Advertisement

Recreational Marijuana Legalization and Adolescent Use of Marijuana, Tobacco, and Alcohol

      Abstract

      Purpose

      Given the rapid expansion of recreational marijuana legalization (RML) polices, it is essential to assess whether such policies are associated with shifts in the use of marijuana and other substances, particularly for adolescents, who are uniquely susceptible to negative repercussions of marijuana use. This analysis seeks to provide greater generalizability, specificity, and methodological rigor than limited prior evidence.

      Methods

      Youth Risk Behavior Survey data from 47 states from 1999 to 2017 assessed marijuana, alcohol, cigarette, and e-cigarette use among adolescents (14–18+ years; N = 1,077,938). Associations between RML and adolescent past-month substance use were analyzed using quasi-experimental difference-in-differences zero-inflated negative binomial models.

      Results

      Controlling for other state substance policies, year and state fixed effects, and adolescent demographic characteristics, models found that RML was not associated with a significant shift in the likelihood of marijuana use but predicted a small significant decline in the level of marijuana use among users (incidence rate ratio = .844, 95% confidence interval [.720–.989]) and a small increase in the likelihood of any e-cigarette use (odds ratio of zero use = .647, 95% confidence interval [.515–.812]). Patterns did not vary over adolescent age or sex, with minimal differences across racial/ethnic groups.

      Conclusions

      Results suggest minimal short-term effects of RML on adolescent substance use, with small declines in marijuana use and increase in the likelihood of any e-cigarette use. Given the delayed rollout of commercial marijuana sales in RML states and rapid expansion of RML policies, ongoing assessment of the consequences for adolescent substance use and related health and behavioral repercussions is essential.

      Keywords

      Implications and Contributions
      Assessing a national sample of high school students, analyses found minimal associations between recreational marijuana legalization and adolescent substance use, with small declines in the frequency of marijuana use among users, and small increases in the likelihood of e-cigarette use, trends which are essential to track as recreational marijuana legalization expands.
      The last decade has seen an historic shift toward legalizing personal marijuana use, with 11 states and the District of Columbia enacting recreational marijuana legalization (RML) policies, most of which provide for commercial marketing and sale of marijuana products [
      ProCon.org
      33 Legal medical marijuana states and D.C.: Laws, fees, and possession limits.
      ]. These laws follow a rash of state policies decriminalizing possession of marijuana and legalizing use for medicinal purposes. It is critical to understand how state laws liberalizing marijuana access affect use of marijuana and other substances, particularly for adolescents. Although not directly targeted by RML policies, which restrict age of access to 21 years, adolescents may be indirectly affected by shifts in norms and beliefs or changes in marijuana access as supply chains transfer from illegal to legal systems [
      • Smart R.
      • Pacula R.L.
      Early evidence of the impact of cannabis legalization on cannabis use, cannabis use disorder, and the use of other substances: Findings from state policy evaluations.
      ].
      Marijuana use among adolescents has steadily decreased in recent decades yet remains common, with one in five U.S. high school students reporting recent marijuana use and more than one-third reporting lifetime use in 2017 [
      CDC
      Trends in the prevalence of marijuana, cocaine, and other illegal drug use national YRBS: 1991—2017.
      ]. Early and heavy initiation of marijuana is linked to heightened use of alcohol, tobacco, and other illicit substances and to a greater likelihood of addiction and abuse [
      • Fergusson D.M.
      • Boden J.M.
      • Horwood L.J.
      Cannabis use and other illicit drug use: Testing the cannabis gateway hypothesis.
      ,
      • Milicic S.
      • Leatherdale S.T.
      The associations between e-cigarettes and binge drinking, marijuana use, and energy drinks mixed with alcohol.
      ]. Furthermore, the still-maturing neurological systems of adolescents make them particularly susceptible to detrimental behavioral and cognitive effects of substance use [
      U.S. Department of Health and Human Services (HHS)
      Facing addiction in America: The Surgeon General’s report on alcohol, drugs, and health.
      ], making understanding early marijuana use a critical public health concern.
      To date, studies have primarily examined the link between medical marijuana legalization (MML) and marijuana decriminalization laws (MDL) and youth marijuana use and have consistently identified either null or small negative links with adolescent marijuana use [
      • Anderson D.M.
      • Hansen B.
      • Rees D.I.
      Medical marijuana laws and teen marijuana use.
      ,
      • Cerdá M.
      • Sarvet A.L.
      • Wall M.
      • et al.
      Medical marijuana laws and adolescent use of marijuana and other substances: Alcohol, cigarettes, prescription drugs, and other illicit drugs.
      ,
      • Coley R.L.
      • Hawkins S.S.
      • Ghiani M.
      • et al.
      A quasi-experimental evaluation of marijuana policies on youth marijuana use.
      ,
      • Johnson J.K.
      • Johnson R.M.
      • Hodgkin D.
      • et al.
      Heterogeneity of state medical marijuana laws and adolescent recent use of alcohol and marijuana: Analysis of 45 states, 1991–2011.
      ,
      • Lynne-Landsman S.D.
      • Livingston M.D.
      • Wagenaar A.C.
      Effects of state medical marijuana laws on adolescent marijuana use.
      ]. A 2018 meta-analysis of 11 studies using rigorous difference-in-differences analyses found nonsignificant associations between MML and adolescent marijuana use [
      • Sarvet A.L.
      • Wall M.M.
      • Fink D.S.
      • et al.
      Medical marijuana laws and adolescent marijuana use in the United States: A systematic review and meta-analysis.
