Journal of Youth and Adolescence

, Volume 38, Issue 3, pp 454–465

Trajectories of Peer Social Influences as Long-term Predictors of Drug Use from Early Through Late Adolescence

Authors

    • Institute for Health Promotion & Disease Prevention ResearchUniversity of Southern California
  • Chih-Ping Chou
    • Institute for Health Promotion & Disease Prevention ResearchUniversity of Southern California
  • Valentina A. Andreeva
    • Institute for Health Promotion & Disease Prevention ResearchUniversity of Southern California
  • Mary Ann Pentz
    • Institute for Health Promotion & Disease Prevention ResearchUniversity of Southern California
Empirical Research

DOI: 10.1007/s10964-008-9310-y

Cite this article as:
Duan, L., Chou, C., Andreeva, V.A. et al. J Youth Adolescence (2009) 38: 454. doi:10.1007/s10964-008-9310-y

Abstract

The present study analyzed the long-term effects of perceived friend use and perceived peer use on adolescents’ own cigarette, alcohol and marijuana use as a series of parallel growth curves that were estimated in two developmental pieces, representing middle and high school (N = 1,040). Data were drawn from a large drug abuse prevention trial, the Midwestern Prevention Project (MPP). Results showed that both perceived peer and friend cigarette use predicted own cigarette use within and across the adolescent years. For own alcohol and marijuana use, peer and friend influences were limited primarily to middle school. The findings suggest that strategies for counteracting peer and friend influences should receive early emphasis in prevention programs that are targeted to middle school. The findings also raise the question of whether cigarette use may represent a symbol of peer group identity that is unlike other drug use, and once formed, may have lasting adverse effects through the adolescent years.

Keywords

PeersAdolescenceDrug useSmokingGrowth curveNormsPiecewise

Introduction

Despite the achievements of many intervention programs that were designed to prevent or decrease adolescent substance use, recent data indicate that such behaviors remain widespread among American youth (Johnston et al. 2006). For example, findings from the 2005 National Youth Risk Behavior Survey documented that over 40% of the U.S. high school students reported drinking alcohol, 23% reported smoking cigarettes, and over 20% reported using marijuana during the previous 30 days alone (Eaton et al. 2006). Even as early as eighth grade, more than a quarter of the students have tried cigarettes, and around half have used alcohol (Johnston et al. 2006). Generally, by the end of high school, more than half of American young people have tried cigarettes, and nearly 80% have consumed alcohol. A common pattern observed in the consumption of cigarettes, alcohol, and marijuana is that it occurs gradually in elementary and middle school, accelerates in late middle and high school, continues to increase during young adulthood, and gradually stabilizes or decreases in the mid-emerging adulthood (Johnston et al. 2006). The use of these drugs in adolescence has been shown to have a significant temporal relationship to use of other drugs later on, and may have a mediator effect on use of drugs such as cocaine use (Pentz and Li 2002). Thus, the term gateway drug has been adopted for reference to the importance of cigarettes, alcohol, and marijuana in predicting later adolescent drug use (Tarter et al. 2006).

Peer Influences

Peer influences are arguably the most significant psychosocial risk factor for adolescent experimentation with drugs (Conrad et al. 1992; Hawkins et al. 1992). These influences include both the perceived influence of peers in general, often referred to as perceived social norms for drug use, as well as perceived use by one’s own friends (Hawkins et al. 1992; Hoffman et al. 2006). Both of these constructs are widely used to represent peer influences in substance use studies with adolescents.

Peer influences have figured prominently in many theories that attempt to explain the development of youth problem behaviors, including drug use (Hoffman et al. 2006), and have been a major feature of effective drug prevention programs (Cuijpers 2002; Gottfredson and Wilson 2003). As a theoretical construct, peer influences are assumed to operate through behavioral modeling and/or through the environment represented by the peer group. For example, Social Cognitive Theory (SCT) (Bandura 1972) emphasizes the reciprocal association among personal/behavioral, social and environmental factors and highlights the importance of modeling the behavior of similar others. In the case of smoking cigarettes, SCT suggests that adolescents’ smoking is partly a product of direct modeling of peer or teacher smoking at school or of a family member’s smoking behavior at home. Cognitive Developmental Theory (CDT) (Bush and Iannotti 1985) proposes that the actual environment is not as important as the youth’s perception of the behavior in that environment. Finally, Social Selection theorists argue that social attachment to peers who are using substances may cause the onset or increase of substance use among adolescents and may also explain adolescents’ self-selection into certain peer groups (Ennett and Bauman 1994; Fisher and Bauman 1988). The Social Development Model reinterprets this type of influence as bonding to deviant peers (Catalano et al. 1996).

