Cannabis Grow

In contrast, all four states relied on industry participation in the development of rules and regulations because of the lack in regulatory models for the novel industry,which led to inadequate adoption of policies to protect public health. For example, in Colorado, industry members of the 2014 working group on pesticide regulation delayed action on a 2013 Colorado Department of Agriculture draft list of allowable pesticides that would have required growers use only nontoxic forms, arguing that the proposed list was too restrictive,resulting in regulatory paralysis. Between 2014 and 2016, pesticide use in Colorado was unregulated during which time producers were reportedly using inappropriate or unsafe chemicals,including Eagle 20, a fungicide used to kill mites, mildew and assorted pests.Eagle 20, which was not among the list of approved pesticides as of 2016, contains myclobutanil, which when burned produces the poisonous gas hydrogen cyanide.In 2016, marijuana industry participation on the rule making board in Alaska led to the adoption of marijuana smoking clubs, despite the risks of exposure to secondhand marijuana smoke on cardiovascular function.A single market, in which all sales are regulated as retail — without a separate medical market — simplifies regulatory efforts, including licensing enforcement, implementation of underage access laws, prevention education programs, and taxation.The existence of a licensing system for medical marijuana in Colorado and Oregon before retail legalization led to regulators developing dual licensing systems for medical and retail; in Colorado marijuana businesses could apply for both.

As of October 2016, Alaska had not developed regulations for its medical marijuana licensing system through for profit dispensaries,rolling grow table though medical marijuana has been legal since 1998. All four US states that had legalized recreational use as of October 2016 maintained dual markets with medical marijuana subject to different rules than retail, including higher possession and cultivation limits, and lower age limits . In 2015, Washington legislators eliminated separate medical marijuana dispensaries. Retail marijuana stores that applied for and received a medical marijuana endorsement from the Department of Health could legally sell both medical and retail marijuana products. Otherwise, the same regulatory inconsistencies still existed in all four states. Medical marijuana is exempt from state and local taxes, which created price differences between the two markets, and likely has contributed to the continued growth of the medical marijuana market.The experience from tobacco is that regulatory complexity ultimately favors corporate interests with the financial resources to manipulate and weaken public health policies to reduce use. The tobacco companies use their extensive legal resources and more detailed information about market structure to take advantage of complexity in regulations to make enforcement more difficult for government regulators, which general move slowly and are subject to political constraints. It is unknown what effect these complex legal environments that often favor large corporate interests, and that complicate implementation of effective public health policies, will have on marijuana prevalence at the population level.While the Netherlands de facto legalized possession and use of marijuana in 1976,in 2014 Uruguay became the first country to legalize the cultivation, processing, distribution, and supply of marijuana for recreational purposes.

Uruguay’s law mandates that the government control the manufacture, distribution, and sale of cannabis under the authority of the Regulation and Cannabis Control Institute a new agency created to oversee cultivation licensees and pharmacies, cannabis clubs, and at-home cultivation.The agency would produce generic, unbranded cannabis, eliminating the incentive to market and advertise competitive products.The state would use its licensing power to grant licenses to qualified professional farmers and limit the number of licenses, depending on demand, to avoid an illegal market. As of November 2016 Uruguay was still developing a government monopoly over marijuana production and distribution system.It had implemented regulations for personal cultivation and the operation of cannabis co-operatives, where Uruguayans pay membership fees to be part of collectives that grow marijuana,but was still in the process of drafting the rules for cultivation licenses for private companies and distribution through pharmacies.Because the four US states elected to adopt commercially-focused marijuana regulatory schemes, it is unlikely that these state governments will legally be able to prohibit all forms of marketing and advertising, particularly if marijuana is legalized at the federal level. While it is theoretically possible for government to limit advertising and promotion in the United States, it is extremely difficult to craft such restrictions in light of how the courts have interpreted the US First Amendment protections of free speech. Other governments have successfully restricted tobacco advertising much more extensively than was possible in the United States. Uruguay in 2013 prohibited all forms of marijuana marketing, advertising, and promotions, modeled on its provisions for tobacco products.As discussed above, the alcohol industry has voluntarily committed not to advertise in mass media outlets where more than 30% of the audience is “reasonably” expected to be under age 21.Also as discussed above, the 30% threshold is high enough to allow the alcohol industry to reach youth with their marketing.In Colorado, event sponsorship, including sporting events and concerts, were permitted as long as less than 30% of the audience is under age twenty-one, whereas in Washington these events were subject to the same location restrictions as traditional mass media advertising . None of the states prohibit online advertising or the use of social media. Colorado prohibits the use of unsolicited pop-up advertisements on the Internet; otherwise in Colorado and Oregon internet advertising is subject to the same 30% threshold for underage exposure. Neither state has specific rules on how the age restriction on internet advertising will be defined or enforced. Washington included an unenforceable guideline suggesting that businesses “use social media with caution and be mindful not to appeal to, or solicit, viewers under the age of 21. If possible, please restrict views to adults age 21 and older.”Indeed, in 2015, within two years of implementation of legalization laws in Colorado and Washington, the marijuana industry was already taking advantage of the weak advertising restrictions. Online advertising is widely used by the marijuana industry. Eighty-five percent in Colorado and 65% in Washington of marijuana companies advertise online through company websites. A little less than half of marijuana companies that had operational websites used age verification systems in Colorado or Washington . Among those with age verification systems,indoor plant table more than half in Colorado and in Washington require viewers hit “yes” to gain access to the website, while only 5% in both states require information on the viewer’s birth date.Advertising restrictions could be designed to protect consumers and vulnerable populations. However, state laws in Colorado and Washington were unable to prevent marijuana companies from using false or misleading health claims to advertise their products online. Among websites of marijuana companies, 61% in Colorado and 44% in Washington made health claims about their products.The most common health claim was on treatment for anxiety and depression , insomnia , and pain management ,even though the scientific literature is either mixed or has low evidence of these therapeutic effects.

None of the states prohibit marijuana companies from using directories or store locator websites which often are de facto advertising.Seventy-five percent in Colorado and 56% in Washington of marijuana companies were listed on WeedMaps. In addition to not requiring age-verification to create an online account, WeedMaps allows marijuana companies to circumvent advertising restrictions by listing product descriptions, prices, price promotions and coupons, and post images of their products without warning statements . In addition, WeedMaps provides a platform for publication of online testimonials in which users make health claims on the therapeutic benefits of marijuana use. Online testimonials undermine enforcement of truth-in-advertising laws that prohibit marijuana companies from making false or misleading claims on their products. The same issues that make age-verification systems for tobacco advertising ineffective have already occurred with online marijuana advertising.Uruguay acted to protect public health by prohibiting marketing and advertising entirely, which would likely prevent the issue of directories and store locator websites.Labels provide information to the consumer on its content, including product potency and serving size. As such, it is important that marijuana labels are accurate so as to avoid marijuana intoxication and accidental use. Poor production and premarketing testing procedures to accurately measure THC concentration contained in a marijuana product had led to inconsistent concentration levels in marijuana edibles.A 2015 study of the accuracy of labels in San Francisco and Los Angeles, California. and Seattle, Washington found that marijuana products were unlikely to be accurately labeled in terms of THC content.While 17% of the sample was accurately labeled, 23% reported THC levels 10% higher than indicated on the label, and 45% reported THC levels 10% below its labeling content.Although not yet in place as of November 2016, Uruguayan authorities had indicated that the IRCCA will develop requirements for generic, non-appealing packaging. None of the four US states require plain packaging, although under the Oregon Liquor Control Commission’s rules, marijuana companies that use generic labels without graphics, pictures, or logos are not required to submit their packages to the OLCC for pre-approval.Colorado prohibits the words “candy” or “candies” on marijuana packaging and Oregon prohibits product packaging that contains “cartoons, including use of comically exaggerated features, attribution of human characteristics to animals, plants or other objects, or the similar use of anthropomorphic technique, or attribution of unnatural or extra-human abilities, such as imperviousness to pain or injury, X-ray vision, tunneling at very high speeds or transformation.”While the four US states prohibit false or misleading health claims on marijuana labeling,6, 364 the regulations do not specify what would invalidate such claim or how the liquor control boards in charge of overseeing packaging regulations would enforce these laws. State laws do not prohibit the use of “natural,” “pure,” “clean,” “additive-free,” “fair trade,” “omega 3, 6, and 9,” or any other descriptor that would increase product appeal or reduce risk perceptions on packaging, labeling, or advertising, which permit marijuana companies to use package design and ingredient lists to circumvent restrictions on health claims. For example, in 2016 the Colorado marijuana company Dixie Elixirs sold an orange zest flavored product labeled as “awakening” and a peppermint flavored product labeled as “relaxing,” and included an ingredient list with supplements including “Siberian ginseng” and “ashwagandha,” an herb promoted as reducing stress and promoting well being, despite the fact that no clinical trials have verified these claims.Another Colorado company that produced Ebbu Raw, used its labeling and package design to “[build] trust with customers.Evergreen Herbal’s 4.20 Bars were labeled with descriptors “fair trade,” “With Omega 3, 6, 9,” and “cacao,” which may signal users that these products are environmentally safe or may produce health benefits. Washington packaging requirements allowed marijuana companies to design marijuana packages with brand names such as Mirth “Relax. It’s Legal” in Rainier Cherry Soda flavor,or Evergreen Herbal’s 4.20 Bars in milk or dark chocolate, and flavored with sea salt, toffee, hazelnut or hemp crunch. One way to prevent marijuana companies from taking advantage of weak language for restrictions on health-related messaging would be to require that all advertising and marketing statements and claims be evidence-based and approved by the health authority, including claims about the product improving sex, energy, sleep, weight reduction, vitamin supplements, among other health-related claims that would increase product appeal. Edibles that lack accurate product labeling pose a serious public health risk to adult consumers as well as children.It appears that the regulations on potency limits, labelling and standardization of dose, and packaging in Colorado and Washington were not strong enough to prevent cannabis-related harm. Adults that have used highly potent products have been increasingly reporting unpleasant psychological experiences such as psychosis, anxiety ,and depression symptoms. There is cause for concern that marijuana edibles in the US states have high THC content, which may responsible for many of these observed effects. In 2015, product regulation laws in Colorado were updated to require clear demarcation or individually wrapped servings .There were no changes to the THC limits per serving size or per package, which could have helped reduce marijuana intoxication in both children and adults. After reviewing the evidence from Colorado and Washington that edibles were causing harm in children and inexperienced users, Oregon reduced its maximum THC limit to 5 mg per serving, and 50 mg per package.This experience illustrates the importance of health authorities having the power to adjust maximum serving size and related packaging as scientific evidence on the harms associated with different doses accumulates.

Based on a procedure reported by Crombie and coworkers, a solution of olivetol and food-grade α-phellandrene in benzene was treated with p-toluenesulfonic acid monohydrate and the mixture was allowed to stir at room temperature for 1 h. Te solvent was removed in vacuo and the residue was purified by silica gel chromatography using a gradient elution to give H2CBD as a dark yellow oil. Spectroscopic data were in full agreement with the literature.Animals were randomly divided into 5 groups of 12 animals per group and received either vehicle , 2:1:17, a positive control or 8,9-dihydrocannabidiol via intraperitoneal injection 1h prior to administration of the convulsant agent pentylenetetrazole to achieve brain cannabinoid Tmax. PTZ was administered intraperitoneally 1h afer drug or vehicle treatment. Seizure activity was video recorded for 30min and video records blinded before ofine review and coding using a modified Racine scale .4,4-Dichlorodiphenyltrichloroethane was used as the internal analytical standard . HPLC grade n-hexane, acetonitrile, water and ascorbic acid were purchased from Sigma Aldrich UK and Fisher Scientifc. Stock standard solutions of CBD, H2CBD and DDT were prepared in acetonitrile and stored at −20 °C until use. These were further diluted in acetonitrile:water , to achieve calibration concentrations of 0.1, 0.2, 0.5, 1, 5, 10 μg mL−1 . Plasma samples were prepared for HPLC using a previously validated method. Briefly, DDT was added to 150 μL of rat plasma sample as internal standard and plasma proteins were precipitated by the addition of ice cold acetonitrile followed by water ,indoor grow shelves with 1min vortexing between additions. n-Hexane was added to each tube and following a 5min vortex, tubes were centrifuged at 1160 × g for 15min at 10 °C and the upper organic layer was carefully decanted by glass pipette and retained.

