Race, friendship networks, and violent delinquency

31
2 HAYNIE F.DOC 10/31/2006 8:18:55 PM CRIMINOLOGY VOLUME 44 NUMBER 4 2006 775 RACE, FRIENDSHIP NETWORKS, AND VIOLENT DELINQUENCY DANA L. HAYNIE * DANIELLE C. PAYNE Department of Sociology The Ohio State University KEYWORDS: race, peer networks, violence, racial heterogeneity Although a growing body of literature emphasizes the role of friendship networks and peer relations for youth involvement in violence and delinquency, little research has examined the role of friendship networks in understanding the varying involvement of different racial-ethnic groups in violence. Using data from approximately 13,000 respondents to the first two waves of the National Longitudinal Study of Adolescent Health (Add Health), we explore the ability of friendship networks to account for the differential rates of violence among racial-ethnic groups. In addition, we evaluate whether race moderates the degree to which friendship characteristics predict adolescent violence. Findings indicate significant differences in the structure and behavioral orientation of friendship networks across racial-ethnic identities. Moreover, incorporating characteristics of friendship networks into multivariate analyses accounts for greater involvement in violence among black and Hispanic youths. Network * We would like to thank Doug Downey, Paul Bellair, and Jim Moody for helpful comments on early versions of the paper. Please direct correspondence to Dana L. Haynie, Department of Sociology, The Ohio State University, 300 Bricker Hall, 190 North Oval Mall, Columbus, OH 43210, e-mail: [email protected]. This research uses data from Add Health, a program designed by J. Richard Udry, Peter S. Bearman, and Kathleen Mullan Harris, and funded by a grant P01-HD31921 from the National Institute of Child Health and Human Development, with cooperative funding from 17 other agencies. Special acknowledgment is due Ronal R. Rindfuss and Barbara Entwisle for assistance in the original design. Persons interested in obtaining data files from Add Health should contact Add Health, Carolina Population Center, 123 W. Franklin Street, Chapel Hill, NC 27516-2524; www.cps.unc.edu/addhealth/contract.html.

Transcript of Race, friendship networks, and violent delinquency

2 HAYNIE F.DOC 10/31/2006 8:18:55 PM

CRIMINOLOGY VOLUME 44 NUMBER 4 2006 775

RACE, FRIENDSHIP NETWORKS, AND VIOLENT DELINQUENCY

DANA L. HAYNIE* DANIELLE C. PAYNE

Department of Sociology The Ohio State University

KEYWORDS: race, peer networks, violence, racial heterogeneity

Although a growing body of literature emphasizes the role of friendship networks and peer relations for youth involvement in violence and delinquency, little research has examined the role of friendship networks in understanding the varying involvement of different racial-ethnic groups in violence. Using data from approximately 13,000 respondents to the first two waves of the National Longitudinal Study of Adolescent Health (Add Health), we explore the ability of friendship networks to account for the differential rates of violence among racial-ethnic groups. In addition, we evaluate whether race moderates the degree to which friendship characteristics predict adolescent violence. Findings indicate significant differences in the structure and behavioral orientation of friendship networks across racial-ethnic identities. Moreover, incorporating characteristics of friendship networks into multivariate analyses accounts for greater involvement in violence among black and Hispanic youths. Network

* We would like to thank Doug Downey, Paul Bellair, and Jim Moody for helpful

comments on early versions of the paper. Please direct correspondence to Dana L. Haynie, Department of Sociology, The Ohio State University, 300 Bricker Hall, 190 North Oval Mall, Columbus, OH 43210, e-mail: [email protected]. This research uses data from Add Health, a program designed by J. Richard Udry, Peter S. Bearman, and Kathleen Mullan Harris, and funded by a grant P01-HD31921 from the National Institute of Child Health and Human Development, with cooperative funding from 17 other agencies. Special acknowledgment is due Ronal R. Rindfuss and Barbara Entwisle for assistance in the original design. Persons interested in obtaining data files from Add Health should contact Add Health, Carolina Population Center, 123 W. Franklin Street, Chapel Hill, NC 27516-2524; www.cps.unc.edu/addhealth/contract.html.

2 HAYNIE F.DOC 10/31/2006 8:18:55 PM

776 HAYNIE AND PAYNE

racial heterogeneity and friends’ popularity also emerge as particular network characteristics that operate differently for black and white youth.

The causes and consequences of adolescent violence continue to garner considerable research interest, especially given that youth commit a significant portion of all violent crime in the United States (Snyder, 2000) and that juveniles are more than twice as likely as adults to be the victim of a violent crime (Baum, 2005). Furthermore, in the United States, racial and ethnic groups are involved in violence to different degrees. Evidence also indicates that these differences first appear in adolescence. For example, in both official and self-report data, violent juvenile offending is more common among African Americans and Hispanics than among whites and much less common among Asian Americans (Hawkins et al., 2000; Huizinga, Loeber, and Thornberry, 1994; La Free, 1995; McNulty and Bellair, 2003; Sampson, Morenoff, and Raudenbush, 2005; Snyder, 1999). Not only does youth violence exert immediate and considerable costs to society but research also suggests that exposure to violence during adolescence may have lasting consequences throughout the life course (Hagan and Foster, 2001). It therefore remains critical to identify the social and structural factors that elevate the risk of violence among particular racial and ethnic groups.

Despite numerous studies at both the individual and community levels of analysis, explanations of the association between race-ethnicity and violence remain controversial and unresolved (McNulty, 2001; Short, 1997). In addition, although numerous studies have examined the individual-, family-, and neighborhood-level correlates of violent behavior, none have examined the role of peer networks in the race-violence association. This is especially surprising given that a growing body of literature emphasizes the role of friendship networks and peer relations for youth involvement in violence and delinquency (Agnew, 1991; Elliott, Huizinga, and Ageton, 1985; Elliott and Menard, 1996; Haynie, 2001, 2002; Warr, 1993a, 1993b, 1996, 2002). To address these shortcomings, this study uses data from the National Longitudinal Study of Adolescent Health to evaluate the role of peer networks in the race-violence association. In particular we ask: first, whether racial-ethnic differences in violent behavior are mediated by exposure to peer networks, and, second, whether race moderates the degree to which friendship characteristics predict adolescent violence.

In addition, we go beyond previous work by examining race-ethnic differences across a wider range of racial and ethnic groups. This is important because, as McNulty and Bellair (2003) point out, theories about racial-ethnic differences in violence have essentially been developed

2 HAYNIE F.DOC 10/31/2006 8:18:55 PM

RACE, NETWORKS, AND DELINQUENCY 777

in reference to comparisons between whites and blacks. This emphasis on black-white differentials has led to an assumption that the causes presumed to underlie differences in black-white violence are also attributable to differences in violence between whites and other minority groups. This assumption is not only unwarranted, but is also likely to generate stereotypical images of minorities as a homogenous population. Therefore, we focus on black, Hispanic, Asian American, and white adolescents when addressing our research questions.

FRIENDSHIP NETWORKS AND VIOLENCE

Interpersonal relations are a critical link between individuals and the social structure because they can provide access to valued resources that shape and direct individual lives. In particular, peer relations and friendship networks in adolescence may have important implications for understanding how particular groups of individuals, such as racial-ethnic minorities, face higher or lower risks of violence. To understand such links, we draw on a social capital and normative influence perspective that allows us to view friendship networks as an important avenue for providing social capital on which adolescents can draw to promote or inhibit violent behavior, depending on the prevalent normative influences of the group.

Peer relations have long been central to the study of delinquency (Short, 1957) and for good reason. Adolescents spend much time with their friends, attribute great importance to them, and are more strongly influenced by them during this period than at any other time (Bell, 1981; Brown, 1990). Socialization theories have proven particularly important in offering explanations for the strong behavioral similarity among peers in their delinquency levels. For instance, Sutherland’s differential association theory and Akers’ extension, differential reinforcement theory, are specifically devoted to the normative influence process (Akers, 1985; Sutherland and Cressey, 1955). Although the generic socialization process applies to any ongoing and close interpersonal association, theories of crime and delinquency have been especially interested in peers as a source of normative influence.1 Thus it is reasonable to expect that, during adolescence, peers begin to occupy a central role in modeling deviant and

1. This emphasis can be traced back to the “cultural deviance” aspects of social disorganization theory (Shaw and McKay, 1942; Thrasher, 1927). It appears in theories that treat strain (Cloward and Ohlin, 1960; Cohen, 1955) or labeling (Tannenbaum, 1938) as the reason that youth join delinquent groups, and it is a major theme of integrated theories (Elliott, Ageton, and Canter, 1979; Thornberry, 1987). In sum, all of these theories view delinquent friends as having direct influence on adolescents’ behavior.

