Date post: | 27-Apr-2015 |
Category: |
Documents |
Upload: | positiveid |
View: | 466 times |
Download: | 2 times |
Murder as Interaction: A Network Approach to Street Gang Homicide1
Andrew V. Papachristos The University of Chicago
A Paper submitted for presentation at the 2004 Annual Conference of the American Sociological Association, San Francisco, CA.
ABSTRACT Using a network approach, this essay argues that there exists a unique structure to gang-related homicides. Viewing murder as an interaction rather than an outcome reveals stable patterns of relations between and among gangs which choreograph homicides in years with high or low murder rates, regardless of the events that trigger individual homicidal episodes. I use homicide records and observation data to reconstruct and analyze patterns of interaction among gangs. The findings suggest a distinct structure of gang homicides that are stable over time and, in fact, seem to predict gang homicides in later years. Furthermore, counter-intuitive to conceptions of gang violence as the by-product of rival gangs battling with each other, gang members are just as likely to be killed by their own gang as an opposing one. Implications for future research, hypothesis testing, and theory construction are also discussed.
Andrew V. Papachristos The University of Chicago Department of Sociology 1126 E. 59th St., Rm. 307 Chicago, IL 60637 773-919-6900 [email protected]
1 I would like to thank Andrew Abbott, Tracey Meares, Jeff Fagan, George Tita, Robert Sampson, Damon Phillips, John Padgett, and Dave Kirk for their comments on different aspects of this work.
INTRODUCTION
Sociologists generally conceive of murder as an outcome, the result of too much poverty and
inequality or too little social capital and social control. Murders and their motives, from
genocide to mercy killing, generally appear circumscribed by the details surrounding an
individual event. Consistent with such perceptions, conventional wisdom associates street
gang homicide with such matters as turf defense, status threats, or retaliation at the micro-
level (Block and Block 1993; Decker and Curry 2002; Short and Strodtbeck 1965) and with
concentrated poverty, inequality, and geographic and social isolation at the macro-level
(Curry and Spergel 1988; Rosenfeld, Bray, and Egley 1999). Moreover, gang homicides
occur in “episodic spurts” (Block and Block 1993) that rise and fall in an unpredictable
fashion stimulated by unknown events or else at the whim of gang members. Gang
homicide is thus perceived as a rather chaotic event that decimates inner city communities in
the form of drive-by shootings and stray bullets. Seldom do we think of gang murders as
having a structure in the sense of patterned relations that determine who kills whom and for
what reason.
The present essay seeks to revise this view of gang homicide as a random and
unpredictable phenomenon. I argue, in fact, that there exists a unique structure to gang-
related homicides. Viewing murder as an interaction rather than an outcome reveals stable
patterns of relations between gangs which choreograph homicides in years with high or low
murder rates, regardless of events or motives that trigger individual homicidal episodes.
Understanding the social structure of gang homicide involves not only rethinking the
dominant approaches to gangs but also re-integrating gang research with developments in
sociology, in particular network analysis and theory. It requires that gangs be viewed
dynamically as both an outcome of neighborhood processes as well as a structure which
3DSDFKULVWRV� 2
shapes and determines the social behaviors of its members and the contours of its social
context.
To explore the social structure of gang homicide in Chicago and discuss its relevance
theoretically and methodologically, I use homicide records and observation data to analyze
patterns of interaction among gangs. Specifically, using a network approach I reconstruct
patterns of gang homicide at the city and community levels and analyze the influence of
these relations on homicide. Consistent with Simmel’s (1908a) formalistic approach to
sociology, I find persistent patterns of antagonistic relations among gangs as well as intricate
alliance structures. Conflictual interactions in the form of murders between gangs appear to
be stable over time although the intensity of interaction varies. Events which intensify
conflictual interaction tend to be highly specific, yet the patterns of such interaction occur
along pre-existing lines of gang-relations. Rarely do gang murders occur between groups
without any prior interaction. Furthermore, counter-intuitive to conceptions of gang
violence as the by-product of rival gangs battling with each other, gang members are just as
likely to be killed by their own gang as an opposing one. My main goal is to analyze the
network of gang homicides to examine how it structures social behaviors, in this case
homicide, and discuss the relevance of such an approach for future research, hypothesis
testing, and theory construction.
Viewing murder as an interaction dynamically shifts the locus of gang research while
simultaneously posing unique new questions for network analysts.2 Gang research for much
of the past two-decades, has largely focused on outcomes or micro-level description,
ignoring the macro- and middle levels of analyses and the processes and structures that link
gangs with other social phenomena. Examining gang homicide from a network approach
2 A network approach to gangs is only in its infancy (e.g. Kennedy, Braga, and Piehl 1997; Tita 2003). The present work can be seen as an application and theoretical extension of such an approach.
3DSDFKULVWRV� 3
resurrects key sociological questions concerning social control, the use of group norms to
mediate behavior, and the power of interaction to affect neighborhood level social
interaction. For network analysts, conceiving of murder as an interaction forces a new
understanding of asymmetric and negative relations in that such interactions appear to create
rather stable and persistent structures. Issues of transitivity and structural balance may also
be revisited from this perspective to understand how such negatives structures are created
and maintained.
INTERACTION AND UNIT OF ANALYSIS: FROM GANG HOMICIDE TO SOCIAL STRUCTURE The use of homicide as a representation of interaction that creates a social structure requires
some justification. As Simmel (1908b) points out, conflict is one of the basic forms of
interaction in that, at the very least, conflict ensures that two parties interact if only to
resolve differences, build group solidarity, achieve some kind of unity, or release tension,
even “if it be through the annihilation of one of the conflicting parties” (p. 70). Thrasher
(1927), Short and Strodtbeck (1965), Klein (1971), and others (Jankowski 1991) demonstrate
that conflict with authority, neighborhood residents, or other gangs tends to solidify gang
cohesion and thus bolster group processes in the gang. Conflict, then, not only appears as a
defining trait of the gang, but also as one of its key forms of interaction. 3 I use a specific
violent event, homicide between gangs, as the main indicator of gang interaction. Gang
homicide, if anything, is a conflictual form of interaction including at a minimum the
interaction of the victim and the offender at the time of the actual event.
