Previous studies have shown that adolescent religious participation is negatively associated with risky health behaviors such as cigarette smoking, alcohol consumption, and illicit drug use. One explanation for these findings is that religion directly reduces risky behaviors because churches provide youths with moral guidance or with strong social networks that reinforce social norms. An alternative explanation is that both religious participation and risky health behaviors are driven by some common unobserved individual trait. We use data from the National Longitudinal Study of Adolescent Health and implement an instrumental variables approach to identify the effect of religious participation on smoking, binge drinking, and marijuana use. Following Gruber (2005), we use a county-level measure of religious market density as an instrument. We find that religious market density has a strong positive association with adolescent religious participation, but not with secular measures of social capital. Upon accounting for unobserved heterogeneity, we find that religious participation continues to have a significant negative effect on illicit drug use. On the contrary, the estimated effects of attendance in instrumental variables models of binge drinking and smoking are statistically imprecise. Copyright © 2010 John Wiley & Sons, Ltd.