Although much is known about the relationships between delinquency and friends’ behavior, only a few studies present detailed information on friendship characteristics among delinquent adolescents. Warr (1996) examined specific features of delinquent subgroups, such as group organization and the instigator role within groups, and determined that the structure of the group, not an individual’s attributes, affects which individual instigates delinquency. Results from his study also indicate that groups are more specialized in terms of delinquency involvement than individuals tend to be, so that most delinquent offenders belong to multiple groups, with each group specializing in a smaller range of offenses. This latter finding also highlights the multifaceted nature of peer groups; individuals in school settings can be members of many different friendship groups and face differing degrees of constraint depending on whether the behavior, norms, and values of the group coincide or diverge. This is consistent with Dunphy’s (1963) finding that most adolescents do not belong to a single, densely knit, isolated friendship group but instead are affiliated with many loosely bounded friendship groups with varying degrees of cohesion and permeability.
Although delinquency is largely a group behavior, there is evidence that some offenses are more likely to occur in groups than others. For instance, offenses including the use of alcohol and marijuana and vandalism are more likely to be committed in groups compared with offenses such as assault and shoplifting, which are among the offenses least likely to be committed in groups (Warr, 1996).
The nature of friendship relations in delinquent versus nondelinquent networks has also been developed in two influential studies. Giordano, Cernkovich, and Pugh (1986) found that various dimensions of friendship relations do not differ markedly between delinquent and nondelinquent adolescents. Both delinquent and nondelinquent youth report similar levels of attachment, intimacy, and contact with friends. Kandel and Davies (1991) also found few differences in the quality of friendship relationships among adolescents who did and did not use illicit drugs.
Despite original emphasis on the importance of exposure to definitions or attitudes favorable to law violation, prior research has consistently indicated that attitude transference is not the primary mechanism through which friends influence one another; instead, adolescents appear more influenced by the behaviors of friends than they are by friends’ attitudes toward crime (Warr & Stafford, 1991). Consistent with social learning explanations of peer influence, these findings suggest that imitation of friends’ behavior and direct reinforcement of behavior by friends are most important (Akers, 1985).
Even though the studies just described are important for considering the role of peer relationships for adolescent delinquency, they were not able to draw on detailed social network data to ask more varied questions about the role of friendship networks. To do this, work using the detailed friendship networks available in theAdd Health data has begun.
Recent work by Haynie illustrates some of the beginning questions that researchers interested in adolescent delinquency can address using network data available from Add Health. A popular issue in the field of criminology has been trying to understand whether adolescents select into delinquent peer groups on the basis of their own behavior (as Gottfredson & Hirschi, 1990, suggested), supporting the common adage that birds of a feather flock together. If this idea is true, then friendship networks should exhibit predominately delinquent or nondelinquent behavior. This, therefore, raises the question: Do adolescents have homogeneous networks in terms of the delinquency of their friends?
Using friendship network information available in the Add Health and a dichotomous measure of delinquency (1 = yes, adolescent engaged in some delinquency during the past year; 0 = no delinquency reported), Haynie (2002) found that adolescents are located in rather heterogeneous networks in terms of the display of delinquent behavior; that is, the most common pattern is for adolescents to have both delinquent and nondelinquent friends in their friendship networks. Specifically, she found that 56% of adolescents are in a mixed network, with both delinquent and nondelinquent peers; 28% are in an entirely delinquent network; and 16% are in an entirely nondelinquent network.
These findings suggest that peer networks are much more heterogeneous in terms of exposure to delinquent friends. Although there is some evidence that delinquents cluster together, most adolescents in schools have both delinquent and nondelinquent friends in their networks of close acquaintances. This is an important finding that is at odds with Gottfredson and Hirschi’s (1990) assertion that self-selection is entirely responsible for the peer–delinquency association, because the assumption is that there are clearly delineated delinquent or nondelinquent networks that adolescents choose to join. Instead, most adolescents are exposed to both delinquent and nondelinquent patterns, and the ratio of these patterns influences behavior. When friendship networks contain access to both delinquent and nondelinquent friends, the network may be less effective in providing clear behavioral guidelines, cohesive norms, and consistent values regarding behavioral expectations.
A second common question concerns whether peer delinquency influences subsequent behavior or instead results from selection processes or the tendency for adolescents to project their own behaviors onto the peers whom they think of as their friends. The Add Health data provide a unique opportunity to address this question, because, as discussed earlier, the methodological structure permits the careful definition of friendship networks.With this approach, results based on the Add Health data suggest that peer delinquency is associated with an adolescent’s subsequent delinquency, controlling for prior delinquency; however, the effect is much smaller than that estimated by prior research that did not incorporate a network method and perspectives (Haynie & Osgood, 2005). This finding suggests that relying on adolescents’ perceptions of friends’ behavior does introduce substantial same-source bias that inflates the correlation between friends’ and adolescents’ behavior.
Third, recent work has been able to ask whether network characteristics condition the strength of the peer– delinquency association. In addition to measures of network behaviors, Add Health data allow for assessment of the structure of peer networks and the location of an adolescent’s position within the friendship network. Three network characteristics in particular appear to shape the degree of influence operating in a friendship network: (1) the density of ties within the network indicating how cohesive the network is, (2) the centrality of the adolescent’s position in the network, and (3) the popularity of the adolescent within the network. Specifically, it is expected that peer delinquency will have a stronger influence on an adolescent’s behavior when the friendship network is very dense (i.e., the adolescent’s friends are friends with one another), when the adolescent is located in a central position (vs. a peripheral position at the edge of the network), or when the adolescent has high prestige in the network (i.e., when he or she is very popular and receives many friendship nominations from others in the school). Findings based on the Add Health data suggest that this is indeed the case. In particular, network density emerges as an important component of the peer–delinquency association, with very cohesive networks promoting greater influence than networks that are less cohesive (Haynie, 2001).
If delinquency is largely a group phenomenon, then what would we expect to find in regard to delinquent behavior for adolescents who are isolated from peers? This is the interesting question that Kreager (2004) tackled using Add Health data. Theories reviewed earlier in this research paper suggest competing hypotheses about these isolated individuals.According to socialization theories (differential association and social learning), isolated adolescents will have limited access to delinquent role models and, as a result, are expected to engage in low or no amount of delinquency. In contrast, social control theory would expect that the lack of attachment to friends would result in individuals who are more inclined to act on their delinquent impulses. Kreager’s results indicate that although isolation from peer friendships is a rare event (less than 5% of the sample were friendless), its relationship to delinquency is more nuanced than socialization or social control theories would predict. Isolates who do not report peer trouble have very low levels of delinquency; however, isolates who also report peer conflict are likely to report higher levels of delinquency. Therefore, the effect of peer isolation on delinquency depends on whether adolescents report peer conflict.
In addition to facilitating examination of the peer– delinquency association, the Add Health data allow researchers to examine network behaviors and network structures as important mediating variables in explaining outcomes of interest (e.g., delinquency). For instance, in the criminology literature, there has been a common finding that girls who experience pubertal development earlier than their peers are at increased risk of engaging in subsequent delinquent behavior. One reason for this increased risk is the differing peer networks in which more developed versus less developed females find themselves. In particular, research using the Add Health data has found that females who experience pubertal development earlier than their peers do have higher levels of delinquency 1 year later, but this is because these girls are at heightened risk of being involved in romantic relationships and because their friends are engaging in risky behaviors (Haynie, 2003). This suggests that peer networks serve as a mechanism that differently place certain groups of adolescents at heightened risk of problem behaviors.