Disrupting Criminal Networks: Network Analysis in Crime Prevention 9781626372573

Citation preview

bichler-fm_ff-fm 1/20/15 10:14 AM Page 1

DISRUPTING CRIMINAL NETWORKS

00_Ekblom_FM_Ekblom 1/20/15 2:47 PM Page ii

Crime Prevention Studies Volume 28 Ronald V. Clarke, series editor

bichler-fm_ff-fm 1/26/15 10:17 AM Page 3

DISRUPTING CRIMINAL NETWORKS Network Analysis in Crime Prevention

edited by

Gisela Bichler Aili E. Malm

bichler-fm_ff-fm 1/30/15 10:14 AM Page 4

Published in the United States of America in 2015 by FirstForumPress A division of Lynne Rienner Publishers, Inc. 1800 30th Street, Boulder, Colorado 80301 www.rienner.com

and in the United Kingdom by FirstForumPress A division of Lynne Rienner Publishers, Inc. 3 Henrietta Street, Covent Garden, London WC2E 8LU

© 2015 by Lynne Rienner Publishers, Inc. All rights reserved

Library of Congress Cataloging-in-Publication Data Disrupting criminal networks : network analysis in crime prevention/ [edited by] Gisela Bichler and Aili E. Malm. (Crime prevention studies; volume 28) Includes bibliographical references and index. ISBN 978-1-62637-227-6 (hc: alk. paper) 1. Crime prevention. 2. Social networks. 3. Crime—Sociological aspects. I. Bichler, Gisela. II. Malm, Aili E. HV7431.D57 2015 364.401'172—dc23 2015000849

British Cataloguing in Publication Data A Cataloguing in Publication record for this book is available from the British Library.

This book was produced from digital files prepared by the author using the FirstForumComposer. Printed and bound in the United States of America

The paper used in this publication meets the requirements of the American National Standard for Permanence of Paper for Printed Library Materials Z39.48-1992. 5 4 3 2 1

bichler-fm_ff-fm 1/20/15 10:14 AM Page 5

To Ronald V. Clarke— Mentor, inspiration, and winner of the 2015 Stockholm Prize in Criminology

Contents

Acknowledgments

ix

1

Why Networks? Aili E. Malm and Gisela Bichler

1

2

Street Gangs and Co-Offending Networks Jean Marie McGloin and Zachary Rowan

9

3

Applying Group Audits to Problem-Oriented Policing Michael Sierra-Arevalo and Andrew V. Papachristos

27

4

Network Stability Issues in a Co-Offending Population Carlo Morselli, Thomas U. Grund, and Rémi Boivin

47

5

Identifying Key Actors in Drug Trafficking Networks David A. Bright

67

6

Predicting Organized Crime Leaders Francesco Calderoni

89

7

Defection from a Fraud Network Robert R. Faulkner and Eric R. Cheney

111

8

Discrediting Vendors in Online Criminal Markets David Décary-Hétu and Dominique Laferrière

129

9

Vulnerabilities in Online Child Exploitation Networks Kila Joffres and Martin Bouchard

153

10 Measuring Disruption in Terrorist Communications Stacy Bush and Gisela Bichler

177

11 Using Space Syntax to Inform Crime Prevention Lucia Summers and Shane D. Johnson

209

vii

viii

Contents

Appendix: Networks in a Nutshell Gisela Bichler and Stacy Bush Bibliography The Contributors Index

233

245 275 279

Acknowledgments

Producing an edited volume is not easy. It takes a dedicated team of contributors and supporters, and for this reason, we have a number of people to acknowledge. The first step in producing an edited volume is to develop a good idea. Shortly after finishing my (Gisela) PhD in 2000, Ronald V. Clarke and I found ourselves in San Diego, California, for the 11th annual Problem-Oriented Policing Conference. At that time he extended an offer, although perhaps it was better described as a challenge. If I ever came up with a clever idea for a volume for the Crime Prevention Studies series, I should send him a short proposal. Needless to say, it took a while to come up with a clever idea. We would like to offer our enduring gratitude to Professor Clarke for seeing the value in this project and affording us the opportunity to contribute a volume to the series. For more than 20 years, Professor Clarke has guided the careers of young academics. This is but one of the many opportunities he has provided. The second step is to assemble an all-star cast of contributors who are able to fit yet another request for a manuscript into their busy research schedules. When we sat down to develop our wish list of contributors we very quickly ran out of room on the page. Knowing what people were currently working on, we selected a group for the first wave of invitations. Nearly everyone accepted with great enthusiasm. We were thrilled at the response, and in the coming months we were continually astounded at the authors’ efforts to meet all of our requests, no matter how demanding. We would like to extend our heartfelt appreciation to the following individuals; without their contributions and support this book would not have been written: Rémi Boivin, Martin Bouchard, David A. Bright, Stacy Bush, Francesco Calderoni, Eric R. Cheney, David Décary-Hétu, Robert R. Faulkner, Thomas U. Grund, Kila Joffres, Shane D. Johnson, Dominique Laferrière, Jean Marie McGloin, Carlo Morselli, Andrew V. Papachristos, Zachary Rowan, Michael Sierra-Arevalo, and Lucia Summers.

ix

x

Acknowledgments

The third step is to assemble a supporting staff. The Crime Prevention Studies series is peer-reviewed. This meant that we needed to assemble a group of subject experts to examine each chapter. Since the review process was blind we cannot name individuals, however, we can say that all reviewers are associated to some extent with one of two working groups: the Illicit Networks Workshop or the Environmental Criminology and Crime Analysis Symposium. Their efforts were instrumental in shaping the contributions. Thank you. Two other people provided invaluable support. Stacy Bush not only contributed material for the book but also helped us with many of the tasks required to pull everything together, for example, integrating the reference lists for each chapter into the bibliography (which also involved finding missing information). Allison Ritto Almstedt stepped in at the eleventh hour to copyedit the entire book. We appreciate her efficiency and resolve to complete the review in record time. And thanks goes to our editor, Andrew Berzanskis. We are indebted to his foresight and generosity in supporting this project. We would also like to thank the rest of the Lynne Rienner Publishers staff for their assistance. Included in this list of supporting staff must be our family and friends whose encouragement and understanding were indispensable. In particular, we would like to thank Illya Robertson, Christine Famega, Jenni Murphy, Nerea Marteache Solans, Dan Dobson, and, last but not least, Aidan Dobson.

1 Why Networks? Aili E. Malm and Gisela Bichler

WHY FOCUS AN ENTIRE VOLUME ON THE UTILITY OF USING network analysis to prevent crime? Why consider networks at all? The answer is simple—crime is dependent on the behavior of individuals interacting with others within a social system. Many crimes are possible because of the opportunities provided through our connections to others, and at different levels of aggregation, this can be viewed as associations among groups, organizations, systems and places. Researching these relationships requires theory, methods and analysis which focus on how people affect one another within the context of the entire group. The people we associate with influence us, and we influence them. Sometimes these influences are indirect. For instance, through a friend of a friend someone may learn of a new application that strips audio files off YouTube videos. As a result they violate copyright laws. This is not a novel way to pass information—criminal or otherwise. Rather, information usually diffuses through a network of people, many of whom do not know each other directly. From this introduction, it is easy to see why network analysis, more specifically, examination of social relations (social network analysis or SNA), is quickly becoming an essential part of the methodological toolbox of criminologists and analysts. To date, criminological applications of network analysis generally focus on describing the social structure rather than explicitly considering how networks can be used to reduce crime. In this volume, we seek to do just that. The chapters presented in this book demonstrate how the associations among individuals, groups, and/or activities can foster criminal opportunity by bringing offenders and targets together in social systems. Just as crime is not random, neither are social 1  

2

Disrupting Criminal Networks

systems; patterns emerge as each person is embedded within a web of interrelated actors (Freeman et al. 1992). We argue that in order to understand the behavior of individuals, the individuals must be viewed within the large social context. This book also begins to extend the way we think about networks. While we have been talking about the relationships between people, these concepts can be applied toward understanding many different phenomena. Though still relatively rare, researchers have begun to explore how network analysis can be used to examine how the interacting parts of a system generate crime opportunity, i.e., street networks and ease of mobility (e.g., Summers and Johnson, Chapter 11 this volume), trade flow (e,g., Bichler and Franquez 2014; Bichler and Malm 2013), and financial exchange mechanisms (Bichler, Bush, and Malm 2013; Malm and Bichler 2013). The crime prevention implications of what Andrew Papachristos (2011) calls a networked-criminology, will expand considerably by pushing beyond direct social interactions, and moving into an exploration of the structure of systems. Placing crime opportunity within a broader context is an accepted practice among criminal justice scholars and practitioners. For instance, geographic information system (GIS) technology opened the door to examining incidents and individuals as they interact with the built environment; much as multi-level modeling offered a mechanism to study the influence of group level characteristics on individuals. A network perspective adds to this exploration of inter-dependence by broadening the range of factors that can be considered as part of the opportunity structure. Network studies are not limited to a single unit of analysis (groups and places can be mixed); models can integrate, time, space, individual attributes, and social relations; and finally, network techniques can demonstrate changes in flow dynamics (i.e., drug market disruption). For these reasons, network analysis is fast becoming an invaluable approach to understanding the situational context of crime. Criminological applications of network analysis already encompass a broad spectrum. Shedding new light on debates about the structure of criminal enterprise (e.g., Calderoni 2011; Morselli and Roy 2008) and victimization (e.g., McCuish, Bouchard, and Corrado 2015; Papachristos, Braga, and Hureau 2012; Randle and Bichler 2015), as well as other types of collaborative crime networks such as gangs (e.g., McGloin and Piquero 2010), terrorism (Everton and Cunningham 2014; Medina and Hepner 2008), computer hacking (Décary-Hétu, Morselli, and Leman-Langlois 2012) and drug trafficking (e.g., Bright and Delaney 2013; Malm and Bichler 2011). Collectively, this research highlights the utility of identifying the central or unifying forces that generate criminal networks. The available

Why Networks?

3

toolkit provides an arsenal of strategies to identify foci for crime prevention activity, in addition to a methodology for evaluating anti-crime efforts. This is important to the field because conventional statistics are not appropriate for examining inter-dependent networks of operatives and processes (Borgatti 2002; Wang et al. 2004-05; Wasserman and Faust 1994).

Book tour To our knowledge, this is the first book dedicated to discussing the crime prevention implications of social network research. While developing a list of contributors, we sought to demonstrate the utility of a network approach by bringing together a wide range of applications, both in terms of topic coverage and methods. This volume is divided into 11 substantive chapters and covers eight different types of crimes from five different data sources. All but two chapters report on original research. What unites this diverse assortment of scholarship is that each of these chapters effectively demonstrates why networks matter to crime prevention. Before moving on to a review of the content of this book, it is important to say something about the editing process. All of the books in the Crime Prevention Studies series are peer-reviewed. When Ronald V. Clarke launched the series, he set out to create a forum that would encourage the dissemination of innovative studies to advance the development of practical and effective crime prevention. By insisting on a peer-review protocol, volume editors were also held accountable for producing high quality scholarship. We are indebted to him for continuing with this mandate for over 20 years, producing 28 volumes. In keeping with this tradition of academic excellence, each chapter was peer-reviewed by at least two scholars. To select reviewers for each chapter, we identified one subject matter expert and one researcher familiar with the methods. Each chapter also received a heavy edit from us. Reviewers were tasked with ensuring not only that the scholarship was sound, but also, that the chapters could be understood by people unfamiliar with social network methods. Faced with the burden of launching network analysis into the field of situational crime prevention, we wanted to make sure that key concepts translated well. To support this crossover, this volume contains an appendix which is really a short primer and glossary of terms. While essential terms are defined within each chapter, we were aware that for many, the lexicon is new, and many people would benefit from including a short explanation of concepts. We also suggest resources that provide a more comprehensive introduction into SNA theory, methods and analytic tools.

4

Disrupting Criminal Networks

Table 1.1 provides our readers with a thumbnail sketch of the remaining chapters by topic area, data sources, social network theory and methods, and primary crime prevention implications. While there is no reason to start at the beginning and work your way to the end of the book, readers should note that the methodological complexity of the research is greater toward the end. The purpose behind crafting this table, and the chapter commentary that follows, is to offer a brief tour, after which readers can select their own spot to jump into the realm of SNA for crime prevention. Chapters 2 through 4 explore gang relations and co-offending. In Chapter 2, McGloin and Rowan examine the utility of adopting a social network approach for crime prevention policy in the context of street gangs and juvenile co-offenders. They argue that using SNA to guide research hypotheses, data collection, and analysis provides insight for crime prevention. The authors offer four crime prevention recommendations in using SNA to study of street gangs: 1) use street gang rivalry and alliance networks to nominate certain gangs for targeted intervention; 2) determine the cohesion among gang members; 3) identify individuals in a position of leverage or vulnerability; and, 4) include non-traditional street gang members, especially those in the legitimate market, when building a profile of the group structure. When considering juvenile co-offending more broadly, it is also important to reduce opportunities for youth to gather and socialize with potential co-offenders. Some of these recommendations are exemplified in Chapter 3 where Sierra-Arevalo and Papachristos present a systematic review of the group audit process using the city of New Haven, Connecticut as a case study. They effectively demonstrate how knowledge of the structure of criminal groups is critical to implementing effective focused-deterrence strategies. Capturing the feuds and rivalries through the analysis and the geo-social visualizations of networks among and between violent groups will improve the success of crime prevention initiatives while avoiding the pitfalls inherent to national and geographic conflation. Conflation is the tendency, common within law enforcement, to unite several smaller, distinct groups into larger and more generally named groups. Conflation weakens crime prevention as the situationally-specific factors inherent to local crime problems are more likely to be overlooked. In Chapter 4, Morselli, Grund, and Boivin, show how crime involvement, frequency, and stability vary within a co-offending population extracted from seven years of arrest data from Quebec, Canada. Varied levels of stability in co-offending partnerships suggest that prevention efforts must be tailored to the social position of active offenders, and that among the individuals who co-offend more than once, there is a high

Why Networks?