      ]. The most recent study found no links with MDL but small significant reductions in adolescent marijuana use in response to MML [
      • Coley R.L.
      • Hawkins S.S.
      • Ghiani M.
      • et al.
      A quasi-experimental evaluation of marijuana policies on youth marijuana use.
      ].
      Reflecting the recency of RML policies (the first two states adopted RML in 2012), there are far fewer studies about their influence and those that exist offer mixed results. A study on the two earliest RML adopters found small increases in marijuana use among eighth and 10th graders in Washington but no significant changes for Washington 12th graders or Colorado adolescents [
      • Cerdá M.
      • Wall M.
      • Feng T.
      • et al.
      Association of state recreational marijuana laws with adolescent marijuana use.
      ], whereas another study on the first four enactor states found no significant changes in recent marijuana use but slight increases in cannabis use disorder [
      • Cerdá M.
      • Mauro C.
      • Hamilton A.
      • et al.
      Association between recreational marijuana legalization in the United States and changes in marijuana use and cannabis use disorder from 2008 to 2016.
      ]. In contrast, an assessment of more than1.4 million adolescents from the Youth Risk Behavior Survey (YRBS) found that RML laws were associated with an 8% decrease in the odds of recent marijuana use and a 9% decrease in the odds of frequent use [
      • Anderson D.M.
      • Hansen B.
      • Rees D.I.
      • Sabia J.J.
      Association of marijuana laws with teen marijuana use: New estimates from the youth risk behavior surveys.
      ]. Although this study had a much larger sample, assessed more RML enactor states, and controlled for other substance use policies, it has been critiqued for using unweighted data and combining state and national YRBS data, methodological choices which the Centers for Disease Control and Prevention specifically report as inappropriate [
      CDC
      Combining YRBS data across years and sites.
      ,
      • Jones C.M.
      • Underwood J.M.
      • Volkow N.D.
      Challenging the association of marijuana laws with teen marijuana use.
      ,
      • Rapoport E.
      • Keim S.A.
      • Adesman A.
      Challenging the association of marijuana laws with teen marijuana use.
      ].
      Given the limited and conflicting evidence and methodological limitations in prior work, additional analysis of the repercussions of RML on adolescent marijuana use is essential. It is further important to consider whether repercussions vary across subgroups. Prior work has not explored racial/ethnic differences, information that is particularly relevant in relation to the ways that current marijuana policies disproportionately burden black and Latinx youth [
      • Resing C.
      Marijuana legalization is a racial justice issue.
      ]. Evidence suggests that those youths most targeted by the criminal justice system, including youth of color and young men, may respond differently to shifts in marijuana restrictions, with MDL and MML policies predicting declines in marijuana use among youth of color and men but not their white or female peers [
      • Coley R.L.
      • Hawkins S.S.
      • Ghiani M.
      • et al.
      A quasi-experimental evaluation of marijuana policies on youth marijuana use.
      ].
      Finally, given the connections between use of marijuana and tobacco, alcohol, and other drugs, it is essential to consider whether there may be spillover effects of marijuana policies in which changes in access to one drug may alter usage of other drugs. Substitution models suggest that when one drug becomes more accessible, use of other drugs may decline [
      • Crost B.
      • Guerrero S.
      The effect of alcohol availability on marijuana use: Evidence from the minimum legal drinking age.
      ,
      • Guttmannova K.
      • Lee C.M.
      • Kilmer J.R.
      • et al.
      Impacts of changing marijuana policies on alcohol use in the United States.
      ]. In contrast, complementarity models argue that increased access to or use of one drug may raise other drug use owing to concurrent use of multiple substances or adoption of a more permissive attitude toward substance use [
      • Guttmannova K.
      • Lee C.M.
      • Kilmer J.R.
      • et al.
      Impacts of changing marijuana policies on alcohol use in the United States.
      ,
      • Williams J.
      • Pacula R.L.
      • Chaloupka F.J.
      • Wechsler H.
      Alcohol and marijuana use among college students: Economic complements or substitutes?.
      ]. Little research has tested these models in relation to marijuana policy shifts among youths. One study found that RML was associated with decreases in alcohol use but increases in sedative use among college students [
      • Alley Z.M.
      • Kerr D.C.R.
      • Bae H.
      Trends in college students’ alcohol, nicotine, prescription opioid and other drug use after recreational marijuana legalization: 2008–2018.
      ]. Other research has linked MML with decreases in cigarette smoking among younger adolescents but increases among older adolescents [
      • Cerdá M.
      • Sarvet A.L.
      • Wall M.
      • et al.
      Medical marijuana laws and adolescent use of marijuana and other substances: Alcohol, cigarettes, prescription drugs, and other illicit drugs.
      ], whereas others found no significant increases in alcohol or cigarette sales in states with RML or MML [
      • Veligati S.
      • Howdeshell S.
      • Beeler-Stinn S.
      • et al.
      Changes in alcohol and cigarette consumption in response to medical and recreational cannabis legalization: Evidence from U.S. state tax receipt data.
      ].
      The current analysis aims to fill gaps in the existing literature, using nationally representative data to examine how state-level RML policy shifts, considered in conjunction with other relevant substance policies, are associated with patterns of adolescent marijuana use. Extending past research, we further assess links between RML and adolescent use of alcohol, cigarettes, and e-cigarettes. Finally, we consider variation in links across adolescent age, sex, and race/ethnicity to identify potential high-risk groups.

