Peer and Community Influences on Adolescent Substance Use in the Context of Adverse Childhood Experiences

Abstract

Both adverse childhood experiences (ACEs) and peer influences consistently predict early tobacco, alcohol, and illicit drug use. However, less research considers how peer and community influences contribute to or modify the association between ACEs and early substance use. This study addresses these gaps in the literature by analyzing multilevel, longitudinal data from the Project on Human Development in Chicago Neighborhoods (PHDCN; N = 1,912). Unstructured socializing and peer substance use largely explained the association between ACEs and drinking, smoking cigarettes, and illicit drug use in the past month. A history of ACEs magnified the association between peer substance use and the number of cigarettes smoked. Collective efficacy also shaped the associations between peer influences, ACEs, and substance use, but in different ways depending on the substance use outcome analyzed.

Introduction and Theoretical Background

Adverse childhood experiences (ACEs) are consistently associated with a wide range of poor mental (Chapman et al. 2004; Dube et al. 2001; Edwards et al. 2003) and physical health outcomes (Anda et al. 2008; Danese and McEwen 2012; Dong et al. 2003) across the life course. The definition of “adverse child experiences” typically encompasses exposure to psychological, physical, and sexual abuse as well as several dimensions of household stressors, including living with someone with a mental illness or substance use disorder, having a household member incarcerated, and witnessing violence toward one’s mother before age 18 (Felitti et al. 1998). One mechanism through which ACEs affect later health is the development of problematic health behaviors such as smoking, drinking alcohol, and illicit drug use (Hostinar et al. 2015). In particular, ACEs are associated with an earlier onset of substance use during adolescence (Anda et al. 1999; Dube et al. 2003; Dube et al. 2006; Rothman et al. 2008), a critical period in which initiation of long-term use often occurs (Chassin et al. 2004). Prior research clearly demonstrates the consequences of peer and community contexts for adolescents’ substance use, but most studies linking ACEs to youth’s substance use conceptualize these behaviors as an individualistic coping mechanism in response to adversity or as an expression of a genetic influence. Few studies address how the combination of peer influences, neighborhood context, and a history of ACEs shape youth’s substance use. Using multilevel data from the Project on Human Development in Chicago Neighborhoods (PHDCN), this study addresses (1) the extent to which peer influences explain the association between ACEs and substance use, (2) whether youth with a history of ACEs are more or less vulnerable to peer influences, and (3) how neighborhood context further shapes this vulnerability.

Mediating Role of Peer Influences in Substance Use

Given that peer influence is one of the strongest and most consistent predictors of adolescent substance use (Kobus 2003; Wills and Cleary 1999), it is plausible that the association between ACEs and substance use operates in part through peer settings. Two dimensions of peer influences commonly studied are the direct normative influence of peers who engage in substance use and the opportunities for substance use that peer groups provide. In particular, spending time with friends without a structured activity and in the absence of authority figures, that is, unstructured socializing, increases the likelihood of substance use independently of the peers’ own substance use (Osgood et al. 1996). Prior research supports both influence-based and opportunity-based theories of the mechanisms linking peers to adolescent substance use (Haynie and Osgood 2005). ACEs may increase the likelihood that youth spend time in such settings for several reasons.

First and most simply, youth tend to seek out friends with similar characteristics and experiences (Brechwald and Prinstein 2011; Poulin and Boivin 2000). Youth who have more ACEs tend to smoke, drink, and use drugs more on average. They may then select peers who also engage in these behaviors (Ennett and Bauman 1994; B. R. Hoffman et al. 2007). In addition, they may bond with friends with similar adverse experiences who also turn to substance use as a coping mechanism. Either way, greater exposure to friends who use substances can increase the opportunities and enhance the social rewards of substance use. In addition, adolescents with more ACEs express greater admiration and desire to imitate antisocial peers (Perez, Jennings, and Baglivio 2018).

Second, children growing up in unstable household environments are more likely to experience school and neighborhood mobility, exposing youth to new, potentially delinquent, peer networks (Fomby and Sennott 2013). Adolescents with more ACEs may not only be more likely to select substance-using peers, but also may spend more unstructured time with their friends. Unstructured socializing without parental monitoring, such as going to parties, in turn is associated with greater substance use.

Third, adolescents with more ACEs by definition experience greater levels of household stress and dysfunction, reflecting greater strain in the parent-child relationship. Greater family instability (Cavanagh et al. 2018), lower levels of relationship warmth (Hair et al. 2008), and less disclosure from the adolescent (Stattin and Kerr 2000) constrain parents’ ability to monitor adolescents’ behavior (Osgood and Anderson 2004). Lastly, according to conventional social control and attachment theories (e.g., Hirschi 1969), adolescents who feel alienated from their families may spend more time with delinquent peers outside of a supervised setting. Thus, the first aim of this study is to test the hypothesis that peer substance use and unstructured socializing mediate the association between ACEs and substance use (the path labeled “a” in Figure 1).

Figure 1.

Conceptual model.

Moderation of Peer Influences by ACEs

In addition to being more likely to spend time in peer contexts conducive to substance use, youth with ACEs may also be more vulnerable to certain peer influences. In statistical terms, ACEs may moderate the association between peer influences and substance use; this hypothesis is represented by the path labeled “b” in Figure 1. One reason for this may be that youth who have experienced ACEs are more likely to live with parents who use drugs or drink alcohol and thus have greater access to these substances. Children may also inherit a genetic predisposition toward substance use, which makes them more vulnerable to peer influences (Chassin et al. 2013).