3 In fact, both Thrasher (1927: 46) and Klein (1971: 111) include such an element of conflict in their definition of the term “gang.”
3DSDFKULVWRV� 4
Definitions of “gang-related” homicide, however, vary drastically, altering
significantly how the event is perceived interactionally (for a review, see Maxson and Klein
1996). Liberal definitions, such as those used in Los Angeles, classify any homicide
involving a gang member as “gang-related.” In contrast, more conservative definitions, such
as those used in Chicago, classify a homicide as “gang-related” only if the crime itself was
motivated by gang activity such as turf defense, drug dealing, or existing gang conflicts. So, a
homicide of one gang member by another gang member because of a personal dispute (e.g.
the infidelity of a lover, intoxication, argument, etc.) would be classified as “gang-related” by
the former definition by not by the latter. The Chicago data used here adheres strictly to the
conservative definition of gang-related.
The use of a conservative definition has a crucial advantage: it ensures that the actual
reason for the interaction (content) is a group one even though the actual event may entail
only two individuals. That is, the homicide is motivated, and often preceded, by an extra-
individual (gang) circumstance. The unit of analysis is the gang and not the gang member as
would be the case in the liberal definition. The drawback, however, is that conservative
definitions tend to underestimate the total number of interactions.4 For the sake the precise
definition of the interaction and unit of analysis, I err on the side of conservatism.
Applying a network framework to gang homicide requires a two-step analytic move.
First, there must exist a distinct patterning of interactions that are not simply random.
Second, and in order for these patterns to constitute a structure, these patterns of interaction
must have an influence on homicide. To these ends, the analysis starts by recreating the
homicide patterns of a single year and then extrapolating to patterns in other years. To
analyze these structures, I (1) examine the role of specific gangs in the network and types of
4 This underestimation is relative. With an average of 182 gang-related homicides a year over the past 10 years, the conservative definition in Chicago provides an adequate sample size for most statistical methods.
3DSDFKULVWRV� 5
interaction using descriptive network analysis and, (2) test the stability of these structures
over time and the randomness of the network structures using Quadratic Assignment
Procedures.
DATA AND METHODS
The primary source of data comes from incident-level records of all homicides in the City of
Chicago from 1994 to 2002 as provided by the Chicago Police Department.5 Data is
recorded at the incident-level and contains demographic, geographic, motive, and gang
information on both the offender and victim. Qualitative data in the form of participant
observation from my on-going field research is used to evaluate the accuracy of the homicide
data and to elaborate on aspects of the network analysis.
The methodological approach of this paper is two-fold, switching levels of analysis
for explanatory purposes. First, I rely on a network analysis of homicide incidents to
represent the structure of gang interactions for the entire city of Chicago for three years,
1994, 1998, and 2002. Standard network measures are calculated using UCINET 6.0
(Borgatti, Everett, and Freeman 2002) with all equations, unless otherwise noted, following
those suggested by Wasserman and Faust (1994). Directed graphs and network diagrams
were created using NetDraw 1.0 (Borgatti 2002). Second, while a network analysis of gangs
in the entire city of Chicago provide a macro-level glimpse of the structure of gang conflict,
like too many other studies it overlooks the relationship between gang interactions and the
neighborhood context. To rectify this potential oversight, I conduct similar network analysis
using as an example from one area of Chicago which has been part of my on-going research.
5 Data was provided by the Chicago Police Department’s division of Research and Development. The analysis of the data reflects the findings and opinions of the author and in no way represents the views of the providing agencies. I am indebted to Tracey Meares and Jeffrey Fagan for their assistance in gathering and use of this data.
3DSDFKULVWRV� 6
The use of qualitative data at this level proves indispensable in explaining specific patterns of
gang interaction as well as understanding the local context of such interactions.6
GANG HOMICIDE IN CHICAGO
I begin by recreating the structure of gang homicide in Chicago by establishing the basic
dyadic homicide interactions for all gang-related homicides for the year 1994, the year with
the highest gang homicide rate in city’s history. In total, there were 284 gang-related
homicides and gang data for both the victim and the offender is available for 226 (79.5
percent) of those incidents.7 Forty-seven different gangs were involved in homicides leaving
a total potential network size of 2,269 possible connections.8 To establish a city-wide
structure, I built up from the basic dyadic homicides by letting each node represent a
different gang and each arc a specific homicide event. Figure 1 displays the directed graph
of all gang-related murders in 1994.
[Figure 1 Here]
Different shaped nodes represent the predominant race of the gang: squares signify
Black gangs, triangles Hispanic gangs, circles White gangs, and diamonds Asian gangs. Each
node is labeled with an abbreviation of the gang name and full names are listed in Appendix
A. Clusters of gangs that are enclosed by a broken circle share a common ancestry or
alliance.9 Arcs represent at least one homicide with the arrow indicating the direction of
victim. Double-headed arrows indicate at least one homicide occurring in both directions—
6 Due to length, much of the qualitative data has been excluded from this present paper. For more details, please contact the author. 7 There was a total of 931 homicides in 1994 with a homicide rate of 33.4 per 100,000. The total gang-related homicide rate was 9.9 for all races, 18.3 for Blacks, and 13.9 for Hispanics. 8 Since self-selection is possible in the case of homicide, meaning members of the same gang kill each other, the total potential network size is g * g. 9 Alliances of Chicago gangs are well known in the gang world, law enforcement, and the media. Gang Nations represent cognitive sub-groups identified by gang members and not necessarily functional sub-groups, such as cliques and blocks, as identified by network analysis
3DSDFKULVWRV� 7
i.e. the gangs killed at least one member of each gang in different events. The thickness of
the line represents the intensity of the conflict as indicated by the number of homicides
between gangs ranging from one to 36 homicides. Isolates represent gangs in which the only
homicide that occurred involved members of the same gang.10
A distinct structure of gang interaction is immediately apparent in Figure 1:
interactions among Black gangs are highly interconnected with intense patterns of conflict
whereas interactions among Hispanic gangs are more diffuse and isolated. That is, Black
gangs interact with a greater number of gangs in their immediate network and to a greater
degree, thus producing a tightly interconnected network structure. In contrast, the network
structure of Hispanic gang conflict is made up of dyads and triads connected by a single
gang, the Latin Kings.
