Current Psychology (2024) 43:3866–3881

https://doi.org/10.1007/s12144-023-04944-z

A meta‑analysis study on peer influence and adolescent substance use

Lara L. Watts1 · Eid Abo Hamza2,3 · Dalia A. Bedewy4,5,6 · Ahmed A. Moustafa1,7,8

Accepted: 30 June 2023 / Published online: 18 July 2023

© The Author(s) 2023

Abstract
The extent to which adolescents are influenced by their peers has been the focus of developmental psychological research
for over 50 years. That research has yielded contradicting evidence and much debate. This study consists of a systematic
review and meta-analysis, with the main aim of quantifying the effect of peer influence on adolescent substance use, as well
as investigation into the factors that moderate this effect. Included studies needed to employ longitudinal designs, provide
the necessary statistics to calculate cross-lagged regression coefficients controlling for target adolescent’s initial substance
use, and comprise participants aged 10–19 years. A search of academic databases and reference lists generated 508 unique
reports, which were screened using Covidence. The final inclusion criteria yielded a total of 99 effect sizes from 27 inde-
pendent studies. A four-level meta-analytic approach with correction to allow the inclusion of multiple effect sizes from
a given study was used to estimate an average effect size. Results revealed a significant effect of peer influence ( β = .147,
p < .001), indicating that adolescents changed their substance use behaviour in accordance with their peers’ perceived or
actual use. Moderation analyses found peer influence effects varied significantly as a function of substance use behaviour
(categorised as alcohol, tobacco, marijuana, or composite substance use) and peer influence measure (perceived vs. actual
peer report); however, no significant effects emerged in the multivariate moderation model simultaneously examining all
five main moderators. These results suggest that adolescent substance use is affected by peer influence processes across
multiple substance use behaviours and both directly and indirectly through perceived norms. This has significant implica-
tions for substance use prevention, including the potential of harnessing peer influence as a positive force and the need to
target misperceptions of substance use.

Keywords Peer influence · Adolescence · Substance use · Meta-analysis

Introduction
* Ahmed A. Moustafa
ahmed.moustafa@bond.edu.au
The extent to which adolescent substance use is influenced
1
School of Psychology, Faculty of Society and Design, Bond by that of their peers has been debated for decades. Although
University, Gold Coast, QLD, Australia adolescent substance use has declined since 2001, many
2
Faculty of Education, Tanta University, Tanta, Egypt young people still engage in substance use behaviours that
3
College of Education, Humanities & Social Sciences, Al Ain
have potentially permanent negative consequences for their
University, Al Ain, UAE physical and psychosocial development (Australian Institute
4
Department of Psychology, College of Humanities
of Health & Welfare, 2022). Especially concerning is the
and Sciences, Ajman University, Ajman, UAE association between substance use and suicidal behaviour,
5
Department of Psychology, Faculty of Education, Tanta
considering both cannabis use and regular tobacco smoking
University, Tanta, Egypt may represent risk factors for the emergence of suicidality in
6
Humanities and Social Sciences Research Center (HSSRC),
adolescents (Serafini et al., 2012). It is therefore important
Ajman University, Ajman, UAE to research and quantify the effect of different processes on
7
Department of Human Anatomy and Physiology,
adolescent substance use behaviour. The aim of this study
the Faculty of Health Sciences, University of Johannesburg, is to synthesise the existing literature on peer influence on
Johannesburg, South Africa adolescent substance use and conduct meta-analyses to
8
Centre for Data Analytics, Bond University, Gold Coast, determine the magnitude of this effect and its moderators.
QLD, Australia

13
Vol:.(1234567890)
Current Psychology (2024) 43:3866–3881 3867

Although previous systematic reviews have investigated the which represents the interactions between the individual’s
effect of peer influence on adolescents, these studies did not microsystems, such as the connections between an adoles-
conduct meta-analyses (e.g., Henneberger et al., 2021; Leung cent’s parents and peers. The mesosystem is encircled by
et al., 2014), only assessed actual peer substance use (e.g., the exosystem, which consists of the larger social systems
Giletta et al., 2021), or were limited to only one substance that operate indirectly with factors within the microsystem
use behaviour (e.g., smoking in Liu et al., 2017; alcohol in but have no direct effect on the individual (e.g., neigh-
Curcio et al., 2012). Additionally, no existing studies have bourhood, school board). Lastly, the macrosystem is the
looked at the moderating effect of perceived vs. actual peer outermost socialisation context which has a cascading
measures, despite evidence suggesting that adolescents tend function on development through the adolescent’s inter-
to be inaccurate when estimating their peers’ substance use actions across all settings. The macrosystem includes cul-
behaviour (Trucco, 2020). The current study is significant in tural values, religion, and laws, and, like the exosystem,
meeting this research gap as it aims to quantify the effect of it operates primarily through the more proximal factors.
peer influence across multiple substance use behaviours as Bronfenbrenner’s Bioecological Model (1979) is a use-
well as investigating any differences between perceived peer ful way to conceptualise adolescent substance use, as
substance use measures and survey data from actual peers. this behaviour does not occur in a vacuum. Rather, it is
both directly and indirectly contributed to by multiple
Theoretical foundations contextual factors with differential salience, with genetic
and neurobiological differences affecting the individual’s
There are several theories that underpin peer influence sensitivity to different contexts (Trucco, 2020).
effects, and it is important to consider substance use behav- Another important theory to consider when studying
iour as having a complex aetiology. Bronfenbrenner’s Bio- peer influence is Social Learning Theory (Bandura, 1977),
ecological Model of Development (1979) suggests that which posits that social environments affect behaviour
adolescent development and health behaviour, including through modelling. It proposes that adolescents develop
substance use behaviour, is shaped by multiple contextual cognitive representations about various attitudes and
factors arranged as socialisation systems surrounding the behaviours through observing influential social referents,
adolescent. Individual characteristics are at the centre of and that these representations are invoked when mak-
this model, and encompass innate factors such as biological ing their own decisions about engaging in those behav-
sex, genetics, and temperament (e.g., inherited differences iours. Bandura (1977) suggested that favourable attitudes
in emotional, attentional, and self-regulation processes; towards substance use are likely to be reinforced when the
Trucco, 2020). These biologically based individual char- adolescent perceives that the role model (a) is rewarded
acteristics are considered the basis of susceptibility to the for enacting those behaviours, (b) is similar to the adoles-
effects of the socialisation contexts. cent, and (c) has greater social status. These factors need
Allen et al. (2022) reinforced the idea that adolescents to be taken into consideration as potential moderators of
are differentially influenced by their peers according to peer influence effects, as they are likely to contribute to
their genetics and traits, such as assertiveness and auton- the degree that an adolescent will change their substance
omy. Interestingly, they deemed peer influence processes use behaviour in accordance with their peer. Integrating
as reflecting an overall adaptive developmental phenom- Social Learning Theory and Bronfenbrenner’s Bioecologi-
enon, despite the potential negative effects, and suggested cal Model is the Social Development Model (Catalano &
that it is the adolescent subculture promoting substance Hawkins, 1996), which posits that there are four contextual
use behaviour that is problematic, rather than peer influ- domains: family, school, peers, and religious and commu-
ence effects themselves. This was supported by the appar- nity institutions. Adolescents develop social bonds across
ent influence processes being both neutral in valence and these domains based on the perceived opportunities and
strongest for adolescents who were assessed as being the rewards for involvement in either antisocial or prosocial
most well-adjusted, contrary to the common assumption that activities. It follows that adolescents who anticipate social
peer influence processes are reflective of poor functioning reward for prosocial behaviours will have a greater likeli-
in the influenced youth (Allen et al., 2022). Bronfenbrenner hood of engaging in these behaviours, as is true for antiso-
(1979) elaborates on the socialisation contexts and classi- cial actions. Additionally, the Social Development Model
fies the concentric circles as the microsystem, mesosystem, suggests that the salience of the socialising agents within
exosystem, and macrosystem. the four domains will change as adolescents mature devel-
The most proximal system to the individual is the opmentally, with parents representing the main socialisa-
microsystem, which encompasses the immediate sociali- tion factor during childhood and early adolescence before
sation factors that affect the youth directly, including shifting to a focus on peers in middle and late adolescence
peers and parents. The next circle is the mesosystem, (Trucco, 2020).