It is not clear from these theories nor related research studies whether an adolescent’s immediate friends have a greater or lesser influence than the perceived norms of their peer group overall (Bot et al. 2005; MacKinnon et al. 1991). For example, some studies have supported CDT; namely, that friends’ perceived substance use exerts a stronger influence than their actual use on adolescents’ own substance use (Bauman and Fisher 1986; Wilks et al. 1989). In contrast, others have observed that actual friends’ smoking might have the same influence on adolescent smoking as do perceived norms (Urberg et al. 1990). Most of the relevant research has assessed either friends’ use or perceived peer norms alone. In the few studies where both constructs were measured simultaneously, actual friends’ use was less influential on substance use when compared to perceived peer norms (Fisher and Bauman 1988; Iannotti and Bush 1992; Wilks et al. 1989). To complicate these findings, more recent longitudinal studies suggest that adolescents’ perception of peer use was not related to subsequent changes in drug use but was predicted by previous drug use (Farrell and Danish 1993; Iannotti et al. 1996).

Overall, such divergent findings leave unclear which peer influences affect the course of adolescent drug use, and how they affect this course. Furthermore, the strength of peer influences, whether measured as perceived friend use or perceived peer use, may differ according to gender and/or ethnicity. For example, females may be more susceptible than males to friend’s influence on tobacco and alcohol use (Dick et al. 2007; White et al. 2002), and whites are more susceptible than African-American adolescents to perceived peer use of tobacco (Hoffman et al. 2006).

Normative school transitions may further contribute to increases in adolescent drug use risk, although the relationship between transition and strength of peer influence is not clear (Costa et al. 1999; Duncan et al. 1994). For example, the transition from middle to high school has been associated with increased engagement with peers and decreased self-esteem and school engagement, all of which are risk factors for drug use and other problem behaviors (Seidman and French 2004). Developmental patterns of tobacco use and other health risk behaviors in general increase after school transition (Graber and Brooks-Gunn 1999; Maggs et al. 1997), as does problem drinking (Guo et al. 2000; Windle et al. 2008). Whether the transition from middle to high school causes a steeper increase in adolescent drug use is not clear. However, theories such as Problem Behavior Theory (Donovan and Jessor 1985) and a few research studies suggest that there may also be a “cliff” effect of step-wise shift in increase that occurs with the high school transition (Pentz 1994). The cliff effect has been hypothesized as being driven by “transition proneness,” (Donovan and Jessor 1985), which makes adolescents more vulnerable to influences of older peers.

Whether peer influences and adolescents’ own use have reciprocal effects over time remains unclear, particularly over the school transition periods occurring during the adolescent years. For the most part, previous studies have either used cross-sectional analyses, or fixed point-to-point longitudinal analyses such as regression to examine the peer influence-drug use relationship (Hoffman et al. 2006).

Growth Curve Analyses

Growth curve models (GCM) may provide a means for a more comprehensive understanding of the developmental trajectories of drug use and its risk factors than conventional approaches such as regression analyses. The latter typically produces change scores which have been criticized for their inefficiency in reflecting the dynamic process of growth (Rogosa et al. 1982; Rogosa and Willett 1985; Willett 1988). Alternatively, GCM rely on initial status and growth slope as latent factors and are capable of accounting for the dynamic growth profile of individuals and examining the factors that affect these growth patterns. Several studies have already applied GCM to the understanding of risk factors for adolescent drug use. For example, Hoffmann et al. (2000) found a significant relationship between growth in stress and growth in drug use during adolescence. Using growth mixture modeling, Soldz and Cui (2002) found that the strength and types of predictors of adolescent smoking depended on the category of smoking examined over time (e.g., non-smoker versus occasional smoker). Riggs et al. (2007) also applied growth mixture modeling to adolescent smoking and found that weekly smoking by age 11, particularly for smoking two or more cigarettes per week, was sufficient to predict nicotine dependence by early adulthood.

