JUSTICE QUARTERLY, 2010, 1–26, iFirst Article
Racial and Ethnic Heterogeneity,
Economic Disadvantage, and
Gangs: A Macro-Level Study
of Gang Membership in
Urban America
David C. Pyrooz, Andrew M. Fox and
Scott H. Decker
[email protected]
Mr
000002009
DavidPyrooz
Justice
10.1080/07418820903473264
RJQY_A_447701.sgm
0741-8825
Original
Taylor
2009
00
and
&
Quarterly
Article
Francis
(print)/1745-9109
Francis
(online)
There is a lack of macro-level gang research. The present study addresses this
shortcoming by providing a theoretically informed analysis of gang membership
in large US cities. More specifically, our goal is to determine whether racial and
ethnic heterogeneity conditions the relationship between economic disadvantage and gang membership. Three separate sources of data are used in this
study: U.S. Census 2000, Law Enforcement Management and Administrative
Services 2000, and National Youth Gang Survey 2002–2006. A series of weighted
least-squares regression models are estimated, finding that both economic
disadvantage and racial and ethnic heterogeneity exhibit independent and
additive effects on gang membership. In addition, the results show that racial
David C. Pyrooz is a doctoral student in the School of Criminology and Criminal Justice at Arizona
State University. He received his B.S. and M.S. in criminology from California State University,
Fresno. His research interests revolve around life-course criminology, gangs and criminal/delinquent
networks, and the intersection between theory and policy. His recent work has appeared in the
Journal of Criminal Justice and the Journal of International Migration and Integration. Andrew M.
Fox is a doctoral student in the School of Criminology and Criminal Justice and a graduate research
assistant for the Center for Violence Prevention and Community Safety at Arizona State University.
He received his B.A. in sociology from Point Loma Nazarene University in 2005 and his M.S. degree in
criminology and criminal justice from Arizona State University in 2008. His research interests
include social network analysis, gangs, drug markets, and communities. Dr. Scott H. Decker is
professor and director in the School of Criminology and Criminal Justice at Arizona State University.
He received his B.A. in social justice from DePauw University and his M.A. and Ph.D. in criminology
from Florida State University. His main research interests are in the areas of gangs, criminal justice
policy, and the offender’s perspective. His most recent books include European Street Gangs and
Troublesome Youth Groups (Winner of the American Society of Criminology, Division of International
Criminology Outstanding Distinguished book award, 2006) and Drug Smugglers on Drug Smuggling:
Lessons from the Inside (Temple University Press, 2008, CHOICE Academic Press Book of the Year).
His forthcoming book (with Hugh Barlow) Criminology and Public Policy: Putting Theory to Work will
be published in 2010 by Temple University Press. Correspondence to: David C. Pyrooz, School of
Criminology and Criminal Justice, Arizona State University, 411 Central Avenue, Phoenix, AZ 85004,
USA. E-mail: [email protected]
ISSN 0741-8825 print/1745-9109 online/10/000001-26
© 2010 Academy of Criminal Justice Sciences
DOI: 10.1080/07418820903473264
2
PYROOZ ET AL.
and ethnic heterogeneity has a conditioning relationship with economic disadvantage. Furthermore, our expanded operationalization of the Blau heterogeneity measure indicates that prior research may have underestimated the effects
of heterogeneity. The authors discuss these findings in the context of existing
gang research and offer directions for future research.
Keywords gang membership; gang formation; racial and ethnic heterogeneity;
Blau; economic disadvantage
Introduction
Economic disadvantage is closely tied with crime and other social problems
(Blau & Blau, 1982; Pratt & Cullen, 2005; Shaw & McKay, 1942). Pratt and
Cullen’s (2005) meta-analysis of predictors of aggregate crime rates found
economic disadvantage to be one of the strongest and most consistent macrolevel predictors of crime. A number of works have found that unsupervised peer
groups (i.e., gangs) are often concentrated in disadvantaged neighborhoods
characterized by economic deprivation, high residential mobility, and racial and
ethnic heterogeneity (Bursik & Grasmick, 1993; Hagedorn, 1991; Hughes &
Short, 2005; Shaw & McKay, 1942; Short & Strodtbeck, 1965; Thrasher, 1927).
Rates of gang membership, however, are not uniform across disadvantaged
communities. In other words, some disadvantaged areas produce gangs while
others do not (see, e.g., Katz & Schnebly, in press). Our knowledge of this
relationship, though, is constrained for two reasons. First, most accounts of
gangs and economic disadvantage are limited because they examine only one
city or a limited geographical area in the USA (e.g., Southern California or
Chicago). Analyses of one gang or one neighborhood result in an “N” of only
one, making the comparative effect of social and economic conditions on gang
membership difficult to assess. In this context, Klein (2005) stressed the need
for comparative gang research across locations (e.g., neighborhoods within
cities, between-cities, and between-countries). While two studies have
employed macro-level methodologies (Jackson, 1991; Wells & Weisheit, 2001),
this research is limited in that it concentrated solely on whether cities had a
gang problem rather than assessing the prevalence of gang membership. Given
the presence of gangs in most large cities in America (Egley, Howell, & Major,
2004), this essentially limits the variation of the dependent variable. A more
comprehensive macro-level study of gang membership is warranted given the
history of macro-level theory (e.g., anomie/strain, social disorganization) in
gang research and its impact on how later theory, research, and policy has
formed (Kornhauser, 1978; Spergel, 1995).
Second, traditional theoretical explanations for gang formation have focused
on strain/anomie and limited opportunity structures as causes of gang formation
(Cloward & Ohlin, 1960; Cohen, 1955; Miller, 1958) while largely ignoring issues
of race and ethnicity. This exclusion is potentially problematic given that some
RACIAL AND ETHNIC HETEROGENEITY AND GANGS
3
research has shown that race and ethnicity play an important role in gang
formation and gang membership (Alonso, 2004; Davis, 2006; Freng & Esbensen,
2007; Thrasher, 1927; Vigil, 1988, 2002). In particular, Alonso (2004, p. 662)
noted that “[t]his apparent omission [of race and ethnicity] leads to some major
shortcomings in both contemporary and traditional gang formation theory.”
Racial and ethnic heterogeneity, hereafter referred to as heterogeneity, has
proven to be an important factor within community contexts (Sampson &
Groves, 1989; Shaw & McKay, 1942; Suttles, 1968), but has rarely been applied
to the gang context. The studies that have employed macro-level methodologies
that could have empirically assessed this relationship included one race or
ethnicity as an independent variable (e.g., percent Black, Hispanic) rather than
examining the composition of ethnic groups across cities.
The present study addresses these limitations by providing a theoretically
informed examination of gang membership across cities. Our goal is to determine whether heterogeneity conditions the relationship between economic
disadvantage and the presence of gang members at the city-level. Evidence in
favor of the moderating capacity that heterogeneity has on gang membership
would legitimize existing qualitative accounts of racial and ethnic conflict as
an explanatory variable. This research will aid in understanding why some
cities have higher levels of gang activity and will bring value to our existing
knowledge of criminological theory and its ability to explain problems at a
macro-level.
Economic Disadvantage and Gangs
Economic disadvantage is an anchor to many macro-level theories of crime.
Pratt and Cullen’s (2005, p. 412) meta-analysis of macro-level research
concluded that the criminogenic importance of economic disadvantage
“appears to be well supported across all macro-level studies of crime.”
Research that examines structural disadvantage concludes that as the level of
affluence or socio-economic status increases, crime rates and other social problems decrease. While this argument is fundamental to a number of theories, the
mechanisms through which economic disadvantage serves to increase crime are
not always agreed upon.
Two major macro-level theories, social disorganization and anomie/strain,
both rely heavily on economic disadvantage as a key predictor of social problems. The social disorganization perspective holds that economic disadvantage
matters insofar as it increases residential mobility and heterogeneity—both key
correlates of crime in transitional areas of a city (Burgess, 1928; Shaw & McKay,
1942). Sampson’s reconstruction of social disorganization theory contends that
these social factors—economic disadvantage, residential instability, and racial
heterogeneity—inhibit a community’s ability to impose social control (Sampson
& Groves, 1989; Sampson, Raudenbush, & Earls, 1997; see also Bursik &
Grasmick, 1993), resulting in higher neighborhood crime rates.
4
PYROOZ ET AL.
On the other hand, a strain/anomie perspective conceptualizes the function
of economic disadvantage differently. This perspective posits that crime is a
product of the discrepancy between the emphasis on economic success and
access to legitimate means to attain such success (Merton, 1938; Messner &
Rosenfeld, 2007). This strain is seen most clearly in a limited or blocked opportunity structure for acceptable means to success. Those in economically
deprived areas experience more strains than those with high economic status.
Crime occurs as people find other means to be successful, or give up altogether
on the goal of traditionally defined success (Chamlin & Cochran, 1995; Merton,
1938; Messner & Rosenfeld, 2007).
