Format and Rules

The seminar is held on Zoom and lasts 60 minutes including questions. A moderator will collect audience questions in chat or using the 'raise hand' function. Please stay muted until called by the moderator and/or the presenter.

Seminars will be recorded and available on this website a few days after. Please note that you may be recorded if you activate your video or speak during the seminar.

Abstract: Incorporating a game-theory based model of social interactions with switch regression, we build a framework for interactions between different types of group members within possible endogenous social groups. We provide conditions for identification and propose the two-step NLS estimation method, which is consistent and asymptotically normal. We apply the model to a unique data set for Shanghai pupils from two kinds of schools, public schools for both local and migrant children, and private ones for migrant children exclusively. We show that there are significantly positive effects on math scores between all types of classmates (peers). Although the influences from peers of the same type are slightly stronger than those from peers of different types, the local-to-migrant and migrant-to-local effects are not significantly different from each other. In addition, for migrant students enrolled in public schools, the treatment effect of studying in public schools for math score is 4.25 points, with more than 50% due to better peers in public schools.