2 HAYNIE F.DOC 10/31/2006 8:18:55 PM

778 HAYNIE AND PAYNE

prosocial behaviors, and that such networks may offer a potent context in which norms governing the use of violence come to operate. These norms about the appropriateness of violence are more likely to develop when adolescents are enmeshed in friendship networks in which members display deviant and violent behavior. Such a prodelinquent reference group (Glaser, 1956; Shibutani, 1955) promotes delinquent conceptions of the self as well as pro-violence attitudes, justifications, and motives to engage in violent behavior (Heimer and Matsueda, 1994). Involvement in friendship networks may thus be particularly important to understanding variation in youth violence (Akers, 1998; Elliott, Huizinga, and Ageton, 1985; Haynie, Silver, and Teasdale, 2006; Heimer, 1997; Thornberry et al., 1993).

A social capital framework also offers a useful way to conceptualize the role of friendship networks in promoting or inhibiting violent behavior among adolescents. Although operationalized in many ways, social capital typically refers to the resources accessed through relationships and often consists of obligations and expectations for behavior (Coleman, 1988, 1990). One mechanism through which social capital is generated is social embeddedness in friendship networks. Social embeddedness in friendship networks not only generates obligations and expectations for behavior, but also encourages the transmission of information, behavioral norms, and sanctions. For adolescents, friendship networks are unique social contexts that generate these different forms of social capital that can be used to fulfill particular needs for social acceptance, personal identity, and a sense of place (Coleman, 1961).

What is unique about viewing adolescent friendships and network structures as facilitating social capital is that this approach bridges the selection-socialization debate that has plagued the study of delinquent peers. This is because conceptualizing friendships as social capital encompasses both selection, choosing friends who have valued resources, and socialization, being influenced by friends’ resources (Crosnoe, Cavanagh, and Elder, 2003).2 From this perspective, the nature of social capital is inherently context-specific in that it can be used to promote either positive or negative behaviors depending on the behavioral and structural orientation of the network. Although most discussions of social capital focus on its ability to promote positive actions, McCarthy and Hagan (2001) argue that social capital need not foster only positive outcomes, but can also lead to negative outcomes such as involvement in crime and violence (see also Haynie, 2002; for a discussion of criminal

2. This also recognizes the growing consensus that peer similarity in behavior reflects both selection and socialization processes (Haynie and Osgood, 2005; Kandel, 1974; Matsueda and Anderson, 1998).

2 HAYNIE F.DOC 10/31/2006 8:18:55 PM

RACE, NETWORKS, AND DELINQUENCY 779

social capital, see Steffensmeier and Ulmer, 2006). This is especially likely if adolescents are embedded in social networks that generate social capital used to facilitate violent encounters. The orientation of the network is thus likely to determine the type of social capital generated and available. When adolescents are enmeshed in networks in which members express deviant or violent behavior, they are likely to have access to resources that make violence especially likely. Such networks may, for example, encourage violent behavior by providing an appreciative audience, decreasing the likelihood of sanctions, reducing the likelihood of outside witnesses, and rewarding the display of violent behavior (Felson et al., 1994).

RACE AND FRIENDSHIP NETWORKS

Because race is a critical element of American society that organizes interactions and influences individual outcomes, it provides a unique lens through which to view variability in relationship dynamics (Crosnoe, Cavanagh, and Elder, 2003). In particular, directing attention to the structural and behavioral characteristics of networks may illuminate why particular racial-ethnic groups are at greater or lesser risk of violence. Although studies of youth development point to the paramount role of peer social networks—broadly defined—in shaping adolescent behavior (Adler and Adler, 1998), surprisingly little is known about racial differences in peer social networks. Most of the extant literature has focused on cross-race interactions and relationships (Epstein, 1985; Hallinan and Williams, 1987; Moody, 2001). This research is important, but it overlooks the fact that most adolescent friendships are intraracial (Giordano, 2003; Moody, 2001). In addition, surprisingly little research has examined the everyday friendship patterns of minority youth (Giordano, 2003).

One reason that rates of violence might vary between different race and ethnic groups is that the structural and behavioral content of their friendship networks differ (Steffensmeier and Ulmer, 2006). In particular, historically disadvantaged ethnic and racial minorities might belong to networks with greater access to social capital oriented toward more problematic behaviors. For instance, black and Hispanic youth may belong to racially homogenous networks in which friends encourage deviance and violence rather than prosocial behaviors, because prosocial behavior does not yield positive resources. Elijah Anderson’s work (1999) supports this argument with regard to black adolescents from inner-city neighborhoods. Many of these youth do not have access to the human and financial capital of mainstream society, thus rely heavily on access to respect, which, for them, constitutes a very valuable form of social capital. Access to respect is important because it “forms the core of the person’s self-esteem,

2 HAYNIE F.DOC 10/31/2006 8:18:55 PM

780 HAYNIE AND PAYNE

particularly when alternative avenues of self-expression are closed or sensed to be” (Anderson 1999: 66). This respect, at least for minority youth such as African Americans, is achieved through toughness and “a reputation for being willing and able to fight” (67). Therefore, for particular groups of racial minorities, for whom social capital in the form of job networks or expectations of educational success are not available, success through behavior such as violence may be more likely.

Related to this idea of attaining social capital through violence rather than prosocial avenues is the idea that certain groups of minority youth might come to trust particular social structures—especially the educational system—to a lesser extent. Trustworthiness of structures is an important form of social capital (Coleman, 1988). When youth come to realize that striving for success within that structure still does not grant them access to legitimate job opportunities,3 they understandably do not trust that structure to facilitate their goal of economic success and status through traditional routes. This might explain why Hispanic and black youths have higher high school dropout rates and lower levels of educational attainment than whites (Castro, Boyer, and Balcazar, 2000; Taylor, Jacobson, and Roberts, 2000). Thus, this lack of trustworthiness in the educational system may explain why minority youth become embedded in peer networks with less educational success; because the educational system has failed them, they may be more likely to achieve their goals through more dependable means, such as violence.

Another reason certain groups of racial minorities may have lower levels of social capital is that they have less intergenerational network closure. Specifically, historically disadvantaged minority youth are more likely to live in single-parent households—61 percent of black youths, for example, live in such environments (Taylor, Jacobson, and Roberts, 2000). The physical presence of adults in the household—their supervision of their children, their knowledge of their children’s friends and classmates and other parents, and their supervision of other youth—is an important, even critical, form of social capital (Coleman, 1988). This lack of social capital in the form of intergenerational closure and adult supervision might lead minority youth to associate with peers who value or encourage violent behavior.

Ultimately, then, this perspective suggests that location in friendship networks that are less integrated into the school environment (or more racially homogenous due to racial segregation within schools), less oriented toward educational attainment, and more oriented toward problem behaviors (like deviance and violence) might explain why black

3. For example, the poverty rate among blacks and Hispanics is 24 and 22, respectively, compared to 8 for non-Hispanic whites (Proctor and Dalaker, 2003).

2 HAYNIE F.DOC 10/31/2006 8:18:55 PM

RACE, NETWORKS, AND DELINQUENCY 781

and Hispanic adolescents are more involved in violence than their Asian and white counterparts. This leads to our mediation hypothesis:

Race and ethnic differences in violence will be explained by the characteristics of friendship networks in which adolescents are enmeshed.

It is also important to examine whether aspects of friendship networks are more or less important for different racial-ethnic groups. For instance, social capital and segregation literature suggest that as racial and ethnic minorities become more integrated into friendship networks and these networks become more integrated into the overall school culture, violence might decline. One way to explore such integration is to look at the racial heterogeneity of networks within schools.

Residential segregation among disadvantaged minorities—particularly blacks and Hispanics—has led to friendship racial segregation within schools4 and among adolescents, and has potentially negative consequences that could extend well into adulthood. For instance, the social and geographic isolation of minorities, especially blacks, has led some minority students to use language increasingly remote from standard American English (Massey and Denton, 1993). This language barrier may pose serious obstacles to success within mainstream society, because U.S. schools rely almost exclusively on the standard dialect for instruction, and because employers require facility with standard English, especially if the employee hopes to advance. Racial segregation of friendship networks within schools may therefore deny minority youth access to the human and social capital they need to achieve traditional success. Embeddedness in heterogeneous peer networks should, then, provide access to a greater variety of social capital and may ultimately result in network members being less likely to engage in violence.

A second way of looking at integration into networks and school culture is to examine how connected adolescents’ friends are to others in the school network (that is, their popularity). Adolescents with friends who are well liked and connected to many others in the school are likely to benefit from the greater social capital generated (that is, adolescents can draw on the social capital that the friends of their friends generate).

Network racial-ethnic heterogeneity and peer popularity are thus two particularly important structural characteristics of friendship networks to consider. They may represent access to positive social capital for disadvantaged minorities. They may also be less important for more advantaged groups whose friends, parents, and teachers are likely to

4. Even when schools have been desegregated (have proportional numbers of various racial-ethnic groups present) friendship networks can remain completely segregated (Moody, 2001).

