Académique Documents
Professionnel Documents
Culture Documents
research-article2015
Article
The influence of
positional and
experienced social
benefits on the
relationship between
peers and alcohol use
Owen Gallupe
Martin Bouchard
Abstract
An assumption of peer influence research is that being connected to an alcoholusing peer group is associated with personal alcohol use. However, most research
assesses peer influence through simply counting the number of peers involved
in a particular behavior or the amount of that behavior within a persons peer
group. Rarely considered is the fact that behavioral pressures may only arise when
the peer group actually provides substantial benefits to adolescents. This study
examines whether adolescents who associate with peers who drink are as likely to
be drinkers themselves when they receive high levels of social benefits as compared
to adolescents who receive lower levels. Specifically, the authors examine how the
effect of peer alcohol use on individual decisions to drink is conditioned by the social
status and power that come with occupying sociometrically optimal positions (high
popularity, centrality, density) and more concretely experienced social benefits
(e.g. spending time/talking with/confiding in friends). Using the longitudinal Add
Corresponding author:
Owen Gallupe, Department of Sociology and Legal Studies, University of Waterloo, 200
University Avenue West, Waterloo, ON N2L 3G1, Canada.
Email: ogallupe@uwaterloo.ca
41
Health data (n=13,351), we find that peer alcohol use is most strongly related to
personal alcohol use when a person is subject to greater social benefits in terms of
both sociometric position and through closer one-on-one interactions with peers.
Keywords
Alcohol, networks, peer influence, social benefits
Social groups tend to be highly valued for a variety of reasons. They have
the power to provide individuals with both emotional and material support
and are a prime means through which desirable relationship characteristics
such as social status can be attained. Peers groups are particularly important
to adolescents (Brown, 2004; Warr, 2002). But not all groups are equal.
They often differ in the types of behaviors that are encouraged by its members (Haynie, 2002). Therefore, what is offensive to members of one group
may not be offensive to members of another group. In this way, certain
groups may encourage academic achievement (Meijs etal., 2010); others
may encourage athletic involvement (Dijkstra etal., 2009); some groups
may promote delinquent behavior and/or substance use (Haynie, 2002). The
behaviors supported by the peer group are important because they play a
major role in determining individual rule-breaking behavior (e.g. Pratt
etal., 2010). However, given the dynamic nature of adolescent behavioral
influences, the effect of peers on behavior is likely to depend as much on the
specific characteristics of a given peer group as on the actions of members
of that group (e.g. Haynie, 2001). In this study, we expand the focus of peer
influence in an attempt to understand the contingencies under which peers
influence drinking behavior. This focus goes beyond simpler approaches
based on counting the number of deviant peers an adolescent has. In doing
so, we examine how benefiting from relationships acts to constrain behaviors to alcohol use in an attempt to maintain access to those benefits. In
essence, it is expected that adolescents make a rational decision (to varying
degrees) regarding the behavior that they feel is likely to secure their access
to the benefits of peer relationships. The term benefits is used to denote
the various ways that a person can view their peers as having value to their
life. It therefore enables us to link together varied concepts such as attachments, supports, time with peers, and social status as they are all potentially
cherished characteristics of peer relations. We choose to focus the examination of these dynamics on alcohol use, given the substantial personal and
societal costs of adolescent alcohol use. Hawkins etal. (1992) provide a
succinct summary of these costs:
42
For the developing young adult, drug and alcohol abuse undermines motivation,
interferes with cognitive processes, contributes to debilitating mood disorders,
and increases risk of accidental injury or death. For the society at large, adolescent
substance abuse extracts a high cost in health care, educational failure, mental
health services, drug and alcohol treatment, and juvenile crime. (p. 64)
43
44
simply be that their group does not really do much for them. It is therefore
unclear which of these general forms of social benefits, positional or experienced, plays a stronger role in constraining behaviors since there are theoretical arguments for both. As such, we will assess various conceptions of
both positional and experienced social benefits to see whether and how they
constrain adolescent behavior to alcohol use when the peer group is more
highly supportive of drinking behavior.
45
46
In this vein, Allen etal. (2005) examined the multifaceted role of popularity in its effect on deviant outcomes. They found that higher levels of
popularity were related to greater levels of positive functioning (ego development, attachment security, friendship competence, and positivity with
mother). But they also found that high popularity was related to increased
levels of alcohol/substance use and minor delinquency, particularly when
their peers offered greater support for misconduct. There was no significant
relationship between popularity and serious forms of delinquency. However,
the most relevant study concerning the antisocial behavioral constraints of
positional benefits is the work by Haynie (2001). Much like this study,
Haynie argued that more popular and centrally located adolescents in the
peer network and those in denser networks were more likely than others to
act in ways supported by the group. Using a large sample of American adolescents, she found that adolescents with these network characteristics were
most delinquent when their network was more highly delinquent. Therefore,
respondents were most likely to match their behavior to that of the network
when the network showed support for delinquency and when the individual
had stronger positioning. Haynies study lends strong support for the constraining effect of positional benefits. However, the positional measures
used in her study to index peer group relations were based on possessing
connections to others without accounting for the content of those relationships. This is a limitation of strict positional measures in general. That is, by
using sociometric measures that index the extent to which a person is
embedded in their network (popularity, centrality, density), it can only be
assumed that benefits are derived from the group through connections to
others since these measures do not get at the features of relations that actually produce benefits for group members. It is conceivable that a person
could have many connections to socially powerful peers while deriving little benefit from the group.
