OREGON MARIJUANA LEGALIZATION Oregon ...

1

RUNNING HEAD: OREGON MARIJUANA LEGALIZATION

Oregon recreational marijuana legalization: Changes in undergraduates’ marijuana use rates from 2008-2016

David C. R. Kerr Harold Bae Andrew L. Koval Oregon State University

Author Note David C. R. Kerr, School of Psychological Science, Oregon State University; Harold Bae, College of Public Health and Human Sciences, Oregon State University; Andrew L. Koval, College of Public Health and Human Sciences, Oregon State University. Correspondence can be directed to [email protected], 213 Reed Lodge, School of Psychological Science, Oregon State University, Corvallis, OR 97331. The study was unfunded and the authors have no conflicts to declare. The authors have not presented these data or narrative interpretations of these data in any previous publications or presentations, or on any websites. The authors wish to thank Mary Hoban for data consultation and assistance.

Kerr, D.C.R., Bae, H., & Koval, A.L. (2018). Oregon recreational marijuana legalization: Changes in undergraduates’ marijuana use rates from 2008-2016. Psychology of Addictive Behaviors. © 2018, American Psychological Association. This paper is not the copy of record and may not exactly replicate the final, authoritative version of the article. Please do not copy or cite without author permission. The final article will be available, upon publication, via its DOI: 10.1037/adb0000385

2

Abstract There have been few studies of marijuana use before and after recreational marijuana legalization (RML) in affected states. We tested whether marijuana use rates increased more among college students in Oregon than in non-RML states following Oregon RML (July, 2015). Repeated cross-sectional National College Health Assessment-II surveys were administered to random samples of students within participating colleges from 2008-2016. Data were from fouryear institutions that participated both before and after Oregon RML. Undergraduates (ages 1826 years) from two institutions in Oregon (n=7,412) and 123 institutions (n=274,340) in nonRML U.S. states self-reported use of marijuana, tobacco, alcohol, and other drugs in the past 30 days. Mixed effects regressions accounted for clustering of participants within institutions and controlled for individual-, contextual-, and institution-level factors as well as secular changes in substance use rates from 2008 to 2016. Following RML Oregon students (compared to non-RML state students) showed relative increases in rates of marijuana use [Odds Ratio (95% Confidence Interval) = 1.29 (1.13-1.48), p = .0002] and decreases in tobacco use rates [OR (95% CI) = 0.71 (0.60-0.85), p < .0001]. Changes in marijuana use after RML did not differ significantly for participants under and over age 21 years. Some study limitations would be addressed with higher survey response rates, inclusion of other Oregon institutions, and controls for marijuana and other substance policies. Still, findings are consistent with an effect of RML on rates of marijuana use among young adult college students, which may have important public health implications. Key words: cannabis, recreational marijuana legalization, early adulthood, Oregon

3

Oregon recreational marijuana legalization: Changes in undergraduates’ marijuana use rates from 2008-2016 The effects of U.S. states’ recreational marijuana legalization (RML) on substance use rates are not well established. Given that marijuana use increases risks for substance use disorders and other health and adjustment problems (Volkow et al., 2014), there may be adverse public health effects if RML increases the prevalence of marijuana use and the frequency of use among users. Young people may be especially sensitive to RML. For example, rates of marijuana use increased among 8th and 10th graders in Washington following RML compared to peers in non-RML states (Cerdá et al., 2016). Surprisingly, rates did not change after RML for Colorado youth or Washington 12th graders. Still, adolescents have shown decreased perceptions of harm from marijuana and increases in favorable attitudes about use after RML (Cerdá et al., 2016; Fleming et al., 2016; University of Colorado, 2015), which may increase onset risk in later adolescence and early adulthood. Young adults may be directly impacted by RML. In early adulthood substance use rates peak (Schulenberg et al., 2017) and 50% of substance use disorder cases show onset (Kessler et al., 2005). Marijuana experimentation, onset of more regular use, and the establishment of chronic, problematic use commonly occur around the transition to adulthood (Fromme et al., 2008), likely due to a convergence of psychosocial changes including increased autonomy (Furstenberg et al., 2003), changes in risk and protective factors such as exposure to peer use and reduced parental monitoring (Napper et al., 2015), and ongoing brain maturation (Sowell et al., 1999; 2001). For young adults attending college, additional behavioral, social, and cultural transitions occur that may influence substance use onset and escalation (Arria et al., 2015), including having more unstructured time and associating with new peers in new situations (e.g.,

