Understanding the mechanisms underlying peer group effects: The role of friendships in determining adolescent outcomes

Jason Fletcher, Stephen Ross 03 November 2013



Over the last decade, research on peer effects in secondary education has flourished – in part because of the within-school/across-cohort design for identifying peer effects popularised in early work by Hoxby (2000), and partly due to the increasing availability of quality data on K-12 students in the US and internationally. The cohort approach to studying peer effects exploits the idea that when choosing schools, parents cannot easily observe the attributes of the specific cohorts to which their children will belong. As a result, variation in exposure to peer composition between different cohorts at the same school can be treated as a randomly assigned treatment. Using this approach, a large number of studies have demonstrated that peer composition affects student outcomes:

  • Lavy and Schlosser (2008) and Lavy, Pasermann, and Schlosser (2012) show that gender and ability composition matter in Israel.
  • Bifulco, Fletcher, and Ross (2010) show that maternal education matters in the US.
  • Friesen and Krauth (2011) show that ethnicity matters in Canada. (See Ross 2011 for a recent literature review.)

Many people have used this evidence to argue that friendships in school could have substantial impacts on children’s academic and life outcomes, and thus suggest that interventions that shape the formation or shift the composition of friendships might be beneficial.

Discipline and disruption

In examining potential mechanisms behind the effects of peer composition from earlier research, several studies have pointed toward discipline and classroom disruption as a major factor – as opposed to the direct effect of social relationships between students. Lavy and Schlosser (2008) find that student outcomes improve in cohorts that have fewer male children, and show that gender composition also influences student-teacher relationships, teachers’ practices, and classroom disruption and violence. Aizer (2008) estimates the impact of having classmates with Attention Deficit Disorder before and after diagnosis, finding that diagnosis improves peer performance. Similarly, Carrell and Hoekstra (2010) find that students from families with domestic violence reports exhibit more disruptive behaviour, and students exposed to peers with domestic violence at home have worse behaviour and academic outcomes relative to their siblings who did not have such exposure. Further, these effects disappear after the reporting of domestic violence to the court (Hoekstra and Carrell 2010). The fact that the peer effects are ameliorated by actions expected to mitigate disruptive behaviour makes these studies especially convincing.

Cohorts and social dynamics

On the other hand, some studies have shown evidence of mechanisms related to aggregate peer effects that suggest social relationships (friendships) may be an important factor in peer effects processes. Babcock (2008) finds that cohorts that have higher connectedness in terms of friendships also have students who obtain more years of schooling relative to other students in the same school. Billings, Deming, and Rockoff (2012) find that high concentrations of economically disadvantaged, African-American male students lead to more crime – potentially due to the greater opportunities for cooperation between high risk students who might be more likely to engage in crime. Carrell, Sacerdote, and West (2013) find that reassigning student groups changed observed peer effects – arguably because it changed the social dynamics of the groups themselves.

Bifulco, Fletcher, and Ross (2010) examine potential behavioural mechanisms for the effect of peer maternal education on high school completion and college attendance, and find no effects – instead, their evidence suggests that the higher rates of completion and attendance might be associated with imitative behaviour. When combined with evidence in Fletcher, Ross, and Zhang (2013) that friendships between students with college-educated parents become more likely as the number of students with college-educated parents increases, these findings might suggest imitation of friends’ choices. Further, Bifulco, Fletcher, Oh, and Ross (2012) find that these effects fade with time as people are separated from their school peers and so from those relationships.

From cohorts to friendships

However, one limitation with most previous studies – and all studies using cohort difference methods – is the focus on aggregated ‘peer groups’ rather than actual friends and social relationships.1 The aggregate peer measures may bias the estimated peer effects downward, as not all students know (or interact with) all of their classmates. For example, Carrell, Fullerton, and West (2009) find large peer effects when examining assignment to work groups of around 30 students who spend most of their time working together – identifying study partners as one potential mechanism.

A recent set of papers have shifted the focus from aggregate peer measures to directly examining the effects of friendships, but have kept the core identification strategies from the prior literature. Fletcher and Ross (2012) find that students who have friends who smoke or drink are more likely to smoke or drink – even when comparing observationally similar students who belonged to different cohorts in the same school, and made exactly the same friendship choices on key student demographics. Fletcher, Ross and Zhang (2013) estimate a model of friendship formation, and use this model to estimate the expected number of friends whose mothers have a college degree for each student. They find that girls have higher grade point averages than very similar students in the same school when they belong to a cohort that implies more friends with a higher level of maternal education – even after controlling for aggregate peer effects associated with maternal education.2

Friendship effects and aggregate peer effects

However, just because friendships matter for student outcomes, this does not necessarily imply that observed friendships are associated with the aggregate peer effects observed in traditional across-cohort studies. For example, Bifulco, Fletcher, and Ross (2010) – using the same data set as Fletcher and Ross (2012) – find no evidence that peer racial or maternal education composition affect drinking or smoking, even though a student’s race and parental education correlate with their own drinking and smoking – so that on average cohort combination should correlate with friendship patterns over drinking and smoking.

