Introduction

There has been a significant increase in the prevalence and costs of antisocial behavior (ASB) committed by American children and adolescents during the last half of the twentieth century, and the cost of youth ASB within the United States has been estimated to exceed one trillion dollars (Anderson 1999). Numerous reviews have examined how youth ASB develops over time; these authors have requested further studies to clarify the etiology and varying subtypes of ASB (e.g., Burke et al. 2002; Loeber et al. 2009). The scope and significant consequences of youth antisocial behavior lead one to ask: How do we define ASB and what causes persistent youth ASB?

Youth ASB has been classified into two subtypes by previous researchers and within the present study (cf. Frick et al. 1993): violent (overt ASB) or sneaky (covert ASB). The validity of this organization is based on a factor analysis of parent-reported disruptive behaviors (Achenbach et al. 1989) and other studies supporting the differentiation of these two categories (Dekovic 2003; Loeber and Schmaling 1985; McEachern and Snyder 2012). The covert subtype of ASB has been defined as disruptive behaviors that are not violent and are committed with the intention of not being observed by authority figures (e.g., stealing or vandalism; Frick et al. 1993; Snyder et al. 2006). On the other hand, overt ASB refers to behaviors that are violent and confrontational (Kazdin 1992; Patterson et al. 2005). Each subtype has been hypothesized to demonstrate a distinct developmental trajectory as well (Loeber and Hay 1997).

Variables that predict ASB have been studied extensively, and three well-established risk factors have been associated with the development of ASB in adolescence (Dodge et al. 2006). Previous researchers have found an association between low parental monitoring (PM) and the development of ASB (e.g., Bacchini et al. 2011; Patterson 2002; Patterson et al. 2005). Another risk factor linked to ASB is an adolescent’s affiliation with deviant peers (e.g., Burnette et al. 2012; Dishion et al. 1994, 1999; Granic and Dishion 2003). A third risk factor found to predict adolescent ASB is adolescents’ exposure to violence within their community (e.g., Barr et al. 2012; Farrell and Bruce 1997; Gorman-Smith et al. 2004).

Although there have been many studies of how each of these risk factors predict the development of adolescent ASB (Dodge et al. 2006; Loeber et al. 2009), few have examined how these risk factors are interrelated and develop when considering multiple outcomes. Moreover, only a small number of researchers have examined how both interpersonal risk factors (e.g., poor PM or deviant peer relationships) and community risk factors interact to predict ASB (Bacchini et al. 2011; Barr et al. 2012; Simons et al. 2005). Instead, most previous work in this area has focused on either how interpersonal variables predict ASB or how community variables predict ASB. Additionally, no study to date has demonstrated how these risk factors of adolescent ASB develop over time with respect to different subtypes of ASB. We attempted to bridge this gap through the investigation of both interpersonal and community contextual risk factors of ASB using a longitudinal data set. Furthermore, we utilized two leading theoretical frameworks to find a model that best explains the relationships among these risk factors. Other theoretical approaches not used in the present study can be reviewed elsewhere (e.g., Burke et al. 2002; Dodge et al. 2006; Farrington 2005; Frick et al. 1993; Krohn and Thornberry 2001; Loeber et al. 2009).

The first guiding framework is the social-interactional model, also known as the coercive family process model, which explains youth antisocial behavior as a consequence of early established (and maintained) parent–child relationships (Compton et al. 2003; Patterson 2002). Patterson (2002) hypothesized that youth ASB is caused by youths’ relations with their parents; aspects of early parent–child dynamics subsequently influence other relationships and risk factors in youths’ lives.

Another well established framework, social contextual theory, presumes that the impact of major developmental influences, such as parenting practices and peers, is influenced by characteristics of the communities in which youth reside (Tolan et al. 2003). Bronfenbrenner’s (1979) theory of nested ecological or contextual niches has been used to describe human development and understand different childhood problem behaviors (Tolan et al. 2003). This theory suggests that when investigating behavior within an ecological context, researchers need to examine both the immediate influences and broader factors. Moreover, individuals are embedded within several layers of contexts, each providing protection against or acting as a risk factor for the development of problem behavior (Bronfenbrenner 1979). Examining the relationship between risk factors utilizing a social-contextual model can provide an explanation of how adolescent ASB develops within multiple interrelated contexts.

