The Impact of Classroom Peers in a Streaming System

This paper investigates educational production with a focus on the influence that socio-economic status of class peers has on academic outcomes of students in a streaming system. Employing the Swiss subsample of the PISA data, I provide evidence that while classroom assignment is not random within schools or tracks it is random within tracks-by-schools. Track-by-school fixed effects therefore render peer group composition conditionally uncorrelated with students' characteristics, while track fixed effects and school fixed effects don’t. Estimates based on track fixed-effects and school fixed-effects approaches are reduced sizably by employing a track-by-school fixed-effects approach while mean effects on test results in mathematics and problem solving remain significant.