In addition, the stress and trauma of ACEs can set in motion the development of maladaptive behavioral patterns and personality traits which can create difficulties in regulating peer influences. For example, in one sample of youth involved in the juvenile justice system, a greater number of ACEs was associated with higher impulsivity, aggression, and admiration of deviant peers (Perez et al. 2018). Experiencing maltreatment also increases the likelihood of developing intrusive or disorganized attachment styles (Cyr et al. 2010; Kendall-Tackett 2002). In two prior studies, for example, Susan D. Hillis and colleagues (2004) hypothesized that the established association between ACEs and risky sexual behavior could be due in part to a desire to seek interpersonal intimacy in the absence of emotional warmth and support at home. Consequently, these youth may experience greater difficulties in saying no to peers and situations that encourage substance use (Agnew 1991). In addition to disrupting the development of healthy attachment styles, negative family experiences and strained household relationships contribute to a lower sense of self-control (Wills and Dishion 2004), lower self-efficacy (Schulenberg et al. 1996), and less autonomy (Allen et al. 2012), all of which limit youth’s ability to set healthy boundaries with peers (Jaccard, Blanton, and Dodge 2005). In other words, at the same time that youth with a greater number of ACEs are more likely to spend times in peer situations conducive to substance use, they may also be more vulnerable to these influences (O’Donnell, Schwab-Stone, and Muyeed 2002). Accordingly, the second aim of this study is to examine whether ACEs magnify youth’s vulnerability to substance use promoting peer influences.

Moderation by Neighborhood Collective Efficacy

Following routine activity theory (Osgood et al. 1996), the larger community context also shapes the opportunities and perceived risks and rewards of substance use. Beyond the sociodemographic composition of a community, contextual characteristic like the aggregate level of parental monitoring (Osgood and Anderson 2004), number of bars and liquor stores (Maimon and Browning 2012), neighborhood norms (Zimmerman and Farrell 2017), and the strength of community ties (Tobler, Komro, and Maldonado-Molina 2009) matter for substance use. To build on prior research examining the contextual effects on delinquency broadly, this study considers how the neighborhood environment shapes the effects of peer influence on youth differentially exposed to ACEs (J. P. Hoffmann 2003).

Collective efficacy, defined as the extent to which the members of the community can monitor and supervise youth and intervene in the presence of risk or physical threat (Sampson, Raudenbush, and Earls 1997), is one such neighborhood characteristic that may impact the strength of peer influences on adolescent substance use (see the pathway labeled “c” in Figure 1). The combination of greater adult supervision in addition to a willingness to intervene can mitigate other environmental or individual risks. For example, David Maimon and Christopher R. Browning (2012) found that the association between alcohol retail outlets and underage drinking was stronger in neighborhoods with low collective efficacy, suggesting that in these environments alcohol is easier and less “expensive” to obtain because youth are less concerned that neighbors will observe and report them. Furthermore, collective efficacy may be particularly protective for youth with a greater number of ACEs: prior research demonstrates that greater collective efficacy attenuated associations between childhood neglect and later externalizing behaviors (Yonas et al. 2010) and between violent victimization and adolescent substance use (Fagan, Wright, and Pinchevsky 2014). In addition to reflecting greater adult supervision and willingness to intervene, collective efficacy is also associated with increased self-efficacy (Dupéré, Leventhal, and Vitaro 2012) and emotional resilience (Jain et al. 2012). Thus, neighborhood collective efficacy may be a source of resilience for youth affected by ACEs, such as improving youth’s ability to decline substance use. In other words, the difference in the strength of the effects of peer influences on substance use by ACE status may not be as large in neighborhoods with greater collective efficacy.

On the other hand, as Christopher R. Browning (2009) points out, most research on social capital (including collective efficacy) tends to assume that social capital facilitates unambiguously positive outcomes. While one aspect of collective efficacy emphasizes the willingness of neighbors to intervene on behalf of the well-being of one another (i.e., informal social control), the other captures trust, cohesion, and reciprocal exchange (i.e., social cohesion). Although both aspects are seemingly positive, the social cohesion aspect may facilitate youth’s unstructured socializing, which in turn increases youth’s risk of substance use. In other words, the social integration inherent in collective efficacy may have countervailing effects on informal social control (Pattillo 1998). If parents perceive the neighborhood to be a protective environment or trust other parents to appropriately supervise, then they may feel comfortable letting their children spend time with peers in unstructured settings (Maimon and Browning 2010), which in turn promotes substance use. In highly integrated social networks, parents may be more accepting of older adolescents in the neighborhood spending time with younger peers, which could provide younger adolescents with models and opportunities for substance use (Harding 2009). For reasons described in the previous section, peer influences may be particularly pronounced for youth who feel alienated from their families. Thus, in higher collective efficacy neighborhoods, the difference in the strength of the relationship between peer influences and substance use by ACE history may be greater. The third aim of this study is to adjudicate between these two hypotheses concerning how collective efficacy shapes the association between peer influences and substance use for youth with varying levels of ACEs.

