Examining Parental Monitoring, Neighborhood Peer Anti-social Behavior, and Neighborhood Social Cohesion and Control as a Pathway to Adolescent Substance Use

Abstract

Adolescence is a critical developmental period in which substance use can have long-term adverse effects. Structural ecosystem theory (SET) argues that community contexts may support or undermine the family’s ability to protect youth from substance use. Specific parenting attributes (e.g., providing structure, monitoring) have consistently been linked to youth substance use. However, less is known about how community factors may be influencing substance use through family and peer dynamics during adolescence. To address this gap, the current study uses five waves (ages 10–17) of data, collected as part of the Pitt Mother and Child Project (N = 228 low-income boys and their parents). This data are used to test a path model that investigates the relationship between neighborhood disadvantage (at age 10) and adolescent boys substance use (at age 17) through parental perceptions of neighborhood process (age 11), parents’ perceptions of monitoring (age 12) and affiliation with anti-social neighborhood peers/best friends (age 15). This study finds support for the relationship between neighborhood disadvantage in late childhood and substance use at age 17 through parental perceptions of neighborhood cohesion, parental monitoring at age 12, and the youths’ association with neighborhood best friends and marijuana use, but limited support for the indirect effect. The findings of this study partially support the assertion that neighborhood factors influence adolescent boys marijuana use by affecting other relationships within their ecological systems, suggesting that more research is needed in this area.

Adolescence is a critical developmental period in which youth become more influenced by their neighborhood environment and peer relationships, relative to earlier age periods when parental influence was more predominant (Steinberg, 2014). During this developmental period, risk and reward pathways are also established, making substance use during this age period a risk factor for substance-use problems later in life (Gil, Wagner, and Tubman, 2004; Griffin, 2017; Merline, O’Malley, Schulenberg, Bachman, and Johnston, 2004). While adolescents may not experience many of the consequences of substance use in the short term, substance use during adolescence is highly predictive of problematic transitions into adulthood (Chassin, Presson, Pitts, and Sherman, 2000; Shedler and Block, 1990), continued substance use, and related health issues (Shedler and Block, 1990).

Eco-developmental Model and Adolescent Substance Use

The eco-developmental model is based on the premise that a child’s development is shaped by interactions with multiple contexts within their environment, interactions that change with children’s emerging social, cognitive, and physical development, in addition to an emerging autonomy from parents (Bronfenbrenner, 1979). Bronfenbrenner (1979) also argued that the other people within a child’s ecological system (e.g., parents, peers, teachers) are also influenced by their own environments, thereby indirectly influencing the child’s development. A child, for example, may never interact with their parents’ employer directly, but still may be affected by their parents’ employer, based on outcomes such as the parent’s income, stress experienced at work, hours spent away from home, and provisions for childcare (e.g., ability to work from home). The eco-developmental model, when applied to adolescent substance use, suggests that it is essential to consider multiple risks and protective factors across multiple levels of an ecological system throughout a child’s life and consider the ways in which these factors interact. In his unified theory of development, Sameroff (2010) argues that a child’s development is impacted by intrinsic characteristics, proximal influences, such as family, and initially distal factors that become increasingly more proximal as children spend more time out of the home interacting with peers and the broader neighborhood as they age (Sitnick et al., 2014). Applying a life-course perspective (Bengtson & Allen, 2009) to an eco-developmental framework requires complex models to capture the increasingly more extensive network of relationships adolescents have relative to younger children, and the relationships between individuals, family, peers, and neighborhood factors over time (Sameroff, 2010).

For example, Szapocznik and Coatsworth (1999) proposed the structural ecosystem theory (SET). SET places the family at the center of the child’s ecological system and argues that community and cultural contexts can either support or undermine the family’s protective function (Szapocznik and Coatsworth, 1999). Thus, while specific parenting attributes (e.g., providing structure, monitoring) have consistently been linked to youth substance use during adolescence, associations between parenting and youth substance use are likely influenced by multiple factors in a youth’s broader ecological system. Accordingly, the authors advocate for research that examines the process through which these factors, including parenting and peer/community factors, relate to each other (Szapocznik and Coatsworth, 1999). This current study applies SET in examining a path model that includes parental perceptions of neighborhood processes in early adolescence, parental monitoring, affiliation with anti-social neighborhood peers, and youth’s substance use in late adolescents.

Literature Review

Discussions of risk and protective factors for substance use among adolescents are typically organized into several broad categories, including: (a) the individual, (b) the family, (c) peers, and (d) community (Hawkins, Catalano, and Miller, 1992; Marschall-Lévesque, Castellanos-Ryan, Vitaro, and Séguin, 2014). These are not often examined conjointly. However, in exception, a large cross-sectional study jointly examined the individual, the family, peers, and community-level protective factors on 30-day substance use (Cleveland et al., 2010). The findings indicated that family (e.g., family attachment and family supervision) and community protective factors (e.g., neighborhood attachment and disorganization) were more salient for 6th–8th graders, while school protective factors were the most salient for 10–12^th graders (Cleveland et al., 2010). In general, these findings suggest potential developmental differences in the impact of protective factors on adolescent substance use across domains. More specifically, the results suggest that neighborhood-level factors might play a critical role in beginning the pathway leading to adolescent substance use in later adolescence.