2 HAYNIE F.DOC 10/31/2006 8:18:55 PM

782 HAYNIE AND PAYNE

provide adequate social capital. This discussion leads to our integration hypothesis:

Racial heterogeneity of the friendship network and friends’ popularity will be associated with a reduction in violence for black and Hispanic youth but will be unassociated with violence among white and Asian youth.

It is important to consider not only whether the structural characteristics of adolescent networks influence violence committed by different racial and ethnic groups differently, but also whether the influence of friends’ behavioral characteristics differs across racial-ethnic groups. A few studies have addressed the question in relation to comparisons between black and white adolescents. Unfortunately, the literature is inconclusive and often contradictory. For example, one perspective argues that the difficulties disadvantaged minority youth face in achieving success along traditional lines increases the salience of peer group interactions. Although some studies support this position, they tend to lack generalizability because they focus on very disadvantaged black youth from the inner city (Anderson, 1989; Savin-Williams and Berndt, 1990; Silverstein and Krate, 1975; Taylor, 1989). We hope to reduce this uncertainty by using a nationally representative sample of adolescents and by examining both prosocial behaviors (educational attainment) and antisocial behaviors (deviance in the forms of truancy and drug use, and violence) of peers. Overall, this position suggests that historically disadvantaged minorities will be influenced to a greater extent by their peers than white and nondisadvantaged minorities (for example, Asian American youth) will be, because for historically disadvantaged minorities peer status is more attainable than other types of status (for example, economic status) (Anderson, 1990). This line of reasoning leads to our compensation hypothesis:

Friends’ behavior (both prosocial and deviant) will have a stronger effect on black and Hispanic than on white and Asian American adolescent involvement in violence.

On the other hand, a few studies suggest that racial and ethnic minorities are more family centered and less likely to be influenced by their friends’ behavior than white adolescents are. This perspective assumes that, as visible minorities within an often hostile and racist environment, disadvantaged minorities may view their families as a particularly important “safe haven,” resulting in less need for peer relations to serve as a central focus (Giordano, Cernkovich, and DeMaris, 1993). This argument is consistent with Furstenberg’s (1993) finding that parents of minority youth, recognizing the dangers of street life for their children, attempt to compensate for these external dangers by limiting

2 HAYNIE F.DOC 10/31/2006 8:18:55 PM

RACE, NETWORKS, AND DELINQUENCY 783

peer interactions and exercising more vigilant supervision. A few studies support this argument. For instance, Iscoe, Williams, and Harvey (1964) report that in an experimental setting black youths are less likely than whites to respond to peer pressure on a task that involved counting a metronome click. Billy and Udry (1985), focusing on sexual behaviors, report that the relationship between the sexual behaviors of same-sex friends is stronger among whites than among blacks. Matsueda and Heimer (1987), using data from a sample of adolescents in junior and senior high schools in one California county, find that associating with delinquent peers is a weaker predictor of delinquency for black than white youth. And, in a sample of 942 adolescents in Toledo, Ohio, Giordano and colleagues (1993) find that black (in contrast to white) youth describe friendships as less intimate, perceive less peer pressure, and report lower need for approval from peers. Therefore, evidence from these studies leads to the fourth and final hypothesis, which we call peer-rejection:

Friends’ behavior (both prosocial and deviant) will have less of an influence on black and Hispanic adolescent involvement in violence than on white and Asian American involvement.

In addition to testing these various hypotheses, we expand on prior research in several important ways. First, we use a nationally representative sample of school-aged adolescents and focus on several racial-ethnic groups (black, Hispanic, Asian, and white adolescents) rather than only on blacks and whites. Second, we take advantage of the detailed social network data available in the Add Health study, which allows us to recreate friendship networks in school and measure friends’ behavior based on the responses of actual friends rather than on potentially biased respondents. Third, we focus on several characteristics of friendship networks that have not been examined in prior literature, including structural characteristics of friendship networks (racial heterogeneity and peer popularity) and prosocial and problematic behavioral characteristics of friendship networks (friends’ grade point average, involvement in deviance, and involvement in violence). Fourth, we control for common correlates of violence and for common explanations of minorities’ higher involvement in violence, including prior violence and deviance, family structure, family SES, self-control, residential mobility, and residence in a rural, suburban, or urban area.

Overall, this strategy allows us to, first, provide a detailed description of racial and ethnic differences in friendship networks for a nationally representative sample of school-aged adolescents in the United States; second, evaluate whether friendship networks can account for the race-violence association; and, third, evaluate whether friendship networks operate similarly across multiple racial-ethnic comparisons.

2 HAYNIE F.DOC 10/31/2006 8:18:55 PM

784 HAYNIE AND PAYNE

DATA AND MEASURES

To address these issues, we use data from the first two waves of the National Longitudinal Study of Adolescent Health. Add Health is a multisurvey, multiwave study of U.S. adolescents, their parents, their friends, and their schools (Bearman, Jones, and Udry, 1997). In the initial in-school survey, conducted in 1994 and 1995, all students attending each of 134 high schools and their “feeder” middle schools (grades 7 through 12) were surveyed (n = 90,118). This sample is the basis for the construction of most of the measures of friendship network characteristics. A randomly sampled subset of about 20,000 of these students was subsequently interviewed at home, as was, in most cases, one of their parents. In the second wave, conducted in 1996, all adolescents in grades 7 through 11 at wave 1 were interviewed a second time (n = 14,738), with a follow-up response rate of 88 percent.5 To preserve correct causal ordering, we take our dependent variable from the second in-home interview and our independent variables from the in-school and initial in-home survey. Our final selection of respondents is limited to those who were attending school at the time of the initial in-home survey, were subsequently interviewed in the second wave (n = 14,738), were assigned valid sampling weights, identified their primary race as white, black, Hispanic, or Asian, nominated or were nominated as a friend by at least one person in their school, and had no missing data on the dependent variable.6 These restrictions result in a sample of 12,912 respondents.

Dependent variable. The dependent variable, which is measured from items collected during the second in-home survey, represents whether the respondent had engaged in serious violence during the previous 12 months (1 = had engaged in violence, 0 = had not engaged in violence). The scale is derived from four items, indicating how often the respondent had gotten into a serious physical fight, hurt someone badly enough to need bandages or care from a doctor or nurse, pulled a knife or gun on someone, or shot or stabbed someone (α = .71). Each item was first recoded into a binary

5. It is important to keep in mind that respondents in grade twelve who participated in the Wave I in-home interview were not eligible to be included in the Wave II in-home survey. Despite this, our reliance on including respondents who participated in both Wave I and Wave II surveys resulted in a sample that contained slightly more white respondents and respondents from two-parent families. These sample selection issues need to be considered when generalizing results.

6. We deleted 1,170 observations due to missing sample weights, 212 of respondents who identified as Native American or Other Race because these two groups combined represented less than 1 percent of our sample, 324 observations because respondents did not have any network ties in their school, and 120 observations because respondents did not respond to the questions assessing violence at Wave II.

2 HAYNIE F.DOC 10/31/2006 8:18:55 PM

RACE, NETWORKS, AND DELINQUENCY 785

response, indicating whether the respondent had engaged in that activity. The final violence scale represents a binary response of whether the respondent had engaged in any of these four activities.7

Race. Race-ethnic group membership is measured as four dichotomous variables for white, black, Hispanic, and Asian. These were created from a series of items asking the respondent to report their race and Hispanic origin. Those who answered that they had Hispanic origins were coded as Hispanic. Those who answered that they were not Hispanic and chose only one race were coded as that race—either white, black–African American, or Asian–Pacific Islander. Non-Hispanics who chose more than one race were coded based on their response to the question asking which single category best described their racial background.

Friendship Network Variables. To determine the structural and behavioral characteristics of adolescent friendship networks, especially as they pertain to youth from minority racial backgrounds, it is important to use network data collected from all nominated members of the respondent’s network. A distinctive characteristic of the Add Health data is that they enable one to measure behavioral characteristics and structural network properties directly from friends, rather than from perceptions of friends derived from the respondent, which often are ill informed and subject to measurement error (Haynie, 2001; Haynie and Osgood, 2005; Jussim and Osgood, 1989). Fortunately, during the in-school questionnaire, students were given rosters of all the students in their school and were asked to list up to five of their closest male and five of their closest female friends using unique identification numbers (for a potential total of ten friends). All of our network variables—two structural characteristics and three behavioral characteristics—use this feature of the Add Health design.

The two structural network characteristics are peer network racial heterogeneity and network popularity. Racial heterogeneity is the diversity of races represented within the respondents’ networks. Formally, this is calculated as:

where A = race; Ak = the number of individuals with trait k in the respondent’s network; en = the number of individuals in the respondent’s network with valid data on A; and n = the total number of traits of A represented in the respondent’s network. A score of 0 indicates that the

7. We also ran supplementary analyses with an ordered logit scale of violence (ranging from 0 to 4) instead of a dichotomous outcome. The results from those analyses are substantively similar to the ones we report here.

