The purpose of the model is to provide theoretical predictions of how supplier responses to the changes in federal enforcement should differ depending on state MML regulations. I can then link these predictions to newly collected empirical evidence on the size of the legal market. The model also provides intuition for the regulatory variables used in the empirical analysis and clarifies the key assumptions upon which my approach relies. As the illegal market for marijuana is not directly observed, modeling the interaction between the legal and illegal markets is helpful in motivating the appropriate empirical framework to estimate how legalizing medical marijuana affects recreational consumption. The model describes the market for marijuana in an MML state as composed of consumers and suppliers who respond to changes in the probability of being prosecuted by state or federal enforcement. Suppliers operating in the legal medical marijuana market also face capacity constraints that vary by how strictly the state regulates producers. Each state’s market is assumed to operate in isolation.7 Aggregate demand D = βd is assumed to be a downward sloping function of potency-adjusted price p. Enforcement by the state and federal government against marijuana users serves as an aggregate demand shifter, with ∂β/∂es ≤ 0, ∂β/∂ef ≤ 0. Users’ expected probability of arrest, penalties, and distaste for illegal activity will all be captured through β. It is worth noting that this setup does not rule out various forms of addiction or habit formation. Both rational and irrational models of consumer behavior imply price and risk responsiveness, but the models differ in the degree of responsiveness to these costs. With enactment of an MML, users who register as medical marijuana patients face new perceived risks from state and federal enforcement. The perceived federal risk of enforcement facing a registered user could exceed that facing an illegal user if, for example,grow racks individuals believed the federal government would use state medical marijuana registries to target individuals for prosecution.
As users can always face the same risks as during the pre-MML period by choosing not to register, I assume aggregate demand in the MML period is no less than demand under prohibition. The magnitude by which demand shifts will depend on the enforcement risk facing a registered compared to an unregistered user, and on how users weight the relative risks of state and federal enforcement. Most MML states have mandatory medical marijuana registration programs: laws requiring medical marijuana users to register in order to receive protection from state arrest.An individual who wants medical marijuana must first obtain a physician’s certification that the individual has a medical condition which could benefit from the use of marijuana. The patient then must submit an application to the state authority, along with a registration fee. If the application is approved, the patient receives documentation providing access to dispensaries and protection from state prosecution. By definition, all other consumption is illegal. The number of registered medical marijuana patients provides an observable measure of the size of the legal market. A limitation faced by previous research has been that state records of registered patient counts are not readily available and are not maintained similarly across states. To overcome this limitation, I collected monthly data from a number of sources, including contact with state officials, state department websites, news articles, and academic papers. Since my outcome variables are annual state-level prevalence measures, I linearly interpolate missing end-of-year registration rates using the nearest available months of registered patient counts. The final measure of market size is the annual registration rate, calculated as the percent of adults registered as medical marijuana patients at the end of December in a given year. I include the voluntary registration data available from California after 2005, but due to lack of data, Maine and Washington are excluded.
From this data, the medical marijuana market in 2013 is estimated to consist of about 1,139,098 legal medical marijuana patients,and industry reports estimate annual retail sales of legal marijuana in 2013 at $1.43 billion .While the legal market size is dwarfed by the estimated $25-$40 billion size of the illicit market , growth in the legal marijuana market over the past decade is twice that of the illicit market. The estimate of the number of registered patients in 2013 represents a more than 300% increase in the number of medical marijuana users from 2007 compared to only a 150% increase in the number of total adult past-month users . The legal market for marijuana is thus rapidly expanding. To study how changes in the legal market affect the illegal market, it is important to first understand the factors driving legal market growth. To isolate the supply-side drivers of growth in legal medical marijuana markets, the ideal policy variation would be such that production costs were exogenously shifted in some MML states but unchanged in others. Based on the model outlined in Section 2.2.2, the Ogden and Cole Memos approximated this ideal by shifting production costs differentially more in MML states with lax supply restrictions compared to those states with strict supply restrictions. Figure 2.1 connects the model’s propositions to the data by documenting registration rate trend breaks at the Ogden and Cole Memos for a sub-sample of states that exemplify the different state production restrictions. Hawaii and New Mexico serve as examples of more restrictive production states; Colorado and Montana are representative of MML states with looser supply restrictions. For all four states, few patients register during the first few years following MML passage. The initial slow growth in medical marijuana take-up suggests that, conditional on federal enforcement remaining high, reduced state risk following MML enactment had little effect on the size of the legal market.
