Estimating the Effects of Friendship Networks on Health Behaviors of Adolescents

Researchers typically examine peer effects by defining the peer group broadly (all classmates, schoolmates, neighbors) because of the lack of friendship information in many data sources as well as to enable the use of plausibly exogenous variation in peer group composition across cohorts in the same school. This paper focuses attention on the effects of friend’s health behaviors on own health behaviors for adolescents combining a cross-cohort, within school design with controls for friendship options through high school fixed effects and friendship choices through the use of “friendship type” fixed effects. This strategy allows us to separate the effect of friends behavior on own behavior from the effect of friends observables attributes on behavior, a key aspect of the reflection problem. Our results suggest that friendship network effects are important in determining adolescent tobacco and alcohol use but are over-estimated in specifications that do not fully take into account the endogeneity of friendship selection by 20-30%  We received valuable comments from seminar participants at Lafayette College, Lehigh University, and Yale University. This research uses data from Add Health, a program project designed by J. Richard Udry, Peter S. Bearman, and Kathleen Mullan Harris, and funded by a grant P01-HD31921 from the National Institute of Child Health and Human Development, with cooperative funding from 17 other agencies. Special acknowledgment is due Ronald R. Rindfuss and Barbara Entwisle for assistance in the original design. Persons interested in obtaining data files from Add Health should contact Add Health, Carolina Population Center, 123 W. Franklin Street, Chapel Hill, NC 27516-2524 ([email protected]). Introduction Individuals in modern societies are socially connected in a multitude of ways. For example, the social networking website Facebook.com has increased its membership by 100 million users in the last eight months (as of July 2009), bringing the total to over 200 million users worldwide. Individuals use their social networks to receive and send information as well as establish, update, and enforce social norms of behavior. Both information acquisition as well as the impacts of social norms within social networks could have large effects on the health behaviors of individuals, particularly adolescents, who are particularly responsive to peer pressure (Brown et al. 1997). This heightening of peer influence also takes place during the developmental stage when many of the most costly health outcomes and behaviors are initiated. This confluence of events during adolescence sets the stage for a critical period of individual decision-making with long-term consequences and also provides a unique window of opportunity for health intervention. While much recent research has shown robust and important correlations in the health behaviors of socially connected individuals, whether the correlations are causal is still largely unknown for many health behaviors. Unlike research on broader peer groups, most work on the effect of friendships and/or the associated networks are identified using information that arises from connections between individuals within broader peer groups assuming that those connections or links are exogenous. . In this paper, we implement a new method of estimating the causal effects of friendship networks. We build on the large literature suggesting that individuals exhibit strong racial, gender, and age preferences when choosing their friends—likes choose likes (Mayer and Puller 2008, Weinberg 2008). Specifically, we use unique data from the Add Health, which contains information both on friendship nominations and health behaviors for a national sample of adolescents. The structure of the Add Health data allows us to control for friendship options and friendship choices using school and friendship choice pattern fixed effects, respectively. Following logic similar to Dale and Krueger (2002), observationally similar individuals who face the same friendship opportunity set must be similar on unobservables if they make similar friendship choices. The inclusion of fixed effects for friendship choices yield consistent estimates of the