2

1HETEROGENEITY 1

nK

iAAen

⎡ ⎤⎛ ⎞⎢ ⎥= − ⎜ ⎟⎢ ⎝ ⎠ ⎥⎣ ⎦∑

2 HAYNIE F.DOC 10/31/2006 8:18:55 PM

786 HAYNIE AND PAYNE

network is racially or ethnically homogenous, and higher scores indicate increasing racial heterogeneity. Peer popularity is measured as the average number of nominations that those in the respondent’s friendship network received from other adolescents within their school (or the feeder school).

We use three behavioral characteristics in these analyses: network GPA, network deviance, and network violence. Network GPA is the average of friends’ grades in math, science, English, and history. Network deviance is a scale that combines the network averages of three variables: mean number of times friends report getting drunk, smoking cigarettes, and skipping school without an excuse (in the previous 12 months). Each item ranges from 0 = “never” to 6 = “nearly every day” (alpha = .76). After each friend’s response to the three questions were summed, all responses to the items were summed, and the final score was divided by the number of friends in the network, yielding a measure of the average deviance reported by the respondent’s friends. Network violence is the average of the number of times the friends report getting into a physical fight in the previous 12 months (0 = “never” to 4 = “more than seven times”).

Control variables. We control for a number of individual-level demographic and behavioral characteristics because previous research indicates that these factors are associated with network characteristics or violence (Clark and Ayers, 1992; Giordano, Cernkovich, and DeMaris, 1993; Hallinan and Williams, 1987; McNulty and Bellair, 2003; South and Haynie, 2004; Williams et al., 1999). We control for sex (male = 1, female = 0) because research has documented greater involvement in violence among males. We also control for age, which is measured in number of years. To control for residential mobility that might disrupt adolescent networks, we include a binary measure indicating whether the respondent has moved within the past 2 years (moved = 1, not moved = 0). We also control for respondent location in an urban, suburban, or rural setting, using three dichotomous variables (with urban as the reference category). Number of out-of-school friendship nominations is measured as the number of friends nominated by the adolescent who are not on their school roster or on the roster of the feeder school with which their school is affiliated.8 Family structure is measured as a dichotomous variable indicating whether the respondent lives in a two-parent family (= 1) or other family type (= 0). Family socioeconomic status is a scale taken from items in the parents’ questionnaire that measure mother’s and father’s education and occupation (each coded as a 5-point scale, from low to

8. Supplementary analyses indicate that the number of out-of-school nominations neither differs across racial categories nor across those reporting violence and those not reporting violence.

2 HAYNIE F.DOC 10/31/2006 8:18:55 PM

RACE, NETWORKS, AND DELINQUENCY 787

high). The SES score is then the sum of the rankings of education and occupation (Bearman and Moody, 2004).

The rest of our control variables are behavioral. We control for whether the respondent has a bad temper (yes = 1, no = 0), based on an item in the parents’ questionnaire. We measure GPA as the average of the respondent’s grades in math, science, English, and history. Unstructured socializing is measured as the number of friends with whom the respondent hangs out with after school. Finally, we control for prior deviance and prior violence. Prior deviance is a summed scale representing the number of times that the respondent got drunk, smoked cigarettes, and skipped school without an excuse during the previous year with each item ranging from 0 = “never” to 6 = “nearly every day” (α = .74). Prior violence indicates the number of times that the respondent got into a physical fight in the previous 12 months where responses ranged from 0 = “never” to 4 = “more than seven times.”

Analytic Strategy. We begin by examining bivariate differences in friendship network characteristics between racial-ethnic minority and white adolescents. We also examine racial-ethnic differences in violence and in the control variables (table 1). We then estimate a series of survey corrected logistic regression equations designed to examine the effects of race-ethnicity on the likelihood of violence, net of our control variables (table 2, model 1), and the degree to which characteristics of friendship networks can account for racial-ethnic differences in violence (table 2, model 2). Following this we examine a series of models (table 3) that incorporate interaction terms between race-ethnicity and each network characteristic that allow us to determine whether the role of network characteristics on violence is conditioned by race. For all analyses, we use the survey correcting procedures available in Stata that produce estimates that adjust for the clustering of observations within schools (Chantala and Tabor, 1999). We use a multiple imputation procedure to handle missing data (Allison, 2001) and sampling weights are applied to achieve national representativeness.

RESULTS

Table 1 presents descriptive statistics on all of the variables separately for each racial-ethnic group (with white adolescents serving as the reference category). Although involvement in violence is relatively rare, rates of violence do vary somewhat across racial-ethnic groups. Compared to white adolescents, for whom 20 percent of the sample report engaging in a violent offense in the past 12 months, 25 percent of blacks and 27 percent of Hispanics report involvement in violence. Asian adolescents, on the other hand, report participating in significantly lower levels of violence

2 HAYNIE F.DOC 10/31/2006 8:18:55 PM

788 HAYNIE AND PAYNE

(15 percent of the sample) compared to whites. These descriptive differences are consistent with earlier findings (McNulty and Bellair, 2003).

Of particular interest to our research questions, minorities also differ significantly from whites in the characteristics of their friendship networks. Again consistent with earlier findings, all racial-ethnic groups have significantly greater racial heterogeneity in their friendship networks than whites do (Moody, 2001). Surprisingly, although white, black, and Asian youth belong to friendship networks with similar levels of peer popularity, Hispanic youth tend to have friends nominated significantly more often by others in the school network than white youth do, though the difference is quite small (5.40 versus 5.26). Focusing on behavioral characteristics reveals that, compared to white youth, black and Hispanic adolescents belong to networks whose members report significantly lower grade point averages (2.61 and 2.63 versus 2.89), and that Asian youth have friends reporting higher grade point averages (3.05 versus 2.89). Examining friends’ involvement in deviance reveals some interesting patterns. Consistent with earlier research, which finds that minority youth are less involved in minor deviance (MacLeod, 1995; Sue and Wagner, 1973), descriptive findings indicate that compared to white youth, blacks and Asians have friends participating in significantly lower levels of deviance (1.66 and 1.62 versus 2.35). Hispanic youth do not differ from whites in the prevalence of deviant behavior in their networks. Perhaps of most interest for this study, black adolescents have friendship networks reporting significantly higher levels of violence than their white counterparts (0.84 versus 0.78), and Asian adolescents have networks reporting the lowest levels (0.63 versus 0.78).

Although this preliminary descriptive evidence, in general, supports the idea that differences in characteristics of friendship networks might account for differential involvement in violence, it is not definitive. Minorities also differ from whites on many of the background variables that serve as controls in this analysis. For instance, black and Hispanic youth are more likely to be recent movers, to belong to a lower socioeconomic group, and to have lower grade point averages than white youth. These significant differences underscore the importance of controlling for these variables in the multivariate analyses that follow.

Table 2 presents the results of the multivariate logistic regression analyses. The first model serves as a baseline model and includes as regressors the dummy variables distinguishing racial-ethnic groups from whites (the reference category) and the control variables.9 The coefficients

9. In supplementary analyses, we considered whether school racial composition affected violence, interacted with individual race, interacted with peer network

2 HAYNIE F.DOC 10/31/2006 8:18:55 PM

RACE, NETWORKS, AND DELINQUENCY 789

characteristics, or determined the degree to which race influenced the effect of network characteristics on adolescent violence (essentially a three-way interaction between racial composition, individual race, and individual network character-istics). Surprisingly, no analyses provided any evidence that school racial composition influenced the pattern of findings.

Tab

le 1

. M

eans

and

Sta

ndar

d D

evia

tions

of I

nclu

ded

Var

iabl

es (

n =

12,9

12)

Var

iabl

e P

oole

d W

hite

B

lack

H

ispa

nic

Asi

an

M

ean

SD

Mea

n SD

M

ean

SD

Mea

n SD

M

ean

SD

Vio

lenc

e .2

2 .4

1 .2

0 .4

0 .2

5* .4

3 .2

7* .4

5 .1

5*.3

6 R

acea

Bla

ck

.15

.36

His

pani

c .1

2 .3

3

A

sian

.0

4 .1

9

C

ontr

ol v

aria

bles

M

ale

.50

.5

0 .4

9 .5

0 .4

8 .5

0 .5

1 .5

0 .5

3 .2

0 A

ge

15.0

6 1.

64

14.9

9 1.

61

15.2

3 1.

66

15.1

6 1.

75

15.2

7 1.

