MEASURING PEER EFFECTS IN ACADEMIC OUTCOMES
Kuersteiner, Guido M
Prucha, Ingmar R
MetadataShow full item record
There is wide belief that a student's behavior and academic outcome are affected by her/his fellow students. This peer effect lies at the center of the debate about education policy. My dissertation focuses on measuring peer effects in academic outcomes. It consists of a theoretical part and an empirical part. In the theoretical part, I study a peer effects model with group-wise equal interactions and random group effects. Identifying peer effects is notoriously challenging due to the reflection problem. Common shocks to the groups also generate spurious peer effects. My model, therefore, controls for the common shocks to the whole group with random group effects. My estimation strategy overcomes the identification problem with spatial econometrics techniques. I develop a quasi-maximum likelihood estimator of the model. Monte-Carlo simulations show that the bias of the estimator decreases with the number of groups and the variation in group size, and increases with group size. Finally, I prove the consistency and asymptotic normality of the estimator under standard assumptions. In the empirical part, I apply the model to Project STAR data to study the peer effects among kindergarten students. Peers constitute an important context for children's academic development. This empirical study measures peer effects on math and reading scores of kindergarten children using data from Project STAR, an experiment in Tennessee that randomly assigned both children and teachers to classrooms of different sizes. It estimates the impact of peers’ scores and characteristics on children’s individual scores, controlling among other things for random class effects. In contrast to most existing studies, the estimated peer effects in the empirical part are small and insignificant. The results are robust when allowing peer effects to be heterogeneous by gender, using data from higher grades or considering alternative specifications. The findings of the empirical study cast doubt on the effectiveness of programs that manipulate peer groups for better educational outcomes.