All models controlled for gender, grade, race/ethnicity, and whether the student attended a nontraditional school. I assessed the parallel trends assumption of the difference-in-difference design by graphing simple prevalence of the marijuana outcomes among Los Angeles and the control group students over the study and visually assessing whether the trends were parallel. The assumption of parallel pretreatment trends was confirmed by the notably parallel trends before Proposition D was enacted. Table 6.2 presents the results of cross tabulation frequencies of the two marijuana measures with student demographic characteristics. Most of the results were consistent with expectations based on the body of literature, such as marijuana use among males being higher than females, use among Asian students being low, and use among students attending non-traditional schools being markedly higher than among students at traditional schools. Among all students the number who reported lifetime marijuana was about one-third but varied considerably by grade, which is not unexpected given that “lifetime” is a longer period of time among eleventh grade students than among ninth grade students. Recent marijuana use was also more common among eleventh grade students, however, with 19% reporting recent use compared to 13% of ninth grade students. These frequencies are consistent with both national and state-level reports. The national Monitoring the Future Study reports that as of 2017 prevalence for lifetime use ranged between 31% for 10th grade students to 45% for 12th grade students,greenhouse racking whereas prevalence of past month use ranged from 16% for 10th grade students to 23% for 12th grade students . State-level reports using a random sample of CHKS participants from 2015-2017 surveys found that 17% of 9th grade students and 32% of 11th grade students reported lifetime marijuana use, whereas prevalence of recent use was 9% and 17%, respectively, in the 2015-2017 survey years.
Table 6.6 presents the difference-in-difference estimates from the Poisson regression comparing marijuana use over time between the City of Los Angeles and the control group cities. The difference-in-difference coefficient is an interaction term comparing the difference in change between the two groups over time and quantifies the impact of Proposition D. Presented below as a risk ratio, the value for the estimate was less than one , which indicated a lower risk of lifetime marijuana use over time and a greater decline relative to the control group cities. That the coefficient was statistically significant means the decline in students’ marijuana use in Los Angeles surpassed the declining countywide trend to such a degree that it is unlikely to have occurred by chance . This result suggests that the decline in rates of lifetime marijuana use among City of Los Angeles high school students is attributable to the stricter regulations enacted with Proposition D and supports H1.2 for lifetime marijuana use. The difference-in-difference coefficients for the covariates presented in Table 6.6 are similarly presented as risk ratios. In this case, they represent the risk of a student within a category reporting lifetime marijuana use relative to the reference group for that category and holding constant all of the other covariates in the model. For example, the risk ratio for males reporting lifetime marijuana were 1.11, or 11% higher than the risk for females reporting lifetime marijuana use. Within racial/ethnic characteristics, students within the African-American, Hispanic, and Other racial/ethnic categories had significantly higher relative risk of reporting lifetime marijuana use than the reference category, Whites. In contrast, Asian students had significantly lower relative risk compared to Whites . The relative risk of eleventh grade students reporting lifetime marijuana use significantly higher than for ninth grade students, as indicated by the 95% confidence interval not including 1. The results for non-traditional schools were in the expected direction as well, with the relative risk of lifetime marijuana among students attending these schools reporting lifetime marijuana was estimated to be almost 75% higher compared to risk of lifetime marijuana use among students attending traditional schools.
The most interesting finding among the covariates was the association between the relative risk of students reporting lifetime marijuana by time. The risk ratios for the preProposition D time periods were all greater than one, indicating that reports of lifetime marijuana use during these periods were significantly greater than baseline . In contrast, the risk ratios for the post-Proposition D time periods were all lower than one, indicating significantly less risk of students reporting lifetime marijuana use during those time periods compared to baseline. The tighter regulations enacted in Los Angeles with Proposition D were followed with lower rates of lifetime marijuana use among high school students when accounting for regional trends and covarying factors. Parallel trends were observed in Los Angeles and the control cities for both lifetime and recent marijuana use, but declines in both these measures were steeper in the City of Los Angeles following enactment of Proposition D. This result supports the hypothesis of a causal effect, although it was not large enough relative to the control group to be statistically significant for recent marijuana use. These results supported Hypothesis 1.2, that cities that enacted more restrictive dispensaries policies would see a trend of declining marijuana use among students attending school there. That a decline in student reports of marijuana use was observed among the control group was unexpected. The similar trend among the control group cities may indicate that marijuana use among high school students is driven less by whether their city allows dispensaries than by secular trends driven by the media, by state and federal laws that impact availability and legal risk for adults. This finding also justified use of the difference-in-difference design to control for background trends in the outcome variable that cannot be attributed to the policy or event of interest. By using the control cities to represent the counterfactual case for student marijuana use trends in Los Angeles had Proposition D not been enacted, I was able to isolate the effect that can be attributed to the policy change and avoid making false conclusions about its impact on students’ marijuana use behaviors. The decline in rates of lifetime and recent marijuana use among the control group cities may have been related to federal enforcement efforts that closed down over 200 dispensaries in the LA County area in 2012 ,indoor cannabis grow system but very little information is available about which dispensaries were closed down in which cities and how many dispensaries were in operation countywide before the raid. It is unknown whether the Federal raids targeted the city of LA and the control cities equally, but if they did than these raids may have played part in the declines in marijuana use that was noted among students in the control group cities as well as among Los Angeles students.