Economic disadvantage plays a major role in macro-level explanations of
crime and a considerable amount of research has documented several correlates
of economic disadvantage,1 including the presence of gangs (Anderson, 1999;
Wilson, 1987). Gang theory,2 akin to macro-level crime theory, has long emphasized economic disadvantage as a chief underlying construct of both social
disorganization and strain/anomie perspectives (Cloward & Ohlin, 1960; Cohen,
1955; Miller, 1958; Thrasher, 1927; Whyte, 1943). Again similar to a macro-level
crime context, the mechanisms by which economic disadvantage fosters gang
formation are operationalized differently by each respective theory. A social
disorganization perspective would argue that poverty creates a breakdown of
informal social controls resulting in a social reorganization, of which gangs are a
product. While a strain/anomie perspective would contend that gangs form in
response to a blocked opportunity structure thus resulting in a realignment of
what constitutes success and legitimate means to obtain that success.
No matter which theory one subscribes to, it is difficult to dispute that gangs
(and their members) are more inclined to form in interstitial, or more economically disadvantaged, areas (Bowker & Klein, 1983; Bursik & Grasmick, 1993;
Curry & Spergel, 1992; Shaw & McKay, 1942; Short & Strodtbeck, 1965;
Thrasher, 1927). City-specific accounts show this to be the case (see Decker &
Van Winkle, 1996; Hagedorn, 1988; Sanchez-Jankowski, 1991; Short & Strodtbeck, 1965; Venkatesh, 1997; Vigil, 1988). Two macro-level studies also found
bivariate associations, although modest, between disadvantage measures and
the presence of gangs (Jackson, 1991; Wells & Weisheit, 2001).
One point of dispute, though, is the breadth of the impact of disadvantaged
areas in producing gangs and gang members. In others words, not all areas of
comparable disadvantage produce equal shares of gangs and gang members. A
recent study in Mesa, Arizona (Katz & Schnebly, in press), illustrated this fact.
Consistent with a disadvantage-based theoretical perspective, the authors found
that economic deprivation and social/familial disadvantage predicted increases
in gang members per capita within Mesa’s 93 census tracts. Most importantly,
their analysis revealed that while gang members tended to live in more
1. For example, early childhood development (Duncan, Brooks-Gunn, & Klebanov, 1994), homicide
(Parker & McCall, 1999), and schooling (Duncan, 1994).
2. By this we mean theoretical explanations for the formation of gangs.
RACIAL AND ETHNIC HETEROGENEITY AND GANGS
5
disadvantaged census tracts, a number of the most disadvantaged and economically deprived census tracts did not have concentrations of gang members
greater than other areas of the city. This leads us to believe that disadvantage
alone may not fully explain gangs and gang membership, suggesting that additional covariates may moderate (or potentially mediate) this relationship.
Relatively recent theoretical developments (Klein, 1995; Vigil, 2002) show
that gang formation is contingent upon multiple factors that simultaneously
converge. Both Vigil and Klein posited that gangs typically form in communities
where an accumulation of different forms of disadvantage (e.g., economic disadvantage, lack of opportunities, family disruption, racial discrimination) come
together. Both the emergence and sustainability of gangs and gang membership
rely on the extent to which these disadvantages are more pervasive in communities. While nearly all cities across America with populations greater than 100,000
experience some type of gang presence (Egley et al., 2004), cities experiencing
increased disadvantage should encounter a larger presence of gang members.
The research above reveals two key issues confounding our existing knowledge about the structural correlates of gangs and gang membership. First,
because different studies examine different units of analysis (e.g., within-city
vs. across cities), it is difficult to understand whether a significant correlate is
limited to one or multiple communities (Short, 1998; see also Esbensen &
Winfree, 1998, p. 506). Moreover, our knowledge is also constrained by the
truncation of the dependent variable, treating the gang-related outcome
dichotomously (no-gang/gang city) or trichotomously (no-gang/stable gang/
chronic gang city). This severely limits the variation of the dependent variable,
particularly in a study that examines large cities, where every jurisdiction
reports the presence of gangs.
Second, scholars have pointed out the multidimensionality of gang problems
(Klein, 1995; Spergel, 1995; Vigil, 1988), meaning that one component (e.g.,
disadvantage) is inadequate for fully explaining the variation in the concentration of gang problems. The role of race and ethnicity has yet to be fully integrated into this equation even though the class Chicago School literature would
suggest doing so. In Alonso’s (2004, p. 662) socio-historical analysis of Black
gang formation in Los Angeles, he noted that “what is most striking about the
corpus of gang formation research is the limited discussion of how race and
structure have worked together to create communities that have produced
gangs.” To the extent that this is true, gang formation theory needs to explore
the degree to which race/ethnicity impacts the formation of gangs and gang
membership in urban communities.
Racial and Ethnic Heterogeneity and Gangs
Across time, population groups in the USA have tended to coalesce around race/
ethnicity (Ihlanfeldt & Scafidi, 2002; Moody, 2001; Omi & Winant, 1994). With
some exceptions (see Howell, Moore, & Egley, 2002; Starbuck, Howell, &
6
PYROOZ ET AL.
Lindquist, 2001), this has also been found to be true among gangs (Esbensen &
Winfree, 1998; Klein, 1995; Knox, 2000; Sanchez-Jankowski, 1991; Thrasher,
1927; Vigil, 2002). This has been the case for over eight decades, as Thrasher
(1927) found that race, ethnicity, and nationality were important factors for
gangs in that neighborhood adolescents aligned with youth of similar backgrounds. The existing problem, however, as Klein and Maxson (2006, p. 221)
stated, is that “[e]thnicity is one of the most widely discussed, and little
studied, aspects of gangs.”
An emphasis on racial and ethnic patterns was evident in Chicago School
research (e.g., Burgess, 1925; Shaw & McKay, 1942; Thrasher, 1927). This
research highlighted the “transition zone” as an area of social disorganization,
subject to relentless disruption, crime, and racial and ethnic turnover. Thrasher
(1927) detailed this turnover and discussed the succession of one ethnic group
to another in certain neighborhoods.3 This led to the unification around ethnic
groupings in response to the ever changing “racial complexion” of neighborhoods, which is consistent with more recent research documenting the importance of gang joining and gang formation for protective purposes (Decker &
Curry, 2000; Esbensen & Lynskey, 2001; Maxson & Whitlock, 2002; Melde,
Taylor, & Esbensen, 2009; Thornberry, Krohn, Lizotte, Smith, & Tobin, 2003).
Thrasher (1927, p. 133) noted how inter-ethnic European rivalries persisted
(e.g., Jewish v. Polish) as “old world antagonisms are carried over into gang
wars.” It appears from Thrasher’s research that race, ethnicity, and nationality
played a significant role in gang formation:
[t]he gang then, to sum up, is one manifestation of the disorganization incident
to cultural conflict among diverse nations and races gathered together in one
place and themselves in contact with a civilization foreign and largely inimical
to them. (1927, p. 154)
This argument has been put forth more recently to explain the gang landscape
in Europe, where changing structural conditions (e.g., immigration) are interacting with more street-level conditions and resulting in the formation of gangs
(Decker, Van Gemert, & Pyrooz, 2009).
While the Chicago School provided much detail on the development and
culture of gangs in Chicago, the other “traditional” gang city—Los Angeles—did
not have a university devoted to studying the ecology of gangs in its urban
regions. Nevertheless, Alonso (2004), Davis (2006), Moore (1985), and Vigil
(2002) have documented the development of gangs in Los Angeles using a variety of sources (e.g., mainstream and ethnic newspapers, ethnography, and
interviews). Their research underscores a common theme: social exclusion,
White privilege, marginalization, and racial conflict as the catalysts behind gang
formation in Los Angeles. In Adamson’s (2000) historical analysis of Black and
White gangs since the late 1700s, a similar theme emerged. He held that gangs
3. For example, Thrasher discussed the Bridgeport community and how the original Irish population
gave way to the Germans, who then later gave way to the Polish.
RACIAL AND ETHNIC HETEROGENEITY AND GANGS
7
served “defensive localism” purposes, in that youth came together when faced
with cultural change (see also Decker et al., 2009; Page & Marcelin, 2003). Much
of the differences between youth gangs, especially Black and White gangs, were
a result of structural factors—a thesis that has yet to be explored in a quantitative context.
Vigil (2002) documented marked differences in cultural change that resulted
in the formation of gangs among four different racial/ethnic groups in Los Angeles: Mexicans, Blacks, Vietnamese, and Salvadorans. Vigil found that these
groups were culturally distinct from one another, yet shared similar routes in
their establishment of youth gangs. These four groups experienced what Vigil
called “multiple marginality,” where breakdowns of social and economic factors
led to a “street socialization” takeover. In Vigil’s words, across these groups:
[w]hat is remarkable is the similarities that underscore how multiple marginality acts and reacts within populations to drive children into the streets and how
immigration or migration adaption is a central part of this process … [a]s children undergo street socialization they form a street subculture, namely, a gang.
(2002, pp. 16-17)
In an examination of Vigil’s marginality framework, Freng and Esbensen (2007)
used peer deviance as a proxy for street socialization and found it to be a significant correlate of Black and Hispanic (but not White) gang membership.