47
48
(although the authors of that study did not directly link their findings to a
social benefits perspective).
This study
To summarize, this study tests whether adolescents who are likely to perceive their peer group as more beneficial (in terms of both positional and
experienced benefits) will be heavier drinkers when their peer group supports drinking. Therefore, the main concern of the study is to examine how
these various conceptions of social benefits interact with peer alcohol use in
its effect on personal alcohol use. The overarching hypothesis is that when
peers provide greater social benefits and are more highly involved in alcohol use, the likelihood of individual alcohol use will be at its peak.
Dependent variable
The outcome variable in this study was a measure of alcohol use in the past
year using the following item: During the past 12months, on how many
days did you drink alcohol (0=never to 6=every day/almost every day)?
49
SE
Minimum
Maximum
1.06
0.01
0.10
0.39
4.62
0.82
60.34
1.04
0.41
14.75
0.50
0.57
6.04
2.84
6.55
0.15
4.71
0.01
0.00
0.03
0.01
0.41
0.01
0.00
0.01
0.00
0.00
0.02
0.01
0.04
0.00
0.01
3.3
0
0
0
0
0
0
10
0
0
0
1
0
0
1
7.7
1
32
4.3
270
6
1
19
1
1
10
4
16
1
5
It was drawn from the first in-home sample (T2). The average respondent
reported quite low levels of alcohol use (mean=1.06) (Table 1).
Independent variables
All independent variables were drawn from the in-school survey (T1).
Positional benefits.Like Haynie (2001), we used density, popularity, and
Bonacich centrality as sociometric independent variables. Popularity is the
number of friendship nominations a person receives from others (in-degree;
mean=4.62).6 Density measures the ratio of the number of ties present in a
network to the number of pairs in that network. In other words, it is a measure of the proportion of all possible connections in a network that are actually made. Networks in which the majority of people are friends with each
other are highly dense (cohesive). In this study, 39% of all possible ties are
made in the average friendship network (using the receive-network7).
Bonacich centrality indexes a persons connectivity as weighted by the connectivity of their peers (Bonacich, 1987). It measures egos ties to socially
powerful others and therefore indicates how socially powerful the individual is through their connection to prominent others (B set to 0.18). People
50
with higher Bonacich centrality scores are therefore those who have more
friends who are themselves highly connected (mean=0.82). Unlike Haynie
(2001), we also included reach in three steps as another measure of centrality. This is a more straightforward measure of centrality than Bonacich centrality9 as it indicates how many others a person can reach within three
ties.10 Those who can reach a greater number of people within a small number of steps are thought to be more centrally located and thus visible (prominent) in their network than individuals who can only reach a small number
of people (mean=55.58).
Experienced benefits. This variable was created by combining three related
measures: spending time with friends, talking with friends on the phone,
and confiding in friends (talking with friends about problems). Spending
time with friends was itself coded from three different sets of questions in
which students were asked how many of their nominated friends they spent
time with after school in the past 7days, spent time with last weekend, and
went to their house in the past 7days. Each student could indicate that they
did any of these with a total of 10 friends (five male and five female students). The number of friends that they indicated doing any of these things
with was summed into a scale ranging from 0 to 30. The idea here is that a
person who has more friends that they spend time with is likely to experience greater value from their friendships than a person who spends time
with fewer friends.
The items confiding in friends and talking on the phone with friends
were structured similarly. Students were asked to indicate how many friends
they talked with about a problem in the last 7days and how many
friends they talked with on the phone. Each of these three measures (spending time with friends, confiding in friends, and talked on the phone with
friends) was then standardized and summed (=0.81). Following Haynie
and Osgood (2005), this sum was then divided by the square root of the
number of nominated friends to create the experienced benefits scale
(mean=0.10).11 The meaning of this scale is therefore that people who have
more friends that they spend time with both in person and on the phone and
who they can confide in are most likely to subjectively experience more
value from their peer relations.12
There is substantial overlap with the measure of unstructured socializing
employed by Haynie and Osgood (2005). They examined interactive effects
of peer delinquency and unstructured socializing on individual delinquency
(and found direct effects but no evidence for moderation). Their unstructured socializing measure included some of the same time with friends
measures that are employed in the experienced benefits scale used here.
However, some key distinctions are made. They combined time with
51
friends with an item asking respondents to indicate how often they just
hang out with friends to clearly elucidate the connection to time that is
unstructured, a core concept in their work. By replacing that item with the
talking to peers on the phone and confiding in peers measures, we push
the scale away from unstructured socializing and instead account for aspects
of friendship that are likely to be subjectively valuable. However, it is a
limitation of the study that we cannot directly assess the fear of losing
access to benefits and therefore must make a conceptual leap between possessing access to benefits and fearing their loss.
Peer alcohol use. The measure of peer alcohol use used in this study was
drawn from self-reports by individuals identified as friends (receivenetwork13). These responses were matched back to the respondent, and the
mean level of alcohol use among network members was determined
(mean=1.04). Using the peers own reported behaviors, the potential overestimation of peer similarity introduced by asking respondents to estimate
the alcohol use of their friends is eliminated.