It is difficult to attribute any impacts on student marijuana use to these enforcement actions due to the limited information available but further study of these events is certainly merited. I was not able to conclude that the decline in recent marijuana use observed to occur in the post-Proposition D period was not due to chance. Recent marijuana use is a less common behavior than lifetime marijuana use, and while the effect was in the expected direction, the smaller number of students reporting this behavior produced a wider confidence interval that included a null effect. Policies take time to have a measurable effect and the City of Los Angeles has experienced significant challenges to enforcing Proposition D’s limits on the number of dispensaries. Hundreds of unlicensed storefront dispensaries continue to operate throughout Los Angeles and each one of them could be expected to weaken the impact of Proposition D. It is possible that given more time and continued investment in enforcement an effect for recent marijuana could be documented as well. An additional explanation for why an effect was not observed for recent marijuana use could be contamination, or “spillover effects”, where people living in other cities in LA County may have obtained marijuana from the many dispensaries located in the City of Los Angeles. Car culture is firmly established in Southern California and vehicle ownership is high; close to 8 million vehicles were registered in LA County last in 2017 , for a county with an estimated population in 2017 of just over 10 million . Given the geographic sprawl of Los Angeles and the many other incorporated cities and unincorporated areas it borders, it is not difficult to imagine that LA County residents who lived outside of Los Angeles obtained marijuana from dispensaries located in Los Angeles if they couldn’t get it in their own city. This could be expected to make events that impact access to marijuana in the City of Los Angeles also have an impact in the other cities. It could also be expected to weaken the impact of dispensary bans altogether, as people could obtain marijuana from other cities if it is banned in theirs.
The ability for high school youth to travel to another city to get marijuana could be less of a concern than for adults but given that youth largely obtain marijuana from adults via the illicit market or their social networks , events impacting adult access could be expected to in turn affect youth access. Additionally, even if a city is successful in enforcing dispensary bans or caps on the number of outlets like Proposition D, policies such as these that restrict access to storefront outlets may still have a limited effect on the availability of marijuana given the many other sources by which residents can obtain it, such as from delivery services or by cultivating their own. In this chapter I will present results from analyses that tested several theories for why city dispensary bans may have an indirect effect on student marijuana use . The results presented in Chapter 5 tested the focal relationship for this dissertation using a cross-sectional sample that included students from 57 cities in LA County. That analyses did not provide evidence of a direct effect between city dispensary bans and high school students’ marijuana use when controlling for student and school characteristics known to be associated with adolescent marijuana use. In contrast, the trend analysis presented in Chapter 6 showed that in the City of Los Angeles enacting and implementing a policy intended to reduce the number of dispensaries and place additional controls on their operation was followed by a decline in lifetime marijuana use among students attending the city’s public high schools. Furthermore, city dispensaries were negatively associated with student marijuana, despite the associations falling short of statistical significance. In this chapter, I therefore conducted a series of mediation analyses to elaborate on the relationship between city dispensary bans and student marijuana use and whether their effect was dependent on some factor I had not accounted for. In this chapter I investigate indirect effects; circumstances where the effect of a variable is dependent on another variable. Identifying dependent relationships is important to elucidate some of the mechanisms by which restrictive city regulations on legal, adult-use products might be effective in preventing substance use among adolescents. The analyses that follow in this chapter tested indirect mechanisms through which I theorized city dispensary policies may influence students’ marijuana behaviors, such as by preventing excessive density of dispensaries in a city, signaling to youth that marijuana use represents a health risk, and/or by preventing dispensaries from operating near their high schools. These analyses will test the hypotheses for Research Questions 2-5. As described in the methods chapter , I used a variation of Baron and Kenney’s Product Method, which described in detail in Zhou et al., 2010, to test for mediation. Baron and Kenney’s Product Method first tests for direct relationship between the independent and dependent variable as a condition of testing for mediation , but this approach has been criticized as a relationship between the independent and dependent variables can be masked by a mediating variable or competing mediators . In the case of this research, the focal relationship is the influence of dispensary bans on student marijuana use and I will investigate the effect of several different mediators on this relationship.