5

tendency to reuse some of their partners. The results also reveal that one’s social position within the co-offending network exhibits some crime specificity: individuals involved in market crimes are more likely to hold core positions in the network; whereas, violent and property crime cooffenders are more peripheral. Taken together, these results demonstrate how social networks constrain and facilitate co-offending differently for subgroups of offenders. Chapters 5 through 10 illustrate how network analysis can be applied to specific types or classes of crime to sharpen the effect of crime prevention efforts or to evaluate the impact of anti-crime policy. In Chapter 5, Bright shows how SNA can be used to help law enforcement identify high value targets whose removal is likely to disrupt or dismantle criminal enterprise, specifically illicit drug production and distribution. The identification of highly-connected actors and/or actors who play critical roles in a criminal organization can contribute to the suite of crime prevention strategies focused on criminal networks. He argues that certain social positions and operational roles within the drug distribution process are more critical than others; efforts to make it more difficult for individuals occupying these positions to act will produce greater overall network disruption than simply removing specific individuals. In Chapter 6, Calderoni analyses meeting attendance of the ‘Ndrangheta organized crime group to map the social hierarchies supporting local criminal activities. He shows that leaders can be identified through meeting attendance, and that these leaders could be targeted by police to disrupt operations and prevent organized crime by restricting the ability of leaders to communicate with their staff. Two general crime prevention strategies are supported by Chapters 5 and 6: 1) using SNA to maximize organized crime group disruption, and 2) using SNA to increase investigative efficiency. While Chapters 7 through 10 cover white-collar crime, computer crime, and terrorism, they are linked through their use of publicly available data. These chapters also show how to use network analysis to most effectively prevent crimes that have been difficult to disrupt using conventional approaches. Faulkner and Cheney (Chapter 7) use historical data from a corporate fraud case to illustrate how social relations can impact a conspirator’s likelihood of defecting and turning state’s evidence. This novel use of SNA demonstrates how white collar criminal investigations and prevention efforts can be aided with knowledge of network devolution.

 

Synopsis of Chapter Content

6

Ch.

Topic Area

Data

SNA Theory and Methods

Crime Prevention Implication

2

Street gangs, co-offending

Secondary data

Cohesion, brokers, social capital

Focused-deterrence, disruption of accomplice generation, and counter-intelligence

3

Criminal gangs

Police data

Centrality, visualization of geo-social networks

Focused-deterrence targeting situationalspecific leverage and vulnerability

4

Co-offending

Police data

Core-periphery, weak ties, affiliation networks

Target social position with situational-specific strategies

5

Illicit drug networks

Court transcripts, police data

Small-world structure, hubs and brokers

Disrupt networks by making it harder to fulfill duties associated with operational roles

6

Organized crime

Court transcripts, police data

Group structure, brokerage, affiliation networks

Target in-person operational meetings that promote leader influence

7

Fraud (white-collar)

Open source

Core-periphery, clustering, network devolution

Target defection-heavy positions for education and auditing to cultivate whistle-blowers

8

Carding (cybercrime)

Embedded web metadata

Inter-group ties

Discredit vendors with strategic use of market mechanisms

9

Child pornography (cybercrime)

Embedded web metadata

Small-world structure, hubs and brokers, fragmentation

Strategic removal of key sites prevents access to materials, deflects offenders, and reduces emotional arousal

10

Terror group messaging

Declassified intelligence

Dynamic network analysis, centrality

Target locally important actors (functional leaders) to prevent rebuilding

11

Offender movement

Street network

Connectivity, integration, local and global structures

Strategically modify urban structure to reduce offender mobility and enhance surveillance

p6_Layout 1 1/20/15 10:52 AM Page 1

Table 1.1

Why Networks?

7

With the dramatic increase in online crime (Holt and Lampke 2010) over the past decade, we would have been remiss to exclude chapters on how SNA is essential in helping prevent crime on the Internet. In Chapter 8, Décary-Hétu and Laferrière use SNA to investigate the formation of trust and business relationships in an online illicit market (carding forum). They then analyze the feasibility of implementing disruption strategies, such as Sybil attacks, to prevent the sale of illegally obtained financial data. This removes the illicit gains associated with hacking. Joffres and Bouchard (Chapter 9) continue the application of SNA to internet crime by examining online child pornography networks—websites which link to each other and facilitate the spread of materials. The authors show how using SNA on data extracted using a web-crawler can reduce the availability of online child pornography. Targeted disruption by removing key sites in a network can impede an individual from accessing illicit materials, deflect offenders to less popular sites, conceal remaining sites, and reduce emotional arousal. In Chapter 10, Bush and Bichler apply dynamic network analysis to publicly available correspondence between Usama Bin Ladin and members of the Al Qa'ida network to determine how information flow and operational efficiency changes after military action. Changes in network structure highlights how criminal networks are resilient when under attack as operational roles are re-staffed when individuals are removed. The authors suggest how tactical responses and prevention strategies can be used together to weaken network structures and decrease social support so as to impede rebuilding efforts. To close the book, Summers and Johnson (Chapter 11) return us to the intersection of geography and networks that was introduced in Chapters 2 and 3. This chapter is also gently leads the reader into the realization that network-oriented research is larger than social network analysis. While network approaches have traditionally been used to study social interactions such as co-offending, gang violence, and criminal organizations, the methodology can also be used to examine systems and spatial mobility. Summers and Johnson introduce us to space syntax, a method developed within the field of urban planning to estimate vehicular and pedestrian flow on street segments. This chapter describes how networks may be used to inform preventative efforts such as the correct positioning of surveillance systems and crime-free urban planning. The Future of Network Analysis and Crime Prevention Situational crime prevention scholars understand that crime is influenced by a context that shifts over time and space. The chapters in this volume unequivocally demonstrate that crime is also influenced by social systems.

8

Disrupting Criminal Networks

Understanding how victims and offenders interact is essential to preventing crime, and network methods are the tools through which we can best understand the interconnectiviey of the social world. 1.

2.

3.

We urge crime researchers to seriously consider the network perspective in their work. Think about how crime is embedded within a social structure: criminals and victims alike are social beings who have friends, families, colleagues, acquaintances, and enemies. These social ties impact people in every aspect of their lives. People are not equally positioned within these networks, nor do they have the same number of connections. This structure is important as it constrains and expands the pool of the opportunities we encounter, and not all of these opportunities are legal. We challenge researchers to stop treating criminals and victims as if they exist in isolation. It is really the interaction between potential offenders and victims that is important. Therefore the basic unit of analysis is the two actors and the link between them. Re-read criminological theories with this in mind, and then consider how network variables and methods could improve our collective understanding of crime. We also advise crime researchers and crime prevention practitioners looking to use network methods to read widely-used books that discuss these methods at length, i.e., Wasserman and Faust (1994). As with all methods, there are limitations associated with SNA and our understanding of these limits is still in its infancy. It is important to go beyond this book and read about network theory and statistics.

The application of network analysis to crime prevention has just begun, and the future is sure to see progress in the following areas: 1) understanding of the intersection of individual agency and the social world; 2) dynamic models capturing structural change; 3) geo-social analysis that integrates spatial properties with social structures; 4) network simulation studies; and, 5) policy evaluation. After reading this volume, we hope you agree with us that network analysis is an essential tool in reducing crime.

2 Street Gangs and Co-Offending Networks Jean Marie McGloin and Zachary Rowan

IN RECENT YEARS, THERE HAS BEEN MARKED GROWTH IN THE use of social network methods in criminology and criminal justice research. Though skeptics may view this growth as yet another example of scholars becoming “seduced” by a new method (Sampson and Laub 2005), it is clear that a social network approach to data collection and analysis is not merely a shallow, “fashionable” trend, but instead provides the opportunity for unique and important insight into questions of criminological interest (Morselli 2009b). Furthermore, though this method undeniably aids in the testing and refinement of theory (e.g., Haynie 2001; Hipp 2008; Hipp and Boessen 2013), it is not limited to this domain. To be sure, adopting a social network approach for the study of criminal networks provides added value that can be converted into policy and intervention guidance; in other words, it can be a powerful tool in the pursuance of “translational criminology” (Laub 2010) by bridging the gap between academics and criminal justice practitioners. Over the past two decades, “the orthodox organized crime doctrine that focuses on more or less stable and hierarchical organizations is slowly giving way to new and more sophisticated paradigmatia, such as…the concept of fluid social networks” (Klerks 2001: 53; see also Clarke and Brown 2003; Eck and Gersh 2000). Perhaps not surprisingly then, several organized crime scholars have explicitly advocated for using network analysis when developing intervention strategies (e.g., Coles 2001; Morselli 2010) and there are numerous examples of where application of this method reveals aspects of the organization, structure and social processes of various criminal networks, including Russian organized crime (Finkenauer and 9  

10

Disrupting Criminal Networks

Waring 1998) and drug trafficking groups (Natarajan 2000) that would be hidden otherwise. The primary benefit of this approach for the study of criminal networks is that, “unlike many approaches that rely on an assumption, speculation, or claim regarding the order of things in a specific criminal setting, social network analysis offers a method to search and assess structure…[that] brings an evidence-based assessment into the process and allows investigators and analysts to identify structure rather than assume it” (Morselli 2010: 383, 390). In doing so, it deepens our understanding of how these criminal networks function and provides insight on the structural vulnerabilities within the networks, highlighting key points for intervention tactics and strategies (Malm and Bichler 2011; Sparrow 1991; Xu and Chen 2008). When turning attention to juvenile delinquency, the two most notable types of criminal networks are street gangs and co-offending networks.1 It has been well established that juveniles more often than not commit crimes together and that most delinquent acts involve more than one individual (Reiss, 1988; Sarnecki 2001; c.f. Stolzenberg and D’Allesio 2008). Street gangs and co-offending networks serve as two examples of this meaningful feature of crime. In the context of these networks, juveniles may find themselves exposed to new skills and techniques to engage in delinquent behavior and an increase in opportunities to offend that ultimately can offer pathways toward deeper entrenchment in criminal lifestyles. Accordingly, these criminal networks should be evaluated to understand juvenile offending careers and for the purpose of supporting intervention efforts directed by law enforcement, school-based initiatives, and other juvenile justice services. Street Gang Networks and Crime Prevention Juvenile street gangs have become increasingly problematic over the past several decades. Communities experienced a growth in gang problems that peaked in the mid-1990s and has stabilized after a period of growth during the early 2000s (Egley, Howell and Major 2004; Howell 2010). Though a plateau is certainly preferable to a continued upward trajectory, the sheer number of gangs that have “stabilized” is troubling. For instance, the Office of Juvenile Justice and Delinquency Prevention (OJJDP) estimates that there are more than 750,000 gang members throughout the United States (Egley and Howell 2013). Unfortunately, even this large figure is likely an underestimate due to the fact OJJDP only surveys approximately 2,500 agencies.2 Further, the locus of street gang problems has expanded both in location and membership characteristics. Street gangs have not only continued to remain problematic in inner-city environments, but have also

Street Gangs and Co-Offending Networks

11

emerged in rural and suburban areas (Hagedorn 2002; Howell 2010; Miller 2001). The composition of gang membership has likewise broadened to include greater age, ethnic and gender heterogeneity (Howell 2010; Howell and Moore 2010; Klein 1995). Though official data on gang-related crime are often unavailable due to the lack of recording a crime as ‘gang-related,’ self-report information from several studies demonstrates that gang members account for a disproportionate amount of serious criminal behavior, most notably violence. For example, 30 percent of the sample in the Rochester Youth Development Study identified as being a gang member was responsible for approximately 68% of all violent offenses reported by the sample (Thornberry 1998). Importantly, such findings are not simply reflective of youth with delinquent or aggressive predispositions actively seeking out or being selectively recruited into gang membership (i.e., “selection”, see Gottfredson and Hirschi 1990). Research demonstrates that active membership in a gang is associated with marked within-individual increases in serious criminal activity when compared to periods of time before and after membership in the gang (Thornberry 1998). For instance, Lacourse et al. (2003) found that gang membership was associated with a notable increase in violent behavior even when accounting for trajectories of delinquency. Ultimately, street gangs not only lead to a bump in crime that is not fully accounted for by selection, but also contribute to serious consequences to the quality of life among the vast array of communities affected by the presence of this type of criminal network. There are several examples of social network analysis (SNA) providing insight on gang relationships—both the relationships among street gangs, and the relationships among members within the gang(s). One of the earliest examples can be found in William Foote Whyte’s (1943) seminal work on the social structure of an “urban slum,” Street Corner Society, in which he used basic sociograms to display the lines of communication in the groups under study. In the decades that followed, many studies conveyed rich observations about street gangs that were easily amenable to network techniques (e.g., Short and Strodtbeck 1965; Suttles 1968; see also Papachristos 2006), but it has not been until relatively recently that SNA has truly taken hold as an approach towards data collection and analysis in gang research. There are now several examples of researchers using social network analysis to better understand the reciprocal dynamics of violence among gangs (Papachristos 2009), to guide pulling lever crime prevention techniques (Engel et al. 2010; Kennedy, Braga and Piehl 1996; Tita et al. 2011), and to understand the relationship among gang members/the internal structure of street gangs (McGloin 2005). From this growing literature, it is clear that SNA has the capacity to provide unique insight into the structure

12

Disrupting Criminal Networks

of street gangs and the positions of importance within these street gangs, and, as a consequence, guide prevention strategies. Crime Prevention Recommendations Given the large number of youth affected by gang membership and the damage to communities that harbor gang related activities, street gangs are clearly a phenomenon worth investigating. Still, despite the large number of gang intervention strategies and policies that have been developed, gang intervention research cannot confidently make a stance on “what works” to reduce or address this phenomenon. As part of a larger effort to evaluate the efficacy of current practices, the application of the principles of social network analysis offers four avenues for developing greater insight into effectively addressing street gang problems. 1.

