Among the small number of cities in LA County that allowed dispensaries, this variation in the number of students from the City of Los Angeles had such a significant influence as to effect estimates of marijuana use among the cities that allowed dispensaries from year to year. Using two combined school years solved this problem and resulted in a greater number of schools being included in the sample, which increased the validity of the geographic analyses in the chapters that follow. Table 5.2 displays results for the cross-tabulations between city dispensary bans and student and school characteristics that I hypothesized would be associated with high school students’ self-report of lifetime marijuana use. Just under a quarter of the students in the cross sectional sample reported having ever used marijuana. Similar numbers of males and female students reported having ever used marijuana, so this marijuana use was not found to differ significantly by gender. Reports of lifetime marijuana use did vary significantly by race/ethnicity and was higher among Hispanic students and African American students than among White students, while use among Asian students was lower than among White students. Attending an after-school program at least 1 day a week was significantly associated with lower rates of lifetime use, while receiving free or reduced-price school meals was significantly associated with higher rates of marijuana use . The proportion of students attending non-traditional schools who reported lifetime marijuana was more than double the proportion reported by students attending traditional schools. Finally, students attending school in a city that banned dispensaries were significantly less likely to have reported lifetime marijuana use than students who attended schools in cities that allowed dispensaries . Table 5.3 displays results for the cross-tabulations between city dispensary bans and student and school characteristics that I hypothesized would be associated with high school students’ self-report of recent marijuana use.

The number of students who reported having used marijuana within the previous 30 days was a little over half the number who reported ever having used it . Similar numbers of males and female students had reported having ever used marijuana,vertical farming equipment but significantly more males reported using marijuana within the last 30 days compared to females. Reports of recent marijuana use also varied significantly by race/ethnicity and was highest among Hispanic students while recent use among Asian students was less than half the proportion reported among the all the other racial/ethnic categories. Attending an after-school program at least 1 day a week was significantly associated with lower rates of recent us . Interestingly, while receiving free or reduced-price school meals was significantly associated with higher rates of lifetime marijuana use , it was associated with significantly lower rates of recent use, indicating that low family income may present a barrier to more frequent use. The proportion of students attending non-traditional schools who reported lifetime marijuana was almost triple the proportion reported by students attending traditional schools . Finally, the proportion of students attending school in a city that banned dispensaries who reported recent marijuana use was greater than the proportion who reported recent use among students who attended school in a city that allowed dispensaries , but the difference fell just short of statistical significance . After assessing bivariate associations, I ran separate multilevel logistic regression models for lifetime and recent marijuana use to assess the impact of city bans on each measure of student marijuana use, while controlling for the student and school characteristics that the bivariate analyses had revealed were significantly associated with each measure of marijuana use. Two level random intercept HGLM models were conducted using PROC GLIMMIX in SAS 9.4 with city as the level-2 variable to account for the multilevel structure of individuals being clustered in cities. School-level weights for the LA Unified School District schools were included in each model, per CHKS documentation .

The multivariate models presented here show that most of the student and school characteristics that I hypothesized would be important influences on students’ marijuana use based on the body of literature were indeed strongly associated with these outcomes. It was therefore important to control for these characteristics when attempting to quantity the effect of dispensary bans in student marijuana use. By controlling for these characteristics, I was able to demonstrate that city dispensary bans do not have a direct association with lower rates of marijuana use among high school students in LA County and conclude that hypotheses H1.1 was not supported. It is possible that city dispensary policies are too many links above students in the marijuana supply chain to directly have an impact on how much marijuana they can access or how reliably. It also possible, however, that the effectiveness of marijuana policies is dependent on factors that I did not include in the models above, such as enforcement practices.This population represents an excellent target for school-based secondary prevention interventions and screening for clinical levels of substance use disorder. The finding that receiving free or reduced-price school meals was significantly associated with higher rates of lifetime marijuana use but was associated with significantly lower rates of recent use is consistent with literature demonstrating that adolescent substance use is responsive to pricing. It’s possible that the adolescents from low income families may have had less money to spend on marijuana than their peers and thus were able to use marijuana less frequently, although the higher rates of lifetime use may indicate that they are just as likely or more likely to have access to it for experimentation or occasional use. That finding that participation in after school programs was protective against both lifetime and recent marijuana is consistent with research showing that youth who participate in after school programs are less likely to report substance use .