To add to knowledge in this area, we investigated a model representing how parenting, peer, and community risk factors predict adolescent ASB. This study advances the literature in several important ways. First, the risk factors we selected addressed multiple settings. Previous studies generally have focused on only one major context (e.g., the family, the community, or peer relationships). Second, in response to the lack of causal studies examining the development of adolescent ASB, we used longitudinal data derived from the National Longitudinal Study of Adolescent Health (Add Health; Udry 1998). Finally, the present study utilized multiple informants across two waves of longitudinal data to investigate the direct and interaction relations between these risk factors and how they influenced the development of antisocial behavior with an adolescent sample.

Interpersonal Risk Factors

Aspects of parenting have been correlated with the development of ASB among youth (Forehand et al. 1997). More specifically, several studies demonstrated how poor parenting (i.e., poor supervision, inconsistent and harsh discipline) leads to ASB (e.g., Burnette et al. 2012; Dishion et al. 1991; Graber et al. 2006; Griffin et al. 2000; Loeber and Hay 1997; Patterson and Stouthamer-Loeber 1984). Although there has been considerable evidence linking poor parenting with future youth ASB (consistent with the social interactional model), PM in particular becomes a crucial predictor of ASB as children move into adolescence and are granted more autonomy (Bacchini et al. 2011; Dishion and McMahon 1998).

Another powerful risk factor for adolescent ASB is the presence of delinquent friends (Dishion et al. 1994, 1997; Poulin et al. 1999). One explanation of this relationship could be related to the modeling effect derived from deviant peer affiliations (Bandura 1973; Salzinger et al. 2006). Thus, externalizing teens who are grouped together will influence each other through modeling of antisocial acts. Also, deleterious peer influences may vary as a function of gender, and may be more indirect for females in the form of relational aggression (Keenan et al. 2010). These explanations of this link support the need to investigate adolescents’ ecological context.

Notably, parenting behaviors and peer affiliations interact in the development of antisocial behavior. As adolescents age, PM has been found to wane in significance relative to the increasing influence of peers (Brendgen et al. 2000; Brown et al. 1993). However, studies have suggested that PM continues to predict ASB directly (Dishion and McMahon 1998) and indirectly by mediating peer influences (Henry et al. 2001).

Community Risk Factors

Exposure to community violence (CV) is another significant risk factor for antisocial behavior, as evidenced by increased aggression and desensitization to violent acts (e.g., Ng-Mak et al. 2004). Extensive evidence supports the link between this risk factor and ASB (Barr et al. 2012; Farrell and Bruce 1997; Farrell and Sullivan 2004; Gorman-Smith et al. 2004; Leventhal and Brooks-Gunn 2000; Shahinfar et al. 2001; Spano et al. 2008, 2009).

Aside from CV’s robust direct relations with ASB, it appears to be interrelated with other risk factors (i.e., interpersonal factors are hypothesized to be embedded within a broader community risk construct). For example, Richards and colleagues (Richards et al. 2004) found that the more time a youth had spent in unmonitored and unstructured contexts with his or her peers, the more likely the youth had been exposed to CV.

Parenting practices have been demonstrated to be related to CV as well. Mazefsky and Farrell (2005) found that parenting practices (i.e., supervision and discipline) mediated the relationship between witnessing violence and later aggressive behavior within a rural population. In contrast, Gorman-Smith et al. (2004) found in their longitudinal study that poor parenting and a strained parent–child relationship were linked to a higher incidence of youth violence and CV.

Parenting practices attenuate the effects of being exposed to CV when predicting ASB. Similar to findings by Mazefsky and Farrell (2005), Miller et al. (1999) not only demonstrated longitudinal relations between youth CV and future ASB, but that the severity of parent–child conflict moderated the relationship between these two variables. These investigators concluded that youths who were exposed to high violence and experienced poor parent–child relations were at greatest risk for antisocial behavior outcomes, whereas youth exposed to high levels of violence and experiencing positive parent–child relations were not at risk for future ASB. Additionally, Barr et al. (2012) found that adolescent witnesses of CV who experienced low family cohesion were approximately twice as likely to behave delinquently in the future than adolescents who had not witnessed violence and experienced family cohesion.