Identify the social network of rivalries and alliances among street gangs. This will nominate certain street gangs for attention when planning and implementing an intervention.

Papachristos (2009: 74) found that the dynamics among gangs can create an “institutionalized network of group conflict” that act to facilitate violence above and beyond any individual-level motives. In other words, the rivalries among street gangs can prompt and sustain violence. Interestingly, Decker’s (1996) analysis of gang members in St. Louis suggested that this conflict among rival gangs can develop a mythical status, such that even benign behaviors can be perceived as slights, insults and invitations for violent responses. Accordingly, mapping or graphing street gang networks to show how members are connected is a useful and fairly simplistic tool that can facilitate effective use of targeted police strategies. Indeed, “using the gang as the unit of analysis permits examination of…how do gang wars or acts of retaliation unfold?...And, do such interactions really influence group or member behaviors?” (Papachristos 2006: 101). For instance, Boston’s Operation Ceasefire noted that certain gangs were deeply enmeshed in rivalry networks that, according to multiple data sources, were driving the violence among youth in the city (Kennedy et al. 1996). Simply put, focusing on gangs with central positions in the local network of hostile relationships pays dividends because “removing” them has a disproportionate effect on violence. Further, Operation Ceasefire also used this rivalry network to anticipate which other gangs may attempt to take advantage of this gang’s vulnerable state while under law enforcement attention and communicate with them that such behavior would not be tolerated. Recently, Radil, Flint and Tita (2010) combined spatial and

Street Gangs and Co-Offending Networks

13

network analysis using what they call “spatialized network data,” allowing them to demonstrate how measuring social position—both in terms of location in the social network of gangs and their geographic territory— could better illuminate patterns of gang violence. This more nuanced analysis may be particularly, if not uniquely, beneficial in locations with many gangs and more complicated patterns of violence. 2.

Determine the street gang network’s cohesion. This will provide guidance regarding whether some prevention techniques may be effective, ineffective, or have iatrogenic effects.

David Kennedy (2008) argues that the group nature of most crime can be leveraged to improve the effectiveness of the pulling levers deterrence strategy. He advocates for including a collective accountability tactic within this strategic framework, in which fellow gang members are essentially held responsible for each other’s behavior. More specifically, street gangs under attention are informed that if any one of their members engages in violence (or whatever the problem behavior of focus is), then all members will be subject to the deep and wide sanctions of the criminal justice system—that is, the collective will be held accountable for the behavior of one. Several Operation Ceasefire strategies have invoked this tactic, including its first use in Boston (Kennedy, Braga and Piehl 2001) and its replication in Hollenbeck, California (Tita et al. 2011). This tactic includes: “using police records so that after a violent incident by a given gang, all members, regardless of who committed the act, were given the highest priority in terms of probation, parole and warrant enforcement” (Rand Corporation 2003; see also Tita, Riley, Ridgeway, Grammich, Abrahamse and Greenwood 2003). Thus, this tactic essentially uses the street gang members as agents of social control by igniting their own selfinterest and using their social connectedness to supervise one another and urge conformity to non-violence. Of course, this assumes that the street gang under focus is sufficiently organized and connected to support the collective accountability tactic—in some situations this may not be the case. When a street gang lacks cohesion (i.e., has limited connectedness), using the collective accountability strategy may not only be ineffective, but it may actually be iatrogenic and amplify criminal behavior. For example, the Los Angeles Group Guidance project attempted to integrate local gang members into pro-social pursuits, such as tutoring and recreational activities, with the aim of reducing violence and gang involvement (Klein 1968). Unfortunately, the program essentially reinforced the gangs’ status as a group and appeared to have increased the cohesion of the rather disorganized gangs over time, resulting in more gang

14

Disrupting Criminal Networks

crime. By conducting a network analysis of the street gangs under focus, officials will have insight on whether treating the gang as a “group” is reasonable, or may increase cohesion among a set of actors with limited connectedness. Importantly, the collective accountability strategy is not fully “off the table” if a gang has low cohesion; this network analysis may identify a smaller sub-group that is driving violence (or whatever criminal behavior is under focus) and is cohesive enough to be amenable to the collective accountability strategy (McGloin 2005). Indeed, several scholars have observed that street gangs generally have rather low cohesion (Curry and Decker 2003; Decker 1996), which underscores the potential usefulness of identifying smaller, dense sub-groups, which may be driving violence. 3.

Identify the “cut points” or brokers within the street gangs. These are key positions of leverage and vulnerability for intervention programs.

In some cases, intervention tactics may be directed at individuals, rather than groups (or sub-groups). Research focused on the individual position within gangs have found that street gangs are often loosely connected, composed a wide range of positions from peripheral to core members of the group (McGloin 2005; Thrasher 1927; Yablonsky 1962). When confronted with an organization without a clear hierarchy or leadership, it can seem difficult to identify points of leverage; yet, social network analyses have identified a small number of people who occupy a “cut point” position in the gang network. These “cut points” serve as the sole connection among individuals and subgroups within the gang. For example, McGloin (2005) analyzed street gangs in Newark and found that cut points existed in each of the street gang networks studied (i.e., Bloods, Crips, Almighty Latin King and Queen Nation, and Netas). As previously discussed, disseminating a message of deterrence among gang members and implementing the “pulling levers” approach were both key strategies in the success of Operation Ceasefire (Kennedy 1998; Kennedy, Braga, and Piehl 1997). As the only connection among individuals or subgroups of individuals, cut points would be natural—if not ideal—precursors for a deterrence message and/or targets for removal to substantially alter the connections among the gang (McGloin 2005, p. 626; see also Ballester, Calvo-Armengol and Zenou 2004). Gladwell (2002) has argued that within social networks, certain positions have a stronger capacity to spread communications and trends. If intervention efforts seek to warn and make all gang members (or any other population of focus) aware of the consequences to problem behaviors, cut points are a clear place to direct resources towards. Acting as a deterrence “contagion agent,” these individuals could quickly disseminate information

Street Gangs and Co-Offending Networks

15

to multiple groups and/or individuals that would otherwise be more difficult and costly to reach (McGloin 2005). Similarly, cut points act as a unique bridging tie to other individuals or subgroups, a tie that may serve to facilitate the spread of criminogenic values, behaviors, and opportunities or even more concrete products such as illicit goods and funds. By removing such an individual from the network, law enforcement agencies could eliminate this individual’s capacity to forward processes of normative influence that make members of a cohesive group more likely to offend (Akers 1998; Haynie 2001; Sutherland 1947) or they could significantly interrupt the gang’s ability to engage in illegal enterprise.3 4.

Allow for the possibility that street gang networks include people who are not traditional “members”. Individuals who exist in the legitimate sphere/market may be key points of intervention.

In using social network analysis to understand and characterize the social organization of a street gang, it is tempting to solely focus on individuals clearly identified as gang members (e.g., McGloin 2005), but such a focus may miss important aspects of and positions within the gang. More specifically, the connection between the criminal and legitimate sectors of society has been well established by research, particularly among the operation of criminal enterprises and organized crime (Block and Chambliss 1981; Jacobs and Peters 2003; Reuter and Haaga 1989). These actors may be viewed as pawns to the street gang; however, legitimate actors quite often serve to facilitate and ‘broker’ resources and relationships among disconnected sectors of the criminal network (Morselli 2010; Morselli and Giguere 2006). Given their financial resources, business experience, and social capital, these actors are well suited to provide “complementary resources” to the activities of criminal networks (Reuter and Haga 1989). As such, these actors may be essential to the continuity of criminal ventures in the face of increased pressure from law-enforcement intervention. A social network perspective can provide a window into the complexity of this criminal and legitimate sphere symbiosis among street gangs. For instance, Morselli (2006) used SNA to study an illegal drug importation network (i.e., The Caviar Network). He found that although most legitimate actors in the network assumed fairly minor roles, a select few were instrumental in both recruitment of additional traffickers and were more likely to be ‘directors’ of relationships within the network; thus, it appeared that legitimate actors within this illegal drug importation network were not only critical to the criminal operations, but were also vital to forging and maintaining the structure of the criminal network. Based upon the evidence-

16

Disrupting Criminal Networks

driven approach of social network analysis, key participants of the network (i.e., cut points, legitimate actors) can be identified. Targeting resources at these individuals may stand to more effectively dismantle or disrupt street gang operations. Co-offending Networks and Crime Prevention As Warr (1996) has argued, co-offending is an essential characteristic of delinquency, which as this section will argue, should make it of particular interest to those interested in crime prevention. Co-offending networks refers to a collection of individuals who have committed crime together. Many types of criminal networks can include relationships among people that are at least partly defined by collective criminal behavior (i.e., group crime), but co-offending networks are somewhat unique because they are “only” defined by that linkage. A youth may belong to a street gang that is comprised of linkages based on a common identity/label, kinship, friendship, shared time socializing, and criminal accomplices, among other sorts of social ties; but, this same youth’s co-offending network would be comprised of only those fellow gang members with whom he had committed crimes, as well as any other criminal accomplices who are not part of the gang. Though this example may help to clarify the nature of cooffending networks, it is important to recognize that for most youth, relationships with criminal accomplices are not a subset of relationships within some larger criminal organization (i.e., street gangs). To be clear, the majority of youth who commit a delinquent act are not gang members, yet the majority of delinquents commit crime with others. Scholars have noted the group nature of juvenile delinquency for nearly a century (Breckenridge and Abbott, 1917) and many of the discipline’s seminal theories consider criminal behavior to be a group phenomenon (Cloward and Ohlin 1960; Matza 1964; Sutherland 1947). It may be tempting to simply consider co-offending an incidental characteristic of juvenile delinquency, but Zimring (1998: 489) has argued that “no fact of adolescent criminality is more important than what sociologists call its ‘group context’.’’ Indeed, co-offending can be a mechanism whereby youth: (1) are first persuaded to engage in delinquency; (2) become more persistent, prolific and violent criminals; and, (3) become more deeply embedded in and committed to a criminal lifestyle. It is not surprising, then, “specific attention to co-offenders can be beneficial in preventing crime…fully understanding the process of co-offending can be a crucial detail in not only understanding and analyzing crime patterns accurately but for crime prevention as well” (Pourheidari and Croisdale 2010: 4).

Street Gangs and Co-Offending Networks

17

Some of the earliest criminological work demonstrating the use of social network analysis in the study of crime focused on co-offending networks. For instance, Sarnecki (1990; see also Sarnecki 2001) used SNA to study the linkages among juveniles suspected of committing crime in Borlange, Sweden; sociograms were constructed from local police data to demonstrate the links between actively delinquent individuals. Sarnecki (1990) found that approximately 45% of the main juvenile population of interest could be linked together through known offenses. Consistent with previous findings on the distribution of offending in the population, a small number of juveniles not only committed the majority of offenses, but were also all linked together in the same network. Sarnecki (1990) was able to highlight the brevity of co-offending relationships and the impact of differential associations on increasing the likelihood of future offending. More notably, these basic sociograms illustrated the fact that delinquency is a group-based phenomenon that warrants further consideration of the nature and impact of co-offending relationships. It is important to acknowledge that research and crime prevention techniques by no means suggest that in order to understand co-offending, one must rely on social network analytic techniques. But, it is fair to suggest that the main goals and assumptions of social network analysis—to illuminate the way individuals or groups are connected because they are presumed to hold insight for social behavior— translate nicely to co-offending and offer the opportunity for added value. Crime Prevention Recommendations Because group crime can introduce youth to and/or further embed them in criminal norms, networks and lifestyles, it is important for those interested in reducing or preventing juvenile delinquency to focus on co-offending. Co-offending groups may be less obvious and are certainly more transient than street gangs—a difficulty that can pose challenges for intervention strategies. However, the potential dividends are notable and certainly worthy of consideration. This section offers three specific recommendations: not all of which are reliant on SNA techniques and modeling approaches but they are undoubtedly rooted in its orientation and approach. 1.

Reduce the opportunities for youth to gather and socialize with potential accomplices.

Theorists often root criminal motivation in the individual, but it can also emerge from the situation—and one of the most powerful situational motivations for crime and deviance is the presence of others. Perhaps the

18

Disrupting Criminal Networks

best-known theory focused on explaining criminal behavior4 that roots motivation in the situation is Osgood et al.’s (1996) extension of routine activity theory. Osgood and colleagues argued that when youth spend time socializing with friends in unstructured and unsupervised settings, this creates natural inducements for deviance that can persuade most (if not all) adolescents to engage in delinquency. When socializing with friends under such situations, a lack of structured activities opens up opportunities for deviance, the void of supervision means that it is unlikely anyone will intervene to prevent or stop deviant activity, and the presence of other adolescents makes delinquency easier and rewarding (Osgood et al. 1996; Thomas and McGloin 2013). Importantly, Osgood et al. (1996) argue that this situational temptation can seduce even those youth who are committed to pro-social goals and would otherwise not engage in deviant behavior. In this way, adolescents who would not commit delinquent acts when alone may be persuaded to take part in group delinquency/crime under certain conditions. This corresponds well with the more general literature on collective behavior and how the presence of others can have a powerful contextual effect on decision making (Postmes and Spears 1998). Rational choice scholars argue that individuals decide to engage in delinquency using the same cognitive methods and processes they rely on when making decisions about other actions, namely by considering the effort, risks, rewards, excuses/rationalizations and provocations associated with the behavioral options at hand (Cornish and Clarke 2003; McCarthy 2002). For instance, the presence of others can heighten the anticipated benefits of deviance, as it amplifies immediate social rewards in the form of social acceptance, a sense of belonging and even praise/status (see Weerman, 2003). The presence of others can also change the perceived risk of formal sanctions, as “the likelihood of being apprehended is smaller the larger number involved” (Granovetter 1978: 1422; Berk 1974; c.f., Hindelang 1976), as well as reduce the anticipated informal costs of guilt and shame, given deindividuation processes such as anonymity and diffusion of responsibility (Diener 1980; Festinger et al. 1952; Latane and Darley 1968; Zimbardo 1969). Further, in such circumstances the risks and costs associated with inaction may be amplified. Among adolescents, social acceptance and status are highly prized; failing to conform to group norms or social pressure can invite social mockery or exclusion. As Zimring (1998: 489-490) argues: When an adult offender commits rape, his motive may be rage or lust or any number of other things. When a teen offender in a group setting commits rape, the motive may well be "I dare you" or its functional equivalent "Don't be a chicken." When an adolescent robs, steals,

Street Gangs and Co-Offending Networks

19

breaks into a house, or shoots another youth in the company of cooffenders, the real motive for his acts may be the explicit or implicit "I dare you" that leads kids to show off and that deters kids from withdrawing from criminal acts. Fear of being called chicken is almost certainly the leading cause of death and injury from youth violence in the United States. "I dare you" is the core reason young persons who would not commit crimes alone do so in groups.

