ONLINE - Misclassification in Group Average Regressions: Theory and Application to Peer Effects Estimation

21.10.2020 14:15 – 15:45

INSTITUTE OF ECONOMICS AND ECONOMETRICS SEMINAR / ABSTRACT

The purpose of this paper is to reconcile the extensive literature on the externalities of disruptive peers. To do this, we first offer a comprehensive, structured, and rigorous synthesis on the empirical literature on peer effects of potentially disruptive students. Second, we develop a conceptual framework that underscores the importance of misclassification and the severity degree of disruptiveness captured to explain the divergence in the peer effect estimates obtained using different proxies of disruptiveness. We show that under random group formation larger misclassification rates bias the peer effect estimate towards zero (or even beyond if the proxy variable is negatively correlated with the latent variable), whereas capturing more severe forms of disruptiveness drives the peer effect estimate away from zero. Under non-random group formation, however, the direction of the bias is a priori unclear. Third, we conduct an empirical analysis that compares peer effects of different proxies under a common institutional and empirical setting. Results of the empirical analysis mostly confirm the theoretical propositions.

Sponsored by Centre LIVES

Lieu

Online

entrée libre