Multiple Contextual Predictors of Antisocial Behavior

Because adolescents are greatly influenced by people and situations outside of their families, it is important to examine adolescents within multiple contexts (e.g., school, community, and peer systems) in addition to the youths’ individual characteristics (e.g., temperament or attribution style) to delineate a model that accurately predicts ASB. Studies that have assessed multiple contexts have predominantly looked at community structure (e.g., concentrated poverty, neighborhood disadvantage, employment rates) or community social organization processes (e.g., collective efficacy; Simons et al. 2005) as predictors of ASB. For example, Rankin and Quane (2002) suggested that ASB seems to have a stronger relation to collective efficacy than to social structural variables, such as residing in a disadvantaged neighborhood.

Although these findings support the social contextual model of the development of ASB, causal inferences require longitudinal data. Tolan et al. (2003) conducted a multi-wave longitudinal study to determine how neighborhood structural characteristics, community collective efficacy, peer relationships, and parenting practices predict youth violence. The investigators demonstrated that interpersonal risk factors mediated the relationship between neighborhood structural characteristics and later violent behavior.

Both interpersonal and community risk factors have been shown to be interrelated when predicting more wide-ranging outcomes such as antisocial behavior rather than violent behavior. Chung and Steinberg (2006) examined how community risk factors (i.e., social organizational and collective efficacy) related to peer and parenting influences in the prediction of delinquent behavior. Utilizing cross-sectional data, these investigators demonstrated that weak neighborhood social organization was indirectly related to delinquency through its association with deviant peer affiliations and parenting practices. The authors stressed that models predicting delinquency are often oversimplified due to the investigation of single contexts. Instead, they suggested that parent and peer influences combined as a joint risk factor. Echoing previous findings from Tolan et al. (2003), Chung and Steinberg (2006) suggested that examining a more specific community factor (e.g., exposure to violence) may provide a more fruitful investigation of interrelations of risk factors and a stronger predictive model of ASB.

Study Overview and Hypotheses

We investigated the relationships among adolescents’ exposure to CV, PM, presence of adolescents’ deviant friends, and adolescents’ commission of antisocial acts. The present study used archived longitudinal data from Add Health (Udry 1998). This stratified national sample of 7th–12th grade students was collected in multiple waves and used multiple informants across different settings. We used Add Health data to predict how adolescent covert and overt antisocial behavior develops within the social context of interpersonal and community risk factors. We conducted eight hierarchical regression analyses (i.e., four models using data from boys and four models using data from girls) to assess relations between the predictor variables and both ASB outcomes. For each gender, two initial models utilized cross-sectional data, and two subsequent models using longitudinal data. Outcomes were divided by gender to assess any differences in the development of ASB.

We expected that PM would directly predict both overt and covert ASB. Specifically, the first research hypothesis (H1) was that poor PM measured at Time 1 would be related to a higher incidence of ASB at both wave 1 and wave 2. Thus, parents’ monitoring was expected to be a strong protective influence (by itself) in the development of both adolescent covert and overt ASB.

For the second hypothesis (H2), we predicted that PM would moderate the relations between the other two predictors of the present study (i.e., deviant peer affiliations and CV) and each subtype of ASB. This hypothesis was based on prior data suggesting that parent–child relations can act as an intervening variable in this regard (e.g., Rankin and Quane 2002; Tolan et al. 2003). Thus, PM was expected to buffer the deleterious effects of access to unsafe environments and deviant peers and mitigate the scope of ASB. We also posited direct relations between both adolescents’ deviant peer affiliations and CV exposure with the development of both adolescent covert and overt ASB.

Method

Data and Participants

We used a subset of the publicly released data (in-home parent surveys, in-home adolescent interviews, and in-school questionnaires) from the Add Health study (Udry 1998). The Add Health study assessed a nationally stratified representative sample of adolescents in grades 7–12 and examined how various social contexts influence adolescents’ health behaviors and psychological adjustment. The sampling design is comprehensively reviewed on the Add Health website (Bearman et al. 1997).

In the present study, samples from wave 1 (home interview, parent survey, and in-school questionnaire, collected between 1994 and 1995) and wave 2 (collected in 1996, containing home interview data only) were used because they both measured the target age group. Add Health’s wave 3 was not used because of its focus on an adult sample (Bearman et al. 1997). The present study incorporated predictor, control, and ASB outcome variables measured in wave 1, and ASB measured in wave 2 (approximately 1–2 years after the wave 1 variables were assessed). Among the adolescents in the original data set, 48 % were boys and 52 % were girls. Adolescents’ ages ranged from 12 to 21 years (M = 16.04, SD = 1.77). Approximately 66 % of the sample was Caucasian American, and approximately 34 % of the sample was an ethnic minority.