4

large parties, bars) within a cultural context that encourages excessive substance use (Dennhardt et al., 2013). In a prior study of college students, greater increases were found in rates of marijuana use following RML at an Oregon university compared to six institutions in non-RML states (Kerr et al., 2017); however, RML effects were only observed among respondents who had reported recent heavy alcohol use. There were no significant RML effects on tobacco or heavy alcohol use, and other drug use was not examined. Another study found that following Washington’s RML marijuana use rates increased more among students at a large public university in that state than was predicted from national trends in use rates (Miller et al., 2017). Given the need for additional studies we tested three research questions in a national sample of college students surveyed cross-sectionally before and after Oregon’s RML. First, we predicted that prevalence and frequency of marijuana use in the past 30 days increased more in Oregon from before to after RML than in states without RML across the same time period. Second, to evaluate whether changes were specific to marijuana use, we also tested whether Oregon students’ rates of 30-day alcohol, tobacco, and other drug use also increased after RML. Third, we tested interactions between RML and two factors that have been associated with stronger increases in marijuana use following RML: reporting recent heavy alcohol use and being younger than the legal drinking/using age of 21 (Kerr et al., 2017; Miller et al., 2017). Method Participants Data were from repeated cross-sectional National College Health Assessment II (NCHA II; American College Health Association (ACHA), 2016) surveys administered at post-secondary institutions across eight academic years [(AY) 2008-2016] (ACHA, n.d.a.; Rahn et al., 2016). Institutions secured approvals from their campus IRB and selected the administration period (fall

5

or spring); most chose spring quarter/semester and many participated biennially, but the particular years of participation differed by institution. We selected data from institutions that used random sampling (ACHA, 2010); however, sample weights were not available. The mean student response proportions from 2008-2016 were low, ranging from 15-36% (mean = 25.6%). However, in 2015-2016, substance use rates from NCHA-II were comparable to those from studies of representative samples of college students (see Supplemental Material). ACHA data use policies and agreements with participating institutions did not permit ACHA to release institutional identities (although de-identified institution codes permitted clustering of responses within institutions in data analyses) or state codes [precluding our coding of states’ Medical Marijuana Legalization (MML) status]. We contacted Oregon institutions for permission to use their data; two public universities agreed, both of which administered the survey every other year from AY2009-10 to 2015-16. ACHA staff then facilitated data transfer, removing data from all other Oregon institutions and institutions in states that had passed RML by Spring 2016 (Washington, Colorado, Alaska, and Washington D.C.). The merged dataset included 919,639 participants from 585 unique institutions. Exclusions. We then selected data from 4-year institutions that participated both prior to [(2008-2015 AY] and after (i.e., 2015-2016 AY) Oregon RML went into effect (7/1/2015), and collected data on the covariates. We further limited the sample to undergraduates aged 18 to 26 years. The final analytic sample included data from the two institutions in Oregon (n = 7,412), and 123 institutions (n = 274,340) in non-RML states 1. Figure 1 shows a timeline of RML events

1

There was some question about whether study validity would be improved by restricting the non-RML comparison group to public universities (73 institutions; total n=194,797), or institutions on the same survey schedule as the two Oregon schools (43 institutions; n=134,572), or both (29 institutions; n=103,150). However, models based on these three sample restrictions (RML odds ratios = 1.29-1.35, p < .001) yielded identical conclusions as the primary model (RML OR = 1.29, p = .0002). Thus, we used the full sample.