13
3868 Current Psychology (2024) 43:3866–3881

Peer socialisation context Adolescence

Concentrating on the peer context domain in the Social Adolescence is a developmental period characterised by sig-
Development Model, peer influence is thought to oper- nificant psychosocial, cognitive, moral, and physical devel-
ate through direct and indirect socialisation mechanisms. opment (Allen et al., 2012). Research suggests that there is a
As an overarching concept, peer influence is defined as normative increase in deviant behaviour in adolescence, with
the social processes by which people change their atti- substance use typically initiated during this period (Veenstra
tudes and behaviours to conform to that of their friends and Laninga-Wijnen, 2022). The World Health Organisa-
(Barnett et al., 2022; DeLay et al., 2023; Leung et al., tion (2021)defines adolescence as the ages between 10 and
2014). Peer influence processes can have both positive 19 years, with pubertal onset typically beginning by age 9
and negative outcomes, with adolescents describing to 12 years. The transition from childhood to adolescence
socialisation pressure to engage in prosocial behaviours is characterised by an increased focus on peer association
as well as antisocial behaviours (Adikini, 2023; Bartolo and acceptance, which is a shift from parents as the primary
et al., 2023; Trucco, 2020; Zedaker et al., 2023). De la socialisation factor to peers (Leung et al., 2014). Adoles-
Haye et al. (2013) suggest that friendships can have a cents are more likely to seek peer approval and internalise
protective effect if friends do not endorse marijuana using the views of their peers, which, when combined with height-
behaviours. This finding aligns in with the concept of ened sensitivity to social reward and increased engagement
peer norms, which suggests that the perception of peer in novel experiences, promotes conforming to perceived
approval of a behaviour promotes that behaviour (Trucco, group norms.
2020). While the idea of adolescents being convinced to
engage in substance use behaviour by coercive and pres- Aims
suring peers is popular, it is likely that adolescents are
more influenced by support and validation than directly The main aim of this study is to quantify the effect that peer
by pressuring behaviour (Allen et al., 2022). This is one influence has on adolescent substance use. The hypothesis
reason why interventions focused on helping adolescents is that both actual and perceived peer substance use for alco-
resist peer pressure are often not effective; only a small hol, tobacco, marijuana, and composite substance use will
part of peer influence may be active persuasion, while significantly predict change in target adolescents' own sub-
most of the effect is contributed to by perception of group stance use over time. Additionally, we predict that perceived
norms, social acceptance, and status (Leung et al., 2014). substance use will have a greater effect size, as adolescents
Additionally, it is difficult to disentangle peer selection tend to erroneously overestimate their peers’ substance use
and peer socialisation processes. Peer selection is consist- behaviour (Helms et al., 2014). Although we predict that
ent with the theory of homophily; i.e., that individuals peer influence effects will be significant for all substance
choose friends who are closely matched to their own atti- use behaviours, we hypothesise that alcohol and composite
tudes and behaviours (Trucco, 2020). Peer selection pro- substance use will have the largest average effect sizes. This
cesses are also invoked by social identity theory, which is because alcohol is the most normalised substance and has
states that a fundamental aspect of psychosocial identity the highest proportion of participants partaking, with only
development is making judgements about the groups you 34% of high school students aged 14–17 never having con-
belong to. Conversely, peer socialisation describes an sumed alcohol (Australian Institute of Health & Welfare,
individual’s decision to modify their attitudes and behav- 2022; Trucco, 2020). Conversely, the same 2017 survey
iours to adapt to social norms (Trucco, 2020). This dis- found that 82% of adolescents aged 12–17 years had never
tinction represents a challenge in research on peer influ- smoked tobacco, and 16% had used cannabis (Australian
ence, as these distinct processes of peer selection (i.e., Institute of Health & Welfare, 2022). This suggests that alco-
adolescent’s own substance use behaviour promotes selec- hol and composite substance use will have the largest and
tion of friends who use substances) and peer socialisation most robust effect sizes, predominantly due to the greater
(i.e., peer group’s substance use behaviour contributes to proportion of adolescents engaging in that behaviour, with
the adolescent’s use) appear the same in cross-sectional cannabis next and tobacco having the smallest effect. Age is
designs. Longitudinal studies are essential to separate the hypothesised to be a moderating factor, with research sug-
two processes, but it must also be considered that they can gesting both linear and curvilinear effects (Trucco, 2020).
operate together and bi-directionally (Leung et al., 2014). Regarding linear effects, we hypothesise that overall sub-
This results in a reinforcing cycle in which substance stance use will increase with age; however, a curvilinear pat-
using adolescents select peers with similar substance use tern is hypothesised to emerge, where peer influence peaks
levels, which promotes normative expectations about sub- in early adolescence (approximately 12–14 years; Giletta
stance use that influence other adolescents. et al., 2021). It is hypothesised that time lag between waves

13
Current Psychology (2024) 43:3866–3881 3869

in the longitudinal studies will have a moderating effect on the following three groups to capture substance use, ado-
peer influence, with shorter time lags being associated with lescents and peer influence: ‘drug addiction / drug abuse /
larger effect sizes. Lastly, gender is not expected to be a sig- substance abuse / substance addiction’ AND ‘adolescents
nificant moderating factor. This is not necessarily because / teenagers / young adults / teen / youth / student / adoles-
there is not a difference in peer influence as a function of cence’ AND ‘peer influence / peer influence on adolescent
gender, rather that opposite effects are commonly found substance abuse’. The search was restricted to peer reviewed
across studies, so we expect that they will mask any real journals published in English between 1972 and 2022. Addi-
effects that may exist (Leung et al., 2014) (Fig. 1). tional records (n = 20; see Fig. 2) were identified using a
‘snowball’ search of reference lists of included articles and
systematic reviews on similar topics. As seen in Fig. 2, these
Methods searches yielded 508 unique reports that were first screened
for eligibility through examination of titles and abstracts,
Search Procedures before evaluating the full texts for suitability.

This literature review consisted of five stages: (1) research Selection Criteria
question development and operational definitions, (2) data-
base search, (3) study screening and evaluation using Covi- Considering that the aim of this meta-analysis was to quan-
dence, (4) data extraction into IBM SPSS Statistics, and (5) tify the degree that peers influence adolescents’ substance
data analysis with R. Eligible studies were identified through use, only studies that used a prospective longitudinal design
electronic searches of the databases Ovid (PsychInfo and were included. While cross-sectional designs are commonly
PsychArticles) and PubMed. To yield the largest amount of used in studies examining peer influence, it is impossible
returned results, the literature search used keywords from to differentiate between peer selection and peer influence

Fig. 1  Bronfenbrenner’s Bio-

ecological Model (1979)

13
3870 Current Psychology (2024) 43:3866–3881

Fig. 2  PRISMA Flow Diagram

of Records Identification of studies via databases and registers

Records removed before

Identification
Records identified from*: screening:
PubMed (n = 37) Duplicate records removed
Ovid (n = 1320) (n = 869)
Other (n = 20) Records removed for other
reasons (n = 0)

Records screened Records excluded**

(n = 508) (n = 264)
Screening

Reports assessed for

Reports excluded: (n = 217)
eligibility (n = 244)
Wrong outcome (n = 77)
Wrong population (n = 53)
Wrong study design (n = 48)
Wrong intervention (n = 39)
Included

Studies included in review

(n = 27)

effects. Concurrent associations between peer and target incorporating interventions or experimental manipulations
adolescent substance use in cross-sectional designs could were excluded, as were studies examining populations in
be a result of peer influence processes; however, it is pos- clinical settings. This exclusion criterion was chosen because
sible that behavioural similarity preceded the peer relation- intervention programs and atypical contexts may have sig-
ship and contributed to its formation. Longitudinal designs nificant effects on peer influence. Studies were included if
enable the observation of substance use change over time in the outcomes measured substance use as alcohol, tobacco,
accordance with peer use, while controlling for past behav- and/or marijuana, with studies that didn’t differentiate being
iour and potential moderators. Although the processes of classified as ‘composite substance use’. The population was
peer influence are too complex to establish direct causal rela- limited to participants aged between 10 and 19 years, as
tionships, longitudinal survey data can reflect influence as this is the age group defined as adolescence (World Health
it exists in everyday interactions over time, which provides Organisation, 2021). Studies were included if they provided
ecological validity (Giletta et al., 2021). Longitudinal stud- the required statistics to compute three zero-order correla-
ies that collected data from at least two time points, with tions, standardised estimates from linear regression models
both peer and target substance use assessed simultaneously adjusting for previous substance use, or path analyses using
at an initial time point, were included. The decision to take a longitudinal actor-partner interdependence model (APIM).
a cross-lagged regression approach whereby effect sizes are
computed from three correlation coefficients across any two Effect Size Calculation
assessments in a longitudinal study (see Effect Size Calcu-
lation), is consistent with previous meta-analyses in peer For most studies, effect sizes were reported as standard-
influence (e.g., Giletta et al., 2021). Only empirical studies ised regression coefficients calculated using three corre-
collecting quantitative survey data were included, with qual- lations: (1) between Time 1 peer substance use and target
itive methods and reviews excluded. Additionally, studies use (baseline concurrent correlation), (2) between Time 1

