Peer effects in sexual initiation: Separating social norms and partner supply

Sexual activity among adolescents is associated with risks such as pregnancy and STDs, and there is substantial policy interest in how peers influence the timing of sexual initiation. This paper measures separate effects for two social mechanisms—peer-group norms and partner availability—using a national sample of high school students. I develop and estimate an equilibrium search and matching model for first sexual encounters that specifies distinct roles for the two mechanisms. Norms are defined based on the share of nonvirgins among same-gender peers, which influences whether an individual searches for a sexual partner. Supply is modeled with an arrival rate for partners, which depends on the search behavior among the opposite gender. The model produces a discrete-time duration to first sex which I estimate with quarterly data on individual virginity status constructed from the Add Health study. The endogeneity of peer behavior with respect to individual behavior is addressed with a combination of strategies. First, I use standard instrumental variables methods to estimate linear regressions for virginity status at the end of each grade. Instruments for group nonvirginity rates are person-specific characteristics such as sibling structure and age of menarche, and the regressions include school-by-grade fixed effects. This analysis demonstrates that school-based social interactions have a large effect on sexual initiation. Second, I estimate the search and matching model via simulated maximum likelihood, in order to decompose this composite effect into separate effects of peer norms and partner availability. Here I control for the endogeneity of peer behavior by (a) defining the norm effect as a function of lagged peer outcomes, (b) including a random effect that is correlated within schools, and (c) using exogenous peer characteristics as supply shifters. I find that peer-group norms have a large effect on the timing of sexual initiation: removing the peer influence on search decisions, 42% fewer boys and 22% fewer girls become sexually active in ninth or tenth grade. Changes in the availability of partners at school (i.e., changes in opposite-gender search behavior) also have a large impact on initiation rates for boys, although not for girls