Data and Method

Data and Analytic Sample

This study used two components of the PHDCN: neighborhood cluster-level data from the Community Survey collected in 1994–1995 and individual data from the Longitudinal Cohort Study collected over three waves in 1994–1997, 1997–1999, and 2000–2001.

Neighborhood clusters consisted of geographically contiguous and sociodemographically homogenous census tracts made up of approximately 8,000 individuals each. For the Community Survey, sampling took place in three stages: city blocks were sampled within each neighborhood cluster, dwelling units were sampled within each block, and then one adult resident was randomly interviewed within each dwelling unit. A total of 8,782 individuals in 343 neighborhood clusters completed interviews regarding several aspects of their neighborhood social environment, with answers aggregated to the neighborhood cluster level.

For the Longitudinal Cohort Study, 80 of the 343 neighborhood clusters were selected for sampling so as to maximize variability in racial/ethnic composition and socioeconomic status. Again, a three-stage sampling design randomly selected block groups within these neighborhood clusters, dwelling units within blocks, and residents with children in seven age groups (within six months of birth, ages 3, 6, 9, 12, 15, and 18). At Wave 1, 6,226 children and their families completed interviews. The sample was limited to youth who were 9, 12, and 15 years old at Wave 1 and approximately 12, 15, and 18 years old at Wave 2 (n = 2,345) to capture the ages when the onset of substance use generally occurs and frequency of use tends to escalate (Chassin et al. 2004). Approximately 85 percent of youth in these cohorts completed Wave 2 interviews. Youth who did not complete the substance use section of the interview, due to attrition or otherwise, were not included in the analysis, resulting in a final analytic sample of 1,912 youth.

Measurement

The independent variable of interest was modeled after Vincent J. Felitti and colleagues’ (1998) ACE Study questionnaire. Based on the availability of questions in the PHDCN, an index of adverse childhood experiences included seven categories derived from primary caregiver reports at Wave 1. First, verbal abuse referred to whether or not the primary caregiver ever insulted or swore at the child, threatened to hit or throw something at the child, or threw, smashed, hit, or kicked something during an argument with the child. Second, physical abuse represented whether or not they ever threw something at the child; pushed, grabbed, or shoved the child; slapped or spanked the child; kicked, bit, or hit child with a fist; hit or tried to hit the child with an object; beat the child up; or burned or scalded the child. Third, maternal abuse was defined as whether the primary caregiver reported that her partner or spouse ever threatened her with or actually enacted physical violence toward her. Fourth, household substance abuse was based on reports of anyone in the household experiencing legal, family, or health problems due to substance use. Fifth, household member mental illness referred to whether anyone in the household ever suffered from severe depression, had frequent legal or disciplinary problems, had problems with their nerves or suffered a nervous breakdown, or ever attempted or committed suicide. The last two items included incarceration of a household member and experiencing parental divorce or separation.

Note that the resulting index ranges from 0 to 7 and represents distinct types of adversity as in past studies (Dong et al. 2004), rather than chronicity or severity. Missingness for each item was generally low (less than 5 percent; see Table 1), except for parental divorce, which was unavailable or children who lived with a primary caregiver other than a parent. Youth with missing data on some but not all items were still included in the analysis, thus their ACE score was a conservative estimate of their actual number of ACEs.

Table 1.

Sample Description (N = 1,912).

Variables	Frequency or M (SD)	Percent missing
Individual variables
Adverse childhood experiences
Any verbal abuse	70.14%	0.00
Any physical abuse	67.47%	0.00
Family/household member substance use	44.71%	1.15
Divorce	17.63%	16.63
Family/household mental illness	43.70%	1.26
Any maternal abuse	32.11%	0.00
Family/household member incarceration	7.96%	4.03
Overall adverse childhood experience index	2.79 (1.55)	0.00
Peer substance use	0.0004 (0.79)	3.40
Unstructured socializing	0.002 (0.87)	0.31
Number of days drank in past month	0.57 (2.08)	7.17
Number of days used drugs in past month	0.67 (3.18)	5.70
Number of cigarettes smoked in past month	10.85 (57.28)	7.64
Female	49.53%	0.00
Race
White	14.91%	0.05
Hispanic	46.94%	0.05
Black	34.43%	0.05
Other	3.72%	0.05
Age cohort
Cohort 9	35.30%	0.00
Cohort 12	35.67%	0.00
Cohort 15	29.03%	0.00
Age of primary caregiver	37.51 (6.32)	6.54
Household socioeconomic status	−0.10 (1.42)	0.78
Family size	5.35 (2.02)	2.30
Neighborhood variables
Collective efficacy	0.02 (1.00)	0.00
Concentrated disadvantage	−0.02 (1.00)	0.00
Immigrant concentration	−0.005 (1.00)	0.00
Residential stability	0.02 (1.00)	0.00
Crime	−0.03 (1.00)	0.00
N	1,912

Youth reported their substance use at Wave 2. Number of cigarettes smoked in the past month ranged from 0 to 638 cigarettes and was created by multiplying the average number of cigarettes smoked per day (less than one, 1–5, 6–15, 16–25, or more than 25) and the number of days smoked in the past month (never, 1–2, 3–5, 6–9, 10–14, 15–20, or more than 20). Before multiplying, categories were recoded to their midpoint (less than one cigarette was recoded to zero cigarettes), and the resulting product was rounded to the nearest integer. Number of days drank in the past month was based on two questions asking if youth ever had a drink of beer, wine, wine cooler, or hard liquor (not including sips or tastes), and if so, how many days have they drank in the past 30 days (never, 1–2, 3–5, 6–9, 10–14, 15–20, or more than 20). As with number of cigarettes, categories were recoded to their midpoint and rounded up; the final variable ranged from 0 to 21. Number of illicit drug use days in the past month was created by summing the responses to how frequently (never, 1–2, 3–5, 6–9, 10–14, 15–20, or more than 20) youth used each of 11 drugs in the past month (marijuana, cocaine, crack, inhalants, hallucinogens, heroin, barbiturates, tranquilizers, amphetamines, steroids, or intravenous drugs). Again, categories were recoded to their midpoint and the sum was rounded to the nearest integer; the final variable ranged from 0 to 33.

