W. Sherman, D. Gottfredson, D. MacKenzie, J. Eck, P. Reuter, & S. Bushway (Eds.), Preventing crime: What works, what doesn’t, what’s promising. Washington: U.S. Department of Justice, National Institute of Justice. Eck, J. (2002). Preventing crime at places. In L. Sherman, D. Farrington, B. Welsh, & D. MacKenzie (Eds.), Evidence-based crime prevention (pp. 241–294). New York: Routledge. Eck, J., & Wartell, J. (1998). Improving the management of rental properties with drug problems: A randomized experiment. In L. G. Mazerolle & J. Roehl (Eds.), Crime prevention studies (Vol. 9, pp. 161–185). Monsey: Criminal Justice Press. Ekblom, P. (1988). Preventing post office robberies in London: Effects and side effects. Journal of Security Administration, 11, 36–43. Elias, B. (2010). Airport passenger screening: Background and issues for Congress. Washington: Congressional Research Service. Enders, W., & Sandler, T. (1993). The effectiveness of antiterrorism policies: A vector-autoregression-intervention analysis. The American Political Science Review, 87(4), 829–844. Enders, W., & Sandler, T. (2000). Is transnational terrorism becoming more threatening? Journal of Conflict Resolution, 44, 307–332. Enders, W., Sandler, T., & Cauley, J. (1990). UN conventions, terrorism, and retaliation in the fight against terrorism: An econometric evaluation. Terrorism and Political Violence, 2(1), 83–105. Englund, W., & Lally, K. (2011). Russia airport blast kills 35. Retrieved July 30, 2011, from http://www.washingtonpost.com/wp-dyn/content/article/2011/01/24/AR2011012406445.html. Farrell, G. (2005). Progress and prospects in the prevention of repeat victimization. In N. Tilley (Ed.), Handbook of crime prevention and community safety (pp. 143–172). Collumpton: Willan Publishing. Farrington, D., Bowen, S., Buckle, A., Burns-Howell, T., Burrows, J., & Speed, M. (1993). An experiment on the prevention of shoplifting. In R. Clarke (Ed.), Crime prevention studies (Vol. 1). Monsey: Criminal Justice Press.

248

C. Lum et al.

Farrington, D., Gill, M., Waples, S., & Argomaniz, J. (2007). The effects of closed-circuit television on crime: Meta-analysis of English national quasi-experimental multi-site evaluation. Journal of Experimental Criminology, 3(1), 21–38. Felson, M. (1994). Crime and everyday life: Insight and implications for society. Thousand Oaks: Pine Forge Press. Felson, M., Berends, R., Richardson, B., & Veno, A. (1997). Reducing pub hopping and related crime. In R. Homel (Ed.), Policing for prevention: Reducing crime, public intoxication and injury (Vol. 7, pp. 115–132). Monsey: Criminal Justice Press. Gabor, T. (1981). The crime displacement hypothesis: An empirical examination. Crime and Delinquency, 26, 390–404. Gil-Garcia, J., Schneider, C., Pardo, T., & Cresswell, A. (2005). Interorganizational information integration in the criminal justice enterprise: Preliminary lessons from state and county initiatives. Proceedings of the 38th Hawaii International Conference on System Sciences. Retrieved on March 1, 2011, from http://www.ctg.albany.edu/publications/journals/hicss_2005_interorganizational/hicss_2005_interorganizational.pdf. Guerette, R., & Bowers, K. (2009). Assessing the extent of crime displacement and diffusion of benefits: A review of situational crime prevention evaluations. Criminology, 47(4), 1331–1368. Hasisi, B., & Weisburd, D. (Forthcoming). Going beyond ascribed identities: The importance of procedural justice in airport security screening in Israel. Law and Society Review. Hesseling, R. (1995). Theft from cars: Reduced or displaced? European Journal of Criminal Policy and Research, 3, 79–92. Jackson, B. (2009). Organizational decisionmaking by terrorist groups. In P. Davis and K. Cragin (eds.), Social science for counterterrorism: Putting the pieces together. Santa Monica, CA: RAND. Kelling, G. L., Pate, A. M., Dieckman, D., & Brown, C. (1974). The Kansas City preventive patrol experiment: Technical report. Washington: Police Foundation. Kenney, D. (1986). Crime on the subways: Measuring the effectiveness of the Guardian Angels. Justice Quarterly, 3, 481–496. Knutsson, J., & Clarke, R. (Eds.). (2006). Putting theory to work: Implementing situational crime prevention and problem-oriented policing. Crime prevention studies (Vol. 20, pp. 89–110). Monsey: Criminal Justice Press. Koper, C. (1995). Just enough police presence: Reducing crime and disorderly behavior by optimizing patrol time in crime hotspots. Justice Quarterly, 12(4), 649–672. LaFree, G. (2011). Generating terrorism event data bases: Results from the Global Terrorism Database, 1970 to 2008. In C. Lum & L. W. Kennedy (Eds.), Evidence-based counterterrorism policy. New York: Springer. LaFree, G., & Dugan, L. (2004). How does studying terrorism compare to studying crime? In M. Deflem (Ed.), Terrorism and counter-terrorism: Criminological perspectives, sociology of crime, law and deviance (Vol. 15, pp. 53–75). Oxford: Elsevier. Landes, W. (1978). An economic study of U.S. aircraft hijackings, 1961–1976. Journal of Law and Economics, 21, 1–31. Langan, P., Greenfeld, L., Smith, S., Durose, M., & Levin, D. (2001). Contacts between police and the public: Findings from the 1999 national survey. Washington: Bureau of Justice Statistics. Lasley, J. (1996). Using traffic barriers to ‘design out’ crime: A program evaluation of LAPD’s Operation Cul-De-Sac. Report to the national institute of justice. Fullerton: California State University. Lasley, J. (1998). “Designing out” gang homicides and street assaults. Washington: National Institute of Justice. Laycock, G. (1985). Property marking: A deterrent to domestic burglary? Crime Prevention Unit Paper no. 3. London: Home Office. Laycock, G. (1991). Operation identification, or the power of publicity? Security Journal, 2, 67–72. Laycock, G., & Austin, C. (1992). Crime prevention in parking facilities. Security Journal, 3, 154–160.

10

Translational Criminology: Assessing Airport Security

249

Lum, C. (2009). Translating police research into practice. Ideas in American policing. Washington: Police Foundation. Lum, C., Beech, D., Connors, M., Crafton, Z., Parsons, R., & Smarr, T. (2007). Examining collateral effects of counterterrorism interventions: Airport security screening. Congressional Briefing, Senate Russell Building, September 10, 2007, Washington. Lum, C., Haberfeld, M., Fachner, G., & Lieberman, C. (2009). Police Activities to Counter Terrorism: What We Know and What We Need To Know. In David Weisburd, Thomas Feucht, Idit Hakimi, Lois Mock and Simon Perry (Eds.), To Protect and to Serve: Police and Policing in an Age of Terrorism - and Beyond. New York, NY: Springer. Lum, C., Kennedy, L., & Sherley, A. (2006). Are counter-terrorism strategies effective? The results of the Campbell Systematic Review on counter-terrorism evaluation research. Journal of Experimental Criminology, 2(4), 489–516. Lum, C., & Koper, C. (2011). Is crime prevention relevant to counter-terrorism? In B. Forst, J. Greene, & J. Lynch (Eds.), Criminologists on terrorism and homeland security. Cambridge: Cambridge University Press. Lum, C., Koper, C., & Telep, C. (2009). Evidence-based policing Matrix online tool. Retrieved on March 1, 2011, from http://gemini.gmu.edu/cebcp/matrix.html. Lum, C., Koper, C., & Telep, C. (2011). The evidence-based policing Matrix. Journal of Experimental Criminology, 7(1), 3–26. Mazerolle, L., Price, J., & Roehl, J. (2000). Civil remedies and drug control: A randomized field trial in Oakland, CA. Evaluation Review, 24(2), 212–241. Mazerolle, L., & Ransley, J. (2005). Third party policing. Cambridge: Cambridge University Press. McCord, J. (2003). Cures that harm: Unanticipated outcomes of crime prevention programs. Annals of the American Academy of Political and Social Sciences, 587, 16–30. McGarrell, E., Chermak, S., Weiss, A., & Wilson, J. (2001). Reducing firearms violence through directed police patrol. Criminology and Public Policy, 1, 119–148. Mythen, G., & Walklate, S. (2006). Criminology and terrorism: Which thesis? Risk society or governmentality? British Journal of Criminology, 46(3), 379–398. Nagin, D. (1998). Criminal deterrence research at the outset of the 21st century. Crime and Justice: A Review of Research, 23, 1–42. National Commission on Terrorist Attacks Upon the United States. (2004). The 9/11 Commission Report. New York: W.W. Norton & Company. National Research Council, The National Academies. (2004). Fairness and effectiveness in policing: The evidence. In W. Skogan & K. Frydl (Eds.), Committee to review research on police policy and practices, committee on law and justice, division of behavioral and social sciences and education. Washington: National Academies Press. National Research Council, The National Academies. (2010). Field evaluation in the intelligence and counterintelligence context: Workshop summary. Washington: The National Academies Press. Newman, O. (1972). Defensible space: Crime prevention through urban design. New York: Macmillan. Newman, O. (1996). Creating defensible space. Washington: U.S. Department of Housing and Urban Development. Poyner, B. (1994). Lessons from Lisson Green: An evaluation of walkway demolition on a British housing estate. In R. Clarke (Ed.), Crime prevention studies (Vol. 3). Monsey: Criminal Justice Press. Poyner, B., & Warner, C. (1986). Violence to staff: A basis for assessment and prevention. London: Tavistock. Putnam, S., Rockett, I., & Campbell, M. (1993). Methodological issues in community-based alcohol-related injury prevention projects: Attribution of program effects. In T. Greenfield & R. Zimmerman (Eds.), Experience with community action projects: New research in the prevention of alcohol and other drug problems (Vol. 14, pp. 31–39). Rockville: Center for Substance Abuse Prevention. Repetto, T. (1976). Crime prevention and the displacement phenomenon. Crime and Delinquency, 22, 166–177. Riley, K., & Hoffman, B. (1995). Domestic terrorism: A national assessment of state and local preparedness. Santa Monica: RAND.

250

C. Lum et al.

Rosenfeld, R. (2004). Terrorism and criminology. In M. Deflem (Ed.), Terrorism and counter-terrorism. Sociology of crime, law and deviance (Vol. 5, pp. 19–32). Oxford: Elsevier. Saltz, R. (1987). The role of bars and restaurants in preventing alcohol-impaired driving: An evaluation of server education. Evaluation in Health Professions, 10, 5–27. Sheehan, I. S. (2011). Assessing and comparing data sources for terrorism research. In C. Lum & L. W. Kennedy (Eds.), Evidence-based counterterrorism policy. New York: Springer. Sherman, L. (1990). Police crackdowns: Initial and residual deterrence. In M. Tonry & N. Morris (Eds.), Crime and justice: An annual review of research (Vol. 12). Chicago: University of Chicago Press. Sherman, L. (1997). Policing for crime prevention. In L. Sherman, D. Gottfredson, D. MacKenzie, J. Eck, P. Reuter, & S. Bushway (Eds.), Preventing crime: What works, what doesn’t, what’s promising (pp. 295–329). Washington: National Institute of Justice. Sherman, L., & Eck, J. (2002). Policing for crime prevention. In L. Sherman, D. Farrington, B. Welsh, & D. MacKenzie (Eds.), Evidence-based crime prevention (pp. 295–329). New York: Routledge. Sherman, L., Farrington, D., Welsh, B., & MacKenzie, D. (2002). Evidence-based crime prevention. New York: Routledge. Sherman, L., Gartin, P., & Buerger, M. (1989). Hot spots of predatory crime. Criminology, 27(1), 27–56. Sherman, L., Shaw, J., & Rogan, D. (1995). The Kansas City gun experiment. Research in brief. Washington: National Institute of Justice, U.S. Department of Justice. Sherman, L., & Weisburd, D. (1995). General deterrent effects of police patrol in crime “hot spots”: A randomized, controlled trial. Justice Quarterly, 12(4), 625–648. Tilley, N., & Webb, J. (1994). Burglary reduction: Findings from safer cities schemes. London: Home Office. Retrieved on March 1, 2011, from http://webarchive.nationalarchives.gov.uk/ 20110220105210/rds.homeoffice.gov.uk/rds/prgpdfs/fcpu51.pdf. Transportation Research Board of the National Academies. (2003). Airport research needs: Cooperative solutions. Special Report 272. Washington: National Academies. Retrieved on March 1, 2011, from http://onlinepubs.trb.org/onlinepubs/sr/sr272.pdf. Tyler, T. (1990). Why people obey the law: Procedural justice, legitimacy, and compliance. New Haven: Yale University. Tyler, T. (2003). Procedural justice, legitimacy, and the effective rule of law. Crime and Justice, 30, 283–357. Tyler, T. (2004). Enhancing police legitimacy. The Annals of the American Academy of Political and Social Science, 593(1), 84–99. Tyler, T. (2011). Toughness vs. fairness: Police policies and practices for managing the risk of terrorism. In C. Lum & L. W. Kennedy (Eds.), Evidence-based counterterrorism policy. New York: Springer. U.S. Department of Homeland Security. (2008). National infrastructure protection plan: Transportation systems sector. Retrieved on January 14, 2011, from http://www.dhs.gov/xlibrary/assets/nipp_ transport.pdf. U.S. Government Accountability Office. (2007). Risk, experience, and customer concerns drive changes to airline passenger screening procedures, but evaluation and documentation of proposed changes could be improved. Retrieved on January 14, 2011, from http://www.gao.gov/ new.items/d07634.pdf. U.S. Government Accountability Office. (2009). DHS and TSA have researched, developed, and begun deploying passenger checkpoint screening technologies, but continue to face challenges. Retrieved on January 14, 2011, from http://www.gao.gov/new.items/d10128.pdf. U.S. Government Accountability Office. (2010). TSA has taken actions to manage risk, improve coordination, and measure performance, but additional actions would enhance its actions. Retrieved January 14, 2011, from http://www.gao.gov/new.items/d10650t.pdf. U.S. Transportation Security Administration. (2010). What we do. Retrieved March 1, 2011, from http://www.tsa.gov/what_we_do/layers/index.shtm. U.S. Transportation Security Administration. (2011). Our history. Retrieved March 1, 2011, from http://www.tsa.gov/research/tribute/history.shtm.

10

Translational Criminology: Assessing Airport Security

251

USA Today/Gallup Poll. (2010). Gallup Poll daily tracking. Retrieved March 1, 2011, from http:// www.pollingreport.com/terror.htm. van Andel, H. (1989). Crime prevention that works: The case of public transport in the Netherlands. British Journal of Criminology, 29, 47–56. Webb, B., & Laycock, G. (1992). Reducing crime on the London underground: An evaluation of three pilot projects (Vol. 30). London: Home Office. Weisburd, D. (1997). Reorienting criminal justice research and policy: From the causes of criminality to the context of crime. Washington: National Institute of Justice. Weisburd, D. (2002). From criminals to criminal contexts: Reorienting crime prevention research and policy. In D. Weisburd & E. Waring (Eds.), Crime & social organization, advances in criminological theory (Vol. 10, pp. 197–216). New Brunswick: Transaction Publishers. Weisburd, D. (2008). Place-based policing. Ideas in American policing. Washington: Police Foundation. Weisburd, D., Bushway, S., Lum, C., & Yang, S. (2004). Trajectories of crime at places: A longitudinal study of street segments in the city of Seattle. Criminology, 42(2), 283–322. Weisburd, D., & Eck, J. (2004). What can the police do to reduce crime, disorder, and fear? The Annals of the American Academy of Political and Social Science, 593, 42–65. Weisburd, D., Feucht, T., Hakimi, I., Mock, L., & Perry, S. (Eds.). (2009). To protect and to serve: Police and policing in an age of terrorism – and beyond. New York: Springer. Weisburd, D., & Green, L. (1995). Policing drug hot spots: The Jersey City drug market analysis experiment. Justice Quarterly, 12(4), 711–735. Weisburd, D., Groff, E., & Yang, S. M. (2009). Understanding developmental crime trajectories at places: Social disorganization and opportunity perspectives at micro units of geography. Washington: National Institute of Justice. Weisburd, D., Jonathan, T., & Perry, S. (2009). The Israeli model for policing terrorism: Goals strategies, and open questions. Criminal Justice and Behavior, 36(12), 1259–1278. Weisburd, D., Lum, C., & Petrosino, A. (2001). Does research design affect study outcomes in criminal justice? Annals of the American Academy of Political and Social Sciences, 578, 50–70. Weisburd, D., Telep, C., Hinkle, J., & Eck, J. (2010). Is problem-oriented policing effective in reducing crime and disorder? Findings from a Campbell systematic review. Criminology and Public Policy, 9(1), 139–172. Weisburd, D., Wyckoff, L., Ready, J., Eck, J., Hinkle, J., & Gajewski, F. (2006). Does crime just move around the corner? A controlled study of spatial displacement and diffusion of crime control benefits. Criminology, 44, 549–592. Welsh, B., & Farrington, D. (2008a). Effects of closed circuit television surveillance on crime. Campbell Systematic Reviews 2008:17. DOI: 10.4073/csr.2008.17. Welsh, B., & Farrington, D. (2008b). Effects of improved street lighting on crime. Campbell Systematic Reviews 2008:13. DOI: 10.4073/csr.2008.13. Welsh, B., Mudge, E., & Farrington, D. (2010). Reconceptualizing public area surveillance and crime prevention: Security guards, place managers, and defensible space. Security Journal, 23(4), 299–319. Wilson, J., & Kelling, G. (1982). Broken windows: The police and neighborhood safety (pp. 29–38). March: Atlantic Monthly.

Part IV

Perspectives in Evaluating Counterterrorism Policy

Chapter 11

Terrorist Finance, Informal Markets, Trade and Regulation: Challenges of Evidence Regarding International Efforts Nikos Passas

Introduction Terrorism is not a new problem, and politically motivated offending has been handled in a variety of ways from ancient Greece to modern times (Passas, 1986). The process of globalization and the cross-border movement of ideas, information, people, goods, and capital, have brought terrorism to the fore of public, media, and government attention. Policymakers, scholars, and others have increased their efforts to understand, detect, and prevent this phenomenon, which defies scientific and universal definition (National Research Council, 2002). A necessary and important piece of this puzzle involves the financing of terrorist activities. In the aftermath of the 11 September 2001 attacks in New York and Washington, DC (hereafter “9/11”), countering of terrorist funding took center stage as a focus of global law enforcement and policy (Giraldo & Trinkunas, 2007; Naylor, 2006; Passas, 2008; Pieth, Thelesklaf, & Ivory, 2009; Warde, 2007). Terrorist fund-raising has been traced almost to every legitimate and criminal source imaginable. Especially when no analytical distinction is made between terrorism and other phenomena, such as insurgency, guerilla warfare, civil war, or independence and liberation movements, the historical record shows that militant groups and organizations are likely to resort to any means available to them for the raising and transfer of funds in support of their operations and activities. The growing infrastructure of anti-money laundering and counter financing of terrorism (hereafter AML/CFT) resulting from 9/11 has highlighted the importance of untangling the myriad of groups involved in terrorist activities and focusing on those representing the highest threat to security and economic interests worldwide. Al Qaeda (hereafter AQ) and AQ-inspired groups came thus to the top of policy and

N. Passas (*) School of Criminology and Criminal Justice, Northeastern University, Boston, MA, USA e-mail: [email protected]

C. Lum and L.W. Kennedy (eds.), Evidence-Based Counterterrorism Policy, Springer Series on Evidence-Based Crime Policy 3, DOI 10.1007/978-1-4614-0953-3_11, © Springer Science+Business Media, LLC 2012

255

256

N. Passas

research priorities. The CFT arsenal, led by the United Nations and United States of America, gradually grew in scope and applicability to a wide range of groups classified by national and international organizations as “terrorist.” The objectives served by CFT are mainly to monitor illicit and militant activities, to prevent terrorist actions, to reduce the harm of terror attacks, and to facilitate the investigative work seeking to identify coconspirators and facilitators. Neither the utility of nor the need for controls against terrorism finance are questioned, but these must be matched with reasonable expectations, in conjunction with other counterterrorism measures, on the basis of evidence and analysis, fairly and in compliance with more general national and international legal standards, including those with respect to universal human rights. Just as AML measures and their effects must be properly evaluated (Levi & Reuter, 2006), so do CFT measures, some of which are quite drastic and invasive. The main questions are: do current policies and measures effectively further the stated AML/ CFT goals? Are the highest risks identified and addressed? Unfortunately, policies and practices over the past 9 years suggest that, while some progress has been made since the first responses to the 9/11 attacks, there is plenty of room for improvement. Our knowledge is incomplete, mainly due to a lack of available reliable data that could be collected and analyzed in a systematic and comprehensive way. Our fragmented collective vision of the social organization of terrorist groups and the financial aspects of it, in particular, has allowed controversies around several issues – the role played by nonprofit organizations and charities, the informal sector (compared to the formal financial system, which is presumed to be well regulated and more transparent), the trade in various commodities (especially precious stones, gold, tobacco, or counterfeit goods, but also narcotics, firearms, counterfeit pharmaceuticals, and even human trafficking), the nexus between terrorist groups, and “organized crime,” especially links between drug trafficking and terror groups, etc. The lack of proper “peer review” of publications about terrorism have allowed inaccurate, wrong, and misleading interpretations to enter into official thinking and policy planning. Research into the areas covered in this paper revealed erroneous descriptions, baseless assumptions, uncritical reproductions of secondary sources and misleading analyses (Brown, 2006; Dandurand & Chin, 2004; Passas, 2004a, 2006a; Passas & Jones, 2006; Warde, 2007). An important implication of such findings is that studies not designed to achieve considerable depth and to “audit” the sources of much of the proliferating terrorism literature are susceptible to rely on and reproduce “facts by repetition.” A large portion of conventional wisdom guiding policy is unfortunately not founded on solid evidence, but reflects perceptions shaped by superficial, incomplete, or wrong information, which has been repeated and regarded as accurate in scholarly and policy documents. Another finding was that serious action is lacking with respect to an area vulnerable to serious abuse: international trade (Passas, 2006b). Trade lends itself not only to significant commercial and economic crimes, but also to practices supportive of corruption, expensive forms of terrorism, proliferation of controlled weapons, and sanctions/ embargo violations (see Passas, Anthony, Deanaz, & Walker, 2010). There is no transparency currently in the area of trade and commerce, where transactions worth trillions of dollars annually can easily serve to obscure illicit fund-raising and -transfer activities.

11

Terrorist Finance, Informal Markets, Trade and Regulation…

257

The consequences of CFT policies not based on evidence and rigorous analysis include missed targets, false positives, false negatives, economic and political damage, and the undermining of national and international security. Further, many control functions have been de facto outsourced to the private sector without clear laws/regulations, proper guidance, and accountability. Rushed actions and uncorrected errors reduce the ability of law enforcement agencies to effectively collaborate with each other and overseas counterparts. Ultimately, the legitimacy of counterterrorism initiatives suffers fueling attitudes sympathetic or conducive to extremism – esp. in the context of additional controversies revolving around due process and respect for human rights in the “war on terror” (see e.g., Arar Report, 2006; Physicians for Human Rights, 2008). Clearly, the potential abuse of informal financial networks and trade activities for terrorist support must be addressed. However, current practices make it less likely that controllers will be able to detect them and avert militant attacks. This issue needs urgent attention. This paper outlines at first the main aspects of CFT and then focuses on informal fund transfer methods illustrating the above points. It then concludes with policy recommendations and suggestions about ways in which the private sector can support more efficiently and productively the objectives of CFT policies, such as transparency and traceability of transactions.

Financial Controls and Terrorism Two developments in social control have progressed significantly since the 1980s with respect to serious crime – mostly of cross-border nature. The first one regards a strong emphasis on a follow-the-money approach targeting profitable underlying offenses. The second one refers to a shift of control functions and responsibilities to the private sector. The follow-the-money approach has been complementing (and occasionally, replacing) efforts to target the predicate profitable misconduct, ever since the application of tax rules against criminal entrepreneurs like Al Capone, but it moved into a new phase when the criminalization of money laundering occurred in the USA and internationally. The priority at that time was the illegal drug trade, as illustrated by the 1988 UN Vienna Convention, and the significant proceeds it generates. The list of predicate offenses has been growing, however, covering most for-profit crimes (Gilmore, 2004; van Duyne & Levi, 2005). Parallel to this development has been a reliance on non-state actors to carry out significant control activities. Financial institutions, in particular, have been charged with duties ranging from due diligence and knowledge of their clients (and their clients’ clients sometimes) to record keeping and filing reports of suspicious activities or transactions to the authorities. In this context, what used to be a professional duty and lauded as “customer confidentiality” gradually turned into crime-facilitative “bank secrecy” (Levi, 1991). Despite criticisms that this approach to combating serious crime is limited and at times counterproductive (Naylor, 1999; van Duyne & de Miranda, 1999), the

258

N. Passas

emphasis on financial aspects and a de facto partial outsourcing or externalization of control responsibilities occurred also with respect to counterterrorism (Levi & Reuter, 2006). Prior to the 9/11 attacks, AML provisions were under review in the USA, as benefits did not seem to convincingly outweigh costs. 9/11, however, completely changed the context and brought about the strengthening and expansion of previous measures with the addition of CFT aims. Interestingly, the first reference to financial controls against terrorism can be found in Europe, when a 1980 Council of Europe Recommendation (R(80) 10) reflected official concern about operations mounted by the Red Army Faction and Red Brigades by citing “acts of criminal violence such as holdups and kidnappings.” This Recommendation referred to customer due diligence practices that later on became part of the core requirements for banks against money laundering. The basis for the current framework countering terrorist finance (CFT) was laid by the following United Nations initiatives. UN Security Council (UNSC) Resolution 1267 (1999) introduced sanctions against AQ, Osama bin Laden, the Taliban, and “associated individuals and entities.” Apart from a travel ban and an arms embargo, the sanctions included the freezing of assets and economic resources of those placed on a “Consolidated List” of individuals and entities. This List is maintained by the UNSC Committee established in order to oversee the implementation of these sanctions, which were further elaborated and reiterated by Resolutions 1333 (2000), 1390 (2002), 1455 (2003), 1526 (2004), 1617 (2005), 1735 (2006), 1822 (2008) and 1963 (2010). On 28 September 2001, going beyond the Taliban and AQ, the UNSC adopted resolution 1373 under Chapter VII of the UN Charter (i.e., legally binding upon UN Member States), which mandated the prevention, suppression, and criminalization of the financing of terrorist acts. It also required that Member States “Freeze without delay funds and other financial assets or economic resources of persons who commit, or attempt to commit, terrorist acts or participate in or facilitate the commission of terrorist acts; of entities owned or controlled directly or indirectly by such persons; and of persons and entities acting on behalf of, or at the direction of such persons and entities, including funds derived or generated from property owned or controlled directly or indirectly by such persons and associated persons and entities” (Para. 1 [c]). The implementation of Resolution 1373 is monitored by the UN Committee against Terrorism. The same Resolution called for the ratification of the International Convention for the Suppression of the Financing of Terrorism, which came into force in April 2002. This UN convention defined the financing of terrorism as follows: 1. Any person commits an offence within the meaning of this Convention if that person by any means, directly or indirectly, unlawfully and willfully, provides or collects funds with the intention that they should be used or in the knowledge that they are to be used, in full or in part, in order to carry out: (a) An act which constitutes an offence within the scope of and as defined in one of the treaties listed in the annex [the pre-existing ‘universal instruments’ against terrorism]; or (b) Any other act intended to cause death or serious bodily injury to a civilian, or to any other person not taking an active part in the hostilities in a situation of armed conflict, when the purpose of such act, by its nature or context, is to intimidate a population, or to compel a government or an international organization to do or to abstain from doing any act (Art. 2).

11

Terrorist Finance, Informal Markets, Trade and Regulation…

259

Interestingly, the last paragraph appears to offer a substantive definition of “terrorism,” albeit in an unsatisfactory fashion, given the strong reliance on subjective elements (intent, purpose), which are open to diverse interpretations. Article 6 mandates that terrorist acts will be “under no circumstances justifiable by considerations of a political, philosophical, ideological, racial, ethnic, religious, or other similar nature.” Nevertheless, the convention goes on to provide that a government may deny an extradition or mutual legal assistance request, if it has “substantial grounds” to believe that it “has been made for the purpose of prosecuting or punishing a person on account of that person’s race, religion, nationality, ethnic origin, or political opinion or that compliance with the request would cause prejudice to that person’s position for any of these reasons” (Art. 15). In any event, the convention mandates the criminalization of terrorist financing as an independent offense, liability for legal persons (criminal, civil, or administrative) and measures enabling the identification, detection, freezing, seizure, and forfeiture of related assets. The convention covers primarily cases with transnational aspects and does not apply to purely national/domestic cases. It has been argued that the introduction of an independent offense of “terrorist financing” is necessary for reasons similar to those of the offense of money laundering: evidence of the principal offense may be absent. That is, terrorist acts may have not occurred or attempted – the sanctioning of terrorist financing would have required otherwise the establishment of “actual guilt for the principal terrorist offense” (Koh, 2006, p. 66). Moreover, the argument has been made that terrorist financing constitutes a continuous and broader harm than that caused by terrorist acts (ibid.). The convention also incorporates a “gatekeeper” role for the private sector, as it requires that states have: Measures requiring financial institutions and other professions involved in financial transactions to utilize the most efficient measures available for the identification of their usual or occasional customers, as well as customers in whose interest accounts are opened, and to pay special attention to unusual or suspicious transactions and report transactions suspected of stemming from a criminal activity (Art 18 [1][b]).

Examples of specific steps States can take are then provided, essentially drawing on Financial Action Task Force (FATF) Recommendations, even though this parallel is not acknowledged in the text (Gilmore, 2004). The FATF was created in 1989 in order to assess cooperation geared toward the prevention of the abuse of the banking system and financial institutions for money laundering and to consider new preventive efforts “including the adaptation of the legal and regulatory systems as to enhance multilateral judicial assistance.” The scope of the FATF’s mandate has widened and it currently includes: • Global surveillance of evolving criminal and terrorist threats. • Responses to new threats which affect the integrity of the financial system such as proliferation finance. • A stronger partnership with the private sector. • Support for global efforts to raise standards, especially in low capacity countries (FATF mandate for 2008–2012).

260

N. Passas

The FATF initially introduced 40 Recommendations, which were revised in 1996 and again in 2003. Even though the Recommendations are not formally binding and constitute “soft law,” their effect has been rather powerful, partly due to the threat of being placed on the list of noncooperating countries and territories (NCCT) and partly because of their use for highly visible country evaluations through FATF processes as well as by the International Monetary Fund, which exert pressures on noncomplying countries. Right after the attacks of 9/11, the FATF introduced eight new Recommendations aiming at terrorist finance and added a ninth in 2004. While these Recommendations parallel the UN Convention and UNSC Resolution 1373, they go beyond them in important ways. For example, they call on States to ensure that terrorist finance is a predicate offense for money laundering, cover what they call “alternative remittance systems” (see more below), nonprofit organizations, wire transfers, and cash couriers. Guidance notes and best practices were also published by the FATF to assist in the process of implementation. These Recommendations did not really add any specific terrorism finance aspects or regulation. Rather, with the exception of charitable organizations, they resuscitated earlier concerns and discussions with respect to money laundering. Further, the Council of Europe 2005 Convention on Laundering, Search, Seizure and Confiscation of the Proceeds from Crime and on the Financing of Terrorism, known as the Warsaw Convention, is the first international instrument to provide for preventive measures, international cooperation and controls against both money laundering and terrorism finance. In the European Union, the Third Money Laundering Directive repealed and replaced the two previous Directives, while specifically addressing terrorist financing as well (Directive 2005/60 EC on the Prevention of the Use of the Financial System for the Purpose of Money Laundering and Terrorist Financing, OJ L 309, 25 November 2005, pp. 15–36; as amended by Directives 2007/64/EC and 2008/20/ EC). This Directive defines terrorism finance as “the provision or collection of funds, by any means directly or indirectly, with the intention that they should be used or in the knowledge that they are to be used, in full or in part in order to carry out any of the offenses that have been defined as terrorism” (Art. 1[4]). The European Union has implemented UNSC Resolutions and FATF Recommendations dealing with sanctions and preventive measures through a series of Common Positions and Regulations. Common Position 2002/402/CFSP adopted the UNSC measures with respect to Taliban and AQ related persons and entities. Regulation 881/2002 repealed earlier regulations, implemented sanctions against Al-Qaida, bin Laden and the Taliban wherever located, and appended the UN List in an Annex. The European Commission is required to update this Regulation and its annex every time the UN List is amended (it has been amended more than a hundred times since). Common Positions 2001/930 and 2001/931 implemented UNSC Resolution 1373. The second Common Position provides for the freezing of assets of persons, groups, or entities listed in the Annex. It also mandates that Member States “through police and judicial cooperation in criminal matters within the framework of Title VI

11

Terrorist Finance, Informal Markets, Trade and Regulation…

261

of the Treaty on European Union, afford each other the widest possible assistance in preventing and combating terrorist acts. To that end they shall, with respect to enquiries and proceedings conducted by their authorities in respect of any of the persons, groups and entities listed in the Annex, fully exploit, upon request, their existing powers in accordance with acts of the European Union and other international agreements, arrangements, and conventions which are binding upon Member States” (Art. 4). This annexed list is updated every 6 months. EU Council Regulation 2580/2001 on specific restrictive measures directed against certain persons and entities with a view to combating terrorism (OJ L 344 of 28 December 2001, p. 70) provided that the Council reviews and amends this list by unanimity. So, the EU has adopted the UN Taliban and AQ lists as well as its own “autonomous” terror lists. While the European Union has been acting in concert with the UNSC, the European Court of Justice has established the principle of review of such lists by taking into consideration both security needs and the protection of fundamental human rights and the principle of legal redress (see Kadi decision: ECJ, 3 September 2008, Kadi et Al Barakaat, C-402/05 P and C-415/05 P). The primacy of UN law is recognized while at the same time it is confirmed that international requirements cannot jeopardize the constitutional principles of the EC Treaty (Labayle & Long, 2009). Given the due diligence, monitoring, record-keeping, and reporting roles the private sector has been called to play – starting with financial institutions, but then with a growing list of nonbank institutions and a range of professions – a critical tool used for bridging public and private sector entities has emerged: Financial Intelligence (or Information) Units (FIUs). These can receive suspicious activity/ transaction reports, analyze them and forward them to investigative agencies for further action. They are well positioned to form a comprehensive picture of money laundering, terrorism finance, and other types of misconduct, which can lead to guidance and feedback to reporting entities in the private sector. They can also use this intelligence for policy analysis, reform, and construction. Thus, FIUs can serve both operational and strategic purposes. In addition, FIUs are vital institutions for international cooperation. The establishment of FIUs has now become a routine requirement in major international instruments (including the UN conventions against transnational organized crime, corruption, and financing of terrorism). In 1995, an informal group of FIUs was established under the name of Egmont Group, in order to further strengthen international cooperation. Since 2007, the Egmont Group, based in Toronto, Canada, convenes annual meetings and supports the establishment, coordination, exchange of information, and increased effectiveness of FIUs around the world. Finally, the World Bank and the International Monetary Fund defined TF as “the financial support, in any form, of terrorism or of those who encourage, plan, or engage in it” (World Bank & IMF, 2006) and have incorporated AML/CFT in their work and operations. CFT was added to AML, and efforts have been intensified in the aftermath of 9/11. AML/CFT is considered as part of their developmental mandate and they support countries through technical assistance, assessments, and policy construction.

262

N. Passas

This work is consistent with the Recommendations and guidance from the FATF and FATF-style regional groups. UNSC Resolutions, initiatives from the European Union, international organizations have raised the profile of financial controls. Combined with other international standards relative to security, transnational organized crime, sanctions, proliferation, corruption, cyber-crime, environment, etc. have brought about a “regulatory tsunami” often overwhelming national governments. CFT laws and measures grew in number and scope significantly at national level and in the private sector. Consequently, a powerful regulatory and control arsenal has been put in place by governments and private sector entities alert to the possibility of being abused by extremists around the world. Several lists were created and circulated, many persons and entities were identified as suspect of terrorism, assets have been frozen around the world, and a great deal of ink has been used for debates on CFT laws, measures, and the need for further international cooperation and action.

Has the CFT Approach Worked So Far? The main questions at this point are: Has it been effective? Has it been fair and commensurate to the threat? In essence, most action has been in the direction of treating CFT as the same or very similar to AML with the addition of designation lists to watch out for. The use of the same approaches to AML and CFT has raised the question of how measures with questionable effectiveness against the very substantial amounts involved in money laundering, could do a better job against the much smaller amounts necessary for terrorist acts and most terror groups. To illustrate1: • Madrid 2004 – about €15,000 (operational costs); explosives obtained through a barter deal for drugs with street value of about €35,000 • Bali nightclub bombings – about $20,000 • US embassy bombings in Kenya and Tanzania – about $10,000 • Istanbul attacks – less than $40,000 • 9/11 – about $320,000 for 19 hijackers over about 2 years • Paris bombs – a few hundred €uro • USS Cole 2000 attack in Aden – less than $10,000 • Bishopsgate IRA attack – £3,000 • London 2005 attacks – a few hundred £pounds • Jakarta 2003 Marriott Hotel bombing – about $30,000 • Chechnya: $4,000 for airplanes; $7,000 for attacks on Kashirskoye Highway and near metro station; Nord-West operation in Beslan $9,500 1

Sources: Personal interviews with investigators and prosecutors from the US, UK, France, Germany, Spain, Turkey, FBI; UN Monitoring Team reports; on Jordan: Air Security International; on Chechnya: Shamil Basaev statement; on US East Africa embassy and Bali bombings, National Commission on Terrorist Attacks upon the US, 2005: 27–28.

11

Terrorist Finance, Informal Markets, Trade and Regulation…

263

• Germany: 2006 train bomb attempt – €200; Cologne bomb $241 • Air India bombings – 3,000 CAD In addition, an official inquiry into the London bombings in 2005 estimated the total cost of overseas and UK trips, bomb-making equipment, rent, car hire, to be less than £8,000. This was funded through defaulted loans, account overdrafts, and checks that eventually bounced (House of Commons, 2006, p. 23). As Levi and Reuter (2006) have pointed out, the evidence suggests that AML has done little against serious crime. Investigative, prosecutorial, and asset seizure objectives have not been furthered as much as expected of the “follow-the-money” approach, which may yield intelligence benefits when applied to terrorism, but also raises the compliance costs to the private sector. Reports from several countries indicate that dealing with CFT in ways similar to AML has caused difficulties and suggest that a rethinking of national and international approaches is in order. UNSC 1267 Monitoring Teams continue to pursue effective, uniform and consistent application of the measures by all Member States (see reports at http://www.un.org/docs/sc/committees/1267/1267mg.htm). Yet, the UN High Level Panel on Threats, Challenges and Change pointed out that the impact of the sanctions on the operation of AQ or al-Qaeda inspired groups has been rather limited (2004, para. 153). This was partly due to an imperfect understanding of AQ and its operations (Burke, 2003). In addition, new laws have been ill-prepared or simply transplanted from other jurisdictions or legal systems. For instance, technical assistance provided by wealthy countries to governments seeking help to cope with the “regulatory tsunami” has not always been of high quality. The US Government Accountability Office has reported that private contractors employed by US agencies in order to assist other countries in drafting laws meeting international standards “did not result in laws meeting FATF standards” or “had substantial deficiencies” (2005, p. 17). More to the point, even when FATF standards are met, there is absolutely no guarantee that they will have the desired effect. It has been observed that even if all of the FATF 40 + 9 Recommendations had been effectively implemented globally before 9/11, the hijackers’ financial activities and transactions would not have raised suspicions or led to Suspicious Activity Reports because they were mundane and indistinguishable in the midst of massive daily flows. Even FIU officials suggest off-the-record that the criteria and indicators offered by the recommendations could not have captured these activities. Hence, the utility of the “terrorist finance” special recommendations is not obvious (Passas, 2006a). As scholarly and policy debates grew heated at times, many have raised concerns about how exactly are suspects’ names placed on designation lists, through what process should innocent parties’ names be removed, how long should names remain on lists without prosecutorial or judicial action and how should frozen assets be used or disposed of. There is no consistent, universal, or harmonized approach to these questions and the guilt or innocence of those subject to executive decisions remains unsettled for many years despite some recent lawsuits and court rulings (e.g., the Kadi case cited earlier). As a synthesis report noted, “Many of the national laws and measures

264

N. Passas

against the finance of terrorism are driven by external and international institutions, the approaches are inconsistent and asymmetrical, the procedures and processes are nontransparent, the criteria are unclear, preventive and temporary measures last for extended periods of time and have a severe impact on those subject to them, errors have been made repeatedly and disagreements on who is and who is not a terrorist are unlikely to be resolved in the immediate future” (Passas, 2008, p. 342). Not only are many measures ineffective, they may also have counterproductive consequences by generating incentives for financial or other flows to shift to channels less regulated or monitored. An important challenge with designation lists has been the vagueness and inaccuracy of information or the inclusion of common names that may affect innocent parties and slow down the whole process. This is well illustrated by the business of Western Union in Dubai, where agents had to check every “Khalid, Mustafa and Omar” against these lists, ask for further documentation and hold up the funds transfer (Gale, 2006). Such practices motivate remitters to turn either to informal channels or to the use of false names. Either way, the objectives of transparency and traceability are defeated. In some instances, such as with respect to list-based sanctions and asset freezes, we have reached the point where it is occasionally impossible for governments to comply both with their own fundamental legal principles and with UNSC requirements, a situation that undermines legitimacy and cooperation without enhancing security. As the UN Monitoring Team pointed out in October 2009: “Such challenges have the potential to undermine the authority of the Security Council to impose sanctions. If States cannot implement decisions taken by the Council under Chapter VII of the Charter of the United Nations without contravening their own laws, the global community will lose the power to take coordinated action against threats to international peace and security” (UN Monitoring Team, 2009, p. 5). The argument in favor of new and specific laws criminalizing terrorism finance to facilitate criminal prosecutions (see above) is undercut not only by the fact that CFT is in practice AML with lists attached to it, but also by the fact that few such cases have been brought with such charges. The dominant opinion among controllers is that, especially when no terrorist act has been committed yet, it is hard to prove terrorism finance beyond reasonable doubt. For similar reasons, terrorism and finance provisions in the United Kingdom under the Prevention of Terrorism (Temporary Provisions) Act did not lead to any prosecutions. As Gold and Levi remarked, “the prosecution would first have to prove ‘beyond reasonable doubt’ that the funds were the proceeds of terrorism, and not the proceeds of any other crime or legitimate business: this may be why there have not been any such prosecutions to date, either in Great Britain or in Northern Ireland: it is often easier to prosecute for the substantive crime” (1994, fn 9; emphasis in original). Instead, other offenses and criminalizations are typically used, when intelligence (most often not revealed to court or public) indicates possible terrorism connections. This is different when funding of a particular listed entity is the subject of the prosecution, but many prosecutions of this type are for supporting groups that are not universally regarded as terrorist (Passas, forthcoming). In any event, it is noteworthy that countries with strong counterterrorism policies and practices, such as

11

Terrorist Finance, Informal Markets, Trade and Regulation…

265

Germany and Italy, sanction terrorism finance not independently, but rather simply as one kind of terrorist offenses, such as participation in or support of terrorist groups (Passas, 2008). In short, CFT has been to a large extent characterized by vague, inconsistent, and empirically unfounded rules and assumptions, while in some respects it requires that countries violate their own laws. The fact that CFT can become less counterproductive, much more effective, and better targeted can be seen even more clearly in the regulation of informal value transfer systems (IVTS), such as hawala and tradefacilitated but hidden financial flows.