Recent developments in GCM allow for incorporating several growth models, each with unique initial status (intercept) and growth rate corresponding to a specific timeframe within the entire follow-up period (Crawford et al. 2003). This strategy is referred to as piecewise GCM. Thus, one can evaluate whether an adolescent’s behavior varies from one developmental transition to another, as well as whether growth occurs overall. For example, Chou et al. (2004) examined growth in adolescent smoking over different developmental transition periods and found that the intervention group had a lower growth rate than the control group in middle school, and that intervention differences were maintained as an intercept difference in high school. It is also possible to estimate the growth in risk factors and growth in drug use simultaneously in a piecewise model. For example, Crawford et al. (2003) modeled sensation seeking as a developmental trajectory of drug use risk that predicted developmental trajectories of students’ own drug use during the transition from middle school to high school. However, this type of strategy has not been applied to the study of social influences.

Hypotheses

In the present study, we applied a piecewise GCM to study the parallel developmental trajectories of adolescent drug use (cigarette, alcohol and marijuana, respectively) and two peer use influences (perceived friends’ use and perceived peer use) from middle through high school. The study is part of a larger drug abuse prevention trial—the Midwestern Prevention Project (MPP), which was based on the Integrative Transactional Theory (ITT) (Pentz 1999), a psychosocial and ecological theory that integrates principles from the theories described earlier, and assumes that peer influences operate on adolescent drug use through both modeling (perceived friends’ use) and perceived norms (perceived peer use). Based on previous studies of the effect of school transitions on both drug use risk and on increase in peer influence, and on studies suggesting reciprocal effects (Iannotti et al. 1996), we advanced three hypotheses. First, peer influences and drug use would increase in growth both in middle and high school, and that the intercepts for both would be significantly larger in high school compared to middle school. Second, the growth in peer influences would predict growth in drug use within each developmental piece and across school transitions. Third, peer influences would have reciprocal influences on drug use, both within and across pieces, such that peer influences would affect drug use, which in turn, would affect subsequent drug use. An exploratory hypothesis was whether the two types of peer influences (perceived friends’ use and perceived peer use) would differ in their trajectories and in the strength of their predictive relationship to drug use.

Method

Data Source

Participants in the MPP were sixth and seventh graders from eight schools in Kansas City, Missouri, who formed a panel for longitudinal follow-up. Details about the MPP can be found elsewhere (Pentz et al. 1989, 1990). Briefly, the selected schools were demographically matched and then randomly assigned to an intervention or a delayed intervention control condition. Participation was 96% (N = 1,606) of the eligible students who had active parental and self consent, with protocols approved by the University of Southern California’s Institutional Review Board. Baseline data were collected in the spring of 1984. The first follow-up was conducted six months later, and then annually thereafter through the end of high school, yielding six waves of data. Retention rates were over 90% (Crawford et al. 2003). Previous studies have shown no overall demographic or drug use differences between experimental groups at baseline, and no attrition or attrition by drug use differences between groups (Fan et al. 2002; Pentz et al. 1989). However, consistent with other longitudinal follow-up studies, a greater proportion of baseline drug users were absent at the end of high school follow-up compared to non-users (Crawford et al. 2003).

Since the number of variables of interest exceeded what could reasonably fit in one survey, a sequential item-sampling design was used, whereby each wave had three forms of a survey, two of which included peer influence items. The forms were randomly distributed to the participants at baseline. Thus, one-third of the full sample were excluded from the current study because their form did not include peer influence items. A total of 1,040 students were subsequently included in the analysis, of whom 50% were male, 81.2% were white, 14.4% were black and 4.4% represented other race/ethnicities, and 54.7% were in the seventh (versus sixth) grade. Of these, 12.8, 7.6, and 2.8% reported monthly use of cigarettes, alcohol, and marijuana respectively; and 7.2, 2.6, and 1.3% reported weekly use of cigarettes, alcohol, and marijuana. By 11th/12th grade, weekly use was 25.6, 28.1, and 11.6%, respectively. Monthly use was 32.2, 49.2, and 18.7%, respectively. These latter monthly use rates are comparable to monthly use rates reported nationally for 12th graders in the same year (1990; 29.4, 57.1, 14.0%) and in 2007 (21.6, 44.4, and 18.8%; Johnston et al. 2007).