Unstructured socializing at Wave 2 was based on the mean response gauging the frequency of five activities (Maimon and Browning 2010): riding in a car or motorcycle for fun, hanging out with friends, going to parties, going out after school or at night, and going on a date. Responses were then standardized. Peer substance use represented the mean number of friends a respondent reported using marijuana, drinking, and smoking. Both of these variables were mean-centered.

The PHDCN Scientific Directors created the neighborhood collective efficacy scale by adding and standardizing responses to two five-item Likert scales measuring social cohesion and informal social control (Sampson et al. 1997). For social cohesion, respondents answered the extent to which they agreed (“very strongly agreed, agreed, neither agreed nor disagreed, disagreed, strongly disagreed”) that they lived in a close-knit neighborhood, neighbors were willing to help each other, neighbors got along with each other, neighbors shared the same values, and neighbors could be trusted. To measure informal social control, respondents reported the likelihood (“very likely, likely, neither likely nor unlikely, unlikely, or very unlikely”) that neighbors would intervene if a group of children skipped school and hung out on the street corner, spray-painted graffiti on a local building, was disrespectful to an adult, if a fight broke outside, or if the city threatened to close down a fire station. These two scales were closely associated across neighborhoods (r = .80; p < .001) and thus combined into a single construct.

Individual-level covariates, all measured at Wave 1, included gender, age, race/ethnicity, family size, primary caregiver’s age, and socioeconomic status constructed from the principal component of the highest education level of the parent or partner, household income, and the highest socioeconomic index of the parent or partner’s occupation (Sampson, Morenoff, and Earls 1999). Neighborhood-level covariates included four variables created by the PHDCN Scientific Directors using data from the 1990 Census as well as from the Chicago Police Department: violent crime, concentrated disadvantage (the first principal component of the percentage of families receiving public assistance, unemployed individuals, female-headed families with children, and percentage of Black residents; Sampson et al. 1997), residential instability, and immigrant population concentration (Kirk and Papachristos 2011; Molnar et al. 2003).

Method

All models were estimated using negative binomial regression. This type of regression is appropriate when the outcome is an over-dispersed count variable in which its variance exceeds its mean. To address the first two aims (mediation and moderation by peer influences), path analysis conducted in Mplus estimated the associations between ACEs, the two peer influence variables, and three substance use outcomes. These models employed robust standard errors to adjust for neighborhood clustering. The Mplus command MODEL INDIRECT tested mediation effects, and interaction terms between peer influences and ACEs tested moderation. For the third aim concerning further moderation by collective efficacy, multilevel random intercepts models were used with individual-level variables and interaction terms specified at the “within” level and neighborhood variables at the “between” level. Iteratively, these models estimated the three-way interaction between the two peer influence variables, collective efficacy, and ACEs for each outcome. All models were estimated in Mplus version 7.31 (Muthén and Muthén 1998–2012) using full-information maximum likelihood procedures to account for the minimal missing data that remained after sample restrictions were applied (Enders and Bandalos 2001).

Results

Sample Description

As displayed in Table 1, ACEs were common among this sample of youth, with youth experiencing an average of about three different types of adversity. The most commonly experienced adverse experiences were verbal and physical abuse, with approximately two-thirds of the sample reporting such experiences. Over two-fifths of the sample reported living with someone with a substance abuse problem or mental illness. About a third of the sample reported abuse of their primary caregiver, 18 percent of youth experienced parental divorce, and about 8 percent experienced the incarceration of a household or family member. For substance use, youth reported a mean of 0.57 days in which they drank and 0.67 days in which they used illicit drugs. Youth smoked an average of 10.85 cigarettes in the past month. However, use varied considerably and the standard deviations of all three variables were quite large relative to their means, thus necessitating the use of negative binomial regression models.

Mediation by Peer Influences

The results displayed in Table 2 address the question, “To what extent do peer influences mediate the association between ACEs and substance use?” All tables show the parameter estimates and can be interpreted as the difference in the logs of expected counts of the outcome variable; exponentiating the coefficients gives the incidence rate ratios. As shown in columns 1, 3, and 5 of Table 2, ACEs were significantly and positively associated with all three substance use outcomes. Each additional ACE was associated with an average increase of 1.12 days drank, 1.21 days used drugs, and 1.51 cigarettes smoked in the past month. Including peer substance use and unstructured socializing as mediators, shown in columns 2, 4, and 6 of Table 2, completely accounted for the observed associations between ACEs and each substance use outcome. These results supported the hypothesis that youth with a higher number of ACEs are more likely to engage in peer groups and settings that increase the frequency of substance use.

