A lottery test of the effect of dispensaries on emergency room visits in Arizona

In August 2012, the Arizona Department of Health Services conducted a lottery to allocate licenses for the state's first medical marijuana dispensaries. The lottery randomly selected an applicant within each of 69 contested Community Health Analysis Areas to open a dispensary. By comparing 36 zip codes with winning applications to 48 zip codes with losing applications and weighting using propensity scores based on the true probability of winning, we estimate the causal effect of the allocation of a dispensary on the emergency room visits of residents of that zip code. Outcomes of interest are emergency room visits for acute symptoms caused by cannabis, opioids, alcohol, and cocaine. Using emergency room discharge data from 2010 to 2016, we find evidence of an increase in visits for acute cannabis-related causes for the winning set of zip codes and weak evidence of an increase in visits for opioid-related causes. The results indicate that in the four years following the lottery, emergency room visits for acute cannabis causes rose by approximately 45% in allocated zip codes relative to non-allocated zip codes. Because of the high likelihood of spillovers to neighboring zip codes, these effects are likely underestimates.