Research that explores gangs and gang membership theoretically tends to
treat race and ethnicity in terms of discrimination (Klein, 1995; Spergel, 1995;
Vigil, 1988). Many of those who have addressed race and ethnicity empirically
have done so in terms of “one-group” prevalence (e.g., percent Black or
Hispanic; Jackson, 1991; Katz & Schnebly, in press; Rosenfeld, Bray, & Egley,
1999; Tita, Cohen, & Engberg, 2005; Wells & Weisheit, 2001). To the extent that
protection and respect are garnered in economically deprived areas by neighborhood youth aligning along ethnic lines in response to conflict, or the threat
of conflict (Adamson, 2000; Decker, 1996; Vigil, 1988), it may be the case that
previous accounts of gang formation theory need to integrate race and ethnicity
as a conditioning factor between economic disadvantage and gangs. If this is the
case, the failure to account for heterogeneity may misrepresent some of the
causes and sustainers of gang formation and gang membership. But at this point
in time, given the scarcity of macro-level gang research, it is necessary to take
a step toward identifying relevant predictors that apply to cities, not just a city.
Current Focus
The present study explores the relationship between racial and ethnic heterogeneity, economic disadvantage, and gang membership. To date, there have
been four quantitative accounts that have looked at (1) whether a city has a
gang problem (Egley, 2005; Jackson, 1991), (2) differences in types of gang
problems between metropolitan and nonmetropolitan areas (Wells & Weisheit,
8
PYROOZ ET AL.
2001), and (3) variations of gang member concentrations within neighborhoods
(Katz & Schnebly, in press). The remainder of the research is typically focused
more qualitatively on a single jurisdiction. Klein (2005, p. 135) argued that
gang research suffers from a lack of cumulative knowledge, and summoned
scholars to conduct their gang research “by coordinated rather than disparate
methods.” The problem is that the evidence in support of race and ethnicity
(and even economic disadvantage) has been limited to observations that are
not capable of generalization beyond the scope of the area of study. As Maxson
(2001) pointed out, a comprehensive macro-level approach is necessary to
understand how larger, social–structural patterns are related to the size of
community gang problems. This will advance the state of the gang literature
from a localized to a generalized starting point for others to test other theoretical contentions. Our goal here is not theory testing per se; rather, it is to
provide a theoretically informed examination of the potential conditioning
relationship between heterogeneity and economic disadvantage across large
cities in the USA.
Methods
Data
Three sources of data are used in this study: U.S. Census Bureau, Law Enforcement Management and Administrative Services (LEMAS; U.S. Department of
Justice, 2000), and the National Youth Gang Survey (NYGS). Across the unit of
analysis—the 100 most populated US cities—we used the above sources of data
for our outcome and explanatory variables. All of our explanatory variables
were extracted from the 2000 U.S Census and the 2000 LEMAS data. The U.S.
Census data were specific to city demographic variables, while the LEMAS data
were used to control for law enforcement-related characteristics within a city.
The Office of Juvenile Justice and Delinquency Prevention (OJJDP) recognized
the lack of systematically collected statistics on the extent of national gang
problems and established the Nation Youth Gang Center in 1994 (Curry &
Decker, 1998). This annual survey administered by the National Youth Gang
Center (NYGC) is the only systematically collected source of gang and gang
member data available (Klein & Maxson, 2006, p. 212). We used the data
collected between 2002 and 2006 for the construction of our dependent variable. This allowed us to establish causal sequencing with our independent variables. Missing data were addressed using PRELIS version 2.30 (Lisrel 8.72;
Jöreskog and Sörbom, 1996). Better known as “hot-decking,” this procedure
uses donor cases with similar response patterns to impute missing values. This
method of imputation has been touted as a superior alternative to mean
replacement and listwise deletion (Gmel, 2001; Little, Yosef, Cain, Nan, &
Harlow, 2008).
RACIAL AND ETHNIC HETEROGENEITY AND GANGS
9
Dependent Variable
The NYGC collects data annually from law enforcement agencies on the extent
to which gang problems exist in cities across the nation (NYGC, 2007). These
data have been used frequently in government publications (e.g., National
Youth Gang Center, 2007) and in peer-reviewed outlets (e.g., Katz, Maguire, &
Roncek, 2002; Wells & Weisheit, 2001). The survey contains information specific
to a city’s onset of gang problems, number of gang homicides, demographics of
gang members, the number of gangs, and the outcome variable of interest in
the present study—the number of gang members.
Law enforcement gang intelligence has been found to be a useful estimate of
city gang activity (Katz, Webb, & Schaefer, 2000). Some have questioned law
enforcement gang information systems on the basis that they are unable to
distinguish between gang and nongang individuals as well as differences in
offenses (McCorkle & Miethe, 1998; Zatz, 1987). Katz et al. (2000), however,
found that gang intelligence exhibited strong discriminant validity for both
gang/nongang youth and offenses. Katz et al. (2000, p. 434) concluded that “as
a measurement tool, gang documentation and the use of data from gang lists
goes beyond mere face validity and reflects the underlying properties and
behaviors associated with gang participation.” Moreover, Katz, Ballance, and
Britt (2005), in an examination of NYGC data and Arizona Youth Gang Survey
data, found data on gang members to be stable and reliable, especially from
law enforcement agencies serving larger populations.4
To further assess the internal reliability and validity of our dependent variable,
we conducted a series of analyses. First, employing a test/re-test method, we
examined the between-year zero-order correlations of gang membership in the
cities in our sample. The mean inter-item correlation across years was .92,
suggesting extremely high reliability. Similarly, to explore the convergent validity
of the measure, we examined the zero-order correlations between the dependent
variable and other NYGC measures (number of gangs; number of gang homicides)
that would be expected to increase as the number of gang members increases.
This analysis revealed that our dependent variable was moderately correlated
(mean inter-item r = .47) with the number of gangs and highly correlated with
number of gang homicides (mean inter-item r = .85). The moderate correlation
of the former is not surprising given that gangs operate within different structures
resulting in larger or smaller numbers of members (Klein & Maxson, 2006, pp.
172-193). Given Katz and colleagues’ assessment of the external reliability (Katz
et al., 2005) and validity (Katz et al., 2000) of the police data and our examination
4. It would be difficult for any study to undertake a macro-level study of gangs/gang membership
without the extensive assistance of law enforcement. The police consistently interact with gangs
and gang members, having the ability to observe, document, and address emerging gang trends on
the street (Decker, 2003; Katz & Webb, 2006). Some empirical research has taken on the task
of collecting nation-level gang information (e.g., Maxson & Klein, 1995; Miller, 1982; Needle &
Stapleton, 1983), but not systemically over time. In addition, the NYGC has been in operation since
1994 and has had the opportunity to refine and bridge some of the gaps between law enforcement
and academia.
10
PYROOZ ET AL.
of the internal reliability of the police data, the use of our dependent variable is
appropriate, as it has passed important measurement tests.
We examined the NYGS data for the 100 largest cities in the USA.5 Our
objective was to construct a measure that most accurately depicted the
number of gang members documented by law enforcement officials in their
city. This was done by creating a multi-year average as our dependent variable, with yearly survey points between 2002 and 2006 to establish temporal
order with our independent variables, using the mean as our measure of
central tendency. This method limits any internal inconsistencies and the
potential for large fluctuations within a city. The natural log was taken to
approximate a normal distribution. A list of the 100 cities included in the analyses can be found in Appendix A. (See Table 1 for the mean, standard deviation, minimum, and maximum values for all the variables included in the
analysis.)
Independent Variables
Heterogeneity
The measure of racial and ethnic heterogeneity traditionally used in macrolevel research was developed by Blau (1977). This measure varies from 0 to
1 and is calculated by taking one minus the squared proportions of the
population in each racial and ethnic group. Traditionally, studies using Blau’s
operationalization of racial and ethnic heterogeneity consist of five or six
major racial/ethnic groups—whites, blacks, Hispanics, Asians, and others. For
example, Salvadorans would be lumped in the Hispanic category.6 As Vigil’s
(2002) and Maxson’s (2009) research underscored, Salvadoran gangs in Los
Angeles (e.g., MS-13) grew in the 1980s, partly in response to the threat
from other racial and ethnic groups. Further, there are substantive differences between many Hispanic groups (e.g., Mexicans, Cubans, Salvadorans,
etc.). Esbensen, Brick, Melde, Tusinski, and Taylor (2008) recommended that
future research be more attentive to gang and gang membership ethnic
differences. With this in mind, for the present study, we expanded the
5. The NYGC typically sends the questionnaire to the same respondent each year and maintains
roughly a 90% response rate. The NYGC asks a law enforcement representative to report information
about “a group of youths or young adults in your jurisdiction that you or other responsible persons in
your agency or community are willing to identify as a ‘gang.’ Motorcycle gangs, hate or ideology
groups, prison gangs, and exclusively adult gangs were excluded” (NYGC, 2007).
6. For instance, if City A is 88% White and 12% Hispanic, this city would receive a heterogeneity
score of .2112, indicating a high level of homogeneity. On the other hand, if City B comprises a
population that is 88% White, 4% Cuban, 4% Mexican, and 4% Salvadoran, the city would have a
heterogeneity score of .2208. While this difference between City A and City B is small, it is a difference that would be masked by combining all Hispanics (or other ethnic groups) into one homogenous
category.