To ensure that what we interpret as peer support for alcohol use as indicated by the level of peer alcohol use is not confounded by the gender composition of the peer group, we control for the proportion of the group that is
male. From the friendship nominations, the number of nominated male
friends was divided by the total number of nominated friends, resulting in a
variable in which a score of 0 indicates that there were no males in the peer
group, and a score of 1 indicates that all members of the group were male
(mean=0.41).
A four-item measure of school problems was controlled by creating a
variable in which responses to the following questions were summed: Since
school started this year, how often have you had trouble: (a) getting along
with your teachers? (b) paying attention in school? (c) getting your homework done? (d) getting along with other students? (all coded 0=never to
4=everyday; =0.837; mean=6.55).14
We also controlled for lack of club involvement by including an item in
which participants responded affirmatively if they do not participate in any
clubs, organizations, or teams at school (0=involved in at least one club/
organization/team, 1=not involved in any club/organization/team;
mean=0.15). This is used as a proxy for delinquent opportunities. The
rationale here is that adolescents who are not involved in any of these activities will have more opportunities to engage in delinquency (see Osgood
etal., 1996). While this is likely to account for much of the variance associated with delinquent opportunities, it is an imperfect proxy as it does not
account for organized activities outside of the school.
52
To account for the fact that adolescents can benefit from family relationships as well as the peer group, we controlled for attachment to parents.
This variable was created from items in which respondents were asked how
much their mother/father cares about them (1=not at all to 5=very much).
The mean score across parents (or for the only parent in the case of singleparent families) was taken (mean=4.71).
Following Bearman etal. (2004) and Young (2011), the socioeconomic
status (SES) measure was based on the educational achievement and occupational status of parents. Education was coded on a six-point scale with the
following categories: 0=never went to school, 1=less than a high
school diploma, 2=high school diploma, 3=some college education
or equivalent, 4=graduated from college, and 5=additional professional education. Likewise, occupation was a six-point scale coded into
0=unemployed, 1=unskilled worker, 2=skilled worker, 3=lowlevel white collar, 4=high-level white collar, and 5=professional.
Scores were then summed to create the measure of family SES, and the
parent with the highest SES score was taken to indicate family SES
(mean=6.04). For single-parent families, the SES score was based on the
responses for that parent alone. Grades were measured by taking the average score across English, history/social science, science, and mathematics
(1=D to 4=A). If a student did not take a particular subject, their score was
based on subjects they did take (mean=2.84). Finally, age (mean=14.75),
race (0=not White, 1=White; mean=0.57), and sex (0=female, 1=male;
mean=0.50) were also controlled.
Analytical methods
The alcohol use outcome measure is a positive count variable with a substantial proportion of cases clustered at the low end and displaying evidence
of overdispersion. Therefore, negative binomial models are ideal. However,
the focus of this analysis is on the multiplicative interactions between peer
alcohol use, experienced benefits, and positional benefits making the use of
negative binomial models inappropriate (see Svensson and Oberwittler
(2010) for a discussion of this issue). We will follow the lead of Svensson
and Oberwittler by using linear regression models while employing corrections to ensure that the model is trustworthy. The main correction was to use
bootstrapped linear models (Efron and Tibshirani, 1993; Mooney and
Duval, 1993; see Russell and Dean (2000) for a discussion of the bootstrap
for moderation analysis). Bootstrapping does not require distributional
assumptions (such as normally distributed errors)[and] can provide more
accurate inferences when the data are not well behaved (Fox, 2008: 587).
53
SE
Experienced benefits
Density
Popularity
Bonacich centrality
Reach in three steps
Peer alcohol use
Proportion male
Age
Male
White
SES
Grades
School problems
Lack of club involvement
Attachment to parents
Intercept
0.226**
0.026*
0.068**
0.005
0.049**
0.351**
0.003
0.212**
0.190**
0.155**
0.045**
0.116**
0.039**
0.063+
0.108**
0.923**
0.013
0.012
0.014
0.016
0.015
0.014
0.013
0.012
0.025
0.024
0.012
0.013
0.011
0.034
0.014
0.020
95% confidence
interval
Lower
Upper
0.201
0.049
0.041
0.025
0.080
0.322
0.029
0.189
0.142
0.108
0.021
0.141
0.016
0.003
0.135
0.885
0.250
0.003
0.096
0.036
0.019
0.379
0.022
0.236
0.238
0.202
0.068
0.091
0.061
0.129
0.081
0.962
Results
Table 2 displays the first-order relationships between the various social benefits measures and alcohol use (including the control variables). It shows
that adolescents who are more likely to benefit from their relationships as
54
SE
95% confidence
interval
Lower
Upper
0.014
0.012
0.015
0.014
0.013
0.041
0.124
0.161
0.072
0.025
0.095
0.077
0.219
0.128
0.077
0.010
0.016
0.018
0.019
0.039
0.069
0.018
0.033
0.001
0.005
0.088
0.043
indicated by the measure combining spending time, talking with, and confiding in friends are likely to have higher levels of alcohol use (experienced
benefits, b=0.226). Of the network position measures, popularity and reach
in three steps are related to alcohol use but in different directions. More
popular adolescents are likely to drink alcohol more often than others
(b=0.068), while those who can reach more peers through their friends tend
to drink less frequently (b=0.049). Additionally, adolescents in more
dense networks report lower levels of alcohol use (b=0.026). Being connected to others with social power (Bonacich centrality, b=0.005) appears
to be unrelated to adolescent alcohol use. Not surprisingly, associating with
alcohol-using peers is related to personal alcohol use (b=0.351). From the
control variables, more frequent alcohol use is related to associating with
being older, male, White, from a higher socioeconomic background, achieving lower grades, having greater amounts of school problems, and reduced
attachments to parents. Associating with greater proportions of males was
not related to alcohol use,15 and club/team involvement display a marginally
significant positive relationship (p<0.1).