6

in relation to NCHA-II survey administration periods. The two Oregon institutions administered the last pre-RML survey in Spring 2014 (AY 2013-14), significantly pre-dating passage of the RML ballot measure in Oregon (November, 2014) and the opening of retail marijuana stores in neighboring Washington state (July, 2014; see Hansen et al., 2017). Both Oregon institutions administered the post-RML survey in Spring 2016 (AY 2015-16), thus post-dating initial legalization (i.e., of possession and use; 7/1/2015) and the start of legal sales of recreational marijuana from retail stores in Oregon (10/1/2015). Measures Substance Use. The primary outcome variable was 30-day marijuana use. Secondary outcome variables (all 30-day) were tobacco use (including cigarettes, tobacco from a water pipe, cigars and clove cigarettes, and smokeless tobacco; e-cigarette use also was included but was first measured in AY 2015-16), alcohol use, and illicit drug use (composite of cocaine, methamphetamine, other amphetamines, sedatives, hallucinogens, anabolic steroids, opiates, inhalants, MDMA, other club drugs, and other illegal drugs). Use was defined as having used (1) or not (0) in the past 30 days. Frequency of marijuana use in the past 30 days was coded as none, 1-2, 3-5, 6-9, 10-19, or ≥20 use days. Recreational Marijuana Legalization (RML). The RML variable indicated whether (1) or not (0) participants were attending an institution in Oregon in Spring 2016. Demographic and Other Covariates. Age was dichotomized to represent legal age of drinking (and marijuana use after RML); participants were coded as ≥21 years (1) or minor (0) at the time of the survey. We coded gender as female (referent), male, or transgender/other, and sexual orientation as heterosexual (referent), bisexual, gay/lesbian, or other. Race/ethnicity included five mutually exclusive categories: white (referent), Asian, Black/African American,

7

Hispanic, and other race/ethnicity including multiracial. Dichotomous codes were created for international (1) or domestic (0) student, first year (1) or other year (0) student, and membership in a fraternity/sorority (1) or not (0). Residence types were categorized as: on-campus/university housing (referent), off-campus housing, parent’s home, Greek (e.g., fraternity), and “other.” The relationship status variable was coded as not in a relationship (referent), in a relationship and living together, or in a relationship but not living together. The survey included several institution-level covariates: survey administration season [fall (1) vs. spring (0)]; enrollment size [categorized into: 20,000 students]; population of the city/town in which the institution is situated [four categories: 500,000]; whether the institution is private (0) or public (1); and U.S. region [West (referent), Midwest, Northeast, South]. Statistical Analysis We used a mixed effects logistic regression model of the following general form: 𝑦𝑦𝑖𝑖𝑖𝑖𝑖𝑖 = 𝑚𝑚(𝛽𝛽0 + 𝛽𝛽1 𝑅𝑅𝑅𝑅𝑅𝑅𝑠𝑠𝑠𝑠 + 𝑋𝑋𝛽𝛽2 + 𝛾𝛾𝑡𝑡 + 𝛼𝛼𝑠𝑠 + 𝜀𝜀𝑖𝑖𝑖𝑖𝑖𝑖 )

, where 𝑖𝑖 denotes an individual, 𝑠𝑠 denotes the institution, and 𝑡𝑡 denotes the academic year (AY).

𝑦𝑦𝑖𝑖𝑖𝑖𝑖𝑖 represents the substance use outcomes for each participant. 𝑅𝑅𝑅𝑅𝑅𝑅𝑠𝑠𝑠𝑠 is the indicator for the

implementation of RML in institution 𝑠𝑠 during year 𝑡𝑡. 𝑋𝑋 is a design matrix for individual- and institution-level covariates. 𝛾𝛾𝑡𝑡 represents a fixed effect for each AY, which captures the time

effect or secular trend (not constrained to be linear); AY 2014-15 was used as the referent year 2.

𝛼𝛼𝑠𝑠 represents a random intercept per each unique institution, which accounts for possible

correlations among the students in the same institution. 𝜀𝜀𝑖𝑖𝑖𝑖𝑖𝑖 represents an error term. 𝑚𝑚(∙) is the 2

The choice of the referent year does not alter the interpretation of the effect of RML, as the role of the AY variable is to estimate the overall secular trend in use among all schools in our data. Institutions did not need to participate in the referent year to be included in the model.