      Methods

      Sample and measures

      Data were derived from the YRBS, which has conducted biennial self-reported surveys since 1991 with state-representative samples of adolescents in grades 9 through 12. Survey weights adjust for the complex sampling design [
      • Brener N.D.
      • Kann L.
      • Shanklin S.
      • et al.
      Methodology of the youth risk behavior surveillance system—2013.
      ,
      • Kann L.
      • McManus T.
      • Harris W.A.
      • et al.
      Youth risk behavior surveillance—United States, 2017.
      ]. We analyzed data from adolescents from 47 states from 1999 to 2017 (N = 1,142,479) to analyze all available state data with survey weights and to focus on more recent years when marijuana policies were shifting most rapidly. We excluded youth who were younger than 14 years (.3%), were missing age, sex, or race/ethnicity data (4.2%), or were missing substance use data (1.7% for marijuana, 7.1% for alcohol, 3.8% for cigarettes, and 74.7% for e-cigarettes, which was only reported in 2015 and 2017), leading to analytic samples ranging from 1,077,938 to 254,729 adolescents. The Institutional Review Board at Boston College reviewed this study and considered it exempt.
      Adolescents self-reported their use of various substances in the prior 30 days, including (1) the number of times they had used marijuana, coded as a count variable ranging from 0 to 40; and the number of days they had (2) drunk alcohol; (3) smoked cigarettes; and (4) used an electronic vapor (e-cigarette) product, all coded 0–30. E-cigarette use was ascertained only in 2015 and 2017; thus, there were only two cohorts of e-cigarette data.
      We linked state-level data on RML, MML, and MDL implementation, using the date that the law became effective (acknowledging that implementation dates may vary across policy components). As most states conduct the YRBS in the spring [
      • Brener N.D.
      • Kann L.
      • Shanklin S.
      • et al.
      Methodology of the youth risk behavior surveillance system—2013.
      ,
      • Kann L.
      • McManus T.
      • Harris W.A.
      • et al.
      Youth risk behavior surveillance—United States, 2017.
      ], adolescents were considered to be exposed if each law was in effect as of April 1 of the cycle year. By 2017, six of the 47 states included in the analyses had RML laws in effect, 24 states had MML laws, and 18 states had MDL laws in effect (see Table 1). We defined binary indicators each year in each state delineating whether RML, MML, and MDL were in effect.
      Table 1State data in YRBS, prevalence of marijuana use, and marijuana policies (N = 1,077,938)
      StateYearsN% Of sample% Use marijuanaRecreational marijuana
      States adopting recreational, medical or decriminalization marijuana laws with effective dates (dates on which the law is in force, as opposed to enacted) after April 1, 2017, indicated in brackets and italics, were coded as non-RML, non-MML, or non-decriminalization states in all models.
      Medical marijuana
      States adopting recreational, medical or decriminalization marijuana laws with effective dates (dates on which the law is in force, as opposed to enacted) after April 1, 2017, indicated in brackets and italics, were coded as non-RML, non-MML, or non-decriminalization states in all models.
      Marijuana decriminalization
      States adopting recreational, medical or decriminalization marijuana laws with effective dates (dates on which the law is in force, as opposed to enacted) after April 1, 2017, indicated in brackets and italics, were coded as non-RML, non-MML, or non-decriminalization states in all models.
      AK03, 07–178,812.2420.9February 2015March 1999May 1975/November 2003/February 2015
      Marijuana (possession of less than four ounces) decriminalized in 1975, re-criminalized in 1990, decriminalized in 2003, re-criminalized in 2006, and decriminalized in 2015.
      AL99–05, 09–1511,1961.6018.8
      AR99, 01, 05–1714,4151.0218.5November 2016
      AZ03–1720,2762.1422.5April 2011
      CA15–173,52012.2422.3November 2016November 1996January 1976
      CO05, 09, 11, 175,5881.6222.2December 2012June 2001December 2012
      CT05–1715,1171.2822.6May 2012July 2011
      DE03–1720,051.2825.3July 2011December 2015
      FL03–1740,9905.9120.4[June 2017]
      GA03–1311,3783.3319.5
      HI175,495.3117.4December 2000[Jan. 2020]
      IA05, 07, 11, 175,8561.1713.6
      ID03–1713,476.6016.3
      IL07–1719,1024.4321.2[January 2020]January 2014July 2016
      IN03–11, 1511,3572.3219.4
      KS05–13, 1711,2411.1014.9
      KY03–1717,2301.4017.2
      LA07–13, 175,3371.2716.4May 2016
      MA99–1731,7212.1626.8December 2016January 2013January 2009
      MD05–17151,9411.8919.9June 2014October 2014
      ME01–1747,727.4622.2January 2017December 1999May 1976
      MI99–1733,2713.6221.0[December 2018]December 2008[December 2018]
      MO99–09, 13–1714,3582.0820.5[December 2018]January 2017
      MS99–03, 07–1513,1931.0218.1July 1977
      MT99–1734,047.3522.4November 2004
      NC01–1729,9303.1121.4July 1977
      ND99–1717,992.2617.2[April 2017][July 2019]
      NE03, 05, 11–1713,798.6514.6January 1979
      NH03–1734,130.4825.2July 2013[September 2017]
      NJ01, 05, 09–138,3982.9421.2July 2010
      NM05–1736,686.7426.6July 2007[May 2019]
      NV99–09, 13–1715,139.8719.9January 2017October 2001October 2001
      NY03–1784,3175.9819.7July 2014July 1977
      OH99,03–07,11,139,6654.7221.6September 2016August 1975
      OK03–1712,6381.3717.8[July 2018]
      PA15, 176,2373.8617.8May 2016
      RI01–1721,745.3525.8January 2006April 2013
      SC99, 05–1713,2801.5020.1
      SD99–1514,092.3217.4
      TN03–13, 1713,8312.1620.2
      TX01, 05–13, 1726,3219.7019.9
      UT99–13, 1714,1631.159.1[December 2018]
      VA11–1715,5422.7917.2[July 2020]
      VT99–11, 1769,279.2326.7[July 2018]July 2004June 2013
      WI99–13, 1719,7132.1419.8
      WV99, 03–1714,002.6321.1[July 2019]
      WY99–1520,345.2118.4
      YRBS = Youth Risk Behavior Survey.
      a States adopting recreational, medical or decriminalization marijuana laws with effective dates (dates on which the law is in force, as opposed to enacted) after April 1, 2017, indicated in brackets and italics, were coded as non-RML, non-MML, or non-decriminalization states in all models.
      b Marijuana (possession of less than four ounces) decriminalized in 1975, re-criminalized in 1990, decriminalized in 2003, re-criminalized in 2006, and decriminalized in 2015.
      We also assessed state policies targeting other substances, including beer taxes (in 2017 dollars) [
      Distilled Spirit Council of the United States
      History of beverage alcohol tax changes.
      ,
      Distilled Spirits Council of the United States
      How high are beer taxes in your state?.
      ], average cigarette taxes (in 2017 dollars) [
      • Orzechowski W.
      • Walker R.C.
      ], and 100% smoke-free restaurant legislation [
      American Nonsmokers’ Rights Foundation
      Chronological table of US population protected by 100% smokefree state or local laws: January 2, 2020.
      ]. We also adjusted for state unemployment rates, drawn from the U.S. Bureau of Labor Statistics.
      Adolescents reported their age (14, 15, 16, 17, 18+ years), sex (female, male), and race/ethnicity (non-Hispanic white, non-Hispanic black, Hispanic, non-Hispanic other). No additional information on student or families' sociodemographic characteristics was collected in the YRBS. We also adjusted for state and year indicators.

      Statistical analysis

      Analyses were estimated using difference-in-differences regressions to compare changes in substance use over time in states that had RML in effect with states that did not [
      • Wing C.
      • Simon K.
      • Bello-Gomez R.A.
      Designing difference in difference studies: Best practices for public health policy research.
      ,
      • Bertrand M.
      • Duflo E.
      • Mullainathan S.
      How much should we trust differences-in-differences estimates?.
      ]. Extending simpler difference-in-differences models that assess a shift at a single time point, this expanded difference-in-differences specification is a rigorous method which has been used extensively in other research to exploit changes in policy domains across multiple states and years [
      • Cerdá M.
      • Sarvet A.L.
      • Wall M.
      • et al.
      Medical marijuana laws and adolescent use of marijuana and other substances: Alcohol, cigarettes, prescription drugs, and other illicit drugs.
      ,
      • Coley R.L.
      • Hawkins S.S.
      • Ghiani M.
      • et al.
      A quasi-experimental evaluation of marijuana policies on youth marijuana use.
      ].