The organic layer was evaporated to dryness under a stream of nitrogen at room temperature and reconstituted in 150μL of the mixture of acetonitrile and water prior to HPLC analysis. For post-mortem brain analysis, brains were weighed and 1.5 x ice-cold solvent  was added followed by homogenization for 1 min. DDT was added to each homogenized brain tissue as internal standard, samples were mixed and allowed to equilibrate overnight at −20 °C. Samples were then centrifuged at 3500 rpm for 15min and the top layer retained. Samples were dried by SpeedVac concentrator at room temperature and reconstituted in 150μL of the mixture of acetonitrile and water for HPLC analysis. An Agilent 1200 series HPLC equipped with a photodiode array detector was used for analysis. 30 μl of all samples were injected and separation was achieved using an ACE C18-PFP 150 mm 4.6 mm, 3 μm particle size column , protected by an ACE C18-PFP 3 μm guard cartridge. The mobile phase was a mixture of acetonitrile and water in a ratio of 62:38 . The flow rate was set at 1 mL min−1 and the column temperature was maintained at 55 °C. The absorbance of the compounds of interest was monitored at 220nm.Youth are regularly exposed to protobacco messaging through a wide variety of media channels, including static tobacco advertising on newspapers and magazines, retail outlets, the Internet,and on television or in the movies.Marketing activities of tobacco industry are a key factor in leading young people to take up tobacco, keeping some users from quitting, and achieving greater consumption among users.The 201257 and 201421 US Surgeon General reports concluded that tobacco industry promotional activities, including branding, imagery, event sponsorship, and marketing campaigns, cause the onset and progression to smoking among young people.

NCI’s smoking and health monograph, The Role of the Media in Promoting and Reducing Tobacco Use, had earlier found a causal relationship between tobacco marketing exposure and youth smoking. Even minimal exposure to tobacco advertising positively influenced youth attitudes and perceptions on smoking, as well as smoking intentions among youth.Causal effects of tobacco marketing on smoking may be stronger among youth than adults as youth are also more likely to be brand loyal. and are more susceptible to tobacco industry marketing.Youth susceptibility to smoking, experimentation, and current use varies by the Tobacco advertising influences youth smoking behavior at multiple levels.Tobacco advertising and promotion affect awareness of smoking, recognition of specific brands, attitudes about smoking, intentions to smoke, and actual smoking behavior among youth and contribute to reduced risk perceptions around tobacco use.Even with prohibitions on youth-targeted marketing, tobacco industry marketing directed at young adults, encourages use and increased consumption within the young adult population,and indirectly impacts youth smoking because youth consider young adults as aspirational role models.The WHO Framework Convention on Tobacco Control recognizes that the most effective strategy to protect public health would be to prohibit tobacco marketing entirely.Tobacco companies use advertising as a marketing technique to create positive imagery and associations with tobacco products, and to attach desirable characteristics, activities, and outcomes with tobacco product use.Branded merchandise helps to establish brand identity and brand loyalty among novice users, which is an integral part of the tobacco industry’s long term economic strategy. Indeed, the tobacco industry specifically targets young adults in clubs using branded promotions and merchandise.For youth, there is evidence that owning cigarette-branded or alcohol-branded items leads to progression to being an established smoker and initiation of drinking.

Among adults, young adults are significantly more likely that older adults to own cigarette-branded items and to be attracted to the advertising of a cigarette brand .In short, promoting products through branded merchandise is a particularly important strategy for companies and they seem to be heavily targeting youth and young adults, who appear to be more susceptible to it than older adults and are the demographic that is susceptible to initiation or escalation of product use. Brand sharing and brand stretching grant another access point for tobacco companies to subliminally advertise and market their products.In addition to using the cigarette package, tobacco companies place brand names and use other design techniques on the actual stick, which is rated by smokers as more attractive than cigarettes without these characteristics.Despite some restrictions in the USA, tobacco companies continue to advertise in magazines with significant youth audiences, and are more likely to advertise youth preferred brands in these magazines.Tobacco companies circumvent partial advertising restrictions by concentrating advertisements in magazines where youth audience composition is near or at the minimum threshold level, thereby still exposing a sizeable number of youth to tobacco ads. For example, in the United States even a 15% threshold, which was the FDA-proposed rule in 1996 for advertising in print media, would have exposed at least two million youth to tobacco industry advertising.Cigarette companies consolidate marketing expenditures for magazine advertisements to brands that are popular among youth, African Americans, and LGBTQ populations .In the 1990s most of the US state attorneys general sued the major cigarette companies alleging, among other things, that the source of pro-tobacco media. Current tobacco use is associated with exposure to static advertising and to on-screen smoking depicted in TV and in movies, both directly and through perception of peer use among youth and young adults.Tobacco advertising influences youth smoking behavior at multiple levels.Tobacco advertising and promotion affect awareness of smoking, recognition of specific brands,grow table attitudes about smoking, intentions to smoke, and actual smoking behavior among youth and contribute to reduced risk perceptions around tobacco use.Even with prohibitions on youth-targeted marketing, tobacco industry marketing directed at young adults, encourages use and increased consumption within the young adult population,and indirectly impacts youth smoking because youth consider young adults as aspirational role models.The WHO Framework Convention on Tobacco Control recognizes that the most effective strategy to protect public health would be to prohibit tobacco marketing entirely.Tobacco companies use advertising as a marketing technique to create positive imagery and associations with tobacco products, and to attach desirable characteristics, activities, and outcomes with tobacco product use.

Branded merchandise helps to establish brand identity and brand loyalty among novice users, which is an integral part of the tobacco industry’s long term economic strategy. Indeed, the tobacco industry specifically targets young adults in clubs using branded promotions and merchandise.For youth, there is evidence that owning cigarette-branded or alcohol-branded items leads to progression to being an established smoker and initiation of drinking. Among adults, young adults are significantly more likely that older adults to own cigarette-branded items and to be attracted to the advertising of a cigarette brand . In short, promoting products through branded merchandise is a particularly important strategy for companies and they seem to be heavily targeting youth and young adults, who appear to be more susceptible to it than older adults and are the demographic that is susceptible to initiation or escalation of product use. Brand sharing and brand stretching grant another access point for tobacco companies to subliminally advertise and market their products. In addition to using the cigarette package, tobacco companies place brand names and use other design techniques on the actual stick, which is rated by smokers as more attractive than cigarettes without these characteristics.Despite some restrictions in the USA, tobacco companies continue to advertise in magazines with significant youth audiences, and are more likely to advertise youth preferred brands in these magazines.Tobacco companies circumvent partial advertising restrictions by concentrating advertisements in magazines where youth audience composition is near or at the minimum threshold level, thereby still exposing a sizeable number of youth to tobacco ads.For example, in the United States even a 15% threshold, which was the FDA-proposed rule in 1996 for advertising in print media, would have exposed at least two million youth to tobacco industry advertising.Cigarette companies consolidate marketing expenditures for magazine advertisements to brands that are popular among youth, African Americans, and LGBTQ populations .In the 1990s most of the US state attorneys general sued the major cigarette companies alleging, among other things, that the companies were advertising to children.The litigation was resolved with the “Master Settlement Agreement,” in which the companies agreed to some restrictions on marketing to children. After the agreements was signed, the percent of total magazine advertisement spending for mentholated brands increased from 13% in 1998 to 76% in 2006, with an associated increase in youth mentholated cigarette smoking .The tobacco industry’s claims that marketing is only used for brand switching and increasing marketing share12 does not make economic sense.For such claim to be economically viable, the number of people switching brands between companies would have to exceed individual tobacco company marketing expenditures, which is unlikely because several brands are sold by a few cigarette companies. Despite claims made by RJ Reynolds in the US in the late 1980s that it did not directly target children, youth were more likely than adults to report previous exposure to RJ Reynolds’ Joe the Camel cartoon character advertising campaign and accurately associated such image with Camel cigarette brand name . Children also found cigarette advertisements that used Joe the Camel as more appealing than adults.The market share for youth use of Camel had also increased from 0.5% in 1988 to 33% in 1991 during the Joe the Camel campaign. Youth receptivity to tobacco marketing is a strong predictor for smoking initiation and consumption patterns independent of other important predictors of smoking behavior .The odds of initiating smoking among youth receptive to tobacco marketing are twice that compared to unreceptive peers .Longitudinal studies show increased odds of progression from initiation of smoking to established smoking among adolescents who both owned cigarette promotional items and had a favorite cigarette advertisement. In the United States, there are few regulations on tobacco industry marketing of other tobacco products , despite rising use among young adult populations. Among young adult bar patrons , marketing receptivity is associated with other tobacco product use, including smokeless tobacco, hookah, cigarillos, and e-cigarettes. Moreover, current smokers receptive to tobacco marketing are also more likely to be poly-tobacco users than among youth not exposed to similar ads.Youth who thought the ads were more effective were more likely to have a positive attitude toward e-cigarettes and greater intentions for future e-cigarette use. Adults in the United States are also influenced by e-cigarette advertising, with adults reporting greater intention to initiate e-cigarette use after exposure to ecigarette advertising.Similar observations on the effects of alcohol marketing on youth substance use behavior are noted in the literature.

The same logic applies to the generation of 4- methylpentanal from VEA, which dominates the distribution of the isomer pair over hexanal. The formation of glyoxal likewise may be enhanced in VEA due to the higher stability of radical intermediates. Diacetyl is thought to be a byproduct of cannabis plants.However, the SIC of C4H6O2 in the vaped THC oil demonstrated multiple isomers of C4H6O2 in that mixture besides diacetyl. From VEA, diacetyl may be generated from the thermal-induced scission of the C−O bond on the acetyl group, which will form acyl radicals that combine to form diacetyl. The formation of a C4H6O2 isomer, 3-oxobutanal, can also be rationalized ; however, there was only one C4H6O2 peak in the vaping aerosol of pure VEA and it has the retention time of diacety lDNPH in our analytical method. Thus, we believe that diacetyl is the main C4H6O2 from VEA, while multiple isomers are likely formed when VEA and THC are vaped together. In some cases, the THC/VEA mixture produced more carbonyl emissions per mg particle mass than the pure compounds . Although this trend is less clear within error, it may suggest some synergetic effects between THC and VEA. Moreover, the THC oil tended to produce a higher amount of acetaldehyde than VEA. VEA degradation may form acetaldehyde , but the unbranched side chain of certain cannabinoids, such as THC and CBG , provides more direct pathways for acetaldehyde formation. Scheme 2 shows that multiple cannabinoids identified in Table 2 may be formed as a result of reactive oxygen addition . While analysis of the THC oil extract did not detect OH-functionalized cannabinoids in the original e-liquid ,flood and drain tray we cannot rule out the possibility that functionalized cannabinoids exist in the unvaped THC oil.

Borille et al.found 123 cannabinoid compounds or metabolites and eight noncannabinoid constituents in the extracts of cannabis plants by ESI-MS, with carbon number ranging from C15 to C55. All molecular formulas of the THC oxidation products in Scheme 2b were also identified in cannabis extracts,suggesting that oxidation from plant metabolism or during extraction could have occurred in addition to vaping. C19H28O3 is identified here as cannabiglendol-C3; C23H34O4 may have multiple isomers ;and C15H16O3/C15H18O3 is identified as cannabispirenone/ cannabispiran.Some compounds in Table 2 still remain unidentified . Due to the uncertainty of the collection efficiency of cannabinoids through the silica cartridge, the quantification reported here for a number of cannabinoids should be considered a lower limit . Cannabinoid emissions per puff increase with temperature at the same e-liquid composition and decrease with VEA addition at the same temperature. Often, chemical emission trends follow the trends in particle mass emission , but yield ratios will indicate if significant chemical transformations occurred during vaping. While the mass yields per mg of particle mass of CBG, Δ9 -THC, CBC, cannabiglendol, OH-cannabinol, and cannflavin B increase with respect to temperature in the vaped THC oil aerosol, it decreases for all of the acids , cannbispirone-A, 10-ethoxy-9-hydroxy THC, and CBN. Compounds such as OH-cannabinol may increase with temperature because they are more efficiently vaporized or are formed through radical-initiated chemistry , adding oxygen functional groups to the carbon skeleton of CBG and THC . The decreasing yield ratio of acids such as THCA is expected with temperature, as they may be degraded through decarboxylation more effectively at higher coil temperatures, a process that also occurs during the initial processing of THC oil, resulting in the formation of neutral cannabinoids .Thus, the large reservoir of THCA present prior to vaping converts to THC while vaping in this third-generation mod device.

The same process occurs from CGBA to CBG, which explains the increasing trend of CBG and THC with temperature. Thermal decarboxylation should be efficient with temperatures above 350 °F ,playing a role to increase the observed yield of neutral cannabinoids. The underlying reasons for the decreasing trend of CBN with temperature is not clear, as CBN is thought to be relatively thermally stable and may be produced from other cannabinoids.Interestingly, when VEA is added, the yields of cannabinoids per mg of particle mass increase at the same temperature for most, but not all, observed compounds. Furthermore, the addition of VEA reverses the temperature trends for some cannabinoids such as THC and CBG, causing their net degradation with temperature. It appears, therefore, that the addition of VEA accelerated both the aerosolization and degradation of many cannabinoids. The reasons for these trends are not apparent, and we cannot rule out that the measurement of particle mass between the two systems introduces sufficient undocumented error to explain these trends. It is possible the cotton wick used for this study exhibited less efficient wicking for the more viscous VEA,which may have caused higher localized temperatures for certain portions of the coil surface despite controlling the temperature in the center of the coil. The temperature, physical integrity of the coil, and the saturation of the wick were monitored to ensure excessive heating did not occur; however, the wicking material remains a limitation of this work as ceramic wicks are more generally used for viscous liquids. For cannabiglendol, OH-cannabinol, and Cannflavin B, a decrease in emission yield was observed when VEA was added. Each of these compounds has multiple polar OH groups that could hydrogen-bond with VEA,which could prevent them from escaping the e-liquid, although it is not clear why this intermolecular interaction would be preferable to those that occur with other cannabinoids. These complex trends may warrant further study. Multiple terpenoids were also quantified in the particles .