RACIAL AND ETHNIC HETEROGENEITY AND GANGS
11
traditional Blau measure to account for smaller ethnic groupings, capturing
more unique differences between cities to operationalize heterogeneity.7
Economic disadvantage
Similar to past research, our measure of economic disadvantage consisted of
five items derived from the 2000 U.S. Census (McNulty & Bellair, 2003; Parker &
Reckdenwald, 2008; Peterson, Krivo, & Harris, 2000). The variables included
were: (1) percentage of female-headed single parent families with children
under 18, (2) percentage of population receiving public assistance, (3) percentage of population with less than high school education, (4) median household
income in 1999, and (5) percentage of population in civilian labor force unemployed. Using principal axis factoring, these items loaded on one factor with an
eigenvalue of 3.06, which explained 61.22% of the variance in the measures.
The lowest factor loading was .65 (percent with less than high school education). The psychometric properties of the scale exhibited strong internal consistency (Cronbach’s alpha = .78; mean inter-item correlation = .60). The five
items were then condensed into a factor score for analytic purposes.
Control Variables
To ensure that any observed relationships between our dependent variable and
key independent variables are not spurious, we control for four city-level, structural characteristics. First, we controlled for two city demographic factors,
youth male population and population density. Gang research has consistently
found that males are more likely to join gangs than females (Klein & Maxson,
2006). Because of this, we control for the percent of the youth male population
between 12 and 24 years old, a standard age interval for gang involvement.
Further, prior macro-level gang research, in addition to other macro-level criminological research, has also consistently controlled for population density
(Baller, Anselin, Messner, Deane, & Hawkins, 2001; Jackson, 1991; Pratt &
Godsey, 2003; Wells & Weisheit, 2001). Second, we control for two indicators of
law enforcement composition and practice, whether the city had a gang unit
and police representativeness. These are important factors to consider when
examining city-level gang membership, especially since the NYGS based on law
enforcement data. By controlling for whether the city has a gang unit we take
into account any potential bias associated with the reporting techniques due to
7. The following racial and ethnic groups were considered for our analyses: Alaskan Natives, American
Indian, Argentinian, Asian Indian, Bangladeshi, non-Hispanic Blacks, Bolivians, Cambodians, Chileans,
Chinese, Columbians, Costa Ricans, Cubans, Dominicans, Ecuadorians, Filipinos, Guatemalans,
Hmong, Honduran, Indonesian, Japanese, Koreans, Laotians, Malaysians, Mexicans, Micronesians,
Multiple Races, Nicaraguans, other Pacific Islander, Pakistani, Panamanian, Paraguayan, Peruvian,
Polynesians, Puerto Ricans, Salvadorans, Spaniards, Sri Lankans, Thai, Uruguayans, Venezuelans, Vietnamese, non-Hispanic Whites.
12
PYROOZ ET AL.
the presence (or lack) of the unit (Katz & Webb, 2006; Katz et al., 2000). In a
similar vein, we control for the racial representativeness of the city’s police
force, based on the assumption that opposite-race officers might profile youth
(Raganella & White, 2004; Walker & Katz, 2007). We do this by dividing the
percentage of the White police force by the percentage of the city’s White
population, with scores distributed close to 1 indicating a racial representativeness while scores greater than 1 point to an over-representativeness of Whites.
Analytic Strategy
We estimate a series of weighted least-squares (WLS) regression models to
determine the relationship between city-level structural characteristics and
gang membership. Consistent with other macro-level studies (see Pratt &
Godsey, 2003; Reisig, Bales, Hay, & Wang, 2007), WLS regression helps correct
for the problems associated with heteroscedasticity that plagues ordinary leastsquares regression (Tabachnick & Fidell, 2007). Each case was weighted by
population in order to induce homoscedasticity among the error variances.
Our strategy for determining whether heterogeneity conditions the relationship between economic disadvantage and gang membership is presented across
four models. In our first model, we regress our dependent variable—gang
members—on economic disadvantage and our control variables to establish
whether an accepted contention within the gang literature (disadvantage
increases the likelihood of gangs/gang membership) is indeed present in a
macro-level, multivariate context (Bowker & Klein, 1983; Curry & Spergel, 1992;
Short & Strodtbeck, 1965). The second model tests, with disadvantage excluded,
whether there is a relationship between heterogeneity and gang membership
across cities. This model will confirm qualitative accounts that ethnicity has
been an overlooked covariate of gang membership (Alonso, 2004; Davis, 2006;
Moore, 1985; Thrasher, 1927; Vigil, 1988, 2002). In our third model, we include
both measures—economic disadvantage and heterogeneity—to examine the
independent effects net of each other (i.e., additive model). The final model
includes an interaction term between our key measures to determine whether
heterogeneity conditions the relationship between economic disadvantage and
gang membership. Our strategy will allow us to assess the independent, additive, and multiplicative effects of our measures, as well as providing the first
examination of structural covariates of gang membership at a macro-level of
analysis.
Results
Table 1 displays the descriptive statistics and zero-order correlations for the
study variables. The cities included in this analysis ranged considerably in population density from 153.40 to 26,403.90 persons per square mile. This item is
RACIAL AND ETHNIC HETEROGENEITY AND GANGS
13
Table 1 Zero-order correlations and summary statistics for study variables (N = 100)
1 Gang members (ln)
2 Gang unit
3 Youth male population
4 Population density
(×103)
1
2
—
.305*
—
.035
−.052
3
5
6
7
Min
Max
3.89 11.10
0.00 1.00
—
.327** .136 −.180
5 Police representativeness .225*
4
6.66 14.69
—
.113
.072
.464**
6 Economic disadvantage
.273** −.007
.096
.342**
7 Heterogeneity
.481** .210* −.090
0.15 26.40
—
.311**
0.83 4.68
—
.422** .296** .175 —
Mean
6.98
.77
9.75
4.68
1.49
.00 .58
Standard deviation
1.38
.42
1.28
4.01
.55
.98 .13
−2.58 2.31
0.22 0.84
*p < .05; **p < .01.
expressed in scientific notation in Tables 1 and 2 for purposes of interpretation.
The proportion of youth male residents ranged from nearly 7% to 15% in cities
included in the sample (mean = 9.75%). Most of the urban police departments
employed a full-time specialized gang unit (mean = .77%) and the police representativeness measure indicated that officers (e.g., White officers), on average,
make up about 49% more of the police force than they do the population in
these cities. Our two key independent variables, economic disadvantage and
heterogeneity, revealed considerable variation across the sample, which can be
seen in Table 1. Before presenting the multivariate findings, it is first necessary
to diagnose whether our models are subject to any issues related to multicollinearity.
Bivariate Correlations and Assessing Multicollinearity
Zero-order correlation coefficients are presented in Table 1. The coefficients in
the first column reveal the bivariate relationships between the independent
variables and gang membership. Gang unit, population density, and police
representativeness positively varied with the dependent variable while youth
population did not. Our two key explanatory variables, economic disadvantage
and heterogeneity, had positive associations with our outcome variable, meaning more disadvantage and greater heterogeneity increase with gang members.
While these bivariate relationships reveal important information about our
study variables, the associations are independent of other influential variables.
In other words, it is necessary to explore the effects in a multivariate context.
Before doing so, it is necessary to determine whether multicollinearity is
present within our data.
14
PYROOZ ET AL.
The problems of multicollinearity can occur when two highly correlated independent variables are both included in a model which may introduce bias that
results in unstable parameter estimates (Farrar & Glauber, 1967; Tabachnick &
Fidell, 2007). As a preliminary precaution, the inter-correlations between independent variables were examined to see if any associations exceeded the .70
threshold deemed as problematic (Fisher & Mason, 1981). This was not the case,
however, problems may still be present when bivariate correlates are large. For
example, population density and police representativeness have a fairly large
bivariate correlation (r = .46), which could potentially be an issue if we based
our diagnostics solely on inter-correlations (Tabachnik & Fidell, 2007). To determine whether this was indeed the case, we examined the squared multiple
correlations (SMC) for each predictor. Variance inflation factors (VIFs) revealed
scores no greater than 3, and conditioning index numbers fell within the standard specified by Belsley, Kuh, and Welsch (2004).8 Based on our examination of
the various collinearity diagnostics (bivariate correlations, tolerance, VIF, and
condition index), we are confident in proceeding to the multivariate models.
Multivariate Models
Table 2 displays the results from our WLS regression models examining the
impact of heterogeneity and economic disadvantage on the number of gang
members across cities.
In Model 1, the purpose is to first establish whether economic disadvantage is
predictive of city-level aggregate gang membership net of controls. The model
is robust, explaining 28% of the variation in the dependent variable. In this
model, the two control variables that are significantly related to gang membership are specialized gang unit (β = .289, p < .01) and population density (β
=.254, p < .05). The insignificance of the police representativeness measure
indicates that the bivariate effect was washed out in the multivariate context.9
Our key measure, economic disadvantage, did in fact have a positive direct
effect on the number of gang members across large cities in the USA. The relationship was moderate (β = .250, p < .05), but the effect size was less than both
specialized gang unit and population density. From a controls-only model
8. We looked at the variation explained in one variable by treating it as an outcome of the remaining independent variables. Typically, statistical packages (e.g., SPSS) convert the SMC to a tolerance
statistic—(1–SMC)—or to a VIF—(1/tolerance)—in reporting whether multicollinearity is present. VIF
scores greater than 4 (i.e., tolerance scores less than .25) are seen as the threshold indicative of
multicollinearity (Fox, 1991). Maddala (1992) pointed out the limitations in basing collinearity diagnostics solely on VIF scores, which is why we examined the conditioning index.