8

function that relates the outcome variable to independent variables. In this model, the primary statistical inference was on the indicator variable for RML exposure; under the null hypothesis, we expect the odds ratio (OR) for RML to be 1. Interaction terms were added to the above model to evaluate whether the effect of RML differed across strata of the moderator variables. When frequency of 30-day marijuana use was modeled as an outcome variable, ordinal logistic regression was used with the cumulative logit link function. Analyses were performed using SAS Studio (release 3.5; build date: Feb 3, 2016). An alpha level of .05 (2-tailed) was used. Results Descriptive statistics and crude prevalence trends Descriptive statistics for analytic variables are summarized in Table 1, stratified by the Oregon and non-RML state institutions. Figure 2 illustrates crude prevalence trends over time for institutions in Oregon and non-RML states; prevalence rates of substance use were generally higher at the Oregon institutions. Opiate use rates were too low for inferential models (crude prevalence rates across years were 0.31% to 0.56% in Oregon and 0.32% to 0.48% in non-RML states). Multiple regressions RML effects on marijuana use. Consistent with the primary hypothesis, the logistic regression (Table 2) showed a positive association between RML and 30-day marijuana use (adjusted OR = 1.29, p = 0.0002). This RML effect was evident when controlling for significant time effects (p < 0.0001; omnibus test for AY) that indicated a national trend of increasing marijuana use over the 8-year period. Thus, students exposed to RML (i.e., Oregon in AY 201516) had 29% increased odds of marijuana use compared to unexposed students while accounting for overall cross-institution trends toward increased use, covariates, and potential correlation

9

among students within each institution. This RML effect is equivalent to a 23% relative increase (i.e. relative risk of 1.23) from predicted marijuana use probability (i.e. back transformed from the adjusted logistic model) among students unexposed to RML. A follow-up regression was run within Oregon institutions only (AY08-09, AY10-11, AY12-13, AY14-15, season, enrollment size, population size, public/private, and region had zero variance and were dropped). Compared to AY13-14 (the referent here) there were significantly higher rates of marijuana use by Oregon students in AY15-16 (i.e., after RML; OR = 1.49, p < .0001), but not in prior administrations (AY09-10 and AY11-12; OR = 1.06 and 1.11, p ≥ .20). RML effects on frequency of marijuana use. Crude prevalence trends for different categories of frequency of 30-day marijuana use suggested greater relative increases following implementation of RML in the more sporadic use categories (1-2 days and 3-5 days) at the Oregon versus non-RML state institutions (see Table 3, and Supplemental Materials). The ordinal logistic regression model (see Supplemental Material) also yielded a significant, positive association between RML and frequency of marijuana use (OR = 1.25, p = 0.0008), consistent with the hypothesis. Compared to other students, those exposed to RML had 25% increased odds of being in the higher frequency category (vs. lower frequency category). RML effects on other substance use. Next we explored RML effects on other substance use rates in the same manner. A significant RML effect indicating decreased tobacco use (OR = 0.71, p < 0.0001) was observed in the presence of a significant time effect (p < 0.0001; omnibus test for AY), indicating a decreasing secular trend of tobacco use. This RML effect is equivalent to 23% relative decrease (i.e. relative risk of 0.77) from predicted tobacco use probability among students unexposed to RML. The crude prevalence rates of frequencies of any tobacco use (based on the form used most often) are shown in the Supplemental Materials, and suggest the

10

relative decreases after RML were limited to less frequent tobacco use patterns (1-9 days in the past month) at the Oregon versus non-RML state institutions, and not to daily use. For alcohol and illicit drug use, no significant RML effects or clear time trends were observed. Table 4 summarizes the adjusted effects of RML and AY for marijuana use (repeated from Table 2) and all other outcome substances. Moderation of RML. Moderators were tested in separate models; none of the interaction terms with RML were significant. Thus, contrary to prior studies, RML effects did not depend on whether respondents reported heavy alcohol use, any 30-day alcohol use (a broader replication test), or were the legal drinking/using age. Missingness on marijuana use variable. RML may have influenced whether individuals were willing to answer the question about marijuana use. At Oregon institutions, missingness rates on the marijuana use question were 0.86%, 0.92%, 0.52%, 0.70% in AY09-10, AY11-12, AY13-14, and AY15-16, respectively. At non-RML institutions, missingness rates increased every year from 0.44% in AY 2008-09 to 0.61% in AY 2013-14 and then were 0.81% and 0.73% in AY 2014-15 and 2015-16, respectively. Thus, patterns were not suggestive of relative changes in willingness to answer the survey item on marijuana use after RML. Discussion Reported marijuana use rates increased more following RML among young adults attending college in Oregon than in non-RML states. Relative increases in crude prevalence of marijuana use frequencies among Oregon students from pre- to post-RML suggested increases in sporadic use (e.g., 1-5 days in the past month), rather than heavier use, as rates did not appear to increase for use on 10 or more days in the past month. Increases in use also were specific to