Notably, only C15 sesquiterpenes were observed, which are of sufficiently low volatility to remain in the particle and which could represent only a small fraction of the terpene diversity and quantity found in the THC oil extract from cannabis.However, it was not possible to perform a quantification of terpenes in the original e-liquid to compare with that found in the particle composition due to issues pertaining to licensure. Many terpenoids, especially the non-functionalized C10 monoterpenes, are considered volatile organic compounds ; they would preferentially be emitted in the gas phase during vaping. Thus, these terpene observations represent a lower limit as the particle filtration and extraction steps may lose terpenes due to volatilization. The hydroxyl groups of cedrol and nerolidol also help reduce vapor pressure; it is not clear from these data if functionalized terpenes were generated through oxidation or were originally present in the e-liquid. While temperature increases the emissions of terpenes per puff, temperature decreases the terpene yield per mg particle collected from the vaped THC oil . Unlike cannabinoids,indoor grow shelves the addition of VEA did not significantly change terpene yield per mg particle. An exception is cedrol, where the yield at medium temperature increased with VEA addition for unknown reasons. The terpene yield data are consistent with other reports of terpene degradation. Meehan-Atrash et al.identified degradation products from myrcene, limonene, and linalool, including methacrolein, hydroxyacetone, and methyl vinyl ketone.Tang et al.found 11 thermal degradation products from a mixture of terpenoids, 7 of them are carbonyls including formaldehyde, acetaldehyde, acetone, acrolein, methacrolein, valeraldehyde, and hexanal. The methacrolein formed from vaped THC oil likely originates from terpene degradation, and its enhancement with VEA addition may be due to the aforementioned acceleration of volatilization or chemistry,indoor garden table as well as its source from VEA.VEA and cannabinoids are observed to chemically react in the e-cigarettevessel in a manner that is consistent with the degradation of PG and VG in conventional e-cigarettes, i.e., via the thermally induced degradation and/or ROS-induced degradation schemes described by Jensen et al.,Li et al.,and Diaz et al.,among others. ROS such as OH radicals have been directly measured and inferred by degradation product analyses,in e-cigarette vessels and aerosol particles. However, OH sources in the vaping process are not well understood mechanistically. It had been suggested that OH can be formed from O2 insertions into organic molecules, or from redox cycling of redox-active organics and/or transition metals.Thermal degradation carbonyls and acids appear to be formed by C−C bond cleavage of the aliphatic side chain of VEA, with one carbonyl moiety formed at the site of each cleaved carbon . This cleavage process produces two aldehydes at an unbranched site and aldehyde/ketone pair at a branched site. The degradation reactions may be initiated by bond homolysis, dehydration, or H-abstraction and addition by radicals such as OH, followed by the rapid reaction with O2 to form peroxy radicals .

The peroxy radicals can react with other RO2 to form carbonyls or alkoxy radicals.Alkoxy radicals may further react to form carbonyls , alcohols , and possibly alkenes .The primary thermal degradation products may go through further oxidation steps and form more thermal degradation products .The RO2-based mechanisms have been well studied and shown to be important in various chemical systems, like the atmosphere, biological redox, or fuel combustion.These mechanism are consistent with observations, as the most abundant carbonyls observed in the VEA aerosol can be rationalized to be formed from the most stable radicals in the first H-abstraction step . The benzylic radicals are stabilized by the conjugation effect from benzene ring and positive hyperconjugation from the adjacent C−H bonds.The proposed thermal degradation pathway is also supported by the detection of alkenes by Riordan-Short et al.and Mikheev et al.since these compounds are predicted in the proposed mechanism. Thus, our observations suggest that the C−C single bonds on the side chain of VEA are easily oxidized and cleaved during the vaping process, which will cause the formation of a series of carbonyls that have VEA-specific structure and also alkenes and alcohols. These primary products may go through a furtherthermal degradation process to generate secondary thermal degradation products like acids and dicarbonyls. OH radical can add to the unsaturated CC bonds of Δ9 THC and CBG to produce oxygen-functionalized products in the vaping aerosol of THC oil . In contrast to VEA, the oxidation of CBG by OH proceeds primarily through the addition of the double bonds in the side chain, consistent with the oxidation of other alkenes.The mechanism for the following steps is similar to the H-abstraction route. CBG may be the source of unique carbonyl products due to its second unsaturated side chain ; the stepwise mechanism is shown in Scheme S2 for C8H14O. The oxidation may also occur on the unsaturated rings of cannabinoids, such as THC . However, unlike CBG, the allylic site of THC also enables substantial Habstraction by OH in addition to the OH addition occurring at the endocyclic CC, preferentially forming the tertiary alkyl radical . Multiple SIC peaks are found at the m/z representing oxidized products of cannabinoids, suggesting that different isomers abound. Our identification results are consistent with those of Carbone et al.,who utilized NMR for identification. Carbone et al. indicated that peroxide products may also be formed during the oxidation process, a mechanism not shown in our schemes but would be consistent with RO2 chemistry.Although the third-generation temperature-programmable mod vaping device used in this work likely protects from excessive formation of toxic pyrolytic byproducts from cannabis extracts,74 a myriad of thermal degradation and oxidation products were observed at the tested temperatures from cannabinoids, VEA, and terpenes under typical operating conditions. The addition of VEA had complex effects on aerosolization efficiency and product formation that is supplemental to temperature. It is clear that the addition of VEA increases the formation of formaldehyde, glyoxal, 4- methylpentanal, methacrolein, and diacetyl, among other carbonyls per unit of particle mass. Self-titration of THC dose by users may enhance their inhalation exposure to VEA products when the VEA fraction in the e-liquid approaches 100%, due to increasingly higher production of certain carbonyls but increasingly lower emissions of THC and total particle mass. However, at the 1:1 mixture, the particle’s THC yield is also enhanced compared to THC oil extract, which may negate increases in some carbonyl emissions for self-titration purposes. At the same time, VEA addition to the e-liquid had no effect on the observed yields of terpenoids, but a complex effect on the cannabinoid yield. Rich oxidative decomposition chemistry was observed for each compound class in the e-liquid. THC has a stronger tendency to degrade compared to VEA.

Shared micromobility services are growing rapidly across the United States and abroad. In 2018, the number of shared micromobility trips in the U.S., including station-based and dockless bike sharing and e-bike sharing and scooter sharing, reached 84 million . Of those trips, 36 million were station-based bike sharing trips, about 500,000 were station based e-bike sharing trips, and about 6 million were dockless e-bike sharing trips. By enabling users to access a fleet of publicly available shared personal transportation devices on an as needed basis, shared micromobility offers on-demand, low-emission public transportation options that can help to reduce congestion and emissions, as well as improve public health within urban areas . Traditional station-based bike sharing systems have been studied in depth, generating some agreement about their positive impact on cycling rates, modal shifts from personal vehicle use, and promotion of public transit ridership through improved first- and last-mile connections to public transit stations . These benefits can contribute to various federal, state, and local objectives to improve mobility, safety, and public health, and reduce congestion, fuel use, and emissions. Innovations to incumbent bike sharing technology and business models are spreading in a new wave of fourth generation bike sharing and scooter sharing, which includes dockless models, demand-responsive pricing and rebalancing, and electric fleets of bicycles, standing scooters,rolling bench and moped-style scooters. The increased geographic coverage and availability of these new service models have great potential to further expand and integrate shared micromobility in the transportation system, as such factors have been identified as significant drivers in traditional bike sharing ridership .

Early analysis of aggregated activity data from dockless and scooter-based models shows substantial expansion of micromobility ridership in urban areas , with an estimated market potential of 8 to 15 percent of all passenger trips under 5 miles . However, concerns about curb management, safety, and the sustainability of the micromobility vehicle supply have become a central focus of the ongoing development of regional and local regulations and permitting programs have . This paper examines the expansion of bike sharing in the City of San Francisco. In January 2018, the San Francisco Municipal Transportation Agency issued a permit for a pilot dockless electric bike sharing system, called JUMP, which began to operate in parallel to the existing regional station-based bike sharing system, Ford GoBike. Around the same time, it was announced that the GoBike system would expand to include electric bicycles as well, which became available in April 2018. The expansion of the docked GoBike system complemented by the newer stationless JUMP bikes necessitates an evaluation of the effectiveness of each system in providing additional mobility in San Francisco and consideration of the spatial distribution of bicycle infrastructure, such as bicycle lanes and public bike racks to support the potential increase in bicycle demand. As a city with a highly concentrated central business district that is surrounded by steep hills with medium density residential and mixed use areas, San Francisco offers an interesting case study for examining the impacts of e-bike and dockless models on bike sharing travel behavior throughout the city.

Our research seeks to understand the impact that the dockless, e-bike sharing model might have on bike sharing users’ travel behavior as compared to traditional docked bike sharing. A primary objective is to characterize the spatial distribution of demand for both dockless electric and docked bike sharing throughout San Francisco. We created a destination choice model using a month of activity data from both JUMP and GoBike to quantify the relative bike sharing attractiveness at a neighborhood scale for users of each system. Our analysis revealed the impact that dockless, e-bike sharing has on the sensitivity of bike sharing users’ destination choices to various exogenous factors such as: bicycle infrastructure, topography, socio-demographics, and land-use variables. This article is organized into four sections. First, we provide background, including a brief history of bike sharing in general and in San Francisco, as well as a discussion of the major factors that impact bike sharing ridership, as revealed in existing literature. Next, we present an overview of our methodology. Third, we present the results of travel behavior analysis, destination choice modeling, and bike sharing suitability analyses. Finally, we highlight our conclusions and future work. Public bike sharing and scooter sharing systems are quickly becoming the most widely adopted and rapidly growing shared miocromobility options across the U.S. . By enabling users to access a fleet of publicly available shared personal transportation devices on an as-needed basis, shared micromobility offers on-demand, low emission public transportation options that can help to reduce congestion and emissions, as well as improve public health within urban areas . As shared micromobility continues to expand and evolve with emerging technology and business models, new insights regarding the unique impacts of electric vehicles and dockless models on ridership and travel behavior is needed to aid cities in understanding how to best manage local micromobility ecosystems to promote a more sustainable and equitable transportation system.

The first public bike sharing system emerged in 1965 in Amsterdam, Netherlands. This innovation has expanded to reach cities across Europe, North America, South America, Asia, and Australia. At present, most bike sharing systems are classified as third generation, characterized by the implementation of information technology for bicycle pick-up, drop-off, and tracking. Bike sharing systems are predominantly “station-based” or docked, where bicycles are located at public docking stations and trips are required to originate and terminate . Docks are typically concentrated in urban areas, creating a network of on-demand bicycles suitable for a variety of trip purposes. Users can instantly unlock an available bicycle from a docking station using a credit/debit card, membership card, key, and/or a smartphone application. There are a variety of fare structures applied in bike sharing systems including: daily, monthly, and annual passes. In most systems, fares tend to cover at least the first 30 minutes of a trip, with overage charges beyond that time. Many systems also allow users to chain multiple trip segments of 30 minutes or less, such that a user can extend their riding time by “ending” a trip segment at a dock and immediately unlocking a bike for another trip segment. Fourth generation bike sharing builds upon the IT-enabled third generation to deliver demand-responsive, multi-modal systems. The dockless, or free floating, bike sharing model is one such innovation, which allows users to pick-up and drop-off bicycles anywhere within a service zone. Demand-responsive bicycle redistribution and value pricing encourages users to participate in the rebalancing of bicycles, facilitating a spatiotemporal distribution of bicycles that closely matches system supply and demand. Bike sharing systems are also becoming more integrated with other transportation modes through mobility as a service models, including: public transit; car sharing ; and ride sourcing/transportation network companies . Uber Technologies, Inc. acquired JUMP in April 2018 . Interestingly,Lyft acquired Motivate, the parent company of GoBike in July 2018. This likely signals that shared mobility companies are interested in becoming multi-modal MAAS platforms consisting of more than one shared mode. In late-2017 and early-2018, bike sharing operators Social Bicycles , Motivate, and Lime, began operating bike sharing systems with electric assist bicycles or e-bikes. E-bikes have an electric motor that reduces the effort required by the rider, allowing for greater speeds and ease in riding uphill. Research on personally owned e-bike use has found that the main reasons people choose to use e-bikes include living or working in hilly areas, medical conditions, fitness,drying rack cannabis and the desire to ride with less effort . E-bikes can mitigate the inconvenience imposed by needing to shower after bicycling, thus providing an attractive alternative to traditional bikes for commute trips.