Table 2.

Direct and Indirect Paths between Peer Influences, ACEs, and Substance Use (N = 1,912).

	Negative Binomial Regression Coefficients
	Days drank in past month		Days used drugs in past month		Cigarettes smoked in past month
Variables	(1)	(2)	(3)	(4)	(5)	(6)
ACEs	0.116* (0.056)	−0.002 (0.052)	0.194* (0.095)	0.003 (0.102)	0.414*** (0.088)	0.083 (0.092)
Peer substance use		1.049*** (0.145)		1.791*** (0.234)		2.090*** (0.230)
Unstructured socializing		0.816*** (0.102)		1.156*** (0.168)		1.298*** (0.203)
Female	−0.260 (0.190)	0.119 (0.179)	−0.497 (0.311)	−0.003 (0.224)	0.042 (0.345)	−0.002 (0.343)
Race (ref: White)
Hispanic	−0.327 (0.271)	0.023 (0.277)	−0.862 (0.485)	−0.158 (0.408)	−0.930 (0.501)	−0.864* (0.398)
Black	−0.877** (0.268)	−0.460* (0.233)	−0.999* (0.438)	−0.096 (0.396)	−1.847*** (0.459)	−1.675** (0.482)
Other	−1.279** (0.436)	−0.627 (0.342)	−1.861** (0.614)	−0.828 (0.553)	−1.011** (0.389)	−0.811 (0.565)
Age cohort (ref: Cohort 9)
Cohort 12	2.964*** (0.602)	1.742** (0.600)	12.699*** (0.332)	10.495*** (0.340)	4.402*** (0.535)	2.021** (0.589)
Cohort 15	4.902*** (0.588)	2.817*** (0.650)	15.152*** (0.228)	11.943*** (0.399)	7.322*** (0.525)	4.840*** (0.627)
Age of primary caregiver	0.390 (0.203)	0.239 (0.187)	−0.406 (0.215)	−0.462 (0.245)	0.829** (0.314)	0.069 (0.236)
Household SES	0.023 (0.075)	0.065 (0.076)	−0.095 (0.096)	−0.090 (0.117)	−0.292* (0.124)	−0.349* (0.142)
Family size	−0.006 (0.062)	0.032 (0.060)	−0.091 (0.074)	−0.066 (0.074)	−0.240** (0.075)	−0.194** (0.070)
Indirect effects
Peer substance use		0.059*** (0.015)		0.101*** (0.025)		0.116*** (0.026)
Unstructured socializing		0.041*** (0.011)		0.058*** (0.017)		0.066** (0.019)
N	1,912	1,912	1,912	1,912	1,912	1,912

Note. ACEs = adverse childhood experiences; SES = socioeconomic status.

^*p < .05.

^**p < .01.

^***p < .001,

two-tailed significance tests.

Moderation of Peer Influences by ACE History

Table 3 shows results for the second analytical aim, which was to determine whether and how ACEs moderate the association between peer influences and substance use. Looking at the first four columns of Table 3, ACEs did not moderate the associations between either of the peer influence variables and number of days drank or used drugs in the past month. ACEs also did not moderate the association between unstructured socializing and number of cigarettes smoked in the past month. However, ACEs did moderate the association between peer substance use and number of cigarettes smoked in the past month. For ease of interpretation, Figure 2 shows the predicted number of cigarettes as a function of the number of ACEs and level of peer substance use. At average and low levels of peer substance use (defined as the mean level of peer substance use and one standard deviation below the mean, respectively), the predicted number of cigarettes smoked did not differ as a function of ACEs history. At high levels of peer substance use (one standard deviation above the mean), the predicted number of cigarettes smoked in the past month increased tenfold from 3.18 cigarettes at zero ACEs to 31.4 cigarettes at the maximum number of seven ACEs. These results partially supported the hypothesis that ACEs magnify youth’s vulnerability to substance use promoting peer influences—specifically, youth with a greater number of ACEs smoked more when they were around substance-using peers than youth without a history of ACEs.

Table 3.

Moderation of Peer Influences on Substance Use by ACEs (N = 1,912).

	Negative Binomial Regression Coefficients
	Days drank in past month		Days used drugs in past month		Cigarettes smoked in past month
Variables	(1)	(2)	(3)	(4)	(5)	(6)
ACEs	0.062 (0.063)	−0.002 (0.067)	0.053 (0.159)	−0.080 −0.097	0.033 (0.107)	0.079 (0.101)
Peer substance use	1.390*** (0.201)	1.054*** (0.149)	2.032*** (0.405)	1.801*** (0.229)	1.094* (0.514)	2.222*** (0.211)
Unstructured socializing	0.809*** (0.099)	0.818*** (0.219)	1.160*** (0.169)	0.623* (0.303)	1.321*** (0.183)	0.889 (0.485)
Peer substance use × ACEs	−0.109 (0.063)		−0.075 (0.587)		0.374* (0.155)
Unstructured socializing × ACEs		−0.001 (0.061)		0.186 −0.105		0.125 (0.146)
N	1,912	1,912	1,912	1,912	1,912	1,912

Note. Models control for child gender, child race/ethnicity, child age, primary caregiver’s age, household socioeconomic status, and family size. ACEs = adverse childhood experiences.

^*p < .05.

^**p < .01.

