Most studies have focused on high school students who are likely to have greater access to marijuana and are more susceptible to social pressures than early adolescents . Meanwhile, young adults different substantially from these younger groups, both in terms of development and access to drugs, being in the peak years of engagement with psychoactive substances during the lifespan . We performed multi-level, serial cross-sectional analyses on 10 annual waves of the U.S. National Survey on Drug Use and Health , from 2004 to 2013. Unlike many prior studies, ours included the key years of 2010–2013—a period of rapid acceleration in the number of states implementing medical marijuana laws , but before state recreational marijuana laws began implementation. In addition, our analyses compared young people across developmentally distinct age groups to account for important age-related heterogeneity in access to marijuana, in the propensity to experiment with psychoactive substances, and in the potential harms of marijuana use.The primary data source was ten annual waves of the NSDUH from 2004 to 2013. Following security clearance and a data use agreement with the U.S. Substance Abuse and Mental Health Services Administration, our team obtained access to individual-level NSDUH data that included the state of residence for each respondent. Each wave of the survey represents the U.S. population in all 50 states and the District of Columbia. During the period studied, no major changes in sampling, data collection, or instruments were made, thus preserving comparability across survey years. Full details of the data collection protocols, informed consent, and the questions asked are available in U.S. Substance Abuse and Mental Health Services Administration methodology reports . This project received an ethics review and was approved by the University of California at San Francisco’s Committee on Human Research. The total sample,vertical grow pooled over 10 years, includes approximately 450,300 individuals. We stratified young people into three discrete age groups: early adolescents , late adolescents , and young adults . Table 1 provides an overview of sample characteristics.

All participant data was provided by the U.S. Substance Abuse and Mental Health Services Administration and is not based upon primary collection of clinical study or patient data requiring individual consent.We examined three dichotomous outcomes at the individual level: self-reports of the accessibility of marijuana, consumption of marijuana within the past month, and initiation or first-time use of marijuana during the past year. The NSDUH framing of the marijuana questions references smoking, edibles, and oils. Individual-level, age-appropriate predictors from the NSDUH dataset were included in the analysis. Across all three age groups, these included sex, race/ethnicity, family income, poor or fair health, and living in an urban area. We included an indicator of poor or fair health status to control for the possibility that participants in medical marijuana states might engage in the legal use of marijuana for health reasons. For early and late adolescents, we also controlled for parental monitoring and participation in group fights, variables that could be indicators of the protective factor of parental involvement and the risk factor of delinquent behavior, respectively. For young adults, additional controls included employment, college attendance, parental status, and marital status. These are strong protective factors mitigating against drug use in this age group . We augmented the NSDUH data with annually updated state-level data on medical marijuana laws and other relevant control variables. For state-level controls, we drew on publicly available sources such as Polidata , including per capita drug courts and whether or not marijuana possession had been decriminalized. We considered a wider range of state-level controls representing demographic, political and religious factors, and aspects of state drug control policies. For the sake of parsimony, we included controls that were most associated with outcome variables. Data collection on state medical marijuana laws included gathering all state statutes and subsequent regulations, and validating information against publicly available data sources and through telephone calls with state officials. Throughout the study, we conducted regular updates to monitor changes in regulations and amendments to state laws . Analyses incorporated a dichotomous measure reflecting whether a state did or did not have a medical marijuana law enacted during any given year of observation. Thus, a law passed or enacted at any point in a calendar year would count that state as a medical marijuana state for that year’s analysis.

We also examined a wide range of characteristics of state laws, such as the amounts of marijuana legally allowed for possession and home cultivation, medical conditions covered, and the number of dispensaries in each state. Through a systematic measurement process, we created and validated a scale capturing the capacity of a given medical marijuana law to control marijuana distribution and diversion into illegal markets .We concatenated 10 annual waves of the NSDUH and all state-level indicators into a single data file. We conducted all analyses using Stata version 13 . For descriptive analyses of each survey year, we used weights to adjust for sampling design effects and non-response ; similar weights were not available for multi-year analyses. Following Williams and others , we accounted for shared variance among participants within states by calculating standard errors clustered at the state level in our regression models. Our analytic approach used logistic regression to predict marijuana consumption and initiation at the individual level separately for early adolescents, late adolescents, and young adults. The regression models incorporated all individual- and state level controls and annual fixed effects. A key analytic concern is that people in states that pass medical marijuana laws hold more permissive views about the drug . These more positive perceptions about marijuana may drive both the passage of the medical marijuana laws and higher rates of consumption . We incorporated that possibility into our uncontrolled comparisons of young people who dwell in states with medical marijuana laws compared to those who do not . By controlling for state-level fixed effects , we were able to examine whether medical marijuana laws have distinct causal impacts on marijuana consumption and initiation . The coefficients for each state controlled for any state-specific confounding not already captured by other control variables in the models. This technique allowed us to rule out the possibility that unobserved state-level confounders account for any associations found between state medical marijuana laws and young people’s consumption and initiation of use.Using the most recent NSDUH survey, 2013, we compared rates of access to marijuana, past-month marijuana use and past-year initiation across early and late adolescent youth and young adults. Table 1 shows pronounced differences in the populations of young people living in states with medical marijuana laws compared with those who were not.