Informal Value Transfer Systems Extremist and terror groups have used the formal banking system and wire transfer facilities (U.S. Department of State, 2004) as well as IVTS to make payments or transfer funds. IVTS refers to ways in which value can be transferred either without leaving easily identifiable traces or entirely outside the formal financial system (Passas, 1999, 2003a). The term encompasses a range of methods from simple cash couriers to internet-facilitated transfer and sophisticated uses of commerce and correspondent banking (see details in Passas, 2003a). Out of these IVTS, this paper focuses on informal remittances, especially hawala, and the illicit use of trade. There is no precise definition of “informal” remittances or fund transfer systems. However, the term is widely employed and part of an extensive socioeconomic, criminological, and policy literature (Pieke, Hear, & Lindley, 2005). Some of the commonly used parameters include the extent to which these activities are regulated, properly supervised, officially recorded, lawful, accountable, and effectively policed. The recent subprime, derivatives, and financial crisis shows that, if such criteria were applied to Wall Street institutions and their activities that contributed to the crisis, these would be considered IVTS too. One would also need to distinguish between state and self-regulation. With respect to hawala, an age-old informal remittance system that originated in the Indian subcontinent, weak and wanting state regulation does not mean there is no regulation at all: self-regulation is efficient, fair, accepted as legitimate, and effective. In any event, this definitional analysis is beyond our scope here. What matters is that hawala is considered everywhere as an informal remittance system, which serves legitimate economic activity and needy populations, while it also facilitates crime, just as do other financial institutions (El-Qorchi, Maimbo, & Wilson, 2003; FATF, 2003; Lee, 2002; Maimbo, 2003; Passas, 2004b, 2004c). The annual remittance flow in 2005 was estimated at $233 billion, of which, $167 billion was received in developing countries. These amounts are growing (they reached $328 billion in 2008; Ratha, Mohapatra, & Silwal, 2009) and are much higher, perhaps double, when informal and unrecorded remittances are also taken into account (Maimbo, Adams, Aggarwal, & Passas, 2005; World Bank, 2006). In many regions, expatriates’ remittance flows are growing fast and exceed those of foreign direct investment and foreign aid. With the significance of these flows well

266

N. Passas

recognized, the development community and policy makers seek further reductions in the cost of remittances, the use of enhanced technologies to increase speed and convenience, and a widening of options and service providers toward a competitive and efficient market. At the same time, both formal and informal remittance channels interface with informal economies and can be used for the commission of offenses ranging from money laundering and tax evasion to fraud and corruption. In the aftermath of the 9/11 attacks, concerns that hawala networks also assist in the financing of terrorism added to pressures for stronger regulation and control. A major policy challenge has been how to best harmonize these economic objectives with crime control and especially AML/CFT efforts. Hawala is at the same time a hierarchical network and market in which fund transfers as a service to retail clients are rather tangential. People in different main occupations engage in currency trading and speculation as well as in commerce. Retail remittance customers provide the necessary liquidity at a low cost, but this is only the more visible (to authorities and supervisors) part of the economic activities involved. The way it works in simple terms is the following: people wish to send money from point A (e.g., the UK) to point B (e.g., Pakistan). Others want to send money from B to A. Intermediaries collect the money and payment instructions from each end and execute the payments accordingly on a daily basis. Delivery can be made at the beneficiary’s doorstep. The exchange rate is much better than what the banks, Western Union, or money changers would offer. The service is fast, reliable, convenient, cheap and, in some locations, the only option. To the extent the amounts they pool together in each jurisdiction are symmetrical, there is no need for either physical or other funds transfer or currency conversion: pounds of migrants will be used to pay exporters to Pakistan, for example, while the rupees of importers will be used to distribute funds to migrants’ families in Pakistan. However, these pools are never symmetrical because people send money to multiple directions and others wish to receive payments in a third country or currency, sometimes on behalf of another client. As a result, there is a need to balance the accounts between A and B at regular intervals. US dollar accounts are typically used for that purpose in the main financial centers (e.g., New York, London, Dubai, Hong Kong, or Singapore). As the networks grow and hawala operators in locations A and B arbitrage and shop around for the best dollar, pound, rupee, or other currency rates, many more intermediaries get involved in complex networks of hawala operators, agents and sub-agents, clients and clients of clients. These clients may be traders or service providers in all kinds of industries and sectors. The more intermediaries get involved in such networks, the less transparent the set of transactions become. On the other hand, it is essential to note that traceability is nonetheless not lost. On the contrary, because each node of these networks maintains records and understanding of the business of their immediate counterparts, it is feasible and quite possibly easier to follow the money in these networks than in Western financial institutional systems. Hence, there is a need to emphasize the difference between transparency (easy access to comparatively mechanized data) and traceability (the ability to find answers to key questions by contacting the information-rich nodes of these networks). If they wish to collaborate, this is a great opportunity for

11

Terrorist Finance, Informal Markets, Trade and Regulation…

267

investigators and intelligence collectors. If they do not wish to collaborate, they can hide their transactions or information about their clients. This is why it is important to engage in outreach and build communication and collaboration bridges in those networks. Informal remittance services and hawala in particular were singled out early on as a crucial target in the policies against AQ. The FATF issued Special Recommendation VI, which stated: “Each country should take measures to ensure that persons or legal entities, including agents, that provide a service for the transmission of money or value, including transmission through an informal money or value transfer system or network, should be licensed or registered and subject to all the FATF Recommendations that apply to banks and non-bank financial institutions. Each country should ensure that persons or legal entities that carry out this service illegally are subject to administrative, civil or criminal sanctions.” The attempt to apply the same rules and measures in formal and informal economic actors was undertaken against advice and warnings that such measures would not work against networks and mechanisms based on trust and rooted in different socioeconomic, political, and cultural contexts. Even though these transfer systems have always catered for legitimate needs of millions of migrants, they are also uniquely able to resist strict controls throughout the world (see El-Qorchi et al., 2003; Passas, 1999, 2003a, 2003b; Pieke et al., 2005). The US Department of State acknowledged, “In 2006, we have a clearer understanding of our vulnerabilities and recognize that anti-money laundering laws and regulations do not always reach alternative and underground systems for moving dirty money, or transferring value, or financing terrorism. New tools and techniques are needed to detect and expose this activity. This is particularly true in the battle against terrorist finance” (US Department of State, 2006). The way in which countries seek to address informal fund transfers varies (Passas, 2005). Know-your-customer (identity verification and client due diligence), record keeping, and reporting standards are very asymmetric. Some countries criminalize all informal remittance operations, others outlaw them and provide for civil penalties, and many subject them to registration or licensing rules, while other countries do not regulate them at all. Such legal asymmetries jeopardize international cooperation, increase compliance costs and force informal operators to go underground even in countries where they are allowed to operate above board. Even at the domestic level, policies have been uncoordinated and unrealistic. The United States’ arrangements may be held as exemplary by some national or international bodies (they received good marks from the FATF), but they are counterproductive. They illustrate how by applying FATF and domestic regulatory guidelines, they have undermined their own AML/CFT and economic objectives. At the federal level, the rules for money service businesses (MSBs) include registration, knowyour-customer, record-keeping, reporting duties, and an obligation to implement an anti-money laundering regime.2 2

MSBs include money transmitters, check cashers, issuer of traveler’s checks, money orders or stored value, sellers or redeemers of traveler’s checks, money orders or stored value, currency dealing or exchange. Updated information is available at http://www.msb.gov/.

268

N. Passas

At the state level, an area frequently neglected by analysts and policy makers, most jurisdictions require licensing for MSBs. At this level, the regulatory framework is a patchwork of diverse and often nonpragmatic, unenforceable provisions. The inconsistency among states and between state and federal authorities has caused uncertainty, confusion, and lack of awareness and understanding of specific requirements. Many states’ requirements are designed primarily for formal institutions of a certain size, but apply to small, ethnic, and informal remitters as well. Cross-state transactions necessitate compliance with the requirements of all states concerned cumulatively, even if the business volume is minimal. As shown elsewhere, bond, capitalization, and other fees entail unaffordable costs to small enterprises and corner shops serving ethnic communities. In addition, remitters must also cover training, auditing, and record-keeping costs, which can be prohibitive to small businesses by themselves. These all add up to hundreds of thousands or often millions of dollars, which are not realistic for MSBs (Passas, 2006a). On the international front, the case of al Barakaat is an example of the regulator who cried wolf. Somalia’s principal export during the last few years of crises has been human labor and remittance flows have a particularly significant impact on human development options there (UNDP, 2001). Al Barakaat used to be the largest remittance provider before it stood accused by the United States of sponsoring and providing logistical support to terrorism. The 9/11 Commission has made a case study out of Barakaat and details can be found there (National Commission on Terrorist Attacks Upon the United States, 2004; see also Passas, forthcoming). In short, the US Treasury Department claimed that Barakaat was funneling about $25 million a year from customer fees to AQ. In November 2001, police raided Barakaat offices in five US states, seized their records and froze their assets. Similar actions took place around the world, including in the UAE, where al Barakaat was headquartered and top executives were arrested. Many years later, there has been no formal indictment or charge relative to terrorism against the owner or operatives of the most investigated money remitter in the world. The only charge (and convictions) in a few cases brought in the US was related to the transfer of funds without license. As noted in the 9/11 Commission’s staff monograph on terrorist finance: “…notwithstanding the unprecedented cooperation of the UAE, significant FBI interviews of the principal players involved in al-Barakaat (including its founder), and complete and unfettered access to its financial records, the FBI could not substantiate any links between al-Barakaat and terrorism” (National Commission on Terrorist Attacks Upon the United States, 2004, fn 39 at p. 84). Somalia had no banking/financial system other than hawala for years, so the action against Barakaat came under criticism. As the United Nations humanitarian representative in Somalia noted, the global action “is having a very, very serious effect…We are at a point where we have to start anticipating a crisis that could be unique in the modern state system – the collapse of an entire national economy.” The crisis was mitigated by coordinated action of international organizations, but it seems it was produced unnecessarily and unfairly. One has to note the paradox that the 9/11 hijackers’ funds flowed through formal Western institutions. The authorities imposed no penalties or sanctions on them,

11

Terrorist Finance, Informal Markets, Trade and Regulation…

269

because there was no way of identifying the transactions as suspicious. Yet, with no evidence that AQ operated through al Barakaat, the authorities devastated the most successful business of Somalia and ruined livelihoods around the world. Even after all this became widely known, the US held out Barakaat as a successful terrorism finance case in a FATF typologies report published in 2005 (see Passas, 2006a). Such practices do not help in the fight against terrorism and harm international cooperation against all types of serious crime. Intensive international investigations have shown that the 19 hijackers engaged in no hawala transactions. AQ did use these methods during the period it was based in Afghanistan, but this is what virtually everyone did there given the lack of any alternative financial system. Those who facilitated AQ hawala payments at the time have been identified. Over several years after 9/11, authorities in the West detected no other instance of hawala transactions by AQ. However, hawala does get used by groups in S. Asia, Middle East and Africa for their fund transfers quite often. We have also seen a few thousand dollars sent from Scandinavia to Northern Iraq in support of the insurgency. Finally, two recent cases (U.S. v. Amina Farah Ali et al. and U.S. v. Mohammad Younis, both in 2010) highlight the use of hawala networks for payments from and to the US. So, hawala can and does get used by AQ and other militant groups, but it does not appear to be the most frequent modus operandi in Europe and N. America. Couriers appear to be a much more preferred method globally.

Risk Assessment? While a great deal of attention, resources, and regulatory action focused on formal and informal remitters, a systematic assessment of threats and vulnerabilities in different economic sectors has not taken place. Research has shown that one particularly risky area left unattended is trade. How much AQ was involved in commerce-assisted finance is not known. It is true that the CFT enterprise was triggered by 9/11, but financial sanctions and vigilance are meant to cover a wide range of different groups that governments decide to qualify as terrorist. Even though these qualifications command no global consensus, most of them are outlawed or considered terrorist in many parts of the world and are included in designation lists. While one must assume that different groups will raise and transfer funds in very different ways and through different channels, providing a typology of terrorist finance is beyond the scope of this paper. It is certain, however, that commodities – and trade in general – have always been part of the political economy of conflict (Ballentine & Nitzschke, 2005). Armed militants typically resort to whatever resources are accessible to them, unless they clash with honestly held religious or ideological positions. Reports about the role of commodities in the financing of terrorist groups cover many types of both legal and illicit trade. Smuggling operations involving commodities such as tobacco, arms, oil, precious stones, gold, silver, farm

270

N. Passas

products, and even humans have been linked to militant groups including the Kosovo Liberation Army (KLA), al Jemaah al Islamiya, the Tamil Tigers (Liberation Tigers of Tamil Eelam or LTTE), Hamas, Hizballah, the Partiya Karkerên Kurdistan (PKK), the Northern Alliance, AQ, Groupe Islamique Armé (GIA), the Irish Republic Army, as well as Armenian, Chechen, and Georgian paramilitary groups. Even though the evidence is sometimes weak with respect to some groups or goods, trade plays actually and potentially an important part in the financing of conflict and terrorism. The significance increases once the role of sympathizers is considered: apart from voluntary contributions made to a cause, businesspeople might assist by letting their companies be used by militants for hidden value transfers or by sharing the know-how with them. Voluntary and forced taxation of businesses has been a part of fund-raising for many groups, such as PLO, FARC and AUC, M19, Christian Phalange in Lebanon, Hizbollah, and LTTE. The relevance of trade also includes the potential logistical or other support for attacks with chemicals or weapons of mass destruction (Passas, 2007; Passas & Jones, 2006). Trafficking and smuggling of goods, such as cigarettes, have also been reported as part of state actor efforts to proliferate in weapons of mass destruction and other illicit cross-border activities (especially with respect to North Korea). Clearly there is cause for concern that the trade in commodities can be used to support terrorist activity. Nevertheless, many public writings and reports lack context as well as a good understanding of the industries on which they focus. Research into the commerce of diamonds, gold, and tobacco revealed misplaced concerns about certain parts or geographic areas of the trade in these commodities, but also identified serious vulnerabilities and possible linkages with militant groups elsewhere (Passas, 2004a). We can take a brief look at findings in the diamonds and tobacco sectors.

Diamonds In recent years, the media, NGOs, the United Nations, the 9/11 Commission, FinCEN and various scholars investigated links between AQ and Hezbollah with (conflict and illicit) diamonds (Campbell, 2002; Gberie, 2001; Global Witness, 2003; UN Monitoring Group, 2002). US law enforcement agencies did not appear to examine systematically or in a coordinated way the allegations involving the commodities trade, including diamonds (GAO, 2003). In the 2002–2006 period, the FBI monitored the situation, followed leads and evidence, issued two (nonpublic) reports, and concluded that there was no strong evidence that AQ actually used the diamonds trade for fund-raising or storage. My own research confirmed this conclusion with respect to conflict diamonds and AQ (Passas & Jones, 2006), but also found several reasons to be concerned as diamonds can be an attractive vehicle for illicit activities such as money laundering and terrorist financing or for simply storing and transferring value: • Diamonds are easy to mine and transport (Dietrich, 2002); this facilitates their secret collection and transfer without being noticed.

11

Terrorist Finance, Informal Markets, Trade and Regulation…

271

• They are small and durable, can be easily hidden, and avoid detection by police dogs. • They can maintain a high value, often easily convertible to cash, which makes them an efficient method to store value.3 • Diamonds are an internationally recognized commodity and can be bartered for arms, drugs, fuel, or other commodities; this makes them a close equivalent to currency for licit and illicit payments (Anderson, 2001). There is evidence that several criminal deals have been settled with diamonds. • Overtime, the diamond industry has been insular, very protective of its privacy, and self-regulating or completely unregulated (Cook, 2003). • Diamonds are almost untraceable, which makes finding their origin virtually impossible (GAO, 2002). It is important to note that the origin of some rough stones may be identified by experts. The same applies to kimberlite diamonds. Rough diamonds (e.g., from Angola, Liberia, etc.) can also be sourced in general (personal interviews). • Their exact value is not always certain or clear to nonexperts; therefore, shipments can be mis-invoiced, thus hiding value transfers behind otherwise ordinary or legal transactions (personal interviews). • In many countries, diamond-related transactions are unrecorded and in cash (some interviewees suggested that large cash transactions for both diamonds and other commodities are occurring in the US, as well). • Informality and trust characterize this trade in many parts of the world. It may be impossible to obtain the cooperation of participants in establishing information about the real price, origins, final destination, names of customers, profit, etc. • Certain types of diamonds may be exempted from duties or taxes in certain countries, rendering them less likely to be inspected or audited by the authorities. • The large volume of trade can be easily used to provide cover for illicit transactions. Billions of dollars worth of diamonds have been imported into the United States. • Kimberley Process (KP) certificates, designed to ensure that conflict diamonds are not used to fuel armed violence in Africa, may be bought in the black market (Zarate, 2005). Thus, in the context of conflict, corruption, transnational crime, and the illegal exploitation of natural resources in parts of Africa, it is reasonable to think militant groups may have been involved in the diamond business. The KP does not explicitly prohibit selling diamonds to militants, so long as they have proper paperwork and are tangibly connected to a member of the KP. If such militant organizations and their associates become too entangled in regulation, they can simply remove themselves from the KP and carry on in the shadows.

3

When precious stones are transferred or transacted in criminal markets, they may not maintain their value. Going through one or two middlemen reduces their value by between 30 and 45%. Gold by contrast does maintain its value, partly because it is flexible and malleable. So, as diamonds become a medium of underground exchanges, they may maintain a high value, but not always the same value.

272

N. Passas

Even if the AQ-diamonds nexus reports were accurate, only a fraction of presumed funds would have been converted and transferred in this way. Other commodities could have been, and may still be, used to do the same. The point is that focusing solely on one type of commodity or one part of the world can lead to inadvertent neglect of other areas with similar or more serious vulnerabilities. The conclusions of our research and analysis are: • Some terrorist groups or persons associated with them appear to have engaged in diamonds transactions (although most likely not AQ). • The amounts involved do not appear to be substantial. • There is certainly vulnerability in the sector for future use by militants and other offenders. • The vulnerability seems to be particularly acute with polished stones (as opposed to the rough diamonds on which most reports focused in the past). • Most identified vulnerabilities also apply to other commodities4 and to trade in general. • Currently, trade is nontransparent and represents a significant threat to all efforts countering terrorist finance, money laundering, or other financial crimes.

Tobacco There are four major forms of illicit cigarette trading: trade diversion, invoice manipulation, smuggling, and counterfeiting of tobacco products. Although all four can occur on their own, they often coincide with each other as well as other illegal activities. The sources of the illicit cigarettes come from several areas: low-tax states, foreign free trade zones, customs bonded warehouses, certain Native American reservations, stolen or hijacked shipments, and manufacturers of counterfeit cigarettes overseas (Snell, 2004). All of these sources are vulnerable to both organized crime and terrorist groups. There is a dearth of literature on trade diversion generally, and an even greater lack of analysis relating to tobacco specifically. Trade diversion is a sophisticated form of fraud and money laundering in which offenders with a distribution network and good understanding of international trade, channel products away from the declared destinations, in order to profit from the price variation in different countries (in different markets or under special agreements). Along the way, they can also launder “dirty” money with a single transaction (Passas, 2003a). It is a method open to both criminal and terrorist use as a way to secretly move and

4

When it comes to gold, for example, there are reported links with FARC and AUC in Colombia, al Qaeda, Jemaah Islamiah and Chechen separatists, while Echo Bay, a gold company made of security payments or “resource taxation arrangements” with Filipino militant groups including Abu Sayyaf (see Snell, 2004).

11

Terrorist Finance, Informal Markets, Trade and Regulation…

273

raise money. Trade diversion is hard to detect, and it is also versatile in that funds can remain in numerous countries without serious inquiry by the authorities (DeKieffer, 2005). A source of vulnerability to terrorist financing is that law enforcement officials have been reluctant to investigate such cases because the victims are often multinational corporations that are partially at fault for failing to conduct proper client and employee due diligence to protect themselves from such activity. Another reason for the small number of criminal proceedings against diversion is the difficulty in obtaining evidence in relatively complex cases due to their occurrence overseas and clever camouflage as legitimate, thus consuming substantial investigative resources (DeKieffer, 2005). Therefore, although trade diversion is a method easily exploitable by terrorist groups, cases are difficult to detect and prove. Without further improvement of trade transparency and law enforcement methods in this area, we cannot know the extent to which such schemes are used to capitalize on the tobacco industry as a method of fund-raising. Invoice manipulation, or mis-invoicing, involves the falsification of trade documents, which enables customs fraud, tax/duty evasion, sanctions violations, money laundering, and illicit fund transfers. Mis-invoicing occurs, for example, when exporters understate the value of their exports in order to deceive authorities and build up a foreign exchange holding. Mis-invoicing also occurs on the import side. When importers “over-invoice” the value of their transactions, they can obtain extra foreign exchange because they receive more foreign exchange than they need to pay for their import bills. This “excess” foreign exchange can be transferred easily to other countries, from a central bank at favorable terms. Under-invoicing and smuggling allow importers to evade customs duties or restrictions and engage in capital flight (Boyce & Ndikumama, 2001; Onwioduokit, 2001). Cross-border trade has been growing and relentlessly promoted by governments, regional, and international organizations (Passas, 2003a). With higher volumes and values, the legitimate trade becomes increasingly vulnerable to abuse by militants and other offenders. As more items are moved from country to country for transit or final use, the ease with which shipments are mis-invoiced in order to hide underlying transactions and value transfers increases. Yet, policy makers have neglected trade-based fund-raising, so we do not know how often mis-invoicing of tobacco shipments is actually used by which terrorist organizations and where. Smuggling is done to avoid duties, excise taxes or to evade rules regarding quotas, sanctions, prohibitions, etc. For example, even though the sale of most foreign brands of cigarettes is forbidden in China, these brands are easily found and advertised in China. Smuggling occurs when consumers purchase tobacco in low-tax jurisdictions in amounts that exceed the limits set by customs regulations for resale in high-tax jurisdictions (Joossens, Chaloupka, Merriman, & Yurekli, 2000) as well as when Native American reservations and military installations sell tax-free cigarettes to non-Native Americans and nonmilitary personnel (Thursby & Thursby, 2000). Cigarette smuggling occurs to different degrees, such as by means of largescale transport, with containers full of cigarettes in ships or trucks, or by means of

274

N. Passas

direct mail, in which case small shipments are made separately but can add up to a significant amount. This is the case with Internet sales.5 Smuggled cigarettes account for 6–8.5% of global consumption. Nearly a fifth of all cigarette production is exported, of which almost one-third – approximately 355 billion cigarettes a year – finds its way into the contraband market (World Health Organization, 2003). According to an ATF report, some cigarette smugglers have ties with terrorist groups, and there are indications that terrorist group involvement in illicit cigarette trafficking may grow in the future because of the large profits that can be made (cited in GAO, 2004). ATF and ICE officials indicate that cigarettes are smuggled into the United States from many countries, including China, Malaysia, Korea, Russia, Latvia, Mexico, Brazil, Paraguay, Uruguay, and the Philippines (GAO). Reports indicate that Russian, Armenian, Ukrainian, Chinese, Taiwanese, Pakistani, Lebanese, Syrian, and other criminal groups are deeply involved in the trafficking of contraband and counterfeit cigarettes and counterfeit tax stamps for profit (GAO, 2004). Other reports suggest that some terrorist groups collaborate with criminal enterprises as well as with drug trafficking groups who have established routes and business contacts. Some criminal cases, interviewees, and secondary sources reveal that tobacco smugglers in the United States and the United Kingdom are providing material support to the Hezbollah and the Real IRA, among other militant groups. Law enforcement reports indicate that groups they associate with AQ, Hamas, PKK (the Kurdish Workers Party), and Islamic Jihad (both Egyptian and Palestinian) are involved in the illicit trafficking of cigarettes (Billingslea, 2004).

Conclusion Financial controls are an essential component of counterterrorism, but these must be based on empirical evidence, risk assessments and analysis of available resources and options. We have seen how an emphasis on financial institutions has led to a neglect of commercial activities where serious vulnerabilities need to be addressed. We have seen that analyses of secondary sources are often misled and thus misleading. Consequently, our findings may serve as a warning that the terrorist finance threat in areas not yet thoroughly researched, such as nonprofit and charitable organizations or other economic spheres may also be not as well understood as generally assumed – a point supported by the 9/11 Commission which raised many questions on the appropriateness and effectiveness of actions taken in this regard. Similar questions for further research and review arise with respect to narcotics, firearms, counterfeit pharmaceuticals, bulk cash smuggling, people trafficking, and organized crime more generally as well. 5

Although smuggling can occur in conjunction with diversion, the latter can also be done without smuggling at all. The most sophisticated form involves proper documentation and paperwork with goods imported and declared at Customs (e.g., as “returned US goods”).

11

Terrorist Finance, Informal Markets, Trade and Regulation…

275

There is no systematic and valid knowledge on precisely how different countries go about controlling the financing of terrorism and how effective their controls are. The international community still does not have a good overall picture of the methods terrorist groups use, their needs, how they shift their modus operandi in response to counterterrorism, and how different funding mechanisms are (or are not) accessible to specific groups or ideological/religious orientations. In other words, we do not know exactly what the nature of terrorism finance is, how countries and regions target it, how we can offer better guidance to the private sector, whose cooperation is vital, and how to improve the effectiveness of our social and legal controls (Passas, 2008). The same applies to the issue of WMD proliferation and financial vigilance against it (Passas et al., 2010). The brief review of diamonds and tobacco illustrates the need for trade transparency with respect to all goods and services. Moreover, because financial and trade flows are not matched and contrasted, we have good reasons to believe that massive irregularities, suspicious transactions, and blatant abuses remain undetected. In addition, current law enforcement practices leave much to be desired. Interviews with Customs officials and analysis of import data pointed to serious gaps in the way the US government deals with trade transactions. Incomplete, erroneous, or illegal documentation was found through routine review of forms filed with Customs. The possibility that substantial value or volumes are traded under regulatory radar screens or without the ability to identify the true contracting parties and beneficiaries is a cause for serious concern. Hence, the most important policy implication is that we must devise and implement measures rendering trade flows more transparent and traceable, in order to prevent and detect financial misconduct underlying international transactions. This goal can be attained through the coordinated efforts of various agencies, the joint review of available data, and the widest possible use of existing technology that facilitates domestic as well as international cooperation. Ultimately, the only way such transparency and traceability can be achieved is by ensuring both financial and commercial flows are identifiable and also match information flows about payments and fund transfers. Regulations with respect to formal and informal financial institutions also need to be reviewed in the light of collected and analyzed evidence on how terrorist groups use these systems and how they change their methods in response to control efforts. Scores of terrorism finance cases have been made in the last 8 years, which should provide plenty of empirical material on which to base updated threat assessments and risk analyses in support of fine-tuned rules and requirements. Once the evidence is available and reviewed, the road is paved for consensus knowledge and consensus formation between regulators and regulated. Some of the guiding principles that should be adopted by the international community are the following: • Regulation ought to reflect the views of all stakeholders, so outreach and consensus building is essential. • Rules must be necessary, practical, and affordable. • Regulation must be enforceable.

276

N. Passas

• Regulation must correspond to available resources and capacity. • Regulation should foster a level-playing field within the industry and across industries (both domestically and internationally). • Efforts at regulation must be consistent across borders. • The effects and effectiveness of regulation must be constantly assessed, measured, and monitored. In conclusion, it should be noted that there is reason to hope that the international community is moving along these suggestions. The FATF has been debating the problems of trade-based money laundering and terrorist finance and some good practices are beginning to emerge. Efforts in the Caribbean and elsewhere are made to raise the bar and self-regulate free zones (e.g., in Curacao) with better due diligence, know-your-customer, good record keeping, and improved cooperation with the authorities. A study reviewing terrorism finance cases is underway to determine the extent to which indicators of suspicious transactions specific to terrorism can be developed for both regulators and the private sector. Finally, arguments in favor evidence-based policy making are gaining momentum. We have seen that there is room for improving the financial controls against terrorism. We now see strong signs that some of the necessary preparatory work and activities are beginning. It is imperative that we support and contribute to these efforts in order to achieve a more secure, peaceful, and prosperous global community.

References Anderson, R., testimony. (2001). Conflict Diamonds: Funding conflict, fueling change. Hearings before the Subcommittee on trade, of the House Committee on Ways and Means, 107th Cong., 46 (2001). Arar Report (Commission of Inquiry into the Actions of Canadian Officials in Relation to Maher Arar). (2006). Report of the events relating to Maher Arar: Analysis and recommendations. Ottawa, Ontario: Public Works and Government Services Canada. Ballentine, K., & Nitzschke, H. (Eds.). (2005). Profiting from peace: Managing the resource dimensions of armed conflict. Boulder: Lynne Rienner. Billingslea, W. (2004). Illicit cigarette trafficking and the funding of terrorism. The Police Chief, 71(2). Retrieved from http://policechiefmagazine.org/magazine/index.cfm?fuseaction=display_arch& article_id=226&issue_id=22004. Boyce, J. K., & Ndikumama, L. (2001). Is Africa a net creditor? New estimates of capital flight from severely indebted Sub-Saharan African countries, 1970–1996. Journal of Development Studies, 38(2), 27–56. Brown, G. (2006). The perils of terrorism: Chinese whispers, Kevin Bacon and al Qaeda in southeast Asia – A review essay. Intelligence and National Security, 21(1), 150–162. Burke, J. (2003). Al-Qaeda: Casting a shadow of terror. London: I.B. Tauris. Campbell, G. (2002). Blood diamonds: Tracing the deadly path of the world’s most precious stones. Boulder: Westview Press. Cook, N. (2003). Diamonds and conflict: Background, policy and legislation (No. RL30751). Washington: Congressional Research Service. Dandurand, Y., & Chin, V. (2004). Links between terrorism and other forms of crime. Vancouver: Report to Foreign Affairs Canada and The United Nations Office on Drugs and Crime.

11

Terrorist Finance, Informal Markets, Trade and Regulation…

277

DeKieffer, D. (2005, November 6). Trade diversion as a fund raising and money laundering technique of terrorist organizations. Unpublished manuscript. Dietrich, C. (2002). Hard currency: The criminalized diamond economy of the Democratic Republic of the Congo and its neighbours (Occasional Working Paper No. 1-894888-044-9). The Diamonds and Human Security Project, Partnership Africa Canada. El-Qorchi, M., Maimbo, S. M., & Wilson, J. F. (2003). Informal funds transfer systems: An analysis of the informal Hawala system (No. IMF Occasional Paper No. 222). Washington: IMF. Farah, D. (2001, November 2). Al Qaeda cash ties to diamond trade. Washington Post. FATF. (2003). Report on money laundering typologies. Financial Action Taskforce on Money Laundering. Gale, I. (2006, July 5). Money transfers are scrutinised to prevent terrorist funding. Gulf News. Retrieved from http://archive.gulfnews.com/business/Banking_and_Finance/10051708.html. (GAO) General Accounting Office. (2002). International trade: Critical issues remain in deterring conflict diamond trade (report to congressional requesters). Washington: United States General Accounting Office. (GAO) General Accounting Office. (2003). Terrorist financing: U.S. agencies should systematically assess terrorists’ use of alternative financing mechanisms. Washington: United States General Accounting Office. (GAO) General Accounting Office. (2004). Cigarette smuggling: Federal law enforcement efforts and seizures increasing: Report to the chairman and ranking minority member, committee on government reform, House of Representatives. Washington: GAO. Gberie, L. (2001). Destabilizing Guinea: Diamonds, Charles Taylor and the potential for wider humanitarian catastrophe. Ottawa: The Diamonds and Human Security Project, Partnership Africa Canada. Gilmore, W. C. (2004). Dirty money: The evolution of money laundering counter-measures (3rd ed.). Strasbourg: Council of Europe Press. Giraldo, J. K., & Trinkunas, H. A. (2007). Terrorism financing and state responses: A comparative perspective. Stanford: Stanford University Press. Gold, M., & Levi, M. (1994). Money laundering in the UK: The impact of suspicion-based reporting. Cardiff: The Police Foundation/University of Wales College of Cardiff. House of Commons. (2006). Report of the official account on the bombings in London on 7th July 2005. London: House of Commons. Joossens, L., Chaloupka, F. J., Merriman, D., & Yurekli, A. (2000). Issues in the smuggling of tobacco products. In P. Jha & F. J. Chaloupka (Eds.), Tobacco control policies in developing countries (pp. 393–406). Oxford: Oxford University Press. Koh, J.-M. (2006). Suppressing terrorist financing and money laundering. Berlin: Springer. Labayle, H., & Long, N. (2009). Overview of European and international legislation on terrorist financing study. Strasbourg: European Parliament. Lee, R. (2002). Terrorist financing: The U.S. and international response. Washington: Congressional Research Service. Levi, M. (1991). Pecunia non Olet: Cleansing the money-launderers from the temple. Crime, Law and Social Change, 16, 217–302. Levi, M., & Reuter, P. (2006). Money laundering. Crime and Justice, 34, 289–375. Maimbo, S., Adams, R., Aggarwal, R., & Passas, N. (2005). Migrant labor remittances in South Asia. Washington: The World Bank. Maimbo, S. M. (2003). The money exchange dealers of Kabul: A study of the Hawala system in Afghanistan. Washington: World Bank. Monitoring Team, U. N. (2009). Tenth report of the Analytical Support and Sanctions Monitoring Team appointed pursuant to resolution 1526 (2004) and extended by Security Council resolution 1822 (2008) concerning Al-Qaida and the Taliban and associated individuals and entities. New York: U.N. Security Council. National Commission on Terrorist Attacks Upon the United States. (2004). Monograph on terrorist financing (staff report to the commission). Washington: National Commission on Terrorist Attacks upon the United States.

278

N. Passas

National Research Council. (2002). Terrorism: Perspectives from the behavioral and social sciences. Washington: National Academies Press. Naylor, R. T. (1999). Wash-out: A critique of follow-the-money methods in crime control policy. Crime, Law and Social Change, 32(1), 1–57. Naylor, R. T. (2006). Satanic purses: Money, myth, and misinformation in the war on terror. Montreal and Kingston: McGill-Queen University Press. Onwioduokit, E. A. (2001). Capital flight from Nigeria: An empirical re-examination. Paper presented at the WIDER development conference on debt relief. http://www.wider.unu.edu/conference/conference-2001-2/poster%20papers/Emmanuel.pdf. Passas, N. (1986). Political crime and political offender: Theory and practice. Liverpool Law Review, 8(1), 23–36. Passas, N. (1999). Informal value transfer systems and criminal organizations: A study into socalled underground banking networks. The Hague: Ministry of Justice. Passas, N. (2003a). Informal value transfer systems, money laundering and terrorism. Washington: Report to the National Institute of Justice (NIJ) and Financial Crimes Enforcement Network (FINCEN). Passas, N. (2003b). Hawala and other informal value transfer systems: How to regulate them? Journal of Risk Management, 5(2), 39–49. Passas, N. (2004a). The trade in diamonds: Vulnerabilities for financial crime and terrorist finance. Vienna: FinCEN, US Treasury Department. Passas, N. (2004b). Informal value transfer systems and criminal activities. The Hague: WODC, Ministry of Justice. Passas, N. (2004c). Indicators of Hawala operations and criminal abuse. Journal of Money Laundering Control, 8(2), 168–172. Passas, N. (2005). Formalizing the informal? Problems in the national and international regulation of Hawala. Regulatory frameworks for Hawala and other remittance systems (pp. 7–16). Washington DC: International Monetary Fund. DOI: 10.5089/9781589064232.072. Passas, N. (2006a). Fighting terror with error: The counter-productive regulation of informal value transfers. Crime, Law and Social Change, 45(4–5), 315–336. Passas, N. (2006b). Setting global CFT standards: A critique and suggestions. Journal of Money Laundering Control, 9(3), 281–292. Passas, N. (2007). Terrorism Financing Mechanisms and Policy Dilemmas. In J. K. Giraldo & H. A. Trinkunas (Eds.), Terrorism Financing and State Responses: A Comparative Perspective (pp. 21–38). Stanford, CA: Stanford University Press. Passas, N. (2008). Terrorism financing general report [for the Intl Association of Penal Law]. Revue Internationale de Droit Penal, 79, 326–343. Passas, N. (forthcoming). Controlling terrorist financing. New York: Springer. Passas, N., Anthony, I., Deanaz, G., & Walker, J. (2010). Prevention of CBRN illicit trafficking and deceptive financial practices. Brussels: Report to the Commission of the European Union. Passas, N., & Jones, K. (2006). The trade in commodities and terrorist financing: Focus on diamonds. European Journal of Criminal Policy and Research, 12, 1–33. Physicians for Human Rights. (2008). Broken laws, broken lives: Medical evidence of the use of torture by US forces. Cambridge: Physicians for Human Rights. Pieke, F., Hear, N. V., & Lindley, A. (2005). Informal remittance systems in Africa, Caribbean and Pacific (ACP) countries. Oxford: ESRC Centre on Migration, Policy and Society. Pieth, M., Thelesklaf, D., & Ivory, R. (2009). Countering terrorist financing: The practitioner’s point of view. Bern, New York: Peter Lang. Ratha, D., Mohapatra, S., & Silwal, A. (2009). Outlook for remittance flows 2009–2011: Remittances expected to fall by 7–10 percent in 2009. Washington: World Bank. Snell, M. B. (2004). The cost of doing business. Sierra Magazine (May/June). Retrieved from http://www.sierraclub.org/sierra/200405/terrorism. Thursby, J. G., & Thursby, M. C. (2000). Interstate cigarette bootlegging: Extent, revenue losses, and effects of federal intervention. National Tax Journal, March 2000. Available at SSRN: http://ssrn.com/abstract=222748.

11

Terrorist Finance, Informal Markets, Trade and Regulation…

279

U.N. Monitoring Group. (2002). Second report of the monitoring group established pursuant to security council resolution 1363 (2001) and extended by resolution 1390. New York: U.N. Security Council. UNDP (United Nations Development Programme). (2001). Human development report. New York: United Nations. United Nations. (2004). High-level panel on threats challenges and change. A more secure world: Our shared responsibility. New York: United Nations [Dept. of Public Information]. US Department of State. (2004). International narcotics control strategy report: Part II money laundering and financial crimes. Washington: Bureau for International Enforcement Affairs. US Government Accountability Office. (2005). Terrorist financing: Better strategic planning needed to coordinate U.S. efforts to deliver counter-terrorism financing training and technical assistance abroad. Washington: GAO. van Duyne, P., & de Miranda, H. (1999). The emperor’s clothes of disclosure. Hot money and suspect disclosures. Crime, Law and Social Change, 27(3), 245–271. van Duyne, P. C., & Levi, M. (2005). Drugs and money: Managing the drug trade and crimemoney in Europe. London: Routledge. Warde, I. (2007). The price of fear: The truth behind the financial war on terror. Berkeley: University of California Press. Witness, G. (2003). For a few dollars more: How al Qaeda moved into the diamond trade. London: Global Witness. World Bank. (2006). Global development finance report. Washington: World Bank. World Bank & International Monetary Fund. (2006). Reference guide to anti-money laundering and combating the financing of terrorism (2nd ed.). Washington: The World Bank and International Monetary Fund. World Health Organization. (2003). Killing for profit: Tobacco industry monitoring report. October 2002 to December 2002. Retrieved from http://www.who.int/tobacco/media/en/tob-ind-monitoring02.pdf. Zarate, J. C. (2005, February 16). Testimony by the assistant secretary terrorist financing and financial crimes. U.S. Department of the Treasury before the United States House Financial Services Committee Subcommittee on Oversight and Investigations. Retrieved from http:// www.ustreas.gov/press/releases/js2256.htm.

Chapter 12

Evaluating the Legal Challenges and Effects of Counterterrorism Policy Linda M. Merola

Since September 11, 2001, a significant portion of US public policy has focused upon the threat of terrorism. Following the attacks, the need for increased attention to counterterrorism resulted in substantial allocations of public funding toward these initiatives. Yet, researchers conducting systematic reviews of the literature surrounding counterterrorism strategies have found that very few rigorous evaluations of such strategies actually exist, notwithstanding these substantial expenditures (Lum, Kennedy, & Sherley, 2006). Even more surprisingly, this shortage of research persists despite the considerable amount of scholarly attention focused on producing other types of publications related to terrorism. An example of this can be seen with respect to the literature surrounding terror-related legal issues. Post-9/11 legal questions have often been central to policy debates regarding terrorism; some examples of key issues include those surrounding harsh interrogation methods, indefinite detentions, warrantless wiretapping, and extraordinary renditions, to name only a few. However, while legal scholars and commentators have energetically published more traditional doctrinal analyses of law, few empirical studies of post-9/11 legal issues have been conducted and even fewer of these have attempted to evaluate specific counterterrorism strategies. One explanation for this may stem from the fact that mainstream legal scholarship has not traditionally employed the methodologies of social science, such as experiments, surveys, or the statistical analysis of data. Additionally, until relatively recently, few scholars have been cross-trained to possess both legal and social scientific expertise. Over time, this is changing, as greater numbers of scholars pursue joint degree programs or opt to embrace the perspectives of fields like law and society and empirical legal studies. In many ways, then, the existence of little empirical research and even fewer evaluation projects related to law and terrorism is perhaps L.M. Merola (*) Department of Criminology, Law and Society, George Mason University, 4400 University Drive, MS 4F4, Fairfax, VA 20110, USA e-mail: [email protected]

C. Lum and L.W. Kennedy (eds.), Evidence-Based Counterterrorism Policy, Springer Series on Evidence-Based Crime Policy 3, DOI 10.1007/978-1-4614-0953-3_12, © Springer Science+Business Media, LLC 2012

281

282

L.M. Merola

merely reflective of the general orientation of legal scholarship. However, given the importance of the topic of counterterrorism and the central role that legal issues have played in the post-9/11 era, the need for research which integrates legal scholarship to a greater degree with the more policy-oriented paradigm of evidencebased counterterrorism is quite pressing. Moreover, though scholars in fields like law and society are producing a great deal of impressive scholarship related to terrorism, few law and society scholars have attempted outcome evaluations of specific terror-related policies. This chapter will explore further some of the post-9/11 empirical research related to law and legal institutions. Though very little of the existing empirical research in this area purports to evaluate specific counterterrorism strategies, this literature will be examined from the viewpoint of building an evidence base that can be utilized by both policy makers and also by scholars interested in future research. Building from the foundation of current empirical work, each section will offer some suggestions for future directions in scholarship. One important note is that this chapter is not meant to represent an exhaustive review of the scholarship related to law and terrorism. Rather, this discussion is meant to provide some examples of areas where the lens of evidence-based counterterrorism might aid interested scholars in thinking about future work.

Building an Evidence Base Through Social Science In highlighting the lack of empirical work in this area, this discussion is not meant to diminish the value of nonempirical legal scholarship. Rather, doctrinal work is important both for its independent contributions and also for the ways that it can inform social science about the law. In the social sciences, much of the empirical scholarship related to legal issues is routinely conducted by nonlawyers. As a result, doctrinal analyses of law are a key foundation for almost any social scientific work. Even where researchers conducting social scientific studies of law possess a high level of legal expertise, doctrinal analyses of legal issues help social scientists to understand the views of experts actually immersed in these issues. For the most positive outcomes from research, these communities must work synergistically, each relying upon the other to broaden its understanding. Yet, there is also a role for social science to play in understanding the operation of law, particularly with respect to a problem like that created by terrorism. When confronting the threat of terrorism, our hope is that law and legal institutions will prove effective tools to aid in a variety of counterterrorism efforts. Additionally, our society seeks to employ counterterrorism strategies without violating our laws or undermining longstanding institutions or legal norms. However, to accomplish any of these goals, law-related terrorism scholarship must be focused not only on issues of “black letter” law, but also on evaluations of law’s operations, effectiveness, and impacts. Without empirical research, we cannot systematically assess how effective our policy choices have been. In making these assessments, there are a number of different types of studies that are needed.