13
Current Psychology (2024) 43:3866–3881 3871

target use and Time 2 target use (substance use stability Pustejovsky & Tipton, 2022). Statistical analyses for the
over time), and (3) between Time 1 peer use and Time mean effect size and moderating effects were based on
2 target use (peer influence). The coefficient represents RVE, with 95% confidence intervals (CI) and Satterth-
the standardised regression weight of peer substance use waite approximation used to find the effective degrees
at Time 1 predicting target substance use at Time 2, con- of freedom in the t-statistics.
trolling for the effect of target use at Time 1. This means The combination of RVE and a four-level random-
that this approach provides information about whether effects regression model explicitly accounts for the poten-
the substance use of peers is associated with changes over tial dependencies among the multiple effect sizes (Giletta
time in target adolescents’ substance use behaviour while et al., 2021). Level 1, random sampling variance, is also
accounting for initial concurrent association between included in traditional random-effects meta-analyses and
peer and target youth substance use. Giletta et al. (2021) describes the variation of the observed effect size around
observes that this advantage in controlling for existing the ‘true’ effect size, as a function of sample size. Level
behaviour has led to cross-lagged approaches with longi- 2 represents within-study within-wave variance (e.g.,
tudinal data becoming the most common method to infer effect sizes for both alcohol and tobacco use measured in
peer influence in observational research. In accordance the same study from Time 1 to Time 2). Level 3 reflects
with Giletta et al. (2021), the software R was used to within-study between-wave variance (e.g., effect sizes for
compute effect sizes as standardised regression coeffi- alcohol use measured in the same study from Time 1 to
cients and sampling variances using the following equa- Time 2, Time 2 to Time 3, and Time 1 to Time 3). Lastly,
tion (Becker, 1992, p. 359): Level 4 is between-study variance (e.g., variation between
ryx1 − ryx2 rx1x2 effect sizes from different studies).
β1 = First, the weighted-mean effect size of peer influence
1 − r2 x1x2 effects was estimated from an unconditional four-level ran-
In Becker’s Eq. (1992), β1 signifies the standardised dom-effects model with RVE correction. Heterogeneity in
regression weight of X1 (i.e., peer substance use at Time the effect sizes was examined, with attention to the distribu-
1) predicting Y (i.e., target adolescent’s substance use at tion of the total variance across the four levels. The median
Time 2), while controlling for X2 (i.e., target’s substance sampling variance was used in heterogeneity calculations.
use at Time 1). Next, each moderator (i.e., substance use behaviour, peer
influence measure, mean age at baseline, time lag, and gen-
der) was analysed separately as a predictor in the four-level
Data Analysis model. Finally, a multivariate conditional model analys-
ing the effects of the main five moderators simultaneously
Of the 27 studies included in the meta-analysis (see was conducted. This was to account for possible correla-
Table 1), 16 studies (60%) could extract multiple effect tion between moderators (e.g., older adolescents reporting
sizes, so a multilevel approach with the robust variance greater substance use) masking effects or yielding illegiti-
estimation (RVE) technique was used (Hedges et al., mate results.
2010; Tipton, 2015). Multiple effect sizes could be cal- The meta-analyses were conducted in R for Mac (Version
culated either in studies that used multiple time points 4.2.2; R Core Team, 2022). The metafor package (Viech-
(i.e., cross-lagged correlation coefficient for correlations tbauer, 2010) was used to estimate multilevel metaregression
between Time 1 to Time 2, Time 2 to Time 3, and Time models, using the restricted maximum likelihood method
1 to Time 3) or multiple substance use behaviours (i.e., (REML) with effect sizes weighted by inverse sampling
separate measures for alcohol and tobacco). This means variance (Giletta et al., 2021). RVE correction was applied
that these effect sizes are nested within the studies and using the packages clubSandwich (Pustejovsky, 2018) and
are not independent, thus violating assumptions of tradi- robumeta (Fisher et al., 2017).
tional meta-analytic approaches (i.e., random effects and Publication bias was examined in two ways: visual
fixed effects models). RVE provides a way to maximise inspection of a funnel plot and moderation analyses of the
data in a meta-analysis when the assumption of inde- inverse sampling variance. Funnel plots are a scatterplot
pendence is violated by correcting the standard errors with the inverse variance on the y-axis and the effect size
(Pustejovsky & Tipton, 2022). Several meta-analyses on the x-axis. In the absence of publication bias, the stud-
have found that the RVE method applied a posteriori ies form a symmetrical funnel shape, with the assumption
yields confidence intervals with acceptable Type I error that studies will scatter centrally around the total overall
rates even if the form of the dependence is unknown estimated effect. Asymmetry in the funnel plot suggests
(Fernández-Castilla et al., 2021; Giletta et al., 2021; that studies with non-significant or negative results were
not published, although it could also indicate inadequate

13
 Table 1  Overview of the Studies Included in this Meta-Analytic Study
3872

Study ID Country Number of % Male Age Range at Baseline Mean Age Substance Use Peer Behaviour Measure Num- Time Lag Number of
Participants at Baseline ber of (months) Effect Sizes

13
(years) Waves

Allen et al., 2012 USA 147 47 13–15 years 14.29 Composite Peer Reported 2 12 1
Allen et al., 2014 USA 179 47 13–15 years 14.29 Alcohol Peer Reported 2 96 1
Allen et al., 2022 USA 157 46.7 Grade 7–8 15.2 Composite Peer Reported 2 12 1
Burk et al., 2011 USA 362 47.5 Grade 10 16 Alcohol Perceived 2 12 1
Cin et al., 2009 USA 6522 - 10–14 years 12.05 Alcohol Perceived 2 8 2
Cleveland et al., 2012
PROSPER Sample USA 7108 49 Grade 6 12.3 Alcohol Perceived 5 12 4
ASAPS Sample USA 5635 44 Grades 7 & 9 12.5 Alcohol Perceived 5 12 4
Curran et al., 1997 USA 363 52 10.5–15.5 years Alcohol Perceived 3 12 3
Deutsch et al., 2015 USA 1192 50 Grades 7–11 15.71 Alcohol, Tobacco & Peer Reported & Per- 2 12 6
Marijuana ceived
Erickson et al., 2000 USA 1503 41.7 Grades 9–12 15.43 Composite Peer Reported 2 12 1
Farrell & Danish, 1993 USA 1256 - Grade 7 Composite Perceived 3 6 12
Halliday-Scher, 2000 USA 290 - Grades 6–12 13.79 Alcohol & Tobacco Peer Reported 2 12 2
Hiatt et al., 2017 Sweden 350 44 Grades 7–12 13.94 Alcohol Peer Reported 2 12 1
Kandel, 1978 USA 783 - - - Marijuana Peer Reported 2 6 1
Larsen et al., 2010 Netherlands 433 44.6 11–15 years 12.3 Alcohol Peer Reported 3 6 3
Li et al., 2017
Sample 1 China 614 51 Grades 7–9 13.3 Tobacco Peer Reported 3 12 6
Sample 2 China 567 47 Grades 10–12 16.7 Tobacco Peer Reported 3 12 6
Madon et al., 2003 USA 505 53.9 Grade 7 - Alcohol Perceived 2 18 1
Mason et al., 2011
Washington State USA 961 49 Grade 7 13.09 Alcohol Perceived 3 12 5
Victoria Australia 984 49 Grade 7 12.93 Alcohol Perceived 3 12 5
Medler, 2000 USA 376 46 - 13.16 Tobacco Peer Reported 3 12 6
Mercken et al., 2007 Netherlands 1763 49.9 Grade 7 12.7 Tobacco Peer Reported 2 12 2
Mounts, 2002 USA 300 40 Grade 9 14.5 Composite Peer Reported 2 6 1
Mrug et al., 2011 USA 126 47 Grades 10–12 15.5 Alcohol & Tobacco Peer Reported 3 12 6
Ouellette et al., 1999 USA 357 48.8 - 15 Alcohol Perceived 2 48 4
Pilgrim, 1998
European American USA 465 44.3 Grades 7, 9 & 11 13.9 Composite Peer Reported 2 12 2
African American USA 175 36 Grades 7, 9 & 11 13.9 Composite Peer Reported 2 12 2
Pomery et al., 2005 USA 225 45 Grade 5 10.5 Composite Perceived 2 24 1
Schelleman-Offermans Netherlands 1286 50.2 13–16 13.6 Alcohol Perceived 3 12 2
2013
Current Psychology (2024) 43:3866–3881

Stice et al., 1998 USA 216 51.9 12–17 14.2 Alcohol Perceived 2 12 1
 Current Psychology (2024) 43:3866–3881 3873

screening in the systematic review or reporting bias. As

(months) Effect Sizes
Time Lag Number of visual inspection might be subjective, it is not considered
a reliable method of estimating publication bias (Liu et al.,
2017). As a more robust measure, moderation analyses of

3
3
the inverse sampling variance were conducted. Publica-
12 tion bias is suggested by a significant association between
12 inverse variance and effect sizes, with the assumption that
small effect sizes with small inverse variances are less
Waves

likely to be reported. If publication bias is detected, the

ber of
Peer Behaviour Measure Num-

validity of the current study’s results is threatened, as the

3
3

results of the meta-analysis may not represent the reality

of peer influence processes.