^***p < .001,

two-tailed significance tests.

Figure 2.

Predicted number of cigarettes smoked in past month, by number of adverse childhood experiences and level of peer substance use.

Moderation of Peer Influences by Neighborhood Collective Efficacy and ACE History

The third aim was to examine how neighborhood collective efficacy shapes the associations between ACEs, peer influences, and substance use. The three-way interactions between ACEs, collective efficacy, and each of the peer influences were not significant for drug use, thus only the results for number of cigarettes and number of days drank are shown. The results for drug use are available from the author upon request.

Table 4 displays the three-way interactions between ACEs, the two peer influence variables, and neighborhood collective efficacy for number of cigarettes smoked and number of days drank in the past month. For ease of interpretation, Figure 3 shows how the associations (expressed as incidence rate ratios) between the two peer influence variables and number of cigarettes smoked changed based on the number of ACEs and level of neighborhood collective efficacy. In Figure 3, “low” and “high” refer to one standard deviation below and above, respectively, the mean neighborhood collective efficacy score. In low collective efficacy neighborhoods, the association between peer substance use and smoking varied considerably by a youth’s number of ACEs. For youth with no ACEs in these neighborhoods, a one point increase in peer substance use was associated with an average expected increase of 1.3 cigarettes smoked in the past month. For an adolescent with five ACEs in the same neighborhood, a one point increase in peer substance use was associated with an increase of 717 cigarettes smoked in the past month (i.e., a little more than a pack a day). This moderation of peer substance use by ACEs was also present in high collective efficacy neighborhoods, but to a smaller extent. The average expected increase of cigarettes smoked in the past month associated with a one point increase in peer substance use ranged from 5.6 cigarettes for youth with no ACEs to 30.7 cigarettes for youth with five ACEs. The moderation of unstructured socializing by ACEs on smoking did not differ by level of neighborhood collective efficacy.

Table 4.

Moderation of Peer Influences on Smoking and Drinking by ACEs and Neighborhood Collective Efficacy (N = 1,912).

	Negative Binomial Regression Coefficients
	Cigarettes smoked in past month			Days drank in past month
Variables	(1)	(2)	(3)	(1)	(2)	(3)
Adverse childhood experiences	0.083 (0.139)	−0.232 (0.218)	−0.073 (0.147)	−0.004 0.051	0.012 (0.065)	−0.015 (0.062)
Peer substance use	3.206*** (0.341)	1.013 (0.665)	2.818*** (0.620)	1.092*** (0.150)	1.335*** (0.208)	1.158*** (0.138)
Unstructured socializing	1.481*** (0.259)	1.470*** (0.245)	0.359 (0.744)	0.798*** (0.102)	0.818*** (0.105)	0.902*** (0.224)
Collective efficacy	0.284 (0.440)	0.292 (0.425)	0.107 (0.476)	0.075 (0.143)	0.024 (0.240)	0.057 (0.186)
Collective efficacy × ACEs		0.175 (0.124)	0.069 (0.052)		0.069* (0.035)	−0.005 (0.042)
Peer substance use × ACEs		0.787*** (0.210)			−0.053 (0.061)
Peer substance use × collective efficacy		0.738 (0.493)			−0.136 (0.128)
Peer substance use × collective efficacy × ACEs		−0.465** (0.150)			−0.047 (0.048)
Unstructured socializing × ACEs			0.334 (0.275)			−0.020 (0.062)
Unstructured socializing × collective efficacy			0.232 (0.445)			−0.400** (0.145)
Unstructured socializing × collective efficacy × ACEs			−0.041 (0.198)			0.149** (0.053)

Note. Models control for child gender, child race/ethnicity, child age, primary caregiver’s age, household socioeconomic status, and family size at the individual level; and concentrated disadvantage, immigrant concentration, residential stability, and violent crime at the neighborhood level. ACEs = adverse childhood experiences.

^*p < .05.

^**p < .01.

^***p < .001,

two-tailed significance tests.

Figure 3.

Associations between peer influences and number of cigarettes smoked in past month by number of ACEs and neighborhood CE.

Note. ACEs = adverse childhood experiences; CE = collective efficacy.

Turning to number of days drank in Figure 4, the moderation of peer substance use by ACEs did not differ by level of neighborhood collective efficacy. However, ACEs did moderate the effect of unstructured socializing on number of days drank differently based on the level of collective efficacy in the youth’s neighborhood. In low collective efficacy neighborhoods, ACEs moderated the effect of unstructured socializing in an opposite direction than in previous models—an increase in ACEs attenuated the association between unstructured socializing and number of days drank. However, the pattern was reversed in high collective efficacy neighborhoods such that ACEs intensified the association between unstructured socializing and drinking. A one point increase in unstructured socializing was associated with an average expected increase of 1.7 days drank in the past month for youth with no ACEs. For youth with five ACEs, this expected increase per one point in unstructured socializing rose to 3.1 days drank in the past month. These results provided partial support to both hypotheses regarding the interactive effects between ACEs, neighborhood collective efficacy, and peer influences, depending on the specific type of peer influence and substance use outcome examined.

Figure 4.

Associations between peer influences and number of days drank in past month by number of ACEs and neighborhood CE.

Note. ACEs = adverse childhood experiences; CE = collective efficacy.