Neighborhood Factors and Adolescent Substance Use

Neighborhood disadvantage has been consistently associated with anti-social behavior, depression, and psychological distress in adolescents (T Leventhal and Brooks-Gunn, 2000; Tama Leventhal and Dupéré, 2019); however, the direct associations between a neighborhood’s level of disadvantage and adolescent substance use are inconsistent (Cambron et al., 2020; Karriker-Jaffe, 2011; Snedker, Herting, and Walton, 2009). For example, (Cambron and colleagues (2020) found a relationship between neighborhood disadvantage and tobacco/alcohol use, but not in marijuana. Theoretically, neighborhoods where children grow up could influence their socialization and social support (Korbin and Coulton, 1997), their exposure to crime and violence (Gibson, Morris, and Beaver, 2009), and their exposure and opportunities to use substances (Crum, Lillie-Blanton, and Anthony, 1996), all of which serve as risk and protective factors for substance use. In a review of the literature, Jackson and colleagues (2014); identified 15 studies that tested the relationship between neighborhood disadvantage and adolescent alcohol use; eight studies found significant effects. In most cases, the associations observed were for recent and frequent use rather than annual or lifetime use. However, the study that was identified as being the most rigorous (i.e., controlled for parental monitoring and deviant peer affiliation) found that neighborhood affluence, rather than disadvantage, was related to more adolescent alcohol use (Maimon and Browning, 2012). Other studies have found neighborhood disadvantage to be positively associated with an adolescent’s monthly consumption of alcohol (Barr, 2018).

Collective efficacy, a combination of social cohesion, social control, and an additional aspect of the neighborhood, is frequently considered as a relational process that occurs at the neighborhood level (Hipp, 2016) and is linked to neighborhood disadvantage (Sampson, 1997). While collective efficacy has been found to protect youth from substance use in the presence of neighborhood-level risk factors, its association with adolescent substance use is also inconsistent (Duncan, Duncan, and Strycker, 2002; Erickson, Harrison, Cook, Cousineau, and Adlaf, 2012; Fagan, Wright, and Pinchevsky, 2015; Maimon and Browning, 2012). In Duncan and colleagues’ study (2002);, they found that neighborhood poverty was negatively related to neighborhood social cohesion, which, in turn, was negatively related to perceived neighborhood problems with youth alcohol and drug use. However, both Fagan and Colleagues (2015); and Maimon and Browning (2012) found no direct relationship between the neighborhood collective efficacy and adolescent substance use. It has been theorized that collective efficacy provides additional supervision (social control) and supports (social cohesion) when youth are in public spaces and transmit a prosocial norm (Catalano and Hawkins, 1996; Laub and Sampson, 1993).

Neighborhood poverty has been consistently related to less collective efficacy (Sampson, 1997), which, in some studies, has been related to less adolescent substance use. Many of these studies, however, were cross-sectional and did not fully assess collective efficacy (social control and cohesion), thus limiting the conclusions that can be drawn from their findings. In Duncan et al. cross-sectional study (2002);, the association between neighborhood poverty and youth drug and alcohol arrests at the neighborhood level was partially explained by social cohesion. In another cross-sectional study conducted in New Zealand, neighborhood disadvantage was related to binge drinking “indirectly” through “collective efficacy” (measured by combining social cohesion and parental knowledge), such that neighborhood disadvantage was related to less collective efficacy, which was related to more binge drinking (Jackson, Denny, Sheridan, Zhao, and Ameratunga, 2016). Although promising in demonstrating associations among neighborhood processes and youth substance use, these studies’ cross-sectional designs and incomplete measures of collective efficacy limit the inferences that can be drawn. In the few studies that fully measured collective efficacy, it was negatively related to adolescent substance use and found to buffer other risk factors within the neighborhood, such as the presence of alcohol stores or violence (Fagan, Wright, and Pinchevsky, 2014; Jackson, Denny, and Ameratunga, 2014; Maimon and Browning, 2012). Longitudinal research that includes the complete construct of collective efficacy is needed to establish the relationships between neighborhood disadvantage, social control, and social cohesion, and adolescent substance use (Booth and Shaw, 2020; Hipp, 2016).

The Role of Peers and Parenting in the Relationship between Neighborhoods and Adolescent Substance Use

Parents act as a liaison between adolescents and their environment, potentially shielding them from neighborhood stressors and interactions with peers that engage in antisocial behavior (Furstenberg, Cook, Eccles, Elder, and Sameroff, 1999). However, living in a disadvantaged neighborhood may undermine parents’ ability to monitor their child’s behavior, and the risk for substance use may be exacerbated by youths’ exposure to anti-social peers (Branstetter and Furman, 2013; Racz and McMahon, 2011; Shaw and McKay, 1943). Several cross-sectional studies have provided evidence that associations with anti-social peers and family functioning explain some of the relationships between neighborhood disadvantage/collective efficacy and substance use (Bernburg, Thorlindsson, and Sigfusdottir, 2009; Su and Supple, 2014; Wen, 2017); however, they are limited in their ability to test the causal associations over time.