9. As an added measure of precaution, we also considered the number of officers per 1,000 citizens
across cities. The effect was not significant in the bivariate (and thus multivariate) model. Including
the item in the multivariate models did not improve model fit, nor did it alter the sign and significance of the other explanatory variables. Moreover, as it is common in macro-level research because
of small sample size, we needed to consider the potential for oversaturating the model (Long,
1997). As a result, we retained gang unit and police representativeness as opposed to officers per
capita due to the obvious contribution the former two variables made in the bivariate models.
RACIAL AND ETHNIC HETEROGENEITY AND GANGS
15
Table 2 WLS regression models predicting gang members (ln) at the city-level (N = 100)
Independent variable
Interaction: Disadvantage–
heterogeneity
Heterogeneity1
Model 1
Model 2
Model 3
—
—
—
—
6.576 (1.114)
0.566**
Economic disadvantage
0.447 (0.180)
—
0.250*
Gang unit
1.368 (.426)
0.784 (.385)
0.289**
0.166*
Youth male population
0.142 (0.153) 0.165 (0.135)
0.090
0.104
Population density (×10−3) 0.049 (0.022) 0.009 (0.021)
0.254*
0.048
Police representativeness −0.044 (0.336) 0.127 (0.287)
0.013
0.038
Model F
7.308**
14.773**
Model R2
0.280
0.440
Model 4
2.357 (0.927)
0.211*
6.598 (1.070) 6.325 (1.046)
0.568**
0.545**
0.453 (0.152) 0.439 (0.148)
0.253**
0.245**
0.856 (0.371) 0.986 (0.364)
0.181*
0.209**
0.124 (0.130) 0.066 (0.128)
0.078
0.042
0.008 (0.021) 0.029 (0.022)
0.043
0.148
−0.091 (0.285) 0.143 (0.292)
0.027
0.043
14.818**
14.370**
0.489
0.522
*p < .05; **p < .01.
1In
Model 4, this variable was mean-centered due to creating an interaction term.
Note. Unstandardized coefficients (β) and standard errors (SE, in parentheses) are located above
the standardized coefficients (β).
(model not reported), the coefficient of multiple determination increased by
nearly 20% (R2∆ = .05, p < .01). More importantly, this finding empirically
confirms the existing literature that economic disadvantage is in fact a significant structural correlate of gang membership at the city-level unit of analysis.
Model 2 examines whether heterogeneity should be considered as a factor in
explaining gang membership. Similar to Model 1, this model is fairly robust and
explains 44% of the variation in the dependent variable. Heterogeneity does in
fact substantially improve the amount of variation explained in comparison to a
controls-only model (R2∆ = .21, p < .01). Heterogeneity exhibited a strong direct
effect (β = .566, p < .01) on gang membership. Population density no longer had
an effect in this model, suggesting that heterogeneity reduced the relationship
to nonsignificance. As a result, both Models 1 and 2 are consistent with our
hypotheses that economic disadvantage and heterogeneity do indeed have
direct effects on gang membership.
To examine whether the findings from Models 1 and 2 are truly “robust,” Model
3 includes both key independent variables to determine whether their effects on
gang membership hold when both measures are in the model. With both predictors in the model, the variation explained increases from that of the first two
models (R2 ∆ = .49; p < .01). This model reveals that both economic disadvantage
(β = .253, p < .01) and heterogeneity (β = .568, p < .01) maintain their significance
and effect size from the independent models. This tells us that both variables are
significant predictors of gang membership at the city-level net of each other.
16
PYROOZ ET AL.
Aside from these additive effects, the standardized coefficients reveal that
heterogeneity exhibits an effect size twice as great as economic disadvantage.
Model 4 examines the research question of whether heterogeneity conditions
the relationship between economic disadvantage and gang membership.
Because a typical multiplicative term would result in multicollinearity, we
employed a technique to create an interaction term that would not produce
biased parameter estimates (Cohen, Cohen, West, & Aiken, 2003; Jaccard &
Turrisi, 2003). We mean-centered heterogeneity (economic disadvantage, a
factor score, was already mean-centered) and then multiplied the terms. For
the interaction model, we can draw a meaningful conclusion about the effect
size of the standardized coefficient. Jaccard and Turrisi (2003, p. 28) pointed
out that we interpret this finding as the “squared semipartial correlation for the
product term holding constant [the interaction term’s] component parts.” The
interactive effect of heterogeneity on economic disadvantage exhibits a moderate overall effect on the dependent variable (β = .211, p < .05), holding the
independent variable (economic disadvantage) and the moderating variable
(heterogeneity) at a constant of zero. In addition, Model 4 reveals that the
interaction term makes an independent contribution to the proportion of the
explained variation in the dependent variable. The coefficient of multiple
determination increased by nearly 7% (R2∆ = .033, p < .01). Figure 1 depicts the
interaction effect on gang membership. The points on the graph reflect
economic disadvantage and heterogeneity at low (1 SD below the mean) and
high (1 SD above the mean) levels (Aiken & West, 1991). The key finding is that
when heterogeneity is low, the effect of increased economic disadvantage is not
particularly large; however, when heterogeneity is high, more economic disadvantage corresponds to a large increase in gang members.10
It is important to note the effect of heterogeneity in the model. With respect
to the traditional Blau (1977) measure, additional analyses revealed that the
more comprehensive (i.e., more ethnic groups included) measure resulted in a
substantially improved model fit (results not reported in table). For example,
comparing the findings in Model 3 to the more commonly used seven-group
measure (Whites, Blacks, Asians, Hispanics, Native Americans, Multi-race, and
Other), the variation explained decreased from .489 to .430—nearly six units or
a 15% difference. This was consistent (no less than 13%) across all four models.
In addition, the standardized coefficients also revealed a moderate, yet meaningful difference. Across the three models, the effect size increased over 15%
(β∆ ≥ .076). These findings are extremely important not only for understanding
the effects of racial and ethnic heterogeneity on gang membership, but it could
also be that there are meaningful differences in effect sizes for studies examining other outcomes using the Blau measure of heterogeneity.
Figure“Low”
Note.
1 Graphical
and “high”
depiction
scores are
of the
computed
interaction
at one
between
standard
heterogeneity
deviation below
and economic
and above
disadvantage.
the mean.
10. The point estimates are as follows: when heterogeneity is low, economic disadvantage results in
gang membership (ln) scores 4.82 (low disadvantage) and 5.06 (high disadvantage) and a slope of
1.05. When heterogeneity is high, economic disadvantage results in gang membership (ln) scores of
5.90 (low disadvantage) and 7.38 (high disadvantage) and a slope of 1.25. The slope differences as
well as the intercept difference when disadvantage is low are particularly revealing.
RACIAL AND ETHNIC HETEROGENEITY AND GANGS
17
Figure 1 Graphical depiction of the interaction between heterogeneity and economic
disadvantage.
Note. “Low” and “high” scores are computed at one standard deviation below and above
the mean.
Discussion
The goal of the present study was to examine the theoretical correlates of gang
membership across cities in the USA. In doing so, our intent was to determine
whether the impact of economic disadvantage on gang membership was conditioned by heterogeneity. We did this by using U.S. Census 2000 data for city
characteristics, LEMAS 2000 data for police department characteristics, and
NYGS 2002–2006 data for gang membership. This study went beyond existing
macro-level gang research (Jackson, 1991; Wells & Weisheit, 2001) by analyzing
gang membership rather than establishing whether a city had a gang problem.
Since gangs are so prevalent in large cities, this is an important step forward in
understanding the role of theory in accounting for differences in gang membership. Based on our findings, there are four major implications that guide this
discussion.
First, racial and ethnic heterogeneity matters. The current study found that
heterogeneity had an independent, additive, and multiplicative effect on gang
membership net of controls. In fact, heterogeneity exhibited the greatest effect
size in every model (.545 ≤ β ≤ .568). This finding is consistent with accounts of
existing gang formation research that underscore the importance of race and
ethnicity (Alonso, 2004; Davis, 2006; Thrasher, 1927; Vigil, 1988, 2002), but this
is the first time this theoretical framework has been applied to this unit of analysis. The support for the moderation effect indicates that greater degrees of
heterogeneity and economic disadvantage are more likely to correspond with
increases in gang activity. While the larger criminological field and its subset of
18
PYROOZ ET AL.
gang scholars have known that economic disadvantage is a correlate of social
malaise (Bursik & Grasmick, 1993; Pratt & Cullen, 2005; Sampson et al., 1997;
Short, 1997), the findings from the present study suggest that those ignoring
heterogeneity in the gang context may be missing a key piece of the gang
formation/gang membership puzzle. It is important to stress that our measure
of heterogeneity is not a simple dichotomy (e.g., White/non-White); rather, this
measure incorporates a larger number of racial and ethnic groups, better
capturing heterogeneity within a city. Moreover, incorporating more groups into
the analysis revealed a noteworthy change in effect size and model fit. It is our
suggestion that researchers consider incorporating more groups into their analysis, or run the risk of underestimating the relationship between heterogeneity
and the outcome of interest.