This finding also supports an important function of the Adult Use Marijuana Act that mandates that a portion of the tax revenue from recreational marijuana be used to support after school programs . The cross-sectional analysis presented here measured the association between city dispensaries and student marijuana while controlling for students being clustered in cities and for potentially confounding student and school characteristics. City dispensary policies, however, are not the exclusive determinant of the actual availability of marijuana in a community. As will be demonstrated in the following chapters, factors such as enforcement and local context also determine access to marijuana in a city. To expand on the relationship between city dispensary policies and adolescent marijuana use,4×4 grow tray the following chapter will investigate the long-term effects of implementing a more restrictive dispensary policy in the City of Los Angeles. Chapter 7 will continue to elaborate on the relationship by testing indirect mechanisms through which 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 risk to their health, and/or by preventing dispensaries from operating near their high schools. Prevention research supports the idea that more convenient access to legal substances for adults often has the end result of creating easier access for youth , which may mean that youth living in or attending school in a city that allows dispensaries can obtain cannabis more easily or more often from adults in their social network. If this is the case, a dispensary policy making access less convenient for adults could have the additional effect of making it less conveniently obtained by teens. Considering that adolescents report older relatives and the illicit market as their primary sources of cannabis tightening dispensary regulations could have a dampening effect on youth marijuana use in Los Angeles even though adolescents are not allowed to access dispensaries directly. To date, little is known about the effectiveness of dispensary bans or other dispensary regulations at preventing youth access to marijuana. By comparing how students’ marijuana use changed over time in a city that allowed dispensaries compared to a group of cities that did not, I hoped to provide some insight into how city policies regulating dispensaries influence marijuana use among youth.

I undertook this task by comparing marijuana use rates among City of Los Angeles high school students before and after the City enacted Proposition D, a voter approved ballot measure that capped the number of outlets allowed to operate in the City at a fraction of their existing number, ordered hundreds of remaining dispensaries to close down, and prohibited all new outlets . I hypothesized that this radical policy change would have an important influence on the availability of marijuana in the City and that rates of adolescent marijuana use in the city would decline after the policy was enacted . To test this hypothesis, I used a difference-in-difference approach to compare change in student marijuana use in the city of Los Angeles, to change in a group of cities that had banned dispensaries throughout the study period. By using the control group to account for any trends in high school students’ marijuana use unrelated to the implementation of Proposition D in the City of Los Angeles I could determine whether the stricter regulations enacted in the City of Los Angeles had a discrete impact on rates of marijuana use among the City’s students . This hypothesis was also a good fit for analysis using a difference-in-difference design given that Proposition D represented a discrete policy change that could be used to clearly distinguish pre- and post-intervention periods in Los Angeles. Difference-in-difference analyses can be performed using different regression techniques. The difference-in-difference coefficient is an interaction term included in a regression equation that compares the difference in change between the two groups over time. In the case of these analyses, it quantifies the impact of Proposition D in Los Angeles relative to the cities where it did not apply. The difference-in-difference coefficients for the covariates presented in in this chapter are presented as risk ratios and can be interpreted as they would be in any Poisson regression model. In this case, they represent the relative risk of a Los Angeles student reporting marijuana use relative to the reference group reporting marijuana use and holding constant all the other covariates in the model. I used robust Poisson regression analyses to test the impact of Proposition D in Los Angeles on the dependent variables, student self-reports of lifetime and recent marijuana use. Although my dependent variables were binary and not count variables, I chose to use a Poisson regression because I was reporting prevalence ratios of behaviors that were not rare, i.e., the prevalence of marijuana use for both measures was over 10%. Under these circumstances using logistic regression and reporting odds ratios can overestimate the prevalence ratio , potentially leading to false conclusions about the volume and statistical significance of intervention effects. I used clustered standard errors in the Poisson regression to account for the grouping of participants in the cities where they attended high school. In such circumstances there may be independence across clusters but correlation within clusters. When this is the case, statistical inference based on the usual assumption of independent observations is no longer appropriate . A common approach to control for clustering is by computing cluster-robust standard errors that control for clustering at the level of the primary theoretical grouping, which in this case was city . I will use this approach to account for the fact that students are nested in cities and there may be unmeasured city effects that influence marijuana use behavior . This approach was used to account for students being clustered within cities rather than a multilevel analysis for simplicity and because the city-level analytical variables used to answer Research Questions 2-5 were not available for the survey years used in the trend analysis. Study weights were included in the descriptive and regression analyses presented below but were available only for LA Unified Students and for the 2016/2017 school year in the datasets I obtained from WestEd. I compared results including and excluding the weights and there were no differences, but I included the school weight where applicable in the descriptive statistics and regression analyses to control for the survey design to the extent that I could.