Similarly, while Fletcher, Ross, and Zhang (2013) find that friends’ maternal education affects girls’ academic outcomes, the equilibrium effect of increases in the number of students whose mothers have a college degree leads to only very small improvements in academic outcomes. These effects are far too small to explain observed peer effects of maternal education on academic outcomes in the same sample (Bifulco, Fletcher, and Ross 2010). On the other hand, the effects on smoking and drinking in Fletcher and Ross (2012) are quite large, and so the relative importance of aggregate peer and friendship effects may depend upon the domain of interest. So, while evidence grows on the importance of friendships and recent progress has been made on the issue, many questions still exist about the overall importance of these effects for aggregate outcomes and their relationship to observed aggregate peer effects.

Authors’ note: Fletcher and Ross are grateful for support by the Eunice Kennedy Shriver National Institute of Child Health & Human Development of the National Institutes of Health under Award Number R21HD066230.


Aizer, Anna (2008), “Peer Effects and Human Capital Accumulation: the Externalities of ADD”, NBER Working Paper 14354.

Babcock, Philip (2008), “From Ties to Gains? Evidence on Connectedness and Human Capital Acquisition”, Journal of Human Capital 2(4): 379–409.

Bifulco, Robert, Jason M Fletcher, and Stephen L Ross (2012), “Do Classmate Effects Fade?”, NBER Working Paper 18648.

Bifulco, Robert, Jason M Fletcher, and Stephen L Ross (2011), “The effect of classmate characteristics on post-secondary outcomes: evidence from the Add Health”, American Economic Journal: Economic Policy 3(1): 25–53.

Billings, Stephen B, David J Deming, and Jonah E Rockoff (2012), “School Segregation, Educational Attainment and Crime: Evidence from the End of Busing in Charlotte- Mecklenburg”, NBER Working Paper 18487.

Carrell, Scott E, Richard L Fullerton and James E West (2009), “Does Your Cohort Matter? Measuring Peer Effects in College Achievement”, Journal of Labor Economics 27(3): 439–464.

Carrell, Scott E and Mark Hoekstra (2010), “Externalities in the Classroom: How Children Exposed to Domestic Violence Affect Everyone’s Kids”, American Economic Journal: Applied Economics 2(1): 211–228.

Carrell, Scott E, Bruce I Sacerdote, and James E West (2013), “From Natural Variation to Optimal Policy? The Importance of Endogenous Peer Group Formation”, Econometrica 81(3): 855–882.

Fletcher, Jason M, and Stephen L Ross (2012), “Estimating the Effects of Friendship Networks on Health Behaviors of Adolescents”, NBER Working Paper 18253.

Fletcher, Jason M, Stephen L Ross and Yuxiu Zhang (2013), “The Determinants and Consequences of Friendship Formation”, NBER Working Paper 19215.

Friesen, Jane and Brian Krauth (2011), “Ethnic enclaves in classrooms”, Labour Economics 18(5): 656–663.

Hoxby, Caroline (2000), “Peer effects in the classroom: Learning from gender and race variation”, NBER Working Paper 7867.

Lavy, Victor, M Daniele Paserman, and Analia Schlosser (2012), “Inside the black-box of peer ability effects: Evidence from variation in high and low achievers in the classroom.” Economic Journal.

Lavy, Victor and Edith Sand (2012), “The Friends Factor: How Students’ Social Networks Affect Their Academic Achievement and Well-Being?”, NBER Working Paper 18430.

Lavy, Victor and Analia Schlosser (2011), “Mechanisms and impacts of gender peer effects at school”, American Economic Journal: Applied Economics 3(2): 1–33.

Nathan, A (2008), “The Effects of Racial and Extracurricular Activity on Friendship and Achievement”, College of the Holy Cross Working Paper.

Ross, Stephen L (2011), “Social interactions within cities: Neighborhood environments and peer relationships”, in Handbook of Urban Economics and Planning (eds. N Brooks, K Donaghy, and G Knapp), Oxford University Press.

1 But also see the literature on the effects of randomly-assigned college roommates on student outcomes, which combines a clear research design with a targeted (rather than aggregate) measure of peers (Sacerdote 2001).

2 Nathan (2008) uses within school variation in aggregate cohort attributes to identify the effect of friends, but cannot distinguish between the broad effect of cohort composition and the effect of friendships. Lavy and Sand (2013) provide an example of a convincing identification strategy that does not exploit across cohort variation treating the failure to honor a student’s request to have their friend attend the same school as a random separation of a friendship and find that students who are separated from friends during a natural transition between schools do worse than students whose friends were assigned to the same upper level school.



Topics:  Education Gender

Tags:  education, Peer Effects, discipline, disruption

Professor of Public Affairs, University of Wisconsin-Madison

Professor of Economics and Phillip A. Austin Chair in Public Policy, University of Connecticut


CEPR Policy Research