67

Rec

ent m

over

.2

1 .4

0 .1

8 .3

9 .2

6* .4

4 .2

7* .4

5 .2

2 .4

2 T

wo-

pare

nt fa

mily

.7

2 .4

5 .7

8 .4

1 .4

3* .4

9 .7

0* .4

6 .8

1 .3

9 F

amily

SE

S 5.

88

2.46

6.

21

2.35

5.

34*

2.51

4.

61*

2.41

6.

32

2.68

G

PA

2.

66

.83

2.73

.8

4 2.

47*

.71

2.41

* .7

9 2.

96*

.77

Uns

truc

ture

d so

cial

izin

g 1.

99

1.01

2.

04

.98

1.92

* 1.

07

1.83

* 1.

08

1.79

* .9

8 T

empe

r .3

0 .4

6 .2

9 .4

5 .3

1 .4

6 .3

4* .4

7 .2

1* .4

1 P

rior

dev

ianc

e 2.

30

2.95

2.

49

3.12

1.

63*

2.26

2.

33

2.70

1.

34*

2.45

P

rior

vio

lenc

e .8

4 .9

5 .8

2 .9

6 .9

5* .9

5 .8

9 .8

7 .5

9* .8

0 O

ut-o

f-sc

hool

nom

inat

ions

1.

27

1.64

1.

26

1.66

1.

29

1.64

1.

29

1.46

1.

31

1.79

Su

burb

an s

choo

l .5

8 .4

9 .6

2 .4

9 .5

7 .5

0 .3

6* .4

8 .7

0 .4

6 R

ural

sch

ool

.25

.44

.18

.38

.30*

.46

.60*

.49

.29

.46

Fri

ends

hip

Net

wor

k

R

acia

l het

erog

enei

ty

.27

.19

.23

.17

.30*

.18

.46*

.14

.39*

.19

Fri

ends

’ pop

ular

ity

5.29

1.

84

5.26

1.

83

5.33

1.

93

5.40

* 1.

74

5.32

1.

88

Fri

ends

’ GPA

2.

82

.44

2.89

.4

3 2.

61*

.40

2.63

* .4

0 3.

05*

.45

Fri

ends

’ dev

ianc

e 2.

20

1.86

2.

35

1.96

1.

66*

1.41

2.

18

1.69

1.

62*

1.62

F

rien

ds’ v

iole

nce

.79

.51

.78

.50

.84*

.54

.79

.50

.63*

.53

a W

hite

is t

he r

efer

ence

cat

egor

y.

* Sig

nifi

cant

dif

fere

nce

from

the

whi

te m

ean

at p

< .0

5.

2 HAYNIE F.DOC 10/31/2006 8:18:55 PM

790 HAYNIE AND PAYNE

for black and Hispanic adolescents are positive and statistically significant; net of the effects of the background control variables, the odds that black youth will engage in violence are 27 percent greater than for white youth (1.27 = e.24) and the odds that Hispanic youth will engage in violence are 34 percent greater than for white youth (1.34 = e.29). Asian youth do not, once background controls are introduced, differ significantly from whites in their odds of participating in violence.

In addition to these significant racial differences, model 1 shows that, consistent with past research, many of the background controls are significantly associated with the odds of participating in violence. For instance, the odds that males will participate in violence are more than two times those for females (2.08 = e.73). Older adolescents, adolescents living in two-parent families, and those with higher grade point averages have reduced risks of violence. Those spending increased time with peers, who have a bad temper, who are more deviant and violent themselves, and who send more out-of-school nominations have greater risks of violence. Overall, the finding of race-ethnic differences in violence in the face of these robust controls is striking.

To evaluate our first question, whether characteristics of friendship networks can account for racial-ethnic differences in violence, we turn to model 2, which adds the five measures of adolescent friendship networks to the baseline model. Results indicate that, on average, adolescents with greater racial heterogeneity in their friendship networks have greater odds of participating in violence, as do those exposed to deviant and violent behavior. In contrast, adolescents exposed to more academically oriented peer networks (as indexed by grade point averages) have lower odds of violence. Friends’ popularity is not associated with violence. More important, controlling for the five network characteristics reduces the coefficient for black and Hispanic youth to nonsignificance. In substantive terms, incorporating characteristics of networks accounts for 29 percent of the apparent effect of black racial group identity on violence (.29 = (.24 – .17) ÷ .24) and 52 percent of the apparent effect of Hispanic racial group identity on violence (.52 = (.29 - .14) ÷ .29). Overall, these results support our mediation hypothesis, which predicted that incorporation of friendship network characteristics would account for racial-ethnic differences in violent behavior.10

10. We also ran supplementary models where we entered each network characteristic separately into the baseline model. Results from these series of models indicated that several network characteristics individually could account for the race-violence association (that is, drive the race coefficients to nonsignificance) including racial heterogeneity of friendship networks, peer GPA, and peer involvement in violence.

2 HAYNIE F.DOC 10/31/2006 8:18:55 PM

RACE, NETWORKS, AND DELINQUENCY 791

Table 2. Survey-Corrected Logistic Regression of Adolescent Violence on Race and Friendship Networks (n = 12,912)

Variable Model 1 Model 2 b (SE) b (SE) Intercept -1.29*** (.36) -1.45*** (.41) Racea Black .24** (.09) .17 (.11) Hispanic .29** (.11) .14 (.12) Asian -.11 (.21) -.18 (.22)

Control variables Male .73*** (.07) .74*** (.07) Age -.07*** (.02) -.06** (.02) Recent mover -.02 (.07) -.04 (.07) Two-parent family -.13* (.07) -.12 (.07) Family SES -.03 (.02) -.02 (.02) GPA -.16*** (.05) -.09 (.05) Unstructured socializing .08** (.03) .08** (.03) Temper .34*** (.06) .31*** (.06) Prior deviance .08*** (.01) .06*** (.01) Prior violence .49*** (.03) .48*** (.03) Out-of-school nominations .13*** (.02) .13*** (.02) Suburban school .17 (.09) .14 (.09) Rural school .19 (.12) .12 (.11)

Friendship network Racial heterogeneity .67** (.22) Friends’ popularity .02 (.02) Friends’ GPA -.25** (.09) Friends’ deviance .04* (.02) Friends’ violence .19** (.06)

* p < .05 ** p < .01 *** p < .001 (two-tailed tests) a White is the reference category.

The effect of friendship networks on adolescent violence may not affect groups in the same way. To evaluate our second question, whether the effect of network characteristics on violence depends on race-ethnicity, we turn to interaction effects. The regression models presented in table 3 evaluate this possibility by adding to the baseline model of table 2 three product terms representing the interaction between racial-ethnic categories and a particular network characteristic. For ease of interpretation, each model focuses on a separate characteristic. Model 1 first examines network racial heterogeneity and asks whether the effect of racial heterogeneity on violence varies by racial-ethnic identity. As we anticipated with our integration hypothesis, we find that among black

2 HAYNIE F.DOC 10/31/2006 8:18:55 PM

792 HAYNIE AND PAYNE

adolescents increasing racial heterogeneity reduces the odds of violence. To better understand the nature of this finding, we graph the effect of racial heterogeneity on violence for blacks and whites.11 Figure 1 illustrates this interaction. It also indicates that among blacks, increasing racial heterogeneity of friends reduces the predicted odds of violence, whereas among whites the slope is diminished (almost flat), indicating a slight positive effect of racial heterogeneity.12 When adolescents are exposed to average racial heterogeneity (mean = .27), the predicted odds of African American youth engaging in violence are 1.5, compared to the white youth odds of 1.2. However, when racial heterogeneity is greater (1 standard deviation above the mean or equal to .46) white youths’ odds are slightly greater.

The next model examines peer popularity. We find that, among blacks, location in a network where friends are more often nominated by others in the school network, as with racial heterogeneity, reduces the odds of violence. Again, to illustrate this finding, we graph the effect of peer popularity on violence for blacks and whites. Figure 2 presents these results and indicates that, among white youth, increasing network popularity is not associated with increased odds of engaging in violence (indicated by the relatively flat slope for whites). Among black youth, increasing network popularity is associated with a reduction in the odds of engaging in violence. In fact, when friends’ popularity is two standard deviations above the mean (8.97), black youths’ odds of engaging in violence are lower than white youths’. Overall, these results provide some support for our integration hypothesis, which predicted that among historically disadvantaged minorities, greater integration into school networks would be associated with reduced risks of violence.

The next set of models (see table 3) addresses the compensation and peer-rejection hypotheses, which offer contrary predictions about the nature of peer behavior similarity for racial-ethnic minorities. Overall, these models offer no evidence that minorities, compared to whites, are influenced to a greater or lesser degree by their friends’ behavior. Instead, results indicate that location in friendship networks where friends report higher grade point averages is associated with a reduction in the odds of violence, and location in a more deviant or violent network is associated with an increase, regardless of race. These findings, therefore, do not support either the compensation or peer-rejection hypotheses.