It also appears that the presence of others can change risk preferences and discounting rates—that is, this situational element can change how one makes decisions, not just the perceived risks and rewards associated with the action under consideration. For instance, Gardner and Steinberg (2005) conducted an experiment in which they randomly varied whether subjects played a video game assessing risky driving behavior alone or in the presence of two peers of similar age. They found that subjects in the group condition took more driving “risks” in order to earn more points in the game; they also found that individuals assigned to the group condition also expressed a greater preference for risks than those subjects randomly assigned to the solo condition. Therefore, when a situation provides access to potential accomplices, this can affect the decision-making process and markedly increase the likelihood that individuals will take part in delinquency (McGloin and Piquero 2009)—again, including individuals who would otherwise not engage in such behavior if alone. Thus, there is a chance for primary intervention—that is, true crime prevention—if one is able to reduce the likelihood of adolescents coming into contact with potential accomplices or situations that are conducive to collective (criminal) behavior. Importantly, this strategy also holds gains for individuals who are already delinquent. For instance, McAndrew (1999: 53) has argued that co-offending can also “lead to sharing of new methods of committing crime, identification of potential targets, information about police activities and opportunities to be part of specific criminal enterprises,” a point introduced by Brantingham and Brantingham (1981). Conway and McCord (2002) found that violence could be transmitted through these temporary linkages defined by shared behavior. More specifically, they found that when juveniles with no history of violence co-offended with accomplices who did have a history of violence, they were more likely to commit violent acts later in their criminal career. More recently, McGloin and Piquero (2010) found that juveniles with non-redundant ego-centric co-offending networks were more likely to demonstrate versatile offending profiles. Their explanation was that nonredundant co-offending networks, that is networks that are less connected and have structural holes, provide a greater chance to learn about varied

20

Disrupting Criminal Networks

norms, skills, and opportunities, facilitating a more versatile criminal repertoire—and, unfortunately, versatile offending during the juvenile years is associated with more serious, prolific antisocial behavior in later life (Loeber 1988). Other research has demonstrated that non-redundant cooffending networks are also associated with higher criminal earnings (Morselli and Tremblay 2004), similarly embedding youth in a criminal lifestyle. Again, one possible means of disrupting these deleterious outcomes of group crime is to address Felson’s (2003) argument that co-offending emerges because potential accomplices are intersect in time and space in shared “offender convergence settings.” These convergence spaces are informal settings that are common to potential offenders’ daily routines, and include schools, parks, and other places where adolescents can come together typically without the interference of capable guardians (see also Andresen and Felson 2010). As Andresen and Felson (2010: 78) argue, “crime rates may be reduced simply by making it more difficult for cooffenders to find one another.” Such a strategy is not without risks, however (Andresen and Felson 2010: 78): If…public officials begin to interfere with offender convergences, that raises real issues about civil liberties and brings to the fore questions about when, where and how society should regulate offender assembling patterns. Indeed, it becomes very important to learn empirically how much harm co-offending produces, whether such harm can be reduced by interfering with offender convergences in search of co-offenders, and how such interferences conflict with the goals of a free society.

Perhaps the best illustration of this difficulty is the history of several U.S. cities wrestling with anti-loitering ordinances that were traditionally focused on reducing gang activity. Such ordinances raise valid and compelling concerns about individual rights and the potential for discriminatory practices (e.g., Poulos 1995; Strosnider 2002). It is possible to leverage a social network approach to identify these criminogenic spaces with more nuance and empirical validity. For example, Bichler et al. (2014) utilized SNA to identify the presence of juvenile offender convergence spaces; specifically, they crafted regional sociograms that depicted the use patterns for facilities frequented by juveniles in their discretionary time. Consistent with other criminological phenomena, Bichler et al. (2014) conclude that that a small number of settings occupy a large percentage of youth activity within their social environment. Many of the settings identified as convergence spaces were not only close to schools, but also private facilities (e.g., shopping complexes, movie theaters) characterized

Street Gangs and Co-Offending Networks

21

by a lack of structured activity and supervision. This finding is hardly surprising in light of Osgood et al.’s (1996) aforementioned argument about the criminogenic effect of unstructured and unsupervised socializing with peers. Once these convergence settings have been identified, one possible manner in which to regulate juveniles’ routines without infringing on civil liberties would be to offer more structured social options for leisure time— that is, attempt to change how they socialize instead of stopping their ability to congregate. Rather than socializing with fellow adolescents with no agenda, if schools, communities and other city agencies offer recreation such as clubs and sports, especially with a supervisor, coach or counselor present, this could reduce the likelihood that the situation will induce and support collective deviant behavior (Osgood et al. 1996). For instance, Mahoney and Stattin (2000) evaluated the effect types of leisure activity among adolescents in Sweden and found that youth who participated in structured activities (e.g., regular meeting times, led by an adult, with similarly aged peers) were less likely to engage in anti-social behavior and were exposed to far fewer delinquent peers. This finding stands in stark contrast to the deleterious effects on adolescent behavior as a result of participation in unstructured youth recreational centers (Mahoney, Stattin and Lord 2004). Thus, given the large percentage of time adolescents spend in discretionary or unsupervised settings (~40% of their waking hours), concerted efforts to provide structured activities during times that would otherwise be used for unstructured socialization could lead to a reduction in anti-social behavior (Bartko and Eccles 2003). In order to enforce such a change, a private-public partnership that requires organizations to mitigate the ‘magnetism’ for juvenile offending in these convergence settings through place managers and performance ordinances should be considered as a potential mechanism for crime prevention (Bichler et al. 2014). 2. Advertise the risks and costs of engaging in group crime. One of the reasons that group crime can tempt even youth who would not otherwise engage in crime is because it offers the opportunity to diffuse responsibility and essentially “side step” or neutralize the anticipated costs of deviance (McGloin and Piquero 2009; see also Clarke, 1980, 1983; Sykes and Matza 1957). Taking strides to strengthen adolescents’ awareness of the shame and guilt that can accompany deviance, or at least remind them of it, may help to “remove the excuses” that they use to rationalize their behavior when presented with the opportunity for group crime. This strategy of making it more difficult for would-be offenders to rationalize their behavior has been incorporated into situational crime prevention (SCP; Clarke and

22

Disrupting Criminal Networks

Eck 2005; Clarke and Homel 1997; see also Wortley 1996), offering some specific guidelines for intervention tactics. Given the fact that the motivation for group crime often resides in the situation (McGloin and Piquero 2009; Osgood et al. 1996), this recommendation fits nicely within the SCP framework because the idea is to induce guilt or shame at the “point when criminal decisions are being made” (Clarke 1997: 17). To be clear, if law enforcement were simply to visit schools or send out flyers reminding youth that “vandalism is illegal and wrong,” any impact this tactic had would likely be directed at propensities or dispositions for deviance. But, again, the notion of collective behavior is that it can tempt even people with low dispositions for deviance because the immediate situation affects decision-making in that particular time and place. Thus, it will take some investments and problem analyses to determine where, when and what sorts of group crime are problematic in the local jurisdiction so that these reminders are delivered in relevant situations, but it certainly possible. For instance, Bichler et al.’s (2014) aforementioned study identified juvenile offender convergence spaces with SNA in order to more thoroughly understand the relationship between setting characteristics and offending behavior. Their results indicated that a small percentage of facilities occupy a ‘highly magnetic’ and relatively stable position in the social realm for the youth studied (Bichler et al. 2014). Consistent with components of other crime prevention programs, targeting these identified spaces could lead to more effective dissemination of information and resources that can be used to disrupt and suppress the formation of accomplice networks. This type of strategy could also be implemented by school-based interventions that seek to alter existing beliefs about risks and costs, develop peer resistance, and promote crime prevention for at-risk students. For instance, if an analysis reveals that youth tend to hang out unsupervised in parking lots after sporting events at local schools, staff can post rules about appropriate behavior, reminding students that any illegal activity on school grounds can lead to suspension, as well as “alert” student consciences by placing signs informing students that fixing any property damage could affect funding for school extracurricular activities (see Clarke and Eck 2005). 3. Focus resources on individuals who occupy central roles in a cooffending network(s). Most police departments have at their disposal the necessary data for constructing social networks of co-offenders (e.g., Sarnecki 2001). Indeed, the requisite information for building these networks (i.e., youth arrested or contacted by police together due to suspected joint criminal activity)

Street Gangs and Co-Offending Networks

23

typically is easier to access than that necessary to construct street gang networks. If departments, or their civilian analysts/academic partners, have the resources to invest in using such data to build and analyze accomplice networks, this would provide insight on which individuals occupy central positions. As Sarnecki (1990: 47) observed, focusing law enforcement attention on these individuals has the potential to provide disproportionate benefits (see also Andresen and Felson 2010): The more interesting issue, however, is the effect of incapacitation of the central figures in the network. This type of incapacitation can mean that the offenses they would otherwise have participated in will not occur at all—even if other members of the network are not incapacitated. Furthermore, these central figures in the network, as has been shown, play an important role in holding the network together at an early stage. One possibility is that if they disappear, the whole network will fall apart, which in turn can significantly decrease the number of offenses committed by the network's other members. In addition, these central figures are responsible for the passing on of the "delinquent tradition" to new generations of juvenile delinquents.

It is possible that these central nodes reflect individuals who occupy, or youth who are transitioning into, the role of criminal mentor. Most notably discussed by Sutherland (1937), criminal mentors can “tutor” less experienced offenders in the act of crime and the recognition of opportunities, as well as serve as the liaison to other actors in the criminal world (Morselli, Tremblay and McCarthy 2006), broadening their connections to and enmeshment in the criminal lifestyle: “Any man who hits the big-time in crime, somewhere or other along the road, became associated with a big-timer who picked him up and educated him” (Sutherland 1937: 23). In his narrative of the delinquent Sidney, Shaw also relayed statements that underscored the role of mentorship “on the job” (i.e., mentorship rooted in criminal collaboration/co-offending): “I would walk behind him and as soon as he would pick up a piece of fruit I was supposed to do likewise. It took lots of practice and he had to set many examples before I could at last gain enough courage to follow suit” (Shaw 1931: 58; see also Sullivan 1989). Incapacitation of such individuals may be possible in circumstances where the relevant legal requirements are suitably met, but there are alternative options as well. The second recommendation in this section focused on “removing excuses” for group behavior, which arguably would be most effective for those youth with lower criminal propensities. For more experienced delinquents occupying central positions in co-offending networks, however, inducing guilt or shame may not be sufficient to alter

24

Disrupting Criminal Networks

their decision-making process. This does not mean that highlighting potential costs and risks associated with group crime would be ineffective, but rather that the content of the message may need to shift. Scholars have long noted the risk of criminal collaboration, which entails accepting the uncertainties of an accomplice’s abilities, intent, and whims (McCarthy, Hagan and Cohen 1998). Co-offenders can greatly complicate criminal endeavors with mistakes or over-reactions, and can always turn each other over to law enforcement (Wright and Decker 1994). It should not be surprising then that research suggests group crime is more likely to be detected than is crime committed alone (Erickson 1973). It would be useful to strongly communicate to the youth who occupy central positions in cooffending networks that these risks and uncertainties are valid and would increase their vulnerability of getting caught, as well as the possibility of being charged with crimes such as conspiracy and incitement. Of course, if these youth are sufficiently motivated, this strategy may simply change the nature of their crime—that is, shift it from group to solo crime. Even so, that would pay dividends by reducing caseloads for criminal justice agencies (see Andresen and Felson 2010), and, more importantly, may impede these youth from recruiting other, perhaps less experienced, youth into the criminal endeavor(s). Concluding Comments The previous sections have described how incorporating social network analysis into current law enforcement operations, school-based interventions, and other juvenile justice services can facilitate additional insight into two important juvenile networks: street gangs and co-offenders. Each of these networks pose sizeable problems to society and requires a significant amount of resources to address. As a result, efficient and effective strategies need to be considered in order to both understand and produce change in tempering anti-social behavior among adolescents. As outlined, social network analysis offers several practical considerations for use in addressing these phenomena. Namely, this approach seeks to aid various agencies and organizations in identifying existing relationships/interactions, direct resources to key individuals, and coordinate meaningful responses to group processes. It is worth reiterating that social network analysis is both a method and a means of driving us towards developing an empirically nuanced and thorough understanding of interactions among subjects of interest. Specifically, this approach reinforces the group nature of offending behavior during adolescence and affirms that it is more than just a coincidental characteristic of the offense. Social network analysis should however not be

Street Gangs and Co-Offending Networks

25

viewed as a panacea to juvenile offending, but should be considered a tool for law enforcement and other stakeholders to use in their efforts to address this type of behavior. It should also be used to assess the impact of any crime prevention strategy, as social networks are dynamic and fluid (Morselli 2010; Sarnecki 1990). Understanding how they change (or do not) in response to intervention efforts is crucial in determining whether the structural characteristics and group dynamics that produced and facilitated anti-social behavior in the first place were substantively affected beyond the point of intervention. Notes 1. Criminology undeniably places much theoretical and empirical attention on the consequences of having a friendship group comprised of youth who engage in and/or endorse deviant behavior, but gangs and co-offending groups are uniquely defined by their criminal orientation and action, making them of particular interest to law enforcement. 2. This sample is meant to be nationally representative, but is nonetheless leaves out more than 10,000 other local police departments in the United States. 3. It is important that one continue to monitor network structure after such a strategy. It is possible for gangs to re-organize quickly; it is also possible that the removal of a powerful individual can result in violent struggles and demonstrations for status within the gang, as it negotiates a new structure (Short and Strodtbeck, 1965). 4. Note that Cohen and Felson’s (1979) routine activity theory explains victimization/crime, but was not intended to explain individual-level offending.