In terms of the original sample, wave 1 included 20,745 adolescents who were interviewed both in their homes and at their schools. Anonymity was provided to participants who listened to these pre-recorded questions on earphones and entered their responses directly into computers (Udry 1998). A total of 17,715 parents also completed an interview. The wave 2 sample included approximately 15,000 of the same students from wave 1. The sample for wave 2 received the same wave 1 in-home interview, except participants who were in the 12th grade during wave 1 were not re-interviewed at wave 2.

The sample (N = 1,196) for specific analyses in the present study used all participants who met the inclusion criteria, and included those adolescents who (a) participated in both wave 1 and 2; (b) had a parent or adult caregiver provide data in wave 1; (c) had at least one peer nominate the respondent to reciprocate friendship in network data in wave 1; and (d) had no missing data for relevant items on either wave of data collection. The selection criteria yielded a sample of 1,196 respondent/parent-adolescent pairs. Among the parents, 21.2 % were men and 78.8 % were women. Parents’ ages ranged from 20 to 80 years (M = 41.66, SD = 6.54); their education ranged from elementary school to professional training beyond college. Among the parents, 80.7 % were mothers of the target adolescent, 3.8 % were fathers, 1.6 % were grandmothers, and 13.9 % were other familial and non-familial caretakers. Sixty-one percent were Caucasian American, 19.4 % were African American, 8.3 % were Hispanic American, 1.3 % were Native American, and 2.7 % were Asian/Pacific Islander American. Demographic data were also collected from the 693 adolescent boys (57.9 %) and 503 adolescent girls (42.1 %), who ranged from 13 to 18 years (M = 15.67, SD = 1.68). Sixty-six percent of these adolescents were Caucasian American, 24.9 % were African American, 2.1 % were Hispanic American, 2.6 % were Native American, and 4.2 % were Asian/Pacific Islander American.

Measures

Demographic Information

These items included the target child’s race (dummy coded as Caucasian American or ethnic minority), gender (boy or girl), and age (at the time of data collection of wave 1).

Overt Antisocial Behavior Outcomes from Waves 1 and 2

Measures of both types of antisocial behavior were derived from adolescents’ responses to a series of items collected during both waves of in-home interviews. The “Overt ASB” measure included items indicating violent or aggressive behavior directed toward people. Examples of the six items utilized for this scale included fighting, injuring another person, and using or threatening to use a weapon. For each, adolescents reported how often they participated in the action during the past 12 months, using a scale ranging from 0 (if youth did not participate during the past year) to 3 (if youth participated in the act several times). Responses to relevant items were added to create this subscale. This six-item subscale had a Cronbach’s alpha of .73 for wave 1 and .75 for wave 2.

Covert Antisocial Behavior Outcomes from Waves 1 and 2

The “Covert ASB” measure included behaviors such as theft, property damage, rule breaking, and deceitfulness. In contrast to Overt ASB, this scale assessed forms of antisociality that did not entail physical aggression. Specific examples of the eight items include painting graffiti, committing property damage, shoplifting, stealing, lying to parents, and selling drugs. More specifically, the items asked adolescents to report how often during the past 12 months they had participated in these activities using a scale ranging from 0 (if youth did not participate during the past year) to 3 (if youth participated in the act several times). Responses to relevant items were added to create this subscale. The eight-item subscale yielded a Cronbach’s alpha of .78 for wave 1 and .75 for wave 2.

Parent- and Adolescent-Reported Parental Monitoring Predictor

This measure assessed both parents’ reports of their monitoring of their adolescent’s activities as well as adolescents’ perceptions of parent monitoring (Grotevant et al. 2006). Items asked parents about their adolescent’s best friend and their involvement in their adolescent’s day to day life, such as: “Have you met this friend in person?” “Have you met this friend’s parents?” “Have you talked about grades with adolescent?” “Have you participated in a school fundraising activity with adolescent?” and “Have you talked about other school activities with teens as well?” Adolescents’ perceptions of PM were measured with several items that assessed whether they participated in different activities with their parents, using a scale of yes (0) or no (1). Representative shared tasks included: “went shopping,” “played a sport,” “talked about a personal problem,” “talked about life,” “talked about school grades,” “went to a religious services,” “went to a movie together,” and “worked on a school project together.” Responses to relevant items were added and reverse coded as needed to create this scale. The 24-item scale had a Cronbach’s alpha of .72 in this sample. This measure aggregates the perceptions of both adolescents and parents, and deletion of any of the items would have lowered the alpha for the scale. Ultimately providing a composite view of the construct, family members’ perceptions were slightly correlated with each other (r = .07, p < .01).