MacArthur et al. found that 80% of sampled e-bike users under the age of 55 and 68% of those 55 and older said that they did not need to shower after using e-bicycles. A report on shared micromobility in 2018 found that, in cities where e-bikes were added to station-based bike sharing fleets, e-bike utilization was about twice that of pedal bikes, on average . Ford GoBike launched in Summer 2017 as a re-branding and expansion of the Bay Area Bike Share system, which launched in San Francisco and San Jose in 2013. GoBike provides access to five cities, 540 stations, and 7,000 bikes . As with many docked bike sharing systems, standard GoBike rides are 30 minutes long, with each additional 15 minutes costing extra. GoBike offers single ride, day pass, and annual membership payment plans, with the day/annual passes providing unlimited standard rides for the duration of their validity. Users can locate and unlock a bicycle using a mobile app, Clipper Card, or by paying on-site using a kiosk. Around the same time as the JUMP pilot launch, in April 2018, GoBike added 250 electric pedal-assist bicycles to its San Francisco fleet followed by an additional 600 in December 2018. However, we note that at the time of the study period, the GoBike fleet comprised solely of standard pedal bicycles. Figure 1 below shows the service areas of GoBike and JUMP during the study period. JUMP Bikes, a program of Social Bicycles, launched in January 2018 after the SFMTA issued the city’s first permit to operate a dockless bike sharing service. As an 18-month pilot program under evaluation by SFMTA, JUMP is committed to providing a “safe, equitable, and accountable” dockless e-bike sharing system . For the duration of the pilot, SFMTA will not issue any other dockless bike sharing permits and aims to develop policy recommendations based on the pilot’s results. The initial pilot allowed for 250 bikes until October, 2018 when an additional 250 bikes were added to the fleet. With integrated onboard Ulocks, JUMP bikes are parked at regular bike racks or locked to a fixed object in the sidewalk “furniture zone,” the portion of sidewalk from the curb to the pedestrian walk zone . Users can locate and unlock the bikes using a smartphone application, password, or radio-frequency identification member card . While the literature on ridership and travel behavior of dockless and electric shared micromobility is limited, there has been extensive research on the use of station-based bike sharing models. The literature reveals three major external factors that impact bike sharing ridership: 1) infrastructure , 2) geography , and 3) user demographics . Much of the literature on internal factors has focused on a-priori and optimization analyses of station location, dock allocation, fleet sizing, and rebalancing algorithms . We thus focus our attention on empirical findings from the literature on the impacts of external factors on bike sharing ridership and travel behavior.Infrastructure indicators for bike sharing ridership relate to the availability and attractiveness of bicycle facilities, such as bike lanes, bike paths, and bike boulevards. Buehler and Pucher show that bike commute ridership correlates positively with the supply of bike paths and lanes, even when controlling for other contributing factors . A destination choice analysis of the Divvy docked bike sharing system in Chicago found that bike sharing users preferred destinations with a greater density of bicycle facilities in the surrounding area, and bike sharing members were more sensitive to this factor than non-members . Finally, a multiple regression analysis of the Capital Bike share system in Washington, D.C., found that the total length of bike lanes within .5 miles of a station was a significant positive factor in the number of rides per day at the station .The quality of the bike infrastructure can also impact the sensitivity of demand for bike sharing. Indeed, route choice modeling of Grid Bike share users in Phoenix, Arizona found that bike-specific facilities increase the preference for a particular route by an amount equivalent to decreasing the travel distance by 44.9% . In Portland, Oregon, route choice modeling revealed that bicyclists prefer bike boulevards and bike paths, which are typically on streets with little and no vehicle traffic, respectively, to bike lanes, which are facilities that share the road with regular traffic . The next critical question is where to supply infrastructure per demand. Although station proximity to bike infrastructure is a top design priority , it is important to note⎯that due to geographical constraints⎯not all origin-destination pairs are equally attractive. Job density, proximity to public transit services, and proximity to recreational areas at the location of a bike sharing station have been found to be positive factors in bike sharing demand . For example, in the Nice Ride program in Minneapolis-St. Paul, stations farther away from central business districts of the twin cities as well as those farther away from parks generated fewer bike sharing trips . For e-Bikes, an elevation change between origin and destination locations is also a positive demand factor.

A second goal was to test whether post-legalization changes in marijuana trajectories would be moderated by demographic or other substance use factors. We expected to see greater increases in use among male participants, and among those who used other substances more frequently. Finally, we explored whether changes in marijuana use frequency following legalization were related to cumulative frequency of use prior to 2018. We recently described the relationship between marijuana and tobacco use in this sample and there is partial overlap in the data used in these studies. The primary differences are the previous study included data only from subjects’ first two years following enrollment, while the current study utilized data from all three years of follow-up, regardless of when subjects enrolled, and the previous study examined trends in use over time without regard to changes in the legal environment, while the current study explicitly examined whether frequency of marijuana use was associated with legalization of recreational sales. Demographics evaluated at baseline included sex, age, racial/ethnic background, and student status. Because the age range was narrow and our interest was in the potential impact of legalization,drying cannabis age was transformed into a time-varying binary variable reflecting whether or not participants were aged 21 at the time of each assessment. Student status was collapsed into a dichotomous variable comparing full-time students to all others.

At annual assessments, participants completed the Timeline Follow Back , on which they reported number of cigarettes, and whether they had used each of e-cigarettes, hookah, cigars, cigarillos, smokeless tobacco, snus, marijuana, and alcohol on each of the previous 14 days. At quarterly assessments, participants completed brief daily surveys, in which they indicated whether they had used each of the same products in the last 24 hours. We created time-varying variables representing constructs of interest. Time reflected the study time point, from baseline to year 3 . For all assessments, we calculated the number of days on which participants used marijuana , cigarettes , e-cigarettes , and alcohol . We counted the number of days at each time point at which assessment occurred to account for the fact that the maximum number of days differed for annual versus quarterly assessments, and that participants may not have provided data for all days during quarterly assessments. We created a binary legalization variable that indicated whether or not each assessment occurred prior to or after January 1, 2018. We also created a post-legalization slope variable that was coded as 0 for all pre-legalization time points, and to reflect time since legalization for post-legalizatin time points . Finally, for each participant we calculated the total number of days prior to January 1, 2018 on which use was assessed , as well as the number of those days on which any marijuana use was reported . We used bivariate tests to evaluate whether demographic variables were related to predictors and outcomes; when associations were significant, we accounted for demographics in hypothesis tests. To test whether frequency of marijuana use changed following legalization we utilized a piece wise or segmented multilevel longitudinal regression model, an approach recommended for evaluating the impact of policy changes . This model included segments for the period prior to January 1, 2018, and for the period from that date onward.

The model tested the temporal trend in frequency of marijuana use, the impact of legalization, and changes in the rate of marijuana use over time following legalization by incorporating the time, legalization, and post-legalization slope variables as predictors. Sex, race/ethnicity, and binary age were included as covariates. Second, we used multilevel longitudinal regression models to evaluate the associations of sex, race/ethnicity, binary age and frequency of alcohol, cigarette and e-cigarette use with frequency of marijuana use before and after legalization. We did so by testing for three-way interactions between the predictors of interest , time, and legalization. Significant three-way interactions would indicate that impact of legalization on the trajectory of marijuana use frequency was moderated by the predictor of interest. All demographic interaction terms were included in one model, and all substance use interaction terms in another. In both cases, non-significant interaction terms were removed in a backward manner and models refit. Finally, we used a similar multilevel modeling approach to test whether time-invariant cumulative marijuana days was associated with time-varying marijuana frequency over time post-legalization. All analyses were conducted using Stata 15.0, with alpha = .05; missing data were not imputed. The proportion of data missing was 0% for the first 3 assessments , and increased with each subsequent assessment, with 3.2% of participants missing outcome data at year 1, 10.5% at year 2, and 14.1% at the final year 3 time point. Missingness increased over time and was most common among participants identifying as White . Missingness was not significantly associated with sex or with frequency of cigarette, e-cigarette, or marijuana use at the previous assessment. The first 5 assesments occurred prior to legalization for all participants. The proportion of the sample for whom assessment occurred after legalization increased with each subsequent assessment, from 1.8% at the first quarterly time point following year 1, to 37.7% at year 2, 80.3% six months after year 2, and 100% at year 3.

Bivariate analyses indicated women tended to use e-cigarettes less frequently , younger participants used marijuana more frequently , and non-White participants reported greater cigarette frequency . Consequently, sex, age, and race/ethnicity were included in subsequent analyses. Table 1 details frequency and likelihood of marijuana, alcohol and tobacco use at baseline and at each annual assessment. The proportion of days on which participants used marijuana remained relatively stable, while the number of participants who reported any marijuana use declined modestly from baseline to year 3. Alcohol use was stable across the three years of observation. Proportion of days using e-cigarettes exhibited a 50% increase, while the proportion of participants with any e-cigarette use was relatively stable. In contrast, use of cigarettes, and consequently overall use of tobacco products, decreased over time.The piece wise regression model is shown in Table 2. Frequency of marijuana use was significantly associated with race/ethnicity and age, such that participants who identified as white and who were under age 21 at the time of assessment reported more days of marijuana use. The main effect of time was not significant,greenhouse benches indicating that days of marijuana use was stable over 3 years of observation, consistent with the descriptive statistics in Table 1. The post-legalization slope term was also not significant, indicating that the trajectory of marijuana use for the post-legalization segment of the model did not differ from the overall trajectory. Table 3 shows the final model evaluating the impact of legalization on associations between demographic variables and frequency of marijuana use over time. We found that age and racial/ethnic identity continued to predict marijuana use frequency, but that the strength of those associations did not change over time or following legalization. In contrast, we found significant interactions of sex with both time and legalization. To better understand these interactions, we removed sex from the model and evaluated associations between time, legalization, and marijuana use frequency separately for men and women. These analyses indicated that marijuana use frequency generally decreased over time for male participants , but also increased non-significantly following legalization . In contrast, female participants reported increasing marijuana use frequency over time overall, but with a non-significant decrease after legalization . Examinat in of adjusted means suggested that, in both cases, the non-significant effect of legalization was a reflection of an initial post-legalization increase followed by a reversion to the previous trend of decreasing use over time for men and increasing use for women. Table 4 shows the results of the model examining substance use predictors. There was a positive association between alcohol frequency and marijuana frequency, but this did not vary by time or after legalization. In contrast, we found that the associations between both cigarette frequency and e-cigarette frequency and marijuana frequency over time were moderated by legalization. To clarify these interactions, we removed legalization from the model and examined associations before and after legalization. These simple effects tests showed that, before legalization, there was a consistent positive association between cigarette and marijuana use frequencies that did not vary over time . However, this association declined over time following legalization .

In contrast, the association between e-cigarette frequency and marijuana frequency was significant at baseline but declined over time prior to legalization . However, following legalization there was a consistent positive association between the two . Finally, we evaluated the extent to which the total number of days of marijuana use prior to legalization predicted days of marijuana use after legalization, and if so whether this varied by time. Age, sex, and race were included as covariates but none were significantly associated with marijuana use after legalization in this model. We found a significant main effect and interaction with time . The former indicates that those who reported more cumulative days of marijuana use prior to 2018 also reported more days of marijuana use at the first assessments they completed in 2018, while the latter indicates that this association grew stronger over subsequent observations. We set out to examine whether frequency of marijuana use changed following legalization of recreational sales in California. We also planned to test whether post-legalization trajectories of marijuana frequency would be associated with sex, age, race/ethnicity, alcohol or tobacco use, or pre-legalization marijuana frequency. We utilized a sample of young adults who were non- and never-daily cigarette smokers at the time of enrollment. This sample has multiple advantages compared with others that are available. Unlike most national datasets, we were able to evaluate change over time in a specific cohort. Additionally, assessment occurred at specific, quarterly intervals. Thus, in addition to providing more assessments within each year, it was possible to pinpoint each assessment to before or after changes in legal status. Additionally, the analytic approach allowed us to include participants who were enrolled at different points prior to legalization and thus had completed varying numbers of assessments at that point. Contrary to our expectations, frequency of marijuana use did not change significantly after legalization, and was stable throughout three years of observation. Participants who were younger and who identified as White reported more days of marijuana use; these associations were consistent over time and did not change with legalization. Sex differences were also noted, with men reporting decreasing and women increasing marijuana use frequency over time, though this association was not significantly related to legalization. This difference is contrary to previous research suggesting greater use among men , though more recent data suggest that this discrepancy is shrinking . Our findings are consistent with evidence that use may escalate more quickly among women . Women appear to be more sensitive to the rewarding effects of cannabis use , and thus may be more vulnerable to increasing use after initiation and/or when barriers to use are reduced. We also found that associations of both cigarette and e-cigarette frequency with marijuana frequency over time were moderated by legalization. More specifically, the association between marijuana and cigarette use became weaker following legalization, while the marijuana-e-cigarette association showed the opposite pattern. Frequency of alcohol consumption was consistently associated with marijuana use over time and did not change with legalization. Finally, we found that those who reported more frequent marijuana use prior to legalization tended to do the same afterward, particularly at later assessment points. Although frequency of of marijuana use was associated with both cigarette and e-cigarette use, the post-legalization findings suggest that co-use of e-cigarettes and marijuana may increase when the latter is legalized. One potential explanation for this could be that many young adults perceive vaping and marijuana use as conferring little risk , in which case legalization may have removed an important barrier to use. In combination with the finding that marijuana use was more common among those under age 21, this suggests that enforcement of minimum age laws may be an important component of limiting use of both marijuana and e-cigarettes. Our finding of no overall change in marijuana frequency is consistent with reports suggesting little impact of medical marijuana laws on use in California .

