Association for Public Policy Analysis & Management

College attendance is an important life decision that occurs during a time when social influences are particularly strong. The goal of this paper is to quantify the impact of social networks, specifically the choices made by friends of an individual, on college application and attendance decisions. Policies that shift incentives related to these choices may have their effects amplified by social influences. Therefore, measuring the existence and magnitude of social effects is important both when generalizing policy evaluations from a small group to a larger population and when examining the pathways through which a given policy operates.

The measurement of social effects is often confounded by several dimensions of endogeneity. Since a pair of friends can simultaneously effect one another, quantifying an effect that runs in only one direction is not straightforward. The process is further complicated by the fact that the researcher does not observe all elements of the environment shared by friends that contribute to the outcomes of interest. Furthermore, the formation of friendships is likely the consequence of yet more unobservables which are correlated with the outcomes.

The relationship between the size of a pool of potential friends (for instance, high school size) and the level of similarity between friends has been documented in previous work. In this paper, an instrumental variables approach uses this feature of social networks to identify a casual effect. This approach is unique in the literature in that it controls for all types of endogeneity discussed above, while past work has primarily ignored the endogeneity of friendship formation.

Student, friendship, and school data from the Wisconsin Longitudinal Study are used for estimation. The IV effects are large and significant for submitting any college application and attending college, while smaller and insignificant for considering college. This evidence suggests that friends have the largest effect when a student is near the margin of attending college but are not able to move students to college attendance if they have never considered it. Consequently, policies that target students near the attendance margin can be expected to benefit most from social multiplier effects, while social interactions may play little or no role in the effects of policies aimed at students who would not otherwise consider college.