Timothy Halliday
Department of Economics
University of Hawaii at Manoa

Sally Kwak
Department of Economics
University of Hawaii at Manoa

Abstract

The potential influence of peers and social networks on individual outcomes is important to a variety of educational policy debates including school vouchers, special education, middle school grade configurations and tracking. Researchers have well-documented the problems associated with credibly estimating peer effects in these settings: endogenous selection, contextual effects and the reflection problem as well as the possible ramifications of measurement error stemming from inaccurate or imprecise definitions of networks.

In this paper, we estimate the effects of peers' smoking, drinking, sexual behavior and educational achievement on a teen's propensity to engage in like-minded behavior. Exploiting the panel nature of the Adolescent Health Survey (Add Health), we address the major identification problems that plague estimation of these effects. Furthermore, we use the well-defined peer groups in the Add Health data to demonstrate that the failure to precisely define the peer network poses a serious impediment to accurate estimation of peer effects.