The empirical evidence from Figure 2.1 is also consistent with the model’s prediction that the Ogden and Cole Memos had far greater effect in states with loose production limits. In Hawaii, where suppliers had strict limits and could only serve a single patient, the Ogden Memo had very little effect on registration rate trends. Correspondingly, there is also little break in trend following the Cole Memo. In New Mexico, which allowed state licensed dispensaries that had higher production limits but faced heavy regulation, the Ogden and Cole Memos also seem to have had small effects. In contrast, the states with lax production limits saw dramatic changes in registration rates with the Ogden and Cole Memos.16 After the Ogden Memo, Colorado experienced the “Green Rush,” a proliferation of dispensaries that arose following decreased fears of federal intervention.17 Alongside the expansion of unregulated dispensaries in Colorado, patient registration rates spiked . Similarly, in Montana, where caregivers were permitted to produce for an unlimited number of patients and receive compensation for their services, the number of caregivers providing for 20 or more patients increased seven-fold within one year following the Ogden Memo. This was accompanied by a nearly six-fold increase in registered patients . In line with the propositions from the model, the Cole Memo reversed this trend.The decline in registration rates in Montana is particularly dramatic because, concurrent with the Cole Memo, Montana’s legislature passed Senate Bill 423 which effectively dismantled the medical marijuana supply industry by establishing caregiver patient limits and preventing caregivers from receiving compensation.To examine this relationship for all states, I exploit the model’s predicted trend reversal between the Ogden period and the Cole period in the relationship between a state’s laxness of medical marijuana supply regulation and registration rates. Intuitively, in states with loose medical marijuana production limits, the Ogden Memo decreased marginal costs over a broader range of production , inducing producer entry and supply increases by existing producers. Column reports regression results that allow federal enforcement changes to also affect MML states with strict restrictions on supply. Consistent with the model predictions,grow table both federal policies had far larger effects on registration rate trends in states where marijuana suppliers were relatively unrestricted. Between the Ogden and Cole Memos, states with strict production limits on average saw an additional 0.2% of the adult population register as medical marijuana patients , which is statistically significant but an order of magnitude smaller than the increases seen in MML states with loose supply regulations . There is no effect of the Cole Memo in strictly regulated MML states, consistent with evidence that federal enforcement related to the Cole Memo targeted large-scale production. Limiting the sample to only those states with MMLs by 2012 in columns and yields coefficients of similar magnitude but with slightly larger standard errors. Unlike initial MML enactment, the federal memos substantially altered the size of the legal market, with larger effects in states with looser supply regulations. The differential response of registration rates to the federal government’s policies in states with lax compared to strict producer restrictions suggests that patient registration rates are driven primarily by supply-side shifters. Therefore, to estimate the causal effect of changes in medical marijuana supply on recreational consumption, my empirical strategy uses the timing of the federal memos and differences in initial MML supply restrictions as instruments.
The main threat to identification is that registration rates and illegal use may be jointly determined by unobservables affecting demand. Time-varying state covariates included that potentially affect recreational marijuana use can be categorized as: demographics influencing recreational marijuana use, economic characteristics, and substance-related policies influencing marijuana consumption. A full listing of covariates is provided in Table 2.3. For all specifications, to account for heteroskedasticity and serial correlation, robust standard errors are clustered at the state level . To account for potential violation of the parallel trends assumption, specifications including state-specific linear trends are also presented as robustness checks in section 2.6. Even after controlling for state and year fixed effects and state-year covariates, identification of β is challenging due to concerns of endogeneity between recreational marijuana use and registration rates. While fixed effects will account for issues of cross sectional endogeneity, β could still be biased due to some omitted state-level time-varying variable that affects both registered users and illicit users. For instance, β will be biased upward if changes in local perceptions regarding the health risks of marijuana use led to changes in both medical and recreational use. To account for potential endogeneity, I instrument for registration rates by way of two-stage-least squares using equation 2.6 as the first-stage specification. The instrumental variable estimates are valid as long as the exclusion restriction is satisfied. In the context of equations 2.6 and 2.7, this occurs if E[εjtZjt|uj , vt , Xjt] = 0, where Zjt is the vector of instruments including the interaction of MML supply restrictions with trend breaks based on the exogenous timing of the Ogden and Cole Memos. As state and year fixed effects are included, the identification is not threatened by level differences between states or by national trends in marijuana consumption . However, the exclusion restriction will be violated if changes in federal enforcement following the Ogden and Cole Memos had differential effects on demand in states with loose compared to strict production restrictions through any channel other than supply. Evidence validating the exclusion restriction is presented in section 2.6. The measures of marijuana consumption come from the National Survey of Drug Use and Health . The NSDUH is an annual survey funded by the Substance Abuse and Mental Health Services Administration of the US population over twelve years of age. The public-use NSDUH provides estimates of the prevalence of past-month use and past-year initiation of marijuana, available separately for age groups of 12-17, 18-25, and 26 years of age and older. Representative state-level data broken down by two-year averages is available from -. Table 2.3 provides summary statistics for the marijuana use measures and state level covariates used in the NSDUH analysis. Comparing mean differences, MML states with a positive number of registered medical marijuana patients have higher levels of past-month cannabis users for all age groups than states without MMLs. However, MML states on average are more likely to have decriminalized marijuana,have higher cigarette taxes, and consist of populations that are on average younger and more male. As such characteristics are potentially correlated with prevalence of recreational cannabis use, all regressions control for the state covariates listed in Table 2.3.