Friendship Formation and Network Effects: Experimental Evidence from Ecuador

• In Event: Identifying Peer Effects: Evidence from Classrooms, Friendships, and Families

Abstract

Friendships are among the most frequent and formative interactions children experience in school and play a central role in shaping behavior, effort, and learning. In elementary school, where children spend most of their time with the same classmates and teachers, friendships are particularly influential. Despite the evidence on peer effects, little is known about whether friendships help or hurt learning and well-being. Friendships can provide emotional support and foster motivation, but they may also distract from academic activities or reinforce negative behaviors. Therefore, understanding whether their social benefits outweigh their academic costs is central to designing classroom environments that support both academic and emotional development.

This paper examines the impact of friendship networks on cognitive and non-cognitive skills during elementary school in Ecuador. Using data from a unique randomized experiment in 202 elementary schools, where students were randomly assigned to classrooms within schools each year from Kindergarten through sixth grade, we exploit exogenous variation in students’ exposure to friends across classrooms. We combine these data with friendship networks elicited in third and sixth grade and rich longitudinal information on child, household, and teacher characteristics. This setting provides a unique opportunity to study friendship networks, a subset of peers representing stronger social ties, in a context where peer exposure is intense and repeated random assignment generates plausibly exogenous variation in social interactions. In addition, the multi-year design allows us to examine how repeated exposure shapes friendship formation and persistence.

In the baseline specification, we include school-by-grade fixed effects and control for the number of friends each student has within the school, comparing students in the same school and grade who differ in the number of their friends assigned to their classroom due to random assignment. I find that having one additional friend in the classroom reduces math test scores by 0.006 standard deviations and executive function by 0.023 standard deviations. These negative effects are concentrated among male students. At the same time, having more friends in the classroom reduces symptoms of depression, suggesting that the effects on cognitive skills are not driven by lower happiness or self-confidence. We show that the negative academic effects are driven by the number of disruptive friends in the classroom, consistent with a distraction or lack of attention channel. These results are robust across alternative specifications, estimation methods, and sensitivity analyses.

The findings suggest that friendship networks generate positive effects on emotional well-being but negative effects on academic performance. While friendships improve mental health, they may hinder learning when they divert attention away from classroom engagement. Given the central role of social interactions in the learning environment, these results highlight the importance of considering not only peer composition but also the structure of social networks when designing classroom assignment policies and education interventions, particularly in settings where classroom environments are a primary source of peer exposure and where policy typically focuses only on observable peer characteristics, overlooking how relationships shape behavior and learning.

Authors

• Nicolas Fuertes Segura, University of Michigan
  Presenting Author

• Yyannú Cruz-Aguayo, Inter-American Development Bank
  Non-Presenting Co-Author