Table 3 displays the results of the interaction models. Each coefficient is
from a different regression model that included all first-order variables that
made up the interactions as well as the control variables. To simplify the
presentation of results, only the interaction coefficients are presented.
55
Full results are available in supplementary tables posted online. Nearly all
of the interactions were significant, which highlights the importance of
assessing the dependence between the various factors predicting alcohol
use. The interactions between peer alcohol use and experienced benefits
(b=0.068), density (b=0.101), popularity (b=0.190), Bonacich centrality (b=0.100), and reach in three steps (b=0.051) were all significantly
related to alcohol use. These interaction effects were plotted to assist interpretation. Figure 1 displays the interaction between peer alcohol use and
experienced benefits.16 It shows that people who report being able to talk
with, spend time with, and confide in friends have higher levels of alcohol
use than others at all levels of peer alcohol use. However, the effect of peer
drinking on individual alcohol use is stronger for people who receive high
levels of benefits from friends. That is, they have the most to lose by not
engaging in peer-supported behaviors such as alcohol use. The difference in
slopes is subtle but significant.
The result of the interaction between peer alcohol use and popularity is
more dramatic. Figure 2 illustrates this (the plots for Bonacich centrality
and reach in three steps are available in the supplementary material online).
It shows that alcohol use is lowest among the most popular adolescents
when there are low levels of drinking among peers. However, popular adolescents appear to be particularly sensitive to peer drinking. The slope of the
relationship between peer and personal alcohol use is much steeper for
56
Figure 2. Interaction between peer alcohol use and popularity on alcohol use (T2).
SD: standard deviation.
highly popular adolescents than for the less popular. The interactions
between peer alcohol use and Bonacich centrality and reach in three steps
parallel this dynamic. The key message is that peer effects are strongest
among adolescents who have more to lose. Interestingly, the benefits associated with being in a dense network operate differently in their effect on
alcohol use. Figure 3 shows that adolescents who are most susceptible to
peer influence are those in low-density groups rather than those in highdensity groups. It therefore appears that dense networks somewhat protect
against alcohol use.
The final section of Table 3 displays the three-way interactions between
peer alcohol use, experienced benefits, and network position. The idea
behind these interactions is that it may be that adolescents who both receive
high levels of experienced benefits and occupy a more desirable position in
the network will drink the most when their peer group supports higher levels of alcohol use. Figure 4 shows the effect of the interaction between peer
alcohol use, experienced benefits, and popularity. This supports the hypothesis that adolescents who are most susceptible to peer influence are those
who have the most to lose in a variety of ways. Here, we see that those who
are the most connected to peers (i.e. spend time, talk with, and confide in
peers), have the highest popularity, and have peers who support drinking
exhibit the highest levels of alcohol use. The three-way interaction involving Bonacich centrality, despite having a different sign, has a similar message; those adolescents who experience higher levels of benefits and are
57
Figure 3. Interaction between peer alcohol use and density on alcohol use (T2).
SD: standard deviation.
58
that account for the reversed sign but that do not alter the fundamental interpretation (figure available in the online supplementary material).
Importantly, it appears to be the case that the interactive effects of social
benefits and peer alcohol use on personal alcohol use cannot be attributed to
opportunities (since a proxy for delinquent opportunities was controlled) or
to the gender composition of the peer group. Nor is it the case that social
benefits are endogenous to school problems (i.e. variability in social benefits is not strictly determined by the extent to which a person has problems
in school) since school problems, a self-control proxy, were included in the
models. In all, the evidence suggesting that adolescents who are most susceptible to peer influence are those who experience higher levels of social
benefits is quite robust.
Discussion
The results of this study show the importance of social benefits provided by
the peer group in establishing the conditions under which adolescent alcohol use occurs. We approach this in a more comprehensive way than much
previous work. By including both positional and experienced social benefits, we sought to determine whether it is the power and social status that
come with positional benefits that most effectively influence behavioral
decision making processes leading to alcohol use or more concrete experiences of benefiting from the network (e.g. spending time, talking with, and
confiding in friends). This focus parallels the work of Lin (1999) who noted
that both mobilized and accessed forms of social capital improved status
attainment outcomes. Like accessed and mobilized social capital, positional
and experienced social benefits are likely to be complementary. Those with
greater positional benefits are also more likely to experience social benefits
in various ways. But the focus of this study was to determine whether these
forms of social benefits perform similarly in their capacity to influence
decisions to engage in alcohol use. In doing so, we support the assertion that
approaches to studying the influence of peers on individual behavior that
rely simply on counting the number of peers that are involved in a particular
behavior may be overly simplistic. Where Sutherland had a keen sense for
the importance of context,17 this has mostly been lost on researchers examining differential association effects who tend to view peer influence as a
unitary phenomenon. By ignoring the benefits provided by the peer group,
researchers must make the assumption that peer groups operate similarly for
all adolescents. This is likely to mask variations in the effect of peers on
behavior.