Peer-Reported Deviant Peer Affiliation Predictor

This measure utilized peer network data from the Add Health study. Adolescent participants had the opportunity to nominate friends during the in-school survey. This survey measured their peer’s reported behavior (wave 1). More specifically, the Add Health study collected reports of a peer’s behavior which was measured by the peers’ actual self-report rather than the respondents’ report. The index of peer deviance was calculated by first defining the respondent’s peer network, which comprised adolescents whom the respondent nominated as a friend and those adolescents who nominated that respondent (e.g., the send-and-receive network). The scale assessed peer participation in minor deviant acts. Peers were asked how often during the past year they had gotten drunk, smoked cigarettes, skipped school without an excuse, and been involved in serious physical fights. Responses to items ranged from 0 (never) to 5 (three to five days a week). Responses to relevant items were added to create this scale. The six-item scale, labeled deviant peer affiliation, had a Cronbach’s alpha of .81 in this sample.

Adolescent-Reported Community Violence Exposure Predictor

The CV measure, derived from the in-home survey conducted at wave 1, included the following items relevant to witnessing or being the victim of CV, each coded on a yes (0) or no (1) basis: “Have you witnessed someone being shot or stabbed?” “Have you had a knife or gun pulled on you?” “Has someone shot or attempted to shoot you?” “Has someone cut, stabbed, or attempted to cut/stab you?” and “Has someone or group of people jumped you?” Adolescents’ responses to relevant items were added to create this scale. This five-item scale, labeled CV exposure, had a Cronbach’s alpha of .71 in this sample.

Results

Overview of data analyses

Table 1 contains descriptive statistics, including the unstandardized means and standard deviations for predictors and outcomes. We examined correlations among the predictors and outcome measures, which let us rule out any potential multicollinearity effects. We then investigated the direct and interactive multivariate relations among PM, CV, and peer deviance in the prediction of ASB over time to investigate the present study’s hypotheses. Both predictor and outcome variables were standardized prior to correlational and multiple regression analyses.

Table 1 Descriptive statistics and correlations among study variables
Full size table

Correlational Analyses

As shown in Table 1, measured correlations indicated that most of the predictors were significantly correlated with ASB (9 of the 12 coefficients calculated). Furthermore, these contextual variables related to both forms of ASB and these associations were significant for both boys and girls. Of note, the association between PM and ASB was found to be in the expected direction (e.g., as PM increases, ASB decreases) for both boys and girls. Exposure to CV was significantly correlated with both subtypes of ASB from both wave 1 and 2 for both boys and girls. Furthermore, exposure to CV had the strongest association with all ASB outcomes relative to the two other predictor variables (r = .24 to r = .60).

Hypothesized Multivariate Relations

We performed a series of three-step hierarchical multiple regressions to identify the significance of PM, deviant peers, and CV exposure in predicting the two subtypes of ASB (i.e., covert and overt) for boys and girls, separately. Step one of all models included age (at wave 1) and race (coded as either Caucasian or not Caucasian). For the four regression models predicting ASB outcomes at wave 2, the corresponding teen-reported ASB at wave 1 was entered as an additional control variable. In step two, the three primary predictor variables were added to the model (i.e., PM, deviant peer affiliations, and CV). Two interaction terms were entered at step three to test for moderated relations (PM × exposure to CV, and PM × deviant peer affiliation). Thus, there were a total of eight regression equations predicting ASB outcomes (each included a specified gender, a subtype of ASB, and a time period).

The predictors as a set were significantly related to all outcome variables and collectively accounted for between 13 and 40 % of the variance associated with the multiple ASB outcomes, as indicated in Tables 2, 3, 4, and 5.

Table 2 Summary of hierarchical regression analyses of variables predicting covert antisocial behavior outcomes at wave 1
Full size table
Table 3 Summary of hierarchical regression analyses of variables predicting overt antisocial behavior outcomes at wave 1
Full size table
Table 4 Summary of hierarchical regression analyses of variables predicting covert antisocial behavior outcomes at wave 2
Full size table
Table 5 Summary of hierarchical regression analyses of variables predicting overt antisocial behavior outcomes at time 2
Full size table

Hypothesis 1 (H1)

To test H1, we examined whether PM directly predicted ASB using multiple regression. Overall, PM was found to be the weakest independent variable, only significantly predicting two of the eight possible outcomes.