A brief glance at this structure highlights the importance of specific gangs which
network measures of centrality can evaluate. Three centrality measures are used: out-degree
centrality, in-degree centrality, and an Eigenvector of overall network centrality. Given the
nature of homicide as an interaction and the structure of the data, care must be given in the
interpretation of such measures.11 Measures of degree centrality generally signal the level of
“activity” of a node in the measured set of relations (Wasserman and Faust 1994). Because
the data is asymmetric, meaning it is possible for one gang to kill a member of another gang
without successful retaliation, separate measures of out-degree and in-degree are used. The
out-degree measures the level at which a specific gang is active in the sense of committing
murders: a gang with a high out-degree is more central because it causes more killings. In-
10 Unfortunately, the di-graph does not include an accurate representation of intra-gang homicides for those gangs connected in the larger network although, as described below, this is an increasingly important phenomenon. 11 Although not addressed here, new network measures might be necessary to explain such negative relations.
3DSDFKULVWRV� 8
degree measures the opposite—it measures how many members of a gang were killed by
other gangs. Thus, out-degree measures the number of homicides committed by the gang
while the in-degree measures the number of members in that gang killed. The Eigenvector
is used to indicate the most central actors in terms of the overall structure with respects both
to local connections and overall network reach by taking into account out-degree, in-degree,
and the total possible number of connections (Wasserman and Faust 1994). Thus, higher
Eigenvectors indicate that actors are “more central” to distances among all actors, i.e. that
such gangs interact not only more frequent with “local” gangs but also connected to other
gangs with high degrees of centrality. Table 1 gives these measures for the 12 gangs with
the highest levels of degree centrality.
The centrality mean of 4.81 implies that on average gangs in the network are
involved in approximately five homicidal interactions with other gangs in 1994 as either
offender or victim. A single gang, the Gangster Disciples, has the highest centrality
measures both for out-degree (46) and in-degree (48) measures, more than ten-times the
network mean. The Latin Kings, one of only two Hispanic gangs in Table 1, are also highly
centralized, more so with regards to killing other gangs (out-degree = 21) than being killed
(in-degree = 14). The four other most central actors with respects to in- and out-degree
measures, the Mickey Cobras, the Black P-Stones, the Four Corner Hustlers, and the Black
Disciples, are all densely connected Black gangs with multiple paths directly to and around
the Gangster Disciples. For the most part, all of the gangs in Table 1 have out- and in-
degree measures greater than the mean suggesting that these 12 gangs are responsible for
much of the interaction found in the larger network.12 Furthermore, with the exception of
12 This pattern is consistent with the findings of Block and Block (1993) which show that four gangs in Chicago (The Black Gangster Disciples, the Vice Lords, the Latin Kings, and the Latin Disciples) were responsible for 55 percent of all gang-related homicides from 1987 to 1990.
3DSDFKULVWRV� 9
the Latin Kings and Mickey Cobras, the interaction of these gangs occurs between other
gangs listed in Table 1.
Table 1: Centrality Measures for Selected Gangs in the 1994 Gang Homicide Network
Gang Out-Degree In-Degree Eigenvector
GD 46 48 .529 LK 21 14 .123 MC 15 4 .254 BPS 18 31 .338 BD 24 30 .221 FCH 18 19 .389 MIVL 10 4 .214 BGD 8 3 .137 BlkSouls 7 7 .158 MLD 6 5 .024 VL 7 17 .399 Ambro 4 1 .008 Network Mean 4.81 4.81 .082
The Eigenvector scores support the central role and overall power of the Gangster
Disciples (.529), followed by other densely connected Black gangs, the Black P-Stones (.338),
the Four Corner Hustlers (.389), and the Vice Lords (.399). The Eigenvector of the Latin
Kings (.123) suggests that it has a considerably low level of overall centrality in regards to the
larger network.
In addition to the importance of individual gangs in the network, Figure 1 suggests
at least two distinct patterns of gang interaction—intra-racial conflict and intra-gang conflict.
First, consistent with other research on violence, almost all of the interaction is intra-racial:
only eight homicides (approximately 3 percent) occur between gangs of different races.
Furthermore, the gang homicide problem in Chicago is almost completely a Black and
Hispanic one. White gangs are only peripherally connected in the larger network and Asian
gangs are completely isolated. Most striking in terms of racial differences, there are almost
3DSDFKULVWRV� 10
twice as many Hispanic gangs as there are Black gangs, 25 as compared to 13 respectively,
but there is more interaction among the fewer Black gangs. Thus, it appears that, in part,
differential homicide rates seem related not to the number of gangs but to the density of interaction among
gangs.13 While it is beyond the scope of this paper, heightened interaction may be linked with
population density or other structural or ecological factors, a matter taken up in the
discussion section.