Measures

For this study, outcome measures were the frequency of weekly use of cigarettes, alcohol, and marijuana, respectively. Weekly use was used as the dependent variable in order to represent a more valid indicator of regular use than lifetime or monthly use. Also, our utilization of weekly use was consistent with findings from a previous study that showed weekly cigarette use in early adolescence to be a significant predictor of nicotine dependence after adolescence (Riggs et al. 2007). In addition, previous studies have shown high internal consistency and high test–retest reliability of weekly with lifetime and monthly use in middle school (Pentz et al. 1989, 1998).

Weekly cigarette use was measured by asking, ‘How many cigarettes have you smoked in the last week?’ Responses ranged from 0 (none) to 5 (>1 pack). The alpha of internal consistency is .84 and 3-week test–retest reliability is .78 according to previously findings (Pentz et al. 1989, 1998).

Weekly alcohol use was measured with ‘How many alcoholic drinks have you had in the last week?’ and the responses ranged from 1 (none) to 5 (more than 14 drinks). One drink was considered as one beer or one glass of wine/wine cooler or one mixed drink/liquor (internal consistency alpha = .86; 3-week test–retest reliability = .53).

Weekly marijuana use was measure by asking ‘How many times have you used marijuana in the last week?’ The responses ranged from 1 (none or once) to 5 (more than 14 times) (internal consistency alpha = .86; 3-week test–retest reliability = .67).

The midpoint from each of these item ranges was then used to provide a meaningful account of weekly cigarettes, alcohol and marijuana use. Because more than 1 pack or more than 20 drinks and more than 20 times had no midpoint, we coded these categories as equal to 21 cigarettes, 15 drinks and 15 times, respectively. Concurrent expired air (CO) samples were also collected at each wave to increase the validity of self-reported cigarette use (Pentz et al. 1989). Similar to other studies, the carbon monoxide level in the expired air sample has been significantly correlated with weekly cigarette smoking in grades 6 and 7 as well as grades 11 and 12 (r = .26 and .58, respectively, p < .01) (Pentz et al. 1998).

Friends’ use was measured by asking how many of the student’s close friends use cigarettes, alcohol or marijuana and responses ranged from 0 (none) to 6 (more than 10) (internal consistency alpha = .83).

Perceived peer use in effect represented perceived peer norms. This construct was assessed for each of the three substances. Each participant was asked, ‘Out of every 100 students your age, how many do you think smoke cigarettes/or drink alcohol/or use marijuana at least once a month?’ Responses were coded on an 11-point scale ranging from 0 (none) to 10 (about 100) (internal consistency alpha = .85). Previous studies have shown these influences together have a high reliability (3-week test–retest alpha = .72) (MacKinnon et al. 1991).

Statistical Analysis

Separate MPP intervention and control group plots of means were used to describe the general trend of adolescents’ substance use and its psychosocial (e.g., peer-related) risk factors: perceived friends’ substance use and perceived peer use. Means were adjusted for gender, ethnicity and grade level. General Linear Models (GLM) were employed in order to obtain adjusted means.