Second, theory building often outpaces empirical testing. This problem exists
in the field of criminology (see, e.g., Hay, 2001), as well as within gang
research. A number of influential works (Cloward & Ohlin, 1960; Cohen, 1955;
Klein, 1995; Miller, 1958; Thrasher, 1927; Vigil, 1988, 2002) have offered explanations for the causes of the formation and persistence of gangs, but there are a
limited number of tests that go beyond lower levels of aggregation. Jackson
(1991) tested Wilson’s (1987) proposition—and Hagedorn’s (1988) application to
gangs—that the US transition from a manufacturing-based to service-based
economy spawned a host of social ills at a macro-level, one of them being
gangs. Katz and Schnebly (in press) included economic disadvantage covariates
to explain concentrations of gang members at a neighborhood level. While data
limitations have surely stifled a number of thoughtful ideas, a lack of testing is
a criticism that has been lodged against gang research (Katz & Jackson-Jacobs,
2004, p. 94). As Klein (2005) pointed out, it has taken over six decades to reach
the present state of understanding of gangs, and much work remains to be
done. Indeed, there is an empirical vacancy that needs to be filled as the accumulation and organization of knowledge is impaired to the extent that existing
theory is not systematically tested (see, e.g., Pratt & Cullen, 2005). This issue is
far from trivial, as policy is rooted in theory (Barlow & Decker, 2009; Cullen,
1988; Huff, 1989; Klein, 1995; Klein & Maxson, 2006; Lilly, Cullen, & Ball, 2007;
Pratt, 2008; Spergel, 1995). While the goal of the present study was to “ignite”
the discussion of gangs in a macro-level context, future research should
continue to test the theoretical applicability of social disorganization, strain,
and integrated hypotheses in examining gang-related outcomes. Moreover, it
would be interesting to see how economic, demographic, and law enforcement
changes over time affect gang-related outcomes.
Third, gang research tells us a lot about criminological theory. Social disorganization (Shaw & McKay, 1942; Thrasher, 1927) and extensions of strain
(Cloward & Ohlin, 1960; Cohen, 1955; Miller, 1958) were influenced largely by
explaining the development of deviant peer groups in a neighborhood context.
The present study found heterogeneity and economic disadvantage—two cornerstones of social disorganization theory and viewed as macro-level covariates of
social problems (Lilly et al., 2007; Sampson & Groves, 1989)—to be central to
RACIAL AND ETHNIC HETEROGENEITY AND GANGS
19
accounting for the levels of gang membership across large American cities.
Gangs provide fertile testing grounds for assessing the capabilities of existing
theory because of their applicability across levels of explanation. Short (1985,
1989, 1998) has noted that the “unit of analysis” problem in the field of criminology is best exemplified in the gang context, where individual-level theory
(e.g., self-control, negative affect), micro-level theory (e.g., differential association), and macro-level theory (e.g., social disorganization, strain/anomie)
have the potential for “talking past one another” while attempting to explain
the same outcome.
Fourth, the void in macro-level gang research has been an unfortunate shortcoming for policy formulation. These results suggest several implications for
responding to gangs. Perhaps the most important finding from our work is that
social and economic variables are more important in explaining levels of gang
membership than are criminal justice system characteristics. This is not a
surprising finding, as Sherman et al. (1997) remind us. The emphasis on the role
of fundamental correlates of crime (economic disadvantage, racial and ethnic
heterogeneity, ecological disparities) is important to emphasize in the context
of our own work and that of others (Xie, 2009). For many forms of crime, the
more proximate criminal justice variables make a small contribution to the
explained variation in the face of other social and economic variables. For
gangs, this appears to be the case as well. This means that any response to
gangs must include a broad range of options; the majority of which are noncriminal justice options. As such, comprehensive models (Howell, 2003, 2009; Klein
& Maxson, 2006) are far more likely to have an impact on gang formation and
gang growth. The comprehensive strategy is an example of such a model, as it
includes an emphasis on creating and enhancing social and economic opportunities, as well as organizing local communities in the response to gangs (Howell,
2003, 2009). Our findings regarding the role of economic disadvantage, particularly its interaction with heterogeneity, provide a strong rationale for cities to
adopt approaches that take a broad view of the gang problem and address its
fundamental causes. Our findings regarding the role of racial/ethnic heterogeneity also have policy relevance. Too often policies are conceived or implemented with a narrow view of the racial or ethnic composition of the groups or
neighborhoods. Because of the large share of gang membership accounted for by
Latinos, accurately specifying race/ethnic heterogeneity is of critical importance. This suggests that responses to gangs should partner with a broadly
representative group of agencies, particularly those agencies that can enhance
the economic prospects of city residents and reduce racial and ethnic isolation
and tensions.
There is great theoretical and practical utility in explaining criminological
outcomes such as gang membership across different units of analysis. Conceptually, it allows for a more robust confirmation when assessing the value of a
specific theory. Practically, it allows for more confident conclusions to inform
policy. Within the scope of the present study, it may be the case that the
effects (independent, additive, and multiplicative) behave differently at a
20
PYROOZ ET AL.
lower unit of aggregation. But as a whole, it provides the potential for integrating concepts. Surely, data limitations have made empirical testing a burdensome task for gang research. The lack of validated measures of the number of
gangs, gang members, and gang crimes introduces potential sources of measurement error. That said, the benefits of comparative research are many, and
contribute to our understanding of gangs as well as broader criminological
theory (Klein, 2005). Future empirical investigations should take this into
account, as it is paramount to advancing gang research (and the potential for
future criminological theory development) to the next stage of knowledge.
References
Adamson, C. (2000). Defensive localism in white and black: A comparative history of
European-American and African-American youth gangs. Ethnic and Racial Studies, 23,
272–298.
Aiken, L. S., & West, S. G. (1991). Multiple regression: Testing and interpreting interactions. Newbury Park, CA: Sage.
Alonso, A. A. (2004). Racialized identities and the formation of black gangs in Los
Angeles. Urban Geography, 25, 658–674.
Anderson, E. (1999). Code of the streets: Decency, violence, and the moral life of the
inner city. New York: W. W. Norton and Company.
Baller, R. D., Anselin, L., Messner, S. F., Deane, G., & Hawkins, D. (2001). Structural
covariates of U.S. county homicide rates: Incorporating spatial effects. Criminology,
39, 561–588.
Barlow, H., & Decker, S. H. (2009). Criminology and public policy: Putting theory to
work. Philadelphia: Temple University Press.
Belsley, D. A., Kuh, E., & Welsch, R. E. (2004). Regression diagnostics: Identifying influential data and sources of collinearity. Hoboken, NJ: Wiley.
Blau, P. M. (1977). Inequality and heterogeneity: A primitive theory of social structure.
New York: Free Press.
Blau, J. R., & Blau, P. M. (1982). The cost of inequality: Metropolitan structure and
violent crime. American Sociological Review, 47, 114–129.
Bowker, L. H., & Klein, M. W. (1983). The etiology of female juvenile delinquency and
gang membership: A test of psychological and social structural explanations. Adolescence, 18, 739–51.
Burgess, E. W. (1925). The growth of the city: An introduction to a research project. In R.
E. Park, E. W. Burgess, & D. McKenzie (Eds.), The city (pp. 47–62). Chicago: University
of Chicago Press.
Burgess, E. W. (1928). Residential segregation in American cities. The Annals of the
American Academy of Political and Social Science, 140, 105–115.
Bursik, R. J., Jr., & Grasmik, H. G. (1993). Neighborhoods and crime: The dimensions of
effective community control. New York: Lexington.
Chamlin, M. B., & Cochran, J. K. (1995). Assessing Messner and Rosenfeld’s institutional
anomie theory: A partial test. Criminology, 33, 411–429.
Cloward, R. A., & Ohlin, L. E. (1960). Delinquency and opportunity: A theory of delinquent gangs. New York: Free Press.
Cohen, A. K. (1955). Delinquent boys: The culture of the gang. New York: Free Press.
Cohen, J., Cohen, P., West, S. G., & Aiken, L. S. (2003). Applied multiple regression/
correlation analysis for the behavioral sciences. Mahwah, NJ: Lawrence Erlbaum.
RACIAL AND ETHNIC HETEROGENEITY AND GANGS
21
Cullen, F. T. (1988). Were Cloward and Ohlin strain theorists? Journal of Research in
Crime and Delinquency, 25, 214–241.
Curry, G. D., & Decker, S. H. (1998). Confronting gangs: Crime and community. Los
Angeles: Roxbury.
Curry, G. D., & Spergel, I. A. (1992). Gang involvement and delinquency among Hispanic
and African-American adolescent males. Journal of Research in Crime and Delinquency, 29, 273–291.
Davis, M. (2006). City of quartz: Excavating the future in Los Angeles. New York: Verso.
Decker, S. H. (1996). Collective and normative features of gang violence. Justice
Quarterly, 13, 243–264.
Decker, S. H. (2003). Policing gangs and youth violence. Belmont, CA: Wadsworth.
Decker, S. H., & Curry, G. D. (2000). Addressing key features of gang membership: Measuring the involvement of young members. Journal of Criminal Justice, 28, 473–482.
Decker, S. H., Van Gemert, F. & Pyrooz, D. C. (2009). Gangs, migration, and crime: The
changing landscape in Europe and the USA. Journal of International Migration and
Integration.
