There are some limitations to this dataset. Firstly, the public-use NSDUH data is only available in two-year averages; thus the registration rate data may not correspond exactly with the marijuana outcome data, leading to a loss of precision for these estimates.Secondly, the NSDUH measures of substance use are self-reported. If individuals are more likely to report marijuana consumption truthfully based on legality, analysis of the effects of legal market size compared to MML enactment should be less subject to reporting bias. However, if reporting is driven by perceived changes in social approval, growth in the legal market size may induce changes in reporting behavior. Section 2.6 and Appendix D- provide supporting evidence that the self-reported marijuana use measure captures true changes in consumption. Despite these problems, the NSDUH is the only publicly available dataset that provides representative state-level estimates of marijuana consumption for all individuals aged 12 and older. Additionally, the NSDUH Restricted-use Data Analysis System provides state-level estimates for perceptions about marijuana,grow rack which are used in section 2.6 to examine the mechanisms by which changes in the legal market affect adolescent marijuana use.Figure 2.2 presents descriptive evidence of changes in marijuana consumption over time. Trends in past-month use and past-year initiation are plotted by age groups and by strictness of state MML supply regulation.
All trends are normalized such that the prevalence measure is zero for 2007-2008, the years just prior to the Ogden Memo. For all age groups, marijuana consumption increases in MML states after the Ogden Memo. For youths and young adults, these increases are largest in MML states with loose supply regulations, while loose and strict MML states saw similar growth rates for older adults. For adolescents aged 12-17, there is a sharp drop in consumption in MML states with loose regulations following the Cole Memo; for adults, consumption does not decrease after the Cole Memo but trends in use flatten. To examine whether these patterns in marijuana consumption are driven by growth in medical marijuana availability, Table 2.4 reports the first-stage and reduced form estimates of the differential effects of the Ogden and Cole Memos in MML states with lax supply restrictions on registration rates and the prevalence of marijuana consumption. The first-stage results of Table 2.4 are replicated from Column of Table 2.2 and are discussed in section 2.2.4. The F-test for excluded instruments takes a value of 52.3, mitigating any concerns of finite sample bias due to weak instruments . The reduced form results for past-month use and past-year initiation by age group have the expected signs, though the instruments have less predictive power for past-year marijuana initiation. Since this approach identifies effects off of changes in a monthly time series, ex-ante the results for past-month use are expected to perform better than the results for past-year initiation. Table 2.5 reports the estimated effects of growth in the legal medical marijuana market on marijuana consumption, showing results separately by age group for preva- lence of past-month marijuana use and past-year initiation.
To test whether these effects are driven by omitted variables influencing both legal and illegal use, the second-stage results of the instrumental variables analysis are compared to the OLS estimates. The results indicate that growth in medical marijuana registration rates significantly increases marijuana consumption for all age groups, with the largest effects for older adults. An additional one percentage point of the adult population registering as medical marijuana patients predicts a significant 6% increase in the share of 12-17 year olds reporting past-month marijuana use, a 7-8% increase for 18-25 year olds, and a 20% increase for older adults. The effects on past-year initiation are similar but smaller in magnitude, indicating a 1-5% increase in the share of 12-17 year-olds reporting past-year initiation, a 6% increase for 18-25 year olds, and a 12-18% increase for adults over age 25. Additionally, these estimates support that using registration rates as an exogenous measure of the size of the legal marijuana market does not produce biased estimates. The IV estimates are neither qualitatively nor statistically different than those from OLS, suggesting that changes in registration rates and recreational use are not being driven by some unobserved factor correlated with both recreational and medical demand. This conclusion will hold if the instruments are indeed excludable, supporting evidence of which is presented next. The main threat to the validity of the instrumental variables approach is a violation of the exclusion restriction.
The exclusion restriction will be violated if changes in federal enforcement due to the Ogden and Cole Memos had differential effects on demand in states with lax compared to strict production restrictions through any channel other than supply. For example, the exclusion restriction would be violated if, when federal enforcement was removed and registration rates increased, local governments in strictly regulated states devoted more resources toward prosecuting users than state governments in states with loose regulations. I provide evidence that this did not occur in Figure 2.3, which plots changes in per user state marijuana possession arrest rates against changes in registration rates. Results are shown separately for juveniles and adults. There is no apparent correlation between legal market growth and state enforcement for either strict or loose regulatory regimes. Another potential violation of the exclusion restriction would occur if, prior to the Ogden Memo, the expected user risks of federal prosecution were higher in states with loose compared to strict supply regulations. To address this, I first show that the actual risk of a marijuana user facing federal enforcement is very low. The federal government has never devoted substantial resources toward prosecuting individuals for simple marijuana possession. Figure 2.4 shows from fiscal year 1996 through 2012, less than 4% of federal prosecutions for marijuana-related offenses have been for simple possession.Between 1996 and 2012, the annual average of federal prosecutions for marijuana possession was 224 compared to 6,259 for sales. Focusing on enforcement against medical marijuana specifically, reports suggest there were fewer than 20 federal prosecutions of medical marijuana users or growers from 1996-2005 and about the same number from 2005-2009; this enforcement was almost exclusively directed toward large-scale producers in California. While the actual probability of federal prosecution was low, uncertainty among potential medical marijuana users about whether the federal government could access registry data may still have stifled demand. This deterrent was likely far greater in California than in other states,hydroponic rack as federal prosecution against suppliers was concentrated in California. Vickovic ’s analysis of news article mentions of medical marijuana showed that almost all articles about federal enforcement prior to the Ogden Memo were about dispensary raids in California, and the majority of these articles were published by local sources. Thus, it may be the case that the Ogden Memo led to a relatively larger shift in demand in California compared to other MML states.To address these concerns, Table 2.6 reports the results of the instrumental variable specification excluding California. The results for past-month use are unchanged, but the effect of legal market growth on past-year initiation for adolescents becomes small and insignificant and shows evidence of endogeneity. These results are consistent with past-month use serving as a better measure to detect short-run changes in marijuana access, while past-year initiation may better reflect longer-run changes in social approval or risk perceptions. There may be a spike in demand from MML enactment due to perceived governmental acceptance of the substance followed by a linear effect of registration rates reflecting changes in availability. To assess this relationship, Table 2.7 compares the effects of medical marijuana market growth on past-month marijuana use to the effects of MML enactment.In the regressions without state-specific trends, both registration rates and the categorical variable for MML enactment have positive effects on marijuana consumption, though the effects of the categorical MML variable are only significant for adults over age 25. However, after including state-specific linear trends to account for state differences in preexisting paths of marijuana use, the estimates for MML passage become small and insignificant and switch sign. In contrast, estimates of the effects of registration rates remain positive and significant for all age groups. The registration rate estimates are less sensitive to trend inclusion than those of the categorical variable.