12

Evaluating the Legal Challenges and Effects of Counterterrorism Policy

283

First, as is clear from the discussion so far, additional empirical legal research which systematically evaluates post-9/11 policy choices is necessary. Since 9/11, our leaders have been consistently challenged by how to adapt legal processes and institutions to conform to the demands of terrorism, yet existing social science research tells us relatively little about the success of these efforts. Some examples of this type of research are discussed below, but, to date, social scientific evaluations of legal policy are exceedingly sparse. There are a few existing examples of this type of research in related fields, such as policing (Huq & Muller, 2008; Kalyvas, 2006; Weisburd, Feucht, Hakimi, Perry, & Mock, 2009), but even accounting for related studies, research of this type is in relatively short supply. In addition to evaluations of legal policy, empirical research is also needed to evaluate other types of policies that have legal implications. For example, an evaluation of new technologies in use by law enforcement for counterterrorism purposes might not seem at first to be an issue for empirical legal researchers. However, the use of such technologies may have implications for privacy rights and also may influence ongoing citizen expectations of privacy. One example of this is the use of license plate recognition technology by law enforcement (Lum, Merola, Willis-Hibdon, & Cave, 2010). Empirical legal scholars have a role to play in comprehensive evaluations of such technologies. To date, few of these types of technological evaluations (integrating empirical legal evaluations) have been conducted. Empirical evaluations of policy can be proactive (for example, completed in an experimental or lab setting) or these studies can involve the retroactive evaluation of policies. However, a greater effort to integrate legal scholarship with other types of evaluative literature would aid counterterrorism efforts and may also make it more likely that legal concerns are considered when policy is developed. In addition to evaluating specific counterterrorism measures, scholars in the field of law and terrorism can also add to the evidence base by focusing on systematically assessing expert opinion about these policies. For example, a tremendous amount of effort has been expended debating the relative merits of policy decisions like the use of Guantanamo Bay as a detention center or the propriety of civilian trials as compared with military tribunals. However, few attempts have been made to systematically assess the views of legal experts (or other types of experts, for that matter) regarding particular policies, nor how these experts would evaluate critical issues. Perhaps paradoxically, the large majority of opinion research about terrorism (even with respect to legal issues) focuses on the opinions of average Americans. Of course, in a democratic system, the study of public opinion is very important, but public opinion studies often tell us little about the relative merits of policy choices. As an example, since 9/11, our society has frequently debated the proper venue for trying terror suspects and the question of whether or not it would be wise to detain these individuals inside the United States to await trial. Yet, to the author’s knowledge, there is only one study (currently in progress) that systematically assesses opinions and gathers policy insights on this topic from those with expertise in the field of corrections (Merola et al., 2011). The vantage point of evidence-based counterterrorism is particularly helpful in focusing scholarship on the ways in which scholars can support policy more effectively. In this way, evaluation research helps to provide a bridge between legal scholarship and policy makers.

284

L.M. Merola

As longer-term questions, researchers might also wonder about the impact of a variety of terror-related legal policies on the continuing legitimacy of legal institutions in the United States or on the maintenance of legal norms. At present, too little empirical research assesses these questions. However, it is important to note here that one study does associate beliefs about policing strategies with the willingness to cooperate with police in their conduct of counterterrorism operations (Tyler, Schulhofer, & Huq, 2010). Also, there is some research (detailed below) that examines developments in public support for the norm of expansive civil liberties in the post-9/11 United States. Of course, Americans hope not to violate the traditional norms of the United States nor to erode the legitimacy of US institutions with counterterrorism operations. However, without empirical evaluations of particular policies or interventions, researchers neglect to make use of one of the most effective tools of systematic assessment that we possess. Greater attention to creating an empirical evidence base upon which policy makers can draw in times of crisis can only increase the real world impact of our scholarship. A fundamental tenet of the evidence-based paradigm is to forge a closer link between researchers and practitioners, with the goal of making policy more effective (Pfeffer & Sutton, 2006). This paradigm demands that practitioners account for the findings of scholarly research while developing policy, but, in this way, the paradigm also provides a bridge for researchers to speak to practitioners and to become a greater force in policy development (Pfeffer & Sutton, 2006; Rousseau, 2006). The next few sections of this chapter will extend this theme further through a discussion of some of the existing empirical research. Though comparatively little empirical research exists on these topics, a number of studies have presented findings that can be useful in framing future evidence-based counterterrorism policy and research.

Empirical Evaluations of Legal Policy in Response to Terrorism Since 9/11, government actors have been faced with a stunning array of policy decisions related to counterterrorism, many of which directly involve law, legal norms, or legal institutions. In fact, many of our most basic rights have been challenged by the events of the post-9/11 era. Additionally, in response to the threat, our nation has put into place a host of new legal provisions, including those contained in the muchdebated USA-PATRIOT Act (2001). It is logical that social science research would be an important tool for evaluating the effectiveness and impacts of these post-9/11 legal policies, particularly when the stakes of counterterrorism are so high. Yet, to date, little empirical research has focused on assessing the legal aspects of policies enacted in response to terrorism. In fact, in a search, the author uncovered only a small number of empirical articles that have assessed the impact or functioning of any law-related policy connected with counterterrorism. As an example, one of these articles examines the passage of state laws in the late 1990s which prohibited the filing of frivolous liens (Chamberlain & Haider-Markel, 2005).

12

Evaluating the Legal Challenges and Effects of Counterterrorism Policy

285

According to the authors, in many cases, such laws were passed in order to deter Common Law groups from engaging in domestic “paper terrorism” by filing frivolous liens against elected officials. These officials usually had been “convicted” of crimes and “sentenced” to pay fines in underground courts established by these groups (Chamberlain & Haider-Markel). Though this study examines pre-9/11 data, the authors argue that their work provides lessons for post-9/11 state-level policy (Chamberlain & Haider-Markel, p. 449). This study provides an interesting example of scholars utilizing pre-9/11 data to conduct studies which may bear upon the post9/11 world. A second article, published in 2008, examines the 9/11 victims compensation fund (Hadfield, 2008). Hadfield highlights a number of concerns about the fund, including “a deeply troubling trade-off between money and a host of nonmonetary values that respondents thought they might obtain from litigation” (Hadfield, p. 647). Though the fund was meant to compensate victims of the attacks, Hadfield finds that those legally entitled to relief through the fund often delayed applying for this relief because they thought they might obtain a nonmonetary good (such as policy change) by waiting or going to court. Still other victims hoped to gain “information from otherwise inaccessible sources, [such as] the decision makers who determined airline and World Trade Center fire safety procedures” through a lawsuit (Hadfield, p. 647). Even though a streamlined legal remedy was designed following 9/11, this research uncovered potential obstacles in its operation and to the operation of future funds of the sort, namely that victims often framed the right to go to court as a responsibility of good citizenship following a national tragedy (Hadfield, p. 673). This study raises an important point with respect to why additional social science research is needed when so much traditional, doctrinal legal research has already been published in relation to the topic of terrorism. Social science is particularly useful for uncovering obstacles or issues that impact the functioning of social processes (in this case, legal processes). It is certainly true that more traditional, nonempirical legal analyses are also useful and even that such analyses may arrive at the same conclusions as an empirical study. Yet, social science allows for the systematic assessment of potential issues, often providing statistically reliable evidence regarding just how frequently a particular problem occurs and also what types of factors influence the process. Another example of this can be seen with respect to an article published in 1993 (Enders & Sandler, 1993). Enders and Sandler used time series analysis to examine whether or not two counterterrorism laws enacted in 1984 actually prevented attacks on US interests (Enders & Sandler). Specifically, the first law put into place more severe penalties for acts of hostage-taking, destroying aircraft or airport facilities, and using a weapon on an aircraft (Enders & Sandler, p. 834). The second law allowed the US attorney general to give rewards to those providing information about terrorism (Enders & Sandler, p. 834). In the end, the results of the analysis suggested that these laws were not effective in reducing terror attacks on US interests in the years following their enactment. Yet, knowledge of this result would not have been possible without utilizing social scientific methods (in this case, statistical analysis) to examine a large dataset of attacks on US interests over the course of multiple years.

286

L.M. Merola

It is clear from this discussion that the research which evaluates specific legal policies enacted for counterterrorism purposes is limited. In many regards, this is not surprising, as others have provided evidence that empirical research which evaluates counterterrorism measures more generally is extremely sparse (Lum et al., 2006). In addition to evaluations of legal policy, research is needed which assesses nonlegal policies that have implications for law or legal norms. In both cases, a greater effort to integrate legal scholarship with other types of policy-relevant outcome evaluations also seems likely to increase the consideration of legal issues when policy is developed. Since empirical research evaluating specific post-9/11 legal policies is so scarce, it will also be necessary to expand the scope of this discussion during the next sections of this chapter. Perhaps there are additional areas of law-related empirical research that (while not evaluating specific laws or policies) provide a useful foundation for scholars wishing to understand and expand the evidence base in this field.

Research Related to Changes in Public Support for Civil Liberties Following 9/11 The largest quantity of existing empirical research related to the topic of law and terrorism investigates the potential for alterations in public support for expansive civil liberties following 9/11. It is not that surprising that researchers have opted to focus on this topic, since questions about the proper scope of post-9/11 civil liberties have been an important part of the public discourse since the attacks. Following 9/11, there existed a sense of concern that Americans might turn away from their historical commitments to broad-based civil liberties in an attempt to bolster security in the face of terrorism. According to a key researcher in this area, “at no other time in American history [did] citizens [believe] that they would have to sacrifice a certain measure of freedom and liberty to be safe and secure. Cherished values of liberty and security, usually unnoticed, taken for granted, now had tangible consequences” (Davis, 2007, p. 218). Researchers interested in examining questions of public support for civil liberties also possessed a fairly extensive pre-9/11 foundation of empirical research upon which to build (Marcus, Sullivan, Theiss-Morse, & Wood, 1995; McClosky & Brill, 1983; Prothro & Grigg, 1963; Stouffer, 1955; Sullivan, Piereson, & Marcus, 1979). Much of this pre-existing empirical research examined what has been termed “political tolerance” or the willingness of the American public to support expansive civil liberties protections for members of groups holding beliefs outside of the mainstream of society. Early work in this field measured Americans’ attitudes toward extending rights to those who did not “conform” to society, most frequently referencing communists, socialists, and atheists (Stouffer, 1955). According to Stouffer, although large numbers of Americans endorsed the ideals of civil liberties in the abstract, the public exhibited resistance to these principles with respect to their application to actual situations and specific nonconforming groups (1955). In fact, the

12

Evaluating the Legal Challenges and Effects of Counterterrorism Policy

287

sheer prevalence of political intolerance present in early data on the subject forced researchers to question the extent to which Americans actually supported the free exercise of civil liberties. Over the course of decades, numerous scholars confirmed these findings, rendering the lack of political tolerance a persistent contradiction within the context of the otherwise robustness of American democracy (McClosky & Brill, 1983; Prothro & Grigg, 1963, Sullivan, Piereson, & Marcus, 1979). Over time, researchers in this field learned quite a bit about the demographic and other individual-level characteristics that are associated with an increased likelihood that an individual will be willing to support robust civil liberties. For example, a number of researchers have provided evidence that better-educated, more politically active citizens are more supportive of civil liberties (Davis, 1975; Gibson, 1987; Lawrence, 1976; McClosky & Brill, 1983; McClosky & Zaller, 1984; Nunn, Crockett, & Williams, 1978; Prothro & Grigg, 1963; Sniderman, Tetlock, Glaser, Green, & Hout, 1989; Stouffer, 1955; Sullivan et al., 1982). Further, those living in more urban areas, who are younger and who are less conservative in ideology, have also been more likely to support robust civil liberties in a number of studies (McClosky & Brill, 1983; McClosky & Zaller, 1984; Sniderman et al., 1989; Stouffer, 1955; Sullivan et al., 1982). However, a number of psychological traits, such as dogmatism (Stouffer, 1955), neuroticism, and extraversion (Marcus et al., 1995), seem to be associated with lowered support for civil liberties. More recently, in addition to examining the level of political tolerance present in the American public and associating variations in tolerance with specific individual characteristics, scholars have also attempted to understand the mental processes which lead to decisions supportive of civil liberties (Bobo & Licari, 1989; Kuklinski, Riggle, Ottati, Schwarz, & Wyer, 1991). For example, following a series of experiments embedded within surveys, one group of scholars provided evidence that the content of information can alter citizen support for the maintenance of broad civil liberties guarantees (Marcus et al., 1995, pp. 78–79). Marcus et al. theorized that this effect occurs when individuals feel threatened by the information that they have encountered regarding a disliked group’s activities. And, since threat is an integral part of the model (Marcus et al., 1995; Stouffer, 1955), this research certainly raises questions about the potential impact of post-9/11 debates surrounding the proper scope of civil liberties on a public confronted with large amounts of threatening information related to terrorism (Davis, 2007; Merola, 2007). Though much of the research in this field was conducted prior to 9/11 (and, as a result, did not refer directly to terrorism), it has provided a strong theoretical foundation for examinations of the impact of terrorism information upon civil liberties during the post-9/11 era (Davis, 2007; Davis & Silver, 2004; Merola, 2007, 2011; Scheufele, Nisbet, & Ostman, 2005). Authors working in this area of post-9/11 law-related scholarship have argued that the terror attacks “dramatically changed the political context” in the United States (Sullivan & Hendriks, 2009, p. 375). This is particularly important in light of the research mentioned above which has demonstrated that decisions about civil liberties are significantly influenced by the information environment (Marcus et al., 1995). Certainly, this is not to say that individual-level traits, socialization,

288

L.M. Merola

or other long-term individual predispositions cease to play a significant role in support for civil liberties in times of crisis. Rather, it has been demonstrated in survey-experimental research that all of these influences – long-term “standing” decisions about civil liberties, individual traits, and the content of information – significantly impact views about the proper scope of rights (Marcus et al., 1995). Yet, the environment is also influential. Despite the fact that Americans are socialized to respect assertions of rights and to value equality before the law, even a single encounter with certain types of negative, group-specific, or other information may lead to qualifications in support for expansive civil liberties, a key norm of the American legal system (Marcus et al., 1995; Nelson, Clawson, & Oxley, 1997). Following 9/11, the American public watched news coverage in record numbers (Althaus, 2002). Given the amount of discussion that has taken place in the media and also in the public debate regarding “balancing civil liberties with security,” it is logical to assume that Americans’ assessments of the proper scope of rights may have undergone some revision. Indeed, the measurements of support for civil liberties taken directly following the attacks seemed to validate concerns that extensive changes had taken place within our society in this regard. For example, in their review of this area of research, Sullivan and Hendriks report the results of a Pew Research Center poll that found that 55% of respondents agreed that sacrificing civil liberties would be necessary in order to curb terrorism, whereas that number was 29% when measured in 1997 (2008, p. 380, citing Pew Research Center, 2001). And, equally consequential in times of crisis, post-9/11 researchers in this field have provided evidence of a substantial relationship between perceptions of terrorist threat and the willingness to sacrifice civil liberties (Davis, 2007; Huddy, Feldman, Capelos, & Provost, 2002). In fact, in post-9/11 models, perception of the threat of terrorism has been shown to be of greater or equal consequence to highly influential individual-level variables, such as respondent ideology, level of education, degree of authoritarianism, and media use (Huddy, Feldman, & Taber, 2005; Nisbet, Ostman, & Shanahan, 2008). Empirical research has also provided evidence that threat cues rose by at least 66% in a random sample of post-9/11 media coverage of civil liberties issues and that this increase continued for several years following the attacks (Merola, 2011). Further, this transmission of threat may be particularly significant, as researchers have also argued that it is a sense of societal (as opposed to personal) threat from terrorism that is most potent when individuals make decisions about the proper scope of civil liberties (Davis, 2007; Gibson & Gouws, 2003; Huddy et al., 2002). These findings suggest the possibility for dramatic changes to key democratic principles following serious attacks. Yet, research completed directly following the attacks also provided some indications that predictions of large scale change should be approached cautiously. For example, even in the context of a willingness on the part of large percentages of Americans to agree to post-9/11 constrictions of civil liberties in principle, the percentage of Americans willing to cede specific rights in order to combat terrorism often diminished in response to more specific survey questions (Davis, 2007). Additionally, measurements taken later in the post-9/11 period seem to suggest that conceptions of rights eventually began to stabilize, at

12

Evaluating the Legal Challenges and Effects of Counterterrorism Policy

289

least in the sense that fewer Americans readily agreed to broad-based constrictions of civil liberties in surveys (Davis, 2007; Sullivan & Hendriks, 2009, p. 387). And, as time passed, greater numbers of Americans likewise expressed concerns about increasing the government’s power too much, as trust in government declined (Davis, 2007). Some explanations for this stabilization suggested by the literature have been this decline in trust in government (Davis), the lessening of perceived threat over time (Nisbet & Shanahan, 2004), and fewer elite cues about threat being transmitted in the popular media as the post-9/11 period progressed (Merola, 2011), to name but a few. Unfortunately, though, polls which have suggested that Americans are “back to normal” in terms of their responses to survey questions about civil liberties may still mask some substantial societal changes. First, as Sullivan and Hendriks argue, these surveys may not account for significant changes in law, which remain in force even after a period of threat has passed (2009). These changes “can shape politics for years to come” (Sullivan & Hendriks, 2009, p. 388) and current crises may also shape expectations about the handling of future crises (Merola, 2007, p. 223). Davis argues that Americans will be more wary of future claims that rights need to be constricted in times of threat, now that they have experienced the post-9/11 era (2007, p. 224). Yet, the opposite may also occur in some cases. Some suggestions made by political leaders regarding the curtailment of rights to meet a crisis situation may no longer be as shocking to the American conscience as they once were (Merola, 2007). Moreover, the notion that conceptions of rights may eventually “rebound” once a threat dissipates may not be useful in the event of a series of attacks or serious crises. Additionally, even where the literature in this field is relatively complete, it has often not been used as a platform for designing and testing interventions nor has it been linked to policy which may curtail some of the effects of terror attacks on the public. For example, a key finding of the political tolerance literature seems to be how significantly the content of information influences the public in times of threat (Marcus et al., 1995; Merola, 2007). Yet, though a number of studies have examined the content of media coverage related to terrorism more generally, little research has investigated the framing of legal issues since 9/11. There is one study which investigates alterations in print and broadcast media content related to civil liberties before and after September 11, 2001 (Merola, 2011). The study focuses specifically upon the types of content (such as threatening information, group-specific information, and contextual information) that prior experimental research has suggested may influence individual decision-making concerning civil liberties. Overall, after examining the post-9/11 media coverage, the study finds the frequent presence of types of information that the experimental literature has suggested tends to decrease support for civil liberties (such as threat language) and the infrequent presence of the types of information which may have a protective effect (such as reminders about the American tradition of expansive civil liberties) (Merola). However, the study also finds significant differences between the types of content routinely found in print media and those contained in broadcast media, with the broadcast far more likely to contain content that appears destructive to the norm of expansive civil liberties in experiments (Merola). These results are also consistent

290

L.M. Merola

with another study that suggests that members of the public who frequently consumed print media remained more supportive of traditional norms of expansive civil liberties than those who frequently watched national television news (Scheufele et al., 2005). Taken together, this research suggests that types of information and media choices may have real consequences for how a populace grapples with the after effects of a terrorist attack. Yet, to date, researchers have largely not linked research on information to support for civil liberties, legal policy, or other effects, such as alterations in the legitimacy of legal institutions in the eyes of the public. The link with information is particularly important, since most members of the US public were not direct victims of the attacks of 9/11, but rather experienced the threat entirely through the media. In order to understand the impacts of terrorism on the public, we must understand the content and influences of messages from media and other elites that reach the public about terrorist events. Additionally, this raises perhaps an even more salient point. If information can negatively impact legal norms, can information also be used to protect communities? Of course, the need for research regarding how to make communities more resilient extends beyond questions related to law and rights. Researchers in other fields have often grappled with this issue since 9/11 (Norris, Tracy, & Galea, 2009; Pfefferbaum, 2009). Yet, in the context of rights, can interested researchers also work proactively to examine the ways in which information or other counter measures may make communities more resilient to terrorism? Currently, this literature provides a great deal of evidence regarding the personality traits and circumstances that correlate with continued support for civil liberties and many of these relationships have persisted following 9/11. Yet, research in the field has generally not been translated into particular policy suggestions or interventions that can protect legal norms, nor has the field examined information’s impacts on opinion about other legal norms or legal institutions. To date, there is one experimental study that has reported a potential protective effect for certain types of information (Merola, 2007). In this case, in a survey-experiment utilizing students, a treatment group exposed to information reminding participants of the American tradition of expansive civil liberties expressed greater support for civil liberties in the context of hypothetical terror attack scenarios than did a group that received information advocating dispensing with civil liberties as a response to terrorism (Merola). In each scenario, this information was delivered to participants via a person-on-the-street interview during a fictional nightly newscast. This study tested only one type of information, but there may be others which could be utilized by government officials to serve a protective function in times of crisis. At least, researchers might be able to provide some guidance to government officials regarding types of information to avoid in order not to exacerbate a crisis. Davis (2007) argues that discussion of many of the policy initiatives promulgated by government following 9/11 – though intended to protect Americans – actually compounded the anxiety that Americans felt. Indeed, it is possible that some of our own actions or perhaps the content of our discourse actually made the impacts of the attacks even more severe than they might otherwise have been. Additional research related to how to utilize information in the most beneficial way (or, at least, not in a harmful way) would be an important contribution.

12

Evaluating the Legal Challenges and Effects of Counterterrorism Policy

291

Also, the methodologies used frequently by pre-9/11 researchers studying support for civil liberties – that of the experiment or the embedded survey-experiment – could be particularly helpful to researchers. In the case of a survey-experiment, in addition to the data derived from the survey itself, an embedded experiment allows researchers to directly test the impact of a particular intervention, such as the influence of specific categories of information on decision-making by members of the public. At present, experiments are not used very frequently to examine law-related research questions (with respect to terrorism or otherwise). Yet, a close causal link between an intervention or event and an outcome can be demonstrated in an experiment. In turn, the ability for researchers to provide strong evidence of such a causal link suggests that experimental methodologies represent an important tool for researchers interested in furthering evidence-based policy.

Elite Discourse, Opinions, and Cues The previous section of this chapter questioned the degree to which we understand the impacts of discourse regarding many of the policy decisions made following 9/11. In the last section, we explored the considerable empirical research devoted to understanding support of civil liberties among members of the general public. Indeed, as discussed above, some researchers have focused specifically on examining public support for civil liberties in the post-9/11 era and have even found that a significant portion of the pre-9/11 research remains valid. In comparison, however, we actually know very little about elite discourse, opinions, or even the role of elites (even legal elites) in shaping support for civil liberties and legal policy in times of crisis (Merola, 2007, 2011; Sullivan & Hendriks, 2009). Generally, this is not surprising, as relatively few elite opinion studies exist within the empirical legal scholarship more broadly. Perhaps paradoxically given the expertise required to understand many terror-related issues, the majority of post-9/11 opinion research has focused on the public. Yet, particularly with respect to an issue like terrorism which raises complex policy questions, it is important to understand elite opinion in order to better comprehend the process by which the public acquires its information and policy preferences about terrorism. Seminal research in the field of communications has argued that “if we are interested in the quality of information reaching the public, we must understand how it is manufactured, which is to say, we must understand the politics of expert communities as they relate to the generation and diffusion of knowledge claims, policy recommendations, and general frames of reference” (Zaller, 1992, p. 319). Legal elites are no exception, yet their role in the framing and diffusion of post-9/11 legal information (such as that related to security and civil liberties) has largely not been assessed systematically. The same is also true for other types of elites, particularly with respect to legal issues. To date, two studies of the post-9/11 behavior and opinions of legal elites have been conducted that are relevant to this discussion. The first study involved a content analysis of the legal community’s discourse about civil liberties, as published in

292

L.M. Merola

law review articles during the post-9/11 period (Merola, 2007, 2008). The goal of the analysis was to examine the role that these elites have played in either supporting or defending civil liberties during the post-9/11 period, since a number of scholarly articles have either explicitly or implicitly argued that elites play an important role in the maintenance of democratic norms in times of crisis (McClosky, 1964; McClosky & Brill, 1983). After examining the language in a large sample of law review articles related to post-9/11 civil liberties, the study found that legal elites have incorporated threat information into this professional discourse with relative frequency, at times even more so than did some media sources (Merola, 2008). This is important both as a barometer of the impact of the terror crisis on the legal community and also as information ultimately communicated to policy makers and to members of the public. For most of the post-9/11 period, this attentiveness to the threat on the part of legal elites was also combined with increased attempts at opinion leadership; yet, on average, law review articles employed relatively deferential language until the year 2006. Following 2006, the language became increasingly less deferential. As a result, the study found evidence to support the hypothesis that legal elites mounted a vigorous defense of civil liberties in their discourse following 9/11. However, this finding must also be interpreted cautiously, given the results with respect to threat and deference that were seen in the years directly following 9/11. The second article investigates the opinions of legal elites directly through a survey on the topics of post-9/11 civil liberties and legal policy (Merola, 2009). Elites in the legal community are often assumed to be unwavering in their defense of robust legal institutions and expansive civil liberties (McClosky & Brill, 1983). However, few studies have specifically examined these questions by surveying an elite sample. Instead, these viewpoints are most often attributed to members of the legal community as a result of their professional training, socialization, and high level of education (McClosky & Brill). Since numerous studies have demonstrated that higher levels of education are associated with stronger support for civil liberties (Davis, 1975; Gibson, 1987; Lawrence, 1976; McClosky & Brill, 1983; McClosky & Zaller, 1984; Nunn et al., 1978; Prothro & Grigg, 1963; Sniderman et al., 1989; Stouffer, 1955; Sullivan et al., 1982), it has been largely assumed that this evidence translates into a depth of knowledge about the views of the legal community. Yet, attorneys were not immune to the threat or anxiety that was so prevalent in the remainder of our society after 9/11. Legal elites, like all Americans, were forced by the ongoing terror crisis to regularly consider the ramifications of “trading” civil liberties for enhanced security. The experimental research in the field of political tolerance would certainly seem to suggest that the commitment of elites (including legal elites) to expansive civil liberties is more durable than the commitment of the general public in the face of threat (McClosky, 1964). And, generally, this survey found legal elites to be very supportive of civil liberties in 2008. However, following an experiment, these individuals also were influenced by the content of terrorism information (Merola, 2009). In fact, despite their high levels of education and the possession of subject-matter expertise, the mere mention of “terrorism” was associated with decreases in support for civil liberties among legal elite respondents.

12

Evaluating the Legal Challenges and Effects of Counterterrorism Policy

293

Though perhaps not to the degree that would be expected in a sample of the general public, these decreases were significant (Merola). This result demonstrated that (much like the nonelite individuals in other political tolerance experiments) the terror threat was still quite salient to members of the legal community and that their education and expertise did not act as a complete barrier to the effects of terrorrelated information. Additionally, the survey yielded some results suggesting that the elite respondents were more willing to accept instances of racial profiling (particularly for those appearing to be Middle Eastern and at airports) than other abridgements of civil liberties discussed on the survey (Merola). Scholars in the field of law and terrorism can add to the evidence base surrounding terror-related legal questions by focusing more attention on the opinions of elites, particularly legal elites. With respect to many post-9/11 policies, a considerable amount of public debate has focused on the relative merits of different choices, such as decisions about where to try terror suspects. However, few attempts have been made to assess the views of legal or other experts systematically, leaving an important potential source of guidance for policy makers underdeveloped. Given their professional expertise, systematic assessments of the views of the legal community seem a valuable resource for policy makers who often do not possess this specialized knowledge. In addition to evaluations of legal policy, systematic investigations of the views of experts may help policy makers to recognize potential issues or to eliminate certain policy choices as infeasible or ineffective. The vantage point of evidence-based counterterrorism research is particularly helpful in focusing scholars on the ways that their work may be linked with decision makers. In this way, evaluation research helps to make scholarship more relevant while making terror policy more effective.

Research Related to Executive Authority and Separation of Powers The scope of executive authority emerged as one of the key constitutional questions during the post-9/11 period. Like many of the topics discussed in this chapter, a good deal of traditional legal scholarship has been published concerning the proper exercise of post-9/11 executive authority. Specifically, there was much discussion in the legal scholarship regarding the president as “unitary executive” (Scheuerman, 2006). The phrase “unitary executive” encompasses a view that all executive authority must be controlled by the President and may not be interfered with by other branches of government, particularly in the area of national security (Scheuerman). The scope of executive authority became a key issue of the post-9/11 era because executive authority was implicated directly in many of the salient policy issues of the time. For example, the question of whether or not the President and armed forces retained exclusive control over the detainees at Guantanamo Bay (or if these individuals could avail themselves of habeas corpus review by federal courts) became an important issue during the administration of George W. Bush.

294

L.M. Merola

A small number of researchers have examined issues related to separation of powers following 9/11 through empirical studies. These scholars have done so through examinations of the functioning of separation of powers during wartime. Historically, the prevailing view of separation of powers research has been that the legislative and judicial branches of the government have a tendency to defer to the executive during times of crisis, particularly in the foreign policy realm (Ducat & Dudley, 1989; Yates & Whitford, 1998). With respect to the question of judicial deference, recent studies of judicial voting have supported this view. For example, Epstein et al. have provided evidence that US Supreme Court justices decide cases related to civil liberties in a more conservative manner when the nation is at war (Epstein, Ho, King, & Segal, 2005). According to the authors, this phenomenon actually occurs to the greatest extent when the case at issue does not directly implicate the war (Epstein et al.). Another study by Clark also confirms some of these findings (2006). Clark demonstrates that judicial preferences seem to undergo a fundamental shift in criminal cases during times of war, but that there is no evidence that judicial preferences change in other types of cases. Rather, this study suggests that judicial decisionmaking in noncriminal cases remains unaltered during times of war (Clark, 2006). For this reason, Clark argues that “concerns about judicial deference to the executive during times of war may not be as serious as conventional wisdom suggests” (Clark, p. 398). Clark’s research goes a long way toward confirming that, at least in terms of noncriminal cases, the separation of powers – a cornerstone principle of American government – remains intact in times of war. Though the two studies described above do not focus solely on changes in judicial decision-making during the post-9/11 period, both of these articles consider the “war on terrorism” within their models. For those researchers interested in evidence-based counterterrorism related to law and legal institutions, these findings represent an important foundation. Though not tests of counterterrorism measures, these studies begin to construct an evidence base regarding how fundamental American legal principles fare when the country is under attack or in grave crisis. A future direction of research might be to expand upon these findings and to think about how these studies and others might inform our thinking regarding the best ways to protect fundamental aspects of our system during times of extreme threat. Empirical researchers might compare the post-9/11 time period explicitly to other wars and assess the differences that occur in the functioning of legal institutions when society is under threat specifically from terrorism. Additionally, as mentioned in the last section, scholars could focus a bit more on assessing issues like the proper scope of executive authority in times of crisis through expert opinion studies. Though many analyses of issues related to separation of powers have been published by legal experts since 9/11, expert opinions about specific policies have generally not been assessed systematically. One existing paper (described in greater detail above) seeks to do that by examining views of policy alternatives through an expert survey of attorneys (Merola, 2009). In a survey of private attorneys and law professors with expertise in constitutional law, more than 83% of respondents indicated that they felt that the scope of executive

12

Evaluating the Legal Challenges and Effects of Counterterrorism Policy

295

authority had been expanded too far since 9/11. Further, similarly large majorities of these respondents indicated that terror suspects should not be treated as illegal enemy combatants and that acts of terror should be treated as crimes and not as acts of war (Merola). As policy makers choose between alternatives – such as continuing the detention of terror suspects at Guantanamo Bay or their relocation to the United States – the systematic assessment of expert opinion on the topic could help to guide decision-making. Another article in progress (also mentioned briefly above) attempts to begin to construct an evidence base in this area through an investigation of the feasibility of policy alternatives to unfettered executive detention at Guantanamo Bay. For example, one alternative that has been quite controversial is the relocation of remaining detainees to US soil. Yet, at this point, social scientific evidence regarding the feasibility of this course of action is extremely limited. For example, there are currently very few evidence-based studies which attempt to assess the challenges of incarcerating religious extremists in civilian prison facilities. Some of the analytical literature surrounding this issue questions the feasibility of housing these prisoners either as a single group or in a configuration where these individuals might encounter other prisoners (Hannah, Clutterbuck, & Rubin, 2008). If housed as an isolated group, the concern is that these prisoners would grow in strength with increased numbers and through isolation from nonextremists. If dispersed throughout a prison containing other inmates, the concern is that these prisoners might convert others to their radical beliefs. Recently, the author of the chapter and a group of undergraduate honors students at George Mason University conducted a national survey of state-level maximum security prisons in the United States with the goal of assessing how prison administrators across the county have addressed these and other issues related to the detention of religious extremist inmates (Merola et al., 2011). Though these prisons do not deal specifically with Guantanamo Bay detainees, they do manage other prisoners who have adopted extremist religious beliefs. Additionally, the survey asked about staff training related to religious extremism, the monitoring of inmate communications, and whether or not valuable antiterrorism information could be collected in prisons. Admittedly, this study addresses only a small part of the issues inherent in any proposed relocation of prisoners from Guantanamo Bay or even the incarceration of any future individuals convicted of terrorism. Knowledge accumulation through social science is always incremental in nature and, in some cases, necessary data may even be inaccessible or infeasible to collect. However, the notion of evidence-based counterterrorism can help sharpen our focus on linking research to policy options.

Research Related to Judicial Decision-Making On the whole, a robust empirical literature exists with respect to the topic of judicial decision-making, particularly at the US Supreme Court level. A variety of competing models have been tested using a wide range of methodologies and data from

296

L.M. Merola

different levels of the court system (Clayton & Gillman, 1998; Epstein & Knight, 1998; Maltzman, Spriggs, & Wahlbeck, 2000; Segal & Spaeth, 2002; Woodford, 1981). Yet, despite the existence of a robust literature in this area and the fact that questions related to the proper scope of judicial authority have been central to post9/11 legal debates, there are only a small number of social science articles which evaluate post-9/11 judicial decision-making. First, as mentioned above, quantitative research in this area has examined the question of whether or not judges alter their decisions during times of war (Clark, 2006; Epstein et al., 2005). Though these projects did not focus specifically on understanding post-9/11 alterations in the judiciary, these studies can serve as models for additional literature on judicial decisionmaking in times of terrorism. In addition to these studies of judicial voting, another existing article utilizes an experimental methodology to examine post-9/11 judicial decision-making. Generally, few researchers in the field of judicial decision-making employ an experimental methodology, but it is worth highlighting the use of this methodology here. Hagan, Farrales, and Jasso (2008) conduct a factorial experiment to examine how Iraqi judges might adjudicate cases in which American military members are accused of torturing suspected terrorists. In reality, Iraqi judges did not preside over such cases, since crimes committed by American military are adjudicated within the American military justice system. However, the researchers utilized the experimental method to shed light on various aspects of the decision-making of these judges. In fact, the researchers provided evidence that hypothetical sentences became stricter when the individual judge was highly concerned that instances of torture would encourage further violence in Iraq (Hagan et al.). According to the researchers, “the judges who were less fearful of violence were more lenient and accommodating of torture by Coalition forces. The implication is that the less fearful judges were freed by an indeterminate law to advance Coalition goals through lenient punishment of torture” (Hagan et al., p. 605). One potential benefit of experimental methodologies is that empirical researchers may use experiments of this type to conduct simulated tests in areas where it might be impossible or impractical to collect “real world” data. This is particularly useful in the case of terrorism research, where data may be inaccessible for many reasons. Though not always as reliable as experiments or empirical analyses based upon real world data, laboratory experiments have been utilized in many fields to simulate justice processes of interest. For example, the research conducted by psychologists in the field of juror decision-making provides an example of this (Diamond & Rose, 2005). In lieu of witnessing the deliberations of actual jurors, psychologists have attempted to recreate juror decision-making under laboratory conditions. Certainly, the conditions under which such experiments take place must be carefully monitored in order to represent the actual process as closely as possible. Admittedly, judicial decision-making may be a difficult area in which to utilize experimental methods under most circumstances, due to the difficulty in replicating the training and socialization of judges among experimental participants. However, it is worth highlighting the Hagan et al. study as one of only very few studies of post-9/11 judicial decision-making. Additionally, it is again worth

12

Evaluating the Legal Challenges and Effects of Counterterrorism Policy

297

emphasizing the utility of experimental studies more generally in the context of terrorism research, as often these can represent creative solutions by researchers who cannot collect real world data.

Conclusion Despite substantial expenditures for the purposes of counterterrorism, few social scientific evaluations of counterterror measures have been published (Lum et al., 2006). This is particularly the case with respect to empirical studies of law-related counterterrorism policy. Traditionally, many legal scholars have not utilized the methods of social science (such as surveys, experiments, or large scale data analysis, for example) in their work, so perhaps this scarcity of empirical research merely reflects the larger orientation of the field. Yet, over time, this orientation is changing, as greater numbers of legal scholars pursue joint degree programs or embrace the perspectives of fields like law and society and empirical legal studies. Regardless, it remains the case that the momentum behind publishing doctrinal studies related to law and terrorism has somehow largely not translated into scholarly focus on empirical studies in this area. The goal of this chapter was to explore the existing empirical legal research related to counterterrorism measures. Though not meant to be an all-encompassing literature review, this chapter surveyed a varied literature related to law and terrorism, finding that only a very small number of authors have actually conducted evaluations of specific post-9/11 legal policies. In fact, the author could only locate a few studies that could be likened to the evidence-based evaluations of policy that take place in other fields. Yet, this should not minimize the importance of other types of studies related to law and terrorism that may be used both independently and as a foundation for building the empirical evidence base further. For example, a comparatively robust empirical literature exists related to public support for the norm of expansive civil liberties during the post-9/11 period. Though not strictly evaluation research, the findings from this literature are fairly comprehensive and may be linked to specific measures in a number of ways. For example, one of the key findings of this literature relates to the impact of information on support for civil liberties. For this reason, it was suggested that researchers might investigate specific measures related to information that could make communities more resilient to terrorism. Additionally, the author located a small number of empirical studies related to legal elite opinions, an area of research that seems likely to be helpful to those making decisions about counterterrorism policy. After all, legal elites (as well as other elites) possess expertise that can effectively guide policy makers, particularly when the opinions of an expert community are assessed systematically. Finally, some additional empirical studies in the areas of separation of powers and judicial decisionmaking were discussed. Here, pioneering studies – such as large scale studies of judicial voting in times of war or a wholly experimental study investigating post-9/11

298

L.M. Merola

judicial decision-making – were described as examples of cutting edge empirical research related to law and terrorism. Among other reasons, these studies are also noteworthy because their authors embraced methodologies novel to their fields when it would aid their research. Currently, a discussion of law-related terrorism research in the context of the evidence-based paradigm is somewhat challenging to accomplish, since so few empirical studies exist and even fewer have connected their investigations to specific policy questions or evaluated specific outcomes. Yet, it is for this reason that a survey of the law and terrorism literature from the vantage point of evidence-based counterterrorism is interesting. An important tenet of the evidence-based paradigm is to forge a closer link between researchers and practitioners with the goal of making policy more effective (Pfeffer & Sutton, 2006). This paradigm demands that practitioners account for the findings of scholarly research while developing policy, but it also suggests a bridge whereby researchers can speak to practitioners and have greater influence on policy development (Pfeffer & Sutton, 2006; Rousseau, 2006). In the end, greater attention to creating an empirical evidence base upon which policy makers can draw in times of crisis can only increase the impact of our scholarship.

References Althaus, S. L. (2002). American news consumption during times of national crisis. PS: Political Science and Politics, 35, 517–521. Bobo, L., & Licari, F. C. (1989). Education and political tolerance: Testing the effects of cognitive sophistication and target group effect. Public Opinion Quarterly, 53, 285–308. Chamberlain, R., & Haider-Markel, D. P. (2005). “Lien on me”: State policy innovation in response to paper terrorism. Political Research Quarterly, 58(3), 449–460. Clark, T. S. (2006). Judicial decision making during wartime. Journal of Empirical Legal Studies, 3, 397–419. Clayton, C., & Gillman, H. (Eds.). (1998). Supreme Court decision-making: New institutionalist approaches. Chicago: University of Chicago Press. Davis, J. A. (1975). Communism, conformity, cohorts, categories: American tolerance in 1954 and 1972–72. American Journal of Sociology, 81, 491–513. Davis, D. W. (2007). Negative liberty: Public opinion and the terrorist attacks on America. New York: Russell Sage Foundation. Davis, D. W., & Silver, B. D. (2004). Civil liberties vs security: Public opinion in the context of the terrorist attacks on America. American Journal of Political Science, 48, 28–46. Diamond, S. S., & Rose, M. R. (2005). Real juries. Annual Review of Law and Social Sciences, 1, 255–284. Ducat, C. R., & Dudley, R. L. (1989). Federal district judges and presidential power during the postwar era. Journal of Politics, 51, 98. Enders, W., & Sandler, S. (1993). The effectiveness of antiterrorism policies: A vector-autoregression-intervention analysis. The American Political Science Review, 87(4), 829–844. Epstein, L., Ho, D., King, G., & Segal, J. (2005). The effect of war on the Supreme Court. New York University Law Review, 8, 1. Epstein, L., & Knight, J. (1998). The choices justices make. Washington: CQ Press. Gibson, J. L. (1987). Homosexuals and the Ku Klux Klan: A contextual analysis of political tolerance. Western Political Quarterly, 40, 427–448.

12

Evaluating the Legal Challenges and Effects of Counterterrorism Policy

299

Gibson, J. L., & Gouws, A. (2003). Overcoming intolerance in South Africa: Experiments in democratic persuasion. New York: Cambridge University Press. Hadfield, G. K. (2008). Framing the choice between cash and the courthouse: Experiences with the 9/11 victim compensation fund. Law and Society Review, 42(3), 645–683. Hagan, J., Farrales, G., & Jasso, G. (2008). How law rules: Torture, terror, and the normative judgments of Iraqi judges. Law and Society Review, 42, 605–643. Hannah, G., Clutterbuck, L., & Rubin, J. (2008). Radicalization or rehabilitation understanding the challenge of extremist and radicalized prisoners. Santa Monica: Rand. Huddy, L., Feldman, S., Capelos, T., & Provost, C. (2002). The consequences of terrorism: Disentangling the effects of personal and national threat. Political Psychology, 23(3), 485–509. Huddy, L., Feldman, S., & Taber, C. (2005). Threat, anxiety, and support of antiterrorism policies. American Journal of Political Science, 49, 593–608. Huq, A. Z., & Muller, C. (2008). The war on crime as precursor to the war on terror. International Journal of Law, Crime, and Justice, 1, 215–229. Kalyvas, S. N. (2006). The logic of violence in civil war. Cambridge: Cambridge University Press. Kuklinski, J. H., Riggle, E., Ottati, V., Schwarz, N., & Wyer, R. S. (1991). The cognitive and affective bases of political tolerance judgments. American Journal of Political Science, 35, 1–27. Lawrence, D. G. (1976). Procedural norms and tolerance: A reassessment. American Political Science Review, 70(1), 70–80. Lum, C., Kennedy, L. W., & Sherley, A. (2006). Are counterterrorism strategies effective? The results of the Campbell systematic review on counterterrorism evaluation research. Journal of Experimental Criminology, 2, 489–516. Lum, C., Merola, L., Willis-Hibdon, J., & Cave, B. (2010). License plate recognition technology: Impact evaluation and community assessment. Final report for the National Institute of Justice. Retrieved from http://gemini.gmu.edu/cebcp/LPR_FINAL.pdf. Accessed on July 29, 2011. Maltzman, F., Spriggs, J. F., & Wahlbeck, P. J. (2000). Crafting law on the Supreme Court: The collegial game. Cambridge: Cambridge University Press. Marcus, G. E., Sullivan, J. L., Theiss-Morse, E., & Wood, S. L. (1995). With malice toward some. Cambridge: Cambridge University Press. McClosky, H. (1964). Consensus and ideology in American politics. American Political Science Review, 58, 361–382. McClosky, H., & Brill, A. (1983). Dimensions of tolerance. New York: Russell Sage Foundation. McClosky, H., & Zaller, J. (1984). The American ethos: Public attitudes toward capitalism and democracy. Cambridge: Harvard University Press. Merola, L. M. (2007). A culture of crisis: Information and the scope of American civil liberties in an era of terrorist threat. Doctoral dissertation. Retrieved from ProQuest Digital Dissertations (publication number AAT 3271064). Merola, L. M. (2008). Threat, leadership and the discourse of legal elites in the post-9/11 period. Presented at the annual meeting of the Law and Society Association, Montreal, Canada. Merola, L. M. (2009). Measuring elite opinion on terrorism and civil liberties: A survey of the legal community. Presented at the annual meeting of the Law and Society Association: Denver, Colorado. Merola, L. M. (2011). Transmitting the threat: Media coverage and the discussion of terrorism and civil liberties since 9/11. Behavioral Sciences of Terrorism and Political Aggression, forthcoming: Washington, D.C. Merola, L. M., Bonometti, J., Day, M., Ellis, T., Hawkins, L., Khan, K., et al. (2011). The continuing policy demands of counterterrorism: The views of criminal justice professionals. Presented at the annual meeting of the American Society of Criminology. Nelson, T. E., Clawson, R. A., & Oxley, Z. M. (1997). Media framing of a civil liberties controversy and its effect on tolerance. American Political Science Review, 91, 567–584. Nisbet, E. C., Ostman, R., & Shanahan, J. (2008). Public opinion toward Muslim Americans: Civil liberties and the role of religiosity, ideology, and media use. In A. Sinno (Ed.), Muslim in western politics (pp. 160–199). Bloomington: Indiana University Press.