Decker, S. H., & Van Winkle, B. (1996). Life in the gang: Family, friends, and violence.
Cambridge: Cambridge University Press.
Duncan, G. J. (1994). Families and neighbors as sources of disadvantage in the schooling decisions of white and black adolescents. American Journal of Education, 103,
20–53.
Duncan, G. J., Brooks-Gunn, J., & Klebanov, P. K. (1994). Economic deprivation and
early childhood development. Child Development, 65, 296–318.
Egley, A. (2005). Highlights of the 2002-2003 National Youth Gang Surveys. Fact Sheet.
Washington DC: Office of Juvenile Justice and Delinquency Prevention.
Egley, A., Jr., Howell, J. C. & Major, A. K. (2004). Recent patterns of gang problems in
the United States: Results from the 1996–2002 National Youth Gang Survey. In F.-A.
Esbensen, S. G. Tibbetts, & L. Gaines (Eds.), American youth gangs at the millennium
(pp. 90–108). Long Grove, IL: Waveland.
Esbensen, F.-A., Brick, B. T., Melde, C., Tusinski, K., & Taylor, T. J. (2008). The role of
race and ethnicity in gang membership. In F. Van Gemert, D. Peterson, & I.-L. Lien
(Eds.), Street gangs, migration, and ethnicity (pp. 117–139). Portland, OR: Willan.
Esbensen, F.-A., & Lynskey, D. P. (2001). Youth gang members in a school survey. In M.
W. Klein, H.-J. Kerner, C. L. Maxson, & E. G. M. Weitekamp (Eds.), The Eurogang
paradox: Street gangs and youth groups in the U.S. and Europe (pp. 93–114).
Dordrecht: Kluwer Academic.
Esbensen, F.-A., & Winfree, L. T., Jr. (1998). Race and gender differences between gang
and non-gang youth: Results from a multisite survey. Justice Quarterly, 15, 505–526.
Farrar, D. E., & Glauber, R. R. (1967). Multicollinearity in regression analysis: The problem revisited. Review of Economics and Statistics, 49, 92–107.
Fisher, J., & Mason, R. (1981). The analysis of multicollinear data in criminology. In J. A.
Fox (Ed.), Methods in quantitative criminology (pp. 99–125). New York: Academic Press.
Fox, J. (1991). Regression diagnostics. Newbury Park, CA: Sage.
Freng, A., & Esbensen, F.-A. (2007). Race and gang affiliation: An examination of multiple marginality. Justice Quarterly, 24, 600–628.
Gmel, G. (2001). Imputation of missing values in the case of a multiple item instrument
measuring alcohol consumption. Statistics in Medicine, 22, 2369–2381.
Hagedorn, J. M. (1991). Gangs, neighborhoods, and public policy. Social Problems, 38,
529–542.
Hagedorn, J. M. (1988). People and folks: Gangs, crime, and the underclass in a rustbelt
city. Chicago: Lake View.
Hay, C. (2001). An exploratory test of Braithwaite’s reintegrative shaming theory.
Journal of Research in Crime and Delinquency, 38, 132–153.
22
PYROOZ ET AL.
Howell, J. C. (2003). Diffusing research into practice using the comprehensive strategy
for serious, violent, and chronic juvenile offenders. Youth Violence and Juvenile
Justice, 1, 219–245.
Howell, J. C. (2009). Preventing and reducing juvenile delinquency: A comprehensive
framework. Thousand Oaks, CA: Sage.
Howell, J. C., Moore, J. P., & Egley, A., Jr. (2002). The changing boundaries of youth
gangs. In C. R. Huff (Ed.), Gangs in America (3rd ed., pp. 3-18). Thousand Oaks, CA: Sage.
Huff, C. R. (1989). Youth gangs and public policy. Crime and Delinquency, 35, 524–537.
Hughes, L. A., & Short, J. F., Jr. (2005). Disputes involving youth street gang members:
Micro-social contexts. Criminology, 43, 43–76.
Ihlanfeldt, K. R., & Scafidi, B. (2002). Black self-segregation as a cause of housing segregation: Evidence from the multi-city study of urban inequality. Journal of Urban
Economics, 51, 366–390.
Jaccard, J., & Turrisi, R. (2003). Interaction effects in multiple regression (2nd ed.).
Newbury Park, CA Sage.
Jackson, P. I. (1991). Crime, youth gangs and urban transitions: The social dislocation of
postindustrial economic development. Justice Quarterly, 8, 379–397.
Katz, C. M., Ballance, S., & Britt, C. (2005). Examining the reliability of the National
Youth Gang Survey. Paper presented at the annual meeting of the Western Society of
Criminology, Honolulu, HI.
Katz, C. M., Maguire, E. R., & Roncek, D. W. (2002). The creation of specialized police
gang units: A macro-level analysis of contingency, social threat and resource dependency explanations. Policing: An International Journal of Police Strategies and
Management, 25, 472–506.
Katz, C. M., & Schnebly, S. M. (in press). Neighborhood variation in gang member
concentrations. Crime and Delinquency. Advance online publication. doi: 10.1177/
0011128708317065.
Katz, C. M., & Webb, V. J. (2006). Policing gangs in America. Cambridge: Cambridge
University Press.
Katz, C. M., Webb, V. J. & Schaefer, D. (2000). The validity of police gang intelligence
lists: Examining differences in delinquency between documented gang members and
non-documented delinquent youth. Police Quarterly, 3, 413–437.
Katz, J., & Jackson-Jacobs, C. (2004). The criminologists’ gang. In C. Sumner (Ed.), The
Blackwell companion to criminology. Malden, MA: Blackwell.
Klein, M. W. (1995). The American street gang: Its nature, prevalence, and control. New
York: Oxford University Press.
Klein, M. W. (2005). The value of comparisons in street gang research. Journal of
Contemporary Criminal Justice, 21, 135–152.
Klein, M. W., & Maxson, C. L. (2006). Street gang patterns and policies. New York:
Oxford University Press.
Knox, G. W. (2000). Introduction to gangs (5th ed.). Peotone, IL: New Chicago School.
Kornhauser, R. R. (1978). Social sources of delinquency: An appraisal of analytic models.
Chicago: University of Chicago Press.
Lilly, J. R., Cullen, F. T., & Ball, R. A. (2007). Criminological theory: Context and consequences. Thousand Oaks, CA: Sage.
Little, R. J., Yosef, M., Cain, K. C., Nan, B., & Harlow, S. D. (2008). A hot-deck multiple
imputation procedure for gaps in longitudinal data on recurrent events. Statistics in
Medicine, 27, 103–120.
Long, J. S. (1997). Regression models for categorical and limited dependent variables.
Thousand Oaks, CA: Sage.
Maddala, G. S. (1992). Introduction to econometrics. New York: MacMillan.
Maxson, C. L. (2001). A proposal for multi-site study European gangs and youth groups. In
M. W. Klein, H.-J. Kerner, C. L. Maxson, & E. G. M. Weitekamp (Eds.), The Eurogang
RACIAL AND ETHNIC HETEROGENEITY AND GANGS
23
paradox: Street gangs and youth groups in the U.S. and Europe (pp. 299–309).
Dordrecht: Kluwer Academic.
Maxson, C. L. (2009). In pursuit of “transnational” gangs: The “special” case of the Mara
Salvatrucha (MS-13). 2008-09 Paul Tappan award address at the Western Society of
Criminology, San Diego, CA.
Maxson, C. L., & Klein, M. W. (1995). Investigating gang structures. Journal of Gang
Research, 3, 33–40.
Maxson, C. L., & Whitlock, M. (2002). Joining the gang: Gender differences in risk factors
for gang membership. In C. R. Huff (Ed.), Gangs in America (3rd ed., pp. 19–35).
Thousand Oaks: Sage.
McCorkle, R. C., & Miethe, T. D. (1998). Political and organizational response to gangs:
An examination of a “moral panic” in Nevada. Justice Quarterly, 15, 41–64.
McNulty, T. L., & Bellair, P. E. (2003). Explaining racial and ethnic differences in serious
adolescent behavior. Criminology, 41, 709–747.
Melde, C., Taylor, T. J., & Esbensen, F.-A. (2009). “I got your back”: An examination of
the protective function of gang membership in adolescence. Criminology, 47, 565–594.
Merton, R. K. (1938). Social structure and anomie. American Sociological Review, 3,
672–682.
Messner, S. F., & Rosenfeld, R. (2007). Crime and the American dream (4th ed.).
Belmont, CA: Wadsworth.
Miller, W. B. (1958). Lower class culture as a generating milieu of gang delinquency.
Journal of Social Issues, 14, 5–19.
Miller, W. B. (1982). Crime by youth gangs and groups in the United States. Washington, DC: U.S. Department of Justice, Office of Juvenile Justice and Delinquency
Prevention.
Moody, J. (2001). Race, school integration, and friendship segregation in America.
American Journal of Sociology, 107, 679–716.
Moore, J. W. (1985). Isolation and stigmatization in the development of an underclass:
The case of Chicago gangs in East Los Angeles. Social Problems, 33, 1–12.
Needle, J. A., & Stapleton, W. V. (1983). Report of the national juvenile justice assessment centers: Police handling of youth gangs. Washington DC: U.S. Department of
Justice.