11

marijuana, as tobacco use rates showed relative decreases following RML, and alcohol and other drug use rates did not show significant relative changes. Overall, findings add to the small number of controlled studies of RML in relation to marijuana use by young people (e.g., Cerdá et al., 2016; Kerr et al., 2017). The present design cannot establish whether RML caused the observed changes in Oregon use rates. One might speculate that RML reduced students’ concerns about the criminal consequences of buying and using marijuana, and made it easier for them to buy the drug at one of the many retail stores. However, the observed RML effect was no weaker for students ages 18-20 for whom purchase and use remains illegal in Oregon. A similar finding was reported in Washington (Miller et al., 2017), whereas another study found that increases in undergraduates’ marijuana use after Oregon RML were stronger for minors than other young adults (Kerr et al., 2017). Such findings may suggest RML has increased rates of two illegal behaviors—use by and supply to minors—which may be of interest to voters and policy makers who likely envisioned a more limited population impact. Additionally, increased marijuana use by those under 21 may be concerning, given evidence for the adverse effects of earlier onset and use (Gruber et al., 2014). Another interpretation of the main finding is that after RML students were no more likely to use marijuana, but instead became willing to admit a previously illegal, stigmatized behavior. On the other hand, minors’ marijuana use also increased in Oregon following RML, even though there was no change in the legality of use for them. Additionally, trends in missingness did not support RML-related changes in students’ willingness to answer a question about use. We detected increases in rates of reported marijuana use following RML in the context of three important trends. First, Oregon rates were higher than rates in other states even before RML. This parallels findings on adolescents’ use and attitudes in MML and non-MML states

12

(Wall et al., 2011). Second, consistent with national trends (Schulenberg et al., 2017) marijuana use rates were increasing both in Oregon and non-RML states before Oregon’s RML. Third, numerous demographic and contextual characteristics that vary by region or institution were powerful independent predictors of marijuana use (e.g., relationship status, Greek system involvement, and being homosexual, White or Black, male or transgender, or a first year or domestic student). Thus, research designs using pre/post measures, relevant comparison conditions, and multiple covariates are needed to evaluate RML effects. Limitations of the present study included low survey participation rates and the lack of sampling weights despite use of random sampling. Of note, however, NCHA-II respondents’ substance use rates were highly comparable to those reported in representative samples of college students (see Supplemental Material). Regarding external validity concerns: findings on Oregon’s RML may not generalize to other states’ RML (Hunt and Miles, 2015); the Oregon students may not be representative of other Oregon or RML state students, given the variation in policies within and between RML states (Hunt and Miles, 2015); and findings may not apply to young adults who are not in college (Schulenberg, 2017). Given that state identifiers were masked in these data, we also were not able to model other policy circumstances that may have impacted substance use. Such policies include: 1) MML, which was legal in Oregon across the years considered here, but shaped RML implementation and markets in Oregon, and may have influenced other MML state students’ use; 2) alcohol or tobacco control or pricing (Grucza et al., 2015; Williams et al., 2004); 3) the 2014 opening of legal retail marijuana markets in Washington which affected use in Oregon (Hansen et al., 2017); 4), the 2015 expansion of the Oregon Indoor Clean Air Act to prohibit the indoor public use of inhalant delivery systems (e.g., vaping); and 5) myriad local factors (e.g., growers; community sales policy) found to be relevant