300

L.M. Merola

Nisbet, E. C., & Shanahan, J. (2004). MSRG special report: Restrictions on civil liberties, views of Islam, and Muslim Americans. Ithaca: Media and Society Research Group, Cornell University, http://www.yuricareport.com/Civil%20Rights/CornellMuslimReportCivilRights.pdf . Accessed on July 29, 2011. Norris, F., Tracy, M., & Galea, S. (2009). Looking for resilience: Understanding the longitudinal trajectories of responses to stress. Social Science and Medicine, 68, 2190–2198. Nunn, C. Z., Crockett, H. J., & Williams, J. A. (1978). Tolerance for nonconformity. San Francisco: Jossey-Bass. Pew Res. Cent. People Press. (2001). American psyche reeling from terror attacks. Survey report, September 19. Washington: Pew Research Center. http://people-press.org/2001/09/19/americanpsyche-reeling-from-terror-attacks/. Accessed on July 29, 2011. Pfeffer, J., & Sutton, R. I. (2006). Evidence-based management. Harvard Business Review, 84(1), 63–74. Pfefferbaum, R. (2009). The role of colleges and universities in building community resilience to disasters. National Social Science Journal, 33, 102–109. Prothro, J. W., & Grigg, C. M. (1963). Fundamental principles of democracy: Bases of agreement and disagreement. Journal of Politics, 22, 276–294. Rousseau, D. M. (2006). Is there such a thing as “evidence-based management?”. Academy of Management Review, 31(2), 256–269. Scheuerman, W. E. (2006). Emergency powers. Annual Review of Law and Social Sciences, 2, 257–277. Scheufele, D. A., Nisbet, M. C., & Ostman, R. E. (2005). September 11 news coverage, public opinion, and support for civil liberties. Mass Communication & Society, 8, 197–218. Segal, J. A., & Spaeth, H. J. (2002). The Supreme Court and the attitudinal model. New York: Cambridge University Press. Sniderman, P. M., Tetlock, P. E., Glaser, J. M., Green, D., & Hout, M. (1989). Principled tolerance and the American mass public. British Journal of Political Science, 19, 25–45. Stouffer, S. (1955). Communism, conformity, and civil liberties. New York: Doubleday. Sullivan, J. L., & Hendriks, H. (2009). Public support for civil liberties pre- and post-9/11. Annual Review of Law and Social Sciences, 5, 375–391. Sullivan, J. L., Piereson, J. E., & Marcus, G. E. (1979). An alternative conceptualization of political tolerance: Illusory increases 1950s–1970s. American Political Science Review, 73, 781–794. Sullivan, J. L., Piereson, J. E., & Marcus, G. E. (1982). Political tolerance and American democracy. Chicago: University of Chicago Press. Tyler, T. R., Schulhofer, S., & Huq, A. Z. (2010). Legitimacy and deterrence effects in counterterrorism policing: A study of Muslim Americans. Law and Society Review, 44(2), 365–402. USA-PATRIOT Act. (1991). Pub. L. No. 107-56, 115 Stat. 272. Weisburd, D., Feucht, T. E., Hakimi, I., Perry, S., & Mock, L. (Eds.). (2009). To protect and serve: Policing in an age of terrorism. New York: Springer. Woodford, H. J. (1981). Courts of appeals in the federal judicial system. Princeton: Princeton University Press. Yates, J., & Whitford, A. (1998). Presidential power and the United States Supreme Court. Political Research Quarterly, 51, 539. Zaller, J. (1992). The nature and origins of mass opinion. New York: Cambridge University Press.

Chapter 13

Public Opinion Research and Evidence-Based Counterinsurgency Clay Ramsay

Introduction In recent years public opinion survey work has played a significant role in parts of the world marked by conflict, and particularly by insurgency, counterinsurgency (CI), and terrorist tactics. In the Israel-Palestinian conflict there has been continuous regular polling of the Palestinian public since the 1990s, often financed by outside sponsors such as the European Union. In both Iraq and Afghanistan, polling has been extensive since the arrival of Western forces – both publicly available polling, organized by international news media or nongovernmental organizations, and polling conducted for the military, intelligence groups, and aid agencies. As some expert Canadian practitioners – responsible for military polling in Kandahar province – have expressed it, “opinion polls are used to formally assess campaign progress along some set of performance metrics (e.g., relying on polls to measure things such as public support for a constituted authority…), to guide information operation campaigns…and for general situational awareness of local attitudes and perceptions” (Vincent, Eles, & Vasiliev, 2009). A contradiction lies within the relationship between evidence-based CI on the one hand, and survey research on the other: the two both converge and diverge. There is convergence because the theory of CI focuses on securing the whole population, and this is understood as the end to which all other means, including both fixed and kinetic military activities, are subordinated. Survey research is focused on whole populations as well. In fact, random sampling cannot take place without a clear understanding of the universe to be sampled – that is, exactly who is within the social whole that is studied and what are its boundaries. Thus both CI and survey research have an essential concern with a social totality.

C. Ramsay (*) Center for International and Security Studies at Maryland (CISSM), University of Maryland, College Park, MD, USA e-mail: [email protected] C. Lum and L.W. Kennedy (eds.), Evidence-Based Counterterrorism Policy, Springer Series on Evidence-Based Crime Policy 3, DOI 10.1007/978-1-4614-0953-3_13, © Springer Science+Business Media, LLC 2012

301

302

C. Ramsay

But there is also divergence between evidence-based CI and survey research. Insurgency and CI create no-go zones and denied areas for each side, and these spots are of intense interest for CI practitioners. These spots are temporary, shifting parts of a larger whole – the full society. It is extremely difficult, perhaps impossible, to set bounds on a conflict area as a universe that can be sampled. Insurgent warfare that uses terrorism is not a total activity that consumes the entire society. Such a conflict is not a “total war” that seeks to embrace all the efforts of the whole population (as in the industrialized combatant countries of World War II). Nor is there anything permanent, or even always verifiable, about the boundaries of a zone that one force can, for the moment, deny to another force. This chapter tries to delineate what is of value and what is dubious in the relations between evidence-based CI and survey research. CI is both a military practice and a burgeoning field of study, and it is beyond this chapter’s scope to treat it inclusively. Instead, this chapter adopts a heuristic method by entering into dialog with one wellrespected recent work, David Kilcullen’s The Accidental Guerrilla (2009). Because of the book’s broad ambitions, its influence on military readers, and Kilcullen’s own military experience, it offers a wide enough canvas to consider a range of key points about survey research. (My intent here is neither to praise nor critique the book on its own terms – a task for which I would be quite ill-qualified.) This chapter is divided into three sections. The first section presents what the body of existing survey research can tell us regarding a series of key topics in the ongoing discussion over CI. The first such topic is the profile that the United States, as a hegemonic actor, has with various publics in the world, above and beyond the setting of the war on terrorism. Further, how does the U.S. public itself view this profile? Is its willingness to change it less, the same, or greater than in the U.S. policy elite? The next topic is how publics in Muslim-majority countries react to U.S. strategy and tactics. Finally, attention is given to national surveys in Iraq, Afghanistan, and Pakistan and how each of these national publics has reacted to specific insurgency-related issues in their countries. The second section argues that survey research cannot be a form of military fieldwork, and tries to clarify what survey research can and cannot do. The third section discusses an appropriate role for survey research that is often underutilized: learning whether and how a national public distinguishes among different groups of militants and types of militancy, and thus providing tools to help disaggregate them. The arguments the chapter makes can be summarized as follows. Polling of national populations can establish – and has already established – many points of information that are central to developing strategies of counterterrorism or CI. Most broadly, terrorist methods are partly techniques to face the massive asymmetry of conventional military power held by the United States, and publics in countries around the world are far more conscious of, and sensitive to, this asymmetry than is sometimes assumed. Interestingly, in the American public, a majority would favor cutting the defense budget in a way that would somewhat lower the

13

Public Opinion Research and Evidence-Based Counterinsurgency

303

profile of U.S. military power in the world. A majority would also tend to rebalance defense spending away from massive conventional and nuclear capabilities and toward the types of capabilities that better assist CI. Polling of publics in Muslim countries shows that, though broad attitudes toward the U.S. have been widely negative in recent years, the al Qaeda-style interpretation of the U.S. as malevolently focused on repressing the Islamic world has not gained a majority, because many instead view the U.S. as simply a selfinterested great power in the classic mold. In recent years the image of al Qaeda in the Islamic world seems to have derived at least as much from attitudes toward the U.S. as from attitudes toward al Qaeda itself. Similarly, in places where the U.S. has an active military presence (for example, Iraq in 2005) militant groups that are also of foreign origin can succeed in looking relatively more “native” by contrast with the U.S. Polling can discern in some instances where a degree of public support for an armed group comes with “caveats” or conditions that may come into play later, for instance, with the departure of another force that the first armed group has been fighting, or the increased possibility of a different political option. On occasion these “caveats” can be spotted through polling before circumstances bring them more fully to light. When a province or minority group is in a conflict with the state, polling the population of the whole country about the grievances of the province or group is both viable and worthwhile. Polling can be used to test whether a specific political accommodation would be acceptable to the public at large. Where a government and an insurgency are in competition, not only through combat, but also through different systems of justice that vie in the same territory, it has been possible to measure through polling the degree of respect enjoyed by the government’s effort at a judicial system. However, it does not seem possible to measure the respect enjoyed by the insurgency’s judicial system or its degree of use by the population. For practices that are formally illegal in one judicial system, but tolerated in another, it is possible to measure how the public’s views vary by region, and hence which judicial system their attitudes are more in tune with. Turning to the limitations of survey research – where a national population has lived with civil conflict for decades through rising and falling levels of intensity, it may be impossible to measure public support for one side or the other. Because many (or even most) respondents have an experience and perspective that includes memories of multiple phases of conflict and different arrangements of allies and enemies, poll questions written by outsiders who are thinking only of the current phase may generate apparently contradictory results. The concept of “support” itself may need to be rethought in a context such as Afghanistan. There is a necessary distinction between military fieldwork that seeks intelligence regarding civilians’ attitudes and public opinion research. Surveying a polity and surveying a population at risk are not the same thing. With the first, the researcher

304

C. Ramsay

begins by accepting the study’s universe as an established social reality with a known time duration (for instance, the geographic boundaries of a country). With the second, the researcher’s own purpose is setting out the universe or object of study. Consequently, there are potential uses of survey techniques in conflict areas that should be avoided because they will be invalid. On the other hand, it is both worthwhile and promising to use public opinion research to learn whether a national population distinguishes between different violent militant groups, where there is a profusion of such groups (as in, for instance, Pakistan). Such research may help to disaggregate such groups, to lessen the tendency to lump them all together, and to offer the public information that is persuasive because it is the information the public actually needs about these groups. It has considerable concrete value for counterterrorism and CI practices, and it should be developed well above its current level. The body of this chapter takes an empirical approach and is riddled with examples from an array of past polls. In most cases, the use of examples is meant to show that a line of inquiry in survey research can work, because here is an instance where it did.

Issues in Counterinsurgency: What Existing Poll Data Can Tell Us Poll data provides a variety of information that can supplement evaluation research and – to a very limited degree – directly test the effects of CI activities. Similar to criminal justice and other social interventions, support for and continued effectiveness (or ineffectiveness) of interventions is often directly entangled in perceptions and beliefs of both targeted and observing populations, making poll data an important component in evaluation. Before exploring the difficulties of this in practice, I explore below numerous topic areas in which poll data has been and can be collected.

The Hegemon as a Rational Threat A primary issue for the United States when it conducts CI (and for its allies who fight alongside it) is the sheer magnitude of the U.S. as a military force relative to others. The U.S.’s bulk can signal its strategic movements in advance and make surprise tactics difficult to execute. More broadly, its potential to overpower others creates “a fundamental mismatch (or asymmetry) between U.S. military capabilities and those of the rest of the world,” in Kilcullen’s words. He argues that “Unless other countries can be assured of America’s benign intent, they must rationally

13

Public Opinion Research and Evidence-Based Counterinsurgency

305

treat the United States as a potential threat.”1 This raises the question of whether regarding the U.S. as a potential threat is an attitude limited to professionals and policy elites, or whether it is also a general attitude shared by national publics. The data suggests that national publics are also sensitive to the “massive asymmetry” posed by the scale of U.S. military power. Over April–July 2009, WorldPublicOpinion.org asked publics in 20 countries worldwide: “Please tell me if you think the United States does or does not use the threat of military force to gain advantages” (WPO, 2009). Majorities in every country said the U.S. does use the threat of military force to gain advantages, the lowest being 61% each in Poland and India – two countries where the public has been distinctly favorable to the U.S. in recent years. The highest was the close ally South Korea, where 92% said the U.S. uses threats in this fashion. The publics of traditional NATO allies concurred (France 72%, Germany 66%, and Great Britain 83%). A more pointed test is offered by a Pew Global Attitudes question, which regrettably has been asked only in Muslim countries. At intervals from 2003 to 2007, Pew has asked: “How worried are you, if at all, that the U.S. could become a military threat to our country someday? Are you very worried, somewhat worried, not too worried, or not at all worried?” In 2007, countries that were very unlikely to become the focus of U.S. military attention – Morocco, Malaysia, Indonesia, and Bangladesh – nonetheless had strikingly high numbers saying they were “very worried” that the U.S. could become a military threat: Morocco 89%, Indonesia 53%, Malaysia 37%, and Bangladesh 81%.

The Hegemon as Rational Threat: The U.S. Public’s View If the size of the hegemonic footprint is itself a problem for conducting CI, what are the prospects for altering this footprint? Kilcullen argues that there is both a “mismatch between military and nonmilitary elements of U.S. national power” and another “substantial mismatch between the capabilities needed for the current international

1

Kilcullen, p. 23. Here is the full passage: The second implication of this massive asymmetry is that because its military superiority gives the United States the capability to destroy any other nation-state on the face of the earth, belief in the fundamentally benign intent of the United States becomes a critical factor in other countries’ strategic calculus. Intelligence threat assessments typically examine the twin factors of capability and intent, focusing on capability because intent is subject to much more rapid and unpredictable change. But the destructive capability of the United States is so asymmetrically huge vis-à-vis every other nation on Earth that it poses what international relations theorists call a “security dilemma.” Unless other countries can be assured of America’s benign intent, they must rationally treat the United States as a potential threat and take steps to balance and contain American power or defend themselves against it. And efforts to improve U.S. military capacity, which American leaders may see as defensive, may therefore have a negative overall effect on U.S. national security because of the responses they generate.

306

C. Ramsay

security environment and those actually present in the U.S. military inventory.” However, he thinks these mismatches are so deeply rooted as to be structural, since the defense sector of the U.S. economy cannot be sustained by supplying less costly “capabilities for irregular or unconventional conflict” (Kilcullen, 2009, pp. 26–27).2 Whether or not this is true, it is worthwhile to ask whether there would be deeprooted public opposition to a transition in which the nonmilitary elements increased, the military elements decreased, and – within the military elements – irregular warfare capabilities increased. This issue has been studied in a series of polls that presented federal and defense budget information to the American public, conducted by the Program on International Policy Attitudes (PIPA), University of Maryland, together with Knowledge Networks in 1996, 2000, 2005, and 2010. When shown the proportions of spending on various areas in the discretionary budget that Congress appropriates each year, majorities cut defense spending by a quarter or more on average, in three different exercises (1996, 2000, and 2005), and by 18% in the 2010 exercise. In 1996, 80% of respondents made some reduction to defense spending (Kull & Destler, 1999, p. 185); in 2000 it was 68% (Kull, 2000); in 2005, 65% (Kull, Ramsay, Subias, Weber, & Lewis, 2005, pp. 7–13); and in 2010, 64% (Kull, Ramsay, & Lewis, 2011, pp. 6–12). It should be noted that the typical poll question on this topic, which asks only whether defense spending should be cut, increased, or kept about the same, does not show majorities wanting to cut defense.3 In a frequently asked Gallup question, from 2002 to 2010 those saying the amount spent on defense is “too little” have remained in a range of 20–33%, while those saying it is “too much” have remained in a range of 17–44%. It is only when respondents are exposed to information about the budget as a whole that most conclude they want to cut defense spending. However, as the federal budget deficit becomes an increasingly compelling issue for policymakers, the spotlight on the topic will tend to bring to light the size of the defense budget relative to the rest of discretionary spending. In the 2005 PIPA/Knowledge Networks study (Kull et al., 2005), we sought to understand how respondents – most of whom had made deep cuts to defense within the

2

Kilcullen’s full argument runs as follows: “Because capabilities for irregular or unconventional conflict are much cheaper to acquire than those for conventional conflict, and require less hardware and industrial capacity, they are paradoxically less likely to be developed. This is because, through the military – industrial complex,’ a substantial portion of the American economy, and numerous jobs in almost every congressional district, are linked to the production of conventional war-fighting capacity. It takes factories, jobs, and industrial facilities to build battleships and bombers, but aid workers, linguists and Special Forces operators are vastly cheaper and do not require the same industrial base. So shifting spending priorities onto currently unconventional forms of warfare would cost jobs and votes in the congressional districts of the very people who control that spending. This makes it structurally difficult…Hence…the pattern of asymmetric warfare, with the United States adopting a basically conventional approach but being opposed by enemies who seek to sidestep American conventional power, is likely to be a long-standing trend” (pp. 26–27). 3 A table showing this Gallup question back to 1987 is available at http://www.pollingreport.com/ defense.htm. Alternately, see the iPoll Databank at the Roper Center, University of Connecticut, http://www.ropercenter.uconn.edu/data_access/ipoll/ipoll.html.

13

Public Opinion Research and Evidence-Based Counterinsurgency

307

exercise – would allocate cuts within the structure of the U.S. military. After doing the general budget exercise discussed above, respondents were shown a wide range of components of the defense budget, divided into (1) current kinds of personnel and hardware; (2) current capabilities; and (3) investment toward future capabilities and hardware. Those respondents who had cut the defense budget in the general exercise (65%) were offered a scale of 0 to −4 for each item, with 0 meaning “not at all” and −4 meaning “reduce a lot.” Thus this group was required to specify real locations for the reductions they had made in the general exercise. Respondents who had maintained (15%) or increased (16%) the defense budget in the general exercise were offered a scale of +4 to −4 for each item, allowing them to increase, maintain or reduce. Respondents’ answers were very clear: they intended to make cuts in high-profile nuclear and classic conventional capabilities. Majorities of 55–65% cut the size of the nuclear arsenal, development of new types of nuclear weapons, large land war capabilities, and new types of destroyers. Smaller majorities (51–53%) cut both development and overall inventory for bombers, submarines, and destroyers. Among the capabilities that majorities protected were urban fighting capabilities, CI, peacekeeping capabilities, troops for special operations, intelligence, and new equipment meant for direct use by infantry and Marines. Thus it appears that the public as a whole is not so attached to the general armory of the superpower, nuclear and conventional, and is more concerned about capabilities relevant to the conflicts in which the U.S. now finds itself. Polling in depth in this way on the public’s preferences (i.e., the taxpayers’ preferences) regarding the U.S. defense profile is important for evidence-based CI, because the insurgent has the goal of wearying its adversary’s home public. Polling can establish whether the home public is wearying specifically of a given CI effort, or whether it is reacting to the size of the overall burden on it imposed by the U.S.’s defense profile in general.

Publics in Majority-Muslim Countries: Assessing the Degree of Alienation from the Hegemon Thus, there is evidence that the U.S. public would be willing to lower the visible power profile of the United States, shrinking the defense establishment while keeping or growing the capabilities that CI and CT need. If this took place, would it make any difference to public attitudes in the Muslim world? Are these publics too anti-American to care if the U.S. handles its security agenda differently? Or are they (in a sense) engaged in an ongoing argument with the U.S., in which changes in the U.S. pattern of actions could matter? Kilcullen speaks of “a population across large parts of the Muslim world…that has been alienated by Western actions in Iraq, Afghanistan and elsewhere and thus provides a recruiting base for guerrilla-style terrorism and a receptive audience for AQ [al-Qaeda] propaganda” (Kilcullen, 2009, p. 33). The term “alienated” that Kilcullen uses suggests a very broad disenchantment. If these populations are truly alienated, they would view the motives of the United States as always suspect, apart from whether

308

C. Ramsay

Table 13.1 Muslim publics’ views of U.S. support for democracy in Muslim countries: Which of these three positions is closest to yours? B. The U.S. favors C. The U.S. favors A. The U.S. democracy in Muslim democracy in Muslim opposes countries, but only countries, whether democracy in if the government is or not the government is Muslim countries cooperative with the U.S. cooperative with the U.S. DK/NS Egypt 37 42 8 14 Jordan 41 40 6 12 Palest. Ter. 35 38 11 15 Turkey 30 49 7 14 Indonesia 17 44 13 27 Pakistan 25 36 10 30 Source: WorldPublicOpinion.org (2008)

specific U.S. actions seem good or bad. This would go beyond opposing the U.S. military’s actions in Iraq and Afghanistan. Instead, it would mean that, U.S. changes in its use of visible power abroad, U.S. development and humanitarian assistance delivered by civilians, or U.S. pressures to promote democratic reforms would be seen as part of a larger negative pattern that is meant to deceive Muslim nations. Indeed, checking for such truly alienated attitudes is a key consideration in evaluating CI effectiveness, because where these attitudes are strong CI strategies will face serious obstacles. Two questions that were asked in late 2008 by WPO in Egypt, Jordan, the Palestinian Territories, Turkey, Pakistan, and Indonesia can help us to consider this. Respondents were asked which came closest to their view: that the U.S. opposes democracy in Muslim countries; that it favors democracy in Muslim countries, but only if the government is cooperative with the U.S.; or that it favors democracy in Muslim countries, whether or not the government is cooperative with the U.S. In these six nations, the perception that the U.S. unconditionally supports democracy in Muslim countries never rose above 13% (Table 13.1). In Egypt, Jordan, and the Palestinian Territories, publics were roughly divided between the most negative view – that the U.S. opposes democracy – and the view of the U.S. as a traditional great power, demanding cooperation and treating democracy as secondary. In Turkey, Pakistan and Indonesia, respondents tended more toward the great-power image of the U.S. and less toward seeing the U.S. as an opponent of democracy. Another question asked about U.S. respect for Muslim countries, using a similar framework: whether the U.S. mostly shows respect to the Islamic world; whether it is often disrespectful toward it, but out of ignorance and insensitivity; or whether the U.S. purposely tries to humiliate the Islamic world. Among Egyptians, Palestinians, and Jordanians, those who thought the U.S. mostly shows respect to the Islamic world did not exceed 20% (Table 13.2). Among Egyptians a majority (56%) thought disrespect was intentional, with 24% taking the more charitable view. Palestinians were similar: 49% thought U.S. disrespect was intentional and a lesser 28% thought it was out of ignorance. Among Jordanians these views were roughly equal (39% intentional, 34% from ignorance). Turks were

13

Public Opinion Research and Evidence-Based Counterinsurgency

309

Table 13.2 Muslim publics’ views on whether the U.S. shows respect to the Islamic world: Which of these three views is closest to yours? B. The U.S. is often A. The U.S. mostly disrespectful to the C. The U.S. purposely shows respect to the Islamic world, but out tries to humiliate the Islamic world of ignorance and insensitivity Islamic world DK/NS Egypt 11 24 56 10 Jordan 16 34 39 12 Palest. Ter. 20 28 49 3 Turkey 8 40 43 9 Indonesia 8 39 30 24 Pakistan 6 22 52 20 Source: WorldPublicOpinion.org (2008)

similarly divided (43% intentional, 40% from ignorance). Outside the Middle East, Pakistanis’ views were similar to Egyptians’ (55% intentional, 34% from ignorance), but Indonesians took the most charitable view among these six publics (23% intentional, 41% from ignorance, and 29% mostly shows respect; Kull, 2011). In this array of results there is a simple underlying pattern. None of these mostly Muslim publics comes close to taking the United States at its own estimation. Instead, there are signs of a debate: whether the U.S. is best understood as truly malevolent toward Muslim countries, or whether it is best understood as a somewhat callous hegemon that functions without much self-awareness in a traditional great-power mold. The fully alienated viewpoint – that the U.S. is malevolent – appears to be at majority levels in Egypt and Pakistan, but not elsewhere. This suggests that Kilcullen’s assessment lacks an important nuance: apart from the alienated view, there is also the disabused view of the U.S. as no different from the European great powers who were full-scale colonialists in the region in the past. For the U.S., this is not pleasant, but it is a far cry from al Qaeda’s analysis of the U.S., and it is a better starting point for eventually changing minds. There is a possible gain for the U.S. from acting in ways that differ from a great power’s regular calculus. Kilcullen also speaks of “…a syndrome that is easily summed up, though extremely hard to counter: AQ moves into remote areas, creates alliances with local traditional communities, exports violence that prompts a Western intervention, and then exploits the backlash against that intervention…Al Qa’ida’s ideology tends to lack intrinsic appeal for traditional societies, and so it draws the majority of its strength from this backlash…” (Kilcullen, 2009, p. 34). Kilcullen’s point raises the possibility that the same thing has happened on an international scale: that in Muslim countries there is some tolerance or passive acceptance for al Qaeda, not because of support for al Qaeda’s notions about governance or its blueprint for the Ummah’s future, but because al Qaeda positions itself as a challenge to Western expeditionary forces and the U.S. bases in the region. (After all, anti-American attitudes in the region did not begin with al Qaeda; rather, al Qaeda is a relatively new entrepreneur, seeking to exploit reserves of anti-Americanism that were once associated primarily with leftist, secular resistance movements.) There is some evidence to support this theory in questions asked for The National Consortium for the Study of Terrorism and Responses to Terrorism (START) by

310

C. Ramsay

WPO in several Muslim countries in 2008 (Kull, Ramsay, Weber, Lewis, & Mohseni, 2009). Respondents heard a variety of goals espoused by al Qaeda and were asked, for each, whether they thought this was in fact a goal of al Qaeda, and whether they agreed with it or not. To quote the report: A primary al Qaeda goal related to U.S. behavior is “to push the U.S. to remove its bases and its military forces from all Islamic countries.” This goal was endorsed by large majorities (Egypt 87%, Indonesia 64%, Pakistan 60%), and rejected by no more than 16% anywhere. In late 2006, 72% of Moroccans also endorsed this goal. Majorities also affirmed that this was probably an al Qaeda goal (Egypt 71%, Indonesia 61%, Pakistan 54%; Morocco 2006, 78%). There were no significant changes from 2007 (p. 20).

One would assume that these strong attitudes about the U.S.’s military profile would carry over to the U.S.’s other forms of influence, but this was not the case. Respondents were less sure that U.S. aid to the Middle East’s quasi-authoritarian governments was a completely nefarious thing – and also less sure that al Qaeda wanted to see an end to this aid: The only goal to not receive strong support was “to push the U.S. to stop providing support to such governments as Egypt, Saudi Arabia, and Jordan.” (In Egypt, only Saudi Arabia and Jordan were mentioned in the question.) Responses to this question have also changed substantially since 2007. Among Egyptians, 56% said they agreed with this goal – up 15 points from 2007 – while 34% disagreed. In Indonesia and Pakistan, only pluralities approved (46% in each country). Since 2007 support has risen in Pakistan by six points, but has fallen in Indonesia by eight points. In late 2006 a modest Moroccan plurality agreed, 42–36%. This was also the only goal that had less than a majority affirming that it indeed was an al Qaeda goal, and responses have been quite unstable relative to 2007 (p. 21).

Thus we see publics wanting a lower U.S. military presence in the region, and identifying this as an al Qaeda goal, but much less sure that they want a drop in U.S. aid, and reluctant to identify the end of U.S. aid as an al Qaeda goal. This suggests that in 2008 at least, the image of al Qaeda was shaped at least as much by attitudes toward the U.S. as it was by attitudes toward, or knowledge of, al Qaeda itself. For Kilcullen, the identity of an “outsider” is not absolute but relative – it is the most foreign actor, compared to others, that gets the outsider stigma. “Although the terrorists may have been seen as outsiders [to a local community]…as soon as foreigners or infidels appear in the area, by comparison the terrorists are able to paint themselves as relative locals”4 (Kilcullen, 2009, pp. 37–38). Thus U.S. military

4

Here is his full argument: “I have noted that during the initial stage of development of an extremist presence, there is usually a local opposition to the terrorist group (albeit often cowed, impotent or intimidated). But during the intervention phase, the entire local dynamic shifts. The presence of the intervening outsiders causes local groups to coalesce in a fusion response, closing ranks against the external threat… A high-profile, violent, or foreign-based intervention tends to increase support for the takfiri terrorists, who can paint themselves as defenders of local people against external influence… Although the terrorists may have been seen as outsiders until this point, their identity as such has been not fixed but “contingent”: as soon as foreigners or infidels appear in the area, by comparison the terrorists are able to paint themselves as relative locals…. The completely understandable (and necessary) imperative for the international community to intervene and prevent extremist contagion can thus act as a provocation, causing the next stage in the process: rejection” (Kilcullen, pp. 37–38).

13

Public Opinion Research and Evidence-Based Counterinsurgency

311

presence can function as a foil against which jihadist militants from other countries seem much less exotic. Although it is hard to know from existing data how widespread this phenomenon has been, Pakistan in 2008 offers a salient case. Pakistanis were asked to rate a wide range of countries and groups as to whether each was a critical threat to the country’s vital interests; an important threat, but not critical; or not an important threat. “The U.S. military presence in Afghanistan” was called a critical threat by 68%, and “the U.S. military presence in Asia” was called a critical threat by 72%. By contrast, the “activities of al Qaeda” were called a critical threat by only 41% – though this was more than the 34% who called the “activities of Islamist militants and local Taliban in the FATA [Federally Administered Tribal Areas] and settled areas” a critical threat (Fair, Ramsay, & Kull, 2008, p. 24). Thus the U.S. was regarded as most dangerous, and al Qaeda as less dangerous than the U.S. but more dangerous than local militant organizations – an ordering by cultural distance that follows Kilcullen’s notion quite well. A less direct test is offered by a question that asked people in the Middle East what they thought the tradeoff was between the U.S.’s capacity to buttress stability and its potential to be a provocation, possibly attracting extremist militants. In a BBC World Service poll (2007) respondents were asked, “Do you think the U.S. military presence in the Middle East is a stabilizing force or provokes more conflict than it prevents?” Large majorities in all four Middle Eastern countries polled said it provokes more conflict than it prevents, including people in Egypt (85% ), Turkey (76% ), Lebanon (77%), and even the UAE (66%). This line of questioning would be an excellent one for future research that would ask about different types of conflict or unrest, internal and external, and what influence U.S. military presence is perceived to have on them.

Testing the Depth of Support for a Nonstate Actor Kilcullen makes an important point about the diversity of violent nonstate actors that also applies to how publics view them. He says: “We need to…distinguish between enemies. Western countries face an extremely diverse threat picture, with multiple adversaries who oppose each other’s interests as well as those of the West” (Kilcullen, 2009, pp. 284–285). The test here is whether there is evidence for an underlying conditionality below apparent public support for militant groups and activities. The end of the Second Intifada, which led to Israel ending its direct military occupation of Gaza, provides one example. In December 2004 – before Israel pulled its military and settler presence out of Gaza, but after the start of a calming-down period – 66% of Gaza’s residents opposed the idea of “disarming the various militias (Al Aqsa, Al Qassam, etc.)” in a Birzeit University poll; 31% supported the idea (Birzeit University, 2004). But after Israel completed its disengagement in September, an October poll showed distinctly less willingness among Gazans to

312

C. Ramsay

give the militias leeway. Now a 50% plurality supported disarming them; 82% wanted their parades in residential areas to stop; and almost everyone (95%) wanted an end to keeping weapon caches in residential areas (Birzeit University, 2005). Clearly Gazans’ support for the militants was conditional in great part on Israeli power being on the ground in Gaza. Turning to Iraq in September 2006: though the strength of various militias at that time showed they must have had bases of support – and a strong government appeared to be only an aspirational goal – the Iraqi public overall much preferred government-assured security (WPO, 2006). Asked “Would you prefer to have a strong government that would get rid of all militias or do you think it would be better to continue to have militias to protect your security?” 77% preferred strong government forces, including majorities of Sunnis, Shias, and Kurds. Only 21% disagreed, with the highest level among Shia (33%). In the same poll only 7% expressed a favorable view of al Qaeda (among Sunnis, no more than 23%) or of bin Laden (8%; 29% among Sunnis). But all these attitudes were compatible with vehement hostility to the U.S. military. Sixty-one percent approved of “attacks on U.S.led forces in Iraq,” up from 47% in January 2006. Anyone who made the assumption that an enemy of U.S. forces was a friend to al Qaeda would have misunderstood the situation. Much of the success of the surge in the next year, which involved Sunni tribes in the west concluding that al Qaeda was more noxious to them than the U.S. military, can be attributed to CI practitioners’ correct assessment in this regard. Both cases illustrate that polling can detect underlying caveats that people may put on their apparently strong support for a militant group – and when circumstances change, these caveats may come into play and lower the public’s level of acceptance of such a group.

Pakistan: A Modern Government for the FATA Another potential use for polling is in checking with the whole population whether or not it would be acceptable to satisfy a grievance or institute a reform in favor of a local group that is in an embattled relationship with the central government. A part of the population may be at loggerheads with the state, but this does not establish whether it is at loggerheads with the society. Kilcullen raises an example from Pakistan: A few months after the January 2006 strike, I spent several hours in conversation with a local politician from the Damadola area [in the FATA], associated with the Jema’ah Islamiya (JI), a pro-Taliban, antigovernment, Deobandi Islamist political party. …This local leader vociferously denied any possible justification for the government attacks on the village, and rejected the implicit paternalism…which he saw as inherent in the traditional government approach to the FATA: “…All of the bodies were of innocent local people, there were no al Qa’ida. The people don’t want to be ruled under the old system by the maliks. Rather they want an elected legislature at the FATA level. FATA is the fifth unit of Pakistan…and the others are all governed by elected democratic representatives: FATA should be too” (Kilcullen, 2009, p. 231).

13

Public Opinion Research and Evidence-Based Counterinsurgency

313

For the purposes of CI, this raises an excellent question: how would the Pakistani public view the proposition of this JI politician? Would this be an acceptable concession? In 2007 WPO raised the issue in a poll of Pakistan’s national urban population (Fair et al., 2008, pp. 15–16). In fact, a large majority favored the phasing out of the FATA’s special legal status (the Frontier Crimes Regulation) and its integration into Pakistan’s legal structure. Respondents were offered three statements about the FATA and asked which came closer to their own views. The least popular statement – chosen by only 8% – was that “the Frontier Crimes Regulation should be left unchanged.” Instead, 72% agreed that these regulations should be modified so that people in the FATA “have the same rights and responsibilities as all other Pakistanis.” Twenty-six percent thought that “the Frontier Crimes Regulation should be abolished,” while another 46% said these regulations should be “modified slowly over time.” Thus although there is wide majority support for changing the FATA legal system, which was codified by the British in 1901, a plurality favors a gradualist approach. Respondents were also asked about the use of military force in FATA by the government. A 58% majority did not think the government should use force to try to make its writ run in FATA; 46% preferred the statement that “the government should not try to exert control over FATA but should try to keep the peace through negotiating deals with local Taliban,” and another 12% thought “the government should withdraw its forces from FATA and leave the people alone.” About a quarter (23%) did believe “Pakistan’s government should exert control over FATA, even if it means using military force to do so.” These two findings show up a paradox that is important for developing an evidence-based CI strategy and could be easy to miss. A majority was in favor of integrating FATA into Pakistan through a phased process to end its unique, and literally “postcolonial,” status. At the same time a majority did not support trying to enforce the type of nominal order that exists in FATA now. Even among those who thought the Frontier Crimes Regulation should be left unchanged, a majority (54% of this group) thought the government should not use force to exert control over FATA. This suggests a CI strategy for Pakistan in changing the center’s relationship with FATA: bet on its integration and capacity for self-government by making it truly part of Pakistan. Polling of both the whole nation, and of FATA, on the issues this process would raise could be valuable in pursuing such a course.

Afghans’ Views of Courts and Judges One of the most difficult problems for the use of polling is probing people’s attitudes toward two competing systems of governance that are in conflict, and that may be both present in a given territory. Afghanistan presents a strong form of this problem. The Kabul government and the Taliban both attempt to run systems of courts, judges, and dispute resolution and they overlap across much of the country.

314

C. Ramsay

Can polls be used to evaluate the Afghan public’s reaction, living as many of them do in the middle of this competition? Kilcullen provides the following account: By mid-2008, the Taliban were operating 13 guerrilla law courts throughout the southern part of Afghanistan—a shadow judiciary that expanded Taliban influence by settling disagreements, hearing civil and criminal matters, and using the provisions of Islamic shari’a law and their own Pashtun code to handle everything from land disputes to capital crimes. Local communities often cry out for external mediation in settling local disputes… Meanwhile the international community is training Supreme Court judges and seeking to build an Afghan legal system based on the post-2001 constitution, but local judges, prosecutors, and police are often known for their love of bribes, and locals see them as giving phony “justice” to whoever can pay most handsomely (Kilcullen, 2009, p. 47).

Surveys of the Afghan public do provide measures of attitudes toward the government’s judicial system that can be usefully compared to the assessments of observers. In The Asia Foundation’s (2010) survey, respondents were divided over whether it was true that “State courts are not corrupt compared to others,” with 49% saying yes and 48% no; there was significant regional variation, running from 60% saying yes in the northwest to only 31% saying this in the southwest (The Asia Foundation, 2010, pp. 130–132). A 55% majority disagreed that “State courts resolve cases timely and promptly” (yes, 42%). Further (to quote the Asia Foundation report), “the majority of respondents that had contact with the state courts in the last year encountered some level of corruption.” The November 2010 survey of Afghans conducted for ABC, BBC, and ARD news organizations asked a startlingly simple question: “Would you say a system of rules and laws that reflects what most people in Afghanistan want exists in this area at this time, or not?” (ABC News, BBC, & ARD, 2010, p. 47). A modest majority (51%) said no, while 40% said yes. Of those who thought such a system did not exist in their area, three quarters (38% of the full sample) said there was no movement toward having one in the future. Although this survey did not ask questions about the courts, it did ask broader questions about corruption. Eighty-five percent described “the issue of corruption among government officials or the police in this area” as a big (50%) or moderate (35%) problem (p. 29). Views were very similar when respondents were asked in turn about “corruption within the government of this province” – which 93% called a big (61%) or moderate (31%) problem – and “corruption at the national level within the government in Kabul,” which 88% called a big (65%) or moderate (22%) problem (p. 30). When asked, “Has it ever happened to you that a representative of the provincial government has asked for money or other payment in exchange for favorable treatment in the performance of his official duties?” 27% of the whole sample said yes; regarding the Afghan National Police, 21% said yes (p. 27). Thus, taking all these findings together: three in ten Afghans recalled paying bribes; three in five regarded corruption as a big problem in the Afghan government, and half thought this was true of their own area; about half saw the state courts as corrupt, and a majority of those who dealt with them found that they were corrupt. This general picture is confirmed by an extensive national survey that focused only on the issue of corruption (Integrity Watch Afghanistan, 2010). Of those respondents who said they had access to the state courts, “25% of these said they felt deprived of

13

Public Opinion Research and Evidence-Based Counterinsurgency

315

justice because of corruption within the courts; 33% said corruption within the courts had a negative impact on their households” (p. 74). However, The Asia Foundation’s poll is also useful for showing what surveys cannot do. Respondents who said they had taken a dispute to some kind of outside authority for resolution were asked who they approached to solve the problem and were invited to give up to three responses (The Asia Foundation, 2010, pp. 123–130). Yet, of this part of the sample, only 1% said they had taken a dispute to the Taliban – an implausibly low level, especially given the low confidence shown that government authorities would deal fairly. Similarly, respondents who had been victims of violence or crime were asked if they had reported the crime to any authority. In 2010, 15% said they did not know whether they had reported the crime or not – an extremely implausible nonresponse which has gone up over time, from 1% in 2007 to 15% in 2010. It is reasonable to guess that some in this group actually reported the crime to the Taliban. It appears that getting the public’s ratings of a judicial system run by an insurgency is outside the capabilities of a survey. However, for practices that are formally illegal in one judicial system, but tolerated in the other, it is possible to measure how the public’s views vary by region, and hence which judicial system their attitudes are more in tune with. Kilcullen expresses hopes that widespread Afghan disapproval of poppy cultivation will provide leverage against the Taliban: By linking themselves so closely to opium, the Taliban have created a vulnerability we can exploit… The Taliban have tied themselves to the poppy, by means of both propaganda and intimidation of farmers; yet only a minority of Afghans consider poppy cultivation acceptable. This suggests that potential public support for CN could be substantial—if handled correctly, that is, with due attention to alternative livelihoods and support to ensure the continued economic wellbeing of the affected population (Kilcullen, 2009, p. 65).

The difficulty with Kilcullen’s characterization is that it refers to a national figure – but the 2010 ABC/BBC/ARD poll is adequate to show that this was not a national attitude in 2010. In that poll 63% nationwide said growing poppy is never acceptable – a level that has remained stable since 2007. In the seven provinces that grow the most poppy, a minority (42%) took this view; in Helmand it was 36% (ABC, BBC, & ARD, 2010, pp. 18–19). Thus the national level of disapproval is not homogeneous enough to put much leverage on regions that depend on poppy for their livelihood (a national sample of only 1,700 was sufficient to test this).

“Power” vs. “Congeniality” of Taliban Kilcullen makes valuable observations about the Afghan population’s relative acceptance of government authority or Taliban authority, that can also show us a limitation of polling – when it applies the concept of “support” in contexts where it may prove meaningless. Kilcullen writes (I have added emphases): There is…a belief, unfounded in reality, that development assistance generates gratitude, or “hope,” in the population and thereby of itself encourages them to support the government.

316

C. Ramsay Field experience in both Afghanistan and Iraq, however, has shown that insurgent intimidation easily overcomes any residual gratitude effect, while historical studies have shown that in civil wars and insurgencies, popular support tends to accrue to locally powerful actors rather than to those actors the population sees as more congenial: the more organized, locally present and better armed a group is, the more likely it is to be able to enforce a consistent system of rules and sanctions, giving the population the order and predictability it craves… (Kilcullen, 2009, pp. 67–68).