The amplitudes of BOLD and CBF percent signal changes, both during rest and task, were calculated as root mean square values by computing the standard deviation across the length of the time courses. Note that for the task-related BOLD and CBF signals, only the last 3 cycles were used in this calculation. As CBF signal changes can be overestimated due to low SNR in ASL CBF data, we removed voxels containing RMS %∆CBF outliers from all analyses. Outliers were de- fined as values lying more than 1.5 times the interquartile range above the third quartile or below the first quartile of all RMS %∆CBF values . Average values for RMS %∆BOLD and RMS %∆CBF were calculated for each subject by first averaging the time courses in the motor cortex ROIs and then calculating RMS values. For comparison with previous research , the evoked BOLD and CBF response amplitudes to finger tapping were also calculated using a GLM analysis on the average %∆BOLD and %∆CBF time courses in the motor cortex. Instead of computing RMS values, task amplitudes were computed as the weights of the stimulus related regressor, which was described in the ROI Definition section. Resting-state BOLD data were also used to measure functional connectivity strength between the right and left motor cortices for each subject. This was done by correlating the average BOLD time course from the left motor ROI with the average BOLD time course from the right motor ROI for a measure of BOLD connectivity. Relationships between BOLD and CBF measures were examined across subjects using correlation analysis.

Linear fits were also computed for each relationship,vertical grow rack and the slopes were compared between the task and rest conditions using an ANCOVA.Relationships between BOLD amplitude and baseline CBF are shown during task and rest in Figure 4.1a. Similar to a previous study using a visual stimulus , we find that the BOLD response to finger tapping displays a significant negative correlation with baseline CBF. In contrast, while the resting BOLD fluctuation amplitude exhibits a trend of inverse dependence on baseline CBF, it is not significant . Furthermore, the linear fit for the finger-tapping data has a significantly steeper slope than for the resting-state data = 9.8, p = 0.004. We did not find a significant relationship between resting-state BOLD connectivity and baseline CBF , however we did find resting-state BOLD connectivity and amplitude measures to be correlated across subjects . Note that the inverse dependence of task-related BOLD signal changes on baseline values of CBF was also found using the GLM-based task amplitudes instead of RMS values . Preliminary work suggests that the inverse dependence of the task BOLD am-plitude on baseline CBF is caused by the direct dependence of %∆BOLD on %∆CBF, which is modeled in the Davis equation . This is because %∆CBF is inversely related to baseline CBF, as CBF0 is the denominator in calculating %∆CBF. Empirically determining the dependence of %∆BOLD on %∆CBF during rest can be challenging because of the inherently low signal to noise ratio in ASL. This can be seen in Figure 4.2, which plots the correlation coefficients between the average motor cortex BOLD and CBF time courses during task and rest. The correlation values are visibly much higher during the task condition, where the solid line represent equality between the two states, reflecting the larger SNR during task than rest.

In Figure 4.2 data points shown in blue represent subjects with insignificant < 0.16, p > 0.05 correlation values during resting state. We explore the relationships between RMS %∆BOLD and RMS %∆CBF during finger tapping and rest in Figure 4.3a, where again subjects with insignificant correlations between their BOLD and CBF time courses are marked in blue. In this plot, lines represent linear fits to the data and correlation values are shown. To investigate the cause of the shallower slope during resting state shown in , we plotted absolute RMS CBF verses baseline CBF in units of ml/ in Figure 4.4a, where solid lines represent linear fits to the data. The absolute CBF fluctuations are independent of baseline CBF during task and rest . In addition, when we compared the slopes of the linear fits to flat lines intersecting the RMS CBF means for task and rest we did not find a significant difference < 0.76, p > 0.39. Note that the mean of the RMS CBF values during task is significantly larger than the mean during rest = 12.2, p = 1.6e-9. The independence of the absolute RMS CBF values on baseline CBF and the significantly smaller mean RMS CBF value during resting state explain the weakened dependence of relative changes in CBF on baseline CBF. To show this, the experimental values of RMS %∆CBF are plotted versus CBF0 in Figure 4.4b. In this plot, solid lines represent simulated RMS %∆CBF assuming a constant absolute RMS CBF of 20.5 for task and 8.4 for rest. These simulated lines closely follow the excursions of measured RMS %∆CBF across the range of baseline CBF, with the simulated task line displaying a visibly steeper slope. To further explore the weakened dependence between RMS %∆BOLD and RMS %∆CBF during resting state , the ratio of RMS %∆BOLD to RMS %∆CBF was determined and plotted in Figure 4.5a for each subject during the rest and task conditions. The solid line represents equality between the two states. A significantly higher BOLD to CBF ratio was found during task than rest = 8.5, p = 2e-7 across all subjects, and also across only those subjects with significant BOLD-CBF correlations shown as black data points = 7.2, p = 2e-5.

Considering the dependence of %∆BOLD on %∆CBF as modeled in Eq. 4.1, we note that the parameter M is not expected to change substantially between task and rest because M depends on experimental parameters and baseline physiology. The experimental parameters are identical for the two scans, while the subject’s baseline physiology is not expected to significantly change between the task and rest scans, which are acquired right next to each other in the scan session. However, a difference in the CBF-CMRO2 coupling factor n could exist between the task and rest conditions, and would give rise to the observed differences in the BOLD to CBF relationship. Estimates of n for each subject are shown for both task and rest in Figure 4.5b, with M assumed to be previously determined 7.2% . We found n to be significantly higher during task = 8.4, p = 3e-7. This result was insensitive to our selection of M, which we varied from 3.4% to 12% in increments of 0.2% < 10.2, 0.03 > p > 2e-8. When only subjects with significant BOLD-CBF correlations were considered, we found that n remains significantly higher during task for M equal to 4.2% and higher < 8.3, 5e-6 < p < 0.01. Using M = 7.2% and the mean n values from Figure 4.5b , we used the Davis model to compute RMS %∆BOLD values and compared those to the measured RMS %∆BOLD values . It can be seen that the smaller n during rest makes the BOLD signal less sensitive to CBF changes. Dashed lines represent model estimates using mean n values calculated from subjects with significant BOLD-CBF correlations. The results presented in this paper are from the two dual-echo ASL/BOLD scans,cannabis grow racks but we also compared the resting-state BOLD measures acquired from the second echo of the resting-state ASL/BOLD scans with those acquired during the single-echo BOLD weighted imaging scans. The same processing methods were applied to the data from the two single-echo BOLD-weighted resting-state scans, with the exception that running average filtering was not applied. Values of RMS %∆BOLD and functional connectivity strength were then averaged across the two runs. We found that RMS %∆BOLD values measured using the dual-echo ASL/BOLD and single-echo BOLD-weighted scans were significantly correlated across subjects . Furthermore, use of the single-echo BOLD-weighted imaging data in place of the dual-echo ASL/BOLD data did not change our overall conclusions. Resting-state fMRI is frequently used to measure patterns of spontaneous neural activity in the brain. However, the BOLD signal provides only an indirect measure of neural activity and is a complex function of blood flow and oxygenation changes.

In this study, we find that inter-subject differences in the amplitude of the BOLD response to finger tapping demonstrate an inverse dependence on baseline CBF, in agreement with previous findings . This suggests that hemodynamic differences between subjects may need to be taken into account when comparing task-related evoked BOLD responses. In contrast, the amplitudes of spontaneous BOLD fluctuations were not significantly related to inter-subject differences in CBF. This reduced dependence appears to be caused by a combination of two factors. The first, and perhaps most dominant, is that percent changes in CBF displayed a weaker inverse dependence on baseline CBF during rest than task, where the BOLD signal is directly linked to relative CBF changes as modeled in the Davis equation . The lower sensitivity of %∆CBF changes on CBF0 is due to the significantly smaller amplitudes of absolute CBF fluctuations during the resting state. As we found absolute CBF fluctuations to be independent of baseline CBF during both finger tapping and rest, dividing the smaller absolute fluctuations by baseline CBF led to a shallower slope between relative changes in CBF and baseline CBF. The independence of absolute functional CBF changes on baseline CBF during visual stimulation has been previously reported , and is consistent with an additive model of functional CBF changes, which are constant for a given stimulus regardless of baseline blood flow. Absolute changes in CBF are presumably smaller during rest because this state is less demanding in terms of energy cost, so a smaller amount of oxygen needs to be delivered to the motor cortex. In addition,as flow-metabolism coupling appears to be tightened during rest, smaller fluctuations in CBF may still accompany fairly large fluctuations in oxygen consumption. The second factor leading to reduced sensitivity of the resting BOLD signal to baseline CBF is the weakened relationship between %∆BOLD and %∆CBF, which seems to be caused by tighter flow-metabolism coupling in the motor cortex during rest. The ratio of the relative changes in CBF and CMRO2 has previously been shown to vary across brain region . In addition, differences in n were found within the visual cortex depending on the subjects’ state of attention . Recent studies have shown that flow-metabolism coupling increases with greater visual stimulus contrast and frequency suggesting that n is modulated by task difficulty. Our results in the motor cortex are consistent with this finding, using the assumption that the resting state represents the lowest level of task difficulty. In contrast to our findings, a previous study did not find significant differences in the BOLD/CBF ratio between rest and a visual task . A potential source of this discrepancy is the use of non-quantitative CBF measures in the prior study. It should be noted that the difference in flow-metabolism coupling observed between task and rest was for a sensorimotor task that elicited a robust task-related response. We expect that similar differences in coupling will also be observed for other paradigms and brain regions with robust task-related responses . However, we expect that there will not be a significant difference in the coupling observed during task and rest conditions for cognitive paradigms and brain regions with relatively weak task-related responses, such as memory-encoding responses in the medial temporal lobe . For these paradigms, the coupling between flow and metabolism is already tighter as compared to the coupling observed in the visual and motor regions . The reason for the tighter coupling is not yet known, but may reflect a difference between brain regions that have adapted to primarily process inputs from the external world versus those that have developed to process intrinsic information from other brain regions. We expect that the tighter coupling for task-related responses in brain regions that handle intrinsic information will also reduce the dependence of the task-related BOLD signal on baseline CBF for paradigms that activate these areas. Further work is needed to test these conjectures. The results of this study suggest that inter-subject differences in resting-state functional connectivity are relatively insensitive to variations in CBF. In a preliminary study, we previously reported that BOLD connectivity strength and baseline CBF were negatively correlated .

Obrig et al. used near infrared spectroscopy to show that hypercapnia also attenuates spontaneous fluctuations in cerebral oxygenation in awake humans. In general, the prior studies suggest that vasodilation reduces the amplitude of spontaneous fluctuations, while vasoconstriction tends to increase fluctuations.Similar to task-related studies, the results of these resting-state studies suggest that the strength of neurovascular coupling between spontaneous neural activity and hemodynamic fluctuations is inversely related to baseline CBF. Furthermore, a restingstate fMRI experiment found that hypercapnia decreased BOLD connectivity measured in the motor cortex in addition to reducing low-frequency BOLD fluctuations . A number of animal studies have found that mild levels of hypercapnia do not affect neural activity . Jones et al. used optical imaging spectroscopy, laser Doppler flowmetry, and multi-channel electrophysiology to measure the hemodynamic and neural responses to whisker stimulation in rats anesthetized with urethane. They found that 5% CO2 did not significantly alter the cerebral metabolic rate of oxygen responses or the EEG responses. In another study with spontaneously breathing isoflurane anesthetized rats, Sicard et al. showed that both baseline CMRO2 levels and CMRO2 increases with forepaw stimulation were not significantly changed by 5% CO2. Zappe et al. found a trend toward hypercapnia reduced spontaneous neural activity as measured with intracortical recordings in the visual cortex of remifentanil anesthetized macaque monkeys,flood and drain hydroponics but it was not significant at 3% CO2 . These findings suggest that a decrease in BOLD connectivity caused by hypercapnia most likely reflects a reduction in the strength of neurovascular coupling, as opposed to an alteration in neural activity.

