Teenagers tend to engage in risky behavior, and many reasons have been identified for this. Still, there are large differences across teenagers, and some literature seems to have identified religiosity as a powerful explanatory variable for less risk taking. But as so often, correlation is not causation. "Nice" kids who follow their parents' advice and behave may also be more likely to follow their parents to church, for example. Perhaps more importantly, individual religiosity may be influence by religious fervor in the social circle beyond the family.
Jennifer Mellor and Beth Freeborn try to disentangle this by using data from the National Longitudinal
Study of Adolescent Health and consider binge drinking, smoking and marijuana use. One may discuss whether the latter is truly risky, but this is not the topic here. Mellor and Freeborn use county-level religious density, a measure that takes into account the proportion of people of a same religious group in the same area, as an instrument to control for the social environment. For religiosity, they use the frequency of church attendance.
Measuring religiosity is tricky business. Density measures depend on the size of the area and how one fragments religious groups (split protestants apart or not, for example). And church attendance may be more a social event than anything else in some areas. I prefer measures like "Do you believe in hell?" to measure this. But let us assume the authors do the right thing (I wished they would supply some robustness exercises).
The results? Marijuana use is robust to taking into account the endogeneity in religiosity. So religious teenager indeed smoke less pot. But smoking cigarettes and binge drinking is a frequent as for less religious teenagers.