Results
Perceived
Perceived

Sample Characteristics

This systematic review identified 27 independent stud-

ies, yielding 99 effect sizes, dating from 1978 to 2022
(Mdn = 2008). Included studies were conducted in five
Substance Use

different countries, most in the USA (n = 22), followed by

Composite
Composite

the Netherlands (n = 3), Sweden (n = 1), Australia (n = 1)

and China (n = 1). A total number of 28,325 participants
(M = 44% boys) were included in the meta-analysis, with
sample sizes differing from 126 to 7,108 participants
at Baseline
Number of % Male Age Range at Baseline Mean Age

across studies. Participants’ mean age at baseline ranged

(years)

from 10.50 to 16 years (M = 13.73, SD = 1.41). The average

12.4
11.5

length of time between follow-up assessment in consecu-

tive waves of the longitudinal studies was 15.25 months
(SD = 16.45, Mdn = 12 months, range = 6–96 months),
with a majority of studies assessing peer influence over
a 12-month period (68.8%). Substance use behaviour was
classified according to the substance investigated, which
Grade 7
Grade 6

comprised alcohol (n = 43), tobacco (n = 26) and mari-

juana (n = 3). Studies that did not distinguish between dif-
ferent substances (e.g., asked participants how often they
used drugs and alcohol), were classified as composite sub-
53
50

stance use (n = 27). One of the aims of this meta-analysis

Participants

was to quantify any significant differences between studies

that used the target adolescent’s perception of their peers’
1190
1277

substance use and studies that used survey data from

actual peers. Across the 99 effect sizes included, 54 used
perceived measures of peer substance use (54.5%) and 45
Country

used actual peer measures (45.5%). See Appendix A for

USA
USA

a summary of the main descriptive characteristics of the

studies included in this review.
Wills & Cleary, 1999
Table 1  (continued)

Weighted‑Mean Effect Size and Heterogeneity

For the total sample of 99 effect sizes, cross-lagged

Sample 1
Sample 2
Study ID

effects of peers' substance use on subsequent target ado-

lescents' use controlling for initial similarity ranged from

13
3874 Current Psychology (2024) 43:3866–3881

Forest Plot

Studies Effect Size

Allen et al., 2022

Composite 0.259

Halliday-Scher, 2000
Alcohol 0.186
Tobacco 0.197

Medler, 2000

Tobacco 0.042
Tobacco 0.015
Tobacco 0.114
Tobacco 0.120
Tobacco 0.115
Tobacco 0.136

Mercken et al., 2007

Tobacco 0.004
Tobacco 0.069

Mounts, 2002
Composite 0.213

Allen et al., 2012

Composite 0.144
Alcohol 0.277

Larsen et al., 2010

Alcohol -0.010
Alcohol 0.062
Alcohol 0.072

Mrug et al., 2011

Alcohol -0.045
Tobacco -0.064
Alcohol -0.064
Alcohol -0.073
Tobacco 0.251
Tobacco -0.072

Pilgrim, 1998
Composite -0.050
Composite 0.000
Composite 0.186
Composite 0.146

Deutsch et al., 2015

Alcohol 0.224
Alcohol 0.154
Tobacco 0.184
Tobacco 0.066
Marijuana 0.153
Marijuana 0.173

Schelleman-Offermans 2013
Alcohol 0.213
Alcohol 0.225

Cleveland et al., 2012

Alcohol 0.182
Alcohol 0.203
Alcohol 0.171
Alcohol 0.176

Cleveland et al., 2012

Alcohol 0.123
Alcohol 0.136
Alcohol 0.161
Alcohol 0.098

Mason et al., 2011

Alcohol 0.104
Alcohol 0.326
Alcohol 0.181
Alcohol 0.169
Alcohol 0.258

Mason et al., 2011

Alcohol 0.220
Alcohol 0.319
Alcohol 0.156
Alcohol 0.199
Alcohol 0.221

Burk et al., 2011

Alcohol 0.302

Cin et al., 2009

Alcohol 0.317
Alcohol 0.340

Erickson et al., 2000

Composite 0.122

Pomery et al., 2005

Composite 0.125

Madon et al., 2003

Alcohol 0.110

Kandel, 1978
Marijuana 0.119

Hiatt et al., 2017

Alcohol 0.290

Li et al., 2017
Tobacco 0.082
Tobacco -0.037
Tobacco 0.078
Tobacco 0.042
Tobacco -0.073
Tobacco -0.041

Li et al., 2017
Tobacco 0.025
Tobacco -0.070
Tobacco 0.118
Tobacco 0.011
Tobacco -0.100
Tobacco 0.052

Wills & Cleary, 1999

Composite 0.184
Composite 0.168
Composite 0.111

Wills & Cleary, 1999

Composite 0.184
Composite 0.134
Composite 0.068

Ouellette et al., 1999

Alcohol 0.208
Alcohol 0.302
Alcohol 0.183
Alcohol 0.252

Stice et al., 1998

Alcohol 0.256

Curran et al., 1997

Alcohol 0.230
Alcohol 0.232
Alcohol 0.135

Farrell & Danish, 1993

Composite 0.063
Composite 0.023
Composite 0.203
Composite 0.099
Composite 0.080
Composite 0.075

Farrell & Danish, 1993

Composite 0.205
Composite 0.360
Composite 0.067
Composite 0.110
Composite 0.017
Composite 0.060

-1 -0.5 0 0.5 1
Effect Size

Fig. 3  Forest Plot for All 99 Effect Sizes

13
Current Psychology (2024) 43:3866–3881 3875

-0.10 for tobacco to 0.36 for composite use (i.e., alcohol Moderator Analyses
and/or tobacco and/or marijuana), which is displayed in
Fig. 3. The multivariate meta-regression model with RVE Substance Use Behaviour
correction to allow for multiple potentially dependent
effect sizes from the same study generated a significant Moderator analyses were conducted to explain the effect size
weighted mean cross-legged regression coefficient, β = heterogeneity. First, substance use behaviour was examined
0.147 (SE = 0.016, 95% CI [0.115, 0.180], p < 0.001). to determine whether the specific substance, categorised as
This result indicates that adolescents changed their sub- alcohol, tobacco, marijuana, or composite substance use in
stance use behaviour over time in the direction of their cases that didn’t differentiate between substances, might
peers’ actual or perceived substance use. affect the magnitude of peer influence processes. Modera-
Examining heterogeneity in the effect sizes revealed tion analyses revealed significant differences between the
significant variation. The median sampling variance was substance use behaviours, F(3, 95) = 6.102, p < 0.001 (see
0.001 and represented 11.24% of the total variance. The Table 2). The average cross-lagged correlation coefficient
Level 2 variance, 0.003, χ2(1) = 61.91, p < 0.001, repre- was significant for alcohol (k = 43, β = 0.182, SE = 0.021,
sented 28.43% of the total variance, indicating that within- 95% CI [0.137, 0.227], p < 0.001), tobacco (k = 26, β =
study within-wave variance (i.e., differences in effect sizes 0.068, SE = 0.023, 95% CI [0.007, 0.128,, p = 0.035), and
between the same wave of a particular study) were larger composite substance use (k = 27, β = 0.134, SE = 0.013, 95%
than expected due to sampling variability alone. The Level CI [0.104, 0.164], p < 0.001).
3 variance, 0.001, χ2(1) = 1.76, p = 0.18, represented 11.42%
of the total variance, although it was not significant, meaning Peer Influence Measure
that within-study between-wave variance (i.e., differences in
effect sizes between different waves of a given study) was Substantial research on peer influence has established that,
within the range expected by random variance. The Level consistent with social norm theories, perceived substance
4 variance, 0.005, χ 2(1) = 12.84, p < 0.001, represented use, acceptance or approval is associated with adopting
48.92% of the total variance, indicating significant between- substance use behaviours (Trucco, 2020). This is especially
study variance (i.e., differences in effect sizes between dif- problematic considering that adolescents tend to overes-
ferent studies). Heterogeneity at both the within-study timate peers’ engagement in substance use, with the size
within-wave and between-study levels suggests that there of these misperceptions amounting to large effect sizes
are likely moderators of the effects observed, which will be (Trucco, 2020). This meta-analysis sought to determine
explored in subsequent moderator analyses. whether there is a significant difference between studies that