Second, much of the intense conflict between Black gangs occurs within the same
gang or between gangs that share a common ancestry or alliance. To simplify, I call the
former intra-gang conflict and the latter intra-nation conflict. Generically, Chicago gangs are
divided into two gang “nations,” the People and the Folks, which consist of gangs with often
complex historical or geographic based alliances (Knox 2001). In addition, gangs that share
some common ancestry often refer to the group of related gangs as a “Nation” (see,
Venkatesh and Levitt 2000). For example, all of the gangs in the cluster around the Gangster
Disciples share a common lineage to a single gang known as the Devil’s Disciples and,
therefore, refer to the conglomerate of “Disciple” gangs as the “Black Gangster Disciples
Nation” (Papachristos 2001). Similarly, the cluster of gangs that share a Vice Lord name
constitute the “Almighty Vice Lord Nation” and share a common ancestry as well as a
geographic point of origin (Knox and Papachristos 2002). These gang federations often
share a common identity and members can often be found to interact more often on social
biases than non-affiliate members.
Of all the gang-related homicides in 1994, 32 were intra-gang (11.3 percent) and 41
were intra-nation (14.4 percent). Overall, a quarter of all homicidal gang interaction in 1994 occurred
13 Between 1992 and 2001, Black gang-related homicides rates have been almost double those of Hispanics suggesting that per capita there are more interaction among Black gangs than Hispanic ones. This might also suggests that even if the number of Black and Hispanic gangs were equal, there would still be a greater number of interactions among Black gangs.
3DSDFKULVWRV� 11
within the same gang or between gangs with some alliance. Large portions of the violent interaction
among gangs in Chicago occurs not from what is traditionally thought of as disturbances
among rival gangs (e.g. turf wars) but from gangs that are supposedly part of the same
Nation.14 This finding suggests that the role of gang Nations and alliances are perhaps more
symbolic than functional. According to gang norms, these alliances are explicitly designed to
structure patterns of interaction among gangs—i.e. to keep members of the same and
affiliated gangs from fighting with each other and to come to the mutual aid of others in the
Nation.
The pattern just described replicates itself almost identically in 1998. Figure 2
provides the structure of gang homicide in 1998 and Table 2 provides the centrality
measures of the 12 most central gangs.
[Figure 2 Here]
In 1998 there were 704 homicides of which approximately 26 percent (n = 186) were
gang-related. Although the network mean of 3.63 is slightly lower than 1994, the patterns of
interaction are almost identical—Black gangs are densely interconnected with interaction
among highly central gangs while Hispanic gangs are configured in a star-like network with a
single gang, the Latin Kings, at the Center. Gangs that were highly central in 1994 remain so
in 1998 as do patterns of intra-gang (12 percent of total gang-related homicides) and intra-
nation homicide (30 percent). More significantly, these patterns remain quite similar in spite
of the fact that the overall and gang-related homicide rates were considerably lower in 1998
14 This pattern of interaction, with the exception of Decker and Curry (2002), is almost completely absent in the gang literature.
3DSDFKULVWRV� 12
than in 1994.15 That is, even though the rate of interactions was less, the patterns of
interaction were the same.
Table 2: Centrality Measures for Selected Gangs in the 1998 Gang Homicide Network
Gang Out-Degree In-Degree Eigenvector GD 33 28 .707 VL 8 4 .141 FCH 8 8 .062 BD 18 17 .606 PP 1 0 .001 BPS 4 3 .058 CVL 3 5 .095 SD 9 3 .011 LK 9 14 .056 TVL 2 7 .116 BGD 3 0 .046 NB 3 5 .052 Network Mean 3.63 3.63 .060
In 2002 the structure of homicide changes slightly for the Hispanic gangs but
remains stable for the Black gangs. Figure 3 displays the structure of gang homicides for
2002 and Table 3 lists the centrality measures of selected gangs.
[Figure 3 Here]
2002 had one of the lowest overall homicide rates in almost a decade. Overall, there were
647 homicides (22.3 per 100,000) of which 132 (20 percent; 4.6 per 100,000) were gang-
related. As seen in the Figure 3 and in the centrality measures in Table 3, even though the
rate of homicide is lower in 2002, the structure of Black gang homicide is remarkably similar
with dense interconnected gangs and specific gangs, most notably the Gangster Disciples
(Eigenvector = .441), the Four Corner Hustlers (.425), and the Black P-Stones (.420), at the
15 In 1994 the overall homicide rate was 33.4 per 100,00 and the gang-related rate was 9.9 compared to 1998 with an overall rate of 24.3 and a gang-related rate of 6.5.
3DSDFKULVWRV� 13
center of much of the interaction. The star-like Hispanic network seen in 1994 and 1998,
however, dissipates into more isolated, non-connected sub-graphs.16 Still, within the
Hispanic network the Latin Kings remain the most central (Eigenvector = .116). High levels
of intra-gang and intra-nation homicides also remain as a crucial pattern of interaction and
appear to have increased over time. Of the 133 gang-related homicides in 2002, 36 (27.1
percent) were intra-gang and 22 (16.5 percent) were intra-nation. Thus, approximately 44
percent of all gang-related homicides in 2002 occurred between members of the same gang or gang Nation,
almost two times higher than in 1994.
Table 3: Centrality Measures for Selected Gangs in 2002 Gang Homicide Network
Gang Out-Degree In-Degree Eigenvector GD 28 29 .441 VL 6 4 .385 FCH 10 12 .425 BD 4 3 .139 PP 3 0 0BPS 5 8 .420 CVL 3 0 .156 SC 3 0 0LK 4 5 .116 TVL 5 5 .338 BGD 2 2 .210 NB 6 4 .175 Network Mean 2.61 2.61 .090
16 One possible reason for the disruption of this network is efforts by law enforcement that target specific Hispanic gangs. It is interesting to note that during these time periods several massive prosecution and other law enforcement efforts have been mounted directly to disrupt the hierarchies of several of these gangs, including the Gangster Disciples, the Black P-Stones, the Mickey Cobras, and the Latin Kings. The network data shows little effect of these prosecutions on Black gangs over time since there is little movement of these gangs’ position in the overall network. In contrast, it appears as though there was considerable change in the Latin Kings position. This might suggest that the disruption of dense, interconnect network merely shifts interaction around whereas the disruption of the star-like network has a more disruptive effect on the overall structure.