Therefore, factors that change baseline CBF may be significant confounds in resting-state fMRI studies, as a decrease in BOLD connectivity due to a reduction in the strength of neurovascular coupling might be incorrectly interpreted as a decrease in neural connectivity. An improved understanding of the effects that vascular changes produce on BOLD connectivity is especially important for clinical populations where disease and medication may alter both neural connectivity and neurovascular coupling. In this study, we examined the effect of a 200 mg caffeine dose on resting-state BOLD measures in the motor cortex. Caffeine is a commonly used neural and metabolic stimulant that readily binds to adenosine receptors, thereby competitively inhibiting adenosine activation and decreasing baseline CBF . As prior work with task-related BOLD fMRI suggests that caffeine may increase the sensitivity of the BOLD signal to stimulated neural activity, we hypothesized that caffeine would also increase the sensitivity of BOLD fluctuations to spontaneous neural activity and therefore produce an increase in resting-state BOLD connectivity. Eleven healthy volunteers participated in this study after providing informed consent. After exclusion of data from 2 subjects due to excessive motion , the sample consisted of 9 subjects . Participants were instructed to refrain from ingesting caffeine for at least 12 hours prior to being scanned. The estimated daily caffeine usage for each subject based on self-reports of coffee, tea, and soda consumption is presented in Table 2.1. The assumed caffeine contents for an 8-oz cup of coffee, an 8-oz cup of tea, and a 12-oz soda were 100 mg, 40 mg, and 20 mg respectively . Each subject participated in two imaging sessions: a caffeine session and a control session, in that order. The two imaging sessions were separated by at least 6 weeks. The caffeine session consisted of a pre-dose and a post-dose imaging section, each lasting around 45 minutes each.

Upon completion of the pre-dose section, participants ingested a 200 mg caffeine pill and then rested for approximately 30 minutes outside of the magnet before starting the post-dose section. This procedure is similar to our previous protocols using caffeine . The first resting-state scan of the post-dose section began approximately 45 minutes after the caffeine pill was ingested to achieve approximately 99% absorption of caffeine from the gastrointestinal tract . Control sessions used the same protocol, but without the administration of caffeine between sections, similar to the protocol used in . Subjects were not given a placebo during the control session. However, for convenience, we will still refer to the two scan sections as the “pre-dose” and “post-dose” sections, even though a dose was not administered. Each scan section consisted of a high-resolution anatomical scan, a bi-lateral finger tapping block design, CBF baseline and quantification scans, and two five-minute resting-state BOLD scans. Bilateral finger tapping was self-paced and the block design run consisted of 20s rest followed by 5 cycles of 30s tapping and 30s resting. Subjects were instructed to tap while a flashing checkerboard was displayed and then to rest during the display of a control image, consisting of a white square situated in the middle of a gray background. During resting-state scans, the control image was displayed for the entirety of the scan and subjects were asked to maintain attention on the white square.Imaging data were collected on a GE Signa 3 Tesla whole body system with an eight channel receive coil. Laser alignment was used to landmark subjects and minimize differences in head position between pre-dose and post-dose sections. Functional data were collected over six oblique 6-mm thick slices prescribed through the primary motor cortex. The finger tapping scan and a CBF baseline scan were acquired with a PICORE QUIPSS II arterial spin labeling sequence with dual echo spiral readout .

The two resting-state BOLD scans were acquired using BOLD-weighted imaging with spiral readout . Additional CBF quantification scans were acquired using the same in-plane parameters as the functional scans, but the number of slices was increased to ensure coverage of the lateral ventricles. These scans included a cerebrospinal fluid reference scan consisting of a single-echo, single repetition sequence , and a minimum contrast scan . The high-resolution anatomical scan was acquired with a magnetization prepared 3D fast spoiled gradient sequence . Cardiac pulse and respiratory effort data were monitored using a pulse oximeter and a respiratory effort transducer , respectively. The pulse oximeter was placed on the subject’s index finger, and the respiratory effort belt was placed around the subject’s abdomen. The pulse oximeter was not worn during the bilateral finger tapping scan. Physiological data were sampled at 40 samples per second using a multichannel data acquisition board . Images from each scan section were co-registered using AFNI software . In addition, the anatomical volume from each post-dose section was aligned to the anatomical volume of its respective pre-dose section, and the rotation and shift matrix used for this alignment was then applied to the post-dose functional images. The outer two slices of the functional data were discarded to minimize partial volume effects associated with the rotation of the post-dose data, and the first 10s of each functional run were not included. In addition,indoor vertical farming voxels from the edge of the brain were not included in the analysis in order to minimize the effects of motion. The second echo data from the finger tapping scans were used to generate BOLD activation maps of the motor cortex. This was accomplished using a general linear model approach for the analysis of ASL data . The stimulus-related regressor was produced by the convolution of the square wave stimulus pattern with a gamma density function . Constant and linear trends were included in the GLM as nuisance regressors. In addition, the data were pre-whitened using an autoregressive model of order 1 . The statistical maps were based on the square root of the F-statistic, which is equal to the t-statistic in the case of one nuisance term . Active voxels were defined using a method based on activation mapping as a percentage of local excitation . In summary, the √ F maps were separated into left and right hemispheric regions. The highest value in each region was identified and then every voxel was converted to a percentage of the peak statistical value for the region ×100. Active voxels were required to exceed an AMPLE value of 45% and a √ F value of 2 . The final activation maps were defined from the intersection of voxels active in both pre-dose and post-dose scan sections. Regions of interest were then defined for the left and right motor cortices from these activation maps. Thus, the same ROIs were used in the comparison of pre-dose and post-dose functional connectivity within an imaging session. Baseline CBF images were calculated from the average difference between the tag and control images in the CBF baseline scan. The mean ASL images were corrected for coil sensitivity and B1 field in homogeneities with the minimum contrast scan , and converted to physiological units, ml/, using the CSF reference scan . Average baseline CBF values were extracted from each subject’s motor cortex, defined as the union of the right and left motor ROIs.

The time series data from the BOLD resting-state runs were used to generate measures of resting-state functional connectivity. First, nuisance parameters were removed from the raw data through linear regression. These regressors included constant and linear trends, physiological noise terms from the measured cardiac and respiratory signals , and six motion parameters obtained during image coregistration. Data were then temporally low-pass filtered using a finite impulse response function with a cutoff frequency of 0.08 Hz. This cutoff frequency was chosen for consistency with previous functional connectivity studies . The average signal from either the left ROI or the right ROI was used to define a reference time course. This reference time course was then correlated with all other voxel time series within the brain to generate a functional connectivity map. The correlation coefficients in these maps were converted to z scores in a method similar to that used by Fox et al. . For a qualitative presentation of functional connectivity changes, average z score maps for the pre-dose and post-dose sections were calculated for each subject by averaging across the two resting-state runs and thresholding at z = 2.58 . Voxels were required to have at least 2 nearest neighbors for inclusion in the resting-state connectivity maps. Two metrics were used to quantitatively assess the strength of resting-state functional connectivity in the motor cortex. For the first metric, the mean z score was extracted from either the right or left ROI , where the average signal from the opposite ROI was used as a reference time course. The second metric, referred to as the percent overlap, was first used by Biswal etal. , and is described in more detail in these references. To summarize, the quantity ¯nLR/nL is the percent of voxels in the left ROI that are significantly correlated with all voxel time courses from the right ROI , where the subscripts R and L refer to right and left, respectively. In this method, ¯nLR/nL provides a measurement of connectivity between hemispheres, with ¯nLR/nL necessarily equal to ¯nRL/nR. These quantities, ¯nLL/nL and ¯nRR/nR , represent the percent of voxels in the left and right ROIs, respectively, that are significantly correlated with all other voxel time series from within the respective ROIs. Two-tailed paired t-tests were performed between the pre-dose and post-dose results to assess changes in these metrics of functional connectivity. In addition, a repeated measures two-way analysis of variance was performed for each metric . Also, the correlation coefficient threshold used in the percent overlap method was varied between 0.24 and 0.42 in increments of 0.02 to assess the robustness of this measurement. For frequency domain analysis, power spectra were calculated from the preprocessed resting-state data using a minimum 4-term Blackman-Harris window. Average power spectra were extracted from the motor cortex in the pre-dose and post-dose sections for each subject. A quantitative measurement of energy in the low-frequency BOLD fluctuations was calculated as the sum of the average power below 0.08 Hz. Paired t tests were performed to assess changes in energy between the pre-dose and post-dose sections. Additional processing steps were taken to determine whether changes in resting-state functional connectivity might be due to changes in low-frequency respiration variations. This was accomplished by adding an extra regressor to the original nuisance matrix. The additional term was either the global signal or the respiration volume per time signal . In order to minimize bias that can be introduced when motor cortex fluctuations are included in the regression , the global signal was extracted from the anterior portion of the brain in the four slices that were analyzed.

Table E.1 compares estimates of the effects of growth in the legal market on past month use restricting the sample to only include Montana and its bordering states in Panel A, to the estimates using the entire sample of states in Panel B. If there are substantial supply spillovers from MML states with large markets to other states, then we would expect the effect sizes in Panel A to exceed those in Panel B. From Table E.1, for youths aged 12-25, the effects of legal market size on past-month use are larger in Panel A than in Panel B, but for adults aged 26 and older they are quite similar. This is consistent with growth in the legal medical marijuana market having supply spillovers across states in the black market, where adolescents and young adults have substantially greater access than older adults. Table E.2 replicates the analysis of Table E.1 using prevalence of past-year initiation as the outcome variable. The results are similar. Thus, there appear to be supply spillovers from medical marijuana markets to recreational marijuana markets used by youths in other states. The differences between the estimates from Montana’s case study to those using the entire sample suggests that the effects of medical marijuana market growth on adolescent and young adult use may be twice as large as shown in the primary results if cross-state supply spillovers are accounted for. If the decision to report marijuana use is more closely related to beliefs about legal penalties or social disapproval compared to availability,flood drain tray then the results from Table 2.8 suggest that the effects of legal market growth on adolescent marijuana use are a true measure of consumption changes and not of reporting behavior.

Tables F.1-F.2 provide additional supporting evidence that the primary results of this paper are not driven by reporting bias. Table F.1 reports estimates for the effects of registration rates on the prevalence of past-month marijuana use by adolescents separately for the time period before the Ogden Memo and after the Cole Memo. If changes in reporting behavior are more likely to be driven by law passage than by legal market size, then registration rates should have no effect on reported past-month except due to the federal government’s memos. As evidenced in Table F.1, the coefficient estimates for adolescent past-month use are not significantly different if examined before the federal policy reduced enforcement with the Ogden Memo, or after the federal government increased enforcement with the Cole Memo. However, adolescent reporting behavior may be more sensitive to changes in risks from social or community disapproval than to changes in perceived disapproval from law enforcement. If this were the case, then changes in state marijuana policy or changes in federal enforcement policy may have less effect on adolescent reporting behavior than changes in perceived social stigma associated with cannabis consumption, which is likely highly correlated with the number of legal users and suppliers visible in the community. To address this potential concern, estimates of the effects of legal medical marijuana market size on juvenile arrests for marijuana possession are shown in Table F.2. Since adolescents for the most part do not qualify as medical marijuana patients, it is unlikely that there were significant state enforcement changes regarding juvenile arrests for marijuana-related crimes, and thus effects of legal market size on adolescent marijuana arrests are likely highly correlated with effects of legal market size on adolescent cannabis use.

Annual data on juvenile arrests from 1994-2012 were obtained from the Uniform Crime Reports County-Level Detailed Arrest Files compiled by the Inter-University Consortium for Political and Social Research. County data were aggregated up to the state level. Table F.2 reports coefficient estimates for the effect of registration rates on the juvenile marijuana possession arrest rates. In Columns -, a log-linear ordinary least-squares specification is employed, with the dependent variable constructed as the natural log of the number of juvenile arrests for marijuana possession per 100,000 of the relevant-aged population for Columns -, or the natural log of the number of juvenile marijuana possession arrests in Columns -. Columns – employ a negative binomial specification. For all model specifications, growth in the legal market size has a positive effect on juvenile arrests for marijuana possession of similar effect size to that found for the effects on adolescent past-month use. This suggests that the observed effects on self-reported use are not driven solely by changes in reporting behavior. According to estimates by Miron , state and federal expenditures on enforcement of marijuana prohibition exceed $7 billion annually. Citing these costs, hundreds of economists signed a petition in 2012 encouraging state and federal officials to rethink marijuana policy in the United States. While a growing number of states are liberalizing the use and distribution of marijuana, the federal government still maintains that prohibition is necessary to limit the costs of increased marijuana use that would occur under a legalized regime . Chapter 2 shows that growth in the size of legal medical marijuana markets significantly increases recreational use, but the welfare effects of this increased consumption alone are ambiguous.

