West Coast Ballroom
Extending difference-in-difference methods to test the impact of state-level marijuana laws on substance use using published prevalence estimates. (306776)
*Christine Mauro, Columbia University Mailman School of Public HealthMelanie Wall, Columbia University Mailman School of Public Health
Nicholas Williams, Weill Cornell Medical College
Keywords: difference-in-difference methods, policy impact, recreational marijuana laws
Colorado and Washington legalized the use of recreational marijuana in 2012; since then eight more states and DC have passed laws. Evaluating the impact of these laws on substance use and substance use disorders is of critical public health importance. Available data are two-year aggregated prevalence and standard error estimates of past-month marijuana, alcohol, and tobacco use from the National Survey on Drug Use and Health, 2006-2007 through 2016-17. We first extend difference-in-difference methods (DID) to assess the impact of these laws on a state-by-state basis using simulated datasets constructed from the published prevalence and standard error estimates. Next, we consider estimating the average effect of law passage from all states with recreational laws using linear mixed-effects models that account for historical trends in use over time borrowing information from states without laws. Lastly, we present a method for visually assessing the parallel-trends assumption when time of exposure varies across units, as is the case in this scenario.