The Social Context of Adolescent Smoking: A Systems Perspective
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
FRAMING HEALTH MATTERS
The Social Context of Adolescent Smoking:
A Systems Perspective
Cynthia M. Lakon, PhD, MPH, John R. Hipp, PhD, and David S. Timberlake, PhD
Furthermore, such models do not provide spe-
We used a systems science perspective to examine adolescents’ personal
cific guidance about mechanisms through which
networks, school networks, and neighborhoods as a system through which
emotional support and peer influence flow, and we sought to determine whether levels of influence relate to outcomes such as
these flows affected past-month smoking at 2 time points, 1994–1995 and 1996. adolescent smoking. There is a need for theo-
To test relationships, we employed structural equation modeling and used retical models that more specifically and holis-
public-use data from the National Longitudinal Study of Adolescent Health tically elaborate features of the social context of
(n = 6504). Personal network properties affected past-month smoking at both adolescent smoking and how they act in concert.
time points via the flow of emotional support. We observed a feedback loop from
personal network properties to emotional support and then to past-month A SYSTEMS MODEL OF THE
smoking. Past-month smoking at time 1 fed back to positively affect in-degree CONTEXT OF ADOLESCENT
centrality (i.e., popularity). Findings suggest that networks and neighborhoods
SMOKING
in this system positively affected past-month smoking via flows of emotional
support. (Am J Public Health. 2010;100:1218–1228. doi:10.2105/AJPH.2009.
In this study, we incorporate valuable in-
167973)
sights from ecological models that theoretically
partition the environment into levels of influ-
Adolescent cigarette smoking remains a com- information about linkages among individ- ence to frame the social context of adolescent
plex public health problem in the United States. uals, and positional characteristics indicate the smoking as a complex social system. We
Although lifetime smoking and current fre- significance of occupying different network employed a systems science14 approach in
quency of smoking among adolescents de- positions. In general, studies find that isolated conceptualizing the social context of adolescent
creased between the late 1990s and 2003, youths are likely to smoke, although some smoking, which emphasizes interdependence in
prevalence remained unchanged from 2003 studies have found that popular youths are complex relationships among people or organi-
through 2005.1 Smoking prevalence among likely to smoke.3–5,9–12 Implicit in each study is zations.15 The defining features of systems are
adolescents is currently estimated to be around the notion that adolescents’ social context of (1) parts or components yielding a whole that is
23%,1 posing ongoing challenges for tobacco- friends and peers plays a critical role in their own greater than the sum of the parts, (2) inputs and
control efforts. smoking behavior. flows coursing through a system, and (3) feed-
Several streams of literature suggest that Given the relevance of adolescents’ social back loops linking parts of a system. Such system
adolescent smoking is inextricably connected context to their smoking behavior, it is impor- features hold promise for informing theoretical
to the social context in which it occurs. Liter- tant to understand how to conceptualize and models that elaborate adolescents’ social net-
ature shows that an adolescent’s smoking be- measure this context. Previous research has works, the dimensionality of their social rela-
havior will tend to be similar to that of his or used ecological models to inform the theoret- tionships, and the interdependence among levels
her peers.2–6 There is a longstanding debate ical specification of the context of adolescent of influence within their social context.
over why this similarity occurs; some studies smoking and other substance-use behaviors.13 We conceptualized 3 key structures in the
suggest that it is caused by peer influence on an Ecological models allow this context to be theo- social environment of adolescent smoking as
individual adolescent’s smoking,7,8 whereas retically partitioned into levels of influence. structural components defining the system un-
others suggest that it is caused by the individual’s Although there are valuable insights yet to be der study: (1) personal networks of friends, (2)
selection of smoking peers,8 and still others gained from such models, more theoretically school networks, and (3) neighborhoods.
attribute the cause to both influence and selec- informed research is necessary to elaborate the School networks were defined as whole net-
tion.2 Some of the literature implicating adoles- complexity of the social context of adolescent works of all students in a school and the social
cents’ social contexts in their smoking behavior smoking. Moreover, because various theories are ties among them. These networks were con-
examines youths’ social networks of friends and often integrated at different levels in such structed from adolescents’ nominations of up to
peers from a structural perspective. Such studies models, it is difficult to ensure that conceptual 10 friends from a roster listing all names of
focus on how structural and positional charac- coherence is achieved across and within levels, adolescents in their own high school and
teristics of these networks relate to adolescent given the possibility that the theories applied a geographically proximal ‘‘sister’’ junior or
smoking. Structural characteristics reflect at each level make incongruous assumptions. senior high school. Adolescents’ personal
1218 | Framing Health Matters | Peer Reviewed | Lakon et al. American Journal of Public Health | July 2010, Vol 100, No. 7FRAMING HEALTH MATTERS
FIGURE 1—Systems model of contextual structures and flows of social processes influencing adolescent smoking.
networks were subsets of whole school net- network level, conceptualized at the level of the Having reciprocated friendships may facilitate
works and comprised friends nominated within adolescent, we focused on (1) adolescents’ popu- the flow of network resources throughout a net-
their schools and their friends’ ties. The neigh- larity, or their in-degree centrality; (2) whether work, as resources are likely exchanged in
borhoods under study were the physical areas their friendships are mutually reciprocated; (3) mutual friendships, both in personal networks
where the adolescents lived. We investigated personal network density, or the extent to which and in school networks. One study found that
how characteristics of these 3 structural com- those named in adolescents’ networks know one adolescents with reciprocated ties with a best
ponents influenced adolescent smoking via another; (4) the social distance between adoles- friend were less likely to smoke at age 11 and 13
flows of 2 social processes: emotional support cents in their networks or the number of path years than those without such ties.4
and peer influence (Figure 1). We hypothesized lengths between them, which are relationship ties Personal network density reflects the extent
that these social processes were mechanisms linking individuals in a network16; and (5) the to which those in an adolescent’s personal
through which the structural components un- number of people they nominated as friends network know one another. The density of ties,
der study influenced smoking. outside of their schools. either in personal networks or whole networks,
In this introduction, we focused on the paths At the school network level, we focused on likely plays a role in the flow of resources
of main theoretical interest depicted in this the network properties of size, density, and the throughout a network by binding people to-
model; we refer to the other constructs in the mutuality of ties, all of which may affect gether and strengthening adherence to pervasive
Methods section only. Throughout this article, smoking behavior.17 Size refers to the number of norms and beliefs. Dense ties can also limit the
we present the model, findings, and discussion students in a school, density is the extent to inflow of resources from outside a network.
