Medical marijuana policy was coded as “1” for years when use of marijuana for medical purposes was legally sanctioned and “0” for years when it was not. When the policy was in place for only part of the year, we coded the value for the fraction of the year during which the policy was in place; for example, if the policy in a state was implemented on July 1, we coded a value of 0.5 for that year. Sources for policy data included Anderson and colleagues (2014) for the years 1990–2007 and Lynne-Landsman and colleagues (2013) for subsequent years. Individual-level covariates extracted from mortality and living population records included state of residence, age, race/ethnicity, and education. Race/ethnicity was coded as non-Hispanic white, non-Hispanic black, Hispanic, and other. Age was grouped into the categories used by Anderson et al (2014): 15–19, 20–29, 30–39, 40–49, 50–59, 60 and above. Education was dichotomized, with individuals classified as having a high-school diploma or less versus having had some post-secondary education.