      The general specification was

      Equation 1:E(yist)=exp(α+γRMLst+β1Zst+β2Xi+δs+θt)


      where yist is the count of substance use by individual i in state s and year t, γ is the primary parameter of interest assessing RML, Zst is a vector of other time-varying state policies and characteristics, Xi is a vector of characteristics for individual i, δs is a state-specific fixed effect, and θt denotes year fixed effects.
      Because the reports of marijuana and other substances were count variables with substantial responses of zero and over-dispersed distributions (The zero counts derived both from respondents who had never used each substance and respondents who had used previously but not in the past month; unfortunately, the YRBS did not consistently ask about lifetime use of each substance across states and years, leaving us unable to reliably distinguish these subgroups. Nonetheless, zero-inflated negative binomial models properly accommodate these two types of zeros), we estimated zero-inflated negative binomial regression models for each outcome. The first set of models assessed associations between RML and adolescent use of marijuana, alcohol, cigarettes, and e-cigarettes, with covariates adjusting for other relevant shifting policy contexts, time-invariant state factors, overall time trends, and adolescent characteristics. Second, we added interactions between the RML indicator and adolescent age, sex, and race/ethnicity indicators to assess whether associations varied across subgroups of adolescents. We conducted all analyses using Stata statistical software, version 15 (StataCorp, College Station, TX), including survey weights (with multiple-year adjustments) [
      • Clark P.A.
      • Capuzzi K.
      • Fick C.
      Medical marijuana: Medical necessity versus political agenda.
      ] to produce representative estimates for the 47 states included in the analysis.
      Before beginning our main analyses, we explored potential bias in data patterns to assure that adolescent substance use in states that had RML in effect were not systemically different than in states that did not implement such laws. We coded the RML policies as a series of six indicator variables for each state and year, indicating whether the RML policy would be in effect within the following year or more than 1, 2, 3, 4, or 5 years in the future (vs. not implementing RML). We included these variables as well as an indicator for having implemented RML and covariates in logistic regression models estimating marijuana, alcohol, and cigarette use (models were not conducted for e-cigarette use as only two waves of data were available). Results (Table 2) rigorously supported the requirement of parallel trends, finding that adolescent use of marijuana, alcohol, and cigarettes did not significantly vary across states that did and did not implement RML in the 5+ years before implementation. These results support our analytic strategy of difference-in-differences models for assessing associations between RML and subsequent shifts in adolescent substance use.
      Table 2Difference-in-differences logistic regression model of years leading up to RML implementation and current substance use
      Marijuana (N = 1,077,938)Alcohol (N = 1,027,313)Cigarettes (N = 1,064,347)
      OR (95% CI)OR (95% CI)OR (95% CI)
      5+ year lead
       No1.000 Referent1.000 Referent1.000 Referent
       Yes.949 (.851–1.059).935 (.852–1.025).953 (.828–1.098)
      4-year lead
       No1.000 Referent1.000 Referent1.000 Referent
       Yes.963 (.828–1.119)1.041 (.929–1.166).991 (.854–1.151)
      3-year lead
       No1.000 Referent1.000 Referent1.000 Referent
       Yes.965 (.834–1.117).973 (.859–1.101).951 (.810–1.118)
      2 year lead
       No1.000 Referent1.000 Referent1.000 Referent
       Yes.970 (.838–1.121).984 (.868–1.115).990 (.805–1.219)
      1-year lead
       No1.000 Referent1.000 Referent1.000 Referent
       Yes.902 (.773–1.053).922 (.785–1.083).919 (.760–1.112)
      No lead
       No1.000 Referent1.000 Referent1.000 Referent
       Yes1.040 (.748–1.445)1.165 (.808–1.679)1.012 (.673–1.522)
      RML in force
       No1.000 Referent1.000 Referent1.000 Referent
       Yes.832 (.638–1.084).887 (.686–1.148).973 (.690–1.370)
      No coefficients are statistically significant (p ≤ .05). Models include adjustment for: age, race/ethnicity, sex, year, state, beer taxes, cigarette taxes, smoke free policies, unemployment rate, MML, and MDL. Models predicted any use of each substance (no/yes) in past month.
      CI = confidence interval; OR = odds ratio; RML = recreational marijuana legalization.

      Results

      Table 3 presents descriptive data, showing that 20.3% of adolescents had used marijuana, 36.5% alcohol, 15.4% cigarettes, and 18.6% e-cigarettes in the prior month. The prevalence of all substances increased significantly with age. White adolescents reported significantly higher rates of use than youth of color for alcohol, cigarettes, and e-cigarettes (except for Hispanics, who showed similar rates of e-cigarette use to whites). In contrast, black and Hispanic adolescents reported higher rates of marijuana use than whites. Male adolescents reported significantly higher use than female adolescents of marijuana, cigarettes, and e-cigarettes but lower use of alcohol.
      Table 3Adolescent characteristics and prevalence of substance use: YRBS, 1999–2017
      % Of sample (N = 1,136,288)Marijuana

      % Use (N = 1,077,938)
      Alcohol

      % Use (N = 1,027,313)
      Cigarette

      % Use (N = 1,064,437)
      ECig

      % Use (N = 254,729)
      Full sample20.336.515.418.6
      Age
       1413.011.6
      differs (p < .05) from 18 years of age.
      23.7
      differs (p < .05) from 18 years of age.
      8.8
      differs (p < .05) from 18 years of age.
      14.2
      differs (p < .05) from 18 years of age.
       1526.515.9
      differs (p < .05) from 18 years of age.
      30.1
      differs (p < .05) from 18 years of age.
      11.816.3
      differs (p < .05) from 18 years of age.
       1626.821.1
      differs (p < .05) from 18 years of age.
      36.8
      differs (p < .05) from 18 years of age.
      15.5
      differs (p < .05) from 18 years of age.
      18.9
      differs (p < .05) from 18 years of age.
       1722.724.5
      differs (p < .05) from 18 years of age.
      43.4
      differs (p < .05) from 18 years of age.
      18.5
      differs (p < .05) from 18 years of age.
      21.0
      differs (p < .05) from 18 years of age.
       1811.026.247.923.826.0
      Race/Ethnicity
       White57.119.840.319.020.6
       Black15.521.7
      differs (p < .05) from whites.
      28.6
      differs (p < .05) from whites.
      9.0
      differs (p < .05) from whites.
      14.6
      differs (p < .05) from whites.
       Hispanic19.721.6
      differs (p < .05) from whites.
      36.1
      differs (p < .05) from whites.
      12.7
      differs (p < .05) from whites.
      19.3
       Other7.717.6
      differs (p < .05) from whites.
      27.2
      differs (p < .05) from whites.
      12.0
      differs (p < .05) from whites.
      16.3
      differs (p < .05) from whites.
      Sex
       Female49.718.437.614.616.8
       Male50.322.1
      differs (p < .05) from women.
      35.9
      differs (p < .05) from women.
      16.7
      differs (p < .05) from women.
      21.2
      differs (p < .05) from women.
      YRBS = Youth Risk Behavior Survey.
      a differs (p < .05) from 18 years of age.
      b differs (p < .05) from whites.
      c differs (p < .05) from women.
      Results from the difference-in-differences zero-inflated negative binomial regression models considering associations between RML and adolescent marijuana, alcohol, cigarette, and e-cigarette use are presented in Table 4, including odds ratios for the inflation portion of the model, which are interpreted as the likelihood that an individual reported zero past month use of that substance, as well as incidence rate ratios for the negative binomial portion, which are interpreted as the percent change in the rate of past month use of each substance. Results indicate that RML was not associated with the likelihood of zero use of marijuana in the past month, but among users, RML predicted a 16% lower rate of marijuana use. RML also predicted a 35% lower odds of zero use of e-cigarettes in the prior month, indicating a higher likelihood of any use, but was not associated with the level of use of e-cigarettes among users. RML was not significantly associated with adolescents' use of alcohol or cigarettes.
      Table 4Main effects of difference-in-differences zero-inflated negative binomial models of RML and substance use
      Marijuana (N = 1,077,938)Alcohol (N = 1,027,313)Cigarettes (N = 1,064,437)E-cigarettes (N = 254,729)
      Inflation model