of the results in terms of the causal direction- which students know one another in a school, Adolescents aged 15 years who had dense local
ality assumed in our model, but only for ease of and mutuality is the extent to which youths’ networks were found to have lower odds of
exposition and not for the purpose of making relationships are mutually reciprocated within recent smoking.4 If adolescents are close to one
any causal claims. the broader school context. School-level network another in a network—that is, if path length is
characteristics may affect the structure and po- low—then social influences and network re-
THE ROLES OF NETWORKS AND sitional attributes of personal networks as larger- sources may be easily transmitted because the
NEIGHBORHOODS scale versions of these constructs. probability of transmission decreases over longer
The theoretical intuition underlying each of path lengths.18 Previous research has found that
For this study, we focused on properties of the aforementioned network properties is what adolescents who were socially proximal to
personal and school networks that are relevant to makes them relevant to adolescent smoking a smoker had increased odds of smoking.4 Lastly,
the flow of network resources throughout the from a systems perspective. Popular youths are the number of people youths nominated as
larger system under study and that have been directly connected to many others and can ‘‘friends’’ outside of their schools captures re-
related to adolescent smoking. Network ties carry quickly and disproportionately transmit and lationships that either reinforce or attenuate the
resources that flow through a network and are receive network resources, such as support. effect of social influences from in-school friends,
exchanged by network members, such as social Being central or popular in networks is posi- depending on the types of influences exerted
influence and social support. At the personal tively related to smoking among youths.11,12 by each of these friends. There is evidence that
July 2010, Vol 100, No. 7 | American Journal of Public Health Lakon et al. | Peer Reviewed | Framing Health Matters | 1219FRAMING HEALTH MATTERS
having ties to friends from outside one’s school is generally extends to feelings of closeness, en- structural components have on smoking is peer
positively related to adolescent smoking.4 couragement, and belonging32—and peer influ- influence. In this study, peer influence was
Although previous research offers insight ence, narrowly defined as the influence exerted conceptualized as the influence exerted by an
into how each of these network properties by adolescents’ friends who smoke, as mecha- adolescent’s friends’ smoking behavior on the
relates to adolescent smoking, smoking behav- nisms linking social network ties and adolescent adolescent’s own smoking behavior. The posi-
ior may affect the structure and position of smoking. Although more inclusive conceptuali- tive relationship between peer influence and
individuals in a network via network processes zations of peer influence exist, we focused only smoking is well-documented.39,40 Peer influ-
such as influence, social support, and selection. on the influence derived from adolescents’ ence processes have been measured in relation
We explored this notion in our study by friends who smoke because they may be partic- to smoking in numerous ways, including the
positing feedback pathways from smoking to ularly proximal to and relevant for adolescents’ number of friends who smoke.41,42 Furthermore,
the network characteristics under study. smoking behavior. studies have found that network characteristics
Finally, we took adolescents’ neighborhoods Studies have found that emotional support is affect the number of friends who smoke. For
into account, utilizing insights from social dis- positively associated with smoking.33 Perhaps example, studies have found that reciprocity43
organization theory.19,20 This ecological theory the closeness generated from an emotionally and density of network ties are associated with
posits that the key structural characteristics of supportive tie reinforces social bonding as friends greater peer influence, arguably through tightly
economic disadvantage, racial/ethnic heteroge- smoke together. An analogous finding comes binding people together in a network and am-
neity, and residential instability lead to an overall from the injection drug use literature, which has plifying social influence.44,45 It is also likely
milieu of social disorganization within certain shown that the provision of emotional support in that peer influence is positively related to
neighborhoods, giving rise to higher levels of relationships mediated the association between being central in networks, because highly central
delinquency. In such neighborhoods, this disor- closeness of ties and needle sharing.34 Among individuals can quickly receive and transmit
ganization limits residents’ ability to provide the injection drug users, needle sharing behavior was influence, and to path length in networks, be-
informal social control that would otherwise likely a symbolic act of solidarity, exclusivity, cause social influences may be quickly trans-
reduce adolescents’ delinquent behavior. Al- and bonding. For adolescents, smoking behavior mitted if individuals are close to one another in
though the bulk of research using this theory has may carry a similar symbolic meaning related to a network.
focused on the generation of criminal forms of social bonding, especially in the context of Although there is likely an influence effect of
delinquency,21–23 recent research has tested emotionally supportive and valued friendships. friends’ smoking behavior on adolescent
whether these structural characteristics also in- Moreover, in some friendships receipt of emo- smoking, there is also likely a selection effect, in
crease the delinquent behavior of cigarette tional support may be contingent upon engaging which persons who smoke are more likely to
smoking.24–26 in a delinquent behavior because failure to choose friends who also smoke. To the extent
smoke and being different from a friend on this that having friends who smoke then affects
EMOTIONAL SUPPORT AND PEER dimension might be perceived as a strike against one’s position in the network, we suggest that
INFLUENCE AS SYSTEM FLOWS the friendship. It is not surprising that emotional there will be a feedback loop from personal
support has potential to compromise health network properties, to friends’ smoking be-
To study how the structural components of because previous research has indicated that havior, to smoking at wave 1 (Figure 1).
interest exert influence on adolescent smoking, social support can reinforce delinquent behav- In sum, we examined direct, mediated, and
we hypothesized that 2 social processes— iors among youths and their network members feedback pathways through which these
emotional support and peer influence—may be via modeling processes.35,36 structural components influenced smoking at 2
mechanisms through which properties of these Beyond the effect of emotional support on time points approximately 1 year apart via flows
structural components influence adolescent smoking behavior, there is reason to expect of emotional support and peer influence.
smoking. Emotional support relates to health that the network characteristics under study Given the systems nature of the study, we
through both direct effects and buffering ef- increase emotional support. It is likely that examined a number of pathways in 2 categories:
fects.27 Social support may be a pathway linking more central individuals are in a position to (1) direct pathways indicating how properties of
social networks and health indicators; studies provide support to others or may be connected personal and school networks and neighbor-
have related structural network characteristics to to others who could also provide support.37 hoods influenced smoking and (2) indirect
health via emotional support and social influence Likewise, mutually reciprocated ties, as well as pathways through which properties of networks
processes.28 The findings of previous research the density of ties,38 likely increase emotional and neighborhoods influenced smoking via
highlight the need for future work to focus on support through increased closeness. Also, emotional support and peer influence exerted
how social influence mechanisms may mediate shorter path lengths in networks likely increase by friends’ smoking behavior. Last, we investi-
the relationship between social networks and emotional support because being proximal to gated 2 feedback loops: (1) a loop from network
health,28–30 particularly the association between others may lead to the provision or receipt of properties to emotional support, to smoking, and
certain types of supportive networks and more emotional support. back to the network variables and (2) a loop
health-compromising behavior.31 Therefore, The second social process hypothesized to from personal network characteristics to past-
we focused on emotional support—which potentially mediate the effect that properties of month smoking via friends’ smoking.