13

to use by adolescents (Paschall et al., 2017; Rusby et al., 2018). Having location and policy data would have permitted the identification of a more optimal control group, such as non-RML states that were more similar to Oregon before RML. Finally, tobacco use trends must be interpreted with caution as measures changed in 2015-16 to include e-cigarettes, and models could not parse out effects of tobacco policies that may have coincided with RML. In conclusion, Oregon’s RML was associated with increased rates of reported 30-day marijuana use and frequency of use among young adult college students, regardless of whether they were old enough to use or buy the drug legally. The increases were specific to marijuana use. Future studies must account for a complex web of regional and subpopulation differences, policy heterogeneity (Pacula et al., 2015), and secular changes in substance use, and should test if and how RML effects change over time (e.g. novelty effects). Studies also should test RML effects on outcomes that are linked to substance abuse and are of particular relevance to young people, such as onset of patterned substance use, motor vehicle accidents, sexual assault, and academic functioning. Finally, substitution (e.g., if marijuana use replaces binge drinking) and facilitation (e.g., marijuana as a “gateway drug”; or co-use of marijuana and tobacco) effects are of critical interest given that non-marijuana substance use is strongly implicated in many leading causes of death, injury, and impairment (Anderson et al., 2013; CDC, 2008; Hall, 2015; 2016; Kolves et al., 2006; Neal and Fromme, 2007).

14

References American College Health Association (2010). American College Health Association-National College Health Assessment Spring 2008 Reference Group Data Report (Abridged). Journal of American College Health, 57, 477-88. American College Health Association (2016). American College Health Association-National College Health Assessment II: Reference Group Executive Summary Spring 2016. Hanover, MD: American College Health Association. American College Health Association (n.d.a.). About ACHA-NCHA: An overview. Retrieved from http://www.acha-ncha.org/overview.html. Anderson, D.M., Hansen, B., Rees, D.I., 2013. Medical marijuana laws, traffic fatalities, and alcohol consumption. J Law Econ, 56, 333-369. Anderson, D.M., Rees, D.L. (2014). The legalization of recreational marijuana: how likely is the worst-case scenario? J Policy Anal Manage, 33, 221-232. Arria, A.M., Caldeira, K.M., Bugbee, B.A., Vincent, K.B., O'Grady, K.E. (2015). Marijuana use trajectories during college predict health outcomes nine years post-matriculation. Drug Alcohol Depend, 159, 158-165. Centers for Disease Control and Prevention (CDC), 2008. Smoking-attributable mortality, years of potential life lost, and productivity losses--United States, 2000-2004. MMWR 57, 1226. Cerdá, M., Wall, M., Feng, T., et al. (2016). Association of state recreational marijuana laws with adolescent marijuana use. JAMA Pediatrics, 171, 142-149. Dennhardt, A.A., Murphy, J.G. (2013). Prevention and treatment of college student drug use: a review of the literature. Addict Behaviors, 38, 2607-2618.

15

Fleming, C.B., Guttmannova, K., Cambron, C., Rhew, I.C., Oesterle, S. (2016). Examination of the divergence in trends for adolescent marijuana use and marijuana-specific risk factors in Washington State. J Adolesc Health, 59, 269-275. Fromme, K., Corbin, W.R., Kruse, M.I. (2008). Behavioral risks during the transition from high school to college. Developmental Psychology, 44, 1497-1504. Furstenberg Jr., F.F., Kennedy, S., McCloyd, V.C., Rumbaut, R.G., Settersten Jr., R.A. (2003). Between adolescence and adulthood: expectations about the timing of adulthood. Network on Transitions to Adulthood and Public Policy Research Network Working Paper No. 1. http://transitions.s410.sureserver.com/wp-content/uploads/2011/08/between.pdf. 2003. Gruber, S.A., Dahlgren, M.K., Sagar, K.A., Gönenç, A., Lukas, S.E. (2014). Worth the wait: effects of age of onset of marijuana use on white matter and impulsivity. Psychopharmacology, 231, 1455-1465. Grucza, R.A., Hur, M., Agrawal, A., Krauss, M.J., Plunk, A.D., Cavazos-Rehg, P.A., et al. (2015). A reexamination of medical marijuana policies in relation to suicide risk. Drug and Alcohol Dependence, 152, 68-72. Hall, W. (2015). What has research over the past two decades revealed about the adverse health effects of recreational cannabis use? Addiction, 110, 19-35. Hall, W. (2016). Alcohol and cannabis: comparing their adverse health effects and regulatory regimes. Int J Drug Policy, 42, 57–62. Hansen, B., Miller, K., & Weber, C. (2017). How extensive is inter-state diversion of recreational marijuana? (No. w23762). National Bureau of Economic Research. Hunt, P.E., Miles, J., 2015. The impact of legalizing and regulating weed: issues with study design and emerging findings in the USA. Curr Top Behav Neuroscience.