11. We focus only on the comparison between blacks and whites because the effect of network racial heterogeneiety is not significantly different between Hispanic, Asians, and whites.

12. The graphs in figures 1 and 2 are limited to a description of the predicted odds of violence when the network characteristic ranges between the 10th and 90th percentiles for both racial groups.

2 HAYNIE F.DOC 10/31/2006 8:18:55 PM

RACE, NETWORKS, AND DELINQUENCY 793

Tab

le 3

. Su

rvey

-Cor

rect

ed L

ogis

tic R

egre

ssio

n of

Ado

lesc

ent V

iole

nce

on R

ace

and

Fri

ends

hip

Net

wor

ks (

n =

12,

912)

V

aria

ble

Mod

el 1

M

odel

2

Mod

el 3

M

odel

4

Mod

el 5

Het

erog

enei

tyP

opul

arit

y G

PA

D

evia

nce

Vio

lenc

e

b (S

E)

b (S

E)

b (S

E)

b (S

E)

b (S

E)

Inte

rcep

t -1

.69**

(.36

) -1

.45**

(.38

) -.

45(.

43)

-1.0

9**(.

37)

-1.9

6**(.

35)

Rac

e a

B

lack

.5

8**(.

15)

.73**

(.19

) .2

4(.

52)

.24

(.14

) .4

0*(.

17)

His

pani

c

.09

(.31

) .0

8 (.

26)

.17

(.56

) .1

6 (.

17)

.18

(.22

) A

sian

-.

03

(.35

) -.

44

(.42

) -.

14(.

58)

-.39

(.

26)

-.18

(.

34)

Fri

ends

hip

netw

ork

Rac

ial h

eter

ogen

eity

1.

04**

(.29

)

F

rien

ds’ p

opul

arit

y

.03

(.02

)

Fri

ends

’ GP

A

-.34

**(.

11)

Fri

ends

’ dev

ianc

e

.06**

(.02

)

Fri

ends

’ vio

lenc

e

.2

9**(.

08)

Inte

ract

ions

Bla

ck *

net

wor

k ch

ar.

-1.3

0**(.

44)

-.09

**(.

03)

-.03

(.20

) .0

2 (.

05)

-.19

(.

16)

His

pani

c *

netw

ork

char

. -.

04

(.63

) .0

4 (.

04)

.02

(.21

) .0

6 (.

05)

.14

(.19

)

Asi

an *

net

wor

k ch

ar.

-.62

(.

82)

.06

(.06

) .0

2(.

33)

.15

(.08

) .1

2 (.

37)

* p <

.05,

** p

< .0

1, **

* p <

.001

(tw

o-ta

iled

test

s)

Not

e: A

ll m

odel

s co

ntai

n co

ntro

l var

iabl

es li

sted

in ta

ble

2.

a W

hite

is t

he r

efer

ence

cat

egor

y.

2 HAYNIE F.DOC 10/31/2006 8:18:55 PM

794 HAYNIE AND PAYNE

Fig

ure

1. T

he E

ffec

t of N

etw

ork

Rac

ial H

eter

ogen

eity

on

Pre

dict

ed O

dds

of V

iole

nce

0

0.51

1.52

2.5

00.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

0.45

0.5

0.55

Net

wor

k R

acia

l Het

erog

enei

ty

Predicted Odds of Violence

Blacks

Whites

2 HAYNIE F.DOC 10/31/2006 8:18:55 PM

RACE, NETWORKS, AND DELINQUENCY 795

Fig

ure

2. T

he E

ffec

t of F

rien

ds’ P

opul

arit

y on

the

Pre

dict

ed O

dds

of V

iole

nce

0

0.2

0.4

0.6

0.81

1.2

1.4

33.75

4.5

5.25

66.75

7.5

8.25

Net

wor

k Po

pula

rity

Predicted Odds of Violence

Blacks

Whites

2 HAYNIE F.DOC 10/31/2006 8:18:55 PM

796 HAYNIE AND PAYNE

DISCUSSION

In this study, we expand on research that has sought to explain the racial gap in violence among adolescents by highlighting the role their friendship networks play. Although a large body of research and theory stresses the importance of peer associations for involvement in delinquency (Agnew, 1991; Akers, 1985; Elliott, Huizinga, and Ageton, 1985; Elliott and Menard, 1996; Short, 1957; Warr, 1993a, 1993b, 1996, 2002), research has neglected to examine how characteristics of these networks might help in understanding the relationship between race-ethnicity and violence. To address this shortcoming, we looked at the extent to which characteristics of friendship networks among adolescents explain violence among four racial groups—whites, blacks, Hispanics, and Asians.

Our results illustrate that, first, black and Hispanic—but not Asian—adolescents engage in significantly more violence than their white counterparts, after individual and family characteristics are controlled. This pattern is consistent with prior research (Hawkins et al., 2000; McNulty and Bellair, 2003) and suggests that addressing racial-ethnic differences in violence remains an important issue. More important, once we take into account the structural and behavioral characteristics of adolescents’ peer networks, these differences in violence are substantially reduced and the race-ethnicity coefficients drop to nonsignificance. This finding supports our mediation hypothesis, which predicted that race and ethnic differences in violence would be explained by the characteristics of adolescent friendship networks. Overall, this suggests that the patterning and structuring of adolescent friendship networks perpetuate racial-ethnic differences in violence.

We also find partial support for our integration hypothesis, which predicted that, as racial heterogeneity and friends’ popularity within peer networks increased, violence for blacks and Hispanics would decline compared to violence for whites. Figures 1 and 2 illustrate that this hypothesis is borne out for black but not for Hispanic adolescents. Among black youth, we argued that increasing racial heterogeneity and exposure to more popular friends increases feelings of integration and provides access to greater and more positive sources of social capital, which in turn inhibit violence. One possible reason we did not find a similar effect for Hispanic youth is that Hispanic youth are not, on average, as disconnected as black youth from others in the school network. Our descriptive findings provide some support for this interpretation: the average Hispanic adolescent has a network in which his or her friends’ popularity is higher than that of other racial and ethnic groups. On the other hand, the practice

2 HAYNIE F.DOC 10/31/2006 8:18:55 PM

RACE, NETWORKS, AND DELINQUENCY 797

of grouping Hispanic youth together is likely to mask considerable variation in groups resulting from differences in national origin. Overall, we find that increased racial heterogeneity is associated with reduced involvement in violence among blacks. However, results from the descriptive and mediation analyses suggest that racial friendship networks are largely homogenous. Ultimately, these findings suggest that such racial homogenieity may be especially detrimental for black youth.

The finding that, among whites, increasing network racial heterogeneity does not have a similar protective effect needs more explanation. The small positive effect of increasing racial heterogeneity on violence among white youth is consistent with earlier predictions by Shaw and McKay (1942), who argued that increasing racial heterogeneity in neighborhoods breaks down informal social controls that operate to keep crime in check. This interpretation, however, is not consistent with our findings for black adolescents, for whom increased network heterogeneity decreases violence. Among white adolescents, increasing racial diversity of friends may be highly correlated with lower socioeconomic status among friends. It is social class differences, rather than racial heterogeneity, that then account for higher involvement in violence.

On the other hand, this finding may simply reflect the availability and type of social capital available to whites in more racially diverse friendship networks. As we discussed earlier, black and Hispanic youth are likely to have access to lower levels and different forms of social capital than their white and Asian counterparts (see also Steffensmeier and Ulmer, 2006). Therefore, as the proportion of historically disadvantaged minorities in white friendship networks increases, white adolescent access to social capital may be constrained. Although this conclusion is certainly controversial and clearly warrants further research, it is also important to recognize that increasing racial heterogeneity has only a slight positive effect among white adolescents (the slope is much diminished compared to the slope for black adolescents). The finding may also simply reflect a period of growing pains, during which a movement toward reaping the positive benefits of increasing racial heterogeneity is occurring, but more quickly for some groups than others. In other words, among whites, as friendship networks become more heterogeneous over time (and thus more normative), we could expect to see increasing positive benefits accrue to whites. Future research should continue to evaluate this possibility.

Although a consistent body of research indicates that exposure to greater racial-ethnic heterogeneity is associated with more positive race relations and educational outcomes (Powers and Ellison, 1995; Sigelman and Welch, 1993), little research has examined other behavioral outcomes associated with having more diverse friends. Future research should

2 HAYNIE F.DOC 10/31/2006 8:18:55 PM

798 HAYNIE AND PAYNE

continue to explore whether and why racial heterogeneity of friendships operates differently for black and white youth. It should also continue to examine other behavioral outcomes associated with greater racial heterogeneity of networks.