1220 | Framing Health Matters | Peer Reviewed | Lakon et al. American Journal of Public Health | July 2010, Vol 100, No. 7FRAMING HEALTH MATTERS
METHODS Structural Equation Modeling models with varying starting values.42 We
We specified the 10 equations in our system also used the Sargan test to determine whether
The data for this study were taken from the as a series of simultaneous equations using our instrumental variables were appropriate,
first wave of the National Longitudinal Study of structural equation modeling. Structural equa- and the test’s nonsignificant results confirmed
Adolescent Health (i.e., Add Health), a school- tion modeling is ideal for our approach because that these instrumental variables were indeed
based longitudinal study consisting primarily of it allows the equations to be estimated simul- independent of the error term in these equa-
in-school and at-home surveys. We also in- taneously with a maximum likelihood estima- tions, as hypothesized. These instrumental
cluded a measure of smoking from the second tor, and it allows reciprocal effects and feedback variables explained a reasonable amount of the
wave of Add Health, which occurred 1 year loops to be specified while appropriately partial variance in the first-stage equation,
after the first wave. The Add Health wave 1 accounting for possible endogeneity. We which is an important indicator of their
surveys were administered to a nationally rep- accounted for clustering within schools by strength.
resentative sample of students in grades 7 computing robust standard errors. Although Note that we cannot estimate the reciprocal
through 12 (and their parents) from September hierarchical linear models handle clustering, effect of the selection-influence relationship
1994 through December 1995; wave 2 sur- they cannot handle reciprocal relationships and between friends’ smoking behavior and ado-
veys were administered from April 1996 to feedback loops, given current software con- lescent’s past-month smoking because we do
August 1996. We used Add Health’s public-use straints. We estimated the model displayed in not have any plausible instrumental variables to
data, which comprise a random sample of Figure 1. Note that the exogenous variables are estimate both of these paths. Thus, having
6504 individuals from the full study. Social depicted on the left-hand side of the figure and friends who smoke likely increases an adoles-
network measures are based on a network are not predicted by any other variables. The cent’s past-month smoking, but those who
elicitation item asking a respondent to nomi- other variables displayed in this figure are smoke are also more likely to associate with
nate up to 5 male friends and 5 female friends; endogenous variables in our system (i.e., each friends who also smoke. Rather than simply
respondents could name persons outside of is predicted by some other variable or vari- assuming that the degree of association between
their school. We also used contextual data from ables). Each of these endogenous variables is these 2 constructs is entirely attributable to an
the 1990 US Census based on the 2407 block represented by an equation based on our influence effect, we adopted a novel technique
groups of residence in the sample (block groups theoretical discussion above. For instance, in- to test the robustness of our system, assuming
had an average population of about 1100 degree centrality is a function of the following various values for the relative proportion of this
persons in 1990). The design sampled first on equation: relationship attributable to the selection ef-
schools and then on students within schools; fect.46 We estimated models in which we fixed
the neighborhoods under study simply arose as ð2Þ in-degree centrality ¼ the selection effect at various values: (1) zero
a byproduct of this sample design because they b1 friends’ smoking behavior þ b2 past selection effect, (2) selection effect one third the
were defined by where adolescents in the -month smoking ðwave1Þ size of the influence effect, (3) equal selection and
sample lived. þC1 neighborhood þ C2 demographics influence effects, and (4) selection effect 3 times
þC3 school network þ f1 ; the size of the influence effect. This technique
Measures has only occasionally been employed.46
The measures used in the analyses are where b1 shows the effect of friends’ smoking Another advantage of structural equation
described in Table 1, along with their summary behavior on the in-degree centrality of the modeling is that it allows us to test the overall
statistics and the levels at which they were respondent, b2 shows the effect of the respon- fit of the model. Structural equation modeling
measured. To aid in identifying the model dent’s past-month smoking on their own pop- allows the specification of a causal model,
(described next), we also included 4 measures ularity, C1 is a vector capturing the effects of the and it can then test how well the model rep-
of smoking risk (measured at the individual neighborhood variables on respondent popu- resents the observed data as a test of causality,
level) that likely affected smoking behavior but larity, C2 is a vector capturing the effects of as has been described elsewhere.47–49 This
not network characteristics. We included de- the respondent’s demographics on his or her tests the similarity of the model’s implied co-
mographic characteristics that were likely re- own popularity, C3 is a vector capturing the variance matrix to the sample covariance matrix.
lated to our endogenous variables. We mea- effects of the school network variables on Although a good model fit would be consistent
sured racial/ethnic heterogeneity on the basis respondent popularity, and f1 is an error term. with our theorized model, it is possible that other
of a dispersion formula: Analogous equations can be written for the models that might be specified could fit these
other endogenous variables. data equally well. Therefore, the causal conclu-
P We allowed for correlated errors among the sions must necessarily be tempered, despite the
K ðN 2 fk2 Þ
ð1Þ D ¼ ; personal network variables and between emo- fact that the specified model is inherently causal.