3DSDFKULVWRV� 14
The patterns of gang homicidal interaction described in the di-graphs and centrality
measures appear to be relatively stable over time with the exception of the dissipation of the
Hispanic network in 2002. These descriptions suggest stability in the form of interaction but
they do not necessarily prove that they influence homicide as an outcome. That is, are these
gang homicide patterns randomly generated? While statistical tests of network properties lag
behind descriptive methods, Quadratic Assignment Procedures (QAP) provide a basic
method for testing the null hypothesis of independence among the gang homicide
networks—that the homicide network at time A is correlated with the homicide patterns at
time B versus the null hypothesis that the networks are randomly created each year. To test
this null hypothesis, I ran QAP correlations using UCINET 6.0 for the 1994, 1998, and 2002
overall networks as well as the race specific networks. For a time series data such as those
used here, the simulations are run to see the relationship between the homicide networks
with the earlier of the years serving as the independent variable and the later year as the
dependent variable. Table 4 provides the results.
Table 4: QAP Correlations for the 1994, 1998, and 2002 Gang-Related Homicide Networks
Overall Network Black Network Hispanic Network
1994 1998 1994 1998 1994 1998
1998 .526*** .500*** .637***
2002 .435*** .561*** .422*** .621*** 0.12 0.072
Includes only those gangs in network all three years. N = 5000 Monte Carlo simulations *** = .001
The QAP correlations provide enough evidence to reject the null hypothesis of
independence for the overall networks, the Black networks, and the early Hispanic networks.
For the overall networks, there is a moderate correlation between the structures from 1994
3DSDFKULVWRV� 15
and 1998 (.526, -level = .001) and 1998 and 2002 (.561, -level = .001). The correlation is
somewhat smaller, although still statistically significant, between 1994 and 1998 (.435, -level
= .001). These findings suggest, even given the change in structure in 2002, that overall
patterns of gang homicide are not random and that the effect is somewhat stronger in years
closer in time. Thus, it appears that the there is a social structure to homicide that helps to
shape homicide patterns over time.
Disaggregated by race, the patterns of the Black network are similar to the overall
network with higher correlations between closer years. Moreover, the correlation between
the 1998 homicide structure and the 2002 structure is highly significant (.621, -level = .001).
Consistent with the di-graphs, for Hispanic gangs the 1994 and 1998 homicide structures are
highly correlated (.637, -level = .001), but the significance is lost in 2002 when the structure
dissipates. That is, there is not enough evidence to reject the null hypothesis of
independence for 2002. The Hispanic network in this year provides the exception—the
previous patterns of interaction do not appear to determine the homicide patterns in 2002.
To summarize, the patterns depicted in the network graphs and tested in the QAP
correlations suggest patterns of interaction between gangs are relatively stable and produce
structures that predict homicide patterns. The analysis also highlights two key attributes
about conflicutal gang interaction: (1) the same gangs are at the center of most homicidal
activity and (2) large portions of interaction occurs between the same gangs or gangs within
the same alliance structure. An examination of a subsection of the larger network in a
confined geographic area not only permits a closer look at such patterns of interaction and
their replication at a smaller level of analysis, but it also provides insight into how such
patterns are influenced by their social context, an important caveat given the rich
ethnographic history of gang research.
3DSDFKULVWRV� 16
GANG CONFLICT ON THE WEST-SIDE: “A VICE LORD THING”
The contours of gang conflict on the West-Side of Chicago are not a secret.17 Residents,
police, gang members, the media, and even scholars readily acknowledge that the West-Side
is home to one of Chicago’s oldest Black street gangs, the Vice Lords. Whereas accounts
from the 1960’s by Keiser (1969) depict inter-racial conflict between the Vice Lords and
White ethnic gangs over turf, present-day gang conflict occurs mainly among different Vice
Lord factions. The comments of “Jimmy,”18 a 45 year-old ranking member of the Unknown
Vice Lords, offer his description of gang interaction on the West-Side:
You see, back in the day, when I was comin’ up, we was fighting against the Whites, over some stupid shit. There was this pool, see, and we had to fight ‘em just so we could hang over there . . . [today], all this craziness up in here, its all Lords trippin’ on each other. You got 4’s [Four Corner Hustlers] trippin’ on Unknowns [Unknown Vice Lords]. You got Insanes [Insane Vice Lords] trippin’ on Travelers [Traveling Vice Lords]. Ain’t nobody care about the GD’s [Gangster Disciples] up in here no more. Man, all this nonsense, it’s all a Vice Lord thing.
Jimmy’s remarks concerning gang conflict among the different Vice Lord gangs are
reflected in both the homicide and the qualitative data. Figure 4 shows two networks of
gang conflict on the West-Side. Figure 4a displays the conflict network based on homicide
data while Figure 4b shows a network based on field data.19 To construct the network in
Figure 4b, I coded fieldnotes for any mention of conflict between gangs on the West-Side.
17 By the “West-Side” of Chicago, I am referring to the North Lawndale, Garfield Park, and Austin Community Areas which are predominately Black (more than 80 percent) and have some of the highest rates of poverty, unemployment, and crime in the city. Similar to Hunter and Suttles’ (1972) concept of the community of “extended liability,” the West-Side represents a large geographic area which posseses some sense of identify. The identity of the West-Side is particularity strong in the gang milieu in contrast with the South-Side, an area consisting largely of Disciple gangs. When asked to describe the boundaries of the “West-Side” during my fieldwork, residents were consistent in indicating the areas west of the University of Illinois until the suburb of Oak Park, north to Chicago, and south until just about 18th Street. 18 Names of all neighborhood residents, gang members, law enforcement officials, and all other individuals mentioned in the ethnographic data have been changed for the sake of anonymity. 19 The homicide data used in this figure is taken from the 11th and 15th police districts, those areas which cover roughly the cognitive areas cited by residents.