16

Kerr, D.C.R., Bae, H., Phibbs, S., Kern, A.C. (2017). Changes in undergraduates’ marijuana, heavy alcohol, and cigarette use following legalization of recreational marijuana use in Oregon. Addiction. Kerr, D.C.R., Owen, L.D., Capaldi, D.M. (2011). The timing of smoking onset, prolonged abstinence and relapse in men: a prospective study from ages 18 to 32 years. Addiction, 106, 2031-2038. Kessler, R.C., Berglund, P., Demler, O., Jin, R., Merikangas, K.R., Walters, E.E. (2005). Lifetime prevalence and age-of-onset distributions of DSM-IV disorders in the National Comorbidity Survey replication. Arch Gen Psychiatry, 62, 593-602. Kolves, K., Värnik, A., Tooding, L.M., Wasserman, D. (2006). The role of alcohol in suicide: a case-control psychological autopsy study. Psychol Med, 36, 923-930. Miller, A.M., Rosenman, R., & Cowan, B.W. (2017). Recreational marijuana legalization and college student use: early evidence. SSM - Population Health, 3, 649-657. Napper, L.E., Hummer, J.F., Chithambo, T.P., LaBrie, J.W. (2015). Perceived parent and peer marijuana norms: the moderating effect of parental monitoring during college. Prev Sci, 16, 364-373. Neal, D.J., Fromme, K. (2007). Event-level covariation of alcohol intoxication and behavioral risks during the first year of college. J Consult Clin Psychol, 75, 294-306. Pacula, R.L., Powell, D., Heaton, P., Sevigny, E L. (2015). Assessing the effects of medical marijuana laws on marijuana use: the devil is in the details. J Policy Anal Manage, 34, 7-31. Paschall, M. J., Grube, J. W., & Biglan, A. (2017). Medical marijuana legalization and marijuana use among youth in Oregon. The Journal of Primary Prevention, 38, 329-341.

17

Rahn, R.N., Pruitt, B., Goodson, P. (2016). Utilization and limitations of the American College Health Association’s National College Health Assessment instrument: a systematic review. J Am Coll Health, 64, 214-37. Rusby, J. C., Westling, E., Crowley, R., & Light, J. M. (2018). Legalization of recreational marijuana and community sales policy in Oregon: Impact on adolescent willingness and intent to use, parent use, and adolescent use. Psychology of Addictive Behaviors, 32, 84. Schulenberg, J.E., Johnston, L.D., O’Malley, P.M., Bachman, J.G., Miech, R.A. & Patrick, M.E. (2017). Monitoring the Future national survey results on drug use, 1975–2016: Volume II, College students and adults ages 19–55. Ann Arbor: Institute for Social Research, The University of Michigan. Sowell, E.R., Thompson, P.M., Holmes, C.J., Jernigan, T.L., Toga, A.W. (1999). In vivo evidence for post-adolescent brain maturation in frontal and striatal regions. Nat Neurosci, 2, 859-861. Sowell, E.R., Thompson, P.M., Tessner, K.D., Toga, A.W. (2001). Mapping continued brain growth and gray matter density reduction in dorsal frontal cortex: inverse relationships during postadolescent brain maturation. J Neurosci, 21, 8819-8829. University of Colorado Denver Community Epidemiology and Program Evaluation Group, 2015. Healthy Kids Colorado Survey 2015 [internet]. Home. Available at: https://www.colorado.gov/pacific/sites/default/files/PF_Youth_HKCS-2015-ExecSummary.pdf (accessed 22 June 2017) (Archived at http://www.webcitation.org/6rPtkrVNC). Volkow, N.D., Baler, R.D., Compton, W.M., Weiss, S.R. (2014). Adverse health effects of marijuana use. N Engl J Med, 370, 2219-2227.