N 2 ðK 1Þ
tional support and friends’ smoking behavior. The comparative fit index of 0.998, the Tucker-
where K is the number of groups, N2 is the We used various techniques to confirm that Lewis index of 0.991, and the root mean square
number of persons squared, and fk is the our model was identified (i.e., there are unique error approximation of 0.008 suggest an excel-
frequency of group k (D ranges from 0 to 1). values for the parameters), including estimating lent fit for our model.49
July 2010, Vol 100, No. 7 | American Journal of Public Health Lakon et al. | Peer Reviewed | Framing Health Matters | 1221FRAMING HEALTH MATTERS
TABLE 1—Descriptions and Summary Statistics of Study Variables: National Longitudinal Study of Adolescent Health, 1994–1996
Variable How Variable Was Measured Range of Values Mean (SD) or %
Smoking and support (individual level)
Past-month smoking (days, logged), wave 1 Number of days the respondent smoked cigarettes 0 = no days to 30 = 30 days 4.921 (10.082)
during the previous month (log transformed)
Past-month smoking (days, logged), wave 2 Number of days the respondent smoked cigarettes 0 = no days to 30 = 30 days 5.182 (10.299)
during the previous month (log transformed)
Emotional support (proportion) Proportion of friends in the respondent’s personal 0–1 0.347 (0.327)
network with whom they discussed a problem
in the previous 7 days
Friends’ smoking behavior Respondent’s perception of how many of their 3 best 0 = no friends to 3 = 3 friends 0.817 (1.066)
friends smoked
Personal network measures (individual level)
In-degree centrality The number of persons in the network who nominated 1–31 people 5.551 (3.692)
the respondent as a friend
Personal network density The number of existing ties in a respondent’s network Theoretically from 0 to 1; actually 0.412 (0.203)
divided by the total possible number of ties from 0.09 to 1
Mean distance to reachable people This is computed by (1) determining all the people the 1–21.39 5.284 (1.620)
respondent could reach in the network either directly
or indirectly, (2) computing the minimum number of
path lengths to reach each person, and (3) computing
the mean of those distances
Ties outside the school Number of people nominated as friends who were not 0–10 1.406 (2.144)
members of the respondent’s school
Best male friend reciprocates Whether the respondent’s best male friend reciprocated 0 = did not reciprocate, 54.4
their tie choice 1 = reciprocate
Best female friend reciprocates Whether the respondent’s best female friend reciprocated 0 = did not reciprocate, 62.7
their tie choice 1 = reciprocated
School network measures (school level)
School network density The proportion of existing ties to the number of possible Theoretically from 0 to 1; actually 0.017 (0.037)
ties in a school from nearly 0 to 0.35
Size of school network Number of persons in the school network 30–2559 students 671.5 (488.5)
Mutuality index The tendency for ties to be reciprocated relative to Theoretically from 0 to 1; actually 0.377 (0.052)
expectations on the basis of chance; higher values from 0.23 to 0.53
indicate more mutuality
Neighborhood measures (block group level)
Median home value Median value of homes in block group, 1990 $15 000–$300 000 95 407 (62 950)
Racial/ethnic heterogeneity Based on a dispersion formula 0–1 0.340 (0.294)
Residential stability The proportion of residents who moved into their housing unit Low, medium, high 1.996 (0.562)
between 1985 and 1990, categorized into 3 groups, with 1
standard deviation above and below the mean as the cutoffs
Demographic characteristics and smoking
risk variables (individual level)
Age Age of the respondent at the time of the survey 10–19 y 14.871 (1.729)
Mother’s education Highest level of mother’s educational achievement 1 = eighth grade or less, 2 = ninth to 5.275 (2.344)
12th grade, 3 = high school graduate
or GED, 4 = vocational school, 5 = some
college, 6 = graduated from college,
7 = professional or graduate training
Continued
1222 | Framing Health Matters | Peer Reviewed | Lakon et al. American Journal of Public Health | July 2010, Vol 100, No. 7FRAMING HEALTH MATTERS
TABLE 1—Continued
Parent smoking Number of parents who smoked 0 = none, 1 = 1 parent smokes, 2 = both 1.050 (0.781)
parents smoke
Wear seatbelts Frequency of wearing seatbelts when riding in a car 0 = never, 1 = rarely, 2 = sometimes, 3.071 (1.190)
3 = most of the time, 4 = always
Motorcycle riding Frequency of riding a motorcycle in the previous 12 months 0 = never, 1 = once or twice, 2 = about 0.363 (0.863)
once a month, 3 = about once a
week, 4 = almost every day
Cigarettes in home Whether cigarettes are easily available in the home 0 = no, 1 = yes 30.5
African American Self-reported race/ethnicity 1 = African American, 0 = not 24.6
Asian Self-reported race/ethnicity 1 = Asian, 0 = not 4.4
Latino Self-reported race/ethnicity 1 = Latino, 0 = not 13.5
Other race Self-reported race/ethnicity 1 = other race, 0 = not 3.3
White (Ref) Self-reported race/ethnicity 1 = White, 0 = not 54.2
Female Self-reported gender 1 = Female, 0 = male 38.2
Note. Data are from the public-use version of the National Longitudinal Study of Adolescent Health (n = 6504).
RESULTS The results for our full simultaneous equa- with more economic resources had more ties
tions model are displayed in Table 2. Given the outside the school (Table 2 equation 5) and
The summary statistics for the variables used complexity of the results, we will now focus on higher mean distance to reachable people
in the analyses are shown in Table 1. At wave 1, key findings based on our theoretical discus- (Table 2 equation 4). Neighborhood economic
70.5% of the entire sample had not smoked, sion, beginning with the system component of resources had a curvilinear relationship with
whereas this figure was 67% by wave 2. The school network measures. School networks the density of personal networks and friends’
mean days of smoking per month for the entire with greater density and mutuality increased smoking behavior: the peak density of personal
sample increased from 4.9 to 5.2 over the 2 the likelihood that the respondents’ personal networks occurred in neighborhoods with
waves. We performed bivariate analyses on network would have greater density and reci- median home values of approximately
our key outcome measures and found that procity (P < .01; Table 2 equations 1 through 3). $151000, which were somewhat above aver-
Whites smoked significantly more, on average, Respondents in school networks with more age (value = –[0.106/(–0.035*2)] =1.514).
than did other groups (5.65 days per month), density (b = 0.289; P < .01) and mutuality By contrast, the number of smokers in adoles-
whereas Blacks and Latinos smoked signifi- (b = 0.714; P < .10) had greater in-degree cen- cents’ networks was lowest in middle-class
cantly less (1.81 and 3.97 days per month, trality. Results for equation 6 in Table 2 show neighborhoods. Residential stability had a di-
respectively; for all P values in this paragraph, that adolescents in the largest and smallest rect negative effect on in-degree centrality
P < .01). We also found that females smoked schools received the fewest nominations. (b = –0.081; P < .01) and increased past-month
fewer days per month than did males (4.16 When we took the first derivative and set it smoking (b = 0.120; P < .05; in Table 2
vs 5.39), as did those whose mothers had equal to zero, we found that students in equation 9).
higher levels of education (3.11). The pattern is schools with about 1580 students had the Regarding the role that emotional support
similar for friends’ smoking behavior: Whites largest in-degree centrality (size = –[0.766/ plays in the system, results for Table 2 equation
had more friends who smoked (0.87 friends (–0.243*2)] =1.576), whereas students in 8 show that several personal network charac-
who smoked), whereas Blacks (0.53), females smaller and larger schools had fewer nomina- teristics affect the amount of emotional support.