The few studies that utilized longitudinal data sets to understand the role of parenting and peers in the relationship between neighborhood influence and substance use have found evidence to support both. Cambron et al. (2020) used longitudinal data collected over four waves from a diverse sample (24% African American, 21% Asian American, and 9% Native American) to investigate the role of neighborhood disadvantage, family functioning, and deviant peers on the growth of substance use from grade 5 to 9. They found that growth in neighborhood disadvantage, low family functioning, and associations with anti-social peers were all directly associated with growth in smoking and alcohol use. The relationship between neighborhood disadvantage and alcohol use was mediated by changes in family functioning and associations with anti-social peers. Significant indirect effects were found between neighborhood disadvantage, anti-social peers, and both smoking and alcohol use. In another longitudinal study that utilized a path model in a sample of primarily White (83%) and middle income (median income 70 K) youth, ages 11–13, neighborhood disadvantage was related to associations with anti-social peers, which in turn was related to alcohol use (Trucco, Colder, Wieczorek, Lengua, and Hawk, 2014). In this study, however, neighborhood disadvantage was unrelated to positive parenting, and positive parenting was not related to alcohol use. Lastly, Chuang et al. (2005), in a sample of 12–14-year-olds assessed three points in time for a year, found that living in a low SES neighborhood was associated with more parental monitoring (reported by the adolescent), and subsequently, fewer associations with substance-using peers and less adolescent substance use. All of these longitudinal studies focused on neighborhood disadvantage and did not assess the role of neighborhood social cohesion and social control on subsequent parental monitoring and affiliation with deviant peers (Jackson et al., 2014).

Parental monitoring is one of the most consistent predictors of adolescent substance use (Lac and Crano, 2009). It is increasingly recognized that parental monitoring may reflect a parent’s ability to solicit information or control behavior and an adolescent’s willingness to disclose information to their parents (Stattin and Kerr, 2000). However, A systematic review of the literature found that parental monitoring and control were more protective against substance use than parental knowledge and communication (Ryan, Roman and Okwany, 2015). In the neighborhood context, neighborhood factors may motivate parental monitoring or inhibit a parent’s ability to monitor their child, and parental monitoring may protect adolescents from exposure to risky spaces and peers (Furstenberg, et al., 1999). Gender differences are observed in how much parents monitor boys and girls and the effect on substance use. Boys are typically monitored less than girls, and while parental monitoring is related to less substance use for boys and girls, it is more strongly related to substance use among boys (Keogh-Clark et al., 2021; Rusby et al., 2018). While adolescents’ perceptions of parental monitoring are consistently related to less substance use (Lac and Crano, 2009), the relationship between parents’ perceptions of parental monitoring and substance use is less clear (Branstetter and Furman, 2013; Augenstein et al., 2016). Villarreal and Nelson (2018), however, found that mothers’ and fathers’ perceptions of parental monitoring were related to less substance use among adolescent boys. Although there are discrepancies in adolescents’ and parents’ perceptions of monitoring, parents’ perceptions of social cohesion and control are related to both adolescents and their parents’ perceptions of parental monitoring (Booth and Shaw, 2020). No studies, to our knowledge, have examined the role of neighborhood factors on parents’ perceptions of parental monitoring and its effect on adolescent boys’ substance use.

While the affiliation with peers that engage in anti-social behavior to adolescent substance use has been well established, very few studies have explicitly focused on the role of anti-social peers that live in the adolescent’s neighborhood. In a notable exception, using the current sample in which friends in the neighborhood and at school were distinguished from one another, youths living in disadvantaged neighborhoods were twice as likely to participate in antisocial behavior with a neighborhood peer, and that peer association was related to escalating anti-social behavior (Ingoldsby et al., 2006).

The Current Study

The direct relationship between neighborhood disadvantage and adolescent substance use has been inconsistent despite clear evidence that there is variation in adolescent substance use at the neighborhood level. These inconsistencies suggest that the critical role of interpersonal mediators, such as parents’ perceptions of neighborhood cohesion and control, parental monitoring, and anti-social peer affiliation, the latter including best friends and other peers in the neighborhood (see Fig. 1). Studies examining the relationship between neighborhood features and substance use are often limited by using administrative data and cross-sectional designs, prohibiting inferences about paths between parents’ perceptions of social process in the neighborhood, parenting, interaction with peers, and adolescent boys substance use across adolescents. In a recent longitudinal study using the present cohort, Booth and Shaw (2020) found that neighborhood social cohesion at youth ages 11 and 12, within the context of low-income neighborhoods, was positively related to parents perceptions of parental monitoring at ages 12 and 15. In contrast, neighborhood social control was negatively linked to later parental monitoring. However, this paper did not test associations between peers, family, and neighborhood in relation to subsequent substance use. Incorporating an eco-developmental model framed by Szapocznik and Coatsworth’s (1999) SET’s perspective, and building on findings from a previous paper with the current sample (Booth and Shaw, 2020), we hypothesized that (see Fig. 1):

Fig. 1.

Theoretical Model

1. Neighborhood disadvantage at age 10, will be negatively related to parental perceptions of neighborhood control and cohesion at age 11.

2. Parental perception of neighborhood social cohesion and control at age 11 would be positively related to parents perceptions of parental monitoring at age 12.

3. Parents perceptions of parental monitoring at age 12 will be negatively associated with neighborhood peers and best friends’ anti-social behavior at age 15.

4. In turn, anti-social behavior among neighborhood peers and best friends will be positively related to adolescent boys’ substance use at age 17.

5. Neighborhood social control and cohesion, parents’ perceptions of parental monitoring and neighborhood peer and best friend antisocial behavior will mediate the relationship between neighborhood disadvantage at age 10 and substance use at age 17.

In addition to these paths, we also tested direct associations between parental monitoring and parents’ perceptions of neighborhood social cohesion and control at age 15 on adolescent substance use at age 17, and the direct relationship between parental perceptions of neighborhood social control and cohesion at age 12 and neighborhood peers and best friends engaging in anti-social behavior at age 15.