The data suggests that Kilcullen’s observations here are, on the whole, insightful. The problem is what is really meant by the term “popular support,” which is a notion that in Afghanistan breaks down under the weight of the data. There is little “support” for either an intransigent war to the finish against the Taliban, or for the Taliban returning to power. “Support” is a term that implies a person feels able to choose a preferred outcome, and then to behave in ways that might help to make that outcome more probable. Afghans’ responses to 2010 poll questions do not accommodate “support” well, but seem to come instead from a pragmatic sense of survival in the midst of uncontrollable forces. First, reports of Taliban activity by respondents point to much more Taliban sway than do respondents’ own assessments of whether or not the Taliban is strong or weak in respondents’ areas. Forty percent said in the ABC/BBC/ARD poll that they were aware of the Taliban “fighting government or foreign troops” in their own areas, and 24% said they were aware of “people giving food or money” to the Taliban (presumably a more discreet activity, unless Taliban members extracted supplies publicly; p. 41). Yet when asked “how much of a presence it has in this area,” only 15% said the Taliban’s local presence was fairly or very strong, and 55% said it had “none” (p. 35). When asked about the Taliban’s “level of support among the people of this area,” only 11% said the Taliban has fairly or very strong support, while 65% said it has no support at all (p. 39). These assessments may arise from wishful thinking or an interviewer effect, but they fit poorly with respondents’ reports of Taliban activity. ABC/BBC/ARD asked respondents whether they themselves supported the presence in Afghanistan of either “jihadi fighters from other countries” or “fighters from the Taliban” (pp. 30–31). Peculiarly, 17% said they supported the presence of foreign jihadis, but only 11% said they supported the presence of fighters from the Taliban – as if they felt more comfortable answering the first question than the second. This suggests that the first question showing 17% support may be the better of the two indications. Otherwise we would have to believe that a militant group that is mostly Afghan is less popular than a comparable group made up of foreigners. When NATO air strikes kill Afghan civilians through error or as collateral damage, the Afghan public tends to spread the blame among all the combatants. In 2010, 35% blamed US and NATO/ISAF forces for mistaken targeting, 32% blamed “antigovernment forces for being among civilians,” and another 32% blamed both sides equally (ABC/BBC/ARD, p. 43; 2009 results were very similar). This is the response of a public that, overall, is reluctant to pick a side in the conflict. There is some evidence that the increase in terrorist methods by the Taliban over 2010 – in both quantity and intensity – brought a response of revulsion among many Afghans. The Asia Foundation poll asked in 2009 and 2010, “Thinking about the

13

Public Opinion Research and Evidence-Based Counterinsurgency

317

reasons the armed opposition used violence during the past year, would you say that you in general have a lot of sympathy, a little sympathy, or no sympathy at all for these armed opposition groups?” (The Asia Foundation, 2010, pp. 49–52). In 2009 a 56% majority had at least a little sympathy with 22% expressing a lot of sympathy; only 36% said they had no sympathy at all. In 2010, though, a 55% majority now had no sympathy at all – up 19 points; 40% had a little (26%) or a lot (14%) of sympathy. However, there is no relationship between this apparent drop in sympathy and any desire to fight a war against the Taliban to a complete conclusion. On the contrary, openness to attempting a negotiated solution rose between 2009 and 2010. The Asia Foundation asked whether people agreed or disagreed with “the Government’s reconciliation efforts and negotiations with the armed opposition”: agreement had risen from 71 to 83%, but strong agreement had risen from 28 to 42% (pp. 45–48). And, agreement in 2010 was very similar across regions, with those strongly agreeing with negotiations varying only within a 12-point range (36–48%). Similarly, in ABC/BBC/ARD’s poll 73% thought “the government in Kabul should negotiate a settlement with Afghan Taliban in which they are allowed to hold political offices if they agree to stop fighting” (up from 65% in late 2009; ABC/ BBC/ARD, 2010, p. 33). A majority drew the line at the idea of dividing control geographically. When asked “What if an agreement to stop the fighting ceded control over certain provinces to the Taliban?” 61% were unwilling to go this far; only 37% were willing. In sum, the Afghan public as a whole seemed reluctant to “add up” what they knew about the Taliban’s presence and activities and to accept its actual scope. The minority that was positive about the presence of foreign “jihadi fighters” was, implausibly, larger than those who said the same of the Taliban. The increasing tempo of a Taliban style of violence that was careless of civilians – if not clearly directed against them – was greatly disliked, but this did not impede a majority desire to attempt a negotiated solution that would share power and keep the country together as a geographic unit. All this does not translate into support for either the government or the Taliban. Rather, it describes a view that believes it will be necessary to come to terms with the Taliban and treat it as a fact of life. This brings into question the use of polling to measure “support” in settings where civilians have lengthy experience of living under civil conflicts. Such a use of polling is an increasingly common practice in CI. The Canadian practitioners quoted at the beginning of this article write that “in assessing the Canadian campaign in Kandahar Province, polling has gone from being an obscure and easily disregarded side show to being sometimes over-emphasized in important operational assessment reports…Mission assessments should always seek the development of other data collection tools” (Vincent et al., p. 125). In circumstances where a national population has lived with civil conflict for decades through rising and falling levels of intensity, it may not be possible to use surveys to measure public support for one side or the other. Because many (or even most) respondents have an experience and perspective that includes memories of multiple phases of conflict and different arrangements of allies and enemies, poll questions written by outsiders who are thinking only of the current phase may generate apparently contradictory results.

318

C. Ramsay

Where Polling Cannot Go There is a necessary distinction between military fieldwork that seeks intelligence regarding civilians’ attitudes and public opinion research. Attempting to survey a population at risk and attempting to survey a polity are not the same thing. In the first case, the researcher’s own purpose outlines an object of study. In the second case, the researcher begins by accepting the baseline of a larger social reality with a longer time duration. A country or province with preexisting boundaries has its own historical, political, cultural, and economic origins that are not interlocked with the assumptions the researcher is making. It is always difficult, and frequently impossible, to use survey research to measure the effects of a particular intervention – and this is just as true in a highly developed society in peacetime. To take a common example, American survey research firms are frequently asked by clients to survey the public before and after a large and expensive public education campaign on some issue. The client assumes that such a campaign, though it directly engages only a small part of the population, will leave a measurable trace. It is possible, though rarely worthwhile, to ask a sample whether they have heard anything about the campaign, and then compare that group’s attitudes to the rest of the sample. But in practice, the kind of measurable change that the client seeks in society comes only in response to major events that enter collective experience, become a talked-about topic, and then inflect some prior attitude that was held by a large number of people. A good example of such an event is Pakistan’s Swat Valley crisis of spring 2009 (discussed below). Nothing less is likely to move the needle in national surveys. In a CI conflict situation, direct effects and side effects of two pervasive strategies – prosecuted by the insurgent and the counterinsurgent – form a mixture to which the public is always responding. Surveys can tell something valuable about the public’s views of overall strategies; they can generally tell little about public reactions to discrete interventions. Kilcullen is a practitioner of “conflict ethnography” or alternately “war zone fieldwork” (he offers both terms). It is only fair to begin with the caveats he makes himself about his approach. The risk, stress and effort inherent in war zone fieldwork also clouds judgment and skews emphasis; researchers tend to place more weight on data we collect with difficulty and danger than on insights gathered in the comfort of a library or archive. This research approach’s deep regional or district focus…is also not necessarily transferable (Kilcullen, 2009, p. 306).

Thus Kilcullen is aware of, and points out, the tendency to prioritize zones of conflict and denied areas over zones that are “safe” or “won” – a tendency that is not holistic, and also prone to temporal shifts in war and politics. Interestingly, the Canadian practitioners in Kandahar Province went through a phase of trying to use their polls to distinguish attitudes in areas inside and outside their own side’s control: Early Kandahar polls attempted to isolate polling results from the area deemed to be the coalition military footprint in the province. One approach to this was taken by reporting results for districts deemed to be in the footprint separately from the others. This was however thought

13

Public Opinion Research and Evidence-Based Counterinsurgency

319

to be too coarse as the actual military footprint did not follow district borders. Another attempt was based on the inclusion of each sampled settlement as either within or outside the footprint, but the poll’s design was not meant for such analysis (Vincent et al., p. 123).

They concluded that this approach was futile and could not explain their polls’ results. Kilcullen summarizes his methodology in a series of points that include the following: – Use documentary sources (including operational and intelligence reports, captured documents, quantitative data, maps and surveys, media content analysis, and the work of other researchers) to create a primary analysis of the environment. – Use this primary analysis to identify a more limited number of “communities” (local areas, population groups, villages, or functional categories) for further detailed personal analysis at the case-study level. – Conduct firsthand, on-the-spot field studies (applying an extended residential fieldwork approach wherever possible) of these secondary communities. – Work from unstructured, face-to-face, open-ended interviews (rather than impersonal questionnaires and surveys) during field work, but integrate this subjective qualitative perspective with quantitative data from the primary analysis (Kilcullen, 2009, pp. 304–305).

The “quantitative data” which is among the “documentary sources” that Kilcullen starts with is sure to include opinion surveys of the general population, if available. And the unstructured interviews – free from the “impersonal” quality of the questionnaire (and from its consistency as well) – are then integrated with “quantitative data from the primary analysis,” including survey data. However, when these two conflict, he admits elsewhere that the greater weight will likely be placed “on data we collect with difficulty and danger” – but apparently when the survey and the fieldwork are in harmony, the survey’s concordance will be welcome. The temptation of making a mishmash of two methods that are very different is suggested by Kilcullen’s borrowing of the term “public opinion” to coin the phrase, “tribal public opinion.” Here it is used to describe Pashtun society: Actions of leaders are sharply circumscribed by group consensus and judgments about what tribal public opinion will bear. In this sense, public opinion is the ultimate sanction, a potent force indeed in one of the most inhospitable regions on earth [i.e., Pashtun areas of Afghanistan] (Kilcullen, 2009, p. 78).

And here it crops up again in a discussion of Iraqi tribes: Key components of tribal-style reconciliation that we identified in Iraq in 2007 included… – The role of social pressure from tribal “public opinion” in enforcing compliance with agreements reached through this process—pressure that does not always succeed in enforcing agreements (Kilcullen, 2009, p. 170).

The concern here is not that Pashtun “tribal public opinion” and Iraqi “tribal public opinion” sound quite different in Kilcullen’s description, with the Pashtun version being the “ultimate sanction” and the Iraqi version just a source of pressure that often does not work. There is no prima facie reason to doubt that these two things are different. The concern is rather that perhaps they need to be understood and named as separate phenomena, instead of getting the same label and both sailing under the colors of “tribal public opinion.”

320

C. Ramsay

There is a misuse of the word “public,” and the idea of the public, that tends to crop up in Kilcullen’s conflict ethnographies. In reality, a public is the human totality of a social whole. Publics are not tied necessarily to nation-states: there are publics of units smaller, or units larger, than a nation. To conduct a survey, there must first be a preexisting way to operationalize this social whole: geographic boundaries, linguistic identities, and (with more difficulty) professions of religious faith are all examples. Otherwise, the researcher’s object of study has been constructed by the researcher – something that happens so often inadvertently that we must at least guard against doing it deliberately. Sometimes Kilcullen’s language betrays an obsession with the part that is incompatible with getting a grasp on the whole. …traditional intelligence services are primarily designed not to find out what is happening but to acquire secrets from other nation states. They are…less suited to nonstate actors--where the problem is to acquire information that is unclassified but may be located in denied, hostile, or inaccessible physical or human terrain. In the Pakistani case, public opinion in Minam Shah or Damadola may matter more to our chances of dealing with AQ senior leadership than any piece of state-based intelligence…Similarly, in Iraq, public perceptions of Coalition and enemy operations were extremely important data for our planning and conduct of CI missions. Again, this was unclassified data that resided in denied areas [emphasis added] (Kilcullen, 2009, p. 293).

But attitudes that CI practitioners think of as “residing in denied areas” are not in fact public attitudes. A public is a pervasive reality. Which geographic or social areas are “held” or “denied” by other forces are likely to change rapidly. One can misunderstand the chessboard (or the go board) by placing too much emphasis on one corner of the board that is under the adversary’s control. This analogy holds good where the public is concerned. Regional variations, while measurable, are often overrated because they appear to line up with differences in who controls different regions. But often these variations in attitudes are only “wrinkles” of 25% or less in a widespread majority attitude that is found throughout the country.

Toward Disaggregating Adversaries On a more positive note, there is another line of research that is undeveloped but could assist the progress of evidence-based CI and CT. It is worthwhile and promising to use public opinion research to learn whether a national population distinguishes between different violent militant groups, where there is a profusion of such groups (as in, for instance, Pakistan). Such research may help to disaggregate such groups, lessen the tendency to lump them all together, and provide the public with information that is persuasive. In this vein, Kilcullen comments: “…there is a need to disaggregate adversaries, separate them from each other, turn them where possible against each other, and deal with those who need to be dealt with in sequence rather than simultaneously” (Kilcullen, 2009, p. 285).

13

Public Opinion Research and Evidence-Based Counterinsurgency

321

Here survey research could be of great value. A public is ultimately the audience for the spectacular events that terrorists produce by working with the means available to them. The public has to bear these events’ psychological impact. Where there are multiple nonstate actors, sometimes working together, sometimes working different sides of the street, it is worthwhile to ask the public in detail about their impressions of each group, its aims, its ideology, and its methods. If the public makes some distinctions between groups, this is crucial information. If the public lumps together groups that in fact have differing aims and methods, it is crucial to know this as well, because were the public aware of these differences, this might lead to a drop in tolerance for a group that is especially antipathetic to the public’s values. If one learns that a major terrorist event has been followed by a strong shift in attitudes – either short or long-lasting – it is crucial to analyze that particular event and what meaning the public drew from it, in order to better explain to the public the toxic nature of nonstate actors that attack civilians. Here are some examples, all drawn from Pakistan, of how poll questions meant to aid the disaggregation of groups can be asked to a national sample. These questions should not be regarded as models; in fact, they are too simplistic. Little thought has been given so far as to how best to use poll questions with a national sample for this purpose. In one experiment in a WPO Pakistan poll in 2007, respondents were reminded that “groups such as Jaish-e-Mohammad, Hizbol Mujahideen and Lashkar-e-Taiba have fought in Indian-controlled Kashmir in the past,” and were asked (for each group separately) whether it was their impression “that [the group] has intentionally targeted civilians in attacks – or do you think it has never intentionally targeted civilians?” (Fair, Ramsay, & Kull, 2008, p. 15). This was, actually, a problematic question that asked about a concrete behavior that respondents could not be sure about. In the results, the sample made no meaningful distinctions made between groups. Overall, 40–42% thought that each group has never intentionally targeted civilians, while 6% thought it had, and a majority said they did not know (46–49%) or refused (5–7%). Nonetheless, these questions have a value: they demonstrate that the groups named did not have any widespread reputation in the public for targeting civilians in their Kashmir operations. Hence, to declare the opposite to the public and present evidence about these groups’ attacks would be a difficult task of persuasion that should be viewed as a long-term effort. In the same poll, respondents were asked to think about a number of militant organizations (tanzeems) as a cluster – “groups like Lashkar-e-Taiba, Jamaat ul Dawa, Hizbol Mujahadeen, and Jasish e Mohammad among other tanzeems.” They were then asked, “Do you think they provide social and community services, or are these not part of their activities?” The researchers’ expectation was that these organizations’ actual community services were fairly well known and an important part of their image. But it turned out that only 23% of urban Pakistanis knew these tanzeems provided such services; 42% thought they did not (the rest gave no answer; Fair, Ramsay, & Kull, 2008, p. 15). Those who knew of tanzeems’ community services were asked to mention in their own words a few services they were aware of. This subgroup overwhelmingly

322

C. Ramsay

Table 13.3 Pakistani public’s threat perceptions of militant groups, 2007 Activities of groups Percentage seeing activity Percentage seeing (September 2007) as at least an important threat activity as a critical threat Activities of al Qaeda 63 41 Activities of militant organizations 61 38 (askari tanzeems) in Pakistan Activities of Islamist militants 60 34 and local Taliban in FATA and settled areas Source: Fair, Ramsay, & Kull (2008)

mentioned schools – both madrassas and other types. But other services (medical, humanitarian, and financial) were known to very few. The value of these questions about militants’ social efforts is twofold. First, they put the seeming success of militant organizations in developing community services into perspective. At the level of the Pakistani public, community services are not what these groups are known for – at least not yet. Second, the questions highlight that tanzeems’ educational offerings are by far their strong suit in approaching the population and presenting themselves as something more than fighting organizations. This raises the larger issue of the severity of the population’s educational needs and the level of government investment toward them. A final example: urban Pakistanis were asked their threat perception of the activities of a range of different types of groups, and whether, for “the vital interests of Pakistan over the next 10 years,” they saw these activities as a critical threat, an important but not critical threat, or not an important threat at all (Fair, Ramsay, & Kull, 2008, pp. 22–23). Their responses can be ranked by the number saying a threat was at least important – since this gives a sense of the size of the potential audience for making the argument that such activities are not only important threats, but increasingly critical ones. Many observers at the time (2007) – including this writer – focused on the Pakistani public’s relatively low perception of threat from al Qaeda and homegrown militant groups, since only slightly more than a third saw such threats as critical. However, the audience that might be persuaded as time went on that these threats were critical was much larger, and included three in five urban Pakistanis (Table 13.3). The size of this potential audience was proven at the time of the Swat Valley struggle between the Army and the Pakistani Taliban in Spring 2009. When polled in May 2009, all those who had seen the threats from militant organizations as at least important now saw them as critical (Ramsay, Kull, Weber, & Lewis, 2009). This effect was lowest in the question about militant organizations in Pakistan overall, because many who were now alarmed by the threat still saw it as essentially localized in the west. However, this was now a large potential audience for making the argument that the rest of Pakistan is not immune from the negative effects of uncontrolled, violent nonstate actors (Table 13.4).

13

Public Opinion Research and Evidence-Based Counterinsurgency

323

Table 13.4 Pakistani public’s threat perceptions of militant groups, 2009 Percentage seeing activity Percentage seeing activity Activities of groups (May 2009) as at least an important threat as a critical threat Activities of Islamist militants 95 (up 35 points) 81 (up 47 points) and local Taliban in FATA and settled areas Activities of al Qaeda 94 (up 31 points) 82 (up 41 points) Activities of militant organizations 85 (up 14 points) 67 (up 29 points) (askari tanzeems) in Pakistan Source: Ramsay, Kull, Weber, & Lewis (2009)

Future research that probes more deeply into a public’s perceptions and assumptions about actual groups, their methods, and their possible appeal is a promising line of inquiry that CT and CI practitioners should ponder.

Conclusion Public opinion surveys have wide relevance for counterterrorism and CI issues of many different “gauges” – the global dimension; the “home” dimension of Western publics that are targets of terrorist actions; Muslim civilization, a different “home” dimension of publics that are both targets and hosts for terrorist groups; and single nations rent by conflicts. At the same time there is a danger to avoid: the borrowing of methods from survey research for the uses of military fieldwork, followed by confusion that distorts the science behind random sampling and leads to invalid conclusions. The most important lesson to draw from this chapter is to weigh this danger when considering how evidence from public opinion surveys can inform CI methods.

References ABC News, BBC, & ARD. (2010). Afghanistan: Where things stand. Retrieved June 10, 2011, from the Langer Research Web site: http://www.langerresearch.com/uploads/1116a1Afghanistan.pdf. BBC World Service, Globescan, & Program on International Policy Attitudes. (2007). World view of U.S. goes from bad to worse. Retrieved June 10, 2011, from WorldPublicOpinion.org Web site: http://www.worldpublicopinion.org/pipa/articles/international_security_bt/306.php? lb=brglm&pnt=306&nid=&id=. Birzeit University Development Studies Center. (2004). Poll no. 20, part 2: An opinion poll concerning the Palestinian living conditions, negotiations and peace process after Arafat. Retrieved June 10, 2011, from Birzeit University Web site: http://home.birzeit.edu/cds/opinionpolls/ poll202/. Birzeit University Development Studies Center. (2005). Poll no. 22: Withdrawal from Gaza, President’s performance, legislative elections. Retrieved June 10, 2011, from Birzeit University Web site: http://home.birzeit.edu/cds/opinionpolls/poll22/.

324

C. Ramsay

Fair, C., Ramsay, C., & Kull, S. (2008). Pakistani public opinion on democracy, Islamist militancy, and relations with the U.S. Washington: WorldPublicOpinion.org and the United States Institute of Peace. Retrieved June 10, 2011, from WorldPublicOpinion.org Web site: http://www.worldpublicopinion.org/pipa/pdf/jan08/Pakistan_Jan08_rpt.pdf. Integrity Watch Afghanistan. (2010). Afghan perceptions and experiences of corruption. Kabul, Afghanistan: Integrity Watch Afghanistan. Retrieved June 10, 2011, from IWA Website site: http://www.iwaweb.org/Reports/PDF/IWA%20corruption%20survey%202010.pdf. Kilcullen, D. (2009). The accidental guerrilla: Fighting small wars in the midst of a big one. Oxford: Oxford University Press. Kull, S. (2000). Americans on federal budget priorities: A study of U.S. public attitudes. Washington: Center on Policy Attitudes and Knowledge Networks. Kull, S. (2011). Feeling betrayed: The roots of Muslim anger at America. Washington: Brookings Institution Press. Kull, S., & Destler, I. M. (1999). Misreading the public: The myth of a new isolationism. Washington: Brookings Institution Press. Kull, S., Ramsay, C., & Lewis, E. (2011). How the American public would deal with the budget deficit. Washington: Program on Public Consultation and Knowledge Networks. Retrieved June 10, 2011, from WorldPublicOpinion.org Web site: http://www.worldpublicopinion.org/ pipa/pdf/feb11/Budget_Feb11_rpt.pdf. Kull, S., Ramsay, C., Subias, S., Weber, S., & Lewis, E. (2005). The federal budget: The public’s priorities. Washington: Program on International Policy Attitudes and Knowledge Networks. Retrieved June 10, 2011, from WorldPublicOpinion.org Web site: http://www.worldpublicopinion.org/pipa/pdf/mar05/FedBudget_Mar05_rpt.pdf. Kull, S., Ramsay, C., Weber, S., Lewis, E., & Mohseni, E. (2009). Public opinion in the Islamic world on terrorism, al Qaeda, and U.S. policies. Retrieved February 25, 2009, from WorldPublicOpinion.org Web site: http://www.worldpublicopinion.org/pipa/pdf/feb09/ STARTII_Feb09_rpt.pdf. Ramsay, C., Kull, S., Weber, S., & Lewis, E. (2009). Pakistani Public Opinion on the Swat Conflict, Afghanistan, and the US. Retrieved July 28, 2011, from WorldPublicOpinion.org Web site: http://www.worldpublicopinion.org/pipa/articles/brasiapacificra/619.php?lb=bras&pnt =619&nid=&id=. The Asia Foundation. (2010). Afghanistan in 2010: A survey of the Afghan people. San Francisco: The Asia Foundation. Retrieved June 10, 2011, from The Asia Foundation Web site: http:// asiafoundation.org/country/afghanistan/2010-poll.php. Vincent, E., Eles, P., & Vasiliev, B. (2009). Opinion polling in support of counterinsurgency. The Cornwallis Group XIV. Analysis of societal conflict and counter-insurgency (pp. 104–125). Retrieved July 28, 2011, from http://www.thecornwallisgroup.org/cornwallis_2009/7-Vincent_ etal-CXIV.pdf. WorldPublicOpinion.org. (2006). The Iraqi public on the US presence and the future of Iraq. Retrieved June 10, 2011, from WorldPublicOpinion.org Web site: http://www.worldpublicopinion.org/pipa/pdf/sep06/Iraq_Sep06_rpt.pdf. WorldPublicOpinion.org. (2008). Poll finds widespread opposition to US bases in Persian Gulf. Retrieved June 10, 2010, from WorldPublicOpinion.org Web site: http://www.worldpublicopinion.org/pipa/articles/international_security_bt/579.php?lb=brglm&pnt=579&nid=&id=. WorldPublicOpinion.org. (2009). As Hu Jin Tao, Obama prepare to meet, world public gives China, US low marks on climate change. Retrieved June 10, 2010, from WorldPublicOpinion.org Web site: http://www.worldpublicopinion.org/pipa/pdf/nov09/WPO_China_Nov09_quaire.pdf.

Chapter 14

Counterinsurgency and Criminology: Applying Routine Activities Theory to Military Approaches to Counterterrorism* Breanne Cave

Counterinsurgents have two distinct but connected problems to manage when responding to an insurgency. First, they must be able to control and reduce insurgent violence. Second, they must assert and maintain the authority of a central government (see Kilcullen, 2009). While scholarly interest in counterinsurgency has expanded over the past decade, there is not much evidence base developed for counterinsurgency and relatively little research attention that focuses on the outcomes of tactical-level military interventions during counterinsurgency, particularly in urban environments. This lack of theoretical, empirical, and evaluative attention is unfortunate, given that a majority of counterinsurgents in current conflicts in Iraq and Afghanistan are military personnel operating on the street level (Belasco et al., 2010; Hoffman, 2007). More generally, such information about counterinsurgency is directly relevant to counterterrorism studies, not only in light of the substantive, political, and rhetorical overlap of the two, but also given that military operations in foreign countries continue to be one major approach used to respond to terrorism (see BrophyBaermann & Conybeare, 1994; Enders & Sandler, 1993, 2000; Enders, Sander, & Cauley, 1990). To add to the research in this area, this paper focuses on two discussions. First, I discuss whether criminological theory about opportunity and crime situations can be applied to explain low-intensity insurgent violence in an urban environment. Second, I hypothesize whether such theories could be extended to study how political, social, and physical context can influence the capacity of social control agents to exert control over time and space during an insurgency. By using opportunity and situational perspectives to understand how violence and control are

*The author is also a Captain, United States Marine Corps (Reserve). Author’s comments do not reflect the views of the U.S. Marine Corps or Department of Defense. B. Cave (*) Department of Criminology, Law and Society, Center for Evidence-Based Crime Policy, George Mason University, Fairfax, VA, USA e-mail: [email protected] C. Lum and L.W. Kennedy (eds.), Evidence-Based Counterterrorism Policy, Springer Series on Evidence-Based Crime Policy 3, DOI 10.1007/978-1-4614-0953-3_14, © Springer Science+Business Media, LLC 2012

325

326

B. Cave

linked to context, U.S. military forces engaged in counterinsurgency may be able to better plan and evaluate interventions at the tactical level.

Insurgency and Counterinsurgency An insurgency is an movement that “uses subversion and violence to reach political ends…to gain power to overthrow or force change of a governing authority” (Joint Publication 3-24, 2009). This goal is what distinguishes insurgency from general terrorism, which may use violence against symbolic targets for a wide variety of purposes (Joint Publication 3-26, 2009). Insurgents are defined more specifically by their intent to destabilize an existing government. Therefore, an effective counterinsurgency is not solely, or even primarily, focused on deterring or capturing individual insurgents, but rather on restoring a failing or vulnerable government’s authority.1 Until recently, studies of insurgents and counterinsurgents were primarily undertaken by military personnel who had direct experience with this type of conflict (e.g., Galula, 1964; Kilcullen, 2009; Leighton & Sanders, 1962; Nagl, 2002; Petraeus, 2006). Over the past decade, interest in this area has expanded dramatically as political scientists, sociologists, and anthropologists have taken a renewed interest in conducting studies in areas that are politically unstable, sometimes partnering with military units that are currently involved in counterinsurgency operations (e.g., see Kelly, Jauregui, Mitchell, & Walton, 2010). For many current U.S. operations, military leaders have actively encouraged such research, noting that the social sciences will play an increased role in future conflicts (U.S. Department of Defense, 2010). Despite this expansion of interest in the social sciences, little research has been applied to understanding and evaluating operations at the tactical level (e.g., see McFate, 2005a). Military organizations continue to emphasize more traditional forms of knowledge – in-depth assessments of known enemies, weapons, and physical terrain – over concerns such as the political motivations, community characteristics, and other social and cultural considerations (Flynn, Pottinger, & Batchelor, 2010). Researchers who are involved in the study of insurgency and counterinsurgency tend to focus on the advising commanders and coordinating intelligence-gathering efforts rather than on theorizing about operations that take place at the tactical, small-unit level (for instance, Human Terrain Teams; see Fawcett, 2009; Kipp, Grau, Prinslow, & Smith, 2006). For most practitioners, military doctrine, or publications that explain the current state of operational knowledge and recommended practices in response to common problems, comprises the majority of available information about insurgency and its political and social context (see for instance, U.S. Department of the Army, 2008).

1

While this paper focuses primarily on the perspective of a military force acting in the capacity of the counterinsurgent, a similar approach might be developed for the insurgent or revolutionary perspective.

14

Counterinsurgency and Criminology: Applying Routine Activities Theory…

327

Military doctrine provides a starting point for developing theory and evaluation in the area of counterinsurgency. Specifically, doctrine can tell us two things about street-level counterinsurgency operations. First, the particular situations that counterinsurgents face vary significantly and are especially variable within an urban area, sometimes changing significantly from street block to street block (Fawcett, 2009; Krulak, 1993). Second, situations and places matter. Successful counterinsurgents have to understand how different situations shape the potential for insurgent violence and community responses to the insurgent and counterinsurgent (U.S. Department of the Army, 2008). Although the doctrine establishes that responding to situational contingencies is important, it does little to describe precisely what is important about situations or how a counterinsurgent operating at a tactical level should adapt in response to situational contingencies. The variability of situations and the importance of context in determining the outcomes of interventions are not unique features of counterinsurgency. Criminologists who study urban crime and policing have found that, like insurgent activity, problems of violence and crime vary significantly from street block to street block, and responses to these problems must be shaped in part by the context in which they occur (Clarke, 1995, 1997; Eck & Weisburd, 1995; Felson, 2002, 2008; Weisburd, 2002, 2008). Further, while counterinsurgents do not have theory that specifically relates situational factors to variations in violence and disorder, criminologists have a number of explanations for the causes of situational variation in crime that may also be relevant to insurgent violence. Such theoretical, practical, and evaluative perspectives could be essential for a more effective, evidence-based approach to counterterrorism or counterinsurgency policy, especially given the importance that both doctrine and theory place on situation and location. There are a variety of criminological theories that explain violence in terms of its situational context. Opportunity theories (see Clarke, 1980, 1995, 1997) and routine activities theories (see Cohen & Felson, 1979; Felson, 2002; Sherman, Gartin, & Buerger, 1989) state that crime varies significantly between situations based on the structures of opportunities that exist at specific places. Criminologists who take such a situational, place-based perspective search for “regularities in relationships between behavior and situations” (Birkbeck & LaFree, 1993, p. 116) and predict the aspects of situations that are likely to lead to criminal offending (Felson, 2002). When suitable targets, lack of guardianship, and crime facilitators such as weapons and alcohol converge in space and time, opportunities can be created that motivated offenders can take advantage of, if given the incentive to do so (Clarke, 1980; Cohen & Felson, 1979). Thus, opportunity theories consider not only how opportunities arise for individuals motivated to commit crime but also how specific places and their characteristics can create the environment to attract, host, and generate opportunities for violence (Sherman et al., 1989). Place-based theories also imply that increasing guardianship, hardening targets, and blocking opportunities through “situational crime prevention” measures may be effective (e.g., see Clarke, 1997; Eck, 2002). Given that opportunities can cluster at places for various reasons, geographic targeting and “hot spots” approaches may also be effective (Sherman et al., 1989). Further, there are growing indications that

328

B. Cave

insurgents and terrorists, like criminals, are responsive to variations of opportunity across space and over time (see for instance, Dugan, Lafree, & Piquero, 2005; Kilcullen, 2009). Situational and place-based theories in criminology may, therefore, shed light on effective counterinsurgency activities at specific places. It is to this discussion that I turn now.

Theoretical Perspectives Linking Opportunities for Violence with Situational Context and Opportunity Individuals and Opportunity Criminological theory has primarily focused on explaining why individuals commit crime rather than why crime occurs at specific places (Weisburd, 2002; Weisburd, Maher, & Sherman, 1992). While some of these theories have focused on early risk factors and dispositional explanations for criminal behavior, other criminological perspectives view individuals as “rational offenders” who take advantage of opportunities to commit crime (Clarke, 1980; Clarke & Felson, 1993; Cornish & Clarke, 1986, 1987). Rational choice perspectives assert that while offenders are perhaps constrained by limited information, when given the opportunities within a particular situation, they still calculate risks and benefits in deciding whether to engage in criminal behavior. While disposition and individual risk factors may play a role, these characteristics do not explain why people offend at certain times or places and not others. It may be the case that some people are more likely to offend simply because they have patterns of behavior that bring them into contact with more criminal opportunities (Clarke, 1992; Clarke & Homel, 1997; Felson, 1997, 2002). Opportunity theory and the rational choice perspective focus on decision making at the time of the crime, rather than on predispositions and risk (or protective) factors. One of the most influential theories about the interaction between individuals, opportunity, and rational choice is routine activities theory (Cohen & Felson, 1979). Routine activities theory states that the opportunity for direct contact predatory crime arises from the convergence in time of a motivated offender, suitable target, and lack of a capable guardian, as illustrated by the “crime triangle” in Fig. 14.1 (Cohen & Felson, 1979; Eck, 2003; Felson, 2008). As Farrell and Pease (2008) discuss, the lifestyle of potential victims may make them more or less likely to come in contact with potential offenders (see also Eck, Clarke, & Guerette, 2007; Farrell, 1995, 2006; Pease, 1998). Additionally, different guardians or agents of crime prevention may be more or less available and able to prevent criminal activity based on their relationships with places, potential offenders, and potential targets (Eck & Wartell, 1997; Felson, 1986; Tillyer & Eck, 2010). Such convergences of offenders, victims, and absent guardians result from the “rhythm,” “tempo,” and “timing” of human activities such as family and work activities or personal lifestyles (Cohen & Felson, 1979, p. 590), which are in turn influenced

14

Counterinsurgency and Criminology: Applying Routine Activities Theory…

329

Fig. 14.1 The crime triangle (adapted from Eck, 2003)

by macrosocial forces (i.e., modernization or changes in employment patterns). Ultimately, these lifestyles influence the frequency, location, and timing of these convergences, which present (or block) opportunities for crime.

Places and Opportunity The initial theoretical descriptions of routine activities and opportunity theories by Cohen and Felson (1979), and Clarke (1995) did not in and of themselves explain why such convergences and opportunities seem to concentrate at particular places, thus leading to crime clustering. Ecological and crime pattern theories attempt to describe how characteristics of specific physical environments could facilitate crime by creating and attracting opportunities, such as the presence of bars or retail stores (Brantingham & Brantingham, 1993, 1995, 2008; Eck & Weisburd, 1995; Newman, 1972; Sherman et al., 1989; Weisburd, 2002, 2008). These environmental and ecological theories of crime link the activities of individuals to their physical context. Sherman et al. (1989) extended the notion of the convergence of opportunity in routine activities by arguing that convergences of motivated offender, suitable target, and lack of guardian might happen more frequently at some places. They argued that places can attract, generate, or facilitate crime, and that characteristics of places play an important role in the prevalence of crime at those locations. More specifically, Sherman and colleagues showed that these places may be very small in size and that crime is incredibly concentrated at “micro” places, such as addresses, street segments, and individual businesses, rather than at larger areas such as neighborhoods or city districts. They discovered that 50% of crime was concentrated at 3% of addresses in the city of Minneapolis. Even within areas that might be considered “disadvantaged” there was an incredible amount of variation of crime across

330

B. Cave

geographic space (see also Weisburd, Morris, & Groff, 2009). These findings about the concentration of crime and place have also been observed in many other studies (see Clarke & Harris, 1992; Groff, Weisburd, & Morris, 2009; Hunter & Jeffrey, 1992; Pease, 1991; Stark, 1987; Weisburd, Bushway, Lum, & Yang, 2004; Weisburd, Groff, & Yang, 2010; Weisburd & Green, 1995b). In addition, despite the variability across places of crime concentrations, the location of these concentrations remains incredibly stable over time (Taylor, 1999; Weisburd et al., 2004).

Place and Opportunity-Based Intervention Strategies The finding that crime concentrates at very specific places and that these concentrations, or “hot spots,” are often stable over time indicates that offenders and opportunities do not easily “move around the corner.” Instead, crime occurs in predictable ways and at specific locations (Weisburd et al., 2004, 2006). These findings have significant implications for place-based interventions to violence. Offenders are not only difficult to locate, but can change their offending patterns over time. On the other hand, physical locations are, by definition, fixed in geographic space and thus more convenient subjects for intervention. There is a growing body of evaluation research that shows that police interventions to improve guardianship, harden targets, and reduce vulnerabilities at micro places can be effective in reducing crime (Clarke, 1997; Eck, 2002; Sherman & Weisburd, 1995; Weisburd & Green, 1995a). Indeed, the National Research Council of the National Academies of Sciences recently stated that “studies that focused police resources on crime hot spots provide the strongest collective evidence of police effectiveness that is now available” (National Research Council, 2004, p. 250; see also Braga, 2007, for a review of hot spot studies). Effective interventions at crime hot spots may take a wide variety of forms, such as situational crime prevention measures, problem-solving, or deterrence strategies. For instance, simply increasing the amount of time that police spend in an area can have a crime reduction effect (Braga & Weisburd, 2010; Eck, 2002; Koper, 1995; Sherman & Eck, 2002; Sherman & Weisburd, 1995; Skogan & Frydl, 2004). Police may also attempt to diagnose places to determine why they seem to experience high rates of crime and then use problem-solving strategies to change the characteristics of the place that are linked to persistent crime, as in problem-oriented policing and situational crime prevention (Braga, 2002; Braga & Bond, 2008; Braga et al., 1999; Clarke, 1995; Hope, 1994). One area of emerging interest to hot spots researchers has been increasing protective factors in places – that is, making places more resistant to crime by involving residents or other organizations (such as churches, youth organizations, and so on) in crime prevention efforts (Braga, Kennedy, Waring, & Piehl, 2001; Mazerolle & Ransley, 2005; Mazerolle, Roehl, & Kadleck, 1998). While these interventions that are often intended to address the capacity to self-regulate have traditionally occurred at the community level (e.g., Reiss, 1986; Sampson, 1997, 1999, 2004; Skogan, 1990;

14

Counterinsurgency and Criminology: Applying Routine Activities Theory…

331

Welsh & Hoshi, 2002), research also suggests that collective efficacy, or the capacity of a community to self-regulate, can significantly vary small units of geography such as the street segment, and that such small units of geography may be a significant unit of social organization (Taylor, 1997; Weisburd, Telep, & Braga, 2010). A frequent objection to these prevention strategies is that offenders could simply displace in response to a change in opportunity caused by an intervention in a particular area (Reppetto, 1976). Displacement could take many forms: Offenders may shift to another type of criminal activity, move to another location, or attempt to commit crimes at another time of day when the police are not present (Clarke & Weisburd, 1994). Research about spatial displacement has generally shown that it seldom occurs and is seldom total, and in fact a diffusion of benefits is just as likely to take place (Braga, 2005, 2007; Guerette & Bowers, 2009; Weisburd et al., 2010). That is, potential offenders are more likely to overestimate the amount of crime prevention activity that is taking place and stop offending in locations that are adjacent to targeted crime hot spots rather than to persist in the face of changing police activity (e.g., see Weisburd et al., 2006). While all of these principals may differ for insurgency, at least for violence, focusing anticrime efforts at very specific places in which opportunities to commit crime exist has had positive effects.

Insurgency, Opportunity, and Situation: Applying a Place-Based, Situational Perspective to Counterinsurgency The development of opportunity theory, routine activities, and place-based criminological perspectives all suggest that variations in the concentrations of crime at specific places might be best understood by the spatial distribution of the convergences of motivated offenders, suitable targets, lack of guardianship, and the presence of crime facilitators and opportunities for crime. Further, evaluation research also suggests that place-based interventions such as situational crime prevention, hot-spot policing, problem-solving at certain locations, or other environmental manipulations can reduce crime at places. But can such ideas also apply to insurgent violence? It is not unreasonable to think so. The notion that insurgent activity or terrorism is analogous to crime is not a new one, although criminologists seem divided in how strong the relationship might be (see Draa, 2008; Johnson & Braithwaite, 2009; Lum & Koper, 2011; Lynch, 2011). Certainly, some forms of insurgent behavior, such as the emplacement of improvised explosive devices and small arms fire attacks on military or civilian targets, may resemble the “direct contact predatory offending,” which is the subject of many opportunity theories (e.g., see Cohen & Felson, 1979). Further, although not examining insurgency, Lum and Koper (2011) note that criminological perspectives can be useful for the study of terrorist violence in that preparation for acts of terrorism and acts of terrorism themselves often involve illegal activities that can be blocked or deterred. The same may be true of insurgent behavior.

332

B. Cave

With regard to insurgency and its analogy to crime situations, neither crime nor insurgency is unique in its tendency to concentrate in specific places. There are many types of social and natural activities and conditions that concentrate across geography, including economic activity (Last, 1997), disease (Ottaviano & Puga, 1997), and functions related to governance (Jonas & Ward, 2007). It would be exceptional if insurgent activity, which bears such a strong resemblance to predatory offending behavior, did not show similar concentration in space and time. Thus, similar to place-based interventions for crime, there may be benefits for a placebased approach to counterinsurgency, in that counterinsurgents are likely to find similar problems in focusing prevention efforts on individuals and neighborhoods, as criminal justice agencies have experienced. Insurgent violence could likely to be characterized by instability of offending behavior among individuals and significant variation (yet stability) in patterns of violence within neighborhoods. Yet, as with crime (see Weisburd, 2002), unwarranted assumptions are often made about the nature of insurgency and counterinsurgency, that can inhibit thinking about these issues in place-based ways. Insurgents are often seen as irrational actors, ready to attack no matter the cost or situation (e.g., see Gurr, 1970). But as with crime, such assumptions require more careful inquiry. Whether one views insurgents as criminals or adversaries in war, both perspectives can imply rational decision-making process. Indeed, targeted places and times are not random – places with large amounts of people (or specific types of people) present are often chosen. Hypothesizing about the rationality of insurgents raises a number of further questions: Do insurgents respond to variations in opportunity in the social and physical environment? Does insurgent activity concentrate at specific places? Do interventions aimed at changing opportunity for insurgent violence work without displacing insurgents to other locations? What kinds of contextual factors are related to spatial concentrations of insurgent violence?