3 Applying Group Audits to Problem-Oriented Policing1 Michael Sierra-Arevalo and Andrew V. Papachristos

THE HUMAN BRAIN HAS A STAGGERING AMOUNT OF STORAGE space—about 2.5 petabytes (or 2.5 million gigabytes). Granted, humans never use up all of the available space, but the fact remains that people are able to store massive amounts of information on topics ranging from song lyrics to the names of bones and muscles in the human body. Not everyone knows as much about a given topic, though; an attorney is unlikely to be able to rattle off the names of the 26 bones in the human foot, just as a physician is unlikely to know the proper legalese for preparing a writ of habeas corpus. Everyone has his or her area of expertise. Unsurprisingly, police and other law enforcement personnel have specialized knowledge of the things and people they have dedicated their careers to learning about: criminals, criminal acts, and the streets in which criminals commit crimes. Unfortunately, much of this detailed information is locked away in the heads, notepads, and cell phones of law enforcement personnel. And while official data sources such as the FBI’s Uniform Crime Report (UCR) or police department arrest and criminal history records certainly provide a great deal of information, these data are prone to biases resulting from under- or selective-reporting (Black 1970). More importantly, official data oftentimes lack the nuanced and seemingly intangible information needed in modern, focused-deterrence policing strategies such as Project Safe Neighborhoods and the Boston Gun Project (Papachristos et al. 2007; Kennedy et al. 1997, 1996; Tita et al. 2010).

27  

28

Disrupting Criminal Networks

In particular, micro-level intelligence of the street, such as who associates with whom or the location of local hangouts, is rarely reflected in institutionally-produced data. It is this kind of intelligence, however, that is almost second nature to police officers and other law enforcement personnel who walk their beats and patrol city streets on a daily basis. Police know, and can describe in great detail, minute details of the people in the communities, their activities, and their associates. Further, it is the daily, street-level interactions with offenders, parolees and probationers, and the local community that allows law enforcement to accrue experiential assets—“experience, observations, local knowledge, and historical perspective” (Kennedy et al. 1997: 220) about how both criminal and noncriminal life unfolds in their municipality. These experiential assets include critical information concerning often unseen phenomena such as the boundaries of gang territory, the relationships that link feuding or allied groups, and the ever-changing nature of street gang membership and structure. Over the past decade, a data collection procedure called the group or gang audit2 has been developed for use in focused-deterrence policing and violence prevention initiatives. Group audits extract this type of experiential, “on the ground” intelligence through focus-group style working sessions with law enforcement and other gang “experts,” such as case workers or members of community organizations. The information obtained through the audit lends itself to use in social network analysis, which is itself particularly useful for linking individual stores of knowledge in order to examine patterns of the whole. Extracting such information from the heads of experts and subjecting it to social network analysis can produce a useful tool in violence prevention and policing efforts—literally, a map of the “unknown universe” of violent groups, network diagrams of the patterns of conflicts and disputes that are driving gun violence in U.S. cities (Kennedy et al. 1997). This chapter describes the audit process, how it relates to focuseddeterrence strategies and social network analysis, and in what ways the group audit, in conjunction with social network analysis, can support and strengthen violence reduction initiatives. We begin by briefly reviewing the audit process, paying particular attention to systematic methods of relational data gathering and analysis. We then summarize the findings from audits in New Haven and demonstrate how the analysis of the networks among violent groups can address some definitional problems surrounding groups and gangs, as well as provide unique insight into groups and their relationships.

Applying Group Audits to Problem-Oriented Policing

29

A Brief History of Focused-Deterrence Strategies In conjunction with social network analysis, the information gathered in the course of the group audit provides a more complete view of the contours of the violence landscape in a city. To understand how this kind of schematic can be used in violence reduction work, it is imperative to first have an understanding of the violence reduction efforts that make use of this streetlevel intelligence. Beginning in the late 1970s, American policing underwent an evolution from the standard model of policing to a more problem-oriented approach (Goldstein 1979; Weisburd et al. 2012). In the former, police activity was enforcement heavy (i.e., random patrols and mass police sweeps), reactionary, and applied in a uniform manner across municipalities despite variation between cities in crime rates and departmental resources (Weisburd and Eck 2004). Problem-oriented policing (POP), on the other hand, takes a more proactive approach by identifying specific underlying issues that can be targeted to reduce crime and delinquency. Further, POP also extends the role of the police beyond that of crime fighters; it stresses a positive relationship between police and community in order to more effectively harness the community’s resources and on-the-ground knowledge to more effectively hinder criminal activity (Goldstein 1979; Weisburd et al. 2010). Indeed, the development of POP developed in tandem with the rise of community-oriented policing and the growing popularity of police-community violence prevention efforts. With the popularization of the POP model throughout the 1980s and 1990s, a variety of crime-reduction strategies have been implemented that seek to capitalize on the POP paradigm. One type of strategy that has found particular traction in cities across the U.S. is based on focused-deterrence theory. Unlike classical deterrence which depends on indiscriminately swift, certain, and severe punishment to dissuade both would be and former offenders from (re)offending (Gibbs 1975; Zimring and Hawkins 1973), focused-deterrence builds on the robust finding that crime is committed by a very small subset of the population in highly localized areas (Braga et al. 2009; Kennedy et al. 1996; Spelman 1995; Weisburd et al. 2004, 2012), and targets these individuals (and groups of individuals) for focused intervention, prevention, and enforcement attention. One of the first steps of a POP process is to identify actionable problems in the municipality towards which its efforts can focus. One problem that has been identified and addressed with focused-deterrence strategies in cities such as Boston, Chicago, and Cincinnati is the issue of gun violence (Braga and Weisburd 2012; Engel et al. 2010; Papachristos et al. 2007). Because the majority of a city’s violence is concentrated among

30

Disrupting Criminal Networks

an exceedingly small percentage of the population (Braga et al. 2008; Kennedy et al. 1996; Papachristos and Wildeman 2014; Papachristos et al. 2012), focused-deterrence strategies targeting gun violence concentrate their efforts towards the individuals most likely to be victims or offenders—those involved in violent street groups. One strategy developed by such focused deterrence efforts employs offender notification forums or call-ins in which law enforcement, social service providers, and community members speak to those individuals and groups involved in ongoing violence. These meetings are used as way to communicate directly with those most actively involved in gun violence (as either potential victim or offender) and to disseminate the program’s message.3 Though the group audit itself is often coordinated and organized by police personnel, the cooperative nature of the call-ins are similar to the goal of modern community policing, whereby police aim to become create collaborative community partnerships to increase trust in police and develop solutions to community issues (U.S. Department of Justice 2012). But before a call-in can be organized and implemented, an important issue remains unresolved: how to identify the groups and individuals within those groups who should be targeted for the program’s efforts. One method for identifying viable candidates for forums and call-ins is the group audit. What is a Group Audit? The idea of the group audit is straightforward: working sessions that bring together law enforcement, outreach workers, and researchers to conduct a “gang/group census.”4 The main data collection objectives of the audits are to gather information on: 1. 2. 3.

Which groups in a given city or jurisdiction are involved in gun violence; Group membership, turf location, and (illegal) activities; and Intergroup relationships— i.e., alliances, disputes, mergers, splits and so on.

The idea of getting detailed information on violent groups and their activities is not a novel one. Informal audits frequently happen within specialized gang or narcotics units as a part of routine police work. In addition, street outreach workers have been trying to create data on gangs and their members since the 1950s (see, Short and Strodtbeck 1965; Tita and Papachristos 2010). Since the 1990s, researchers have been formalizing this process of information gathering and sharing into a methodology that elucidates the often hidden world of violent streets. During the height of a

Applying Group Audits to Problem-Oriented Policing

31

youth crime epidemic in Boston, David Kennedy and colleagues (1997) described one of the first academic-lead group audits as a process of “mapping the unknown universe” of street violence—something that law enforcement and violence prevention workers knew intimately, but only rarely tried to organize and systematically analyze. The innovation of the group audit is the systematic pooling of various and often disconnected information streams from law enforcement5, followed by systematically combining that information into a more centralized, complete, and accessible repository of data on groups, turf, and networks of feuds and alliances. One of the end goals of this audit process is to develop a set of data amenable to social network analysis, hereafter SNA. SNA refers to both a theoretical lens and a set of methods for examining the relationships between actors (Wasserman and Faust 1994). In terms of theory, SNA emphasizes the relationships, the connections, between social actors. As such, SNA asserts the importance of the interdependence between entities, be they individuals or groups, and interprets the social world as one composed of patterned interactions between multiple entities (Wasserman and Faust, 1994; Wellman, 1983). Criminologists have used SNA in a variety of research contexts, such as analyzing the structure of criminal groups like gangs (Fleisher 2006; McGloin 2007; Papachristos 2009; Tita et al. 2010), organized crime syndicates (Klerks and Smeets 2001; McIllwain 1999; Morselli 2003) terrorist cells (Pedahzur and Perliger 2006; Xu and Chen 2003), and whitecollar conspiracies (Baker and Faulkner 1993, 2003). While the criminological community has been making use of SNA for some time, law enforcement and violence reduction practitioners are only just beginning to harness the analytic and theoretical strengths of a “network approach” to crime and violent groups. The group audit is a methodological tool that facilitates the collection of the relational data needed to use SNA and, in so doing, enhances our knowledge of the underlying structure of violent groups. The types of data obtained from the audit process are inherently relational—they describe not just the groups themselves, but the relationships among the groups. Network methods and statistics can then be used to analyze patterns that would have otherwise remained in the heads and notebooks of individual police and gang experts. It is the combination of multiple information streams and the systematization of the collection process that is the real improvement on methods that date back to classical works in anthropology. Similar to how anthropologists have recorded the relationships of those who broke bread with one another or how young brides-to-be were selected or traded

32

Disrupting Criminal Networks

between tribal families (for a review, see White and Johansen 2005) researchers in the group audit record the relationships between violent street groups. The process differs in that it is not what a researcher observes in the field that is recorded, but instead the combination of multiple, expert perspectives on what the complex set of intergroup relationships are. These relationships are then codified and centralized6, creating a markedly more complete and nuanced body of knowledge on violent street groups. Performing a Group Audit: The New Haven Case Despite a population7 smaller than Bridgeport and Hartford, New Haven leads Connecticut in violent crime. In 2010, Hartford had 1,624 violent crimes and Bridgeport had 1,412, New Haven experienced 1,992, outpacing Bridgeport and Hartford by about 27 percent and 40 percent. In the same year, New Haven’s violent victimization rate of 15.95 per 1,000 outpaced or closely approximated the victimization rate of larger cities across the country, including Baltimore, MD (14.6), Memphis, TN (15.4), and Oakland, CA (15.3). While aggregate crime statistics show that violence victimization decreased in New Haven between 2010 and 2011 from a rate of 15.95 to 13.44 per 1,000, the most recent crime statistics reported by the New Haven Police Department indicate that, between 2011 and 2012, there has been a 9 percent increase from 1,423 violent crimes to 1,555.8 Recently, New Haven has moved to re-engage the community in policing efforts and, more specifically, is part of a state-wide project employing a focused-deterrence strategy to reduce gun violence by targeting a specific audience in high-violence areas. This focused-deterrence program, called Project Longevity, hinges on disseminating a customized message concerning the cessation of gun violence to those most likely to be shooters and victims. In the remainder of this chapter, we use an extended case study of how group audits are currently being used in a focused-deterrence program in New Haven to illustrate how the audit procedure and network analysis techniques are being used. We also discuss how a network approach and perspective on violent groups can address some common problems encountered while conducting an audit. Before beginning, it is important to stress that the audit process is iterative and dynamic; it changes as groups change and as perceptions that police have of these groups change. As such, the New Haven audit should be seen as a particular audit done in a particular city—as an example and not necessarily an ideal.