If individuals are rational and fully anticipate the potential negative consequences of marijuana consumption on future utility, then any increase in use induced by liberalization increases consumer welfare . If, however, individuals underestimate the potential negative consequences or make mistakes in their consumption choices triggered by environmental cues, this increased marijuana use may decrease social welfare . Under any theory of individual decision-making, if marijuana use generates negative externalities, then the socially optimal level of consumption is below the individually optimal level of consumption achieved under a free market regime. This paper contributes toward understanding the potential welfare consequences of legalization by studying how growth in the legal market for medical marijuana affects traffic fatalities and deaths related to alcohol and opioid poisonings. These outcomes will reflect direct externalities from marijuana use due to impaired driving,flood and drain tray as well as indirect externalities resulting from substitution or complementarity with alcohol and opioids. Since past research has found evidence of heterogeneity by age in the elasticity of substitution between marijuana and other substances , particular attention is given to differences in these outcomes by age. In the aggregate, I find that greater medical marijuana access decreases mortality from traffic accidents and substance-related poisonings. However, the aggregate effect masks an important welfare trade-off generated by age differences in the elasticity of substitution between marijuana and alcohol. For older adults aged 45-64, increased medical marijuana availability has positive health consequences, as growth in registration rates reduces mortality related to alcohol and opioid poisonings by 7-11% and 12-16% respectively. In contrast, for youths, greater marijuana access generates negative externalities in the form of a 6% increase in traffic fatalities caused by drivers aged 15-20, with large and significant effects on alcohol-related accidents. These results are consistent with complementarity between alcohol and marijuana among youth and substitution among older adults. The paper proceeds as follows. Section 3.2 provides a brief description of medical marijuana markets in the United States, and section 3.3 discusses the literature on marijuana’s role as a substitute or complement for other additive substances. The data and empirical framework are described in sections 3.4 and 3.5 respectively. Section 3.6 presents the empirical results of the effects of growth in medical marijuana availability on traffic fatalities and substance-related poisonings. Finally, sections 3.7 and 3.8 conclude with discussion and directions for future work. The past few decades have seen a growing movement, both worldwide but particularly in the United States, away from a strict policy of marijuana prohibition. Over half of the U.S. population now live in states with medical marijuana laws , which provide legal protections for the medical use of marijuana by qualifying patients and allow for the legal supply and distribution of medical marijuana. As of 2013, in all but three of the twenty MML states, an individual who wants medical marijuana must obtain a physician’s certification that the individual has a medical condition which could benefit from marijuana use and register with the state.The share of adults registered as patients reflects the extent of medical marijuana participation in a given state and provides a measure of market penetration.

Variation in how production is regulated has led to heterogeneity in medical marijuana take-up across states, and data on state counts of registered medical marijuana patients highlights this variation. Figure 3.1 plots the percent of the state adult population registered as legal medical marijuana users in December 2014 against months since the effective date of MML enactment. As expected, the size of the market is positively correlated with the length of time the MML has been in place. However, another important determinant of market size is the strictness of regulations facing suppliers. States that allowed large-scale production with little oversight or monitoring show growth in market size above trend, while states without operational dispensaries have seen little growth.While Figure 3.1 presents a static snapshot of the current state of MML market size, the structure and size of state medical marijuana markets have undergone substantial changes over the past decade. Although several MML policies had established legal protections for large-scale producers prior to 2009, the threat of federal prosecution and product seizure served as a sufficient barrier to entry in the state-legal market. However, in October 2009, the federal government released the Ogden Memo which formally de-prioritized prosecution of users and producers in MML states who were compliant with state law . This decrease in perceived federal enforcement risk resulted in a significant shift in the structure of the medical marijuana industry. Prior to 2009, California was the only state which had seen the rise of large scale medical marijuana production collectives . However, as detailed in Table 3.1, in states where caregivers could legally produce for multiple patients, the number of operational large-scale producers increased rapidly after the Ogden Memo. For example, in Colorado, very few medical marijuana commercial operations had opened between initial MML enactment and 2009; by mid-2010, over 900 operational dispensaries were identified by law enforcement . In Montana, which did not have state-licensed dispensaries but allowed caregivers to produce for and sell marijuana to an unlimited number of patients, the number of caregivers serving ten or more patients grew from 84 in October 2009 to more than 480 by September 2010. Trends in the number of registered medical marijuana patients reflect this rapid expansion in the size of the market for medical marijuana following the Ogden Memo. Figure 3.2 depicts trends in the total number of registered patients, aggregated over all states with registration rate data that had MMLs effective prior to the Ogden Memo. Consistent with the increase in medical marijuana production, the number of registered patients spiked following the announcement of decreased federal enforcement. The flattening in trend in 2011 was driven by the federal government’s reversal in stance and re-prioritization of involvement in MML states . As documented in Chapter 2, growth in medical marijuana market size significantly increases marijuana consumption among all age groups. The welfare implications of this increased marijuana use will depend largely on whether marijuana consumption itself generates externalities, and on the extent to which marijuana is a substitute or complement to other addictive substances. While prior work has sought to address the effects of medical marijuana liberalization on traffic fatalities, alcohol consumption, and opioid use, there is little agreement as yet. Alcohol and marijuana are the drugs most frequently detected in fatally injured drivers , and the effect of marijuana legalization on drugged driving is a potential negative externality of primary concern in the current policy debate. While a large body of research has established the role of alcohol in increasing crash risk , the effects of cannabis on driving impairment are less clear. Cognitive studies show that marijuana use impairs a number of tasks associated with driving ability , but experimental research has found the effects of marijuana on driver impairment to be only modest when compared to the effects of alcohol. Still, the consumption of both alcohol and marijuana has an additive or even multiplicative effect on driver impairment , with one study finding these effects to be particularly strong during nighttime driving simulations .

By reducing the perceived risk of federal prosecution for legal producers in compliance with state law, the Ogden Memo should have increased benefits to patients by increasing medical marijuana availability. The Cole Memo should have had the opposite effect. If supply-side factors are an important determinant of the relative value of medical marijuana participation, To measure medical marijuana participation, I collected data on the number of registered medical marijuana patients for all states with mandatory registration programs as of 2014. The full listing of data sources for each state — which include direct contact with state officials, state department reports and websites , academic papers, and local news articles — is provided in Appendix A. This paper uses monthly data from 1999-2014, and Table 1.2 presents count tabulations of data availability by year and state. The measure of interest is the registration rate, calculated as the percent of the resident adult population registered as medical marijuana patients.16 As shown in Table 1.2, data availability on registered patient counts varies across states. Some states provide monthly statistics, while others collect data quarterly or annually. For states with smaller registration programs , administrative records were not made available and had to be collected from older news articles and archived web pages. For states with more developed registration systems, statistics could be found starting from the program’s inception,2×4 flood tray but the frequency of data collection increased substantially following the Ogden Memo in 2009.

For months with missing data, registered patient counts were linearly interpolated using the two closest months of data. This new dataset presents the most comprehensive state panel of medical marijuana participation made available as yet. The solid line in Figure 1.2 plots the total number of individuals registered as medical marijuana patients from 1997-2014 in states that required patient registration. As the data show, registered patient counts were relatively flat during the period of federal intervention from 1997-2008, but the Ogden Memo led to a rapid increase in medical marijuana patient participation. The spike in patient take-up coincided with significant growth in the number of legal medical marijuana producers. According to estimates by Sevigny et al. , from 2008-2010 the number of medical marijuana dispensaries increased from around 1,400 to 3,800, and the number of legal producers grew from less than 20,000 to almost 90,000. As shown in Figure 1.2, medical marijuana participation stalled following the Cole Memo. Patient registration rates resumed growth in mid-2013 when Deputy Attorney General James Cole released a second memorandum re-clarifying that federal enforcement resources should focus on large-scale marijuana operations only if they are suspected of engaging in certain criminal activities such as trafficking across states lines, distributing to minors, and supporting cartels . While the aggregate data suggest that these federal memos significantly affected trends in medical marijuana participation, the magnitude of these changes varied widely across states.

To illustrate this variation, Figure 1.3 graphs trends in adult per capita patient registration rates for states with effective registry dates prior to 2010. Some states saw exponential growth in registration rates following the Ogden Memo and declines in registered patient counts at the time of the first Cole Memo. Other states show an up-tick in patient registration with the Ogden Memo but appear to have been relatively unaffected by the Cole Memos. Finally, a few states have seen relatively flat trends in medical marijuana participation since program enactment. Summary statistics for the variables used in this paper’s regression analysis are presented in Table 1.3. Columns and show the mean and standard deviation in monthly medical marijuana registration rate data and for the other included control variables in the models. Column presents the standard deviation across state averages, such that comparing columns and indicates how much of the data variation comes from differences across versus within states. Based on the conceptual framework outlined in section 1.3, this study considers the effects of federal policy changes, state regulations, and their interactions on medical marijuana participation. Due to data limitations, I take a reduced-form approach and do not separately model eligibility, take-up conditional on eligibility, or entry and exit. then the federal memos may have influenced patient take-up through their effects on medical marijuana access. The magnitude of these effects will depend on the regulatory framework for legal production established by state MML policy. These findings suggest that the effects of the federal memos on medical marijuana suppliers was an important driver of patient registration. For Colorado, there is sufficient data to disaggregate registered patient counts by those patients with and without a designated caregiver.

For Colorado, Figure 1.4 shows that, indeed, the most substantial growth in registered patient counts was seen by patients reporting a primary caregiver as their source of marijuana; similarly, the Cole Memo led a larger reduction in registered patient counts among patients with caregivers compared to patients without caregivers. Figure 1.5 provides further evidence that interest in medical marijuana flows from producers to patients. The graph shows quarterly data for Google search interest in the phrases “how to become a patient” and “how to become a caregiver.” Data was collected from Google Trends, which measures relative search interest over time for these phrases from a sample of total searches. The spike in search interest for becoming a caregiver occurs at the time of the Ogden Memo, and it clearly precedes that of search interest in becoming a patient. This suggests that producers responded more rapidly to the announcement effects of the Ogden Memo than users, and is consistent with evidence that incentives to obtain information about a program are influenced by the expected net benefit of participating . To assess the relative role of supply and demand in driving medical marijuana patient registration, ideally one would have detailed state-level time series data on potency-adjusted marijuana prices. Unfortunately, since marijuana remains illegal at the federal level, accurate price data is highly limited. High Times is an online magazine where users can submit the price they paid for their last marijuana purchase. The magazine reports individual price submissions by city and strain of cannabis. Priceofweed.com is a website that collects user-submitted data in real-time on the price of marijuana purchases and classifies them into “high”, “medium”, or “low” quality. For completeness, I present evidence based on this crowd-sourced data, but they are intended only as suggestive evidence and should be interpreted with caution. Table 1.6 presents estimates for the effects of registration rates on the natural log of price per ounce of high-potency marijuana. For the regressions, data on high quality marijuana prices was aggregated at the state-quarter level and converted to price-per-ounce. Outlying price values were dropped.17 The results from Table 1.6 show that increases in registration rates significantly predict lower prices. This suggests that,flood and drain table even if higher medical marijuana participation rates to some extent reflect increased demand, they reflect even larger effects on supply. A number of studies have exploited state-time variation in the enactment of MMLs to estimate their effects on marijuana use in the general population. Findings have varied substantially, with estimates ranging from significantly negative, to statistically insignificant, to significantly positive for an excellent review. However, the standard difference-in-differences approach employed in these studies implicitly assumes that the “treatment effect” of MML enactment is dichotomous, i.e. the policy change occurs at a specified date, and it is implemented completely and equally across states. Whether this assumption holds will depend on the mechanisms by which MMLs induce changes in behavior. According to deterrence theory, by reducing the perceived severity of legal or informal sanctions associated with marijuana consumption, MML enactment should ceteris paribus increase demand. Since the passage of MMLs provided similar legal protections and represented a shift in either governmental or social acceptance of marijuana, ex-ante these effects should occur simultaneously with law enactment and be similar across states. This prediction relies on three conditions: that the statutory policy change is actually implemented, that no offsetting changes in enforcement occur simultaneously, and that the public is aware of the change in policy . Since MMLs provide protection from state-level but not federal prosecution, citizens may be even less likely to update their expectations about potential prosecution until it is observed or known that the federal government will not intervene.

