The analytical design used both here and in the former study corresponds to a quasi-experiment that controls for state differences in suicide rates, as well as both national and state-specific secular trends (Angrist and Pischke, 2008). This approach can be very powerful in that it analyzes within-state changes in an outcome in relation to policy change. However, confounding can be an issue when state characteristics that change over time and are correlated with both policy and outcome. We showed that adoption of medical marijuana policy was associated with shifts toward populations that were older and higher in percentages of minorities and women (Table S56), which is problematic because minorities and women have much lower suicide rates than whites and men, respectively (Centers for Disease Control and Prevention, 2013; Crosby et al., 2011). Medical marijuana states also trended toward stronger tobacco control policies and higher per-capita mental health spending. Our own recent work suggests that implementation of strong tobacco control policies is associated with reductions in suicide risk (Grucza et al., 2014). These findings are corroborated by analyses presented here (Table