Piecewise GCM was used to assess the parallel growth trajectories of adolescents’ substance use and its peer-related psychosocial risk factors (Li et al. 2001; Muthén and Muthén 1999). In the current study, piecewise GCM included two pieces (stages), one representing the middle school period and the other piece representing the high school period. There were three waves of measurement during middle school (waves 1–3) and four waves of measurement through the high school years (waves 4–7). Distinct initial status and growth rates were estimated for adolescents’ substance use, friends’ perceived substance use and perceived peer use during middle and high school. The equation used to estimate the growth curves can be specified as the following 2-level model (Chou et al. 2004):
$$ y_{{ij}} = S_{{ms}} a_{{ms\_j}} + t_{{ms\_ij}} b_{{ms\_j}} + S_{{hs}} a_{{hs\_j}} + t_{{hs\_ij}} b_{{hs\_j}} + e_{{ij}} \quad {\text{Level-1 model}} $$
where the subscripts ms and hs represent the two pieces pertaining to the middle school and high school periods. For example, \( a_{{ms\_j}} \) and \( b_{{ms\_j}} \) denote initial status and growth rate for middle school. \( S_{{ms}} \) and \( t_{{ms\_ij}} \) versus \( S_{{hs}} \) and \( t_{{hs\_ij}} \) are mutually exclusive; \( S_{{ms}} \) and \( S_{{hs}} \) have the values of either 1 or 0, indicating the corresponding period, while \( t_{{ms\_ij}} \) and \( t_{{hs\_ij}} \) represent the time lag between each two waves. Because waves 1 and 2 were spaced by 6 months and all other waves were distanced annually, the indicators for middle school were fixed at 0, 0.5, and 1 (\( S_{{hs}} \) and \( t_{{hs\_ij}} \) remain 0), and the indicators for high school were fixed at 0, 1, 2, and 3 (\( S_{{ms}} \) and \( t_{{ms\_ij}} \) remain 0).
The level-2 equations are specified as:
$$ b_{{ms\_j}} = \gamma _{{10}} + \gamma _{{11}} GENDER + \gamma _{{12}} WHITE + \gamma _{{13}} PRORGAM + u_{{1j}} $$
$$ a_{{hs\_j}} = \gamma _{{20}} + \gamma _{{21}} GENDER + \gamma _{{22}} WHITE + \gamma _{{23}} PRORGAM + u_{{2j}} $$
$$ b_{{hs\_j}} = \gamma _{{30}} + \gamma _{{31}} GENDER + \gamma _{{32}} WHITE + \gamma _{{33}} PRORGAM + u_{{3j}} $$
where \( \gamma _{{00}}, \)\( \gamma _{{10}},\)\( \gamma _{{20}},\) and \( \gamma _{{30}} \) stand for the adjusted means of initial status and growth rate of the two stages, respectively. GENDER (0 = male, 1 = female), WHITE (1 = white, 0 = non-white), and PROGRAM (1 = intervention group, 0 = control group) are the covariates, and all the other γ associated with them have the same meaning as usual regression slopes.

All parallel GCM were estimated in EQS (Bentler 2004), controlling for MPP program condition, gender and ethnicity. Missing data were handled by a built-in option in EQS, the maximum likelihood estimator based on case-wise covariances (Bentler 2004; Yuan and Bentler 2000).

Results

Descriptive Characteristics

Figures 13 show the plots of means over time for adolescents’ cigarette, alcohol and marijuana use and the corresponding perceived friends’ use and perceived peer norms, respectively. In general, we found an overall increasing trend in the two psychosocial (e.g., peer related) risk factors and in the use of all substances. In addition, adolescents’ perceived peer use (peer norms) for all substances increase most sharply during the transition from middle school to high school. Among the three substances, perceived friends’ use and perceived peer use of alcohol showed the greatest growth rate. Adolescents reported that 40% of their peers used alcohol at the beginning of middle school, which increased to 80% at the end of high school.
https://static-content.springer.com/image/art%3A10.1007%2Fs10964-008-9310-y/MediaObjects/10964_2008_9310_Fig1_HTML.gif
Fig. 1

Growth curves of adolescents’ cigarette use and parallel social influences. Notes: Y-axis for own use = number of cigarettes used in the past week; Y-axis for peer use = number of close friends who use cigarettes; Y-axis for peer norm = perceived percentage of peers who use cigarettes

https://static-content.springer.com/image/art%3A10.1007%2Fs10964-008-9310-y/MediaObjects/10964_2008_9310_Fig2_HTML.gif
Fig. 2

Growth curves of adolescents’ alcohol use and parallel social influences. Notes: Y-axis for own use = number of alcoholic drinks used in the past week; Y-axis for peer use = number of close friends who use alcohol; Y-axis for peer norm = perceived percentage of peers who use alcohol

https://static-content.springer.com/image/art%3A10.1007%2Fs10964-008-9310-y/MediaObjects/10964_2008_9310_Fig3_HTML.gif
Fig. 3

Growth curves of adolescents’ marijuana use and parallel social influences. Notes: Y-axis for own use = number of times of marijuana use in the past week; Y-axis for peer use = number of close friends who use marijuana; Y-axis for peer norm = perceived percentage of peers who use marijuana

We found the largest increase in use related to smoking cigarettes during both middle school and high school. Compared to cigarette use, the level of marijuana use remained relatively stable in middle school, whereas alcohol use increased slightly during high school. Among the three substances, cigarette use increased most sharply during the transition period from middle school to high school, but the growth rate then slowed down through the high school period. Conversely, the adolescents showed an increase of marijuana use during this period.