Table 2  Results from Multilevel Meta-Regression Models

Effect size Moderator effect
Moderator k β SE β 95% CI p β SE β 95% CI p F df p

Substance 99 6.102 3, 95 < 0.001

Alcohol 43 0.182 0.021 [0.137, 0.227] < 0.001
Tobacco 26 0.068 0.023 [0.007, 0.128] 0.035
Marijuana 3 0.134 0.015 [-0.025, 0.293] 0.060
Composite 27 0.134 0.013 [0.104, 0.164] < 0.001
Peer Measure 99 12.695 1, 97 < 0.001
Perceived 54 0.179 0.014 [0.148, 0.210] < 0.001
Actual 45 0.095 0.021 [0.050, 0.141] < 0.001
Mean Age 85
Linear -0.004 0.013 [-0.032, 0.025] 0.782 0.106 1, 83 0.746
Quadratic -0.005 0.003 [-0.015, 0.004] 0.165 0.634 2, 82 0.533
Time Lag 99 1.010 2, 96 0.368
Linear -0.001 0.022 [-0.049, 0.047] 0.961
Quadratic 0.003 0.004 [-0.008, 0.016] 0.444
Gender 94 0.272 0.104 [-0.002, 0.546] 0.051 11.813 1, 92 < 0.001
(% boys)

k = number of effect sizes; β = average effect size; β = regression coefficient; CI = confidence interval, df = degrees of freedom.

13
3876 Current Psychology (2024) 43:3866–3881

used perceived measures of peer substance use (i.e., target larger sample sizes or smaller standard errors have a nar-
adolescent’s estimations of their peers’ substance use) and rower dispersion. Visually inspecting the funnel plot reveals
studies that used actual peer measures (i.e., nominated peers that the effect sizes (indicated as dots on the plot) largely
record their own substance use). A significant moderation fall in the inverted funnel shape. For a more robust evalu-
effect was found, F(1, 97) = 12.695, p < 0.001 (see Table 2). ation, the moderating effect of the inverse sampling vari-
Both perceived and actual peer measures were found to be ation was calculated, which yielded nonsignificant results
significant, with perceived measures having a larger average (β = 0.000013, SE = 0.00001, 95% CI [-0.00002, 0.00004],
effect size (k = 54, β = 0.179, SE = 0.014, 95% CI [0.148, p = 0.264). Taken together, the funnel plot and inverse sam-
0.210], p < 0.001) than actual peer measures (k = 45, β = pling variance moderation suggest that publication bias is
0.095, SE = 0.021, 95% CI [0.050, 0.141], p < 0.001). not significantly contributing to the results.

Mean Age at Baseline

Discussion
To examine the moderating effect of participants’ age, a lin-
ear and a quadratic model were executed. This was due to As hypothesised, the results revealed a significant positive
evidence suggesting that peer influence effects operate in effect ( β = 0.147) of peer influence on adolescent substance
both a linear manner, with an overall increase from child- use, suggesting that adolescents will change their sub-
hood to adolescence, and in a curvilinear way, peaking dur- stance use behaviour in the same direction as their peers.
ing mid-adolescence (Trucco, 2020). As shown in Table 2, It is important to recognise that this finding also implies
neither model yielded significant results. This suggests that that target adolescents with peers using substances at low
the strength of peer influence effects does not differ with levels are likely to decrease their substance use behaviour
participant age. over time (Allen et al., 2022). No significant results emerged
to suggest that publication bias is prevalent in this study.
Time Lag In line with the hypotheses, significant differences emerged
between the substance use behaviours. As expected, target
The time lag between assessments was examined following adolescents significantly changed their alcohol, tobacco, and
the Lag-as-Moderator Meta-Analysis approach (Card, 2019). composite substance use over time in accordance with the
This model yields both a linear (i.e., peer influence becom- level of substance use behaviour of their peers. Alcohol use
ing larger or smaller with longer time spans) and quadratic emerged as having the largest average cross-lagged corre-
(i.e., peer influence reaching a maximum effect at a par- lation coefficient ( β = 0.182), which was predicted due to
ticular lag length) moderation of lag, which is centred on it being the most normalised and prevalent substance use
the weighted mean. These analyses revealed no significant behaviour. The smallest effect size was for tobacco ( β =
effect of time lag for neither linear nor quadratic term (see 0.068), which is likely due to small populations of partici-
Table 2). pants engaging in tobacco smoking behaviours. Only a few
studies (k = 3) differentiated marijuana use, with majority
Multivariate Moderation Model of studies that used a composite substance use measure
including marijuana. Composite substance use emerged
A multivariate moderation model simultaneously examining as significant with the same effect size as marijuana ( β =
all five main moderators (i.e., substance use behaviour, peer 0.134), although marijuana use was not significantly differ-
substance use measure, time lag, age, and gender) revealed ent. Recent Australian survey results found marijuana to be
no significant differences. the most commonly used illicit substance among 12–17 year
olds, with 8.2% of 14–17 year olds reporting recently using
Publication Bias marijuana in 2019, which was an increase from 7.9% in
2016. It is likely that marijuana use not being significant is
Publication bias is a potential threat to all systematic due to the small number of studies examining it, and that the
reviews, despite efforts to locate unpublished effect sizes. To composite measure is a better representation of peer influ-
evaluate and quantify the impact of publication bias, inverse ence effects on marijuana use.
sampling variance moderation analyses were conducted, and Consistent with the hypotheses, a significant moderation
a funnel plot was assessed (see Fig. 4). In the absence of effect was revealed for perceived vs. actual peer measures,
publication bias, the plot should present a symmetrical fun- and perceived measures estimated a larger average cross-
nel shape centred on the mean effect size. This means that lagged correlation coefficient ( β = 0.179) than actual peer
studies with smaller sample sizes or larger standard errors measures ( β = 0.095). This is consistent with social norm
will scatter widely at the bottom of the plot, while those with theories, which suggest that adolescents tend to overestimate

13
Current Psychology (2024) 43:3866–3881 3877

Fig. 4  Funnel Plot of All 99

8579.493
Effect Sizes Included

6461.515
Inverse Variance

4343.538
2225.56
107.582

-0.1 0 0.1 0.2 0.3 0.4

Effect size

their peers’ substance use behaviour and that this perception Strengths and Limitations
is sufficient to influence target adolescent behaviour (Trucco,
2020). This was reinforced by Helms et al. (2014) who found While great effort went into ensuring this study was a com-
that adolescents significantly misperceived the substance prehensive synthesis and robust meta-analysis, there are sig-
use of popular peers; however, it contradicts Urberg et al. nificant limitations to consider. Firstly, this study addressed
(2003) who found that adolescents would only conform to peer influence processes in general, rather than specifically
peer substance use behaviour if the friendship was positive investigating whether the nature of the peer relationship
and reciprocated. This contradiction captures many of the moderates the effect. Considering that the included stud-
debates about peer influence and lends support to Allen ies used measures from peers of varying degrees of friend-
et al.’s (2022) theory that peer influence effects are norma- ship to the target adolescent, there remains the question of
tive and adaptive; different from the maladaptive effects that whether the closeness of friendship impacts peer influence
can occur from interaction with a problematic subculture or processes. Additionally, broader group influence processes
norms. Contrary to the hypothesis, the results showed no lin- may behave differently, with groups of peers exerting an
ear or quadratic effect of age on peer influence. This finding amplifying effect on behaviour. Secondly, the study did
may be due to the limited variance in age range and varia- not consider differences in the degree to which a peer may
tions in the age of puberty, rather than no actual difference. actively attempt to be influential. It is suggested that ado-
Additionally, no significant effects emerged for time lag or lescents are more influenced by supportive peers, rather
gender. The multivariate moderation model which examined than those who are more coercive and pressuring; however,
all five moderators simultaneously revealed no significant it may be that coercive peer behaviours have a short-term
differences, which was contrary to the univariate analyses for impact on behaviour in a given situation but aren’t captured
substance use behaviour and peer influence measures. This by the time lags in the included studies (Allen et al., 2022).
contradictory result raises questions about the robustness Thirdly, while care was taken to collate studies from a vari-
of the moderation analysis results and suggests that peer ety of nationalities, almost all the records included in the
influence is a complex phenomenon with interacting factors. meta-analysis are from cultures that value individualism

