Social Science Research 50 (2015) 367–381

Contents lists available at ScienceDirect

Social Science Research

journal homepage: www.elsevier.com/locate/ssresearch

Extracurricular associations and college enrollment

Benjamin G. Gibbs a,b,⇑, Lance D. Erickson b, Mikaela J. Dufur b, Aaron Miles c
a
Department of Sociology, 2032 JFSB, Brigham Young University, Provo, UT 84602, USA
b
Brigham Young University, Provo, UT 84602, USA
c
The Ohio State University, Columbus, OH 43210, USA

a r t i c l e i n f o a b s t r a c t

Article history: There is consistent evidence that student involvement in extracurricular activities (EAs) is
Received 10 November 2012 associated with numerous academic benefits, yet understanding how peer associations
Revised 11 August 2014 within EAs might influence this link is not well understood. Using Add Health’s compre-
Accepted 17 August 2014
hensive data on EA participation across 80 schools in the United States, we develop a novel
Available online 6 September 2014
measure of peer associations within EA activities. We find that EA participation with high
achieving peers has a nontrivial link to college enrollment, even after considering individ-
Keywords:
ual, peer, and school-level factors. This suggests that school policies aimed at encouraging
Extracurricular activities
GPA
student exposure to high achieving peers in EAs could have an important impact on a stu-
Peers dent’s later educational outcomes.
College Enrollment Ó 2014 Elsevier Inc. All rights reserved.

1. Introduction

A college education is critical for socioeconomic mobility and other important life outcomes (U.S. Department of
Education, 2013; Hout, 2012; Arum and Roksa, 2011), but the transition from high school to college is fraught with impor-
tant decision-making and overall uncertainty (Niu and Tienda, 2008; Grodsky and Jones, 2007). Although family resources
and other related factors are important determinants of a student’s likelihood of college enrollment (e.g. Sandefur et al.,
2006), growing evidence has also been able to document the association between college attendance and a student’s peer
networks (Fletcher and Tienda, 2009; Flether, 2012; DiMaggio and Garip, 2012). We add to this research by examining
the extent to which involvement with high achieving peers in extracurricular activities (EAs) associates with a student’s like-
lihood for attending college.
Extracurricular activities are an embedded feature of an U.S. education; more than three-fourths of students participate in
some kind of extracurricular activity (Feldman and Matjasko, 2007; Mahoney et al., 2005). The impact of EAs on academic
outcomes has been debated for decades (see Coleman, 1961; Spady, 1970), but recent work has documented a consistently
positive relationship between participation in EAs and educational outcomes (Eccles et al., 2003; Covay and Carbonaro,
2010; Kaufman and Gabler, 2004; McNeal, 1995). Yet, despite the ubiquity of EA participation and its potentially positive
effects, how participation in EAs is beneficial is still uncertain (Feldman and Matjasko, 2005; Broh, 2002).
Research suggests that EAs are social contexts that can foster positive peer associations (Moody, 2001; McNeal, 1998)—
students interact with each other in ways that can mutually influence their educational experiences. For example, peers can
encourage effective study practices and high educational expectations, as well as an overall positive orientation toward
school (Grodsky and Riegle-Crumb, 2010; Riegle-Crumb and Callahan, 2009; Merolla et al., 2012). Because peers in EAs typ-
ically have disproportionately higher GPAs, come from higher socioeconomic backgrounds, and are less delinquent (Feldman

⇑ Corresponding author at: Brigham Young University, Provo, UT 84602, USA.

E-mail address: Benjamin_gibbs@byu.edu (B.G. Gibbs).

http://dx.doi.org/10.1016/j.ssresearch.2014.08.013
0049-089X/Ó 2014 Elsevier Inc. All rights reserved.
368 B.G. Gibbs et al. / Social Science Research 50 (2015) 367–381

and Matjasko, 2007; Coleman, 1961; Covay and Carbonaro, 2010), it is reasonable to think that EAs can influence these out-
comes, in part, due to positive peer influences within activities. Yet, to date, there simply has not been sufficient data to
explore this explanation empirically. We examine the relationship between the characteristics of peers participating in
EAs and college enrollment using the National Longitudinal Study of Adolescent Health (Add Health). Add Health contains
EA information from all students across the 80 sampled schools in the data. This allows us to know the average GPA of
all students in each EA in each school.
This approach adds to an important task of social scientists, to reveal the myriad ways students are influenced in their
transition to college. Surprisingly, despite acknowledgements that peers play a likely role in why EAs matter (Eccles and
Barber, 1999; Fredricks and Eccles, 2006; Feldman and Matjasko, 2005), we have yet to find any formal assessments of
how the composition of advantaged peers in EAs might influence a student’s educational futures. Because it is difficult to
demonstrate causality, our first step then is to document whether any link is evident, and then to consider whether the size
of the association warrants increased attention among scholars and practitioners of education.

2. Background

2.1. Extracurricular activities

In The Adolescent Society (1961), Coleman argued that EAs undermined educational outcomes by involving students in
activities that competed with academic goals.1 But a lot has changed since—decades of research on EAs have demonstrated
a net benefit to schooling. A student’s participation in EAs has been linked to self-confidence (Mahoney et al., 2003; Gore
et al., 2001), commitment to school values (Kaufman and Gabler, 2004; Marsh and Kleitman, 2002), educational expectations
(Troutman and Dufur, 2007), improved grades (Marsh, 1992; Feijin, 1994; Eccles et al., 2003), higher standardized test scores
(Landers and Landers, 1978; Broh, 2002), better cognitive and noncognitive skills (Covay and Carbonaro, 2010), retention
(McNeal, 1995), and long-term outcomes such as college enrollment (Eccles et al., 2003; Kaufman and Gabler, 2004) and college
completion (Eccles et al., 2003; Marsh, 1992; Marsh and Kleitman, 2002). Participation in extracurricular activities has also been
linked to lower levels of antisocial behavior (Gilman et al., 2004) and, for some students, less involvement in other risky behav-
iors such as drug and alcohol consumption (Eccles et al., 2003; Fredricks and Eccles, 2006).2 Generally, EA participation has a
positive association across a number of outcomes regardless of activity type (e.g. sports, performing arts, school involvement, or
academic clubs) (Eccles et al., 2003; Barber et al., 2001).
To date, it is unclear exactly how participation in EAs may influence these positive outcomes. There is some indirect evi-
dence that participation promotes work ethic (Broh, 2002), structures idle time (Mahoney and Stattin, 2000; Gilman et al.,
2004), and develops a positive self-identity (Broh, 2002; Eccles and Barber, 1999; see Fredricks and Eccles, 2006). We argue
that another potential pathway is through peer associations fostered in EAs.
It is well known that peer networks in adolescence, ranging from romantic partnerships to delinquent peer groups, can
have a powerful influence on student outcomes (Giordano, 2003; Haynie and Wayne Osgood, 2005). In criminological
research, peers strongly affect an adolescent’s illicit substance use (Andrews et al., 2002) as well as the probability of com-
mitting violent crime (Zimmerman and Messner, 2011). Peer associations via friendship networks and school course work
can also be positive, with evidence that peer associations in school are linked to higher academic achievement (Choi
et al., 2008; Crosnoe et al., 2003, 2008). Thus, one possibility is that EAs foster the kinds of peer associations that creates
an environment conducive to positive academic outcomes (Eccles and Barber, 1999; Fredricks and Eccles, 2006; Feldman
and Matjasko, 2005).
To detail this possibility theoretically, we organize the following perspectives from broad to specific—beginning with how
the composition of peers in a given context can influence an individual student, then how this process can work specifically
in EAs, and finally how peer associations in EAs might influence a participant’s chances for college enrollment.