      OR (95% CI)
      NB model

      IRR (95% CI)
      Inflation model

      OR (95% CI)
      NB model

      IRR (95% CI)
      Inflation model

      OR (95% CI)
      NB model

      IRR (95% CI)
      Inflation model

      OR (95% CI)
      NB model

      IRR (95% CI)
      RML1.037 (.803–1.340).844 (.720–.989).904 (.638–1.281).933 (.745–1.169).910 (.673–1.230).798 (.517–1.233).647 (.515–.812).889 (.713–1.108)
      MML1.58 (.983–1.138)1.032 (.976–1.092)1.031 (1.029–1.208).977 (.917–1.040)1.012 (.936–1.093).985 (.931–1.041)1.668 (1.213–2.293).921 (.679–1.248)
      MDL.967 (.876–1.068).969 (.898–1.048)1.041 (.931–1.164).980 (.906–1.061)1.027 (.938–1.124).960 (.867–1.063)1.501 (1.111–2.028)1.160 (.941–1.429)
      Age
       142.742 (2.567–2.928).663 (.607–.724)4.356 (3.808–4.982).721 (.665–.782)3.034 (2.804–3.284).686 (.642–.733)2.431 (1.863–3.174).793 (.632–.995)
       151.879 (1.764–2.001).704 (.665–.745)2.868 (2.534–3.246).716 (.691–.742)2.351 (2.235–2.472).771 (.736–.808)2.082 (1.681– 2.579).841 (.698–1.013)
       161.309 (1.230–1.394).801 (.773–.830)1.932 (1.709–2.184).813 (.775–.852)1.713 (1.617–1.815).896 (.865–.928)1.689 (1.400–2.037).916 (.719–1.167)
       171.075 (1.024–1.129).897 (.856–.941)1.217 (1.117–1.326).848 (.822–.874)1.365 (1.297–1.437).951 (.919–.984)1.380 (1.182–1.612).866 (.695–1.079)
       181.000 Ref1.000 Ref1.000 Ref1.000 Ref1.000 Ref1.000 Ref1.000 Ref1.000 Ref
      Race/Ethnicity
       White1.000 Ref1.000 Ref1.000 Ref1.000 Ref1.000 Ref1.000 Ref1.000 Ref1.000 Ref
       Black.855 (.814–.898).968 (.938- 1.000)2.069 (1.950–2.196).818 (.777–.861)2.554 (2.448–2.665).768 (.741–.795)1.566 (1.289–1.903).751 (.588–.960)
       Hispanic.931 (.861–1.007)1.000 (.939–1.064)1.096 (.971–1.236).969 (.938–1.001)1.130 (1.058–1.206).736 (.685–.792)1.120 (.877–1.431).776 (.652–.924)
       Other1.227 (1.082–1.391)1.088 (1.016–1.165)2.009 (1.809–2.231)1.070 (.986–1.160)1.254 (1.158–1.358).983 (.917–1.055)1.635 (1.214–2.204).964 (.802–1.160)
      Sex
       Female1.000 Ref1.00 Ref1.000 Ref1.000 Ref1.000 Ref1.00 Ref1.000 Ref1.000 Ref
       Male.861 (.834–.889)1.539 (1.486–1.594)1.557 (1.483–1.636)1.445 (1.407–1.484).869 (.842–.896)1.088 (1.053–1.123).884 (.781–1.001)1.838 (1.591- 2.125)
      Other State Policies
       Beer taxes1.080 (.932–1.250).999 (.882–1.132).944 (.808–1.103).972 (.854–1.106).986 (.833–1.168)1.003 (.895–1.124)1.081 (.682–1.712).812 (.567–1.162)
       Smoke free.990 (.927–1.058).967 (.921–1.015).994 (.921–1.074).970 (.933–1.007)1.055 (.993–1.121).986 (.948–1.025)1.917 (.695–5.286)1.224 (.539–2.777)
       Cigarette taxes.986 (.946–1.028)1.010 (.978–1.042).981 (.933–1.033)1.019 (.989–1.049)1.048 (1.008–1.091).991 (.964–1.019)1.011 (.829–1.234).955 (.785–1.163)
       Unemploy-ment.984 (.963–1.005)1.011 (.995–1.026).997 (.973–1.021)1.024 (1.009–1.038).998 (.976–1.020)1.009 (.997–1.022).899 (.749–1.079)1.041 (.899–1.206)
      Values in bold type are statistically significant (p ≤ .05). Models includes adjustments for year and state.
      CI = confidence interval; IRR = incidence rate ratio; MDL = marijuana decriminalization laws; MML = medical marijuana legalization; NB = negative binomiall; RML = recreational marijuana legalization.
      Table 5 presents results of models assessing whether RML associations with substance use varied across key demographic characteristics, showing the significance of the omnibus tests of each set of interaction terms (Wald test) in addition to interaction odds ratios and incidence rate ratios for RML among each subgroup. Results show limited evidence of moderation. Specifically, the omnibus tests of the interactions between RML and age and between RML and sex were nonsignificant across all four substances, indicating consistent patterns across ages 14–18 years and between male and female adolescents. In contrast, significant omnibus tests emerged between RML and race/ethnicity for marijuana, cigarette and e-cigarette use. RML led to 2.14 greater odds of zero uses of marijuana in adolescents identifying as other race/ethnicity in comparison to white adolescents, with no race/ethnicity differences in use among users. For cigarettes, RML was associated with greater levels of use among users of color, predicting 2.20 times higher levels of use of cigarettes among Hispanics and 2.96 times higher levels of use among other racially/ethnically identified adolescents in comparison with whites. In contrast, RML predicted levels of use of e-cigarettes 69% lower among blacks than among whites.
      Table 5Interaction effects of zero-inflated negative binomial models of RML and current substance use
      Marijuana (N = 1,077,938)Alcohol (N = 1,027,313)Cigarettes (N = 1,064,347)E-Cigarettes (N = 254,729)
      Inflation Model