Discussion and Conclusion

ACEs consistently predict worse mental and physical health outcomes throughout the life course, in part through the development of risky health behaviors. In particular, youth with adverse experiences are more likely to ever drink, smoke, and use drugs (Dube et al. 2006), use these substances in greater quantities (Anda et al. 1999) and to initiate use at earlier ages (Dube et al. 2003). Peer influences, particularly the increased availability and perceived rewards of substance use within certain peer groups, also consistently predict adolescents’ own substance use. Given these findings, more research needs to consider how peers shape substance use behaviors for youth affected by ACEs in particular. This study fills in these gaps in the literature by considering the extent to which peer influences explain the association between ACEs and substance use as well as how ACEs moderate the established relationship between peer influences and substance use. Community contexts also shape opportunities for substance use and the extent to which peer groups can facilitate substance use. Thus, this study also considers how peer influences operate differentially for youth depending on both their history of ACEs and neighborhood context.

As hypothesized, peer substance use and unstructured socializing mediated the association between ACEs and all three substance use outcomes examined. Furthermore, ACEs strengthened the association between peer substance use and number of cigarettes smoked in the past month. Lastly, two three-way interactions between neighborhood collective efficacy, peer influences, and ACEs revealed that the strength of the link between peer influences and substance use differed not only by a history of ACEs but also by community context and the type of substance use examined. Specifically, the strengthening effect of ACEs on the association between peer substance use and smoking was more pronounced in low collective efficacy neighborhoods. The opposite pattern emerged for drinking and unstructured socializing—ACEs attenuated the association between unstructured socializing and drinking in low collective efficacy neighborhoods but strengthened it in high collective efficacy neighborhoods. These results point to three main themes.

First, youth with a history of ACEs were more likely to spend time in peer settings that are associated with increased substance use; specifically, they are more likely to engage in unstructured socializing and have friends who drink, smoke, and use drugs. A major methodological concern in the literature on peer effects on adolescent substance use is the distinction between peer selection and peer influence. In other words, do youth who already drink or smoke seek out peers who engage in similar behaviors (i.e., selection) or do substance-using peers encourage their friends to partake (i.e., influence)? This study cannot precisely disentangle whether youth with a greater number of ACEs simply gravitate toward friends who have similar substance use habits or if their friends’ influence exerts a causal effect. Both mechanisms likely shape adolescents’ risk of substance use (Haynie 2001; Kandel 1978; Krohn et al. 1996; Matsueda and Anderson 1998; Thornberry 1987); a better understanding of both how youth with ACEs select peer groups and how these peers affect them can aid in prevention and outreach efforts. Another limitation of estimating peer effects is that respondents may overestimate how much their friends actually use substances. However, this study supports prior research (e.g., Haynie and Osgood 2005) indicating that both unstructured socializing and peer substance are independently associated with youth’s own substance use, particularly for youth with a history of ACEs. New methodological advances such as the use of intensive longitudinal modeling (e.g., Weerman, Wilcox, and Sullivan 2018) can help to better elucidate the proximate processes linking peer settings and influences to youth’s substance use.

Second, a history of ACEs was associated with greater vulnerability to peer substance use regarding number of cigarettes smoked. This result corresponds to Julie S. Olson and Robert Crosnoe’s (2018) finding that the association between peer drinking and youth’s own drinking was stronger for youth who had binge-drinking parent(s) compared with youth whose parents who had not recently binge drank. Going beyond parental substance use, this study considers multiple types of family experiences that may set the stage for adolescent substance use. The intensification may apply to cigarette smoking in particular because cigarettes are easier to procure (especially among older adolescents who can buy them legally) compared with alcohol or illicit drugs. Smoking may be an attractive and easily available tool for emotional regulation among youth with ACEs who spend time with peers who also smoke.

One limitation of these data is that the ACE index was derived from primary caregiver reports of their own behaviors and household experiences; thus, this measurement likely represents an underreport of ACEs if primary caregivers are reluctant to disclose abusive behavior. In this case, estimates reported in this study are conservative. Another limitation is that this measure of ACEs only captures whether or not youth ever experienced each type of adversity, rather than chronicity and/or severity. Nevertheless, this measure of ACEs improves over many past studies by capturing more temporally proximate experiences, rather than relying on adults’ retrospective reports of their childhoods. Next steps for this line of research include more sophisticated measurement of ACEs, such as explicit consideration of the severity and chronicity of maltreatment, as well as the cumulative and interactive effects of multiple types of trauma. Other research also expands the definition of ACEs to consider exposure to violence in the community, foster care involvement, and other adverse experiences at school and in the neighborhood (e.g., Cronholm et al. 2015; Fagan et al. 2014).

Third, the extent to which peer influences matter for youth’s substance use depended not only on their history of ACEs but also their neighborhood context. Not only do neighborhood characteristics moderate parental influences on adolescent substance use (Zimmerman and Farrell 2017), but this study suggests that they also shape the influence of peers on adolescent substance use, and in nuanced ways depending on an adolescent’s ACEs history and type of substance. Collective efficacy limited the interactive effect between peer substance use and ACEs on smoking. In other words, youth with a high number of ACEs and substance-using peers smoked less in neighborhoods with high collective efficacy compared with their peers with the same number of ACEs and same level of peer substance use but in low collective efficacy neighborhoods. This finding aligns with the hypothesis that collective efficacy can have a protective effect for youth in vulnerable situations and is consistent with Abigail A. Fagan and colleagues’ (2014) finding that violence-exposed youth are less likely to turn to substance use in high collective efficacy neighborhoods compared with their peers in lower collective efficacy neighborhoods. This could be because adults in high collective efficacy neighborhoods are more likely to monitor adolescents and intervene to prevent negative health behaviors like smoking.

