“…An obvious response to potential violations of SUTVA is to move the analysis to a more aggregate level, i.e., a classroom, family, organization, or local labor market, at which SUTVA can more plausibly be maintained and estimate macro treatment effects at that level (e.g., Moffitt 2005, Morgan & Winship 2007, Smith 2003. Empirically, this corresponds to sociologists' interests in group-level rather than individuallevel interventions (see Axinn & Barber 2001;Smith 2005;Hong & Raudenbush 2006Sampson 2008), and on a larger scale this is also consistent with Coleman's (1990) theoretical position of sociology as the study of social systems. However, the more inconvenient implication is that actual generative mechanisms will be left unspecified in the analysis, and, to the extent that they are built on social interactions, cannot readily be addressed in the standard statistical framework.…”