3DSDFKULVWRV� 17
The results were cross-checked and a link between gangs was established when any two
different sources mentioned the same pattern of conflict.20
Key similarities exist between these networks. First, consistent with the city-wide
network, almost all of the conflict is either intra-gang or intra-nation accounting for 60
percent of the interaction in Figure 4a and 70 percent in Figure 4b. More importantly,
with the exception of three gangs (the Gangster Disciples, the New Breed, and the Black
Souls), gang conflict occurs mainly among Vice Lord gangs—gang violence on the West-
Side is indeed “a Vice Lord thing.”
Second, a single gang, the Four Corner Hustlers21, is at the center of either conflict
network occupying the core of a star-like structure. Unlike the city-wide networks, the
Gangster Disciples play only a peripheral role in gang interaction on the West-Side, mostly
because of geography and history: the Vice Lords began on the West-Side whereas the
Gangster Disciples originated on the South-Side. While the Gangster Disciples do exist and
operate on the West-Side, they do not occupy as central a role in this part of the city. The
West-Side is so dominated by Vice Lord gangs that a portion of it, near 18th and Lawndale, is
referred to affectionately by gang members as “Vice Lord City.” If similar methods were
applied to the South-Side, one might expect to find the roles of the Gangster Disciples and
the Vice Lords reversed. Like the city structure, the networks in Figure 4 show that specific
gangs are at the center of interaction and that interaction between Black gangs is dense with
much of the interaction occurring within the same gang alliance. In these respects, the larger
structure is recreated at this level of analyses.
20 The fieldwork data used in Figure 4b differs from the homicide data in two important ways. First, it does not distinguish directionality of the conflict. While most informants could identify conflict patterns, often with explicit detail, they could not distinguish in which direction the hostility flowed. Second, the fieldwork data is not limited to homicides but includes any conflicts known to informants. 21 Although not obvious by in its name, as is the case with other Vice Lord gangs, the Four Corner Hustlers are in fact one of the key Vice Lord factions (also, Knox and Papachristos 2002).
3DSDFKULVWRV� 18
CONCLUSION
The strength of a network approach rests in its view of expressing the social environment as
patterns of relationships among interacting units which produce a persistent structure that
can affect behavior or action. Many of the unanswered questions in gang research result
from the failure of the dominant approaches to consider, either theoretically or
methodologically, gang-relations—how gangs interact and respond to each other and their
social environs. The findings of the present study suggest that interaction between gangs
structure violence in Chicago by generating persistent types and patterns of interaction over
time. The theoretical and methodological implications of these findings not only address
many of the shortcomings in gang research but also potentially offer a link between the field
and other areas of sociology.
Theoretically, gangs have been viewed as important structural entities in cities since
the work of Thrasher (1927) and Shaw and McKay (1942). Underclass theories extend these
views by arguing that gangs have become structurally embedded entities in extreme poverty
areas, meaning that gangs now function as quasi-institutions that influence other behaviors,
regulate norms, and fulfill key social functions (Bursik 2002; Bursik and Grasmick 1993;
Hagedorn 1988; Venkatesh 1997). At the macro- or middle-levels of analysis, how gang-
relations and activities are shaped by extra-gang forces such as processes of social
disorganization or de-industrialization is unknown as these claims remain relatively untested
empirically.
The network approached used here provides a method for testing such social
disorganization and underclass hypotheses. If gangs are in fact embedded structures, then
patterns of gang interaction should either facilitate or hinder social behaviors other than
3DSDFKULVWRV� 19
violence. Take for example, the issue of social control. If gangs are important structural
actors, gang conflict should decrease levels of social control and, conversely, increase incidences
of violence or deviant behavior. Similarly, if, as argued by Venkatesh (1997), gangs are
central actors in the social, physical, and economic context of underclass communities, then
patterns of gang-relations should have either a positive or negative influence on the
distribution of social capital or the development of collective efficacy (e.g. Sampson,
Morenoff, and Earls 1999).
Methodologically, linking gang-relation data with other demographic or survey data
provides a platform from which to test such hypotheses. A methodological direction can be
borrowed from the application of network methods in organizational sociology which have
used network measures to assess the effect of social structures on outcomes such as
organizational competition and collaboration (e.g. Podolny, Stuart, and Hannan 1996) and
patterns and intensity of stock trading (e.g. Baker 1984), to name just a few examples. Such
studies link network data with other multi-variate analysis techniques, including regression
and maximum likelihood estimates. Similarly, work by Gould (1995) incorporates
autoregressive indicators of social networks into other contextual or neighborhood based
statistical models.
Thinking of homicide as an interaction also has implications for network analysis and
theory, particularly structural balance theory. Homicides between gangs occur not only
between rival gangs but also between gangs within the same alliance structures apparently
breaking the “enemy of my enemy is my friend” rule associated with structural balance
theories (e.g. Davis 1967). Given that these gang Nation structures were intentionally
created to form unified alliances that would respond to rivals in a cohesive fashion, intra-
nation homicide appears to directly contradict principles of structural balance. Thinking of it
3DSDFKULVWRV� 20
as an issue of network transitivity, such findings beg the question: is murder transitive?
Additionally, homicide, as well as other negative interactions such as hostile take-overs or
war, require a reformulation and interpretation of standard network measures of centrality,
power, and cohesive sub-groupings.