(0.74), and those with more educated mothers tions. Although there was no evidence that A 10% increase in in-degree centrality led to
(0.62) had fewer friends who smoked. For these school network measures directly af- a 0.087 proportionate increase in persons
emotional support, we found that Whites and fected adolescents’ past-month smoking (on the providing emotional support. Likewise, those
females had more (38% and 44% of their basis of ancillary models not presented, in with more ties outside the school, and those
friends provided emotional support, respec- which we added these variables to Table 2 with reciprocation from the best friend (espe-
tively), whereas Latinos (30%) and Blacks equations 9 and 10), adolescents in larger cially females), received an increased amount
(31.5%) had less. Finally, Whites had higher in- school networks did have fewer friends of emotional support. Only personal network
degree centrality (named by 5.95 ties), as did who smoked (b = –0.251; P < .01; Table 2 density showed a negative effect on emotional
females (5.73 vs 5.35 for males), whereas equation 7). support.
Blacks (5.06) and Latinos (4.77) had fewer For the structural component of neighbor- There is also evidence that adolescents with
social ties. hood properties, adolescents from block groups more emotional support engaged in more
July 2010, Vol 100, No. 7 | American Journal of Public Health Lakon et al. | Peer Reviewed | Framing Health Matters | 1223TABLE 2—Simultaneous Equations Model Testing Pathways Among Characteristics of Social Networks and Neighborhoods, Emotional Support,
Peer Influence, and Smoking: National Longitudinal Study of Adolescent Health, 1994–1996
Equation 2: Equation 3:
Equation 1: Best Female Best Male Equation 4: Equation 5: Equation 6: Equation 7: Equation 8: Equation 9: Equation 10:
Personal Friend Friend Distance to Ties Outside In-Degree Friends’ Emotional Past-Month Past-Month
Network Reciprocates, Reciprocates, Reachable the School, Centrality, Smoking Support, Smoking, Smoking,
Density, b (SE) b (SE) b (SE) Others, b (SE) b (SE) b (SE) Behavior, b (SE) Proportion, b (SE) Time 1, b (SE) Time 2, b (SE)
Past-month smoking 0.005 (0.005) 0.002 (0.007) 0.007 (0.008) 0.034 (0.021) 0.009 (0.021) 0.023*** (0.007) 0.085 (NA) 0.553*** (0.018)
Emotional support (proportion) 0.381** (0.155) 0.660*** (0.169)
Friends’ smoking behavior –0.047 (0.039) 0.012 (0.052) –0.106** (0.054) –0.169 (0.158) 0.139 (0.171) –0.153*** (0.057) 0.765*** (0.050) 0.323*** (0.045)
Personal network measures
In-degree centrality 0.233** (0.118) 0.087*** (0.010)
Ties outside the school –0.019 (0.035) 0.032*** (0.003)
Mean distance to reachable people 0.046 (0.066) 0.001 (0.004)
Best male friend reciprocates 0.188 (0.225) 0.022 (0.015)
Best female friend reciprocates –0.324 (0.232) 0.050*** (0.018)
1224 | Framing Health Matters | Peer Reviewed | Lakon et al.
Personal network density 0.083 (0.199) –0.052*** (0.015)
School network measures
School network density 0.117*** (0.025) 0.058*** (0.022) 0.051** (0.023) –0.899*** (0.153) 0.019 (0.161) 0.289*** (0.064) –0.137* (0.080) –0.009 (0.015)
Size of school network 0.024 (0.036) 0.040 (0.038) 0.035 (0.038) 0.138 (0.184) 0.096 (0.247) 0.766*** (0.208) –0.251*** (0.071) 0.019 (0.018)
Size of school network squared –0.243*** (0.063)
Mutuality index 0.467* (0.244) 0.582** (0.238) 0.947*** (0.232) 0.365 (1.682) 1.542 (1.437) 0.714** (0.354) 0.164 (0.511) 0.229* (0.123)
Neighborhood (block group) measures
Median home value 0.106*** (0.031) 0.029 (0.026) 0.015 (0.025) 0.399*** (0.132) 0.217** (0.102) –0.005 (0.036) –0.125** (0.055) –0.010 (0.011) –0.002 (0.125) –0.009 (0.096)
Median home value squared –0.035* (0.019) 0.080*** (0.028)
Racial/ethnic heterogeneity 0.063** (0.031) 0.077** (0.036) –0.035 (0.051) 0.138 (0.137) –0.098 (0.207) –0.008 (0.057) 0.014 (0.060) 0.040** (0.020) 0.099 (0.126) –0.208 (0.165)
Residential stability 0.001 (0.016) –0.032* (0.018) –0.002 (0.021) –0.069 (0.060) 0.016 (0.067) –0.080*** (0.023) 0.019 (0.025) –0.008 (0.011) 0.120** (0.055) –0.111* (0.061)
FRAMING HEALTH MATTERS
Demographic characteristics
African American –0.060** (0.024) –0.115*** (0.034) –0.091** (0.036) 0.023 (0.108) 0.489*** (0.153) –0.110*** (0.038) –0.218*** (0.056) –0.062*** (0.013) –0.967*** (0.115) –0.790*** (0.119)
Asian 0.000 (0.033) 0.028 (0.049) –0.004 (0.054) 0.327* (0.174) –0.132 (0.226) –0.094 (0.064) –0.044 (0.067) –0.036 (0.026) –0.438** (0.198) –0.184 (0.242)
Latino 0.031 (0.032) –0.074** (0.036) –0.075* (0.044) 0.088 (0.082) –0.026 (0.126) –0.104** (0.042) –0.036 (0.058) –0.052*** (0.013) –0.460** (0.190) –0.305 (0.203)
Other race –0.035 (0.030) –0.070* (0.038) –0.063 (0.040) 0.039 (0.063) 0.188 (0.123) –0.101*** (0.036) 0.031 (0.053) 0.008 (0.013) 0.072 (0.134) –0.362*** (0.138)
Age 0.026*** (0.006) 0.025*** (0.007) 0.029*** (0.008) 0.056*** (0.021) 0.098*** (0.032) 0.001 (0.010) 0.087*** (0.013) 0.023*** (0.003) 0.213*** (0.029) 0.019 (0.029)
Female 0.048*** (0.015) 0.315*** (0.020) –0.156*** (0.021) –0.013 (0.078) 0.495*** (0.068) 0.111*** (0.025) 0.068 (0.082) 0.164*** (0.011) –0.131 (0.090) –0.058 (0.076)
Mother’s education –0.028*** (0.007) –0.027 (0.019) –0.029 (0.021)
Parent smoking 0.101*** (0.015) 0.270*** (0.041) 0.209*** (0.043)
Wear seatbelts –0.092*** (0.013) –0.259*** (0.037) –0.091** (0.037)
Motorcycle riding 0.094*** (0.012) 0.323*** (0.050) 0.062 (0.041)
Cigarettes in home 0.207*** (0.037) 0.729*** (0.096) –0.032 (0.101)
R2 0.050 0.134 0.033 0.324 0.035 0.048 0.254 0.338 0.437 0.313
Note. Selection effect of friends’ smoking on individual smoking is fixed at same size as influence effect on individual smoking. Models were estimated using maximum likelihood estimation, with standard errors corrected for clustering within
schools. c258 = 79.6; P = .03. Data are from the public-use version of the National Longitudinal Study of Adolescent Health (n = 6504).