2.2. Extracurricular associations: a compositional approach

To understand how associations with peers in EAs might influence participants, we begin with a compositional frame-
work. Generally ‘‘composition’’ in the education literature refers to the demographic profile of students within a given con-
text (e.g. socioeconomic status, race, gender), often within schools (see Crosnoe, 2009). The study of peer composition and its
link to academic outcomes is largely due to the legacy of James Coleman. With particular attention to racial composition, the
Coleman Report concluded that ‘‘the social composition of the student body is more highly related to achievement, indepen-
dent of the student’s own social background, than is any school factor’’ (Coleman et al., 1966:325). Coleman’s compositional
interpretation of schooling reframed the education debate (Jencks, 1972) and became an important catalyst for racial deseg-
regation policies in schools (e.g., busing) (Mickelson, 2010). More recently there has been a turn in composition studies to

1
Coleman, however, focused most of his analysis on popularity and school context rather than educational outcomes. Within his analysis, the ‘‘Leading
Crowd’’ were the most likely students to have college expectations. He also argued that some activities (i.e. sports), if representing the most popular students in
school, would have a positive relationship with educational outcomes because it would effectively increase one’s commitment to school.
2
But not all studies of EAs uncover ‘‘positive’’ effects. EAs have also been linked to more alcohol use and increased sexual activity, especially among boys in
high school sports (Feldman and Matjasko, 2005).
 B.G. Gibbs et al. / Social Science Research 50 (2015) 367–381 369

examine socioeconomic status and other forms of advantage across schools (see Crosnoe, 2009; Mickelson, 2010). But like
the Coleman Report (1966), conventional approaches often assess compositional differences between schools (e.g., Crosnoe,
2009), where the peer composition of one school is compared to another as a way to understand its impact on students’ edu-
cational outcomes. Although this is an important way to capture composition, it can obfuscate the diversity of experiences
within a school (Gamoran and Berends, 1987; Choi et al., 2008; Moody, 2001).
Several studies confirm just how important within-school composition considerations may be. Moody’s 2001 study pro-
vides an important link between peer associations and EAs, but for a nonacademic outcome—friendship integration by race.
He found that friendship integration within schools was directly linked to the level of integration within EAs. In fact, he cal-
culated that a one standard deviation increase in the racial mixing within EAs decreased school-level racial segregation by a
one-third standard deviation. Thus, it appears that EAs produce a context of peer interactions that is conducive to interracial
friendships. As friendship formation is often the result of preferences for sameness (Blau, 1977; Coleman, 1961), Moody
warns that, ‘‘without organizational structures [such as EAs] to ensure status equality within the school, minority concen-
tration may lead to greater friendship segregation, even as the absolute number of cross-race ties increases’’ (2001:706).
Adapting Moody’s use of racial-mixing theories to explain why EAs might foster interracial friendships, we anticipate a
similar process for extracurricular associations with advantaged peers. First, EAs can organize academically diverse students
around common interests that can supersede achievement-based friendship preferences. Second, EAs can provide opportu-
nities for social mixing (Blau, 1977), where students are given a structural opportunity to meet and build friendships with
others with whom they would not normally have interactions. Finally, borrowing from contact theory, EAs can be a setting of
status equality where students employ cooperative interdependence to achieve specified goals (e.g. winning a game, per-
forming in a school assembly) (see Moody, 2001).
Thus, if EAs can shape friendship segregation by race—as Moody (2001) demonstrates—might they also influence educa-
tional outcomes by exposing participants to high achieving peers in much the same way? In essence, we argue that EAs func-
tion as ‘‘sites of associations’’ that facilitate meaningful integration with academically diverse peers who provide important
social resources needed for successful educational attainment (Mahoney et al., 2005; McNeal, 1998). These sites are unique
from other social settings in schools because EAs are a formal context where peer associations are subject to more external
regulation than friendship networks. For example, institutional agents (school administrators) are primarily responsible for
creating, shaping, and sustaining EAs within a given school. Consequently, school administrators can set guidelines for par-
ticipation like grade eligibility and attendance requirements that, when violated, could block students from further partic-
ipation (Moody, 2001; Quiroz et al., 1996). These formal structures thus influence both the type of peers students will
encounter and the quality of the environment in EAs. Of course, EAs are also informal contexts where peer associations
are less regulated by institutional actors than other school settings (e.g. classrooms) and can even approach the kind of infor-
mal student interactions and norms typically found in voluntary peer associations (Moody, 2001). Thus, EAs represent an
intersection of both formal and informal forms of peer association that may be an important reason why EAs are linked
to so many educational benefits.

2.2.1. Social capital, local positions, and brokerage

How might ‘‘extracurricular associations’’ influence student participants? We draw upon social capital theory, Frank
et al.’s (2008) concept of ‘‘local positions,’’ and Small’s (2009) concept of ‘‘brokerage’’ to consider this relationship. The types
of social capital Coleman (1988) identified as being particularly important to academic outcomes—norms, information, obli-
gation—are also related to students’ aspirations and ability to attend college. Of the three, norms may represent the clearest
mechanism within EAs. Peer norms can emphasize the desirability of higher education and the expectations that students
can and should attend college, transmitted across peer ties in the kinds of networks built through various activities, clubs,
and teams. In such an environment, peers may not be asking each other whether or not they plan to attend college, but where.
This can create a normative environment where college is assumed rather than questioned. This may be why the educational
aspirations of friends (including the educational level of the parents of friends) are important for college completion, as
Cherng et al. (2013) demonstrate.
Of course, social capital may also work through the exchange of information and the influence of social obligations fos-
tered in EAs. For example, information about how to apply to, pay for, and navigate college might be more effective coming
from these more informal networks than from other sources (Tierney and Venegas, 2006). Social obligations can occur
through the often repetitive, teambuilding nature of many EAs—creating a sense of reciprocity where students feel a com-
mitment to extend help and resources to other participants. As these forms of social capital exchange are possible, given that
EA environments are less formal and less time-consuming than other school settings (e.g. classrooms, see Choi et al., 2008;
Fredricks, 2012), it is likely that both information and social obligations are less salient (but still important) explanations for
how peer social capital operates in EAs when compared with a normative peer culture of college-going.
So how might a normative culture influence participants in EAs? Students often create tight, closed networks within EA
groups where important social capital may be shared (Broh, 2002; Johnson et al., 2001) and activities have the potential to
create a culture of achievement where high educational expectations emerge and are fostered (Guest and Schneider, 2003)—
what Harris (2010) describes as ‘‘group-based contagion.’’ This normative culture of academic expectations is a likely result
of shared interests of participants that are strongly associated with their socioeconomic backgrounds (Feldman and
Matjasko, 2005), but EAs might also facilitate some degree of socioeconomic integration (e.g. Moody, 2001) where partici-
pants take on new views and behaviors regarding school as a result.
370 B.G. Gibbs et al. / Social Science Research 50 (2015) 367–381

Specifically, Frank et al.’s (2008) term ‘‘local positions’’ is a helpful way to understand how EAs might encourage college
enrollment through peer associations. Local positions are defined as ‘‘a set of students within a school who participate in a
set of [activities] that differentiate them from other students in the school’’ (p. 7). Though they use the term ‘‘positions,’’
Frank et al. are describing social interactions between and across actors, a mechanism through which social capital is cre-
ated. Thus, EAs provide the setting or local positions where forms of capital can be transmitted across peers.
Small (2009) further explains how local positions can enhance social capital with his concept of ‘‘organizational embedd-
edness’’—or in our case students embedded within a context of EAs that provide constraints, imperatives, and opportunities
for individuals to interact and share resources (2009). Thus, extracurricular activities in a school can effectively broker the
development and exchange of social capital between participants. Small describes brokerage as ‘‘the general process by
which an organization connects an individual to another individual, to another organization, or to the resources they con-
tain’’ (Small, 2009:19). Small demonstrates this with childcare centers, where a mother’s economic and emotional well-
being is associated with friendships formed with other mothers of enrolled children.
Likewise, EAs provide a structure where positive peer interactions are possible. Activities such as band, tennis, and debate,
allow for brokerage simply by structuring time for students to interact. They also create social positions within the school
where teachers, coaches, parent volunteers, and students have clearly-defined, shared goals and formal motivations to trans-
mit pro-educational norms. When EAs are understood this way, it also suggests that activities may not be equally effective in
brokering positive peer associations, especially for activities that attract fewer advantaged peers than other activities.
To date, the standard ways EAs have been measured and modeled have not captured the concept of extracurricular asso-
ciations—how the role of EAs’ ‘‘local positions’’ can broker positive peer social capital. We introduce a method that better
approximates this process by examining the peer composition of EAs. Specifically, we examine whether associating with aca-
demically advantaged peers in EAs is linked to college enrollment.