18

Wall, M.M., Poh, E., Cerdá, M., Keyes, K.M., Galea, S., & Hasin, D.S. (2011). Adolescent marijuana use from 2002 to 2008: higher in states with medical marijuana laws, cause still unclear. Annals of Epidemiology, 21, 714-716. Williams, J., Pacula, R.L., Chaloupka, F.J., Wechsler, H. (2004). Alcohol and marijuana use among college students: economic complements or substitutes? Health Econ, 13, 825–843.

19

Table 1. Descriptive Statistics for Variables in non-RML schools and Oregon schools, collapsed across all academic years of survey participation.

N No. of schools Gender (%) Female Male Transgender Sexual Orientation (%) Heterosexual Bisexual Gay/Lesbian Others Race/Ethnicity (%) White Asian Black Hispanic Others First Year (%) Legal Age (% ≥ 21 years) Greek membership (%) International Student (%) Relationship Status (%) Not in a relationship In a relationship & living together In a relationship & Not living together Residence Type (%) University housing Off-campus housing Parents' home Greek house Others Season of Survey Administration (Fall) (%) Enrollment Size (%) > 20,000

non-RML Schools 274,340 123

Oregon Schools 7,412 2

66.4 33.0 0.6

60.1 39.5 0.5

89.1 4.1 2.6 4.2

90.7 3.8 2.2 3.3

69.6 12.8 5.1 8.6 4.0 25.6 45.3 12.3 8.6

77.9 11.1 1.0 4.0 6.0 21.4 50.9 15.7 7.1

53.4 9.3

51.9 12.2

37.4

36.0

45.8 35.0 15.0 1.9 2.4

24.4 64.1 4.0 5.7 1.8

19.8

0.0

51.9

a

20

5,000 - 20,000 < 5,000 Population (i.e. size of the city) (%) > 500,000 250,000 - 500,000 50,000 - 250,000 < 50,000 Private Institutions (%) US Geographical Region (%) Midwest Northeast South West a Identifying data removed.

27.9 20.2 a

24.4 11.4 34.3 29.9 31.7

0.0

15.4 23.1 31.6 29.9

0 0 0 100

21

Table 2. Logistic Regression: Adjusted Odds Ratios for 30-day Marijuana Use Comparison Variable Beta SE t Categories RML 0.257 0.0695 3.7 Academic Year AY08-09 vs. AY14-15 -0.326 0.0368 -8.8 Academic Year AY09-10 vs. AY14-15 -0.211 0.0319 -6.6 Academic Year AY10-11 vs. AY14-15 -0.165 0.0344 -4.8 Academic Year AY11-12 vs. AY14-15 -0.160 0.0318 -5.0 Academic Year AY12-13 vs. AY14-15 -0.130 0.0329 -4.0 Academic Year AY13-14 vs. AY14-15 -0.137 0.0318 -4.3 Academic Year AY15-16 vs. AY14-15 -0.083 0.0297 -2.8 Season Fall vs. Spring -0.102 0.0308 -3.3 Gender Male vs. Female 0.425 0.0114 37.4 Transgender vs. Gender 0.213 0.0640 3.3 Female Bisexual vs. Sexual Orientation 0.743 0.0229 32.4 Heterosexual Gay/Lesbian vs. Sexual Orientation 0.312 0.0302 10.3 Heterosexual Other vs. Sexual Orientation 0.264 0.0259 10.2 Heterosexual Race/Ethnicity Asian vs. White -0.891 0.0219 -40.8 Race/Ethnicity Black vs. White -0.146 0.0259 -5.6 Race/Ethnicity Hispanic vs. White -0.180 0.0210 -8.6 Race/Ethnicity Other vs. White -0.044 0.0281 -1.6 First Year 0.056 0.0150 3.7 Legal Status (age ≥ 21) -0.274 0.0132 -20.8 Fraternity/Sorority 0.496 0.0159 31.1 International Student -0.450 0.0239 -18.9

1.29 0.72 0.81 0.85 0.85 0.88 0.87 0.92 0.90 1.53

95% LL 1.13 0.67 0.76 0.79 0.80 0.82 0.82 0.87 0.85 1.50

95% UL 1.48 0.78 0.86 0.91 0.91 0.94 0.93 0.98 0.96 1.59

0.0002