By drawing on the rational choice perspective of Hechter (1987) to
develop a social benefits framework, we find that it is important to account
59
for the degree to which adolescents benefit from their relationships to their
peers. Results suggest that adolescents level of alcohol use is dependent
upon the extent to which they get something that is valued from their peer
relations. Adolescents are more likely to adhere to the norms of the peer
group when it seems like the group is acting in their favor (regardless of
whether or not the individual actually achieves greater returns because of
the group). Therefore, when the group supports drinking, adolescents are
more likely to engage in alcohol use if they risk losing their social benefits
by not doing so. While a person has a choice to engage in a particular behavior or not in any given situation, these dynamics are thought to increase the
odds that they will choose the deviant option. Social networks take time to
form, and adolescents are likely to be keenly aware that they are dependent
upon their school-based network for group interaction and accessing the
benefits provided by peer groups (since schools are the primary venue for
friendship maintenance see Warr, 2002). Therefore, by not acting in ways
supported by the group, adolescents may be jeopardizing their access to the
social status and power that come with popularity and occupying a central
position in the network.
All of the forms of social benefits examined in this study appear to be
valued and therefore play a role in establishing the likelihood of alcohol use.
When adolescents have friends to talk to and hang out with and confide in,
risking these relationships by not acting in ways supported the group may
be incentive enough to ensure behavioral conformity. In other words, if
there is a concern that peers will react negatively to specific behavioral
decisions, then people are likely to adapt their behavior in ways most likely
to elicit positive reactions. When peers are more heavily involved in alcohol
use, peer support for alcohol use is likely to be higher. If these peer relationships are meaningful to the individual, they are likely to drink more often,
regardless of how much they actually enjoy it.
Positional benefits exhibit very similar conditioning effects on alcohol
use. Positional benefits are a product of a persons pattern of relationships
to the other members of the group. Social status (from a sociometric perspective) is a limited commodity in any network, and, as such, some people
will have more friends than others and some people will be located in the
center while others are at the periphery. Since positional benefits equate to
social status and power (Hanneman and Riddle, 2011; Wasserman and
Faust, 1994), they tend to be highly valued, given the inherent desirability
associated with being at the top of the status hierarchy. While all forms of
network position act to constrain behavior, it is popularity and Bonacich
centrality that play the strongest role in influencing the decision to engage
in alcohol use when peers drink more heavily. These are the measures most
60
associated with social power. Adolescents who have many friends and who
are connected to other socially prominent individuals are highly visible
within the network and tend to play a greater role in determining what
actions are acceptable in the peer group. But as such, they also have the
most to lose by not engaging in those same behaviors. Given the desirability
of their social location, adolescents who are popular and connected to influential others are unlikely to do anything to jeopardize it. If this means drinking alcohol on a regular basis, they are more likely to do so, more so than
other adolescents who may regularly talk with friends but who are not necessarily popular or prominent in the network.
Ego network density also interacts with peer alcohol use in its effect on
individual alcohol use but not in the expected way. Instead of stronger constraints toward alcohol use occurring in tightly knit groups of drinkers, what
we find is that higher levels of alcohol use occur when the individual is part
of a more loosely connected group of drinkers. In highly insular friendship
groups where members are closely tied to each other, behaviors are less
likely to be noticed by people outside of the group (to the extent that network cohesion is accompanied by exclusion of outsiders). As such, the
broader network cannot offer rewards for these behaviors since they are not
known. A potentially more plausible explanation is that people who are connected more broadly than to a single insular group are exposed to a greater
variety of behaviors and influences and are therefore more likely to be
involved with peers who support drinking.
There are some interesting differences between the results of this study
and other research that has examined similar dynamics. Haynie (2001)
found significant positive interaction effects between peer delinquency and
popularity, centrality, and density in their relationship to personal delinquency. These match with the current findings for the interactions between
peer behavior and popularity and centrality, but the density interaction differs. Haynie found that adolescents were most highly delinquent when their
peer group was delinquent but cohesive. We find that low density, in interaction with higher levels of peer behavior, is most strongly related to individual behavior. Certain methodological distinctions could account for
these differences,18 but it may be no more complicated than the fact that different outcomes were examined (Haynie focused on a general measure of
delinquency; we focused on alcohol use). Alcohol use is most commonly a
social activity for adolescents. While other forms of delinquency may or
may not be social, they are unlikely to be events in the same way that adolescents commonly engage in alcohol use (e.g. at parties). Therefore, visibility beyond the primary peer group (i.e. low density) may promote alcohol
use in ways that do not hold for other deviant behavior.
61
Another closely related study is the work by Haynie and Osgood (2005)
who examined the extent to which unstructured socializing and peer attachments moderated the effect of peer delinquency on personal delinquency.
Using statistically sophisticated analytical methods, they did not find evidence for either effect. There is some overlap in their unstructured socializing measure and the experienced benefits measure used in this study.
Haynie and Osgood created their measure of unstructured socializing by
combining time spent with friends after school and on the weekend. These
are both included in our measure of experienced benefits. However, it is in
the other items that comprised the scales that the differences lay. Haynie and
Osgood included an item asking how often respondents just hang out with
friends (p. 1117) that specifically gets at the concept of unstructured socializing. We included items that get at deeper levels of connection between
peers by incorporating measures of talking on the phone with friends and
confiding in friends (talking about problems).19 These differences ensure
that the measures are suitable for the distinct interests of the studies but also
mean that they are not directly comparable. Therefore, it is not entirely surprising that the results differ between our study and that of Haynie and
Osgood. The fact that the two studies are concerned with different behaviors
is also likely to account for much of the differences in the interactions found
(or not in the case of Haynie and Osgood).