      OR (95% CI)
      NB Model

      IRR (95% CI)
      Inflation Model

      OR (95% CI)
      NB Model

      IRR (95% CI)
      Inflation Model

      OR (95% CI)
      NB Model

      IRR (95% CI)
      Inflation Model

      OR (95% CI)
      NB Model

      IRR (95% CI)
      RML × Age
       14.882 (.421–1.844)1.042 (.486–2.236)1.271 (.431–3.752)1.144 (.714–1.833)1.821 (.661–5.020).833 (.269–2.575).692 (.293–1.636).512 (.260–1.009)
       151.231 (.657–2.307)1.092 (.622–1.915)1.075 (.389–2.975).898 (.669–1.204)2.085 (.937–4.641)1.605 (.719–3.581).742 (.366–1.504).594 (.303–1.164)
       161.210 (.719–2.038).938 (.654–1.345)1.336 (.484–3.684).993 (.594–1.661)2.215 (1.179–4.160)1.715 (.674–4.368).940 (.519–1.700).882 (.452–1.723)
       17.928 (.575–1.499)1.038 (.752–1.432).541 (.262–1.116).805 (.597–1.084)1.710 (.982–2.977)1.798 (.736–4.395).728 (.376–1.411).687 (.331–1.426)
       181.000 Ref1.000 Ref1.000 Ref1.000 Ref1.000 Ref1.000 Ref1.000 Ref1.000 Ref
      Wald p-Value.698.982.124.489.097.255.795.340
      RML × Race/ethnicity
       White1.000 Ref1.000 Ref1.000 Ref1.000 Ref1.000 Ref1.000 Ref1.000 Ref1.000 Ref
       Black1.063 (.621–1.822).906 (.514–1.596)2.143 (.900–5.103)1.017 (.479–2.160).959 (.306–3.012).569 (.249–1.301).630 (.296–1.341).309 (.175–.547)
       Hispanic1.365 (.976–1.908)1.087 (.797–1.483).983 (.542–1.784).926 (.685–1.252).945 (.584–1.529)2.201 (1.105–4.384).927 (.575–1.494).654 (.407–1.051)
       Other2.138 (1.342–3.404)1.328 (.855–2.062)1.694 (.978–2.934)1.255 (.839–1.878)1.203 (.652–2.219)2.962 (1.242–7.066)1.109 (.544–2.261).766 (.490–1.200)
       Wald p-Value.008.542.145.474.905<.000.594.001
      RML × Sex
       Female1.000 Ref1.000 Ref1.000 Ref1.000 Ref1.000 Ref1.000 Ref1.000 Ref1.000 Ref
       Male1.106 (.827–1.478).960 (.728–1.266)1.243 (.797–1.938).876 (.692–1.111).826 (.488–1.397).843 (.463–1.535)1.073 (.735–1.567).998 (.686–1.453)
       Wald p-Value.498.771.337.275.476.576.715.993
      Values in bold type are statistically significant (p ≤ .05). Models include adjustments for: age, race/ethnicity, sex, year, state, beer taxes, cigarette taxes, smoke free policies, unemployment rate, MML, and MDL.
      CI = confidence interval; IRR = incidence rate ratio; MDL = marijuana decriminalization laws; MML = medical marijuana legalization; NB = negative binomiall; RML = recreational marijuana legalization.
      In considering the role of other substance use policies (Table 4), we found that MML predicted 1.67 greater odds of adolescents reporting zero e-cigarette use, whereas MDL predicted 1.50 greater odds of zero e-cigarette use, reinforcing the pattern that more liberal marijuana laws show limited, and primarily negative links with adolescent substance use. Considering tax policies, results replicated prior studies finding negative associations between cigarette taxes and any use of cigarettes. Smoke-free legislation and beer taxes were not significantly associated with youth substance use, whereas the unemployment rate predicted heightened rates of alcohol use among users.
      We estimated additional model specifications to test the robustness of our results. Specifically, we estimated our main effect models excluding Colorado, given that Colorado implemented RML far earlier than other states in our sample. We also estimated models excluding California, given the large population size and resulting ability to unduly affect results. No substantive differences in results emerged (available on request).