13
3878 Current Psychology (2024) 43:3866–3881

more highly. A meta-analysis by Liu et al. (2017) found a influence processes through social media and identifying the
significant effect of individualism-collectivism, with ado- factors that moderate both problematic and productive social
lescents from cultures scoring more highly on collectivism media use would be an important area for future research.
being more likely to conform to normative influence from Additionally, another concerning trend is that of smoking
peers. The current review only included one study from a electronic cigarettes. While cigarette smoking is declining
country that scored equal to or above 50 on collectivism, as overall, with only 5% of Australian secondary school stu-
per Liu et al. (2017) measure, which was Li et al.’s (2017) dents aged 12–17 reporting being current smokers in 2017,
investigation into tobacco use by Chinese adolescents. This 14% of those students had tried e-cigarettes and vaping
suggests that the results attained in this study may not be e-cigarettes is significantly increasing (Australian Institute
representative of peer influence effects in more collectivis- of Health & Welfare, 2022). Future research into e-cigarette
tic cultures. Additionally, individualism-collectivism is not use in adolescents and whether normative peer influences
the only cultural consideration, as peer influence processes are greater for vaping than for traditional tobacco smoking
may operate differently among adolescents from different would be valuable.
countries due to variations in the prevalence of substance As is common in psychological research, the systematic
use, the availability of substances, laws and policies, social review yielded few studies in non-Western countries. As
norms, and other cultural factors. mentioned previously, a significant limitation of this study
While the results suggested that publication bias was not was the overreliance on studies based in Western individ-
significantly impacting the results, it is an important consid- ualistic cultures, with Li et al.’s (2017) investigation into
eration in systematic reviews, as many studies that find no tobacco use in China being the only non-Western record
significant results or negative results are not published. This included. A review by Liu et al. (2017) revealed significant
‘file drawer problem’ leads to effect sizes generated in meta- differences in smoking initiation and continuation based on
analysis being overestimated. Therefore, it is best practice the country’s collectivism-individualism measure, indicating
when conducting a systematic review and meta-analysis to a need for studies with diverse populations. Lastly, future
conduct a ‘snowball’ search of reference lists and contact research that conceptualises peer influence as an overall
prominent authors on the research topic and request any adaptive phenomenon with a neutral valence and investi-
unpublished studies. Unfortunately, due to the time restraints gates the broader subcultures that may be contributing to
and scope of this study, previous authors were not contacted, adolescent substance use would be worthwhile (Allen et al.,
although several unpublished manuscripts were identified 2022). With the wealth of research on peer influence and
and included through reference list searches. It is therefore findings that peer influence effects can often overcome tar-
possible that unpublished data that would contribute to this geted interventions, looking past the question of the magni-
topic was missed. tude of these processes, and instead focusing on the adaptive
and maladaptive norms within peer contexts may provide
Future Research better information about how to prevent adolescents from
engaging in substance use behaviours.
As social media continues to grow and evolve, future
research should look at the processes of online peer influ-
ence. Social media has created a new context through which Conclusion
adolescents interact with their peers, as well as perceive peer
behaviours and norms (Choukas-Bradley & Nesi, 2020). The main aim of this study was to systematically review
Additionally, social media is now populated by ‘influenc- the large body of literature on peer influence from the past
ers’ who earn money based on their ability to impact the 50 years and quantify the magnitude of peer influence on
purchasing decisions of their audience. Influencers present adolescent substance use behaviour using the best available
an interesting case, as the influence processes occur exclu- methodologies. The results revealed a significant positive
sively online and often without any direct interaction or effect ( β = 0.147), indicating that adolescents whose peers
reciprocal exchange with users. Research into the impact engage in greater or lesser substance use behaviour are
of social media influencers on adolescents’ decisions to significantly likely to alter their own substance use accord-
engage in risky behaviours would be worthwhile. Hamilton ingly over time. These findings support the existing litera-
et al. (2022) noted the potential for online peer influence ture which identified peer influence processes as contrib-
to be both positive and negative in the context of COVID- uting to adolescents’ decisions to engage in substance use
19, with images and videos shared by peers practicing safe behaviours. Significant heterogeneity was found between
social distancing practices influencing adolescents to also the effect sizes across multiple levels, suggesting that unex-
engage in pro-social health behaviours, while the opposite amined contextual, individual, and methodological factors
was also true. Developing a more nuanced understanding of may modify peer influence effects. Moderation analyses

13
Current Psychology (2024) 43:3866–3881 3879

revealed significant differences between different substance References

use behaviours and between studies that used perceived vs.
actual peer measures; however, when assessed simultane-
ously no moderators emerged as significant. This suggests References included in the meta‑analysis are
that peer influence works through complex processes across marked with an asterisk
substance use behaviours, and that attempting to quantify its
magnitude may be innately flawed. Overall, findings from Adikini, S. (2023). Peer pressure, substance abuse and risky sexual
this meta-analysis reveal a significant and robust positive behavior among students at Makerere University (Doctoral dis-
sertation, Makerere University).
effect for peer influence on adolescent substance use. Defini- *Allen, J. P., Chango, J., Szwedo, D., Schad, M., & Marston, E. (2012).
tively establishing the impact of peer influence across multi- Predictors of susceptibility to peer influence regarding substance
ple substance use behaviours and peer measures of substance use in adolescence. Child Development, 83(1), 337–350. https://​
use has significant implications for adolescent substance use doi.​org/​10.​1111/j.​1467-​8624.​2011.​01682.x
*Allen, J. P., Chango, J., & Szwedo, D. (2014). The adolescent rela-
prevention. This includes the potential of harnessing peer tional dialectic and the peer roots of adult social functioning.
influence as a positive force and the need to target misper- Child Development, 85(1), 192–204. https://​doi.​org/​10.​1111/​
ceptions of substance use, as well as providing an opportu- cdev.​12106
nity to focus on the adolescent subculture and norms that *Allen, J. P., Loeb, E. L., Kansky, J., & Davis, A. A. (2022). Beyond
susceptibility: Openness to peer influence is predicted by adaptive
are contributing to risky health behaviour. Research has social relationships. International Journal of Behavioral Develop-
demonstrated that interventions targeting peer influence are ment, 46(3), 180–189. https://d​ oi.o​ rg/1​ 0.1​ 177/0​ 16502​ 54209​ 22616
largely ineffective in preventing substance use behaviour in Australian Institute of Health and Welfare. (2022). Alcohol, tobacco &
adolescents, potentially due to peer influence representing other drugs in Australia. Retrieved from https://w ​ ww.a​ ihw.g​ ov.a​ u/​
repor​ts/​alcoh​ol/​alcoh​ol-​tobac​co-​other-​drugs-​austr​alia
an overall adaptive developmental phenomenon. Therefore, Bandura, A. (1977). Social learning theory. Prentice-Hall.
the emphasis of future studies in preventing dangerous sub- Barnett, N. P., DiGuiseppi, G. T., Tesdahl, E. A., & Meisel, M. K.
stance use behaviour should look less at peer influence and (2022). Peer selection and influence for marijuana use in a com-
more at the adolescent subcultures that support problematic plete network of first-year college students. Addictive Behaviors,
124, 107087.
behaviour. Bartolo, M. G., Palermiti, A. L., Servidio, R., Musso, P., Tenuta, F.,
Amendola, M. F., ... & Inguglia, C. (2023). The relationship
between parental monitoring, peer pressure, and motivations for
Funding Open Access funding enabled and organized by CAUL and responsible drinking among Italian adolescents: the mediating
its Member Institutions role of positive alcohol expectancies. The Journal of Genetic
Psychology, 184(1), 23–41.
Data Availability All data are all available upon request from the first Becker, B. J. (1992). Using results from replicated studies to estimate
author of this article. linear models. Journal of Educational Statistics, 17(4), 341–362.
https://​doi.​org/​10.​2307/​11651​28
Declarations Blanton, H., Gibbons, F. X., Gerrard, M., Conger, K. J., & Smith, G.
E. (1997). Role of family and peers in the development of proto-
Ethical Statement N/A types associated with substance use. Journal of Family Psychol-
ogy, 11(3), 271–288. https://​doi.​org/​10.​1037/​0893-​3200.​11.3.​271
Informed Consent N/A Bronfenbrenner, U. (1979). The Ecology of Human Development: Exper-
iments by Nature and Design. UK: Harvard University Press.
*Burk, L. R., Armstrong, J. M., Goldsmith, H. H., Klein, M. H., Strau-
Conflict of Interest The authors declared not having any conflict of man, T. J., Costanzo, P., & Essex, M. J. (2011). Sex, tempera-
interest. ment, and family context: How the interaction of early factors
differentially predict adolescent alcohol use and are mediated by
Open Access This article is licensed under a Creative Commons Attri- proximal adolescent factors. Psychology of Addictive Behaviors,
bution 4.0 International License, which permits use, sharing, adapta- 25(1), 1–15https://​doi.​org/​10.​1037/​a0022​349
tion, distribution and reproduction in any medium or format, as long Card, N. A. (2012). Applied meta-analysis for social science research.
as you give appropriate credit to the original author(s) and the source, Guilford Press.
provide a link to the Creative Commons licence, and indicate if changes Card, N. A. (2019). Lag as moderator meta-analysis: A methodological
were made. The images or other third party material in this article are approach for synthesizing longitudinal data. International Journal
included in the article's Creative Commons licence, unless indicated of Behavioral Development, 43(1), 80–89. https://d​ oi.o​ rg/1​ 0.1​ 177/​
otherwise in a credit line to the material. If material is not included in 01650​25418​773461
the article's Creative Commons licence and your intended use is not Catalano, R. F., & Hawkins, J. D. (1996). The social development
permitted by statutory regulation or exceeds the permitted use, you will model: A theory of antisocial behavior. In J. D. Hawkins (Ed.),
need to obtain permission directly from the copyright holder. To view a Delinquency and crime: Current theories (pp. 149–197). Cam-
copy of this licence, visit http://​creat​iveco​mmons.​org/​licen​ses/​by/4.​0/. bridge University Press.
Catalano, R. F., Kosterman, R., Hawkins, J. D., Newcomb, M. D., &
Abbott, R. D. (1996). Modeling the etiology of adolescent sub-
stance use: A test of the social development model. Journal of