However, this pattern did not hold true in the models for drinking. Collective efficacy intensified the interactive effect between ACEs and unstructured socializing on number of days drank in the past month. The magnification of the interaction between ACEs and unstructured socializing on drinking in the context of high neighborhood collective efficacy observed in this study stands in contrast to David Maimon and Christopher R. Browning’s (2010) findings, in which greater collective efficacy attenuated the association between unstructured socializing and violent behavior. Consistent with routine activity and negotiated coexistence theory (Browning 2009), these findings reflect the idea that the deviant behavior may flourish in what may typically be thought of as prosocial environments (Osgood et al. 1996). Specifically, the greater trust among adults and youth may actually facilitate adolescent drinking.

These findings raise the question as to why collective efficacy operated so differently for the two different substance use outcomes. The effect of collective efficacy on youth’s substance use may depend on local norms among adults (Ahern et al. 2009) whose intervention can either facilitate or prevent adolescent substance use. Most parents, regardless of socioeconomic status, likely do not approve of adolescent smoking. However, many parents may approve of youth drinking in moderation, particularly in wealthier communities (Snedker, Herting, and Walton 2009). One study found that collective efficacy was associated with more drinking for older adolescents (ages 16–19), but less drinking for young adolescents (under age 16) (Jackson et al. 2016). Adults may see it as appropriate for older adolescents to drink and may facilitate this drinking. Furthermore, collective efficacy may reflect the trust among adults in the community that facilitates opportunities for vulnerable youth to engage with substance-using peers in unstructured, unsupervised settings. A previous study demonstrates that youth’s own perception of safety is also positively associated with binge-drinking initiation (Tucker et al. 2013). The combination of high collective efficacy and high unstructured socializing which predicts more drinking among youth with ACEs may also reflect youth’s own perceptions of neighborhood safety. As the authors of the study note, youth’s perception of safety likely refers to safety from being punished or stopped by adults, rather than from neighborhood crime or violence. In high collective efficacy neighborhoods, adults are willing to intervene to stop adolescents from engaging in unhealthy or unsafe behavior—but they might not perceive drinking as such.

Another limitation of this study is that unstructured socializing and substance use may not necessarily occur within the bounds of an adolescent’s neighborhood. Collective efficacy may be confounded with some other characteristic of a youth’s school or home environment; however, the analysis controls for several neighborhood-level covariates such as crime, concentrated disadvantage, and residential instability to minimize this possibility. In addition, collective efficacy was measured three years prior to youth’s substance use and peer settings and the data do not capture the characteristics of a different neighborhood if youth moved between waves. However, prior research indicates that the social environment of a neighborhood generally does not change drastically in a short period of time, nor do families usually move into significantly different types of neighborhoods (Sampson 2011). Future research using new methodological tools such as ecological momentary assessment (e.g., Roberts et al. 2019) and theoretical concepts such as activity spaces and ecological networks (e.g., Browning, Soller, and Jackson 2015) can clarify how place matters for youth’s substance use.

It is important to note that the PHDCN sample is not nationally representative and is disproportionately made up of non-white adolescents living in urban areas. Although this group may experience more ACEs than the general population, ACEs are common across all socioeconomic groups (Merrick et al. 2018). Black and Hispanic adolescents are less likely to smoke cigarettes, use illicit drugs, and drink heavily compared with white adolescents (although these differences have narrowed in recent years; Johnston et al. 2017). Furthermore, the association between economic deprivation and substance use is weaker for Black and Hispanic youth (Bachman et al. 2011). Thus, these results may be an underestimate of the associations between ACEs, peer influences, neighborhood context, and substance use. A major advantage of using the PHDCN is its rich data on neighborhood contexts that is not available with nationally representative data sets like Monitoring the Future. Nevertheless, future national data collection efforts regarding adolescent substance use may benefit from questions regarding youth’s neighborhood context. More research should also pay attention to adolescents in rural areas, particularly those most affected by the opioid epidemic.

Linking insights from criminology, sociology, and developmental psychology, this study demonstrates the relevance of examining family and peer contexts as crucial influences on adolescent health behaviors that may persist into adulthood. It also offers a start to the examination of the neighborhood contexts of substance use trajectories for youth affected by ACEs. In sum, peer settings explained a large portion of the association between ACEs and adolescent substance use; additionally, youth with a history of ACEs were more vulnerable to problematic peer influences with regard to cigarette smoking. Last, collective efficacy further modified the interactive effect of ACEs with peer influences, but in different ways depending on the type of substance use examined. These results invite further investigation into the proximate school, peer, and community factors influencing the substance use behaviors of youth affected by ACEs.

Author Biography

Haley Stritzel received her PhD in Sociology from the University of Texas at Austin and is a current post-doctoral scholar at the University of North Carolina at Chapel Hill. She studies the family and community contexts of youth well-being, with a particular emphasis on children involved with the child welfare system. Her current research focuses on foster children affected by parental substance use and kinship caregivers.