Methods

This study uses data collected as part of the Pitt Mother & Child Project, a longitudinal study of risk and vulnerability of low-income boys that began in 1991 when the children were between 12 and 18 months of age (Shaw et al., 2003). Families were initially recruited from Women, Infants, and Children Nutritional Supplement Programs (WIC) clinics surrounding the Pittsburgh metropolitan area. At the time of recruitment (1991) families were eligible to receive WIC if their income was <185% of the established poverty line, which in 1991 was $24,790 annually for a family of four. Participants who had complete responses at age 10 (N = 228), were included in this study. Families that were included in this analysis did not differ significantly from those in the sample that were not included, on anti-social behavior (t (237) = 0.04, p = 0.96), SES(t (240) = 0.19, p = 0.85), and reported parental monitoring(age 12: t (234) = 0.97, p = 0.33; age 15: t (251) = −0.16, p = 0.88) or adolescent substance use (tobacco: t (246) = −1.28, p = 0.20; marijuana: t (246) = −0.34, p = 0.73; alcohol:: t (246) = −1.76, p = 0.08). Adolescents in the sample lived in 130 different census tracts, with between one and six respondents in each tract. Census tracts are the administrative boundaries used by the US government when measuring geographical areas. On average, they include 4000 residents and are typical used to measure “neighborhoods.” In this study, <14–16% of the sample lived in census tracts with three or more participants, and only 2–8% lived in the same tracts as four respondents, depending upon the year of assessment. Thirty-one percent of the sample moved from one census tract to another at least once during the study period.

Description of the Sample

All participants in this study were male and the majority primary care givers reported being a mother. At age 11,93.5% of the parent sample were mothers, 3.3% were fathers, and 3.3% were otherwise related (e.g., grandparents, aunts). Primary caregiver characteristics were consistent overall time points with, with 92.5% of primary care givers at age 12 reporting being a mother, 93.4% at age 15, and 89.6% at age 17. At age 11, 45.7% of primary care givers reported that they had completed at least 1 year of college, 35.9% reported having a high school diploma, and 9.8% reported graduating from a 4-year college. The average annual family income in the sample was $28,789 at age 11, $32,268 at age 12, $34,251 at age 15 and $37,437 at age 17. At age 11, 46.91% reported being married, and this remained consistent across all ages (age 12, 46.48%, age 15, 45.53%, and age 17, 44.6%). Most of the youths reported being White (56.4%), 34.6% reported being African American, and 8.9% were more than one race. Similarly, 64.8% of the caregivers reported being White, 34.2% reported being African American, and less than one percent reported being more than one race.

Measures

Adolescents’ substance use was measured using five items from the Adolescents Activities Questionnaire (Elliott, Ageton, and Huizinga, 1982) which was completed by the adolescents when they were 15 and 17. Both cigarette smoking and marijuana use were measured with a single item that asked: “How often have you (a) secretly used tobacco, (b) smoked marijuana. Alcohol use was measured using three items asking adolescents: “How often have you (a) secretly taken a sip of beer, (b) secretly taken a sip of wine, and (c) secretly taken a sip of liquor. All questions had three possible responses: never (0), once/twice (1), more often (2). A mean score of the alcohol-use questions was created. In the model use of each substance was included as individual variables and allowed to covary. It should be noted that these measures are only assessing substance use that is done in secret which may not include all the adolescent’s substance use behavior.

Neighborhood cohesion and control

Neighborhood cohesion and control were both assessed in the parental survey when the youths were ages 11, 12, and 15; each was measured using a set of five items (Sampson, Morenoff & Earls, 1999). To measure social cohesion, parents were asked to indicate their level of agreement with the following items: (a) people around here are willing to help their neighbors, (b) this is a close-knit neighborhood, (c) people in this neighborhood can be trusted, (d) people in this neighborhood generally don’t get along with each other, and (e) people in this neighborhood do not share the same values. This 5-item Likert scale included possible responses ranging from strongly agree (5) to strongly disagree (1). The neighborhood cohesion factor had strong internal consistency at all time points: age 11 α = 0.84, age 12 α = 0.87, and age 15 α = 0.87. Neighborhood social control was assessed by asking respondents to indicate how likely it was that their neighbors could be counted on to intervene if the following occurred: (a) adolescents were skipping school and hanging out on a street corner, (b) adolescents were spray-painting graffiti on a local building, (c) adolescents were showing disrespect to an adult, (d) a fight broke out in front of their house, and (e) the fire station closest to their home was threatened with budget cuts. Again, possible responses ranged from very unlikely (1) to very likely (5). Neighborhood social control had strong internal consistency at all time points: age 11 α = 0.84, age 12 α = 0.82 and age 15 α = 0.88.

Neighborhood disadvantage

Neighborhood disadvantage was measured at age ten. The adolescent’s home address was assessed at age 10 and geo-coded into census tracts. Neighborhood disadvantage was assessed with four items: (a) the percentage of individuals living below the poverty line, (b) the percentage of unemployed, (c) the percentage on public assistance, and (d) the percentage of single-parent households. This scale had good internal consistency, α = 0.78. All items were standardized, and a mean of standardized scores was created. All scores are in relation to the mean level of neighborhood poverty, unemployment, public assistance, and single-parent households within the sample.

Parents’ perceptions of parental monitoring were measured based on parental reports when the adolescents were ages 12 and 15. Parental perception of monitoring was assessed using a measure originally developed by Dishion et al. (1991) and adapted by Moilanen et al. (2009). The scale included questions: (a) How often do you think your adolescent goes to places that you ask him not to go? (b) When your adolescent is going to a friend’s house, how often do you check to see if a parent or another adult will be present? (c) When your adolescent goes out of the house for more than a few minutes, how often are you aware of what he is doing? (d) How often are there rules when your adolescent is home without an adult? Response options ranged from (1) never to (5) always or almost always. The first item in the scale was reverse coded. The internal consistencies of the scale α = 0.59 (average interim correlation = 0.26) at age 12 and α = 0.52 (average interim correlation = 0.21) at age 15.

