Recreational cannabis legalization : predicting local policy adoption and estimating the associated effects on population cannabis use

Cannabis is undergoing a remarkable transformation from a regulated drug to a recreationally legal one in the United States (U.S.). Yet, in states that have legalized recreational cannabis, there is substantial geographic variability in actual cannabis policies and the effects of cannabis legalization are still being debated. This dissertation addresses these modern scientific issues of the recreational cannabis landscape. The population under study primarily includes non-institutionalized U.S. civilian residents, sampled and assessed in successive waves of the National Survey on Drug Use and Health (NSDUH) starting in 2008 through 2019. Estimates on drug use and mental illness prevalences are aggregated to the county level for the first aim, and to the state level for the second and third aims. In the first aim, the county-level data are linked to several other publicly available sources of information on all 3,142 U.S. counties including the 2010 Census, 2012 presidential election, and recreational cannabis sales policies. I then used these data to train a machine learning algorithm to predict which counties allowed for the recreational sale of cannabis in 2014. In the second aim, I used state-level estimates of cannabis incidence in an event study model to estimate the effects of legalizing recreational cannabis on cannabis use onsets for persons under and over the legal minimum age of 21. The final aim focuses specifically on 21 year-olds to better understand the implications for setting a legal minimum age drug policy on age-specific patterns of incidence and proposes a theoretical framework that may help understand these findings. For the first aim, the model-averaging predictions classified almost 94% of the U.S. counties correctly. The main factors associated with county-level recreational cannabis laws were the prevalences of past-month cannabis use and past-year cocaine use. In the second aim, I found that for those who were legally able to purchase cannabis (21 and older), cannabis legalization did not appear to affect incidence in the first year following legalization. Even so, between two and four years after legalization, the difference in differences modeling disclosed statistically robust increases of 0.6% for this sub-population of adults. After four years, the estimated increase is 1.3%. The corresponding estimates for underage persons who were ineligible to legally purchase cannabis show no appreciable differences in the occurrence in past-year cannabis use incidence. Finally, the age-specific incidence estimates for 21-year-olds show a rise after the passage of recreational cannabis laws (RCL) and are suggestive of the arrival of a new pattern of age-specific incidence. Taken together, the work and results of this dissertation point toward four potential conclusions. First, cannabis legalization might depend on a predictable process driven in part by prior drug use in each jurisdiction. Second, once implemented, recreational cannabis legalization might not have effects on adolescent onset newly incident cannabis use. Third, for adults permitted to buy cannabis without penalty, the occurrence of newly incident cannabis use might increase. Fourth, a tentative conclusion is that legalization of retail sales to adults removes a barrier for adults who had been interested in trying cannabis, but did not do so, perhaps due to concerns about legal or social consequences faced before legalization.