Two essays on social interactions

Social scientists have become increasingly interested in effects of social interactions on individual decisions. Social interactions occur when an individual's behaviors are influenced by the behaviors and/or characteristics of the members of her reference groups. Endogenous interactions, where the behavior of an individual depends on the behaviors of the other members of her group, generate multipliers, which can help to understand large variations across areas and over time. Because of the reflection problem and omitted variables problem, identification of effects of social interactions is difficult. This dissertation proposes a strategy to identify and estimate endogenous and exogenous social interactions, and applies it to study effects of social interactions on women's welfare participation.;This first essay studies how women's welfare participation is affected by the participation of other women in her reference group. It takes advantage of the variations in benefit levels across the county and in neighborhood demographic characteristics to construct an instrument to identify endogenous social interactions in welfare participation. An individual's responsiveness to welfare benefits varies with her observable characteristics. Because of neighborhood selection, neighborhoods vary in their demographic composition, and therefore benefit-responsiveness. The first essay uses interactions between benefits and the estimated neighborhood benefit responsiveness using exogenous characteristics of the neighbors as an instrument for neighborhood welfare participation rate, and finds that neighborhood welfare participation has large positive effects on a woman's welfare participation.;The second essay proposes a two-step method to identify and estimate endogenous and exogenous social interactions in Manski (1993) and Brock and Durlauf's (2001) binary choice model with omitted group variables. Taking advantage of large social groups, we first estimate a probit model with group fixed-effects. We use two-stage least squares to estimate endogenous and exogenous group effects using the estimates of group fixed-effects. The method does not depend on distribution assumptions of unobserved group variables, and are computationally simple. When there are multiple equilibria, we can take advantage of large group sizes to estimate endogenous and exogenous group effects without specifying equilibrium selection mechanisms.