*P < .05 (1-tail test); **P < .05 (2-tail test); ***P < .01 (2-tail test).
American Journal of Public Health | July 2010, Vol 100, No. 7FRAMING HEALTH MATTERS past-month smoking. This effect was positive selection. For example, the effect of adolescent influence from friends’ smoking behavior was (0.381) in the equation predicting past-month past-month smoking on in-degree centrality not affected by personal or school-level net- smoking at the same time point (Table 2 was 0.014 if we assumed no selection effect, work characteristics, it did affect past-month equation 9), and in Table 2 equation 10 for 0.018 if 25% of the relationship was attribut- smoking. Findings pertaining to the school past-month smoking at the next time point able to selection, 0.023 if 50% was attributable component indicate that school-level density (0.66), even after control for past-month to selection, and 0.030 if 75% was attributable and size affected personal network character- smoking at the previous time point. This im- to selection (for all effects, P < .05). Likewise, istics, density, reciprocation of ties from best plies that emotional support plays a long-term the effect of friends’ smoking behavior on in- female friend, mean distance to reachable role in mediating the relationship between the degree centrality ranged from –0.128 to people, in-degree centrality, and number of various personal network measures and –0.171 (for all effects, P < .05). The other nominations outside of respondents’ schools. smoking the following year. For example, those parameters in the model showed even more Findings related to the neighborhood compo- with more social ties inside and outside the stability over these various parameterizations, nent of the system indicate that median home school, and those with reciprocated ties with indicating that our results are robust regardless value negatively affected past-month smoking their best friend, had more emotional support, of the portion of the relationship that was at time 1. Neighborhood characteristics also which led to more smoking 1 year later. attributable to selection and the portion at- affected network characteristics, including There is little evidence that friends’ smoking tributable to influence. density, reciprocation, and ties outside the behavior mediated the relationship between We also assessed whether emotional support school. Findings indicate a feedback loop from personal network measures and past-month and friends’ smoking worked multiplicatively at personal network characteristics, to emotional smoking. On the one hand, friends’ smoking the level of social ties. We tested this by support, to past-month smoking, and then behavior increased past-month smoking (Table constructing a measure that multiplied the back to personal network characteristics, in- 2 equation 9) and past-month smoking at the emotional support score and the smoking cluding the provocative finding that past-month next wave, after control for time 1 smoking behavior for each of an individual’s friendship smoking affected in-degree centrality. (Table 2 equation 10). Each additional friend ties and then computing the average of these Our finding that personal network charac- who smoked increased the number of days values among the ties of an individual. In teristics increased emotional support, which smoked per month by 77%. On the other hand, ancillary models including personal and school then increased past-month smoking, supports there is evidence in Table 2 equation 7 that network measures and demographics, we previous research identifying emotional sup- those who were more popular (i.e., those with tested the effect of this variable on past-month port as a mechanism through which networks high in-degree centrality) had more friends smoking behavior at time 1 and time 2 and relate to health and to risk behavior.29,34 These who smoked. The other network measures had found no significant effects (results not pre- findings were not surprising because being no effect on friends’ smoking behavior. sented; available upon request). Finally, we central in a network likely affords opportunities Although there is little evidence that these estimated a school-level fixed-effects model to for giving and receiving emotional support. personal network measures increased friends’ control for unobserved differences across Second, having friendships outside of school may smoking behavior, there is evidence in Table 2 schools, and the substantive results were very increase the number of friends an adolescent equation 6 that the popularity of adolescents similar to those presented in the text (results has, thus increasing the probability of receiving (in-degree centrality) was affected both by their available upon request). support from any one friend, and may also own past-month smoking and by their friends’ suggest that friends outside of school are close smoking behavior. A 1% increase in past- DISCUSSION friends who provide emotional support. Third, month smoking increased in-degree centrality having a female friend reciprocate friendship by 2.3% (b = 0.023: P < .01). However, there Our findings indicate that when a system of may increase the emotional support exchanged was a countervailing effect if one had friends pathways between characteristics of personal in a mutually reciprocated friendship tie, given who smoked: each additional friend who networks, school networks, and neighborhoods that females may often be viewed as sources of smoked reduced in-degree centrality by 15.3% is taken into account, together with its constit- emotional support. Last, the negative effect of (b = –0.153; P < .01). There is little evidence uent flows of emotional support and the influ- density on emotional support may be explained that past-month smoking or friends’ smoking ence exerted through friends’ smoking behavior, by the numerous relationship obligations and behavior affected the other personal network important insights into the complexity of the constraints that densely connected ties can im- measures (Table 2 equations 1 through 5). social context of adolescent smoking can be pose, constituting a great demand on one’s We performed sensitivity tests of our model. gained. Personal network characteristics—being personal resources. Moreover, if density of ties As described already, we set the parameter for central in a network, having ties outside the limits the resources entering from outside the the effect of adolescents’ past-month smoking school, having a best female friend reciprocate network, this can further limit the amount and on friends’ smoking behavior to various values friendship, and the density of ties—influenced diversity of personal resources to expend as and assessed the robustness of the system. In the flow of emotional support, which in turn emotional support to others. short, the system appeared to be relatively influenced past-month smoking at both time It is notable that the only school-level char- robust, regardless of the ratio of influence to points. We found that although the flow of acteristic to increase emotional support was the July 2010, Vol 100, No. 7 | American Journal of Public Health Lakon et al. | Peer Reviewed | Framing Health Matters | 1225