The Relevance of Opportunity and Routine Activities Theory to Insurgency As previously mentioned, the popular image of insurgents is that, they are persistently motivated offenders who will carry out attacks regardless of the situation because of some radical ideology or affiliation with a larger organization that directs their activities. Research about insurgent behavior, however, suggests that this image is not accurate for all insurgents. For instance, Kilcullen’s (2009) model of the “accidental guerrilla” depicts an insurgent who is driven by local concerns, opportunities to offend, or a reaction to poor intervention strategies from a counterinsurgent. When situations change, interventions become less of a perceived challenge to local norms or sovereignty, or opportunities to attack local government interests are reduced, this less dedicated group of insurgents becomes less likely to offend. Although their focus was not on insurgent activity, Dugan et al. (2005) found similar rational, opportunity-based responses to deterrence strategies on the part of terrorists when examining hijacking. In short, research suggests that insurgents and

14

Counterinsurgency and Criminology: Applying Routine Activities Theory…

333

terrorists are rational actors who are responsive to variations in opportunity and context. Insurgents are not the only rational actors in this equation. The individuals and communities who live in countries that are experiencing an insurgency also exhibit rationality. The support of the population within which insurgents operate has long been recognized as a key factor in the success (or failure) of insurgency (e.g., see Cassidy, 2004; Mao, 1937; Petraeus, 2006). Communities are able to set norms of behavior for its members and are able to reinforce or counteract control efforts on the part of a central government (Scott, 1985), thereby, shaping the capacity of insurgents to operate in different areas. Anthropologists, economists, and political scientists have variously examined the effect of programs intended to promote public participation and perceptions of civil authorities (Gurr, 1970; Horowitz, 1985), investment in small civil affairs projects (Berman, Shapiro, & Felter, 2008), and government access to financial resources on individual and community support for the authority of the government (Collier & Hoeffler, 2004; Sambanis, 2003). As Galula notes, individual-level support for government authority may also vary significantly across space and time (1964). Criminologists have similarly observed that structural and cultural factors may fundamentally shape the opportunity and incentives for committing crime at certain places (e.g., see Clarke, 1980; Cloward & Ohlin, 1960; Cohen, 1955). If insurgents do, in fact, act more rationally than assumed, and if their motivation to offend varies by opportunity, situation, context, and routines, then according to routine activities and place-based theories, areas that offer more opportunity for violence would be more heavily targeted by insurgents, more receptive to insurgent activity, or attract more violence. This in turn may lead to a high concentration of insurgent activities at specific places. In fact, observers have noted that terrorist violence and insurgency are highly concentrated in time and place (Johnson & Braithwaite, 2009; Townsley, Johnson, & Ratcliffe, 2009), among cities (Behlendorf, Lafree, & Lagault, 2010), or in provinces (Berman et al., 2008) and regions (GlobalSecurity.org, 2009). This may imply that environmental factors may be operating at places to attract, generate, and host insurgent activity in a similar way that opportunity in places influences the generation of crime. Furthermore, similar to place-based crime prevention, interventions that attempt to block insurgent opportunity to offend have been successful in reducing many different types of terrorist and insurgent violence. For example, Kagan (2006) discusses the construction of sand berms (earth barriers that surround a city and reduce outsider access to it) around Iraqi cities that cut off insurgent access to easy targets and led to an apparent reduction of violence. Bynam (2006) discusses the construction of a fence along the Israeli-Palestinian border that led to a reduction in attacks by Palestinian terrorists. These deterrent effects also apply to transnational terrorism; for instance, many scholars have found that enhanced security at airports resulted in a reduction of terrorist violence that have targeted airlines (see Clarke & Newman, 2006; Dugan et al., 2005; Enders & Sandler, 1993; Enders et al., 1990; Landes, 1978). While these interventions do not follow the “hot spots” model of intervention, they do support the notion that place-based variation in opportunity is important in terms of predicting variations in insurgent activity.

334

B. Cave

Limitations of Using Place-Based Theories for Insurgency There are, however, limitations to describing insurgencies using a place-based, situational, or routine activities perspective. Although the base of theory and research in counterinsurgency seems to support the idea that a situational perspective is worth investigating, there are significant gaps in the body of research surrounding insurgent violence. First, counterinsurgents should attempt to determine whether opportunity for insurgent violence, like that for crime, varies at the micro as well as the macro level. Second, research about the correlates of insurgent activity, or the factors that are linked to violence at micro places, needs to be conducted in order to establish which environmental and social factors appear to be most related to spatial trends of persistent violence. Third, more rigorous evaluations should be conducted to determine whether opportunity-blocking approaches actually reduce violence during an insurgency. If spatial patterns of insurgent violence are predictable, and if concentrations of insurgent activity can be described by theories similar to those developed for other forms of predatory offending, then it is likely that intervention strategies similar to those that have been developed for crime at places could be applied to urban insurgency. One basic shortcoming of research about interventions intended to reduce crime opportunity is that there is not enough of it. As Kennedy, Caplan, and Piza (2010) observed, some place-based crime prevention efforts backfire. There is also relatively little known about the effects of short-term changes in arrest patterns in terms of the long-term impacts on future offenses (e.g., see Taylor et al., 2009), its effect on resident perceptions of the police (Kochel, in press; Rosenbaum, 2006), or its potential to invoke “defiance” or other negative attitudes from offenders (as in Sherman, 1993). These concerns do not negate the effectiveness of hot spots and situational strategies on reducing crime and violence, but they do strongly suggest that police and other social control agencies must develop a nuanced understanding of crime problems and take into account potential for harm that could occur as a result of interventions intended to reduce violence. As mentioned previously, Kilcullen’s (2009) model of insurgency implies that opportunity plays a role in at least some insurgent activities. However, while some insurgents may be motivated by opportunity in the environment, others are actually a part of insurgent organization and may be more likely to seek out opportunities in order to advance that organization’s agenda (as in the corporate model proposed by Lynch, 2011). This may mean that situational responses to insurgent activity may be more effective in terms of reducing violence without displacement in some areas than others, depending on the state of insurgent organization and the type of activity targeted (see Lum, Kennedy, & Sherley, 2006, for a discussion of displacement as a result of counterterrorism interventions). Opportunity theories assume that the social and physical environment in micro places can explain a significant amount of the variation in crime and violence that occur in those places. However, counterinsurgents operate in a wide variety of environments, some of which might be significantly altered by the presence of an insurgency. Conflicts within a city may displace residents, disrupt normal patterns of

14

Counterinsurgency and Criminology: Applying Routine Activities Theory…

335

activity, and cause disorder and physical damage in larger geographical areas (Taw & Hoffman, 1994; U.S. Department of the Army, 2008). In those circumstances, there may be more homogeneity in opportunity for insurgent violence at larger units of geography enabled by a weaker community that lacks the ability or inclination to attempt to control the actions of its members.

A Further Limitation of Applying Criminological Theories to Insurgency and Counterinsurgency: The Role of Stable Governance in Crime Prevention There is, however, a more fundamental concern in applying criminological theories of opportunity, routine activities of places, and the criminology of place to counterinsurgency. Many of these theories were developed to explain crime in stable, modern democracies – in particular, the United States and the United Kingdom. Criminologists operating in such environments do not often consider how social and political instability – or even competition over political power and authority – may influence crime or the individuals and agencies that may legitimately control it (Lum, 2009). Recall the earlier description of routine activities theory, which suggests that crime is the result of the intersection of a motivated offender, suitable target, and a lack of capable guardianship (Cohen & Felson, 1979). To understand the relevance of this theory to counterinsurgency, we first have to examine the terms that are used to describe the interaction between the individuals involved in the “crime triangle.” There is a target – that is, a person or object that has some characteristic that makes it desirable to a potential offender. There is a motivated offender, who attempts to access the suitable target by illegitimate means. Then there is the capable guardian, who is motivated to exert some type of force, coercive or otherwise, in order to prevent the offender’s illegitimate activities. In opportunity and situational theories, the lack of guardianship is a key element in the chemistry of crime (Felson, 2002). However, these distinctions between different situational actors may become blurred during an insurgency, especially with regard to guardianship. Capability of guardianship implies the guardian also has a specific motivation or reason to exert watchfulness. According to routine activities, opportunities to commit a crime at a place emerge when a guardian is missing, scared, or weakened. Situational theories assume that there is a supply of guardians, just as there is a supply of offenders. Guardianship organically emerges as people try to protect their persons and property. But “guardianship” is a notion that means different things within different political contexts. Indeed, the Weberian idea of the state as the entity with the monopoly of the legitimate use of force (Hobbes, 1651; Klockars, 1985; Manning, 1977; Weber, 1946) is directly challenged during insurgency. The perceptions about illegitimate and legitimate users of force – the difference between the guardian and the offender – are precisely what are being contested by insurgency (Drapeau, Hurley, & Armstrong, 2008).

336

B. Cave

In fact, one way to think about “success” from the perspective of the insurgent is to say that the population no longer distinguishes between the use of force by the counterinsurgent and the insurgent, or better yet, that an insurgent is considered to have more legitimate authority to use coercion than the counterinsurgent. A lack of differentiation between the “offender” and the “guardian” is how the struggle between the government and the insurgency for the control of the population may appear from the perspective of an observer at the street level. This lack of differentiation could be attributed to a number of different causes, including the success of some insurgent program, the failure of the counterinsurgent to provide security, or harsh treatment of the population by the counterinsurgent (Shafer, 1988). In stable societies, especially where legal boundaries, property rights, and the social contract are clear, the notion of guardianship is also fairly clear. However, insurgency occurs within environments in which political stability is not certain. Indeed, insurgencies seek to capitalize on the lack of political and social order. Before, during, and after an insurgency (or any political transition), the definitions of crime and guardianship may be distorted or unstable themselves. That is, different behaviors may be considered socially prohibited, and different groups may be considered legitimate enforcers of the law. The notion of the instability of guardians and guardianship is implied in the “hearts and minds” perspectives of counterinsurgency. Specifically, while the prevention of violence is an important goal, counterinsurgents largely agree that it is a secondary concern relative to the task of gaining the support of the population for the central government and their cooperation with the activities of counterinsurgents (Galula, 1964; Kilcullen, 2009; Petraeus, 2006; U.S. Department of the Army, 2008). Not only does this mean that counterinsurgents must be seen as legitimate guardians, but increasing the supply of stable guardians and a consensus about who they are in times of crisis is a key goal. In terms of the discussion of guardianship and routine activities theory, this support produces consensus about authority to prevent crime and, more broadly to regulate behavior. Opportunity and routine activity theories in contemporary criminology do not speak directly to these issues in unstable or transitioning nations. The shifts in macro forces that affect everyday routines that Cohen and Felson (1979) wrote about were still within a relatively stable political eras in the United States. This has important implications when applying perspectives such as opportunity, routine activities, or criminology of place theories to insurgency. Can place-based interventions work if the assumption about who ought to respond to violence is challenged? Do they work when the definition of crime itself is open to debate? Relative to research that focuses on victims or offenders in routine activities theory, criminologists have devoted little attention to the nature of stability in guardianship across various political contexts. In military doctrine and research on counterinsurgency, the behaviors of the population and insurgents have been explained in many ways (e.g., see Berman et al., 2008; Kilcullen, 2009). Advice on how to best engage communities, local politicians, and insurgents of different origin certainly abounds (Sepp, 2005; U.S. Department of the Army, 2008). But similar attention has not been paid to explaining the behavior

14

Counterinsurgency and Criminology: Applying Routine Activities Theory…

337

of counterinsurgents themselves – how social control agents operate during times of significant social instability, how they foster other guardians in the population, and what might predict their success or failure in establishing or reestablishing state authority in a given location (for an example of an exception, see Drapeau et al., 2008). A situational, place-based perspective for counterinsurgency would have to incorporate these notions about the role of police-like agencies in establishing and maintaining political authority as well as the challenge of relying on and fostering natural guardians within the population. From a Durkheimian perspective, the state of the society may be more anomic during insurgency, which may in turn challenge sociological theories (and subsequently interventions) that rely on notions of stable forms of formal and informal control. Durkheim argued that societies exist on a continuum between regulation by official laws established by a central authority that emerges through the process of modernization or based on informally enforced norms of behavior (Durkheim, 1893; Macfarlane, 2002). Using this framework, guardianship might be described as contingent on the authority of the guardian and the guardian’s capacity to intervene effectively in a specific context. Given Durkheim’s suggestion that regulation of behavior can be achieved through both formal and informal sources of authority, formal guardianship can therefore be viewed as contingent on three factors: (1) the state of the law and legal authority for a particular group or individual to intercede to prevent violence; (2) the presence and strength of shared norms of appropriate and inappropriate behavior; and (3) the actual capacity of guardians to exert control. Without some degree of stability in these factors in a given setting, the concept of guardianship, one of the cornerstones of routine activities and other theories of crime prevention, may be inapplicable. That is, conflict over the content of the law, normative expectations, and power of legitimate control agencies may mean that the criminal justice analogies that we would like to draw with insurgency and counterinsurgency are invalid. However, as illustrated in the following discussion, these factors may also vary over place and time and concentrate in geographic space. While the examples below focus on the influence of these factors on formal guardians, such as police and military organizations, these concepts might also be generalized to more informal sources of control.

Authority Regardless of the conditions surrounding a counterinsurgency (for instance, whether the counterinsurgent is an occupying force or the current regime), there must be some consensus on legitimate political authority in the long term for the concept of guardianship to be stable. The successful establishment of local police agencies might be viewed as one of the symptoms of this consensus (e.g., see Bayley, 2001, 2005; Deflem & Sutphin, 2006). One can think of police authority being a result of both “top-down” and “bottom-up” influences – established from both the actions of the state and the consensus of the population.

338

B. Cave

The perspective that is prevalent in counterinsurgency literature – that peacekeeping and policing functions should operate on democratic principles – highlights the importance of public consensus about the legitimacy of central authority (e.g., see Galula, 1964; Henriksen, 2010; Kilcullen, 2009; Shafer, 1988; U.S. Department of the Army, 2008). Therefore, the long-term success of a counterinsurgent relies on its adherence to international laws and its ability to convince the population that it is acting in a legitimate manner (Galula, 1964; Kilcullen, 2009). This concern about legitimacy is a one shared with democratic police of all varieties; as Bayley (2001, p. 59) notes, a “local police force cannot be created by command. It requires the consent of politicians, the public, and the police.” The “consent of the population,” however, is not easily measured, and different potential counterinsurgents may enjoy different levels of public support (e.g., see Ramsey, 2011). Political science and criminological research, furthermore, indicates that the state of this consensus varies among individuals, neighborhoods, and even on the street-block level. Although not a perfect measure of consensus about government authority, political participation may help to illustrate this variation. Participation in civil society and trust in government authority are related concepts. As Keele (2007) states: “When citizens disengage from civic life and its lessons of social reciprocity, they are unable to trust the institutions that govern political life” (p. 241). The relationship between civil society and government authority, however, seems complex. Booth and Richard (1998), for instance, suggest civic organizations may be divisive under certain circumstances when they advocate activities that are radically different from government policy. This notion is certainly supported by one of the common adaptations to a weak central authority in Iraq: the establishment of local governments and militias (Deflem & Sutphin, 2006). Thus, the type of political participation in a particular area may influence perceptions about the authority of counterinsurgents to intervene in particular situations. More generally, however, distrust or cynicism about the government and a lack of civic participation appear to be related to one another. For instance, Sampson and Bartusch (1998) find that cynicism about police and local government varies significantly between neighborhoods and is tied to low rates of participation in voting. Weisburd, Groff, and Yang (2010) also find that voting behavior varies significantly among street segments and is related to the amount of crime that occurred in the segment. Therefore, the amount of political participation may also be indicative of perceptions on the authority of a government and the legitimacy of counterinsurgents that support it. These variations in participation and cynicism about government can help inform us about the strength of the state of formal law and authority in places. Considered alongside spatial analysis of insurgent activity or crime more generally, mapping micro-level variations in participation in governance might be helpful in assessing not only the characteristics of places that are linked to high concentrations of violence, but also those that are linked to strong support for functions of governance. This may also be important in understanding both how the prevalence of insurgent activity is linked to broader perceptions about government authority and how amenable a place might be to interventions from agencies that are tied to the central government.

14

Counterinsurgency and Criminology: Applying Routine Activities Theory…

339

Norms Continuing with the Durkheimian frame, we expect that informal enforcement of informal standards of behavior plays an important role in the regulation of societies more generally and is particularly relevant in nonmodernized societies. Taking a criminological perspective on this issue, one could use social disorganization theory to relate the presence of effective social norms to a lower level of crime and delinquency. The absence of such norms, as in socially disorganized places, may facilitate offending behavior (Bursik, 1988; Sampson, 1999; Shaw & McKay, 1942). Informal social control in the form of norms of behavior may influence not only the ability of the community to self-regulate, but also the ability of formal control agencies to assess community expectations and perform law enforcement and crime prevention activities. Similarly, Galula (1964), among others, suggests that the ability of communities to identify potential insurgents and regulate their behavior is a key factor in the success of counterinsurgents. The notion of social norms also indirectly implicates communities as natural guardians who are just, if not more important than, policelike guardians, who cannot be present constantly. This capacity for self regulation and collective efficacy may influence formal guardians in several ways. Under the model of democratic policing posited earlier, some level of responsiveness to public will and norms is required from the police or peacekeeping forces. A guardian’s actions, therefore, are likely to be influenced by the nature of the community as well as local norms and expectations of behavior. Criminologists have found that police and other social control agents significantly vary their behavior, for example, in response to norms surrounding minor criminal behavior, juvenile activity, or the racial composition of places (see Klinger, 2004; Lum, 2010; Rengert & Pelfrey, 1997). Further, one important community norm – the comfort associated with being able to alert those exerting legitimate social control – is directly connected to the success of peacekeepers or law enforcement (e.g., Bittner, 2005). The strength of community norms in regulating behavior also influences individuals within the community and their willingness to either provide guardianship or cooperate with other individuals who are involved in the enforcement of norms. Social disorganization theory assumes that, in most places, the presence of community norms discourages juvenile delinquency and other forms of criminal behavior (Shaw & McKay, 1942). The likelihood that potential offenders will offend is shaped by their relationship to families, friends, and neighbors who may encourage or discourage criminal offending (as in Eck, 1994; Felson, 1995; Reiss, 1986). Furthermore, as Felson (1986) discusses, every offender has “handles” or ways that their behavior can be shaped by others. Informal guardianship in neighborhoods thus provides another potential set of handlers who may shape the behavior of potential insurgents and reduce the likelihood that they will become involved in violence (e.g., Reynald, 2010; Tillyer & Eck, 2010). The strength of community control over individual behavior has further implications for counterinsurgents. To extend Felson’s (2006) metaphor, communities

340

B. Cave

themselves may also have “handles” or mechanisms with which their collective behavior can be influenced – for instance, through investment in community needs (e.g., Berman et al., 2008). By identifying and manipulating such handles, counterinsurgents may be able to influence the behavior of many individuals and also influence those individuals to see themselves as guardians or facilitators of opportunity. This might also imply that a community that is more poorly organized as a result of population heterogeneity or high residential turnover not only provides a venue for crime and disorder but also requires more effort from law enforcement or peacekeeping forces. It has less consensus about its values and interests, fewer “handles” that can be grasped by formal control agents, and less capacity to shape the behavior of the individuals who live within it.

Capability The ability of guardians to actually prevent violence may also vary from place to place and shape the overall state of guardianship in a given context. The previous discussion effective guardianship as it relates to community social control, but formal organizations themselves may also vary in their ability to control behavior over geographic space. There is a wide range of factors that could contribute to the variation of police or peacekeeping organizations to exert effective social control, and this topic cannot be discussed exhaustively here (although see for instance Huntington, 1968 for a discussion of institutionalization and social control). Reynald (2010) however, provides a basis for discussion of “capability” of guardianship in her identification of the willingness to supervise, ability to detect offenses, and willingness to intervene as key features of guardianship. Like authority and community norms, each of these elements of guardian capability may also vary over time across geographic space. One of the basic elements of guardianship is the willingness to supervise an area in order to deter violence or crime. Counterinsurgents may vary significantly in their willingness to meaningfully supervise an area – that is, not only to be present in an armored vehicle or reinforced outpost, but also to interact with the local population (Galula, 1964). If this notion of “willingness to supervise” is extended to include “willingness to provide services,” then one can also include civil organizations in this discussion. That is, although a particular location may require some sort of humanitarian assistance, the level of violence in an area may prohibit the population’s access to humanitarian aid (Gompert, Kelly, Lawson, Parker, & Colloton, 2009). Practically speaking, it seems important to understand how this amount of oversight and ability to access different areas shape the capacity of organizations to actually deter insurgent violence and meet community needs. The capability of guardians to detect offenses might also vary significantly across space. Ronceck (1981), for instance, suggests that the ability of police to distinguish legitimate users of space from illegitimate users – for instance, potential offenders – is one of the key elements of capable guardianship. Urban features and geography may also play a role in the ability of counterinsurgents to identify potential

14

Counterinsurgency and Criminology: Applying Routine Activities Theory…

341

insurgents, as different types of urban forms may make it more or less possible to have oversight of public areas (as in Hunter & Jeffrey, 1992; Newman, 1972). The definition of deviant behavior may also vary, with some behaviors considered prohibited in one area but permissible in others (Birkbeck & Lafree, 1993). The capacity of counterinsurgents to analyze the environment and identify opportunities for insurgents could also be considered one component of their capability to detect offenses. Finally, the willingness of peacekeeping or police forces to intervene in different ways may vary over space and between organizations. Felson (2006) argues that the capability of an effective guardian is also related to an understanding of his or her role in preventing crime (see also Reynald, 2010). One way that this understanding might vary is in the type of activity that these agencies consider appropriate to respond to insurgent activity – for instance, it might not be apparent to some counterinsurgents why humanitarian or community-based interventions may be preferable to coercive responses (for a discussion of this issue, see Arreguin-Toft, 2001; Downes, 2008). The relationship of counterinsurgents to different communities, their cultural distance from different communities within an area, and their understanding of community needs and interests may also condition their willingness to respond to incidents of violence as well as persistent conditions that cause opportunity for violence (McFate, 2005b). Although these concepts of formal authority, community norms, and capability are presented separately, it is certainly reasonable to believe that they operate in concert. For instance, in order for legal authority to have a meaning, it must be at least symbolically enforced and have some relevance to community norms (Meares & Kahan, 1998). While the discussion of capability primarily focused on formal guardianship, concepts of collective efficacy illustrate that in order for shared norms to be effective, they must have some force behind them, although this control might be very informal (Eck, 1994; Shaw & McKay, 1942). In the long term, in order for force to be effective, it must have some sort of normative or legal appeal (ArreguinToft, 2001; Bayley, 2001). It is also important to distinguish this paper’s discussion of the prevalence of authority, norms, and organizational capability throughout an environment from larger issues of political and cultural context. As Heyer (in press) notes, one of the challenges of implementing police reform in developing nations is that the local political and cultural context does not support democratic models of policing. This implies that models of crime and crime prevention must be fundamentally shaped by local context and that many intervention strategies that have been developed in Western, democratic nations are not generalizable. The perspective taken here, however, assumes that regardless of the content of political and cultural forces, the effects of these forces will vary across places. Further, these forces will influence similar places in similar and predictable ways. While important, a larger understanding of the political and social characteristics of places will not assist us with understanding how opportunities for crime or control diffuse, for instance, throughout a complex urban environment. The research that exists concerning crime, social order, and places generally suggests that there will be consistencies in the effects of opportunity for crime and collective capacity for

342

B. Cave

self-regulation on violence and civil order in places that may hold true across cultural contexts.

Conclusion The relevance of opportunity at places for counterinsurgents may not be immediately discernable. Counterinsurgents may use any force with which they are equipped to deliver, in any place that they are capable of accessing, in ways that deny or restrict opportunity for insurgents to act in places. If we accept the model of situational intervention posed by routine activities theory for counterinsurgents, then we have to make a distinction between “using force” and “providing guardianship.” However formal or informal the process that establishes it, legitimacy is what makes a guardian, and legitimate control over coercion is what makes a government. When government control is weakened, as during an insurgency, the question of who may use which types of coercion under which circumstances is called into question. Inasmuch as the prevention of violence can only be achieved through the intervention of legitimate guardians, then engaging in police activities without being attentive to these variations in perceptions about legitimacy may not only waste resources but could also backfire and increase violence. It is not possible to survey the public at every incident where a counterinsurgent intervenes to find out what bystanders think about the legitimacy of the action that occurred and the agency that performed it. However, it may be possible to predict, on average, the influences of various social and ecological features on patterns of perceptions about a counterinsurgent’s activities and hypothesize about the potential for guardianship at a location. A counterinsurgent may then attempt to manipulate these factors or simply take into account the fact that in places where the state of authority and norms are unstable, perceptions of guardianship will be unstable, and reactions to interventions will be less predictable. The long-term outcomes from situational interventions – including responses like defiance – may not be predictable (Sherman, 1993). The extent to which situational intervention can enhance perceptions of legitimacy of either a social control agency or a government more generally over the long term is still open to debate (although, see Tyler, 1990, 2004). Research certainly shows, however, that actions taken on the street level can negatively affect the public’s perceptions of the police (e.g., see Mastrofski, Reisif, & McClusky, 2002; Tyler, 2011). To help street-level practitioners intervene effectively (or at least do no harm), formal control agencies that are involved in a counterinsurgency must give those practitioners relevant information about context that is directly related to the scale of geography where they operate, and can inform their expectations of the outcomes of their activities. This is why thinking about factors such as authority, community norms, and capacity to intervene in terms of very small units of geography is valuable, particularly in an urban environment where these may vary widely.

14

Counterinsurgency and Criminology: Applying Routine Activities Theory…

343

In sum, theories about situations, places, and crime relate the opportunity for individuals to commit crime with the environment. The existing research about insurgent behavior seems to indicate that, like those committing crime, many insurgents are responsive to variations in opportunity. By taking a theory-driven, scientific approach to situations, counterinsurgents may identify measures of varying opportunity at specific places so as to develop interventions to prevent insurgent violence. But much more knowledge is required before the extent of influence of opportunity on insurgent violence can really be understood. Furthermore, opportunity does not solely influence the behavior of insurgents. Counterinsurgents may also have an “opportunity structure” (Clarke, 1997) that shapes their ability to effectively intervene across geographic space in ways that support a central government’s claim to legitimate coercive power. This structure of opportunity likely varies in terms of perceptions of the community about the authority of the counterinsurgent, the strength and type of norms of behavior in particular communities, and the actual capability of counterinsurgents to observe, detect, and respond to insurgent activity. Very little research, however, has focused on establishing how specific contextual factors influence the capacity of counterinsurgents to successfully exert different types of control at specific places. To develop a situational theory for counterinsurgency, researchers need specific data at small units of analysis about the location of insurgent violence, the activities of counterinsurgents, urban form, and about citizen perceptions of counterinsurgent organizations, government, and other agencies involved in insurgent conflicts. As Ramsey (2011) discusses, survey data may help counterinsurgents understand the distinctions that the public makes about state and non-state actors who are involved in insurgency and counterinsurgency. If surveys can be conducted in a manner that allows for fine distinctions of opinion across spaces (or determine if such distinctions exist), then this information may be of critical interest to the streetlevel practitioner, who may then be able to intervene in ways that address specific perceptions and concerns held by the community. Such data and theory are of little use unless they inform practice. Evaluation research suggests that the process for implementing research matters. Criminologists have proposed that crime prevention agencies should use a scientific, evidencebased model to determine whether practices are effective (Sherman, 1998). According to this model, practitioners use high-quality research to identify the types of interventions that are supported by evidence and implement them in a way that allows the effectiveness of interventions to be tested. This evidence-based approach should be (1) theory driven, (2) should clearly establish on what basis interventions are considered to be effective or ineffective, and (3) establish reciprocity between practice and theory. These same precepts can be applied to counterinsurgency studies. With regard to theory, the purpose of theory is not only to serve as a basis for developing hypotheses about how one intervention or another may work and then testing a hypothesis in the field. A base of theory also allows researchers and practitioners to build a base of evidence that can then be applied to other problems and conditions. One of the

344

B. Cave

major problems of counterinsurgency is that the lessons learned from previous conflicts are frequently misunderstood or misapplied to current conflicts (Cassidy, 2004). Theory can help to reduce these misunderstandings by ensuring that there is a logical framework under which particular types of analysis and intervention can be clearly understood. By developing interventions that are theory based, counterinsurgents can determine not only “what works” but also the types of interventions that have proven ineffective at reducing violence. Crime control research is rife with examples of well-intentioned and well-executed interventions that actually made crime problems worse (McCord, 2003). Counterinsurgents need to understand not only the circumstances under which interventions are successful but also when they are unsuccessful. A systematic method of developing and testing interventions is essential to produce this information. Furthermore, developing and testing specific hypothesis discredits ideas that are not very useful and prevents them from being continuously recycled from conflict to conflict. The relationship between theory and practice should not be a static one. In crime prevention and criminal justice, the movement toward evidence-based practices is characterized by developing a set of interventions that are based on theoretical explanations for crime that are then tested by practitioners. This ensures that theory remains relevant to real-world problems and that practitioners are operating based on the most rigorous research on crime prevention and causation. One could certainly argue that the “state of the art” for research in counterinsurgency should be moving toward theories and research methodologies that more precisely define the relationship between risks of violence, intervention, and outcomes. Although the majority of day-to-day U.S. counterinsurgency operations are carried out by military personnel operating on the street level, the information that does exist about the effects of military intervention on insurgent violence and the establishment of political authority is of relatively low methodological rigor, and focused on relatively large units of analysis such as neighborhoods or cities. Scholars who are interested in counterinsurgency should be moving toward theoretical models and methodologies that analyze counterinsurgency in ways that allow for valid judgments about the effectiveness of particular programs and interventions on the part of counterinsurgents (see Weisburd, 2003, for a discussion of this issue in crime and criminal justice). In military operations like those the United States is engaged in Iraq and Afghanistan, opportunities for this type of research is likely unfortunately more plentiful than not. A decade of conflict has illustrated that some counterinsurgency interventions might be effective, but others might be extremely costly failures. Given the cost in terms of lives of civilians, military personnel, and others who are involved in these conflicts, learning to more precisely identify effective and ineffective practices should be a priority for research in this area. By taking a situational perspective to questions of insurgent violence and counterinsurgent response, those who are interested in counterinsurgency may be able to make these distinctions more confidently, even in relatively complex environments.

14

Counterinsurgency and Criminology: Applying Routine Activities Theory…

345

References Arreguin-Toft, I. (2001). How the weak win wars: A theory of asymmetric conflict. International Security, 26(1), 93–128. Bayley, D. H. (2001). Democratizing the police abroad: What to do and how to do it. Washington: U.S. Department of Justice, National Institute of Justice. Bayley, D. H. (2005). Changing the guard: Developing democratic police abroad. New York: Oxford University Press. Behlendorf, B., LaFree, G., & Legault, R. (2010). Spurring them on: Spatio-temporal dependency and the cycle of terrorist violence. Paper presented at the annual meeting of the ASC annual meeting, San Francisco. Belasco, A., Else, D. H., Lindsay, B. R., Margesson, R., Nakamura, K. H., Taft-Mroales, M., et al. (2010). FY2010 supplemental for wars, disaster assistance, Haiti relief, and court cases. Washington: Congressional Research Service. Retrieved July 15, 2011, from http://www.dtic. mil/cgibin/GetTRDoc?AD=ADA523595&Location=U2&doc=GetTRDoc.pdf. Berman, E., Shapiro, J. N., & Felter, J. H. (2008). Can hearts and minds be bought? The economics of counterinsurgency in Iraq. NBER Working Paper Series No. 14606. Birkbeck, C., & Lafree, G. (1993). The situational analysis of crime and deviance. Annual Review of Sociology, 19, 113–137. Bittner, E. (2005). Florence Nightingale in pursuit of Willie Sutton: A theory of police. In T. Newburn (Ed.), Policing: Key readings (pp. 150–172). Collumpton: Willan Publishing. (Original work published 1983). Booth, J. A., & Richard, P. B. (1998). Civil society, political capital, and democratization in Central America. The Journal of Politics, 60(3), 780–800. Braga, A. A. (2002). Problem-oriented policing and crime prevention. Monsey: Criminal Justice Press. Braga, A. A. (2005). Hot spots policing and crime prevention: A systematic review of randomized controlled trials. Journal of Experimental Criminology, 1, 317–342. Braga, A. A. (2007). The effects of hot spots policing on crime. The Campbell Collaboration. Retrieved June 10, 2010, from www.campbellcollaboration.org/lib/download/117/. Braga, A. A., & Bond, B. J. (2008). Policing crime and disorder hot sports: A randomized controlled trial. Criminology, 46, 577–607. Braga, A. A., Kennedy, D. M., Waring, E. J., & Piehl, A. M. (2001). Problem-oriented policing, deterrence, and youth violence: An evaluation of Boston’s operation ceasefire. Journal of Research in Crime and Delinquency, 38(3), 195–225. Braga, A. A., Weisburd, D., Waring, E. J., Green Mazerolle, L., Spelman, W., & Gajewski, F. (1999). Problem-oriented policing in violent crime places: A randomized controlled experiment. Criminology, 37, 541–580. Braga, A. A., & Weisburd, D. L. (2010). Policing problem places: Crime hot spots and effective crime prevention. Oxford: Oxford University Press. Brantingham, P. L., & Brantingham, P. J. (1993). Environment, routine and situation: Towards a pattern theory of crime. In R. V. Clarke & M. Felson (Eds.), Routine activities theory and rational choice. Advances in criminological theory: Vol. 5 (pp. 259–288). New Brunswick: Transaction Publishers. Brantingham, P. L., & Brantingham, P. J. (1995). Criminality of place: Crime generators and crime attractors. European Journal of Criminal Policy and Research, 3, 5–26. Brantingham, P. L., & Brantingham, P. J. (2008). Crime pattern theory. In R. Wortley & L. Mazerolle (Eds.), Environmental criminology and crime analysis (pp. 78–93). Collumpton, Devon: Willan Publishing. Brophy-Baermann, B., & Conybeare, J. A. (1994). Retaliating against terrorism: Rational expectations and the optimality of rules versus discretion. American Journal of Political Science, 38(1), 196–210. Bursik, R. J. (1988). Social disorganization and theories of crime and delinquency: Problems and prospects. Criminology, 26, 519–551.

346

B. Cave

Bynam, D. (2006). Do targeted killings work? Foreign Affairs, 85(2), 95–111. Cassidy, R. M. (2004). Back to the street without joy: Counterinsurgency lessons from Vietnam and other small wars. Parameters, Summer, 73–83. Clarke, R. V. (1980). “Situational” crime prevention: Theory and practice. British Journal of Criminology, 20(2), 136–147. Clarke, R. V. (1992). Situational crime prevention: Successful case studies. Albany: Harrow & Heston. Clarke, R. V. (1995). Situational crime prevention. Crime and Justice: A Review of Research, 19, 91–150. Clarke, R. V. (1997). Situational crime prevention: Successful case studies (2nd ed.). Albany: Harrow & Heston. Clarke, R. V., & Felson, M. (1993). Routine activity and rational choice. Advances in criminological theory: Vol. 5. New Brunswick: Transaction Publishers. Clarke, R. V., & Harris, P. M. (1992). Auto theft and its prevention. Crime and Justice: A Review of Research, 16(1), 1–54. Clarke, R. V., & Homel, R. (1997). A revised classification of situational crime prevention techniques. In S. P. Lab (Ed.), Crime prevention at the cross roads (pp. 17–27). Cincinnati: Anderson. Clarke, R. V., & Newman, G. R. (2006). Outsmarting the terrorists. Westport: Praeger Security International. Clarke, R. V., & Weisburd, D. (1994). Diffusion of crime control benefits: Observations on the reverse of displacement. In R. V. Clarke (Ed.), Crime prevention studies: Vol. 2. Monsey: Criminal Justice Press. Cloward, R. A., & Ohlin, L. E. (1960). Delinquency and opportunity: A theory of delinquent gangs. New York: The Free Press. Cohen, A. K. (1955). Delinquent boys: The culture of the gang. Glencoe: The Free Press. Cohen, L., & Felson, M. (1979). Social change and crime rate trends: A routine activity approach. American Sociological Review, 44(4), 588–608. Collier, P., & Hoeffler, A. (2004). Greed and grievance in civil war. Oxford Economic Papers, 56, 563–595. Cornish, D. B., & Clarke, R. V. (1986). The reasoning criminal: Rational choice perspectives on offending. New York: Springer. Cornish, D. B., & Clarke, R. V. (1987). Understanding crime displacement: An application of rational choice theory. Criminology, 25(4), 933–947. Council, N. R. (2004). Fairness and effectiveness in policing: The evidence. In W. Skogan & K. Frydl (Eds.), Committee to review research on police policy and practices. Committee on Law and Justice, Division of Behavioral and Social Sciences and Education. Washington: National Academies Press. Deflem, M., & Sutphin, S. (2006). Policing post-war Iraq: Insurgency, civilian police, and the reconstruction of society. Retrieved June 10, 2010, from http://www.cas.sc.edu/socy/faculty/ deflem/ziraqpolice.pdf. Downes, A. B. (2008). Targeting civilians in war. New York: Cornell University Press. Draa, R. M. (2008). Thinks like a cop, fights like a Marine. Quantico: Marine Corps Command and Staff College. Drapeau, M. D., Hurley, P. C., & Armstrong, R. E. (2008). So many zebras, so little time: Ecological models and counterinsurgency operations. Defense Horizons, 62, 1–8. Dugan, L., Lafree, G., & Piquero, A. R. (2005). Testing a rational choice model of airline hijackings. Criminology, 43, 1031–1066. Durkheim, E. (1893). The division of labor in society (G. Simpson, Trans.). Glencoe: Free Press. Durlauf, S. N., & Nagin, D. S. (2011). Imprisonment and crime: Can both be reduced? Criminology and Public Policy, 10(1), 55–62. Eck, J. E. (1994). Drug markets and drug places: A case-control study of the spatial structure of illicit drug dealing. Doctoral dissertation, University of Maryland, College Park. Eck, J. E. (2002). Preventing crime at places. In L. W. Sherman, D. Farrington, B. Welsh, & D. L. MacKenzie (Eds.), Evidence-based crime prevention (pp. 241–294). New York: Routledge.

14

Counterinsurgency and Criminology: Applying Routine Activities Theory…

347

Eck, J. E. (2003). Police problems: The complexity of problem theory, research, and evaluation. Crime prevention studies: Vol. 15 (pp. 79–113). Monsey: Criminal Justice Press. Eck, J. E., Clarke, R. V., & Guerette, R. T. (2007). Risky facilities: Crime concentration in homogenous sets of establishments and facilities. In G. Farrell, K. J. Bowers, S. D. Johnson, & M. Townsley (Eds.), Imagination in crime prevention. Cullompton: Willan Publishing. Eck, J. E., & Wartell, J. (1997). Reducing crime and drug dealing by improving place management: A randomized experiment. Report to the U.S. National Institute of Justice. San Diego: San Diego Police Department. Eck, J. E., & Weisburd, D. (1995). Crime and place. Crime prevention studies: Vol. 4. Monsey: Criminal Justice Press. Enders, W., & Sandler, T. (1993). The effectiveness of antiterrorism policies: A vector-autoregression-intervention analysis. The American Political Science Review, 87(4), 829–844. Enders, W., & Sandler, T. (2000). Is transnational terrorism becoming more threatening? Journal of Conflict Resolution, 44, 307–332. Enders, W., Sandler, T., & Cauley, J. (1990). UN conventions, terrorism, and retaliation in the fight against terrorism: An econometric evaluation. Terrorism and Political Violence, 2(1), 83–105. Farrell, G. (1995). Preventing repeat victimization. In M. Tonry & D. P. Farrington (Eds.), Building a safe society: Strategic approaches to crime prevention, crime and justice (pp. 469–534). Chicago: University of Chicago Press. Farrell, G. (2006). Progress and prospects in prevention of repeat victimization. In N. Tilley (Ed.), Handbook of crime prevention and community safety. Cullompton: Willan Publishing. Farrell, G., & Pease, K. (2008). The sting in the British crime survey tail: Multiple victimizations. In M. G. Maxfield & M. Hough (Eds.), Surveying crime in the 21st century. Cullompton: Willan Publishing. Fawcett, G. S. (2009). Cultural understanding in counterinsurgency: Analysis of the human terrain system. Fort Leavenworth, Kansas: United States Army Command and General Staff College. Felson, M. (1986). Linking criminal choices, routine activities, informal control, and criminal outcomes. In D. B. Cornish & R. V. Clarke (Eds.), The reasoning criminal: Rational choice perspectives on offending (pp. 119–128). New York: Springer. Felson, M. (1995). Those who discourage crime. In J. E. Eck & D. Weisburd (Eds.), Crime prevention studies: Vol. 4. Crime and place (pp. 53–66). Monsey: Criminal Justice Press. Felson, M. (1997). Reconciling Hirschi’s 1969 control theory with the general theory of crime. In S. P. Lab (Ed.), Crime prevention at a crossroads. Cincinnati: Anderson. Felson, M. (2002). Crime and everyday life (3rd ed.). Thousand Oaks: Sage and Pine Press. Felson, M. (2006). Crime and nature. Thousand Oaks: Sage. Felson, M. (2008). Routine activity approach. In R. Wortley & L. Mazerolle (Eds.), Environmental criminology and crime analysis (pp. 70–77). Cullompton, Devon: Willan Publishing. Flynn, M. T., Pottinger, M., & Batchelor, P. D. (2010). Fixing intel: A blueprint for making intelligence relevant in Afghanistan. Retrieved March 8, 2010, from http://www.cnas.org/files/ documents/publications/AfghanIntel_Flynn_Jan2010_code507_voices.pdf. Galula, D. (1964). Counterinsurgency warfare: Theory and practice. Westport: Praeger Security International. Global Security.org. (2009). OIF Iraq significant activities. Retrieved January 31, 2011, from http://www.globalsecurity.org/military/ops/iraq_sigacts.htm. Gompert, D. C., Kelly, T. K., Lawson, B. S., Parker, M., & Colloton, K. (2009). Reconstruction under fire: Unifying civil and military counterinsurgency. Santa Monica: RAND. Groff, E., Weisburd, D., & Morris, N. A. (2009). Where the action is at places: Examining spatiotemporal patterns of juvenile crime at places using trajectory analysis and GIS. In D. Weisburd, W. Bernasco, & G. J. N. Bruinsma (Eds.), Putting crime in its place: Units of analysis in geographic criminology (pp. 61–86). New York: Springer. Guerette, R. T., & Bowers, K. J. (2009). Assessing the extent of crime displacement and diffusion of benefits: A review of situational crime prevention evaluations. Criminology, 47(4), 1331–1368.

348

B. Cave

Gurr, T. R. (1970). Why men rebel. Princeton: Princeton University Press for the Center for International Studies, Princeton University. Henriksen, T. H. (2010). Afghanistan, counterinsurgency, and the indirect approach. Hurlburt Field: Joint Special Operations University. Heyer, G. (In press). The role of civilian police in peacekeeping 1999. Washington, DC: Police Foundation. Hobbes, T. (1651). Leviathan or the matter, form, & power of a commonwealth ecclesiastical and civil. Retrieved July 15, 2011, from the Project Gutenberg E-book archive: http://www.gutenberg. org/files/3207/3207-h/3207-h.htm. Hoffman, F. G. (2007). Neo-classical counterinsurgency? Parameters, Summer, 71–87. Hope, T. (1994). Problem-oriented policing and drug-market locations: Three case studies. In R. V. Clarke (Ed.), Crime prevention studies: Vol. 2 (pp. 5–32). Monsey: Criminal Justice Press. Horowitz, D. L. (1985). Ethnic groups in conflict. Berkeley: University of California. Hunter, R. D., & Jeffrey, C. R. (1992). Preventing convenience store robbery through environmental design. In R. V. Clarke (Ed.), Situational crime prevention: Successful case studies. Guilderland: Harrow and Heston. Huntington, S. P. (1968). Political order in changing societies. New Haven: Yale University Press. Johnson, S. D., & Braithwaite, A. (2009). Spatio-temporal modelling of insurgency in Iraq. In J. D. Frielich & G. Newman (Eds.), Reducing terrorism through situational crime prevention (pp. 9–32). New York: Criminal Justice Press. Joint Publication 3-24. (2009). Counterinsurgency operations. Retrieved March 13, 2011, from http://www.dtic.mil/doctrine/new_pubs/jp3_24.pdf. Joint Publication 3-26. (2009). Counterterrorism. Retrieved March 13, 2011, from http://www. dtic.mil/doctrine/new_pubs/jp3_26.pdf. Jonas, E. G., & Ward, K. (2007). Introduction into a debate on city-regions: New geographies of governance, democracy, and social reproduction. International Journal of Urban and Regional Research, 31, 169–178. Kagan, F. W. (2006). A plan for victory in Iraq. The Weekly Standard. Retrieved February 1, 2011, from http://www.aei.org/docLib/20060714_KaganArticle.pdf. Keele, L. (2007). Social capital and the dynamics of trust in government. American Journal of Political Science, 51(2), 241–254. Kelly, J. D., Jauregui, B., Mitchell, S. T., & Walton, J. (2010). Anthropology and global counterinsurgency. Chicago: University of Chicago Press. Kennedy, L. W., Caplan, J. M., & Piza, E. (2010). Risk clusters, hot spots, and spatial intelligence: Risk terrain modeling as an algorithm for police resource allocation strategies. Journal of Quantitative Criminology (online first). http://www.springerlink.com.mutex.gmu.edu/content/ n77546101pj45764/fulltext.pdf. Kilcullen, D. (2009). The accidental guerilla: Fighting small wars in the midst of a big one. New York: Oxford University Press. Kipp J., Grau, L., Prinslow, K., & Smith, D. (2006). The human terrain system: A CORDS for the 21st century. Military Review, September–October, 8–15. Klinger, D. A. (2004). Environment and organization: Reviving a perspective on the police. The Annals of the American Academy of Political and Social Science, 593(1), 119–136. Klockars, C. B. (1985). The idea of police. Thousand Oaks: Sage. Kochel, T. R. (2011, Online first). Constructing hot spots policing: Unexamined consequences for disadvantaged populations and for police legitimacy. Criminal Justice Policy Review. Retrieved July 15, 2011, from http://cjp.sagepub.com/content/early/2010/06/28/0887403410376233. abstract. Koper, C. (1995). Just enough police presence: Reducing crime and disorderly behavior by optimizing patrol time in crime hotspots. Justice Quarterly, 12(4), 649–672. Krulak, C. C. (1993). The strategic corporal: Leadership in the Three Block War. Retrieved March 10, 2010, from http://www.au.af.mil/au/awc/awcgate/usmc/strategic_corporal.htm.