13
3880 Current Psychology (2024) 43:3866–3881

Drug Issues, 26(2), 429–455. https://d​ oi.o​ rg/1​ 0.1​ 177/0​ 02204​ 2696​ and directions for future research. Perspectives on Psychological Sci-
02600​207 ence, 17(3), 662–679. https://​doi.​org/​10.​1177/​17456​91621​10141​89
Choukas-Bradley, S., & Nesi, J. (2020). Applying developmental the- Hedges, L. V., Tipton, E., & Johnson, M. C. (2010). Robust variance
ory to adolescent peer influence processes in the social media con- estimation in meta-regression with dependent effect size esti-
text. In N. Van Zalk & C. P. Monks (Eds.), Online peer engage- mates. Research Synthesis Methods, 1(1), 39–65. https://​doi.​org/​
ment in adolescence: Positive and negative aspects of online 10.​1002/​jrsm.5
social interaction (pp. 140–162). Routledge/Taylor & Francis Helms, S. W., Choukas-Bradley, S., Widman, L., Giletta, M., Cohen,
Group. https://​doi.​org/​10.​4324/​97804​29468​360-9 G. L., & Prinstein, M. J. (2014). Adolescents misperceive and are
*Cin, S. D., Worth, K. A., Gerrard, M., Gibbons, F. X., Stoolmiller, M., influenced by high-status peers’ health risk, deviant, and adaptive
Wills, T. A., & Sargent, J. D. (2009). Watching and drinking: Expec- behavior. Developmental Psychology, 50(12), 2697–2714. https://​
tancies, prototypes, and friends’ alcohol use mediate the effect of doi.​org/​10.​1037/​a0038​178
exposure to alcohol use in movies on adolescent drinking. Health Henneberger, A. K., Mushonga, D. R., & Preston, A. M. (2021). Peer
Psychology, 28(4), 473–483. https://​doi.​org/​10.​1037/​a0014​777 influence and adolescent substance use: A systematic review of
*Cleveland, M. J., Feinberg, M. E., & Jones, D. E. (2012). Predicting dynamic social network research. Adolescent Research Review,
alcohol use across adolescence. Psychology of Addictive Behav- 6(1), 57–73. https://​doi.​org/​10.​1007/​s40894-​019-​00130-0
iors, 26(4), 703–713. https://​doi.​org/​10.​1037/​a0027​583 *Hiatt, C., Laursen, B., Stattin, H., & Kerr, M. (2017). Best friend
Curcio, A. L., Mak, A. S., & George, A. M. (2012). Do adolescent influence over adolescent problem behaviors: Socialized by the
delinquency and problem drinking share psychosocial risk fac- satisfied. Journal of Clinical Child and Adolescent Psychology,
tors? A Literature Review. Addictive Behaviors, 38(4), 2003–2013. 46(5), 695–708. https://​doi.​org/​10.​1080/​15374​416.​2015.​10507​23
https://​doi.​org/​10.​1016/j.​addbeh.​2012.​12.​004 *Kandel, D. B. (1978). Homophily, selection, and socialization in ado-
*Curran, P. J., Stice, E., & Chassin, L. (1997). The relation between lescent friendships. The American Journal of Sociology, 84(2),
adolescent alcohol use and peer alcohol use: A longitudinal 427–436. https://​doi.​org/​10.​1086/​226792
random coefficients model. Journal of Consulting and Clinical *Larsen, H., Overbeek, G., Vermulst, A., Granic, I., & Engels, R. C. M.
Psychology, 65(1), 130–140. https://​doi.​org/​10.​1037/​0022-​006X.​ (2010). Initiation and continuation of best friends and adolescents’
65.1.​130 alcohol consumption: Do self-esteem and self-control function as
de la Haye, K., Green, H. D., Jr., Kennedy, D. P., Pollard, M. S., & moderators? International Journal of Behavioral Development,
Tucker, J. S. (2013). Selection and influence mechanisms associ- 34(5), 406–416. https://​doi.​org/​10.​1177/​01650​25409​350363
ated with marijuana initiation and use in adolescent friendship Leung, R. K., Toumbourou, J. W., & Hemphill, S. A. (2014). The effect
networks. Journal of Research on Adolescence, 23(3), 474–486. of peer influence and selection processes on adolescent alcohol
https://​doi.​org/​10.​1111/​jora.​12018 use: A systematic review of longitudinal studies. Health Psychol-
DeLay, D., Shen, M., Cook, R. E., Zhao, S., Logis, H., & French, D. ogy Review, 8(4), 426–457. https://​doi.​org/​10.​1080/​17437​199.​
C. (2023). Peers influence the tobacco and alcohol use of Chinese 2011.​587961
adolescents. Journal of Research on Adolescence. *Li, L., Lu, T., Niu, L., Feng, Y., Jin, S., & French, D. C. (2017).
*Deutsch, A. R., Chernyavskiy, P., Steinley, D., & Slutske, W. S. Tobacco use by middle and high school Chinese adolescents and
(2015). Measuring peer socialization for adolescent substance their friends. Journal of Youth and Adolescence, 46(6), 1262–
use: A comparison of perceived and actual friends’ substance use 1274. https://​doi.​org/​10.​1007/​s10964-​016-​0563-6
effects. Journal of Studies on Alcohol and Drugs, 76(2), 267–277. Liu, J., Zhao, S., Chen, X., Falk, E., & Albarracín, D. (2017). The influ-
https://​doi.​org/​10.​15288/​jsad.​2015.​76.​267 ence of peer behavior as a function of social and cultural close-
*Erickson, K. G., Crosnoe, R., & Dornbusch, S. M. (2000). A social ness: A meta-analysis of normative influence on adolescent smok-
process model of adolescent deviance: Combining social control ing initiation and continuation. Psychological Bulletin, 143(10),
and differential association perspectives. Journal of Youth and 1082–1115. https://​doi.​org/​10.​1037/​bul00​00113
Adolescence, 29(4), 395–425. https://​doi.​org/​10.​1023/A:​10051​ *Madon, S., Guyll, M., Spoth, R. L., Cross, S. E., & Hilbert, S. J.
63724​952 (2003). The self-fulfilling influence of mother expectations on
*Farrell, A. D., & Danish, S. J. (1993). Peer drug associations and emo- children’s underage drinking. Journal of Personality and Social
tional restraint: Causes or consequences of adolescents’ drug use? Psychology, 84(6), 1188–1205. https://​doi.​org/​10.​1037/​0022-​
Journal of Consulting and Clinical Psychology, 61(2), 327–334. 3514.​84.6.​1188
https://​doi.​org/​10.​1037/​0022-​006X.​61.2.​327 *Mason, W. A., Toumbourou, J. W., Herrenkohl, T. I., Hemphill, S. A.,
Fernández-Castilla, B., Aloe, A. M., Declercq, L., Jamshidi, L., Catalano, R. F., & Patton, G. C. (2011). Early age alcohol use and
Beretvas, S. N., Onghena, P., & Van den Noortgate, W. (2021). later alcohol problems in adolescents: Individual and peer media-
Estimating outcome-specific effects in meta-analyses of multiple tors in a bi-national study. Psychology of Addictive Behaviors,
outcomes: A simulation study. Behavior Research Methods, 53(2), 25(4), 625–633. https://​doi.​org/​10.​1037/​a0023​320
702–717. https://​doi.​org/​10.​3758/​s13428-​020-​01459-4 *Medler, S. M. (2000). Adolescent smoking behavior: The dynamics of
Fisher, Z., Tipton, E., & Zhipeng, H. (2017). Robumeta: Robust Vari- parental and peer influence, attitude, and the active role of the ado-
ance Meta-Regression. R package version 2.0. https://​CRAN.R-​ lescent (Publication No. 9992242) [Doctoral dissertation, Wayne
proje​ct.​org/​packa​ge=​robum​eta State University]. ProQuest Dissertations & Theses Global.
Giletta, M., Choukas-Bradley, S., Maes, M., Linthicum, K. P., Card, *Mercken, L., Candel, M., Willems, P., & de Vries, H. (2007). Disentan-
N. A., & Prinstein, M. J. (2021). A meta-analysis of longitudinal gling social selection and social influence effects on adolescent smok-
peer influence effects in childhood and adolescence. Psychologi- ing: The importance of reciprocity in friendships. Addiction, 102(9),
cal Bulletin, 147(7), 719–747. https://d​ oi.o​ rg/1​ 0.1​ 037/b​ ul000​ 0329 1483–1492. https://​doi.​org/​10.​1111/j.​1360-​0443.​2007.​01905.x
*Halliday-Scher, K. (2000). A model of parent and friend relationship *Mounts, N. S. (2002). Parental management of adolescent peer relation-
effects on adolescent substance use (Publication No. 9966143) ships in context: The role of parenting style. Journal of Family Psy-
[Doctoral dissertation, Wayne State University]. ProQuest Dis- chology, 16(1), 58–69. https://​doi.​org/​10.​1037/​0893-​3200.​16.1.​58
sertations & Theses Global. *Mrug, S., Borch, C., & Cillessen, A. H. (2011). Other-sex friend-
Hamilton, J. L., Nesi, J., & Choukas-Bradley, S. (2022). Reexamining ships in late adolescence: Risky associations for substance use
social media and socioemotional well-being among adolescents and sexual debut? Journal of Youth and Adolescence, 40(7),
through the lens of the COVID-19 pandemic: A theoretical review 875–888. https://​doi.​org/​10.​1007/​s10964-​010-​9605-7