FRAMING HEALTH MATTERS mutuality index. Perhaps a whole network in-degree centrality and best male friend recip- they smoked, but did not hang out with fellow structure with a large proportion of mutually rocation suggest that having friends who smoke smokers—were the most popular, based on in- reciprocated ties increases the possibility that actually decreases popularity and the reciproca- degree centrality. Alternatively, smokers who emotional support will be exchanged in any tion of friendship ties. Overall, such findings affiliated with friends who smoked were gener- one of these close ties. We observed that suggest that having friends who smoke was not ally no more popular than average adolescents. neither school-level density nor size had an well received among the greater social milieu of Our findings suggest the need to examine effect on the flow of emotional support; the youths in our study. how the pathways represented in the systems former finding is consistent with the idea that Our findings suggest evidence of a feedback model under study might differ across gender density may limit support in a network. The loop: personal network characteristics in- and racial/ethnic groups, given the possible latter finding may indicate that larger schools creased the emotional support received by group differences. Also, future studies should promote anonymity and consequently fewer adolescents, which then appears to have led to examine how other types of social support, support resources, leading to more diffuse more smoking at both waves. In addition, such as confidant support, might function in networks and fewer close and supportive ties. adolescent smoking at time 1 flowed back in the lieu of emotional support in our study model. Our finding that emotional support influ- other direction through the system by bringing Confidant support has been associated with ences past-month smoking at both time points about more friends who smoked (through positive health outcomes50,51and is relevant is consistent with previous work showing that a selection effect) and then leading to greater given the notable effects of reciprocated ties and emotional support positively relates to smok- in-degree centrality. This greater in-degree emotional support, both likely characteristics of ing.33 Perhaps the effect between emotional centrality and greater distance to reachable a confidant relationship, on adolescent smoking support and smoking is more likely to occur people then led to more emotional support, and in this study. between close friends in emotionally supportive thus the loop begins again. Our findings are relationships. The persistence of this relation- indicative of a feedback process that encom- Limitations ship at the second time point may indicate its passes the amplifying effects of personal net- Our study has some limitations. First, the strength and stability over time. work characteristics on emotional support, the network elicitation items were limited in the Although we found modest effects for our reinforcing effect of emotional support on number of friendship nominations. Capping neighborhood component, findings of note smoking, and the effect of smoking on popu- friendship nominations is a common strategy, were the curvilinear relationships that neigh- larity and distance to reachable others. Such though it is a potential drawback among studies borhood economic resources had with the a ‘‘reinforcing’’ loop might suggest that smoking utilizing network generator items. It remains density of personal networks and with friends’ brings social gains in the way of emotional unclear how social position and network smoking behavior. Adolescents living in mid- support and popularity in the social system structure would differ if the number of nomi- dle-income neighborhoods had networks with under study. nations were not capped at this level. Second, the highest density, suggesting a relative co- Our findings also have implications for ex- network data were not collected for the full hesion among their personal ties. At the same tant and future studies that employ the general national sample at wave 2; therefore, we could time, adolescents in middle-income neighbor- strategy of examining relationships between not account for network variables at time 2 hoods had the fewest smokers in their net- network characteristics and smoking among in our models. It is unclear how the inclusion of works, suggesting a relatively low effect from youths. Previous research found a positive these variables might have changed our results. influence of friends’ smoking behavior. It is association between the popularity of students Future studies should include network vari- notable that adolescents in both low- and high- (as measured by in-degree centrality) and ables at multiple time points, to permit obser- income neighborhoods had networks with adolescent smoking, and it has been assumed vation of the evolution of the system. Third, more smokers than adolescents in middle-in- conceptually that the direction proceeded from because we conducted a secondary analysis, we come neighborhoods. popularity to smoking,12 but we have specified were restricted in the types of network vari- Friends’ smoking behavior was not affected a system that allows this directionality to proceed ables, social processes, and outcomes available by any of the network characteristics under in either direction. As a consequence, we were for study. Nevertheless, we investigated theo- study, but it did increase past-month smoking able to detect more evidence that smoking retically informed pathways composing a larger at the first time point. The lack of any effects of behavior and peers who smoked affected one’s system of adolescent smoking. Lastly, what network characteristics on friends’ smoking popularity, rather than the reverse. This finding constitutes a friendship tie is of note here behavior suggests that although these charac- has the potential to inform future models used to because it is unclear whether there was uni- teristics may be important for promoting investigate the relationship between in-degree formity in the strength, duration, and fre- smoking behavior among youths,11,12 they are centrality and smoking among youths. More quency of contact in friendship ties. not important for the smoking behavior of broadly, this finding suggests that a social youths’ friends. This finding runs counter to the behavior—cigarette smoking—could alter an im- Implications for Prevention many studies indicating homogeneity in the portant positional attribute of a social network. It In spite of these limitations, our findings smoking status of friends. The findings that is notable that the individuals who showed provide insight into the importance of the friends’ smoking behavior reduced both a relatively high degree of autonomy—in that strength of reciprocated friendships and the 1226 | Framing Health Matters | Peer Reviewed | Lakon et al. American Journal of Public Health | July 2010, Vol 100, No. 7
FRAMING HEALTH MATTERS
emotional support they can transact to help behavior was not. We found evidence of 4. Ennett ST, Bauman KE, Hussong A, et al. The peer
context of adolescent substance use: findings from
adolescents support each other in remaining a feedback process, as past-month smoking had
social network analysis. J Res Adolesc. 2006;16(2):
nonsmokers or in quitting smoking. These a direct effect on the popularity of students 159–186.
friendship pairs could be targeted for a school- (in-degree centrality). Overall, our findings 5. Pearson M, West P. Drifting smoke rings: social
based intervention, either to help both adoles- suggest complexity in the social context of network analysis and Markov processes in a longitudinal
study of friendship groups and risk-taking. Connections.
cents in a pair remain nonsmokers or so that adolescent smoking and the need for theory to
2003;25:59–76.
they could help each other stop smoking. account for it. j
6. Kirke DM. Chain reactions in adolescents’ cigarette,
This could be done by teaching youths in these alcohol and drug use: similarity through peer influence or
pairs how to use emotional support as re- the patterning of ties in peer networks? Soc Networks.
inforcement for helping one another remain About the Authors 2004;26(1):3–28.