2.2.2. EAs and college enrollment

Several studies find that EA involvement is associated with college enrollment (Zaff et al., 2003; Gardner et al., 2008;
Marsh and Kleitman, 2002; Marsh, 1992). It is well established that peers matter in other contexts for college enrollment
(e.g. Sokatch, 2006; Terenzini et al., 1994) and many researchers suggest that exposure to advantaged peers may help explain
some of the EA effect (Eccles et al., 2003, Gardner et al., 2008; Kaufman and Gabler, 2004). In much the same way we con-
ceptualize EAs, Choi et al. (2008) show how these two research traditions overlap in settings similar to EAs—the classroom.
Choi et al. (2008) find that the socioeconomic characteristics of grouped course-mates are associated with a student’s
likelihood of college enrollment—when students are exposed to a high proportion of peers whose parents are college edu-
cated in their classes, they work harder on assignments and are more likely to enroll in college. They speculate that inter-
acting with advantaged course-mates is linked to college enrollment because these peers have high educational expectations
and aspirations that likely influence the normative views college for other course-mates, they likely provide information
about college, and they may even encourage other students to pursue demanding coursework in ways that help prepare
them for the demands of the college application process (2008, see also Bankston and Caldas, 1998).
Similarly, Frank et al. (2008) find that advancement in math (especially for girls) depends partly on the course-taking
norms of their peer groups, creating pressures to conform. They find that social positions also matter in the pathways that
lead to college enrollment. Thus, as with coursework, the peer composition of EAs may also be an important mechanism for
understanding the link between EAs and college enrollment. Together, these studies demonstrates that thinking about col-
lege is a much more complex decision-making process than the weighing of income rewards; nor is the acquisition of a col-
lege degree simply the result of a student’s human capital (Manksi, 1990; Beattie, 2002; McLeod and Kaiser, 2004).

2.2.3. Causality
Of course, students may sort themselves into EA’s based on pre-existing characteristics and/or attitudes. In particular,
there is a tendency for already advantaged students to choose activities with a high ratio of advantaged peers (Feldman
and Matjasko, 2005). This complicates the task of producing an estimate of how participation in EA’s with other advantaged
participants might influence college enrollment. It may be that, instead of extracurricular associations influencing a student’s
academic outcomes, the ‘‘effect’’ is simply a by-product of homophily—‘‘birds of a feather flock together.’’ For example, does
participation in EAs with advantaged peers increase the likelihood of college enrollment or does being advantaged with the
accompanying high likelihood of enrolling in college increase the likelihood of participation in EAs with advantaged peers?
Disentangling the order of causation matters for potential policy efforts that would use EAs as an intervention to increase the
likelihood of college enrollment.
We account for some of this uncertainty by considering alternative causes of college enrollment (Choi et al., 2008; Covay
and Carbonaro, 2010; Eccles et al., 2003; Wimer et al., 2008; Broh, 2002; Eitle and Eitle, 2002). We have identified three spe-
cific concerns relevant for this study that would challenge causal interpretations of our findings, namely that a measure of
peer composition of EAs really proxy unmeasured individual characteristics, preexisting peer friendships who also happen to
participate in the same EA,3 or school-level composition effects. As we describe below we control for a variety of individual

3
We are grateful for an anonymous reviewer for helping us develop the method necessary to account for this issue.
 B.G. Gibbs et al. / Social Science Research 50 (2015) 367–381 371

characteristics, remove participant’s friends from EA rosters, and model school-level fixed effects. These approaches can help to
minimize—but not eliminate—concerns about causality.

3. Data and methods

3.1. Data

We use the National Longitudinal Study of Adolescent Health (Add Health). Add Health is a nationally representative
study of adolescents in Grades 7–12 in the United States in 1994. The data include in-depth interviews with adolescents
and their parents, which provide detailed information on child outcomes, family and peer relationships, and school and
neighborhood characteristics. Add Health uses a multistage, stratified, school-based cluster sampling design. Included in
the sample are students from 80 high schools (both public and private) and a corresponding feeder junior high or middle
school. The study sampled some race-ethnic groups in proportion to their size within the U.S. population; other race-ethnic
groups were oversampled. Additional information on the study can be found in Harris et al. (2003).
Add Health data in our study include the in-school survey (1994), the school administrator survey (1994), the parent sur-
vey (1995), and the first and fourth waves of in-home interviews (1995 and 2008). The in-school survey is particularly
important for this study because it surveyed all students in the sample schools who attended the day of the survey (approx-
imately 90,000). Therefore, we can account for the composition of students in each EA across each school.
Attrition across waves was modest, with 90 percent of sampled respondents in Wave 1 participating in Wave 2, 73 per-
cent in Wave 3 and 76 percent in Wave 4. Although attrition may compromise the generalizability of the data, analysis else-
where suggests that the data accurately represent the population of students in each year sampled (Chantala et al., 2004).
Because the nature of many EAs changes as students transition from junior to senior high school (e.g., sports teams
become more competitive), our analyses only include respondents who were in high school (grades 9–12) at Wave 1. We
also limited the analytic sample to (1) those who answered questions about participation in EAs (2) those who were followed
from high school to college age and (3) those who reported participating in 10 or fewer EAs. We excluded respondents with
11 or more activities over concerns their data were not accurate or a special case of EA participation (this applied to 60 indi-
viduals in the data). After excluding respondents who did not satisfy these criteria, 8087 cases remained. There were no
missing data on college enrollment for these remaining cases.

3.2. Measures

3.2.1. Dependent variable

At Wave 4, individual ages ranged from 26 to 32.4 This provides enough time to include respondents who may not have
enrolled in college directly out of high school. At Wave 4, respondents were asked, ‘‘What is the highest level of education
you have achieved to date?’’ College enrollment is coded 0 for never enrolled in post-secondary education, 1 for enrolled in a
vocational or technical college, and 2 for enrolled in a 4-year college (responses of ‘‘some’’ experience with vocational or 4-year
college were considered enrolled as well as those who graduated). We separate technical from traditional 4-year colleges
because these schools have important differences—selection processes vary because of tuition costs, structure of course-taking,
and vocational training, to name a few (see Stephan et al., 2009). As a result, it is important to separate these two forms of col-
lege enrollment as EAs and other covariates may operate differently based on in which type of college a student has chosen to
enroll. By using multinomial logistic regression, we can then examine if exposure to advantaged peers in EAs is similarly linked
to enrollment in technical and more traditional college programs when compared with students who did not attend college.

3.2.2. Extracurricular activities and associations

Respondents reported which EAs, if any, they participated in during the in-school survey. The question read, ‘‘Here is a list
of clubs, organizations, and teams found at many schools. Darken the oval next to any of them that you are participating in
this year, or that you plan to participate in later in the school year.’’ This wording is necessary to capture participation when
the survey was completed outside of the school year and activities that take place later in the school year, such as seasonal
sports. It is possible that some respondents marked an activity they intended to participate in and then did not. EA partic-
ipation is coded 1 if the respondent reported participating in any of 31 EAs. The variable is coded 0 if no EA participation was
reported. We excluded ‘‘Other’’ and ‘‘Other sport’’ categories because we could not compile a roster of other students in each
school for an open-ended category of ‘‘other.’’ Finally, because there is evidence that the quantity of EAs a student partici-
pates in is also linked to educational outcomes (Gardner et al., 2008), we calculated a count of the number of EAs a respondent
reported.