Since awareness about laws and enforcement policies will be diffused through social networks, personal experience, and the mass media, the federal memos and their coverage by the media and marijuana advocacy groups may have had far greater effects on public perception than MML enactment alone. To provide suggestive evidence that knowledge about MMLs was limited prior to the Ogden Memo, Table 1.7 presents state-representative data on MML awareness from the National Survey of Drug Use and Health , which starting in 2002, asked respondents the following: “In your state, has marijuana been approved for medical use?” Table 1.7 reports cross-sectional variation in the percent of youths and adults who responded “yes” to this question, comparing the 2008-2009 and 2010-2011 for each state with an MML prior to 2009.18 Although these are not causal effects, they provide some useful insights. The first striking feature of Table 1.7 is the wide range of awareness across MML states. Oregon was the only state in 2008-2009 where over half of adult respondents correctly reported that the state had an MML. In contrast, less than 18% of adults in Nevada correctly responded that their state had an MML. On average, youths aged 12-17 are less aware of MML existence, but there is similar variation across states. The share of adolescents correctly reporting their state had an MML in 2008-2009 ranged from 25% in Vermont to 47% in Oregon. This variation in awareness is not explained by differences in how long the MML has been in effect. Also of note is the substantial increase in awareness of MML status following the Ogden Memo for Colorado, Montana, and Michigan. In two years, the share of adults who correctly responded that their state law allowed for the use of medical marijuana nearly doubled. These states also show the largest increase in awareness among youths. From Tables 1.1 and 1.5, these were also states with MMLs allowing caregivers to serve multiple patients and experiencing the greatest growth in medical marijuana patient participation following the Ogden Memo. The evidence from Table 1.7 suggests that the perceptual effect of state medical marijuana liberalization was relatively unrealized until after the Ogden Memo. It is thus unsurprising that studies only covering a time period prior to 2009 find insignificant effects of MML enactment on use for both adolescents and adults.19 Another mechanism by which MMLs may affect marijuana consumption is through increasing availability or decreasing prices. Research by Pacula et al. recognized that, if these are important channels through which MMLs generate spillovers, estimation needs to account for heterogeneity in the specifics of MML provisions. Accordingly, more recent studies have, in addition to using a binary measure of MML enactment, also included indicator variables for allowance for the legal operation of retail dispensaries and allowance for home cultivation by patients and/or caregivers. Still, findings have varied . While this approach is an improvement to treating MMLs as a homogeneous set of policies, it still relies on the assumptions of the DID approach and thus suffers from similar limitations. For example, including a categorical measure of “home cultivation allowance” implicitly assumes that all home cultivation laws are created equal. As Table 1.1 shows, this is clearly not the case. A binary variable for whether a state allowed patients or caregivers to cultivate would take a value of one for both Colorado and Vermont. However, a caregiver in Vermont was limited to growing for only one patient, and thus could only legally cultivate three plants; a caregiver in Colorado could grow for an unlimited number of patients, and could thus theoretically be legally protected for maintaining a large-scale grow operation with hundreds of plants. Additionally, it is unclear whether the dummy variables for any specific policy measure should “turn on” when the law is passed, when it becomes statutorily effective, or when it becomes effective “on the ground” . This is especially problematic for the measurement of dispensary legalization. As shown in Table 1.1, some states did not explicitly permit dispensaries but they did not explicitly prohibit them either.

Tobacco policies at the state, county, and city jurisdiction levels had similar degrees and patterns of co-occurrence among policies. For example, comprehensive clean-air laws for bars and comprehensive clean-air laws for restaurants frequently co-occurred at the state, county, and city levels . Most policy measures were positively correlated, but we also found pockets of negative correlations. For example, country-years with child tax credits tended not to have child tax allowances . The heat maps also revealed groups of co-occurring and independent policies. For example, labor policies requiring licensing for different professions frequently co-occurred, but this set was relatively independent of policies regarding collective bargaining and minimum wages .Most of the variability in policy measures across jurisdictions and times was explained by the other policies in the same database. Figure 4 displays the distributions of R2 values: the higher the R2, the less unique variation there is for an individual policy, to a maximum of 1.0. The impacts of policy co-occurrence on identifiability were generally substantial: of all 502 policies considered, 65% had R2 values greater than 0.90 when regressed on other policies in the same database. Child benefits had the lowest R2 distribution, with a median of 0.19; policies on poverty and social welfare, family leave, fertility/immigration, firearms, cannabis, alcohol, state tobacco control, and county tobacco control had R2 distributions with medians of approximately 0.9 or greater. In some cases,rolling benches correlations between predictor policy variables were so strong that the statistical software forced certain variables from the model .

Policy co-occurrence substantially reduced the precision of possible effect estimates in all cases . Across policy measures, databases, and simulation iterations, policy co-occurrence effectively increased the variance of effect estimates by a median of 57-fold. Across policies, the lowest degree of variance inflation observed was 7% for country child tax rebates. For other policies, particularly family leave, variance inflation was so substantial as to render estimates effectively indeterminate. Again, some predictors were dropped from models due to strong correlations with other predictors .We analyzed 13 social policy databases drawn from contemporary research in top epidemiology, clinical, and social science journals. These exemplar databases represented diverse policy domains, geographies, and times to describe the pervasiveness and impacts of policy co-occurrence on estimation of health effects. We found that high degrees of co-occurrence were the norm rather than the exception. For a majority of policies, greater than 90% of the variation across jurisdictions and times was explained by other related policies in the same database. Unbiased studies attempting to isolate individual policy effects must control for these related policies, so for many applications, there may be little independent variation left with which to study the policy of interest. Consistent with this, we found that adequate control for co-occurring policies is also likely to substantially reduce the precision of estimated effects, often so dramatically that informative effect estimates are unlikely to be derived.Several factors make the pervasiveness and consequences of policy co-occurrence likely to be even greater than we have estimated. First, we only examined policy cooccurrence within domain-specific databases.

Yet social policy changes may happen in multiple domains simultaneously. For health outcomes affected by diverse types of policies , researchers must consider policy co-occurrence across domains, which likely will indicate even more severe co-occurrence. Second, each policy database we considered included only 1 jurisdictional level, but true policy environments involve complex overlays of national, state or province, county, municipal, employer, and/or school policies. Third, we did not incorporate lagged effects or nonlinear relationships between variables. Fourth, policy variables that perfectly or near-perfectly predicted one another were dropped from the regression models. Finally, we did not consider the multitude of social, economic, political, or societal factors that may also co-occur with policies of primary interest, including changes in social norms, implementation, or enforcement that can be conflated with policy changes. Some such confounders can be controlled with jurisdiction or time fixed effects; measured confounders that are jurisdiction-specific and time-varying could be evaluated using the same methods illustrated here. This is a formidable task; data sharing efforts would facilitate its assessment and handling. We found that the overall degree of policy co-occurrence varied across databases, ranging from very high for state level recreational cannabis policies to low for country-level sexual minority rights policies. Several different factors may drive this variation. Our finding that tobacco policies at the state, county, and city levels had similar degrees and patterns of co-occurrence among similar sets of policies suggests that co-occurrence may be a characteristic of the domain. Political polarization may result in greater co-occurrence for certain policy domains versus others . Databases with rarer policies, fewer umbrella policies , or more nested policies also tended to have more co-occurrence. Databases with more unique policies also generally had more cooccurrence; with a fixed number of jurisdictions and times of observation, considering more policies creates more opportunities for alignment.

Importantly, these patterns highlight that the measured degree of co-occurrence depends not only on the policies themselves but also on the investigator’s choices of policy measures. Furthermore, policies that could be considered alternatives rather than complements co-occurred less frequently and may offer the opportunity for more robust studies of causal impacts. Differences in the ways that policies are adopted across different political systems and different jurisdictional levels may also matter. In our examples, country-level policies appeared to co-occur less frequently than state-level policies, implying that estimating causal effects of country-level policies may be more feasible. Similar considerations apply to the temporal scale of analyses as well: The feasibility of estimating health effects may depend on whether analyses are conducted at the level of the year, month, or even election cycle. Our analysis could not determine which of these factors drives variation in policy cooccurrence; this would be a fruitful area for future research.Several other limitations of this study must be noted. Certain policy domains were not covered, either because no social policy studies for that domain were sampled or because no corresponding policy database was identified or accessed. We did not review all potentially relevant journals. Our results may not generalize to policy domains or journals not examined. Our approach also assumes that all the policies in each domain-specific database are relevant to the health outcome of interest; this is plausible for social interventions that likely affect a broad range of health outcomes, but for some outcomes, only a subset of the policies in a database may need to be controlled to isolate the effect of the index policy. In addition, our approach is only relevant when a database of the relevant policies exists or can be constructed. Developing policy databases is often an arduous task requiring systematic review of legal language. We did not consider the quality of the underlying databases. Our selections serve to illustrate the policy co-occurrence problem, but for applied researchers, the optimal policy database may differ from the one used here. The problem of correlated exposures arises in many domains, including environmental health, and although social policies are distinct in important regards, methods in other domains may nonetheless prove helpful. Furthermore, our analysis did not examine the distinctions between policy adoption, implementation, promulgation, or changes in norms that precede or follow from policy changes,grow tray but these considerations are essential in applied policy research. Finally, data sparsity arising from co-occurring policies can lead to bias, not just imprecision. Our simulations did not incorporate this because this type of bias is less relevant to studies of the health effects of social policies and is highly context specific. Simulation results on the magnitude of bias from positivity violations are therefore unlikely to be generalizable. Specifically, bias arising from positivity problems depends on the estimation method. For methods that rely on modeling the outcome , positivity-related bias arises from model-based extrapolation. For methods that involve modeling the exposure mechanism , bias can result from disproportionate reliance on the experiences of a just a few units or on the absence of certain confounder strata . Because our simulations were based on outcome regressions—the most common approach for differences-in differences, panel fixed effects, and related designs—bias would only arise from model-based extrapolation. However, for the vast majority of policies identified in this study, measures were binary, and thus extrapolation cannot occur. For continuous policy measures , model-based extrapolation is possible but application dependent. Thus, simulating the potential degrees of bias resulting from model-based extrapolation requires either tenuous generalizations or substantive assumptions about each policy area. We suspect that extremely nonlinear relationships that would lead to large extrapolation bias are rare for policy effects, but this remains an open question.Researchers should be cautious when seeking to make causal inferences about the health effects of single social policies using methodological approaches premised on arbitrary or quasi-random variation in policies across jurisdictions and time. Not every policy change offers a valid differences-in-differences or panel fixed-effects study design. These methods are most compelling when policy implementation is staggered across jurisdictions and dates independently from other policies and for plausibly like-random or arbitrary reasons. For example, there could be differing timing of elections, legislative sessions, crises that provoke specific policy changes, or lottery-type roll outs. In these cases, such research can be very persuasive, or at least constrain the set of co-occurring policies.

Our finding of pervasive policy co-occurrence across numerous databases suggests that many policies do not fit this criterion. Inadequate control for co-occurring policies or differences in the set of policies controlled may explain surprising or conflicting results in previous studies. Investigators should base interpretations of social policy research on the plausibility that policy adoption is distributed arbitrarily with respect to other uncontrolled policies or social changes, a phenomenon that, in reality, may be rare. This evaluation should be based on deep content knowledge of law, politics,and society—a compelling argument for involving policymakers in the design and interpretation of studies.We illustrate an approach for researchers to assess whether the effects of individual policies can be estimated. Although other simulation-based methods for assessing positivity exist , the approach we propose is tailored to the policy co-occurrence problem and facilitates examining how a full set of policies substantively occur together. For a given application, if the heat map indicates high correlations, and estimated R2 values and variance inflation are high, it may be necessary to alter the research question and corresponding analytic approach. Researchers have applied numerous analytic approaches to address the challenge of highly co-occurring policies, ranging from machine-learning algorithms that identify policy measures most strongly related to an outcome of interest to methods that characterize overall policy environments based on expert panels. The second article in this issue on policy co-occurrence provides a systematic assessment of available methods. We briefly review 3 promising analytic options here, and refer the reader to the other article for more detail. One approach is to focus on outcomes that are affected by the index policy of interest but not the co-occurring policies. For example, changes in state Earned Income Tax Credits co-occur with changes in other social welfare policies . Rehkopf et al. took advantage of seasonality in the disbursement of EITC cash benefits versus benefits without the same seasonal dispersal pattern, to examine the association of EITC with health using a differences-in differences design. By comparing health outcomes that can change monthly for EITC-eligible versus non-eligible individuals in months of income supplementation versus non-supplementation, the authors measured potential short-term impacts of EITC independent of other social welfare policies. Another approach is to move beyond binary measures of policy adoption to more detailed characterizations . These measures may co-occur less frequently with related policies or provide opportunities to examine dose– response effects among jurisdictions adopting the policy. For example, the adoption of certain unemployment benefit policies co-occurs frequently with other social welfare policies across state-years. Researchers have successfully assessed these policies’ health impacts by comparing varying levels of unemployment benefit generosity—measured as the total maximum allowable benefit per bout of unemployment—across states and years . Heatmaps like those presented in this study may help researchers identify specific policy measures that are more independent from related policies. A final option is to conceptualize policy clusters, instead of individual policies, as exposures. This is promising if policies are typically adopted as a group, as is the case with the large omnibus bills that are increasingly common at the state and federal levels.