Table 1 summarizes the relationships between gender, ethnicity, the parallel growth trajectories of adolescents’ substance use, perceived friends’ use and perceived peer norms, and model fit statistics. In general, males had higher intercepts and slopes of drug use than females and higher rates of perceived friends’ use, while females reported higher rates of perceived peer norms. Whites reported higher use, and higher friend use, but lower growth trajectories of overall peer influence. Overall, the three parallel GCMs showed good model fitting despite significant p-values. According to Hu and Bentler (1999), it is difficult to get a non-significant p-value when the sample size is large. In this case, the comparative fit index (CFI) and the root-mean-square error of approximation (RMSEA) were the alternative criteria for estimating model fit. For all drug use models in the current study, the CFI was greater than 0.96 and the RMSEA was no more than 0.035, indicating good fit to the observed data. During middle school, the growth rates for perceived friends’ use of cigarettes and alcohol in the MPP intervention condition were lower than those in the control condition. In other words, the former successfully slowed the growth rate of perceived friends’ use of cigarettes and alcohol and peer norms for marijuana use. During high school, the MPP intervention group reported lower initial levels of students’ own cigarette and alcohol use than the control group. Additionally, during middle school, the MPP changed the growth rate of adolescents’ cigarette use during high school (β = 0.416, standard error = 0.146).
Table 1

Beta (se) estimates of MPP program condition, gender and ethnicity on parallel trajectories and model fit statistics

 

Cigarettes

Alcohol

Marijuana

Own use

Friends’ use

Peer norms

Own use

Friends’ use

Peer norms

Own use

Friends’ use

Peer norms

Middle school

Intercept on

    Gender

−.374 (.124)*

−.081 (.077)

.602 (.165)*

−.039 (.042)

−.319 (.081)*

.148 (.163)

−.049 (.031)

−.122 (.052)*

.381 (.131)*

    White

.053 (.159)

.068 (.098)

.790 (.210)*

.104 (.054)

.236 (.103)*

1.001 (.208)

−.011 (.040)

.007 (.066)

.279 (.167)*

Slope on

    Gender

.270 (.209)

.133 (.084)

−.370 (.192)*

−.107 (.076)

.313 (.113)*

.191 (.169)

−.041 (.045)

.147 (.077)*

.145 (.153)

    White

.390 (.264)

.314 (.107)*

−1.105 (.248)*

.129 (.095)

.465 (.139)*

−.313 (.231)

.051 (.055)

.107 (.095)

−.766 (.190)

High school

Intercept on

    Gender

−.264 (.266)

.029 (.180)

−.184 (.080)*

    White

1.292 (.386)*

.734 (.268)*

.129 (.187)

Slope on

    Gender

.067 (.146)

−.332 (.098)*

−.021 (.048)

    White

.309 (.212)

−.057 (.145)

.072 (.101)

Model fit

χ2

199.201

194.305

223.145

df

70

70

70

p

<.05

<.05

<.05

CFI

0.981

0.971

0.967

RMSEA

0.035

0.034

0.035

Notes: * Significance level p < 0.05. All models adjusted for MPP program group

Figure 4 shows the standardized parameter estimates for the latent GCM of cigarette use and its parallel risk factors. The intercept of perceived peer use and perceived friends’ use of cigarettes positively predicted the growth rate (slope) of adolescents’ own use of cigarettes (standardized β = 0.438 and 0.140, respectively, p < 0.05). These results provide direct support for peer influence on one’s own use over time. The path from the adolescent use intercept to perceived peer norms was also significant (standardized β = 0.180, p < 0.05), providing support for a reciprocal effect between peer influence and one’s own drug use. As expected, the intercept and slope of adolescent use during middle school predicted the intercept of adolescent use during high school. Moreover, the intercept and slope of perceived friends’ use during middle school also had a significant impact on the slope of adolescent’s use during high school (standardized β = 0.418 and 0.483, for friends’ use intercept and slope, respectively; p < 0.05), suggesting that early peer influences extend beyond the immediate context of middle school to affect drug use risk in the high school context.
https://static-content.springer.com/image/art%3A10.1007%2Fs10964-008-9310-y/MediaObjects/10964_2008_9310_Fig4_HTML.gif
Fig. 4