4
The nature of EA participation and the role of EA student composition on college enrollment is most likely to be important for ‘‘timely’’ enrollees rather than
for those who enroll some years later as non-traditional students. However, the Add Health does not have data specific to the timing of enrollment available for
all respondents who enrolled in a post-secondary educational program. Consequently the measure we employ assesses having ever enrolled. To the extent that
non-EA factors influence non-traditional college enrollment (i.e., enrollment is a result of seeking a higher credential to advance in one’s existing career), our
estimates are likely biased downward, suggesting that the results we present here are somewhat conservative.
372 B.G. Gibbs et al. / Social Science Research 50 (2015) 367–381

The activity type measure aggregates the responses into sports teams (baseball/softball, basketball, field hockey, football,
ice hockey, soccer, swimming, tennis, track, volleyball, and wrestling), academic activities (French, German, Latin, Spanish,
book club, computer, debate, history, math, and science), performance activities (drama, band, cheerleading/dance, chorus or
choir, and orchestra), and school activities (newspaper, honor society, student council, and yearbook). These groupings fit
conventional categorizations of EA types (Fredricks and Eccles, 2006).
Using data on the grades of EA participants, we created an EA peer composition measure.5 First, we calculated the mean
score of math, English, history, and science grades reported by each respondent in the in-school survey. Respondents reported
their grade as D = 1, C = 2, B = 3, and A = 4. Second, we calculated the mean GPA of all participants in all of the EAs in which the
respondent reported participating. Because high school sports are gender-segregated, male and female respondents who
reported playing the same sport were considered to be in different EAs. We did not consider EAs with three or fewer partici-
pants. To account for the potential that pre-existing friendships might affect the respondent’s decision to participate in EAs
and to avoid (as much as possible) confounding the role of peer relationships that are limited to EAs and those whose scope
goes beyond EAs, we excluded the GPA of peers who respondents identified as ‘‘friends’’ who also participated with them in
a given EA when calculating this measure.
We should note one unique strength of our EA peer composition measure. This measure is adjusted for each school. We use
data about participation in school-based EAs (e.g., the actual co-participants in a respondent’s EAs) rather than the national-
level composition of the EA (e.g., the mean composition measure of all participants in all schools sampled). This allows for an
unprecedented assessment of EAs—rather than take the national average of GPA of participants in a given EA, we instead
create a school specific GPA of participants in each activity. This approach is not only more precise, but it allows for the very
real variation in peer composition that undoubtedly occurs across the United States. In some schools, high achieving stu-
dents may flock to certain activities that in others they may avoid.

3.2.3. Other factors

We include several measures that are well-documented predictors of college enrollment, taken from the Wave 1 in-home
survey (except where noted). Gender is measured as female (1 = female, 0 = male). Age is measured in years (ranging from
ages 13–19 in Wave 1 and 26–32 in Wave 4). We include individual-level measures for race-ethnicity (including categories
for white, black, Asian, Hispanic, and Native American), GPA, expectations for college enrollment (included in models as a
categorical variable ranging from 1 ‘‘low’’ to 5 ‘‘high’’), and the educational level of the child’s parents. To approximate cog-
nitive skill, we use the Add Health Picture Vocabulary Test (PVT)—an abridged version of the Peabody Picture Vocabulary
Test—Revised (Dunn, 1981)—that was designed to measure vocabulary for Standard American English (scores were stan-
dardized by age to have a mean of 100 and standard deviation of 15).
Intact family was coded as 1 if the respondent lived with both biological parents and 0 in all other cases. Parental employ-
ment was coded 0 for unemployed, 1 for manual and 2 for non-manual.6 This variable was the father’s employment status if
available and employed. Otherwise the mother’s employment status was used. Because of its skewed distribution, we included
the log of 1+ household income (from parent report).7 Welfare was coded as 1 if the responding parent reported ‘‘receiving pub-
lic assistance, such as welfare’’ and 0 otherwise. Finally, we included the mean GPA of each respondent’s nominated peers. Col-
linearity diagnostics show measures to be sufficiently independent from each other.

4. Analysis

All analyses were performed in Stata 13.1 (Stata Corp, 2013). Many of the variables had missing values ranging between 1
and 25 percent of the sample (see Table 1). Methodologists currently suggest that analysts avoid traditional missing data
treatments, like listwise deletion, because biased parameter estimates may result (Allison, 2001; Enders, 2010). We gener-
ated 20 imputed data sets, each using all variables from the models reported in the study, and analyzed the subsequent data
as implemented in Stata’s mi routines. There was no missing for the dependent variable.
First, we present a scatterplot of the mean GPA of EA members and the percent of EA members who enrolled in college for
all 31 EAs to demonstrate the relationship between peer composition of EAs and participants’ rate of college enrollment.
Second, we estimate a multinomial logistic regression model predicting college enrollment (i.e., not enrolled, enrolled in

5
It is important to note that the EA peer composition variable is missing for students who do not participate in EAs. Without adjustments, any analysis that
includes the peer composition measure would necessarily exclude nonparticipants. To overcome this problem, we followed Allison’s (2001, see footnote 4)
suggestion and assigned the mean composition value of participants to all nonparticipants. This allows all nonparticipants to be included in the analysis but
assigns each a value that does not influence estimates for peer composition. Then, to account for nonparticipants, a participant dummy variable is included for
analysis estimating the peer composition measure.
6
Manual occupations include restaurant worker or personal service, such as waitress, housekeeper; craftsperson, such as toolmaker, woodworker;
construction worker, such as carpenter, crane operator; mechanic, such as electrician, plumber, machinist; factory worker or laborer, such as assembler, janitor;
transportation, such as bus driver, taxi driver; military or security, such as police officer, soldier, fire fighter; and farm or fishery worker. Non-manual
occupations include professional 1, such as doctor lawyer, scientist; professional 2, such as teacher, librarian, nurse; manager, such as executive, director;
technical, such as computer specialist, radiologist; office worker, such as bookkeeper, office clerk, secretary; and sales worker, such as insurance agent, store
clerk. We also estimated models without collapsing given responses into manual and non-manual. The results were substantively similar so estimates with the
more general categories are shown for parsimony of presentation.
7
The log of 0 is undefined, it is common practice to add 1.
 B.G. Gibbs et al. / Social Science Research 50 (2015) 367–381 373

Table 1
Means (or proportions), standard deviations, minimum, maximum and percent imputed of study variables. Source: National longitudinal study of adolescent
health.

Meana SD Min Max Percent imputed

EA Participator .71 .45 .00 1.00 0
EA composition
Mean GPA of EA peers 2.93 .25 1.50 3.83 0
EA type
Sportsb .49 .50 .00 1.00 0
Baseball/Softball .17 .37 .00 1.00 0
Basketball .16 .37 .00 1.00 0
Track .12 .33 .00 1.00 0
Football .11 .31 .00 1.00 0
Volleyball .07 .25 .00 1.00 0
Soccer .06 .25 .00 1.00 0
Swimming .04 .19 .00 1.00 0
Tennis .04 .20 .00 1.00 0
Wrestling .03 .17 .00 1.00 0
Field hockey .01 .08 .00 1.00 0
Ice hockey .01 .10 .00 1.00 0
Performanceb .30 .46 .00 1.00 0
Band .10 .31 .00 1.00 0
Cheerleading/Dance .09 .28 .00 1.00 0
Chorus/Choir .09 .29 .00 1.00 0
Drama .07 .26 .00 1.00 0
Orchestra .02 .13 .00 1.00 0
Academicb .27 .45 .00 1.00 0
Honor society .10 .30 .00 1.00 0
Spanish .07 .26 .00 1.00 0
French .04 .19 .00 1.00 0
Math .03 .17 .00 1.00 0
Science .03 .18 .00 1.00 0
Latin .02 .13 .00 1.00 0
Computer .02 .13 .00 1.00 0
Debate .02 .15 .00 1.00 0
FFA .02 .15 .00 1.00 0
German .01 .11 .00 1.00 0
Book .01 .08 .00 1.00 0
History .01 .09 .00 1.00 0
Schoolb .16 .37 .00 1.00 0
Student council .08 .27 .00 1.00 0
Yearbook .08 .27 .00 1.00 0
Newspaper .04 .20 .00 1.00 0
Number of EAs 1.77 1.77 .00 10.00 0
College enrollment
None (High School Degree only) .20 .40 .00 1.00 0
Vocational College .10 .30 .00 1.00 0
4-Year College .71 .45 .00 1.00 0
Individual-level controls
Female .54 .50 .00 1.00 0
Age (in years) 16.32 1.22 13.00 21.00 0
White .52 .50 .00 1.00 0
Black .42 .82 .00 2.00 0
Asian .23 .81 .00 3.00 0
Hispanic .69 1.51 .00 4.00 0
Native American .11 .73 .00 5.00 0
Intact Family .54 .50 .00 1.00 0
Parent educational attainment
No high school degree .12 .32 .00 1.00 0
High school degree .51 .87 .00 2.00 0
Some college .85 1.35 .00 3.00 0
College degree 1.38 1.90 .00 4.00 0
Parent occupation
Unemployed .08 .27 .00 1.00 0
Manual .28 .45 .00 1.00 0
Non-Manual .64 .48 .00 1.00 0
Logged household income 3.56 .88 .00 6.91 25
GPA 2.78 .75 1.00 4.00 0

(continued on next page)

374 B.G. Gibbs et al. / Social Science Research 50 (2015) 367–381

Table 1 (continued)

Meana SD Min Max Percent imputed

PVT 100.87 14.81 9.00 135.00 5
Likely to attend college
1 – Low .04 .21 .00 1.00 0
2 .04 .21 .00 1.00 0
3 .13 .34 .00 1.00 0
4 .20 .40 .00 1.00 0
5 – High .58 .49 .00 1.00 0
Mean friend GPA 2.79 .55 1.00 4.00 19

Note: N = 8087.
a
Categorical data are presented as proportions.
b
The proportion in the aggregate club type is not the sum of the individual types in its category because respondents could participate in more than one
individual activity (e.g., basketball and track).

a vocational college, or enrolled in a 4-year college), controlling for a variety of individual-level controls and including
school-level fixed effects. Robust standard errors were used to account for the clustered nature of the Add Health data
(i.e., students nested within schools). To facilitate a complete interpretation of our results, we present the results of the
model in three different ways. We first present results as relative risk ratios, which are the exponentiated coefficients.
We report average marginal effects all covariates (except the school fixed effects). Finally, we report marginal effects at dif-
ferent values of mean GPA of EA members holding other covariates of their respective means.

