A variable is generally considered a mediator if it carries the influence of a given independent variable to a given dependent variable . Mediation is often assessed by how much the independent variable affects the mediator, and how much the mediator affects the dependent variable, and whether the effect of the independent variable on the dependent variable is reduced upon the addition of the mediator to the model, i.e., by controlling for the mediator. This would be the case if complementary mediation is occurring, but this is not the only pattern of mediated relationships. Instead, Zhou et al. describe three patterns consistent with mediation. The first, complementary mediation, is described above. The second, competitive mediation, occurs when the direct effect and the mediated effect both exist but point in opposite directions. The third pattern is indirect-only mediation, where a mediated effect exists, but no direct effect . In other words, the independent variable is dependent on the mediator to have an effect. This third pattern, indirect mediation, is what I will be testing for in this chapter. In cases of indirect mediation, adding the mediator to the model would be expected to strengthen the association rather than reduce it . The number of dispensaries in a city could be a measure of how strict a dispensary policy is if it accurately reflects the number of dispensaries that the city has allowed. It could also reflect how effectively the policy is enforced, for example if the number of dispensaries is greater than the number allowed.
I therefore compared three related measures of the density of dispensaries within a city: the number of verified licensed dispensaries,equipment for growing weed the number of verified unlicensed dispensaries, and the overall number of verified dispensaries for the city. Each of these measures were normalized by city population and converted to rates of dispensaries per 10,000 city residents. The first research question that this chapter will address is Research Question 2: “Is the effect of dispensary bans on student marijuana use dependent on the number of dispensaries operating in a city?”. The hypotheses associated with RQ2 propose that H2.1) dispensary bans are negatively associated with the number of dispensaries operating in a city, H2.2) the number of dispensaries in a city is positively associated with greater odds of marijuana use among students, and H2.3) the number of dispensaries in a city mediates the relationship between city policies banning dispensaries and high school students’ marijuana use behaviors. The need to account for the actual number of dispensaries in a city rather than assuming cities that had dispensary bans had no dispensaries in operation was evident when I mapped the location of the dispensaries throughout the County based on the addresses I had had obtained from commercial dispensary listings . In support of reports of numerous unlicensed outlets from city and county governments , I verified 433 unlicensed outlets and 146 licensed outlets operating in the County in September of 2016, a ratio of unlicensed to licensed outlets of over 3 to 1. To test the hypotheses associated with Research Question 2, I used the number of verified dispensaries located within a city per 10,000 residents as the mediating variable and controlled for factors known to be associated with marijuana use among adolescents, such as gender, race/ethnicity, and social/economic status.
The independent variable of whether a city had a dispensary ban was determined by the city policy that was in effect when the count of dispensaries per city was obtained in September 2016. The outcome variable was self-reported student marijuana use . As previously presented in Chapter 5, there was a not a statistically significant negative relationship between dispensary bans and lifetime marijuana use . I theorized that this relationship might depend on the number of dispensaries operating in a city, given that the reality of dispensary density in many cities did not reflect what one would expect based on the city dispensary policy. I tested this theory by regressing the indicator variable for dispensary bans on the number of dispensaries in a city using a hierarchical linear regression model with the number of dispensaries per 10,000 residents as the dependent variable and whether the city had a dispensary ban as the independent variable. I included the covariates mentioned earlier and a random intercept for city. As shown in the t-test results comparing the mean rate of dispensaries per 10,000 residents by whether the city had a dispensary in September 2016 , there was a statistically significant negative association between dispensary bans and the number of dispensaries in a city. This result was replicated using a single-level linear regression . The significant association between dispensary bans and less dispensaries per city resident supported H2.1. The number of dispensaries per 10,000 residents was then regressed on the likelihood of lifetime marijuana use , which revealed that the rate of dispensaries per residents in the city was not significantly associated with student reports of lifetime marijuana use . This finding refuted H2.2, where I theorized that the number of dispensaries was the key determinant of marijuana use among students. In the final model, which accounts for path c’, the dispensary ban indicator variable was regressed on both lifetime marijuana use and the rate of dispensaries in the city.