Standardized estimates for the parallel latent growth curve models of friends’ use (fcig) and perceived peer norms (ncig) predicting weekly cigarette (cig) use during middle schools (ms) and high schools (hs) in Kansas City, MO. cig1–cig7, measures of weekly cig use for waves 1–7; fcig1–fcig7, measures of friends’ cigarette use for waves 1–7; ncig1–ncig7, measures of perceived peer use of cigarette for waves 1–7; Intercept, intercept growth factors for weekly cig use in ms and hs and for friends’ cigarette use and perceived peer use of cigarette in ms. Paths for the intercept and slope factors regressed on covariates (MPP program condition, gender and ethnicity) are not shown (see Table 1 for covariate estimates). Correlations among disturbance or errors, or non-significant regression weights also not shown. Factor loadings for indicators were fixed at the value indicated. p < 0.05

The latent GCM of alcohol use and marijuana use are presented in Figs. 5 and 6. We fit the same piecewise linear GCM for friends’ use and perceived norms predicting alcohol and marijuana use in middle school and high school. For alcohol use, the intercept of perceived norms in middle school predicted the slope of adolescent use of alcohol during the same developmental period (standardized β = 0.310, p < 0.05), representing a proximal relationship. However, perceived norms for alcohol use in middle school did not carry over to alcohol use in high school. Similarly, we found that the intercept of perceived norms and perceived friends’ use during middle school influenced the contemporaneous growth rate of adolescent use of marijuana during the middle school years. Initial levels of perceived friends’ use led to a faster increase of adolescents’ use of marijuana during middle school (standardized β = 0.426, p < 0.05). Alternatively, adolescents who reported high perceived peer use in middle school had a relatively slower growth rate of own marijuana use during the same period (standardized β = −0.501, p < 0.05) during middle school.
https://static-content.springer.com/image/art%3A10.1007%2Fs10964-008-9310-y/MediaObjects/10964_2008_9310_Fig5_HTML.gif
Fig. 5

Standardized estimates for the parallel latent growth curve models of friends’ use (falc) and perceived peer norms (nalc) predicting weekly alcohol (alc) use during middle schools (ms) and high schools (hs) in Kansas City, MO. alc1–alc7, measures of weekly alc use for waves 1–7; falc1–falc7, measures of friends’ alcohol use for waves 1–7; nalc1–nalc7, measures of perceived peer use of alcohol for waves 1–7; Intercept, intercept growth factors for weekly alc use in ms and hs and for friends’ alcohol use and perceived peer use of alcohol in ms. Paths for the intercept and slope factors regressed on covariates (MPP program condition, gender and ethnicity) are not shown (see Table 1 for covariate estimates). Correlations among disturbance or errors, or non-significant regression weights also not shown. Factor loadings for indicators were fixed at the value indicated. p < 0.05

https://static-content.springer.com/image/art%3A10.1007%2Fs10964-008-9310-y/MediaObjects/10964_2008_9310_Fig6_HTML.gif
Fig. 6

Standardized estimates for the parallel latent growth curve models of friends’ use (fmaj) and perceived peer norms (nmaj) predicting weekly marijuana (maj) use during middle schools (ms) and high schools (hs) in Kansas City, MO. maj1–maj7, measures of weekly maj use for waves 1–7; fmaj1–fmaj7, measures of friends’ marijuana use for waves 1–7; nmaj1–nmaj7, measures of perceived peer use of marijuana for waves 1–7; Intercept, intercept growth factors for weekly maj use in ms and hs and for friends’ marijuana use and perceived peer use of marijuana in ms. Paths for the intercept and slope factors regressed on covariates (MPP program condition, gender and ethnicity) are not shown (see Table 1 for covariate estimates). Correlations among disturbance or errors, or non-significant regression weights also not shown. Factor loadings for indicators were fixed at the value indicated. p < 0.05