Applied to gangs, such an approach possesses tremendous theoretical explanatory
power by linking gang network measures, at multiple levels of analysis, to social structural
variables such as population density, structural disadvantage, social mobility, and social
control. At the neighborhood level, this can be done by deriving the context specific gang
network and including such network measures in multi-variate models as either predictor or
outcome variables. For example, the link between gangs and neighborhood level social
control, as hypothesized by Bursik (2002; Bursik and Grasmick 1993), can be tested by
including context specific gang network measures using autoregressive parameters with
standard social disorganization models. In addition to understanding the relationship
between gangs and social control, such analyses can also shed light on the relationship
between neighborhood characteristics such as population density, poverty, or mobility on
the development of gang activity. Such analyses will not only help to evaluate seventy-five
years of relatively untested theoretical claims about gangs but they can also provide an
understanding of the role gangs in key sociological processes and structures.
REFERNCES Baker, Wayne E. 1984. “The Social Structure of a National Securities Market.” American Journal of Sociology 89: 775-809.
Block, Richard and Carolyn R. Block. 1993. “Street Gang Crime in Chicago.” Research in Brief. National Institute of Justice, Office of Justice Programs, U.S. Department of Justice.
Borgatti, S.P. NetDraw 1.0. Harvard: Analytic Technologies
Borgatti, S.P., Everett, M.G. and Freeman, L.C. 2002. Ucinet for Windows: Software for Social Network Analysis. Harvard: Analytic Technologies.
Bursik, Robert J. 2002. “The Systemic Model of Gang Behavior,” In Gangs in America III, edited by Huff, C. Ronald. Thousand Oaks, CA: Sage.
Bursik, Robert J. and Harold G. Grasmick. 1993. Neighborhoods and Crime: The Dimensions of Effective Community Control. New York: Lexington Books.
3DSDFKULVWRV� 21
Curry, David and Irving A. Spergel. 1988. “Gang Homicide, Delinquency, and Community” Criminology, 26 (August): 381-405.
Davis, James A. 1967. “Clustering and Structural Balance in Graphs,” Human Relations 20: 181-187.
Decker, Scott and G. David Curry. 2002. “Gangs, Gang Homicides, and Gang Loyalty: Organized Crime or Disorganized Criminals,” Journal of Criminal Justice 30: 343-352.
Gould, Roger V. 1995. Insurgent Identities. Chicago: University of Chicago Press.
Hagedorn, John. 1988. People and Folks: Gangs, Crime, and the Underclass in a Rustbelt City. Chicago: Lake View.
Hunter, Albert and Gerald D. Suttles 1972. “The Expanding Community of Limited Liability.” Pp. 44-81. In The Social Construction of Communities. Edited by Gerald D. Suttles. Chicago: University of Chicago Press.
Jankowski, Martin Sanchez. 1991. Islands in the Street: Gangs and American Urban Society. Berkeley: University of California Press.
Kennedy, David M., Anthony A. Braga, and Anne M. Piehl. 1997. “The (Un)Known Universe: Mapping Gangs and Gang Violence in Boston,” In Crime Mapping and Crime Prevention, edited by David Weisburd and J. Thomas McEwen. New York: Criminal Justice Press.
Keiser, Lincoln. 1969. Vice Lords: Warriors of the Street. Chicago: Holt, Rinehart, and Winston.
Klein, Malcolm W. 1971. Street Gangs and Street Workers. Englewood Cliffs, NJ: Prentice-Hall
Knox, George W. 2001. An Introduction to Gangs (5th edition). Peotone, IL: New Chicago School Press.
Knox, George W. and Andrew V. Papachristos 2002. The Vice Lords: A Gang Profile Analysis. Peotone, IL: New Chicago School Press.
Maxson, Cheryl and Malcolm Klein. 1996. “Defining Gang Homicide: An Updated Look at Member and Motive Approaches,” In Gangs in American (2nd edition), edited by C. Ronald Huff. Thousand Oaks, CA: Sage.
Papachristos, Andrew V. 2001. A.D., The Neighborhood Impact of Federal Gang Prosecution. Peotone, IL: New Chicago School Press.
Podolny, Joel M., Toby Stuart, and Michael Hannan. 1996. “Networks, Knowledge, and Niches: Competition in the World-Side Semiconductor Industry, 1984 -1991.” American Journal of Sociology 102 (3): 659-689
Putnam, Robert D. 1993. “The Prosperous Community: Social Capital and Community Life.” The American Prospect 35-42.
Rosenfeld, Richard, Timothy M. Bray, and Arlen Egley. 1999. “Facilitating Violence: A Comparison
of Gang-Motivated, Gang-Affiliated, and Nongang Youth Homicides,” Journal of Quantitative
Criminology, Vol. 15: 495-516.
Sampson, Robert J., Jeffrey D. Morenoff, and Felton Earls. 1999. “Beyond Social Capital: Spatial Dynamics of Collective Efficacy for Children,” American Sociological Review, v64: 633-700.
Shaw, Clifford R. & McKay, Henry D. 1942. Juvenile Delinquency and Urban Areas. Chicago: University of Chicago Press.
Short, James and Fred Strodtbeck. 1965. Group Process and Gang Delinquency. Chicago: University of Chicago Press.
Simmel, Georg. 1908a. [1971] “The Problem of Sociology.” In Georg Simmel on Individuality and Social Forms, edited by Donald N. Levine. Chicago: University of Chicago Press.
-------------------. 1908b. [1971]. “Conflict.” In Georg Simmel on Individuality and Social Forms, edited by Donald N. Levine. Chicago: University of Chicago Press.
Thrasher, Fredric M. 1927 [2001]. The Gang: A Study of 1,313 Gangs in Chicago. (unabridged edition). Peotone, IL: New Chicago School Press, Inc.
Tita, George. 2003. “Beyond Pin Maps: Mapping the Social and Geographical Dimensions of Gang Crime.” Presentation at the Project Safe Neighborhoods National Conference, Philadelphia.
