Peer effects identified through social networks. Evidence from Uruguayan schools

This paper provides evidence on peer effects in educational achievement exploiting for the first time a unique data set on social networks within primary schools in Uruguay. The relevance of peer effects in education is still largely debated due to the identification challenges that the study of social interactions poses. I adopt a recently developed identification method that exploits detailed information on social networks, i.e. individual-specific peer groups. This method enables me to disentangle endogenous effects from contextual effects via instrumental variables that emerge naturally from the network structure. Correlated effects are controlled, to some extent, by classroom fixed effects. I find significant endogenous effects in standardized tests for reading and math. A one standard deviation increase in peers’ test score increases the individual’s test score by 40% of a standard deviation. This magnitude is comparable to the effect of having a mother that completed college. By means of a simulation I illustrate that when schools are stratified by socioeconomic status peer effects may operate as amplifiers of educational inequalities.