Importantly, these findings cannot be explained away by other criminological correlates. The multivariate analysis shows that the interaction
effects hold even when controlling for socio-demographics and other important criminological correlates (e.g. proxies for self-control and delinquent
opportunities).
Overall, this study provides support for the work of Hechter (1987). He
argued that a person is likely to behave in ways supported by the group to
the extent that they have something to lose by going against the norms of the
peer group. However, Hechters work has received little attention in the
study of substance use. We show that his theory provides a simple, useful
way to explain variation in peer influence effects. Future work should consider other applications of this framework. For example, it might be particularly applicable to heavy drugusing subcultures. People in these groups
may associate predominantly with others who are members of the same
subculture. They may also be ostracized by more conventional individuals
and groups. In such cases, their only access to social status and power may
come from their membership in the drug-using group at the expense of any
outside sources. Therefore, the possibility of giving up access to their sole
source of social status might act to prevent them from leaving the group. In
this way, social benefits might play less of a role in promoting deviant
62
63
Furthermore, this study does not account for the inherently pleasurable
effects of alcohol use. For some people, alcohol use will be strictly a product of these non-social factors and not the interpersonal dynamics assessed
here. Time 1 measures of individual alcohol use are not suitable controls for
this as it would not distinguish those who are motivated to drink by social
factors as opposed to non-social factors while introducing potentially severe
estimation problems (see Achen (2000) for an in-depth discussion of the
issues associated with this approach). Non-social drinkers are likely to be
the heaviest drinkers. However, replicating models excluding the heaviest
drinkers showed very similar results.
Finally, this study does not account for more complicated bidirectional
dynamics (e.g. as outlined in social exchange theory). This is not to suggest
that exchange dynamics are not important in determining behavior but
rather that the focus of this study was on the viability of the more unidirectional social benefits aspect of behavioral decision making (i.e. by examining the behavioral effect of what people derive from their peer group rather
than what they provide to it).
Conclusion
The premise of this study was that simplistic assessments of the peer influence effect on alcohol use are likely to mask the reality under which peers
have an effect on adolescent drinking. We suggested that adolescents are
most likely to engage in alcohol use when the normative behaviors of the
group support these actions and when the group does something for the
adolescent that they particularly value. Results supported these hypotheses.
Importantly, we examined differences in peer group benefits stemming
from the social status conferred by occupying certain positions in the network (e.g. being in the center of the network) to more concretely experienced benefits (e.g. spending time/talking with and confiding in friends)
and found that both forms of benefits played a role in establishing conditions under which peer alcohol use is likely to be particularly influential for
individual drinking behavior.
Acknowledgements
Special acknowledgment is due to Ronald R. Rindfuss and Barbara Entwisle for
assistance in the original design. Information on how to obtain the Add Health data
files is available on the Add Health website (http://www.cpc.unc.edu/addhealth).
The authors would like to thank Patrick Lussier, Garth Davies, Dana Haynie, Brian
Krauth, Douglas D. Heckathorn, and the anonymous reviewers for their helpful
comments on earlier versions of this article.
64
Funding
This research uses data from Add Health, a program project directed by Kathleen
Mullan Harris and designed by J. Richard Udry, Peter S. Bearman, and Kathleen
Mullan Harris at the University of North Carolina at Chapel Hill and funded by grant
P01-HD31921 from the Eunice Kennedy Shriver National Institute of Child Health
and Human Development, with cooperative funding from 23 other federal agencies
and foundations. No direct support was received from grant P01-HD31921 for this
analysis. Owen Gallupe would like to acknowledge support from the Social Sciences
and Humanities Research Council of Canada (SSHRC) via a Joseph-Armand
Bombardier Canada Graduate Scholarship.
Notes
1. While this perspective overlaps to a certain degree with social exchange theory
in terms of its focus on social relations, it differs in important ways. Where
theorists agree that social exchange involves a series of interactions that generate obligations (Cropanzano and Mitchell, 2005: 874; see also Emerson,
1976), the constraining effect of social benefits has less to do with obligatory
exchanges than it does with a more one-way flow of resources. That is, we are
not concerned with what a person provides to the group; what matters is what a
person receives from the group in terms of creating psychosocial pressures that
influence their decision to act in particular ways to avoid losing access to those
resources. This is not to deny the influence of more bidirectional influences
but rather to investigate the effect on behavioral decision making of this one
particular aspect of peer relations (i.e. social benefits).
2. Although valued sociometric data have the potential to do so as it can be used to
index the strength of ties within a sociometric data set. This is essentially what
is done in this study by using the send-network to measure experienced social
benefits.
3. Even those students who are beyond the age at which they are required to
attend school are generally pressured into completing high school.
4. Some prominent work has outlined the criminogenic effect of time use. Osgood
etal. (1996) suggest that delinquency is a product of situational factors such
that adolescents who spend more unsupervised time with peers are likely to
have more opportunities to commit delinquent acts and that the presence of
peers makes delinquency easier and more rewarding. This does not preclude
our use of time with peers as a social benefit. Spending time with friends is
likely to be simultaneously experienced as something emotionally desirable as
well as providing opportunities for delinquency. While time with peers is multifaceted, we distinguish our use of it as a social benefit from Osgood etal.s
use by scaling it with items measuring talking with, and confiding in, peers.