Venkatesh, Sudhir A. 1997. “The Social Organization of Street Gang Activity in an Urban Ghetto,” American Journal of Sociology, v103: 82-111.
Venkatesh, Sudhir A. and Steve D. Levitt. (2000). “Are We a Family or a Business? History and Disjuncture in the Urban American Street Gang.” Theory and Society 29: 427-462.
Wasserman, Stanley and Katherine Faust. 1994. Social Network Analysis: Methods and Applications. New York: Cambridge University Press.
Figure 1: Chicago Gang Conflict Network Based on Gang-Related Homicides, 1994
Notes:Square represents Black GangTriangle represents Hispanic GangDiamond represents Asian GangCircle represents White GangDirection of Arrow indicates direction of homicideThickness of line indicates number of homicides between gangsCircled nodes represent gangs that share an alliance or ancestry
GD
BGD
BD
ImpG
T itPStn
BPSFCH
MIVL
T VLVL
UnkVL
CVL
BlkSouls
MC
GayLrds
S imCtyRoy
PP
LtnBroOrg
LtnEgls
Pauch
SpnLrds
StnDis
LaFam
LaRaza
Ambro
PRStn
LK
26Bys
22Bys
Brzrs
SpnGD
LtnDis
Bshps
AllLov
LtnJiv
CD
SpnCbr
CNts
LtnCnts
LtnLvrs
InsDrg
LtnDrg
CmpBys
InsUnk
MLD
RedScrp FlyDrgn
Notes:Square represents Black GangTriangle represents Hispanic GangDiamond represents Asian GangCircle represents White GangDirection of Arrow indicates direction of homicideThickness of line indicates number of homicides between gangsCircled nodes represent gangs that share an alliance or ancestry
Figure 2: Chicago Gang Conflict Network Based on Gang-Related Homicides, 1998
GD
BGDBD
ImpG
T itPStn
BPS
FCH
MIVL
T VL
VL
CVL
BlkSouls
MC SimCtyRoy PP
LtnEgls
StnDis
LaFam
LaRaza
Ambro
LK
26Bys
LtnDis
SpnCbr
LtnCnts
InsDrg
LtnDrg
MLD
BlkStns
AshVikings
NB
CiInsVL
LtnSaints
InsPopes
KingCbrs
SpnVL
MilKings
LtnSouls
UndrVL
RVL
InsDuc
Notes:Square represents Black GangTriangle represents Hispanic GangDiamond represents Asian GangCircle represents White GangDirection of Arrow indicates direction of homicideThickness of line indicates number of homicides between gangsCircled nodes represent gangs that share an alliance or ancestry
Figure 3: Chicago Gang Conflict Network Based on Gang-Related Homicides, 2002
GD
BGD
BD
NB
BlkSt
BPS
T itPS
VL
CVL
T VL
FCH
CiVL
MIVL
BS
MC
LK
PRSt
LtDrg
LtEgl
22Boys
MilKngs
SD
CD
SGDLaFam
ISC
CB
MLD
LtJiv
SC
ImpIG
LaRaz
PP
MB
Bis
OALtDis
LtCnts26Boys
GDNB
FCH
CVL
T VL
VL
MIVL
BS
UknVL
FCH
T VL
UnkVLUnderVL
GDCVL
MIVL
Figure 4: Gang Conflict Networks on the West-Side
Figure 4a: Conflict Network Based on HomicideData
Figure 4b: Conflict Network Based on Field Data
Notes:Square represents Black GangDirection of Arrow indicates direction of homicideThickness of line indicates number of homicides between gangsCircled nodes represent gangs that share an alliance or ancestry
Appendix A Gang Names and Abbreviations
African American Gangs
Hispanic Gangs
White & Asian Gangs
BPS – Black P-Stones FCH – Four Corner Hustlers GD – Gangster Disciples BGD – Black Gangster Disciples BD – Black Disciples BlkSt. – Black Stones CVL – Conservative Vice Lords TVL – Traveling Vice Lords BS or BlkSouls – Black Souls CiVL – Cicero Insane Vice Lords VL – Vice Lords MIVL – Mafia Insane Vice Lords TitPS – Titanic P-Stones NB – New Breed MC – Mikey Cobras UnkVL – Unknown Vice Lords ImpG – Imperial Gangsters UnderVL – Undertaker Vice Lords RVL – Renegade Vice Lords
LK – Latin Kings SD or StnDis – Satan’s Disciples ISC – Insane Spanish Cobras MLD – Maniac Latin Disciples PP – Party People PRSt. – Puerto Rican Stones LDrg – Latin Dragons LE – Latin Eagles CD – Cullerton Dueces 22B – Two-Two Boys MilK. – Milwaukee Kings LaFam. – La Familia CB or CmpBys– Campbell Boys LJ or LtnJiv – Latin Jivers SC or SpnCbr – Spanish Cobras ImIG – Imperial Insane Gangsters LaRaz. – La Raza MB. – Morgan Boys Bis. – Bishops OA – Orchastra Albany LD or LtnDis – Latin Disciples LC or LtnCnts– Latin Counts 26Boys – Two-Six Boys SGD or SpnGD– Spanish Gangster Disciples Pauch — Pauchucos LtnBroOrg – Latin Brothers Organization InsUnk – Insane Unknowns SpnLrds – Spanish Lords LtnLvrs – Latin Lovers InsDrg – Insane Dragons AllLov –Allport Lovers Cnts – C –Notes Ambro – Ambrose Brzrs – Brazers SpnVL – Spanish Vice Lords KngCbrs – King Cobras InsPopes – Insane Popes
GayLrds – Gay Lords SimCtyRoy – Simon City Royals RedScrp – Red Scorpions FlyDrg – Flying Dragons