Neighborhood Best Friends and Peer Groups Antisocial Behaviors

Neighborhood Best Friends and Peer Groups Antisocial Behaviors were measured using the Children’s Friend’s Form: “My Neighborhood Friends” instrument at age 15 (Ingoldsby et al. 2006) that asked the adolescents’ how often their “neighborhood best friend” and “neighborhood overall peer group” participated in 19 anti-social behaviors. All 19 items were assessed referring to the adolescents: (a) best friend in their neighborhood (BFN) (α = 0.90) and (b) peer group in their neighborhood (PGN) (α = 0.94). Responses were assessed on a 4-point scale, with possible responses ranging from never (0) to a lot/always (3). Anti-social behaviors assessed included: (a) skipping school without an excuse, (b) purposely damaging or destroying something that is not yours, (c) stealing something from a store or from someone else, (d) hitting an adult parent or teacher, (e) hitting other kids or physical fights, (f) drinking alcohol, beer, wine or liquor, (g) using marijuana, (h) involved in gang activity and/or (i) carried a weapon. A sum score was calculated for both, best friends and peer groups.

In this study, our models were estimated controlling for the effect of youths’ socioeconomic status and their parents’ report of anti-social behavior at age 10 when predicting parental monitoring, neighborhood peers, and best friends engaging in anti-social behavior, and substance use. Socioeconomic status (SES) at age 10 was measured using the Hollingshead (1975) index, which incorporates the parents’ education, occupation, sex, and marital status; a separate standardized measure was used for the family’s annual income. The adolescent’s level of anti-social behavior was measured using the anti-social behavior subscale from the Child Behavior Checklist at age 10 (α = 0.68) (Achenbach, 1991). This checklist was a parental report of the adolescent’s anti-social behavior. SES was included as a control variable because it could plausibly account for the relationship between neighborhood factors, parental monitoring, interaction with neighborhood anti-social peers/best friends, and substance use. Anti-social behavior at age 10 was also included as a control to account for monitoring behavior that may be in response to a high level of problematic behavior, acknowledging the bidirectional relationship that has been found between anti-social behavior and parental monitoring in previous studies (Jang & Smith, 1997).

Data Analysis Strategy

To understand the relationship between parental perceptions of neighborhood characteristics, parental monitoring, adolescents’ neighborhood peers and best friends engaging in anti-social behavior, and substance use, path models were estimated in MPlus using maximum likelihood estimation. Missing data was imputed using FIML. Path models were not estimated in a multilevel framework due to the low rate of clusters in the data (participants were not sampled to assess neighborhood effects); however, standard errors and chi square test of model fit were adjusted to account for clustering, using the census tracts that youths lived in when they were 10 years old, with a total of 148 clusters identified. Additionally, and MLR estimator was used because it is robust to non-normality of observations. Path model fit was assessed using several fixed indices: chi-squared (χ²); comparative fit index (CFI); Tucker-Lewis index (TLI); and the root mean square error of approximation (RMSEA). The model was considered good fitting when the chi-squared was not significant (Kline, 1998), CFI, and TLI greater than 0.95 and an RMSEA, <0.05 (Wang & Wang, 2019). Using the Preacher & Coffman (2006) tool to calculate the power in a SEM, given the 65 degrees of freedom in our model, a sample of 209 is required to have the power (Beta = 0.80, α = 0.05) to detect a good model fit (RMSEA = 0.05). The first model estimated, tested all the paths hypothesized, including direct and indirect paths. Based on these findings, an additional model was estimated to investigate the direct relationship between parental monitoring at age 12 and substance use at age 17. All variables measured within a one-time point were allowed to covary with all other measures in that time point. Standardized betas are reported to provide ease of interpretation. All paths predicting parental monitoring, neighborhood peers and best friends engaging in anti-social behavior and adolescent substance use, controlled for the parent’s SES (measured using the Hollingshead index), adolescent anti-social behavior at age 10 and neighborhood disadvantage at age 10. All paths predicting parental perceptions of neighborhood social control or cohesion were estimated, controlling for neighborhood disadvantage at age 10. All paths predicting adolescent substance use at age 17 were estimated, controlling for the youths’ reported use of that same substance at age 15. The indirect relationships through paths ways, found to be statistically significant, were also tested, including the full hypothesized path and the shorter paths within it using bootstrapping method, drawing 1000 samples. Based on the findings, Wald’s tests were conducted to test the differences in the association between parental monitoring at age 12 and associating with anti-social peers and best friends at age 15, and between associating with anti-social peers and best friends at age 15 and substance use at age 17.

Results

At age 17, 24.6% of participants reported ever using tobacco, 37.1% reported ever using marijuana, and 45.2% reported ever using alcohol. The youths’ use of all substances was correlated, with the highest correlation being between marijuana and alcohol use (r = 0.43, p < 0.001) (see Table 1). Adolescents in the sample reported their neighborhood peers participated in significantly more antisocial behavior (M = 8.95, SD = 10.73) than their neighborhood best friends (M = 5.61, SD = 7.18) (t (240) = −6.17, p < 0.001), and these two groups’ involvement in anti-social behavior was highly, but not overwhelmingly correlated (r = 0.61, p < 0.001), indicating variability in anti-social behavior across peer context.

Table 1.