The coefficient for dispensary bans after controlling for the rate of dispensaries in the city changed very little and remained negative and statistically insignificant . These results indicated that while there is a statistically significant relationship between dispensary bans and the number of dispensaries in a city , the number of dispensaries actually operating in the city is not a significant determinant of lifetime marijuana use among the city’s high school students . Furthermore, the relationship between dispensary bans and lifetime marijuana use among the high school students in a city is not dependent on dispensary bans limiting the number of dispensaries operating in a city, which negated hypothesis 2.3. I repeated the steps described above for recent marijuana, with similar outcomes, which are also reported Table 7.3. There was not a statistically significant relationship between dispensary bans and student reports of recent marijuana use , Table 7.3 presents the results of the analysis of the number of dispensaries in a city as a potential mediator of the effectiveness of dispensary bans in preventing recent marijuana use among high school students policy. As reported above, there was a statistically significant negative association between dispensary bans and the average number of unlicensed dispensaries in a city ,grow tables 4×8 which supported H2.1. The number of active dispensaries was then regressed on recent marijuana use. The number of dispensaries operating in the city was not significantly positively associated with student reports of recent marijuana use . This finding refuted H2.2 for recent marijuana use, which proposed that there would be a direct positive association between the number of dispensaries in a city and the likelihood of students attending school in that city to report recent marijuana use. In the final model , the dispensary bans variable was regressed on the number of dispensaries in the city and self-reported recent marijuana use among the students. The estimate for dispensary bans remained positive but not statistically significant after controlling for the number of dispensaries in the city and changed very little in magnitude . This suggests that the impact of dispensary bans on the likelihood of a student reporting having recently used marijuana is neither significant nor dependent on the association between dispensary bans and lower dispensary density in a city . Included as a measure of the actual exposure to dispensaries in communities, the number of dispensaries per 10,000 city residents had surprisingly little influence on the outcomes of interest for this study. As youth are not allowed to access these storefront outlets directly, the presence of dispensaries in their city may have little impact on the availability of marijuana within their social circles.
The finding that the rate of dispensaries per 10,000 residents in their community had no effect on high school students’ marijuana use was in line with research indicating that adolescents generally do not get marijuana directly from dispensaries, but rather from social sources like relatives or friends . I hypothesized that a greater number of dispensaries located within a city could create more convenient access for the adults that act as a conduit of marijuana to adolescents. However, creating easy access for adults through legitimate sources like dispensaries may have also shrunk the illicit market as a source for adolescents. Further investigation of this effect is needed, such as to determine if effects differ within cities or are confounded by any variables that were not measured here. Attitudes toward drugs and alcohol are known to be powerful predictors of adolescent substance use , and changing attitudes to perceive cannabis use as more acceptable and less risky have been noted among youth populations . For example, qualitative research with at-risk youth in LA County indicates that many view marijuana use as having fewer negative consequences than alcohol use . A community assessment conducted in LA County also found that the risks of cannabis use were rated much lower among cannabis users than among non-users , indicating a potentially important relationship between perceptions of the risk of marijuana use and the willingness to use it. Table 7.6 presents the results of the perceived risk mediation analysis for lifetime and recent marijuana use. Whether a city had a dispensary ban was then regressed on whether students perceived great risk from frequent marijuana use using a HGLM model controlling for the covariates and with a random intercept for city. There association between dispensary bans and student perceptions of risk was positive but not statistically significant , refuting my hypotheses that dispensary bans would have a significant positive association with perceived risk among the students who attend school there. Perceived risk was then regressed on the likelihood of lifetime marijuana use, which revealed a statistically significant negative association between perceived risk and lifetime marijuana use , confirming the hypotheses that perceived risk is a determinant of student marijuana use behavior . Finally, the regression analysis of dispensary and lifetime marijuana use was repeated including perceived risk in the model. Adding perceived risk to the model changed the relationship between dispensary bans and lifetime marijuana use very little. Although including perceived risk in the regression of dispensary bans on lifetime marijuana changed the coefficient only slightly, adding it to the model strengthened the relationship, indicating a small and not statistically significant dependent relationship between dispensary bans and lifetime use, and at indirect mediation . These steps were repeated for the recent marijuana use outcome, with similar results, reported below. Table 7.8 presents the results of the perceived risk mediation analysis for lifetime and recent marijuana use. The relationship between dispensary bans and recent marijuana use was negative and non-significant . Whether a city had a dispensary ban was then regressed on whether students perceived great risk from frequent marijuana use using a HGLM model controlling for the covariates and with a random intercept for city. There association between dispensary bans and student perceptions of risk was positive but not statistically significant , refuting my hypotheses that dispensary bans would have a significant positive association with perceived risk among the students who attend school there. Perceived risk was then regressed on the likelihood of recent marijuana use, which revealed a statistically significant negative association between perceived risk and recent marijuana use , confirming the hypotheses that perceived risk is a determinant of student marijuana use behavior . Finally, the regression analysis of dispensary and recent marijuana use was repeated including perceived risk in the model. Adding perceived risk to the model changed the relationship between dispensary bans and recent marijuana use very little. Although including perceived risk in the regression of dispensary bans on recent marijuana changed the coefficient only slightly, adding it to the model strengthened the relationship, indicating a small and not statistically significant dependent relationship between dispensary bans and recent use, i.e., a non-significant degree of indirect mediation .