14

Counterinsurgency and Criminology: Applying Routine Activities Theory…

349

Landes, W. M. (1978). An economic study of U.S. aircraft hijackings, 1961–1976. Journal of Law and Economics, 21, 1–31. Last, J. (1997). Human-induced ecological determinants of infectious disease. Ecosystem Health, 4(2), 83–91. Leighton, R. M., & Sanders, R. (1962). Insurgency and counterinsurgency: An anthology. Washington: Industrial College of the Armed Forces. Lum, C. (2009). Community policing or zero tolerance? Preferences of police officers from 22 countries in transition. British Journal of Criminology, 49, 788–809. Lum, C. (2010). Does the “race of places” influence police officer decision making? Final report for the DuBois Fellowship. Washington: National Institute of Justice. Lum, C., Kennedy, L., & Sherley, A. (2006). Are counter-terrorism strategies effective? The results of the Campbell Systematic Review on counter-terrorism evaluation research. Journal of Experimental Criminology, 2(4), 489–516. Lum, C., & Koper, C. (2011). Is crime prevention relevant to counter-terrorism? In B. Forst, J. Greene, & J. Lynch (Eds.), Criminologists on terrorism and homeland security. New York: Cambridge University Press. Lynch, J. P. (2011). Implications of opportunity theory for combating terrorism. In B. Forst, J. R. Green, & J. P. Lynch (Eds.), Criminologists on terrorism and homeland security (pp. 151–182). New York: Cambridge University Press. Macfarlane, A. (2002). Emile Durkheim and the riddle of the modern world. Cambridge: King’s College. Manning, P. K. (1977). Police work: The social organization of policing. Cambridge: MIT Press. Mao, T. (1937). On guerrilla warfare (S. Griffen, Trans.). Chicago: University of Illinois. Mastrofski, S. D., Reisif, M. D., & McCluskey, J. D. (2002). Police disrespect toward the public: An encounter-based analysis. Criminology, 40(3), 519–552. Mazerolle, L., & Ransley, J. (2005). Third party policing. Cambridge: Cambridge University Press. Mazerolle, L., Roehl, J., & Kadleck, C. (1998). Controlling social disorder using civil remedies: Results from a randomized field experiment in Oakland, California. Crime Prevention Studies, 9, 141–159. McCord, J. (2003). Cures that harm: Unanticipated outcomes of crime prevention programs. The Annals of the American Academy of Political and Social Science, 587(1), 16–30. McFate, M. (2005a). Anthropology and counterinsurgency: The strange story of their curious relationship. Military Review, March–April, 24–38. McFate, M. (2005b). The military utility of understanding adversary culture. Joint Forces Quarterly, 38, 42–48. Meares, T. L., & Kahan, D. M. (1998). Law and (norms of) order in the inner city. Law and Society Review, 32(4), 805–838. Nagl, J. (2002). Counterinsurgency lessons from Malaya and Vietnam: Learning to eat soup with a knife. Westport: Praeger. Newman, O. (1972). Defensible space: Crime prevention through urban design. New York: Macmillan. Ottaviano, G. I. P., & Puga, D. (1997). Agglomeration in the global economy: A survey of the “new economic geography.” Centre for Economic Performance, Discussion Paper No. 356. Pease, K. (1991). The Kirkhold Project: Preventing burglary on a British public housing estate. Security Journal, 2, 73–77. Pease, K. (1998). Repeat victimization: Taking stock (Crime Detection and Prevention Series No. Paper 90). London, UK: Home Office. Petraeus, D. (2006). Learning counterinsurgency: Observations from soldiering in Iraq. Military Review, Jan–Feb, 80–87. Ramsey, C. (2011). Public opinion research and evidence-based counterterrorism. In C. Lum & L. Kennedy (Eds.), Evidence-based counterterrorism policy. New York: Springer. Reiss, A. J. (1986). Co-offender influences on criminal careers. In A. Blumstein, J. Cohen, J. Roth, & C. Visher (Eds.), Criminal careers and career criminals: Vol. 2. Washington: National Academy Press.

350

B. Cave

Rengert, G., & Pelfrey, W., Jr. (1997). Cognitive mapping of the city center: Comparative perceptions of dangerous places. In D. Weisburd & T. McEwen (Eds.), Crime mapping and crime prevention (pp. 193–218). Monsey: Criminal Justice Press/Willow Tree Press. Reppetto, T. A. (1976). Crime prevention and the displacement phenomenon. Crime & Delinquency, 22, 166–177. Reynald, D. M. (2010). Guardians on guardianship: Factors affecting the willingness to supervise, the ability to detect potential offenders, and willingness to intervene. Journal of Research in Crime and Delinquency, 47(3), 358–390. Ronceck, D. W. (1981). Dangerous places: Crime and residential environment. Social Forces, 60(1), 74–96. Rosenbaum, D. P. (2006). The limits of hot spots policing. In D. Wesburd & A. A. Braga (Eds.), Police innovation: Contrasting perspectives. Cambridge: Cambridge University Press. Sambanis, N. (2003). Expanding economic models of civil war using case studies. New Haven: Yale. Sampson, R. J. (1997). Collective regulation of adolescent misbehavior: Validation results from eighty Chicago neighborhoods. Journal of Adolescent Research, 12, 227–244. Sampson, R. J. (1999). What “community” supplies. In R. F. Ferguson & W. T. Dickens (Eds.), Urban problems and community (pp. 241–292). Washington: Brookings Institution Press. Sampson, R. J. (2004). Neighborhood and community: Collective efficacy and community safety. New Economy, 11, 106–113. Sampson, R. J., & Bartusch, D. J. (1998). Legal cynicism and (subcultural?) tolerance of deviance: The neighborhood context of racial differences. Law and Society Review, 32, 777–804. Scott, J. C. (1985). Weapons of the weak: Everyday forms of peasant resistance. New Haven: Yale University Press. Sepp, K. I. (2005). Best practices in counterinsurgency. Military Review, May–June, 8–12. Shafer, D. M. (1988). The unlearned lessons of counterinsurgency. Political Science Quarterly, 103(1), 57–80. Shaw, C., & McKay, H. (1942). Juvenile delinquency and urban areas. Chicago: University of Chicago Press. Sherman, L. W. (1993). Defiance, deterrence, and irrelevance: A theory of criminal sanction. Journal of Research in Crime and Delinquency, 30(4), 445–473. Sherman, L. W. (1998). Evidence-based policing. Ideas in American policing series. Washington, DC: The Police Foundation. Sherman, L. W., & Eck, J. E. (2002). Policing for crime prevention. In L. W. Sherman, D. P. Farrington, B. C. Welsh, & D. L. MacKenzie (Eds.), Evidence-based crime prevention (pp. 295–329). New York: Routledge. Sherman, L. W., Gartin, P., & Buerger, M. (1989). Hot spots of predatory crime: Routine activities and the criminology of place. Criminology, 27(1), 27–56. Sherman, L. W., & Weisburd, D. (1995). General deterrent effects of police patrol in crime hot spots: A randomized, controlled trial. Justice Quarterly, 12, 625–648. Skogan, W. G. (1990). Disorder and decline: Crime and the spiral of decline in American neighborhoods. New York: Free Press. Skogan, W. G., & Frydl, K. (2004). Fairness and effectiveness in policing: The evidence. Washington: National Academy of Science. Stark, R. (1987). Deviant places: A theory of the ecology of crime. Criminology, 25(4), 893–909. Taw, J., & Hoffman, N. (1994). The urbanization of insurgency: The potential challenge to U.S. army operations. Santa Monica: RAND. Taylor, R. B. (1997). Social order and disorder of streetblocks and neighborhoods: Ecology, microecology and the systemic model of social disorganization. Journal of Research in Crime and Delinquency, 33, 113–155. Taylor, R. B. (1999). Crime, grime, fear, and decline: A longitudinal look. Research in Brief. Washington: National Institute of Justice, U.S. Department of Justice.

14

Counterinsurgency and Criminology: Applying Routine Activities Theory…

351

Taylor, R. B., Harris, P. W., Jones, P. R., Weiland, D., Garcia, R. M., & McCord, E. S. (2009). Short-term changes in adult arrest rates influence later short-term changes in serious male delinquency prevalence: A time-dependent relationship. Criminology, 47(3), 657–697. Tillyer, M. S., & Eck, J. E. (2010). Getting a handle on crime: A further extension of routine activities theory. Security Journal. Retrieved July 15, 2011, from http://www.palgrave-journals.com/ sj/journal/vaop/ncurrent/abs/sj20102a.html Townsley, M., Johnson, S. D., & Ratcliffe, J. H. (2009). Space time dynamics of insurgent activity in Iraq. Security Journal, 21, 139–146. Tyler, T. R. (1990). Why people obey the law. New Haven: Yale University Press. Tyler, T. R. (2004). Enhancing police legitimacy. The Annals of the American Academy of Political and Social Science, 593(1), 84–99. Tyler, T. R. (2011). Toughness vs. fairness: Police policies and practices for managing the risk of terrorism. In C. Lum & L. Kennedy (Eds.), Evidence-based counterterrorism policy. New York: Springer. U.S. Department of Defense. (2010). Quadrennial defense review. Retrieved March 14, 2011, from http://www.defense.gov/qdr/QDR%20as%20of%2026JAN10%200700.pdf United States Department of the Army. (2008). Counterinsurgency field manual (FM 3–24). Chicago: University of Chicago Press. Weber, M. (1946). Politik als beruf. In H. H. Gerth & C. Wright Mills (Trans. and Eds.), From Max Weber: Essays in sociology (pp. 77–128). New York: Oxford University Press. (Original work published 1919) Weisburd, D. (2002). From criminals to criminal contexts: Reorienting crime prevention research and policy. In D. Weisburd & E. Waring (Eds.), Crime & social organization, Advances in criminological theory: Vol. 10 (pp. 197–216). New Brunswick: Transaction Publishers. Weisburd, D. (2003). Ethical practice and evaluation of interventions in crime and justice: The moral imperative for randomized trials. Evaluation Review, 27(3), 336–354. Weisburd, D. (2008). Place-based policing. Ideas in American Policing Series. Washington, DC: Police Foundation. Weisburd, D., Bushway, S., Lum, C., & Yang, S. M. (2004). Trajectories of crime at places: A longitudinal study of street segments in the city of Seattle. Criminology, 42(2), 283–321. Weisburd, D., Groff, E. R., & Yang, S. (2010). The criminology of place: Street segments and our understanding of the crime problem. Oxford: Oxford University Press. Weisburd, D., Maher, L., & Sherman, L. (1992). Contrasting crime general and crime specific theory: The case of hot-spots of crime. Advances in Criminological Theory, 4, 45–70. Weisburd, D., Morris, N. A., & Groff, E. R. (2009). Hotspots of juvenile crime: A longitudinal study of arrest incidents at street segments in Seattle, Washington. Journal of Quantitative Criminology, 25(4), 443–467. Weisburd, D., Telep, C. W., & Braga, A. A. (2010). The importance of place in policing: Empirical evidence and policy recommendations. Stockholm, Sweden: Swedish National Council for Crime Prevention, Information and Publications. Weisburd, D., Wycoff, L. A., Ready, J., Eck, J. E., Hinkle, J. C., & Gajewski, F. (2006). Does crime just move around the corner? A controlled study of spatial displacement and diffusion of crime control benefits. Criminology, 44, 549–592. Weisburd, D. W., & Green, L. (1995a). Policing drug hot spots: The Jersey City dug market analysis experiment. Justice Quarterly, 12, 711–735. Weisburd, D. W., & Green, L. (1995b). Measuring immediate spatial displacement: Methodological issues and problems. In J. Eck & D. Weisburd (Eds.), Crime and place (pp. 349–361). Monsey: Criminal Justice. Welsh, B. C., & Hoshi, A. (2002). Communities and crime prevention. In L. W. Sherman, D. P. Farrington, B. C. Welsh, & D. L. MacKenzie (Eds.), Evidence-based crime prevention. New York: Routledge.

Chapter 15

Toughness vs. Fairness: Police Policies and Practices for Managing the Risk of Terrorism Tom R. Tyler

When societies feel under threat they often respond by displays of force, using police and judicial power to project authority and threaten consequences for both those people who engage in non-normative actions motivated by the desire to undermine the state and those who are members of the communities which shelter and support those individuals. American reactions to recent terrorist actions directed against America, in particular the September 11, 2001 World Trade Center/Pentagon attacks followed this pattern and were initially highly punitive, involving active efforts to identify, find, and neutralize/punish those who had engaged in/supported or might engage in or support terrorism. Many of these initial actions were military and directed at foreign targets. However, there were also police efforts to identify, monitor, and intimidate people within the United States who were perceived as posing a security risk. In extreme cases people were detained or deported, while suspect people and communities were generally made the targets of widespread and myriad forms of police surveillance and other forms of social control. On one level these projections of force over suspect individuals and groups seem natural, reasonable, and necessary. When society is at risk those creating that risk need to be identified and stopped. While this may seem straightforward, recent research suggests that this strategy may nonetheless be nonoptimal and even counterproductive. This research-based analysis develops from an examination of the actions not of terrorists themselves but of those in the communities whose support terrorists rely upon for success. When people are seeking to overthrow existing governments through violence, as with terrorism, they need popular support that shelters and aids them. (“The guerrilla must move among the people as a fish swims in the sea,” Mao Tse-Tung.) Consequently, actions taken against terrorism are and should be linked to attempts to

T.R. Tyler (*) Department of Psychology, New York University, New York, NY, USA e-mail: [email protected]

C. Lum and L.W. Kennedy (eds.), Evidence-Based Counterterrorism Policy, Springer Series on Evidence-Based Crime Policy 3, DOI 10.1007/978-1-4614-0953-3_15, © Springer Science+Business Media, LLC 2012

353

354

T.R. Tyler

shape the behaviors of potentially supportive communities in ways that discourage those communities from providing aid and assistance to terrorists and terrorist organizations. This aid can range from providing money and other forms of material support to hiding terrorists and shielding their activities from the attention of the legal authorities to helping them to disseminate their ideological message. In recent anti-terror efforts those suspect communities have been the Muslim minority communities within targeted countries. While the majority of Muslims in countries such as the United States and the United Kingdom are supportive of the government and opposed to terrorism, ties of history and shared culture/religion link members of this community to many of those who commit acts of terror. Hence, suspicions about their potential willingness to report upon terrorists are natural, even if largely unfounded. How should such suspect communities be policed? This question can be addressed by focusing first upon identifying the goal that should guide policing efforts. That goal is to obtain community cooperation with police efforts to identify and prevent terror actions. The community includes all members of society, but especially those people within communities within which terrorists might hide, in this case the Muslim community. This goal should shape which police forces are given primary responsibility both for managing terror threats and for suggesting police policies and practices. A 2008 empirically based report from the RAND Corporation (Jones & Libicki, 2008) points to the value of cooperation when it argues that terrorism is most effectively approached as a policing problem since the local police are best able to build relationships with members of the communities that shelter terrorists and provide a potential source of individuals who can be recruited to become terrorists themselves. The RAND reports recommends: “Law enforcement officers should actively encourage and cultivate cooperation by building stronger ties with community leaders, including elected officials, civil servants, clerics, businessmen, and teachers, among others, and thereby enlist their assistance and support (p. x in the introduction).” In like manner, a former senior advisor to U.S. forces in Iraq recommends a focus on “local partnerships and local security forces that protect communities and guard against extremist presence” based on his combat experience (Kilcullen, 2009). A similar argument is made in the U.S. Army – Marine Corps Counterinsurgency Field Manual (2007), with a focus on community members, i.e. “the civilian population [is] … the deciding factor in the struggle, (p. xxv)” with the key issue being the ability to secure their “support.” Even in foreign areas of operation, in short, conventional military tactics, though still favored by some, are increasingly deemphasized in support of measures linked to domestic policing because cooperation is viewed as a key operational objective. In any account of policing against terrorism, the cooperation of local communities has significance in terms of results. In comparison to nonideological crime of the kind police generally address, terrorism is a relatively dispersed and infrequent phenomenon. Accurate and timely information to separate genuine threats from background noise therefore has special value. To the extent that terrorist groups either seek to recruit within or to hide within co-religionist communities, cooperation has the potential to provide police with information at lower cost than coercive

15

Toughness vs. Fairness: Police Policies and Practices for Managing…

355

or intrusive forms of intelligence gathering. The September 2001 attackers, for example, came into the sphere of indigenous Muslim-American communities (National Commission on Terrorist Attacks Upon the United States, 2004: 216–217). The goal of the present study is to assess quantitatively the potential gains/costs of intrusive intelligence gathering and to identify ways for police to maximize the voluntary provision of information by Muslim-American communities in circumstances of this kind.

Does Deterrence Work? It is often assumed that deterrence models generalize to the terrorism context and deterrence-based models “have long dominated both criminal justice and counterterrorist policies on responding to violence” (LaFree, Dugan, & Korte, 2009, p. 17). Deterrence theory suggests that people will cooperate with authorities when they view such actions as in their self-interest (Kalyvas, 2006). In the case of the threat of contemporary terrorism, self-interested motives might prompt people to cooperate for two reasons. First, they may anticipate rewards in terms of safety from identifying terrorists and ending a terrorist threat. Second, they may act in an effort to lower police intrusions into their community pre-emptively and avoid confrontations with police in their homes, on the streets, or in places of worship and community centers. On a personal level, people can cooperate to avoid pressure or punishment by the police, as well as for the financial rewards gained through cooperation. The first possible explanation for public cooperation with police is instrumental and grounded in a rational-choice model of human decision-making. In An Introduction to the Principles of Morals and Legislation, Bentham outlined an account of punishment as justified when the expected cost of breaking the law outweighed the expected benefits of the crime (Bentham, 1996). This account implies that people cooperate with law enforcement in expectation of net gains from compliance – e.g., increased safety – or of net losses from noncooperation – e.g., increased unwelcome and burdensome attention from law enforcement (Becker, 1976; Posner, 1985). Police following the instrumental model encourage cooperative behavior by making community residents’ cooperation more rewarding, for example by showing police are effective in fighting crime (Kelling & Coles, 1996), by punishing more rule breakers (Bayley & Mendelsohn, 1968; Nagin, 1998), or by directing unwelcome policing resources and attention toward uncooperative communities. An instrumental approach to policing based on a simple rationale of deterrence dominated British policing policy through the 1980s (Hough, 2007; Mclaughlin, Muncie, & Hughes, 2001). In the 1990s, instrumental logic motivated the British government’s model of “new public managerialism.” This approach to policing emphasized tangible results, targets, league tables, costing, and market testing of activities (McLaughlin et al.). In a quintessentially instrumental approach to crime control, new public managerialism “pric[ed]” offenses to calibrate optimal sanctions (Garland, 2001, p. 130).

356

T.R. Tyler

A countervailing view in the terrorism literature, however, warns of the potential of intrusive measures to stimulate terrorist recruitment and ideological estrangement in the targeted communities (Donohue, 2008). They can also prompt law-abiding individuals to withhold cooperation out of fear that suspicions, if reported, will trigger overreaction and unjust treatment of innocents (as can occur with ordinary crime, see Sherman, 1993). A recent study of Britain’s anti-terror campaign in Northern Ireland (LaFree et al., 2009), provides empirical confirmation of this risk. These authors identified six highly visible British interventions aimed at reducing terrorist violence in Northern Ireland from the 1970s on, and assessed whether each intervention diminished subsequent attacks or instead increased the frequency or intensity of terrorism. One of the six measures, a highly intrusive military maneuver, did have a deterrence effect. But two others had no statistically significant impact, suggesting that any deterrence gains were overwhelmed by backlash effects. More tellingly, two of the intrusive new deterrence-based policies resulted in significant increases in violence (also see Lum, Kennedy, & Sherley, 2006). Lafree et al. hypothesize that erroneous arrests and the adoption of internment without trial contributed to this backlash effect by undermining the legitimacy of British anti-terrorism efforts. Several studies conducted in Iraq have also found that perceived injustice on the part of US forces is a strong predictor of support for resistance among Iraqis (Fischer, Harb, Al-Sarraf, & Nashabe, 2008; Harb, Al Hafedh, & Fischer, 2006). As LaFree and Ackerman observe: “To the extent that government-based counterterrorism strategies outrage participants or energize a base of potential supporters, such strategies may increase the likelihood of further terrorist strikes” (2009, p. 15). Because of this, government management of terrorist threats may be as important as terrorism itself in determining future levels of violence (Kilcullen, 2009; McCauley, 2006; Sharp, 1973). These recent efforts notwithstanding, policing and military approaches to terrorism on the local level have not been unified in strategy or tactics. Different agencies and individuals vary in goals and behavior. Inconsistencies flow from ambivalence about the gains associated with various forms of policing against crime and against terrorism (see Bayley & Weisburd, 2009; Hasisi, Alpert, & Flynn, 2009; Oliver, 2006) and complexities in the relationship between local and federal law enforcement (Lum, Haberfield, Fachner, & Lieberman, 2009). These findings are in accord with the evidence on policing against ordinary crime, which in the domestic American context consistently suggests that deterrence effects on compliance and cooperation, when found, tend to be weak, and are associated with negative side effects (see Tyler, 2009, for a review). Their result is supported by Berrebi and Klor’s analysis of dynamic interactions between terrorism in the Israel-Palestine conflict and electoral outcomes, which found no correlation between more aggressive policies and reductions in terrorism levels (Berrebi & Klor, 2006). Notably, however, the sixth intervention studied by LaFree et al. raises an important caveat. That measure involved a shift from military methods to locally administered criminal justice procedures, treating captured terrorists as ordinary criminal suspects in an effort to delegitimate their cause. The detainees responded with a

15

Toughness vs. Fairness: Police Policies and Practices for Managing…

357

hunger strike to obtain a return to “prisoner-of-war” status. Further anti-British animosity resulted, and levels of violence subsequently increased. That finding underscores the point that both military and policing approaches alike have the potential to either lessen or increase subsequent community support for terrorism and terrorists.

Legitimacy as an Alternative Basis for Securing Cooperation The alternative model emphasizes self-regulatory, normative motivations. It posits that people comply and cooperate when they believe authorities are legitimate and entitled to be obeyed (Tyler, 2007, 2008). Research identifies strong evidence that when authorities are viewed as more legitimate, their rules and decisions are more likely to be accepted (Tyler, 2006a, 2006b). Research further links the legitimacy of institutions to the concept of procedural justice (Sunshine & Tyler, 2003; Tyler, 2006a; Tyler & Fagan, 2008). The fairness of police procedures depends, for example, on: the manner street stops are conducted; whether the police are neutral and transparent in their application of legal rules; whether they explain their actions and seek input from community members before making decisions; and whether they treat people with dignity and respect. Judgments about procedural justice have been found to influence the perceived legitimacy of law enforcement and thus to affect willingness to comply and to cooperate (Tyler, 2009). An extension of this approach to anti-terror policing would be based upon the view that “policy makers are involved in a battle with opponents over the fairness of governments and their policies” (LaFree & Ackerman, 2009, p. 15). To win this battle the government must win legitimacy by displaying fairness. The self-regulatory model has been widely supported in the studies of ordinary crime (see Tyler, 2009, for a review). In principle, the positive effects of legitimacy and procedural justice upon cooperation observed in ordinary law enforcement could well apply to policing against terrorism within Muslim-American populations after the September 2001 attacks. Three factors, however, counsel against taking that relationship for granted. First, terrorism differs from the crime that police typically address because terrorist acts are politically or ideologically motivated in ways that are distinct from the more idiosyncratic emotional motivations for crime (English, 2009; Katz, 1988). Traditionally conceived instrumental motivations (the desire for material gain; the fear of punishment) therefore may not be as significant for terrorists as for ordinary criminals (Varshney, 2003). Additionally, members of a co-religionist or co-ethnic community may share political or ideological views with those who commit acts of terrorism in a way that is not usually observed with criminal conduct. As a consequence, they may be unwilling to undermine terrorists due to feelings of solidarity. Terrorism thus involves distinct values that could interact in different ways with conceptions of legitimacy.

358

T.R. Tyler

Second, the core terrorism-related concern of policy makers is al Qaeda: an organization that adduces religious justifications for both its methods and goals (Kepel, 2002). Even if this explicit appeal to religiously grounded motivations is rarely successful, it raises the question whether the degree of religiosity among members of communities targeted by al Qaeda alters the effect of legitimacy or procedural justice on cooperation. Prior research suggests that moral and religious values can act to undermine the effect of legitimacy and procedural justice upon deference to government, with people less willing to defer to actions that are contrary to their values (Napier & Tyler, 2008; Skitka & Mullen, 2002). Historically religious authority has often been in conflict with the authority of the state, with people placing loyalty to their moral and religious values above duties to the law and the government (Kelman & Hamilton, 1989). Third, it is not safe to assume that legitimacy and procedural justice effects persist across different national cultures, or between a dominant national culture and immigrant sub-groups. On the contrary, while such effects are widespread, the literature suggests that they are not found in all societies. For example, studies conducted in China find the people do not react as strongly as in other cultures to whether or not procedures are fair (Brockner et al., 2001; Tyler, Lind, & Huo, 2000). Other studies suggest that the experience of procedural injustice associated with repressive governments is a major motivator of terrorism and political violence, since people find conventional means of participation blocked (Crenshaw, 1981, 1983; Krueger & Maleckova, 2003; Smelser, 2007; Voigt, 2005). Research suggests that after experiencing procedural injustice people become “radicalized” and focus upon violent means of achieving their goals. Because many recent immigrants in Muslim-American communities have spent significant time living in countries ruled by such governments, their background may affect their judgments of legitimacy or its importance for their behavior. Extended experience with repressive government is therefore another factor that could alter the connection between legitimacy on the one hand, and compliance and cooperation on the other. Tankebe raises a concern not unlike that just noted, suggesting that in some societies the procedural justice-legitimacy-cooperation model may not hold. His work in Ghana suggests that the legacy of colonialism has created a different relationship between the public and the police that is instrumentally based rather than linked to procedural justice or legitimacy (Tankebe, 2009a). Even in Ghana, however, procedural justice is linked to whether people supported vigilante violence (Tankebe, 2009b). A similar effect may be framed in terms of religion. It may be that a cluster of interlinked religious beliefs correlate with a distinctive conception of authority, and that conception might alter the effects of legitimacy on cooperation. If a substantial part of the Muslim-American community understands their faith tradition to impel an autocratic conception of religious authority, people with that belief may evaluate issues of fairness, participation, and equality of treatment differently and may be less affected by concerns about legitimacy and procedural justice when dealing with authorities (Davis & Silver, 2004). So, for example, people who do not expect to be consulted by their leaders before decisions are made may not feel that they are treated unfairly when they are not consulted by the police about police policies.

15

Toughness vs. Fairness: Police Policies and Practices for Managing…

359

The Effectiveness of Policing Strategies in the US and the UK Recent efforts to examine the role of the factors outlined in shaping the willingness of the members of the Muslim minority communities in the United States and the United Kingdom to cooperate with the police in identifying and combating terror threats suggest that the procedural justice of policing efforts is the primary factor shaping willingness to cooperate. This is true both in the United States (Tyler, Schulhofer, & Huq, 2010) and in the United Kingdom (Huq, et al. 2011). Both studies utilized the same approach. In each, a random sample of Muslims living within the appropriate geographical area (New York City or London) were identified and then interviewed over the telephone concerning their beliefs, attitudes, and cooperative behaviors. Two types of behaviors were of particular concern: the willingness to work with the police to combat terrorism and the willingness to alert the police to possible terror threats. In both settings the study suggested that Muslims were strongly influenced by the fairness of the procedures used by government. Two aspects of such procedure were found to matter. First, the fairness of the procedures through which the authorities formulated anti-terror policing strategies. This involved consultation with and consideration of the views of the Muslim community. Second, the fairness of the approaches used by the police in implementing such strategies. This included the fairness of police decision-making procedures and the fairness of the interpersonal treatment people received from the police. Both aspects of procedural fairness shaped cooperation. In contrast, many of the factors typically thought to shape cooperation had little influence. These included religious practices and identity; views about government policy; and instrumental factors such as the seriousness of terror and the effectiveness of the police. In general, none of the ethnic or religious issues that might set Muslims apart from other minority group members were found to be central to willingness to cooperate with the police. Instead, that willingness to cooperate was primarily influenced by judgments about how the police exercised their authority. Past studies of procedural justice have emphasized the mediating role of legitimacy, arguing that procedural justice gives the police legitimacy which motivates subsequent cooperation. In the studies it was noted that the connection between procedural justice and cooperation was more direct. In the United States legitimacy partially mediated the relationship between procedural justice and cooperation, while in England legitimacy shaped cooperation among those who identified with the United Kingdom, but not among those who did not. Hence, for this subsample the connection between the police and public behavior was more direct and not as strongly related to broader relations to the law and legal authority as has been found in the past studies of legitimacy in the context of ordinary crime. The results of these studies also suggest that people are willing to accept police intrusions into their community in efforts to combat terrorism. This is particularly true when the threat is viewed as serious and the police are considered effective. However, under no circumstances were people found to be willing to accept the

360

T.R. Tyler

police acting in procedurally unfair ways when exercising authority. Muslims were found to distinguish between reasonable and appropriate policing practices to combat terrorism, which were accepted, and procedural unfairness or harassment, which was not accepted and led to lower levels of legitimacy and cooperation.

Comparing Terrorism to Ordinary Crime In addition to examining the views of Muslims about cooperating in the fight against terrorism, it is important to consider whether terrorism is viewed differently than is ordinary crime. Huq, Tyler, and Schulhofer (in press). This study interviewed Muslims about ordinary crime and interviewed non-Muslims about terrorism. Its overall conclusion was that people reacted to the police and policing activities in similar ways irrespective of whether the police were policing against crime or against terrorism and/or whether those involved were Muslim or non-Muslim or White vs. other minorities (African-Americans; Hispanics). Although minor variations were identified, the overall conclusion was that people focus upon the fairness of the procedures used by the police (i.e., on procedural justice) irrespective of the type of crime involved. Further, this was true of people irrespective of their ethnic or cultural background. This finding echoes the earlier conclusion of Tyler and Huo (2002) that procedural justice is the key to the reactions of both White and minority group members to policing against ordinary crime. It suggests that the “war on terrorism” can be best conceptualized as similar to the war on crime and argues that the procedural justice models that have been successfully applied to crime should also be applied to terrorism and to anti-terror policing. Irrespective of whether the crime that the police are dealing with is ordinary crime or terror and irrespective of whether the community is largely White, minority, or Muslim, people’s primary reason for voluntary deference as well as cooperation is that they experience the police as acting procedurally fairly when creating and implementing policing policies (Huq et al., in press-a). A second important conclusion from the study of non-Muslims is that people who are not the targets of policing efforts nonetheless react negatively when they believe that the police harass or otherwise treat Muslims unfairly. The police lose legitimacy in the non-Muslim community when they are believed to be mistreating Muslims. This echoes the earlier finding the arena of anti-crime policing that the police lose legitimacy in the White community when they are viewed as racially profiling or otherwise mistreating minority group members (Tyler & Wakslak, 2004). In both cases the targets of policing are minority groups. In the case of ordinary crime the police often focus their attention upon poor and largely minority communities and engage in street stops and other forms of intrusion into the community in an effort to identify criminals and observe crimes. With terrorism similar actions are directed toward the members of the Muslim community. Whether such actions are undertaken depends in part upon how the police believe their behavior will be viewed in the broader community. Will that larger community be upset if they

15

Toughness vs. Fairness: Police Policies and Practices for Managing…

361

believe the police harass ethnic or religious minority group members? Research suggests that the larger community views the police as less legitimate when they believe that they engage in unfair practices toward minorities.

Summary There is a long history of societies reacting to perceived and real threats by directing force and sanctions toward suspect minorities groups. For example during World War II Japanese Americans were placed in camps to prevent them from being able to provide aid to the Japanese in their war against the United States. While these actions were extreme they were not different than many other actions taken against Germans and German organizations during both World Wars I and II. The arguments made here address this ongoing issue of how to deal with potentially dangerous minority groups in the context of recent actions toward the Muslim community in the wake of recent attacks on the United States. The literature reviewed supports several conclusions. First, deterrence strategies do not seem particularly effective in motivating cooperation in fighting terror. This parallels similar earlier finding in the arena of policing against crime (Tyler & Fagan, 2008). Second, procedural justice approaches are more effective as a basis for securing cooperation from the public. Third, people are willing to accept police intrusions and the police do not necessarily lose legitimacy by intervening into people’s lives and communities. However, the police need to intervene in procedurally fair ways. This includes consultation with the community about policing policies and implementation of those policies through fair procedures. Overall, these findings point to the need to rethink how the government and the police respond to the next emergency. Although it may be a natural and instinctive response to project force in response to threat, that response is nonetheless nonoptimal if the goal is to effectively respond to the external threat. Rather the police and other governmental authorities need to engage in the potentially less emotionally satisfying but more useful approach of building support among those populations that might be potential sources of support for terrorists. This involves active consultation with the members of those communities regarding how to respond to the threat and policies and practices that ensure that members of those communities experience fair and respectful anti-terror policing when dealing with police officers and other government agents.

References Bayley, D., & Mendelsohn, H. (1968). Minorities and the police. New York: Free Press. Bayley, D., & Weisburd, D. (2009). Cops and Spooks: The role of the police in counterterrorism. In D. Weisburd et al. (Eds.), To protect and to serve. New York: Springer.

362

T.R. Tyler

Becker, G. (1976). The economic approach to human behavior. Chicago: University of Chicago Press. Bentham, J. (1996). An introduction to the principles of morals and legislation. New York: Oxford University Press. Berrebi, C., & Klor, E. (2006). On terrorism and electoral outcomes. Journal of Conflict Resolution, 50, 899–925. Brockner, J., Ackerman, G., Greenberg, J., Gelfand, M. J., Francesco, A. M., Chen, Z. X., et al. (2001). Culture and procedural justice: The influence of power distance on reactions to voice. Journal of Experimental Social Psychology, 37, 300–315. Counterinsurgency Field Manual: U.S. Army Marine Corps. (2007). Chicago: The manual was reprinted as a book by the University of Chicago press. Crenshaw, M. (1981). The causes of terrorism. Comparative Politics, 13, 379–399. Crenshaw, M. (1983). Terrorism, legitimacy, and power. Middleton: Wesleyan University Press. Davis, D. W., & Silver, B. D. (2004). Civil liberties vs. security: Public opinion in the context of the terrorist attacks on America. American Journal of Political Science, 48, 28–46. Donohue, L. K. (2008). The cost of counterterrorism: Power, politics, and liberty. Cambridge: Cambridge University Press. English, R. F. (2009). Terrorism: How to respond. Oxford: Oxford University Press. Fischer, R., Harb, C., Al-Sarraf, S., & Nashabe, O. (2008). Support for resistance among Iraqi students. Basic and Applied Social Psychology, 30, 1–9. Garland, D. (2001). The culture of control: Crime and social order in contemporary society. Chicago: University of Chicago Press. Harb, C., Al Hafedh, A., & Fischer, R. (2006). Understanding support for resistance in Iraq. Paper, 18th congress of the international association of cross-cultural psychology, July 11–15, 2006. Spetses, Greece. Hasisi, B., Alpert, G. P., & Flynn, D. (2009). The impact of policing terrorism on society. In D. Weisburd et al. (Eds.), To protect and to serve. New York: Springer. Hough, H. (2007). Policing, new public management, and legitimacy in Britain. In T. R. Tyler (Ed.), Legitimacy and criminal justice: International perspectives. New York: Russell Sage Foundation. Huq, A. Z., Tyler, T. R., & Schulhofer, S., et al. (2011). Why does the public cooperate with law enforcement? The influence of the purposes and targets of policing. Psychology, Public Policy and Law, 17(3), 419–450. Huq, A. Z., Tyler, T. R., & Schulhofer, S. (in press). Mechanisms for eliciting cooperation in counterterrorism policing: Evidence from the United Kingdom. Journal of Empirical Legal Studies. Jones, S. G., & Libicki, M. C. (2008). How terrorist groups end: Lessons for countering al Qa’ida. Arlington: RAND. Kalyvas, S. N. (2006). The logic of violence in civil war. Cambridge: Cambridge University Press. Katz, J. (1988). Seductions of crime. New York: Basic Books. Kelling, G. L., & Coles, C. M. (1996). Fixing broken windows. New York: Touchstone. Kelman, H. C., & Hamilton, V. L. (1989). Crimes of obedience. New Haven: Yale. Kepel, G. (2002). Jihad: The trail of political Islam. Cambridge: Harvard. Kilcullen, D. (2009). The accidental guerrilla: Fighting small wars in the midst of a big one. New York: Oxford University Press. Krueger, A. B., & Maleckova, J. (2003). Education, poverty and terrorism. Journal of Economic Perspectives, 17, 119–144. LaFree, G., & Ackerman, G. (2009). The empirical study of terrorism: Social and legal research. Annual Review of Law and Social Science, 5, 347–374. LaFree, G., Dugan, L., & Korte, R. (2009). The impact of British counterterrorist strategies on political violence in Northern Ireland: Comparing deterrence and backlash models. Criminology, 47, 17–45.

15

Toughness vs. Fairness: Police Policies and Practices for Managing…

363

Lum, C., Haberfield, M., Fachner, G., & Lieberman, C. (2009). Police activities to counter terrorism. In D. Weisburd et al. (Eds.), To protect and to serve. New Jersey: Springer. Lum, C., Kennedy, L. W., & Sherley, A. J. (2006). Are counterterrorism strategies effective? The results of the Campbell systematic review on counter-terrorism evaluation research. Journal of Experimental Criminology, 2, 489–516. McCauley, C. (2006). Psychological issues in understanding terrorism and the response to terrorism. In B. Bongar, L. M. Brown, L. E. Beutler, J. N. Breckenridge, & P. G. Zimbardo (Eds.), Psychology of terrorism (pp. 13–31). Oxford: Oxford University Press. McLaughlin, E., Muncie, J., & Hughes, G. (2001). The permanent revolution: New labor, new public management and the modernization of criminal justice. Criminology and Criminal Justice, 1, 301–318. Nagin, D. S. (1998). Criminal deterrence at the onset of the 21st century. Crime and Justice, 23, 1–42. Napier, J., & Tyler, T. R. (2008). Does moral conviction really override concerns about procedural justice. Social Justice Research, 21, 509–528. Oliver, W. M. (2006). Homeland security for policing. Upper Saddle River: Prentice-Hall. Posner, R. (1985). An economic theory of the criminal law. Columbia Law Review, 85, 1193–1231. Sharp, G. (1973). The politics of non-violent actions. Boston: Porter Sargent. Sherman, L. (1993). Defiance, deterrence, and irrelevance: A theory of criminal sanction. Journal of Research in Crime and Delinquency, 30, 445–473. Skitka, L., & Mullen, E. (2002). Understanding judgments of fairness in a real-world political context: A test of the value protection model of justice reasoning. Personality and Social Psychology Bulletin, 28, 1419–1429. Smelser, N. J. (2007). The faces of terrorism: Social and psychological dimensions. Princeton: Princeton University Press. Sunshine, J., & Tyler, T. R. (2003). The role of procedural justice and legitimacy in shaping public support for policing. Law and Society Review, 37, 513–547. Tankebe, J. (2009a). Public cooperation with the police in Ghana: Does procedural fairness matter? Criminology, 47(4), 1265–1293. Tankebe, J. (2009b). Self-help, policing, and procedural justice: Ghanaian vigilantism and the rule of law. Law and Society Review, 43, 245–268. Tyler, T. R. (2006a). Psychological perspectives on legitimacy and legitimation. Annual Review of Psychology, 57, 375–400. Tyler, T. R. (2006b). Why people obey the law. Princeton: Princeton University Press. Tyler, T. R. (2007). Psychology and the design of legal institutions. Nijmegen: Wolf Legal Publishers. Tyler, T. R. (2008). Psychology and institutional design. Review of Law and Economics (symposium issue on Law and Social Norms). The Berkeley Electronic Press, 4(3), 801–887. Tyler, T. R. (2009). Legitimacy and criminal justice: The benefits of self-regulation. Ohio State Journal of Criminal Law, 7, 307–359. Tyler, T. R., & Fagan, J. (2008). Legitimacy and cooperation: Why do people help the police fight crime in their communities? Ohio State Journal of Criminal Law, 6, 231–275. Tyler, T. R., Lind, E. A., & Huo, Y. J. (2000). Cultural values and authority relations. Psychology, Public Policy, and Law, 6(4), 1138–1163. Tyler, T. R., & Huo, Y. J. (2002). Trust in the Law. N.Y.: Russell-Sage. Tyler, T. R., Schulhofer, S., & Huq, A. Z. (2010). Legitimacy and deterrence effects in counterterrorism policing: A study of Muslim Americans. Law and Society Review, 44, 365–402. Tyler, T. R., & Wakslak, C. (2004). Profiling and the legitimacy of the police: Procedural justice, attributions of motive, and the acceptance of social authority. Criminology, 42, 13–42. Varshney, A. (2003). Nationalism, ethnic conflict, and rationality. Perspectives on Politics, 1(1), 85–99. Voigt, S. (2005). Islam and the institutions of a free society. The Independent Review, 10, 59–82.

Part V

Epilogue

Chapter 16

The Next Steps: A Need for a Research Infrastructure for Evaluating Counterterrorism Cynthia Lum and Leslie W. Kennedy

This volume emphasizes a central theme in crime prevention research and public interventions: Interventions addressing issues of great public importance must still be evidence-based. Not only must we continue to add more to the growing research on terrorism etiology, but we also need to develop an equal level of interest in evaluating counterterrorism interventions. Unlike crime prevention more generally, there is a research lacuna in counterterrorism studies. Indeed, there are special challenges in advancing counterterrorism research that are different from crime prevention studies more generally, which contribute to the scant supply of counterterrorism research. Such challenges include rare and difficult-to-study units of analysis, an absence of knowledge about base rates, lack of access to government activities and information, biases in research and politics, and unresolved definitional debates. While all of these are compelling, the values of democratic governance, which include accountability, legitimacy, legality, safety, and cost-effectiveness of government actions, necessitate the development of a more evidence-based approach to programs and interventions, whether they focus on “everyday” crime or terroristic violence. Toward this end, a main goal of evidence-based counterterrorism policy is to expand the way we think about data, methods, and perspectives of counterterrorism evaluation research. Each chapter in this volume contributes individually and collectively toward this end. However, this outcome is only possible if a research infrastructure exists that supports and funds knowledge building, provides opportunities for new ideas to flourish, allows for access to collecting data, and demands high-quality

C. Lum (*) Center for Evidence-Based Crime Policy, Department of Criminology, Law & Society, George Mason University, Fairfax, VA, USA e-mail: [email protected] L.W. Kennedy Rutgers University School of Criminal Justice, Rutgers University, Newark, NJ, USA Rutgers Center for the Study of Public Security, Rutgers University, Newark, NJ, USA C. Lum and L.W. Kennedy (eds.), Evidence-Based Counterterrorism Policy, Springer Series on Evidence-Based Crime Policy 3, DOI 10.1007/978-1-4614-0953-3_16, © Springer Science+Business Media, LLC 2012

367

368

C. Lum and L.W. Kennedy

and open evaluation. In summarizing the contributions made by the papers in this volume, we first consider why knowledge from evaluation research in the area of counterterrorism is lacking. We then explore what a research infrastructure might need to support further scientific inquiry and evaluation in this important public policy arena. Drawing upon experiences from evidence-based policing, we reflect on lessons learned in that endeavor, and how they might be applied to counterterrorism.