5. Results

Table 1 presents basic descriptive statistics for EA participation. As expected, EA participation is normative, with 71 per-
cent of students indicating involvement with at least one activity. The mean number of EAs any given student participated in
was nearly two (1.77). Using the EA type measure, 49 percent of students participated in sports, 30 percent in performance,
27 percent in academic, and 16 percent in school leadership. Baseball/softball and basketball had the highest proportion of
participants (17 percent) followed by track (12 percent), football (11 percent each), band and honor society (10 percent).
Field and ice hockey, as well as German, history and book/reading clubs had the lowest overall participation rate at 1 percent
each. We also see that a large percentage (71%) of students enroll in college. Mean GPA of students is 2.78 with most all stu-
dents anticipating enrolling in college (78% indicated either a 4 or 5, with 5 representing that the likelihood of attending col-
lege was high).
Fig. 1 reports plots of the 31 EAs in the data based on two variables: the mean GPA of EA members across all schools and
corresponding percent EA participants who enroll in a 4-year college. EAs are labeled by specific activity and type (i.e., sports,
performance, academic, and school). These aggregated EA-level results are not surprising—participating in EAs that have

100
EA Members Who Later Enrolled in

Honor society

Orchestra Debate
90 Student council Math
Science
German
Tennis
a 4-Year College,

Band Newspaper
Book Latin Drama
Soccer French
Track Swimming
History Volleyball
80 Field hockey Chorus/Choir
Percent

Cheerleading/Dance Spanish Yearbook

Computer
Basketball Ice hockey
Baseball/Softball
Wrestling
70 Football

FFA

60
Non-participants

50
2.7 2.8 2.9 3 3.1 3.2
Mean GPA of EA Members
Sports Performance Academic School
Note: Non-participants are positioned at the sample mean for average GPA of EA members
althought they do not have corresponding value.

Fig. 1. Scatterplot of EAs: the mean GPA of EA members and the percent of EA peers who enrolled in college.
 B.G. Gibbs et al. / Social Science Research 50 (2015) 367–381 375

high mean GPAs is related to a higher probability of enrolling in college. However, it is important to note that while the rela-
tionship between mean EA GPA and college enrollment follows a near linear relationship, EAs do not cluster by their types. In
other words, EAs with the highest mean GPA and probability of enrolling in college represent diverse activities such as hon-
ors, tennis, debate, and orchestra; and individual EAs from the global types (e.g., Sports: tennis, soccer, track, football, etc.)
are spread across the distribution. This suggests that traditional activity groupings do not sufficiently capture the peer com-
position link to college enrollment and that peer composition may be one important way to understand how EAs link to col-
lege enrollment. A more rigorous examination of this relationship requires modeling individual-level data controlling for
non-EA peer influences, as well as other individual- and school-level factors of college enrollment.
Table 2 reports relative risk ratios (RRRs) from a multinomial logistic regression model predicting college enrollment in
vocational and 4-year colleges compared with no enrollment. We should note that although ‘‘risk’’ is often associated with a
problematic outcome, risk in this context is considered desirable. We include four measures of EA participation: any partic-
ipation, EA type, number of EAs, and our composition measure of peer GPA’s in EAs—Mean GPA of EA peers. Although they
are not reported in the table, school-level fixed effects were included.
The relative risk of enrollment in vocational schools compared with non-enrollment in vocational or 4-year colleges is
1.718, although not statistically significant (first column in Table 2). None of the measures of EA, including the mean GPA
of EA peers, is significant. This suggests that there may not be a critical link between extracurricular associations, regardless
of its operationalization, and vocational school enrollment. However, the relative risk ratio of the mean GPA of EA peers for

Table 2
Extracurricular activities and college enrollment: relative risk ratios from multinomial logistic regression. Source: National
longitudinal study of adolescent health.

Vocational vs. no college 4-Year vs. no college

EA participator 1.060 1.302*
Mean GPA of EA peers 1.718 2.273**
EA type
Sports .943 1.273
Academic .991 1.366**
Performance 1.137 1.216
School .808 1.095
Number of EAs 1.042 1.016
Controlsa
Female 1.369** 1.502***
Age (in years) 1.041 .964
Race-ethnicity
White – –
Black 1.260 1.599**
Asian 1.582* 1.595*
Hispanic 1.253 1.412*
Native American .858 .987
Intact Family 1.003 1.132
Highest parent education
No high school degree – –
High school degree 1.112 1.010
Some college 1.123 1.118**
College degree 1.086 1.218***
Parent employment
Unemployed – –
Manual .978 .886
Non-manual 1.118 1.333
Logged household income 1.068 1.134*
Welfare .907 .915
GPA 1.318*** 1.921***
PVT 1.017*** 1.033***
Likely to attend college
1 – Low – –
2 .662* .818
3 1.118 1.533*
4 1.405 2.502***
5 – High 1.546* 4.426***
Mean friend GPA 1.042 1.365***

Note: N = 8087.
a
Fixed effects for school included but not shown.
*
p < .05.3 two-tailed tests.
**
p < .01 two-tailed tests.
***
p < .001 two-tailed tests.
376 B.G. Gibbs et al. / Social Science Research 50 (2015) 367–381

enrolling in a 4-year college relative to not enrolling in any college, 2.273, is significant even after fairly rigorous controls
(second column of Table 2). The relative risk of enrolling in a 4-year college relative to not enrolling in any college more than
doubles for a 1-point increase in mean GPA of EA peers. Given that a 1 unit increase in GPA is a large change in the compo-
sition of peers in EAs, we also estimated the model with the mean GPA of EA peers measure scaled so that a one-unit increase
is .25 of the natural metric of GPA (also roughly equivalent to the standard deviation of the measure in this sample). When
converted this way, the coefficient was 1.228.
Other EA measures were also important for 4-year college enrollment. Participation in any EA had a relative risk ratio of
1.302 suggesting the risk of enrolling in a 4-year college is expected to increase by a factor of 1.3 relative to not enrolling in
any college. Of the different types of EA programs, we only found that student participation in an Academic EA increased the
risk of enrolling in a 4-year college relative to not enrolling in any college, by a factor of 1.366. Participation in Sports, Per-
formance, or School EAs did not increase the risk of enrolling over and above participation alone.
A weakness in the relative risk ratios, although commonly used to present results from multinomial logistic regression
models, is that the size of the ratios depend on how the variables are scaled and their interpretation should be sensitive
to their distribution in the data. To facilitate a more thorough interpretation of the relationship between mean GPA of EA
peers and college enrollment we first present a histogram of the mean GPA of EA peers measure followed by two alternatives
to relative risk ratios—average marginal effects and marginal effects at specific values. The histogram of mean GPA of EA
peers in Fig. 2 indicates that the distribution is quite symmetric and normally distributed. Although the minimum and max-
imum of the distribution are 1.50 and 3.83, the figure indicates a few cases where the mean GPA of EA peers is below 2.0.
Fig. 3 plots the average marginal effects (AMEs) and 95% confidence intervals for all covariates in the analysis using school
fixed effects. AMEs are the probability that a respondent will experience a particular category of the outcome for a one-unit
change in the independent variable. Probabilities are estimated for each covariate for each category of the outcome, with 95%
confidence intervals that do not contain zero being statistically significant. AMEs facilitate interpreting the results somewhat
compared to the RRRs because they are not relative to one of the other possible outcomes. However, it is important to note
that because AMEs are not relative to another outcome, patterns of statistical significance should not necessarily be the same
as they are for RRRs.
The AMEs for enrolling in a vocational college, similar to the RRR results, indicate that none of the EA measures are related
to the probability of enrollment. Four of the AMEs for the probability of enrolling in a 4-year college appear to be large rel-
ative to other variables in the model—mean GPA of EA peers, parent education, GPA, and the perceived likelihood of attend-
ing college. We should note that comparability is not necessarily straightforward because of varying metrics across
covariates—each AME representing the change in probability for a one-unit increase in the covariate. In most cases, a
one-unit increase should be considered a substantial change. For example, a one-point increase in GPA, the unit for the mean
GPA of EA peers and GPA variables, might mean moving from a GPA of 2.0 to 3.0. The AME for GPA is close to .1 while the
AME for mean GPA of EA peers is about .2. For the likelihood of attending college variable, the approximately .2 AME for
‘‘high’’ is the change in the probability of enrollment in a 4-year college compared to the ‘‘low’’ likelihood, which is, again,
a large change in the independent variable.
Because our focus is the mean GPA of EA peers, we present marginal effects for this variable for more specific values from
2.0 to 3.5, holding other covariates at their respective means (Fig. 4). We chose this range of values for mean GPA of EA peers
to capture the observed range of the distribution in these data (98% of the sample falls in this range). These marginal effects
are also presented as probabilities but because the mean GPA of EA peers as a continuous variable, we estimated probabil-
ities for each outcome across the range of observed values at intervals of .25, which is the standard deviation of the measure
in this sample. Ninety-five percent confidence bands are also presented. In the figure, the confidence bands are substantial in
500
400
Frequency
300
200
100
0