As demonstrated in Table 2, PM for girls was significantly associated with Covert ASB at wave 1, β = −.12, t = −3.30, p < .001. This inverse relationship suggests that greater PM may serve as protective influence for girls’ initial covert antisocial behavior, as predicted in H1. The significance of this gender difference was examined in a post hoc regression including all participants. The gender × PM interaction term was not significant; thus, somewhat qualifying the interpretation. As shown in Table 4, PM also significantly predicted boys’ Covert ASB at wave 2, β = .09, t = 2.58, p < .05. The significance of this gender difference was verified in a post hoc regression including all participants. The gender × PM interaction term indicated significant moderation effects, β = −.20, t = −2.66, p < .01. Thus, increases in PM over time were linked with increasing levels of covert antisociality for boys. Finally, no significant relationships were found between PM and all four Overt ASB outcomes. This is in contrast with the significant bivariate correlations between PM and overt ASB, as described earlier.

Hypothesis 2 (H2)

We also tested H2 (i.e., PM moderates the relationship between the other two contextual predictors and antisocial behavior) in the regressions described above. These analyses indicated that deviant peer affiliations and CV exposure both had significant relationships with ASB when other predictors were controlled. In general, these results were stronger for wave 1 outcomes than those assessed at wave 2. Exposure to CV consistently emerged as the most robust predictor and was significant in all equations.

To test for moderation as hypothesized, we created two interaction terms to enter into each regression equation (CV × PM and deviant peers × PM) in the third step of each analysis (cf. Baron and Kenny 1986). The CV × PM interaction term was significant only for predicting boys’ self-reported Covert ASB at wave 2, β = .08, t(542) = 2.37, p < .05. Following the procedures described by Aiken and West (1991), simple slopes were plotted for the relation between exposure to CV and boys’ reported Covert ASB at wave 2 for high levels, mean levels, and low levels of PM (high = +1 SD; mean = 0; low = −1 SD). These simple slopes were then analyzed post hoc to determine whether they were significantly different from the mean (due to the predictors being standardized the mean value for PM was zero).

As shown in Fig. 1, the simple slope for high PM significantly differed from zero (p < .001), the simple slope for the mean level of PM was the threshold for statistical significance (p < .001), and the simple slope for low PM differed significantly from zero as well (p < .01). Based on this procedure, those adolescents who were exposed to high amounts of violence within their community and increasing levels of PM across time displayed the greatest amount of covert antisocial behavior. This suggests that CV raises the likelihood of adolescent boys participating in nonviolent, delinquent acts, but parents’ increasing efforts to monitor their sons’ serves as a marker for increasing covert antisociality rather than as an effective means for reducing such behaviors.

Fig. 1
figure 1

Simple slopes of boys’ covert antisocial behavior at wave 2 predicted by exposure to community violence (CV) at two levels of parental monitoring (PM). Both predictors were centered, thus their mean equals zero. High PM one SD above the mean (+1); low PM one SD below the mean (−1). Values depicted are non-standardized regression coefficients (B; *p < .01; **p < .001)

Full size image

We also tested whether PM served as a mediator between the two other predictor variables (deviant peers and CV exposure) and ASB on a post hoc basis. However, we found minimal evidence for its mediating role. Specifically, PM was only associated with antisocial behavior in two out of the eight regressions, as described above. In both of these analyses, the exogenous variables maintained a stronger association with the outcome variable than did the proposed mediator.

Discussion

Using a large longitudinal data set, we juxtaposed the social interactional and social contextual models for predicting antisocial behavior for both boys and girls. The results from the present investigation addressed several questions. First, does PM predict either subtype of ASB? Second, does the gender of the adolescent affect how PM influences the development of ASB? Third, how salient is the social contextual model when conceptualizing the development of adolescents’ ASB? Finally, does PM act as a moderating variable, protecting teens from other stressors that may lead to ASB?

Parental Monitoring and Its Direct Influence on ASB Outcomes

Parental monitoring was found to be a relatively weak predictor of ASB within the present study, diverging from previous investigations (Dishion and McMahon 1998). For example, Sandstrom and Coie (1999) found that parents affect children’s interpersonal relationships through monitoring their youths’ social activities. Furthermore, they found that poor monitoring was linked to higher rates of delinquency and externalizing behavior.