Applying Group Audits to Problem-Oriented Policing

33

Project Longevity: Background and Structure of Audit Project Longevity debuted in New Haven near the end of 2012. Based on a similar focused-deterrence intervention strategy used in Boston’s Ceasefire, (Braga, Kennedy, Waring, et al. 2001; Braga, Kennedy, Piehl, et al. 2001), Project Longevity set out to reduce gun violence driven by violent groups. A key part of the intervention in New Haven is the use of call-ins (also referred to as “forums”) where high-risk individuals hear a strong, unified message from law enforcement, social service providers, and community members concerning the need for gun violence to stop. As a lead up to the first call-in, Project Longevity leadership called for a group audit in light of a lack of systematic, up-to-date information on violent groups within the city. At the onset of Project Longevity, there was a widely held belief on the part of law enforcement that much of the gun violence in New Haven was the result of interpersonal disputes among young men involved in gangs or other street-oriented groups. Group audits were one of the initial methods used to determine: (1) which groups were involved in violence; (2) the membership and activities of identified groups; and (3), the patterns of relationships among these groups. The first in a series of audits took place in a large meeting room at the U.S. Attorney’s office in downtown New Haven.9 The location for an audit largely comes down to a matter of coordination and convenience, and the law enforcement driven nature of the audit in New Haven made the centrally located U.S. Attorney’s office an ideal location. The audit gathered law enforcement from a variety of agencies and of varying ranks, and included both patrol and high-ranking police officers of the New Haven Police Department, parole and probation officers, the Connecticut Department of Corrections, investigators from the U.S. Attorney’s Office, and agents from the DEA, FBI, and ATF. All totaled, there were approximately 100 members of the law enforcement community present, with 10 researchers on hand for the data collection. The audit began, much as this chapter did, with a brief presentation by an informed law enforcement official on what focused-deterrence is about, how strategies like Ceasefire and Project Safe Neighborhoods target specific individuals instead of entire communities, and how the forthcoming audit was going to help researchers assist law enforcement in their efforts to reduce gun violence and support the Project Longevity initiative. Of the points covered during the introductory presentation, the definition of “gang” or “group” is paramount to the success of the audit. In New Haven the working definition for a gang or group was: “three or more individuals engaging together in violence and/or high risk behavior”. As noted by the

34

Disrupting Criminal Networks

presenter, this is a relatively wide definition for a group/gang, and makes no reference to popular conceptions of gangs. After the preliminary presentation, participants split up into groups based on which geographic area of the city they worked in, had parole/probation/supervision cases in, or where they had particular expertise. New Haven is split up into 10 police districts10 (see Figure 3.1), and the audit was performed with the district as the unit for each group discussion, although participants were asked about relationships that extended beyond their specific geographic area. Audit groups ranged in size from as few as four officers to as many as 10. Some groups had only patrol officers while others had both patrol and high-ranking officers from NHPD, federal agencies, and parole and probation. Each group had three researchers assigned to it; two acted as coders to take down group information, and the third acted as a moderator that guided participants through the audit process, asking questions to gather the necessary information on groups and gangs. Each audit group was also provided with a large, laminated map of the particular police district they were familiar with, along with several dry erase markers of different colors to mark the geographic location of group hang-outs and turf. Figure 3.1

New Haven Police Districts

Applying Group Audits to Problem-Oriented Policing

35

The process for identifying the groups in the district began with questions from the moderating researcher, such as, “Who is the most violent group in this district?” While the goal of the group audit is to capture information on all groups in the area, beginning with a prominent group can help spark discussion. In the course of group identification, the moderator and participants marked street faces and locations on the map identified as being group turf or hangouts, and each identified area was labeled with the group name.11 Along with information on turf or hangout locations, the moderator also asked a series of questions to fulfill the second goal of the audit: gathering information about the identified group’s activities. Such questions included whether or not the group is involved in shootings or homicides, whether the group’s level of violence appears to be changing, whether or not the group sells drugs (and if so, what kind), and in what other types of criminal behaviors the group might be involved (e.g., robberies, home invasions, etc.). In addition, audits were flexible enough to allow the moderator to change or adapt questions to try and gather whatever information the violence reduction strategy and law enforcement leadership felt would be beneficial in the aims of the program, or most accurately described the nature of the group and its activities. To achieve the third goal of the audit (collect information on intergroup relationships), moderators asked questions pertaining to conflicts and alliances between groups. In general, broad questions like “Does this group have a beef or feud with anyone else in the city?” or “Is this group allied with anyone else in the city?” are good starting points for gathering this information. This is also an effective way to begin to shed light on other groups that are present in the district, their relationships with groups within that district, and their relationship to groups and activities in other areas of the city. Furthermore, these questions also provide a way of cross-checking what groups were identified by one group of participants but not another, providing an important check of internal validity. This process of asking targeted questions to uncover what groups exist in the district, where they claim turf, and who they are feuding or allied with is repeated until participants cannot identify any more new groups. To be sure, the length of time to conduct such an audit varies greatly. The audit in district 1, for example, (see Figure 3.1), took very little time since there were no identified groups. This is likely because district 1 is largely composed of local businesses, Yale, government buildings, and a large, public green. In high-violence districts such as district 7, however, many groups existed, and great time and care had to be taken to exhaustively probe for all the groups in the area, as well as the complex set of relations among them.

36

Disrupting Criminal Networks

Social Network Analysis of Violent Groups The initial New Haven audit, as well as a follow-up audits targeting specific districts, identified 52 active groups, with at least one group or gang identified in nine of the ten police districts. Looking to Figure 3.2, the line features and points mark street faces and geocoded addresses that police identified as being group turf or hangouts.12 Figure 3.3 details the number of identified groups in each police district—clearly, not all districts have an equal prevalence of gangs. As such, 60% of districts have four or fewer identified gangs, and only one district has more than 10. Moving beyond simple counts of groups per district, relational data its analysis provides a wealth of information that is often obfuscated by prevailing assumptions about violent street groups and their activities. The data gathered on the feuds and alliances between groups shows that, at the time of the audit, there were 60 alliances and 21 feuds among these identified groups. Like the distribution of gangs, however, the distribution of feuds was also uneven across gangs—only 40 percent of the identified groups were reported as being in a feud. Figure 3.2

New Haven Police Districts and Identified Group Turf and Hangouts

Applying Group Audits to Problem-Oriented Policing

Identified Groups/Gangs

Figure 3.3

37

Frequency Distribution of Groups by Police District

14 12 10 8 6 4 2 0 1

2

3

4

5

6

7

8

9

10

Police District Number

One of the key outputs from the audit process is a social network map of the group feuds and alliances in a city (Kennedy et al. 1997). Figure 3.4 illustrates these two types of networks by plotting the feuds and alliances among all groups that law enforcement identified. Each node represents a unique group, and the size of each node is proportional to that node’s degree centrality, the number of ties the node has to other groups. The labeled nodes in the top section of the figure are the three groups with highest degree centrality. For comparison, the same three groups are labeled in the alliances sociogram. The distribution of feuds and alliances within the network reveals that just as groups concentrate in different geographic areas, so too do they concentrate in social space. With only 21 nodes in the total feud network, this indicates that less than half (45.6%) of all identified groups are involved in any feuds. Of those who are involved in at least one feud, only a handful of groups are involved in more than a few feuds. The majority of nodes in the network (61.9%) have only one or two feud ties, a few have three (23.8%), and only three outliers are involved in six or seven feuds. Clearly, with over half of all groups not involved in feuds, and with most who are not being involved in very many at one time, groups’ feuds and their resultant violence are not distributed equally among the groups. In short, only a few groups drive the majority of feuds in New Haven. The insight that it is only a handful of groups and individuals that account for the lion’s share of a city’s violence is only one insight that social network analysis can give to law enforcement and others working to reduce urban violence. A common problem that social network analysis can help alleviate is the issue of nation conflation. Nation conflation refers to a

38

Disrupting Criminal Networks

tendency for law enforcement (and others) to conflate several smaller, distinct groups into larger and more generally named groups or gang “federations” such as the Blood and Crip federations of California, or the People and Folks nations of Chicago. This is especially problematic because labeling a small group as “Crips” de facto assigns that small group the same organizational history, image, and notoriety of the larger and criminally sophisticated California Crip organization. In other words, assigning a small group a big label biases the audit (and officers) into thinking a small, less organized group is in fact a much larger and more organized than it actually is. Figure 3.4

New Haven Feud and Alliance Networks

Applying Group Audits to Problem-Oriented Policing

Figure 3.5

39

R1, R2, and WR2 Feuds

For instance, in one of New Haven’s northern police districts (district 7), officers adamantly claimed that that the northern section was largely controlled by the “Crips.” After pushing the officers for more detail, it was revealed that what was at first a large section of territory simply controlled by Crips was actually home to three distinct groups or subgroups: R1, R2, and WR2. As told by officers, these three sets do indeed affiliate themselves as being “Crips”, though the officers also stipulate that there is no evidence to suggest that they have any real ties back to California-based Crips.13 More importantly, all three groups operate uniquely and independently. While they share the claim to Crip-hood and connection to local geography, they do not share any clear organizational structures, nor do they coordinate their activities. Figure 3.5 illustrates this point. As shown, while R1, R2, and WR2 do have some common feuds, they by no means perfectly overlap. While all three have a feud with node e, R2 has three feuds (with a, b, and c) that WR2 and R1 are not involved in. Similarly, while only R2 and WR2 are both feuding with d, only WR2 and R1 are in conflict with f. If these identified groups really were all part of some larger, cohesive Crip organization, we would expect them to feud with the same rival sets—this is clearly not the case. It is precisely the ability of social network analysis to visualize the “real” situation on the street that makes it such a valuable tool

40

Disrupting Criminal Networks

for combating problems like nation conflation, as well as providing law enforcement and violence reduction strategies with more accurate information. Another common problem that the group audit and social network analysis can help alleviate is the problem of geographic conflation. Just as smaller, independent groups are often lumped into larger, more general groups or gang nations such as the Bloods or Crips, so too are groups sometimes conflated based on common geographic location. In short, it is often assumed that groups that share geography also share some organizational connection. Continuing to use the groups identified in district 7 as an example, not only did officers continually referred to them as “Crips”, but they were also discussed by officers as being part of an area called “The R”, or as being part of a larger group called “The R”. It was not until well into the audit process that the research team discovered that The R referred not to a group or gang, but instead referred to a geographically defined neighborhood— The R is a shortening of “Read Street,” a large street that spans the length of the territory controlled by the three groups (see Figure 3.5). In this case a single street became the common category by which law enforcement defined unique groups as a single, homogenous entity. Figure 3.6

R1, R2, and WR2 Turf

Applying Group Audits to Problem-Oriented Policing

41

The geographic areas that police referred to in lieu of more accurate, gang-specific names were much larger than the actual territory controlled by any one gang. Although Read Street does indeed run through the entirety of the territory controlled by R1, R2, and WR2, each group controls specific sections along Read Street. This can be seen in greater detail in Figure 3.6. Note that WR2 does not have its own territory. This is because the WR2’s turf is dictated by where R2 claims territory. This is likely because of the power dynamic that has WR2 serving as R2’s “piss boys.”14 As with nation conflation, it was the group audit’s systematic gathering of relational data that made the micro-level, geographic distinction shown in Figure 3.6 possible. What’s more, it is possible to combine the information shown in a basic sociogram (Figure 3.5) with that of a geographic map (Figure 3.2), effectively being able to show how relationships between groups exist in both social and geographic space. Figure 3.7 is an example of how relational and geographic data can be combined to provide a more nuanced view of intergroup dynamics. Figure 3.7

Read Street Groups and Feuds

42

Disrupting Criminal Networks

The grey nodes from Figure 3.5 have been plotted on the group turf map shown in figure 3.2. Note that the grey nodes in Figure 3.7 are the three Read Street groups, and the lighter nodes are all those groups that have are involved in a feud with at least one of the Read Street groups. All nodes are plotted over their corresponding group turf and hangout locations. Despite the fact that all of the information that makes up the above figure has already been presented in some fashion, the combination of geographic and relational data yields new insights. First, it is apparent that the feuds the Read Street groups are involved in are not restricted to a particular geographic area. Instead, the groups feuding with R1, R2, and WR2 are spread out over four different districts, with only district 8 having more than a single one of those groups. Note also that the conflicts the Read Street groups are involved in cross large areas of the city, and groups that are much closer, such as those also in district 7, are not in conflict with Read Street. By the same token, while the Read Street groups are certainly involved in several violent feuds, they are by no means involved with them indiscriminately—the vast majority of groups in the city have no qualms with anyone from the R. The pattern of relationships seen in Figure 3.7 suggests several things. First, the conflicts in which R1, R2, and WR2 are involved in are not likely motivated by competition over turf. If this were the case, their most likely adversaries would be the groups that are one or two blocks away, but instead they feud with groups miles away. Second, this figure also speaks to the issues of nation and geographic conflation. As shown in previous figures, there are three unique groups in the northern section of district 7, and each group has a unique set of relationships (though they do have feuds in common with the other Read Street groups). When these relationships are plotted geographically, however, the spatial differences in the feuds between the Read Street groups and other groups in the city become apparent. Only WR2 has a feud that extends as far out as district 5, and only R2 and WR2 have feuds with far away groups in district 8. In contrast, R1’s feuds don’t extend as far geographically, and are restricted to districts 4 and 6. While it’s difficult to say why R2 and WR2 seem to be more willing to engage in feuds with more distant groups, providing a geographic view of the social network analysis highlights one more dimension along which these groups differ. This further strengthens the claim that, although they share a neighborhood and even claim a common Crip allegiance, the groups that call Read Street home are unique entities that do not necessarily act in concert or towards a common goal.

