Using the identification strategy proposed by Graham and Hahn (2005), the authors estimate the magnitude of classmate effects on math scores using Brazilian data from 2005. In addition, they provide a detailed discussion about the identification of endogenous peer effects in the linear in means models.

The authors' results show that both peer characteristics (exogenous peer effects) – like race, socioeconomic status and gender – and peer actions (endogenous peer effects) are important determinants of students’ outcomes in the fifth grade of elementary school. Their estimates of endogenous peer effects are about 0.008 of 1 SD of math test scores, which can be interpreted as evidence of a so-called ‘conformist’ individual behaviour, under which students face large costs to exert effort levels that are distant from what is believed to be the norm in the classroom. Those estimates of endogenous peer effects imply a social multiplier of about 1.67.