This inconsistency could be due to the previous studies not simultaneously considering the influence of community contexts. Specifically, we expected that parenting would explain the greatest amount of variance in ASB because of the large amount of research that has supported the social interactional framework (e.g., Patterson 2002). However, many previous investigations have not incorporated community predictors which, when included, appear to remove large amounts of influence usually attributed to PM in the prediction of ASB. The relations found between PM and violent or aggressive subtypes of ASB extend the thesis that multiple contexts are salient when conceptualizing ASB.

Parental monitoring was found to predict covert ASB (in some instances) but not overt ASB across outcomes. Adolescents’ self-reported perpetration of aggressive acts, initially and 1 year later, were not influenced by their parents’ monitoring. These findings were unexpected and inconsistent with the previous literature (e.g., Dishion and McMahon 1998).

A plausible explanation for this finding is that the reactive and impulsive nature of most aggressive acts may not be as amenable to influence by parental control strategies (Dodge et al. 2006). In contrast, covert behaviors might be less impulsive and proactive, and possibly more subject to parent control.

Gender Differences in the Influence of Parental Monitoring

Another important finding that supports the social contextual model was the detection of gender differences in the relationship between PM and ASB. Boys who were highly monitored by their parents were found to be more antisocial a year later. This effect appears to be inconsistent with previous findings (i.e., Laird et al. 2003). Alternatively, girls were found to respond to PM by reporting less covert ASB at Time 1. However, PM was found to be significantly and negatively related with Overt ASB when examining their bivariate relations.

These gender differences may reflect different expectations for the appropriate amount of independence granted to each gender by their caregivers. For example, boys might expect and receive more autonomy from their caregivers relative to girls. Vieno et al. (2009) suggested that closeness in parent–child relations and perceived control contribute to differential outcomes. Girls were found to respond more favorably to being close and monitored by parents, whereas boys were found to be less close to their parents and viewed PM as a control tactic, which was linked with their perpetration of more ASB. Gorman-Smith and Loeber (2005) similarly found that girls’ covert ASB outcomes were significantly and negatively associated with PM, whereas boys’ covert and overt ASB outcomes were not.

Alternatively, these gender differences may have been driven by the adolescents’ propensity for ASB rather than the parents’ reaction to adolescent ASB. For example, parents may increase the monitoring of adolescents whom they believe are at greater risk of committing antisocial acts. However, their monitoring may not be effective at reducing ASB over time (thus, the increase in acting out behaviors).

Relevance of the Social Contextual Model

Exposure to community was the most powerful and pervasive predictor of ASB outcomes when controlling for both interpersonal and demographic risk factors, consistent with a social contextual framework for conceptualizing the development of adolescent ASB. Even after controlling for adolescents’ previous ASB from wave 1, their exposure to CV still strongly predicted subsequent ASB. This finding supports the need to consider distal or ecological risk factors when conceptualizing the development of ASB.

The present study replicated previous findings that adolescents’ exposure to CV significantly predicted overt antisocial behavior (Gorman-Smith et al. 2004; Trentacosta et al. 2009); however, this research further indicated that CV can be a risk factor for covert or sneaky antisocial behavior as well, consistent with previous studies (e.g., Miller et al. 1999). The present study also demonstrated that the relations between risk factors and outcomes were influenced by gender, time of measurement, and the type of ASB. Other research has likewise supported this notion (e.g., Miller et al. 1999; Scaramella et al. 2002; Tolan et al. 2003).

Parental Monitoring as an Intervening Variable in Predicting ASB

Overall, some evidence was found indicating that PM acted as an intervening variable in the development of ASB. Specifically, PM functioned as a moderator in this regard for boys, albeit in a different manner than anticipated. Specifically, PM interacted with CV exposure in the prediction of covert ASB. In other words, boys who lived in a dangerous neighborhood and had high amounts of PM demonstrated higher levels of covert antisocial behavior. No other moderating effects were indicated by the analyses.