Correlations and descriptive statistics

	2	3	4	5	6	7	8	9	10	11	12	13	14	15	16	17	18	19	M(SD)
Tobacco Use 17 (1)	0.36	0.41	0.32	0.26	0.36	0.16	0.19	−0.04	−0.05	−0.06	−0.02	−0.05	−0.02	0.01	−0.01	−0.10	0.06	0.04	0.35(0.67)
Marijuana Use 17 (2)	–	0.43	0.18	0.36	0.23	0.33	0.30	−0.08	−0.04	−0.07	−0.03	−0.05	−0.05	−0.18	−0.09	0.06	0.27	0.11	0.54(0.77)
Alcohol Use 17 (3)		–	0.06	0.11	0.44	0.13	0.15	−0.03	0.06	0.05	0.04	0.02	0.00	−0.03	−0.05	−0.03	0.01	0.11	0.41(0.57)
Tobacco Use 15 (4)			–	0.40	0.27	0.20	0.09	0.12	0.05	0.04	0.06	0.05	0.07	0.01	−0.11	0.05	0.24	−0.04	0.19(0.50)
Marijuana Use 15 (5)				–	0.24	0.29	0.25	−0.08	−0.15	−0.08	−0.07	−0.09	−0.03	−0.17	−0.20	0.07	0.36	−0.11	0.19(0.48)
Alcohol Use 15 (6)					–	0.20	0.09	−0.01	−0.03	0.01	−0.02	−0.04	0.03	0.02	−0.02	−0.08	0.07	0.11	0.21(0.49)
Best-friend Antisocial 15 (7)						–	0.60	−0.04	−0.08	−0.06	−0.09	−0.04	−0.14	−0.15	−0.19	−0.02	0.19	0.11	5.61(7.18)
Peer Antisocial 15 (8)							–	−0.13	−0.15	−0.11	−0.11	−0.11	−0.22	−0.14	−0.10	0.01	0.18	0.04	8.95(10.73)
Social Cohesion 11 (9)								–	0.63	0.46	0.63	0.48	0.37	0.17	0.11	−0.14	−0.12	0.14	17.42(4.18)
Social Cohesion 12 (10)									–	0.56	0.53	0.69	0.50	0.21	0.19	−0.23	−0.25	0.12	17.35(4.25)
Social Cohesion 15 (11)										–	0.38	0.41	0.72	0.07	0.11	−0.11	−0.17	0.03	17.03(4.33)
Social Control 11 (12)											–	0.63	0.40	0.07	0.10	−0.08	−0.02	0.13	19.36(4.67)
Social Control 12 (13)												–	0.44	0.26	0.10	−0.08	−0.14	0.10	18.98(4.70)
Social Control 15 (14)													–	0.07	0.14	−0.15	−0.16	0.03	18.27(5.38)
Parental Monitoring 12 (15)														–	0.50	−0.13	−0.23	0.11	4.29(0.67)
Parental Monitoring 15 (16)															–	−0.13	−0.19	0.09	4.02(0.72)
Neighborhood Dis 10 (17)																–	0.20	−0.18	14.31(12.46)
Anti-social Behavior 10 (18)																	–	−0.02	1.97(2.78)
SES 10 (19)																		–	30.79(9.63)

All pairwise correlations with p values < 0.05 are bolded

The final model testing our hypothesized path model demonstrated good model fit X²(65) = 74.38, p = 0.20, RMSEA = 0.03; CFI = 0.99; TLI = 0.98; SRMR = 0.05 (see Fig. 2). The first hypothesized relationship in the path model was partially supported; a significant relationship was found between neighborhood disadvantage when the youths were age 10 and parental perception of neighborhood cohesion at age 11 (β(SE) = −0.15(0.06), p < 0.05). However, there was no significant association between neighborhood disadvantage at age 10 and parental perceptions of neighborhood social control. There was also partial support for the second hypothesis as a positive relationship was found between parental perception of neighborhood social cohesion at age 11, and parental monitoring at age 12 (β(SE) = 0.21(0.09), p < 0.05). However, there was no support for the relationship between parental perception of neighborhood social control and parental monitoring. There was also partial support for the third hypothesis: parental monitoring at age 12 would be negatively associated with the adolescents’ neighborhood best friends and peers antisocial behavior. In our model, there was a negative relationship between parental monitoring at age 12 and the youths’ neighborhood best friends’ anti-social behavior (β(SE = −0.14(0.07), p < 0.05), however, no significant association between parental monitoring and neighborhood peers’ anti-social behaviors. However, a Wald’s test indicated there was no statistically significant difference in the paths between associating with anti-social peers and best friends and parental monitoring.

Fig. 2.

Full model testing the hypothesized path model and direct effects

Lastly, there was partial support for the fourth hypothesis: neighborhood best friends and peers’ anti-social behavior at age 15 would be related to substance use at age 17. Specifically, there was a significant relationship between neighborhood peers’ antisocial behaviors and tobacco use (β(SE = 0.20(0.08), p < 0.01) and alcohol use (β(SE = 0.19(0.08), p < 0.05). A significant relationship was also found between neighborhood best friends’ antisocial behavior at age 15 and marijuana use at age 17 (B(SE = 0.20(0.08), p < 0.01). A Wald’s test indicated there was no difference between the role of anti-social neighborhood best-friends and peers at age 15 and marijuana use or tobacco use, and a marginally significant difference for alcohol use (Wald = 3.13, p = 0.08). In the case of both tobacco and alcohol, neighborhood peers’ antisocial behavior was related to substance use; however, neighborhood best friend anti-social behavior was not. There was also a difference in the role of neighborhood best friends on marijuana use when compared to tobacco use (Wald = 11.00, p < 0.001) and between marijuana and alcohol use (Wald = 11.78, p < 0.001), but not between alcohol and tobacco use. There was no difference in the role of neighborhood peers on substance use across all substances.