These demographic differences—especially ones associated with drug-related attitudes—underscore the importance of applying individual-level controls in the analysis. For example, in 2013, individuals living in medical marijuana states were disproportionately white and Hispanic. Young adults living in medical marijuana states were comparatively less likely to be married and to have children. Figure 1 shows a positive age gradient in rates of reporting that marijuana is easily accessible and in past-month marijuana use: The highest prevalence occurred among young adults at 19.1%, then 11.9% of late adolescents and 2.2% of early adolescents . In contrast, initiation of marijuana use in the past year was most common among late adolescents , with young adults the next most likely to initiate marijuana use and early adolescents the least likely to have tried marijuana for the first time in the past year .Table 2 shows logistic regression models predicting past month marijuana consumption that include all individual and state-level controls,vertical outdoor farming and annual and state-level fixed effects. Results provided no evidence of a causal relationship between living in a state where medical marijuana was legal and the past month use of marijuana. Across all age groups, the odds ratio associated with medical marijuana state residence was not statistically significant. Table 3 provides similar fully controlled results for logistic regression analyses predicting past-year initiation of marijuana use. Results show that young adults dwelling in states that have legalized medical marijuana are significantly more likely to initiate marijuana use than counterparts in non-medical marijuana states . Such a relationship is not evident for early or late adolescents . We performed additional analyses to rule out several alternate explanations of these findings. Incorporating the amount of time since the passage of the medical marijuana law into our models produced similar results regardless of duration of the law. To rule out the possibility that young adults are more likely to initiate marijuana use due to mental health conditions, which in some states are legally allowed indications for a medical marijuana prescription, we estimated an alternate version of the models that included additional mental health-related variables, specifically, past-year use of mental health treatment and past-year unmet need for mental health treatment. After introduction of these additional controls, the effect of living in medical marijuana state remained statistically significant for young adults . We also considered the possibility that states with less restrictive medical marijuana laws could have more significant impacts on young people. We repeated all analyses with these variables and a summary scale reflecting the strength of controls on medical marijuana distribution, but these analyses failed to produce statistically significant results for any age group .Prior studies focusing on whether medical marijuana laws impact young people’s consumption of marijuana have produced mixed results. In the absence of robust evidence that medical marijuana laws are not adversely impacting young people, the number of states passing these laws has accelerated. The analyses presented here found that medical marijuana laws are not causally associated with recent marijuana consumption in young people.

However, we did find that medical marijuana laws impact the initiation of marijuana use, but that this is confined to young adults and does not include the more vulnerable populations of early and late adolescents.Prior research has largely focused on how medical marijuana laws impact rates of marijuana consumption, placing less emphasis on the initiation of marijuana use. But the potential for these laws to impact the age-at-first use of marijuana has considerable public health significance. Initiating marijuana use in early adolescence is an important prognosticator for subsequent drug dependence . Younger age at initiation is one of the strongest predictors of drug dependence and related problems later in life . Although we found a positive age gradient in the rates of consuming marijuana during the past month, patterns were different for initiating marijuana use. This finding is consistent with developmental theories suggesting that high-school age youth are uniquely prone to act on social messages and to experimentation with drugs . Fully controlled regression analyses showed that medical marijuana laws significantly increase the likelihood of trying marijuana for the first time among young adults, but not younger age groups. Young adults are in the peak years of engagement with illicit drugs during the life course . Compared to early and late adolescents, young adults have heightened availability and opportunities to use illicit drugs. This age group is past the peak age for initiating marijuana use and is therefore at a reduced risk for developing persistent marijuana-use disorders. However, our findings suggest that some of the young adults in this study might never have tried marijuana had they not been in a state that legalized medical marijuana. Future research should disentangle the mechanisms that account for why young adults in states with medical marijuana laws are more prone to initiating use. This finding is consistent with the notion that medical marijuana products may be diverted into illegal markets, thus increasing marijuana’s availability and driving down its price . Increased accessibility of illicit drugs is an important factor predicting the likelihood that individuals will initiate use . Where larger numbers are using marijuana, whether for medical or non-medical reasons, individuals interested in trying the drug can more easily access information on how to obtain it . Another possibility is that, in states with relatively lax enforcement of existing medical marijuana regulations, young adults are more willing to try marijuana because they perceive that the risk of arrest is low or generally perceive the drug as less risky.