It is likely that adolescents who regularly talk and spend time with peers and
can discuss problems with them will subjectively value their peer relations to
a greater degree than others. This scaling approach is detailed in the variable
description below.
65
5. The discrepancy between 13,466 and 13,351 is due to missing values on sex
(n=87) and the dependent variable (alcohol usen=28), which were not
imputed. Missing values on the other variables were imputed using chained
equations averaged over 100 iterations. A number of independent variables had
over 5% of cases with imputed values. The grades variable had the greatest
proportion of missing values (20.3%), followed by socioeconomic status (SES;
7.5%), attachment to parents (6.8%), and school problems (5.7%). Imputing
did not change the distributional properties for any variable (very similar
means, standard deviation, skew, and kurtosis between the imputed and unimputed versions).
6. The average respondent nominated 4.59 friends (out-degree).
7. Since social status (in terms of popularity) is based on received friendship nominations, it makes sense that perceived peer network norms are based around
the same set of alters. For this reason, the effect of density is examined using
the receive-network. It is conceivable, however, that norms emanating from
people perceived to be friends (the send-network) or the full ego network (the
send- and receive-network) will differ. For that reason, supplementary models
using the send-network and send- and receive-network measures for density
and peer alcohol use were examined. Results were found to be highly similar
(not displayed).
8. Setting B >0 leads to the interpretation that higher scores are derived from
associating with peers who are more highly connected (Bonacich, 1987).
Setting B<0 means that associating with people who have fewer connections
will lead to higher scores. Since we are interested in the benefits of associating
with more socially prominent others, B=0.1 is appropriate.
9. While Bonacich centrality is often considered the best measure of network
social status, the combination of multiple concepts makes precise interpretation less intuitive. For example, there is no easy answer to the question of who
has greater social status: a person with two highly connected friends or a person
with five poorly/moderately connected friends.
10. For example, consider the network connections A-B-C-D-E. Person A can
reach Persons B, C, and D within three steps, but not Person E as they are four
steps away.
11. As described by Haynie and Osgood (2005),
Dividing by the square root of the number of friends is a compromise
between the sum across friends and the average across friends. The former
is unduly correlated with the number of friends, while the latter makes
the unreasonable assumption that respondents with many friends spend no
more time socializing than those with few friends. (p. 1126)
12. To increase confidence in this scale, we examined its relation to other variables
in an attempt to establish criterion validity. We found that adolescents who
feel close to people at their school, feel part of their school, and feel socially
accepted have significantly higher levels of experienced benefits than others.
We would expect that adolescents who perceive high levels of benefit from
66
13.
14.
15.
16.
17.
18.
19.
References
Achen CH (2000) Why lagged dependent variables can suppress the explanatory
power of other independent variables. Paper presented at the Annual Meeting
of the Political Methodology Section of the American Political Science
Association, Los Angeles, CA, 2022 July.
Aiken LS and West SG (1991) Multiple Regression: Testing and Interpreting
Interactions. Thousand Oaks, CA: SAGE.
Allen JP, Porter MR, McFarland FC, etal. (2005) The two faces of adolescents success with peers: adolescent popularity, social adaptation, and deviant behavior.
Child Development 76(3): 747760.
67
Bearman PS, Moody J and Stovel K (2004) Chains of affection: the structure of adolescent romantic and sexual networks. American Journal of Sociology 110(1):
4499.
Birkbeck C and LaFree G (1993) The situational analysis of crime and delinquency.
Annual Review of Sociology 19: 113137.
Bonacich P (1987) Power and centrality: a family of measures. American Journal of
Sociology 92(5): 11701182.
Brown BB (2004) Adolescents relationships with peers. In: Lerner RM and
Steinberg L (eds) Handbook of Adolescent Psychology, vol. 2. Hoboken, NJ:
Wiley, pp. 363394.
Brown BB and Lohr MJ (1987) Peer-group affiliation and adolescent self-esteem:
an Integration of ego-identity and symbolic-interaction theories. Journal of
Personality and Social Psychology 52(1): 4755.
Camarena PM, Sarigiani PA and Petersen AC (1990) Gender-specific pathways to
intimacy in early adolescence. Journal of Youth and Adolescence 19(1): 1932.
Carrington PJ and Scott J (2011) Introduction. In: Scott J and Carrington PJ (eds)
The SAGE Handbook of Social Network Analysis. Thousand Oaks, CA: SAGE,
pp. 18.
Coleman JS (1988) Social capital in the creation of human capital. American Journal
of Sociology 94(Suppl. 1): S95S120.
Coleman JS (1990) Foundations of Social Theory. Cambridge, MA: Harvard
University Press.
Cropanzano R and Mitchell MS (2005) Social exchange theory: an interdisciplinary
review. Journal of Management 31(6): 874900.
Dijkstra JK, Lindenberg S, Verhulst FC, etal. (2009) The relation between popularity and aggressive, destructive, and norm-breaking behaviors: moderating
effects of athletic abilities, physical attractiveness, and prosociality. Journal of
Research on Adolescence 19(3): 401413.
Dodge KA (1983) Behavioral antecedents of peer social status. Child Development
54(6): 13861399.
