Identifying network effects is notoriously hard. My approach? I take advantage of the exogenous nature of shale formations, plus I isolate predetermined variation in social interactions due to historical migration. This works pretty well (cf. appendix C), fracking did not even exist back then! 4/7