Applying Group Audits to Problem-Oriented Policing

43

Conclusion Good police officers know the streets they patrol. They are keenly aware of the social pulse of their beat, and they know the people who call those streets home. Importantly, good cops know who the troublemakers are. They can recite the names and nicknames of known offenders, what they have been arrested for in the past, where they hang out, who they hang out with, and countless other pieces of information. And, they put that information to use every day in the course of their work. However, while each officer has a wealth of knowledge about the places they protect and serve, that knowledge is often unconnected across officers. Group audits are a powerful tool for compiling the experiential assets of law enforcement personnel into a more complete, fine-grained picture of reality on the street. By systematically compiling intelligence provided by the men and women that patrol city streets, the experiential assets of law enforcement personnel can be repackaged, analyzed, and incorporated into ongoing law enforcement action and efforts to reduce gun violence. The group audit technique described in this chapter is a method that facilitates the systematic collection of such data to help analyze the violence landscape, especially among violent groups and gangs. The methods behind the audit are rather straightforward and the output is readily amenable to the techniques of social network analysis—including methods described in other chapters in this volume. The audit process we conducted in New Haven—and the results it produced—represent a fairly typical experience in a medium-sized city. And, as can been seen in some of the descriptive analysis, compiling such data through audits helps provide useful and actionable insights for law enforcement and violence prevention efforts. Some fairly common problems arise during the audit process, as they did in New Haven. Problems stemming from preconceptions about gangs resulted in the consolidation of small, independent groups into larger “nations” or into broad neighborhood categories. Thankfully, social network analysis of the relational data gathered during the group audit allows for researchers to visualize the street-level reality that law enforcement has an implicit understanding of. Once the complex web of feuds, alliances, and turf is simplified into network diagrams and geographic maps, it is very clear that groups that were once all simply “Crips” or who were conflated into a group denoted only by a street or neighborhood name are in fact unique sets or gangs. Unique insights like these make social network analysis a valuable complement to the process of the group audit, and ultimately provide law enforcement and violence reduction strategists with more complete, higher quality information upon which to base their operational decisions. Granted, no matter how well researchers employ

44

Disrupting Criminal Networks

these techniques to mitigate methodological issues of the group audit, audits will always be a work in progress. As gangs change over time, so too must audits and targeted audits must also be repeated to ensure that the data is as close to reflecting street-level reality as possible. This is no small task, but the ability for group audits not only to generate empirically useful data, but also to help law enforcement better serve their communities, makes them a promising endeavor for researchers and police departments alike to pursue. Just as importantly, the audit process also provides a new intelligence repository for the police by collecting, cleaning, and synthesizing data on gangs and their behavior from multiple law enforcement sources. Data sources that were once spread out and segregated from one another were combined and used to build a more accurate and more complete picture of the gang and violence profile. From patrol officers and parole/probation case workers, to U.S. Attorney investigators and federal agents will now have access to this newly information source. It will hopefully not only allow more effective policing and investigation, but will also help police and other bodies of law enforcement save lives.

Notes 1. The authors would like to thank all of those involved with Project Longevity in Connecticut, Tracey Meares, David Kennedy, Tony Cheng, Robin Engel and the University of Cincinnati Policing Institute, and our research partners at The University of New Haven. This project was supported by Cooperative Agreement Number 2012-CK-WX-K039 awarded by the Office of Community Oriented Policing Services, U.S. Department of Justice. The opinions contained herein are those of the author(s) and do not necessarily represent the official position or policies of the U.S. Department of Justice. References to specific agencies, companies, products, or services should not be considered an endorsement by the author(s) or the U.S. Department of Justice. Rather, the references are illustrations to supplement discussion of the issues. 2. While a great amount of effort has been spent debating the definition of a “gang,” much less effort has focused on what differentiates a gang from other delinquent groups (for a recent review, see Howell 2012). Some definitions of a “gang” stress involvement in criminal activity, such as drug use or sales ( Fagan 1989; Klein and Crawford,1967), while other definitions stress common “turf” (Hagedorn and Macon, 1988). The definition used in this study revolved around “groupness,” not popularized gang-related behaviors like hand signs, colors, or even formalized group names (see pg. 9). Thus, we prefer the term “group audit.” The objective of the audit is to cast the net broadly, and then refine analytic focus with subsequent investigation and interpretation. In short, while not all groups are gangs, gangs are groups. 3. More detailed descriptions of how the architecture and implementation of such forums has been described in several cities including Boston (Kennedy

Applying Group Audits to Problem-Oriented Policing

45

et al.1997), Chicago (Papachristos et al. 2007), High Point, NC (Corsaro et al. 2012), and Cincinnati (Engel et al. 2010). 4. While the call-ins and forums include representatives from law enforcement, social service providers, and the community, the group audit itself overwhelmingly relies on the expertise and knowledge on the part of local police and federal agents. Although the audits in New Haven were composed entirely of law enforcement and researchers, audits in some cities also include community outreach workers and service providers. 5. Some municipalities—including New Haven—limit audit participants to law enforcement, although other municipalities include social outreach workers, community stakeholders, or local pastors in their audits. See Papachristos (2012). 6. To analyze this new collection of data, we made use of several tools in compiling, analyzing, and visualizing the data gathered during the group audit. All geographic maps of New Haven were generated using Esri’s ArcMap 10.1 program in conjunction with geospatial vector data provided by the New Haven Police Department. Group turf was digitized using ArcMap 10.1 by drawing vector lines that corresponded to the hand drawn sections of turf collected on laminated maps during the audit. The data on groups, their activities, and the relationships to other groups was recorded with pen and paper during the audit, and was later entered into Microsoft Excel for cleaning and data management purposes. The tables on group relationships were converted into .csv format for manipulation with the freeware program R, as well as the igraph package for social network analysis. Once converted into a usable graph file (i.e., a .gml file), we exported the graph (a collection of nodes and the edges connecting them) to another freeware program, Gephi, which we used to customize the appearance and readability of the network diagrams. 7. 129,774 as reported by the U.S. Census Bureau. 8. The New Haven Police Department included this information in the FBI’s yearly UCR report. http://www.cityofnewhaven.com/police/statistics.asp. 9. The first audit in New Haven was led by a team of researchers from the University of Cincinnati and included researchers from The University of New Haven and Yale University. All subsequent audits were led by researchers from Yale University and The University of New Haven. 10. Police districts are the most common geographic distinction used by the New Haven Police Department. Because of its ubiquity, the police district was used as the unit of analysis when performing the audit. Most of the districts are no more than a few square miles, which was useful for ensuring a high level of detail in information gathered during the audit. The police districts correspond to neighborhoods within New Haven (e.g., District 7 corresponds to the Newhallville neighborhood, District 4 corresponds to the Dwight/Kensington neighborhoods), though the administratively drawn district boundaries do not necessarily perfectly reflect the reality of informal neighborhood boundaries. Despite this, police were very knowledgeable about what parts of the city police district boundaries represented, and were aware not only of activities within those districts, but could point to officers that could better speak to group activities in those areas they were not as familiar with. 11. As per the wide definition of group used in this audit, many of the groups identified did not have a specific name they called themselves. A

46

Disrupting Criminal Networks

common technique for identifying these “nameless” groups is to use a street corner they inhabit or the name of a local landmark such as a park. 12. Some groups that were identified do not appear on the turf map because there was insufficient intelligence to denote a definite section of turf for certain gangs. This imperfect knowledge is normal, and is addressed through the readministration of audits over time. This allows for gaps in knowledge to be filled, as well as for the ever-changing landscape of groups in a city to be recorded. 13. The first groups to claim the title “Crips” originated in Los Angeles in the late 1960s, though their roots can be traced to 1950s gangs in South Central Los Angeles. For more information on this gang, and their historical (and popularized) rivals, the Bloods, see Vigil (2010). 14. Being a “piss boy” entails things like stashing drugs or weapons, delivering messages, running a weapon or drug package from one place to another, or even carrying out a shooting for the older set. Interestingly, police stipulate that WR2 members are strictly forbidden from entering R1 territory because of R1’s worry that young, impulsive WR2 members will bring police attention with needless violence.

4 Network Stability Issues in a Co-Offending Population Carlo Morselli, Thomas U. Grund, and Rémi Boivin

THIS CHAPTER FORMULATES THE IMPLICATIONS EXTENDING from the application of a social network framework to the study of cooffending and overall crime. The focus is on how crime involvement and frequency vary within a co-offending population and, more specifically, on the stability of co-offending relationships. Our overall objective is to demonstrate why networks matter in our general understanding of crime patterns. Theoretically, the study is based on a long-standing tradition highlighting the social basis of crime. Analyses are based on a complete cooffending population for the province of Quebec (Canada) over a seven-year period. While early co-offending research typically turned to the removal of prolific offenders as a key intervention strategy, this study highlights the underlying network mechanisms that are in place in a general co-offending population by assessing relationships between three segments: the core (or most connected) participants; the peripheral participants who are directly affiliated to that core; and the remaining mass of sporadic offenders who were generally short-term and opportunistic in their co-offending experiences. Such subsets are directly linked to offending patterns and, although it is found that the most connected and those directly connected to this core in past co-offending relationships emerge as the most active offenders, stability across such relationships tends to decrease crime commission frequency. This final finding is counter intuitive to expectations from early co-offending research, thus suggesting that, when taking a step 47  

48

Disrupting Criminal Networks

away from the prolific offender stance, it is not the removal of individuals, but the breaking of relationships that matters most. The Co-Offending Tradition The idea that offenders, young and old, are required to associate with others in varying offences has been of increasing interest to criminologists. Early research in this area found that most crimes were committed by more than one person and that most offenders were more likely to commit crimes with a co-offender (Breckinridge and Abbott 1917; Shaw and McKay 1931). Such findings were also consistent in later research conducted during the 1960s and 1970s. The co-offending issue and the rise in research in this field were heavily influenced by Reiss’s work in this area (1986; see also Reiss and Farrington 1991). He found that half of burglaries were committed by two or more offenders. This proportion increased to 67% if we examine the proportion of offenders who committed burglaries with two or more people. Findings on robbery were also consistent. Just less than one-half of robberies were committed by two or more offenders, while three-quarters of robbers co-offended during their crimes (pp.124-125). Group prevalence was also found for violent crimes (aside from homicide and rapes) (p.134). Reiss also found that 17% of offenders always committed crimes with cooffenders. Based on past research, he hypothesized a positive relationship between offending and co-offending. While Reiss focused on specific crimes and revealed a strong prevalence of co-offending, more recent research on general crime has found much less co-offending. Studies of mass sets of official data have found that co-offending varies between 10 and 20 percent across crime events (Hodgson 2007; Stolzenberg and D’Alessio 2008; van Mastrigt and Farrington 2009), while the percentage of individuals taking part in cooffending varies between 20 to 45 percent. In a study of Canadian arrest records, Carrington (2002) concluded that 24% of offenders were linked to co-offending events, with this proportion being much higher amongst youths (44%) than adults (20%). Van Mastrigt and Farrington (2009) found 30% of offenders to be involved in co-offending. Slightly higher proportions were found in other research by Hodgson (2007), who found 35% of offenders to be involved in co-offending research, and in McCord and Conway’s (2002) study of youth patterns (40%). Reiss’s research objective was guided by a strict prevention focus and, more particularly, a selective incapacitation framework that was prominent in mainstream criminology of the 1970s and 1980s. His suggestion was that selective incapacitation efforts, which are designed to assess actual prison

Network Stability Issues in a Co-Offending Population

49

sentences in accordance with an offender’s offending rate and risk of recidivism, should include an individual’s co-offending scope. He argued that crime prevention efforts relied specifically on a better understanding of “the group status of the offender and the behavior of co-offenders and their affiliated offending groups” (p.121). This led Reiss to focus on recruitment issues within co-offending circles: “delinquents are continually signalling their interest in locating others with whom they may engage in offending” (p.140); “It is possible that high-λ offenders who frequently change co-offenders may actually be composed of subpopulations of ‘joiners’ and ‘recruiters’” (p.142). This specific joiner/recruitment distinction would subsequently be refuted by Warr (1996), who demonstrated the transient nature of such roles in cooffending settings. The same observation would be substantiated by McGloin and Nguyen (2012), who found some evidence of offending instigation across types of crime. Reiss also studied the link between age and co-offending. The recruitment assumption also guided this proposition, with Reiss arguing that, as with typical persistent offenders that participated in both solo and group offending, the more prevalent young offenders were likely those who took part in solo offending, but who also associated with wider groups of offenders that served as a “reference group” and “a resource for recruiting accomplices for offending” (p. 149). Furthermore, that co-offending decreases with age is a pattern that has been confirmed across most research in this area (see Piquero, Blumstein, and Farrington 2007; van Mastrigt and Farrington 2009; Carrington 2009; Andresen and Felson 2010). One of the first to propose an alternative to Reiss’ outlook on the cooffending issue was Tremblay (1993), who reformulated the problem from a criminal opportunity perspective. Tremblay’s approach was not policy oriented, but he was able to integrate an element that was overlooked by Reiss. By following opportunity theories, such as Cohen and Felson’s (1979) routine activity approach, Tremblay was able to pursue the cooffending problem as a rational, intelligible, process that resulted in a variety of outcomes for individual searches of accomplices. Unlike Reiss, who was more concerned with the extensions of co-offending, Tremblay (1993) was focused on the patterns that emerged from the co-offending phenomenon. Following Cohen and Felson, he argued that attractive targets often rely on access to a “loose” network of criminally-apt contacts that render the opportunity executable (p.19). This network focus was partially addressed by Reiss, who found that offenders were bound in “a web of affiliation or network of contacts and exchanges” (p.137). Another key distinction between Tremblay’s work and Reiss’s work on co-offending was the risk and trust trade-off that underlies co-offending