      Discussion

      The legal landscape controlling commercial sales and personal use of marijuana is shifting rapidly, along with public opinion regarding the dangers of marijuana versus the social and economic costs and inequities related to the criminalization of marijuana use [
      • Smart R.
      • Pacula R.L.
      Early evidence of the impact of cannabis legalization on cannabis use, cannabis use disorder, and the use of other substances: Findings from state policy evaluations.
      ,
      • Resing C.
      Marijuana legalization is a racial justice issue.
      ]. It is of central public health concern to delineate the repercussions of this shifting policy landscape, particularly on the populations most vulnerable to detrimental effects of substance use, most notably adolescents, whose brains and health behaviors are rapidly maturing. To begin examining the potential role marijuana policy shifts play in adolescent drug use behavior, we analyzed data on more than 1 million high school students from 47 states weighted to be representative and hence present generalizable results. Rigorous difference-in-differences models assessed associations between RML and adolescent use of marijuana, alcohol, cigarettes, and e-cigarettes. Importantly, models adjusted for other marijuana policies (MML and MDL) as well as policies target alcohol and tobacco consumption, adolescent demographic characteristics, state unemployment rates, and state and year fixed effects. Inclusion of these covariates was essential to help control for possible spurious effects related to overlapping policy changes, shifting demographics, changes in social attitudes and norms, macroeconomic forces, and unmeasured characteristics of states that may have been associated with adolescent substance use and marijuana policies. Moreover, we verified the central requirement of parallel trends showing empirically that patterns of marijuana, alcohol, and tobacco use did not vary in states that implemented versus did not implement RML in the years leading up to policy enactment. Nonetheless, we caution that data were not derived from a randomized experiment, limiting causal interpretations.
      From these models, three central results emerged. First, we found no evidence that RML was associated with increased likelihood or level of marijuana use among adolescents. Rather, among adolescents who reported any use of marijuana in the past month, the frequency of use declined by 16% after RML, with no differences in patterns across males and females or adolescents ranging from 14 to 18 years of age. It is essential to highlight the small size of this shift, which translated into about .5 fewer uses of marijuana per month. These results are in partial agreement with those of the other recent analysis of the YRBS data [
      • Anderson D.M.
      • Hansen B.
      • Rees D.I.
      • Sabia J.J.
      Association of marijuana laws with teen marijuana use: New estimates from the youth risk behavior surveys.
      ] in finding declines in frequent use of marijuana but methodologically improves on this study through the correct use of weighted YRBS data. Yet, our results conflict with those of other studies of smaller sets of early state adopters of RML which found slight increases in marijuana use [
      • Cerdá M.
      • Wall M.
      • Feng T.
      • et al.
      Association of state recreational marijuana laws with adolescent marijuana use.
      ] and no difference in use but small increases in cannabis use disorder [
      • Cerdá M.
      • Sarvet A.L.
      • Wall M.
      • et al.
      Medical marijuana laws and adolescent use of marijuana and other substances: Alcohol, cigarettes, prescription drugs, and other illicit drugs.
      ] among younger adolescents. It is unclear whether the differences in results are related to the inclusion of more recent state adopters of RML that may differ systematically from the earliest adopters (although our results remained unchanged when we omitted the earliest RML state in our data, Colorado), the far larger sample of adolescents, differences in sampling methods, or the adjustment for a broader array of marijuana and other substance use policies and state and year fixed effects in this study. Taken together, the growing evidence of RML policies suggest that short-term associations with adolescent use of marijuana are small, if statistically significant.
      As RML policies spread across additional states and lead to increasing commercial availability, it is essential to continue to assess connections with adolescent marijuana use and to increase understanding of why shifts may occur. Anderson et al [
      • Anderson D.M.
      • Hansen B.
      • Rees D.I.
      • Sabia J.J.
      Association of marijuana laws with teen marijuana use: New estimates from the youth risk behavior surveys.
      ] argued that the changing marijuana markets, with sales shifting from illegal markets which disregard age restrictions to legal markets with enhanced enforcement of age 21-year restrictions, may decrease adolescents' access to marijuana products. Alternately, changes may be due to interpersonal factors: prior research finding small declines in adolescent marijuana use after MML argued that shifting attitudes and knowledge of marijuana's accessibility may alter parental supervision and oversight, limiting youth substance use [
      • Coley R.L.
      • Hawkins S.S.
      • Ghiani M.
      • et al.
      A quasi-experimental evaluation of marijuana policies on youth marijuana use.
      ,
      • Hasin D.S.
      • Wall M.
      • Keyes K.M.
      • et al.
      Medical marijuana laws and adolescent marijuana use in the USA from 1991 to 2014: Results from annual, repeated cross-sectional surveys.
      ].
      A second key result from this study suggests the possibility of spillover effects, with RML linked to small increases in the likelihood of e-cigarette use among adolescents, although no significant link emerged with the frequency of use. It is essential to highlight that only 2 years of data were available for e-cigarette use, coinciding with dramatic increases in use and allowing a limited window into shifts in use over time. The link between RML enactment and e-cigarette use was small and did not vary across sex or age strata, translating to a 9% increase in the likelihood of e-cigarette use or about .2 more days of e-cigarette use a month. Moreover, we uncovered no links between RML and adolescent cigarette or alcohol use.
      This is the first study to our knowledge to have assessed whether RML is associated with other types of substance use among adolescents, or across key demographic strata. The link with a rising prevalence of e-cigarettes is particularly concerning, coinciding with an exponential rise of this new type of tobacco product [
      • Gentzke A.S.
      • Creamer M.
      • Cullen K.A.
      Vital signs: Tobacco product use among middle and high school students- United States, 2011–2018.
      ]. While a new federal law prohibiting the sale of e-cigarettes and other tobacco products to youth under the age 21 and partial ban of flavored e-cigarette products [
      FDA, Center for Tobacco Products
      Tobacco 21.
      ] may help stem the increase, the swift rise in such products has reversed a decades-long decline in adolescent tobacco use and raised notable concerns over negative health effects [
      CDC, Office on Smoking and Health
      Outbreak of lung injury associated with the use of E-cigarette, or vaping, products.
      ]. Moreover, research documents a connection between initiation of e-cigarette use and later marijuana use among adolescents [
      • Chadi N.
      • Schroeder R.
      • Jensen J.W.
      • Levy S.
      Association between electronic cigarette use and marijuana use among adolescents and young adults. A systematic review and meta-analysis.
      ], raising concerns over the possibility of longer-term repercussions of RML on adolescent marijuana use. It is worth noting that the e-cigarette use item in the YRBS does not specify use for smoking tobacco or marijuana, both of which could be used through this modality.
      A third finding from this research relates to consistency in results across sex and age strata, but some differential effects of RML polices across racial and ethnic groups. Given the heightened levels of marijuana use among youth of color, in addition to the inequitable legal repercussions of drug use among black and Latinx populations, understanding the response of these populations to policy shifts is key. Our results provided some suggestion that RML was associated with small declines in marijuana and e-cigarette use among adolescents of color but slightly increased levels of cigarette use, in comparison with their white peers.
      In interpreting the results of this study, it is essential to acknowledge limitations. Most importantly, although we analyzed the most recent national data available, assessing substance use through 2017, we had data before and after RML implementation from only six states, which included only one of the earliest RML states (Colorado) but excluded two others, Washington and Oregon, which did not participate in the YRBS. Moreover, in five of the included RML states, there was only one cohort of data post-RML implementation, often collected before the wide rollout of commercial marijuana dispensaries. As such, it is essential for future research to explore the longer-term repercussions of RML as commercial dispensaries open and the laws become more culturally ingrained. Variability in RML policies across states (such as the number of dispensaries or the amount of personal marijuana allowed in homes) was not assessed, neither were we able to attend to variability in policies at the municipal level, important issues for future research. And, only 2 years of data on e-cigarettes were available. As such, this is an initial view of short-term repercussions of state level RML, and it is essential to evaluate whether trends strengthen or shift over time. As RML spreads to more states and commercial sales dramatically increase, both availability and acceptance of marijuana use may expand. Moreover, we cannot rule out reporting bias from adolescent self-reports of substance use, sample bias from the exclusion of high school drop outs and others not attending school, or unmeasured heterogeneity bias driven by other factors.
      Beyond these cautions, results from this study suggest minimal effects of RML on adolescent substance use, with evidence of small declines in the frequency of marijuana use among users, and small increases in the likelihood of e-cigarette use, trend which are essential to track as RML matures and spreads.

      Acknowledgments

      The study sponsors had no role in the study design; the collection, analysis, and interpretation of data; the writing of the report; or the decision to submit the manuscript for publication.

      Funding Sources

      Research reported in this publication was supported by a Boston College Research Across Departments and Schools (RADS) grant to R.L.C.