1.5 2 2.5 3 3.5 4

Mean GPA of EA Peers
Note: Only EA participants included: N = 5,764.
Source: National Longitudinal Study of Adolescent Health.

Fig. 2. Histogram of mean GPA of EA peers with normal overlay.

 B.G. Gibbs et al. / Social Science Research 50 (2015) 367–381 377

No college Vocational 4-Year

Extracurricular Activities
EA participator

Mean GPA of EA peers

EA type
Sports
Academic
Performance
School

Number of EAs
Controls
Female

Age
Race
a
White
Black
Hispanic
Asian
Native American

Intact family
Highesta parent education
No high school diploma
High school diploma
Some college
College degree

Parent employment
a
Unemployed
Manual
Non-Manual

Logged household income

Welfare

GPA

PVT
Likely to attend college
a
Low - 1
2
3
4
High - 5

Mean friend GPA

-.2 0 .2 .4 -.2 0 .2 .4 -.2 0 .2 .4

a
Note: Reference group. N = 8087.
Source: National Longitudinal Study of Adolescent Health.

Fig. 3. Extracurricular activities and college enrollment: mean marginal effects and 95% confidence intervals from multinomial logistic regression.

.8
Probability

.6

.4

.2

0
2 2.5 3 3.5
Mean GPA of EA Peers
No college Vocational 4-Year

Note: All model covariates other than mean GPA of EA peers held at sample
means. N = 8,087.
Source: National Longitudinal Study of Adolescent Health.

Fig. 4. Mean GPA of peers in extracurricular activities: marginal effects across the distribution with 95% confidence bands.
378 B.G. Gibbs et al. / Social Science Research 50 (2015) 367–381

the lower range of the distribution which is the result of few EAs having mean GPAs in this range. The figure illustrates that
the probability of enrolling in a 4-your college increases through the range of observed values. Note that, because the multi-
nomial logistic regression model is a nonlinear model, the probabilities derived from the model are also nonlinear (see Long,
1997 for further explanation). The relationship between mean GPA of EA peers and enrollment in a 4-year college is stron-
gest until about the 3.0 GPA mark and, although it continues to be positive, is not as strong when EA peers have a mean GPA
between 3.0 and 3.5.

6. Conclusion

Our research extends previous work regarding adolescents’ social environments to show that extracurricular activities
and associations can positively influence desirable outcomes. In addition to a thorough investigation into EA participation
in its various forms, we add the role of associations within EAs and show how they have an important relationship to a stu-
dents’ risk of college enrollment beyond individual characteristics and other peer and school factors—a 1 point increase in
the GPA of peers in an EA more than doubles the risk of 4-year college enrollment.
In addition, we find that of EA types, only participation in academic activities is associated with 4-year college enrollment
once other factors are taken into account. To use an example, this suggests that participation in debate team is linked to a
greater likelihood to enroll in a 4-year college because of an individual’s association with high achieving peers and the
unique properties of debate—whether debate fosters work ethic that increases a student’s confidence for attending college
or it signals to admissions that the student is prepared for college.
And surprisingly, we find no evidence that multiple EA participation has an added impact on the risk of college enroll-
ment. This finding may have a somewhat nuanced fit in the literature. As Lareau argues that class advantage filters through
numerous activities that foster ‘‘concerted cultivation’’ and subsequent child well-being (Lareau, 2011), it appears that any
advantage might be occurring though nonschool-based activities. For example, Mahoney and Vest (2012) find a net benefit
for intensity of participation in activities out-of-school across several outcomes, including educational attainment, whereas
Fredricks (2012) finds a nonlinear association between EA (school-based activities) intensity and breadth and school out-
comes that cap out at 5 or more activities or activities that require more than 10 h a week. In other words, these findings
support a conclusion that there may be a limited benefit to participating in multiple school-based EAs (as we find), but
potentially no such limit in non-school activities.
Together, our findings provide new insights into a long-held debate: what are the mechanisms through which EAs are
associated with educational outcomes? While previous research has suggested the possibility that social capital might be
accrued through EA participation (Johnson et al., 2001; Broh, 2002; Guest and Schneider, 2003), our analyses take the liter-
ature one step closer by explicitly testing the link between exposure to advantaged peers in school-related activities and col-
lege enrollment. Although the precise causal mechanisms operating within extracurricular associations are still uncertain,
finding that ‘‘extracurricular associations’’ are strongly correlated with college enrollment suggests that EAs may be mean-
ingful ‘‘local positions’’ that increase the risk of college enrollment for students from all GPA levels. School administrators
might use these activities as a means of brokering meaningful interactions with academically strong students (Frank
et al., 2008; Moody, 2001). In these high-achieving EA environments, peers likely play a role in modeling high achievement,
creating norms of college-going and providing paths to information about how post-secondary education works, how to
apply, and what opportunities are available to youth like themselves.
Our results are among the strongest evidence to date that the social connections adolescents make at school—and spe-
cifically in extracurricular activities—can aid in their educational futures. As school boards across the United States contem-
plate the funding and development of extracurricular activities, our results reveal that extracurricular activities and the
associations they foster may likely play a nontrivial policy role in increasing students’ academic success. Although we cannot
be definitive regarding causality, these results support an often overlooked perspective regarding any EA effect, that they are
not just about what you do, but also who you do it with.
We also show that EA type—at least for college enrollment—is relatively an unimportant distinction with one exception;
academic activities have an additive link to college enrollment above and beyond general participation in EAs. This is intu-
itive in some ways—academic activities are often geared toward college preparation, and are part of the kinds of activities
college admissions often reward (Kaufman and Gabler, 2004). This finding, however, contradicts studies claiming that types
of EAs have unique benefits of educational outcomes. This may still be so for other outcomes, but our results suggest that any
benefits of EA types for 4-year college enrollment are indistinguishable from each other—only academic club participation as
a category appears to have added benefits.

