Essays on Peer Effects
This dissertation considers the relationship between peer and individual student interaction. The central finding is that self reported friends play a crucial role in individual behaviors, a role that is more significant than other students in their school. Also, using the network of friendships within a school it is possible to construct new peer effect measures and account for endogenous peer group formation. It is however important to distinguish these peer measures from unobserved individual characteristics that may also influence behavior.
The first chapter examines the effect of potentially misidentifying the reference group on peer effect estimates. The differential impact of school, grade and friend level peer effects on student decisions to smoke and drink are calculated. Friendship nominations come from the Add Health dataset, where students can list up to 10 friends from the school. The bias due to endogenous peer group formation and simulteneity are considered using various instrumenting strategies. Peer effects are found to be large and significant at the friends level for both delinquency variables. It is possible to show that misidentifying the peer group can result in peer effect estimates that are understated by as much as 40\%.
The second chapter of the dissertation further examines the role of peer interactions, this time considering the effect of popularity on student academic achievement. Recent work has found a strong positive relationship between these variables. In this chapter I ascertain the robustness of these previous findings to controls for unobserved student heterogeneity using and instrumenting technique and a structural model. The results indicate that popularity influences academic achievement positively in the baseline model. However, instrumenting for popularity or including measures of unobserved student characteristics results in a large drop in the effect of popularity, and leads to a significantly negative coefficient in the majority of cases. Interestingly, popularity influences future earnings and attitudes positively, where this effect is robust to the inclusion of unobserved type. Policy simulations where students are redistributed based on race or income indicate that the predicted number of friendships and popularity fall but academic achievement increases. Since student popularity increases happiness and earnings, the overall effect of the redistribution policies have to be considered before implementation.
This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 3.0 United States License.
Rights for Collection: Duke Dissertations