50

Disrupting Criminal Networks

relationships. Reiss was more concerned with whether co-offending increased an individual’s risk of apprehension. While he did not maintain a direct relationship between the two, his discussion did suggest that the link may be an indirect one that is mediated by the additional fact that cooffending increases one’s prevalence in crime, therefore increasing one’s risk of apprehension. Rather than focus exclusively on risk, Tremblay entered the element of trust and followed findings from social network research, particularly Granovetter’s (1973) weak tie argument. Tremblay argued that weak ties (relationships that are low on trust, but high on opportunities) were more instrumental, while strong ties (relationships that are high on trust, but low on opportunities) brought greater security. Such an insight had a direct influence on subsequent research that examined the choice-structuring properties in co-offending with greater detail (McCarthy, Hagan, and Cohen 1998; Weerman 2003) and more general constructs for co-offending (see Felson 2003). Finally, Tremblay also revised the traditional unit of analysis that was used for co-offending research. He extended directly from Reiss (1988) in establishing his own definition of co-offending: “Many offenses may be committed on a solo basis but nevertheless depend on the availability of other offenders. The term, co-offenders...is given, therefore, a larger definitional scope and refers not only to the subset of an offender’s pool of accomplices but rather to all those other offenders he must rely on before, during, and after the crime event in order to make the contemplated crime possible or worthwhile” (p.20). By taking this path, Tremblay directed us toward a network or structural model of co-offending that maintained that the quality (suitability) of co-offenders mediated a relationship between the availability (quantity) of potential co-offenders and the likelihood of committing crime with co-offenders. Co-offending from a Social Network Perspective While the network outlook represents one of the rare common themes uniting Tremblay’s and Reiss’ respective work in the co-offending field, empirical research efforts have only begun to uncover key patterns from such a perspective. The consistency and clarity of results and extending propositions have been met with some important criticism. The strongest critique of a network approach in co-offending research was formulated by Felson (2003), who set out the challenge facing any researcher in this area. One of the basic criticisms made against efforts to apply network applications in co-offending and general criminology concerns the lack of insights provided in regard to how the presence of a network structure shapes the phenomenon under analysis. Felson’s call was indeed clear:

Network Stability Issues in a Co-Offending Population

51

A social network for crime, as important as it might be, generates a serious problem, since a network has no clear boundaries and is difficult to measure, analyze, or use to predict what happens on the ground. Somebody has got to specify the facts about a delinquency network and show that it has an ongoing structure producing criminal cooperation. (pp.155)

While Felson’s call for more concrete results and propositions was well founded, there has been some work that has already addressed his main points. The first and most ambitious study in this area was conducted by Sarnecki (2001), who pursued the social network framework developed in his earlier work in Borlänge (Sarnecki 1990) and extended it to the larger confines of Stockholm and its immediate outskirts. Consistent with the Borlänge study, the bulk of Sarnecki’s (2001) analysis focused on the central network that emerged from official arrest data. Network analyses led to the identification of an ensemble of 15,426 direct and indirect sets of cooffending relationships (3,979 individuals) that accounted for a considerable proportion of offences in Stockholm. Sarnecki maintained that the identification of this central network did not constitute the youth branch of a criminal underworld. The main reason for this was that relations were too short-lived. Also, while co-offending was a common component in most criminal events, such relations were generally restricted to an average two participants per crime and concentrated within tight geographical proximities. As with crime participation patterns in general, the substantial majority of youths involved in co-offending rarely participated beyond more than one crime. Sarnecki’s (2001) study remains one of the more exhaustive efforts to gather network data on delinquency, but little effort was devoted to applying the measures and analyses that were available from the wide social network analysis repertoire. A core network was identified, but little was provided regarding the structural components of this central piece. Allusions to network density were made, but minimal effort was provided to assess how density could generate new insights in understanding the structure of delinquency. In more recent years, additional research, designed primarily by Jean McGloin, has developed some of the key structural dimensions of cooffending networks. McGloin, Sullivan, Piquero, and Bacon (2008) examined youth co-offending data from a Philadelphia-based longitudinal data set and found that co-offending relationships were generally shortterm—co-offending relationships tended not to be re-used. Restricted to the “juvenile years” and consistent with Sarnecki’s earlier findings when assessing their general sample, the transient feature of co-offending was less

52

Disrupting Criminal Networks

of a feature for the more active offenders in McGloin et al’s (2008) study. Another important study was conducted by McGloin and Piquero (2010), who examined the link between non-redundant networking and offending versatility. Using individual (or egocentric) level density as a main indicator of network redundancy, they found that individuals with lower density (or less redundancy) in their personal networks were more likely to be versatile in their group offences. Such a finding was consistent with other research on criminal networks that demonstrated the benefits of brokerage (aka nonredundant networking) for increasing offenders’ earnings (Morselli and Tremblay 2004) and reputation (Morselli 2009b). In a recent appraisal of co-offending research, McGloin and Nguyen (2014) formulate the main points that emphasize how this tradition would benefit from a closer focus on network structure. First, knowing how the size of co-offending networks varies and affects crime patterns is telling in itself. Second, determining a co-offending network’s stability is also a crucial step toward approaching it from a practical outlook. Engaging in a stable network is radically different than engaging in a volatile network and such a variable is crucial for understanding the availability of crime opportunities and the level of commitment that is required by an individual. Third, studying the network structure of co-offending offers insight into the openness of a general crime setting, with redundant (or closed) networks calling for higher levels of trust and non-redundant (open) networks focused more typically on resource-exchange. Finally, the openness or closure of a network is also telling in regard to the frequency and versatility of crime (both variables are higher in open or non-redundant networks). What McGloin and Nguyen’s examination suggests is that a cooffending network does not have to be stable or closed within a fixed set of parameters for it to be important for criminological and preventive purposes. Transient and volatile co-offending networks are as important facts as their stable counterpart. The network framework allows us to determine the size, shape, and scope of a co-offending network at various dimensions (overall, for specific segments, or at the individual level). These structural features of the co-offending phenomenon are the main concern in the present study. In addressing Felson’s challenge and remaining cognizant of the important nuances set forward in more recent research, we examine the basic components of a complete co-offending network for the province of Quebec over a seven-year period. Our analysis begins with a comparison of co-offenders and solo offenders in the province and moves on to examine how traditional criminological variables, such as age, crime commission frequency, and types of crime, compare within segments of the co-offending population. The analysis then proceeds toward a dyad-level assessment, with a specific focus on stability issues.

Network Stability Issues in a Co-Offending Population

53

Data and Method The study is based on official records found in the Module d’information policière (MIP) and which were provided by the Sureté du Québec (SQ), Quebec’s provincial police. Such information is slightly different from the more traditional Uniform Crime Reports (UCR) data in that the MIP provides data on all crime events in which there is sufficient proof for one or more individuals to be arrested or accused, without necessarily being convicted. In contrast to the UCR data, MIP data provides systematic information on suspects or offenders across crimes. All data were coded to assure the confidentiality of individual identities. While the original data set that was transferred to us by the SQ was based on a twenty year period (1990 to 2009), analyses are based only date from 2003 through 2009. Depending on the type of crime marking the event, the MIP archives are regularly filtered to eliminate old cases. Some offences, such as homicides, are never deleted from the data. Others, such as illegal drug trafficking cases, are retained for a ten year period. More general crimes are kept for a seven year period. Because of this inconsistency, we were obliged to restrict the time period of the study. Nevertheless, the resulting 7 year timeframe is longer than what has been used in prior research (with the noted exception of Carrington (2002). Official data provide information on three analytical units: offences, offenders, and offence participations (van Mastrigt and Farrington 2009). An offence is a criminal event that has several characteristics, such as a location, a date of occurrence, and a type (e.g., violent crime, property crime). An offender is an individual that has been arrested for an offence. There may also be more than one offender involved in a given offence. Offence participations represent the criminal involvement of one person in one incident (Carrington 2009; Frank and Carrington 2007). If an offence was committed by only one offender, than there is only one offence participation; if more than one offender was involved, the number of offence participations per offence is equal to the number of offenders involved in the offence. Out of the 811,586 offences we have data on, 41% (n=333,274) were classified as violent crimes (or crimes against the person). This was followed by property crimes, which accounted for 36% (n= 290,923) of all offences, and market crimes, which account for 23% (n=187,389) of all offences. Most crimes were committed by a single offender (solo offences; n=732,474; 90.3%); the remainder were committed by multiple offenders (co-offences; n=79,112; 9.7%).

 

54

Disrupting Criminal Networks

Creating the Co-offending Network Based on information about criminal events and the participants in these events, a social network was created by linking those 113,591 offenders who were involved in at least one co-offending event over the seven year period. From a conceptual point of view one can illustrate the original data structure as depicted in Figure 4.1 where the color of the events represent the type of criminal activity. Links between individuals and events mean that individuals participate in a specific criminal event. Each event has exactly one type. A straightforward way to handle such event-affiliation (or two-mode) network data is generating an offender-to-offender (or one-mode network) projection (see Figure 4.2). Such a transformation effectively collapses the information into a single level. One only focuses on individuals and translates the relationships between individuals and events into relationships between individuals, where an individual-individual relationship means that two individuals co-offended with each other in the same event. Normally, one-mode projections need to be performed with caution as often information is lost. In fact, a lot of effort has recently been put into the development of techniques for un-projected two-mode data. However, when events have a unique ordering we can proceed without information loss. This dataset includes 113,591 unique individuals who had been arrested for co-offending. The co-offending relationships are defined as quadruples which represent a co-offending relationship between individuals i and j at time t of type k. Keeping the information on time (not depicted in Figure 4.2) and type of criminal event (depicted as color of the co-offending ties in Figure 4.2) allows seamless transformation between two-mode and oneprojection in both directions. In total, there are 171,869 undirected quadruples in the data defining the time and type-conserved relationships of the one-mode projection. These can be collapsed into 136,448 undirected ties between individual offenders. Figure 4.1

Illustration of Original Data Structure

Network Stability Issues in a Co-Offending Population

Figure 4.2

55

One-Mode Projection of Individual-Event Data from Figure 4.1

2 5 4

1

3

6

This network is the largest to be compiled for a co-offending study, surpassing Sarnecki’s Stockholm network of over 22,000 youths suspected of delinquent acts and their adult co-participants in close to 30,000 offences and over 35,000 offence participations between 1991 and 1995. In contrast to Sarnecki’s (2001) study, the Quebec network under analysis is more heavily focused on adults. For youth protection reasons, the SQ was not authorized to provide information on cases in which anyone under eighteen years of age was suspected or accused of a crime. Thus, contrary to Sarnecki, who designed a network of youths and the adults who were directly connected to them during their crimes, our data is based on adults and only those juvenile offenders who were connected to them during their crimes. The proportion of youths (26%) in this network (the population of individuals who co-offend) is consequently underestimated and the average age is an overestimate of the real population. While this important segment of the population is lacking in our data, important distinctions nevertheless emerge when comparing co-offenders and solo offenders (see Table 4.1). Co-offenders were younger than solo offenders (27 years compared to 36 years). The male/female proportion was very similar across both segments of this population. Co-offenders were, however, also more active in crime than solo offenders. On average, co-offenders committed 3.5 crimes, while solo offenders committed 1.8 crimes. When examining the types of crime, co-offenders were more involved in market and property crimes, but less involved in violent crimes: 60% of individuals who co-offended were involved in property crimes (compared to 34% of solo offenders); 42% of co-offenders were involved in market crimes (25% for solo offenders); and 45% of co-offenders were involved in violent crimes (62% for solo offenders).

56

Disrupting Criminal Networks

Table 4.1

Comparison of Co-offenders and Solo Offenders Co-offenders

Solo offenders

Age

27.4

36.0

Gender (% male)

78.0

79.0

3.5

1.8

Violent crimes (%)

45.0

62.0

Property crimes (%)

60.0

34.0

Market crimes (%)

42.0

25.0

Demographic Characteristics

Criminal Behavior Avg. number of crimes

Thus, while important limits, such as the lack of youth in the data, do have an important impact on our findings, the data does offer a rich opportunity to assess features of co-offending that are rarely integrated in past research. In order to advance our assessment of the Quebec cooffending network, the overall distribution of offenders was subsequently divided into three new analytical units that were designed for this specific study. Following Sarnecki (2001), we focused on the core of the network, designated as the top 5% of co-offenders (n=5,408 individuals; 4.7%) with the most co-offending links. These 5,408 individuals account for 29.5% of all co-offending relationships. The remaining 95% of the network was subsequently divided in two segments (the periphery and mass). The periphery was constructed by identifying those co-offenders who were not amongst the top 5% segment of co-offenders, but who were directly connected to one or more of the individuals in the core of the population. This periphery was comprised of 10,063 individuals (8.9% of the network).1 The remaining 86% of the population (n=98,120 individuals) was designated as the mass of the population. Features of the Co-offending Network One of the more important challenges facing anyone handling such a massive data set is the selection of an initial analytical path. A typical network analysis of such a population would have us starting with a

Network Stability Issues in a Co-Offending Population

57

traditional sociogram, however, a network illustration of over 113,000 people is rarely effective for illustrative purposes. We therefore begin with a straightforward distribution of the number of direct co-offending links within the population. Figure 4.3 shows the degree distribution for the number of co-offenders. What we find is the positive-asymmetrical curve that has become a classic to students in criminology and typically found when examining the age-crime curve or individual crime commission patterns. For co-offending, we are faced with a similar pattern in which the largest amount of individuals (55%) only had one co-offending partner; 95% of all co-offenders (our segment of the population that is placed classified beyond the core) have less than seven co-offending partners. Figure 4.3

The Number of Unique Co-offenders (Degree Centrality Distribution)

% of co-offenders

60 50 40 30 20 10 0 1

6

11 16 21 26 31 36 41 46 51 56 61 66 71 76 Number of unique co-offenders

Overall, offenders who co-offended at least once had, on average, 2.4 different co-offenders across the seven years. The highest count documented in the Quebec data set was for one person who was directly linked to 77 cooffenders. What we have is an extremely high volume of offenders with very few co-offenders and a small proportion of offenders with a relatively high volume of co-offenders. Only marginal relationships are found between the number of cooffenders and the more traditional criminological variables. One pattern that was assessed within this population concerned the relationship between the offender’s sex and number of co-offenders. On average, males did have more co-offenders, but this difference was only marginal (2.49 versus 2.10 co-offenders for females). The number of unique co-offenders (degree centrality) is weakly correlated with the average age of offenders over all

58

Disrupting Criminal Networks

their offenses (r=.02, α