This could be due to the non-monotonicity of trends in registered patient counts or the categorical MML measure confounding preexisting trends with the dynamic effects of the policy . Additional robustness checks of sample selection and model specification in Appendix D- support that the binary MML measure misses heterogeneous effects across states and the dynamics of these policies.There are a number of mechanisms through which MML policy, promoted as legislation to protect cannabis use for a relatively small number of individuals with chronic health conditions, might affect marijuana consumption by adolescents. If registered adult users are more visible to adolescents, youths may increase consumption due to lower perceived risks associated with social disapproval, formal sanctions, or health consequences. If growth in the legal market increases total availability, increased adolescent use may be driven by lower search costs, increased product variety, or declines in the quality-adjusted price of marijuana following the shift in supply . Sections 2.2 and 2.5 suggest that supply channels are particularly important in determining MMLs’ effects on adolescent use. Spillover of medical marijuana supply to the illegal market serving adolescents could occur through resale or sharing of medical marijuana by legal users.Recent national surveys of high-school seniors show that almost 18% of past-year users report receiving some marijuana from another’s medical marijuana “prescription” ; focusing only on high-school seniors in states with MMLs, 34% of past-year users report one of their sources as another person’s medical marijuana “prescription” . This indicates substantial diversion from the legal market supplying medical marijuana patients to the illegal market supplying youths. Table 2.8 provides evidence of the mechanisms by which growth in the legal market influences cannabis consumption by youths aged 12-17 using state-level data from the NSDUH R-DAS. Columns – report effects on risk perceptions, and columns – report effects on measures of adolescent access. The outcome variable for each column is the share of youths aged 12-17 who report that: using marijuana monthly poses a “great risk,” their friends “somewhat” or “strongly” disapprove of monthly marijuana use, the maximum penalty for possession of one ounce of marijuana in their state is prison time, marijuana is “somewhat” or “very” easy to obtain, most students their age use marijuana, and they purchased marijuana in the past year. Since these measures are likely determined endogenously, these results are intended only to provide suggestive evidence. While growth in the legal market affects adolescent perceptions of both risk and availability, the results of Table 2.8 suggest that the legal market primarily affects adolescent use through changing access. A one percentage-point rise in registration rates significantly increases the share of 12-17 year-olds who believe marijuana is easy to obtain , who believe most students their age use marijuana , and who report buying marijuana in the past year ; these estimates are jointly significant with a p-value of 0.04. In contrast, estimates of the effects on risk perceptions are smaller and jointly insignificant with a p-value of 0.13. These results thus provide supporting evidence that the increases in adolescent marijuana use following medical marijuana market growth are due to supply spillovers rather than decreased risks of legal penalties or social disapproval. According to Department of Justice National Drug Threat Assessment reports, there is substantial internal movement of domestically produced marijuana, and Figure E.1 clearly illustrates that the sources of much of the marijuana trafficked domestically are the western MML states of Washington, Oregon, California, and Arizona. As marijuana markets are not isolated , supply shocks in legal medical marijuana state markets likely have spillover effects on price and availability in states without MMLs or in MML states with high production costs. For the empirical results, this should bias the estimates of the effects of legal market size toward zero. Tables E.1 and E.2 use Montana as a case study to assess whether spillover of marijuana supply from states with large legal markets to other states is biasing the estimates from the primary results downward. Presently, Montana and Michigan are the only MML states in the continental United States that do not share a border with another MML state.28 In addition, Figure E.1 shows that Montana and its bordering states are relatively unimportant trafficking routes for marijuana; cannabis smuggled through these states is primarily sourced from Washington and Oregon and is destined for larger, more urban centers in the midwest. Therefore, growth in the legal medical marijuana market in Montana is unlikely to have significant supply spillovers to its neighboring states.