7. Limitations and future work

Our study has certain limitations. Although traditional regression approaches attempt to control for alternative causes
(Covay and Carbonaro, 2010; Eccles et al., 2003; Wimer et al., 2008; Broh, 2002; Eitle and Eitle, 2002) as we have done here,
future work should explore methods to better account for selection. Although methods such as propensity score matching
are promising, it is not clear how to identify EA participation as a ‘‘treatment,’’ especially when students engage in multiple
activities.
 B.G. Gibbs et al. / Social Science Research 50 (2015) 367–381 379

Also, by using the Add Health data we are unable to consider the actual content or intensity of the EAs. It is possible that
some activities are more effective in fostering skills and experiences necessary for college preparation than others, but are
simply unmeasured in our data. Additionally, we are assuming that exposure to higher-achieving peers in EAs provides
access to social resources that influence or create a culture that encourages further education. No data with which we are
familiar, including the data we use here, can test whether such pro-education norms are actually being transmitted from
peer to peer. Future research of within-school processes may yield more insight (see Gamoran and Berends, 1987 for a dis-
cussion of such approaches).
With these limitations in mind, a compositional approach to understanding EAs has several exciting possibilities for
future work. For example, the overlapping role of friendship networks and EA participation is a context of adolescent social-
ization that informed much of Coleman’s work in Adolescent Society (1961). We think, if thoroughly examined, this kind of
analysis could reveal how friendships and other student attributes, such as popularity, link to extracurricular associations.
Also, in line with Moody’s work on friendship networks, research that connects race-based and gender-based friendship net-
works within EAs to academic and nonacademic outcomes may provide new insights into the consequences of school friend-
ships. While we focus on college enrollment, extracurricular associations may also influence delinquency, academic
achievement, school dropout, and college completion to name a few—all important possible extensions.
Taken together, our results document that not only does participation in EAs likely improve the ‘‘risk’’ of college enroll-
ment, but that the average GPA of peers within EAs is a strong component in explaining the association. This suggests that
school policies aimed at encouraging student exposure to high achieving peers in EAs could have an important impact on a
student’s later educational outcomes.

Acknowledgments

This research used data from Add Health, a program project designed by J. Richard Udry, Peter S. Bearman, and Kathleen
Mullan Harris and funded by Grant P01-HD31921 from the Eunice Kennedy Shriver National Institute of Child Health and
Human Development, with cooperative funding from 17 other agencies. Special acknowledgment is due to Ronald R. Rind-
fuss and Barbara Entwisle for assistance in the original design. Persons who are interested in obtaining data files from Add
Health should contact Add Health, Carolina Population Center, 123 West Franklin Street, Chapel Hill, NC 27516-2524 (add-
health@unc.edu). No direct support was received from grant P01-HD31921 for this analysis.
We gratefully acknowledge Niki Weight, Marie Cornwall, William Carbonaro, Joshua Klugman, Jennifer Flashman, and
anonymous reviewers for their thoughtful comments as the manuscript developed.

References

Allison, Paul D., 2001. Missing Data. Sage Publications, Thousand Oaks, CA.
Andrews, Judy A., Tildesley, Elizabeth, Hops, Hyman, Li, Fuzhong, 2002. The influence of peers on young adult substance abuse. Health Psychol. 21, 349–357.
Arum, Richard, Roksa, Josipa, 2011. Academically Adrift: Limited Learning on College Campuses. University of Chicago Press, Chicago.
Bankston, Carl L., Caldas, Stephen J., 1998. Family structure, schoolmates, and racial inequalities in school achievement. J. Marriage Family 60, 715–723.
Barber, Bonnie L., Eccles, Jacquelynne S., Stone, Margaret R., 2001. Whatever happened to the ‘‘Jock,’’ the ‘‘Brain,’’ and the ‘‘Princess’’?: young adult pathways
linked to adolescent activity involvement and social identity. J. Adolescent Res. 16, 429–455.
Beattie, Irenee R., 2002. Are all ‘Adolescent Econometricians’ created equal? racial, class, and gender differences in college enrollment. Sociol. Educ. 75, 19–
43.
Blau, Peter M., 1977. Inequality and Heterogeneity: A Primitive Theory of Social Structure. Free Press, New York.
Broh, Beckett A., 2002. Linking extracurricular programming to academic achievement: who benefits and why? Sociol. Educ. 75, 69–95.
Chantala, Kim, Kalsbeek, William, Andraca, Eugenio., 2004. Non-Response in Wave III of the Add Health Study. <http://www.cpc.unc.edu/projects/
addhealth/data/guides/W3nonres.pdf/view>.
Cherng, Hua-Yu Sebastian, Calarco, Jessica McCrory, Kao, Grace, 2013. Along for the ride: best Friends’ Resources and Adolescents’ college completion. Am.
Educational Res. J. 50, 76–106.
Choi, Kate H., Kelly Raley, R., Muller, Chandra, Riegle-Crumb, Catherine, 2008. Class composition: socioeconomic characteristics of coursemates and college
enrollment. Social Sci. Q. 89, 846–866.
Coleman, James S., 1961. The Adolescent Society. Free Press of Glencoe, New York.
Coleman, James S., 1988. Social capital in the creation of human capital. Am. J. Sociol. 94, S95–S120.
Coleman, James S., Campbell, Ernest Q., Hobson, Carol F., McPartland, James M., Mood, Alexander M., Weinfeld, Frederic D., York, Robert L., 1966. Equality of
Educational Opportunity. U.S. Government Printing Office, Washington, DC.
Covay, Elizabeth, Carbonaro, William, 2010. After the bell: participation in extracurricular activities, classroom behavior, and academic achievement. Sociol.
Educ. 83, 20–45.
Crosnoe, Robert, 2009. Low-income students and the socioeconomic composition of public high schools. Am. Sociol. Rev. 74, 709–730.
Crosnoe, Robert, Cavanagh, Shannon, Elder Jr., Glen H., 2003. Adolescent friendships as academic resources: the intersection of friendship, race, and school
disadvantage. Sociol. Perspect. 46, 331–352.
Crosnoe, Robert, Riegle-Crumb, Catherine, Frank, Kenneth, Field, Sam, Muller, Chandra, 2008. Peer group contexts of girls’ and boys’ academic experiences.
Child Dev. 79, 139–155.
DiMaggio, Paul, Garip, Filiz, 2012. Network effects and social inequality. Annu. Rev. Sociol. 38, 93–118
Dunn, Lloyd M., 1981. Peabody Picture Vocabulary Test, Revised. American Guidance Service. Circle Pines, MN.
Eccles, Jacquelynne S., Barber, Bonnie L., 1999. Student council, volunteering, basketball, or marching band: what kind of extracurricular involvement
matters? J. Adolesc. Res. 14, 10–43.
Eccles, Jacquelynne S., Barber, Bonnie L., Stone, Margaret, Hunt, James, 2003. Extracurricular activities and adolescent development. J. Social Issues 59, 865–
889.
Eitle, Tamela M., Eitle, David J., 2002. Race, cultural capital, and the educational effects of participation in sports. Sociol. Educ. 75, 123–146
380 B.G. Gibbs et al. / Social Science Research 50 (2015) 367–381