Divergent findings have been reported in previous investigations regarding how PM interacts with risk factors for ASB to attenuate their deleterious effects. Some researchers have suggested that skilled parenting practices are not as protective as once had been thought, especially when adolescents are exposed to extensive violence within their community (Miller et al. 1999). In contrast, Mazefsky and Farrell (2005) found that parenting practices moderate the relationship between witnessing violence and aggressive behavior. The data from the present study, however, only provided evidence of a bivariate relationship between PM and adolescent aggressive behavior. After examining different subtypes of ASB and gender patterns, parenting practices appear to not be as effective at attenuating risk factors of ASB as some have documented.

Limitations and Future Research

There are several limitations of the present investigation that suggest directions for future research. First, this longitudinal study measured variables of interest at two points in time. These risk factors could be measured at additional times to provide further evidence of causation as well as more specific data regarding the trajectory of ASB throughout adolescence. In addition, the relations among the Time 1 variables are cross-sectional.

Second, due to the constraints associated with the collection and use of the archival data, we were unable to examine the risk factors measured at Time 2 (i.e., only home interview questions were collected during wave 2). The present investigation may have benefitted from analyzing differences between the adolescents’ level of risk between these two time periods. Additionally, assessing adolescents’ initial level of risk (i.e., in terms of PM, deviant peer affiliations, or CV exposure) can serve as a control in analyzing the longitudinal impact of these predictors. These control variables could similarly help to rule out the effects of the risk factors across time (e.g., a teen that no longer has deviant peers or a teen that moved away from a dangerous neighborhood). Connected to limitations regarding this archival data set, only a subset of the participants had peer-informant data, which reduced the valid N in some analyses. Examining these relations with different data sets may obviate this issue.

A third limitation pertains to the use of only self-report measures of adolescent ASB, which may provide a potential underestimate of such behaviors. One way to extend the present study’s findings would be to measure ASB using institutional data (e.g., police reports). Similarly, information from multiple reporters, including parental or peer reports, could be used for a composite index of adolescents’ antisocial behavior.

On a related note, CV exposure was the most consistent predictor of antisocial behavior, but both were based on adolescent report. Although there are issues associated with shared method variance, there is value to this measurement approach as well. Specifically, often individuals can provide a more accurate appraisal of their environments based on their lived experience. For instance, it is likely that adolescents have distinct experiences in their neighborhoods about which their parents are not aware or privy. The same holds true for parenting and peer variables. Perhaps if these constructs were assessed using self-report, more robust associations with ASB would have been found as well.

A fourth limitation relates to our exclusive focus on interpersonal and contextual risk factors. Including additional, established risk factors (see Loeber et al. 2009) into models may clarify their relations with the subtypes of ASB. Incorporating adolescents’ personal traits (i.e., personality features, cognitive attributions or intellectual deficits, genetic factors, executive functioning deficits) would be essential to include as well. Also, examining gender-specific risk factors, like callousness or relational aggression in females, could be examined within this framework (Keenan et al. 2010). Extensions of this study should examine additional dimensions of parenting, such as discipline, parent–child communication, parental warmth, and overall parent–child relationship measures to broaden the scope of the results.

Implications for Reducing Adolescent ASB

Results from this investigation have implications for treating antisocial youth. First, social workers in treatment settings can expand their assessment to include community or contextual factors, as they have bearing on the presenting problems (e.g., inquiring about community resources or problems, such as gang activity, as a routine part of the intake process). It is important that professionals not only assess adolescents’ relationships and family dynamics, but also aspects of the community in which they reside as well. Social workers have emphasized the importance of other predictors of ASB including adolescents’ temperament, socioeconomic status, low academic performance in conjunction with school failure, large family size, and parental criminality or mental health problems as well (Gentle-Genitty 2010). The number of accumulated risk factors appears to be more important than what types of risk factors youth face (De Mey et al. 2009).

Therefore, social workers can increase contacts with other agencies and institutions in coordinating services (e.g., schools, community organizations). For example, McKay (2010) suggested that school social workers implement critical service learning projects meant to foster compassion, empathy, sense of social responsibility and connection to community in youth at risk for ASB. In another example, social workers executing the STOP4-7 program intervened on the individual level by providing children with social skills training, home environment by offering parent management training, and school environment by providing teachers with classroom management training to reduce ASB (De Mey et al. 2009). Two other examples of interventions that use a contextual perspective are multisystemic therapy (Henggeler et al. 2009) and functional family therapy (Sexton and Alexander 2005). The overarching implication is that social workers may increase their effectiveness when they construe both overt and covert antisocial behavior from an ecological perspective, and consequently assess and intervene on multiple levels.