NYGC (National Youth Gang Center). (2007). National Youth Gang Survey analysis.
Retrieved April 14, 2008, from http://www.iir.com/nygv/nygsa/
Omi, M., & Winant, H. (1994). Racial formation in the United States: From the 1960s to
the 1990s (2nd ed). New York: Routledge.
Page, B. J., & Marcelin, H. L. (2003). Formation of gangs and involvement in drug use
among marginalized youth: Uses of the anthropological view. Free Inquiry in Creative
Sociology, 31, 175–182.
Parker, K. F., & McCall, P. L. (1999). Structural conditions and racial homicide patterns:
A look at the multiple disadvantages in urban areas. Criminology, 37, 447–478.
Parker, K. F., & Reckenwald, A. (2008). Concentrated disadvantage, traditional male
role models, and African American juvenile violence. Criminology, 46, 711–735.
Peterson, R. D., Krivo, L. J., & Harris, M. A. (2000). Disadvantage and neighborhood
violent crime: Do local institutions matter? Journal of Research in Crime and Delinquency, 37, 31–63.
Pratt, T. C. (2008). Rational choice theory, crime control policy, and criminological relevance. Criminology and Public Policy, 7, 43–52.
Pratt, T. C., & Cullen, F. T. (2005). Assessing macro-level predictors and theories of
crime: A meta-analysis. In M. Tonry (Ed.), Crime and justice: A review of research
(Vol. 32, pp. 373–450). Chicago: University of Chicago Press.
Pratt, T. C., & Godsey, T. W. (2003). Social support, inequality, and homicide. A crossnational test of an integrated theoretical model. Criminology, 41, 101–133.
24
PYROOZ ET AL.
Raganella, A. J., & White, M. D. (2004). Race, gender, and motivation for becoming a
police officer: Implication for building a representative police department. Journal of
Criminal Justice, 32, 501–513.
Reisig, M. D., Bales, W. D., Hay, C., & Wang, X. (2007). The effect of racial inequality on
black male recidivism. Justice Quarterly, 24, 408–434.
Rosenfeld, R., Bray, T. M., & Egley, A. (1999). Facilitating violence: A comparison of
gang-motivated, gang-affiliated, and non-gang youth homicides. Journal of Quantitative Criminology, 15, 495–516.
Sampson, R. J., & Groves, W. B. (1989). Community structure and crime: Testing social
disorganization theory. American Journal of Sociology, 94, 774–802.
Sampson, R. J., Raudenbush, S. W., & Earls, F. (1997). Neighborhoods and violent crime:
A multilevel study of collective efficacy. Science, 277, 918–924.
Sanchez-Jankowski, M. (1991). Islands in the street: Gangs and American urban society.
Berkeley: University of California Press.
Shaw, C. R., & McKay, H. D. (1942). Juvenile delinquency and urban areas. Chicago:
University of Chicago Press.
Sherman, L. W., Gottfredson, D. C., MacKenzie, D. L., Eck, J., Reuter, P., & Bushway, S.
D. (1997). Preventing crime: What works, what doesn’t, what’s promising (Research
in brief). Washington, DC: U.S. Department of Justice, National Institute of Justice.
Short, J. F. (1997). Poverty, ethnicity, and violent crime. Boulder, CO: Westview Press.
Short, J. F., Jr. (1985). The level of explanation problem in criminology. In R. F. Meier
(Ed.), Theoretical methods in criminology. Beverly Hills: Sage.
Short, J. F., Jr. (1989). Exploring integration of theoretical levels of explanations: Notes
on gang delinquency. In S. F. Messner, M. D. Krohn, & T. P. Thornberry (Eds.),
Theoretical integration in the study of deviance and crime: Problems and prospects
(pp. 243–261). Albany, NY: State University of New York Press.
Short, J. F., Jr. (1998). The level of explanation problem revisited: The American Society
of Criminology 1997 Presidential Address. Criminology, 36, 3–36.
Short, J. F., Jr., & Strodtbeck, F. L. (1965). Group process and gang delinquency.
Chicago: University of Chicago Press.
Spergel, I. (1995). The youth gang problem. New York: Oxford University Press.
Starbuck, D., Howell, J. C., & Lindquist, D. J. (2001). Hybrid and other modern gangs.
Washington, DC: Office of Juvenile Justice and Delinquency Prevention.
Suttles, G. (1968). The social order of the slum. Chicago: University of Chicago Press.
Tabachnick, B. G., & Fidell, L. S. (2007). Using multivariate statistics (5th ed.). Boston:
Pearson.
Thornberry, T., Krohn, M. D., Lizotte, A. J., Smith, C. A., & Tobin, K. (2003). Gangs and
delinquency in developmental perspective. Cambridge: Cambridge University Press.
Thrasher, F. M. (1927). The gang: A study of 1,313 gangs in Chicago. Chicago: University
of Chicago Press.
Tita, G. E., Cohen, J., & Engberg, J. (2005). An ecological study of gang “set space.”
Social Problems, 52, 272–299.
U.S. Department of Justice. Bureau of Justice Statistics. (2000). Law Enforcement
Management and Administrative Statistics (LEMAS). Sample Survey of Law Enforcement Agencies [Computer file]. ICPSR03565-v2. Ann Arbor, MI: Inter-university
Consortium for Political and Social Research [distributor], 2008-12-04. doi:10.3886/
ICPSR03565
Venkatesh, S. (1997). The social organization of street gang activity in an urban ghetto.
American Journal of Sociology, 103, 82–111.
Vigil, J. D. (1988). Barrio gangs: Street life and identity in Southern California. Austin:
University of Texas Press.
Vigil, J. D. (2002). A rainbow of gangs: Street cultures in the mega-city. Austin:
University of Texas Press.
RACIAL AND ETHNIC HETEROGENEITY AND GANGS
25
Walker, S., & Katz, C. M. (2007). The police in America: An introduction (6th ed.). New
York: McGraw Hill.
Wells, L. E., & Weisheit, R. A. (2001). Gang problems in nonmetropolitan areas: A longitudinal assessment. Justice Quarterly, 18, 791–823.
Whyte, W. F. (1943). Street corner society: The social structure of an Italian slum.
Chicago: University of Chicago Press.
Wilson, W. J. (1987). The truly disadvantaged: The inner city, the underclass, and public
policy. Chicago: University of Chicago Press.
Xie, M. (2009). The effects of multiple dimensions of residential segregation on black and
Hispanic homicide victimization. Journal of Quantitative Criminology.
Zatz, M. S. (1987). Chicano youth gangs and crime: The creation of moral panic. Contemporary Crisis, 11, 129–158.
26
PYROOZ ET AL.
Appendix A.
Akron, Ohio
Albuquerque, New Mexico
Anaheim, California
Anchorage municipality,
Alaska
Arlington, Texas
Atlanta, Georgia
Augusta-Richmond, Georgia
Aurora, Colorado
Austin, Texas
Bakersfield, California
Baltimore, Maryland
Baton Rouge, Louisiana
Birmingham, Alabama
Boston, Massachusetts
Buffalo, New York
Charlotte, North Carolina
Chesapeake, Virginia
Chicago, Illinois
Cincinnati, Ohio
Cleveland, Ohio
Colorado Springs, Colorado
Columbus, Ohio
Corpus Christi, Texas
Dallas, Texas
Denver, Colorado
Des Moines, Iowa
Detroit, Michigan
List of Cities Included in Sample
Glendale, California
Grand Rapids, Michigan
Greensboro, North Carolina
Hialeah, Florida
Philadelphia, Pennsylvania
Phoenix, Arizona
Pittsburgh, Pennsylvania
Plano, Texas
Portland, Oregon
Raleigh, North Carolina
Richmond, Virginia
Riverside, California
Rochester, New York
Sacramento, California
San Antonio, Texas
San Diego, California
San Francisco, California
San Jose, California
Santa Ana, California
Scottsdale, Arizona
Seattle, Washington
Shreveport, Louisiana
Spokane, Washington
St. Louis, Missouri
St. Paul, Minnesota
St. Petersburg, Florida
Stockton, California
Tacoma, Washington
Tampa, Florida
Toledo, Ohio
Tucson, Arizona
El Paso, Texas
Fort Wayne, Indiana
Fort Worth, Texas
Honolulu CDP, Hawaii
Houston, Texas
Indianapolis, Indiana
Irving, Texas
Jacksonville, Florida
Jersey, New Jersey
Kansas, Missouri
Las Vegas, Nevada
Lexington-Fayette, Kentucky
Lincoln, Nebraska
Long Beach, California
Los Angeles, California
Louisville, Kentucky
Lubbock, Texas
Madison, Wisconsin
Memphis, Tennessee
Mesa, Arizona
Miami, Florida
Milwaukee, Wisconsin
Minneapolis, Minnesota
Mobile, Alabama
Montgomery, Alabama
Nashville-Davidson,
Tennessee
New Orleans, Louisiana
New York, New York
Newark, New Jersey
Fremont, California
Fresno, California
Garland, Texas
Glendale, Arizona
Norfolk, Virginia
Oakland, California
Oklahoma, Oklahoma
Omaha, Nebraska
Tulsa, Oklahoma
Virginia Beach, Virginia
Washington, District of
Columbia
Wichita, Kansas
Yonkers, New York