Efron B and Tibshirani RJ (1993) An Introduction to the Bootstrap, vol. 57. New
York: Chapman & Hall.
Emerson RM (1976) Social exchange theory. Annual Review of Sociology 2: 335
362.
Ennett ST, Faris R, Hipp J, etal. (2008) Peer smoking, other peer attributes, and
adolescent cigarette smoking: a social network analysis. Prevention Science
9(2): 8898.
Fox J (2008) Applied Regression Analysis and Generalized Linear Models. 2nd ed.
Thousand Oaks, CA: SAGE.
Hanneman RA and Riddle M (2011) Concepts and measures for basic network
analysis. In: Scott J and Carrington PJ (eds) The SAGE Handbook of Social
Network Analysis. Thousand Oaks, CA: SAGE, pp. 340369.
Hannon L and Knapp P (2003) Reassessing nonlinearity in the urban disadvantage/
violent crime relationship: an example of methodological bias from log transformation. Criminology 41(4): 14271448.
68
Harris KM, Halpern CT, Whitsel E, etal. (2009) The National Longitudinal Study
of Adolescent Health: research design (www document). Available at: http://
www.cpc.unc.edu/projects/addhealth/design (accessed 7 December 2010).
Hawkins JD, Catalano RF and Miller JY (1992) Risk and protective factors for alcohol and other drug problems in adolescence and early adulthood: implications
for substance abuse prevention. Psychological Bulletin 112(1): 64105.
Haynie DL (2001) Delinquent peers revisited: does network structure matter?
American Journal of Sociology 106(4): 10131057.
Haynie DL (2002) Friendship networks and delinquency: the relative nature of peer
delinquency. Journal of Quantitative Criminology 18(2): 99134.
Haynie DL and Osgood DW (2005) Reconsidering peers and delinquency: how do
peers matter? Social Forces 84(2): 11091130.
Hechter M (1987) Principles of Group Solidarity. Berkeley, CA: University of
California Press.
Kolenikov S (2010) Resampling variance estimation for complex survey data. The
Stata Journal 10(2): 165199.
Kreager DA (2004) Strangers in the halls: isolation and delinquency in school networks. Social Forces 83(1): 351390.
Lin N (1999) Social networks and status attainment. Annual Review of Sociology
25: 467487.
Meijs N, Cillessen AHN, Scholte RHJ, etal. (2010) Social intelligence and academic achievement as predictors of adolescent popularity. Journal of Youth
and Adolescence 39(1): 6272.
Mooney CZ and Duval RD (1993) Bootstrapping: A Nonparametric Approach to
Statistical Inference. Newbury Park, CA: SAGE.
Newcomb AF and Bagwell CL (1995) Childrens friendship relations: a metaanalytic review. Psychological Bulletin 177(2): 306347.
Osgood DW, Wilson JK, OMalley PM, etal. (1996) Routine activities and individual deviant behavior. American Sociological Review 61(4): 635655.
Pratt TC, Cullen FT, Sellers CS, etal. (2010) The empirical status of social learning
theory: a meta-analysis. Justice Quarterly 27(6): 765802.
Rebellon CJ (2006) Do adolescents engage in delinquency to attract the social attention of peers? An extension and longitudinal test of the social reinforcement
hypothesis. Journal of Research in Crime and Delinquency 43(4): 387411.
Russell C and Dean M (2000) To log or not to log: bootstrap as an alternative to the
parametric estimation of moderation effects in the presence of skewed dependent variables. Organizational Research Methods 3(2): 166185.
Sabongui AG, Bukowski WM and Newcomb AE (1998) The peer ecology of popularity: the network embeddedness of a childs friend predicts the childs subsequent popularity. New Directions for Child and Adolescent Development 81:
8391.
StataCorp (2011) Stata Statistical Software: Release 12. College Station, TX:
StataCorp LP.
Sutherland EH and Cressey DR (1978) Criminology, vol. 10. Philadelphia, PA: J.
B. Lippincott & Co.
69
Svensson R and Oberwittler D (2010) Its not the time they spend, its what they do:
the interaction between delinquent friends and unstructured routine activity on
delinquency: Findings from two countries. Journal of Criminal Justice 38(5):
10061014.
Vaughn BE and Waters E (1981) Attention structure, sociometric status, and dominance: interrelations, behavioral correlates, and relationships to social competence. Developmental Psychology 17(3): 275288.
Warr M (2002) Companions in Crime. New York: Cambridge University Press.
Wasserman S and Faust K (1994) Social Network Analysis: Methods and
Applications. New York: Cambridge University Press.
Young JTN (2011) How do they end up together? A social network analysis of
self-control, homophily, and adolescent relationships. Journal of Quantitative
Criminology 27(3): 251273.
Appendix 1
Spearman correlation matrix.
1
(1) Alcohol use
(T2)
(2) E xperienced
benefits
(3) Density
(4) Popularity
(5) B
onacich
centrality
(6) Reach in
three steps
(7) Peer alcohol
use
1.000
0.230**
1.000
0.040**
0.102**
0.021*
0.072**
0.262**
0.190**
1.000
0.275**
0.077**
1.000
0.430**
1.000
0.043**
0.169**
0.024**
0.378**
0.783** 1.000
0.353**
0.205**
0.264**
0.103** 0.142**
0.015
**p<0.01; *p<0.05.