Finally, our results do not support the compensation and peer-rejection hypotheses. These predicted racial differences in the importance of friendship network characteristics for violence. Instead, our results indicate that each of the racial-ethnic groups examined are influenced by friends’ grade point average as well as their involvement in deviant or violent behavior. This suggests that minority friendships are no more or less influential in shaping behavior than white friendships are.13

This paper has important implications for the study of youth violence. Racial differences in youth violence continues to be an important issue for policy makers, for two reasons. One, self-report data illustrate that nearly twice as many black as white youth continue violent offending into early adulthood (Elliott, 1994). Two, official data demonstrate that black youths are overrepresented in juvenile arrests for violent crimes (Snyder, 2005). Our results thus underscore the importance of taking race-ethnicity into account in building theories of adolescent development and in understanding how social networks influence important outcomes such as involvement in violence.

Our study found no evidence that school racial composition or urbanicity influenced the pattern of findings reported. Future research, however, should continue to explore school characteristics that are likely to make race-ethnicity more or less salient for adolescents. For example, school policies such as tracking and the opportunity to become involved in extracurricular activities are likely to shape racial relationships in schools as well as dynamics of friendship networks (Hallinan and Williams, 1987; Moody, 2001). In addition, future research should explore whether the pattern of findings reported in our study applies to other behaviors that show evidence of race-ethnicity differences including school achievement, drug and alcohol use, and victimization experiences.

Finally, this study was hindered by an important limitation. Due to the nature of the Add Health study design, we were unable to examine properties of out-of-school friendships. This is an important extension that future research might undertake, especially given that characteristics of out-of-school friendship networks might vary across different racial groups of adolescents. The school environment is clearly important for the

13. These findings are particularly interesting for what they imply for Ogbu’s (1990) theory of oppositional culture. An oppositional culture argument could imply that peer’s academic achievements are less important for black compared to white youth, yet our findings indicate no support for this thesis.

2 HAYNIE F.DOC 10/31/2006 8:18:55 PM

RACE, NETWORKS, AND DELINQUENCY 799

development of peer associations. Other environments, however, such as the community, are also important. For example, if youth are bused to school in a different community but maintain close ties to peers in their home neighborhood, these out-of-school relationships are likely to have consequences—among them, the odds of engaging in behaviors such as violence.

Despite this limitation, our study makes an important contribution toward identifying the reasons for racial disparities in adolescent violence. We expand on the current body of research that seeks to explain these differences by incorporating measures of friendship network characteristics into a framework that merges differential association theory and social capital theory. Moreover, by using network data on actual friends within networks, we overcome the problems of projection and same-source bias inherent in most earlier studies of adolescent delinquency. We further extend research that explores the race-violence relationship by moving beyond a strict two-group dichotomy of black-white differences and examining differences among Hispanic and Asian American youths as well. Ultimately, our results suggest that racial-ethnic differences in youth violence are significantly related to race differences in friendship networks, and as such, it is essential that we continue to give attention to this particular social context.

REFERENCES

Adler, Patricia A., and Peter Adler. 1998. Peer Power: Preadolescent Culture and Identity. New Brunswick, NJ: Rutgers University Press.

Akers, Ronald L. 1985. Deviant Behavior: A Social Learning Approach, 3rd ed. Belmont, CA: Wadsworth Publishing.

Akers, Ronald L. 1998. Social Learning and Social Structure: A General Theory of Crime and Deviance. Boston, MA: Northeastern University Press.

Agnew, Robert. 1991. The interactive effects of peer variables on delinquency. Criminology 29:47–72.

Allison, Paul D. 2001. Missing Data. Thousand Oaks, CA: Sage Publications.

Anderson, Elijah. 1989. Sex codes and family life among poor inner-city youth. Annals of the American Academy of Political and Social Science 501:59–78.

Anderson, Elijah. 1990. Streetwise: Race, Class, and Change in an Urban Community. Chicago: University of Chicago Press.

2 HAYNIE F.DOC 10/31/2006 8:18:55 PM

800 HAYNIE AND PAYNE

Anderson, Elijah. 1999. Code of the Street: Decency, Violence, and the Moral Life of the Inner City. New York: W. W. Norton.

Baum, Katrina. 2005. Juvenile Victimization and Offending, 1993–2003. Washington, DC: U.S. Department of Justice, Office of Justice Programs, Bureau of Justice Statistics.

Bearman, Peter S., and James Moody. 2004. Suicide and friendships among American adolescents. American Journal of Public Health 94:89–95.

Bearman, Peter S., Jo Jones, and J. Richard Udry. 1997. The National Longitudinal Study of Adolescent Health: Research Design. Chapel Hill, NC: Carolina Population Center.

Bell, Robert R. 1981. Friendships of women and of men. Psychology of Women Quarterly 5:402–17.

Billy, John O.G., and J. Richard Udry. 1985. Patterns of adolescent friendship and effects on sexual behavior. Social Psychology Quarterly 48:27–41.

Brown, Bradford B. 1990. Peer groups and peer cultures. In At the Threshold: The Developing Adolescent, eds. S. Shirley Feldman and Glen R. Elliott. Cambridge, MA: Harvard University Press.

Castro, Felipe Gonzales, Gina R. Boyer, and Hector G. Blacazar. 2000. Healthy adjustment in Mexican American and other Hispanic adolescents. In Adolescent Diversity in Ethnic, Economic, and Cultural Contexts, eds. Raymond Montemayor, Gerald Adams, and Thomas P. Gullotta. Thousand Oaks, CA: Sage Publications.

Chantala, Kim, and Joyce Tabor. 1999. Strategies to perform a design-based analysis using the add health data. Chapel Hill: Carolina Publication Center, University of North Carolina. http://www.cpc.unc.edu/ projects/ addhealth/strategies.html.

Clark, M. L., and Marla Ayers. 1992. Friendship similarity during early adolescence: Gender and racial patterns. The Journal of Psychology 126(4): 393–405.

Cloward, Richard A., and Lloyd E. Ohlin. 1960. Delinquency and Opportunity: A Theory of Delinquent Gangs. Glencoe, IL: The Free Press.

Cohen, Albert K. 1955. Delinquent Boys: The Culture of the Gang. Glencoe, IL: The Free Press.

Coleman, James S. 1961. The Adolescent Society: The Social Life of the Teenager and Its Impact on Education. New York: Free Press of Glencoe.

2 HAYNIE F.DOC 10/31/2006 8:18:55 PM

RACE, NETWORKS, AND DELINQUENCY 801

Coleman, James S. 1988. Social capital in the creation of human capital. American Journal of Sociology 94:95–120.

Coleman, James S. 1990. Foundations of Social Theory. Cambridge, MA: Belknap Harvard.

Crosnoe, Robert, Shannon Cavanagh, and Glen H. Elder, Jr. 2003. Adolescent friendships as academic resources: The intersection of friendship, race, and school disadvantage. Sociological Perspectives 46:31–52.

Elliott, Delbert S. 1994. Serious violent offenders: Onset, developmental course, and termination. Criminology 32:1–21.

Elliott, Delbert S., and Scott Menard. 1996. Delinquent friends and delinquent behavior: Temporal and developmental patterns. In Delinquency and Crime: Current Theories, ed. David J. Hawkins. New York: Cambridge University Press.

Elliott, Delbert, Susan Ageton, and Rachelle Cantor. 1979. An integrated theoretical perspective on delinquent behavior. Journal of Research on Crime and Delinquency 16:3–27.

Elliott, Delbert, David Huizinga, and Susan Ageton. 1985. Explaining Delinquency and Drug Use. Beverly Hills, CA: Sage Publications.

Epstein, Joyce L. 1985. After the bus arrives: Resegregation in desegregated schools. Journal of Social Issues 41:23–43.

Felson, Richard B., Allen E. Liska, Scott J. South, and Thomas L. McNulty. 1994. The subculture of violence and delinquency: Individual vs. school context effects. Social Forces 73: 155–73.

Furstenberg, Frank. 1993. How families manage risk and opportunity in dangerous neighborhoods. In Sociology and the Public Agenda, ed. W. Wilson. Newbury Park, CA: Sage Publications.

Giordano, Peggy C. 2003. Relationships in adolescence. Annual Review of Sociology 29:257–81.

Giordano, Peggy C., Stephen A. Cernkovich, and Alfred DeMaris. 1993. The family and peer relations of black adolescents. Journal of Marriage and the Family 55:277–87.

Glaser, Daniel. 1956. Criminality theories and behavioral images. American Journal of Sociology 61:433–44.

Hagan, John, and Holly Foster. 2001. Youth violence and the end of adolescence. American Sociological Review 66:874–99.

Hallinan, Maureen T., and Richard A. Williams. 1987. The stability of students’ interracial friendships. American Sociological Review 52:653–64.

2 HAYNIE F.DOC 10/31/2006 8:18:55 PM

802 HAYNIE AND PAYNE

Hallinan, Maureen T., and Richard A. Williams. 1989. Interracial friendship choices in secondary schools. American Sociological Review 54:76–78.