13
Current Psychology (2024) 43:3866–3881 3881

*Ouellette, J. A., Gerrard, M., Gibbons, F. X., & Reis-Bergan, M. *Stice, E., Barrera, M., & Chassin, L. (1998). Prospective differential
(1999). Parents, peers, and prototypes: Antecedents of adolescent prediction of adolescent alcohol use and problem use: Examin-
alcohol expectancies, alcohol consumption, and alcohol-related ing the mechanisms of effect. Journal of Abnormal Psychology,
life problems in rural youth. Psychology of Addictive Behaviors, 107(4), 616–628. https://​doi.​org/​10.​1037/​0021-​843X.​107.4.​616
13(3), 183–197. https://​doi.​org/​10.​1037/​0893-​164X.​13.3.​183 Tipton, E. (2015). Small sample adjustments for robust variance esti-
Page, M. J., McKenzie, J. E., Bossuyt, P. M., Boutron, I., Hoffmann, T. mation with meta-regression. Psychological Methods, 20(3),
C., Mulrow, C. D., Shamseer, L., Tetzlaff, J. M., Akl, E. A., Bren- 375–393. https://​doi.​org/​10.​1037/​met00​00011
nan, S. E., Chou, R., Glanville, J., Grimshaw, J. M., Hróbjartsson, Trucco, E. M. (2020). A review of psychosocial factors linked to ado-
A., Lalu, M. M., Li, T., Loder, E. W., Mayo-Wilson, E., McDon- lescent substance use. Pharmacology, Biochemistry and Behavior,
ald, S., … Moher, D. (2021). The PRISMA 2020 statement: An 196, Article 172969. https://​doi.​org/​10.​1016/j.​pbb.​2020.​172969
updated guideline for reporting systematic reviews. BMJ, n71. Urberg, K. A., Luo, Q., Pilgrim, C., & Degirmencioglu, S. M. (2003).
https://​doi.​org/​10.​1136/​bmj.​n71 A two-stage model of peer influence in adolescent substance use:
*Pilgrim, C. (1998). Close friend's drug use, authoritative parent- Individual and relationship-specific differences in susceptibility
ing and sensation seeking effects on drug use among European- to influence. Addictive Behaviors, 28(7), 1243–1256. https://​doi.​
American and African-American adolescents (Publication No. org/​10.​1016/​S0306-​4603(02)​00256-3
9815360) [Doctoral dissertation, Wayne State University]. Pro- Veenstra, R., & Laninga-Wijnen, L. (2022). Peer network studies and
Quest Dissertations & Theses Global. interventions in adolescence. Current Opinion in Psychology, 44,
*Pomery, E. A., Gibbons, F. X., Gerrard, M., Cleveland, M. J., Brody, 157–163.
G. H., & Wills, T. A. (2005). Families and risk: Prospective analy- Viechtbauer, W. (2010). Conducting meta-analyses in R with the meta-
ses of familial and social influences on adolescent substance use. for package. Journal of Statistical Software, 36(3), 1–48. https://​
Journal of Family Psychology, 19(4), 560–570. https://d​ oi.o​ rg/1​ 0.​ doi.​org/​10.​18637/​jss.​v036.​i03
1037/​0893-​3200.​19.4.​560 *Wills, T. A., & Cleary, S. D. (1999). Peer and adolescent substance
Pustejovsky, J. E., & Tipton, E. (2022). Meta-analysis with robust use among 6th-9th graders: Latent growth analyses of influence
variance estimation: Expanding the range of working models. versus selection mechanisms. Health Psychology, 18(5), 453–
Prevention Science, 23(3), 425–438. https://​doi.​org/​10.​1007/​ 463. https://​doi.​org/​10.​1037/​0278-​6133.​18.5.​453
s11121-​021-​01246-3 World Health Organisation. (2021, November 17). Adolescent mental
Pustejovsky, J. (2018). ClubSandwich: Cluster-robust (sandwich) vari- health [Fact Sheet]. https://​www.​who.​int/​news-​room/​fact-​sheets/​
ance estimators with small-sample corrections. R package version detail/​adole​scent-​mental-​health
0.3.2. https://​CRAN.R-​proje​ct.​org/​packa​ge=​clubS​andwi​ch Zedaker, S., Fansher, A. K., & Jin, H. R. (2023). Adolescent Romance
R Core Team. (2022). R: A language and environment for statistical Or Partners In Crime? The Influence Of Romantic Relationships
computing (Version 4.2.2) [Computer software] R Foundation for And Peers On Criminal Behavior In Adolescence. Journal of
Statistical Computing. https://​www.R-​proje​ct.​org School Violence, 1–14.
*Schelleman-Offermans, K., Knibbe, R. A., & Kuntsche, E. (2013).
Are the effects of early pubertal timing on the initiation of weekly Publisher's note Springer Nature remains neutral with regard to
alcohol use mediated by peers and/or parents? A longitudinal jurisdictional claims in published maps and institutional affiliations.
study. Developmental Psychology, 49(7), 1277–1285. https://​doi.​
org/​10.​1037/​a0029​880
Serafini, G., Pompili, M., Innamorati, M., Rihmer, Z., Sher, L., &
Girardi, P. (2012). Can cannabis increase the suicide risk in psy-
chosis? A Critical Review. Current Pharmaceutical Design, 18(32),
5165–5187. https://​doi.​org/​10.​2174/​13816​12128​02884​663

13