Discussion

In this study, we applied a two-piece GCM to cigarette, alcohol, and marijuana use, respectively, to examine the parallel social (e.g., peer) influences on adolescents’ drug use from middle school through high school. To study the mechanisms of social influence on adolescents’ substance use, we included perceived friends’ use and perceived peer use simultaneously and compared their relationships with adolescents’ own substance use behavior. We found that perceived friends’ use and perceived peer norms increased over time, and such increases were associated with the growth in adolescents’ own drug use within each developmental period (Epstein et al. 2007; Henry et al. 2005; Musher-Eizenman et al. 2003; Stanton et al. 2002). Across developmental periods, however, only perceived friends’ use in middle school influenced the growth rate of cigarette use later on in high school. Supporting the alternative explanation (e.g., that behavior leads to changes in normative expectations rather than the reverse) in the current study we observed that the initial status of cigarette use in middle school influenced changes in perceived peer use within the same developmental period (Farrell and Danish 1993; Iannotti et al. 1996). This reciprocal effect did not appear for alcohol or marijuana use.

Overall, the results supported hypotheses 1 and 2, and partially supported hypothesis 3 in that reciprocal effects were shown for cigarette use, but not alcohol or marijuana use. The pattern of growth for both drug use and peer influences suggest a cliff as well as a growth effect, which supports both Problem Behavior (Donovan and Jessor 1985) and SCT theories (Bandura 1972). The significant relationships of peer influences to drug use support CDT (Bush and Iannotti 1985). The exploratory hypothesis supported that overall, perceived friends’ use was a greater peer influence on drug use than perceived peer norms.

The pattern of results suggests several avenues for developing or refining drug prevention programs in the future. One is that programs might consider tailoring the relative emphasis on a particular drug depending on whether an adolescent is in middle or high school. Interventions for cigarette and alcohol use may apply to adolescents in middle school while interventions for marijuana use may be better targeted in high school. Another route pertains to deliberately addressing the apparently reciprocal relationship between peer influence and one’s own drug use, particularly for cigarette use. Notable is the finding that cigarette use, perhaps more than the other substances, appeared to represent a symbol of identity with a particular peer group that might extend beyond a straightforward modeling influence (Eckert 1983). Prosocial, healthier alternatives to cigarette use as a symbol of identity and bonding could be addressed in prevention programs.

Limitations of this study should be noted. First, data were based on self-reports. However, we also measured carbon monoxide levels coincident with survey administration, which is a common practice used to increase the validity of self-reported drug use, and correlations of CO with self-reported use were significant as noted earlier (Pentz et al. 1989). Second, our study used weekly drug use, which has a somewhat low prevalence in middle and high school. The small variation in alcohol and marijuana use during this period could explain why few paths in the parallel analysis were significant. As a confirmatory analysis, we also analyzed monthly use, which showed patterns of results similar to those reported here. Based on control group values to generate population-based prevalence rates, weekly use of cigarettes, alcohol, and marijuana use increased respectively from 7.2 to 25.6, 2.6 to 28.1, and 1.3 to 11.6%, from 6th/7th grade to 11th/12th grade. Monthly use rates were comparable to those reported in national studies, for both the time periods in which data were collected for this study, as well as for current 2007 adolescent drug use prevalence rates (Johnston et al. 2007). Extending the analyses into early adulthood, when weekly use prevalence would be expected to be higher, represents a future direction for research.

Despite the limitations, this study provides important information about the patterns of perceived friends’ use and perceived peer use over time, which appears to grow and influence growth in adolescents’ own drug use. In particular, peers have particularly strong proximal and distal influences on adolescent cigarette use that extend from middle through high school. Altogether, the findings from this study suggest that drug prevention programs could be tailored even further for high risk adolescents who report high levels of friends’ substance use, and that early peer and own use of cigarettes may indicate a type of social bonding that could be re-directed in a prevention program.

Copyright information

© Springer Science+Business Media, LLC 2008