A significant negative relationship was also found between parental perception of neighborhood social cohesion at age 15 and adolescents’ tobacco use at age 17 (β(SE = −0.19(0.09), p < 0.05). However, significant associations were not evident between parental monitoring at age 15 and adolescent substance use at age 17, or between parental perceptions of neighborhood social cohesion and control at age 12 and youths’ neighborhood best friends or peers engaging in anti-social behaviors at age 15.

Based on these findings, the indirect effect of social cohesion at age 11, parental monitoring at age 12, and associating with neighborhood best friends at age 15 on marijuana use at age 17 was estimated. The indirect effect for the entire pathway, tested using bootstrapped standard errors was not statistically significant (β(SE = 0.00(0.00), p = 0.34), marginally significant indirect effects were found for the relationships between (a) neighborhood disadvantage at age 10, social cohesion at age 11, and parental monitoring at age 12 (β(SE = −0.04(0.02), p = 0.06); (b) neighborhood social cohesion at age 11, parental monitoring at age 12, and association with a best friend that engages antisocial behavior at age 15 (β(SE = −0.03(0.02), p = 0.09) and (3) parental monitoring at age 12, association with a best friend that engages in antisocial behavior at age 15 and substance use at age 17 (β(SE = −0.03(0.02), p = 0.13).

The second model estimated excluded parental monitoring at age 15 and adolescents’ neighborhood peers and best friends’ antisocial behaviors, and tested the direct effect of parental monitoring at age 12 on substance use at age 17. This model demonstrated excellent model fit X²DF = 58.51(49), p = 0.17, RMSEA = 0.02; CFI = 0.99; TLI = 0.97; SRMR = 0.06. No statistically reliable associations were found between parental monitoring at age 12 and any substance (i.e., only marginal trend with age 15 marijuana use). Similarly, no direct effects were found between parental monitoring at age 15 and substance use at age 17.

Discussion

Based on ecological systems theory (Bronfenbrenner, 1979) and SET (Szapocznik and Coatsworth, 1999), and motivated by inconsistent associations between neighborhood disadvantage and adolescent substance use in past research, this study sought to test the associations between neighborhood disadvantage and adolescent boys substance use through parental perceptions of social process in the neighborhood, parental monitoring, and neighborhood peers/friends’ anti-social behavior. This study improves upon previous research that had sought to unpack these complex relationships by utilizing data collected over five-time points between ages 10 and 17. Our study found partial support for a developmental process by which neighborhood disadvantage in late middle childhood is associated with adolescent marijuana use through parental perceptions of social cohesion, parental monitoring, and adolescents’ interaction with neighborhood best friends’ who engage in anti-social behavior. While all the hypothesized direct relationships were statistically significant, the indirect effect of the full path was not. Despite the lack of significant indirect effect, the findings support the importance of considering how meso systems (i.e., neighborhood attributes) are directly and indirectly related to youth by influencing people throughout their ecological systems, such as their parents, peers, and best friends. The results suggest that controlling for family and peer risk and protective factors in studies seeking to understand the role of neighborhood dynamics in adolescent substance use may be masking its effects by not considering the ways that neighborhood characteristics might influence those relationships.

Consistent with previous literature that found family and community factors to be salient in early adolescence and peer influences to become more salient in later adolescence (Cleveland et al., 2010), this study found that parental monitoring at age 12, rather than age 15, was related with associating with antisocial peers at 15 and substance use at 17. This finding indicates that parental monitoring may be most influential in early adolescence by limiting youths’ exposure to peers that engage in anti-social behavior. The lack of a direct relationship between parental monitoring at age 15 and the use of all substances at age 17 reinforces the need to examine risk and protective factors from a life course perspective. These results support the importance of considering both family and peer influence to understand the relationship between neighborhood-level factors and substance use. In line with Bernburg et al. (2009) study that found anti-social peers to explain the effect of family functioning on substance use, our findings suggest that family processes, like monitoring, is associated with marijuana use by influencing these interactions with neighborhood best friends that engage in antisocial behavior.

This study showed evidence that neighborhood disadvantage may indirectly influence adolescent boys’ marijuana use by promoting or deterring protective families and peer processes. Previous studies have found that parents living in disadvantaged neighborhoods may increase their monitoring to protect their adolescents from environmental stressors (Fauth, Leventhal, & Brooks-Gunn, 2007). However, this study found that neighborhood disadvantage was not directly related to parental monitoring, indicating that parents did not feel a need to increase monitoring of their adolescent boys due to environmental stressors. This finding is partially in line with the literature that indicates that boys are generally monitored less than girls (Keogh-Clark et al., 2021). In this study, neighborhood disadvantage is related to less social cohesion, which in turn was related to less parental monitoring, lending more support to the idea that neighborhood influence monitoring by facilitating the practice rather than necessitating it. While we are not able to test gender differences in this study, this is a finding that may be unique to adolescent boys. These findings suggest that adolescent boys in disadvantaged neighborhoods may be more exposed to stressors because poverty undermines social cohesion and parental monitoring, placing them at elevated risk for interacting with substance-using peers and using substances themselves.