      References

        • ProCon.org
        33 Legal medical marijuana states and D.C.: Laws, fees, and possession limits.
        (Available at:)
        • Smart R.
        • Pacula R.L.
        Early evidence of the impact of cannabis legalization on cannabis use, cannabis use disorder, and the use of other substances: Findings from state policy evaluations.
        Am J Drug Alcohol Abuse. 2019; 45: 644-663
        • CDC
        Trends in the prevalence of marijuana, cocaine, and other illegal drug use national YRBS: 1991—2017.
        (Available at:)
        • Fergusson D.M.
        • Boden J.M.
        • Horwood L.J.
        Cannabis use and other illicit drug use: Testing the cannabis gateway hypothesis.
        Addiction. 2006; 101: 556-569
        • Milicic S.
        • Leatherdale S.T.
        The associations between e-cigarettes and binge drinking, marijuana use, and energy drinks mixed with alcohol.
        J Adolesc Health. 2017; 60: 320-327
        • U.S. Department of Health and Human Services (HHS)
        Facing addiction in America: The Surgeon General’s report on alcohol, drugs, and health.
        HHS, Washington, DC2016: 6 (Available at:)
        https://www.ncbi.nlm.nih.gov/books/NBK424857/
        Date accessed: February 20, 2020
        • Anderson D.M.
        • Hansen B.
        • Rees D.I.
        Medical marijuana laws and teen marijuana use.
        Am Law Econ Rev. 2015; 17: 495-528
        • Cerdá M.
        • Sarvet A.L.
        • Wall M.
        • et al.
        Medical marijuana laws and adolescent use of marijuana and other substances: Alcohol, cigarettes, prescription drugs, and other illicit drugs.
        Drug Alcohol Depend. 2018; 183: 62-68
        • Coley R.L.
        • Hawkins S.S.
        • Ghiani M.
        • et al.
        A quasi-experimental evaluation of marijuana policies on youth marijuana use.
        Am J Drug Alcohol Abuse. 2019; 45: 292-303
        • Johnson J.K.
        • Johnson R.M.
        • Hodgkin D.
        • et al.
        Heterogeneity of state medical marijuana laws and adolescent recent use of alcohol and marijuana: Analysis of 45 states, 1991–2011.
        Subst Abus. 2018; 39: 247-254
        • Lynne-Landsman S.D.
        • Livingston M.D.
        • Wagenaar A.C.
        Effects of state medical marijuana laws on adolescent marijuana use.
        Am J Public Health. 2013; 103: 1500-1506
        • Sarvet A.L.
        • Wall M.M.
        • Fink D.S.
        • et al.
        Medical marijuana laws and adolescent marijuana use in the United States: A systematic review and meta-analysis.
        Addiction. 2018; 113: 1003-1016
        • Cerdá M.
        • Wall M.
        • Feng T.
        • et al.
        Association of state recreational marijuana laws with adolescent marijuana use.
        JAMA Pediatr. 2017; 171: 142-149
        • Cerdá M.
        • Mauro C.
        • Hamilton A.
        • et al.
        Association between recreational marijuana legalization in the United States and changes in marijuana use and cannabis use disorder from 2008 to 2016.
        JAMA Psychiatry. 2020; 77: 165-171
        • Anderson D.M.
        • Hansen B.
        • Rees D.I.
        • Sabia J.J.
        Association of marijuana laws with teen marijuana use: New estimates from the youth risk behavior surveys.
        JAMA Pediatr. 2019; 173: 879-881
        • CDC
        Combining YRBS data across years and sites.
        (2. Available at:)
        • Jones C.M.
        • Underwood J.M.
        • Volkow N.D.
        Challenging the association of marijuana laws with teen marijuana use.
        JAMA Pediatr. 2020; 174: 99
        • Rapoport E.
        • Keim S.A.
        • Adesman A.
        Challenging the association of marijuana laws with teen marijuana use.
        JAMA Pediatr. 2020; 174: 98-99
        • Resing C.
        Marijuana legalization is a racial justice issue.
        (Available at:)
        • Crost B.
        • Guerrero S.
        The effect of alcohol availability on marijuana use: Evidence from the minimum legal drinking age.
        J Health Econ. 2012; 31: 112-121
        • Guttmannova K.
        • Lee C.M.
        • Kilmer J.R.
        • et al.
        Impacts of changing marijuana policies on alcohol use in the United States.
        Alcohol Clin Exp Res. 2016; 40: 33-46
        • Williams J.
        • Pacula R.L.
        • Chaloupka F.J.
        • Wechsler H.
        Alcohol and marijuana use among college students: Economic complements or substitutes?.
        Health Econ. 2004; 13: 825-843
        • Alley Z.M.
        • Kerr D.C.R.
        • Bae H.
        Trends in college students’ alcohol, nicotine, prescription opioid and other drug use after recreational marijuana legalization: 2008–2018.
        Addict Behav. 2019; 102: 106212
        • Veligati S.
        • Howdeshell S.
        • Beeler-Stinn S.
        • et al.
        Changes in alcohol and cigarette consumption in response to medical and recreational cannabis legalization: Evidence from U.S. state tax receipt data.
        Int J Drug Policy. 2020; 75: 102585
        • Brener N.D.
        • Kann L.
        • Shanklin S.
        • et al.
        Methodology of the youth risk behavior surveillance system—2013.
        MMWR Morb Mortal Wkly Rep. 2013; 62: 1-20
        • Kann L.
        • McManus T.
        • Harris W.A.
        • et al.
        Youth risk behavior surveillance—United States, 2017.
        MMWR Surveill Summ. 2018; 67: 1
        • Distilled Spirit Council of the United States
        History of beverage alcohol tax changes.
        (Available at:)
        http://www.discus.org
        Date accessed: October 20, 2017
        • Distilled Spirits Council of the United States
        How high are beer taxes in your state?.
        (Available at:)
        https://taxfoundation.org/beer-taxes-state/
        Date accessed: August 20, 2019
        • Orzechowski W.
        • Walker R.C.
        The tax burden on tobacco. Historical compilation. vol. 48. Centers for Disease Control and Prevention, Arlington, VA2018 (Arlington, VA)
        • American Nonsmokers’ Rights Foundation
        Chronological table of US population protected by 100% smokefree state or local laws: January 2, 2020.
        (Available at:)
        • Wing C.
        • Simon K.
        • Bello-Gomez R.A.
        Designing difference in difference studies: Best practices for public health policy research.
        Annu Rev Public Health. 2018; 39: 453-469
        • Bertrand M.
        • Duflo E.
        • Mullainathan S.
        How much should we trust differences-in-differences estimates?.
        Q J Econ. 2004; 119: 249-275
        • Clark P.A.
        • Capuzzi K.
        • Fick C.
        Medical marijuana: Medical necessity versus political agenda.
        Med Sci Monit. 2011; 17: RA249
        • Hasin D.S.
        • Wall M.
        • Keyes K.M.
        • et al.
        Medical marijuana laws and adolescent marijuana use in the USA from 1991 to 2014: Results from annual, repeated cross-sectional surveys.
        Lancet Psychiatry. 2015; 2: 601-608
        • Gentzke A.S.
        • Creamer M.
        • Cullen K.A.
        Vital signs: Tobacco product use among middle and high school students- United States, 2011–2018.
        MMWR Morb Mortal Wkly Rep. 2019; 68: 157-164
        • FDA, Center for Tobacco Products
        Tobacco 21.
        (Available at:)
        • CDC, Office on Smoking and Health
        Outbreak of lung injury associated with the use of E-cigarette, or vaping, products.
        (Available at:)
        • Chadi N.
        • Schroeder R.
        • Jensen J.W.
        • Levy S.
        Association between electronic cigarette use and marijuana use among adolescents and young adults. A systematic review and meta-analysis.
        JAMA Pediatr. 2019; 173: e192574