Enders, Craig K., 2010. Applied Missing Data Analysis. The Guilford Press, New York.
Feijin, Naomi, 1994. Participation in high school competitive sports: a subversion of school mission or contribution to academic goals? Sociol. Sport J. 11,
211–230.
Feldman, Amy F., Matjasko, Jennifer L., 2005. The role of school-based extracurricular activities in adolescent development: a comprehensive review and
future directions. Rev. Educational Res. 75, 159–210.
Feldman, Amy F., Matjasko, Jennifer L., 2007. Profiles and portfolios of adolescent school-based extracurricular activity participation. J. Adolescence 30, 313–
332.
Fletcher, Jason M., Tienda, Marta, 2009. High school classmates and college success. Sociol. Educ. 82, 287–314.
Flether, Jason M., 2012. Similarity in peer college preferences: new evidence from texas. Soc. Sci. Res. 41, 321–330.
Frank, Kenneth A., Muller, Chandra, Schiller, Kathryn S., Riegle-Crumb, Catherine, Mueller, Anna Strassman, Crosnoe, Robert, Pearson, Jennifer, 2008. The
social dynamics of mathematics coursetaking in high school. Am. J. Sociol. 113, 1645–1696.
Fredricks, Jennifer A., 2012. Extracurricular participation and academic outcomes: testing the over-scheduling hypothesis. J. Youth Adolesc. 41, 295–306.
Fredricks, Jennifer A., Eccles, Jacquelynne S., 2006. Is extracurricular participation associated with beneficial outcomes? Concurrent and longitudinal
relations. Dev. Psychol. 42, 698–713.
Gamoran, Adam, Berends, Mark, 1987. The effects of stratification in secondary schools: synthesis of survey and ethnographic research Rev. Educational Res.
57, 415–435.
Gardner, Margo, Roth, Jodie, Brooks-Gunn, Jeanne, 2008. Adolescents’ participation in organized activities and developmental success 2 and 8 years after
high school: do sponsorship, duration, and intensity matter? Dev. Psychol. 44, 814–830.
Gilman, Rich, Meyers, Joel, Perez, Laura, 2004. Structured extracurricular activities among adolescents: findings and implications for school psychologists.
Psychol. Schools 41, 31–41.
Giordano, Peggy C., 2003. Relationships in adolescence. Annu. Rev. Sociol. 29, 257–281.
Gore, Susan, Farrell, Florence, Gordon, Jennifer, 2001. Sports involvement as protection against depressed mood. J. Res. Adolescence 11, 119–130.
Grodsky, Eric, Jones, Melanie T., 2007. Real and imagined barriers to college entry: perceptions of cost. Soc. Sci. Res. 36, 745–766.
Grodsky, Eric, Riegle-Crumb, Catherine, 2010. Those who choose and those who don’t: social background and college orientation. Ann. Am. Acad. Political
Social Sci. 627, 14–35.
Guest, Andrew, Schneider, Barbara, 2003. Adolescents’ extracurricular participation in context: the mediating effects of schools, communities, and identity.
Sociol. Educ. 76, 89–109.
Harris, Douglas N., 2010. How do school peers influence student educational outcomes? Theory and evidence from economics and other social sciences
Teachers Coll. Rec. 112, 1163–1197.
Harris, Kathleen Mullan, Francesca Florey, Joyce Tabor, Bearman, Peter S., Jo Jones, Richard Udry, J., 2003. The National Longitudinal Study of Adolescent
Health: Research Design. <http://www.cpc.unc.edu/projects/addhealth/design>.
Haynie, Dana, Wayne Osgood, D., 2005. Reconsidering peers and delinquency: how do peers matter? Soc. Forces 84, 1109–1130.
Hout, Michael, 2012. Social and economic returns to college education in the united states. Annu. Rev. Sociol. 38, 379–400.
Jencks, Christopher, 1972. Inequality: A Reassessment of the Effect of Family and Schooling in America. Basic Books, New York.
Johnson, Monica Kirkpatrick, Crosnoe, Robert, Elder, Glen H., 2001. Students’ attachment and academic engagement: the role of race and ethnicity. Sociol.
Educ. 74, 318–340.
Kaufman, Jason, Gabler, Jay, 2004. Cultural capital and the extracurricular activities of girls and boys in the college attainment process. Poetics 32, 145–168.
Landers, Daniel, Landers, Donna, 1978. Socialization via interscholastic athletics: its effects on delinquency. Sociol. Educ. 51, 299–303.
Lareau, Annette, 2011. Unequal Childhoods: Class, Race, and Family Life, second ed. University of California Press, Berkeley, CA.
Long, J. Scott, 1997. Regression Models for Categorical and Limited Dependent Variables. Sage, Thousand Oaks, CA.
Mahoney, Joseph L., Stattin, Hakan, 2000. Leisure activities and adolescent antisocial behavior: the role of structure and social context. J. Adolescence 23,
113–127.
Mahoney, Joseph L., Vest, Andrea E., 2012. The over-scheduling hypothesis revisited: intensity of organized activity participation during adolescence and
young adult outcomes. J. Res. Adolescence 22, 409–418.
Mahoney, Joseph L., Cairns, Beverley D., Farmer, Thomas W., 2003. Promoting interpersonal competence and educational success through extracurricular
activity participation. J. Educ. Psychol. 95, 409–418.
Mahoney, Joseph L., Larson, Reed W., Eccles, Jacquelynne S., 2005. Organized Activities as Contexts Of Development: Extracurricular Activities, After-School
and Community Programs. Lawrence Erlbaum Associates, Inc., Publishers, Mahwah, NJ.
Manksi, Charles F., 1990. Adolescent econometricians: how do youth infer the returns to schooling In: Clotfetter, Charles T., Rothschild, Michael (Eds.),
Studies of Supply and Demand in Higher Education. University of Chicago Press, Chicago, IL, pp. 952–991.
Marsh, Herbert W., 1992. Extracurricular activities: beneficial extension of the traditional curriculum or subversion of academic goals? J. Educ. Psychol. 84,
553–562.
Marsh, Herbert W., Kleitman, Sabrina, 2002. Extracurricular school activities: the good, the bad, and the nonlinear. Harvard Educational Rev. 72, 464–514.
McLeod, Jane D., Kaiser, Karen, 2004. Childhood emotional and behavioral problems and educational attainment. Am. Sociol. Rev. 69, 636–658.
McNeal, Ralph, 1995. Extracurricular activities and high school dropouts. Sociol. Educ. 68, 62–80.
McNeal, Ralph, 1998. High school extracurricular activities: closed structures and stratifying patterns of participation. J. Educational Res. 91, 183–191.
Merolla, David M., Serpe, Richard T., Stryker, Sheldon, Wesley Schultz, P., 2012. Structural precursors to identity processes: the role of proximate social
structures. Social Psychol. Q. 75, 149–172.
Mickelson, Roslyn Arlin, 2010. Goals, grades, fears, and peers. Introductory essay for special issues on the effects of school and classroom racial and SES
composition on educational outcomes. Teachers Coll. Rec. 112, 961–977.
Moody, James, 2001. Race, school integration, and friendship segregation in America. Am. J. Sociol. 107, 679–716.
Niu, Sunny X., Tienda, Marta, 2008. Choosing colleges: identifying and modeling choice sets. Soc. Sci. Res. 37, 416–433.
Quiroz, Pamela Anne, Gonzalez, Nilda Flores, Frank, Kenneth A., 1996. Carving a niche in the high school social structure: formal and informal constraints on
participation in the extra curriculum. Res. Sociol. Ed. 11, 93–120.
Riegle-Crumb, Catherine, Callahan, Rebecca M., 2009. Exploring the academic benefits of friendship ties for latino boys and girls. Social Sci. Q. 90, 611–
631.
Sandefur, Gary D., Meier, Ann M., Campbell, Mary E., 2006. Family resources, social capital, and college attendance. Soc. Sci. Res. 35, 525–553.
Small, Mario, 2009. Unanticipated Gains: Origins of Network Inequality in Everyday Life. Oxford University Press, New York.
Sokatch, Andrew, 2006. Peer influences on the college-going decisions of low socioeconomic status urban youth. Ed. Urban Soc. 39, 128–146.
Spady, William G., 1970. Dropouts from higher education: an interdisciplinary review and synthesis. Interchange 1, 64–85.
Stata Corp., 2013. Stata Statistical Software: Release 13. College Station, TX: StataCorp LP.
Stephan, Jennifer L., Rosenbaum, James E., Person, Ann E., 2009. Stratification in college entry and completion. Soc. Sci. Res. 38, 572–593.
Terenzini, Patrick T., Rendon, Laura I., Lee Upcraft, M., Millar, Susan B., Allison, Kevin W., Gregg, Patricia L., Jalomo, Romero, 1994. The transition to college:
diverse students, diverse stories. Res. High. Educ. 35, 57–73.
Tierney, William G., Venegas, Kristan M., 2006. Fictive kin and social capital: the role of peer groups in applying and paying for college. Am. Behavioral
Scientist 49, 1687–1702.
Troutman, Kelly P., Dufur, Mikaela J., 2007. From high school jocks to college grads: assessing the long-term effects of high school sport participation on
females’ educational attainment. Youth Soc. 38, 443–462.
 B.G. Gibbs et al. / Social Science Research 50 (2015) 367–381 381

U.S. Department of Education, National Center for Education Statistics. 2013. The Condition of Education 2013 (NCES 2013–037), Annual Earning of Young
Adults.
Wimer, Christopher, Simpkins, Sandra Denise, Dearing, Eric, Bouffard, Suzanne M., 2008. Predicting youth out-of-school time participation: multiple risks
and developmental differences. Merril-Palmer Q. 54, 179–207.
Zaff, Jonathan A., Moore, Kristin A., Papillo, Angela Romano, Williams, Stephanie, 2003. Implications of extracurricular activity participation during
adolescence on positive outcomes. J. Adolesc. Res. 18, 599–630.
Zimmerman, Gregory M., Messner, Steven F., 2011. Neighborhood context and nonlinear peer effects on adolescent violent crime. Criminology 49, 873–903.