Showing Effectiveness: The Lack of Evaluation Why is it that the public and policy makers settle for ineffective or unevaluated responses to problems? Part of the answer to this question resides in the general lack of awareness of how evaluation can correct erroneous assumptions about what are believed to be “effective” or “successful” interventions. To avoid risk and blame, policy makers often wish to be seen as responding aggressively to major social problems. There may be a misguided belief about the public’s expectation for immediate action, or its lack of interest or receptivity toward scientifically based approaches in assessing social programs. Also, ideologies often drive decisionmaking, which can overshadow appeals to science, reason, and facts. The time frame of the usefulness of an evaluation is connected to this avoidance of risk and blame in security, law enforcement, and counterterrorism efforts. Although thoughtful review, debate, contemplation, and evaluation may be seen as more feasible in long-standing policy initiatives such as those in social welfare or education, this is less the case in concerns like crime and terrorism, which generate fear and a sense of urgency. However, as we will argue, avoiding thoughtful assessment because of this urgency is a bad and dangerous counterterrorism or anti-crime habit. The stakes in these cases are so high in terms of how interventions create both public safety and government legitimacy, that ironically, to not know whether interventions “work” (or worse, whether they harm society further) damages the core intentions of the intervention. In some cases, avoiding evaluation is neither about avoiding risk and blame, or how urgency calls us to immediate action. There may also be an assertion on the part of counterterrorism practitioners that certain types of programs simply cannot – or should not be – subjected to scientific evaluation because of national security risks such assessment could create. This view presents at least two disturbing implications. First, it suggests that the type of intervention (countering terrorism) and its target (terrorists, violence) are so secretive or important that lower, or no, standards of assessment are acceptable. Although there are some very rare counterterrorism interventions – like killing Osama bin Laden – that will be out of the sight of the public eye or scientific scrutiny, most government actions for homeland security are fairly straightforward and nonsecretive. These include security at airports and other transportation hubs as discussed by Costigan and Lum et al.’s chapters, or efforts to reduce public support for terrorism as discussed by Ramsay and Cave. Counterterrorism financing and anti-money laundering are not so secretive that they

16

The Next Steps: A Need for a Research Infrastructure…

369

cannot be examined, as Passas advocates in his chapter. And, there is a long tradition of statistically examining the impact of interventions on terror attacks over time, using existing and new methods such as those developed by Dugan, Yang, CioffiRevillo, and Porter and colleagues. Indeed, there is much publically available data for evaluation as Sheehan, LaFree, Carter, and Chermak discuss. Even the bin Laden assassination can be evaluated for the effects of its consequences on future risk of terrorism events and backlash. This first implication of not evaluating because counterterrorism is “too important and therefore immune from accountability” is a dangerous and slippery slope in security democracies. This is especially problematic if this determination is not based in fact or reality, but in political ideology, unsupported assumptions or fears, or simply a security bureaucrat’s need for generating self importance by creating a mythology around their position. Secondly, a view that certain interventions are so important to avoid scrutiny misjudges the efficacy and usefulness of field applications of the tools of social science in counterterrorism, and fails to realize the consequences of using weaker evaluation approaches. Weisburd, Lum, and Petrosino (2001) found that there was an inverse relationship between the quality of methodological rigor in evaluations in criminal justice and the likelihood of a positive effect of an intervention. In other words, more positive and “successful” conclusions about interventions are found more often in evaluations using weaker designs, such as a before and after comparison of an intervention without an appropriate comparison group. The problem with weaker designs is that they threaten internal validity of studies – results are less believable and could be excessively optimistic or even opposite to reality. However, as many have advocated, there is compelling support for an increased role of scientific inquiry in crime policy, and the same precept holds true with counterterrorism. This view does not necessarily arise from scientists or the counterterrorism community, but from victims and their surrounding communities. Recently, concerned victim groups have demanded and received more information on government accountability regarding knowledge about terrorism and counterterrorism. The 9-11 Commission (National Commission on Terrorist Attacks Upon the United States, 2004), reviews by the Government Accountability Office (see U.S. GAO, 2007, 2009, 2010), and other assessments (see Home Office, 2006; Lum, Kennedy, & Sherley, 2006; National Research Council, 2010) have all called for more effective and comprehensive programs to respond to terrorism. The Department of Homeland Security’s Science and Technology Directorate has spearheaded a number of research efforts, including the START Center at the University of Maryland, as well as some research projects discussed in this volume. However, much more rigorous assessment is needed and across different sectors related to counterterrorism. No clear evaluation standard has emerged to guide the public discussion of the effectiveness of these counter-terrorism responses. Further as Lum et al. (2006) point out, the research supporting claims of terrorism intervention effectiveness is scarce, the results uneven, and the standards of assessment are low. Prior to the Lum et al. Campbell review of counterterrorism evaluations, only a few general appraisals of terrorism studies had been undertaken (see, e.g., Halkides, 1995; Hoffman, 1992; Miller, 1988; Romano, 1984; Schmid &

370

C. Lum and L.W. Kennedy

Jongman, 1988), but none were systematically focused on examining the state and quality of evaluation studies. This was true in spite of the proliferation of antiterrorism programs and policies. As a consequence, we were left with a limited understanding of the effectiveness of any of these programs. The authors throughout this volume provide a number of reasons why this might be the case and we summarize them below.

Scarcity, Discrepancy, and Silos of Data First, the rarity of terrorism events makes them difficult to study (as are the groups and individuals who commit terroristic violence). Unlike mainstream criminological research that can employ large-scale data sets compiled by police and corrections agencies containing frequently occurring events, terrorism happens less frequently and is not compiled with the same level of accuracy and standardization. This not only makes descriptions of the genesis of terrorism difficult, but it also complicates any effort to establish the responsiveness of terrorism trends to interventions. As LaFree and Sheehan emphasize in this volume, there is also a variety of data sources that use different definitions and standards for data collection. Information may also be hoarded in “data silos” that include detailed records of incidents that are not available to all researchers or even all law enforcement groups (an exception would be the openness of the START Center’s Global Terrorism Database). Sheehan’s comparison of datasets is an excellent exercise that can provide more clarity and exposure of what information is available and why. Furthermore, efforts to share information, such as the development of the Fusion Centers that Carter and Chermak write about, can also reduce scarcity, discrepancy, and hoarding of data. Data quality and increased awareness of problems may also directly result from sharing. Lessons learned from the development of the Federal Bureau of Investigation’s Uniform Crime Reporting (UCR) system is one case in point. Today’s voluntary submission to the UCR by almost all law enforcement agencies in the U.S. is testament to the useful role that data sharing plays in creating standards and more open data cultures for purposes of analysis of crime or evaluation of interventions. Whether data from new or other data sources, for example, the Nationwide Suspicious Activity Reporting Initiative1, tip lines, transportation hubs, or financial institutions will be available to researchers is a great uncertainty. However, such information is necessary not only to validate other information sources but to better understand how both citizens and law enforcement interpret what they believe to be terrorism activities, and the programs responding to those interpretations.

1

See http://nsi.ncirc.gov/.

16

The Next Steps: A Need for a Research Infrastructure…

371

Secrecy The evaluation of counter-terrorism programs confronts another problem related to availability and silos of data: secrecy. With concerns about potential offenders learning from interventions and developing new forms of attack, counter-terrorism planners and officials keep their programs hidden behind the barriers of secret classifications and cultures. As a result, implementation strategies and information about incidents and outcomes are difficult to compile for scientific purposes. Even some of the chapters published in this volume received close scrutiny prior to publication. However, security interventions, whether local or national, have one foot in governance and the other in society and therefore are reliant upon the stability of the give and take that characterizes the complex environment that modern democracies must negotiate. While transparency, trust, and accountability by law enforcement agencies can improve citizens’ tolerance of the collateral effects of crime prevention policies, such dynamics may be different in the case of counterterrorism. Government interests in national security may block these concerns from having significant effects, and citizens may agree to sacrifice transparency and accountability for security and secrecy in a high-fear, low information, environment. The obstacles that both the realities and the mythologies of secrecy present are not insurmountable. There continues to be a strong tradition of time series and longitudinal analysis in counterterrorism literature that can evaluate visible interventions using public information. Although information on the implementation of these programs is shielded from public access, the occurrence of them is not. Scientists know when a military campaign occurred, when metal detectors were put in place, or when special security measures were implemented around an embassy. Thus, much of the methodological literature, including those in this volume (see Dugan and Yang, Porter et al., Cioffi-Revillo and Rusnak et al.) focus on improving methods to conduct time series analysis of interventions or forecasting of future events. This method bias is an excellent example of how secrecy and scarcity of data and programs, whether warranted or not, have influenced the type of research topics and methods used. A case can also be made, we believe, that counterterrorism programs can be offered for public evaluation and scrutiny in ways that promote accountability and improve outcomes without compromising aspects of the counterterrorism efforts that must remain classified. Police evaluations were certainly able to proceed in this way through the use of human subjects protections and confidentiality agreements between researchers and agencies. Importantly, the development of evaluation research in policing since the 1970s has led to important discoveries about the ineffectiveness of commonly used interventions once assumed to be effective, such as rapid response to 911, random preventative patrol, reactive investigations, and drug abuse education programs (National Research Council, 2004). And, as the National Research Council (2004) report also shows, evaluations in the area of racial profiling and police discretion have also led to important reforms and increased awareness about practices that reduce government and police legitimacy. Providing evidence of cost-effectiveness could assist in gaining greater public investment in homeland security programs and

372

C. Lum and L.W. Kennedy

would be helpful in fine-tuning their implementation. Arguments rejecting assessment in the name of “national security” have to themselves be evaluated, especially when the implementation of programs requires public assistance.

Resistance to Experimentation at the Programmatic Level Another challenge to evidence-based counterterrorism policy is the unwillingness to engage in experimentation that might test differential program effects for actual security measures. Most evaluations that have been conducted were implemented ex post facto, leaving doubt about the real effects of treatment. Quasi-experimental methods such as ARIMA interrupted-time-series designs (see Campbell & Stanley, 1966; Cook & Campbell, 1979) have been used and can overcome the problem of ex-post facto assessment. However, this does not imply that attempts at experimentation should be abandoned. Researchers have to confront the real fear by practitioners and citizens that experimentation leaves some places or individuals unprotected because they are not receiving treatment. However, as with many experimentally tested policing interventions, “no treatment” may actually be the current status quo, which weakens such arguments. For example, hot spots policing, or concentrating police patrol in small locations with high levels of crime, has been evaluated in multiple randomized controlled trials (for a review, see Braga, 2005) compared with “business as usual” beat policing. In countering terrorism, “status quo” might mean the nonexistence of a new program or new technology that was not used in the first place but which is now being tested. For example, one might imagine evaluating a security measure such as randomly assigning behavior detection to passengers at airport checkpoints, but still subject everyone to the same second layer (i.e., metal detectors) that they would have gone through anyway. Full body scans might be implemented at randomly selected checkpoints, but after those selected go through, they are still required to go through a metal detector. Again, these examples show that there can be experimentally rigorous tests to check the effectiveness of the one layer of security without compromising security by implementing a second, pre-existing layer. Arguments that experimentation may inconvenience passengers by adding more time to their commute must also be assessed for validity. In a study of passengers at a major airport hub, Lum and colleagues2 found that the time through screening was often shorter than expected and fairly easy to negotiate. Indeed, almost all passengers approached had enough time to fill out a survey after they had gone through security. As with experimental evaluation in the social sciences more generally, contextual information that reveals the multitude of factors that influence events is impor2

See http://gemini.gmu.edu/cebcp/Briefings/LUMNew.pdf for a description of this unpublished research project.

16

The Next Steps: A Need for a Research Infrastructure…

373

tant to generate. As demonstrated in Rusnak et al.’s study on Turkey, factors that can be utilized to identify concentrations of high-risk locations for terrorism can be manipulated, and the relative effects of these factors can be judged over time, providing the bases for forecasting future events from something other than past events or through simulations.

Limited Relationships and Limited Perspectives Finally, the lack of a strong culture of evaluation of counterterrorism interventions may be due to a weak relationship between the intelligence communities and academic researchers with regard to examining data needs, developing analytical approaches, and implementing long-term evaluation programs. Yet there is much hope as shown in the history of police scholarship. Secrecy, scarcity of data, and resistance to evaluation and experimentation were overcome in studies of law enforcement through institutionalizing and developing a culture of positive interactions between researchers and the police. However, the successes that have been witnessed in police research have not occurred without great difficulties. At the core of this working relationship is a trust that is assumed in democratic theories of governance but requires nurturing in practice. Evaluation and research for any agency is a direct challenge to its legitimacy, transparency, authority, and accountability, and will likely not occur easily. Further confounding this relationship, as discussed by Ramsey, is the role of public opinion, which weighs heavily in all sorts of regimes and can also reinforce closed governance in this particular research area, depending on the citizenry’s level of fear. And, as Passas discusses, given that counterterrorism efforts often extend across borders with vastly different government systems, evidence collection becomes even more difficult in an international context.

Developing a Research Infrastructure for Evaluation All of these challenges point to the need for greater support for evaluation research in counterterrorism, in the form of more access, better understanding, increased research opportunities, and stronger receptivity to research by practice. These challenges in counterterrorism research are not new or unique to evidence-based crime policy but can be seen in many other sectors of society and governance, including in education, social welfare, and public health. The development of research agendas, regardless of the topic, require more than just enthusiastic scientists claiming democratic or academic values; entire research infrastructures need to be built and fostered to help evaluation research along. Changing the evidence base surrounding counterterrorism requires creating an environment that is friendlier to evaluation research and the one in which evaluation culture is institutionalized. Many efforts are already under way, including the funding of research by the Departments of

374

C. Lum and L.W. Kennedy

Justice, Homeland Security, Defense, and State. In addition, many scholars have already spent lifetimes preparing the field in terms of data collection, theory development, and definition setting for the new generation of evaluation scholars. To envision such an infrastructure and supportive base for evaluation research in counterterrorism, we should again look to lessons learned in the policing world. As Lum (2009) describes, the reason for the relative ease by which police scholars today can generate research knowledge of policing and evaluations of police tactics is the base on which this endeavor stands. This foundation includes work by early researchers and practitioners who forged good relationships; development of the research knowledge itself; technological advancements that help with method and data development; improvements in police–citizen relations, which foster a culture of accountability and transparency; increased expectations for chief executives to deliver effective strategies, not just believed “best practices”; third parties who could facilitate police–research relationships; and funding efforts by governmental agencies interested in analyzing cost-effectiveness. In policing, each of these separately and together helped create an environment that makes it easier for current scholars to evaluate interventions in law enforcement. To achieve similar accomplishments in counterterrorism research, a research foundation based on cooperation and transparency is needed in order to create an environment in which rigorous evaluation research on counterterrorism and security efforts can be accomplished. Developing such an infrastructure and environment requires effort by all involved and likely will involve the following activities: • Changing the Departments of Homeland Security, Justice, Treasury and Defense’s focus and research portfolios related to terrorism away from primarily technology development and more toward methodologically rigorous outcome evaluations of existing interventions. • Continuing to explore alternative methods of evaluation, especially for rare events. • Developing discourse and research around evaluation goals, including standards for data reporting, definitional analysis, systematic reviews of research, and finding ways to link etiological research to evaluation research. • Creating access routes for researchers to both classified and unclassified data. • Improving and increasing dialog between researchers, policy makers, and practitioners, including overcoming myths, fears, and uncertainties about the intentions and capabilities of all three groups. • Developing a clear understanding of, and research base for, assessing collateral effects of interventions. • Building into public policies and laws requirements for the evaluation of those policies. • Developing mechanisms of both delivery and interpretation of research evaluation results to practitioners and policy makers. • Basing agent and bureaucratic performance on measurable outcomes and third party assessments. • Encouraging more discourse about evaluation among leaders in the field.

16

The Next Steps: A Need for a Research Infrastructure…

375

• Enhancing academic training and support for new scholars who are interested in studying within these areas but are fearful of risks to publication and tenure. • Developing research dissemination and review frameworks to make sense of growing amounts of research in this area. Of course, there is much more work to be done for such a research infrastructure to be developed. There are a number of process-based obstacles and restrictions to moving toward building the capacity for research. For example, the political nature of data and the control over evaluation results can inhibit good science, as been seen in other fields (for example, gun crime research). Even if data meet scientific standards, the highly charged debates about threats to security cause action to take priority over assessment in this area. The arguments presented here assert that both can co-exist if there is carefully protected and thoughtful regulation of third-party objective evaluation that can guide the implementation of intervention programs. This must be done, given that there is rarely political will by the general citizenry for an evidence-based approach. In a related way, there are often pressures to do something, regardless of the fiscal or political costs, to enhance public safety. In a variation on the precautionary principle, this view suggests that the dire consequences of a successful terrorist attack makes it necessary to avoid all possible conditions in which this might occur (see van Brunschot & Kennedy, 2008 for a discussion of the precautionary principle in the context of balancing risk in the public domain). In this argument, there is little room for evaluation, as the only result of importance is the nonoccurrence of the attack. There may also be an unbalanced investment in technology and data management over a focus on consequences of interventions. Such investments are guided by the view that if the technology is fully developed to detect patterns of threat and direct action against the protagonists, all risks can be effectively mitigated. This approach downplays the uncertainties that are inherent in all security and anti-terrorism programs and devalues evaluation. It also rejects the reality and complexity of the constant balance and compromises that public policies confront in free societies. Finally, there may be poor skill development concerning evaluation in policy circles. The problems with evaluation extend to the limited skill base in the techniques of evaluation among individuals who are involved in policy and program development in counterterrorism. This problem could be addressed in the improvement of relationships between academic researchers and policy makers, as well as conscious efforts by leaders within the various cabinets to educate personnel on high-quality evaluation practices. The current efforts by the Department of Justice, for example, in its Evidence-Integration Initiative (E2I)3 are an excellent example. The bottom line: Despite obstacles, politics, and real concerns about evaluating counterterrorism interventions, we still must make the effort to fulfill the mandates of good governance, civil and free society, and accountability. It is our hope that the community of scholars and policy makers working in this area will embrace these 2

Spearheaded by Assistant Attorney General Laurie Robinson, see “Evidence-Integration Initiative (E2I)” at the Office of Justice Programs (http://www.ojp.usdoj.gov/index.htm).

376

C. Lum and L.W. Kennedy

ideas as we move forward to a more accountable and transparent assessment of the efficacy of the programs that are developed to make us safe.

References Braga, A. (2005). Hot spots policing and crime prevention: A systematic review of randomized controlled trials. Journal of Experimental Criminology, 1, 317–342. Campbell, D. T., & Stanley, S. C. (1966). Experimental and quasi-experimental designs for research. Chicago: Rand McNally. Cook, T., & Campbell, D. (1979). Quasi-experimentation: Design and analysis issues for field settings. Chicago: Rand McNally. Council, N. R., & Academies, T. N. (2004). Fairness and effectiveness in policing: The evidence. In W. Skogan & K. Frydl (Eds.), Committee to review research on police policy and practices, Committee on Law and Justice, Division of Behavioral and Social Sciences and Education. Washington: National Academies Press. Council, N. R., & Academies, T. N. (2010). Field evaluation in the intelligence and counterintelligence context: Workshop summary. Washington: The National Academies Press. Halkides, M. (1995). How not to study terrorism. Peace Review: A Journal of Social Justice, 7(3), 253–260. Hoffman, B. (1992). Current research on terrorism and low-intensity conflict. Studies in Conflict and Terrorism, 15, 25–37. London Resilience (for the Home Office). (2006, September). Addressing lessons from the emergency response to the 7 July 2005 London bombings: What we learned and what we are doing about it. Retrieved May 17, 2011, from http://www.londonprepared.gov.uk/downloads/homeoffice_lessonslearned.pdf. Lum, C. (2009). Translating police research into practice. Ideas in American policing. Washington: Police Foundation. Lum, C., Kennedy, L., & Sherley, A. (2006). Are counter-terrorism strategies effective?: The results of the Campbell Systematic Review on counter-terrorism evaluation research. Journal of Experimental Criminology, 2(4), 489–516. Miller, R. (1988). The literature of terrorism. Terrorism, 11, 63–87. National Commission on Terrorist Attacks Upon the United States. (2004). The 9/11 commission report. New York: W.W. Norton & Company. Romano, T. (1984). Terrorism: An analysis of the literature. Unpublished doctoral dissertation. Fordham University, Department of Sociology, Criminology and Penology, New York. Schmid, A. P., & Jongman, A. J. (1988). Political terrorism: A new guide to actors, authors, concepts, databases, theories and literature. Amsterdam: North-Holland Publishing Company. U.S. Government Accountability Office. (2007). Risk, experience, and customer concerns drive changes to airline passenger screening procedures, but evaluation and documentation of proposed changes could be improved. Retrieved July 15, 2011 from http://www.gao.gov/new. items/d07634.pdf. U.S. Government Accountability Office. (2009). DHS and TSA have researched, developed, and begun deploying passenger checkpoint screening technologies, but continue to face challenges. Retrieved July 15, 2011 from http://www.gao.gov/new.items/d10128.pdf. U.S. Government Accountability Office. (2010). TSA has taken actions to manage risk, improve coordination, and measure performance, but additional actions would enhance its actions. Retrieved July 15, 2011 from http://www.gao.gov/new.items/d10650t.pdf. van Brunschot, E., & Kennedy, L. W. (2008). Risk balance and security. Thousand Oaks: Sage. Weisburd, D., Lum, C., & Petrosino, A. (2001). Does research design affect study outcomes in criminal justice? Annals of the American Academy of Political and Social Sciences, 578, 50–70.

Biographies

Joel M. Caplan, Ph.D., is an Assistant Professor at Rutgers University School of Criminal Justice and Associate Director of the Rutgers Center on Public Security. He researches in the area of computational criminology, a multidisciplinary approach that takes the strengths of several disciplines and builds new methods and techniques for the analysis of crime and crime patterns. Jeremy Carter, Ph.D., is an Assistant Professor in the Department of Criminology & Criminal Justice at the University of North Florida (UNF). Dr. Carter’s research interests include law enforcement intelligence, policing, organizational behavior, and policy implementation. Dr. Carter is a principal investigator for a project funded by The National Consortium for the Study of Terrorism and Responses to Terrorism (START) and is currently a member of a project funded by the National Institute of Justice to evaluate information sharing practices among law enforcement fusion centers. Breanne Cave, M.A., is a researcher and doctoral student in the Center for Evidence-Based Crime Policy, Department of Criminology, Law and Society at George Mason University. She was commissioned as an officer in the United States Marine Corps and served on active duty from 2005 to 2008, deploying to the Al Anbar Province, Iraq in support of Operation Iraqi Freedom. Her research interests include policing, terrorism, insurgency, and crime and place. Steven Chermak, Ph.D., is Professor in the School of Criminal Justice at Michigan State University. His main research interests of late are in the areas of domestic terrorism, focusing on the organizational characteristics of violent white supremacists groups, participation in terrorist organizations, and lone-wolf terrorism. Claudio Cioffi-Revilla, D.Sc.Pol., Ph.D., is Director of the Center for Social Complexity and Professor of Computational Social Science at George Mason University. He conducts research on theory and methods in quantitative and computational models of complex systems. His research is funded by the National Science

C. Lum and L.W. Kennedy (eds.), Evidence-Based Counterterrorism Policy, Springer Series on Evidence-Based Crime Policy 3, DOI 10.1007/978-1-4614-0953-3, © Springer Science+Business Media, LLC 2012

377

378

Biographies

Foundation and the Office of Naval Research. He serves as a National Academies’ Jefferson Science Fellow at the U.S. Department of State, Office of Geographic and Global Issues. Tracy E. Costigan, Ph.D., is a Principal Research Scientist at the American Institutes for Research (AIR) in Washington, D.C. Her work focuses on quantitative methodology, instrument development and validation, and program evaluation applied to modeling human behavior across a wide range of scenarios, including the implementation of national security programs. Laura Dugan, Ph.D., is an Associate Professor in the Department of Criminology and Criminal Justice at the University of Maryland. She is a member of the National Center for the Study of Terrorism and the Response to Terrorism. Her research examines the consequences of violence and the efficacy of violence prevention/ intervention policy and practice. She also designs methodological strategies to overcome data limitations inherent in the social sciences. She received her Ph.D. in Public Policy and Management from Carnegie Mellon University in 1999. Eldivan, Dr., is a member of the Turkish National Police and a recent graduate of the Rutgers School of Criminal Justice. He is an expert on religious extremism and its relationship to terrorism. Charlotte Gill, Ph.D., is a Senior Research Associate and codirector of the Systematic Reviews research program in the Center for Evidence-Based Crime Policy, Department of Criminology, Law and Society at George Mason University. She is also the Managing Editor of the Campbell Collaboration Crime and Justice Group, an international network of researchers that produces systematic reviews of evaluation research in crime and justice. Her research interests include crime prevention, juvenile justice, place-based criminology, quantitative and evaluation methods, and research synthesis. Julie Hibdon, M.A., is a Research Associate for the Center for Evidence-Based Crime Policy and a doctoral candidate in the Department of Criminology, Law and Society at George Mason University. Her research interests include crime and place, environmental criminology, and policing. Leslie W. Kennedy, Ph.D., is University Professor at Rutgers University School of Criminal Justice and served as its Dean from 1998 to 2007. He is also the Director of the Rutgers Center on Public Security. Dr. Kennedy has published extensively in the areas of fear of crime, victimology, and violence. Gary LaFree, Ph.D., is the Director of the National Center for the Study of Terrorism and Responses to Terrorism (START) and professor in the Department of Criminology and Criminal Justice at the University of Maryland. Much of his recent research has dealt with national and international macro-level trends in political and criminal violence.

Biographies

379

Cynthia Lum, Ph.D., is the Deputy Director and Associate Professor of the Center for Evidence-Based Crime Policy, Department of Criminology, Law and Society at George Mason University. She researches in the area of evidence-based policing, democratization and justice, security and counterterrorism, and crime and place. Lorraine Mazerolle, Ph.D., is a Research Professor in the Institute for Social Science Research (ISSR) at the University of Queensland and an Australian Research Council Laureate Fellow. She is also the Foundation Director and a Chief Investigator in the Australian Research Council (ARC) Centre of Excellence in Policing and Security (CEPS), a Chief Investigator in the Drug Policing Modeling Program, and the ISSR “Policing and Security” Program Director. Her research interests include community regulation, problem-oriented policing, police technologies, civil remedies, street-level drug enforcement, and policing public housing sites. Linda M. Merola, J.D., Ph.D., is the Director of the Evidence-Based Legal Policy Research Program within the Center for Evidence-Based Crime Policy and an Assistant Professor in the Department of Criminology, Law and Society at George Mason University. She received her Ph.D. in Government from Georgetown University and her J.D. from George Washington University Law School. Her academic interests relate to civil liberties, terrorism, and the measurement of public and expert opinion about law and courts. Nikos Passas, Ph.D., is a Professor in the School of Criminal Justice at Northeastern University. Dr. Passas has researched extensively in the fields of terrorism, whitecollar crime, corruption, organized crime, and international crime. He is a consultant to the World Bank, the International Monetary Fund, United Nations Centre for International Crime Prevention and the Office on Drugs and Crime, the UN Development Program, the Commission of the European Union, the German Parliament, and government agencies from various countries, including the U.S. Department of Treasury and the U.S. Department of Justice. Michael D. Porter, Ph.D., is a Principal Research Scientist at GeoEye Analytics. His research areas include crime and terrorism modeling, dynamic network analysis, anomaly detection, point processes, predictive modeling, and statistical forensics. Clay Ramsay, Ph.D., is the research director at the Program on International Policy Attitudes (PIPA) at the University of Maryland, and a fellow of the Center for International and Security Studies at Maryland (CISSM). He cofounded PIPA in 1992. For 10 years, he has conducted international and US research on public opinion regarding counterterrorism issues. The views expressed are the author’s own. Danielle M. Rusnak, M.A., is a doctoral student and research assistant at Rutgers University’s School of Criminal Justice. She researches in the area of risk assessment, terrorism, crime and geography, sexual offending/offenders, and prisoner reentry. Ms. Rusnak has done previous work in the area of Forensic Psychology on criminal behavior and assessment.

380

Biographies

Ivan Sascha Sheehan, Ph.D., is the Director of the graduate program in Negotiation and Conflict Management at the University of Baltimore, where he is an Assistant Professor in the School of Public and International Affairs. His research focuses on transnational terrorism, counterterrorism, and international conflict management. Dr. Sheehan is the author of When Terrorism and Counterterrorism Clash: The War on Terror and the Transformation of Terrorist Activity (2007). Tom R. Tyler, Ph.D., is a University Professor at New York University. He studies the legitimacy of authority and public cooperation with the police. His particular concern is with public views about the justice and injustice of varying police policies and practices. David Weisburd, Ph.D., is Distinguished Professor and the Director of the Center for Evidence-Based Crime Policy in the Department of Criminology, Law and Society at George Mason University. He is the 2010 Stockholm Prize recipient for his work on crime places and displacement. Gentry White, Ph.D., is a Research Fellow at The Institute for Social Science Research at the University of Queensland and at The Australian Research Council Centre of Excellence in Policing and Security. He researches in quantitative methodologies associated with intelligence analysis. Sue-Ming Yang, Ph.D., is an assistant professor in the Department of Criminal Justice at Georgia State University and National Chung Cheng University. Her research interests include place-based criminology, criminological theory testing, experimental research methods, analysis of longitudinal terrorism patterns, and understanding the relationship between disorder and crime over time.

Index

A Ackerman, G., 354 Afghanistan, 3, 47, 58, 63, 120, 121, 267, 299–301, 305, 306, 309, 311, 312, 314, 317, 323, 342 Airport, 4, 132, 186, 207–211, 215–231, 233, 235–243, 283, 370 Al Qaeda, 110, 114, 120, 121, 126, 127, 167, 168, 253, 261, 270, 301, 305, 307–310, 320, 321, 356 Alt, J.E., 134 Analytic products, 73, 78, 80, 81, 84, 86 Arce, D.G., 32

B Barbieri, R., 102 Barros, C.P., 92, 94 Bartholomew, D.J., 154 Bayley, D.H., 67 Bennett, T., 236 Bentham, J., 353 Berrebi, C., 354 Boba, R., 239 Booth, J.A., 336 Boraz, S., 73 Brady, H.E., 13 Braga, A.A., 336 Brown, E., 102 Bushway, S.D., 123 Bynam, D., 331

C Campbell, D.T., 198, 367 Caplan, J.M., 165–182 Carter, D.L., 69

Carter, J.G., 65, 69 Cauley, J., 212, 232, 240 Cave, B., 207–243 Change-point detection, Chermak, S., 234 Chermak, S.M., 61, 65 Cioffi-Revilla, C., 149–162 Civil liberties, 71, 76, 84, 211, 282, 284–292, 295, 377 Clarke, R., 212, 213, 215, 234 Clark, T.S., 292 Cohen, J., 196 Cohen, L., 212, 327, 334 Collier, D., 13, 22 Cook, T.D., 198 Cooperation, 6, 70, 208, 210, 211, 216–217, 219, 222, 224, 227–231, 233, 238–243, 257–260, 262, 265–267, 269, 273, 274, 306, 334, 352–359, 372, 378 Cornish, D., 213, 215 Costigan, T.E., 185–205 Counterinsurgency (COIN), 7, 161, 299–321, 323–342, 352 Counterterrorism, 3–8, 19, 25, 61, 62, 66–71, 74, 76, 85, 86, 91–111, 149–161, 165, 168, 177–180, 185–187, 194, 199, 200, 210–213, 216, 219, 235, 239–242, 254–256, 262, 279–296, 300, 302, 323–341, 354, 365–373, 376–378 Counterterrorism policy evaluations, 3–8, 150, 151, 279–296, 365, 369 Crenshaw, M., 115 Crime and place, 68, 130, 212–216, 219, 224, 234, 235, 237–240, 242, 260, 282, 283, 294, 325, 328–336, 338, 369, 370, 375, 376 381

382 Criminology (use of in understanding counterterrorim), 116, 207–243, 323–341, 375–378 Criteria, 14, 18–21, 24–27, 33–35, 37, 43, 61, 115, 116, 123, 124, 152, 188, 189, 205, 261–263 Cutter, S., 47

D Data, 4, 13, 41, 65, 91, 115, 149, 165, 188, 211, 254, 279, 302, 341, 365, 376 Databases, 13–15, 17–19, 21, 24–26, 28–35, 41–63, 85, 178 Davis, L.M., 68 Dugan, L., 47, 92, 94, 113–145, 211, 212, 235, 330, 354 Durlauf, S., 215, 235

E Eck, J., 216, 230, 234, 239 Eckstein, Z., 169 Eden, U., 102 Eggleston, E.P., 130 Eldivan, I.S., 165–182 Enders, W., 32, 92 Epstein, L., 292 Event data, 46, 61, 95, 99, 134, 138, 152 Evidence-based, 3–8, 65–87, 150, 178–180, 208, 210, 217–219, 223, 230–233, 274, 280–282, 289, 291–293, 295, 296, 299–321, 325, 341, 342, 365, 366, 369, 371, 373, 375–378

F Fachner, G., 68, 84 Farrales, G., 294 Farrell, G., 326 Farrington, D., 234, 236 Fawcett, G.S., 327, 339 Felson, M., 212, 327, 334 Framework, 6, 14, 19–21, 24, 35, 60, 92–94, 98, 99, 110, 177, 178, 186, 187, 204, 212–213, 217, 256, 258, 266, 306, 335, 342 Frank, L., 102 Freilich, J.D., 61 Fusion centers, 6, 65–86, 239, 368, 375

G Galula, D., 331 Gaza strip, 63

Index Geographic information systems (GIS), 160 Gerring, J., 14, 20 Gill, C., 207–243 Global Terrorism Database (GTD), 5, 15, 17, 25, 26, 41–63, 99, 115, 211, 368 Goertz, G., 22 Gold, M., 262 Grant, J.A., 200 Graphia-Joyal, R., 82, 83 Groff, E.R., 336 Group-based trajectory analysis. See Trajectory analysis GTD. See Global Terrorism Database Guardianship, 209, 212–215, 221–237, 241, 242, 325, 328, 329, 333–335, 337–340

H Haberfield, M., 68 Hagan, J., 294 Hamm, M.S., 170 Harries, K., 178 Hasisi, B., 241 Hazard models, 139 Heyer, G., 339 Hibdon, J., 207–243 Ho, D., 292 Holden, R.T., 135 Holloway, K., 236 Huang, J., 132, 135 Huntington, S.P., 338 Huq, A.Z., 358

I Im, E., 212, 232, 240 Indonesia, 63, 92, 99–101, 110, 111, 118, 120, 129, 303, 306–308 Information sharing, 65–78, 85, 86, 216, 239, 375 Innovation, 5–8, 116 Instrument validation, 6, 185–187, 191, 193–195, 197, 198, 200, 201, 203, 376 Insurgency, 7, 151, 253, 267, 299–301, 313, 323–326, 329–335, 375 Intelligence, 6, 14, 15, 33, 37, 42, 43, 61, 65–86, 114, 115, 160, 165, 197, 239, 259, 261, 262, 265, 299, 301, 303, 305, 316–318, 324, 353, 370, 375, 378 Intervention analysis, 92, 94, 101–106, 110 Iraq, 3, 45, 47, 51, 52, 59, 61, 63, 120, 121, 267, 294, 299–301, 305, 306, 310, 314, 317, 318, 323, 336, 342, 352, 354, 375

Index J Jasso, G., 294 Jenkins, B.M., 53 Jones, S.G., 73 Jongman, A.J., 31, 41

K Kagan, F.W., 331 Keele, L., 336 Kennedy, L.W., 3, 74, 116, 165–182, 185, 210, 279, 332, 367 Kilcullen, D., 300, 302–305, 307–310, 312–314, 316–318, 330, 332 King, G., 134, 292 Kline, M., 159 Klor, E., 354 Koper, C., 329 Korte, R., 94, 354 Krieger, T., 179 Krohn, M.D., 123 Krueger, A.B., 22 Kull, S., 320, 321

L LaFree, G., 41, 47, 51, 92, 94, 115, 117–119, 123, 129, 132, 140, 202, 211, 330, 354 Laitin, D.D., 22 Land, K.C., 117 Laub, J.H., 130 Law, 3, 6, 7, 43, 65–87, 116, 131, 150, 152–154, 156–158, 194, 207, 211, 212, 215, 216, 222, 224, 227, 235, 236, 238–241, 253, 255, 258, 259, 268, 271–273, 279–292, 294–296, 312, 323, 334–338, 352–357, 366, 368, 371, 372, 375–378 Legal challenges, 279–296 Legitimacy, 4, 7, 237, 240, 241, 255, 262, 282, 288, 336, 340, 354–359, 365, 366, 369, 371, 378 Levi, M., 261, 262 Lewis, E., 321 Lieberman, C., 68 Lipscy, P., 73 Li, Q., 115 Location quotients (LQs), 165–167, 175–177 Longitudinal analysis, 369 Lum, C., 3, 68, 74, 84, 116, 185, 207–243, 279, 329, 351, 365–367, 370, 371 Lynch, J.P., 61

383 M Mahon, J.E., Jr., 22 Marcus, G.E., 285 Mazerolle, L., 91 McCauley, C., 132 McGarrell, E., 234 Meeker, J., 239 Meierriek, D., 179 Merari, A., 23, 42 Merola, M., 279 Methodology, 6, 74, 149, 179, 198, 204, 219, 294, 317, 376 Military, 3, 4, 7, 38, 42, 54, 55, 170, 172, 271, 281, 294, 299–306, 308–311, 316, 317, 323–342, 351, 352, 354, 355, 369 Modeling, 5, 6, 84, 91, 94, 98, 99, 113–145, 150, 151, 155, 160, 166, 178, 376, 377 Morris, N.A., 117, 118, 336 Mudge, E., 234 Muslim community, 352, 357–359

N Nagin, D.S., 117, 123, 125, 129, 130, 215, 235 National Center for the Study of Terrorism and Responses to Terrorism (START), 6, 26, 29, 37, 43, 62, 115, 151, 211, 307, 367, 368, 375, 376 Newman, O., 213 Norman, D.A., 193

P Pakistan, 3, 47, 58, 120, 264, 300, 302, 306–311, 316, 318–321 Parkin, W.S., 61 Partnerships, 4, 66, 67, 75, 76, 83–84, 86, 87, 110, 216, 352 Passas, N., 253 Pease, K., 326 Perl, R., 91 Petrosino, A., 236, 367 Piquero, A.R., 92, 211, 330 Point process, 92–102, 105, 108, 110 Policing matrix, 217–219, 223 Policing policies, 358, 359 Policy evaluation methodology, 150, 151 Political tolerance, 284, 285, 287, 290, 291 Porter, M.D., 91 Procedural justice, 240, 355–359 Public opinion, 36, 281, 299–321, 371, 377

384 Q Qualitative, 5, 13, 14, 20, 35, 113, 114, 122, 154, 156, 158, 223, 317 Quantitative, 5, 14, 20, 35, 91, 94, 96, 99, 116, 117, 151, 153, 154, 189, 294, 317, 353, 375, 376, 378

R Ramsay, C., 320, 321, 341 Rapoport, D.C., 119 Research, 3, 13, 41, 65, 91, 113, 151, 166, 185, 208, 253, 279, 299, 323, 351, 365 Research infrastructure, 8, 365–373 Richard, P.B., 336 Risk, 3, 13, 46, 72, 91, 115, 154, 165, 190, 209, 254, 301, 326, 351, 366 Robson, C., 19 Rogan, D., 234 Rojek, J.R., 81 Ronceck, D.W., 338 Rossmo, D.K., 178 Routine activities theory, 211, 323–342 Rusnak, D.M., 165–182

S Sampson, R.J., 130, 336 Sandler, T., 32, 33, 92, 94 Schmid, A., 31, 32, 41 Schulhofer, S., 358 Seawright, J., 13 Security, 3, 15, 62, 65, 131, 158, 170, 185, 207, 253, 284, 303, 331, 351, 366 Security screening, 185–205, 221, 226 Segal, J., 292 September 11th, 2001, 3, 4, 6, 44, 65, 66, 68, 69, 110, 207, 279, 287, 351, 353, 355 Series hazard model, 117, 131–145 Shaw, J., 234 Sheehan, I.S., 13, 202 Sheldon, K.M., 159 Shellman, S.M., 134 Sherley, A.J., 3, 74, 116, 185, 210, 367 Sherman, L.W., 216, 231, 234, 327 Signorino, C.S., 134 Singh P., 47 Situational crime prevention, 209, 211, 213–215, 218, 219, 221, 230, 232, 234– 238, 240–242, 325, 328, 329 Social contract theory, 334

Index Social science (use of in understanding counterterrorim), 7, 15, 20, 62, 159–161, 208, 279–283, 293–295, 324, 367, 370, 375–378 Spatial analysis, 61, 165–181, 336 START. See National Center for the Study of Terrorism and Responses to Terrorism Stouffer, S., 284 Survey research, 68, 299–302, 316, 319, 321 Survival analysis, 92, 94, 117

T Taliban, 52, 58, 121, 127, 256, 258, 259, 309–315, 320, 321 Tankebe, J., 356 Telep, C.W., 336 Terrorism, 3, 13, 41, 66, 91, 113, 149, 165, 185, 208, 253, 279, 300, 323, 351, 365 Terrorism events, 3, 14, 15, 17, 18, 22, 24, 25, 149, 150, 156, 158, 367, 368 Terrorism finance (and counterterrorism finance), 7, 254, 258, 259, 262, 263, 267, 273, 274 Terrorist organizations, 42, 44, 51–52, 113–122, 125–130, 132, 144, 170, 271, 352, 375 Terrorist strategies, 113, 179 Thornberry, T.P., 123 Time-series, 94, 370 Trajectory analysis, 113–145 Translational criminology, 207–243 Transportation, 4, 7, 54, 55, 171, 172, 186, 202, 207–209, 366 Transportation Security Administration (TSA), 4, 186, 194, 202, 207–243 Trends, 5, 19, 24, 33, 44, 48, 57, 59, 114, 116–119, 128–130, 144, 240, 332, 368, 376 Treverton, G.F., 73 TSA. See Transportation security administration Tsiddon, D., 169 Turkey, 6, 16, 47, 63, 120, 121, 165–168, 170, 173, 175, 177–179, 306, 307, 309, 370 Tyler, T.R., 240, 351, 358

U U.S. Defense budget, 300, 304, 305

Index V Validity, 7, 13, 14, 19–21, 173, 175, 177–179, 185–205, 367 Van Brakle, M., 51

W Weber, S., 321 Weisburd, D., 67, 207–243, 336, 367 Weiss, A., 234 Welsh, B., 234, 236

385 White, G., 91 Wigle, J., 33 Wilson, J., 234

X Xie, M., 47

Y Yang, S.-M., 113–145 Young, J., 115