Cynthia M. Lakon and David S. Timberlake are with the 7. Flay BR, Hu FB, Siddiqui O, et al. Differential
nonsmokers (among nonsmoking pairs) and for Department of Population Health and Disease Prevention, influence of parental smoking and friends’ smoking on
considering quitting (among smoking pairs). Program in Public Health, University of California, Irvine. adolescent initiation and escalation and smoking. J Health
Second, adolescents could learn self-regulatory John R. Hipp is with the Department of Criminology, Law Soc Behav. 1994;35(3):248–265.
and Society and the Department of Sociology, University
techniques (e.g., journaling) to help one another of California, Irvine.
8. Hoffman BR, Monge PR, Chou CP, Valente TW.
Perceived peer influence and peer selection on adoles-
identify cues in the social environment that Correspondence should be sent to Cynthia Lakon, De-
cent smoking. Addict Behav. 2007;32(8):1546–1554.
elicit interest in smoking or smoking associa- partment of Population Health and Disease Prevention,
Program in Public Health, University of California, Irvine, 9. Abel G, Plumridge L, Graham P. Peers, networks or
tions. All participating adolescent pairs could 101 Theory Drive, Suite 250, Irvine, CA 92697-3957 relationships: strategies for understanding social dynam-
form task forces in schools and lead smoking (e-mail: clakon@uci.edu). Reprints can be ordered at ics as determinants of smoking behaviour. Drugs Educ
http://www.ajph.org by clicking the ‘‘Reprints/Eprints’’ link. Prev Policy. 2002;9(4):325–338.
awareness campaigns. Reciprocated relation-
This article was accepted November 30, 2009. 10. Fang X, Li X, Stanton B, Dong Q. Social network
ship pairs would become a channel through
positions and smoking: experimentation among Chinese
which antismoking messages permeate per- adolescents. Am J Health Behav. 2003;27(3):257–267.
Contributors
sonal and school networks. C. M. Lakon conceptualized the study, interpreted the 11. Alexander C, Piazza M, Mekos D, Valente T. Peers,
Findings suggest the need to target adoles- findings, and led the writing of the article. J. R. Hipp schools, and adolescent cigarette smoking. J Adolesc
conducted the statistical analyses, wrote the Methods Health. 2001;29(1):22–30.
cents who smoke, have nonsmoking friends,
section, and reviewed drafts of the article. D. S. Timber- 12. Valente TW, Unger JB, Johnson CA. Do popular
and are not yet popular. Research suggests that lake assisted with modeling tobacco use outcomes. students smoke? The association between popularity and
popular youths can set norms in a school smoking among middle school students. J Adolesc Health.
context.12 A corollary is that if popular youths Acknowledgments 2005;37(4):323–329.
smoke, others will emulate them. Building on This research uses data from Add Health, a program 13. Bronfenbrenner U. Toward an experimental ecology
previous research12 suggesting that popular project directed by Kathleen Mullan Harris and designed of human development. Am Psychol. 1977;32(7):513–
by J. Richard Udry, Peter S. Bearman, and Kathleen 531.
youths will need to adopt antismoking norms in Mullan Harris at the University of North Carolina at 14. Midgley G. Systems Thinking. Vol 1–4. Thousand
order for programs to become effective, we Chapel Hill and funded by grant P01-HD31921 from the Oaks, CA: Sage; 2003.
suggest that interventions should target youths Eunice Kennedy Shriver National Institute of Child
15. Leischow SJ, Milstein B. Systems thinking and
Health and Human Development, with cooperative
who smoke before they become popular. Perhaps funding from 23 other federal agencies and foundations.
modeling for public health practice. Am J Public Health.
these youths are not yet frequent smokers, 2006;96(3):403–405.
Special acknowledgment is due Ronald R. Rindfuss and
given that they affiliate with nonsmoking friends, Barbara Entwisle for assistance in the original design of 16. Wasserman S, Faust K. Social Network Analysis:
the Add Health Study. Information on how to obtain the Methods and Applications. New York, NY: Cambridge
and thus may be tolerant of antismoking norms. Add Health data files is available on the Add Health Web University Press; 1994.
Such adolescents could be educated about the site (http://www.cpc.unc.edu/addhealth). No direct sup- 17. Ennett ST, Faris R, Hipp JR, et al. Peer smoking,
risks of smoking with the hope that they would port was received from grant P01-HD31921 for this other peer attributes, and adolescent cigarette smoking:
analysis. a social network analysis. Prev Sci. 2008;9(2):88–98.
adopt antismoking norms, which their non-
smoking friends might reinforce. This interven- 18. Moody J. The importance of relationship timing for
Human Participant Protection diffusion. Soc Forces. 2002;81(1):25–56.
tion would be disseminated through adolescents’ No protocol approval was necessary because data were 19. Shaw C, McKay HD. Juvenile Delinquency and Urban
personal networks and would have the potential obtained from publicly available secondary sources. Areas. Chicago, IL: University of Chicago Press; 1942.
to solidify antismoking norms over time, as 20. Sampson RJ, Groves WB. Community structure and
these messages spread from personal to school References crime: testing social-disorganization theory. Am J Sociol.
networks. 1. Centers for Disease Control and Prevention. Ciga- 1989;94(4):774–802.
rette use among high school students–United States, 21. Osgood DW, Anderson AL. Unstructured socializing
This study demonstrates the merit of using 1991–2005. MMWR Morb Mortal Wkly Rep. 2006; and rates of delinquency. Criminol. 2004;42(3):519–550.
a systems science approach to conceptualize 55(26):724–726.
22. Sampson RJ. Family management and child de-
complexity in the social context of adolescent 2. Ennett ST, Bauman KE. The contribution of in- velopment: insights from social disorganization theory.
smoking. We found evidence of direct path- fluence and selection to adolescent peer group homoge- In: McCord J, ed. Facts, Frameworks, Forecasts: Advances
neity: the case of adolescent cigarette smoking. J Pers Soc in Criminological Theory. New Brunswick, NJ: Transac-
ways and feedback processes. Emotional sup-
Psychol. 1994;67(4):653–663. tion; 1992:63–93.
port was a pathway linking personal network
3. Ennett ST, Bauman KE. Peer group structure and 23. Gottfredson DC, Mcneil RJ III, Gottfredson GD.
characteristics and past-month smoking, but adolescent cigarette smoking: a social network analysis. Social area influences on delinquency: a multilevel anal-
the peer influence process of friends’ smoking J Health Soc Behav. 1993;34(3):226–236. ysis. J Res Crime Delinq. 1991;28(2):197–226.
July 2010, Vol 100, No. 7 | American Journal of Public Health Lakon et al. | Peer Reviewed | Framing Health Matters | 1227You can also read