Hawkins, Darnell F., John H. Laub, Janet L. Lauritsen, and Lynn Cothern. 2000. Race, Ethnicity, and Serious and Violent Juvenile Offending. Washington, DC: U.S. Department of Justice, Office of Justice Programs, Office of Juvenile Justice and Delinquency Prevention.

Haynie, Dana L. 2001. Delinquent peers revisited: Does network structure matter? American Journal of Sociology 106:1013–57.

Haynie, Dana L. 2002. Friendship networks and adolescent delinquency: The relative nature of peer delinquency. Journal of Quantitative Criminology 18:99–134.

Haynie, Dana L., and D. Wayne Osgood. 2005. Reconsidering peers and delinquency: How do peers matter? Social Forces 84:1107–128.

Haynie, Dana L., Eric Silver, and Brent Teasdale. 2006. Neighborhood characteristics, peer influence, and adolescent violence. Journal of Quantitative Criminology 22(2): 147–69.

Heimer, Karen. 1997. Socioeconomic status, subcultural definitions, and violent delinquency. Social Forces 74:799–833.

Heimer, Karen, and Ross L. Matsueda. 1994. Role-taking, role commitment, and delinquency: A theory of differential social control. American Sociological Review 59:365–90.

Huizinga, David, Rolf Loeber, and Terrence P. Thornberry. 1994. Urban Delinquency and Substance Abuse: Initial Findings. Washington, DC: U.S. Department of Justice, Office of Justice Programs, Office of Juvenile Justice and Delinquency Prevention.

Iscoe, Ira, Martha Williams, and Jerry Harvey. 1964. Age, intelligence, and sex as variables in the conformity behavior of Negro and white children. Child Development 35:451–60.

Jussim, Lee, and D. Wayne Osgood. 1989. Influence and similarity among friends: An integrative model applied to incarcerated adolescents. Social Psychology Quarterly 52:98–112.

Kandel, Denise B. 1974. Interpersonal influences on adolescent illegal drug use. The Epidemiology of Drug Abuse, eds. E. Josephson and E. Carroll. Washington, DC: Winston.

LaFree, Gary. 1995. Race and crime trends in the United States, 1946–1990. In Ethnicity, Race, and Crime: Perspectives Across Time and

2 HAYNIE F.DOC 10/31/2006 8:18:55 PM

RACE, NETWORKS, AND DELINQUENCY 803

Place, ed. Darnell F. Hawkins. Albany: State University of New York Press.

MacLeod, Jay. 1995. Ain’t No Makin It: Aspirations and Attainment in a Low-Income Neighborhood. Boulder, CO: Westview Press.

Massey, Douglas, and Nancy Denton. 1993. American Apartheid: Segregation and the Making of the Underclass. Cambridge, MA: Harvard University Press.

Matsueda, Ross L., and Kathleen Anderson. 1998. The dynamics of delinquent peers and delinquent behavior. Criminology 36:269–308.

Matsueda, Ross L., and Karen Heimer. 1987. Race, family structure, and delinquency: A test of differential association and social control theories. American Sociological Review 52:826–40.

McCarthy, Bill, and John Hagan. 2001. When crime pays: Capital, competence, and criminal success. Social Forces 79:1035–60.

McNulty, Thomas L. 2001. Assessing the race-violence relationship at the macro level: The assumption of racial invariance and the problem of restricted distributions. Criminology 39:301–24.

McNulty, Thomas L., and Paul E. Bellair. 2003. Explaining racial and ethnic differences in serious adolescent violent behavior. Criminology 41:701–30.

Moody, James. 2001. Race, school integration, and friendship segregation in America. American Journal of Sociology 107:679–716.

Ogbu, John U. 1990. Minority education in comparative perspective. The Journal of Negro Education 59:45–57.

Powers, Daniel A., and Christopher G. Ellison. 1995. Interracial contact and black racial attitudes: The contact hypothesis and selectivity bias. Social Forces 74:205–26.

Proctor, Bernadette D., and Joseph Dalaker. 2003. Poverty in the United States: 2002. U.S. Census Bureau, Current Population Reports. P60-222. Washington, DC: U.S. Government Printing Office.

Sampson, Robert J., Jeffrey D. Morenoff, and Stephen Raudenbush. 2005. Social anatomy of racial and ethnic disparities in violence. American Journal of Public Health 95:224–32.

Savin-Williams, R. C., and T. J. Berndt. 1990. Friendship and peer relations. In At the Threshold: The Developing Adolescent, eds. S. Shirley Feldman and Glen R. Elliott. Cambridge, MA: Harvard University Press.

Shaw, Clifford R., and Henry D. McKay. 1942. Juvenile Delinquency and Urban Areas. Chicago: University of Chicago Press.

2 HAYNIE F.DOC 10/31/2006 8:18:55 PM

804 HAYNIE AND PAYNE

Shibutani, Tamotsu. 1955. Reference groups as perspectives. American Journal of Sociology 60:562-69.

Short, James F., Jr. 1957. Differential association and delinquency. Social Problems 4:233–39.

Short, James F., Jr. 1997. Poverty, Ethnicity, and Violent Crime. Boulder, CO: Westview Press.

Sigelman, Lee, and Susan Welch. 1993. The contact hypothesis revisited: Black-white interaction and positive racial attitudes. Social Forces 71:781–95.

Silverstein, Barry, and Ronald Krate. 1975. Children of the Dark Ghetto: A Developmental Psychology. New York: Praeger.

Snyder, Howard N. 1999. Juvenile Arrests, 1998. Washington, DC: U.S. Department of Justice, Office of Justice Programs, Office of Juvenile Justice and Delinquency Prevention.

Snyder, Howard N. 2000. Juvenile Arrests, 1999. Washington, DC: U.S. Department of Justice, Office of Justice Programs, Office of Juvenile Justice and Delinquency Prevention.

Snyder, Howard N. 2005. Juvenile Arrests, 2003. Washington, DC: U.S. Department of Justice, Office of Justice Programs, Office of Juvenile Justice and Delinquency Prevention.

South, Scott, and Dana L. Haynie. 2004. Friendship networks of mobile adolescents. Social Forces 83:315–50.

Steffensmeier, Darrell, and Jeffery T. Ulmer. 2006. Black and white control of numbers gambling: A cultural assets–social capital view. American Sociological Review 71:1–37.

Sutherland, Edwin H., and Donald R. Cressey. 1955. Principles of Criminology, 5th ed. Philadelphia: J.B. Lippincott.

Sue, Stanley, and Nathanial N. Wagner, eds. 1973. Asian Americans: Psychological Perspectives. Palo Alto, CA: Science and Behavior Books.

Tannenbaum, Frank. 1938. Crime and Community. New York: Ginn.

Taylor, Ralph L. 1989. African American inner-city youth and the subculture of disengagement. Urban League Review 12:15–24.

Taylor, Ronald D., Leanne Jacobson, and Debra Roberts. 2000. Ecological correlates of the social and emotional adjustment of African American adolescents. In Adolescent Diversity in Ethnic, Economic, and Cultural Contexts, eds. Raymond Montemayor, Gerald R. Adams, and Thomas P. Gullotta. Thousand Oaks, CA: Sage Publications.

2 HAYNIE F.DOC 10/31/2006 8:18:55 PM

RACE, NETWORKS, AND DELINQUENCY 805

Thornberry, Terence P. 1987. Toward an interactional theory of delinquency. Criminology 25:863–92.

Thornberry, Terence P., Marvin D. Krohn, A. J. Lizotte, and D. Chard-Wierschem. 1993. The role of juvenile gangs in facilitating delinquent behavior. Journal of Research in Crime and Delinquency 30:55–87.

Thrasher, Frederick M. 1927. The Gang. Chicago: University of Chicago Press.

Warr, Mark. 1993a. Age, peers, and delinquency. Criminology 31:17–40.

Warr, Mark. 1993b. Parents, peers, and delinquency. Social Forces 72:247–64.

Warr, Mark. 1996. Organization and instigation in delinquent groups. Criminology 34:11–37.

Warr, Mark. 2002. Companions in Crime: The Social Aspects of Criminal Conduct. Cambridge: Cambridge University Press.

Williams, James Herbert, Charles D. Ayers, Robert D. Abbott, J. David Hawkins, and Richard F. Catalano. 1999. Racial differences in risk factors for delinquency and substance use among adolescents. Social Work Research 23:241–55. Dana L. Haynie is an associate professor of sociology at The Ohio State

University. Her current research examines friendship networks and delinquency, the connection between adolescent residential mobility and problem behaviors, and the developmental implications of exposure to violence in adolescence.

Danielle C. Payne is ABD in sociology at The Ohio State University. Her research interests include peers and delinquency, social capital, and behavioral variations in crime and delinquency across the urbanicity continuum.