Interestingly, our study found that living in a disadvantaged neighborhood only influenced parents’ perceptions of social cohesion and not social control. This study’s findings are consistent with previous studies that focused on the role of social cohesion rather than social control or the full measure of collective efficacy in understanding the relationship between neighborhood features, family processes, and adolescent substance use (Duncan et al., 2002; Jackson et al., 2016). While it is unclear if these studies omitted measures of social control or did not present them because they were unrelated to adolescent substance use, our study found that parental perceptions of neighborhood social cohesion, rather than neighborhood social control, was related to parental monitoring, which in turn was associated with having a neighborhood best friend that engaged in antisocial behavior.

While the indirect effect of the full pathway was not statistically significant, marginally significant indirect effects indicated that when parents lived in a disadvantaged neighborhood the reported less social cohesion and when they reported less social cohesion the also reported less parental monitoring. Similarly marginally significant indirect effects indicated that when parents report more social cohesion, they report more parental monitoring and adolescents therefor report that their neighborhood best friends’ engage in less anti-social behavior. The lack of relationship between parent report of social control and adolescents reporting that their neighborhood best friends engages in anti-social behaviors is inconsistent with theories of neighborhood disorganization (Snedker et al., 2009), but similar to previous research indicating a limited protective effect of neighborhood social control (Byrnes et al., 2013; Musick, Seltzer, and Schwartz, 2008) (Stritzel, 2022). Parental perception of social cohesion was associated with adolescent tobacco use, making it is plausible to consider that this is another indirect pathway through which neighborhood disadvantage may lead to adolescent tobacco use.

The differential influence of an adolescent’s neighborhood best friend and peer engaging in anti-social behavior on adolescent substance use, is of note. While many studies have considered the role of deviant peers in the relationship between neighborhood factors and substance use, few have considered the context in which the relationship takes place, or if the person is considered a peer or a best friend. In this study, the anti-social behaviors of neighborhood peers and best friends were related but were not perfectly correlated and the difference between them was statistically significant, indicating these measures were assessing distinct relationships. Notably, when an adolescents’ reported that their neighborhood peers’ engaged in anti-social behavior they also reported more was alcohol use but this was not the case when associating with neighborhood best friends’ that engaged in anti-social behavior. This study showed that those adolescents who had best friends who were engaging in more anti-social behavior tended to use more marijuana, though this pattern was seen for neighborhood friends more broadly. While this difference was not statistically significant, the findings suggest that considering the level of intimacy and context in which peer relationships take place may be necessary for understanding their relationship to adolescent substance use. Youths may, for instance, feel more comfortable using marijuana, a substance that was illegal at the time of this data collection, with people they considered best friends, rather than a peer. The differential relationship between neighborhood best friends and peers’ anti-social behavior and the use of tobacco, marijuana, and alcohol, although not always statically significant, indicates that this may be a fruitful area of inquiry for future research.

Limitations of the Study

The results of this study need to be considered in light of several limitations. First, the internal consistency of the measure of parents perceptions of monitoring at both ages, 12 and 15, was low. The measure of monitoring also did not account for bidirectional processes or other aspects of the child-parent relationship (such as closeness) that may be critical to the concept of monitoring. Additionally, this measure only captured the parent’s perception of monitoring, which may differ from an adolescent’s perception. Secondly, in this study, both marijuana and tobacco use were measured with a single item asking youths how often they had ever participated in using the substance. While controlling for substance use at age 15 ruled out the possibility that the participant used the substance when they were younger and did not engage in the time period of interest, asking about their substance use throughout their life is less specific than asking about it the previous year or 30 days. Additionally, the items assessing alcohol and tobacco use asked participants to report how often they “secretly” participated in substance use. While it could be argued that secret use is inherently more problematic, it is possible that substance use that was done with their parents’ permission was not measured in this item. While power calculation indicated that the study was adequately powered to fit the model (Preacher & Coffman, 2006), the small sample size may have limited our ability to detect small indirect effects. Lastly, respondents were not clustered in a manner that we could use multilevel models to parse out the amount of variance in neighborhood cohesion and control occurring at the neighborhood level versus the individual level. This paper, therefore, is only testing the relationship between parental perceptions of these constructs. Future research should sample more participants per neighborhood, allowing the research to distinguish between neighborhood-level variation and individual-level variation in perceptions of neighborhood factors.

Conclusions

The findings of this study demonstrate plausible pathways through which neighborhood disadvantage in childhood can lead to adolescent marijuana and tobacco use. In line with eco-developmental models, the findings of this study support the importance of understanding the influence of meso systems on interpersonal relationships, such as parental monitoring and associating with peers that engage in antisocial behavior, and their contributions to how adolescents outcome change over time. This study also found some support for the assertion that the role of peers on substance use may differ based on how the adolescent describes the relationship, i.e., a peer or a best friend. Lastly, in line with SET, the findings of this study suggest that parental perceptions of the neighborhood environment may support or deter parenting practices that prevent substance use, and therefore are important points of assessment to inform prevention and intervention efforts. While not conclusive, the findings of this study suggest the need to further investigate the associations between adolescents’ environments and their relationships with peers and parents throughout adolescence to better understand the role of neighborhood environments in substance use behavior.

Highlights.

• Neighborhood disadvantage in late middle childhood is associated with adolescent marijuana use at 17 through parents’ perceptions of social cohesion, parental monitoring, and adolescents’ neighborhood best friend.

• Meso systems (i.e., neighborhood attributes) are directly and indirectly related to youth by influencing people in their ecological systems.

• Controlling for family and peer risk and protective factors in studies seeking to understand the role of neighborhood dynamics in adolescent substance use may be masking its effects.